perf intel-pt: Intel PT fixes
Adrian Hunter
reviewercover perf intel-pt: Intel PT fixes
Hi
Here are 3 non-urgent fixes for Intel PT.
Adrian Hunter (3):
perf intel-pt: Fix instructions sampling rate perf intel-pt: Fix improved sample timestamp perf intel-pt: Fix sample timestamp wrt non-taken branches
.../perf/util/intel-pt-decoder/intel-pt-decoder.c | 31 +++++++++++++++++----- 1 file changed, 24 insertions(+), 7 deletions(-)
Regards
Adrian
Adrian Hunter
reviewerPATCH 1/3 perf intel-pt: Fix instructions sampling rate
The timestamp used to determine if an instruction sample is made, is an estimate based on the number of instructions since the last known timestamp. A consequence is that it might go backwards, which results in extra samples. Change it so that a sample is only made when the timestamp goes forwards.
Note this does not affect a sampling period of 0 or sampling periods specified as a count of instructions.
Example:
Before:
$ perf script --itrace=i10us ls 13812 [003] 2167315.222583: 3270 instructions:u: 7fac71e2e494 __GI___tunables_init+0xf4 (/lib/x86_64-linux-gnu/ld-2.28.so) ls 13812 [003] 2167315.222667: 30902 instructions:u: 7fac71e2da0f _dl_cache_libcmp+0x2f (/lib/x86_64-linux-gnu/ld-2.28.so) ls 13812 [003] 2167315.222667: 10 instructions:u: 7fac71e2d9ff _dl_cache_libcmp+0x1f (/lib/x86_64-linux-gnu/ld-2.28.so) ls 13812 [003] 2167315.222667: 8 instructions:u: 7fac71e2d9ea _dl_cache_libcmp+0xa (/lib/x86_64-linux-gnu/ld-2.28.so) ls 13812 [003] 2167315.222667: 14 instructions:u: 7fac71e2d9ea _dl_cache_libcmp+0xa (/lib/x86_64-linux-gnu/ld-2.28.so) ls 13812 [003] 2167315.222667: 6 instructions:u: 7fac71e2d9ff _dl_cache_libcmp+0x1f (/lib/x86_64-linux-gnu/ld-2.28.so) ls 13812 [003] 2167315.222667: 14 instructions:u: 7fac71e2d9ff _dl_cache_libcmp+0x1f (/lib/x86_64-linux-gnu/ld-2.28.so) ls 13812 [003] 2167315.222667: 4 instructions:u: 7fac71e2dab2 _dl_cache_libcmp+0xd2 (/lib/x86_64-linux-gnu/ld-2.28.so) ls 13812 [003] 2167315.222728: 16423 instructions:u: 7fac71e2477a _dl_map_object_deps+0x1ba (/lib/x86_64-linux-gnu/ld-2.28.so) ls 13812 [003] 2167315.222734: 12731 instructions:u: 7fac71e27938 _dl_name_match_p+0x68 (/lib/x86_64-linux-gnu/ld-2.28.so) ...
After:
$ perf script --itrace=i10us ls 13812 [003] 2167315.222583: 3270 instructions:u: 7fac71e2e494 __GI___tunables_init+0xf4 (/lib/x86_64-linux-gnu/ld-2.28.so) ls 13812 [003] 2167315.222667: 30902 instructions:u: 7fac71e2da0f _dl_cache_libcmp+0x2f (/lib/x86_64-linux-gnu/ld-2.28.so) ls 13812 [003] 2167315.222728: 16479 instructions:u: 7fac71e2477a _dl_map_object_deps+0x1ba (/lib/x86_64-linux-gnu/ld-2.28.so) ...
--- tools/perf/util/intel-pt-decoder/intel-pt-decoder.c | 13 ++++++++++--- 1 file changed, 10 insertions(+), 3 deletions(-)
tools/perf/util/intel-pt-decoder/intel-pt-decoder.c +10 -3
@@ -888,16 +888,20 @@ static uint64_t intel_pt_next_period(struct intel_pt_decoder *decoder) timestamp = decoder->timestamp + decoder->timestamp_insn_cnt; masked_timestamp = timestamp & decoder->period_mask;if (decoder->continuous_period) { -if (masked_timestamp != decoder->last_masked_timestamp) +if (masked_timestamp > decoder->last_masked_timestamp)return 1 ; }else { timestamp +=1 ; masked_timestamp = timestamp & decoder->period_mask; -if (masked_timestamp != decoder->last_masked_timestamp) { +if (masked_timestamp > decoder->last_masked_timestamp) { decoder->last_masked_timestamp = masked_timestamp; decoder->continuous_period = true; } } + +if (masked_timestamp < decoder->last_masked_timestamp) +return decoder->period_ticks; +return decoder->period_ticks - (timestamp - masked_timestamp); }
@@ -926,7 +930,10 @@ static void intel_pt_sample_insn(struct intel_pt_decoder *decoder)case INTEL_PT_PERIOD_TICKS : timestamp = decoder->timestamp + decoder->timestamp_insn_cnt; masked_timestamp = timestamp & decoder->period_mask; - decoder->last_masked_timestamp = masked_timestamp; +if (masked_timestamp > decoder->last_masked_timestamp) + decoder->last_masked_timestamp = masked_timestamp; +else + decoder->last_masked_timestamp += decoder->period_ticks;break ;case INTEL_PT_PERIOD_NONE :case INTEL_PT_PERIOD_MTC :
Signature · 1 line
2.17.1
tip-bot for Adrian Hunter
merge bot[tip:perf/core] perf intel-pt: Fix instructions sampling rate
perf intel-pt: Fix instructions sampling rate
Commit-ID: 7ba8fa20e26eb3c0c04d747f7fd2223694eac4d5 Gitweb: https://git.kernel.org/tip/7ba8fa20e26eb3c0c04d747f7fd2223694eac4d5 Author: Adrian Hunter <[email protected]> AuthorDate: Fri, 10 May 2019 15:41:41 +0300 Committer: Arnaldo Carvalho de Melo <[email protected]> CommitDate: Thu, 16 May 2019 14:17:23 -0300
perf intel-pt: Fix instructions sampling rate
The timestamp used to determine if an instruction sample is made, is an estimate based on the number of instructions since the last known timestamp. A consequence is that it might go backwards, which results in extra samples. Change it so that a sample is only made when the timestamp goes forwards.
Note this does not affect a sampling period of 0 or sampling periods specified as a count of instructions.
Example:
Before:
$ perf script --itrace=i10us ls 13812 [003] 2167315.222583: 3270 instructions:u: 7fac71e2e494 __GI___tunables_init+0xf4 (/lib/x86_64-linux-gnu/ld-2.28.so) ls 13812 [003] 2167315.222667: 30902 instructions:u: 7fac71e2da0f _dl_cache_libcmp+0x2f (/lib/x86_64-linux-gnu/ld-2.28.so) ls 13812 [003] 2167315.222667: 10 instructions:u: 7fac71e2d9ff _dl_cache_libcmp+0x1f (/lib/x86_64-linux-gnu/ld-2.28.so) ls 13812 [003] 2167315.222667: 8 instructions:u: 7fac71e2d9ea _dl_cache_libcmp+0xa (/lib/x86_64-linux-gnu/ld-2.28.so) ls 13812 [003] 2167315.222667: 14 instructions:u: 7fac71e2d9ea _dl_cache_libcmp+0xa (/lib/x86_64-linux-gnu/ld-2.28.so) ls 13812 [003] 2167315.222667: 6 instructions:u: 7fac71e2d9ff _dl_cache_libcmp+0x1f (/lib/x86_64-linux-gnu/ld-2.28.so) ls 13812 [003] 2167315.222667: 14 instructions:u: 7fac71e2d9ff _dl_cache_libcmp+0x1f (/lib/x86_64-linux-gnu/ld-2.28.so) ls 13812 [003] 2167315.222667: 4 instructions:u: 7fac71e2dab2 _dl_cache_libcmp+0xd2 (/lib/x86_64-linux-gnu/ld-2.28.so) ls 13812 [003] 2167315.222728: 16423 instructions:u: 7fac71e2477a _dl_map_object_deps+0x1ba (/lib/x86_64-linux-gnu/ld-2.28.so) ls 13812 [003] 2167315.222734: 12731 instructions:u: 7fac71e27938 _dl_name_match_p+0x68 (/lib/x86_64-linux-gnu/ld-2.28.so) ...
After:
$ perf script --itrace=i10us ls 13812 [003] 2167315.222583: 3270 instructions:u: 7fac71e2e494 __GI___tunables_init+0xf4 (/lib/x86_64-linux-gnu/ld-2.28.so) ls 13812 [003] 2167315.222667: 30902 instructions:u: 7fac71e2da0f _dl_cache_libcmp+0x2f (/lib/x86_64-linux-gnu/ld-2.28.so) ls 13812 [003] 2167315.222728: 16479 instructions:u: 7fac71e2477a _dl_map_object_deps+0x1ba (/lib/x86_64-linux-gnu/ld-2.28.so) ...
--- tools/perf/util/intel-pt-decoder/intel-pt-decoder.c | 13 ++++++++++--- 1 file changed, 10 insertions(+), 3 deletions(-)
tools/perf/util/intel-pt-decoder/intel-pt-decoder.c +10 -3
@@ -888,16 +888,20 @@ static uint64_t intel_pt_next_period(struct intel_pt_decoder *decoder) timestamp = decoder->timestamp + decoder->timestamp_insn_cnt; masked_timestamp = timestamp & decoder->period_mask;if (decoder->continuous_period) { -if (masked_timestamp != decoder->last_masked_timestamp) +if (masked_timestamp > decoder->last_masked_timestamp)return 1 ; }else { timestamp +=1 ; masked_timestamp = timestamp & decoder->period_mask; -if (masked_timestamp != decoder->last_masked_timestamp) { +if (masked_timestamp > decoder->last_masked_timestamp) { decoder->last_masked_timestamp = masked_timestamp; decoder->continuous_period = true; } } + +if (masked_timestamp < decoder->last_masked_timestamp) +return decoder->period_ticks; +return decoder->period_ticks - (timestamp - masked_timestamp); }
@@ -926,7 +930,10 @@ static void intel_pt_sample_insn(struct intel_pt_decoder *decoder)case INTEL_PT_PERIOD_TICKS : timestamp = decoder->timestamp + decoder->timestamp_insn_cnt; masked_timestamp = timestamp & decoder->period_mask; - decoder->last_masked_timestamp = masked_timestamp; +if (masked_timestamp > decoder->last_masked_timestamp) + decoder->last_masked_timestamp = masked_timestamp; +else + decoder->last_masked_timestamp += decoder->period_ticks;break ;case INTEL_PT_PERIOD_NONE :case INTEL_PT_PERIOD_MTC :
Adrian Hunter
reviewerPATCH 2/3 perf intel-pt: Fix improved sample timestamp
In reply to Adrian HunterThe decoder uses its current timestamp in samples. Usually that is a timestamp that has already passed, but in some cases it is a timestamp for a branch that the decoder is walking towards, and consequently hasn't reached. The intel_pt_sample_time() function decides which is which, but was not handling TNT packets exactly correctly. In the case of TNT, the timestamp applies to the first branch, so the decoder must first walk to that branch. That means intel_pt_sample_time() should return true for TNT, and this patch makes that change. However, if the first branch is a non-taken branch (i.e. a 'N'), then intel_pt_sample_time() needs to return false for subsequent taken branches in the same TNT packet. To handle that, introduce a new state INTEL_PT_STATE_TNT_CONT to distinguish the cases.
Note that commit 3f04d98e972b5 ("perf intel-pt: Improve sample timestamp") was also a stable fix and appears, for example, in v4.4 stable tree as commit a4ebb58fd124 ("perf intel-pt: Improve sample timestamp").
--- tools/perf/util/intel-pt-decoder/intel-pt-decoder.c | 13 ++++++++++--- 1 file changed, 10 insertions(+), 3 deletions(-)
tools/perf/util/intel-pt-decoder/intel-pt-decoder.c +10 -3
@@ -58,6 +58,7 @@ enum intel_pt_pkt_state {INTEL_PT_STATE_NO_IP ,INTEL_PT_STATE_ERR_RESYNC ,INTEL_PT_STATE_IN_SYNC , +INTEL_PT_STATE_TNT_CONT ,INTEL_PT_STATE_TNT ,INTEL_PT_STATE_TIP , INTEL_PT_STATE_TIP_PGD,
@@ -72,8 +73,9 @@ static inline bool intel_pt_sample_time(enum intel_pt_pkt_state pkt_state)case INTEL_PT_STATE_NO_IP :case INTEL_PT_STATE_ERR_RESYNC :case INTEL_PT_STATE_IN_SYNC : -case INTEL_PT_STATE_TNT : +case INTEL_PT_STATE_TNT_CONT :return true; +case INTEL_PT_STATE_TNT :case INTEL_PT_STATE_TIP :case INTEL_PT_STATE_TIP_PGD :case INTEL_PT_STATE_FUP :
@@ -1261,7 +1263,9 @@ static int intel_pt_walk_tnt(struct intel_pt_decoder *decoder)return -ENOENT ; } decoder->tnt.count -=1 ; -if (!decoder->tnt.count) +if (decoder->tnt.count) + decoder->pkt_state =INTEL_PT_STATE_TNT_CONT ; +else decoder->pkt_state =INTEL_PT_STATE_IN_SYNC ; decoder->tnt.payload <<=1 ; decoder->state.from_ip = decoder->ip;
@@ -1292,7 +1296,9 @@ static int intel_pt_walk_tnt(struct intel_pt_decoder *decoder)if (intel_pt_insn.branch ==INTEL_PT_BR_CONDITIONAL ) { decoder->tnt.count -=1 ; -if (!decoder->tnt.count) +if (decoder->tnt.count) + decoder->pkt_state =INTEL_PT_STATE_TNT_CONT ; +else decoder->pkt_state =INTEL_PT_STATE_IN_SYNC ;if (decoder->tnt.payload &BIT63 ) { decoder->tnt.payload <<=1 ;
@@ -2372,6 +2378,7 @@ const struct intel_pt_state *intel_pt_decode(struct intel_pt_decoder *decoder) err =intel_pt_walk_trace (decoder);break ;case INTEL_PT_STATE_TNT : +case INTEL_PT_STATE_TNT_CONT : err =intel_pt_walk_tnt (decoder);if (err == -EAGAIN ) err =intel_pt_walk_trace (decoder);
Signature · 1 line
2.17.1
tip-bot for Adrian Hunter
merge bot[tip:perf/core] perf intel-pt: Fix improved sample timestamp
perf intel-pt: Fix improved sample timestamp
Commit-ID: 61b6e08dc8e3ea80b7485c9b3f875ddd45c8466b Gitweb: https://git.kernel.org/tip/61b6e08dc8e3ea80b7485c9b3f875ddd45c8466b Author: Adrian Hunter <[email protected]> AuthorDate: Fri, 10 May 2019 15:41:42 +0300 Committer: Arnaldo Carvalho de Melo <[email protected]> CommitDate: Thu, 16 May 2019 14:17:23 -0300
perf intel-pt: Fix improved sample timestamp
The decoder uses its current timestamp in samples. Usually that is a timestamp that has already passed, but in some cases it is a timestamp for a branch that the decoder is walking towards, and consequently hasn't reached.
The intel_pt_sample_time() function decides which is which, but was not handling TNT packets exactly correctly.
In the case of TNT, the timestamp applies to the first branch, so the decoder must first walk to that branch.
That means intel_pt_sample_time() should return true for TNT, and this patch makes that change. However, if the first branch is a non-taken branch (i.e. a 'N'), then intel_pt_sample_time() needs to return false for subsequent taken branches in the same TNT packet.
To handle that, introduce a new state INTEL_PT_STATE_TNT_CONT to distinguish the cases.
Note that commit 3f04d98e972b5 ("perf intel-pt: Improve sample timestamp") was also a stable fix and appears, for example, in v4.4 stable tree as commit a4ebb58fd124 ("perf intel-pt: Improve sample timestamp").
--- tools/perf/util/intel-pt-decoder/intel-pt-decoder.c | 13 ++++++++++--- 1 file changed, 10 insertions(+), 3 deletions(-)
tools/perf/util/intel-pt-decoder/intel-pt-decoder.c +10 -3
@@ -58,6 +58,7 @@ enum intel_pt_pkt_state {INTEL_PT_STATE_NO_IP ,INTEL_PT_STATE_ERR_RESYNC ,INTEL_PT_STATE_IN_SYNC , +INTEL_PT_STATE_TNT_CONT ,INTEL_PT_STATE_TNT ,INTEL_PT_STATE_TIP , INTEL_PT_STATE_TIP_PGD,
@@ -72,8 +73,9 @@ static inline bool intel_pt_sample_time(enum intel_pt_pkt_state pkt_state)case INTEL_PT_STATE_NO_IP :case INTEL_PT_STATE_ERR_RESYNC :case INTEL_PT_STATE_IN_SYNC : -case INTEL_PT_STATE_TNT : +case INTEL_PT_STATE_TNT_CONT :return true; +case INTEL_PT_STATE_TNT :case INTEL_PT_STATE_TIP :case INTEL_PT_STATE_TIP_PGD :case INTEL_PT_STATE_FUP :
@@ -1261,7 +1263,9 @@ static int intel_pt_walk_tnt(struct intel_pt_decoder *decoder)return -ENOENT ; } decoder->tnt.count -=1 ; -if (!decoder->tnt.count) +if (decoder->tnt.count) + decoder->pkt_state =INTEL_PT_STATE_TNT_CONT ; +else decoder->pkt_state =INTEL_PT_STATE_IN_SYNC ; decoder->tnt.payload <<=1 ; decoder->state.from_ip = decoder->ip;
@@ -1292,7 +1296,9 @@ static int intel_pt_walk_tnt(struct intel_pt_decoder *decoder)if (intel_pt_insn.branch ==INTEL_PT_BR_CONDITIONAL ) { decoder->tnt.count -=1 ; -if (!decoder->tnt.count) +if (decoder->tnt.count) + decoder->pkt_state =INTEL_PT_STATE_TNT_CONT ; +else decoder->pkt_state =INTEL_PT_STATE_IN_SYNC ;if (decoder->tnt.payload &BIT63 ) { decoder->tnt.payload <<=1 ;
@@ -2372,6 +2378,7 @@ const struct intel_pt_state *intel_pt_decode(struct intel_pt_decoder *decoder) err =intel_pt_walk_trace (decoder);break ;case INTEL_PT_STATE_TNT : +case INTEL_PT_STATE_TNT_CONT : err =intel_pt_walk_tnt (decoder);if (err == -EAGAIN ) err =intel_pt_walk_trace (decoder);
Adrian Hunter
reviewerPATCH 3/3 perf intel-pt: Fix sample timestamp wrt non-taken branches
In reply to Adrian HunterThe sample timestamp is updated to ensure that the timestamp represents the time of the sample and not a branch that the decoder is still walking towards. The sample timestamp is updated when the decoder returns, but the decoder does not return for non-taken branches. Update the sample timestamp then also.
Note that commit 3f04d98e972b5 ("perf intel-pt: Improve sample timestamp") was also a stable fix and appears, for example, in v4.4 stable tree as commit a4ebb58fd124 ("perf intel-pt: Improve sample timestamp").
--- tools/perf/util/intel-pt-decoder/intel-pt-decoder.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)
tools/perf/util/intel-pt-decoder/intel-pt-decoder.c +4 -1
@@ -1318,8 +1318,11 @@ static int intel_pt_walk_tnt(struct intel_pt_decoder *decoder)return 0 ; } decoder->ip += intel_pt_insn.length; -if (!decoder->tnt.count) +if (!decoder->tnt.count) { + decoder->sample_timestamp = decoder->timestamp; + decoder->sample_insn_cnt = decoder->timestamp_insn_cnt;return -EAGAIN ; + } decoder->tnt.payload <<=1 ;continue ; }
Signature · 1 line
2.17.1
tip-bot for Adrian Hunter
merge bot[tip:perf/core] perf intel-pt: Fix sample timestamp wrt non-taken branches
perf intel-pt: Fix sample timestamp wrt non-taken branches
Commit-ID: 1b6599a9d8e6c9f7e9b0476012383b1777f7fc93 Gitweb: https://git.kernel.org/tip/1b6599a9d8e6c9f7e9b0476012383b1777f7fc93 Author: Adrian Hunter <[email protected]> AuthorDate: Fri, 10 May 2019 15:41:43 +0300 Committer: Arnaldo Carvalho de Melo <[email protected]> CommitDate: Thu, 16 May 2019 14:17:24 -0300
perf intel-pt: Fix sample timestamp wrt non-taken branches
The sample timestamp is updated to ensure that the timestamp represents the time of the sample and not a branch that the decoder is still walking towards. The sample timestamp is updated when the decoder returns, but the decoder does not return for non-taken branches. Update the sample timestamp then also.
Note that commit 3f04d98e972b5 ("perf intel-pt: Improve sample timestamp") was also a stable fix and appears, for example, in v4.4 stable tree as commit a4ebb58fd124 ("perf intel-pt: Improve sample timestamp").
--- tools/perf/util/intel-pt-decoder/intel-pt-decoder.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)
tools/perf/util/intel-pt-decoder/intel-pt-decoder.c +4 -1
@@ -1318,8 +1318,11 @@ static int intel_pt_walk_tnt(struct intel_pt_decoder *decoder)return 0 ; } decoder->ip += intel_pt_insn.length; -if (!decoder->tnt.count) +if (!decoder->tnt.count) { + decoder->sample_timestamp = decoder->timestamp; + decoder->sample_insn_cnt = decoder->timestamp_insn_cnt;return -EAGAIN ; + } decoder->tnt.payload <<=1 ;continue ; }
Arnaldo Carvalho de Melo
AppliedEm Fri, May 10, 2019 at 03:41:40PM +0300, Adrian Hunter escreveu:
Quoted 3 lines
Hi
Here are 3 non-urgent fixes for Intel PT.
Thanks, applied.
- Arnaldo
Quoted 12 lines
Adrian Hunter (3):
perf intel-pt: Fix instructions sampling rate perf intel-pt: Fix improved sample timestamp perf intel-pt: Fix sample timestamp wrt non-taken branches.../perf/util/intel-pt-decoder/intel-pt-decoder.c | 31 +++++++++++++++++----- 1 file changed, 24 insertions(+), 7 deletions(-)Regards
Adrian
Signature · 1 line
- Arnaldo