Re: [PATCH] perf probe: Fix failure to probe events on arm

[hekuang <hekuang@zoho.com>]

hi, Arnaldo

On 06/15/2015 10:49 PM, Arnaldo Carvalho de Melo wrote:
> Em Mon, Jun 15, 2015 at 08:06:53AM +0000, He Kuang escreveu:
>> Fix failure to probe events on arm, problem is introduced by commit
>> 5a51fcd1f30c ("perf probe: Skip kernel symbols which is out of
>> .text"). For some architectures, label '_etext' is not in the .text
>> section(in .notes section for arm/arm64). Label out of .text section is
>> not loaded as symbols and we got a zero value when look up its address,
>> which causes all events be wrongly skiped.
>>
>> This patch uses kernel map->end when failed to get the address of
>> '_etext' and fixes the problem.
> Masami, can't we always use map->end then? Can you please take a look at
> this patch and ack/nack it?
>
> - Arnaldo

I think _etext is more accurate than kernel map->end, because
__map_groups__fixup_end() is called at the end of
dso__load_sym(), which fixes map->end to next_map->start.

Comparative result as this:
   etext_addr=ffffffff819a1b85, map->end=ffffffff81ff1000.

So if possible, we should use _etext.

Thanks.
>   
>> Problem can be reproduced on arm as following:
>>
>>    # perf probe --add='generic_perform_write'
>>    generic_perform_write+0 is out of .text, skip it.
>>    Probe point 'generic_perform_write' not found.
>>      Error: Failed to add events. Reason: No such file or directory (Code: -2)
>>
>> After this patch:
>>
>>    # perf probe --add='generic_perform_write'
>>    Added new event:
>>      probe:generic_perform_write (on generic_perform_write)
>>
>>    You can now use it in all perf tools, such as:
>>
>>      perf record -e probe:generic_perform_write -aR sleep 1
>>
>> Signed-off-by: He Kuang <hekuang@huawei.com>
>> ---
>>   tools/perf/util/probe-event.c | 16 +++++++++++++++-
>>   1 file changed, 15 insertions(+), 1 deletion(-)
>>
>> diff --git a/tools/perf/util/probe-event.c b/tools/perf/util/probe-event.c
>> index daa24a2..ee26961 100644
>> --- a/tools/perf/util/probe-event.c
>> +++ b/tools/perf/util/probe-event.c
>> @@ -575,8 +575,22 @@ static int post_process_probe_trace_events(struct probe_trace_event *tevs,
>>   		pr_warning("Relocated base symbol is not found!\n");
>>   		return -EINVAL;
>>   	}
>> -	/* Get the address of _etext for checking non-probable text symbol */
>> +	/* Get the address of _etext for checking non-probable text symbol,
>> +	   for some architectures (e.g. arm, arm64), _etext is out of .text
>> +	   section and not loaded as symbols, use kernel map->end instead.
>> +	 */
>>   	etext_addr = kernel_get_symbol_address_by_name("_etext", false);
>> +	if (etext_addr == 0) {
>> +		struct map *map;
>> +
>> +		map = kernel_get_module_map(NULL);
>> +		if (!map) {
>> +			pr_err("Failed to get a map for kernel\n");
>> +			return -EINVAL;
>> +		}
>> +
>> +		etext_addr = map->end;
>> +	}
>>   
>>   	for (i = 0; i < ntevs; i++) {
>>   		if (tevs[i].point.address && !tevs[i].point.retprobe) {
>> -- 
>> 1.8.5.2
> --
> 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/
>