linux-kernel - Re: [PATCH v3 5/5] block: use chunk_sectors when evaluating stacked atomic write limits

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: <4098ac0d-1558-4ca5-a873-275e6630ea40@oracle.com>
Date: Wed, 9 Jul 2025 14:16:24 +0100
From: John Garry <john.g.garry@...cle.com>
To: "Martin K. Petersen" <martin.petersen@...cle.com>
Cc: Mikulas Patocka <mpatocka@...hat.com>, axboe@...nel.dk, agk@...hat.com,
        snitzer@...nel.org, song@...nel.org, yukuai3@...wei.com, hch@....de,
        nilay@...ux.ibm.com, dm-devel@...ts.linux.dev,
        linux-kernel@...r.kernel.org, linux-raid@...r.kernel.org,
        linux-block@...r.kernel.org, ojaswin@...ux.ibm.com
Subject: Re: [PATCH v3 5/5] block: use chunk_sectors when evaluating stacked
 atomic write limits

On 09/07/2025 02:39, Martin K. Petersen wrote:
> 
> John,
> 
>> For io_min/opt, maybe reduce to a factor of the stripe size / width
>> (and which fits in a unsigned int).
>>
>> I am not sure if it is even sane to have such huge values in io_min
>> and the bottom disk io_min should be used directly instead.
> 
> The intent for io_min was to convey the physical_block_size in the case
> of an individual drive. And for it to be set to the stripe chunk size in
> stacking scenarios that would otherwise involve read-modify-write (i.e.
> RAID5 and RAID6).

And the same is done for md raid0/dm stripe; I suppose that the idea is 
to encourage larger than chunk size writes to see the performance 
advantage in striping.

 > > io_opt was meant to communicate the stripe width. Reporting very large
> values for io_opt is generally counterproductive since we can't write
> multiple gigabytes in a single operation anyway.
>

Right, and so it seems counterproductive to have chunk size much bigger 
than bottom device io_opt

> logical <= physical <= io_min <= io_opt <= max_sectors <= max_hw_sectors
>