Re: [PATCH 06/13] perf tools: Use tid in mmap/mmap2 events to find maps
Em Wed, Mar 19, 2014 at 02:30:25PM +0100, Jiri Olsa escreveu: > On Wed, Mar 19, 2014 at 09:19:45AM -0400, Don Zickus wrote: > > On Wed, Mar 19, 2014 at 09:13:38AM +0200, Adrian Hunter wrote: > > > On 18.03.2014 23:25, Arnaldo Carvalho de Melo wrote: > > > > From: Don Zickus > > > > > > > > Now that we can properly synthesize threads system-wide, make sure the > > > > mmap and mmap2 events use tids instead of pids to locate their maps. > > > > > > This is a good change but it cannot go yet because it is dependent on > > > Jiri's > > > "perf tools: Share map groups within process" series. > > > > No it isn't. :-) Commit bfd66cc71a3f831ba7c2116d79416cfb8883f6cf is the > > reason why. It synthesizes thread events to make this work correctly. > > ok, I haven't realized this one was pulled in as well ;-) > it's ok then.. > > Arnaldo, I'll rebase the perf/core_mmaps (shared map groups) to your > latest perf/core Thanks! -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [PATCH 06/13] perf tools: Use tid in mmap/mmap2 events to find maps
On 19/03/2014 4:43 p.m., Don Zickus wrote: On Wed, Mar 19, 2014 at 04:27:25PM +0200, Adrian Hunter wrote: On 19.03.2014 15:30, Jiri Olsa wrote: On Wed, Mar 19, 2014 at 09:19:45AM -0400, Don Zickus wrote: On Wed, Mar 19, 2014 at 09:13:38AM +0200, Adrian Hunter wrote: On 18.03.2014 23:25, Arnaldo Carvalho de Melo wrote: From: Don Zickus Now that we can properly synthesize threads system-wide, make sure the mmap and mmap2 events use tids instead of pids to locate their maps. This is a good change but it cannot go yet because it is dependent on Jiri's "perf tools: Share map groups within process" series. No it isn't. :-) Commit bfd66cc71a3f831ba7c2116d79416cfb8883f6cf is the reason why. It synthesizes thread events to make this work correctly. ok, I haven't realized this one was pulled in as well ;-) it's ok then.. No, you mustn't split up the maps between threads. We currently assume CLONE_VM is always used to create threads. But that doesn't work when you try to synthesize threads which was the problem I was trying to solve. It worked if you looked up the maps on the pid not the tid. -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [PATCH 06/13] perf tools: Use tid in mmap/mmap2 events to find maps
On Wed, Mar 19, 2014 at 04:27:25PM +0200, Adrian Hunter wrote: > On 19.03.2014 15:30, Jiri Olsa wrote: > > On Wed, Mar 19, 2014 at 09:19:45AM -0400, Don Zickus wrote: > >> On Wed, Mar 19, 2014 at 09:13:38AM +0200, Adrian Hunter wrote: > >>> On 18.03.2014 23:25, Arnaldo Carvalho de Melo wrote: > From: Don Zickus > > Now that we can properly synthesize threads system-wide, make sure the > mmap and mmap2 events use tids instead of pids to locate their maps. > >>> > >>> This is a good change but it cannot go yet because it is dependent on > >>> Jiri's > >>> "perf tools: Share map groups within process" series. > >> > >> No it isn't. :-) Commit bfd66cc71a3f831ba7c2116d79416cfb8883f6cf is the > >> reason why. It synthesizes thread events to make this work correctly. > > > > ok, I haven't realized this one was pulled in as well ;-) > > it's ok then.. > > No, you mustn't split up the maps between threads. We currently assume > CLONE_VM is always used to create threads. But that doesn't work when you try to synthesize threads which was the problem I was trying to solve. Cheers, Don -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [PATCH 06/13] perf tools: Use tid in mmap/mmap2 events to find maps
On 19.03.2014 15:30, Jiri Olsa wrote: > On Wed, Mar 19, 2014 at 09:19:45AM -0400, Don Zickus wrote: >> On Wed, Mar 19, 2014 at 09:13:38AM +0200, Adrian Hunter wrote: >>> On 18.03.2014 23:25, Arnaldo Carvalho de Melo wrote: From: Don Zickus Now that we can properly synthesize threads system-wide, make sure the mmap and mmap2 events use tids instead of pids to locate their maps. >>> >>> This is a good change but it cannot go yet because it is dependent on Jiri's >>> "perf tools: Share map groups within process" series. >> >> No it isn't. :-) Commit bfd66cc71a3f831ba7c2116d79416cfb8883f6cf is the >> reason why. It synthesizes thread events to make this work correctly. > > ok, I haven't realized this one was pulled in as well ;-) > it's ok then.. No, you mustn't split up the maps between threads. We currently assume CLONE_VM is always used to create threads. > > Arnaldo, I'll rebase the perf/core_mmaps (shared map groups) to your > latest perf/core > > jirka > > -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [PATCH 06/13] perf tools: Use tid in mmap/mmap2 events to find maps
On Wed, Mar 19, 2014 at 09:19:45AM -0400, Don Zickus wrote: > On Wed, Mar 19, 2014 at 09:13:38AM +0200, Adrian Hunter wrote: > > On 18.03.2014 23:25, Arnaldo Carvalho de Melo wrote: > > > From: Don Zickus > > > > > > Now that we can properly synthesize threads system-wide, make sure the > > > mmap and mmap2 events use tids instead of pids to locate their maps. > > > > This is a good change but it cannot go yet because it is dependent on Jiri's > > "perf tools: Share map groups within process" series. > > No it isn't. :-) Commit bfd66cc71a3f831ba7c2116d79416cfb8883f6cf is the > reason why. It synthesizes thread events to make this work correctly. ok, I haven't realized this one was pulled in as well ;-) it's ok then.. Arnaldo, I'll rebase the perf/core_mmaps (shared map groups) to your latest perf/core jirka -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [PATCH 06/13] perf tools: Use tid in mmap/mmap2 events to find maps
On Wed, Mar 19, 2014 at 09:13:38AM +0200, Adrian Hunter wrote: > On 18.03.2014 23:25, Arnaldo Carvalho de Melo wrote: > > From: Don Zickus > > > > Now that we can properly synthesize threads system-wide, make sure the > > mmap and mmap2 events use tids instead of pids to locate their maps. > > This is a good change but it cannot go yet because it is dependent on Jiri's > "perf tools: Share map groups within process" series. > > perf tools assume that threads share virtual memory. That means that a > mapping created by one thread is visible to all threads in the thread group. > That was implemented by putting all the maps on the thread group leader > (pid == tid). The alternative is to share the map groups instead, which is > what Jiri is doing. right, this needs the sharing map groups changes (or the 1/2 patch Don sent prior to this patch) jirka -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [PATCH 06/13] perf tools: Use tid in mmap/mmap2 events to find maps
On 18.03.2014 23:25, Arnaldo Carvalho de Melo wrote:
> From: Don Zickus
>
> Now that we can properly synthesize threads system-wide, make sure the
> mmap and mmap2 events use tids instead of pids to locate their maps.
This is a good change but it cannot go yet because it is dependent on Jiri's
"perf tools: Share map groups within process" series.
perf tools assume that threads share virtual memory. That means that a
mapping created by one thread is visible to all threads in the thread group.
That was implemented by putting all the maps on the thread group leader
(pid == tid). The alternative is to share the map groups instead, which is
what Jiri is doing.
>
> Signed-off-by: Don Zickus
> Cc: Jiri Olsa
> Cc: Stephane Eranian
> Link:
> http://lkml.kernel.org/r/1393429527-167840-3-git-send-email-dzic...@redhat.com
> Signed-off-by: Arnaldo Carvalho de Melo
> ---
> tools/perf/tests/hists_link.c | 1 +
> tools/perf/util/machine.c | 4 ++--
> 2 files changed, 3 insertions(+), 2 deletions(-)
>
> diff --git a/tools/perf/tests/hists_link.c b/tools/perf/tests/hists_link.c
> index 2b6519e0e36f..7ccbc7b6ae77 100644
> --- a/tools/perf/tests/hists_link.c
> +++ b/tools/perf/tests/hists_link.c
> @@ -101,6 +101,7 @@ static struct machine *setup_fake_machine(struct machines
> *machines)
> .mmap = {
> .header = { .misc = PERF_RECORD_MISC_USER, },
> .pid = fake_mmap_info[i].pid,
> + .tid = fake_mmap_info[i].pid,
> .start = fake_mmap_info[i].start,
> .len = 0x1000ULL,
> .pgoff = 0ULL,
> diff --git a/tools/perf/util/machine.c b/tools/perf/util/machine.c
> index d280bf210183..a53cd0b8c151 100644
> --- a/tools/perf/util/machine.c
> +++ b/tools/perf/util/machine.c
> @@ -1027,7 +1027,7 @@ int machine__process_mmap2_event(struct machine
> *machine,
> }
>
> thread = machine__findnew_thread(machine, event->mmap2.pid,
> - event->mmap2.pid);
> + event->mmap2.tid);
> if (thread == NULL)
> goto out_problem;
>
> @@ -1075,7 +1075,7 @@ int machine__process_mmap_event(struct machine
> *machine, union perf_event *event
> }
>
> thread = machine__findnew_thread(machine, event->mmap.pid,
> - event->mmap.pid);
> + event->mmap.tid);
> if (thread == NULL)
> goto out_problem;
>
>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
