Re: [RFP 2/3] Fix unmap_vma() bug related to mmu_notifiers

[Robin Holt <holt@sgi.com>]

On Fri, Jan 29, 2010 at 12:54:26PM -0800, Andrew Morton wrote:
> On Thu, 28 Jan 2010 13:56:29 -0600
> Robin Holt <holt@sgi.com> wrote:
> 
> > 
> > unmap_vmas() can fail to correctly flush the TLB if a
> > callout to mmu_notifier_invalidate_range_start() sleeps.
> > The mmu_gather list is initialized prior to the callout. If it is reused
> > while the thread is sleeping, the mm field may be invalid.
> > 
> > If the task migrates to a different cpu, the task may use the wrong
> > mmu_gather.
> 
> I don't think that description is complete.
> 
> There might be ways in which we can prevent this task from being
> migrated to another CPU, but that doesn't fix the problem because the
> mmu_gather is a per-CPU resource and might get trashed if another task
> is scheduled on THIS cpu, and uses its mmu_gather.

I couldn't reword it to make it any more clear.  The third paragraph of
Jack's original changelog says basically what you said.

Robin

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>