This is the mail archive of the systemtap@sourceware.org mailing list for the systemtap project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [QUESTION] Is it safe to use lock in systemtap tapset?


Thank you for your reply!
Frank Ch. Eigler <fche@redhat.com> 于2018年11月17日周六 上午5:24写道:
>
>
> spacewanderlzx wrote:
>
> > I am searching for some information about lock usage in the systemtap
> > tapset.
>
> The general advice is to avoid explicit locking in embedded-C functions,
> since they may in principle run in any context, including atomic.  I am
> not sure how explicitly we spell this out in [man stap] or
> tapset/DEVGUIDE.
>
>
> > Is it safe to use lock in systemtap tapset?
> > Maybe the answer is different for tapset used in different context
> > (timers, kernel.function, process.function, etc.)?
> > Which kind of lock should I use in the tapset?
>
> It depends on the shared resource you are concerned about, in what
> probe/context you want to access them, whether you want to read or
> write, and whether you can tolerate corrupt data or not.
>
> Note: systemtap's own script-level global variables (and hidden runtime
> state) are all already protected automatically.  There's no need to
> initialize or worry about locking them.  (This is a unique feature in
> systemtap, in contrast to almost every other tracing/probing system.)
>
>
> - FChE


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]