[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20160425162706.GB25218@kernel.org>
Date: Mon, 25 Apr 2016 13:27:06 -0300
From: Arnaldo Carvalho de Melo <acme@...nel.org>
To: Alexei Starovoitov <alexei.starovoitov@...il.com>
Cc: David Ahern <dsahern@...il.com>,
Arnaldo Carvalho de Melo <arnaldo.melo@...il.com>,
Peter Zijlstra <peterz@...radead.org>,
Frederic Weisbecker <fweisbec@...il.com>,
Ingo Molnar <mingo@...nel.org>,
Adrian Hunter <adrian.hunter@...el.com>,
Brendan Gregg <brendan.d.gregg@...il.com>,
Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
Alexei Starovoitov <ast@...nel.org>,
He Kuang <hekuang@...wei.com>, Jiri Olsa <jolsa@...hat.com>,
Masami Hiramatsu <mhiramat@...nel.org>,
Milian Wolff <milian.wolff@...b.com>,
Namhyung Kim <namhyung@...nel.org>,
Stephane Eranian <eranian@...gle.com>,
Thomas Gleixner <tglx@...utronix.de>,
Vince Weaver <vincent.weaver@...ne.edu>,
Wang Nan <wangnan0@...wei.com>, Zefan Li <lizefan@...wei.com>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH/RFC v2] perf core: Allow setting up max frame stack depth
via sysctl
Em Mon, Apr 25, 2016 at 01:14:25PM -0300, Arnaldo Carvalho de Melo escreveu:
> Em Fri, Apr 22, 2016 at 03:18:08PM -0700, Alexei Starovoitov escreveu:
> > On Fri, Apr 22, 2016 at 04:05:31PM -0600, David Ahern wrote:
> > > On 4/22/16 2:52 PM, Arnaldo Carvalho de Melo wrote:
> > > >+int sysctl_perf_event_max_stack __read_mostly = PERF_MAX_STACK_DEPTH;
> > > >+
> > > >+static size_t perf_callchain_entry__sizeof(void)
> > > >+{
> > > >+ return sizeof(struct perf_callchain_entry) +
> > > >+ sizeof(__u64) * sysctl_perf_event_max_stack;
> > > >+}
> > > >+
>
> > > To Alexei's comment, a max_stack of 0 still has a non-zero alloc size so
> > > that should be ok.
>
> > > > static DEFINE_PER_CPU(int, callchain_recursion[PERF_NR_CONTEXTS]);
> > > > static atomic_t nr_callchain_events;
> > > > static DEFINE_MUTEX(callchain_mutex);
> > > >@@ -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,
>
> > right... and looking into it further, realized that the patch is broken,
> > since get_callchain_entry() is doing:
> > return &entries->cpu_entries[cpu][*rctx];
> > whereas it should be dynamic offset based on sysctl_perf_event_max_stack*8
> > So definitely needs another respin.
>
> Huh? Can you elaborate a bit more?
>
> Are you saying this is a bug introduced by this patch or something
> pre-existing?
When we alloc we alloc:
/*
* Number of contexts where an event can trigger:
* task, softirq, hardirq, nmi.
*/
#define PERF_NR_CONTEXTS 4
callchain_cpus_entries[0 .. (nr_cpu_ids - 1)][in_nmi() (3)]
callchain_cpus_entries[0 .. (nr_cpu_ids - 1)][in_irq() (2)]
callchain_cpus_entries[0 .. (nr_cpu_ids - 1)][in_sortirq() (1)]
callchain_cpus_entries[0 .. (nr_cpu_ids - 1)][(0)]
And all of these have this type:
struct perf_callchain_entry {
__u64 nr;
__u64 ip[0]; /* /proc/sys/kernel/perf_event_max_stack */
};
So, it will return a struct_callchain_entry pointer to a 8-byte sized
chunk, with just perf_callchain_entry->nr, and will not try to touch
perf_callchain_entry->ip[] since sysctl_perf_event_max_stack is zero.
But perf_callchain_entry->ip is not a pointer... Got it ;-\
Touché, respinning...
Brendan, this probably affected you results...
- Arnaldo
Powered by blists - more mailing lists