[PATCH] powerpc/perf/hv-gpci: Fix the logic to compute counter value from the hcall result buffer.
Nageswara Sastry
rnsastry at linux.ibm.com
Thu Aug 19 16:15:51 AEST 2021
On 13/08/21 1:51 pm, Kajol Jain wrote:
> H_GetPerformanceCounterInfo (0xF080) hcall returns the counter data in the
> result buffer. Result buffer has specific format defined in the PAPR
> specification. One of the field is counter offset and width of the counter
> data returned.
>
> Counter data are returned in a unsigned char array. To
> get the final counter data, these values should be left shifted
> byte at a time. But commit 220a0c609ad17 ("powerpc/perf: Add support
> for the hv gpci (get performance counter info) interface") made the
> shifting bitwise. Because of this, hcall counters values could end up
> in lower side, which messes the counter prev vs now calculation. This
> lead to huge counter value reporting
>
> [command]#: perf stat -e hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
> -C 0 -I 1000
> time counts unit events
> 1.000078854 18,446,744,073,709,535,232 hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
> 2.000213293 0 hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
> 3.000320107 0 hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
> 4.000428392 0 hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
> 5.000537864 0 hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
> 6.000649087 0 hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
> 7.000760312 0 hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
> 8.000865218 16,448 hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
> 9.000978985 18,446,744,073,709,535,232 hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
> 10.001088891 16,384 hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
> 11.001201435 0 hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
> 12.001307937 18,446,744,073,709,535,232 hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
>
> Patch here fixes the shifting logic to make is byte-wise with which no more the issue seen.
>
> Fixes: e4f226b1580b3 ("powerpc/perf/hv-gpci: Increase request buffer size")
> Reported-by: Nageswara R Sastry<rnsastry at linux.ibm.com>
> Signed-off-by: Kajol Jain <kjain at linux.ibm.com>
Tested-by: Nageswara R Sastry<rnsastry at linux.ibm.com>
Now not seeing huge numbers.
# perf stat -e
hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/ -C 0 -I 1000
# time counts unit events
1.001023931 26,624
hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
2.002176767 0
hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
3.003296382 0
hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
4.004385311 33,280
hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
> ---
> arch/powerpc/perf/hv-gpci.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/arch/powerpc/perf/hv-gpci.c b/arch/powerpc/perf/hv-gpci.c
> index d48413e28c39..c756228a081f 100644
> --- a/arch/powerpc/perf/hv-gpci.c
> +++ b/arch/powerpc/perf/hv-gpci.c
> @@ -175,7 +175,7 @@ static unsigned long single_gpci_request(u32 req, u32 starting_index,
> */
> count = 0;
> for (i = offset; i < offset + length; i++)
> - count |= arg->bytes[i] << (i - offset);
> + count |= (u64)(arg->bytes[i]) << ((length - 1 - (i - offset)) * 8);
>
> *value = count;
> out:
>
--
Thanks and Regards
R.Nageswara Sastry
More information about the Linuxppc-dev
mailing list