Resend of patch sent back in January 2011 in light of recent confusion around
unsupported events for a given platform.
Improve sys_perf_event_open ENOENT return handling in top and record, just
like 5a3446b does for stat.
Retry of Arnaldo's patch using error instead of die which allows the fallback
from hardware cycles to software clock.
Signed-off-by: David Ahern <[email protected]>
---
tools/perf/builtin-record.c | 6 ++++++
tools/perf/builtin-top.c | 7 +++++++
2 files changed, 13 insertions(+), 0 deletions(-)
diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
index 6febcc1..73fa8b7 100644
--- a/tools/perf/builtin-record.c
+++ b/tools/perf/builtin-record.c
@@ -282,6 +282,8 @@ try_again:
else if (err == ENODEV && cpu_list) {
die("No such device - did you specify"
" an out-of-range profile CPU?\n");
+ } else if (err == ENOENT) {
+ error("%s event is not supported.", event_name(pos));
} else if (err == EINVAL && sample_id_all_avail) {
/*
* Old kernel, no attr->sample_id_type_all field
@@ -307,6 +309,10 @@ try_again:
attr->config = PERF_COUNT_SW_CPU_CLOCK;
goto try_again;
}
+
+ if (err == ENOENT)
+ exit(1);
+
printf("\n");
error("sys_perf_event_open() syscall returned with %d (%s). /bin/dmesg may provide additional information.\n",
err, strerror(err));
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index 80c9e06..ab27693 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -852,6 +852,9 @@ try_again:
die("Permission error - are you root?\n"
"\t Consider tweaking"
" /proc/sys/kernel/perf_event_paranoid.\n");
+ else if (err == ENOENT)
+ error("%s event is not supported.", event_name(counter));
+
/*
* If it's cycles then fall back to hrtimer
* based cpu-clock-tick sw counter, which
@@ -867,6 +870,10 @@ try_again:
attr->config = PERF_COUNT_SW_CPU_CLOCK;
goto try_again;
}
+
+ if (err == ENOENT)
+ exit(1);
+
printf("\n");
error("sys_perf_event_open() syscall returned with %d "
"(%s). /bin/dmesg may provide additional information.\n",
--
1.7.4
Em Fri, Mar 25, 2011 at 12:05:50PM -0600, David Ahern escreveu:
> Resend of patch sent back in January 2011 in light of recent confusion around
> unsupported events for a given platform.
>
> Improve sys_perf_event_open ENOENT return handling in top and record, just
> like 5a3446b does for stat.
>
> Retry of Arnaldo's patch using error instead of die which allows the fallback
> from hardware cycles to software clock.
Please use ui__warning(""...) as it will work in the TUI too.
> Signed-off-by: David Ahern <[email protected]>
> ---
> tools/perf/builtin-record.c | 6 ++++++
> tools/perf/builtin-top.c | 7 +++++++
> 2 files changed, 13 insertions(+), 0 deletions(-)
>
> diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
> index 6febcc1..73fa8b7 100644
> --- a/tools/perf/builtin-record.c
> +++ b/tools/perf/builtin-record.c
> @@ -282,6 +282,8 @@ try_again:
> else if (err == ENODEV && cpu_list) {
> die("No such device - did you specify"
> " an out-of-range profile CPU?\n");
> + } else if (err == ENOENT) {
> + error("%s event is not supported.", event_name(pos));
> } else if (err == EINVAL && sample_id_all_avail) {
> /*
> * Old kernel, no attr->sample_id_type_all field
> @@ -307,6 +309,10 @@ try_again:
> attr->config = PERF_COUNT_SW_CPU_CLOCK;
> goto try_again;
> }
> +
> + if (err == ENOENT)
> + exit(1);
> +
> printf("\n");
> error("sys_perf_event_open() syscall returned with %d (%s). /bin/dmesg may provide additional information.\n",
> err, strerror(err));
> diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
> index 80c9e06..ab27693 100644
> --- a/tools/perf/builtin-top.c
> +++ b/tools/perf/builtin-top.c
> @@ -852,6 +852,9 @@ try_again:
> die("Permission error - are you root?\n"
> "\t Consider tweaking"
> " /proc/sys/kernel/perf_event_paranoid.\n");
> + else if (err == ENOENT)
> + error("%s event is not supported.", event_name(counter));
> +
> /*
> * If it's cycles then fall back to hrtimer
> * based cpu-clock-tick sw counter, which
> @@ -867,6 +870,10 @@ try_again:
> attr->config = PERF_COUNT_SW_CPU_CLOCK;
> goto try_again;
> }
> +
> + if (err == ENOENT)
> + exit(1);
> +
> printf("\n");
> error("sys_perf_event_open() syscall returned with %d "
> "(%s). /bin/dmesg may provide additional information.\n",
> --
> 1.7.4
On 03/25/11 12:08, Arnaldo Carvalho de Melo wrote:
> Em Fri, Mar 25, 2011 at 12:05:50PM -0600, David Ahern escreveu:
>> Resend of patch sent back in January 2011 in light of recent confusion around
>> unsupported events for a given platform.
>>
>> Improve sys_perf_event_open ENOENT return handling in top and record, just
>> like 5a3446b does for stat.
>>
>> Retry of Arnaldo's patch using error instead of die which allows the fallback
>> from hardware cycles to software clock.
>
> Please use ui__warning(""...) as it will work in the TUI too.
error() is consistent with current pattern. Is ui__warning the
preference for new messages? Is that for perf-top only?
Em Fri, Mar 25, 2011 at 12:26:41PM -0600, David Ahern escreveu:
> On 03/25/11 12:08, Arnaldo Carvalho de Melo wrote:
> > Em Fri, Mar 25, 2011 at 12:05:50PM -0600, David Ahern escreveu:
> >> Resend of patch sent back in January 2011 in light of recent confusion around
> >> unsupported events for a given platform.
> >>
> >> Improve sys_perf_event_open ENOENT return handling in top and record, just
> >> like 5a3446b does for stat.
> >>
> >> Retry of Arnaldo's patch using error instead of die which allows the fallback
> >> from hardware cycles to software clock.
> >
> > Please use ui__warning(""...) as it will work in the TUI too.
>
> error() is consistent with current pattern. Is ui__warning the
> preference for new messages? Is that for perf-top only?
Yes, it is, it takes care of using TUI of doing it like error() does.
Eventually error(), etc will go away.
- Arnaldo