HTM trace details are captured as-is in PERF_RECORD_AUXTRACE
records. To present htm entries as samples, create an event
with name as "htm" and type PERF_TYPE_SYNTH.
Add perf_synth_id, "PERF_SYNTH_POWERPC_HTM" as config value for the
event. Create a sample id to be a fixed offset from evsel id.
Invoke powerpc_htm_create_sample() using the logical address
as sample ip.
This will help in understanding hot logical address from the
traces.
Usage:
# perf record -C 1 -e htm/nodalchipindex=2,nodeindex=0,htm_type=1/ -o
perf_1.data ls 1>out
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 257.504 MB perf_1.data ]
# ./perf report -i perf_1.data
# Samples: 8M of event 'htm'
# Event count (approx.): 8515199
#
# Children Self Command Shared Object Symbol
# ........ ........ ....... ................ ......................
#
0.61% 0.61% swapper [unknown] [.] 0x00000006fd567fe0
0.33% 0.33% swapper [unknown] [.] 0x00000006fc194b20
0.20% 0.20% swapper [unknown] [.] 0x0000000066113f80
0.18% 0.18% swapper [unknown] [.] 0x00000007fd888f20
0.15% 0.15% swapper [unknown] [.] 0x00000006fd567fc0
0.08% 0.08% swapper [unknown] [.] 0x00000006fc194b00
0.05% 0.05% swapper [unknown] [.] 0x00000007fd888f00
0.03% 0.03% swapper [unknown] [.] 0x0000000422510700
0.03% 0.03% swapper [unknown] [.] 0x0000000422510820
0.03% 0.03% swapper [unknown] [.] 0x00000007fd888b80
0.02% 0.02% swapper [unknown] [.] 0x000000000a0ece40
0.02% 0.02% swapper [unknown] [.] 0x000000000a0ed2e0
0.01% 0.01% swapper [unknown] [.] 0x00000007fd888c40
0.01% 0.01% swapper [unknown] [.] 0x00000007fd889000
0.01% 0.01% swapper [unknown] [.] 0x00000007fd5bc200
0.01% 0.01% swapper [unknown] [.] 0x00000007fd61c200
0.01% 0.01% swapper [unknown] [.] 0x00000007fd28c200
0.01% 0.01% swapper [unknown] [.] 0x00000007fd22c200
0.01% 0.01% swapper [unknown] [.] 0x00000007fd1fc200
Signed-off-by: Athira Rajeev <[email protected]>
---
tools/perf/util/event.h | 1 +
tools/perf/util/powerpc-htm.c | 110 +++++++++++++++++++++++++++++++++-
2 files changed, 109 insertions(+), 2 deletions(-)
diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h
index 2ea83fdf8a03..f5aa7eb9f5b7 100644
--- a/tools/perf/util/event.h
+++ b/tools/perf/util/event.h
@@ -118,6 +118,7 @@ enum perf_synth_id {
PERF_SYNTH_INTEL_EVT,
PERF_SYNTH_INTEL_IFLAG_CHG,
PERF_SYNTH_POWERPC_VPA_DTL,
+ PERF_SYNTH_POWERPC_HTM,
};
/*
diff --git a/tools/perf/util/powerpc-htm.c b/tools/perf/util/powerpc-htm.c
index 83253850870c..050fbceac71e 100644
--- a/tools/perf/util/powerpc-htm.c
+++ b/tools/perf/util/powerpc-htm.c
@@ -42,6 +42,7 @@ struct powerpc_htm {
char trans_file[64];
int htm_mem_entries;
int mem_maps;
+ u64 sample_id;
};
struct htm_mem {
@@ -128,6 +129,43 @@ static int run_htmdecode(const char *input_file, const
char *output_file)
return 0;
}
+static int powerpc_htm_create_sample(unsigned long addr, struct perf_session
*session,
+ struct powerpc_htm *htm)
+{
+ struct perf_sample sample;
+ union perf_event event;
+
+ if (dump_trace)
+ return 0;
+
+ memset(&sample, 0, sizeof(sample));
+ sample.cpumode = PERF_RECORD_MISC_USER;
+
+ if (!addr)
+ return 0;
+
+ if (addr >= 0xc000000000000000)
+ sample.cpumode = PERF_RECORD_MISC_KERNEL;
+
+ sample.ip = addr;
+ sample.period = 1;
+ sample.cpu = 0;
+ sample.id = htm->sample_id;
+ sample.callchain = NULL;
+ sample.branch_stack = NULL;
+ memset(&event, 0, sizeof(event));
+ event.sample.header.type = PERF_RECORD_SAMPLE;
+ event.sample.header.misc = sample.cpumode;
+ event.sample.header.size = sizeof(struct perf_event_header);
+
+ if (perf_session__deliver_synth_event(session, &event, &sample)) {
+ pr_debug("Failed to create sample for htm entry\n");
+ return -1;
+ }
+
+ return 0;
+}
+
static void *safe_realloc(void *ptr, size_t new_size)
{
void *tmp = realloc(ptr, new_size);
@@ -309,7 +347,7 @@ static struct addr_map *process_trace_file(const char
*trace_file,
return maps;
}
-static int create_mem_maps(struct powerpc_htm *htm)
+static int create_mem_maps(struct perf_session *session, struct powerpc_htm
*htm)
{
off_t file_size;
void *htmdata, *mapped_data;
@@ -422,6 +460,7 @@ static int create_mem_maps(struct powerpc_htm *htm)
maps[i].event,
maps[i].phys_addr,
(unsigned long)maps[i].logical_addr);
+ powerpc_htm_create_sample(maps[i].logical_addr, session, htm);
}
free(maps);
@@ -581,7 +620,7 @@ static int powerpc_htm_process_event(struct perf_session
*session __maybe_unused
}
/* Only for power bus traces, we decode traces */
if (config == 1)
- create_mem_maps(htm);
+ create_mem_maps(session, htm);
}
return 0;
@@ -636,6 +675,69 @@ static void powerpc_htm_print_info(__u64 *arr)
fprintf(stdout, powerpc_htm_info_fmts[POWERPC_HTM_TYPE],
arr[POWERPC_HTM_TYPE]);
}
+static void set_event_name(struct evlist *evlist, u64 id,
+ const char *name)
+{
+ struct evsel *evsel;
+
+ evlist__for_each_entry(evlist, evsel) {
+ if (evsel->core.id && evsel->core.id[0] == id) {
+ if (evsel->name)
+ zfree(&evsel->name);
+ evsel->name = strdup(name);
+ if (!evsel->name) {
+ pr_err("Failed to allocate memory for event
name\n");
+ return;
+ }
+ break;
+ }
+ }
+}
+
+static int
+powerpc_htm_synth_events(struct powerpc_htm *htm, struct perf_session *session)
+{
+ struct evlist *evlist = session->evlist;
+ struct evsel *evsel;
+ struct perf_event_attr attr;
+ bool found = false;
+ u64 id;
+ int err;
+
+ evlist__for_each_entry(evlist, evsel) {
+ if (strstarts(evsel->name, "htm")) {
+ found = true;
+ break;
+ }
+ }
+
+ if (!found) {
+ pr_debug("No selected events with HTM trace data\n");
+ return 0;
+ }
+
+ memset(&attr, 0, sizeof(struct perf_event_attr));
+ attr.size = sizeof(struct perf_event_attr);
+ attr.sample_type = evsel->core.attr.sample_type;
+ attr.sample_id_all = evsel->core.attr.sample_id_all;
+ attr.type = PERF_TYPE_SYNTH;
+ attr.config = PERF_SYNTH_POWERPC_HTM;
+
+ /* create new id val to be a fixed offset from evsel id */
+ id = evsel->core.id[0] + 1000000000;
+ if (!id)
+ id = 1;
+
+ err = perf_session__deliver_synth_attr_event(session, &attr, id);
+ if (err)
+ return err;
+
+ htm->sample_id = id;
+ set_event_name(evlist, id, "htm");
+
+ return 0;
+}
+
int powerpc_htm_process_auxtrace_info(union perf_event *event,
struct perf_session *session)
{
@@ -698,6 +800,10 @@ int powerpc_htm_process_auxtrace_info(union perf_event
*event,
if (err)
goto err_free_queues;
+ err = powerpc_htm_synth_events(htm, session);
+ if (err)
+ goto err_free;
+
return 0;
err_free_queues:
--
2.52.0