Messages in this thread | | | Date | Thu, 21 Apr 2016 12:42:48 -0300 | From | Arnaldo Carvalho de Melo <> | Subject | Re: [PATCH/RFC] perf core: Allow setting up max frame stack depth via sysctl |
| |
Em Thu, Apr 21, 2016 at 12:17:07PM -0300, Arnaldo Carvalho de Melo escreveu: > Em Thu, Apr 21, 2016 at 12:48:58PM +0200, Peter Zijlstra escreveu: > > On Wed, Apr 20, 2016 at 07:47:30PM -0300, Arnaldo Carvalho de Melo wrote: > > > +++ b/kernel/events/callchain.c > > > > @@ -73,7 +81,7 @@ static int alloc_callchain_buffers(void) > > > if (!entries) > > > return -ENOMEM;
> > > - size = sizeof(struct perf_callchain_entry) * PERF_NR_CONTEXTS; > > > + size = perf_callchain_entry__sizeof() * PERF_NR_CONTEXTS;
> > > for_each_possible_cpu(cpu) { > > > entries->cpu_entries[cpu] = kmalloc_node(size, GFP_KERNEL,
> > And this alloc _will_ fail if you put in a decent sized value..
> > Should we put in a dmesg WARN if this alloc fails and > > perf_event_max_stack is 'large' ? > > Unsure, it already returns -ENOMEM, see, some lines above, i.e. it > better have error handling up this, ho-hum, call chain, I'm checking...
So, it is, using the tools, since it fits this bill:
[root@jouet acme]# perf probe get_callchain_buffers Added new event: probe:get_callchain_buffers (on get_callchain_buffers)
You can now use it in all perf tools, such as:
perf record -e probe:get_callchain_buffers -aR sleep 1
[root@jouet acme]# perf trace -e perf_event_open --event probe:get_callchain_buffers/call-graph=fp/ -- perf record -g usleep 1 [ perf record: Woken up 1 times to write data ] 26.740 ( 0.008 ms): perf/1264 perf_event_open(attr_uptr: 0x1e6b618, pid: 1265, group_fd: -1, flags: FD_CLOEXEC) ... 26.740 ( ): probe:get_callchain_buffers:(ffffffff811a9480)) get_callchain_buffers+0xfe200001 ([kernel.kallsyms]) SYSC_perf_event_open+0xfe2003bc ([kernel.kallsyms]) sys_perf_event_open+0xfe200009 ([kernel.kallsyms]) do_syscall_64+0xfe200062 ([kernel.kallsyms]) return_from_SYSCALL_64+0xfe200000 ([kernel.kallsyms]) syscall+0xffff0179d3c22019 (/usr/lib64/libc-2.22.so) cmd_record+0xffffffffff8005fc (/home/acme/bin/perf) run_builtin+0xffffffffff800061 (/home/acme/bin/perf) main+0xffffffffff800672 (/home/acme/bin/perf) __libc_start_main+0xffff0179d3c220f0 (/usr/lib64/libc-2.22.so) 26.749 ( 0.016 ms): perf/1264 ... [continued]: perf_event_open()) = 4 <SNIP the perf_event_open syscalls for the other 3 CPUs in this system)
So, if we put some too big value there and the allocation fails, people asking for callchains will get that -ENOMEM back in their face.
I'll do this test, i.e. put some huge value there and ask for callchains...
- Arnaldo
| |