andy at openmoko.com
Mon Jul 21 02:18:44 CEST 2008
-----BEGIN PGP SIGNED MESSAGE-----
Somebody in the thread at some point said:
| I just thought of a condition that isn't well tested here. It is quite
| possible that my sdcard corruption occurs on suspend or partial suspend.
| Is a block guaranteed to have been written and settled before a suspend
| can take place? Is there any confirmation of a write completing before
| the write returns? How about settle time for a write? Is there one for
| memory cards?
Wurrr... no we score badly on it.
pcf50633 is about the first down in suspend and joyfully pulls the plug
on SD card power then. Only later does Glamo / MCI driver go down.
It can make the trouble you're seeing if you had just written a block
but the SD Card hadn't truly digested it yet.
MCI stack tries to send CMD7 to the card to deselect it I guess to solve
this, but pcf50633 suspend action has occurred long before and we
removed SD card power then.
Tomorrow I'll try making the Glamo a child device of pcf50633, but this
is going to have radical suspend ordering ramifications.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org
-----END PGP SIGNATURE-----
More information about the openmoko-kernel