NULL pointer dereference at s3cmci

[Andy Green]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Somebody in the thread at some point said:
| Andy Green escreveu:
|> I was upleveling our tracking branch to 2.6.27-rc1 yesterday and I
|> noticed major changes went in there for s3cmci, just a FYI.
|
| If I'm looking correctly at the gitk output, the major change is
| actually the addition of the driver itself (which wasn't there before).
| However, the problematic code seems to be the same.

That's right, but the driver they added is not the same as what we had,
there are many minor edits and they changed from counting bytes to words
in the main struct.  I know this because I went through a full diff of
it inline after our failed original patch.  I chose the mainline version
of pretty much everything but still I noticed these changes.

|> Looking at the backtrace I think you have the right idea, somehow async
|> change detect comes before probe completes, something like that.
|
| Which would mean it's very timing dependent, since the oops depends on
| the scheduled work getting scheduled before the task executing the
| s3cmci initialization.

Yes, it would seem to need something like the scheduled work to
reschedule itself from inside if an "init complete" flags is not set then.

- -Andy
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org

iEYEARECAAYFAkiV7CQACgkQOjLpvpq7dMqRyQCeKB4fotDD81xR7rdzkDTT24VB
VBkAnRPP/ge+07fYM+ayiBYaA/cqfnLn
=eJDl
-----END PGP SIGNATURE-----