linux-kernel - Re: [PATCH] perf tools: Add missing Intel CPU events to parser

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: <fc3c4dee-981e-4c39-566a-4163ee0bcc02@intel.com>
Date:   Wed, 25 Mar 2020 19:10:44 +0200
From:   Adrian Hunter <adrian.hunter@...el.com>
To:     Jiri Olsa <jolsa@...hat.com>
Cc:     Arnaldo Carvalho de Melo <arnaldo.melo@...il.com>,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH] perf tools: Add missing Intel CPU events to parser

On 25/03/20 5:22 pm, Jiri Olsa wrote:
> On Wed, Mar 25, 2020 at 04:24:58PM +0200, Adrian Hunter wrote:
>> On 25/03/20 4:22 pm, Arnaldo Carvalho de Melo wrote:
>>> Em Wed, Mar 25, 2020 at 02:53:50PM +0100, Jiri Olsa escreveu:
>>>> On Wed, Mar 25, 2020 at 10:15:49AM -0300, Arnaldo Carvalho de Melo wrote:
>>>>> Em Wed, Mar 25, 2020 at 11:33:45AM +0100, Jiri Olsa escreveu:
>>>>>> On Tue, Mar 24, 2020 at 05:04:43PM +0200, Adrian Hunter wrote:
>>>>>>> perf list expects CPU events to be parseable by name, e.g.
>>>>>
>>>>>>>     # perf list | grep el-capacity-read
>>>>>>>       el-capacity-read OR cpu/el-capacity-read/          [Kernel PMU event]
>>>>>
>>>>>>> But the event parser does not recognize them that way, e.g.
>>>>>
>>>>>>>     # perf test -v "Parse event"
>>>>>>>     <SNIP>
>>>>>>>     running test 54 'cycles//u'
>>>>>>>     running test 55 'cycles:k'
>>>>>>>     running test 0 'cpu/config=10,config1,config2=3,period=1000/u'
>>>>>>>     running test 1 'cpu/config=1,name=krava/u,cpu/config=2/u'
>>>>>>>     running test 2 'cpu/config=1,call-graph=fp,time,period=100000/,cpu/config=2,call-graph=no,time=0,period=2000/'
>>>>>>>     running test 3 'cpu/name='COMPLEX_CYCLES_NAME:orig=cycles,desc=chip-clock-ticks',period=0x1,event=0x2/ukp'
>>>>>>>     -> cpu/event=0,umask=0x11/
>>>>>>>     -> cpu/event=0,umask=0x13/
>>>>>>>     -> cpu/event=0x54,umask=0x1/
>>>>>>>     failed to parse event 'el-capacity-read:u,cpu/event=el-capacity-read/u', err 1, str 'parser error'
>>>>>>>     event syntax error: 'el-capacity-read:u,cpu/event=el-capacity-read/u'
>>>>>>>                            \___ parser error test child finished with 1
>>>>>>>     ---- end ----
>>>>>>>     Parse event definition strings: FAILED!
>>>>>
>>>>>>> Fix by adding missing Intel CPU events to the event parser.
>>>>>>> Missing events were found by using:
>>>>>
>>>>>>>     grep -r EVENT_ATTR_STR arch/x86/events/intel/core.c
>>>>>>>
>>>>>>> Signed-off-by: Adrian Hunter <adrian.hunter@...el.com>
>>>>>>
>>>>>> Acked-by: Jiri Olsa <jolsa@...hat.com>
>>>>>
>>>>> So, I'm not being able to reproduce this, what an I missing?
>>>>
>>>> I think you need to be on some really recent intel
>>>> which defines events which we did not covered yet
>>>> like el-capacity-write in icelake
>>>
>>> That is why I tried with el-capacity, which is moved to the parser as
>>> well, I've replaced el-capacity-read, which I don't have in this Kaby
>>> Lake machine, with el-capacity, that is present:
>>>
>>> [root@...enth ~]# perf list | grep el-capacity
>>>   el-capacity OR cpu/el-capacity/                    [Kernel PMU event]
>>> [root@...enth ~]#
>>
>> I just checked that and it seems to be a "feature" of the parser that it
>> gets confused between el-capacity and el-capacity-read.
>>
>> Making them explicit in parse-events.l makes the problem go away, but I
>> wonder now if the parser could be better in this regard.
> 
> so we have that PRE/SUFFIX logic that allows us
> to specify any sysfs event term as standalone event
> 
> the lexer in this case below was used to handle special cases..
> and IIUC think having more than one '-' is one of them
> 
> could you check if the patch below will fix that for you?
> 
> jirka
> 
> 
> ---
> diff --git a/tools/perf/util/parse-events.l b/tools/perf/util/parse-events.l
> index 7b1c8ee537cf..347eb3e6794a 100644
> --- a/tools/perf/util/parse-events.l
> +++ b/tools/perf/util/parse-events.l
> @@ -342,11 +342,15 @@ bpf-output					{ return sym(yyscanner, PERF_TYPE_SOFTWARE, PERF_COUNT_SW_BPF_OUT
>  	 * Because the prefix cycles is mixed up with cpu-cycles.
>  	 * loads and stores are mixed up with cache event
>  	 */
> -cycles-ct					{ return str(yyscanner, PE_KERNEL_PMU_EVENT); }
> -cycles-t					{ return str(yyscanner, PE_KERNEL_PMU_EVENT); }
> -mem-loads					{ return str(yyscanner, PE_KERNEL_PMU_EVENT); }
> -mem-stores					{ return str(yyscanner, PE_KERNEL_PMU_EVENT); }
> -topdown-[a-z-]+					{ return str(yyscanner, PE_KERNEL_PMU_EVENT); }
> +cycles-ct				|
> +cycles-t				|
> +mem-loads				|
> +mem-stores				|
> +topdown-[a-z-]+				|
> +tx-capacity-read			|
> +tx-capacity-write			|
> +el-capacity-read			|
> +el-capacity-write			{ return str(yyscanner, PE_KERNEL_PMU_EVENT); }

Yes that works, as does

tx-capacity-[a-z-]+
el-capacity-[a-z-]+

Do we have an explanation for why we cannot make it accept 3-part names
without handling them as special cases?

I just tried but it didn't work.