LKML Archive on lore.kernel.org
help / color / mirror / Atom feed
From: Arnaldo Carvalho de Melo <acme@kernel.org>
To: Ingo Molnar <mingo@kernel.org>
Cc: Clark Williams <williams@redhat.com>,
	linux-kernel@vger.kernel.org, linux-perf-users@vger.kernel.org,
	David Miller <davem@davemloft.net>,
	Don Zickus <dzickus@redhat.com>, Jiri Olsa <jolsa@redhat.com>,
	Joe Mario <jmario@redhat.com>,
	Arnaldo Carvalho de Melo <acme@redhat.com>
Subject: [PATCH 26/28] perf tools: Don't clone maps from parent when synthesizing forks
Date: Wed, 31 Oct 2018 13:45:06 -0300	[thread overview]
Message-ID: <20181031164508.4784-27-acme@kernel.org> (raw)
In-Reply-To: <20181031164508.4784-1-acme@kernel.org>

From: David Miller <davem@davemloft.net>

When synthesizing FORK events, we are trying to create thread objects
for the already running tasks on the machine.

Normally, for a kernel FORK event, we want to clone the parent's maps
because that is what the kernel just did.

But when synthesizing, this should not be done.  If we do, we end up
with overlapping maps as we process the sythesized MMAP2 events that
get delivered shortly thereafter.

Use the FORK event misc flags in an internal way to signal this
situation, so we can elide the map clone when appropriate.

Signed-off-by: David S. Miller <davem@davemloft.net>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Joe Mario <jmario@redhat.com>
Link: http://lkml.kernel.org/r/20181030.222404.2085088822877051075.davem@davemloft.net
[ Added comment about flag use in machine__process_fork_event(),
  use ternary op in thread__clone_map_groups() as suggested by Jiri ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 include/uapi/linux/perf_event.h       |  2 ++
 tools/include/uapi/linux/perf_event.h |  2 ++
 tools/perf/util/event.c               |  1 +
 tools/perf/util/machine.c             | 19 ++++++++++++++++++-
 tools/perf/util/thread.c              | 13 +++++--------
 tools/perf/util/thread.h              |  2 +-
 6 files changed, 29 insertions(+), 10 deletions(-)

diff --git a/include/uapi/linux/perf_event.h b/include/uapi/linux/perf_event.h
index f35eb72739c0..9de8780ac8d9 100644
--- a/include/uapi/linux/perf_event.h
+++ b/include/uapi/linux/perf_event.h
@@ -646,10 +646,12 @@ struct perf_event_mmap_page {
  *
  *   PERF_RECORD_MISC_MMAP_DATA  - PERF_RECORD_MMAP* events
  *   PERF_RECORD_MISC_COMM_EXEC  - PERF_RECORD_COMM event
+ *   PERF_RECORD_MISC_FORK_EXEC  - PERF_RECORD_FORK event (perf internal)
  *   PERF_RECORD_MISC_SWITCH_OUT - PERF_RECORD_SWITCH* events
  */
 #define PERF_RECORD_MISC_MMAP_DATA		(1 << 13)
 #define PERF_RECORD_MISC_COMM_EXEC		(1 << 13)
+#define PERF_RECORD_MISC_FORK_EXEC		(1 << 13)
 #define PERF_RECORD_MISC_SWITCH_OUT		(1 << 13)
 /*
  * These PERF_RECORD_MISC_* flags below are safely reused
diff --git a/tools/include/uapi/linux/perf_event.h b/tools/include/uapi/linux/perf_event.h
index f35eb72739c0..9de8780ac8d9 100644
--- a/tools/include/uapi/linux/perf_event.h
+++ b/tools/include/uapi/linux/perf_event.h
@@ -646,10 +646,12 @@ struct perf_event_mmap_page {
  *
  *   PERF_RECORD_MISC_MMAP_DATA  - PERF_RECORD_MMAP* events
  *   PERF_RECORD_MISC_COMM_EXEC  - PERF_RECORD_COMM event
+ *   PERF_RECORD_MISC_FORK_EXEC  - PERF_RECORD_FORK event (perf internal)
  *   PERF_RECORD_MISC_SWITCH_OUT - PERF_RECORD_SWITCH* events
  */
 #define PERF_RECORD_MISC_MMAP_DATA		(1 << 13)
 #define PERF_RECORD_MISC_COMM_EXEC		(1 << 13)
+#define PERF_RECORD_MISC_FORK_EXEC		(1 << 13)
 #define PERF_RECORD_MISC_SWITCH_OUT		(1 << 13)
 /*
  * These PERF_RECORD_MISC_* flags below are safely reused
diff --git a/tools/perf/util/event.c b/tools/perf/util/event.c
index bc646185f8d9..e9c108a6b1c3 100644
--- a/tools/perf/util/event.c
+++ b/tools/perf/util/event.c
@@ -308,6 +308,7 @@ static int perf_event__synthesize_fork(struct perf_tool *tool,
 	event->fork.pid  = tgid;
 	event->fork.tid  = pid;
 	event->fork.header.type = PERF_RECORD_FORK;
+	event->fork.header.misc = PERF_RECORD_MISC_FORK_EXEC;
 
 	event->fork.header.size = (sizeof(event->fork) + machine->id_hdr_size);
 
diff --git a/tools/perf/util/machine.c b/tools/perf/util/machine.c
index 8ee8ab39d8ac..8f36ce813bc5 100644
--- a/tools/perf/util/machine.c
+++ b/tools/perf/util/machine.c
@@ -1708,6 +1708,7 @@ int machine__process_fork_event(struct machine *machine, union perf_event *event
 	struct thread *parent = machine__findnew_thread(machine,
 							event->fork.ppid,
 							event->fork.ptid);
+	bool do_maps_clone = true;
 	int err = 0;
 
 	if (dump_trace)
@@ -1736,9 +1737,25 @@ int machine__process_fork_event(struct machine *machine, union perf_event *event
 
 	thread = machine__findnew_thread(machine, event->fork.pid,
 					 event->fork.tid);
+	/*
+	 * When synthesizing FORK events, we are trying to create thread
+	 * objects for the already running tasks on the machine.
+	 *
+	 * Normally, for a kernel FORK event, we want to clone the parent's
+	 * maps because that is what the kernel just did.
+	 *
+	 * But when synthesizing, this should not be done.  If we do, we end up
+	 * with overlapping maps as we process the sythesized MMAP2 events that
+	 * get delivered shortly thereafter.
+	 *
+	 * Use the FORK event misc flags in an internal way to signal this
+	 * situation, so we can elide the map clone when appropriate.
+	 */
+	if (event->fork.header.misc & PERF_RECORD_MISC_FORK_EXEC)
+		do_maps_clone = false;
 
 	if (thread == NULL || parent == NULL ||
-	    thread__fork(thread, parent, sample->time) < 0) {
+	    thread__fork(thread, parent, sample->time, do_maps_clone) < 0) {
 		dump_printf("problem processing PERF_RECORD_FORK, skipping event.\n");
 		err = -1;
 	}
diff --git a/tools/perf/util/thread.c b/tools/perf/util/thread.c
index 2048d393ece6..3d9ed7d0e281 100644
--- a/tools/perf/util/thread.c
+++ b/tools/perf/util/thread.c
@@ -330,7 +330,8 @@ static int thread__prepare_access(struct thread *thread)
 }
 
 static int thread__clone_map_groups(struct thread *thread,
-				    struct thread *parent)
+				    struct thread *parent,
+				    bool do_maps_clone)
 {
 	/* This is new thread, we share map groups for process. */
 	if (thread->pid_ == parent->pid_)
@@ -341,15 +342,11 @@ static int thread__clone_map_groups(struct thread *thread,
 			 thread->pid_, thread->tid, parent->pid_, parent->tid);
 		return 0;
 	}
-
 	/* But this one is new process, copy maps. */
-	if (map_groups__clone(thread, parent->mg) < 0)
-		return -ENOMEM;
-
-	return 0;
+	return do_maps_clone ? map_groups__clone(thread, parent->mg) : 0;
 }
 
-int thread__fork(struct thread *thread, struct thread *parent, u64 timestamp)
+int thread__fork(struct thread *thread, struct thread *parent, u64 timestamp, bool do_maps_clone)
 {
 	if (parent->comm_set) {
 		const char *comm = thread__comm_str(parent);
@@ -362,7 +359,7 @@ int thread__fork(struct thread *thread, struct thread *parent, u64 timestamp)
 	}
 
 	thread->ppid = parent->tid;
-	return thread__clone_map_groups(thread, parent);
+	return thread__clone_map_groups(thread, parent, do_maps_clone);
 }
 
 void thread__find_cpumode_addr_location(struct thread *thread, u64 addr,
diff --git a/tools/perf/util/thread.h b/tools/perf/util/thread.h
index 36c09a9904e6..30e2b4c165fe 100644
--- a/tools/perf/util/thread.h
+++ b/tools/perf/util/thread.h
@@ -89,7 +89,7 @@ struct comm *thread__comm(const struct thread *thread);
 struct comm *thread__exec_comm(const struct thread *thread);
 const char *thread__comm_str(const struct thread *thread);
 int thread__insert_map(struct thread *thread, struct map *map);
-int thread__fork(struct thread *thread, struct thread *parent, u64 timestamp);
+int thread__fork(struct thread *thread, struct thread *parent, u64 timestamp, bool do_maps_clone);
 size_t thread__fprintf(struct thread *thread, FILE *fp);
 
 struct thread *thread__main_thread(struct machine *machine, struct thread *thread);
-- 
2.14.4


  parent reply	other threads:[~2018-10-31 16:46 UTC|newest]

Thread overview: 44+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-10-31 16:44 [GIT PULL 00/24] perf/urgent improvements and fixes Arnaldo Carvalho de Melo
2018-10-31 16:44 ` [PATCH 01/28] tools include uapi: Grab a copy of linux/fs.h Arnaldo Carvalho de Melo
2018-10-31 16:44 ` [PATCH 02/28] perf beauty: Add a generator for MS_ mount/umount's flag constants Arnaldo Carvalho de Melo
2018-10-31 16:44 ` [PATCH 03/28] perf beauty: Switch from GPL v2.0 to LGPL v2.1 Arnaldo Carvalho de Melo
2018-10-31 16:44 ` [PATCH 04/28] perf beauty: Introduce strarray__scnprintf_flags() Arnaldo Carvalho de Melo
2018-10-31 16:44 ` [PATCH 05/28] perf trace beauty: Allow syscalls to mask an argument before considering it Arnaldo Carvalho de Melo
2018-10-31 16:44 ` [PATCH 06/28] perf trace beauty: Beautify mount/umount's 'flags' argument Arnaldo Carvalho de Melo
2018-10-31 16:44 ` [PATCH 07/28] perf trace: Consider syscall aliases too Arnaldo Carvalho de Melo
2018-10-31 16:44 ` [PATCH 08/28] perf trace: Beautify the umount's 'name' argument Arnaldo Carvalho de Melo
2018-10-31 16:44 ` [PATCH 09/28] perf trace: Beautify mount's first pathname arg Arnaldo Carvalho de Melo
2018-10-31 16:44 ` [PATCH 10/28] perf top: Allow disabling the overwrite mode Arnaldo Carvalho de Melo
2018-10-31 16:44 ` [PATCH 11/28] perf top: Do not use overwrite mode by default Arnaldo Carvalho de Melo
2018-10-31 16:44 ` [PATCH 12/28] perf unwind: Take pgoff into account when reporting elf to libdwfl Arnaldo Carvalho de Melo
2018-10-31 16:44 ` [PATCH 13/28] perf cs-etm: Correct CPU mode for samples Arnaldo Carvalho de Melo
2018-10-31 16:44 ` [PATCH 14/28] perf callchain: Honour the ordering of PERF_CONTEXT_{USER,KERNEL,etc} Arnaldo Carvalho de Melo
2018-10-31 16:44 ` [PATCH 15/28] tools include uapi: Update linux/fs.h copy Arnaldo Carvalho de Melo
2018-10-31 16:44 ` [PATCH 16/28] tools arch uapi: Update asm-generic/unistd.h and arm64 unistd.h copies Arnaldo Carvalho de Melo
2018-10-31 16:44 ` [PATCH 17/28] tools include uapi: Update asound.h copy Arnaldo Carvalho de Melo
2018-10-31 17:00   ` Takashi Iwai
2018-10-31 17:27     ` Joe Perches
2018-10-31 18:32       ` Takashi Iwai
2018-10-31 19:29       ` Arnaldo Carvalho de Melo
2018-11-01 11:54         ` Takashi Sakamoto
2018-11-01 13:04           ` Arnaldo Carvalho de Melo
2018-11-01 13:57             ` Takashi Sakamoto
2018-10-31 19:28     ` Arnaldo Carvalho de Melo
2018-10-31 16:44 ` [PATCH 18/28] perf beauty: Add a generator for MAP_ mmap's flag constants Arnaldo Carvalho de Melo
2018-10-31 16:44 ` [PATCH 19/28] perf beauty: Wire up the mmap flags table generator to the Makefile Arnaldo Carvalho de Melo
2018-11-05 15:44   ` Guenter Roeck
2018-11-05 17:11     ` Arnaldo Carvalho de Melo
2018-11-05 18:46       ` Arnaldo Carvalho de Melo
2018-11-05 19:13         ` Arnaldo Carvalho de Melo
2018-11-05 21:18           ` Guenter Roeck
2018-11-06 19:18         ` [tip:perf/urgent] perf beauty: Use SRCARCH, ARCH=x86_64 must map to "x86" to find the headers tip-bot for Arnaldo Carvalho de Melo
2018-10-31 16:45 ` [PATCH 20/28] perf trace beauty: Use the mmap flags table generated from headers Arnaldo Carvalho de Melo
2018-10-31 16:45 ` [PATCH 21/28] tools include uapi: Update linux/mmap.h copy Arnaldo Carvalho de Melo
2018-10-31 16:45 ` [PATCH 22/28] tools headers: Sync the various kvm.h header copies Arnaldo Carvalho de Melo
2018-10-31 16:45 ` [PATCH 23/28] tools headers uapi: Update linux/netlink.h header copy Arnaldo Carvalho de Melo
2018-10-31 16:45 ` [PATCH 24/28] tools headers uapi: Update linux/if_link.h " Arnaldo Carvalho de Melo
2018-10-31 16:45 ` [PATCH 25/28] perf top: Start display thread earlier Arnaldo Carvalho de Melo
2018-10-31 16:45 ` Arnaldo Carvalho de Melo [this message]
2018-10-31 16:45 ` [PATCH 27/28] perf intel-pt: Insert callchain context into synthesized callchains Arnaldo Carvalho de Melo
2018-10-31 16:45 ` [PATCH 28/28] perf intel-pt/bts: Calculate cpumode for synthesized samples Arnaldo Carvalho de Melo
2018-10-31 21:56 ` [GIT PULL 00/24] perf/urgent improvements and fixes Ingo Molnar

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20181031164508.4784-27-acme@kernel.org \
    --to=acme@kernel.org \
    --cc=acme@redhat.com \
    --cc=davem@davemloft.net \
    --cc=dzickus@redhat.com \
    --cc=jmario@redhat.com \
    --cc=jolsa@redhat.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-perf-users@vger.kernel.org \
    --cc=mingo@kernel.org \
    --cc=williams@redhat.com \
    --subject='Re: [PATCH 26/28] perf tools: Don'\''t clone maps from parent when synthesizing forks' \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).