lkml.org 
[lkml]   [2021]   [Oct]   [11]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [PATCH 1/2] tools/perf: Add reverse_fn to handle branch_stack endian issue
From
Date

On 10/7/21 11:12 PM, Jiri Olsa wrote:
> On Thu, Sep 30, 2021 at 12:34:09PM +0530, Madhavan Srinivasan wrote:
>> branch_stack struct has bit field definition which
>> produces different bit ordering for big/little endian.
>> Because of this, when branch_stack sample is collected
>> in a BE system and viewed/reported in a LE system, bit
>> fields of the branch stack are not presented properly.
>> To address this issue, a evsel__reverse64_branch_stack_flags()
>> is defined and introduced in evsel__parse_sample.
>>
>> Signed-off-by: Madhavan Srinivasan <maddy@linux.ibm.com>
>> ---
>> tools/perf/util/evsel.c | 64 +++++++++++++++++++++++++++++++++++++++--
>> tools/perf/util/evsel.h | 5 ++++
>> 2 files changed, 67 insertions(+), 2 deletions(-)
>>
>> diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
>> index dbfeceb2546c..007be66b69a2 100644
>> --- a/tools/perf/util/evsel.c
>> +++ b/tools/perf/util/evsel.c
>> @@ -2221,6 +2221,46 @@ void __weak arch_perf_parse_sample_weight(struct perf_sample *data,
>> data->weight = *array;
>> }
>>
>> +u64 evsel__reverse64_branch_stack_flags(u64 value)
>> +{
>> + u64 new_val = 0;
>> +
>> + /*
>> + * branch_stack flag (u64)
>> + * union {
>> + * u64 values;
>> + * struct {
>> + * mispred:1 //target mispredicted
>> + * predicted:1 //target predicted
>> + * in_tx:1 //in transaction
>> + * abort:1 //transaction abort
>> + * cycles:16 //cycle count to last branch
>> + * type:4 //branch type
>> + * reserved:40
>> + * }
>> + * }
>> + */
> please describe in comment how the bitfield is swapped

Sure will do.

>> + if (bigendian()) {
>> + new_val = reverse_64(value, 0, 1);
>> + new_val |= reverse_64(value, 1, 1);
>> + new_val |= reverse_64(value, 2, 1);
>> + new_val |= reverse_64(value, 3, 1);
>> + new_val |= reverse_64(value, 4, 16);
>> + new_val |= reverse_64(value, 20, 4);
>> + new_val |= reverse_64(value, 24, 40);
>> + } else {
>> + new_val = reverse_64(value, 63, 1);
>> + new_val |= reverse_64(value, 62, 1);
>> + new_val |= reverse_64(value, 61, 1);
>> + new_val |= reverse_64(value, 60, 1);
>> + new_val |= reverse_64(value, 44, 16);
>> + new_val |= reverse_64(value, 40, 4);
>> + new_val |= reverse_64(value, 0, 40);
>> + }
>> +
>> + return new_val;
>> +}
>> +
>> int evsel__parse_sample(struct evsel *evsel, union perf_event *event,
>> struct perf_sample *data)
>> {
>> @@ -2408,6 +2448,8 @@ int evsel__parse_sample(struct evsel *evsel, union perf_event *event,
>> if (type & PERF_SAMPLE_BRANCH_STACK) {
>> const u64 max_branch_nr = UINT64_MAX /
>> sizeof(struct branch_entry);
>> + struct branch_entry *e;
>> + unsigned int i;
>>
>> OVERFLOW_CHECK_u64(array);
>> data->branch_stack = (struct branch_stack *)array++;
>> @@ -2416,10 +2458,28 @@ int evsel__parse_sample(struct evsel *evsel, union perf_event *event,
>> return -EFAULT;
>>
>> sz = data->branch_stack->nr * sizeof(struct branch_entry);
>> - if (evsel__has_branch_hw_idx(evsel))
>> + if (evsel__has_branch_hw_idx(evsel)) {
>> sz += sizeof(u64);
>> - else
>> + e = &data->branch_stack->entries[0];
>> + } else {
>> data->no_hw_idx = true;
>> + e = (struct branch_entry *)&data->branch_stack->hw_idx;
> hum, why do we convert hw_idx? it's the same struct as entries?
> please explain this in comment as well

No. IIUC, hw_idx is valid only if PERF_SAMPLE_BRANCH_HW_INDEX is applied.
If not, then only nr and entries[] will be output-ed by kernel
42bbabed09ce ('perf tools: Add hw_idx in struct branch_stack')

>
>> + }
>> +
>> + if (swapped) {
>> + /*
>> + * struct branch_flag does not have endian specific
>> + * bit field definition. And bswap will not resolve the
>> + * issue, since these are bit fields.
>> + *
>> + * evsel__reverse64_branch_stack_flags() uses a reverse64
>> + * macro to reverse the bit position based on the host
>> + * endians.
>> + */
>> + for (i = 0; i < data->branch_stack->nr; i++, e++)
>> + e->flags.value = evsel__reverse64_branch_stack_flags(e->flags.value);
>> + }
>> +
>> OVERFLOW_CHECK(array, sz, max_size);
>> array = (void *)array + sz;
>> }
>> diff --git a/tools/perf/util/evsel.h b/tools/perf/util/evsel.h
>> index 1f7edfa8568a..1127c23710cf 100644
>> --- a/tools/perf/util/evsel.h
>> +++ b/tools/perf/util/evsel.h
>> @@ -482,4 +482,9 @@ struct evsel *evsel__leader(struct evsel *evsel);
>> bool evsel__has_leader(struct evsel *evsel, struct evsel *leader);
>> bool evsel__is_leader(struct evsel *evsel);
>> void evsel__set_leader(struct evsel *evsel, struct evsel *leader);
>> +
>> +#define reverse_64(src, pos, size) \
>> + ((((src) >> (pos)) & ((1ull << (size)) - 1)) << (63 - ((pos) + (size) - 1)))
> hum, is this reversing anything?
> could you please add comment describing what this is doing?

Sure will do.
Thanks for review
Maddy


> thanks, jirka
>> +
>> +u64 evsel__reverse64_branch_stack_flags(u64 value);
>> #endif /* __PERF_EVSEL_H */
>> --
>> 2.31.1
>>

\
 
 \ /
  Last update: 2021-10-11 16:31    [W:0.080 / U:0.276 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site