Unknown delay during boot (i.mx253) when DMA is enabled.

[PeterKardos]

Hi folks.

On your custom board based on i.mx253 running a stock 2.6.39.4 kernel, i've noticed a strange behavior when DMA is enabled.

For some reason the booting stalls for ~60s, and resumes without any error and/or warnings.

From what i could find out, the DMA in mx25 is the same as in mx35. I'm using a firmware originally designed for mx35 with the 2.6.39.4 kernel, "stock defaults" with the 3.x.x kernels.

I've tried different kernels; 3.0.8, 3.1.2, 3.2.0 all behaving the same way...

Is this a normal behavior? Am i doing something wrong, am i missing something?

Is there a way to improve the speed?

<snip>

[    3.386247] mousedev: PS/2 mouse device common for all mice
[    3.390370] input: gpio-keys as /devices/platform/gpio-keys/input/input0
[    3.397442] input: TSC2007 Touchscreen as /devices/virtual/input/input1
[    3.404424] rtc-ds1307 0-0068: SET TIME!
[    3.409180] rtc-ds1307 0-0068: rtc core: registered ds1339 as rtc0
[    3.413267] i2c /dev entries driver
[   64.181134] imx-sdma imx-sdma: initialized
[   64.184593] sdhci: Secure Digital Host Controller Interface driver
[   64.188044] sdhci: Copyright(c) Pierre Ossman
[   64.190758] _regulator_get: sdhci-esdhc-imx.1 supply vmmc not found, using dummy regulator
[   64.196746] mmc0: SDHCI controller on platform [sdhci-esdhc-imx.1] using DMA

</snip>

Thanx for your help in advance.

Cheers, Kardy

[FlorentAuger]

Hello,

Could this help?

http://sw-git.freescale.net/cgi-bin/gitweb.cgi?p=linux-2.6-imx.git;a=commit;h=c037ead48aeb3cbc54fb15...

[PeterKardos]

Hi Robert,

It seem to be only DMA related. If i disable DMA (kernel config=>Device Drivers=>DMA engine support), then the kernel starts within ~5s...

It could be that the firmware is "not the right fit" (even thou it should be compatible). Is there a tool for generating the FW?

-Kardy

[fabio_estevam]

Yes, there is a tool for generating the the sdma firmware that the mainline kernel requires at:

http://git.pengutronix.de/?p=imx/sdma-firmware.git;a=summary

If you provide the firmware generated by this tool, you will not experience the long wait timeout.

Regards,

Fabio Estevam

[RobertSchwebel]

Dear Kardy,

if it happens with 3.2.0 as well, I'd say it is a bug and should be reported on the ARM Linux Kernel mailinglist, with Cc: to kernel@pengutronix.de. I don't remember having seen anything like this on the mainline before. You could try to disable some of the drivers "around" the breakage (which is i2c and imx-sdma) and check if it is really the sdma driver.

rsc