Commit 20aa3970 authored by Adrian Hunter's avatar Adrian Hunter Committed by Arnaldo Carvalho de Melo

perf intel-pt: Fix premature IPC

The code assumed a change in cycle count means accurate IPC. That is not
correct, for example when sampling both branches and instructions, or at
a FUP packet (which is not CYC-eligible) address. Fix by using an explicit
flag to indicate when IPC can be sampled.

Fixes: 5b1dc0fd ("perf intel-pt: Add support for samples to contain IPC ratio")
Signed-off-by: default avatarAdrian Hunter <adrian.hunter@intel.com>
Reviewed-by: default avatarAndi Kleen <ak@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: linux-kernel@vger.kernel.org
Link: https://lore.kernel.org/r/20210205175350.23817-3-adrian.hunter@intel.comSigned-off-by: default avatarArnaldo Carvalho de Melo <acme@redhat.com>
parent 03fb0f85
...@@ -2814,9 +2814,18 @@ const struct intel_pt_state *intel_pt_decode(struct intel_pt_decoder *decoder) ...@@ -2814,9 +2814,18 @@ const struct intel_pt_state *intel_pt_decode(struct intel_pt_decoder *decoder)
} }
if (intel_pt_sample_time(decoder->pkt_state)) { if (intel_pt_sample_time(decoder->pkt_state)) {
intel_pt_update_sample_time(decoder); intel_pt_update_sample_time(decoder);
if (decoder->sample_cyc) if (decoder->sample_cyc) {
decoder->sample_tot_cyc_cnt = decoder->tot_cyc_cnt; decoder->sample_tot_cyc_cnt = decoder->tot_cyc_cnt;
decoder->state.flags |= INTEL_PT_SAMPLE_IPC;
decoder->sample_cyc = false;
}
} }
/*
* When using only TSC/MTC to compute cycles, IPC can be
* sampled as soon as the cycle count changes.
*/
if (!decoder->have_cyc)
decoder->state.flags |= INTEL_PT_SAMPLE_IPC;
} }
decoder->state.timestamp = decoder->sample_timestamp; decoder->state.timestamp = decoder->sample_timestamp;
......
...@@ -17,6 +17,7 @@ ...@@ -17,6 +17,7 @@
#define INTEL_PT_ABORT_TX (1 << 1) #define INTEL_PT_ABORT_TX (1 << 1)
#define INTEL_PT_ASYNC (1 << 2) #define INTEL_PT_ASYNC (1 << 2)
#define INTEL_PT_FUP_IP (1 << 3) #define INTEL_PT_FUP_IP (1 << 3)
#define INTEL_PT_SAMPLE_IPC (1 << 4)
enum intel_pt_sample_type { enum intel_pt_sample_type {
INTEL_PT_BRANCH = 1 << 0, INTEL_PT_BRANCH = 1 << 0,
......
...@@ -1381,6 +1381,7 @@ static int intel_pt_synth_branch_sample(struct intel_pt_queue *ptq) ...@@ -1381,6 +1381,7 @@ static int intel_pt_synth_branch_sample(struct intel_pt_queue *ptq)
sample.branch_stack = (struct branch_stack *)&dummy_bs; sample.branch_stack = (struct branch_stack *)&dummy_bs;
} }
if (ptq->state->flags & INTEL_PT_SAMPLE_IPC)
sample.cyc_cnt = ptq->ipc_cyc_cnt - ptq->last_br_cyc_cnt; sample.cyc_cnt = ptq->ipc_cyc_cnt - ptq->last_br_cyc_cnt;
if (sample.cyc_cnt) { if (sample.cyc_cnt) {
sample.insn_cnt = ptq->ipc_insn_cnt - ptq->last_br_insn_cnt; sample.insn_cnt = ptq->ipc_insn_cnt - ptq->last_br_insn_cnt;
...@@ -1431,6 +1432,7 @@ static int intel_pt_synth_instruction_sample(struct intel_pt_queue *ptq) ...@@ -1431,6 +1432,7 @@ static int intel_pt_synth_instruction_sample(struct intel_pt_queue *ptq)
else else
sample.period = ptq->state->tot_insn_cnt - ptq->last_insn_cnt; sample.period = ptq->state->tot_insn_cnt - ptq->last_insn_cnt;
if (ptq->state->flags & INTEL_PT_SAMPLE_IPC)
sample.cyc_cnt = ptq->ipc_cyc_cnt - ptq->last_in_cyc_cnt; sample.cyc_cnt = ptq->ipc_cyc_cnt - ptq->last_in_cyc_cnt;
if (sample.cyc_cnt) { if (sample.cyc_cnt) {
sample.insn_cnt = ptq->ipc_insn_cnt - ptq->last_in_insn_cnt; sample.insn_cnt = ptq->ipc_insn_cnt - ptq->last_in_insn_cnt;
...@@ -1983,14 +1985,8 @@ static int intel_pt_sample(struct intel_pt_queue *ptq) ...@@ -1983,14 +1985,8 @@ static int intel_pt_sample(struct intel_pt_queue *ptq)
ptq->have_sample = false; ptq->have_sample = false;
if (ptq->state->tot_cyc_cnt > ptq->ipc_cyc_cnt) {
/*
* Cycle count and instruction count only go together to create
* a valid IPC ratio when the cycle count changes.
*/
ptq->ipc_insn_cnt = ptq->state->tot_insn_cnt; ptq->ipc_insn_cnt = ptq->state->tot_insn_cnt;
ptq->ipc_cyc_cnt = ptq->state->tot_cyc_cnt; ptq->ipc_cyc_cnt = ptq->state->tot_cyc_cnt;
}
/* /*
* Do PEBS first to allow for the possibility that the PEBS timestamp * Do PEBS first to allow for the possibility that the PEBS timestamp
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment