Re: ipc/testmsg GPF.

From: Dave Jones
Date: Thu Mar 07 2013 - 17:03:53 EST


On Thu, Mar 07, 2013 at 01:45:27PM -0800, Linus Torvalds wrote:
> Adding Davem to the list, since this looks like networking.
>
> That said, I'm also starting to wonder if you're just triggering odd
> bugs because of some use-after-free corruption of random kernel data
> structures?

Plausible. The sheer volume of crashes I've started seeing in the last 24 hours
since adding those small features to trinity are making me really suspicious.
If I had this volume of crashes reported to me by a Fedora user, I'd be
suspecting their hardware ;)

Tangentally related: We have a *ton* of weird bug reports against Fedora
that have been open for a while that look like random memory corruption
(linked list corruption, 'bad page table' scribbles etc).
I'd *love* to think that the bugs I'm now seeing are related, on the off-chance
we can finally get some resolution on those.

> I seem to have a dim memory of you usually doing your
> trinity runs with more debugging options, but based on just the oops
> message you don't have DEBUG_PAGEALLOC enabled, for example. Maybe
> that and SLUB debugging might turn up the culprit more quickly..

My kernel-hacker hive-mind implant is working. Just started a rebuild with that enabled.
Had slub debug already.

> [ Just checked - yes, you often do have DEBUG_PAGEALLOC in your oops
> reports, but not this time. list debugging too? ]

Yep.

Dave

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/