linux-kernel - Re: [PATCH 01/10] migrate: add migrate_entry_wait

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <1364271900-byj82fk2-mutt-n-horiguchi@ah.jp.nec.com>
Date:	Tue, 26 Mar 2013 00:25:00 -0400
From:	Naoya Horiguchi <n-horiguchi@...jp.nec.com>
To:	Michal Hocko <mhocko@...e.cz>
Cc:	linux-mm@...ck.org, Andrew Morton <akpm@...ux-foundation.org>,
	Mel Gorman <mel@....ul.ie>, Hugh Dickins <hughd@...gle.com>,
	KOSAKI Motohiro <kosaki.motohiro@...fujitsu.com>,
	Andi Kleen <andi@...stfloor.org>,
	Hillf Danton <dhillf@...il.com>, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 01/10] migrate: add migrate_entry_wait_huge()

On Mon, Mar 25, 2013 at 11:13:40AM +0100, Michal Hocko wrote:
> On Fri 22-03-13 16:23:46, Naoya Horiguchi wrote:
> > When we have a page fault for the address which is backed by a hugepage
> > under migration, the kernel can't wait correctly until the migration
> > finishes. This is because pte_offset_map_lock() can't get a correct
> > migration entry for hugepage. This patch adds migration_entry_wait_huge()
> > to separate code path between normal pages and hugepages.
> 
> The changelog is missing a description what is the effect of the bug. I
> assume that we end up busy looping on the huge page page fault until
> migration finishes, right?

Right. I'll add it in the description.

> Should this be applied to the stable tree or the current usage of the huge
> page migration (HWPOISON) is not spread enough?

Yes, it's better to also send it to stable.

> I like how you got rid of the duplication but I think this still doesn't
> work for all archs/huge page sizes.
> 
> [...]
> > diff --git v3.9-rc3.orig/mm/hugetlb.c v3.9-rc3/mm/hugetlb.c
> > index 0a0be33..98a478e 100644
> > --- v3.9-rc3.orig/mm/hugetlb.c
> > +++ v3.9-rc3/mm/hugetlb.c
> > @@ -2819,7 +2819,7 @@ int hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
> >  	if (ptep) {
> >  		entry = huge_ptep_get(ptep);
> >  		if (unlikely(is_hugetlb_entry_migration(entry))) {
> > -			migration_entry_wait(mm, (pmd_t *)ptep, address);
> > +			migration_entry_wait_huge(mm, (pmd_t *)ptep, address);
> 
> e.g. ia64 returns pte_t *

We need arch-independent fix.

> [...]
> > +void migration_entry_wait_huge(struct mm_struct *mm, pmd_t *pmd,
> > +				unsigned long address)
> > +{
> > +	spinlock_t *ptl = pte_lockptr(mm, pmd);
> > +	__migration_entry_wait(mm, (pte_t *)pmd, ptl);
> 
> So you are trying to get lockptr from pmd but you have pte in fact. No
> biggy for !USE_SPLIT_PTLOCKS but doesn't work otherwise. So you probably
> need a arch specific huge_pte_lockptr callback for USE_SPLIT_PTLOCKS.

I must fix it, thanks.
And it might be good to generalize the idea of USE_SPLIT_PTLOCKS to pud and pmd.

> Or am I missing something here? All the pte usage in hugetlb is one
> giant mess and I wouldn't be surprised if there were more places
> hardcoded to pmd there.

Yes, that's a big problem on hugetlb and we need many clean-ups.

Thanks,
Naoya
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/