linux-kernel - Re: [PATCH v6 10/12] perf tools: Improve IBS error handling

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: <ec3ce0b5-00bf-6deb-1b71-7a215104099d@amd.com>
Date:   Wed, 16 Mar 2022 16:46:56 +0530
From:   Ravi Bangoria <ravi.bangoria@....com>
To:     Stephane Eranian <eranian@...gle.com>
Cc:     peterz@...radead.org, kim.phillips@....com, acme@...hat.com,
        jolsa@...hat.com, songliubraving@...com,
        linux-kernel@...r.kernel.org, Ravi Bangoria <ravi.bangoria@....com>
Subject: Re: [PATCH v6 10/12] perf tools: Improve IBS error handling



On 16-Mar-22 4:37 PM, Ravi Bangoria wrote:
> Stephane,
> 
> On 16-Mar-22 5:33 AM, Stephane Eranian wrote:
>> On Tue, Mar 15, 2022 at 12:46 AM Ravi Bangoria <ravi.bangoria@....com> wrote:
>>>
>>> Stephane,
>>>
>>> On 09-Feb-22 2:46 AM, Stephane Eranian wrote:
>>>> From: Kim Phillips <kim.phillips@....com>
>>>>
>>>> improve the error message returned on failed perf_event_open() on AMD when
>>>> using IBS.
>>>>
>>>> Output of executing 'perf record -e ibs_op// true' BEFORE this patch:
>>>>
>>>> The sys_perf_event_open() syscall returned with 22 (Invalid argument)for event (ibs_op//u).
>>>> /bin/dmesg | grep -i perf may provide additional information.
>>>>
>>>> Output after:
>>>>
>>>> AMD IBS cannot exclude kernel events.  Try running at a higher privilege level.
>>>>
>>>> Output of executing 'sudo perf record -e ibs_op// true' BEFORE this patch:
>>>>
>>>> Error:
>>>> The sys_perf_event_open() syscall returned with 22 (Invalid argument) for event (ibs_op//).
>>>> /bin/dmesg | grep -i perf may provide additional information.
>>>>
>>>> Output after:
>>>>
>>>> Error:
>>>> AMD IBS may only be available in system-wide/per-cpu mode.  Try using -a, or -C and workload affinity
>>>
>>> This patch seems to be causing regression to perf python test.
>>>
>>> Without this patch:
>>>
>>>   $ sudo ./perf test -vvv 19
>>>    19: 'import perf' in python                                         :
>>>   --- start ---
>>>   test child forked, pid 145391
>>>   python usage test: "echo "import sys ; sys.path.append('python'); import perf" | '/usr/bin/python2' "
>>>   test child finished with 0
>>>   ---- end ----
>>>   'import perf' in python: Ok
>>>
>>> With this patch:
>>>
>>>   $ sudo ./perf test -vvv 19
>>>    19: 'import perf' in python                                         :
>>>   --- start ---
>>>   test child forked, pid 144415
>>>   python usage test: "echo "import sys ; sys.path.append('python'); import perf" | '/usr/bin/python2' "
>>>   Traceback (most recent call last):
>>>     File "<stdin>", line 1, in <module>
>>>   ImportError: python/perf.so: undefined symbol: perf_env__cpuid
>>>   test child finished with -1
>>>   ---- end ----
>>>   'import perf' in python: FAILED!
>>>
>>
>> The fix I sent you is just to prevent a potential SEGFAULT in
>> is_amd_ibs(). I bet the test fails some perf_event_open()
>> and ends up in strerror function and from there I don't see how the
>> patch could impact the test, given you'd segfault
>> otherwise.
>>
>> I tried on my side and with or without this patch this test fails. I
>> think this looks like an unrelated issue.
> 
> That's strange. IIUC, python/perf.so needs to know about util/evsel.c

I meant util/env.c not util/evsel.c

> which can be done by adding an entry in util/python-ext-sources.
> 
> In any case, do we really need this patch now? For both the example given
> in description, I see no difference with or without patch:
> 
> Without patch:
> 
>   $ sudo ./perf record -e ibs_op// true
>   Error:
>   Invalid event (ibs_op//) in per-thread mode, enable system wide with '-a'.
> 
>   $ ./perf record -e ibs_op// true
>   Error:
>   Invalid event (ibs_op//u) in per-thread mode, enable system wide with '-a'.
> 
> Same o/p with the patch as well.
> 
> Thanks,
> Ravi