-
Notifications
You must be signed in to change notification settings - Fork 3k
ONSEMI: Fix a few issues related to I2C #5394
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
Three main issues: 1) The 0x13 special case section in write data in ncs36510_i2c.c didn't have a write++ command. 2) In the same write function, the WDAT8 command was put before the 0x13 section and this is not correct 3) Needed to add wait_us(0) before and after the register writes for apparent clock domain crossing delay times until registers are stable in HW There were also a handful of other tweaks related to general code maintenance and moving some status register checks to the proper locations.
+1 for testing and seeing these fixes. If you can also add/paste here the test results for this patch , to share what tests have been run. |
// The wait_us(0) command is needed so the I2C state machines have enough | ||
// time for data to settle across all clock domain crossings in their | ||
// synchronizers, both directions. | ||
#define SEND_COMMAND(cmd) wait_us(0); obj->membase->CMD_REG = cmd; wait_us(0); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It was wait_us(1), now it is 0? What is intention with 0 as a time waiting period?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Hello @0xc0170, I ran the I2C tests for the CI test shield, all the mbed tests, and our own test program below with 0 failures after many runs, even at different system clock divider frequencies.
For the delay going to 0 instead of 1us, only 0 is needed. The system just needs a handful of extra clock cycles so giving the MCU something to do so the internal state machines can synchronize the data. There are clock domain crossings as the data crosses from APB bus to internal and then back. It's not exactly a bug but not exactly what was intended either.
Here is the test code. I've ran tens of thousands of transactions and no longer see any failures.
// check if I2C is supported on this device
#if !DEVICE_I2C
#error [NOT_SUPPORTED] I2C not supported on this platform, add 'DEVICE_I2C' definition to your platform.
#endif
#include "mbed.h"
#include "I2CEeprom.h"
#include <string.h>
#define I2C_ADDRESS 0xA0
#define SIZE_DATA 33
#define ITERATIONS 512
int total_failures = 0;
int total_num_tests = 0;
Serial pc(USBTX, USBRX);
I2CEeprom memory(I2C_SDA, I2C_SCL, I2C_ADDRESS, 32, 0, 100000);
// Fill array with random characters
void init_string(char* buffer, int len)
{
srand(time(NULL));
int x = 0;
for(x = 0; x < len; x++){
buffer[x] = 'A' + (rand() % 26);
}
buffer[len-1] = 0; // add \0 to end of string
//pc.printf("\r\n****\r\nBuffer Len = `%d`, String = `%s`\r\n****\r\n",len,buffer);
}
void flash_WR(PinName I2C_SDA, PinName I2C_SCL, int size, int eeprom_address)
{
//I2CEeprom memory(I2C_SDA, I2C_SCL, eeprom_address, 32, 0);
int num_read = 0;
int num_written = 0;
volatile char test_string[size] = {0};
volatile char read_string[size] = {0};
init_string((char *)test_string,size); // populate test_string with random characters
for(int x = 0; x< size;x++){
read_string[x] = 0;
}
//pc.printf("\r\n****\r\n Test String = `%s` \r\n****\r\n",test_string);
num_written = memory.write(eeprom_address,(char *)test_string,size);
num_read = memory.read(eeprom_address,(char *)read_string,size);
pc.printf("\r\n****\r\n eeprom_address = 0x%x\r\n Len = %d\r\n Num Bytes Written = %d \r\n Num Bytes Read = %d \r\n Written String = %s \r\n Read String = %s \r\n****\r\n",eeprom_address,size,num_written,num_read,test_string,read_string);
if(num_written != num_read) {
total_failures++;
pc.printf("\r\n****\r\n ERROR: Wrote %d bytes but only read back %d bytes...\r\n****\r\n", num_written, num_read);
return;
}
if(strcmp((const char *)test_string, (const char *)read_string) != 0) {
total_failures++;
pc.printf("\r\n****\r\n ERROR: Strings don't match!...\r\n****\r\n ");
return;
}
}
// Test single byte R/W
void single_byte_WR_commands(PinName I2C_SDA, PinName I2C_SCL, int eeprom_address)
{
srand(time(NULL));
char test = 0;
char read;
int r = 0;
int w = 0;
for(int i = 0; i < 0x20; i++) {
test = i;
total_num_tests++;
w = memory.write(eeprom_address,test);
r = memory.read(eeprom_address,read);
pc.printf("\r\n****\r\n eeprom_address = 0x%x\r\n Num Bytes Read = %d \r\n Num Bytes Written = %d \r\n Read byte = 0x%x \r\n Written Byte = 0x%x \r\n****\r\n",eeprom_address,r,w,read,test);
if(w != r) {
total_failures++;
pc.printf("\r\n****\r\n ERROR: Wrote %d bytes but only read back %d bytes...\r\n****\r\n", w, r);
return;
}
if(test != read) {
total_failures++;
pc.printf("\r\n****\r\n ERROR: Bytes don't match! Sent 0x%x, received 0x%x...\r\n****\r\n ", test, read);
return;
}
}
}
// Test single byte R/W
void single_byte_WR(PinName I2C_SDA, PinName I2C_SCL, int eeprom_address)
{
srand(time(NULL));
char test = rand() % 0xFF;
char read;
int r = 0;
int w = 0;
total_num_tests++;
w = memory.write(eeprom_address,test);
r = memory.read(eeprom_address,read);
pc.printf("\r\n****\r\n eeprom_address = 0x%x\r\n Num Bytes Read = %d \r\n Num Bytes Written = %d \r\n Read byte = 0x%x \r\n Written Byte = 0x%x \r\n****\r\n",eeprom_address,r,w,read,test);
if(w != r) {
total_failures++;
pc.printf("\r\n****\r\n ERROR: Wrote %d bytes but only read back %d bytes...\r\n****\r\n", w, r);
return;
}
if(test != read) {
total_failures++;
pc.printf("\r\n****\r\n ERROR: Bytes don't match! Sent 0x%x, received 0x%x...\r\n****\r\n ", test, read);
return;
}
}
int genRandAddress() {
srand(time(NULL));
return (rand() % 0x7FFF);
}
// Entry point into the tests
int main()
{
pc.printf("\r\n****\r\n Starting single byte tests... \r\n****\r\n ");
single_byte_WR_commands(I2C_SDA, I2C_SCL, genRandAddress());
if(1) {
for(int i=0; i < ITERATIONS; i++) {
total_num_tests++;
single_byte_WR(I2C_SDA, I2C_SCL, genRandAddress());
pc.printf("\r\n Test iteration = %d \r\n", total_num_tests);
}
}
pc.printf("\r\n****\r\n Starting multi-byte tests... \r\n****\r\n");
if(1) {
for(int i=0; i < ITERATIONS; i++) {
total_num_tests++;
flash_WR(I2C_SDA, I2C_SCL, SIZE_DATA, genRandAddress());
pc.printf("\r\n Test iteration = %d \r\n", total_num_tests);
}
}
pc.printf("\r\n****\r\n Done... \r\n****\r\n");
pc.printf("\r\n****\r\n FAILED %d tests out of %d total tests \r\n****\r\n", total_failures, total_num_tests);
}
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks, that explains it
0xc0170 are you happy with this now ? |
Nice work Dan! |
/morph build |
Build : SUCCESSBuild number : 413 Triggering tests/morph test |
Test : SUCCESSBuild number : 206 |
This should not have got merged to mbed-os-5.6! PRs always go to master first.... |
Whoops. I did not notice that. I'll revert this pronto |
I already sent the fix to master. see the ref above. As this was just a day ago, can be removed also. |
Reverted. |
Three main issues were found and corrected:
There were also a handful of other tweaks related to general code maintenance and moving some status register checks to the proper locations.
Description
This pull request addresses some I2C issues that were found by CI test shield and by a customer. This pull request solves #5007 and #3511.
Status
Fully tested by extensive ON Semiconductor tests based on CI test shield hardware as well as fully verified on CI test shield with CI test shield tests for I2C.
@jacobjohnson-ON
@maclobdell