SPI issue on iMX28 when change cpu clock

[victorien]

Hi,

I add a custom driver for cpu clock control on linux 3.14 (based on freescale's driver from 2.6.35). This driver works but I have an issue. If I reduce the CPU clock (261.819, 360 or 392.728 MHz) I have some data error when I receive data on SPI bus (number 2). The SPI bus doesn't miss data but modify some bytes (but not all of them) : bxxxxxx01 is changed to bxxxxxx00 and bxxxxxx10 is changed to bxxxxxx11 . At 454.737Mhz (max clock) I have no problem.

I checked waveform of data bus with an oscilloscope and the shape and timing are correct.

Someone knows this problem ?

Regards

Victorien

[wolfgangfraedri]

Hi all,

I stumbled upon the same issue after analyzing problems with our SPI.

In my case the error rate correlated highly with the temperature of the i.MX28: After startup of the system we had no errors. Heating up the i.MX28 caused the error rate to go up, reaching about 30 - 40% of the affected data (where the last two bits are different). After turning off the heating the error rate dropped again to virtual zero.

Switching to ref_xtal worked also for me.

Maybe this information is helpful for others looking for temperature issues...

Best regards,

Wolfgang

[lategoodbye]

I know this issue and i can reproduce it without cpufreq driver. Until now i didn't find a better solution than switching the clock parent to ref_xtal:

SPI receive error with mxs-spi

[victorien]

Thanks for this answer ; I'll test with the clock switching.

Is it possible that it's a bug of iMX28 ?

[fabio_estevam]

What about using the standard cpufreq driver instead?

[victorien]

I use standard cpufreq driver. The custom part is on power driver and voltage regulator. This drivers don't exist in kernel mainline for iMX28

[fabio_estevam]

It would be nice if you could send some patches for that (even if they are RFC) so that people could review the changes in the mailing lists.

Stefan Wahren has started to work on this sometime ago.

[victorien]

I based my development on Stefan Wahren's driver. I don't send patch to kernel mainline because the driver is dirty (commit not in english, copy files from Stefan Wahren driver instead of add his patch) and writed for 3.14. I don't yet have time to clean it and modify for 4.x kernel.

I work on standby to but there are bugs.

I can share patchs on forum if some people is interested.

[Yuri]

Hello,

  Do You use SPI master mode ? PIO or DMA ?

Have a great day,
Yuri

-----------------------------------------------------------------------------------------------------------------------
Note: If this post answers your question, please click the Correct Answer button. Thank you!
-----------------------------------------------------------------------------------------------------------------------

[victorien]

During my tests I read on SPI flash memory. I made the same test on my custom board and on evaluation board and I have the same issu on both board. The difference is that I have more byte change with evaluation board.

[victorien]

I use SPI master mode with DMA