On Tue, Mar 27, 2012 at 09:53:41PM -0700, Hugh Dickins wrote:

On Wed, 28 Mar 2012, Sasha Levin wrote:

quoted

On Tue, Mar 27, 2012 at 1:17 AM, Andrew Morton
[off-list ref] wrote:

quoted

The task is waiting for IO to complete against a page, and it isn't
happening.

There are quite a lot of things which could cause this, alas. VM,
readahead, scheduler, core wait/wakeup code, IO system, interrupt
system (if it happens outside KVM, I guess).

So.... ugh. Hopefully someone will hit this in a situation where it
can be narrowed down or bisected.

I've only managed to reproduce it once, and was unable to get anything
useful out of it due to technical reasons.

The good part is that I've managed to hit something similar (although
I'm not 100% sure it's the same problem as the one in the original
mail).

I don't think this one has anything to do with the first you posted,
but it does look like a good catch against current linux-next, where
pagemap_pte_range() appears to do a spin_lock(&walk->mm->page_table_lock)
which should have been removed by "thp: optimize away unnecessary page
table locking".  Some kind of mismerge perhaps: Horiguchi-san added to Cc.

Thanks for reporting.
This spin_lock() also exists in mainline, so we need a fix on it.
I'll post later for -stable tree.

Thanks,
Naoya

--
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/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>