linux-kernel - Re: [PATCHv3, RESEND 5/8] khugepaged: Allow to collapse a page shared across fork

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: <ca114751-1a19-5b8a-5187-a0150716d3fa@nvidia.com>
Date:   Tue, 14 Apr 2020 19:42:39 -0700
From:   John Hubbard <jhubbard@...dia.com>
To:     "Kirill A. Shutemov" <kirill@...temov.name>
CC:     "Kirill A. Shutemov" <kirill.shutemov@...ux.intel.com>,
        <akpm@...ux-foundation.org>,
        Andrea Arcangeli <aarcange@...hat.com>,
        Zi Yan <ziy@...dia.com>, Yang Shi <yang.shi@...ux.alibaba.com>,
        Ralph Campbell <rcampbell@...dia.com>,
        William Kucharski <william.kucharski@...cle.com>,
        <linux-mm@...ck.org>, <linux-kernel@...r.kernel.org>
Subject: Re: [PATCHv3, RESEND 5/8] khugepaged: Allow to collapse a page shared
 across fork

On 2020-04-14 14:35, Kirill A. Shutemov wrote:
> On Mon, Apr 13, 2020 at 01:48:22PM -0700, John Hubbard wrote:
> 
> [Thanks for all your suggestions and corrections]
> 
>>> +	if (IS_ENABLED(CONFIG_DEBUG_VM) && expected_refcount > refcount) {
>>> +		pr_err("expected_refcount: %d, refcount: %d\n",
>>> +				expected_refcount, refcount);
>>> +		dump_page(page, "Unexpected refcount");
>>
>>
>> I see two issues with the pr_err() and the dump_page() call:
>>
>> 1. You probably want to rate limit this, otherwise you'll have a big
>> problem if lots of pages are pinned!
> 
> Nope. Only if kernel is buggy. See below.
> 
>> 2. Actually, I don't think you'd want to print anything at all here, even with
>> rate limiting, because doing so presumes that "unexpected" means "wrong". And I
>> think this patch doesn't expect to have GUP pins (or pin_user_pages() pins, ha),
>> but that doesn't mean that they're wrong to have.
> 
> See condition. We only do it if refcount is *below* expected refcount. It
> should never happen. Pinned page would have refcount above expected.
> 

Yes, you are right. I misread the condition. This actually is just right. :)

thanks,
-- 
John Hubbard
NVIDIA