linux-kernel - Re: NVME performance regression in Linux 5.x due to lack of block level IO queueing

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: <a00e3a0d-8dff-c845-ed3b-9dc435187ecd@nvidia.com>
Date:   Fri, 1 Apr 2022 02:59:00 +0000
From:   Chaitanya Kulkarni <chaitanyak@...dia.com>
To:     Michael Marod <michael@...haelmarod.com>
CC:     Keith Busch <kbusch@...nel.org>,
        Christoph Hellwig <hch@...radead.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "linux-block@...r.kernel.org" <linux-block@...r.kernel.org>
Subject: Re: NVME performance regression in Linux 5.x due to lack of block
 level IO queueing


> Thanks for the info. I don't know of anything block or nvme specific that might
> explain an order of magnitude perf difference.
> 
> Could you try the same test without the filesytems? You mentioned using mdraid,
> so try '--filename=/dev/mdX'. If that also shows similiar performance
> difference, try using one of your nvme member drives directly, like
> '--filename=/dev/nvme1n1'. That should isolate which subsystem is contributing
> to the difference.
> 

With Keith's suggestions you can always take the perf numbers and
compare before and after mainly for nvme_queue_rq() and nvme_irq().

-ck