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:   Tue, 26 Mar 2019 09:02:40 +0200
From:   Alexander Shishkin <alexander.shishkin@...ux.intel.com>
To:     Ben Gainey <Ben.Gainey@....com>,
        "mingo\@kernel.org" <mingo@...nel.org>,
        "peterz\@infradead.org" <peterz@...radead.org>,
        "acme\@redhat.com" <acme@...hat.com>,
        Will Deacon <Will.Deacon@....com>
Cc:     "linux-arm-kernel\@lists.infradead.org" 
        <linux-arm-kernel@...ts.infradead.org>,
        "linux-kernel\@vger.kernel.org" <linux-kernel@...r.kernel.org>,
        alexander.shishkin@...ux.intel.com
Subject: Re: BUG in "perf: Suppress AUX/OVERWRITE records"?

Ben Gainey <Ben.Gainey@....com> writes:

> Hi all
>
> Regarding commit 1627314fb54a33ebd23bd08f2e215eaed0f44712 "perf:
> Suppress AUX/OVERWRITE records", I have found that I no longer receive
> PERF_RECORD_AUX on context switch when collecting data from the arm_spe
> PMU driver. This is because, on context switch, the arm_spe driver
> calls perf_aux_output_end with `handle->aux_flags == 0`, failing the
> test added in this commit.
>
> This is a problem as it means when capturing data for multiple threads
> (using perf_event_open) where AUX data is written to a per-cpu buffer,
> I can no longer accurately attribute SPE AUX data to an individual
> thread.

Sounds like PERF_RECORD_SWITCH should be sufficient for your purposes,
have you considered that?

> If I read the intent of the commit as to remove OVERWRITE AUX records,
> then it seems the added if condition is incorrect and should probably
> be formulated as:
>
>     if ((handle->aux_flags & ~(u64)PERF_AUX_FLAG_OVERWRITE) || !handle-
>>aux_flags)
>
> Is this correct (and would you like a patch?), or is my use of
> PERF_RECORD_AUX incorrect in this case?

No, the point of AUX records is to communicate useful things about the
data in AUX buffer. It was an unintentional side effect that it also
happened to coincide with context switches in the overwrite mode.

Thanks,
--
Alex

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ