v2.6.24-rc2-409-g9418d5d: attempt to access beyond end of device

Bug 9370 Matthew Wilcox 1 participant 1 message Nov 13, 2007 2 messages not in the archive SCSI

lore

Share & tools

lore

Matthew Wilcox

author

2007-11-13 22:08 UTC #2 lore raw

In reply to a message not in the archive

Quoted 9 lines · On Tue, Nov 13, 2007 at 01:13:29PM -0800, [email protected]…

Probably caused by:
commit 6f5391c283d7fdcf24bf40786ea79061919d1e1d
Author: Matthew Wilcox <[email protected]>
Date:   Tue Sep 25 12:42:04 2007 -0400
    [SCSI] Get rid of scsi_cmnd->done
http://git.kernel.org/gitweb.cgi?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=6f5391c283d7fdcf24bf40786ea79061919d1e1d

Since the reporter says that reverting this commit fixes the problem, it must have caused the problem. For the benefit of those on linux-scsi who are just tuning into this report, the original mail is at
http://lkml.org/lkml/2007/11/13/250 and the report that reverting this commit fixes it is a followup to it.

I've stared at the patch, trying to figure out what's wrong. The only thing which makes any sense to me is that a request comes in, tagged as REQ_TYPE_FS. Later, it becomes REQ_TYPE_BLOCK_PC. This would have caused us to call sr_done() before the patch, and after the patch, we would not call sr_done(). Further eyes are welcome.

James has suggested that the problem may be that ide-scsi does just this. Thomas, can you confirm whether or not you were using ide-scsi?

Signature · 4 lines

Intel are signing my paycheques ... these opinions are still mine
"Bill, look, we understand that you're interested in selling us this
operating system, but compare it to ours.  We can't possibly take such
a retrograde step."