Messages in this thread |  | | | Date | Fri, 18 Jan 2008 15:20:21 -0800 | | From | Arjan van de Ven <> | | Subject | Re: [patch 2/3] Latencytop instrumentations part 1 | |
Frank Ch. Eigler wrote:
> Hi -
>
> On Fri, Jan 18, 2008 at 02:33:34PM -0800, Arjan van de Ven wrote:
>> [...]
>>> Can you suggest of some reason why all this instrumentation could
>>> not be in the form of standard markers (perhaps conditionally
>>> compiled out if necessary)?
>> sure. Every instrumentation you see is of the nested kind (since the lowest level
>> of nesting is already automatic via wchan).
>> If markers can provide me the following semantics, I'd be MORE than happy to use markers:
>> [...]
>> If markers can provide that semantics ... you sold me.
>
> Further to what acme said, markers are semantics-free. Callback
> functions that implement your entry & exit semantics can be attached
> at run time, at your pleasure. (So can systemtap probes, for that
> matter.) The main difference would be that these callback functions
> would have manage the per-thread LIFO data structures themselves,
> instead of allocating backpointers on the kernel stack. (Bonus marks
> for not modifying task_struct. :-)
modifying task struct to have storage space is no big deal...
|  |