linux-kernel - Re: [PATCH 3/3] apple-nvme: defer cache flushes by a specified amount

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: <84447986-6cd7-499c-834e-9aa01fad98da@kernel.dk>
Date: Tue, 11 Feb 2025 12:50:02 -0700
From: Jens Axboe <axboe@...nel.dk>
To: Alyssa Rosenzweig <alyssa@...enzweig.io>, Hector Martin
 <marcan@...can.st>, Sven Peter <sven@...npeter.dev>,
 Keith Busch <kbusch@...nel.org>, Christoph Hellwig <hch@....de>,
 Sagi Grimberg <sagi@...mberg.me>, Philipp Zabel <p.zabel@...gutronix.de>
Cc: asahi@...ts.linux.dev, linux-arm-kernel@...ts.infradead.org,
 linux-nvme@...ts.infradead.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 3/3] apple-nvme: defer cache flushes by a specified amount

On 2/11/25 11:25 AM, Alyssa Rosenzweig wrote:
> From: Jens Axboe <axboe@...nel.dk>
> 
> Cache flushes on the M1 nvme are really slow, taking 17-18 msec to
> complete. This can slow down workloads considerably, pure random writes
> end up being bound by the flush latency and hence run at 55-60 IOPS.
> 
> Add a deferred flush work around to provide better performance, at a
> minimal risk. By default, flushes are delayed at most 1 second, but this
> is configurable.
> 
> With this work-around, a pure random write workload runs at ~12K IOPS
> rather than 56 IOPS.

I knew this one would bite in the ass at some point down the line ;-)

I do think the feature is sane, and to my knowledge we haven't had any
issues with it since I originally wrote it 3 years ago. But I also
think it should probably go in the block layer proper, as other
devices might benefit from it.

That said, I'm fine with parking this in the apple nvme driver for
now, as we don't have to deal with multiple namespaces etc. Can
always get migrated to a core feature later, if desired.

-- 
Jens Axboe