2+4+2 timing seemed to disrupt access to the sdcard

[neo at el-hennig.de]

> All this crashing caused some filesystem trouble. And then the
> second problem showed up.
> 
> I got i/o errors reading from the sdcard. Sometimes, I didn't get
> tiles for tangogps or pictures for my contacts.
> 
> Unmounting the sdcard and running fsck ought to fix this. But the fix
> was temporary. fsck would claim to fix some issues, and when I remounted 
> the card, I had normal access to the files. But it was
> all in the page cache - it wasn't real. Reboot, and exactly the same
> errors came up again. fsck would give exactly the same messages
> while running. And things would work for a while, but another reboot
> would always bring the same problems back.
> 
> It was as if writes to the sdcard failed - but failed silently. So the
> kernel didn't notice. But nothing actually got written, it just stayed
> in the page cache so everything seemed ok.
> 
> I also had a lot of hangs - probably because I have swap on the sdcard.
> If swapping to the card silently fail, then surely there is
> trouble when the stuff is swapped back in.
> 
> 
> Do the sdcard and the glamo share a bus? Maybe glamo can handle a
> speed that the sdcard cannot?

This is not related to the timing settings. I still use the "default" timing with new SHR-U and I have the same problem. I already opened a ticket in shr trac: http://trac.shr-project.org/trac/ticket/1143

The problem came up woth the new kernel, so I guess there is a problem with the sd driver in 2.6.32

--
neo at el-hennig.de

 Lubarsky's Law of Cybernetic Entomology:
   There's always one more bug.