Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / tools / perf / builtin-timechart.c
diff --git a/kernel/tools/perf/builtin-timechart.c b/kernel/tools/perf/builtin-timechart.c
new file mode 100644 (file)
index 0000000..e50fe11
--- /dev/null
@@ -0,0 +1,2011 @@
+/*
+ * builtin-timechart.c - make an svg timechart of system activity
+ *
+ * (C) Copyright 2009 Intel Corporation
+ *
+ * Authors:
+ *     Arjan van de Ven <arjan@linux.intel.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; version 2
+ * of the License.
+ */
+
+#include <traceevent/event-parse.h>
+
+#include "builtin.h"
+
+#include "util/util.h"
+
+#include "util/color.h"
+#include <linux/list.h>
+#include "util/cache.h"
+#include "util/evlist.h"
+#include "util/evsel.h"
+#include <linux/rbtree.h>
+#include "util/symbol.h"
+#include "util/callchain.h"
+#include "util/strlist.h"
+
+#include "perf.h"
+#include "util/header.h"
+#include "util/parse-options.h"
+#include "util/parse-events.h"
+#include "util/event.h"
+#include "util/session.h"
+#include "util/svghelper.h"
+#include "util/tool.h"
+#include "util/data.h"
+#include "util/debug.h"
+
+#define SUPPORT_OLD_POWER_EVENTS 1
+#define PWR_EVENT_EXIT -1
+
+struct per_pid;
+struct power_event;
+struct wake_event;
+
+struct timechart {
+       struct perf_tool        tool;
+       struct per_pid          *all_data;
+       struct power_event      *power_events;
+       struct wake_event       *wake_events;
+       int                     proc_num;
+       unsigned int            numcpus;
+       u64                     min_freq,       /* Lowest CPU frequency seen */
+                               max_freq,       /* Highest CPU frequency seen */
+                               turbo_frequency,
+                               first_time, last_time;
+       bool                    power_only,
+                               tasks_only,
+                               with_backtrace,
+                               topology;
+       /* IO related settings */
+       u64                     io_events;
+       bool                    io_only,
+                               skip_eagain;
+       u64                     min_time,
+                               merge_dist;
+       bool                    force;
+};
+
+struct per_pidcomm;
+struct cpu_sample;
+struct io_sample;
+
+/*
+ * Datastructure layout:
+ * We keep an list of "pid"s, matching the kernels notion of a task struct.
+ * Each "pid" entry, has a list of "comm"s.
+ *     this is because we want to track different programs different, while
+ *     exec will reuse the original pid (by design).
+ * Each comm has a list of samples that will be used to draw
+ * final graph.
+ */
+
+struct per_pid {
+       struct per_pid *next;
+
+       int             pid;
+       int             ppid;
+
+       u64             start_time;
+       u64             end_time;
+       u64             total_time;
+       u64             total_bytes;
+       int             display;
+
+       struct per_pidcomm *all;
+       struct per_pidcomm *current;
+};
+
+
+struct per_pidcomm {
+       struct per_pidcomm *next;
+
+       u64             start_time;
+       u64             end_time;
+       u64             total_time;
+       u64             max_bytes;
+       u64             total_bytes;
+
+       int             Y;
+       int             display;
+
+       long            state;
+       u64             state_since;
+
+       char            *comm;
+
+       struct cpu_sample *samples;
+       struct io_sample  *io_samples;
+};
+
+struct sample_wrapper {
+       struct sample_wrapper *next;
+
+       u64             timestamp;
+       unsigned char   data[0];
+};
+
+#define TYPE_NONE      0
+#define TYPE_RUNNING   1
+#define TYPE_WAITING   2
+#define TYPE_BLOCKED   3
+
+struct cpu_sample {
+       struct cpu_sample *next;
+
+       u64 start_time;
+       u64 end_time;
+       int type;
+       int cpu;
+       const char *backtrace;
+};
+
+enum {
+       IOTYPE_READ,
+       IOTYPE_WRITE,
+       IOTYPE_SYNC,
+       IOTYPE_TX,
+       IOTYPE_RX,
+       IOTYPE_POLL,
+};
+
+struct io_sample {
+       struct io_sample *next;
+
+       u64 start_time;
+       u64 end_time;
+       u64 bytes;
+       int type;
+       int fd;
+       int err;
+       int merges;
+};
+
+#define CSTATE 1
+#define PSTATE 2
+
+struct power_event {
+       struct power_event *next;
+       int type;
+       int state;
+       u64 start_time;
+       u64 end_time;
+       int cpu;
+};
+
+struct wake_event {
+       struct wake_event *next;
+       int waker;
+       int wakee;
+       u64 time;
+       const char *backtrace;
+};
+
+struct process_filter {
+       char                    *name;
+       int                     pid;
+       struct process_filter   *next;
+};
+
+static struct process_filter *process_filter;
+
+
+static struct per_pid *find_create_pid(struct timechart *tchart, int pid)
+{
+       struct per_pid *cursor = tchart->all_data;
+
+       while (cursor) {
+               if (cursor->pid == pid)
+                       return cursor;
+               cursor = cursor->next;
+       }
+       cursor = zalloc(sizeof(*cursor));
+       assert(cursor != NULL);
+       cursor->pid = pid;
+       cursor->next = tchart->all_data;
+       tchart->all_data = cursor;
+       return cursor;
+}
+
+static void pid_set_comm(struct timechart *tchart, int pid, char *comm)
+{
+       struct per_pid *p;
+       struct per_pidcomm *c;
+       p = find_create_pid(tchart, pid);
+       c = p->all;
+       while (c) {
+               if (c->comm && strcmp(c->comm, comm) == 0) {
+                       p->current = c;
+                       return;
+               }
+               if (!c->comm) {
+                       c->comm = strdup(comm);
+                       p->current = c;
+                       return;
+               }
+               c = c->next;
+       }
+       c = zalloc(sizeof(*c));
+       assert(c != NULL);
+       c->comm = strdup(comm);
+       p->current = c;
+       c->next = p->all;
+       p->all = c;
+}
+
+static void pid_fork(struct timechart *tchart, int pid, int ppid, u64 timestamp)
+{
+       struct per_pid *p, *pp;
+       p = find_create_pid(tchart, pid);
+       pp = find_create_pid(tchart, ppid);
+       p->ppid = ppid;
+       if (pp->current && pp->current->comm && !p->current)
+               pid_set_comm(tchart, pid, pp->current->comm);
+
+       p->start_time = timestamp;
+       if (p->current && !p->current->start_time) {
+               p->current->start_time = timestamp;
+               p->current->state_since = timestamp;
+       }
+}
+
+static void pid_exit(struct timechart *tchart, int pid, u64 timestamp)
+{
+       struct per_pid *p;
+       p = find_create_pid(tchart, pid);
+       p->end_time = timestamp;
+       if (p->current)
+               p->current->end_time = timestamp;
+}
+
+static void pid_put_sample(struct timechart *tchart, int pid, int type,
+                          unsigned int cpu, u64 start, u64 end,
+                          const char *backtrace)
+{
+       struct per_pid *p;
+       struct per_pidcomm *c;
+       struct cpu_sample *sample;
+
+       p = find_create_pid(tchart, pid);
+       c = p->current;
+       if (!c) {
+               c = zalloc(sizeof(*c));
+               assert(c != NULL);
+               p->current = c;
+               c->next = p->all;
+               p->all = c;
+       }
+
+       sample = zalloc(sizeof(*sample));
+       assert(sample != NULL);
+       sample->start_time = start;
+       sample->end_time = end;
+       sample->type = type;
+       sample->next = c->samples;
+       sample->cpu = cpu;
+       sample->backtrace = backtrace;
+       c->samples = sample;
+
+       if (sample->type == TYPE_RUNNING && end > start && start > 0) {
+               c->total_time += (end-start);
+               p->total_time += (end-start);
+       }
+
+       if (c->start_time == 0 || c->start_time > start)
+               c->start_time = start;
+       if (p->start_time == 0 || p->start_time > start)
+               p->start_time = start;
+}
+
+#define MAX_CPUS 4096
+
+static u64 cpus_cstate_start_times[MAX_CPUS];
+static int cpus_cstate_state[MAX_CPUS];
+static u64 cpus_pstate_start_times[MAX_CPUS];
+static u64 cpus_pstate_state[MAX_CPUS];
+
+static int process_comm_event(struct perf_tool *tool,
+                             union perf_event *event,
+                             struct perf_sample *sample __maybe_unused,
+                             struct machine *machine __maybe_unused)
+{
+       struct timechart *tchart = container_of(tool, struct timechart, tool);
+       pid_set_comm(tchart, event->comm.tid, event->comm.comm);
+       return 0;
+}
+
+static int process_fork_event(struct perf_tool *tool,
+                             union perf_event *event,
+                             struct perf_sample *sample __maybe_unused,
+                             struct machine *machine __maybe_unused)
+{
+       struct timechart *tchart = container_of(tool, struct timechart, tool);
+       pid_fork(tchart, event->fork.pid, event->fork.ppid, event->fork.time);
+       return 0;
+}
+
+static int process_exit_event(struct perf_tool *tool,
+                             union perf_event *event,
+                             struct perf_sample *sample __maybe_unused,
+                             struct machine *machine __maybe_unused)
+{
+       struct timechart *tchart = container_of(tool, struct timechart, tool);
+       pid_exit(tchart, event->fork.pid, event->fork.time);
+       return 0;
+}
+
+#ifdef SUPPORT_OLD_POWER_EVENTS
+static int use_old_power_events;
+#endif
+
+static void c_state_start(int cpu, u64 timestamp, int state)
+{
+       cpus_cstate_start_times[cpu] = timestamp;
+       cpus_cstate_state[cpu] = state;
+}
+
+static void c_state_end(struct timechart *tchart, int cpu, u64 timestamp)
+{
+       struct power_event *pwr = zalloc(sizeof(*pwr));
+
+       if (!pwr)
+               return;
+
+       pwr->state = cpus_cstate_state[cpu];
+       pwr->start_time = cpus_cstate_start_times[cpu];
+       pwr->end_time = timestamp;
+       pwr->cpu = cpu;
+       pwr->type = CSTATE;
+       pwr->next = tchart->power_events;
+
+       tchart->power_events = pwr;
+}
+
+static void p_state_change(struct timechart *tchart, int cpu, u64 timestamp, u64 new_freq)
+{
+       struct power_event *pwr;
+
+       if (new_freq > 8000000) /* detect invalid data */
+               return;
+
+       pwr = zalloc(sizeof(*pwr));
+       if (!pwr)
+               return;
+
+       pwr->state = cpus_pstate_state[cpu];
+       pwr->start_time = cpus_pstate_start_times[cpu];
+       pwr->end_time = timestamp;
+       pwr->cpu = cpu;
+       pwr->type = PSTATE;
+       pwr->next = tchart->power_events;
+
+       if (!pwr->start_time)
+               pwr->start_time = tchart->first_time;
+
+       tchart->power_events = pwr;
+
+       cpus_pstate_state[cpu] = new_freq;
+       cpus_pstate_start_times[cpu] = timestamp;
+
+       if ((u64)new_freq > tchart->max_freq)
+               tchart->max_freq = new_freq;
+
+       if (new_freq < tchart->min_freq || tchart->min_freq == 0)
+               tchart->min_freq = new_freq;
+
+       if (new_freq == tchart->max_freq - 1000)
+               tchart->turbo_frequency = tchart->max_freq;
+}
+
+static void sched_wakeup(struct timechart *tchart, int cpu, u64 timestamp,
+                        int waker, int wakee, u8 flags, const char *backtrace)
+{
+       struct per_pid *p;
+       struct wake_event *we = zalloc(sizeof(*we));
+
+       if (!we)
+               return;
+
+       we->time = timestamp;
+       we->waker = waker;
+       we->backtrace = backtrace;
+
+       if ((flags & TRACE_FLAG_HARDIRQ) || (flags & TRACE_FLAG_SOFTIRQ))
+               we->waker = -1;
+
+       we->wakee = wakee;
+       we->next = tchart->wake_events;
+       tchart->wake_events = we;
+       p = find_create_pid(tchart, we->wakee);
+
+       if (p && p->current && p->current->state == TYPE_NONE) {
+               p->current->state_since = timestamp;
+               p->current->state = TYPE_WAITING;
+       }
+       if (p && p->current && p->current->state == TYPE_BLOCKED) {
+               pid_put_sample(tchart, p->pid, p->current->state, cpu,
+                              p->current->state_since, timestamp, NULL);
+               p->current->state_since = timestamp;
+               p->current->state = TYPE_WAITING;
+       }
+}
+
+static void sched_switch(struct timechart *tchart, int cpu, u64 timestamp,
+                        int prev_pid, int next_pid, u64 prev_state,
+                        const char *backtrace)
+{
+       struct per_pid *p = NULL, *prev_p;
+
+       prev_p = find_create_pid(tchart, prev_pid);
+
+       p = find_create_pid(tchart, next_pid);
+
+       if (prev_p->current && prev_p->current->state != TYPE_NONE)
+               pid_put_sample(tchart, prev_pid, TYPE_RUNNING, cpu,
+                              prev_p->current->state_since, timestamp,
+                              backtrace);
+       if (p && p->current) {
+               if (p->current->state != TYPE_NONE)
+                       pid_put_sample(tchart, next_pid, p->current->state, cpu,
+                                      p->current->state_since, timestamp,
+                                      backtrace);
+
+               p->current->state_since = timestamp;
+               p->current->state = TYPE_RUNNING;
+       }
+
+       if (prev_p->current) {
+               prev_p->current->state = TYPE_NONE;
+               prev_p->current->state_since = timestamp;
+               if (prev_state & 2)
+                       prev_p->current->state = TYPE_BLOCKED;
+               if (prev_state == 0)
+                       prev_p->current->state = TYPE_WAITING;
+       }
+}
+
+static const char *cat_backtrace(union perf_event *event,
+                                struct perf_sample *sample,
+                                struct machine *machine)
+{
+       struct addr_location al;
+       unsigned int i;
+       char *p = NULL;
+       size_t p_len;
+       u8 cpumode = PERF_RECORD_MISC_USER;
+       struct addr_location tal;
+       struct ip_callchain *chain = sample->callchain;
+       FILE *f = open_memstream(&p, &p_len);
+
+       if (!f) {
+               perror("open_memstream error");
+               return NULL;
+       }
+
+       if (!chain)
+               goto exit;
+
+       if (perf_event__preprocess_sample(event, machine, &al, sample) < 0) {
+               fprintf(stderr, "problem processing %d event, skipping it.\n",
+                       event->header.type);
+               goto exit;
+       }
+
+       for (i = 0; i < chain->nr; i++) {
+               u64 ip;
+
+               if (callchain_param.order == ORDER_CALLEE)
+                       ip = chain->ips[i];
+               else
+                       ip = chain->ips[chain->nr - i - 1];
+
+               if (ip >= PERF_CONTEXT_MAX) {
+                       switch (ip) {
+                       case PERF_CONTEXT_HV:
+                               cpumode = PERF_RECORD_MISC_HYPERVISOR;
+                               break;
+                       case PERF_CONTEXT_KERNEL:
+                               cpumode = PERF_RECORD_MISC_KERNEL;
+                               break;
+                       case PERF_CONTEXT_USER:
+                               cpumode = PERF_RECORD_MISC_USER;
+                               break;
+                       default:
+                               pr_debug("invalid callchain context: "
+                                        "%"PRId64"\n", (s64) ip);
+
+                               /*
+                                * It seems the callchain is corrupted.
+                                * Discard all.
+                                */
+                               zfree(&p);
+                               goto exit;
+                       }
+                       continue;
+               }
+
+               tal.filtered = 0;
+               thread__find_addr_location(al.thread, cpumode,
+                                          MAP__FUNCTION, ip, &tal);
+
+               if (tal.sym)
+                       fprintf(f, "..... %016" PRIx64 " %s\n", ip,
+                               tal.sym->name);
+               else
+                       fprintf(f, "..... %016" PRIx64 "\n", ip);
+       }
+
+exit:
+       fclose(f);
+
+       return p;
+}
+
+typedef int (*tracepoint_handler)(struct timechart *tchart,
+                                 struct perf_evsel *evsel,
+                                 struct perf_sample *sample,
+                                 const char *backtrace);
+
+static int process_sample_event(struct perf_tool *tool,
+                               union perf_event *event,
+                               struct perf_sample *sample,
+                               struct perf_evsel *evsel,
+                               struct machine *machine)
+{
+       struct timechart *tchart = container_of(tool, struct timechart, tool);
+
+       if (evsel->attr.sample_type & PERF_SAMPLE_TIME) {
+               if (!tchart->first_time || tchart->first_time > sample->time)
+                       tchart->first_time = sample->time;
+               if (tchart->last_time < sample->time)
+                       tchart->last_time = sample->time;
+       }
+
+       if (evsel->handler != NULL) {
+               tracepoint_handler f = evsel->handler;
+               return f(tchart, evsel, sample,
+                        cat_backtrace(event, sample, machine));
+       }
+
+       return 0;
+}
+
+static int
+process_sample_cpu_idle(struct timechart *tchart __maybe_unused,
+                       struct perf_evsel *evsel,
+                       struct perf_sample *sample,
+                       const char *backtrace __maybe_unused)
+{
+       u32 state = perf_evsel__intval(evsel, sample, "state");
+       u32 cpu_id = perf_evsel__intval(evsel, sample, "cpu_id");
+
+       if (state == (u32)PWR_EVENT_EXIT)
+               c_state_end(tchart, cpu_id, sample->time);
+       else
+               c_state_start(cpu_id, sample->time, state);
+       return 0;
+}
+
+static int
+process_sample_cpu_frequency(struct timechart *tchart,
+                            struct perf_evsel *evsel,
+                            struct perf_sample *sample,
+                            const char *backtrace __maybe_unused)
+{
+       u32 state = perf_evsel__intval(evsel, sample, "state");
+       u32 cpu_id = perf_evsel__intval(evsel, sample, "cpu_id");
+
+       p_state_change(tchart, cpu_id, sample->time, state);
+       return 0;
+}
+
+static int
+process_sample_sched_wakeup(struct timechart *tchart,
+                           struct perf_evsel *evsel,
+                           struct perf_sample *sample,
+                           const char *backtrace)
+{
+       u8 flags = perf_evsel__intval(evsel, sample, "common_flags");
+       int waker = perf_evsel__intval(evsel, sample, "common_pid");
+       int wakee = perf_evsel__intval(evsel, sample, "pid");
+
+       sched_wakeup(tchart, sample->cpu, sample->time, waker, wakee, flags, backtrace);
+       return 0;
+}
+
+static int
+process_sample_sched_switch(struct timechart *tchart,
+                           struct perf_evsel *evsel,
+                           struct perf_sample *sample,
+                           const char *backtrace)
+{
+       int prev_pid = perf_evsel__intval(evsel, sample, "prev_pid");
+       int next_pid = perf_evsel__intval(evsel, sample, "next_pid");
+       u64 prev_state = perf_evsel__intval(evsel, sample, "prev_state");
+
+       sched_switch(tchart, sample->cpu, sample->time, prev_pid, next_pid,
+                    prev_state, backtrace);
+       return 0;
+}
+
+#ifdef SUPPORT_OLD_POWER_EVENTS
+static int
+process_sample_power_start(struct timechart *tchart __maybe_unused,
+                          struct perf_evsel *evsel,
+                          struct perf_sample *sample,
+                          const char *backtrace __maybe_unused)
+{
+       u64 cpu_id = perf_evsel__intval(evsel, sample, "cpu_id");
+       u64 value = perf_evsel__intval(evsel, sample, "value");
+
+       c_state_start(cpu_id, sample->time, value);
+       return 0;
+}
+
+static int
+process_sample_power_end(struct timechart *tchart,
+                        struct perf_evsel *evsel __maybe_unused,
+                        struct perf_sample *sample,
+                        const char *backtrace __maybe_unused)
+{
+       c_state_end(tchart, sample->cpu, sample->time);
+       return 0;
+}
+
+static int
+process_sample_power_frequency(struct timechart *tchart,
+                              struct perf_evsel *evsel,
+                              struct perf_sample *sample,
+                              const char *backtrace __maybe_unused)
+{
+       u64 cpu_id = perf_evsel__intval(evsel, sample, "cpu_id");
+       u64 value = perf_evsel__intval(evsel, sample, "value");
+
+       p_state_change(tchart, cpu_id, sample->time, value);
+       return 0;
+}
+#endif /* SUPPORT_OLD_POWER_EVENTS */
+
+/*
+ * After the last sample we need to wrap up the current C/P state
+ * and close out each CPU for these.
+ */
+static void end_sample_processing(struct timechart *tchart)
+{
+       u64 cpu;
+       struct power_event *pwr;
+
+       for (cpu = 0; cpu <= tchart->numcpus; cpu++) {
+               /* C state */
+#if 0
+               pwr = zalloc(sizeof(*pwr));
+               if (!pwr)
+                       return;
+
+               pwr->state = cpus_cstate_state[cpu];
+               pwr->start_time = cpus_cstate_start_times[cpu];
+               pwr->end_time = tchart->last_time;
+               pwr->cpu = cpu;
+               pwr->type = CSTATE;
+               pwr->next = tchart->power_events;
+
+               tchart->power_events = pwr;
+#endif
+               /* P state */
+
+               pwr = zalloc(sizeof(*pwr));
+               if (!pwr)
+                       return;
+
+               pwr->state = cpus_pstate_state[cpu];
+               pwr->start_time = cpus_pstate_start_times[cpu];
+               pwr->end_time = tchart->last_time;
+               pwr->cpu = cpu;
+               pwr->type = PSTATE;
+               pwr->next = tchart->power_events;
+
+               if (!pwr->start_time)
+                       pwr->start_time = tchart->first_time;
+               if (!pwr->state)
+                       pwr->state = tchart->min_freq;
+               tchart->power_events = pwr;
+       }
+}
+
+static int pid_begin_io_sample(struct timechart *tchart, int pid, int type,
+                              u64 start, int fd)
+{
+       struct per_pid *p = find_create_pid(tchart, pid);
+       struct per_pidcomm *c = p->current;
+       struct io_sample *sample;
+       struct io_sample *prev;
+
+       if (!c) {
+               c = zalloc(sizeof(*c));
+               if (!c)
+                       return -ENOMEM;
+               p->current = c;
+               c->next = p->all;
+               p->all = c;
+       }
+
+       prev = c->io_samples;
+
+       if (prev && prev->start_time && !prev->end_time) {
+               pr_warning("Skip invalid start event: "
+                          "previous event already started!\n");
+
+               /* remove previous event that has been started,
+                * we are not sure we will ever get an end for it */
+               c->io_samples = prev->next;
+               free(prev);
+               return 0;
+       }
+
+       sample = zalloc(sizeof(*sample));
+       if (!sample)
+               return -ENOMEM;
+       sample->start_time = start;
+       sample->type = type;
+       sample->fd = fd;
+       sample->next = c->io_samples;
+       c->io_samples = sample;
+
+       if (c->start_time == 0 || c->start_time > start)
+               c->start_time = start;
+
+       return 0;
+}
+
+static int pid_end_io_sample(struct timechart *tchart, int pid, int type,
+                            u64 end, long ret)
+{
+       struct per_pid *p = find_create_pid(tchart, pid);
+       struct per_pidcomm *c = p->current;
+       struct io_sample *sample, *prev;
+
+       if (!c) {
+               pr_warning("Invalid pidcomm!\n");
+               return -1;
+       }
+
+       sample = c->io_samples;
+
+       if (!sample) /* skip partially captured events */
+               return 0;
+
+       if (sample->end_time) {
+               pr_warning("Skip invalid end event: "
+                          "previous event already ended!\n");
+               return 0;
+       }
+
+       if (sample->type != type) {
+               pr_warning("Skip invalid end event: invalid event type!\n");
+               return 0;
+       }
+
+       sample->end_time = end;
+       prev = sample->next;
+
+       /* we want to be able to see small and fast transfers, so make them
+        * at least min_time long, but don't overlap them */
+       if (sample->end_time - sample->start_time < tchart->min_time)
+               sample->end_time = sample->start_time + tchart->min_time;
+       if (prev && sample->start_time < prev->end_time) {
+               if (prev->err) /* try to make errors more visible */
+                       sample->start_time = prev->end_time;
+               else
+                       prev->end_time = sample->start_time;
+       }
+
+       if (ret < 0) {
+               sample->err = ret;
+       } else if (type == IOTYPE_READ || type == IOTYPE_WRITE ||
+                  type == IOTYPE_TX || type == IOTYPE_RX) {
+
+               if ((u64)ret > c->max_bytes)
+                       c->max_bytes = ret;
+
+               c->total_bytes += ret;
+               p->total_bytes += ret;
+               sample->bytes = ret;
+       }
+
+       /* merge two requests to make svg smaller and render-friendly */
+       if (prev &&
+           prev->type == sample->type &&
+           prev->err == sample->err &&
+           prev->fd == sample->fd &&
+           prev->end_time + tchart->merge_dist >= sample->start_time) {
+
+               sample->bytes += prev->bytes;
+               sample->merges += prev->merges + 1;
+
+               sample->start_time = prev->start_time;
+               sample->next = prev->next;
+               free(prev);
+
+               if (!sample->err && sample->bytes > c->max_bytes)
+                       c->max_bytes = sample->bytes;
+       }
+
+       tchart->io_events++;
+
+       return 0;
+}
+
+static int
+process_enter_read(struct timechart *tchart,
+                  struct perf_evsel *evsel,
+                  struct perf_sample *sample)
+{
+       long fd = perf_evsel__intval(evsel, sample, "fd");
+       return pid_begin_io_sample(tchart, sample->tid, IOTYPE_READ,
+                                  sample->time, fd);
+}
+
+static int
+process_exit_read(struct timechart *tchart,
+                 struct perf_evsel *evsel,
+                 struct perf_sample *sample)
+{
+       long ret = perf_evsel__intval(evsel, sample, "ret");
+       return pid_end_io_sample(tchart, sample->tid, IOTYPE_READ,
+                                sample->time, ret);
+}
+
+static int
+process_enter_write(struct timechart *tchart,
+                   struct perf_evsel *evsel,
+                   struct perf_sample *sample)
+{
+       long fd = perf_evsel__intval(evsel, sample, "fd");
+       return pid_begin_io_sample(tchart, sample->tid, IOTYPE_WRITE,
+                                  sample->time, fd);
+}
+
+static int
+process_exit_write(struct timechart *tchart,
+                  struct perf_evsel *evsel,
+                  struct perf_sample *sample)
+{
+       long ret = perf_evsel__intval(evsel, sample, "ret");
+       return pid_end_io_sample(tchart, sample->tid, IOTYPE_WRITE,
+                                sample->time, ret);
+}
+
+static int
+process_enter_sync(struct timechart *tchart,
+                  struct perf_evsel *evsel,
+                  struct perf_sample *sample)
+{
+       long fd = perf_evsel__intval(evsel, sample, "fd");
+       return pid_begin_io_sample(tchart, sample->tid, IOTYPE_SYNC,
+                                  sample->time, fd);
+}
+
+static int
+process_exit_sync(struct timechart *tchart,
+                 struct perf_evsel *evsel,
+                 struct perf_sample *sample)
+{
+       long ret = perf_evsel__intval(evsel, sample, "ret");
+       return pid_end_io_sample(tchart, sample->tid, IOTYPE_SYNC,
+                                sample->time, ret);
+}
+
+static int
+process_enter_tx(struct timechart *tchart,
+                struct perf_evsel *evsel,
+                struct perf_sample *sample)
+{
+       long fd = perf_evsel__intval(evsel, sample, "fd");
+       return pid_begin_io_sample(tchart, sample->tid, IOTYPE_TX,
+                                  sample->time, fd);
+}
+
+static int
+process_exit_tx(struct timechart *tchart,
+               struct perf_evsel *evsel,
+               struct perf_sample *sample)
+{
+       long ret = perf_evsel__intval(evsel, sample, "ret");
+       return pid_end_io_sample(tchart, sample->tid, IOTYPE_TX,
+                                sample->time, ret);
+}
+
+static int
+process_enter_rx(struct timechart *tchart,
+                struct perf_evsel *evsel,
+                struct perf_sample *sample)
+{
+       long fd = perf_evsel__intval(evsel, sample, "fd");
+       return pid_begin_io_sample(tchart, sample->tid, IOTYPE_RX,
+                                  sample->time, fd);
+}
+
+static int
+process_exit_rx(struct timechart *tchart,
+               struct perf_evsel *evsel,
+               struct perf_sample *sample)
+{
+       long ret = perf_evsel__intval(evsel, sample, "ret");
+       return pid_end_io_sample(tchart, sample->tid, IOTYPE_RX,
+                                sample->time, ret);
+}
+
+static int
+process_enter_poll(struct timechart *tchart,
+                  struct perf_evsel *evsel,
+                  struct perf_sample *sample)
+{
+       long fd = perf_evsel__intval(evsel, sample, "fd");
+       return pid_begin_io_sample(tchart, sample->tid, IOTYPE_POLL,
+                                  sample->time, fd);
+}
+
+static int
+process_exit_poll(struct timechart *tchart,
+                 struct perf_evsel *evsel,
+                 struct perf_sample *sample)
+{
+       long ret = perf_evsel__intval(evsel, sample, "ret");
+       return pid_end_io_sample(tchart, sample->tid, IOTYPE_POLL,
+                                sample->time, ret);
+}
+
+/*
+ * Sort the pid datastructure
+ */
+static void sort_pids(struct timechart *tchart)
+{
+       struct per_pid *new_list, *p, *cursor, *prev;
+       /* sort by ppid first, then by pid, lowest to highest */
+
+       new_list = NULL;
+
+       while (tchart->all_data) {
+               p = tchart->all_data;
+               tchart->all_data = p->next;
+               p->next = NULL;
+
+               if (new_list == NULL) {
+                       new_list = p;
+                       p->next = NULL;
+                       continue;
+               }
+               prev = NULL;
+               cursor = new_list;
+               while (cursor) {
+                       if (cursor->ppid > p->ppid ||
+                               (cursor->ppid == p->ppid && cursor->pid > p->pid)) {
+                               /* must insert before */
+                               if (prev) {
+                                       p->next = prev->next;
+                                       prev->next = p;
+                                       cursor = NULL;
+                                       continue;
+                               } else {
+                                       p->next = new_list;
+                                       new_list = p;
+                                       cursor = NULL;
+                                       continue;
+                               }
+                       }
+
+                       prev = cursor;
+                       cursor = cursor->next;
+                       if (!cursor)
+                               prev->next = p;
+               }
+       }
+       tchart->all_data = new_list;
+}
+
+
+static void draw_c_p_states(struct timechart *tchart)
+{
+       struct power_event *pwr;
+       pwr = tchart->power_events;
+
+       /*
+        * two pass drawing so that the P state bars are on top of the C state blocks
+        */
+       while (pwr) {
+               if (pwr->type == CSTATE)
+                       svg_cstate(pwr->cpu, pwr->start_time, pwr->end_time, pwr->state);
+               pwr = pwr->next;
+       }
+
+       pwr = tchart->power_events;
+       while (pwr) {
+               if (pwr->type == PSTATE) {
+                       if (!pwr->state)
+                               pwr->state = tchart->min_freq;
+                       svg_pstate(pwr->cpu, pwr->start_time, pwr->end_time, pwr->state);
+               }
+               pwr = pwr->next;
+       }
+}
+
+static void draw_wakeups(struct timechart *tchart)
+{
+       struct wake_event *we;
+       struct per_pid *p;
+       struct per_pidcomm *c;
+
+       we = tchart->wake_events;
+       while (we) {
+               int from = 0, to = 0;
+               char *task_from = NULL, *task_to = NULL;
+
+               /* locate the column of the waker and wakee */
+               p = tchart->all_data;
+               while (p) {
+                       if (p->pid == we->waker || p->pid == we->wakee) {
+                               c = p->all;
+                               while (c) {
+                                       if (c->Y && c->start_time <= we->time && c->end_time >= we->time) {
+                                               if (p->pid == we->waker && !from) {
+                                                       from = c->Y;
+                                                       task_from = strdup(c->comm);
+                                               }
+                                               if (p->pid == we->wakee && !to) {
+                                                       to = c->Y;
+                                                       task_to = strdup(c->comm);
+                                               }
+                                       }
+                                       c = c->next;
+                               }
+                               c = p->all;
+                               while (c) {
+                                       if (p->pid == we->waker && !from) {
+                                               from = c->Y;
+                                               task_from = strdup(c->comm);
+                                       }
+                                       if (p->pid == we->wakee && !to) {
+                                               to = c->Y;
+                                               task_to = strdup(c->comm);
+                                       }
+                                       c = c->next;
+                               }
+                       }
+                       p = p->next;
+               }
+
+               if (!task_from) {
+                       task_from = malloc(40);
+                       sprintf(task_from, "[%i]", we->waker);
+               }
+               if (!task_to) {
+                       task_to = malloc(40);
+                       sprintf(task_to, "[%i]", we->wakee);
+               }
+
+               if (we->waker == -1)
+                       svg_interrupt(we->time, to, we->backtrace);
+               else if (from && to && abs(from - to) == 1)
+                       svg_wakeline(we->time, from, to, we->backtrace);
+               else
+                       svg_partial_wakeline(we->time, from, task_from, to,
+                                            task_to, we->backtrace);
+               we = we->next;
+
+               free(task_from);
+               free(task_to);
+       }
+}
+
+static void draw_cpu_usage(struct timechart *tchart)
+{
+       struct per_pid *p;
+       struct per_pidcomm *c;
+       struct cpu_sample *sample;
+       p = tchart->all_data;
+       while (p) {
+               c = p->all;
+               while (c) {
+                       sample = c->samples;
+                       while (sample) {
+                               if (sample->type == TYPE_RUNNING) {
+                                       svg_process(sample->cpu,
+                                                   sample->start_time,
+                                                   sample->end_time,
+                                                   p->pid,
+                                                   c->comm,
+                                                   sample->backtrace);
+                               }
+
+                               sample = sample->next;
+                       }
+                       c = c->next;
+               }
+               p = p->next;
+       }
+}
+
+static void draw_io_bars(struct timechart *tchart)
+{
+       const char *suf;
+       double bytes;
+       char comm[256];
+       struct per_pid *p;
+       struct per_pidcomm *c;
+       struct io_sample *sample;
+       int Y = 1;
+
+       p = tchart->all_data;
+       while (p) {
+               c = p->all;
+               while (c) {
+                       if (!c->display) {
+                               c->Y = 0;
+                               c = c->next;
+                               continue;
+                       }
+
+                       svg_box(Y, c->start_time, c->end_time, "process3");
+                       sample = c->io_samples;
+                       for (sample = c->io_samples; sample; sample = sample->next) {
+                               double h = (double)sample->bytes / c->max_bytes;
+
+                               if (tchart->skip_eagain &&
+                                   sample->err == -EAGAIN)
+                                       continue;
+
+                               if (sample->err)
+                                       h = 1;
+
+                               if (sample->type == IOTYPE_SYNC)
+                                       svg_fbox(Y,
+                                               sample->start_time,
+                                               sample->end_time,
+                                               1,
+                                               sample->err ? "error" : "sync",
+                                               sample->fd,
+                                               sample->err,
+                                               sample->merges);
+                               else if (sample->type == IOTYPE_POLL)
+                                       svg_fbox(Y,
+                                               sample->start_time,
+                                               sample->end_time,
+                                               1,
+                                               sample->err ? "error" : "poll",
+                                               sample->fd,
+                                               sample->err,
+                                               sample->merges);
+                               else if (sample->type == IOTYPE_READ)
+                                       svg_ubox(Y,
+                                               sample->start_time,
+                                               sample->end_time,
+                                               h,
+                                               sample->err ? "error" : "disk",
+                                               sample->fd,
+                                               sample->err,
+                                               sample->merges);
+                               else if (sample->type == IOTYPE_WRITE)
+                                       svg_lbox(Y,
+                                               sample->start_time,
+                                               sample->end_time,
+                                               h,
+                                               sample->err ? "error" : "disk",
+                                               sample->fd,
+                                               sample->err,
+                                               sample->merges);
+                               else if (sample->type == IOTYPE_RX)
+                                       svg_ubox(Y,
+                                               sample->start_time,
+                                               sample->end_time,
+                                               h,
+                                               sample->err ? "error" : "net",
+                                               sample->fd,
+                                               sample->err,
+                                               sample->merges);
+                               else if (sample->type == IOTYPE_TX)
+                                       svg_lbox(Y,
+                                               sample->start_time,
+                                               sample->end_time,
+                                               h,
+                                               sample->err ? "error" : "net",
+                                               sample->fd,
+                                               sample->err,
+                                               sample->merges);
+                       }
+
+                       suf = "";
+                       bytes = c->total_bytes;
+                       if (bytes > 1024) {
+                               bytes = bytes / 1024;
+                               suf = "K";
+                       }
+                       if (bytes > 1024) {
+                               bytes = bytes / 1024;
+                               suf = "M";
+                       }
+                       if (bytes > 1024) {
+                               bytes = bytes / 1024;
+                               suf = "G";
+                       }
+
+
+                       sprintf(comm, "%s:%i (%3.1f %sbytes)", c->comm ?: "", p->pid, bytes, suf);
+                       svg_text(Y, c->start_time, comm);
+
+                       c->Y = Y;
+                       Y++;
+                       c = c->next;
+               }
+               p = p->next;
+       }
+}
+
+static void draw_process_bars(struct timechart *tchart)
+{
+       struct per_pid *p;
+       struct per_pidcomm *c;
+       struct cpu_sample *sample;
+       int Y = 0;
+
+       Y = 2 * tchart->numcpus + 2;
+
+       p = tchart->all_data;
+       while (p) {
+               c = p->all;
+               while (c) {
+                       if (!c->display) {
+                               c->Y = 0;
+                               c = c->next;
+                               continue;
+                       }
+
+                       svg_box(Y, c->start_time, c->end_time, "process");
+                       sample = c->samples;
+                       while (sample) {
+                               if (sample->type == TYPE_RUNNING)
+                                       svg_running(Y, sample->cpu,
+                                                   sample->start_time,
+                                                   sample->end_time,
+                                                   sample->backtrace);
+                               if (sample->type == TYPE_BLOCKED)
+                                       svg_blocked(Y, sample->cpu,
+                                                   sample->start_time,
+                                                   sample->end_time,
+                                                   sample->backtrace);
+                               if (sample->type == TYPE_WAITING)
+                                       svg_waiting(Y, sample->cpu,
+                                                   sample->start_time,
+                                                   sample->end_time,
+                                                   sample->backtrace);
+                               sample = sample->next;
+                       }
+
+                       if (c->comm) {
+                               char comm[256];
+                               if (c->total_time > 5000000000) /* 5 seconds */
+                                       sprintf(comm, "%s:%i (%2.2fs)", c->comm, p->pid, c->total_time / 1000000000.0);
+                               else
+                                       sprintf(comm, "%s:%i (%3.1fms)", c->comm, p->pid, c->total_time / 1000000.0);
+
+                               svg_text(Y, c->start_time, comm);
+                       }
+                       c->Y = Y;
+                       Y++;
+                       c = c->next;
+               }
+               p = p->next;
+       }
+}
+
+static void add_process_filter(const char *string)
+{
+       int pid = strtoull(string, NULL, 10);
+       struct process_filter *filt = malloc(sizeof(*filt));
+
+       if (!filt)
+               return;
+
+       filt->name = strdup(string);
+       filt->pid  = pid;
+       filt->next = process_filter;
+
+       process_filter = filt;
+}
+
+static int passes_filter(struct per_pid *p, struct per_pidcomm *c)
+{
+       struct process_filter *filt;
+       if (!process_filter)
+               return 1;
+
+       filt = process_filter;
+       while (filt) {
+               if (filt->pid && p->pid == filt->pid)
+                       return 1;
+               if (strcmp(filt->name, c->comm) == 0)
+                       return 1;
+               filt = filt->next;
+       }
+       return 0;
+}
+
+static int determine_display_tasks_filtered(struct timechart *tchart)
+{
+       struct per_pid *p;
+       struct per_pidcomm *c;
+       int count = 0;
+
+       p = tchart->all_data;
+       while (p) {
+               p->display = 0;
+               if (p->start_time == 1)
+                       p->start_time = tchart->first_time;
+
+               /* no exit marker, task kept running to the end */
+               if (p->end_time == 0)
+                       p->end_time = tchart->last_time;
+
+               c = p->all;
+
+               while (c) {
+                       c->display = 0;
+
+                       if (c->start_time == 1)
+                               c->start_time = tchart->first_time;
+
+                       if (passes_filter(p, c)) {
+                               c->display = 1;
+                               p->display = 1;
+                               count++;
+                       }
+
+                       if (c->end_time == 0)
+                               c->end_time = tchart->last_time;
+
+                       c = c->next;
+               }
+               p = p->next;
+       }
+       return count;
+}
+
+static int determine_display_tasks(struct timechart *tchart, u64 threshold)
+{
+       struct per_pid *p;
+       struct per_pidcomm *c;
+       int count = 0;
+
+       p = tchart->all_data;
+       while (p) {
+               p->display = 0;
+               if (p->start_time == 1)
+                       p->start_time = tchart->first_time;
+
+               /* no exit marker, task kept running to the end */
+               if (p->end_time == 0)
+                       p->end_time = tchart->last_time;
+               if (p->total_time >= threshold)
+                       p->display = 1;
+
+               c = p->all;
+
+               while (c) {
+                       c->display = 0;
+
+                       if (c->start_time == 1)
+                               c->start_time = tchart->first_time;
+
+                       if (c->total_time >= threshold) {
+                               c->display = 1;
+                               count++;
+                       }
+
+                       if (c->end_time == 0)
+                               c->end_time = tchart->last_time;
+
+                       c = c->next;
+               }
+               p = p->next;
+       }
+       return count;
+}
+
+static int determine_display_io_tasks(struct timechart *timechart, u64 threshold)
+{
+       struct per_pid *p;
+       struct per_pidcomm *c;
+       int count = 0;
+
+       p = timechart->all_data;
+       while (p) {
+               /* no exit marker, task kept running to the end */
+               if (p->end_time == 0)
+                       p->end_time = timechart->last_time;
+
+               c = p->all;
+
+               while (c) {
+                       c->display = 0;
+
+                       if (c->total_bytes >= threshold) {
+                               c->display = 1;
+                               count++;
+                       }
+
+                       if (c->end_time == 0)
+                               c->end_time = timechart->last_time;
+
+                       c = c->next;
+               }
+               p = p->next;
+       }
+       return count;
+}
+
+#define BYTES_THRESH (1 * 1024 * 1024)
+#define TIME_THRESH 10000000
+
+static void write_svg_file(struct timechart *tchart, const char *filename)
+{
+       u64 i;
+       int count;
+       int thresh = tchart->io_events ? BYTES_THRESH : TIME_THRESH;
+
+       if (tchart->power_only)
+               tchart->proc_num = 0;
+
+       /* We'd like to show at least proc_num tasks;
+        * be less picky if we have fewer */
+       do {
+               if (process_filter)
+                       count = determine_display_tasks_filtered(tchart);
+               else if (tchart->io_events)
+                       count = determine_display_io_tasks(tchart, thresh);
+               else
+                       count = determine_display_tasks(tchart, thresh);
+               thresh /= 10;
+       } while (!process_filter && thresh && count < tchart->proc_num);
+
+       if (!tchart->proc_num)
+               count = 0;
+
+       if (tchart->io_events) {
+               open_svg(filename, 0, count, tchart->first_time, tchart->last_time);
+
+               svg_time_grid(0.5);
+               svg_io_legenda();
+
+               draw_io_bars(tchart);
+       } else {
+               open_svg(filename, tchart->numcpus, count, tchart->first_time, tchart->last_time);
+
+               svg_time_grid(0);
+
+               svg_legenda();
+
+               for (i = 0; i < tchart->numcpus; i++)
+                       svg_cpu_box(i, tchart->max_freq, tchart->turbo_frequency);
+
+               draw_cpu_usage(tchart);
+               if (tchart->proc_num)
+                       draw_process_bars(tchart);
+               if (!tchart->tasks_only)
+                       draw_c_p_states(tchart);
+               if (tchart->proc_num)
+                       draw_wakeups(tchart);
+       }
+
+       svg_close();
+}
+
+static int process_header(struct perf_file_section *section __maybe_unused,
+                         struct perf_header *ph,
+                         int feat,
+                         int fd __maybe_unused,
+                         void *data)
+{
+       struct timechart *tchart = data;
+
+       switch (feat) {
+       case HEADER_NRCPUS:
+               tchart->numcpus = ph->env.nr_cpus_avail;
+               break;
+
+       case HEADER_CPU_TOPOLOGY:
+               if (!tchart->topology)
+                       break;
+
+               if (svg_build_topology_map(ph->env.sibling_cores,
+                                          ph->env.nr_sibling_cores,
+                                          ph->env.sibling_threads,
+                                          ph->env.nr_sibling_threads))
+                       fprintf(stderr, "problem building topology\n");
+               break;
+
+       default:
+               break;
+       }
+
+       return 0;
+}
+
+static int __cmd_timechart(struct timechart *tchart, const char *output_name)
+{
+       const struct perf_evsel_str_handler power_tracepoints[] = {
+               { "power:cpu_idle",             process_sample_cpu_idle },
+               { "power:cpu_frequency",        process_sample_cpu_frequency },
+               { "sched:sched_wakeup",         process_sample_sched_wakeup },
+               { "sched:sched_switch",         process_sample_sched_switch },
+#ifdef SUPPORT_OLD_POWER_EVENTS
+               { "power:power_start",          process_sample_power_start },
+               { "power:power_end",            process_sample_power_end },
+               { "power:power_frequency",      process_sample_power_frequency },
+#endif
+
+               { "syscalls:sys_enter_read",            process_enter_read },
+               { "syscalls:sys_enter_pread64",         process_enter_read },
+               { "syscalls:sys_enter_readv",           process_enter_read },
+               { "syscalls:sys_enter_preadv",          process_enter_read },
+               { "syscalls:sys_enter_write",           process_enter_write },
+               { "syscalls:sys_enter_pwrite64",        process_enter_write },
+               { "syscalls:sys_enter_writev",          process_enter_write },
+               { "syscalls:sys_enter_pwritev",         process_enter_write },
+               { "syscalls:sys_enter_sync",            process_enter_sync },
+               { "syscalls:sys_enter_sync_file_range", process_enter_sync },
+               { "syscalls:sys_enter_fsync",           process_enter_sync },
+               { "syscalls:sys_enter_msync",           process_enter_sync },
+               { "syscalls:sys_enter_recvfrom",        process_enter_rx },
+               { "syscalls:sys_enter_recvmmsg",        process_enter_rx },
+               { "syscalls:sys_enter_recvmsg",         process_enter_rx },
+               { "syscalls:sys_enter_sendto",          process_enter_tx },
+               { "syscalls:sys_enter_sendmsg",         process_enter_tx },
+               { "syscalls:sys_enter_sendmmsg",        process_enter_tx },
+               { "syscalls:sys_enter_epoll_pwait",     process_enter_poll },
+               { "syscalls:sys_enter_epoll_wait",      process_enter_poll },
+               { "syscalls:sys_enter_poll",            process_enter_poll },
+               { "syscalls:sys_enter_ppoll",           process_enter_poll },
+               { "syscalls:sys_enter_pselect6",        process_enter_poll },
+               { "syscalls:sys_enter_select",          process_enter_poll },
+
+               { "syscalls:sys_exit_read",             process_exit_read },
+               { "syscalls:sys_exit_pread64",          process_exit_read },
+               { "syscalls:sys_exit_readv",            process_exit_read },
+               { "syscalls:sys_exit_preadv",           process_exit_read },
+               { "syscalls:sys_exit_write",            process_exit_write },
+               { "syscalls:sys_exit_pwrite64",         process_exit_write },
+               { "syscalls:sys_exit_writev",           process_exit_write },
+               { "syscalls:sys_exit_pwritev",          process_exit_write },
+               { "syscalls:sys_exit_sync",             process_exit_sync },
+               { "syscalls:sys_exit_sync_file_range",  process_exit_sync },
+               { "syscalls:sys_exit_fsync",            process_exit_sync },
+               { "syscalls:sys_exit_msync",            process_exit_sync },
+               { "syscalls:sys_exit_recvfrom",         process_exit_rx },
+               { "syscalls:sys_exit_recvmmsg",         process_exit_rx },
+               { "syscalls:sys_exit_recvmsg",          process_exit_rx },
+               { "syscalls:sys_exit_sendto",           process_exit_tx },
+               { "syscalls:sys_exit_sendmsg",          process_exit_tx },
+               { "syscalls:sys_exit_sendmmsg",         process_exit_tx },
+               { "syscalls:sys_exit_epoll_pwait",      process_exit_poll },
+               { "syscalls:sys_exit_epoll_wait",       process_exit_poll },
+               { "syscalls:sys_exit_poll",             process_exit_poll },
+               { "syscalls:sys_exit_ppoll",            process_exit_poll },
+               { "syscalls:sys_exit_pselect6",         process_exit_poll },
+               { "syscalls:sys_exit_select",           process_exit_poll },
+       };
+       struct perf_data_file file = {
+               .path = input_name,
+               .mode = PERF_DATA_MODE_READ,
+               .force = tchart->force,
+       };
+
+       struct perf_session *session = perf_session__new(&file, false,
+                                                        &tchart->tool);
+       int ret = -EINVAL;
+
+       if (session == NULL)
+               return -1;
+
+       symbol__init(&session->header.env);
+
+       (void)perf_header__process_sections(&session->header,
+                                           perf_data_file__fd(session->file),
+                                           tchart,
+                                           process_header);
+
+       if (!perf_session__has_traces(session, "timechart record"))
+               goto out_delete;
+
+       if (perf_session__set_tracepoints_handlers(session,
+                                                  power_tracepoints)) {
+               pr_err("Initializing session tracepoint handlers failed\n");
+               goto out_delete;
+       }
+
+       ret = perf_session__process_events(session);
+       if (ret)
+               goto out_delete;
+
+       end_sample_processing(tchart);
+
+       sort_pids(tchart);
+
+       write_svg_file(tchart, output_name);
+
+       pr_info("Written %2.1f seconds of trace to %s.\n",
+               (tchart->last_time - tchart->first_time) / 1000000000.0, output_name);
+out_delete:
+       perf_session__delete(session);
+       return ret;
+}
+
+static int timechart__io_record(int argc, const char **argv)
+{
+       unsigned int rec_argc, i;
+       const char **rec_argv;
+       const char **p;
+       char *filter = NULL;
+
+       const char * const common_args[] = {
+               "record", "-a", "-R", "-c", "1",
+       };
+       unsigned int common_args_nr = ARRAY_SIZE(common_args);
+
+       const char * const disk_events[] = {
+               "syscalls:sys_enter_read",
+               "syscalls:sys_enter_pread64",
+               "syscalls:sys_enter_readv",
+               "syscalls:sys_enter_preadv",
+               "syscalls:sys_enter_write",
+               "syscalls:sys_enter_pwrite64",
+               "syscalls:sys_enter_writev",
+               "syscalls:sys_enter_pwritev",
+               "syscalls:sys_enter_sync",
+               "syscalls:sys_enter_sync_file_range",
+               "syscalls:sys_enter_fsync",
+               "syscalls:sys_enter_msync",
+
+               "syscalls:sys_exit_read",
+               "syscalls:sys_exit_pread64",
+               "syscalls:sys_exit_readv",
+               "syscalls:sys_exit_preadv",
+               "syscalls:sys_exit_write",
+               "syscalls:sys_exit_pwrite64",
+               "syscalls:sys_exit_writev",
+               "syscalls:sys_exit_pwritev",
+               "syscalls:sys_exit_sync",
+               "syscalls:sys_exit_sync_file_range",
+               "syscalls:sys_exit_fsync",
+               "syscalls:sys_exit_msync",
+       };
+       unsigned int disk_events_nr = ARRAY_SIZE(disk_events);
+
+       const char * const net_events[] = {
+               "syscalls:sys_enter_recvfrom",
+               "syscalls:sys_enter_recvmmsg",
+               "syscalls:sys_enter_recvmsg",
+               "syscalls:sys_enter_sendto",
+               "syscalls:sys_enter_sendmsg",
+               "syscalls:sys_enter_sendmmsg",
+
+               "syscalls:sys_exit_recvfrom",
+               "syscalls:sys_exit_recvmmsg",
+               "syscalls:sys_exit_recvmsg",
+               "syscalls:sys_exit_sendto",
+               "syscalls:sys_exit_sendmsg",
+               "syscalls:sys_exit_sendmmsg",
+       };
+       unsigned int net_events_nr = ARRAY_SIZE(net_events);
+
+       const char * const poll_events[] = {
+               "syscalls:sys_enter_epoll_pwait",
+               "syscalls:sys_enter_epoll_wait",
+               "syscalls:sys_enter_poll",
+               "syscalls:sys_enter_ppoll",
+               "syscalls:sys_enter_pselect6",
+               "syscalls:sys_enter_select",
+
+               "syscalls:sys_exit_epoll_pwait",
+               "syscalls:sys_exit_epoll_wait",
+               "syscalls:sys_exit_poll",
+               "syscalls:sys_exit_ppoll",
+               "syscalls:sys_exit_pselect6",
+               "syscalls:sys_exit_select",
+       };
+       unsigned int poll_events_nr = ARRAY_SIZE(poll_events);
+
+       rec_argc = common_args_nr +
+               disk_events_nr * 4 +
+               net_events_nr * 4 +
+               poll_events_nr * 4 +
+               argc;
+       rec_argv = calloc(rec_argc + 1, sizeof(char *));
+
+       if (rec_argv == NULL)
+               return -ENOMEM;
+
+       if (asprintf(&filter, "common_pid != %d", getpid()) < 0)
+               return -ENOMEM;
+
+       p = rec_argv;
+       for (i = 0; i < common_args_nr; i++)
+               *p++ = strdup(common_args[i]);
+
+       for (i = 0; i < disk_events_nr; i++) {
+               if (!is_valid_tracepoint(disk_events[i])) {
+                       rec_argc -= 4;
+                       continue;
+               }
+
+               *p++ = "-e";
+               *p++ = strdup(disk_events[i]);
+               *p++ = "--filter";
+               *p++ = filter;
+       }
+       for (i = 0; i < net_events_nr; i++) {
+               if (!is_valid_tracepoint(net_events[i])) {
+                       rec_argc -= 4;
+                       continue;
+               }
+
+               *p++ = "-e";
+               *p++ = strdup(net_events[i]);
+               *p++ = "--filter";
+               *p++ = filter;
+       }
+       for (i = 0; i < poll_events_nr; i++) {
+               if (!is_valid_tracepoint(poll_events[i])) {
+                       rec_argc -= 4;
+                       continue;
+               }
+
+               *p++ = "-e";
+               *p++ = strdup(poll_events[i]);
+               *p++ = "--filter";
+               *p++ = filter;
+       }
+
+       for (i = 0; i < (unsigned int)argc; i++)
+               *p++ = argv[i];
+
+       return cmd_record(rec_argc, rec_argv, NULL);
+}
+
+
+static int timechart__record(struct timechart *tchart, int argc, const char **argv)
+{
+       unsigned int rec_argc, i, j;
+       const char **rec_argv;
+       const char **p;
+       unsigned int record_elems;
+
+       const char * const common_args[] = {
+               "record", "-a", "-R", "-c", "1",
+       };
+       unsigned int common_args_nr = ARRAY_SIZE(common_args);
+
+       const char * const backtrace_args[] = {
+               "-g",
+       };
+       unsigned int backtrace_args_no = ARRAY_SIZE(backtrace_args);
+
+       const char * const power_args[] = {
+               "-e", "power:cpu_frequency",
+               "-e", "power:cpu_idle",
+       };
+       unsigned int power_args_nr = ARRAY_SIZE(power_args);
+
+       const char * const old_power_args[] = {
+#ifdef SUPPORT_OLD_POWER_EVENTS
+               "-e", "power:power_start",
+               "-e", "power:power_end",
+               "-e", "power:power_frequency",
+#endif
+       };
+       unsigned int old_power_args_nr = ARRAY_SIZE(old_power_args);
+
+       const char * const tasks_args[] = {
+               "-e", "sched:sched_wakeup",
+               "-e", "sched:sched_switch",
+       };
+       unsigned int tasks_args_nr = ARRAY_SIZE(tasks_args);
+
+#ifdef SUPPORT_OLD_POWER_EVENTS
+       if (!is_valid_tracepoint("power:cpu_idle") &&
+           is_valid_tracepoint("power:power_start")) {
+               use_old_power_events = 1;
+               power_args_nr = 0;
+       } else {
+               old_power_args_nr = 0;
+       }
+#endif
+
+       if (tchart->power_only)
+               tasks_args_nr = 0;
+
+       if (tchart->tasks_only) {
+               power_args_nr = 0;
+               old_power_args_nr = 0;
+       }
+
+       if (!tchart->with_backtrace)
+               backtrace_args_no = 0;
+
+       record_elems = common_args_nr + tasks_args_nr +
+               power_args_nr + old_power_args_nr + backtrace_args_no;
+
+       rec_argc = record_elems + argc;
+       rec_argv = calloc(rec_argc + 1, sizeof(char *));
+
+       if (rec_argv == NULL)
+               return -ENOMEM;
+
+       p = rec_argv;
+       for (i = 0; i < common_args_nr; i++)
+               *p++ = strdup(common_args[i]);
+
+       for (i = 0; i < backtrace_args_no; i++)
+               *p++ = strdup(backtrace_args[i]);
+
+       for (i = 0; i < tasks_args_nr; i++)
+               *p++ = strdup(tasks_args[i]);
+
+       for (i = 0; i < power_args_nr; i++)
+               *p++ = strdup(power_args[i]);
+
+       for (i = 0; i < old_power_args_nr; i++)
+               *p++ = strdup(old_power_args[i]);
+
+       for (j = 0; j < (unsigned int)argc; j++)
+               *p++ = argv[j];
+
+       return cmd_record(rec_argc, rec_argv, NULL);
+}
+
+static int
+parse_process(const struct option *opt __maybe_unused, const char *arg,
+             int __maybe_unused unset)
+{
+       if (arg)
+               add_process_filter(arg);
+       return 0;
+}
+
+static int
+parse_highlight(const struct option *opt __maybe_unused, const char *arg,
+               int __maybe_unused unset)
+{
+       unsigned long duration = strtoul(arg, NULL, 0);
+
+       if (svg_highlight || svg_highlight_name)
+               return -1;
+
+       if (duration)
+               svg_highlight = duration;
+       else
+               svg_highlight_name = strdup(arg);
+
+       return 0;
+}
+
+static int
+parse_time(const struct option *opt, const char *arg, int __maybe_unused unset)
+{
+       char unit = 'n';
+       u64 *value = opt->value;
+
+       if (sscanf(arg, "%" PRIu64 "%cs", value, &unit) > 0) {
+               switch (unit) {
+               case 'm':
+                       *value *= 1000000;
+                       break;
+               case 'u':
+                       *value *= 1000;
+                       break;
+               case 'n':
+                       break;
+               default:
+                       return -1;
+               }
+       }
+
+       return 0;
+}
+
+int cmd_timechart(int argc, const char **argv,
+                 const char *prefix __maybe_unused)
+{
+       struct timechart tchart = {
+               .tool = {
+                       .comm            = process_comm_event,
+                       .fork            = process_fork_event,
+                       .exit            = process_exit_event,
+                       .sample          = process_sample_event,
+                       .ordered_events  = true,
+               },
+               .proc_num = 15,
+               .min_time = 1000000,
+               .merge_dist = 1000,
+       };
+       const char *output_name = "output.svg";
+       const struct option timechart_options[] = {
+       OPT_STRING('i', "input", &input_name, "file", "input file name"),
+       OPT_STRING('o', "output", &output_name, "file", "output file name"),
+       OPT_INTEGER('w', "width", &svg_page_width, "page width"),
+       OPT_CALLBACK(0, "highlight", NULL, "duration or task name",
+                     "highlight tasks. Pass duration in ns or process name.",
+                      parse_highlight),
+       OPT_BOOLEAN('P', "power-only", &tchart.power_only, "output power data only"),
+       OPT_BOOLEAN('T', "tasks-only", &tchart.tasks_only,
+                   "output processes data only"),
+       OPT_CALLBACK('p', "process", NULL, "process",
+                     "process selector. Pass a pid or process name.",
+                      parse_process),
+       OPT_STRING(0, "symfs", &symbol_conf.symfs, "directory",
+                   "Look for files with symbols relative to this directory"),
+       OPT_INTEGER('n', "proc-num", &tchart.proc_num,
+                   "min. number of tasks to print"),
+       OPT_BOOLEAN('t', "topology", &tchart.topology,
+                   "sort CPUs according to topology"),
+       OPT_BOOLEAN(0, "io-skip-eagain", &tchart.skip_eagain,
+                   "skip EAGAIN errors"),
+       OPT_CALLBACK(0, "io-min-time", &tchart.min_time, "time",
+                    "all IO faster than min-time will visually appear longer",
+                    parse_time),
+       OPT_CALLBACK(0, "io-merge-dist", &tchart.merge_dist, "time",
+                    "merge events that are merge-dist us apart",
+                    parse_time),
+       OPT_BOOLEAN('f', "force", &tchart.force, "don't complain, do it"),
+       OPT_END()
+       };
+       const char * const timechart_subcommands[] = { "record", NULL };
+       const char *timechart_usage[] = {
+               "perf timechart [<options>] {record}",
+               NULL
+       };
+
+       const struct option timechart_record_options[] = {
+       OPT_BOOLEAN('P', "power-only", &tchart.power_only, "output power data only"),
+       OPT_BOOLEAN('T', "tasks-only", &tchart.tasks_only,
+                   "output processes data only"),
+       OPT_BOOLEAN('I', "io-only", &tchart.io_only,
+                   "record only IO data"),
+       OPT_BOOLEAN('g', "callchain", &tchart.with_backtrace, "record callchain"),
+       OPT_END()
+       };
+       const char * const timechart_record_usage[] = {
+               "perf timechart record [<options>]",
+               NULL
+       };
+       argc = parse_options_subcommand(argc, argv, timechart_options, timechart_subcommands,
+                       timechart_usage, PARSE_OPT_STOP_AT_NON_OPTION);
+
+       if (tchart.power_only && tchart.tasks_only) {
+               pr_err("-P and -T options cannot be used at the same time.\n");
+               return -1;
+       }
+
+       if (argc && !strncmp(argv[0], "rec", 3)) {
+               argc = parse_options(argc, argv, timechart_record_options,
+                                    timechart_record_usage,
+                                    PARSE_OPT_STOP_AT_NON_OPTION);
+
+               if (tchart.power_only && tchart.tasks_only) {
+                       pr_err("-P and -T options cannot be used at the same time.\n");
+                       return -1;
+               }
+
+               if (tchart.io_only)
+                       return timechart__io_record(argc, argv);
+               else
+                       return timechart__record(&tchart, argc, argv);
+       } else if (argc)
+               usage_with_options(timechart_usage, timechart_options);
+
+       setup_pager();
+
+       return __cmd_timechart(&tchart, output_name);
+}