Skip to contentSkip to conversation

perf_event: introduce 'inject' event and get HZ

PATCH Stijn Devriendt 2 participants 2 messages Dec 29, 2009

lore
Share & tools
lore
#0 lore raw
Quoted 2 lines

'inject' event is a very useful feature and it's suggested by Ingo [ See http://lkml.org/lkml/2009/12/28/31 ]

Is it possible to, instead of injecting a specific inject_event enum, pass a type/config pair (+task/cpu?) to inject the value of that specific event? Then the HZ could have it's own perf_event type.

This integrates better with what I'm struggling with right now:
counters that are not ever incrementing but rather indicate a discrete value inside the kernel that might change both up and down. My dynticks knowledge is rather limited but if HZ might vary then this approach is definately worth it.

Regards,
Stijn

#1 lore raw
Quoted 7 lines · Stijn Devriendt
Quoted 2 lines

'inject' event is a very useful feature and it's suggested by Ingo [ See http://lkml.org/lkml/2009/12/28/31 ]

Is it possible to, instead of injecting a specific inject_event enum, pass a type/config pair (+task/cpu?) to inject the value of that specific event? Then the HZ could have it's own perf_event type.

Until now, 'inject' event just trigger parameter/query events, if you need sample cpu/task, you can use other event.

Quoted 6 lines

This integrates better with what I'm struggling with right now:
counters that are not ever incrementing but rather indicate a discrete value inside the kernel that might change both up and down. My dynticks knowledge is rather limited but if HZ might vary then this approach is definately worth it.

dynticks not touch HZ at all as i know.

Thanks,
Xiao