Skip to contentSkip to conversation

perf tools pt: Remove obsolete paragraph in intel-pt.c

PATCH In review Andi Kleen 3 participants 7 messages Nov 9, 2016 – Nov 10, 2016 perf events

lore
Share & tools
lore

Andi Kleen

author
#0 lore raw

PATCH 1/1 perf tools pt: Remove obsolete paragraph in intel-pt.c

From: Andi Kleen <[email protected]>

Since the unprivileged sched switch event was added in perf, PT doesn't need need perf_event_paranoid=-1 anymore for per cpu decoding. So remove the obsolete paragraph in the documentation.

---
 tools/perf/Documentation/intel-pt.txt | 10 ----------
 1 file changed, 10 deletions(-)
tools/perf/Documentation/intel-pt.txt +0 -10
index c6c8318e38a2..c7f817fd3611 100644
@@ -550,16 +550,6 @@ Unless /proc/sys/kernel/perf_event_paranoid is set to -1, unprivileged users
 have memory limits imposed upon them.  That affects what buffer sizes they can
 have as outlined above.
 
-Unless /proc/sys/kernel/perf_event_paranoid is set to -1, unprivileged users are
-not permitted to use tracepoints which means there is insufficient side-band
-information to decode Intel PT in per-cpu mode, and potentially workload-only
-mode too if the workload creates new processes.
-
-Note also, that to use tracepoints, read-access to debugfs is required.  So if
-debugfs is not mounted or the user does not have read-access, it will again not
-be possible to decode Intel PT in per-cpu mode.
-
-
 sched_switch tracepoint
 -----------------------
 
Signature · 1 line
2.5.5
#1 lore raw

Em Tue, Nov 08, 2016 at 04:11:00PM -0800, Andi Kleen escreveu:

Quoted 6 lines

From: Andi Kleen <[email protected]>

Since the unprivileged sched switch event was added in perf, PT doesn't need need perf_event_paranoid=-1 anymore for per cpu decoding. So remove the obsolete paragraph in the documentation.

Thanks for pointing that out, I'll do something slightly different tho, pointing out that from kernel X.Y.Z, when the unprivileged PERF_RECORD_SWITCH metadata event was introduced, this is no longer an issue, having to be considered only on older kernels.

- Arnaldo

Quoted 29 lines
---
 tools/perf/Documentation/intel-pt.txt | 10 ----------
 1 file changed, 10 deletions(-)
tools/perf/Documentation/intel-pt.txt +0 -10
index c6c8318e38a2..c7f817fd3611 100644
@@ -550,16 +550,6 @@ Unless /proc/sys/kernel/perf_event_paranoid is set to -1, unprivileged users
 have memory limits imposed upon them.  That affects what buffer sizes they can
 have as outlined above.
 
-Unless /proc/sys/kernel/perf_event_paranoid is set to -1, unprivileged users are
-not permitted to use tracepoints which means there is insufficient side-band
-information to decode Intel PT in per-cpu mode, and potentially workload-only
-mode too if the workload creates new processes.
-
-Note also, that to use tracepoints, read-access to debugfs is required.  So if
-debugfs is not mounted or the user does not have read-access, it will again not
-be possible to decode Intel PT in per-cpu mode.
-
-
 sched_switch tracepoint
 -----------------------
 
Signature · 1 line
2.5.5
#2 lore raw

Em Wed, Nov 09, 2016 at 10:14:26AM -0300, Arnaldo Carvalho de Melo escreveu:

Quoted 8 lines

Em Tue, Nov 08, 2016 at 04:11:00PM -0800, Andi Kleen escreveu:

Quoted 6 lines

From: Andi Kleen <[email protected]>

Since the unprivileged sched switch event was added in perf, PT doesn't need need perf_event_paranoid=-1 anymore for per cpu decoding. So remove the obsolete paragraph in the documentation.

Quoted 4 lines

Thanks for pointing that out, I'll do something slightly different tho, pointing out that from kernel X.Y.Z, when the unprivileged PERF_RECORD_SWITCH metadata event was introduced, this is no longer an issue, having to be considered only on older kernels.

It ended up as:

tools/perf/Documentation/intel-pt.txt +12 -0
index c6c8318e38a2..4d12db118476 100644
@@ -546,6 +546,18 @@ mode by using the --per-thread option.
 Privileged vs non-privileged users
 ----------------------------------
 
+The v4.2 kernel introduced support for a context switch metadata event,
+PERF_RECORD_SWITCH, which allows unprivileged users to see when their processes
+are scheduled out and in, just not by whom, which is left for the
+PERF_RECORD_SWITCH_CPU_WIDE, that is only accessible in system wide context,
+which in turn requires CAP_SYS_ADMIN.
+
+Please see the 45ac1403f564 ("perf: Add PERF_RECORD_SWITCH to indicate context
+switches") commit, that introduces these metadata events for further info.
+
+When working with kernels < v4.2, the following considerations must be taken,
+as the sched:sched_switch tracepoints will be used to receive such information:
+
 Unless /proc/sys/kernel/perf_event_paranoid is set to -1, unprivileged users
 have memory limits imposed upon them.  That affects what buffer sizes they can
 have as outlined above.
reviewer
#3 lore raw
Quoted 29 lines · Arnaldo Carvalho de Melo

Em Wed, Nov 09, 2016 at 10:14:26AM -0300, Arnaldo Carvalho de Melo escreveu:

Quoted 8 lines

Em Tue, Nov 08, 2016 at 04:11:00PM -0800, Andi Kleen escreveu:

Quoted 6 lines

From: Andi Kleen <[email protected]>

Since the unprivileged sched switch event was added in perf, PT doesn't need need perf_event_paranoid=-1 anymore for per cpu decoding. So remove the obsolete paragraph in the documentation.

Quoted 4 lines

Thanks for pointing that out, I'll do something slightly different tho, pointing out that from kernel X.Y.Z, when the unprivileged PERF_RECORD_SWITCH metadata event was introduced, this is no longer an issue, having to be considered only on older kernels.

It ended up as:

tools/perf/Documentation/intel-pt.txt +5 -0
index c6c8318e38a2..4d12db118476 100644
@@ -546,6 +546,18 @@ mode by using the --per-thread option.
 Privileged vs non-privileged users
 ----------------------------------
 
+The v4.2 kernel introduced support for a context switch metadata event,
+PERF_RECORD_SWITCH, which allows unprivileged users to see when their processes
+are scheduled out and in, just not by whom, which is left for the
+PERF_RECORD_SWITCH_CPU_WIDE, that is only accessible in system wide context,
+which in turn requires CAP_SYS_ADMIN.

tools/perf/Documentation/intel-pt.txt mode by using the --per-thread option. · line 563

Quoted 10 lines
tools/perf/Documentation/intel-pt.txt +7 -0
+
+Please see the 45ac1403f564 ("perf: Add PERF_RECORD_SWITCH to indicate context
+switches") commit, that introduces these metadata events for further info.
+
+When working with kernels < v4.2, the following considerations must be taken,
+as the sched:sched_switch tracepoints will be used to receive such information:
+
 Unless /proc/sys/kernel/perf_event_paranoid is set to -1, unprivileged users
 have memory limits imposed upon them.  That affects what buffer sizes they can
 have as outlined above.

Maybe put that last paragraph about memory limits above the new text.

#4 lore raw

Em Wed, Nov 09, 2016 at 04:01:12PM +0200, Adrian Hunter escreveu:

Quoted 34 lines · Arnaldo Carvalho de Melo
Quoted 33 lines

Em Wed, Nov 09, 2016 at 10:14:26AM -0300, Arnaldo Carvalho de Melo escreveu:

Quoted 8 lines

Em Tue, Nov 08, 2016 at 04:11:00PM -0800, Andi Kleen escreveu:

Quoted 6 lines

From: Andi Kleen <[email protected]>

Since the unprivileged sched switch event was added in perf, PT doesn't need need perf_event_paranoid=-1 anymore for per cpu decoding. So remove the obsolete paragraph in the documentation.

Quoted 4 lines

Thanks for pointing that out, I'll do something slightly different tho, pointing out that from kernel X.Y.Z, when the unprivileged PERF_RECORD_SWITCH metadata event was introduced, this is no longer an issue, having to be considered only on older kernels.

It ended up as:

tools/perf/Documentation/intel-pt.txt +9 -0
index c6c8318e38a2..4d12db118476 100644
@@ -546,6 +546,18 @@ mode by using the --per-thread option.
 Privileged vs non-privileged users
 ----------------------------------
 
+The v4.2 kernel introduced support for a context switch metadata event,
+PERF_RECORD_SWITCH, which allows unprivileged users to see when their processes
+are scheduled out and in, just not by whom, which is left for the
+PERF_RECORD_SWITCH_CPU_WIDE, that is only accessible in system wide context,
+which in turn requires CAP_SYS_ADMIN.
+
+Please see the 45ac1403f564 ("perf: Add PERF_RECORD_SWITCH to indicate context
+switches") commit, that introduces these metadata events for further info.
+
Quoted 8 lines

tools/perf/Documentation/intel-pt.txt mode by using the --per-thread option. · line 563

Quoted 6 lines
tools/perf/Documentation/intel-pt.txt +3 -0
+When working with kernels < v4.2, the following considerations must be taken,
+as the sched:sched_switch tracepoints will be used to receive such information:
+
 Unless /proc/sys/kernel/perf_event_paranoid is set to -1, unprivileged users
 have memory limits imposed upon them.  That affects what buffer sizes they can
 have as outlined above.

Maybe put that last paragraph about memory limits above the new text.

Ok, as it is not affected by the new text, thanks.

- Arnaldo

maintainer Acked-by Adrian Hunter
#5 lore raw
In reply to Adrian Hunter

Em Wed, Nov 09, 2016 at 04:01:12PM +0200, Adrian Hunter escreveu:

Quoted 16 lines · Arnaldo Carvalho de Melo
Quoted 15 lines

Em Wed, Nov 09, 2016 at 10:14:26AM -0300, Arnaldo Carvalho de Melo escreveu:
+++ b/tools/perf/Documentation/intel-pt.txt

(quoted diff) +9 -0
@@ -546,6 +546,18 @@ mode by using the --per-thread option.
 Privileged vs non-privileged users
 ----------------------------------
 
+The v4.2 kernel introduced support for a context switch metadata event,
+PERF_RECORD_SWITCH, which allows unprivileged users to see when their processes
+are scheduled out and in, just not by whom, which is left for the
+PERF_RECORD_SWITCH_CPU_WIDE, that is only accessible in system wide context,
+which in turn requires CAP_SYS_ADMIN.
+
+Please see the 45ac1403f564 ("perf: Add PERF_RECORD_SWITCH to indicate context
+switches") commit, that introduces these metadata events for further info.
+
Quoted 8 lines

Review mode by using the --per-thread option. · line 563

Quoted 6 lines
(quoted diff) +3 -0
+When working with kernels < v4.2, the following considerations must be taken,
+as the sched:sched_switch tracepoints will be used to receive such information:
+
 Unless /proc/sys/kernel/perf_event_paranoid is set to -1, unprivileged users
 have memory limits imposed upon them.  That affects what buffer sizes they can
 have as outlined above.

Maybe put that last paragraph about memory limits above the new text.

Ok, adding your Acked-by, it now stands as, please ack:

commit c955ced680270b95a3de6a2434311befaeaea948

Author: Arnaldo Carvalho de Melo <[email protected]>
Date:   Wed Nov 9 11:04:05 2016 -0300

perf intel-pt: Update documentation about context switch events

Since the unprivileged sched switch event was added in perf, PT doesn't need need perf_event_paranoid=-1 anymore for per cpu decoding.

Add a note stating that that is only needed for kernels < 4.2.

Report-link: http://lkml.kernel.org/r/http://lkml.kernel.org/n/[email protected]

Fixes 45ac1403f564 ("perf: Add PERF_RECORD_SWITCH to indicate context switches")
tools/perf/Documentation/intel-pt.txt +17 -2
index c6c8318e38a2..b0b3007d3c9c 100644
@@ -550,6 +550,18 @@ Unless /proc/sys/kernel/perf_event_paranoid is set to -1, unprivileged users
 have memory limits imposed upon them.  That affects what buffer sizes they can
 have as outlined above.
 
+The v4.2 kernel introduced support for a context switch metadata event,
+PERF_RECORD_SWITCH, which allows unprivileged users to see when their processes
+are scheduled out and in, just not by whom, which is left for the
+PERF_RECORD_SWITCH_CPU_WIDE, that is only accessible in system wide context,
+which in turn requires CAP_SYS_ADMIN.
+
+Please see the 45ac1403f564 ("perf: Add PERF_RECORD_SWITCH to indicate context
+switches") commit, that introduces these metadata events for further info.
+
+When working with kernels < v4.2, the following considerations must be taken,
+as the sched:sched_switch tracepoints will be used to receive such information:
+
 Unless /proc/sys/kernel/perf_event_paranoid is set to -1, unprivileged users are
 not permitted to use tracepoints which means there is insufficient side-band
 information to decode Intel PT in per-cpu mode, and potentially workload-only
@@ -564,8 +576,11 @@ sched_switch tracepoint
 -----------------------
 
 The sched_switch tracepoint is used to provide side-band data for Intel PT
-decoding.  sched_switch events are automatically added. e.g. the second event
-shown below
+decoding in kernels where the PERF_RECORD_SWITCH metadata event isn't
+available.
+
+The sched_switch events are automatically added. e.g. the second event shown
+below:
 
 	$ perf record -vv -e intel_pt//u uname
 	------------------------------------------------------------
reviewer
#6 lore raw
Quoted 26 lines · Arnaldo Carvalho de Melo

Em Wed, Nov 09, 2016 at 04:01:12PM +0200, Adrian Hunter escreveu:

Quoted 16 lines · Arnaldo Carvalho de Melo
Quoted 15 lines

Em Wed, Nov 09, 2016 at 10:14:26AM -0300, Arnaldo Carvalho de Melo escreveu:
+++ b/tools/perf/Documentation/intel-pt.txt

tools/perf/Documentation/intel-pt.txt +9 -0
@@ -546,6 +546,18 @@ mode by using the --per-thread option.
 Privileged vs non-privileged users
 ----------------------------------
 
+The v4.2 kernel introduced support for a context switch metadata event,
+PERF_RECORD_SWITCH, which allows unprivileged users to see when their processes
+are scheduled out and in, just not by whom, which is left for the
+PERF_RECORD_SWITCH_CPU_WIDE, that is only accessible in system wide context,
+which in turn requires CAP_SYS_ADMIN.
+
+Please see the 45ac1403f564 ("perf: Add PERF_RECORD_SWITCH to indicate context
+switches") commit, that introduces these metadata events for further info.
+
Quoted 8 lines

tools/perf/Documentation/intel-pt.txt Unless /proc/sys/kernel/perf_event_paranoid is set to -1, unprivileged users

Quoted 6 lines
tools/perf/Documentation/intel-pt.txt +3 -0
+When working with kernels < v4.2, the following considerations must be taken,
+as the sched:sched_switch tracepoints will be used to receive such information:
+
 Unless /proc/sys/kernel/perf_event_paranoid is set to -1, unprivileged users
 have memory limits imposed upon them.  That affects what buffer sizes they can
 have as outlined above.

Maybe put that last paragraph about memory limits above the new text.

Quoted 1 line

Ok, adding your Acked-by, it now stands as, please ack:

Yes that is good. Thank you!

Quoted 57 lines

commit c955ced680270b95a3de6a2434311befaeaea948

Author: Arnaldo Carvalho de Melo <[email protected]>
Date:   Wed Nov 9 11:04:05 2016 -0300

perf intel-pt: Update documentation about context switch events

Since the unprivileged sched switch event was added in perf, PT doesn't need need perf_event_paranoid=-1 anymore for per cpu decoding.

Add a note stating that that is only needed for kernels < 4.2.

Report-link: http://lkml.kernel.org/r/http://lkml.kernel.org/n/[email protected]

Fixes 45ac1403f564 ("perf: Add PERF_RECORD_SWITCH to indicate context switches")
tools/perf/Documentation/intel-pt.txt +17 -2
index c6c8318e38a2..b0b3007d3c9c 100644
@@ -550,6 +550,18 @@ Unless /proc/sys/kernel/perf_event_paranoid is set to -1, unprivileged users
 have memory limits imposed upon them.  That affects what buffer sizes they can
 have as outlined above.
 
+The v4.2 kernel introduced support for a context switch metadata event,
+PERF_RECORD_SWITCH, which allows unprivileged users to see when their processes
+are scheduled out and in, just not by whom, which is left for the
+PERF_RECORD_SWITCH_CPU_WIDE, that is only accessible in system wide context,
+which in turn requires CAP_SYS_ADMIN.
+
+Please see the 45ac1403f564 ("perf: Add PERF_RECORD_SWITCH to indicate context
+switches") commit, that introduces these metadata events for further info.
+
+When working with kernels < v4.2, the following considerations must be taken,
+as the sched:sched_switch tracepoints will be used to receive such information:
+
 Unless /proc/sys/kernel/perf_event_paranoid is set to -1, unprivileged users are
 not permitted to use tracepoints which means there is insufficient side-band
 information to decode Intel PT in per-cpu mode, and potentially workload-only
@@ -564,8 +576,11 @@ sched_switch tracepoint
 -----------------------
 
 The sched_switch tracepoint is used to provide side-band data for Intel PT
-decoding.  sched_switch events are automatically added. e.g. the second event
-shown below
+decoding in kernels where the PERF_RECORD_SWITCH metadata event isn't
+available.
+
+The sched_switch events are automatically added. e.g. the second event shown
+below:
 
 	$ perf record -vv -e intel_pt//u uname
 	------------------------------------------------------------