linux-ext4 - Re: [RFC 0/5] Ext4: Add support for blocksize < pagesize for dioread

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:   Thu, 24 Oct 2019 06:42:20 +0530
From:   Ritesh Harjani <riteshh@...ux.ibm.com>
To:     "Theodore Y. Ts'o" <tytso@....edu>
Cc:     jack@...e.cz, linux-ext4@...r.kernel.org,
        linux-fsdevel@...r.kernel.org, mbobrowski@...browski.org
Subject: Re: [RFC 0/5] Ext4: Add support for blocksize < pagesize for
 dioread_nolock

Hello Ted,

On 10/24/19 4:56 AM, Theodore Y. Ts'o wrote:
> Hi Ritesh,
> 
> I haven't had a chance to dig into the test failures yet, but FYI....
> when I ran the auto test group in xfstests, I saw failures for
> generic/219, generic 273, and generic/476 --- these errors did not
> show up when running using a standard 4k blocksize on x86, and they
> also did not show up when running dioread_nolock using a 4k blocksize.
> 

Sorry about that. Were these 3 the only tests you saw to be failing,
or there were more?


> So I tried running "generic/219 generic/273 generic/476" 30 times,
> using in a Google Compute Engine VM, using gce-xfstests, and while I
> wasn't able to get generic/219 to fail when run in isolation,
> generic/273 seems to fail quite reliably, and generic/476 about a
> third of the time.
> 
> How much testing have you done with these patches?

I did test "quick" group of xfstests & ltp/fsx tests with the posted
version. And had tested a full suite of xfstests with one of my previous 
version.
I guess I was comparing against 1K blocksize without dioread_nolock.
But as I think more about it, I may need to compare against vanilla
kernel with 1K blocksize even without dioread_nolock. Since there may
be some changes in blocksize < pagesize path with this patch.


Also I see these tests(generic/273 & generic/476) are not part
of quick group. Let me check more about these failing tests at my end.

Thanks for your inputs.

-ritesh


> 
> Thanks,
> 
> 							- Ted
> 
> TESTRUNID: tytso-20191023144956
> KERNEL:    kernel 5.4.0-rc3-xfstests-00005-g39b811602906 #1244 SMP Wed Oct 23 11:30:25 EDT 2019 x86_64
> CMDLINE:   --update-files -C 30 -c dioread_nolock_1k generic/219 generic/273 generic/476
> CPUS:      2
> MEM:       7680
> 
> ext4/dioread_nolock_1k: 90 tests, 42 failures, 10434 seconds
>    Failures: generic/273 generic/273 generic/273 generic/273
>      generic/476 generic/273 generic/476 generic/273 generic/273
>      generic/273 generic/476 generic/273 generic/476 generic/273
>      generic/476 generic/273 generic/476 generic/273 generic/273
>      generic/273 generic/273 generic/273 generic/476 generic/273
>      generic/273 generic/273 generic/273 generic/476 generic/273
>      generic/476 generic/273 generic/476 generic/273 generic/273
>      generic/273 generic/273 generic/273 generic/273 generic/273
>      generic/476 generic/273 generic/476
> Totals: 90 tests, 0 skipped, 42 failures, 0 errors, 10434s
>