linux-kernel - Re: [PATCH RFC 06/10] xfs: iomap CoW-based atomic write support

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: <6959eb95-acd2-45b5-be40-39892219c0d5@oracle.com>
Date: Fri, 7 Feb 2025 11:48:01 +0000
From: John Garry <john.g.garry@...cle.com>
To: "Darrick J. Wong" <djwong@...nel.org>
Cc: brauner@...nel.org, cem@...nel.org, dchinner@...hat.com, hch@....de,
        linux-xfs@...r.kernel.org, linux-fsdevel@...r.kernel.org,
        linux-kernel@...r.kernel.org, ojaswin@...ux.ibm.com,
        ritesh.list@...il.com, martin.petersen@...cle.com
Subject: Re: [PATCH RFC 06/10] xfs: iomap CoW-based atomic write support

On 06/02/2025 21:44, Darrick J. Wong wrote:
>>> What is this checking?  That something else already created a mapping in
>>> the COW fork, so we want to bail out to get rid of it?
>> I want to check if some data is shared. In that case, we should unshare.
> Why is it necessary to unshare?  Userspace gave us a buffer of new
> contents, and we're already prepared to write that out of place and
> remap it.

fine, as long as the remap does what we need, then I won't bother with 
this explicit unshare.

> 
>> And I am not sure if that check is sufficient.
>>
>> On the buffered write path, we may have something in a CoW fork - in that
>> case it should be flushed, right?
> Flushed against what?  Concurrent writeback or something?  The directio
> setup should have flushed dirty pagecache, so the only things left in
> the COW fork are speculative preallocations.  (IOWs, I don't understand
> what needs to be flushed or why.)

ah, ok, as long as DIO would have flushed dirty relevant pagecache, then 
we should be good.

Cheers,
John