Android systrace: Optimize memory dumps.

systrace/atrace_helper/jni/atrace_process_dump.cc

Index: systrace/atrace_helper/jni/atrace_process_dump.cc
diff --git a/systrace/atrace_helper/jni/atrace_process_dump.cc b/systrace/atrace_helper/jni/atrace_process_dump.cc
new file mode 100644
index 0000000000000000000000000000000000000000..0a6b9f449155c836dd14754b6708f3cceab03f0f
--- /dev/null
+++ b/systrace/atrace_helper/jni/atrace_process_dump.cc
@@ -0,0 +1,215 @@
+// Copyright 2017 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "atrace_process_dump.h"
+
+#include <sys/time.h>
+#include <sys/timerfd.h>
+#include <stdint.h>
+
+#include <memory>
+#include <sstream>
+
+#include "logging.h"
+
+void AtraceProcessDump::SetFullDumpConfig(const std::string& config) {
+  self_pid_ = (int) getpid();

[Primiano Tucci (use gerrit), 2017/06/26 10:07:19]

static_cast<int>(getpid()). c-style cast are not allowed in our code style:

https://google.github.io/styleguide/cppguide.html#Casting

+  if (config == "all") {
+    full_dump_all_ = true;
+  } else if (config == "apps") {
+    full_dump_apps_ = true;
+  } else {
+    full_dump_whitelist_.clear();
+    std::istringstream ss(config);
+    std::string entry;
+    while (std::getline(ss, entry, ',')) {
+      full_dump_whitelist_.insert(entry);
+    }
+  }
+}
+
+void AtraceProcessDump::SerializeSnapshot(
+    const InstantProcessInfo* snapshot, const std::set<int>& smaps_pids) {
+
+  fprintf(out_, "{\"ts\":\"%llu\",\"memdump\":{\n", snapshot->timestamp());
+  const InstantProcessInfo::ProcessSnapshotMap* processes =
+      snapshot->processes();
+  for (auto pit = processes->begin(); pit != processes->end();) {

[Primiano Tucci (use gerrit), 2017/06/26 10:07:20]

I thought this was a typo (pit -> pid) but then realized that you meant iterator
here.
Eitehr just |it|, or |proc_iter|, or |proc_it|.

+    const InstantProcessInfo::ProcessSnapshot* process = pit->second.get();
+    const ProcessMemoryStats* mem = &process->memory;
+    fprintf(out_, "\"%d\":{", process->pid);
+
+    fprintf(out_, "\"vm\":%llu,\"rss\":%llu",
+            mem->virt_kb(), mem->rss_kb());
+
+    fprintf(out_, ",\"oom_sc\":%d,\"oom_sc_adj\":%d"
+                  ",\"min_flt\":%lu,\"maj_flt\":%lu"
+                  ",\"utime\":%lu,\"stime\":%lu",
+            process->oom_score, process->oom_score_adj,
+            process->minor_faults, process->major_faults,
+            process->utime, process->stime);
+
+    if (mem->full_stats_available()) {
+      fprintf(out_, ",\"pss\":%llu,\"swp\":%llu"
+                    ",\"pc\":%llu,\"pd\":%llu,\"sc\":%llu,\"sd\":%llu",
+              mem->pss_kb(), mem->swapped_kb(),
+              mem->private_clean_kb(), mem->private_dirty_kb(),
+              mem->shared_clean_kb(), mem->shared_dirty_kb());
+    }
+
+    if (mem->gpu_stats_available()) {
+      fprintf(out_, ",\"gpu_egl\":%llu,\"gpu_egl_pss\":%llu"
+                    ",\"gpu_gl\":%llu,\"gpu_gl_pss\":%llu"
+                    ",\"gpu_etc\":%llu,\"gpu_etc_pss\":%llu",
+              mem->gpu_graphics_kb(), mem->gpu_graphics_pss_kb(),
+              mem->gpu_gl_kb(), mem->gpu_gl_pss_kb(),
+              mem->gpu_other_kb(), mem->gpu_other_pss_kb());
+    }
+
+    if (mem->full_stats_available() && smaps_pids.count(process->pid)) {
+      fprintf(out_, ", \"mmaps\":[");
+      size_t n_mmaps = mem->mmaps_count();
+      for (size_t k = 0; k < n_mmaps; ++k) {
+        const ProcessMemoryStats::MmapInfo* mm = mem->mmap(k);
+        fprintf(out_,
+                "{\"vm\":\"%llx-%llx\",\"file\":\"%s\",\"flags\":\"%s\","
+                "\"pss\":%llu,\"rss\":%llu,\"swp\":%llu,"
+                "\"pc\":%llu,\"pd\":%llu,"
+                "\"sc\":%llu,\"sd\":%llu}",
+                mm->start_addr, mm->end_addr, mm->mapped_file, mm->prot_flags,
+                mm->pss_kb, mm->rss_kb, mm->swapped_kb,
+                mm->private_clean_kb, mm->private_dirty_kb,
+                mm->shared_clean_kb, mm->shared_dirty_kb);
+        if (k < n_mmaps - 1)
+          fprintf(out_, ", ");
+      }
+      fprintf(out_, "]");
+    }
+
+    // TODO(kraynov): Return OOM numbers.
+
+    ++pit;
+    if (pit != processes->end()) {
+      fprintf(out_, "},\n");
+    } else {
+      fprintf(out_, "}}\n");
+    }
+  }
+  fprintf(out_, "}");
+}
+
+void AtraceProcessDump::SerializeProcessInfo(
+    const PersistentProcessInfo::ProcessMap* processes) {
+
+  fprintf(out_, "\"processes\":{");
+  for (auto pit = processes->begin(); pit != processes->end();) {
+    const PersistentProcessInfo::ProcessInfo* process = pit->second.get();
+    fprintf(out_, "\"%d\":{", process->pid);
+    fprintf(out_, "\"name\":\"%s\"", process->name.c_str());
+
+    if (!process->in_kernel) {
+      fprintf(out_, ",\"exe\":\"%s\",", process->exe.c_str());
+      fprintf(out_, "\"threads\":{\n");
+      const PersistentProcessInfo::ThreadMap* threads = &process->threads;
+      for (auto tit = threads->begin(); tit != threads->end();) {
+        const PersistentProcessInfo::ThreadInfo* thread = tit->second.get();
+        fprintf(out_, "\"%d\":{", thread->tid);
+        fprintf(out_, "\"name\":\"%s\"", thread->name.c_str());
+
+        ++tit;
+        if (tit != threads->end()) {
+          fprintf(out_, "},\n");
+        } else {
+          fprintf(out_, "}\n");
+        }
+      }
+      fprintf(out_, "}");
+    }
+
+    ++pit;
+    if (pit != processes->end()) {
+      fprintf(out_, "},\n");
+    } else {
+      fprintf(out_, "}\n");
+    }
+  }
+  fprintf(out_, "}");
+}
+
+void AtraceProcessDump::RunAndPrintJson(FILE* stream) {
+  using ProcessInfo = PersistentProcessInfo::ProcessInfo;
+  out_ = stream;
+
+  std::set<int> smaps_pids;
+  std::unique_ptr<ProcessDumpManager> mgr(new ProcessDumpManager());

[Primiano Tucci (use gerrit), 2017/06/26 10:07:19]

Copy/pasting to what I said in the previous review (I didn't get any reply):
Architecturally I am not sure I understand the difference between
atrace_process_dump.cc and ProcessDumpManager. Both of them seem to somehow
coordinate dumps. Can you pick one?

+  mgr->SetFullDumpPredicate([this](const ProcessInfo* process) -> bool {
+    if (full_dump_all_)
+      return !process->in_kernel && (process->pid != self_pid_);
+    if (full_dump_apps_)
+      return process->is_app;
+    if (full_dump_whitelist_.empty())
+      return false;
+    return full_dump_whitelist_.count(process->name) > 0;
+  });
+  if (graphics_stats_)
+    mgr->EnableGraphicsStats();
+
+  SetupTimer();
+  fprintf(out_, "{\"start_ts\": \"%llu\", \"snapshots\":[\n", GetTimestamp());
+  for (int dump_number = 0; dump_number < dump_count_; dump_number++) {
+    if (dump_number > 0) {
+      fprintf(out_, ",\n");
+      if (!WaitForTimer())
+        break;  // Interrupted by signal.
+    }
+    mgr->TakeSnapshot();
+
+    // Smaps are too heavy to serialize. Using only in whitelist mode.
+    if (print_smaps_ && !full_dump_whitelist_.empty()) {
+      smaps_pids.clear();
+      auto processes = mgr->all_processes();
+      for (auto pit = processes->begin(); pit != processes->end(); ++pit) {
+        if (full_dump_whitelist_.count(pit->second->name))
+          smaps_pids.insert(pit->second->pid);
+      }
+    }
+    SerializeSnapshot(mgr->last_snapshot(), smaps_pids);
+    fflush(out_);
+  }
+  fprintf(out_, "],\n");
+  SerializeProcessInfo(mgr->all_processes());
+  CloseTimer();
+
+  fprintf(out_, "}\n");
+  fclose(out_);
+}
+
+void AtraceProcessDump::SetupTimer() {
+  timer_fd_ = timerfd_create(CLOCK_MONOTONIC, 0);
+  CHECK(timer_fd_ >= 0);
+  int sec = dump_interval_ms_ / 1000;
+  int nsec = (dump_interval_ms_ % 1000) * 1000000;
+  struct itimerspec ts = {};
+  ts.it_value.tv_nsec = nsec;
+  ts.it_value.tv_sec = sec;
+  ts.it_interval.tv_nsec = nsec;
+  ts.it_interval.tv_sec = sec;
+  CHECK(timerfd_settime(timer_fd_, 0, &ts, nullptr) == 0);
+}
+
+bool AtraceProcessDump::WaitForTimer() {
+  uint64_t stub = 0;
+  int res = read(timer_fd_, &stub, sizeof(stub));
+  if (res < 0 && errno == EBADF)
+    return false;  // Received SIGINT/SIGTERM signal.
+  CHECK(res > 0);
+  return true;
+}
+
+void AtraceProcessDump::CloseTimer() {
+  if (timer_fd_ < 0)
+    return;
+  close(timer_fd_);
+  timer_fd_ = 0;
+}