linux-kernel - Re: [PATCH v8 10/10] nvme: 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: <20240618064907.GA29009@lst.de>
Date: Tue, 18 Jun 2024 08:49:07 +0200
From: Christoph Hellwig <hch@....de>
To: John Garry <john.g.garry@...cle.com>
Cc: Kanchan Joshi <joshi.k@...sung.com>, axboe@...nel.dk, kbusch@...nel.org,
	hch@....de, sagi@...mberg.me, jejb@...ux.ibm.com,
	martin.petersen@...cle.com, viro@...iv.linux.org.uk,
	brauner@...nel.org, dchinner@...hat.com, jack@...e.cz,
	djwong@...nel.org, linux-block@...r.kernel.org,
	linux-kernel@...r.kernel.org, linux-nvme@...ts.infradead.org,
	linux-fsdevel@...r.kernel.org, tytso@....edu, jbongio@...gle.com,
	linux-scsi@...r.kernel.org, ojaswin@...ux.ibm.com,
	linux-aio@...ck.org, linux-btrfs@...r.kernel.org,
	io-uring@...r.kernel.org, nilay@...ux.ibm.com,
	ritesh.list@...il.com, willy@...radead.org, agk@...hat.com,
	snitzer@...nel.org, mpatocka@...hat.com, dm-devel@...ts.linux.dev,
	hare@...e.de, Alan Adamson <alan.adamson@...cle.com>
Subject: Re: [PATCH v8 10/10] nvme: Atomic write support

On Mon, Jun 17, 2024 at 07:04:23PM +0100, John Garry wrote:
>> Nit: I'd cache blk_rq_bytes(req), since that is repeating and this
>> function is called for each atomic IO.
>
> blk_rq_bytes() is just a wrapper for rq->__data_len. I suppose that we 
> could cache that value to stop re-reading that memory, but I would 
> hope/expect that memory to be in the CPU cache anyway.

Yes, that feels a bit pointless.

> Only NVMe supports an LBA space boundary, so that part is specific to NVMe.
>
> Regardless, the block layer already should ensure that the atomic write 
> length and boundary is respected. nvme_valid_atomic_write() is just an 
> insurance policy against the block layer or some other component not doing 
> its job.
>
> For SCSI, the device would error - for example - if the atomic write length 
> was larger than the device supported. NVMe silently just does not execute 
> the write atomically in that scenario, which we must avoid.

It might be worth to expand the comment to include this information to
help future readers.