linux-kernel - Re: [xfs] ab23a77687: aim7.jobs-per-min -18.5% regression

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]

Date:   Mon, 13 Sep 2021 09:32:19 +1000
From:   Dave Chinner <dchinner@...hat.com>
To:     kernel test robot <oliver.sang@...el.com>
Cc:     "Darrick J. Wong" <djwong@...nel.org>,
        LKML <linux-kernel@...r.kernel.org>, lkp@...ts.01.org,
        lkp@...el.com, ying.huang@...el.com, feng.tang@...el.com,
        zhengjun.xing@...ux.intel.com
Subject: Re: [xfs]  ab23a77687:  aim7.jobs-per-min -18.5% regression

On Thu, Sep 09, 2021 at 10:23:55PM +0800, kernel test robot wrote:
> 
> 
> Hi Dave Chinner,
> 
> we reported "[xfs]  6df693ed7b:  aim7.jobs-per-min -15.7% regression" as [1]
> when this change is still on
> https://git.kernel.org/cgit/linux/kernel/git/djwong/xfs-linux.git xfs-5.15-merge
> 
> now this change is on mainline.
> so we report again to highlight we still observe similar performance regression
> on mainline.
> 
> previously we also tried to test by turning off RWSEM_SPIN_ON_OWNER per your
> guidance, but still observed similar regression, though we are not sure if our
> method to turn off RWSEM_SPIN_ON_OWNER is enough. the detail result is in [2]
> 
> [1] https://lore.kernel.org/all/20210809064248.GB5761@xsang-OptiPlex-9020/
> [2] https://lore.kernel.org/all/20210818085248.GA28771@xsang-OptiPlex-9020/

Did you look at the config that was generated before running the
tests again?

> #
> # Automatically generated file; DO NOT EDIT.
> # Linux/x86_64 5.14.0-rc4 Kernel Configuration
> #
....
> CONFIG_ARCH_SUPPORTS_ATOMIC_RMW=y
> CONFIG_MUTEX_SPIN_ON_OWNER=y
> CONFIG_RWSEM_SPIN_ON_OWNER=y
> CONFIG_LOCK_SPIN_ON_OWNER=y
....


So these tests were still run with the same RWSEM_SPIN_ON_OWNER=y
configuration, and so unsurprisingly the result was the same with
spin contention on the directory inode rwsem.

You'll save yourself (and everyone else) a lot of time by validating
that your config changes are valid before re-running tests...

Cheers,

Dave.
-- 
Dave Chinner
dchinner@...hat.com