linux-kernel - Re: [PATCH 03/10] soft-offline: use migrate_pages() instead of migrate_huge

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>] [day] [month] [year] [list]

Message-ID: <20130329162337.GA30407@dhcp22.suse.cz>
Date:	Fri, 29 Mar 2013 17:23:51 +0100
From:	Michal Hocko <mhocko@...e.cz>
To:	"Aneesh Kumar K.V" <aneesh.kumar@...ux.vnet.ibm.com>
Cc:	Naoya Horiguchi <n-horiguchi@...jp.nec.com>, 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 03/10] soft-offline: use migrate_pages() instead of
 migrate_huge_page() (fwd)

Forwarding off-list email

On Fri 29-03-13 18:24:01, Aneesh Kumar K.V wrote:
> Michal Hocko <mhocko@...e.cz> writes:
> 
> > On Fri 29-03-13 10:56:00, Aneesh Kumar K.V wrote:
> >> Michal Hocko <mhocko@...e.cz> writes:
> > [...]
> >> > Little bit offtopic:
> >> > Btw. hugetlb migration breaks to charging even before this patchset
> >> > AFAICS. The above put_page should remove the last reference and then it
> >> > will uncharge it but I do not see anything that would charge a new page.
> >> > This is all because regula LRU pages are uncharged when they are
> >> > unmapped. But this a different story not related to this series.
> >> 
> >> 
> >> But when we call that put_page, we would have alreayd move the cgroup
> >> information to the new page. We have
> >> 
> >> 	h_cg = hugetlb_cgroup_from_page(oldhpage);
> >> 	set_hugetlb_cgroup(oldhpage, NULL);
> >> 
> >> 	/* move the h_cg details to new cgroup */
> >> 	set_hugetlb_cgroup(newhpage, h_cg);
> >> 
> >> 
> >> in hugetlb_cgroup_migrate
> >
> > Yes but the res counter charge would be missing for the newpage after
> > put_page
> >
> 
> Moving this to private as i didn't get what you meant here. 
> 
> We unchage hugepage via 
> 
> __put_compound_page -> free_huge_page -> hugetlb_cgroup_uncharge_page
> 
> That does
> 
> 	h_cg = hugetlb_cgroup_from_page(page);
> 	if (unlikely(!h_cg))
> 		return;
> 
> During migrate we do 
> 	set_hugetlb_cgroup(oldhpage, NULL);
> 
> 	/* move the h_cg details to new cgroup */
> 	set_hugetlb_cgroup(newhpage, h_cg);
> 
> So when we do a put_page(oldhpage), we won't be finding any cgroup
> attached to it and hence won't be updating the cgroup res counter value.
> 
> So not sure what you mean by res counter change would be missing for the
> newpage after put_page. I may be missing something here. Can you explain
> ?

Dang! You are right of course. I have missed the !h_cg check in
hugetlb_cgroup_uncharge_page. 

Sorry about the confusion. I am so distracted by many issues these
days...

Thanks!

-- 
Michal Hocko
SUSE Labs
--
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/