Re: [PATCH 1/4] perf tools: unwind: ensure unwind hooks return negative errorno.

[Wang Nan <wangnan0@huawei.com>]

Hi folks,

I'm rethinking --map-adjustment now, and I believe what we need should be something
like 'perf inject', which allows us to inject fake mmap events into perf.data to
make 'perf report' believe some //anon memory are file based mapping. Patch 2/4 - 4/4
seem not useful now. However, patch 1/4 is still useful because it is a bugfix. Could
you please drop the other 3 patches and merge this one?

Thank you.

On 2015/4/1 20:41, Wang Nan wrote:
> On 2015/4/1 20:12, Jiri Olsa wrote:
>> On Wed, Apr 01, 2015 at 10:33:12AM +0000, Wang Nan wrote:
>>> According to man pages of libunwind, unwind hooks should return
>>> 'negative value of one of the unw_error_t error-codes', they are
>>> different from generic error code. In addition, access_dso_mem()
>>> returns '!(size == sizeof(*data))', compiler never ensure it is
>>> negative when failure, which causes libunwind get undesire value
>>> when accessing //anon memory.
>>>
>>> This patch fixes this problem by force returning negative value when
>>> error, instead of returning 'ret' itself when it is non-zero.
>>
>> hum, how about find_proc_info callback.. should it follow the same rules?
>>
> 
> Yes, but it only returns -EINVAL and dwarf_search_unwind_table(....). The latter
> one is part of libunwind so we can trust it returns negative when fail.
> 
>> thanks,
>> jirka
>>
> 
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
>