S3C6410 SDHCI driver crash

S3C6410 SDHCI driver crash

[Maurus Cuelenaere]

Hi,

I'm currently experiencing an SDHCI bug in 2.6.31 and linux-next, which 
didn't
seem to appear in 2.6.31-rc8.

This is on a new S3C6410 platform (SmartQ 7, see [1]) and it seems to be
consistently happening right after displaying "Loading hardware drivers" (in
user-space; Ubuntu-based).

The relevant kernel log lines are:

blk: request botched
kernel BUG at drivers/mmc/host/sdhci.c:298!

This is on 2.6.31-next-20090923, but I've also had this on regular 
2.6.31, Ben
Dooks latest tree and linux-next with the ADMA patches applied.

Any ideas?

[1] 
http://gitorious.org/mcuelenaere-smartq-kernel/mcuelenaere-smartq-kernel/commits/experimental-2.6.31

Regards,

Maurus Cuelenaere

S3C6410 SDHCI driver crash

[Ben Dooks]

On Thu, Sep 24, 2009 at 05:32:37PM +0200, Maurus Cuelenaere wrote:
> Hi,
> 
> I'm currently experiencing an SDHCI bug in 2.6.31 and linux-next, which 
> didn't
> seem to appear in 2.6.31-rc8.
> 
> This is on a new S3C6410 platform (SmartQ 7, see [1]) and it seems to be
> consistently happening right after displaying "Loading hardware drivers" (in
> user-space; Ubuntu-based).
> 
> The relevant kernel log lines are:
> 
> blk: request botched
> kernel BUG at drivers/mmc/host/sdhci.c:298!

Having a look at the code in mainline, this is being triggered in 
sdhci_write_block_pio(), by the following code:

296     while (blksize) {
297             if (!sg_miter_next(&host->sg_miter))
298                     BUG();

Which suggests that the size of the transfer was bigger than the sg list
passed to the transfer, which is worrying as this seems to suggest there
is a problem with the PIO code writing insufficient data to the card.

-- 
Ben

Q:      What's a light-year?
A:      One-third less calories than a regular year.

S3C6410 SDHCI driver crash

[Maurus Cuelenaere]

Op 24-09-09 21:00, Ben Dooks schreef:
 > On Thu, Sep 24, 2009 at 05:32:37PM +0200, Maurus Cuelenaere wrote:
 >> Hi,
 >>
 >> I'm currently experiencing an SDHCI bug in 2.6.31 and linux-next, which
 >> didn't
 >> seem to appear in 2.6.31-rc8.
 >>
 >> This is on a new S3C6410 platform (SmartQ 7, see [1]) and it seems to be
 >> consistently happening right after displaying "Loading hardware 
drivers" (in
 >> user-space; Ubuntu-based).
 >>
 >> The relevant kernel log lines are:
 >>
 >> blk: request botched
 >> kernel BUG at drivers/mmc/host/sdhci.c:298!
 >
 > Having a look at the code in mainline, this is being triggered in
 > sdhci_write_block_pio(), by the following code:
 >
 > 296     while (blksize) {
 > 297             if (!sg_miter_next(&host->sg_miter))
 > 298                     BUG();
 >
 > Which suggests that the size of the transfer was bigger than the sg list
 > passed to the transfer, which is worrying as this seems to suggest there
 > is a problem with the PIO code writing insufficient data to the card.
 >

An update on this: apparently this issue was caused by using the EXT4 driver
for loading an EXT3 FS, so it has nothing to do with the SDHCI driver...

Sorry for the false alarm.

I also want to mention that the ADMA patch from kki_ap/linux-2.6-samsung.git
seems to be working smoothly here.


Regards,

Maurus Cuelenaere