From: Masami Hiramatsu (Google) <mhiramat@kernel.org>
To: Andrii Nakryiko <andrii.nakryiko@gmail.com>
Cc: Andrii Nakryiko <andrii@kernel.org>,
linux-trace-kernel@vger.kernel.org, rostedt@goodmis.org,
peterz@infradead.org, oleg@redhat.com, bpf@vger.kernel.org,
linux-kernel@vger.kernel.org, jolsa@kernel.org
Subject: Re: [PATCH v3] uprobes: turn trace_uprobe's nhit counter to be per-CPU one
Date: Thu, 5 Sep 2024 00:02:43 +0900 [thread overview]
Message-ID: <20240905000243.c8549b30be33a3ad73eabf05@kernel.org> (raw)
In-Reply-To: <CAEf4BzbHuHPuLHT5E619crJ1Gdvr8LnQpR7w=vrx8Gb6NHAeZA@mail.gmail.com>
On Thu, 29 Aug 2024 10:28:24 -0700
Andrii Nakryiko <andrii.nakryiko@gmail.com> wrote:
> On Tue, Aug 27, 2024 at 8:55 PM Masami Hiramatsu <mhiramat@kernel.org> wrote:
> >
> > On Tue, 13 Aug 2024 13:34:09 -0700
> > Andrii Nakryiko <andrii@kernel.org> wrote:
> >
> > > trace_uprobe->nhit counter is not incremented atomically, so its value
> > > is questionable in when uprobe is hit on multiple CPUs simultaneously.
> > >
> > > Also, doing this shared counter increment across many CPUs causes heavy
> > > cache line bouncing, limiting uprobe/uretprobe performance scaling with
> > > number of CPUs.
> > >
> > > Solve both problems by making this a per-CPU counter.
> > >
> >
> > Looks good to me. Let me pick it to linux-trace probes/for-next.
> >
>
> Thanks! I just checked linux-trace repo, doesn't seem like this was
> applied yet, is that right? Or am I checking in the wrong place?
Sorry, I missed to push probes/for-next. Let me push it.
Thank you,
>
> > Thank you,
> >
> >
> > > Reviewed-by: Oleg Nesterov <oleg@redhat.com>
> > > Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
> > > ---
> > > kernel/trace/trace_uprobe.c | 24 +++++++++++++++++++++---
> > > 1 file changed, 21 insertions(+), 3 deletions(-)
> > >
> > > diff --git a/kernel/trace/trace_uprobe.c b/kernel/trace/trace_uprobe.c
> > > index c98e3b3386ba..c3df411a2684 100644
> > > --- a/kernel/trace/trace_uprobe.c
> > > +++ b/kernel/trace/trace_uprobe.c
> > > @@ -17,6 +17,7 @@
> > > #include <linux/string.h>
> > > #include <linux/rculist.h>
> > > #include <linux/filter.h>
> > > +#include <linux/percpu.h>
> > >
> > > #include "trace_dynevent.h"
> > > #include "trace_probe.h"
> > > @@ -62,7 +63,7 @@ struct trace_uprobe {
> > > char *filename;
> > > unsigned long offset;
> > > unsigned long ref_ctr_offset;
> > > - unsigned long nhit;
> > > + unsigned long __percpu *nhits;
> > > struct trace_probe tp;
> > > };
> > >
> > > @@ -337,6 +338,12 @@ alloc_trace_uprobe(const char *group, const char *event, int nargs, bool is_ret)
> > > if (!tu)
> > > return ERR_PTR(-ENOMEM);
> > >
> > > + tu->nhits = alloc_percpu(unsigned long);
> > > + if (!tu->nhits) {
> > > + ret = -ENOMEM;
> > > + goto error;
> > > + }
> > > +
> > > ret = trace_probe_init(&tu->tp, event, group, true, nargs);
> > > if (ret < 0)
> > > goto error;
> > > @@ -349,6 +356,7 @@ alloc_trace_uprobe(const char *group, const char *event, int nargs, bool is_ret)
> > > return tu;
> > >
> > > error:
> > > + free_percpu(tu->nhits);
> > > kfree(tu);
> > >
> > > return ERR_PTR(ret);
> > > @@ -362,6 +370,7 @@ static void free_trace_uprobe(struct trace_uprobe *tu)
> > > path_put(&tu->path);
> > > trace_probe_cleanup(&tu->tp);
> > > kfree(tu->filename);
> > > + free_percpu(tu->nhits);
> > > kfree(tu);
> > > }
> > >
> > > @@ -815,13 +824,21 @@ static int probes_profile_seq_show(struct seq_file *m, void *v)
> > > {
> > > struct dyn_event *ev = v;
> > > struct trace_uprobe *tu;
> > > + unsigned long nhits;
> > > + int cpu;
> > >
> > > if (!is_trace_uprobe(ev))
> > > return 0;
> > >
> > > tu = to_trace_uprobe(ev);
> > > +
> > > + nhits = 0;
> > > + for_each_possible_cpu(cpu) {
> > > + nhits += per_cpu(*tu->nhits, cpu);
> > > + }
> > > +
> > > seq_printf(m, " %s %-44s %15lu\n", tu->filename,
> > > - trace_probe_name(&tu->tp), tu->nhit);
> > > + trace_probe_name(&tu->tp), nhits);
> > > return 0;
> > > }
> > >
> > > @@ -1512,7 +1529,8 @@ static int uprobe_dispatcher(struct uprobe_consumer *con, struct pt_regs *regs)
> > > int ret = 0;
> > >
> > > tu = container_of(con, struct trace_uprobe, consumer);
> > > - tu->nhit++;
> > > +
> > > + this_cpu_inc(*tu->nhits);
> > >
> > > udd.tu = tu;
> > > udd.bp_addr = instruction_pointer(regs);
> > > --
> > > 2.43.5
> > >
> >
> >
> > --
> > Masami Hiramatsu (Google) <mhiramat@kernel.org>
> >
--
Masami Hiramatsu (Google) <mhiramat@kernel.org>
prev parent reply other threads:[~2024-09-04 15:02 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-08-13 20:34 [PATCH v3] uprobes: turn trace_uprobe's nhit counter to be per-CPU one Andrii Nakryiko
2024-08-21 16:38 ` Andrii Nakryiko
2024-08-26 13:50 ` Jiri Olsa
2024-08-28 3:53 ` Masami Hiramatsu
2024-08-28 3:54 ` Masami Hiramatsu
2024-08-29 17:28 ` Andrii Nakryiko
2024-09-04 15:02 ` Masami Hiramatsu [this message]
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=20240905000243.c8549b30be33a3ad73eabf05@kernel.org \
--to=mhiramat@kernel.org \
--cc=andrii.nakryiko@gmail.com \
--cc=andrii@kernel.org \
--cc=bpf@vger.kernel.org \
--cc=jolsa@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-trace-kernel@vger.kernel.org \
--cc=oleg@redhat.com \
--cc=peterz@infradead.org \
--cc=rostedt@goodmis.org \
/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
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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).