Re: 2.6.5-rc2-aa4

[Andrea Arcangeli <andrea@suse.de>]

On Mon, Mar 29, 2004 at 03:36:38PM +0300, Sergey S. Kostyliov wrote:
> Hello Andrea,
> 
> On Friday 26 March 2004 11:21, Andrea Arcangeli wrote:
> > Fixup an hugetlbfs prio-tree truncate bug.
> > 
> > 	http://www.us.kernel.org/pub/linux/kernel/people/andrea/kernels/v2.6/2.6.5-rc2-aa4.gz
> > 	http://www.us.kernel.org/pub/linux/kernel/people/andrea/kernels/v2.6/2.6.5-rc2-aa4/
> > 
> > Files 2.6.5-rc2-aa3/extraversion and 2.6.5-rc2-aa4/extraversion differ
> > 
> > 	Rediffed.
> > 
> > Files 2.6.5-rc2-aa3/prio-tree.gz and 2.6.5-rc2-aa4/prio-tree.gz differ
> > 
> > 	Avoid missing vmas starting at pg_off > truncate offset in
> > 	hugetlbfs truncate. From Rajesh Venkatasubramanian.
> > -
> > 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/
> > 
> 
> Here is what I get after one day of uptime:
> 
> ------------[ cut here ]------------
> kernel BUG at mm/objrmap.c:111!
> invalid operand: 0000 [#1]
> PREEMPT
> CPU:    0
> EIP:    0060:[<c0142ed7>]    Not tainted
> EFLAGS: 00010282   (2.6.5-rc2-aa4)
> EIP is at find_pte+0xa7/0xc0

this this is a false positive, my mistake:

out_wrong_vma:
	BUG_ON(vma->vm_file);
	goto out;


it had to be a PageAnon check instead, this is likely a MAP_PRIVATE with
an anonymous space, and the vma has been splitted.

I will change the bug check to BUG_ON(PageAnon(page)), let's see if it
triggers again then.

Many thanks for helping fixing it!