[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20181120020421.GE20153@sejong>
Date: Tue, 20 Nov 2018 11:04:21 +0900
From: Namhyung Kim <namhyung@...nel.org>
To: Jiri Olsa <jolsa@...nel.org>
Cc: Arnaldo Carvalho de Melo <acme@...nel.org>,
lkml <linux-kernel@...r.kernel.org>,
Ingo Molnar <mingo@...nel.org>,
Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
Peter Zijlstra <a.p.zijlstra@...llo.nl>,
David Miller <davem@...emloft.net>, kernel-team@....com
Subject: Re: [PATCH 05/12] perf top: Moving lost events warning to helpline
On Mon, Nov 19, 2018 at 01:20:09PM +0100, Jiri Olsa wrote:
> We can't display the UI box saying that we are slow in reader
> thread. That will make perf top even slower and user even more
> angry ;-)
>
> Moving the UI box message out of the reader thread into UI thread
> and changing it into helpline, so there's no 'press any key'
> necessary.
>
> Link: http://lkml.kernel.org/n/tip-0rpmmt3omlait889ewp1cl61@git.kernel.org
> Signed-off-by: Jiri Olsa <jolsa@...nel.org>
> ---
> tools/perf/builtin-top.c | 16 +++++-----------
> 1 file changed, 5 insertions(+), 11 deletions(-)
>
> diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
> index 047e99de09f3..1d77aa7650da 100644
> --- a/tools/perf/builtin-top.c
> +++ b/tools/perf/builtin-top.c
> @@ -553,8 +553,6 @@ static void perf_top__sort_new_samples(void *arg)
> struct perf_evsel *evsel = t->sym_evsel;
> struct hists *hists;
>
> - perf_top__reset_sample_counters(t);
> -
> if (t->evlist->selected != NULL)
> t->sym_evsel = t->evlist->selected;
>
> @@ -571,6 +569,11 @@ static void perf_top__sort_new_samples(void *arg)
>
> hists__collapse_resort(hists, NULL);
> perf_evsel__output_resort(evsel, NULL);
> +
> + if (t->lost)
> + ui_helpline__fpush("Too slow to read ring buffer (change period (-c/-F) or limit CPUs (-C)\n");
In general, ui_helpline__[f]push() requires corresponding __pop()
which handles timeout logic (or something can dismiss the message).
The ui_helpline__show() is to show the message and overwrite the old
one.
But we don't use it strictly and I think just pr_err() or pr_warning()
is ok here (which call ui_helpline__show() internally).
Anyway, I found that tui_helpline__push() should use ui__lock.
Otherwise it could race with display thread and break rendering.
A bug from the beginning but no one reported... :)
Thanks,
Namhyung
>From a5c3a0c88b9d13fedc65b61a45d99fc3abef0089 Mon Sep 17 00:00:00 2001
From: Namhyung Kim <namhyung@...nel.org>
Date: Tue, 20 Nov 2018 10:56:03 +0900
Subject: [PATCH] perf ui/tui: Fix possible UI rendering breakage
The tui_helpline__push() should acquire/release the ui__lock when it
deals with screen setting. Otherwise it could race with display thread
and screen rendering would not be handled properly.
Fixes: e6e904687949 ("perf ui: Introduce struct ui_helpline")
Signed-off-by: Namhyung Kim <namhyung@...nel.org>
---
tools/perf/ui/tui/helpline.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/tools/perf/ui/tui/helpline.c b/tools/perf/ui/tui/helpline.c
index 4ca799aadb4e..ff38b997f457 100644
--- a/tools/perf/ui/tui/helpline.c
+++ b/tools/perf/ui/tui/helpline.c
@@ -20,11 +20,13 @@ static void tui_helpline__push(const char *msg)
{
const size_t sz = sizeof(ui_helpline__current);
+ pthread_mutex_lock(&ui__lock);
SLsmg_gotorc(SLtt_Screen_Rows - 1, 0);
SLsmg_set_color(0);
SLsmg_write_nstring((char *)msg, SLtt_Screen_Cols);
SLsmg_refresh();
strncpy(ui_helpline__current, msg, sz)[sz - 1] = '\0';
+ pthread_mutex_unlock(&ui__lock);
}
static int tui_helpline__show(const char *format, va_list ap)
--
2.19.0
Powered by blists - more mailing lists