linux_dsm_epyc7002/tools/perf/util/thread.c

243 lines
4.9 KiB
C
Raw Normal View History

#include "../perf.h"
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include "session.h"
#include "thread.h"
#include "thread-stack.h"
#include "util.h"
#include "debug.h"
#include "comm.h"
#include "unwind.h"
int thread__init_map_groups(struct thread *thread, struct machine *machine)
{
struct thread *leader;
pid_t pid = thread->pid_;
if (pid == thread->tid || pid == -1) {
thread->mg = map_groups__new(machine);
} else {
perf machine: Protect the machine->threads with a rwlock In addition to using refcounts for the struct thread lifetime management, we need to protect access to machine->threads from concurrent access. That happens in 'perf top', where a thread processes events, inserting and deleting entries from that rb_tree while another thread decays hist_entries, that end up dropping references and ultimately deleting threads from the rb_tree and releasing its resources when no further hist_entry (or other data structures, like in 'perf sched') references it. So the rule is the same for refcounts + protected trees in the kernel, get the tree lock, find object, bump the refcount, drop the tree lock, return, use object, drop the refcount if no more use of it is needed, keep it if storing it in some other data structure, drop when releasing that data structure. I.e. pair "t = machine__find(new)_thread()" with a "thread__put(t)", and "perf_event__preprocess_sample(&al)" with "addr_location__put(&al)". The addr_location__put() one is because as we return references to several data structures, we may end up adding more reference counting for the other data structures and then we'll drop it at addr_location__put() time. Acked-by: David Ahern <dsahern@gmail.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Borislav Petkov <bp@suse.de> Cc: Don Zickus <dzickus@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/n/tip-bs9rt4n0jw3hi9f3zxyy3xln@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-07 06:43:22 +07:00
leader = __machine__findnew_thread(machine, pid, pid);
if (leader)
thread->mg = map_groups__get(leader->mg);
}
return thread->mg ? 0 : -1;
}
struct thread *thread__new(pid_t pid, pid_t tid)
{
char *comm_str;
struct comm *comm;
struct thread *thread = zalloc(sizeof(*thread));
if (thread != NULL) {
thread->pid_ = pid;
thread->tid = tid;
thread->ppid = -1;
thread->cpu = -1;
INIT_LIST_HEAD(&thread->comm_list);
if (unwind__prepare_access(thread) < 0)
goto err_thread;
comm_str = malloc(32);
if (!comm_str)
goto err_thread;
snprintf(comm_str, 32, ":%d", tid);
comm = comm__new(comm_str, 0, false);
free(comm_str);
if (!comm)
goto err_thread;
list_add(&comm->list, &thread->comm_list);
atomic_set(&thread->refcnt, 0);
perf machine: Protect the machine->threads with a rwlock In addition to using refcounts for the struct thread lifetime management, we need to protect access to machine->threads from concurrent access. That happens in 'perf top', where a thread processes events, inserting and deleting entries from that rb_tree while another thread decays hist_entries, that end up dropping references and ultimately deleting threads from the rb_tree and releasing its resources when no further hist_entry (or other data structures, like in 'perf sched') references it. So the rule is the same for refcounts + protected trees in the kernel, get the tree lock, find object, bump the refcount, drop the tree lock, return, use object, drop the refcount if no more use of it is needed, keep it if storing it in some other data structure, drop when releasing that data structure. I.e. pair "t = machine__find(new)_thread()" with a "thread__put(t)", and "perf_event__preprocess_sample(&al)" with "addr_location__put(&al)". The addr_location__put() one is because as we return references to several data structures, we may end up adding more reference counting for the other data structures and then we'll drop it at addr_location__put() time. Acked-by: David Ahern <dsahern@gmail.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Borislav Petkov <bp@suse.de> Cc: Don Zickus <dzickus@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/n/tip-bs9rt4n0jw3hi9f3zxyy3xln@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-07 06:43:22 +07:00
RB_CLEAR_NODE(&thread->rb_node);
}
return thread;
err_thread:
free(thread);
return NULL;
}
void thread__delete(struct thread *thread)
{
struct comm *comm, *tmp;
perf machine: Protect the machine->threads with a rwlock In addition to using refcounts for the struct thread lifetime management, we need to protect access to machine->threads from concurrent access. That happens in 'perf top', where a thread processes events, inserting and deleting entries from that rb_tree while another thread decays hist_entries, that end up dropping references and ultimately deleting threads from the rb_tree and releasing its resources when no further hist_entry (or other data structures, like in 'perf sched') references it. So the rule is the same for refcounts + protected trees in the kernel, get the tree lock, find object, bump the refcount, drop the tree lock, return, use object, drop the refcount if no more use of it is needed, keep it if storing it in some other data structure, drop when releasing that data structure. I.e. pair "t = machine__find(new)_thread()" with a "thread__put(t)", and "perf_event__preprocess_sample(&al)" with "addr_location__put(&al)". The addr_location__put() one is because as we return references to several data structures, we may end up adding more reference counting for the other data structures and then we'll drop it at addr_location__put() time. Acked-by: David Ahern <dsahern@gmail.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Borislav Petkov <bp@suse.de> Cc: Don Zickus <dzickus@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/n/tip-bs9rt4n0jw3hi9f3zxyy3xln@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-07 06:43:22 +07:00
BUG_ON(!RB_EMPTY_NODE(&thread->rb_node));
thread_stack__free(thread);
if (thread->mg) {
map_groups__put(thread->mg);
thread->mg = NULL;
}
list_for_each_entry_safe(comm, tmp, &thread->comm_list, list) {
list_del(&comm->list);
comm__free(comm);
}
unwind__finish_access(thread);
free(thread);
}
struct thread *thread__get(struct thread *thread)
{
perf machine: Protect the machine->threads with a rwlock In addition to using refcounts for the struct thread lifetime management, we need to protect access to machine->threads from concurrent access. That happens in 'perf top', where a thread processes events, inserting and deleting entries from that rb_tree while another thread decays hist_entries, that end up dropping references and ultimately deleting threads from the rb_tree and releasing its resources when no further hist_entry (or other data structures, like in 'perf sched') references it. So the rule is the same for refcounts + protected trees in the kernel, get the tree lock, find object, bump the refcount, drop the tree lock, return, use object, drop the refcount if no more use of it is needed, keep it if storing it in some other data structure, drop when releasing that data structure. I.e. pair "t = machine__find(new)_thread()" with a "thread__put(t)", and "perf_event__preprocess_sample(&al)" with "addr_location__put(&al)". The addr_location__put() one is because as we return references to several data structures, we may end up adding more reference counting for the other data structures and then we'll drop it at addr_location__put() time. Acked-by: David Ahern <dsahern@gmail.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Borislav Petkov <bp@suse.de> Cc: Don Zickus <dzickus@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/n/tip-bs9rt4n0jw3hi9f3zxyy3xln@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-07 06:43:22 +07:00
if (thread)
atomic_inc(&thread->refcnt);
return thread;
}
void thread__put(struct thread *thread)
{
if (thread && atomic_dec_and_test(&thread->refcnt)) {
list_del_init(&thread->node);
thread__delete(thread);
}
}
struct comm *thread__comm(const struct thread *thread)
{
if (list_empty(&thread->comm_list))
return NULL;
return list_first_entry(&thread->comm_list, struct comm, list);
}
struct comm *thread__exec_comm(const struct thread *thread)
{
struct comm *comm, *last = NULL;
list_for_each_entry(comm, &thread->comm_list, list) {
if (comm->exec)
return comm;
last = comm;
}
return last;
}
int __thread__set_comm(struct thread *thread, const char *str, u64 timestamp,
bool exec)
{
struct comm *new, *curr = thread__comm(thread);
int err;
/* Override the default :tid entry */
if (!thread->comm_set) {
err = comm__override(curr, str, timestamp, exec);
if (err)
return err;
} else {
new = comm__new(str, timestamp, exec);
if (!new)
return -ENOMEM;
list_add(&new->list, &thread->comm_list);
if (exec)
unwind__flush_access(thread);
}
thread->comm_set = true;
return 0;
}
const char *thread__comm_str(const struct thread *thread)
{
const struct comm *comm = thread__comm(thread);
if (!comm)
return NULL;
return comm__str(comm);
}
/* CHECKME: it should probably better return the max comm len from its comm list */
int thread__comm_len(struct thread *thread)
perf tools: Bind callchains to the first sort dimension column Currently, the callchains are displayed using a constant left margin. So depending on the current sort dimension configuration, callchains may appear to be well attached to the first sort dimension column field which is mostly the case, except when the first dimension of sorting is done by comm, because these are right aligned. This patch binds the callchain to the first letter in the first column, whatever type of column it is (dso, comm, symbol). Before: 0.80% perf [k] __lock_acquire __lock_acquire lock_acquire | |--58.33%-- _spin_lock | | | |--28.57%-- inotify_should_send_event | | fsnotify | | __fsnotify_parent After: 0.80% perf [k] __lock_acquire __lock_acquire lock_acquire | |--58.33%-- _spin_lock | | | |--28.57%-- inotify_should_send_event | | fsnotify | | __fsnotify_parent Also, for clarity, we don't put anymore the callchain as is but: - If we have a top level ancestor in the callchain, start it with a first ascii hook. Before: 0.80% perf [kernel] [k] __lock_acquire __lock_acquire lock_acquire | |--58.33%-- _spin_lock | | | |--28.57%-- inotify_should_send_event | | fsnotify [..] [..] After: 0.80% perf [kernel] [k] __lock_acquire | --- __lock_acquire lock_acquire | |--58.33%-- _spin_lock | | | |--28.57%-- inotify_should_send_event | | fsnotify [..] [..] - Otherwise, if we have several top level ancestors, then display these like we did before: 1.69% Xorg | |--21.21%-- vread_hpet | 0x7fffd85b46fc | 0x7fffd85b494d | 0x7f4fafb4e54d | |--15.15%-- exaOffscreenAlloc | |--9.09%-- I830WaitLpRing Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> LKML-Reference: <1256246604-17156-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-23 04:23:23 +07:00
{
if (!thread->comm_len) {
const char *comm = thread__comm_str(thread);
if (!comm)
perf tools: Bind callchains to the first sort dimension column Currently, the callchains are displayed using a constant left margin. So depending on the current sort dimension configuration, callchains may appear to be well attached to the first sort dimension column field which is mostly the case, except when the first dimension of sorting is done by comm, because these are right aligned. This patch binds the callchain to the first letter in the first column, whatever type of column it is (dso, comm, symbol). Before: 0.80% perf [k] __lock_acquire __lock_acquire lock_acquire | |--58.33%-- _spin_lock | | | |--28.57%-- inotify_should_send_event | | fsnotify | | __fsnotify_parent After: 0.80% perf [k] __lock_acquire __lock_acquire lock_acquire | |--58.33%-- _spin_lock | | | |--28.57%-- inotify_should_send_event | | fsnotify | | __fsnotify_parent Also, for clarity, we don't put anymore the callchain as is but: - If we have a top level ancestor in the callchain, start it with a first ascii hook. Before: 0.80% perf [kernel] [k] __lock_acquire __lock_acquire lock_acquire | |--58.33%-- _spin_lock | | | |--28.57%-- inotify_should_send_event | | fsnotify [..] [..] After: 0.80% perf [kernel] [k] __lock_acquire | --- __lock_acquire lock_acquire | |--58.33%-- _spin_lock | | | |--28.57%-- inotify_should_send_event | | fsnotify [..] [..] - Otherwise, if we have several top level ancestors, then display these like we did before: 1.69% Xorg | |--21.21%-- vread_hpet | 0x7fffd85b46fc | 0x7fffd85b494d | 0x7f4fafb4e54d | |--15.15%-- exaOffscreenAlloc | |--9.09%-- I830WaitLpRing Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> LKML-Reference: <1256246604-17156-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-23 04:23:23 +07:00
return 0;
thread->comm_len = strlen(comm);
perf tools: Bind callchains to the first sort dimension column Currently, the callchains are displayed using a constant left margin. So depending on the current sort dimension configuration, callchains may appear to be well attached to the first sort dimension column field which is mostly the case, except when the first dimension of sorting is done by comm, because these are right aligned. This patch binds the callchain to the first letter in the first column, whatever type of column it is (dso, comm, symbol). Before: 0.80% perf [k] __lock_acquire __lock_acquire lock_acquire | |--58.33%-- _spin_lock | | | |--28.57%-- inotify_should_send_event | | fsnotify | | __fsnotify_parent After: 0.80% perf [k] __lock_acquire __lock_acquire lock_acquire | |--58.33%-- _spin_lock | | | |--28.57%-- inotify_should_send_event | | fsnotify | | __fsnotify_parent Also, for clarity, we don't put anymore the callchain as is but: - If we have a top level ancestor in the callchain, start it with a first ascii hook. Before: 0.80% perf [kernel] [k] __lock_acquire __lock_acquire lock_acquire | |--58.33%-- _spin_lock | | | |--28.57%-- inotify_should_send_event | | fsnotify [..] [..] After: 0.80% perf [kernel] [k] __lock_acquire | --- __lock_acquire lock_acquire | |--58.33%-- _spin_lock | | | |--28.57%-- inotify_should_send_event | | fsnotify [..] [..] - Otherwise, if we have several top level ancestors, then display these like we did before: 1.69% Xorg | |--21.21%-- vread_hpet | 0x7fffd85b46fc | 0x7fffd85b494d | 0x7f4fafb4e54d | |--15.15%-- exaOffscreenAlloc | |--9.09%-- I830WaitLpRing Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> LKML-Reference: <1256246604-17156-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-23 04:23:23 +07:00
}
return thread->comm_len;
perf tools: Bind callchains to the first sort dimension column Currently, the callchains are displayed using a constant left margin. So depending on the current sort dimension configuration, callchains may appear to be well attached to the first sort dimension column field which is mostly the case, except when the first dimension of sorting is done by comm, because these are right aligned. This patch binds the callchain to the first letter in the first column, whatever type of column it is (dso, comm, symbol). Before: 0.80% perf [k] __lock_acquire __lock_acquire lock_acquire | |--58.33%-- _spin_lock | | | |--28.57%-- inotify_should_send_event | | fsnotify | | __fsnotify_parent After: 0.80% perf [k] __lock_acquire __lock_acquire lock_acquire | |--58.33%-- _spin_lock | | | |--28.57%-- inotify_should_send_event | | fsnotify | | __fsnotify_parent Also, for clarity, we don't put anymore the callchain as is but: - If we have a top level ancestor in the callchain, start it with a first ascii hook. Before: 0.80% perf [kernel] [k] __lock_acquire __lock_acquire lock_acquire | |--58.33%-- _spin_lock | | | |--28.57%-- inotify_should_send_event | | fsnotify [..] [..] After: 0.80% perf [kernel] [k] __lock_acquire | --- __lock_acquire lock_acquire | |--58.33%-- _spin_lock | | | |--28.57%-- inotify_should_send_event | | fsnotify [..] [..] - Otherwise, if we have several top level ancestors, then display these like we did before: 1.69% Xorg | |--21.21%-- vread_hpet | 0x7fffd85b46fc | 0x7fffd85b494d | 0x7f4fafb4e54d | |--15.15%-- exaOffscreenAlloc | |--9.09%-- I830WaitLpRing Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> LKML-Reference: <1256246604-17156-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-23 04:23:23 +07:00
}
size_t thread__fprintf(struct thread *thread, FILE *fp)
{
return fprintf(fp, "Thread %d %s\n", thread->tid, thread__comm_str(thread)) +
map_groups__fprintf(thread->mg, fp);
}
void thread__insert_map(struct thread *thread, struct map *map)
perf tools: Use rb_tree for maps Threads can have many and kernel modules will be represented as a tree of maps as well. Ah, and for a perf.data with 146607 samples: Before: [root@doppio ~]# perf stat -r 5 perf report > /dev/null Performance counter stats for 'perf report' (5 runs): 699.823680 task-clock-msecs # 0.991 CPUs ( +- 0.454% ) 74 context-switches # 0.000 M/sec ( +- 1.709% ) 2 CPU-migrations # 0.000 M/sec ( +- 17.008% ) 23114 page-faults # 0.033 M/sec ( +- 0.000% ) 1381257019 cycles # 1973.721 M/sec ( +- 0.290% ) 1456894438 instructions # 1.055 IPC ( +- 0.007% ) 18779818 cache-references # 26.835 M/sec ( +- 0.380% ) 641799 cache-misses # 0.917 M/sec ( +- 1.200% ) 0.705972729 seconds time elapsed ( +- 0.501% ) [root@doppio ~]# After Performance counter stats for 'perf report' (5 runs): 691.261451 task-clock-msecs # 0.993 CPUs ( +- 0.307% ) 72 context-switches # 0.000 M/sec ( +- 0.829% ) 6 CPU-migrations # 0.000 M/sec ( +- 18.409% ) 23127 page-faults # 0.033 M/sec ( +- 0.000% ) 1366395876 cycles # 1976.670 M/sec ( +- 0.153% ) 1443136016 instructions # 1.056 IPC ( +- 0.012% ) 17956402 cache-references # 25.976 M/sec ( +- 0.325% ) 661924 cache-misses # 0.958 M/sec ( +- 1.335% ) 0.696127275 seconds time elapsed ( +- 0.377% ) I.e. we see some speedup too. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: "H. Peter Anvin" <hpa@zytor.com> LKML-Reference: <20090928174846.GA3361@ghostprotocols.net> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-09-29 00:48:46 +07:00
{
map_groups__fixup_overlappings(thread->mg, map, stderr);
map_groups__insert(thread->mg, map);
}
static int thread__clone_map_groups(struct thread *thread,
struct thread *parent)
{
int i;
/* This is new thread, we share map groups for process. */
if (thread->pid_ == parent->pid_)
return 0;
if (thread->mg == parent->mg) {
pr_debug("broken map groups on thread %d/%d parent %d/%d\n",
thread->pid_, thread->tid, parent->pid_, parent->tid);
return 0;
}
/* But this one is new process, copy maps. */
for (i = 0; i < MAP__NR_TYPES; ++i)
if (map_groups__clone(thread->mg, parent->mg, i) < 0)
return -ENOMEM;
return 0;
}
int thread__fork(struct thread *thread, struct thread *parent, u64 timestamp)
{
int err;
if (parent->comm_set) {
const char *comm = thread__comm_str(parent);
if (!comm)
return -ENOMEM;
err = thread__set_comm(thread, comm, timestamp);
if (err)
return err;
}
thread->ppid = parent->tid;
return thread__clone_map_groups(thread, parent);
}
void thread__find_cpumode_addr_location(struct thread *thread,
enum map_type type, u64 addr,
struct addr_location *al)
{
size_t i;
const u8 const cpumodes[] = {
PERF_RECORD_MISC_USER,
PERF_RECORD_MISC_KERNEL,
PERF_RECORD_MISC_GUEST_USER,
PERF_RECORD_MISC_GUEST_KERNEL
};
for (i = 0; i < ARRAY_SIZE(cpumodes); i++) {
thread__find_addr_location(thread, cpumodes[i], type, addr, al);
if (al->map)
break;
}
}