Re: [PATCH] mmc: at91_mci: add multiwrite cap

[Pierre Ossman]

On Thu, 09 Aug 2007 16:09:19 +0200
Nicolas Ferre <nicolas.ferre@xxxxxxxxxxxxx> wrote:

> 
> Ok thank you : it was the point.
> 
> Results : in brief :
> - there is work to be done ;-)
> - multiwrite test result is : OK
> 

I'm starting to get the feeling that writing this test driver was a
good idea. :)

> I had to modify the mmc_test to workaround the driver being stuck.
> I attach the patch so you can figure out the testcase I ran :
> 
> diff -u b/drivers/mmc/card/mmc_test.c b/drivers/mmc/card/mmc_test.c
> --- b/drivers/mmc/card/mmc_test.c
> +++ b/drivers/mmc/card/mmc_test.c
> @@ -149,7 +149,8 @@
>  	printk(KERN_INFO "%s: Testing reading power of two block
> sizes...\n", mmc_hostname(card->host));
>  
> -	for (i = 1;i <= 512;i <<= 1) {
> +//	for (i = 1;i <= 512;i <<= 1) {
> +	for (i = 512;i <= 512;i <<= 2) { /*must have size%4 == 0*/
>  		memset(&mrq, 0, sizeof(struct mmc_request));
>  
>  		mrq.cmd = &cmd;

These "fixes" shouldn't be needed with the recent patch by Marc Pignat.

> 
> And here is the output :
> 
> With a SD card :
> root@at91sam9263ek:~$ mmc0: card is read-write
> mmc0: new SD card at address 0002
> mmc0: About to test mmc subsystem
> mmc0: Testing writing power of two block sizes...
> .<7>mmc0: starting CMD16 arg 00000200 flags 00000015
> 
> mmc0: Result: OK
> mmc0: Testing reading power of two block sizes...
> mmc0: Result: OK
> mmc0: Testing correct bytes_xfered for a single block...
> mmc0: Result: OK
> mmc0: Testing correct bytes_xfered for multiple blocks...
> mmc0: Result: OK

This looks ok (although this test doesn't verify the more exotic cases
of updating bytes_xfered).

> I do not know if the OOps is due to a bad behavior of the driver
> during the wait for busy test...
> 

I don't suppose you could do a bit more digging? I'm not getting that
crash here.

> And with a MMC : quite the same results but with :
> 
> mmc0: Testing correct bytes_xfered for a single block...
> mmc0: Result: FAIL
> 

This is very odd. Could you test some more MMC and SD cards and see if
this is just a fluke? This test shouldn't be card dependent.

Rgds
Pierre

Attachment: signature.asc
Description: PGP signature