This patch set implements out of the box support of perf tools for persistent events. For this the kernel must provide necessary information about existing persistent events via sysfs to userland. Persistent events are provided by the kernel with readonly event buffers. To allow the independent usage of the event buffers by any process without limiting other processes, multiple users of a single event must be handled too.
The basic concept is to use a pmu as an event container for persistent events. The pmu registers events in sysfs and provides format and event information for the userland. The persistent event framework requires to add events to the pmu dynamically.
With the information in sysfs userland knows about how to setup the perf_event attribute of a persistent event. Since a persistent event always has the persistent flag set, a way is needed to express this in sysfs. A new syntax is used for this. With 'attr<num>:<mask>' any bit in the attribute structure may be set in a similar way as using 'config<num>', but <num> is an index that points to the u64 value to change within the attribute.
For persistent events the persistent flag (bit 23 of flag field in struct perf_event_attr) needs to be set which is expressed in sysfs with "attr5:23". E.g. the mce_record event is described in sysfs as follows:
/sys/bus/event_source/devices/persistent/events/mce_record:persistent,config=106 /sys/bus/event_source/devices/persistent/format/persistent:attr5:23
Note that perf tools need to support the 'attr<num>' syntax that is added in a separate patch set. With it we are able to run perf tool commands to read persistent events, e.g.:
# perf record -e persistent/mce_record/ sleep 10 # perf top -e persistent/mce_record/
In general the new syntax is flexible to describe with sysfs any event to be setup by perf tools.
First patches contain also fixes and reworks made after reviewing code. Could be applied separately.
Patches base on Boris' patches which I have rebased to latest tip/perf/core. All patches can be found here:
git://git.kernel.org/pub/scm/linux/kernel/git/rric/oprofile.git persistent-v2
I wonder if this patch set could be applied to a tip/perf topic branch? This would avoid reposting already reviewed patches.
Note: The perf tools patch set do not need to be reposted.
-Robert
Changes for V2: * Merged minor changes into Boris' patches * Included Boris' patches for review * Document attr<index> syntax in sysfs ABI * Adding cpu check to perf_get_persistent_event_fd() * Rebased to latest tip/perf/core
Borislav Petkov (4): perf, ring_buffer: Use same prefix perf: Add persistent events perf: Add persistent event facilities MCE: Enable persistent event
Robert Richter (10): perf, persistent: Rework struct pers_event_desc perf, persistent: Remove rb_put() perf, persistent: Introduce get_persistent_event() perf, persistent: Reworking perf_get_persistent_event_fd() perf, persistent: Protect event lists with mutex perf, persistent: Avoid adding identical events perf, persistent: Implementing a persistent pmu perf, persistent: Name each persistent event perf, persistent: Exposing persistent events using sysfs perf, persistent: Allow multiple users for an event
.../testing/sysfs-bus-event_source-devices-format | 43 ++- arch/x86/kernel/cpu/mcheck/mce.c | 19 ++ include/linux/perf_event.h | 9 +- include/uapi/linux/perf_event.h | 3 +- kernel/events/Makefile | 2 +- kernel/events/core.c | 56 ++-- kernel/events/internal.h | 3 + kernel/events/persistent.c | 320 +++++++++++++++++++++ kernel/events/ring_buffer.c | 7 +- 9 files changed, 419 insertions(+), 43 deletions(-) create mode 100644 kernel/events/persistent.c