[PATCH 7/9] perf intel-pt: Add support for synthesizing branch stacks for regular events

From: Adrian Hunter
Date: Wed Apr 29 2020 - 11:07:46 EST


Use the new thread_stack__br_sample_late() function to create a thread
stack for regular events.

Example:

# perf record --kcore --aux-sample -e '{intel_pt//,cycles:ppp}' -c 10000 uname
Linux
[ perf record: Woken up 2 times to write data ]
[ perf record: Captured and wrote 0.743 MB perf.data ]
# perf report --itrace=Le --stdio | head -30 | tail -18

# Samples: 11K of event 'cycles:ppp'
# Event count (approx.): 11648
#
# Overhead Command Source Shared Object Source Symbol Target Symbol Basic Block Cycles
# ........ ....... .................... .............................................. .............................................. ..................
#
5.49% uname libc-2.30.so [.] _dl_addr [.] _dl_addr -
2.41% uname ld-2.30.so [.] _dl_relocate_object [.] _dl_relocate_object -
2.31% uname ld-2.30.so [.] do_lookup_x [.] do_lookup_x -
2.17% uname [kernel.kallsyms] [k] unmap_page_range [k] unmap_page_range -
2.05% uname ld-2.30.so [k] _dl_start [k] _dl_start -
1.97% uname ld-2.30.so [.] _dl_lookup_symbol_x [.] _dl_lookup_symbol_x -
1.94% uname [kernel.kallsyms] [k] filemap_map_pages [k] filemap_map_pages -
1.60% uname [kernel.kallsyms] [k] __handle_mm_fault [k] __handle_mm_fault -
1.44% uname [kernel.kallsyms] [k] page_add_file_rmap [k] page_add_file_rmap -
1.12% uname [kernel.kallsyms] [k] vma_interval_tree_insert [k] vma_interval_tree_insert -
0.94% uname [kernel.kallsyms] [k] perf_iterate_ctx [k] perf_iterate_ctx -

Signed-off-by: Adrian Hunter <adrian.hunter@xxxxxxxxx>
---
tools/perf/util/intel-pt.c | 73 ++++++++++++++++++++++++++++++++++----
1 file changed, 66 insertions(+), 7 deletions(-)

diff --git a/tools/perf/util/intel-pt.c b/tools/perf/util/intel-pt.c
index 248a39fd4d0e..7fb807b91f73 100644
--- a/tools/perf/util/intel-pt.c
+++ b/tools/perf/util/intel-pt.c
@@ -71,6 +71,7 @@ struct intel_pt {
bool use_thread_stack;
bool callstack;
unsigned int br_stack_sz;
+ unsigned int br_stack_sz_plus;
int have_sched_switch;
u32 pmu_type;
u64 kernel_start;
@@ -129,6 +130,7 @@ struct intel_pt {
unsigned int range_cnt;

struct ip_callchain *chain;
+ struct branch_stack *br_stack;
};

enum switch_state {
@@ -910,6 +912,44 @@ static void intel_pt_add_callchain(struct intel_pt *pt,
sample->callchain = pt->chain;
}

+static struct branch_stack *intel_pt_alloc_br_stack(struct intel_pt *pt)
+{
+ size_t sz = sizeof(struct branch_stack);
+
+ sz += pt->br_stack_sz * sizeof(struct branch_entry);
+ return zalloc(sz);
+}
+
+static int intel_pt_br_stack_init(struct intel_pt *pt)
+{
+ struct evsel *evsel;
+
+ evlist__for_each_entry(pt->session->evlist, evsel) {
+ if (!(evsel->core.attr.sample_type & PERF_SAMPLE_BRANCH_STACK))
+ evsel->synth_sample_type |= PERF_SAMPLE_BRANCH_STACK;
+ }
+
+ pt->br_stack = intel_pt_alloc_br_stack(pt);
+ if (!pt->br_stack)
+ return -ENOMEM;
+
+ return 0;
+}
+
+static void intel_pt_add_br_stack(struct intel_pt *pt,
+ struct perf_sample *sample)
+{
+ struct thread *thread = machine__findnew_thread(pt->machine,
+ sample->pid,
+ sample->tid);
+
+ thread_stack__br_sample_late(thread, sample->cpu, pt->br_stack,
+ pt->br_stack_sz, sample->ip,
+ pt->kernel_start);
+
+ sample->branch_stack = pt->br_stack;
+}
+
static struct intel_pt_queue *intel_pt_alloc_queue(struct intel_pt *pt,
unsigned int queue_nr)
{
@@ -928,10 +968,7 @@ static struct intel_pt_queue *intel_pt_alloc_queue(struct intel_pt *pt,
}

if (pt->synth_opts.last_branch) {
- size_t sz = sizeof(struct branch_stack);
-
- sz += pt->br_stack_sz * sizeof(struct branch_entry);
- ptq->last_branch = zalloc(sz);
+ ptq->last_branch = intel_pt_alloc_br_stack(pt);
if (!ptq->last_branch)
goto out_free;
}
@@ -1962,7 +1999,7 @@ static int intel_pt_sample(struct intel_pt_queue *ptq)
thread_stack__event(ptq->thread, ptq->cpu, ptq->flags,
state->from_ip, state->to_ip, ptq->insn_len,
state->trace_nr, pt->callstack,
- pt->br_stack_sz,
+ pt->br_stack_sz_plus,
pt->mispred_all);
} else {
thread_stack__set_trace_nr(ptq->thread, ptq->cpu, state->trace_nr);
@@ -2608,6 +2645,8 @@ static int intel_pt_process_event(struct perf_session *session,
if (event->header.type == PERF_RECORD_SAMPLE) {
if (pt->synth_opts.add_callchain && !sample->callchain)
intel_pt_add_callchain(pt, sample);
+ if (pt->synth_opts.add_last_branch && !sample->branch_stack)
+ intel_pt_add_br_stack(pt, sample);
}

if (event->header.type == PERF_RECORD_AUX &&
@@ -3369,13 +3408,33 @@ int intel_pt_process_auxtrace_info(union perf_event *event,
goto err_delete_thread;
}

- if (pt->synth_opts.last_branch)
+ if (pt->synth_opts.last_branch || pt->synth_opts.add_last_branch) {
pt->br_stack_sz = pt->synth_opts.last_branch_sz;
+ pt->br_stack_sz_plus = pt->br_stack_sz;
+ }
+
+ if (pt->synth_opts.add_last_branch) {
+ err = intel_pt_br_stack_init(pt);
+ if (err)
+ goto err_delete_thread;
+ /*
+ * Additional branch stack size to cater for tracing from the
+ * actual sample ip to where the sample time is recorded.
+ * Measured at about 200 branches, but generously set to 1024.
+ * If kernel space is not being traced, then add just 1 for the
+ * branch to kernel space.
+ */
+ if (intel_pt_tracing_kernel(pt))
+ pt->br_stack_sz_plus += 1024;
+ else
+ pt->br_stack_sz_plus += 1;
+ }

pt->use_thread_stack = pt->synth_opts.callchain ||
pt->synth_opts.add_callchain ||
pt->synth_opts.thread_stack ||
- pt->synth_opts.last_branch;
+ pt->synth_opts.last_branch ||
+ pt->synth_opts.add_last_branch;

pt->callstack = pt->synth_opts.callchain ||
pt->synth_opts.add_callchain ||
--
2.17.1