Hi,
For some reason my 4GB Kingston SDHC class 6 cards stopped functioning
properly with 2.6.28 on my beagleboard (arm based). The very same cards (I
have two of them) worked properly with 2.6.27. What happened was that if I
booted from the card I would get errors -110.
They are similar to the ones reported at lkml last sept in
http://lkml.org/lkml/2008/9/23/390
Fortunately later on in this thread was a fix:
http://lkml.org/lkml/2008/10/20/124
This did not help me as the limit was already at 300000. However, I
decided to raise the limit to 500000. Still no go so I also doubled the
read limit from 100000 to 200000. After that the cards work without
problem. Changing the write timing back to 300000 brought the problem back
so apparently both timings need to be extended. Attached is the patch I
used for this. As it is a timeout limit value, it should not harm anyone,
and a longer timeout at least allows more cards to be used.
Signed-off-by: Frans Meulenbroeks <[email protected]>
---
What somewhat troubles me is that this worked in .27, so it might be this
patch does not address the root cause. Then again this is the best I can
do. If someone has a better solution, I am more than happy to test it.
Frans.
Index: git/drivers/mmc/core/core.c
===================================================================
--- git.orig/drivers/mmc/core/core.c
+++ git/drivers/mmc/core/core.c
@@ -284,9 +284,9 @@ void mmc_set_data_timeout(struct mmc_dat
* The limit is really 250 ms, but that is
* insufficient for some crappy cards.
*/
- limit_us = 300000;
+ limit_us = 500000;
else
- limit_us = 100000;
+ limit_us = 200000;
/*
* SDHC cards always use these fixed values.
On Wed, 25 Mar 2009 23:03:28 +0100 (CET)
Frans Meulenbroeks <[email protected]> wrote:
>
> This did not help me as the limit was already at 300000. However, I
> decided to raise the limit to 500000. Still no go so I also doubled the
> read limit from 100000 to 200000. After that the cards work without
> problem. Changing the write timing back to 300000 brought the problem back
> so apparently both timings need to be extended. Attached is the patch I
> used for this. As it is a timeout limit value, it should not harm anyone,
> and a longer timeout at least allows more cards to be used.
>
Unfortunately some controllers cannot cope with huge timeouts and will
complain. And such a huge timeout shouldn't be needed (and they weren't
any larger in .27).
>
> What somewhat troubles me is that this worked in .27, so it might be this
> patch does not address the root cause. Then again this is the best I can
> do. If someone has a better solution, I am more than happy to test it.
>
Indeed. Have you enabled MMC_DEBUG and checked that the controller
actually follows the configured timeouts?
Rgds
--
-- Pierre Ossman
Linux kernel, MMC maintainer http://www.kernel.org
rdesktop, core developer http://www.rdesktop.org
WARNING: This correspondence is being monitored by the
Swedish government. Make sure your server uses encryption
for SMTP traffic and consider using PGP for end-to-end
encryption.