perf: support enable and disable commands in stat and record modes
v2 cover perf: support enable and disable commands in stat and record modes
Changes in v2:
- renamed resume and pause commands to enable and disable ones, renamed CTL_CMD_RESUME and CTL_CMD_PAUSE to CTL_CMD_ENABLE and CTL_CMD_DISABLE to fit to the appropriate ioctls and avoid mixing up with PAUSE_OUTPUT ioctl;
- factored out event handling loop into a handle_events() for stat mode;
- separated -D,--delay=-1 into separate patches for stat and record modes;
v1: https://lore.kernel.org/lkml/[email protected]/
repo: tip of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux.git perf/core
The patch set implements handling of 'start disabled', 'enable' and 'disable' external control commands which can be provided for stat and record modes of the tool from an external controlling process. 'start disabled' command can be used to postpone enabling of events in the beginning of a monitoring session. 'enable' and 'disable' commands can be used to enable and disable events correspondingly any time after the start of the session.
The 'start disabled', 'enable' and 'disable' external control commands can be used to focus measurement on specially selected time intervals of workload execution. Focused measurement reduces tool intrusion and influence on workload behavior, reduces distortion and amount of collected and stored data, mitigates data accuracy loss because measurement and data capturing happen only during intervals of interest.
A controlling process can be a bash shell script [1], native executable or any other language program that can directly work with file descriptors, e.g. pipes [2], and spawn a process, specially the tool one.
-D,--delay <val> option is extended with -1 value to skip events enabling in the beginning of a monitoring session ('start disabled' command). --ctl-fd and --ctl-fd-ack command line options are introduced to provide the tool with a pair of file descriptors to listen to control commands and reply to the controlling process on the completion of received commands.
The tool reads two byte control command message from ctl-fd descriptor, handles the command and optionally replies two bytes acknowledgement message to fd-ack descriptor, if it is specified on the command line. 'enable' command is recognized as 'e' character message and 'disable' command is recognized as 'd' character message both received from ctl-fd descriptor. Completion message is 'a''\n' and sent to fd-ack descriptor.
example bash script demonstrating simple use case:
#!/bin/bash
ctl_dir=/tmp/
ctl_fifo=${ctl_dir}perf_ctl.fifo
test -p ${ctl_fifo} && unlink ${ctl_fifo}
mkfifo ${ctl_fifo} && exec {ctl_fd}<>${ctl_fifo}
ctl_ack_fifo=${ctl_dir}perf_ctl_ack.fifo
test -p ${ctl_ack_fifo} && unlink ${ctl_ack_fifo} mkfifo ${ctl_ack_fifo} && exec {ctl_fd_ack}<>${ctl_ack_fifo}
perf stat -D -1 -e cpu-cycles -a -I 1000 \
--ctl-fd ${ctl_fd} --ctl-fd-ack ${ctl_fd_ack} \
-- sleep 40 &perf_pid=$!
sleep 5 && echo 'e' >&${ctl_fd} && read -u ${ctl_fd_ack} e1 && echo "enabled(${e1})" sleep 10 && echo 'd' >&${ctl_fd} && read -u ${ctl_fd_ack} d1 && echo "disabled(${d1})" sleep 5 && echo 'e' >&${ctl_fd} && read -u ${ctl_fd_ack} e2 && echo "enabled(${e2})" sleep 10 && echo 'd' >&${ctl_fd} && read -u ${ctl_fd_ack} d2 && echo "disabled(${d2})"
exec {ctl_fd_ack}>&- && unlink ${ctl_ack_fifo}
exec {ctl_fd}>&- && unlink ${ctl_fifo}
wait -n ${perf_pid}
exit $?
Script output:
[root@host acme] example
Events disabled
# time counts unit events
1.001101062 <not counted> cpu-cycles
2.002994944 <not counted> cpu-cycles
3.004864340 <not counted> cpu-cycles
4.006727177 <not counted> cpu-cycles
Events enabled
enabled(a)
4.993808464 3,124,246 cpu-cycles
5.008597004 3,325,624 cpu-cycles
6.010387483 83,472,992 cpu-cycles
7.012266598 55,877,621 cpu-cycles
8.014175695 97,892,729 cpu-cycles
9.016056093 68,461,242 cpu-cycles
10.017937507 55,449,643 cpu-cycles
11.019830154 68,938,167 cpu-cycles
12.021719952 55,164,101 cpu-cycles
13.023627550 70,535,720 cpu-cycles
14.025580995 53,240,125 cpu-cycles
disabled(a)
14.997518260 53,558,068 cpu-cycles
Events disabled
15.027216416 <not counted> cpu-cycles
16.029052729 <not counted> cpu-cycles
17.030904762 <not counted> cpu-cycles
18.032073424 <not counted> cpu-cycles
19.033805074 <not counted> cpu-cycles
Events enabled
enabled(a)
20.001279097 3,021,022 cpu-cycles
20.035044381 6,434,367 cpu-cycles
21.036923813 89,358,251 cpu-cycles
22.038825169 72,516,351 cpu-cycles
# time counts unit events
23.040715596 55,046,157 cpu-cycles
24.042643757 78,128,649 cpu-cycles
25.044558535 61,052,428 cpu-cycles
26.046452785 62,142,806 cpu-cycles
27.048353021 74,477,971 cpu-cycles
28.050241286 61,001,623 cpu-cycles
29.052149961 61,653,502 cpu-cycles
disabled(a)
30.004980264 82,729,640 cpu-cycles
Events disabled
30.053516176 <not counted> cpu-cycles
31.055348366 <not counted> cpu-cycles
32.057202097 <not counted> cpu-cycles
33.059040702 <not counted> cpu-cycles
34.060843288 <not counted> cpu-cycles
35.000888624 <not counted> cpu-cycles [root@host acme]#
[1] http://man7.org/linux/man-pages/man1/bash.1.html
[2] http://man7.org/linux/man-pages/man2/pipe.2.html
---
Alexey Budankov (11):
perf evlist: introduce control file descriptors
perf evlist: implement control command handling functions perf stat: factor out event handling loop into a function perf stat: extend -D,--delay option with -1 value
perf stat: implement control commands handling
perf stat: introduce --ctl-fd[-ack] options
perf docs: extend stat mode docs with info on --ctl-fd[-ack] options perf record: extend -D,--delay option with -1 value
perf record: implement control commands handling
perf record: introduce --ctl-fd[-ack] options
perf docs: extend record mode docs with info on --ctl-fd[-ack] options
tools/perf/Documentation/perf-record.txt | 39 +++++++ tools/perf/Documentation/perf-stat.txt | 40 +++++++ tools/perf/builtin-record.c | 39 ++++++- tools/perf/builtin-stat.c | 137 ++++++++++++++++------- tools/perf/builtin-trace.c | 2 +- tools/perf/util/evlist.c | 103 +++++++++++++++++ tools/perf/util/evlist.h | 18 +++ tools/perf/util/record.h | 4 +- tools/perf/util/stat.h | 4 +- 9 files changed, 339 insertions(+), 47 deletions(-)
Signature · 1 line
2.24.1
PATCH v2 1/11 perf evlist: introduce control file descriptors
Define and initialize control file descriptors.
--- tools/perf/util/evlist.c | 3 +++ tools/perf/util/evlist.h | 3 +++ 2 files changed, 6 insertions(+)
tools/perf/util/evlist.c +3 -0
@@ -63,6 +63,9 @@ void evlist__init(struct evlist *evlist, struct perf_cpu_map *cpus,perf_evlist__set_maps (&evlist->core, cpus, threads); evlist->workload.pid =-1 ; evlist->bkw_mmap_state =BKW_MMAP_NOTREADY ; + evlist->ctl_fd =-1 ; + evlist->ctl_fd_ack =-1 ; + evlist->ctl_fd_pos =-1 ; }struct evlist *evlist__new (void )
tools/perf/util/evlist.h +3 -0
@@ -74,6 +74,9 @@ struct evlist {pthread_t th;volatile int done; } thread; +int ctl_fd; +int ctl_fd_ack; +int ctl_fd_pos; };struct evsel_str_handler {
Signature · 1 line
2.24.1
PATCH v2 2/11 perf evlist: implement control command handling functions
In reply to Alexey BudankovImplement functions of initialization, finalization and processing of control commands coming from control file descriptors.
--- tools/perf/util/evlist.c | 100 +++++++++++++++++++++++++++++++++++++++ tools/perf/util/evlist.h | 12 +++++ 2 files changed, 112 insertions(+)
tools/perf/util/evlist.c +100 -0
@@ -1707,3 +1707,103 @@ struct evsel *perf_evlist__reset_weak_group(struct evlist *evsel_list, }return leader; } + +int perf_evlist__initialize_ctlfd (struct evlist *evlist,int ctl_fd,int ctl_fd_ack) +{ +if (ctl_fd ==-1 ) { +pr_debug ("Control descriptor is not initialized\n" ); +return 0 ; + } + + evlist->ctl_fd_pos =perf_evlist__add_pollfd (&evlist->core, ctl_fd,NULL ,POLLIN ); +if (evlist->ctl_fd_pos <0 ) { + evlist->ctl_fd_pos =-1 ; +pr_err ("Failed to add ctl fd entry: %m\n" ); +return -1 ; + } + + evlist->ctl_fd = ctl_fd; + evlist->ctl_fd_ack = ctl_fd_ack; + +return 0 ; +} + +int perf_evlist__finalize_ctlfd (struct evlist *evlist) +{ +if (evlist->ctl_fd_pos ==-1 ) +return 0 ; + + evlist->core.pollfd.entries[evlist->ctl_fd_pos].fd =-1 ; + evlist->ctl_fd_pos =-1 ; + evlist->ctl_fd_ack =-1 ; + evlist->ctl_fd =-1 ; + +return 0 ; +} + +static int perf_evlist__ctlfd_recv (struct evlist *evlist,enum evlist_ctl_cmd *cmd) +{ +int err; +char buf[2 ]; + + err =read (evlist->ctl_fd, &buf,sizeof (buf)); +if (err >0 ) + *cmd = buf[0 ]; +else if (err ==-1 ) +pr_err ("Failed to read from ctlfd %d: %m\n" , evlist->ctl_fd); + +return err; +} + +static int perf_evlist__ctlfd_ack (struct evlist *evlist) +{ +int err; +char buf[2 ] = {CTL_CMD_ACK ,'\n' }; + +if (evlist->ctl_fd_ack ==-1 ) +return 0 ; + + err =write (evlist->ctl_fd_ack, buf,sizeof (buf)); +if (err ==-1 ) +pr_err ("failed to write to ctl_ack_fd %d: %m\n" , evlist->ctl_fd_ack); + +return err; +} + +int perf_evlist__ctlfd_process (struct evlist *evlist,enum evlist_ctl_cmd *cmd) +{ +int err =0 ; +int ctlfd_pos = evlist->ctl_fd_pos; +struct pollfd *entries = evlist->core.pollfd.entries; + +if (!entries[ctlfd_pos].revents) +return 0 ; + +if (entries[ctlfd_pos].revents &POLLIN ) { + err =perf_evlist__ctlfd_recv (evlist, cmd); +if (err >0 ) { +switch (*cmd) { +case CTL_CMD_ENABLE : +evlist__enable (evlist); +break ; +case CTL_CMD_DISABLE : +evlist__disable (evlist); +break ; +case CTL_CMD_ACK : +case CTL_CMD_UNSUPPORTED : +default : +pr_debug ("ctlfd: unsupported %d\n" , *cmd); +break ; + } +if (!(*cmd ==CTL_CMD_ACK || *cmd ==CTL_CMD_UNSUPPORTED )) +perf_evlist__ctlfd_ack (evlist); + } + } + +if (entries[ctlfd_pos].revents & (POLLHUP |POLLERR )) +perf_evlist__finalize_ctlfd (evlist); +else + entries[ctlfd_pos].revents =0 ; + +return err; +}
tools/perf/util/evlist.h +12 -0
@@ -358,4 +358,16 @@ void perf_evlist__force_leader(struct evlist *evlist);struct evsel *perf_evlist__reset_weak_group (struct evlist *evlist,struct evsel *evsel,bool close); + +enum evlist_ctl_cmd { +CTL_CMD_UNSUPPORTED =0 , +CTL_CMD_ENABLE ='e' , +CTL_CMD_DISABLE ='d' , +CTL_CMD_ACK ='a' +}; + +int perf_evlist__initialize_ctlfd (struct evlist *evlist,int ctl_fd,int ctl_fd_ack); +int perf_evlist__finalize_ctlfd (struct evlist *evlist); +int perf_evlist__ctlfd_process (struct evlist *evlist,enum evlist_ctl_cmd *cmd); +#endif /* __PERF_EVLIST_H */
Signature · 1 line
2.24.1
Re: [PATCH v2 02/11] perf evlist: implement control command handling functions
Em Wed, May 06, 2020 at 09:19:22PM +0300, Alexey Budankov escreveu:
Quoted 123 lines
Implement functions of initialization, finalization and processing of control commands coming from control file descriptors.
Signed-off-by Alexey Budankov <[email protected]>--- tools/perf/util/evlist.c | 100 +++++++++++++++++++++++++++++++++++++++ tools/perf/util/evlist.h | 12 +++++ 2 files changed, 112 insertions(+)tools/perf/util/evlist.c +100
-0@@ -1707,3 +1707,103 @@ struct evsel *perf_evlist__reset_weak_group(struct evlist *evsel_list, }return leader; } + +int perf_evlist__initialize_ctlfd (struct evlist *evlist,int ctl_fd,int ctl_fd_ack) +{ +if (ctl_fd ==-1 ) { +pr_debug ("Control descriptor is not initialized\n" ); +return 0 ; + } + + evlist->ctl_fd_pos =perf_evlist__add_pollfd (&evlist->core, ctl_fd,NULL ,POLLIN ); +if (evlist->ctl_fd_pos <0 ) { + evlist->ctl_fd_pos =-1 ; +pr_err ("Failed to add ctl fd entry: %m\n" ); +return -1 ; + } + + evlist->ctl_fd = ctl_fd; + evlist->ctl_fd_ack = ctl_fd_ack; + +return 0 ; +} + +int perf_evlist__finalize_ctlfd (struct evlist *evlist) +{ +if (evlist->ctl_fd_pos ==-1 ) +return 0 ; + + evlist->core.pollfd.entries[evlist->ctl_fd_pos].fd =-1 ; + evlist->ctl_fd_pos =-1 ; + evlist->ctl_fd_ack =-1 ; + evlist->ctl_fd =-1 ; + +return 0 ; +} + +static int perf_evlist__ctlfd_recv (struct evlist *evlist,enum evlist_ctl_cmd *cmd) +{ +int err; +char buf[2 ]; + + err =read (evlist->ctl_fd, &buf,sizeof (buf)); +if (err >0 ) + *cmd = buf[0 ]; +else if (err ==-1 ) +pr_err ("Failed to read from ctlfd %d: %m\n" , evlist->ctl_fd); + +return err; +} + +static int perf_evlist__ctlfd_ack (struct evlist *evlist) +{ +int err; +char buf[2 ] = {CTL_CMD_ACK ,'\n' }; + +if (evlist->ctl_fd_ack ==-1 ) +return 0 ; + + err =write (evlist->ctl_fd_ack, buf,sizeof (buf)); +if (err ==-1 ) +pr_err ("failed to write to ctl_ack_fd %d: %m\n" , evlist->ctl_fd_ack); + +return err; +} + +int perf_evlist__ctlfd_process (struct evlist *evlist,enum evlist_ctl_cmd *cmd) +{ +int err =0 ; +int ctlfd_pos = evlist->ctl_fd_pos; +struct pollfd *entries = evlist->core.pollfd.entries; + +if (!entries[ctlfd_pos].revents) +return 0 ; + +if (entries[ctlfd_pos].revents &POLLIN ) { + err =perf_evlist__ctlfd_recv (evlist, cmd); +if (err >0 ) { +switch (*cmd) { +case CTL_CMD_ENABLE : +evlist__enable (evlist); +break ; +case CTL_CMD_DISABLE : +evlist__disable (evlist); +break ; +case CTL_CMD_ACK : +case CTL_CMD_UNSUPPORTED : +default : +pr_debug ("ctlfd: unsupported %d\n" , *cmd); +break ; + } +if (!(*cmd ==CTL_CMD_ACK || *cmd ==CTL_CMD_UNSUPPORTED )) +perf_evlist__ctlfd_ack (evlist); + } + } + +if (entries[ctlfd_pos].revents & (POLLHUP |POLLERR )) +perf_evlist__finalize_ctlfd (evlist); +else + entries[ctlfd_pos].revents =0 ; + +return err; +}tools/perf/util/evlist.h +0
-0
tools/perf/util/evlist.h void perf_evlist__force_leader(struct evlist *evlist); · line 367
Quoted 10 lines
tools/perf/util/evlist.h +7
-0@@ -358,4 +358,16 @@ void perf_evlist__force_leader(struct evlist *evlist);struct evsel *perf_evlist__reset_weak_group (struct evlist *evlist,struct evsel *evsel,bool close); + +enum evlist_ctl_cmd { +CTL_CMD_UNSUPPORTED =0 , +CTL_CMD_ENABLE ='e' , +CTL_CMD_DISABLE ='d' , +CTL_CMD_ACK ='a' +};
Can we make this a string, I think we'll eventually ask for lots more stuff, like asking for a --switch-output snapshot with --overwrite, reconfiguring events to increase/decrease frequency, etc, interfacing with PERF_EVENT_IOC_MODIFY_ATTRIBUTES, PERF_EVENT_IOC_SET_FILTER, etc.
This will also allow us to have parameters, etc, wdyt?
Also please since these are events that deal with 'struct evlist', name them with the evlist__ prefix, not the perf_evlist__ one, as those should be used with 'struct perf_evlist', i.e. the one in libperf (tools/lib/perf/).
Right now this is inconsistent, we did it that way to minimize disruption of the codebase when moving things from tools/perf/ to tools/lib/perf/, but this confuses things and I just did a s/perf_evsel__/evsel__) for things dealing with 'struct evsel', so lets not add new ones with the wrong prefix, eventually we'll have perf_ only for things in libperf.
Quoted 10 lines
tools/perf/util/evlist.h +5
-0+ +int perf_evlist__initialize_ctlfd (struct evlist *evlist,int ctl_fd,int ctl_fd_ack); +int perf_evlist__finalize_ctlfd (struct evlist *evlist); +int perf_evlist__ctlfd_process (struct evlist *evlist,enum evlist_ctl_cmd *cmd); +#endif /* __PERF_EVLIST_H */ Signature · 1 line
2.24.1
Signature · 1 line
- Arnaldo
Re: [PATCH v2 02/11] perf evlist: implement control command handling functions
Quoted 10 lines · Arnaldo Carvalho de Melo
Em Wed, May 06, 2020 at 09:19:22PM +0300, Alexey Budankov escreveu:
Quoted 9 lines
Implement functions of initialization, finalization and processing of control commands coming from control file descriptors.
Signed-off-by Alexey Budankov <[email protected]>--- tools/perf/util/evlist.c | 100 +++++++++++++++++++++++++++++++++++++++ tools/perf/util/evlist.h | 12 +++++ 2 files changed, 112 insertions(+)
<SNIP>
Quoted 9 lines
Quoted 9 lines
tools/perf/util/evlist.h +1
-0@@ -358,4 +358,16 @@ void perf_evlist__force_leader(struct evlist *evlist);struct evsel *perf_evlist__reset_weak_group (struct evlist *evlist,struct evsel *evsel,bool close); +
Quoted 8 lines
tools/perf/util/evlist.h void perf_evlist__force_leader(struct evlist *evlist); · line 367
Quoted 6 lines
tools/perf/util/evlist.h +6
-0+enum evlist_ctl_cmd { +CTL_CMD_UNSUPPORTED =0 , +CTL_CMD_ENABLE ='e' , +CTL_CMD_DISABLE ='d' , +CTL_CMD_ACK ='a' +};Can we make this a string, I think we'll eventually ask for lots more
Like this?
#define EVLIST__CTL_CMD_ENABLE "enable"
#define EVLIST__CTL_CMD_DISABLE "disable"
#define EVLIST__CTL_CMD_ACK "ack"
Quoted 5 lines
stuff, like asking for a --switch-output snapshot with --overwrite, reconfiguring events to increase/decrease frequency, etc, interfacing with
PERF_EVENT_IOC_MODIFY_ATTRIBUTES,PERF_EVENT_IOC_SET_FILTER, etc.This will also allow us to have parameters, etc, wdyt?
Being a part of this patch the extension will implement configurability that potentially could never be used.
Switch from int to string commands of variable length belongs to the patches also implementing usage of that string commands.
Quoted 5 lines
Also please since these are events that deal with 'struct evlist', name them with the
evlist__prefix, not theperf_evlist__one, as those should be used with 'structperf_evlist', i.e. the one in libperf (tools/lib/perf/).
Accepted in v3.
Quoted 19 lines
Right now this is inconsistent, we did it that way to minimize disruption of the codebase when moving things from
tools/perf/totools/lib/perf/, but this confuses things and I just did a s/perf_evsel__/evsel__) for things dealing with 'struct evsel', so lets not add new ones with the wrong prefix, eventually we'll haveperf_only for things in libperf.Quoted 10 lines
tools/perf/util/evlist.h +5
-0+ +int perf_evlist__initialize_ctlfd (struct evlist *evlist,int ctl_fd,int ctl_fd_ack); +int perf_evlist__finalize_ctlfd (struct evlist *evlist); +int perf_evlist__ctlfd_process (struct evlist *evlist,enum evlist_ctl_cmd *cmd); +#endif /* __PERF_EVLIST_H */ Signature · 1 line
2.24.1
Re: [PATCH v2 02/11] perf evlist: implement control command handling functions
Em Thu, May 07, 2020 at 11:32:53AM +0300, Alexey Budankov escreveu:
Quoted 33 lines · Arnaldo Carvalho de Melo
Quoted 10 lines
Em Wed, May 06, 2020 at 09:19:22PM +0300, Alexey Budankov escreveu:
Quoted 9 lines
Implement functions of initialization, finalization and processing of control commands coming from control file descriptors.
Signed-off-by Alexey Budankov <[email protected]>--- tools/perf/util/evlist.c | 100 +++++++++++++++++++++++++++++++++++++++ tools/perf/util/evlist.h | 12 +++++ 2 files changed, 112 insertions(+)<SNIP>
Quoted 9 lines
Quoted 9 lines
tools/perf/util/evlist.h +1
-0@@ -358,4 +358,16 @@ void perf_evlist__force_leader(struct evlist *evlist);struct evsel *perf_evlist__reset_weak_group (struct evlist *evlist,struct evsel *evsel,bool close); +Quoted 8 lines
tools/perf/util/evlist.h void perf_evlist__force_leader(struct evlist *evlist); · line 367
Quoted 6 lines
tools/perf/util/evlist.h +6
-0+enum evlist_ctl_cmd { +CTL_CMD_UNSUPPORTED =0 , +CTL_CMD_ENABLE ='e' , +CTL_CMD_DISABLE ='d' , +CTL_CMD_ACK ='a' +};Can we make this a string, I think we'll eventually ask for lots more
Quoted 5 lines
Like this?
#define
EVLIST__CTL_CMD_ENABLE"enable"
#defineEVLIST__CTL_CMD_DISABLE"disable"
#defineEVLIST__CTL_CMD_ACK"ack"
Yeah
Quoted 11 lines
Quoted 5 lines
stuff, like asking for a --switch-output snapshot with --overwrite, reconfiguring events to increase/decrease frequency, etc, interfacing with
PERF_EVENT_IOC_MODIFY_ATTRIBUTES,PERF_EVENT_IOC_SET_FILTER, etc.This will also allow us to have parameters, etc, wdyt?
Being a part of this patch the extension will implement configurability that potentially could never be used.
Switch from int to string commands of variable length belongs to the patches also implementing usage of that string commands.
Well, at that point we would have to support both, i.e. the way you're doing now with integers, and as strings, otherwise 3rd party tooling (vtune? :)) using this interface would break.
I.e. this is like the syscall interface.
So if we have "enable" now we can go ahead and forever understand that as "please enable this evlist", but in the future we can extend it and pass parameters to it, to control how that enablement will take place, perhaps with a delay, etc.
Quoted 26 lines
Quoted 4 lines
Also please since these are events that deal with 'struct evlist', name them with the
evlist__prefix, not theperf_evlist__one, as those should be used with 'structperf_evlist', i.e. the one in libperf (tools/lib/perf/).Accepted in v3.
Quoted 19 lines
Right now this is inconsistent, we did it that way to minimize disruption of the codebase when moving things from
tools/perf/totools/lib/perf/, but this confuses things and I just did a s/perf_evsel__/evsel__) for things dealing with 'struct evsel', so lets not add new ones with the wrong prefix, eventually we'll haveperf_only for things in libperf.Quoted 10 lines
tools/perf/util/evlist.h +5
-0+ +int perf_evlist__initialize_ctlfd (struct evlist *evlist,int ctl_fd,int ctl_fd_ack); +int perf_evlist__finalize_ctlfd (struct evlist *evlist); +int perf_evlist__ctlfd_process (struct evlist *evlist,enum evlist_ctl_cmd *cmd); +#endif /* __PERF_EVLIST_H */ Signature · 1 line
2.24.1
Signature · 1 line
- Arnaldo
Re: [PATCH v2 02/11] perf evlist: implement control command handling functions
Quoted 34 lines · Arnaldo Carvalho de Melo
Em Thu, May 07, 2020 at 11:32:53AM +0300, Alexey Budankov escreveu:
Quoted 33 lines · Arnaldo Carvalho de Melo
Quoted 10 lines
Em Wed, May 06, 2020 at 09:19:22PM +0300, Alexey Budankov escreveu:
Quoted 9 lines
Implement functions of initialization, finalization and processing of control commands coming from control file descriptors.
Signed-off-by Alexey Budankov <[email protected]>--- tools/perf/util/evlist.c | 100 +++++++++++++++++++++++++++++++++++++++ tools/perf/util/evlist.h | 12 +++++ 2 files changed, 112 insertions(+)<SNIP>
Quoted 17 lines
Quoted 5 lines
tools/perf/util/evlist.h +0
-0tools/perf/util/evlist.h void perf_evlist__force_leader(struct evlist *evlist); · line 367
Quoted 10 lines
tools/perf/util/evlist.h +7
-0@@ -358,4 +358,16 @@ void perf_evlist__force_leader(struct evlist *evlist);struct evsel *perf_evlist__reset_weak_group (struct evlist *evlist,struct evsel *evsel,bool close); + +enum evlist_ctl_cmd { +CTL_CMD_UNSUPPORTED =0 , +CTL_CMD_ENABLE ='e' , +CTL_CMD_DISABLE ='d' , +CTL_CMD_ACK ='a' +};Can we make this a string, I think we'll eventually ask for lots more
Quoted 7 lines
Quoted 5 lines
Like this?
#define
EVLIST__CTL_CMD_ENABLE"enable"
#defineEVLIST__CTL_CMD_DISABLE"disable"
#defineEVLIST__CTL_CMD_ACK"ack"Yeah
Well, ok. Accepted in v3.
Command becomes of variable length of chars + \n,
in comparison to current single char + \n.
Quoted 51 lines
Quoted 5 lines
Quoted 5 lines
stuff, like asking for a --switch-output snapshot with --overwrite, reconfiguring events to increase/decrease frequency, etc, interfacing with
PERF_EVENT_IOC_MODIFY_ATTRIBUTES,PERF_EVENT_IOC_SET_FILTER, etc.Quoted 7 lines
Quoted 1 line
This will also allow us to have parameters, etc, wdyt?
Being a part of this patch the extension will implement configurability that potentially could never be used.
Switch from int to string commands of variable length belongs to the patches also implementing usage of that string commands.
Well, at that point we would have to support both, i.e. the way you're doing now with integers, and as strings, otherwise 3rd party tooling (vtune? :)) using this interface would break.
I.e. this is like the syscall interface.
So if we have "enable" now we can go ahead and forever understand that as "please enable this evlist", but in the future we can extend it and pass parameters to it, to control how that enablement will take place, perhaps with a delay, etc.
Quoted 26 lines
Quoted 4 lines
Also please since these are events that deal with 'struct evlist', name them with the
evlist__prefix, not theperf_evlist__one, as those should be used with 'structperf_evlist', i.e. the one in libperf (tools/lib/perf/).Accepted in v3.
Quoted 19 lines
Right now this is inconsistent, we did it that way to minimize disruption of the codebase when moving things from
tools/perf/totools/lib/perf/, but this confuses things and I just did a s/perf_evsel__/evsel__) for things dealing with 'struct evsel', so lets not add new ones with the wrong prefix, eventually we'll haveperf_only for things in libperf.Quoted 10 lines
tools/perf/util/evlist.h +5
-0+ +int perf_evlist__initialize_ctlfd (struct evlist *evlist,int ctl_fd,int ctl_fd_ack); +int perf_evlist__finalize_ctlfd (struct evlist *evlist); +int perf_evlist__ctlfd_process (struct evlist *evlist,enum evlist_ctl_cmd *cmd); +#endif /* __PERF_EVLIST_H */ Signature · 1 line
2.24.1
PATCH v2 3/11 perf stat: factor out event handling loop into a function
In reply to Alexey BudankovFactor out event handling loop into handle_events() function.
--- tools/perf/builtin-stat.c | 85 +++++++++++++++++++++++---------------- 1 file changed, 50 insertions(+), 35 deletions(-)
tools/perf/builtin-stat.c +50 -35
@@ -371,6 +371,16 @@ static void process_interval(void)print_counters (&rs,0 ,NULL ); } +static bool print_interval_and_stop (struct perf_stat_config *config,int *times) +{ +if (config->interval) { +process_interval (); +if (interval_count && !(--(*times))) +return true; + } +return false; +} +static void enable_counters (void ) {if (stat_config.initial_delay)
@@ -436,6 +446,42 @@ static bool is_target_alive(struct target *_target,return false; } +static int handle_events (pid_t pid,struct perf_stat_config *config) +{ +pid_t child =0 ; +bool res, stop = false; +struct timespec time_to_sleep; +int sleep_time, status =0 , times = config->times; + +if (config->interval) + sleep_time = config->interval; +else if (config->timeout) + sleep_time = config->timeout; +else + sleep_time =1000 ; + + time_to_sleep.tv_sec = sleep_time /MSEC_PER_SEC ; + time_to_sleep.tv_nsec = (sleep_time %MSEC_PER_SEC ) *NSEC_PER_MSEC ; + +do { +if (pid !=-1 ) + child =waitpid (pid, &status,WNOHANG ); +if (child || stop || done) +break ; +nanosleep (&time_to_sleep,NULL ); +if (pid ==-1 ) + stop = !is_target_alive (&target, evsel_list->core.threads); +if (config->timeout) { + stop = !stop ? true : stop; + }else { + res =print_interval_and_stop (config, ×); + stop = !stop ? res : stop; + } + }while (1 ); + +return status; +} +enum counter_recovery {COUNTER_SKIP ,COUNTER_RETRY ,
@@ -494,12 +540,10 @@ static enum counter_recovery stat_handle_error(struct evsel *counter)static int __run_perf_stat (int argc,const char **argv,int run_idx) {int interval = stat_config.interval; -int times = stat_config.times;int timeout = stat_config.timeout;char msg[BUFSIZ ];unsigned long long t0, t1;struct evsel *counter; -struct timespec ts;size_t l;int status =0 ;const bool forks = (argc >0 );
@@ -508,17 +552,6 @@ static int __run_perf_stat(int argc, const char **argv, int run_idx)int i, cpu;bool second_pass = false; -if (interval) { - ts.tv_sec = interval /USEC_PER_MSEC ; - ts.tv_nsec = (interval %USEC_PER_MSEC ) *NSEC_PER_MSEC ; - }else if (timeout) { - ts.tv_sec = timeout /USEC_PER_MSEC ; - ts.tv_nsec = (timeout %USEC_PER_MSEC ) *NSEC_PER_MSEC ; - }else { - ts.tv_sec =1 ; - ts.tv_nsec =0 ; - } -if (forks) {if (perf_evlist__prepare_workload (evsel_list, &target, argv, is_pipe, workload_exec_failed_signal) <0 ) {
@@ -675,16 +708,9 @@ static int __run_perf_stat(int argc, const char **argv, int run_idx)perf_evlist__start_workload (evsel_list);enable_counters (); -if (interval || timeout) { -while (!waitpid (child_pid, &status,WNOHANG )) { -nanosleep (&ts,NULL ); -if (timeout) -break ; -process_interval (); -if (interval_count && !(--times)) -break ; - } - } +if (interval || timeout) +handle_events (child_pid, &stat_config); +if (child_pid !=-1 ) {if (timeout)kill (child_pid,SIGTERM );
@@ -701,18 +727,7 @@ static int __run_perf_stat(int argc, const char **argv, int run_idx)psignal (WTERMSIG (status), argv[0 ]); }else {enable_counters (); -while (!done) { -nanosleep (&ts,NULL ); -if (!is_target_alive (&target, evsel_list->core.threads)) -break ; -if (timeout) -break ; -if (interval) { -process_interval (); -if (interval_count && !(--times)) -break ; - } - } +handle_events (-1 , &stat_config); }disable_counters ();
Signature · 1 line
2.24.1
PATCH v2 4/11 perf stat: extend -D,--delay option with -1 value
In reply to Alexey BudankovExtend -D,--delay option with -1 value to start monitoring with events disabled to be enabled later by enable command provided via control file descriptor.
--- tools/perf/builtin-stat.c | 18 ++++++++++++++---- tools/perf/util/evlist.h | 3 +++ tools/perf/util/stat.h | 2 +- 3 files changed, 18 insertions(+), 5 deletions(-)
tools/perf/builtin-stat.c +14 -4
@@ -383,16 +383,26 @@ static bool print_interval_and_stop(struct perf_stat_config *config, int *times)static void enable_counters (void ) { -if (stat_config.initial_delay) +if (stat_config.initial_delay <0 ) { +pr_info (PERF_EVLIST__DISABLED_MSG ); +return ; + } + +if (stat_config.initial_delay >0 ) { +pr_info (PERF_EVLIST__DISABLED_MSG );usleep (stat_config.initial_delay *USEC_PER_MSEC ); + }/* * We need to enable counters only if: * - we don't have tracee (attaching to task or cpu) * - we have initial delay configured */ -if (!target__none (&target) || stat_config.initial_delay) +if (!target__none (&target) || stat_config.initial_delay) {evlist__enable (evsel_list); +if (stat_config.initial_delay >0 ) +pr_info (PERF_EVLIST__ENABLED_MSG ); + } }static void disable_counters (void )
@@ -929,8 +939,8 @@ static struct option stat_options[] = { "aggregate counts per thread", AGGR_THREAD),OPT_SET_UINT (0 ,"per-node" , &stat_config.aggr_mode,"aggregate counts per numa node" ,AGGR_NODE ), -OPT_UINTEGER ('D' ,"delay" , &stat_config.initial_delay, -"ms to wait before starting measurement after program start" ), +OPT_INTEGER ('D' ,"delay" , &stat_config.initial_delay, +"ms to wait before starting measurement after program start (-1: start with events disabled" ),OPT_CALLBACK_NOOPT (0 ,"metric-only" , &stat_config.metric_only,NULL ,"Only print computed metrics. No raw values" , enable_metric_only), OPT_BOOLEAN(0, "topdown", &topdown_run,
tools/perf/util/evlist.h +3 -0
@@ -366,6 +366,9 @@ enum evlist_ctl_cmd { CTL_CMD_ACK = 'a' }; +#define PERF_EVLIST__ENABLED_MSG "Events enabled\n" +#define PERF_EVLIST__DISABLED_MSG "Events disabled\n" +int perf_evlist__initialize_ctlfd (struct evlist *evlist,int ctl_fd,int ctl_fd_ack);int perf_evlist__finalize_ctlfd (struct evlist *evlist);int perf_evlist__ctlfd_process (struct evlist *evlist,enum evlist_ctl_cmd *cmd);
tools/perf/util/stat.h +1 -1
@@ -113,7 +113,7 @@ struct perf_stat_config {FILE *output;unsigned int interval;unsigned int timeout; -unsigned int initial_delay; +int initial_delay;unsigned int unit_width;unsigned int metric_only_len;int times;
Signature · 1 line
2.24.1
Re: [PATCH v2 04/11] perf stat: extend -D,--delay option with -1 value
Em Wed, May 06, 2020 at 09:20:54PM +0300, Alexey Budankov escreveu:
Quoted 10 lines
Extend -D,--delay option with -1 value to start monitoring with events disabled to be enabled later by enable command provided via control file descriptor.
Signed-off-by Alexey Budankov <[email protected]>--- tools/perf/builtin-stat.c | 18 ++++++++++++++---- tools/perf/util/evlist.h | 3 +++ tools/perf/util/stat.h | 2 +-
You forgot to update tools/perf/Documentation/perf-stat.h
Quoted 77 lines
3 files changed, 18 insertions(+), 5 deletions(-)tools/perf/builtin-stat.c +14
-4@@ -383,16 +383,26 @@ static bool print_interval_and_stop(struct perf_stat_config *config, int *times)static void enable_counters (void ) { -if (stat_config.initial_delay) +if (stat_config.initial_delay <0 ) { +pr_info (PERF_EVLIST__DISABLED_MSG ); +return ; + } + +if (stat_config.initial_delay >0 ) { +pr_info (PERF_EVLIST__DISABLED_MSG );usleep (stat_config.initial_delay *USEC_PER_MSEC ); + }/* * We need to enable counters only if: * - we don't have tracee (attaching to task or cpu) * - we have initial delay configured */ -if (!target__none (&target) || stat_config.initial_delay) +if (!target__none (&target) || stat_config.initial_delay) {evlist__enable (evsel_list); +if (stat_config.initial_delay >0 ) +pr_info (PERF_EVLIST__ENABLED_MSG ); + } }static void disable_counters (void )@@ -929,8 +939,8 @@ static struct option stat_options[] = { "aggregate counts per thread", AGGR_THREAD),OPT_SET_UINT (0 ,"per-node" , &stat_config.aggr_mode,"aggregate counts per numa node" ,AGGR_NODE ), -OPT_UINTEGER ('D' ,"delay" , &stat_config.initial_delay, -"ms to wait before starting measurement after program start" ), +OPT_INTEGER ('D' ,"delay" , &stat_config.initial_delay, +"ms to wait before starting measurement after program start (-1: start with events disabled" ),OPT_CALLBACK_NOOPT (0 ,"metric-only" , &stat_config.metric_only,NULL ,"Only print computed metrics. No raw values" , enable_metric_only), OPT_BOOLEAN(0, "topdown", &topdown_run,tools/perf/util/evlist.h +3
-0@@ -366,6 +366,9 @@ enum evlist_ctl_cmd { CTL_CMD_ACK = 'a' }; +#define PERF_EVLIST__ENABLED_MSG "Events enabled\n" +#define PERF_EVLIST__DISABLED_MSG "Events disabled\n" +int perf_evlist__initialize_ctlfd (struct evlist *evlist,int ctl_fd,int ctl_fd_ack);int perf_evlist__finalize_ctlfd (struct evlist *evlist);int perf_evlist__ctlfd_process (struct evlist *evlist,enum evlist_ctl_cmd *cmd);tools/perf/util/stat.h +1
-1@@ -113,7 +113,7 @@ struct perf_stat_config {FILE *output;unsigned int interval;unsigned int timeout; -unsigned int initial_delay; +int initial_delay;unsigned int unit_width;unsigned int metric_only_len;int times;Signature · 1 line
2.24.1
Signature · 1 line
- Arnaldo
Re: [PATCH v2 04/11] perf stat: extend -D,--delay option with -1 value
Quoted 6 lines · Arnaldo Carvalho de Melo
Em Wed, May 06, 2020 at 09:20:54PM +0300, Alexey Budankov escreveu:
Quoted 4 lines
Extend -D,--delay option with -1 value to start monitoring with events disabled to be enabled later by enable command provided via control file descriptor.
Quoted 7 lines
Quoted 5 lines
Signed-off-by Alexey Budankov <[email protected]>--- tools/perf/builtin-stat.c | 18 ++++++++++++++---- tools/perf/util/evlist.h | 3 +++ tools/perf/util/stat.h | 2 +-You forgot to update
tools/perf/Documentation/perf-stat.h
Good catch. Addressed in v3. Thanks!
~Alexey
PATCH v2 5/11 perf stat: implement control commands handling
In reply to Alexey BudankovImplement handling of 'enable' and 'disable' control commands coming from control file descriptor.
--- tools/perf/builtin-stat.c | 48 +++++++++++++++++++++++++++++---------- 1 file changed, 36 insertions(+), 12 deletions(-)
tools/perf/builtin-stat.c +36 -12
@@ -460,8 +460,9 @@ static int handle_events(pid_t pid, struct perf_stat_config *config) {pid_t child =0 ;bool res, stop = false; -struct timespec time_to_sleep; -int sleep_time, status =0 , times = config->times; +int time_to_sleep, sleep_time, status =0 , times = config->times; +enum evlist_ctl_cmd cmd =CTL_CMD_UNSUPPORTED ; +struct timespec time_start, time_stop, time_diff;if (config->interval) sleep_time = config->interval;
@@ -470,22 +471,45 @@ static int handle_events(pid_t pid, struct perf_stat_config *config)else sleep_time =1000 ; - time_to_sleep.tv_sec = sleep_time /MSEC_PER_SEC ; - time_to_sleep.tv_nsec = (sleep_time %MSEC_PER_SEC ) *NSEC_PER_MSEC ; + time_to_sleep = sleep_time;do {if (pid !=-1 ) child =waitpid (pid, &status,WNOHANG );if (child || stop || done)break ; -nanosleep (&time_to_sleep,NULL ); -if (pid ==-1 ) - stop = !is_target_alive (&target, evsel_list->core.threads); -if (config->timeout) { - stop = !stop ? true : stop; - }else { - res =print_interval_and_stop (config, ×); - stop = !stop ? res : stop; +clock_gettime (CLOCK_MONOTONIC , &time_start); +if (!(evlist__poll (evsel_list, time_to_sleep) >0 )) {/* poll timeout or EINTR */ +if (pid ==-1 ) + stop = !is_target_alive (&target, evsel_list->core.threads); +if (config->timeout) { + stop = !stop ? true : stop; + }else { + res =print_interval_and_stop (config, ×); + stop = !stop ? res : stop; + } + time_to_sleep = sleep_time; + }else {/* fd revent */ +if (perf_evlist__ctlfd_process (evsel_list, &cmd) >0 ) { +switch (cmd) { +case CTL_CMD_ENABLE : +pr_info (PERF_EVLIST__ENABLED_MSG ); + stop =print_interval_and_stop (config, ×); +break ; +case CTL_CMD_DISABLE : + stop =print_interval_and_stop (config, ×); +pr_info (PERF_EVLIST__DISABLED_MSG ); +break ; +case CTL_CMD_ACK : +case CTL_CMD_UNSUPPORTED : +default : +break ; + } + } +clock_gettime (CLOCK_MONOTONIC , &time_stop); +diff_timespec (&time_diff, &time_stop, &time_start); + time_to_sleep -= time_diff.tv_sec *MSEC_PER_SEC + + time_diff.tv_nsec /NSEC_PER_MSEC ; } }while (1 );
Signature · 1 line
2.24.1
PATCH v2 6/11 perf stat: introduce --ctl-fd[-ack] options
In reply to Alexey BudankovIntroduce --ctl-fd[-ack] options to pass open file descriptors numbers from command line.
--- tools/perf/builtin-stat.c | 10 ++++++++++ tools/perf/util/stat.h | 2 ++ 2 files changed, 12 insertions(+)
tools/perf/builtin-stat.c +10 -0
@@ -187,6 +187,8 @@ static struct perf_stat_config stat_config = { .metric_only_len = METRIC_ONLY_LEN, .walltime_nsecs_stats = &walltime_nsecs_stats, .big_num = true, + .ctl_fd = -1, + .ctl_fd_ack = -1 }; static inline void diff_timespec(struct timespec *r, struct timespec *a,
@@ -984,6 +986,10 @@ static struct option stat_options[] = {"Use with 'percore' event qualifier to show the event " "counts of one hardware thread by sum up total hardware " "threads of same physical core"), +OPT_INTEGER (0 ,"ctl-fd" , &stat_config.ctl_fd, +"Listen on fd descriptor for command to control measurement ('e': enable events, 'd': disable events)" ), +OPT_INTEGER (0 ,"ctl-fd-ack" , &stat_config.ctl_fd_ack, +"Send control command completion ('a') to fd ack descriptor" ), OPT_END() };
@@ -2180,6 +2186,8 @@ int cmd_stat(int argc, const char **argv)signal (SIGALRM , skip_signal);signal (SIGABRT , skip_signal); +perf_evlist__initialize_ctlfd (evsel_list, stat_config.ctl_fd, stat_config.ctl_fd_ack); + status =0 ;for (run_idx =0 ; forever || run_idx < stat_config.run_count; run_idx++) {if (stat_config.run_count !=1 && verbose >0 )
@@ -2199,6 +2207,8 @@ int cmd_stat(int argc, const char **argv)if (!forever && status !=-1 && !interval)print_counters (NULL , argc, argv); +perf_evlist__finalize_ctlfd (evsel_list); +if (STAT_RECORD ) {/* * We synthesize the kernel mmap record just so that older tools
tools/perf/util/stat.h +2 -0
@@ -130,6 +130,8 @@ struct perf_stat_config {struct perf_cpu_map *cpus_aggr_map;u64 *walltime_run;struct rblist metric_events; +int ctl_fd; +int ctl_fd_ack; };void update_stats (struct stats *stats,u64 val);
Signature · 1 line
2.24.1
PATCH v2 7/11 perf docs: extend stat mode docs with info on --ctl-fd[-ack] options
In reply to Alexey BudankovExtend perf-stat.txt file with --ctl-fd[-ack] options description. Document possible usage model introduced by --ctl-fd[-ack] options by providing example bash shell script.
--- tools/perf/Documentation/perf-stat.txt | 40 ++++++++++++++++++++++++++ 1 file changed, 40 insertions(+)
tools/perf/Documentation/perf-stat.txt +40 -0
@@ -164,6 +164,46 @@ with it. --append may be used here. Examples: 3>results perf stat --log-fd 3 -- $cmd 3>>results perf stat --log-fd 3 --append -- $cmd +--ctl-fd:: +--ctl-fd-ack:: + +Listen on ctl-fd descriptor for command to control measurement ('e': enable events, +'d': disable events). Optionally send control command completion ('a') to fd-ack +descriptor to synchronize with the controlling process. Example of bash shell script +to enable and disable events during measurements: + +#!/bin/bash + +ctl_dir=/tmp/ + +ctl_fifo=${ctl_dir}perf_ctl.fifo +test -p ${ctl_fifo} && unlink ${ctl_fifo} +mkfifo ${ctl_fifo} +exec {ctl_fd}<>${ctl_fifo} + +ctl_ack_fifo=${ctl_dir}perf_ctl_ack.fifo +test -p ${ctl_ack_fifo} && unlink ${ctl_ack_fifo} +mkfifo ${ctl_ack_fifo} +exec {ctl_fd_ack}<>${ctl_ack_fifo} + +perf stat -D -1 -e cpu-cycles -a -I 1000 \ + --ctl-fd ${ctl_fd} --ctl-fd-ack ${ctl_fd_ack} \ + -- sleep 30 & +perf_pid=$! + +sleep 5 && echo 'e' >&${ctl_fd} && read -u ${ctl_fd_ack} e1 && echo "enabled(${e1})" +sleep 10 && echo 'd' >&${ctl_fd} && read -u ${ctl_fd_ack} d1 && echo "disabled(${d1})" + +exec {ctl_fd_ack}>&- +unlink ${ctl_ack_fifo} + +exec {ctl_fd}>&- +unlink ${ctl_fifo} + +wait -n ${perf_pid} +exit $? + + --pre:: --post:: Pre and post measurement hooks, e.g.:
Signature · 1 line
2.24.1
Re: [PATCH v2 07/11] perf docs: extend stat mode docs with info on --ctl-fd[-ack] options
Em Wed, May 06, 2020 at 09:22:55PM +0300, Alexey Budankov escreveu:
Quoted 4 lines
Extend perf-stat.txt file with --ctl-fd[-ack] options description. Document possible usage model introduced by --ctl-fd[-ack] options by providing example bash shell script.
Please update the docs in the same patch that introduces the options being documented.
- Arnaldo
Quoted 60 lines
Signed-off-by Alexey Budankov <[email protected]>--- tools/perf/Documentation/perf-stat.txt | 40 ++++++++++++++++++++++++++ 1 file changed, 40 insertions(+)tools/perf/Documentation/perf-stat.txt +40
-0@@ -164,6 +164,46 @@ with it. --append may be used here. Examples: 3>results perf stat --log-fd 3 -- $cmd 3>>results perf stat --log-fd 3 --append -- $cmd +--ctl-fd:: +--ctl-fd-ack:: + +Listen on ctl-fd descriptor for command to control measurement ('e': enable events, +'d': disable events). Optionally send control command completion ('a') to fd-ack +descriptor to synchronize with the controlling process. Example of bash shell script +to enable and disable events during measurements: + +#!/bin/bash + +ctl_dir=/tmp/ + +ctl_fifo=${ctl_dir}perf_ctl.fifo +test -p ${ctl_fifo} && unlink ${ctl_fifo} +mkfifo ${ctl_fifo} +exec {ctl_fd}<>${ctl_fifo} + +ctl_ack_fifo=${ctl_dir}perf_ctl_ack.fifo +test -p ${ctl_ack_fifo} && unlink ${ctl_ack_fifo} +mkfifo ${ctl_ack_fifo} +exec {ctl_fd_ack}<>${ctl_ack_fifo} + +perf stat -D -1 -e cpu-cycles -a -I 1000 \ + --ctl-fd ${ctl_fd} --ctl-fd-ack ${ctl_fd_ack} \ + -- sleep 30 & +perf_pid=$! + +sleep 5 && echo 'e' >&${ctl_fd} && read -u ${ctl_fd_ack} e1 && echo "enabled(${e1})" +sleep 10 && echo 'd' >&${ctl_fd} && read -u ${ctl_fd_ack} d1 && echo "disabled(${d1})" + +exec {ctl_fd_ack}>&- +unlink ${ctl_ack_fifo} + +exec {ctl_fd}>&- +unlink ${ctl_fifo} + +wait -n ${perf_pid} +exit $? + + --pre:: --post:: Pre and post measurement hooks, e.g.:Signature · 1 line
2.24.1
Signature · 1 line
- Arnaldo
Re: [PATCH v2 07/11] perf docs: extend stat mode docs with info on --ctl-fd[-ack] options
Quoted 8 lines · Arnaldo Carvalho de Melo
Em Wed, May 06, 2020 at 09:22:55PM +0300, Alexey Budankov escreveu:
Quoted 4 lines
Extend perf-stat.txt file with --ctl-fd[-ack] options description. Document possible usage model introduced by --ctl-fd[-ack] options by providing example bash shell script.
Please update the docs in the same patch that introduces the options being documented.
Addressed in v3.
~Alexey
PATCH v2 8/11 perf record: extend -D,--delay option with -1 value
In reply to Alexey BudankovExtend -D,--delay option with -1 to start collection with events disabled to be enbled later by enable command provided via control file descriptor.
--- tools/perf/builtin-record.c | 12 ++++++++---- tools/perf/builtin-trace.c | 2 +- tools/perf/util/record.h | 2 +- 3 files changed, 10 insertions(+), 6 deletions(-)
tools/perf/builtin-record.c +8 -4
@@ -1704,8 +1704,12 @@ static int __cmd_record(struct record *rec, int argc, const char **argv) }if (opts->initial_delay) { -usleep (opts->initial_delay *USEC_PER_MSEC ); -evlist__enable (rec->evlist); +pr_info (PERF_EVLIST__DISABLED_MSG ); +if (opts->initial_delay >0 ) { +usleep (opts->initial_delay *USEC_PER_MSEC ); +evlist__enable (rec->evlist); +pr_info (PERF_EVLIST__ENABLED_MSG ); + } }trigger_ready (&auxtrace_snapshot_trigger);
@@ -2413,8 +2417,8 @@ static struct option __record_options[] = {OPT_CALLBACK ('G' ,"cgroup" , &record.evlist,"name" ,"monitor event in cgroup name only" , parse_cgroups), -OPT_UINTEGER ('D' ,"delay" , &record.opts.initial_delay, -"ms to wait before starting measurement after program start" ), +OPT_INTEGER ('D' ,"delay" , &record.opts.initial_delay, +"ms to wait before starting measurement after program start (-1: start with events disabled )" ),OPT_BOOLEAN (0 ,"kcore" , &record.opts.kcore,"copy /proc/kcore" ),OPT_STRING ('u' ,"uid" , &record.opts.target.uid_str,"user" , "user to profile"),
tools/perf/builtin-trace.c +1 -1
@@ -4773,7 +4773,7 @@ int cmd_trace(int argc, const char **argv) "per thread proc mmap processing timeout in ms"),OPT_CALLBACK ('G' ,"cgroup" , &trace,"name" ,"monitor event in cgroup name only" , trace__parse_cgroups), -OPT_UINTEGER ('D' ,"delay" , &trace.opts.initial_delay, +OPT_INTEGER ('D' ,"delay" , &trace.opts.initial_delay,"ms to wait before starting measurement after program " "start" ), OPTS_EVSWITCH(&trace.evswitch),
tools/perf/util/record.h +1 -1
@@ -60,7 +60,7 @@ struct record_opts {const char *auxtrace_snapshot_opts;const char *auxtrace_sample_opts;bool sample_transaction; -unsigned initial_delay; +int initial_delay;bool use_clockid;clockid_t clockid;u64 clockid_res_ns;
Signature · 1 line
2.24.1
PATCH v2 9/11 perf record: implement control commands handling
In reply to Alexey BudankovImplement handling of 'enable' and 'disable' control commands coming from control file descriptor.
--- tools/perf/builtin-record.c | 18 +++++++++++++++++- 1 file changed, 17 insertions(+), 1 deletion(-)
tools/perf/builtin-record.c +17 -1
@@ -1496,6 +1496,7 @@ static int __cmd_record(struct record *rec, int argc, const char **argv)bool disabled = false, draining = false;int fd;float ratio =0 ; +enum evlist_ctl_cmd cmd =CTL_CMD_UNSUPPORTED ;atexit (record__sig_exit);signal (SIGCHLD , sig_handler);
@@ -1793,8 +1794,23 @@ static int __cmd_record(struct record *rec, int argc, const char **argv)* Propagate error, only if there's any. Ignore positive * number of returned events and interrupt error. */ -if (err >0 || (err <0 && errno ==EINTR )) +if (err >0 || (err <0 && errno ==EINTR )) { err =0 ; +if (perf_evlist__ctlfd_process (rec->evlist, &cmd) >0 ) { +switch (cmd) { +case CTL_CMD_ENABLE : +pr_info (PERF_EVLIST__ENABLED_MSG ); +break ; +case CTL_CMD_DISABLE : +pr_info (PERF_EVLIST__DISABLED_MSG ); +break ; +case CTL_CMD_ACK : +case CTL_CMD_UNSUPPORTED : +default : +break ; + } + } + } waking++;if (evlist__filter_pollfd (rec->evlist,POLLERR |POLLHUP ) ==0 )
Signature · 1 line
2.24.1
Re: [PATCH v2 09/11] perf record: implement control commands handling
Em Wed, May 06, 2020 at 09:29:05PM +0300, Alexey Budankov escreveu:
Quoted 28 lines
Implement handling of 'enable' and 'disable' control commands coming from control file descriptor.
Signed-off-by Alexey Budankov <[email protected]>--- tools/perf/builtin-record.c | 18 +++++++++++++++++- 1 file changed, 17 insertions(+), 1 deletion(-)tools/perf/builtin-record.c +2
-1@@ -1496,6 +1496,7 @@ static int __cmd_record(struct record *rec, int argc, const char **argv)bool disabled = false, draining = false;int fd;float ratio =0 ; +enum evlist_ctl_cmd cmd =CTL_CMD_UNSUPPORTED ;atexit (record__sig_exit);signal (SIGCHLD , sig_handler);@@ -1793,8 +1794,23 @@ static int __cmd_record(struct record *rec, int argc, const char **argv)* Propagate error, only if there's any. Ignore positive * number of returned events and interrupt error. */ -if (err >0 || (err <0 && errno ==EINTR )) +if (err >0 || (err <0 && errno ==EINTR )) { err =0 ;
tools/perf/builtin-record.c static int __cmd_record(struct record *rec, int argc, const char **argv) · line 1808
Quoted 10 lines
tools/perf/builtin-record.c +10
-0+if (perf_evlist__ctlfd_process (rec->evlist, &cmd) >0 ) { +switch (cmd) { +case CTL_CMD_ENABLE : +pr_info (PERF_EVLIST__ENABLED_MSG ); +break ; +case CTL_CMD_DISABLE : +pr_info (PERF_EVLIST__DISABLED_MSG ); +break ; +case CTL_CMD_ACK : +case CTL_CMD_UNSUPPORTED :
Shouldn't we have a pr_debug() or even pr_err() for the unsupported one?
Quoted 12 lines
tools/perf/builtin-record.c +5
-0+default : +break ; + } + } + } waking++;if (evlist__filter_pollfd (rec->evlist,POLLERR |POLLHUP ) ==0 )Signature · 1 line
2.24.1
Signature · 1 line
- Arnaldo
Re: [PATCH v2 09/11] perf record: implement control commands handling
Quoted 33 lines · Arnaldo Carvalho de Melo
Em Wed, May 06, 2020 at 09:29:05PM +0300, Alexey Budankov escreveu:
Quoted 32 lines
Implement handling of 'enable' and 'disable' control commands coming from control file descriptor.
Signed-off-by Alexey Budankov <[email protected]>--- tools/perf/builtin-record.c | 18 +++++++++++++++++- 1 file changed, 17 insertions(+), 1 deletion(-)tools/perf/builtin-record.c +6
-1@@ -1496,6 +1496,7 @@ static int __cmd_record(struct record *rec, int argc, const char **argv)bool disabled = false, draining = false;int fd;float ratio =0 ; +enum evlist_ctl_cmd cmd =CTL_CMD_UNSUPPORTED ;atexit (record__sig_exit);signal (SIGCHLD , sig_handler);@@ -1793,8 +1794,23 @@ static int __cmd_record(struct record *rec, int argc, const char **argv)* Propagate error, only if there's any. Ignore positive * number of returned events and interrupt error. */ -if (err >0 || (err <0 && errno ==EINTR )) +if (err >0 || (err <0 && errno ==EINTR )) { err =0 ; +if (perf_evlist__ctlfd_process (rec->evlist, &cmd) >0 ) { +switch (cmd) { +case CTL_CMD_ENABLE : +pr_info (PERF_EVLIST__ENABLED_MSG );
Quoted 8 lines
tools/perf/builtin-record.c static int __cmd_record(struct record *rec, int argc, const char **argv) · line 1808
Quoted 6 lines
tools/perf/builtin-record.c +6
-0+break ; +case CTL_CMD_DISABLE : +pr_info (PERF_EVLIST__DISABLED_MSG ); +break ; +case CTL_CMD_ACK : +case CTL_CMD_UNSUPPORTED :Shouldn't we have a
pr_debug()or evenpr_err()for the unsupported one?
It already exists on lower level, at perf_evlist__ctlfd_process() (see patch 02/11).
~Alexey
Re: [PATCH v2 09/11] perf record: implement control commands handling
Em Thu, May 07, 2020 at 11:58:45AM +0300, Alexey Budankov escreveu:
Quoted 44 lines · Arnaldo Carvalho de Melo
Quoted 33 lines
Em Wed, May 06, 2020 at 09:29:05PM +0300, Alexey Budankov escreveu:
Quoted 32 lines
Implement handling of 'enable' and 'disable' control commands coming from control file descriptor.
Signed-off-by Alexey Budankov <[email protected]>--- tools/perf/builtin-record.c | 18 +++++++++++++++++- 1 file changed, 17 insertions(+), 1 deletion(-)tools/perf/builtin-record.c +6
-1@@ -1496,6 +1496,7 @@ static int __cmd_record(struct record *rec, int argc, const char **argv)bool disabled = false, draining = false;int fd;float ratio =0 ; +enum evlist_ctl_cmd cmd =CTL_CMD_UNSUPPORTED ;atexit (record__sig_exit);signal (SIGCHLD , sig_handler);@@ -1793,8 +1794,23 @@ static int __cmd_record(struct record *rec, int argc, const char **argv)* Propagate error, only if there's any. Ignore positive * number of returned events and interrupt error. */ -if (err >0 || (err <0 && errno ==EINTR )) +if (err >0 || (err <0 && errno ==EINTR )) { err =0 ; +if (perf_evlist__ctlfd_process (rec->evlist, &cmd) >0 ) { +switch (cmd) { +case CTL_CMD_ENABLE : +pr_info (PERF_EVLIST__ENABLED_MSG );Quoted 8 lines
tools/perf/builtin-record.c static int __cmd_record(struct record *rec, int argc, const char **argv) · line 1808
Quoted 6 lines
tools/perf/builtin-record.c +6
-0+break ; +case CTL_CMD_DISABLE : +pr_info (PERF_EVLIST__DISABLED_MSG ); +break ; +case CTL_CMD_ACK : +case CTL_CMD_UNSUPPORTED :Shouldn't we have a
pr_debug()or evenpr_err()for the unsupported one?
Quoted 1 line
It already exists on lower level, at
perf_evlist__ctlfd_process()(see patch 02/11).
oic
Quoted 1 line
~Alexey
Signature · 1 line
- Arnaldo
PATCH v2 10/11 perf record: introduce --ctl-fd[-ack] options
In reply to Alexey BudankovIntroduce --ctl-fd[-ack] options to pass open file descriptors numbers from command line.
--- tools/perf/builtin-record.c | 9 +++++++++ tools/perf/util/record.h | 2 ++ 2 files changed, 11 insertions(+)
tools/perf/builtin-record.c +9 -0
@@ -1704,6 +1704,8 @@ static int __cmd_record(struct record *rec, int argc, const char **argv)perf_evlist__start_workload (rec->evlist); } +perf_evlist__initialize_ctlfd (rec->evlist, opts->ctl_fd, opts->ctl_fd_ack); +if (opts->initial_delay) {pr_info (PERF_EVLIST__DISABLED_MSG );if (opts->initial_delay >0 ) {
@@ -1850,6 +1852,7 @@ static int __cmd_record(struct record *rec, int argc, const char **argv)record__synthesize_workload (rec, true); out_child: +perf_evlist__finalize_ctlfd (rec->evlist);record__mmap_read_all (rec, true);record__aio_mmap_read_sync (rec);
@@ -2331,6 +2334,8 @@ static struct record record = { }, .mmap_flush = MMAP_FLUSH_DEFAULT, .nr_threads_synthesize = 1, + .ctl_fd = -1, + .ctl_fd_ack = -1, }, .tool = { .sample = process_sample_event,
@@ -2526,6 +2531,10 @@ static struct option __record_options[] = {OPT_UINTEGER (0 ,"num-thread-synthesize" , &record.opts.nr_threads_synthesize,"number of threads to run for event synthesis" ), +OPT_INTEGER (0 ,"ctl-fd" , &record.opts.ctl_fd, +"Listen on fd descriptor for command to control measurement ('e': enable events, 'd': disable events)" ), +OPT_INTEGER (0 ,"ctl-fd-ack" , &record.opts.ctl_fd_ack, +"Send control command completion ('a') to fd ack descriptor" ), OPT_END() };
tools/perf/util/record.h +2 -0
@@ -69,6 +69,8 @@ struct record_opts {int mmap_flush;unsigned int comp_level;unsigned int nr_threads_synthesize; +int ctl_fd; +int ctl_fd_ack; };extern const char *const *record_usage;
Signature · 1 line
2.24.1
PATCH v2 11/11 perf docs: extend record mode docs with info on --ctl-fd[-ack] options
In reply to Alexey BudankovExtend perf-record.txt file with --ctl-fd[-ack] options description. Document possible usage model introduced by --ctl-fd[-ack] options by providing example bash shell script.
--- tools/perf/Documentation/perf-record.txt | 39 ++++++++++++++++++++++++ 1 file changed, 39 insertions(+)
tools/perf/Documentation/perf-record.txt +39 -0
@@ -613,6 +613,45 @@ appended unit character - B/K/M/G The number of threads to run when synthesizing events for existing processes. By default, the number of threads equals 1. +--ctl-fd:: +--ctl-fd-ack:: +Listen on ctl-fd descriptor for command to control measurement ('e': enable events, +'d': disable events. Optionally send control command completion ('a') to fd-ack +descriptor to synchronize with the controlling process. Example of bash shell script +to enable and disable events during measurements: + +#!/bin/bash + +ctl_dir=/tmp/ + +ctl_fifo=${ctl_dir}perf_ctl.fifo +test -p ${ctl_fifo} && unlink ${ctl_fifo} +mkfifo ${ctl_fifo} +exec {ctl_fd}<>${ctl_fifo} + +ctl_ack_fifo=${ctl_dir}perf_ctl_ack.fifo +test -p ${ctl_ack_fifo} && unlink ${ctl_ack_fifo} +mkfifo ${ctl_ack_fifo} +exec {ctl_fd_ack}<>${ctl_ack_fifo} + +perf record -D -1 -e cpu-cycles -a \ + --ctl-fd ${ctl_fd} --ctl-fd-ack ${ctl_fd_ack} \ + -- sleep 30 & +perf_pid=$! + +sleep 5 && echo 'e' >&${ctl_fd} && read -u ${ctl_fd_ack} e1 && echo "resumed(${e1})" +sleep 10 && echo 'd' >&${ctl_fd} && read -u ${ctl_fd_ack} d1 && echo "paused(${d1})" + +exec {ctl_fd_ack}>&- +unlink ${ctl_ack_fifo} + +exec {ctl_fd}>&- +unlink ${ctl_fifo} + +wait -n ${perf_pid} +exit $? + + SEE ALSO -------- linkperf:perf-stat[1], linkperf:perf-list[1], linkperf:perf-intel-pt[1]
Signature · 1 line
2.24.1
Re: [PATCH v2 11/11] perf docs: extend record mode docs with info on --ctl-fd[-ack] options
Em Wed, May 06, 2020 at 09:30:50PM +0300, Alexey Budankov escreveu:
Quoted 4 lines
Extend perf-record.txt file with --ctl-fd[-ack] options description. Document possible usage model introduced by --ctl-fd[-ack] options by providing example bash shell script.
Ditto, combine this patch with the one that introduces these options.
Quoted 59 lines
Signed-off-by Alexey Budankov <[email protected]>--- tools/perf/Documentation/perf-record.txt | 39 ++++++++++++++++++++++++ 1 file changed, 39 insertions(+)tools/perf/Documentation/perf-record.txt +39
-0@@ -613,6 +613,45 @@ appended unit character - B/K/M/G The number of threads to run when synthesizing events for existing processes. By default, the number of threads equals 1. +--ctl-fd:: +--ctl-fd-ack:: +Listen on ctl-fd descriptor for command to control measurement ('e': enable events, +'d': disable events. Optionally send control command completion ('a') to fd-ack +descriptor to synchronize with the controlling process. Example of bash shell script +to enable and disable events during measurements: + +#!/bin/bash + +ctl_dir=/tmp/ + +ctl_fifo=${ctl_dir}perf_ctl.fifo +test -p ${ctl_fifo} && unlink ${ctl_fifo} +mkfifo ${ctl_fifo} +exec {ctl_fd}<>${ctl_fifo} + +ctl_ack_fifo=${ctl_dir}perf_ctl_ack.fifo +test -p ${ctl_ack_fifo} && unlink ${ctl_ack_fifo} +mkfifo ${ctl_ack_fifo} +exec {ctl_fd_ack}<>${ctl_ack_fifo} + +perf record -D -1 -e cpu-cycles -a \ + --ctl-fd ${ctl_fd} --ctl-fd-ack ${ctl_fd_ack} \ + -- sleep 30 & +perf_pid=$! + +sleep 5 && echo 'e' >&${ctl_fd} && read -u ${ctl_fd_ack} e1 && echo "resumed(${e1})" +sleep 10 && echo 'd' >&${ctl_fd} && read -u ${ctl_fd_ack} d1 && echo "paused(${d1})" + +exec {ctl_fd_ack}>&- +unlink ${ctl_ack_fifo} + +exec {ctl_fd}>&- +unlink ${ctl_fifo} + +wait -n ${perf_pid} +exit $? + + SEE ALSO -------- linkperf:perf-stat[1], linkperf:perf-list[1], linkperf:perf-intel-pt[1]Signature · 1 line
2.24.1
Signature · 1 line
- Arnaldo
Re: [PATCH v2 11/11] perf docs: extend record mode docs with info on --ctl-fd[-ack] options
Quoted 7 lines · Arnaldo Carvalho de Melo
Em Wed, May 06, 2020 at 09:30:50PM +0300, Alexey Budankov escreveu:
Quoted 4 lines
Extend perf-record.txt file with --ctl-fd[-ack] options description. Document possible usage model introduced by --ctl-fd[-ack] options by providing example bash shell script.
Ditto, combine this patch with the one that introduces these options.
Addressed in v3.
~Alexey