linux-kernel - Re: [PATCH 2/4] mm/memory-failure: introduce get_hwpoison

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

Message-ID: <20150513001141.GA14599@hori1.linux.bs1.fc.nec.co.jp>
Date:	Wed, 13 May 2015 00:11:41 +0000
From:	Naoya Horiguchi <n-horiguchi@...jp.nec.com>
To:	Andrew Morton <akpm@...ux-foundation.org>
CC:	Andi Kleen <andi@...stfloor.org>, Tony Luck <tony.luck@...el.com>,
	"Kirill A. Shutemov" <kirill@...temov.name>,
	"linux-mm@...ck.org" <linux-mm@...ck.org>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	Naoya Horiguchi <nao.horiguchi@...il.com>
Subject: Re: [PATCH 2/4] mm/memory-failure: introduce get_hwpoison_page()
 for consistent refcount handling

On Tue, May 12, 2015 at 03:00:17PM -0700, Andrew Morton wrote:
> On Tue, 12 May 2015 09:46:47 +0000 Naoya Horiguchi <n-horiguchi@...jp.nec.com> wrote:
> 
> > memory_failrue() can run in 2 different mode (specified by MF_COUNT_INCREASED)
> > in page refcount perspective. When MF_COUNT_INCREASED is set, memory_failrue()
> > assumes that the caller takes a refcount of the target page. And if cleared,
> > memory_failure() takes it in it's own.
> > 
> > In current code, however, refcounting is done differently in each caller. For
> > example, madvise_hwpoison() uses get_user_pages_fast() and hwpoison_inject()
> > uses get_page_unless_zero(). So this inconsistent refcounting causes refcount
> > failure especially for thp tail pages. Typical user visible effects are like
> > memory leak or VM_BUG_ON_PAGE(!page_count(page)) in isolate_lru_page().
> > 
> > To fix this refcounting issue, this patch introduces get_hwpoison_page() to
> > handle thp tail pages in the same manner for each caller of hwpoison code.
> > 
> > There's a non-trivial change around unpoisoning, which now returns immediately
> > for thp with "MCE: Memory failure is now running on %#lx\n" message. This is
> > not right when split_huge_page() fails. So this patch also allows
> > unpoison_memory() to handle thp.
> >
> > ...
> >
> >  /*
> > + * Get refcount for memory error handling:
> > + * - @page: raw page
> > + */
> > +inline int get_hwpoison_page(struct page *page)
> > +{
> > +	struct page *head = compound_head(page);
> > +
> > +	if (PageHuge(head))
> > +		return get_page_unless_zero(head);
> > +	else if (PageTransHuge(head))
> > +		if (get_page_unless_zero(head)) {
> > +			if (PageTail(page))
> > +				get_page(page);
> > +			return 1;
> > +		} else {
> > +			return 0;
> > +		}
> > +	else
> > +		return get_page_unless_zero(page);
> > +}
> 
> This function is a bit weird.
> 
> - The comment looks like kerneldoc but isn't kerneldoc

OK, will fix it.

> - Why the inline?  It isn't fastpath?

No, so I'll drop 'inline'.

> - The layout is rather painful.  It could be
> 
> 	if (PageHuge(head))
> 		return get_page_unless_zero(head);
> 
> 	if (PageTransHuge(head)) {
> 		if (get_page_unless_zero(head)) {
> 			if (PageTail(page))
> 				get_page(page);
> 			return 1;
> 		} else {
> 			return 0;
> 		}
> 	}
> 
> 	return get_page_unless_zero(page);

OK, will do like this.

> - Some illuminating comments would be nice.  In particular that code
>   path where it grabs a ref on the tail page as well as on the head
>   page.  What's going on there?

We can't call get_page_unless_zero() directly for thp tail pages because
that breaks thp's refcounting rule (refcount of tail pages is stored in
->_mapcount.) This code intends to comply with the rule in hwpoison code too.
So I'll comment the point.

Hmm, I found just now that I forget to put_page(head) in if (PageTail) block,
which leaks head page after thp split.
So it'll be fixed in the next version.

Thanks,
Naoya Horiguchi--
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/