Skip to contentSkip to conversation

Slab corruption in 2.6.16-rc5-mm2

DISCUSSION Chuck Ebbert 4 participants 7 messages Mar 8, 2006

lore
Share & tools
lore
#0 lore raw

In-Reply-To: <[email protected]>

Quoted 10 lines · Linus Torvalds
Quoted 8 lines

When someone converted the *buffer* allocation to kzalloc they also removed the the memset for the *packet_cmmand* struct.

The

memset(&cgc, 0, sizeof(struct packet_command));

should be added back I think.

Good eyes. I bet that's it.

Heh. This exact fix was posted to linux-kernel by Lee Schermerhorn three weeks ago:

Date: Wed, 15 Feb 2006 14:07:37 -0500
From: Lee Schermerhorn <[email protected]>
Subject: [PATCH] 2.6.16-rc3-mm1 - restore zeroing of packet_command
struct in sr_ioctl.c
To: linux-kernel <[email protected]>
Cc: Andrew Morton <[email protected]>
Message-ID: <[email protected]>

Signature · 2 lines
Chuck
"Penguins don't come from next door, they come from the Antarctic!"
#1 lore raw
Quoted 20 lines · Chuck Ebbert

In-Reply-To: <[email protected]>

Quoted 10 lines · Linus Torvalds
Quoted 8 lines

When someone converted the *buffer* allocation to kzalloc they also removed the the memset for the *packet_cmmand* struct.

The

memset(&cgc, 0, sizeof(struct packet_command));

should be added back I think.

Good eyes. I bet that's it.

Heh. This exact fix was posted to linux-kernel by Lee Schermerhorn three weeks ago:

Quoted 9 lines

Date: Wed, 15 Feb 2006 14:07:37 -0500
From: Lee Schermerhorn <[email protected]>
Subject: [PATCH] 2.6.16-rc3-mm1 - restore zeroing of packet_command
struct in sr_ioctl.c
To: linux-kernel <[email protected]>
Cc: Andrew Morton <[email protected]>
Message-ID: <[email protected]>

It isn't Andrew's job to make sure a patch gets to the right place until it is safely in -mm, and even then he's not always going to know the severity and importance unless he's told.

If it was a patch to "restore" a regression in behaviour, CCs should at least have gone to the author of the patch that broke it, and the subsystem maintainers / list / etc as well.

Signature · 2 lines
SUSE Labs, Novell Inc.
Send instant messages to your online friends http://au.messenger.yahoo.com 
#2 lore raw
Quoted 32 lines · Nick Piggin
Quoted 20 lines · Chuck Ebbert

In-Reply-To: <[email protected]>

Quoted 10 lines · Linus Torvalds
Quoted 8 lines

When someone converted the *buffer* allocation to kzalloc they also removed the the memset for the *packet_cmmand* struct.

The

memset(&cgc, 0, sizeof(struct packet_command));

should be added back I think.

Good eyes. I bet that's it.

Heh. This exact fix was posted to linux-kernel by Lee Schermerhorn three weeks ago:

Quoted 9 lines

Date: Wed, 15 Feb 2006 14:07:37 -0500
From: Lee Schermerhorn <[email protected]>
Subject: [PATCH] 2.6.16-rc3-mm1 - restore zeroing of packet_command
struct in sr_ioctl.c
To: linux-kernel <[email protected]>
Cc: Andrew Morton <[email protected]>
Message-ID: <[email protected]>

Quoted 3 lines

It isn't Andrew's job to make sure a patch gets to the right place until it is safely in -mm, and even then he's not always going to know the severity and importance unless he's told.

Is too!

Quoted 3 lines

If it was a patch to "restore" a regression in behaviour, CCs should at least have gone to the author of the patch that broke it, and the subsystem maintainers / list / etc as well.

I actually merged Lee's patch into -mm, copied James on it and then I dropped it when I saw that it spat rejects against an updated version of James's tree, assuming that it had been merged.

Often I'll check that a patch reverts successfully from the upstream tree before dropping it, but for an obvious one like that I guess I didn't bother, and assumed that James had taken it. Only he hadn't - instead he'd
gone and merged something else, hence the rejects. Oh well.

#3 lore raw
Quoted 10 lines · Andrew Morton
Quoted 3 lines · Nick Piggin

It isn't Andrew's job to make sure a patch gets to the right place until it is safely in -mm, and even then he's not always going to know the severity and importance unless he's told.

Is too!

OK, partially. As this case illustrates, everybody makes mistakes and you obviously can't go back and verify you got all the patches because.

The guy who hits the bug and/or writes the patch can easily see it is still not merged and shout.

Quoted 6 lines
Quoted 4 lines

If it was a patch to "restore" a regression in behaviour, CCs should at least have gone to the author of the patch that broke it, and the subsystem maintainers / list / etc as well.

Quoted 9 lines

I actually merged Lee's patch into -mm, copied James on it and then I dropped it when I saw that it spat rejects against an updated version of James's tree, assuming that it had been merged.

Often I'll check that a patch reverts successfully from the upstream tree before dropping it, but for an obvious one like that I guess I didn't bother, and assumed that James had taken it. Only he hadn't - instead he'd
gone and merged something else, hence the rejects. Oh well.

You do a great job, but "push the work out to the end nodes", right? That's how we get this network to scale. It is trivial for people to verify their important patches have propogated as the release approaches.

(A little harder for part-timers who aren't in the loop about exactly when the release will happen, thanks to our -ridiculous-count release system, but still easy compared with your having to double check everything).

Signature · 2 lines
SUSE Labs, Novell Inc.
Send instant messages to your online friends http://au.messenger.yahoo.com 
#4 lore raw
Quoted 7 lines · Nick Piggin
Quoted 6 lines

Often I'll check that a patch reverts successfully from the upstream tree before dropping it, but for an obvious one like that I guess I didn't bother, and assumed that James had taken it. Only he hadn't - instead he'd
gone and merged something else, hence the rejects. Oh well.

Quoted 8 lines

You do a great job, but "push the work out to the end nodes", right? That's how we get this network to scale. It is trivial for people to verify their important patches have propogated as the release approaches.

(A little harder for part-timers who aren't in the loop about exactly when the release will happen, thanks to our -ridiculous-count release system, but still easy compared with your having to double check everything).

Well yes, Lee sent the fix to the guy who he got the kernel release from in the reasonable expectation that I'd take care of getting it to where it needed to be.

Problem is, a) I screwed up, b) James screwed up and c) someone just happened to change those few lines of code in that place within a few-day window.

That triple-combo doesn't happen very often.

#5 lore raw
Quoted 10 lines · Andrew Morton

Well yes, Lee sent the fix to the guy who he got the kernel release from in the reasonable expectation that I'd take care of getting it to where it needed to be.

Problem is, a) I screwed up, b) James screwed up and c) someone just happened to change those few lines of code in that place within a few-day window.

That triple-combo doesn't happen very often.

I guess what I'm advocating isn't foolproof either: the guy who wrote the patch might die (knock on wood) ;)

Carry on.

Signature · 2 lines
SUSE Labs, Novell Inc.
Send instant messages to your online friends http://au.messenger.yahoo.com 
#6 lore raw
Quoted 10 lines · Nick Piggin
Quoted 8 lines · Andrew Morton

Well yes, Lee sent the fix to the guy who he got the kernel release from in the reasonable expectation that I'd take care of getting it to where it needed to be.

Problem is, a) I screwed up, b) James screwed up and c) someone just happened to change those few lines of code in that place within a few-day window.

Quoted 5 lines
Quoted 2 lines

That triple-combo doesn't happen very often.

I guess what I'm advocating isn't foolproof either: the guy who wrote the patch might die (knock on wood) ;)

Thanks, Nick. I'll remember that... See you at OLS?

Lee

P.S. ;-)