On Fri, 27 Apr 2012 10:42:26 -0700
Ying Han <yinghan@xxxxxxxxxx> wrote:
> I am resending this patch orginally from Mel, and the reason we spotted this
> is due to the next patch where I am adding the mlock stat into per-memcg
> meminfo. We found out that it is impossible to update the counter if the page
> is in the freeing patch w/ mlocked bit set.
>
> Then we started wondering if it is possible at all. It shouldn't happen that
> freeing a mlocked page without going through munlock_vma_pages_all(). Looks
> like it did happen few years ago, and here is the patch introduced it
>
> commit 985737cf2ea096ea946aed82c7484d40defc71a8
> Author: Lee Schermerhorn <lee.schermerhorn@xxxxxx>
> Date: Sat Oct 18 20:26:53 2008 -0700
>
> mlock: count attempts to free mlocked page
>
> There are two ways to persue and I would like to ask people's opinion:
>
> 1. revert the patch totally and the page will get into bad_page(). Then we
> get the report as well.
>
> 2. fix up the page like the patch does but put on warn_once() to report the
> problem.
>
> People might feel more confident by doing step by step which adding the
> warn_on() first and then revert it later. So I resend the patch from Mel and
> here is the patch:
>
> When a page is freed with the PG_mlocked set, it is considered an unexpected
> but recoverable situation. A counter records how often this event happens
> but it is easy to miss that this event has occured at all. This patch warns
> once when PG_mlocked is set to prompt debuggers to check the counter to
> see how often it is happening.
The changelog is kinda confusing and tl;dr, but the idea seems good ;)
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -599,6 +599,11 @@ out:
> */
> static inline void free_page_mlock(struct page *page)
> {
> + WARN_ONCE(1, KERN_WARNING
> + "Page flag mlocked set for process %s at pfn:%05lx\n"
> + "page:%p flags:%#lx\n",
> + current->comm, page_to_pfn(page),
> + page, page->flags|__PG_MLOCKED);
> __dec_zone_page_state(page, NR_MLOCK);
> __count_vm_event(UNEVICTABLE_MLOCKFREED);
> }
KERN_WARNING seems wimpy: we want to shout about this, so KERN_ERR. Or
just leave it empty.
Also, the patch duplicates bad_page()/dump_page(). Can we use them?
They use KERN_ALERT, btw.
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. 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@xxxxxxxxx";> email@xxxxxxxxx </a>