sdcard report

Andy Green andy at
Tue Jun 24 12:28:28 CEST 2008

Hash: SHA1

Somebody in the thread at some point said:
| I recently worked with showing a demo of software on 2 gta02v5 phones in
| asia. The demo does a good amount of I/O to an ext3 partition on an
| sdcard split into fat16 and ext3. Combined with the experience I've had
| here I have:

Sounds cool.  (BTW OM Email seems to have been down until now, I just
got this).

| 2GB A-Data sdcard fails
| 2GB Patriot sdcard works
| 512GB Kingston sdcard fails
| 512MB Sandisk sdcard fails (I think it was 512. Maybe 1GB)
| 1GB Transcend sdcard works
| At times, there were no failures indicated by the driver at all, but
| there are corruptions.

Hum.  We actually ship 512MB Sandisk, I write to it a lot on the VFAT
partition replacing the kernel there, literally several hundred times
and there is zero failure... (I would notice this very clearly because
it would make me have to take the card out and nuke it in a card reader
on the laptop.)

| The application creates multiple directories and files. There are
| consistency checks in the application that end up failing because the
| UID/GID of directories are incorrect. The odd thing is if you reboot and
| look they are correct. Perhaps there is some sort of cache incoherency
| going on here?

Hum.  Are we suspending during all this?  Currently we do something a
bit dirty on suspend to work around another pcf50633 resume ordering
issue.  If it turns out it makes trouble, we can work around it another way.

| I would like to see more complete testing of the sdcard read/write. If I
| have time, I will try to implement something simple to recreate the
| problem.


In the meanwhile, what has happened about the GSM resume crashes?  I
didn't hear any more about them bad or good and I can't recreate them.
Have they gone away or they are alive and well?

- -Andy
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Fedora -


More information about the openmoko-kernel mailing list