Re: perf top -p broken for multithreaded processes since 5.19

[Jiri Olsa <olsajiri@gmail.com>]

On Sat, Sep 03, 2022 at 10:14:25AM +0300, Adrian Hunter wrote:
> On 2/09/22 22:17, Arnaldo Carvalho de Melo wrote:
> > Em Fri, Sep 02, 2022 at 05:50:22PM +0300, Adrian Hunter escreveu:
> >> On 2/09/22 17:46, Tomáš Trnka wrote:
> >>> Hello,
> >>>
> >>> A bug in perf v5.19 and newer completely breaks monitoring multithreaded
> >>> processes using "perf top -p". The tool fails to start with "Failed to mmap
> >>> with 22 (Invalid argument)". It still seems to work fine on single-threaded
> >>> processes. "perf record" is also unaffected.
> >>
> >> It has been reported here:
> >>
> >> 	https://bugzilla.kernel.org/show_bug.cgi?id=216441
> > 
> > If I do:
> > 
> > ⬢[acme@toolbox perf-urgent]$ git log -2
> > commit dfeb0bc60782471c293938e71b1a1117cfac2cb3 (HEAD -> perf/urgent)
> > Author: Arnaldo Carvalho de Melo <acme@redhat.com>
> > Date:   Fri Sep 2 16:15:39 2022 -0300
> > 
> >     Revert "libperf evlist: Check nr_mmaps is correct"
> > 
> >     This reverts commit 4ce47d842d4c16c07b135b8a7975b8f0672bcc0e.
> > 
> >     Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
> > 
> > commit 78cd283f6b8ab701cb35eafd5af8140560a88f16
> > Author: Arnaldo Carvalho de Melo <acme@redhat.com>
> > Date:   Fri Sep 2 16:13:41 2022 -0300
> > 
> >     Revert "libperf evlist: Allow mixing per-thread and per-cpu mmaps"
> > 
> >     This reverts commit ae4f8ae16a07896403c90305d4b9be27f657c1fc.
> > 
> >     Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
> > ⬢[acme@toolbox perf-urgent]$
> > 
> > It works again, Tomáš can you please try doing this to see if this works
> > for you?
> > 
> 
> This is the fix I have so far.  I would like to test it some more though.
> 
> From: Adrian Hunter <adrian.hunter@intel.com>
> Date: Sat, 3 Sep 2022 10:05:08 +0300
> Subject: [PATCH] libperf evlist: Fix per-thread mmaps for multi-threaded
>  targets
> 
> Offending commit did not consider the different set-output rules for
> per-thread mmaps i.e. in the per-thread case set-output is used for
> mmaps of the same thread not the same cpu.
> 
> This was not immediately noticed because it only happens with
> multi-threaded targets and we do not have a test for that yet.
> 
> Fixes: ae4f8ae16a07 ("libperf evlist: Allow mixing per-thread and per-cpu mmaps")
> Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>

SNIP

>  static int
>  mmap_per_cpu(struct perf_evlist *evlist, struct perf_evlist_mmap_ops *ops,
>  	     struct perf_mmap_param *mp)
> @@ -528,6 +571,8 @@ mmap_per_cpu(struct perf_evlist *evlist, struct perf_evlist_mmap_ops *ops,
>  	int nr_mmaps = 0;
>  	int cpu, thread;
>  
> +	pr_debug("%s: nr cpu values %d nr threads %d\n", __func__, nr_cpus, nr_threads);
> +
>  	for (cpu = 0; cpu < nr_cpus; cpu++) {
>  		int output = -1;
>  		int output_overwrite = -1;
> @@ -569,6 +614,7 @@ int perf_evlist__mmap_ops(struct perf_evlist *evlist,
>  			  struct perf_evlist_mmap_ops *ops,
>  			  struct perf_mmap_param *mp)
>  {
> +	const struct perf_cpu_map *cpus = evlist->all_cpus;
>  	struct perf_evsel *evsel;
>  
>  	if (!ops || !ops->get || !ops->mmap)
> @@ -588,6 +634,9 @@ int perf_evlist__mmap_ops(struct perf_evlist *evlist,
>  	if (evlist->pollfd.entries == NULL && perf_evlist__alloc_pollfd(evlist) < 0)
>  		return -ENOMEM;
>  
> +	if (perf_cpu_map__empty(cpus))
> +		return mmap_per_thread(evlist, ops, mp);
> +

could we just enable per-cpu mapping in top? I'd think it should be
enabled anyway because we need ring buffers on all cpus for sampling

jirka


---
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index e89208b4ad4b..d6be82315dd8 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -1422,6 +1422,7 @@ int cmd_top(int argc, const char **argv)
 			.freq		= 4000, /* 4 KHz */
 			.target		= {
 				.uses_mmap   = true,
+				.default_per_cpu = true,
 			},
 			/*
 			 * FIXME: This will lose PERF_RECORD_MMAP and other metadata