Re: [PATCH v2] perf: disable PERF_RECORD_MMAP2 support

[David Ahern <dsahern@gmail.com>]

On 10/17/13 9:57 AM, Stephane Eranian wrote:
> On Thu, Oct 17, 2013 at 5:54 PM, David Ahern <dsahern@gmail.com> wrote:
>> On 10/17/13 8:28 AM, Stephane Eranian wrote:
>>>
>>>
>>> For now, we disable the extended MMAP record support (MMAP2).
>>> We have identified cases where it would not report the correct
>>> mapping information, clone(VM_CLONE) but with separate pids.
>>> We will revisit the support once we find a solution for this case.
>>>
>>> The patch changes the kernel to return EINVAL if attr->mmap2
>>> is set. The patch also modifies the perf tool to use regular
>>> PERF_RECORD_MMAP
>>> for synthetic events and it also prevents the tool from requesting
>>> attr->mmap2
>>> mode because the kernel would reject it.
>>>
>>> The support will be revisited once the kenrel interface is updated.
>>
>>
>> Why not disable mmap2 as well:
>>
>> diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
>> index 0ce9febf1ba0..289f34dbe970 100644
>> --- a/tools/perf/util/evsel.c
>> +++ b/tools/perf/util/evsel.c
>> @@ -678,7 +678,7 @@ void perf_evsel__config(struct perf_evsel *evsel,
>>          attr->sample_type   |= PERF_SAMPLE_WEIGHT;
>>
>>      attr->mmap  = track;
>> -   attr->mmap2 = track && !perf_missing_features.mmap2;
>> +   /* attr->mmap2 = track && !perf_missing_features.mmap2; */
>>      attr->comm  = track;
>>
> It is disabled automatically later on. I checked that. Wanted to minimize
> the changes.
>

Understood that the fallback kicks in. Why fail the perf_event_open 
every single time when we know it is not going to succeed?

David