On Fri, 28 Feb 2003, Petr Vandrovec wrote:
> for some time I'm using patch attached at the end of this email,
> which modifies check_poison function to not only verify that
> last byte is POISON_END, but also that all preceeding bytes are
> either POISON_BEFORE or POISON_AFTER bytes.
>
> And now when I returned from my month vacation and upgraded
> from 2.5.52 to 2.5.63, when dselect/apt updates dozens of packages,
> I'm getting memory corruption reports as shown below - 22nd byte
> in vm_area_struct - which looks like that VM_ACCOUNT in vm_flags
> is set after vma is freeed... Any clue? Setting VM_ACCOUNT
> in mremap.c:move_vma after calling do_unmap() looks suspicious
> to me, but as I know almost nothing about MM...
Petr, do you have a corner I can hide in? Andrew will whip me.
He nags me from time to time about that code, he's sceptical
whereas I believed it fragile but okay for now. Looks like
you've caught me out, perhaps I should just take your POISON.
Hmm, it's really not what I want to be thinking about right now:
the real fix is to split_vma explicitly before the do_unmap, but
hard to get into that without opening up other cans of worms
(what to do if split_vma fails for lack of memory). Many thanks
for the report, I won't be able to fix it in the next three days,
but will get back to you when I have a patch to try.
Hugh
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Fri Feb 28 2003 - 22:00:50 EST