2016-10-06 13:21:30 +02:00
// Copyright 2015-2016 Espressif Systems (Shanghai) PTE LTD
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
# include "esp32-hal-i2c.h"
2016-10-11 14:11:51 +02:00
# include "esp32-hal.h"
2016-10-06 13:21:30 +02:00
# include "freertos/FreeRTOS.h"
# include "freertos/task.h"
2016-10-11 13:20:02 +02:00
# include "freertos/semphr.h"
2018-06-27 09:01:06 +02:00
# include "freertos/event_groups.h"
2016-10-06 13:21:30 +02:00
# include "rom/ets_sys.h"
2017-09-29 15:17:13 +02:00
# include "driver/periph_ctrl.h"
2016-10-06 13:21:30 +02:00
# include "soc/i2c_reg.h"
2016-10-11 14:11:51 +02:00
# include "soc/i2c_struct.h"
2016-10-06 13:21:30 +02:00
# include "soc/dport_reg.h"
2018-06-27 09:01:06 +02:00
# include "esp_attr.h"
2019-01-09 10:07:54 +01:00
# include "esp32-hal-cpu.h" // cpu clock change support 31DEC2018
2016-10-11 14:11:51 +02:00
//#define I2C_DEV(i) (volatile i2c_dev_t *)((i)?DR_REG_I2C1_EXT_BASE:DR_REG_I2C_EXT_BASE)
//#define I2C_DEV(i) ((i2c_dev_t *)(REG_I2C_BASE(i)))
2016-10-06 13:21:30 +02:00
# define I2C_SCL_IDX(p) ((p==0)?I2CEXT0_SCL_OUT_IDX:((p==1)?I2CEXT1_SCL_OUT_IDX:0))
# define I2C_SDA_IDX(p) ((p==0)?I2CEXT0_SDA_OUT_IDX:((p==1)?I2CEXT1_SDA_OUT_IDX:0))
2017-03-25 23:27:27 +01:00
# define DR_REG_I2C_EXT_BASE_FIXED 0x60013000
# define DR_REG_I2C1_EXT_BASE_FIXED 0x60027000
2016-10-11 13:20:02 +02:00
2018-07-24 19:43:45 +02:00
/* Stickbreaker ISR mode debug support
ENABLE_I2C_DEBUG_BUFFER
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
Enable debug interrupt history buffer , fifoTx history buffer .
Setting this define will result in 2570 bytes of RAM being used whenever CORE_DEBUG_LEVEL
is higher than WARNING . Unless you are debugging a problem in the I2C subsystem ,
I would recommend you leave it commented out .
2018-07-24 19:43:45 +02:00
*/
//#define ENABLE_I2C_DEBUG_BUFFER
# if (ARDUHAL_LOG_LEVEL >= ARDUHAL_LOG_LEVEL_INFO) && (defined ENABLE_I2C_DEBUG_BUFFER)
# define INTBUFFMAX 64
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
# define FIFOMAX 512
2018-07-24 19:43:45 +02:00
static uint32_t intBuff [ INTBUFFMAX ] [ 3 ] [ 2 ] ;
static uint32_t intPos [ 2 ] = { 0 , 0 } ;
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
static uint16_t fifoBuffer [ FIFOMAX ] ;
static uint16_t fifoPos = 0 ;
2018-07-24 19:43:45 +02:00
# endif
2018-06-27 09:01:06 +02:00
// start from tools/sdk/include/soc/soc/i2c_struct.h
typedef union {
struct {
uint32_t byte_num : 8 ; /*Byte_num represent the number of data need to be send or data need to be received.*/
uint32_t ack_en : 1 ; /*ack_check_en ack_exp and ack value are used to control the ack bit.*/
uint32_t ack_exp : 1 ; /*ack_check_en ack_exp and ack value are used to control the ack bit.*/
uint32_t ack_val : 1 ; /*ack_check_en ack_exp and ack value are used to control the ack bit.*/
uint32_t op_code : 3 ; /*op_code is the command 0： ： ： ：
uint32_t reserved14 : 17 ;
uint32_t done : 1 ; /*When command0 is done in I2C Master mode this bit changes to high level.*/
} ;
uint32_t val ;
} I2C_COMMAND_t ;
typedef union {
struct {
uint32_t rx_fifo_full_thrhd : 5 ;
uint32_t tx_fifo_empty_thrhd : 5 ; //Config tx_fifo empty threhd value when using apb fifo access * /
uint32_t nonfifo_en : 1 ; //Set this bit to enble apb nonfifo access. * /
uint32_t fifo_addr_cfg_en : 1 ; //When this bit is set to 1 then the byte after address represent the offset address of I2C Slave's ram. * /
uint32_t rx_fifo_rst : 1 ; //Set this bit to reset rx fifo when using apb fifo access. * /
// chuck while this bit is 1, the RX fifo is held in REST, Toggle it * /
uint32_t tx_fifo_rst : 1 ; //Set this bit to reset tx fifo when using apb fifo access. * /
// chuck while this bit is 1, the TX fifo is held in REST, Toggle it * /
uint32_t nonfifo_rx_thres : 6 ; //when I2C receives more than nonfifo_rx_thres data it will produce rx_send_full_int_raw interrupt and update the current offset address of the receiving data.* /
uint32_t nonfifo_tx_thres : 6 ; //when I2C sends more than nonfifo_tx_thres data it will produce tx_send_empty_int_raw interrupt and update the current offset address of the sending data. * /
uint32_t reserved26 : 6 ;
} ;
uint32_t val ;
} I2C_FIFO_CONF_t ;
2018-07-24 19:43:45 +02:00
typedef union {
struct {
uint32_t rx_fifo_start_addr : 5 ; /*This is the offset address of the last receiving data as described in nonfifo_rx_thres_register.*/
uint32_t rx_fifo_end_addr : 5 ; /*This is the offset address of the first receiving data as described in nonfifo_rx_thres_register.*/
uint32_t tx_fifo_start_addr : 5 ; /*This is the offset address of the first sending data as described in nonfifo_tx_thres register.*/
uint32_t tx_fifo_end_addr : 5 ; /*This is the offset address of the last sending data as described in nonfifo_tx_thres register.*/
uint32_t reserved20 : 12 ;
} ;
uint32_t val ;
} I2C_FIFO_ST_t ;
2018-06-27 09:01:06 +02:00
// end from tools/sdk/include/soc/soc/i2c_struct.h
// sync between dispatch(i2cProcQueue) and worker(i2c_isr_handler_default)
typedef enum {
//I2C_NONE=0,
I2C_STARTUP = 1 ,
I2C_RUNNING ,
I2C_DONE
} I2C_STAGE_t ;
typedef enum {
I2C_NONE = 0 ,
I2C_MASTER ,
I2C_SLAVE ,
I2C_MASTERSLAVE
} I2C_MODE_t ;
// internal Error condition
typedef enum {
// I2C_NONE=0,
I2C_OK = 1 ,
I2C_ERROR ,
I2C_ADDR_NAK ,
I2C_DATA_NAK ,
I2C_ARBITRATION ,
I2C_TIMEOUT
} I2C_ERROR_t ;
// i2c_event bits for EVENTGROUP bits
// needed to minimize change events, FreeRTOS Daemon overload, so ISR will only set values
// on Exit. Dispatcher will set bits for each dq before/after ISR completion
# define EVENT_ERROR_NAK (BIT(0))
# define EVENT_ERROR (BIT(1))
# define EVENT_ERROR_BUS_BUSY (BIT(2))
# define EVENT_RUNNING (BIT(3))
# define EVENT_DONE (BIT(4))
# define EVENT_IN_END (BIT(5))
# define EVENT_ERROR_PREV (BIT(6))
# define EVENT_ERROR_TIMEOUT (BIT(7))
# define EVENT_ERROR_ARBITRATION (BIT(8))
# define EVENT_ERROR_DATA_NAK (BIT(9))
# define EVENT_MASK 0x3F
// control record for each dq entry
typedef union {
struct {
uint32_t addr : 16 ; // I2C address, if 10bit must have 0x7800 mask applied, else 8bit
uint32_t mode : 1 ; // transaction direction 0 write, 1 read
uint32_t stop : 1 ; // sendStop 0 no, 1 yes
uint32_t startCmdSent : 1 ; // START cmd has been added to command[]
uint32_t addrCmdSent : 1 ; // addr WRITE cmd has been added to command[]
uint32_t dataCmdSent : 1 ; // all necessary DATA(READ/WRITE) cmds added to command[]
uint32_t stopCmdSent : 1 ; // completed all necessary commands
uint32_t addrReq : 2 ; // number of addr bytes need to send address
uint32_t addrSent : 2 ; // number of addr bytes added to FIFO
uint32_t reserved_31 : 6 ;
} ;
uint32_t val ;
} I2C_DATA_CTRL_t ;
// individual dq element
typedef struct {
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
uint8_t * data ; // data pointer for read/write buffer
2018-06-27 09:01:06 +02:00
uint16_t length ; // size of data buffer
uint16_t position ; // current position for next char in buffer (<length)
2018-07-24 19:43:45 +02:00
uint16_t cmdBytesNeeded ; // used to control number of I2C_COMMAND_t blocks added to queue
2018-06-27 09:01:06 +02:00
I2C_DATA_CTRL_t ctrl ;
EventGroupHandle_t queueEvent ; // optional user supplied for Async feedback EventBits
} I2C_DATA_QUEUE_t ;
2017-10-21 14:26:46 +02:00
2016-10-11 13:20:02 +02:00
struct i2c_struct_t {
i2c_dev_t * dev ;
2016-11-18 14:07:25 +01:00
# if !CONFIG_DISABLE_HAL_LOCKS
2016-10-11 13:20:02 +02:00
xSemaphoreHandle lock ;
2016-11-18 14:07:25 +01:00
# endif
2016-10-11 13:20:02 +02:00
uint8_t num ;
2018-06-27 09:01:06 +02:00
int8_t sda ;
int8_t scl ;
I2C_MODE_t mode ;
I2C_STAGE_t stage ;
I2C_ERROR_t error ;
EventGroupHandle_t i2c_event ; // a way to monitor ISR process
// maybe use it to trigger callback for OnRequest()
intr_handle_t intr_handle ; /*!< I2C interrupt handle*/
I2C_DATA_QUEUE_t * dq ;
2018-07-24 19:43:45 +02:00
uint16_t queueCount ; // number of dq entries in queue.
uint16_t queuePos ; // current queue that still has or needs data (out/in)
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
int16_t errorByteCnt ; // byte pos where error happened, -1 devId, 0..(length-1) data
2018-07-24 19:43:45 +02:00
uint16_t errorQueue ; // errorByteCnt is in this queue,(for error locus)
2018-06-27 09:01:06 +02:00
uint32_t exitCode ;
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
uint32_t debugFlags ;
2016-10-11 13:20:02 +02:00
} ;
enum {
I2C_CMD_RSTART ,
I2C_CMD_WRITE ,
I2C_CMD_READ ,
I2C_CMD_STOP ,
I2C_CMD_END
} ;
2016-11-18 14:07:25 +01:00
# if CONFIG_DISABLE_HAL_LOCKS
# define I2C_MUTEX_LOCK()
# define I2C_MUTEX_UNLOCK()
static i2c_t _i2c_bus_array [ 2 ] = {
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
{ ( volatile i2c_dev_t * ) ( DR_REG_I2C_EXT_BASE_FIXED ) , 0 , - 1 , - 1 , I2C_NONE , I2C_NONE , I2C_ERROR_OK , NULL , NULL , NULL , 0 , 0 , 0 , 0 , 0 } ,
{ ( volatile i2c_dev_t * ) ( DR_REG_I2C1_EXT_BASE_FIXED ) , 1 , - 1 , - 1 , I2C_NONE , I2C_NONE , I2C_ERROR_OK , NULL , NULL , NULL , 0 , 0 , 0 , 0 , 0 }
2016-11-18 14:07:25 +01:00
} ;
# else
2019-01-09 10:07:54 +01:00
# define I2C_MUTEX_LOCK() do {} while (xSemaphoreTakeRecursive(i2c->lock, portMAX_DELAY) != pdPASS)
# define I2C_MUTEX_UNLOCK() xSemaphoreGiveRecursive(i2c->lock)
2016-10-11 13:20:02 +02:00
static i2c_t _i2c_bus_array [ 2 ] = {
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
{ ( volatile i2c_dev_t * ) ( DR_REG_I2C_EXT_BASE_FIXED ) , NULL , 0 , - 1 , - 1 , I2C_NONE , I2C_NONE , I2C_ERROR_OK , NULL , NULL , NULL , 0 , 0 , 0 , 0 , 0 , 0 } ,
{ ( volatile i2c_dev_t * ) ( DR_REG_I2C1_EXT_BASE_FIXED ) , NULL , 1 , - 1 , - 1 , I2C_NONE , I2C_NONE , I2C_ERROR_OK , NULL , NULL , NULL , 0 , 0 , 0 , 0 , 0 , 0 }
2016-10-11 13:20:02 +02:00
} ;
2016-11-18 14:07:25 +01:00
# endif
2016-10-11 13:20:02 +02:00
2018-06-27 09:01:06 +02:00
/*
* index - command index ( 0 to 15 )
* op_code - is the command
* byte_num - This register is to store the amounts of data that is read and written . byte_num in RSTART , STOP , END is null .
* ack_val - Each data byte is terminated by an ACK bit used to set the bit level .
* ack_exp - This bit is to set an expected ACK value for the transmitter .
* ack_check - This bit is to decide whether the transmitter checks ACK bit . 1 means yes and 0 means no .
* */
/* Stickbreaker ISR mode debug support
*/
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
static void IRAM_ATTR i2cDumpCmdQueue ( i2c_t * i2c )
2018-06-27 09:01:06 +02:00
{
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
# if (ARDUHAL_LOG_LEVEL >= ARDUHAL_LOG_LEVEL_ERROR)&&(defined ENABLE_I2C_DEBUG_BUFFER)
static const char * const cmdName [ ] = { " RSTART " , " WRITE " , " READ " , " STOP " , " END " } ;
2018-06-27 09:01:06 +02:00
uint8_t i = 0 ;
while ( i < 16 ) {
I2C_COMMAND_t c ;
c . val = i2c - > dev - > command [ i ] . val ;
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
log_e ( " [%2d] \t %c \t %s \t val[%d] \t exp[%d] \t en[%d] \t bytes[%d] " , i , ( c . done ? ' Y ' : ' N ' ) ,
cmdName [ c . op_code ] ,
2018-06-27 09:01:06 +02:00
c . ack_val ,
c . ack_exp ,
c . ack_en ,
c . byte_num ) ;
i + + ;
}
# endif
}
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
/* Stickbreaker ISR mode debug support
2018-06-27 09:01:06 +02:00
*/
2018-11-29 11:34:55 +01:00
# if (ARDUHAL_LOG_LEVEL >= ARDUHAL_LOG_LEVEL_INFO)
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
static void i2cDumpDqData ( i2c_t * i2c )
{
2018-11-29 11:34:55 +01:00
# if defined (ENABLE_I2C_DEBUG_BUFFER)
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
uint16_t a = 0 ;
char buff [ 140 ] ;
I2C_DATA_QUEUE_t * tdq ;
int digits = 0 , lenDigits = 0 ;
a = i2c - > queueCount ;
while ( a > 0 ) {
digits + + ;
a / = 10 ;
}
while ( a < i2c - > queueCount ) { // find maximum number of len decimal digits for formatting
if ( i2c - > dq [ a ] . length > lenDigits ) lenDigits = i2c - > dq [ a ] . length ;
a + + ;
}
a = 0 ;
while ( lenDigits > 0 ) {
a + + ;
lenDigits / = 10 ;
}
lenDigits = a ;
a = 0 ;
while ( a < i2c - > queueCount ) {
tdq = & i2c - > dq [ a ] ;
char buf1 [ 10 ] , buf2 [ 10 ] ;
sprintf ( buf1 , " %0*d " , lenDigits , tdq - > length ) ;
sprintf ( buf2 , " %0*d " , lenDigits , tdq - > position ) ;
log_i ( " [%0*d] %sbit %x %c %s buf@=%p, len=%s, pos=%s, ctrl=%d%d%d%d%d " , digits , a ,
( tdq - > ctrl . addr > 0x100 ) ? " 10 " : " 7 " ,
( tdq - > ctrl . addr > 0x100 ) ? ( ( ( tdq - > ctrl . addr & 0x600 ) > > 1 ) | ( tdq - > ctrl . addr & 0xff ) ) : ( tdq - > ctrl . addr > > 1 ) ,
( tdq - > ctrl . mode ) ? ' R ' : ' W ' ,
( tdq - > ctrl . stop ) ? " STOP " : " " ,
tdq - > data ,
buf1 , buf2 ,
tdq - > ctrl . startCmdSent , tdq - > ctrl . addrCmdSent , tdq - > ctrl . dataCmdSent , ( tdq - > ctrl . stop ) ? tdq - > ctrl . stopCmdSent : 0 , tdq - > ctrl . addrSent
) ;
uint16_t offset = 0 ;
while ( offset < tdq - > length ) {
memset ( buff , ' ' , 140 ) ;
buff [ 139 ] = ' \0 ' ;
uint16_t i = 0 , j ;
j = sprintf ( buff , " 0x%04x: " , offset ) ;
while ( ( i < 32 ) & & ( offset < tdq - > length ) ) {
char ch = tdq - > data [ offset ] ;
sprintf ( ( char * ) & buff [ ( i * 3 ) + 41 ] , " %02x " , ch ) ;
if ( ( ch < 32 ) | | ( ch > 126 ) ) {
ch = ' . ' ;
}
j + = sprintf ( ( char * ) & buff [ j ] , " %c " , ch ) ;
buff [ j ] = ' ' ;
i + + ;
offset + + ;
}
log_i ( " %s " , buff ) ;
}
a + + ;
}
2018-11-29 11:34:55 +01:00
# else
log_i ( " Debug Buffer not Enabled " ) ;
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
# endif
}
2018-11-29 11:34:55 +01:00
# endif
# if ARDUHAL_LOG_LEVEL >= ARDUHAL_LOG_LEVEL_INFO
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
static void i2cDumpI2c ( i2c_t * i2c )
{
log_e ( " i2c=%p " , i2c ) ;
log_i ( " dev=%p date=%p " , i2c - > dev , i2c - > dev - > date ) ;
# if !CONFIG_DISABLE_HAL_LOCKS
log_i ( " lock=%p " , i2c - > lock ) ;
# endif
log_i ( " num=%d " , i2c - > num ) ;
log_i ( " mode=%d " , i2c - > mode ) ;
log_i ( " stage=%d " , i2c - > stage ) ;
log_i ( " error=%d " , i2c - > error ) ;
log_i ( " event=%p bits=%x " , i2c - > i2c_event , ( i2c - > i2c_event ) ? xEventGroupGetBits ( i2c - > i2c_event ) : 0 ) ;
log_i ( " intr_handle=%p " , i2c - > intr_handle ) ;
log_i ( " dq=%p " , i2c - > dq ) ;
log_i ( " queueCount=%d " , i2c - > queueCount ) ;
log_i ( " queuePos=%d " , i2c - > queuePos ) ;
log_i ( " errorByteCnt=%d " , i2c - > errorByteCnt ) ;
log_i ( " errorQueue=%d " , i2c - > errorQueue ) ;
log_i ( " debugFlags=0x%08X " , i2c - > debugFlags ) ;
if ( i2c - > dq ) {
i2cDumpDqData ( i2c ) ;
}
}
2018-11-29 11:34:55 +01:00
# endif
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
2018-11-29 11:34:55 +01:00
# if (ARDUHAL_LOG_LEVEL >= ARDUHAL_LOG_LEVEL_INFO)
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
static void i2cDumpInts ( uint8_t num )
{
2018-11-29 11:34:55 +01:00
# if defined (ENABLE_I2C_DEBUG_BUFFER)
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
uint32_t b ;
log_i ( " %u row \t count \t INTR \t TX \t RX \t Tick " , num ) ;
for ( uint32_t a = 1 ; a < = INTBUFFMAX ; a + + ) {
b = ( a + intPos [ num ] ) % INTBUFFMAX ;
if ( intBuff [ b ] [ 0 ] [ num ] ! = 0 ) {
log_i ( " [%02d] \t 0x%04x \t 0x%04x \t 0x%04x \t 0x%04x \t 0x%08x " , b , ( ( intBuff [ b ] [ 0 ] [ num ] > > 16 ) & 0xFFFF ) , ( intBuff [ b ] [ 0 ] [ num ] & 0xFFFF ) , ( ( intBuff [ b ] [ 1 ] [ num ] > > 16 ) & 0xFFFF ) , ( intBuff [ b ] [ 1 ] [ num ] & 0xFFFF ) , intBuff [ b ] [ 2 ] [ num ] ) ;
}
}
# else
log_i ( " Debug Buffer not Enabled " ) ;
# endif
}
2018-11-29 11:34:55 +01:00
# endif
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
# if (ARDUHAL_LOG_LEVEL >= ARDUHAL_LOG_LEVEL_INFO)&&(defined ENABLE_I2C_DEBUG_BUFFER)
2018-11-29 11:34:55 +01:00
static void IRAM_ATTR i2cDumpStatus ( i2c_t * i2c ) {
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
typedef union {
struct {
uint32_t ack_rec : 1 ; /*This register stores the value of ACK bit.*/
uint32_t slave_rw : 1 ; /*when in slave mode 1：
uint32_t time_out : 1 ; /*when I2C takes more than time_out_reg clocks to receive a data then this register changes to high level.*/
uint32_t arb_lost : 1 ; /*when I2C lost control of SDA line this register changes to high level.*/
uint32_t bus_busy : 1 ; /*1:I2C bus is busy transferring data. 0:I2C bus is in idle state.*/
uint32_t slave_addressed : 1 ; /*when configured as i2c slave and the address send by master is equal to slave's address then this bit will be high level.*/
uint32_t byte_trans : 1 ; /*This register changes to high level when one byte is transferred.*/
uint32_t reserved7 : 1 ;
uint32_t rx_fifo_cnt : 6 ; /*This register represent the amount of data need to send.*/
uint32_t reserved14 : 4 ;
uint32_t tx_fifo_cnt : 6 ; /*This register stores the amount of received data in ram.*/
uint32_t scl_main_state_last : 3 ; /*This register stores the value of state machine for i2c module. 3'h0: SCL_MAIN_IDLE 3'h1: SCL_ADDRESS_SHIFT 3'h2: SCL_ACK_ADDRESS 3'h3: SCL_RX_DATA 3'h4 SCL_TX_DATA 3'h5:SCL_SEND_ACK 3'h6:SCL_WAIT_ACK*/
uint32_t reserved27 : 1 ;
uint32_t scl_state_last : 3 ; /*This register stores the value of state machine to produce SCL. 3'h0: SCL_IDLE 3'h1:SCL_START 3'h2:SCL_LOW_EDGE 3'h3: SCL_LOW 3'h4:SCL_HIGH_EDGE 3'h5:SCL_HIGH 3'h6:SCL_STOP*/
uint32_t reserved31 : 1 ;
} ;
uint32_t val ;
} status_reg ;
status_reg sr ;
sr . val = i2c - > dev - > status_reg . val ;
log_i ( " ack(%d) sl_rw(%d) to(%d) arb(%d) busy(%d) sl(%d) trans(%d) rx(%d) tx(%d) sclMain(%d) scl(%d) " , sr . ack_rec , sr . slave_rw , sr . time_out , sr . arb_lost , sr . bus_busy , sr . slave_addressed , sr . byte_trans , sr . rx_fifo_cnt , sr . tx_fifo_cnt , sr . scl_main_state_last , sr . scl_state_last ) ;
}
2018-11-29 11:34:55 +01:00
# endif
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
# if (ARDUHAL_LOG_LEVEL >= ARDUHAL_LOG_LEVEL_INFO)&&(defined ENABLE_I2C_DEBUG_BUFFER)
2018-11-29 11:34:55 +01:00
static void i2cDumpFifo ( i2c_t * i2c ) {
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
char buf [ 64 ] ;
uint16_t k = 0 ;
uint16_t i = fifoPos + 1 ;
i % = FIFOMAX ;
while ( ( fifoBuffer [ i ] = = 0 ) & & ( i ! = fifoPos ) ) {
i + + ;
i % = FIFOMAX ;
}
if ( i ! = fifoPos ) { // actual data
do {
if ( fifoBuffer [ i ] & 0x8000 ) { // address byte
if ( fifoBuffer [ i ] & 0x100 ) { // read
if ( fifoBuffer [ i ] & 0x1 ) { // valid read dev id
k + = sprintf ( & buf [ k ] , " READ 0x%02X " , ( fifoBuffer [ i ] & 0xff ) > > 1 ) ;
} else { // invalid read dev id
k + = sprintf ( & buf [ k ] , " Bad READ 0x%02X " , ( fifoBuffer [ i ] & 0xff ) ) ;
}
} else { // write
if ( fifoBuffer [ i ] & 0x1 ) { // bad write dev id
k + = sprintf ( & buf [ k ] , " bad WRITE 0x%02X " , ( fifoBuffer [ i ] & 0xff ) ) ;
} else { // good Write
k + = sprintf ( & buf [ k ] , " WRITE 0x%02X " , ( fifoBuffer [ i ] & 0xff ) > > 1 ) ;
}
}
} else k + = sprintf ( & buf [ k ] , " % 4X " , fifoBuffer [ i ] ) ;
i + + ;
i % = FIFOMAX ;
bool outBuffer = false ;
if ( fifoBuffer [ i ] & 0x8000 ) {
outBuffer = true ;
k = 0 ;
}
if ( ( outBuffer ) | | ( k > 50 ) | | ( i = = fifoPos ) ) log_i ( " %s " , buf ) ;
outBuffer = false ;
if ( k > 50 ) {
k = sprintf ( buf , " -> " ) ;
}
} while ( i ! = fifoPos ) ;
}
}
2018-11-29 11:34:55 +01:00
# endif
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
static void IRAM_ATTR i2cTriggerDumps ( i2c_t * i2c , uint8_t trigger , const char locus [ ] ) {
# if (ARDUHAL_LOG_LEVEL >= ARDUHAL_LOG_LEVEL_INFO)&&(defined ENABLE_I2C_DEBUG_BUFFER)
if ( trigger ) {
log_i ( " %s " , locus ) ;
if ( trigger & 1 ) i2cDumpI2c ( i2c ) ;
if ( trigger & 2 ) i2cDumpInts ( i2c - > num ) ;
if ( trigger & 4 ) i2cDumpCmdQueue ( i2c ) ;
if ( trigger & 8 ) i2cDumpStatus ( i2c ) ;
if ( trigger & 16 ) i2cDumpFifo ( i2c ) ;
}
# endif
}
// end of debug support routines
2019-01-09 10:07:54 +01:00
/* Start of CPU Clock change Support
*/
static void i2cApbChangeCallback ( void * arg , apb_change_ev_t ev_type , uint32_t old_apb , uint32_t new_apb ) {
i2c_t * i2c = ( i2c_t * ) arg ; // recover data
if ( i2c = = NULL ) { // point to peripheral control block does not exits
2019-02-14 13:09:05 +01:00
return ;
2019-01-09 10:07:54 +01:00
}
uint32_t oldFreq = 0 ;
switch ( ev_type ) {
case APB_BEFORE_CHANGE :
if ( new_apb < 3000000 ) { // too slow
log_e ( " apb speed %d too slow " , new_apb ) ;
break ;
}
I2C_MUTEX_LOCK ( ) ; // lock will spin until current transaction is completed
break ;
case APB_AFTER_CHANGE :
oldFreq = ( i2c - > dev - > scl_low_period . period + i2c - > dev - > scl_high_period . period ) ; //read old apbCycles
if ( oldFreq > 0 ) { // was configured with value
oldFreq = old_apb / oldFreq ;
i2cSetFrequency ( i2c , oldFreq ) ;
}
I2C_MUTEX_UNLOCK ( ) ;
break ;
default :
log_e ( " unk ev %u " , ev_type ) ;
I2C_MUTEX_UNLOCK ( ) ;
}
return ;
}
/* End of CPU Clock change Support
*/
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
static void IRAM_ATTR i2cSetCmd ( i2c_t * i2c , uint8_t index , uint8_t op_code , uint8_t byte_num , bool ack_val , bool ack_exp , bool ack_check )
{
I2C_COMMAND_t cmd ;
cmd . val = 0 ;
cmd . ack_en = ack_check ;
cmd . ack_exp = ack_exp ;
cmd . ack_val = ack_val ;
cmd . byte_num = byte_num ;
cmd . op_code = op_code ;
i2c - > dev - > command [ index ] . val = cmd . val ;
}
2018-06-27 09:01:06 +02:00
static void IRAM_ATTR fillCmdQueue ( i2c_t * i2c , bool INTS )
{
2018-07-24 19:43:45 +02:00
/* this function is called on initial i2cProcQueue() or when a I2C_END_DETECT_INT occurs
2018-06-27 09:01:06 +02:00
*/
uint16_t cmdIdx = 0 ;
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
uint16_t qp = i2c - > queuePos ; // all queues before queuePos have been completely processed,
// so look start by checking the 'current queue' so see if it needs commands added to
// hardware peripheral command list. walk through each queue entry until all queues have been
// checked
2018-06-27 09:01:06 +02:00
bool done ;
bool needMoreCmds = false ;
bool ena_rx = false ; // if we add a read op, better enable Rx_Fifo IRQ
bool ena_tx = false ; // if we add a Write op, better enable TX_Fifo IRQ
while ( ! needMoreCmds & & ( qp < i2c - > queueCount ) ) { // check if more possible cmds
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
if ( i2c - > dq [ qp ] . ctrl . stopCmdSent ) { // marks that all required cmds[] have been added to peripheral
2018-06-27 09:01:06 +02:00
qp + + ;
} else {
needMoreCmds = true ;
}
}
//log_e("needMoreCmds=%d",needMoreCmds);
done = ( ! needMoreCmds ) | | ( cmdIdx > 14 ) ;
while ( ! done ) { // fill the command[] until either 0..14 filled or out of cmds and last cmd is STOP
//CMD START
I2C_DATA_QUEUE_t * tdq = & i2c - > dq [ qp ] ; // simpler coding
if ( ( ! tdq - > ctrl . startCmdSent ) & & ( cmdIdx < 14 ) ) { // has this dq element's START command been added?
2018-07-24 19:43:45 +02:00
// (cmdIdx<14) because a START op cannot directly precede an END op, else a time out cascade occurs
2018-06-27 09:01:06 +02:00
i2cSetCmd ( i2c , cmdIdx + + , I2C_CMD_RSTART , 0 , false , false , false ) ;
tdq - > ctrl . startCmdSent = 1 ;
}
//CMD WRITE ADDRESS
2018-07-24 19:43:45 +02:00
if ( ( ! done ) & & ( tdq - > ctrl . startCmdSent ) ) { // have to leave room for continue(END), and START must have been sent!
2018-06-27 09:01:06 +02:00
if ( ! tdq - > ctrl . addrCmdSent ) {
i2cSetCmd ( i2c , cmdIdx + + , I2C_CMD_WRITE , tdq - > ctrl . addrReq , false , false , true ) ; //load address in cmdlist, validate (low) ack
tdq - > ctrl . addrCmdSent = 1 ;
done = ( cmdIdx > 14 ) ;
ena_tx = true ; // tx Data necessary
}
}
/* Can I have another Sir?
ALL CMD queues must be terminated with either END or STOP .
If END is used , when refilling the cmd [ ] next time , no entries from END to [ 15 ] can be used .
AND the cmd [ ] must be filled starting at [ 0 ] with commands . Either fill all 15 [ 0 ] . . [ 14 ] and leave the
2018-07-24 19:43:45 +02:00
END in [ 15 ] or include a STOP in one of the positions [ 0 ] . . [ 14 ] . Any entries after a STOP are IGNORED by the StateMachine .
2018-06-27 09:01:06 +02:00
The END operation does not complete until ctr - > trans_start = 1 has been issued .
2018-07-24 19:43:45 +02:00
So , only refill from [ 0 ] . . [ 14 ] , leave [ 15 ] for a continuation ( END ) if necessary .
As a corollary , once END exists in [ 15 ] , you do not need to overwrite it for the
2018-06-27 09:01:06 +02:00
next continuation . It is never modified . But , I update it every time because it might
actually be the first time !
23 NOV17 START cannot proceed END . if START is in [ 14 ] , END cannot be in [ 15 ] .
2018-07-24 19:43:45 +02:00
So , if END is moved to [ 14 ] , [ 14 ] and [ 15 ] can no longer be used for anything other than END .
2018-06-27 09:01:06 +02:00
If a START is found in [ 14 ] then a prior READ or WRITE must be expanded so that there is no START element in [ 14 ] .
*/
2018-07-24 19:43:45 +02:00
if ( ( ! done ) & & ( tdq - > ctrl . addrCmdSent ) ) { //room in command[] for at least One data (read/Write) cmd
uint8_t blkSize = 0 ; // max is 255
2018-06-27 09:01:06 +02:00
while ( ( tdq - > cmdBytesNeeded > tdq - > ctrl . mode ) & & ( ! done ) ) { // more bytes needed and room in cmd queue, leave room for END
2018-07-24 19:43:45 +02:00
blkSize = ( tdq - > cmdBytesNeeded > 255 ) ? 255 : ( tdq - > cmdBytesNeeded - tdq - > ctrl . mode ) ; // Last read cmd needs different ACK setting, so leave 1 byte remainder on reads
tdq - > cmdBytesNeeded - = blkSize ;
2018-06-27 09:01:06 +02:00
if ( tdq - > ctrl . mode = = 1 ) { //read mode
i2cSetCmd ( i2c , ( cmdIdx ) + + , I2C_CMD_READ , blkSize , false , false , false ) ; // read cmd, this can't be the last read.
ena_rx = true ; // need to enable rxFifo IRQ
} else { // write
i2cSetCmd ( i2c , cmdIdx + + , I2C_CMD_WRITE , blkSize , false , false , true ) ; // check for Nak
ena_tx = true ; // need to enable txFifo IRQ
}
done = cmdIdx > 14 ; //have to leave room for END
}
if ( ! done ) { // buffer is not filled completely
if ( ( tdq - > ctrl . mode = = 1 ) & & ( tdq - > cmdBytesNeeded = = 1 ) ) { //special last read byte NAK
i2cSetCmd ( i2c , ( cmdIdx ) + + , I2C_CMD_READ , 1 , true , false , false ) ;
// send NAK to mark end of read
tdq - > cmdBytesNeeded = 0 ;
done = cmdIdx > 14 ;
ena_rx = true ;
}
}
tdq - > ctrl . dataCmdSent = ( tdq - > cmdBytesNeeded = = 0 ) ; // enough command[] to send all data
if ( ( ! done ) & & ( tdq - > ctrl . dataCmdSent ) ) { // possibly add stop
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
if ( ! tdq - > ctrl . stopCmdSent ) {
if ( tdq - > ctrl . stop ) { //send a stop
i2cSetCmd ( i2c , ( cmdIdx ) + + , I2C_CMD_STOP , 0 , false , false , false ) ;
done = cmdIdx > 14 ;
}
2018-06-27 09:01:06 +02:00
tdq - > ctrl . stopCmdSent = 1 ;
}
}
}
if ( ( cmdIdx = = 14 ) & & ( ! tdq - > ctrl . startCmdSent ) ) {
// START would have preceded END, causes SM TIMEOUT
// need to stretch out a prior WRITE or READ to two Command[] elements
done = false ; // reuse it
uint16_t i = 13 ; // start working back until a READ/WRITE has >1 numBytes
cmdIdx = 15 ;
while ( ! done ) {
i2c - > dev - > command [ i + 1 ] . val = i2c - > dev - > command [ i ] . val ; // push it down
if ( ( ( i2c - > dev - > command [ i ] . op_code = = 1 ) | | ( i2c - > dev - > command [ i ] . op_code = = 2 ) ) ) {
2018-07-24 19:43:45 +02:00
/* add a dummy read/write cmd[] with num_bytes set to zero,just a place holder in the cmd[]list
2018-06-27 09:01:06 +02:00
*/
i2c - > dev - > command [ i ] . byte_num = 0 ;
done = true ;
} else {
if ( i > 0 ) {
i - - ;
} else { // unable to stretch, fatal
log_e ( " invalid CMD[] layout Stretch Failed " ) ;
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
i2cDumpCmdQueue ( i2c ) ;
2018-06-27 09:01:06 +02:00
done = true ;
}
}
}
}
if ( cmdIdx = = 15 ) { //need continuation, even if STOP is in 14, it will not matter
// cmd buffer is almost full, Add END as a continuation feature
i2cSetCmd ( i2c , ( cmdIdx ) + + , I2C_CMD_END , 0 , false , false , false ) ;
2018-07-24 19:43:45 +02:00
i2c - > dev - > int_ena . end_detect = 1 ;
i2c - > dev - > int_clr . end_detect = 1 ;
2018-06-27 09:01:06 +02:00
done = true ;
}
if ( ! done ) {
if ( tdq - > ctrl . stopCmdSent ) { // this queue element has been completely added to command[] buffer
qp + + ;
if ( qp < i2c - > queueCount ) {
tdq = & i2c - > dq [ qp ] ;
} else {
done = true ;
}
}
}
} // while(!done)
2018-07-24 19:43:45 +02:00
if ( INTS ) { // don't want to prematurely enable fifo ints until ISR is ready to handle them.
2018-06-27 09:01:06 +02:00
if ( ena_rx ) {
i2c - > dev - > int_ena . rx_fifo_full = 1 ;
}
if ( ena_tx ) {
i2c - > dev - > int_ena . tx_fifo_empty = 1 ;
}
}
}
static void IRAM_ATTR fillTxFifo ( i2c_t * i2c )
{
2018-07-24 19:43:45 +02:00
/*
2018-06-27 09:01:06 +02:00
12 / 01 / 2017 The Fifo ' s are independent , 32 bytes of tx and 32 bytes of Rx .
overlap is not an issue , just keep them full / empty the status_reg . xx_fifo_cnt
tells the truth . And the INT ' s fire correctly
*/
uint16_t a = i2c - > queuePos ; // currently executing dq,
bool full = ! ( i2c - > dev - > status_reg . tx_fifo_cnt < 31 ) ;
uint8_t cnt ;
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
bool rxQueueEncountered = false ;
2018-06-27 09:01:06 +02:00
while ( ( a < i2c - > queueCount ) & & ! full ) {
I2C_DATA_QUEUE_t * tdq = & i2c - > dq [ a ] ;
cnt = 0 ;
// add to address to fifo ctrl.addr already has R/W bit positioned correctly
if ( tdq - > ctrl . addrSent < tdq - > ctrl . addrReq ) { // need to send address bytes
if ( tdq - > ctrl . addrReq = = 2 ) { //10bit
if ( tdq - > ctrl . addrSent = = 0 ) { //10bit highbyte needs sent
if ( ! full ) { // room in fifo
i2c - > dev - > fifo_data . val = ( ( tdq - > ctrl . addr > > 8 ) & 0xFF ) ;
cnt + + ;
tdq - > ctrl . addrSent = 1 ; //10bit highbyte sent
}
}
full = ! ( i2c - > dev - > status_reg . tx_fifo_cnt < 31 ) ;
if ( tdq - > ctrl . addrSent = = 1 ) { //10bit Lowbyte needs sent
if ( ! full ) { // room in fifo
i2c - > dev - > fifo_data . val = tdq - > ctrl . addr & 0xFF ;
cnt + + ;
tdq - > ctrl . addrSent = 2 ; //10bit lowbyte sent
}
}
} else { // 7bit}
if ( tdq - > ctrl . addrSent = = 0 ) { // 7bit Lowbyte needs sent
if ( ! full ) { // room in fifo
i2c - > dev - > fifo_data . val = tdq - > ctrl . addr & 0xFF ;
cnt + + ;
tdq - > ctrl . addrSent = 1 ; // 7bit lowbyte sent
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
# if (ARDUHAL_LOG_LEVEL >= ARDUHAL_LOG_LEVEL_INFO) && (defined ENABLE_I2C_DEBUG_BUFFER)
uint16_t a = 0x8000 | tdq - > ctrl . addr | ( tdq - > ctrl . mode < < 8 ) ;
fifoBuffer [ fifoPos + + ] = a ;
fifoPos % = FIFOMAX ;
# endif
2018-06-27 09:01:06 +02:00
}
}
}
}
// add write data to fifo
if ( tdq - > ctrl . mode = = 0 ) { // write
if ( tdq - > ctrl . addrSent = = tdq - > ctrl . addrReq ) { //address has been sent, is Write Mode!
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
uint32_t moveCnt = 32 - i2c - > dev - > status_reg . tx_fifo_cnt ; // how much room in txFifo?
while ( ( moveCnt > 0 ) & & ( tdq - > position < tdq - > length ) ) {
# if (ARDUHAL_LOG_LEVEL >= ARDUHAL_LOG_LEVEL_INFO) && (defined ENABLE_I2C_DEBUG_BUFFER)
fifoBuffer [ fifoPos + + ] = tdq - > data [ tdq - > position ] ;
fifoPos % = FIFOMAX ;
# endif
2018-06-27 09:01:06 +02:00
i2c - > dev - > fifo_data . val = tdq - > data [ tdq - > position + + ] ;
cnt + + ;
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
moveCnt - - ;
2018-06-27 09:01:06 +02:00
}
}
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
} else { // read Queue Encountered, can't update QueuePos past this point, emptyRxfifo will do it
if ( ! rxQueueEncountered ) {
rxQueueEncountered = true ;
if ( a > i2c - > queuePos ) {
i2c - > queuePos = a ;
}
}
2018-06-27 09:01:06 +02:00
}
2018-07-24 19:43:45 +02:00
# if (ARDUHAL_LOG_LEVEL >= ARDUHAL_LOG_LEVEL_INFO) && (defined ENABLE_I2C_DEBUG_BUFFER)
2018-06-27 09:01:06 +02:00
// update debug buffer tx counts
cnt + = intBuff [ intPos [ i2c - > num ] ] [ 1 ] [ i2c - > num ] > > 16 ;
intBuff [ intPos [ i2c - > num ] ] [ 1 ] [ i2c - > num ] = ( intBuff [ intPos [ i2c - > num ] ] [ 1 ] [ i2c - > num ] & 0xFFFF ) | ( cnt < < 16 ) ;
# endif
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
full = ! ( i2c - > dev - > status_reg . tx_fifo_cnt < 31 ) ;
2018-06-27 09:01:06 +02:00
if ( ! full ) {
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
a + + ; // check next buffer for address tx, or write data
2018-06-27 09:01:06 +02:00
}
}
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
if ( a > = i2c - > queueCount ) { // disable TX IRQ, all tx Data has been queued
2018-07-24 19:43:45 +02:00
i2c - > dev - > int_ena . tx_fifo_empty = 0 ;
2018-06-27 09:01:06 +02:00
}
i2c - > dev - > int_clr . tx_fifo_empty = 1 ;
}
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
2018-06-27 09:01:06 +02:00
static void IRAM_ATTR emptyRxFifo ( i2c_t * i2c )
{
uint32_t d , cnt = 0 , moveCnt ;
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
2018-06-27 09:01:06 +02:00
moveCnt = i2c - > dev - > status_reg . rx_fifo_cnt ; //no need to check the reg until this many are read
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
2018-07-24 19:43:45 +02:00
while ( ( moveCnt > 0 ) & & ( i2c - > queuePos < i2c - > queueCount ) ) { // data to move
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
I2C_DATA_QUEUE_t * tdq = & i2c - > dq [ i2c - > queuePos ] ; //short cut
while ( ( tdq - > position > = tdq - > length ) & & ( i2c - > queuePos < i2c - > queueCount ) ) { // find were to store
i2c - > queuePos + + ;
tdq = & i2c - > dq [ i2c - > queuePos ] ;
}
if ( i2c - > queuePos > = i2c - > queueCount ) { // bad stuff, rx data but no place to put it!
log_e ( " no Storage location for %d " , moveCnt ) ;
// discard
while ( moveCnt > 0 ) {
d = i2c - > dev - > fifo_data . val ;
moveCnt - - ;
cnt + + ;
}
break ;
}
if ( tdq - > ctrl . mode = = 1 ) { // read command
2018-07-24 19:43:45 +02:00
if ( moveCnt > ( tdq - > length - tdq - > position ) ) { //make sure they go in this dq
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
// part of these reads go into the next dq
2018-06-27 09:01:06 +02:00
moveCnt = ( tdq - > length - tdq - > position ) ;
}
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
} else { // error
log_e ( " RxEmpty(%d) call on TxBuffer? dq=%d " , moveCnt , i2c - > queuePos ) ;
// discard
while ( moveCnt > 0 ) {
d = i2c - > dev - > fifo_data . val ;
moveCnt - - ;
cnt + + ;
}
break ;
2018-06-27 09:01:06 +02:00
}
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
while ( moveCnt > 0 ) { // store data
d = i2c - > dev - > fifo_data . val ;
moveCnt - - ;
cnt + + ;
tdq - > data [ tdq - > position + + ] = ( d & 0xFF ) ;
}
moveCnt = i2c - > dev - > status_reg . rx_fifo_cnt ; //any more out there?
2018-07-24 19:43:45 +02:00
}
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
2018-07-24 19:43:45 +02:00
# if (ARDUHAL_LOG_LEVEL >= ARDUHAL_LOG_LEVEL_INFO)&& (defined ENABLE_I2C_DEBUG_BUFFER)
2018-06-27 09:01:06 +02:00
// update Debug rxCount
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
cnt + = ( intBuff [ intPos [ i2c - > num ] ] [ 1 ] [ i2c - > num ] ) & 0xffFF ;
2018-06-27 09:01:06 +02:00
intBuff [ intPos [ i2c - > num ] ] [ 1 ] [ i2c - > num ] = ( intBuff [ intPos [ i2c - > num ] ] [ 1 ] [ i2c - > num ] & 0xFFFF0000 ) | cnt ;
# endif
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
}
2018-06-27 09:01:06 +02:00
static void IRAM_ATTR i2cIsrExit ( i2c_t * i2c , const uint32_t eventCode , bool Fatal )
{
switch ( eventCode ) {
case EVENT_DONE :
i2c - > error = I2C_OK ;
break ;
case EVENT_ERROR_NAK :
i2c - > error = I2C_ADDR_NAK ;
break ;
case EVENT_ERROR_DATA_NAK :
i2c - > error = I2C_DATA_NAK ;
break ;
case EVENT_ERROR_TIMEOUT :
i2c - > error = I2C_TIMEOUT ;
break ;
case EVENT_ERROR_ARBITRATION :
i2c - > error = I2C_ARBITRATION ;
break ;
default :
i2c - > error = I2C_ERROR ;
}
uint32_t exitCode = EVENT_DONE | eventCode | ( Fatal ? EVENT_ERROR : 0 ) ;
2018-07-24 19:43:45 +02:00
if ( ( i2c - > queuePos < i2c - > queueCount ) & & ( i2c - > dq [ i2c - > queuePos ] . ctrl . mode = = 1 ) ) {
2018-06-27 09:01:06 +02:00
emptyRxFifo ( i2c ) ; // grab last few characters
}
2018-07-24 19:43:45 +02:00
i2c - > dev - > int_ena . val = 0 ; // shut down interrupts
2018-07-12 15:18:26 +02:00
i2c - > dev - > int_clr . val = 0x1FFF ;
2018-06-27 09:01:06 +02:00
i2c - > stage = I2C_DONE ;
i2c - > exitCode = exitCode ; //true eventcode
portBASE_TYPE HPTaskAwoken = pdFALSE , xResult ;
// try to notify Dispatch we are done,
2018-07-24 19:43:45 +02:00
// else the 50ms time out will recover the APP, just a little slower
2018-06-27 09:01:06 +02:00
HPTaskAwoken = pdFALSE ;
xResult = xEventGroupSetBitsFromISR ( i2c - > i2c_event , exitCode , & HPTaskAwoken ) ;
if ( xResult = = pdPASS ) {
if ( HPTaskAwoken = = pdTRUE ) {
portYIELD_FROM_ISR ( ) ;
// log_e("Yield to Higher");
}
}
}
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
static void IRAM_ATTR i2c_update_error_byte_cnt ( i2c_t * i2c )
{
2018-07-24 19:43:45 +02:00
/* i2c_update_error_byte_cnt 07/18/2018
Only called after an error has occurred , so , most of the time this function is never used .
This function obliterates the need to interrupt monitor each byte transferred , at high bitrates
the byte interrupts were overwhelming the OS . Spurious Interrupts were being generated .
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
it updates errorByteCnt , errorQueue .
2018-07-24 19:43:45 +02:00
*/
uint16_t a = 0 ; // start at top of DQ, count how many bytes added to tx fifo, and received from rx_fifo.
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
int16_t bc = 0 ;
2018-07-24 19:43:45 +02:00
I2C_DATA_QUEUE_t * tdq ;
i2c - > errorByteCnt = 0 ;
while ( a < i2c - > queueCount ) { // add up all bytes loaded into fifo's
tdq = & i2c - > dq [ a ] ;
i2c - > errorByteCnt + = tdq - > ctrl . addrSent ;
i2c - > errorByteCnt + = tdq - > position ;
a + + ;
}
// now errorByteCnt contains total bytes moved into and out of FIFO's
// but, there may still be bytes waiting in Fifo's
i2c - > errorByteCnt - = i2c - > dev - > status_reg . tx_fifo_cnt ; // waiting to go out;
i2c - > errorByteCnt + = i2c - > dev - > status_reg . rx_fifo_cnt ; // already received
// now walk thru DQ again, find which byte is 'current'
bc = i2c - > errorByteCnt ;
i2c - > errorQueue = 0 ;
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
while ( i2c - > errorQueue < i2c - > queueCount ) {
2018-07-24 19:43:45 +02:00
tdq = & i2c - > dq [ i2c - > errorQueue ] ;
if ( bc > 0 ) { // not found yet
if ( tdq - > ctrl . addrSent > = bc ) { // in address
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
bc = - 1 ; // in address
break ;
2018-07-24 19:43:45 +02:00
} else {
bc - = tdq - > ctrl . addrSent ;
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
if ( tdq - > length > bc ) { // data nak
break ;
2018-07-24 19:43:45 +02:00
} else { // count down
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
bc - = tdq - > length ;
2018-07-24 19:43:45 +02:00
}
}
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
} else break ;
2018-07-24 19:43:45 +02:00
i2c - > errorQueue + + ;
}
i2c - > errorByteCnt = bc ;
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
}
2018-06-27 09:01:06 +02:00
static void IRAM_ATTR i2c_isr_handler_default ( void * arg )
{
i2c_t * p_i2c = ( i2c_t * ) arg ; // recover data
2018-07-12 15:18:26 +02:00
uint32_t activeInt = p_i2c - > dev - > int_status . val & 0x7FF ;
2018-06-27 09:01:06 +02:00
2018-07-12 15:18:26 +02:00
if ( p_i2c - > stage = = I2C_DONE ) { //get Out, can't service, not configured
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
p_i2c - > dev - > int_ena . val = 0 ;
p_i2c - > dev - > int_clr . val = 0x1FFF ;
2018-07-24 22:06:50 +02:00
# if ARDUHAL_LOG_LEVEL >= ARDUHAL_LOG_LEVEL_VERBOSE
2018-07-24 19:43:45 +02:00
uint32_t raw = p_i2c - > dev - > int_raw . val ;
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
log_w ( " eject raw=%p, int=%p " , raw , activeInt ) ;
2018-07-24 22:06:50 +02:00
# endif
2018-06-27 09:01:06 +02:00
return ;
}
while ( activeInt ! = 0 ) { // Ordering of 'if(activeInt)' statements is important, don't change
2018-07-24 19:43:45 +02:00
# if (ARDUHAL_LOG_LEVEL >= ARDUHAL_LOG_LEVEL_INFO) && (defined ENABLE_I2C_DEBUG_BUFFER)
2018-06-27 09:01:06 +02:00
if ( activeInt = = ( intBuff [ intPos [ p_i2c - > num ] ] [ 0 ] [ p_i2c - > num ] & 0x1fff ) ) {
intBuff [ intPos [ p_i2c - > num ] ] [ 0 ] [ p_i2c - > num ] = ( ( ( intBuff [ intPos [ p_i2c - > num ] ] [ 0 ] [ p_i2c - > num ] > > 16 ) + 1 ) < < 16 ) | activeInt ;
} else {
intPos [ p_i2c - > num ] + + ;
intPos [ p_i2c - > num ] % = INTBUFFMAX ;
intBuff [ intPos [ p_i2c - > num ] ] [ 0 ] [ p_i2c - > num ] = ( 1 < < 16 ) | activeInt ;
intBuff [ intPos [ p_i2c - > num ] ] [ 1 ] [ p_i2c - > num ] = 0 ;
}
intBuff [ intPos [ p_i2c - > num ] ] [ 2 ] [ p_i2c - > num ] = xTaskGetTickCountFromISR ( ) ; // when IRQ fired
# endif
2018-07-12 15:18:26 +02:00
2018-06-27 09:01:06 +02:00
if ( activeInt & I2C_TRANS_START_INT_ST_M ) {
if ( p_i2c - > stage = = I2C_STARTUP ) {
p_i2c - > stage = I2C_RUNNING ;
}
activeInt & = ~ I2C_TRANS_START_INT_ST_M ;
p_i2c - > dev - > int_ena . trans_start = 1 ; // already enabled? why Again?
p_i2c - > dev - > int_clr . trans_start = 1 ; // so that will trigger after next 'END'
}
if ( activeInt & I2C_TXFIFO_EMPTY_INT_ST ) { //should this be before Trans_start?
fillTxFifo ( p_i2c ) ; //fillTxFifo will enable/disable/clear interrupt
activeInt & = ~ I2C_TXFIFO_EMPTY_INT_ST ;
}
if ( activeInt & I2C_RXFIFO_FULL_INT_ST ) {
emptyRxFifo ( p_i2c ) ;
p_i2c - > dev - > int_clr . rx_fifo_full = 1 ;
p_i2c - > dev - > int_ena . rx_fifo_full = 1 ; //why?
activeInt & = ~ I2C_RXFIFO_FULL_INT_ST ;
}
if ( activeInt & I2C_ACK_ERR_INT_ST_M ) { //fatal error, abort i2c service
if ( p_i2c - > mode = = I2C_MASTER ) {
2018-07-24 19:43:45 +02:00
i2c_update_error_byte_cnt ( p_i2c ) ; // calc which byte caused ack Error, check if address or data
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
if ( p_i2c - > errorByteCnt < 0 ) { // address
2018-06-27 09:01:06 +02:00
i2cIsrExit ( p_i2c , EVENT_ERROR_NAK , true ) ;
} else {
2018-07-24 19:43:45 +02:00
i2cIsrExit ( p_i2c , EVENT_ERROR_DATA_NAK , true ) ; //data
2018-06-27 09:01:06 +02:00
}
}
return ;
}
if ( activeInt & I2C_TIME_OUT_INT_ST_M ) {
// let Gross timeout occur, Slave may release SCL before the configured timeout expires
// the Statemachine only has a 13.1ms max timout, some Devices >500ms
p_i2c - > dev - > int_clr . time_out = 1 ;
activeInt & = ~ I2C_TIME_OUT_INT_ST ;
2018-07-24 19:43:45 +02:00
// since a timeout occurred, capture the rxFifo data
emptyRxFifo ( p_i2c ) ;
p_i2c - > dev - > int_clr . rx_fifo_full = 1 ;
p_i2c - > dev - > int_ena . rx_fifo_full = 1 ; //why?
2018-06-27 09:01:06 +02:00
}
if ( activeInt & I2C_TRANS_COMPLETE_INT_ST_M ) {
2018-07-24 19:43:45 +02:00
p_i2c - > dev - > int_clr . trans_complete = 1 ;
2018-06-27 09:01:06 +02:00
i2cIsrExit ( p_i2c , EVENT_DONE , false ) ;
return ; // no more work to do
/*
// how does slave mode act?
if ( p_i2c - > mode = = I2C_SLAVE ) { // STOP detected
// empty fifo
// dispatch callback
*/
}
if ( activeInt & I2C_ARBITRATION_LOST_INT_ST_M ) { //fatal
i2cIsrExit ( p_i2c , EVENT_ERROR_ARBITRATION , true ) ;
return ; // no more work to do
}
if ( activeInt & I2C_SLAVE_TRAN_COMP_INT_ST_M ) {
p_i2c - > dev - > int_clr . slave_tran_comp = 1 ;
// need to complete this !
}
if ( activeInt & I2C_END_DETECT_INT_ST_M ) {
p_i2c - > dev - > int_ena . end_detect = 0 ;
p_i2c - > dev - > int_clr . end_detect = 1 ;
p_i2c - > dev - > ctr . trans_start = 0 ;
fillCmdQueue ( p_i2c , true ) ; // enable interrupts
p_i2c - > dev - > ctr . trans_start = 1 ; // go for it
activeInt & = ~ I2C_END_DETECT_INT_ST_M ;
}
if ( activeInt ) { // clear unhandled if possible? What about Disabling interrupt?
p_i2c - > dev - > int_clr . val = activeInt ;
log_e ( " unknown int=%x " , activeInt ) ;
// disable unhandled IRQ,
p_i2c - > dev - > int_ena . val = p_i2c - > dev - > int_ena . val & ( ~ activeInt ) ;
}
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
// activeInt = p_i2c->dev->int_status.val; // start all over if another interrupt happened
// 01AUG2018 if another interrupt happened, the OS will schedule another interrupt
// this is the source of spurious interrupts
2018-06-27 09:01:06 +02:00
}
}
/* Stickbreaker added for ISR 11/2017
functional with Silicon date = 0x16042000
*/
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
static i2c_err_t i2cAddQueue ( i2c_t * i2c , uint8_t mode , uint16_t i2cDeviceAddr , uint8_t * dataPtr , uint16_t dataLen , bool sendStop , bool dataOnly , EventGroupHandle_t event )
2018-06-27 09:01:06 +02:00
{
// need to grab a MUTEX for exclusive Queue,
2018-07-24 19:43:45 +02:00
// what about if ISR is running?
2018-06-27 09:01:06 +02:00
if ( i2c = = NULL ) {
return I2C_ERROR_DEV ;
}
I2C_DATA_QUEUE_t dqx ;
dqx . data = dataPtr ;
dqx . length = dataLen ;
dqx . position = 0 ;
dqx . cmdBytesNeeded = dataLen ;
dqx . ctrl . val = 0 ;
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
if ( dataOnly ) {
/* special case to add a queue data only element.
START and devAddr will not be sent , this dq element can have a STOP .
allows multiple buffers to be used for one transaction .
sequence : normal transaction ( sendStop = = false ) , [ dataonly ( sendStop = = false ) ] , dataOnly ( sendStop = = true )
* * * Currently only works with WRITE , final byte NAK an READ will cause a fail between dq buffer elements . ( in progress 30 JUL2018 )
*/
dqx . ctrl . startCmdSent = 1 ; // mark as already sent
dqx . ctrl . addrCmdSent = 1 ;
} else {
dqx . ctrl . addrReq = ( ( i2cDeviceAddr & 0xFC00 ) = = 0x7800 ) ? 2 : 1 ; // 10bit or 7bit address
}
2018-06-27 09:01:06 +02:00
dqx . ctrl . addr = i2cDeviceAddr ;
dqx . ctrl . mode = mode ;
dqx . ctrl . stop = sendStop ;
dqx . queueEvent = event ;
if ( event ) { // an eventGroup exist, so, initialize it
xEventGroupClearBits ( event , EVENT_MASK ) ; // all of them
}
if ( i2c - > dq ! = NULL ) { // expand
//log_i("expand");
I2C_DATA_QUEUE_t * tq = ( I2C_DATA_QUEUE_t * ) realloc ( i2c - > dq , sizeof ( I2C_DATA_QUEUE_t ) * ( i2c - > queueCount + 1 ) ) ;
if ( tq ! = NULL ) { // ok
i2c - > dq = tq ;
memmove ( & i2c - > dq [ i2c - > queueCount + + ] , & dqx , sizeof ( I2C_DATA_QUEUE_t ) ) ;
} else { // bad stuff, unable to allocate more memory!
log_e ( " realloc Failure " ) ;
return I2C_ERROR_MEMORY ;
}
} else { // first Time
//log_i("new");
i2c - > queueCount = 0 ;
i2c - > dq = ( I2C_DATA_QUEUE_t * ) malloc ( sizeof ( I2C_DATA_QUEUE_t ) ) ;
if ( i2c - > dq ! = NULL ) {
memmove ( & i2c - > dq [ i2c - > queueCount + + ] , & dqx , sizeof ( I2C_DATA_QUEUE_t ) ) ;
} else {
log_e ( " malloc failure " ) ;
return I2C_ERROR_MEMORY ;
}
}
return I2C_ERROR_OK ;
}
i2c_err_t i2cAddQueueWrite ( i2c_t * i2c , uint16_t i2cDeviceAddr , uint8_t * dataPtr , uint16_t dataLen , bool sendStop , EventGroupHandle_t event )
{
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
return i2cAddQueue ( i2c , 0 , i2cDeviceAddr , dataPtr , dataLen , sendStop , false , event ) ;
2018-06-27 09:01:06 +02:00
}
i2c_err_t i2cAddQueueRead ( i2c_t * i2c , uint16_t i2cDeviceAddr , uint8_t * dataPtr , uint16_t dataLen , bool sendStop , EventGroupHandle_t event )
{
//10bit read is kind of weird, first you do a 0byte Write with 10bit
// address, then a ReSTART then a 7bit Read using the the upper 7bit +
// readBit.
// this might cause an internal register pointer problem with 10bit
2018-07-24 19:43:45 +02:00
// devices, But, Don't have any to test against.
2018-06-27 09:01:06 +02:00
// this is the Industry Standard specification.
if ( ( i2cDeviceAddr & 0xFC00 ) = = 0x7800 ) { // ten bit read
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
i2c_err_t err = i2cAddQueue ( i2c , 0 , i2cDeviceAddr , NULL , 0 , false , false , event ) ;
2018-06-27 09:01:06 +02:00
if ( err = = I2C_ERROR_OK ) {
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
return i2cAddQueue ( i2c , 1 , ( i2cDeviceAddr > > 8 ) , dataPtr , dataLen , sendStop , false , event ) ;
2018-06-27 09:01:06 +02:00
} else {
return err ;
}
}
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
return i2cAddQueue ( i2c , 1 , i2cDeviceAddr , dataPtr , dataLen , sendStop , false , event ) ;
2018-06-27 09:01:06 +02:00
}
i2c_err_t i2cProcQueue ( i2c_t * i2c , uint32_t * readCount , uint16_t timeOutMillis )
{
/* do the hard stuff here
install ISR if necessary
setup EventGroup
handle bus busy ?
*/
//log_e("procQueue i2c=%p",&i2c);
if ( readCount ) { //total reads accomplished in all queue elements
* readCount = 0 ;
}
if ( i2c = = NULL ) {
return I2C_ERROR_DEV ;
}
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
if ( i2c - > debugFlags & 0xff000000 ) i2cTriggerDumps ( i2c , ( i2c - > debugFlags > > 24 ) , " before ProcQueue " ) ;
2018-06-27 09:01:06 +02:00
if ( i2c - > dev - > status_reg . bus_busy ) { // return error, let TwoWire() handle resetting the hardware.
/* if multi master then this if should be changed to this 03/12/2018
if ( multiMaster ) { // try to let the bus clear by its self
uint32_t timeOutTick = millis ( ) ;
while ( ( i2c - > dev - > status_reg . bus_busy ) & & ( millis ( ) - timeOutTick < timeOutMillis ( ) ) ) {
delay ( 2 ) ; // allow task switch
}
}
if ( i2c - > dev - > status_reg . bus_busy ) { // still busy, so die
*/
log_i ( " Bus busy, reinit " ) ;
return I2C_ERROR_BUSY ;
}
I2C_MUTEX_LOCK ( ) ;
/* what about co-existence with SLAVE mode?
Should I check if a slaveMode xfer is in progress and hang
until it completes ?
if i2c - > stage = = I2C_RUNNING or I2C_SLAVE_ACTIVE
*/
i2c - > stage = I2C_DONE ; // until ready
2018-07-24 19:43:45 +02:00
# if (ARDUHAL_LOG_LEVEL >= ARDUHAL_LOG_LEVEL_INFO) && (defined ENABLE_I2C_DEBUG_BUFFER)
2018-06-27 09:01:06 +02:00
for ( uint16_t i = 0 ; i < INTBUFFMAX ; i + + ) {
intBuff [ i ] [ 0 ] [ i2c - > num ] = 0 ;
intBuff [ i ] [ 1 ] [ i2c - > num ] = 0 ;
intBuff [ i ] [ 2 ] [ i2c - > num ] = 0 ;
}
intPos [ i2c - > num ] = 0 ;
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
fifoPos = 0 ;
memset ( fifoBuffer , 0 , FIFOMAX ) ;
2018-06-27 09:01:06 +02:00
# endif
// EventGroup is used to signal transmission completion from ISR
// not always reliable. Sometimes, the FreeRTOS scheduler is maxed out and refuses request
// if that happens, this call hangs until the timeout period expires, then it continues.
if ( ! i2c - > i2c_event ) {
i2c - > i2c_event = xEventGroupCreate ( ) ;
}
if ( i2c - > i2c_event ) {
xEventGroupClearBits ( i2c - > i2c_event , 0xFF ) ;
} else { // failed to create EventGroup
log_e ( " eventCreate failed=%p " , i2c - > i2c_event ) ;
I2C_MUTEX_UNLOCK ( ) ;
return I2C_ERROR_MEMORY ;
}
i2c_err_t reason = I2C_ERROR_OK ;
i2c - > mode = I2C_MASTER ;
i2c - > dev - > ctr . trans_start = 0 ; // Pause Machine
i2c - > dev - > timeout . tout = 0xFFFFF ; // max 13ms
2018-07-24 19:43:45 +02:00
i2c - > dev - > int_clr . val = 0x1FFF ; // kill them All!
2018-07-12 15:18:26 +02:00
2018-06-27 09:01:06 +02:00
i2c - > dev - > ctr . ms_mode = 1 ; // master!
i2c - > queuePos = 0 ;
2018-07-24 19:43:45 +02:00
i2c - > errorByteCnt = 0 ;
i2c - > errorQueue = 0 ;
2018-06-27 09:01:06 +02:00
uint32_t totalBytes = 0 ; // total number of bytes to be Moved!
// convert address field to required I2C format
while ( i2c - > queuePos < i2c - > queueCount ) { // need to push these address modes upstream, to AddQueue
I2C_DATA_QUEUE_t * tdq = & i2c - > dq [ i2c - > queuePos + + ] ;
uint16_t taddr = 0 ;
if ( tdq - > ctrl . addrReq = = 2 ) { // 10bit address
taddr = ( ( tdq - > ctrl . addr > > 7 ) & 0xFE )
| tdq - > ctrl . mode ;
2018-11-28 00:35:43 +01:00
taddr = ( taddr < < 8 ) | ( tdq - > ctrl . addr & 0xFF ) ;
2018-06-27 09:01:06 +02:00
} else { // 7bit address
taddr = ( ( tdq - > ctrl . addr < < 1 ) & 0xFE )
| tdq - > ctrl . mode ;
}
tdq - > ctrl . addr = taddr ; // all fixed with R/W bit
2018-07-24 19:43:45 +02:00
totalBytes + = tdq - > length + tdq - > ctrl . addrReq ; // total number of byte to be moved!
2018-06-27 09:01:06 +02:00
}
i2c - > queuePos = 0 ;
fillCmdQueue ( i2c , false ) ; // don't enable Tx/RX irq's
// start adding command[], END irq will keep it full
//Data Fifo will be filled after trans_start is issued
i2c - > exitCode = 0 ;
2018-07-24 19:43:45 +02:00
I2C_FIFO_CONF_t f ;
f . val = i2c - > dev - > fifo_conf . val ;
f . rx_fifo_rst = 1 ; // fifo in reset
f . tx_fifo_rst = 1 ; // fifo in reset
f . nonfifo_en = 0 ; // use fifo mode
f . nonfifo_tx_thres = 31 ;
// need to adjust threshold based on I2C clock rate, at 100k, 30 usually works,
// sometimes the emptyRx() actually moves 31 bytes
// it hasn't overflowed yet, I cannot tell if the new byte is added while
// emptyRX() is executing or before?
// let i2cSetFrequency() set thrhds
// f.rx_fifo_full_thrhd = 30; // 30 bytes before INT is issued
// f.tx_fifo_empty_thrhd = 0;
f . fifo_addr_cfg_en = 0 ; // no directed access
i2c - > dev - > fifo_conf . val = f . val ; // post them all
f . rx_fifo_rst = 0 ; // release fifo
f . tx_fifo_rst = 0 ;
i2c - > dev - > fifo_conf . val = f . val ; // post them all
2018-06-27 09:01:06 +02:00
i2c - > stage = I2C_STARTUP ; // everything configured, now start the I2C StateMachine, and
// As soon as interrupts are enabled, the ISR will start handling them.
// it should receive a TXFIFO_EMPTY immediately, even before it
// receives the TRANS_START
2018-07-24 19:43:45 +02:00
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
uint32_t interruptsEnabled =
2018-06-27 09:01:06 +02:00
I2C_ACK_ERR_INT_ENA | // (BIT(10)) Causes Fatal Error Exit
I2C_TRANS_START_INT_ENA | // (BIT(9)) Triggered by trans_start=1, initial,END
I2C_TIME_OUT_INT_ENA | //(BIT(8)) Trigger by SLAVE SCL stretching, NOT an ERROR
I2C_TRANS_COMPLETE_INT_ENA | // (BIT(7)) triggered by STOP, successful exit
I2C_ARBITRATION_LOST_INT_ENA | // (BIT(5)) cause fatal error exit
I2C_SLAVE_TRAN_COMP_INT_ENA | // (BIT(4)) unhandled
I2C_END_DETECT_INT_ENA | // (BIT(3)) refills cmd[] list
I2C_RXFIFO_OVF_INT_ENA | //(BIT(2)) unhandled
I2C_TXFIFO_EMPTY_INT_ENA | // (BIT(1)) triggers fillTxFifo()
I2C_RXFIFO_FULL_INT_ENA ; // (BIT(0)) trigger emptyRxFifo()
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
i2c - > dev - > int_ena . val = interruptsEnabled ;
2018-06-27 09:01:06 +02:00
if ( ! i2c - > intr_handle ) { // create ISR for either peripheral
// log_i("create ISR %d",i2c->num);
uint32_t ret = 0 ;
2018-09-21 08:40:01 +02:00
uint32_t flags = ESP_INTR_FLAG_IRAM | //< ISR can be called if cache is disabled
ESP_INTR_FLAG_LOWMED | //< Low and medium prio interrupts. These can be handled in C.
ESP_INTR_FLAG_SHARED ; //< Reduce resource requirements, Share interrupts
2018-06-27 09:01:06 +02:00
if ( i2c - > num ) {
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
ret = esp_intr_alloc_intrstatus ( ETS_I2C_EXT1_INTR_SOURCE , flags , ( uint32_t ) & i2c - > dev - > int_status . val , interruptsEnabled , & i2c_isr_handler_default , i2c , & i2c - > intr_handle ) ;
2018-06-27 09:01:06 +02:00
} else {
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
ret = esp_intr_alloc_intrstatus ( ETS_I2C_EXT0_INTR_SOURCE , flags , ( uint32_t ) & i2c - > dev - > int_status . val , interruptsEnabled , & i2c_isr_handler_default , i2c , & i2c - > intr_handle ) ;
2018-06-27 09:01:06 +02:00
}
if ( ret ! = ESP_OK ) {
log_e ( " install interrupt handler Failed=%d " , ret ) ;
I2C_MUTEX_UNLOCK ( ) ;
return I2C_ERROR_MEMORY ;
}
2019-01-09 10:07:54 +01:00
if ( ! addApbChangeCallback ( i2c , i2cApbChangeCallback ) ) {
log_e ( " install apb Callback failed " ) ;
I2C_MUTEX_UNLOCK ( ) ;
return I2C_ERROR_DEV ;
}
2018-06-27 09:01:06 +02:00
}
//hang until it completes.
2018-07-24 19:43:45 +02:00
// how many ticks should it take to transfer totalBytes through the I2C hardware,
// add user supplied timeOutMillis to Calculated Value
2018-06-27 09:01:06 +02:00
portTickType ticksTimeOut = ( ( totalBytes * 10 * 1000 ) / ( i2cGetFrequency ( i2c ) ) + timeOutMillis ) / portTICK_PERIOD_MS ;
i2c - > dev - > ctr . trans_start = 1 ; // go for it
# if ARDUHAL_LOG_LEVEL >= ARDUHAL_LOG_LEVEL_ERROR
portTickType tBefore = xTaskGetTickCount ( ) ;
# endif
2018-07-24 19:43:45 +02:00
// wait for ISR to complete the transfer, or until timeOut in case of bus fault, hardware problem
2018-06-27 09:01:06 +02:00
uint32_t eBits = xEventGroupWaitBits ( i2c - > i2c_event , EVENT_DONE , pdFALSE , pdTRUE , ticksTimeOut ) ;
# if ARDUHAL_LOG_LEVEL >= ARDUHAL_LOG_LEVEL_ERROR
portTickType tAfter = xTaskGetTickCount ( ) ;
# endif
2018-07-24 19:43:45 +02:00
2018-06-27 09:01:06 +02:00
// if xEventGroupSetBitsFromISR() failed, the ISR could have succeeded but never been
// able to mark the success
if ( i2c - > exitCode ! = eBits ) { // try to recover from O/S failure
// log_e("EventGroup Failed:%p!=%p",eBits,i2c->exitCode);
eBits = i2c - > exitCode ;
}
2018-07-24 19:43:45 +02:00
if ( ( eBits & EVENT_ERROR ) | | ( ! ( eBits & EVENT_DONE ) ) ) { // need accurate errorByteCnt for debug
i2c_update_error_byte_cnt ( i2c ) ;
}
2018-06-27 09:01:06 +02:00
if ( ! ( eBits = = EVENT_DONE ) & & ( eBits & ~ ( EVENT_ERROR_NAK | EVENT_ERROR_DATA_NAK | EVENT_ERROR | EVENT_DONE ) ) ) { // not only Done, therefore error, exclude ADDR NAK, DATA_NAK
# if ARDUHAL_LOG_LEVEL >= ARDUHAL_LOG_LEVEL_INFO
i2cDumpI2c ( i2c ) ;
i2cDumpInts ( i2c - > num ) ;
# endif
}
if ( eBits & EVENT_DONE ) { // no gross timeout
switch ( i2c - > error ) {
case I2C_OK :
reason = I2C_ERROR_OK ;
break ;
case I2C_ERROR :
reason = I2C_ERROR_DEV ;
break ;
case I2C_ADDR_NAK :
reason = I2C_ERROR_ACK ;
break ;
case I2C_DATA_NAK :
reason = I2C_ERROR_ACK ;
break ;
case I2C_ARBITRATION :
reason = I2C_ERROR_BUS ;
break ;
case I2C_TIMEOUT :
reason = I2C_ERROR_TIMEOUT ;
break ;
default :
reason = I2C_ERROR_DEV ;
}
} else { // GROSS timeout, shutdown ISR , report Timeout
i2c - > stage = I2C_DONE ;
i2c - > dev - > int_ena . val = 0 ;
i2c - > dev - > int_clr . val = 0x1FFF ;
2018-07-24 19:43:45 +02:00
i2c_update_error_byte_cnt ( i2c ) ;
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
if ( ( i2c - > errorByteCnt = = 0 ) & & ( i2c - > errorQueue = = 0 ) ) { // Bus Busy no bytes Moved
2018-06-27 09:01:06 +02:00
reason = I2C_ERROR_BUSY ;
eBits = eBits | EVENT_ERROR_BUS_BUSY | EVENT_ERROR | EVENT_DONE ;
2018-07-24 19:43:45 +02:00
# if ARDUHAL_LOG_LEVEL >= ARDUHAL_LOG_LEVEL_DEBUG
log_d ( " Busy Timeout start=0x%x, end=0x%x, =%d, max=%d error=%d " , tBefore , tAfter , ( tAfter - tBefore ) , ticksTimeOut , i2c - > error ) ;
2018-06-27 09:01:06 +02:00
i2cDumpI2c ( i2c ) ;
i2cDumpInts ( i2c - > num ) ;
# endif
} else { // just a timeout, some data made it out or in.
reason = I2C_ERROR_TIMEOUT ;
eBits = eBits | EVENT_ERROR_TIMEOUT | EVENT_ERROR | EVENT_DONE ;
2018-07-24 19:43:45 +02:00
# if ARDUHAL_LOG_LEVEL >= ARDUHAL_LOG_LEVEL_DEBUG
log_d ( " Gross Timeout Dead start=0x%x, end=0x%x, =%d, max=%d error=%d " , tBefore , tAfter , ( tAfter - tBefore ) , ticksTimeOut , i2c - > error ) ;
2018-06-27 09:01:06 +02:00
i2cDumpI2c ( i2c ) ;
i2cDumpInts ( i2c - > num ) ;
# endif
}
}
2018-07-24 19:43:45 +02:00
/* offloading all EventGroups to dispatch, EventGroups in ISR is not always successful
11 / 20 / 2017
if error , need to trigger all succeeding dataQueue events with the EVENT_ERROR_PREV
07 / 22 / 2018
Need to use the queueEvent value to identify transaction blocks , if an error occurs ,
all subsequent queue items with the same queueEvent value will receive the EVENT_ERROR_PREV .
But , ProcQue should re - queue queue items that have a different queueEvent value ( different transaction )
This change will support multi - thread i2c usage . Use the queueEvent as the transaction event
identifier .
*/
uint32_t b = 0 ;
2018-06-27 09:01:06 +02:00
while ( b < i2c - > queueCount ) {
if ( i2c - > dq [ b ] . ctrl . mode = = 1 & & readCount ) {
* readCount + = i2c - > dq [ b ] . position ; // number of data bytes received
}
if ( b < i2c - > queuePos ) { // before any error
if ( i2c - > dq [ b ] . queueEvent ) { // this data queue element has an EventGroup
xEventGroupSetBits ( i2c - > dq [ b ] . queueEvent , EVENT_DONE ) ;
}
} else if ( b = = i2c - > queuePos ) { // last processed queue
if ( i2c - > dq [ b ] . queueEvent ) { // this data queue element has an EventGroup
xEventGroupSetBits ( i2c - > dq [ b ] . queueEvent , eBits ) ;
}
} else { // never processed queues
if ( i2c - > dq [ b ] . queueEvent ) { // this data queue element has an EventGroup
xEventGroupSetBits ( i2c - > dq [ b ] . queueEvent , eBits | EVENT_ERROR_PREV ) ;
}
}
b + + ;
}
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
if ( i2c - > debugFlags & 0x00ff0000 ) i2cTriggerDumps ( i2c , ( i2c - > debugFlags > > 16 ) , " after ProcQueue " ) ;
2018-06-27 09:01:06 +02:00
I2C_MUTEX_UNLOCK ( ) ;
return reason ;
}
static void i2cReleaseISR ( i2c_t * i2c )
{
if ( i2c - > intr_handle ) {
esp_intr_free ( i2c - > intr_handle ) ;
i2c - > intr_handle = NULL ;
2019-01-09 10:07:54 +01:00
if ( ! removeApbChangeCallback ( i2c , i2cApbChangeCallback ) ) {
log_e ( " unable to release apbCallback " ) ;
}
2018-06-27 09:01:06 +02:00
}
}
static bool i2cCheckLineState ( int8_t sda , int8_t scl ) {
2018-08-18 08:50:59 +02:00
if ( sda < 0 | | scl < 0 ) {
return false ; //return false since there is nothing to do
2018-06-27 09:01:06 +02:00
}
2018-08-18 08:50:59 +02:00
// if the bus is not 'clear' try the cycling SCL until SDA goes High or 9 cycles
2018-06-27 09:01:06 +02:00
digitalWrite ( sda , HIGH ) ;
digitalWrite ( scl , HIGH ) ;
2018-08-18 08:50:59 +02:00
pinMode ( sda , PULLUP | OPEN_DRAIN | INPUT ) ;
pinMode ( scl , PULLUP | OPEN_DRAIN | OUTPUT ) ;
2018-06-27 09:01:06 +02:00
if ( ! digitalRead ( sda ) | | ! digitalRead ( scl ) ) { // bus in busy state
2018-08-18 08:50:59 +02:00
log_w ( " invalid state sda(%d)=%d, scl(%d)=%d " , sda , digitalRead ( sda ) , scl , digitalRead ( scl ) ) ;
2018-06-27 09:01:06 +02:00
digitalWrite ( scl , HIGH ) ;
for ( uint8_t a = 0 ; a < 9 ; a + + ) {
delayMicroseconds ( 5 ) ;
digitalWrite ( scl , LOW ) ;
delayMicroseconds ( 5 ) ;
digitalWrite ( scl , HIGH ) ;
2018-08-18 08:50:59 +02:00
if ( digitalRead ( sda ) ) { // bus recovered
log_d ( " Recovered after %d Cycles " , a + 1 ) ;
break ;
}
2018-06-27 09:01:06 +02:00
}
}
if ( ! digitalRead ( sda ) | | ! digitalRead ( scl ) ) { // bus in busy state
2018-08-18 08:50:59 +02:00
log_e ( " Bus Invalid State, TwoWire() Can't init sda=%d, scl=%d " , digitalRead ( sda ) , digitalRead ( scl ) ) ;
2018-06-27 09:01:06 +02:00
return false ; // bus is busy
}
return true ;
}
2016-10-11 14:11:51 +02:00
i2c_err_t i2cAttachSCL ( i2c_t * i2c , int8_t scl )
2016-10-06 13:21:30 +02:00
{
2018-06-27 09:01:06 +02:00
if ( i2c = = NULL ) {
2016-10-11 14:11:51 +02:00
return I2C_ERROR_DEV ;
2016-10-11 13:20:02 +02:00
}
2018-06-27 09:01:06 +02:00
digitalWrite ( scl , HIGH ) ;
pinMode ( scl , OPEN_DRAIN | PULLUP | INPUT | OUTPUT ) ;
2016-10-06 13:21:30 +02:00
pinMatrixOutAttach ( scl , I2C_SCL_IDX ( i2c - > num ) , false , false ) ;
pinMatrixInAttach ( scl , I2C_SCL_IDX ( i2c - > num ) , false ) ;
2016-10-11 14:11:51 +02:00
return I2C_ERROR_OK ;
2016-10-06 13:21:30 +02:00
}
2016-10-11 14:11:51 +02:00
i2c_err_t i2cDetachSCL ( i2c_t * i2c , int8_t scl )
2016-10-06 13:21:30 +02:00
{
2018-06-27 09:01:06 +02:00
if ( i2c = = NULL ) {
2016-10-11 14:11:51 +02:00
return I2C_ERROR_DEV ;
2016-10-11 13:20:02 +02:00
}
2016-10-06 13:21:30 +02:00
pinMatrixOutDetach ( scl , false , false ) ;
pinMatrixInDetach ( I2C_SCL_IDX ( i2c - > num ) , false , false ) ;
2018-06-27 09:01:06 +02:00
pinMode ( scl , INPUT | PULLUP ) ;
2016-10-11 14:11:51 +02:00
return I2C_ERROR_OK ;
2016-10-06 13:21:30 +02:00
}
2016-10-11 14:11:51 +02:00
i2c_err_t i2cAttachSDA ( i2c_t * i2c , int8_t sda )
2016-10-06 13:21:30 +02:00
{
2018-06-27 09:01:06 +02:00
if ( i2c = = NULL ) {
2016-10-11 14:11:51 +02:00
return I2C_ERROR_DEV ;
2016-10-11 13:20:02 +02:00
}
2018-06-27 09:01:06 +02:00
digitalWrite ( sda , HIGH ) ;
pinMode ( sda , OPEN_DRAIN | PULLUP | INPUT | OUTPUT ) ;
2016-10-06 13:21:30 +02:00
pinMatrixOutAttach ( sda , I2C_SDA_IDX ( i2c - > num ) , false , false ) ;
pinMatrixInAttach ( sda , I2C_SDA_IDX ( i2c - > num ) , false ) ;
2016-10-11 14:11:51 +02:00
return I2C_ERROR_OK ;
2016-10-06 13:21:30 +02:00
}
2016-10-11 14:11:51 +02:00
i2c_err_t i2cDetachSDA ( i2c_t * i2c , int8_t sda )
2016-10-06 13:21:30 +02:00
{
2018-06-27 09:01:06 +02:00
if ( i2c = = NULL ) {
2016-10-11 14:11:51 +02:00
return I2C_ERROR_DEV ;
2016-10-11 13:20:02 +02:00
}
2016-10-06 13:21:30 +02:00
pinMatrixOutDetach ( sda , false , false ) ;
pinMatrixInDetach ( I2C_SDA_IDX ( i2c - > num ) , false , false ) ;
2018-06-27 09:01:06 +02:00
pinMode ( sda , INPUT | PULLUP ) ;
2016-10-11 14:11:51 +02:00
return I2C_ERROR_OK ;
2016-10-06 13:21:30 +02:00
}
/*
2018-06-27 09:01:06 +02:00
* PUBLIC API
2016-10-06 13:21:30 +02:00
* */
2018-06-27 09:01:06 +02:00
// 24Nov17 only supports Master Mode
2019-01-09 10:07:54 +01:00
i2c_t * i2cInit ( uint8_t i2c_num , int8_t sda , int8_t scl , uint32_t frequency ) {
2018-07-24 19:43:45 +02:00
log_v ( " num=%d sda=%d scl=%d freq=%d " , i2c_num , sda , scl , frequency ) ;
2018-06-27 09:01:06 +02:00
if ( i2c_num > 1 ) {
return NULL ;
2016-11-16 23:46:56 +01:00
}
2016-10-11 13:20:02 +02:00
2018-06-27 09:01:06 +02:00
i2c_t * i2c = & _i2c_bus_array [ i2c_num ] ;
2016-10-06 13:21:30 +02:00
2019-01-09 10:07:54 +01:00
// pins should be detached, else glitch
2018-06-27 09:01:06 +02:00
if ( i2c - > sda > = 0 ) {
i2cDetachSDA ( i2c , i2c - > sda ) ;
2016-10-11 13:20:02 +02:00
}
2018-06-27 09:01:06 +02:00
if ( i2c - > scl > = 0 ) {
i2cDetachSCL ( i2c , i2c - > scl ) ;
2017-09-29 15:17:13 +02:00
}
2018-06-27 09:01:06 +02:00
i2c - > sda = sda ;
i2c - > scl = scl ;
2017-09-29 15:17:13 +02:00
2018-06-27 09:01:06 +02:00
# if !CONFIG_DISABLE_HAL_LOCKS
if ( i2c - > lock = = NULL ) {
2019-01-09 10:07:54 +01:00
i2c - > lock = xSemaphoreCreateRecursiveMutex ( ) ;
2018-06-27 09:01:06 +02:00
if ( i2c - > lock = = NULL ) {
return NULL ;
}
}
# endif
I2C_MUTEX_LOCK ( ) ;
2016-10-08 16:32:11 +02:00
2018-06-27 09:01:06 +02:00
i2cReleaseISR ( i2c ) ; // ISR exists, release it before disabling hardware
2016-10-10 13:54:36 +02:00
2018-06-27 09:01:06 +02:00
if ( frequency = = 0 ) { // don't change existing frequency
frequency = i2cGetFrequency ( i2c ) ;
if ( frequency = = 0 ) {
frequency = 100000L ; // default to 100khz
2016-10-06 13:21:30 +02:00
}
2018-06-27 09:01:06 +02:00
}
2016-12-14 01:41:22 +01:00
2018-06-27 09:01:06 +02:00
if ( i2c_num = = 0 ) {
DPORT_SET_PERI_REG_MASK ( DPORT_PERIP_RST_EN_REG , DPORT_I2C_EXT0_RST ) ; //reset hardware
DPORT_SET_PERI_REG_MASK ( DPORT_PERIP_CLK_EN_REG , DPORT_I2C_EXT0_CLK_EN ) ;
DPORT_CLEAR_PERI_REG_MASK ( DPORT_PERIP_RST_EN_REG , DPORT_I2C_EXT0_RST ) ; // release reset
} else {
DPORT_SET_PERI_REG_MASK ( DPORT_PERIP_RST_EN_REG , DPORT_I2C_EXT1_RST ) ; //reset Hardware
DPORT_SET_PERI_REG_MASK ( DPORT_PERIP_CLK_EN_REG , DPORT_I2C_EXT1_CLK_EN ) ;
DPORT_CLEAR_PERI_REG_MASK ( DPORT_PERIP_RST_EN_REG , DPORT_I2C_EXT1_RST ) ;
}
i2c - > dev - > ctr . val = 0 ;
i2c - > dev - > ctr . ms_mode = 1 ;
i2c - > dev - > ctr . sda_force_out = 1 ;
i2c - > dev - > ctr . scl_force_out = 1 ;
i2c - > dev - > ctr . clk_en = 1 ;
2016-10-10 13:54:36 +02:00
2018-06-27 09:01:06 +02:00
//the max clock number of receiving a data
i2c - > dev - > timeout . tout = 400000 ; //clocks max=1048575
//disable apb nonfifo access
i2c - > dev - > fifo_conf . nonfifo_en = 0 ;
2016-10-10 13:54:36 +02:00
2018-06-27 09:01:06 +02:00
i2c - > dev - > slave_addr . val = 0 ;
I2C_MUTEX_UNLOCK ( ) ;
2016-10-10 13:54:36 +02:00
2018-06-27 09:01:06 +02:00
i2cSetFrequency ( i2c , frequency ) ; // reconfigure
2016-10-08 16:32:11 +02:00
2018-06-27 09:01:06 +02:00
if ( ! i2cCheckLineState ( i2c - > sda , i2c - > scl ) ) {
return NULL ;
2016-10-06 13:21:30 +02:00
}
2018-06-27 09:01:06 +02:00
if ( i2c - > sda > = 0 ) {
i2cAttachSDA ( i2c , i2c - > sda ) ;
2017-10-21 14:26:46 +02:00
}
2018-06-27 09:01:06 +02:00
if ( i2c - > scl > = 0 ) {
i2cAttachSCL ( i2c , i2c - > scl ) ;
}
return i2c ;
2017-10-21 14:26:46 +02:00
}
2018-06-27 09:01:06 +02:00
void i2cRelease ( i2c_t * i2c ) // release all resources, power down peripheral
2016-10-06 13:21:30 +02:00
{
2016-10-11 13:20:02 +02:00
I2C_MUTEX_LOCK ( ) ;
2018-06-27 09:01:06 +02:00
if ( i2c - > sda > = 0 ) {
i2cDetachSDA ( i2c , i2c - > sda ) ;
}
if ( i2c - > scl > = 0 ) {
i2cDetachSCL ( i2c , i2c - > scl ) ;
2017-09-29 15:17:13 +02:00
}
2018-06-27 09:01:06 +02:00
i2cReleaseISR ( i2c ) ;
2016-10-08 16:32:11 +02:00
2018-06-27 09:01:06 +02:00
if ( i2c - > i2c_event ) {
vEventGroupDelete ( i2c - > i2c_event ) ;
i2c - > i2c_event = NULL ;
2017-10-21 21:14:14 +02:00
}
2018-06-27 09:01:06 +02:00
i2cFlush ( i2c ) ;
2016-10-08 16:32:11 +02:00
2018-06-27 09:01:06 +02:00
// reset the I2C hardware and shut off the clock, power it down.
if ( i2c - > num = = 0 ) {
DPORT_SET_PERI_REG_MASK ( DPORT_PERIP_RST_EN_REG , DPORT_I2C_EXT0_RST ) ; //reset hardware
DPORT_CLEAR_PERI_REG_MASK ( DPORT_PERIP_CLK_EN_REG , DPORT_I2C_EXT0_CLK_EN ) ; // shutdown hardware
} else {
DPORT_SET_PERI_REG_MASK ( DPORT_PERIP_RST_EN_REG , DPORT_I2C_EXT1_RST ) ; //reset Hardware
DPORT_CLEAR_PERI_REG_MASK ( DPORT_PERIP_CLK_EN_REG , DPORT_I2C_EXT1_CLK_EN ) ; // shutdown Hardware
}
2016-12-14 01:41:22 +01:00
2018-06-27 09:01:06 +02:00
I2C_MUTEX_UNLOCK ( ) ;
}
2016-10-10 13:54:36 +02:00
2018-06-27 09:01:06 +02:00
i2c_err_t i2cFlush ( i2c_t * i2c )
{
if ( i2c = = NULL ) {
return I2C_ERROR_DEV ;
}
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
i2cTriggerDumps ( i2c , i2c - > debugFlags & 0xff , " FLUSH " ) ;
2018-06-27 09:01:06 +02:00
// need to grab a MUTEX for exclusive Queue,
// what out if ISR is running?
i2c_err_t rc = I2C_ERROR_OK ;
if ( i2c - > dq ! = NULL ) {
// log_i("free");
// what about EventHandle?
free ( i2c - > dq ) ;
i2c - > dq = NULL ;
}
i2c - > queueCount = 0 ;
i2c - > queuePos = 0 ;
// release Mutex
return rc ;
}
2016-10-10 13:54:36 +02:00
2018-06-27 09:01:06 +02:00
i2c_err_t i2cWrite ( i2c_t * i2c , uint16_t address , uint8_t * buff , uint16_t size , bool sendStop , uint16_t timeOutMillis ) {
2019-01-10 21:37:13 +01:00
if ( ( i2c = = NULL ) | | ( ( size > 0 ) & & ( buff = = NULL ) ) ) { // need to have location to store requested data
return I2C_ERROR_DEV ;
}
2018-06-27 09:01:06 +02:00
i2c_err_t last_error = i2cAddQueueWrite ( i2c , address , buff , size , sendStop , NULL ) ;
2016-12-14 01:41:22 +01:00
2018-06-27 09:01:06 +02:00
if ( last_error = = I2C_ERROR_OK ) { //queued
if ( sendStop ) { //now actually process the queued commands, including READs
last_error = i2cProcQueue ( i2c , NULL , timeOutMillis ) ;
if ( last_error = = I2C_ERROR_BUSY ) { // try to clear the bus
if ( i2cInit ( i2c - > num , i2c - > sda , i2c - > scl , 0 ) ) {
last_error = i2cProcQueue ( i2c , NULL , timeOutMillis ) ;
2017-10-21 14:26:46 +02:00
}
2016-10-08 16:32:11 +02:00
}
2018-06-27 09:01:06 +02:00
i2cFlush ( i2c ) ;
} else { // stop not received, so wait for I2C stop,
last_error = I2C_ERROR_CONTINUE ;
}
}
return last_error ;
}
2016-10-08 16:32:11 +02:00
2018-06-27 09:01:06 +02:00
i2c_err_t i2cRead ( i2c_t * i2c , uint16_t address , uint8_t * buff , uint16_t size , bool sendStop , uint16_t timeOutMillis , uint32_t * readCount ) {
2019-01-10 21:37:13 +01:00
if ( ( size = = 0 ) | | ( i2c = = NULL ) | | ( buff = = NULL ) ) { // hardware will hang if no data requested on READ
return I2C_ERROR_DEV ;
}
2018-06-27 09:01:06 +02:00
i2c_err_t last_error = i2cAddQueueRead ( i2c , address , buff , size , sendStop , NULL ) ;
if ( last_error = = I2C_ERROR_OK ) { //queued
if ( sendStop ) { //now actually process the queued commands, including READs
last_error = i2cProcQueue ( i2c , readCount , timeOutMillis ) ;
if ( last_error = = I2C_ERROR_BUSY ) { // try to clear the bus
if ( i2cInit ( i2c - > num , i2c - > sda , i2c - > scl , 0 ) ) {
last_error = i2cProcQueue ( i2c , readCount , timeOutMillis ) ;
2017-10-21 14:26:46 +02:00
}
}
2018-06-27 09:01:06 +02:00
i2cFlush ( i2c ) ;
} else { // stop not received, so wait for I2C stop,
last_error = I2C_ERROR_CONTINUE ;
2016-10-06 13:21:30 +02:00
}
}
2018-06-27 09:01:06 +02:00
return last_error ;
2016-10-06 13:21:30 +02:00
}
2019-01-09 10:07:54 +01:00
# define MIN_I2C_CLKS 100 // minimum ratio between cpu and i2c Bus clocks
# define INTERRUPT_CYCLE_OVERHEAD 16000 // number of cpu clocks necessary to respond to interrupt
2016-10-11 14:11:51 +02:00
i2c_err_t i2cSetFrequency ( i2c_t * i2c , uint32_t clk_speed )
2016-10-06 13:21:30 +02:00
{
2018-06-27 09:01:06 +02:00
if ( i2c = = NULL ) {
2016-10-11 14:11:51 +02:00
return I2C_ERROR_DEV ;
2016-10-11 13:20:02 +02:00
}
2018-12-20 08:56:02 +01:00
uint32_t apb = getApbFrequency ( ) ;
uint32_t period = ( apb / clk_speed ) / 2 ;
if ( ( apb / 8192 > clk_speed ) | | ( apb / MIN_I2C_CLKS < clk_speed ) ) { //out of bounds
2019-01-09 10:07:54 +01:00
log_d ( " i2c freq(%d) out of bounds.vs APB Clock(%d), min=%d, max=%d " , clk_speed , apb , ( apb / 8192 ) , ( apb / MIN_I2C_CLKS ) ) ;
2018-12-20 08:56:02 +01:00
}
if ( period < ( MIN_I2C_CLKS / 2 ) ) {
period = ( MIN_I2C_CLKS / 2 ) ;
clk_speed = apb / ( period * 2 ) ;
2019-01-09 10:07:54 +01:00
log_d ( " APB Freq too slow, Reducing i2c Freq to %d Hz " , clk_speed ) ;
2018-12-20 08:56:02 +01:00
} else if ( period > 4095 ) {
period = 4095 ;
clk_speed = apb / ( period * 2 ) ;
2019-01-09 10:07:54 +01:00
log_d ( " APB Freq too fast, Increasing i2c Freq to %d Hz " , clk_speed ) ;
2018-12-20 08:56:02 +01:00
}
2019-01-09 10:07:54 +01:00
log_v ( " freq=%dHz " , clk_speed ) ;
2018-12-20 08:56:02 +01:00
2016-12-14 01:41:22 +01:00
uint32_t halfPeriod = period / 2 ;
uint32_t quarterPeriod = period / 4 ;
2016-10-11 13:20:02 +02:00
I2C_MUTEX_LOCK ( ) ;
2018-07-12 15:18:26 +02:00
2018-12-20 08:56:02 +01:00
I2C_FIFO_CONF_t f ;
2018-07-12 15:18:26 +02:00
f . val = i2c - > dev - > fifo_conf . val ;
2019-01-09 10:07:54 +01:00
/* Adjust Fifo thresholds based on differential between cpu frequency and bus clock.
The fifo_delta is calculated such that at least INTERRUPT_CYCLE_OVERHEAD cpu clocks are
available when a Fifo interrupt is triggered . This allows enough room in the Fifo so that
interrupt latency does not cause a Fifo overflow / underflow event .
*/
log_v ( " cpu Freq=%dMhz, i2c Freq=%dHz " , getCpuFrequencyMhz ( ) , clk_speed ) ;
uint32_t fifo_delta = ( INTERRUPT_CYCLE_OVERHEAD / ( ( getCpuFrequencyMhz ( ) * 1000000 / clk_speed ) * 10 ) ) + 1 ;
if ( fifo_delta > 24 ) fifo_delta = 24 ;
f . rx_fifo_full_thrhd = 32 - fifo_delta ;
f . tx_fifo_empty_thrhd = fifo_delta ;
2018-07-12 15:18:26 +02:00
i2c - > dev - > fifo_conf . val = f . val ; // set thresholds
2019-01-09 10:07:54 +01:00
log_v ( " Fifo delta=%d " , fifo_delta ) ;
2018-07-12 15:18:26 +02:00
2016-12-14 01:41:22 +01:00
//the clock num during SCL is low level
2017-01-05 12:54:40 +01:00
i2c - > dev - > scl_low_period . period = period ;
2016-12-14 01:41:22 +01:00
//the clock num during SCL is high level
2016-10-06 13:21:30 +02:00
i2c - > dev - > scl_high_period . period = period ;
2016-12-14 01:41:22 +01:00
//the clock num between the negedge of SDA and negedge of SCL for start mark
i2c - > dev - > scl_start_hold . time = halfPeriod ;
//the clock num between the posedge of SCL and the negedge of SDA for restart mark
i2c - > dev - > scl_rstart_setup . time = halfPeriod ;
2016-10-06 13:21:30 +02:00
2016-12-14 01:41:22 +01:00
//the clock num after the STOP bit's posedge
i2c - > dev - > scl_stop_hold . time = halfPeriod ;
//the clock num between the posedge of SCL and the posedge of SDA
i2c - > dev - > scl_stop_setup . time = halfPeriod ;
2016-10-06 13:21:30 +02:00
2016-12-14 01:41:22 +01:00
//the clock num I2C used to hold the data after the negedge of SCL.
i2c - > dev - > sda_hold . time = quarterPeriod ;
//the clock num I2C used to sample data on SDA after the posedge of SCL
i2c - > dev - > sda_sample . time = quarterPeriod ;
2016-10-11 13:20:02 +02:00
I2C_MUTEX_UNLOCK ( ) ;
2016-10-11 14:11:51 +02:00
return I2C_ERROR_OK ;
2016-10-06 13:21:30 +02:00
}
uint32_t i2cGetFrequency ( i2c_t * i2c )
{
2018-06-27 09:01:06 +02:00
if ( i2c = = NULL ) {
2016-10-11 13:20:02 +02:00
return 0 ;
}
2018-06-27 09:01:06 +02:00
uint32_t result = 0 ;
uint32_t old_count = ( i2c - > dev - > scl_low_period . period + i2c - > dev - > scl_high_period . period ) ;
if ( old_count > 0 ) {
2018-12-20 01:57:32 +01:00
result = getApbFrequency ( ) / old_count ;
2018-06-27 09:01:06 +02:00
} else {
result = 0 ;
}
return result ;
2016-10-06 13:21:30 +02:00
}
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
uint32_t i2cDebug ( i2c_t * i2c , uint32_t setBits , uint32_t resetBits ) {
if ( i2c ! = NULL ) {
i2c - > debugFlags = ( ( i2c - > debugFlags | setBits ) & ~ resetBits ) ;
return i2c - > debugFlags ;
2016-12-14 01:41:22 +01:00
}
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
return 0 ;
}
uint32_t i2cGetStatus ( i2c_t * i2c ) {
if ( i2c ! = NULL ) {
return i2c - > dev - > status_reg . val ;
2017-09-29 15:17:13 +02:00
}
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
else return 0 ;
2017-09-29 15:17:13 +02:00
}
2019-01-09 10:07:54 +01:00
2018-06-27 09:01:06 +02:00
/* todo
2018-07-24 19:43:45 +02:00
22 JUL18
need to add multi - thread capability , use dq . queueEvent as the group marker . When multiple threads
transactions are present in the same queue , and an error occurs , abort all succeeding unserviced transactions
with the same dq . queueEvent value . Succeeding unserviced transactions with different dq . queueEvent values
can be re - queued and processed independently .
Wire ReSTART fix, with others (#1717)
* ReSTART fix, Sequencing fix
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Update esp32-hal-i2c.c
* Update Wire.cpp
* ReSTART, Sequencing
pr #1665 introduce a problem with ReSTART, when solving this problem I found an interaction between the TxFifo refill, RxFifo empty and CMD[] fill. during certain sequences a dataqueue command would be skipped, this skipping resulted in a mismatch between the contents of the TxFifo and the i2c command sequence. The problem manifested as an ACK error.
In addition to this required bug fix I propose:
* `Wire.begin()` be changed from a `void` to a `bool` this will allow the reset functionality of `Wire.begin()` to be reported. Currently `Wire.begin()` attempts to reset the i2c Peripheral, but cannot report success/failure.
* `Wire.busy()` be added. this `bool` function returns the hardware status of the bus. This status can be use in multi-master environments for application level interleaving of commands, also in single master environment, it can be used to detect a 'hung' bus. With the functional change to `Wire.begin()` this allows app level recover of a hung bus.
* `Wire.lastError()` value updated for all errors, previously when interleaving `Wire.endTransmission(false)` and `Wire.readTransmission(false)`, the 128 byte `Wire.write()` buffer was exhausted without generating and error(very exotic). I discovered this error when I created a sequence of directed reads to a EEPROM. Each directed read used 2 bytes of the 128 byte `write()` buffer, so after 64 consecutive ReSTART writes with ReSTART reads, `Wire()` had no room to record the directed address bytes. It generated just a NAK check without setting the EEPROMs internal register address. The succeeding ReSTART read succeeded at incorrect address.
* Changes to the HAL layer:
** added `i2cGetStatus()` which returns the i2c peripheral status word, used to detect bus_busy currently
** added `i2cDebug()` programmatic control of debug buffer output
** changed `i2cAddQueue()` to allow data_only queue element this will allow a i2c transaction to use multiple data pointers.
** removed direct access to DumpInts(), DumpI2c() from app, use i2cDebug() to set trigger points
*
* Forgot DebugFlags Return
@andriyadi found this, total brain fade on my part.
2018-08-14 11:51:15 +02:00
30 JUL18 complete data only queue elements , this will allow transfers to use multiple data blocks ,
2018-06-27 09:01:06 +02:00
*/
