On Fri, Aug 14, 2009 at 4:19 PM, Greg Freemyer<greg.freemyer@xxxxxxxxx> wrote: > On Fri, Aug 14, 2009 at 7:12 AM, Mohammed Gamal<m.gamal005@xxxxxxxxx> wrote: >> On Fri, Aug 14, 2009 at 11:25 AM, Mulyadi >> Santosa<mulyadi.santosa@xxxxxxxxx> wrote: >>> On Fri, Aug 14, 2009 at 12:15 PM, Michael >>> Blizek<michi1@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> wrote: >>>> Hi! >>>> >>>> On 00:08 Fri 14 Aug , Mohammed Gamal wrote: >>>> >>>> ... >>>> >>>>> As far as I understood, Michi's answer explains why and when the >>>>> kernel *can* get preempted, however what I really want to know is when >>>>> and where kernel preemption is *triggered*. Please correct me if I did >>>>> misunderstand anything. >>>> >>>> It is triggered by the timer interrupt. This is an interrupt which fires >>>> periodically on configureable intervals. It does not only preempt the kernel, >>>> but user space processes as well. If it fires, the kernel will enter the >>>> scheduler and decide what to run next. >>> >>> In addition, IIRC: >>> - every return from hard interrupt, either back to kernel space (if >>> kernel level preemption is enabled) or to user space triggers >>> preemption checking. >>> >>> - going back from kernel space to user space triggers preemption checking. >>> >>> >>> -- >>> regards, >>> >>> Mulyadi Santosa >>> Freelance Linux trainer >>> blog: the-hydra.blogspot.com >>> >> >> Thanks guys, >> Any pointers to where this is done in the kernel source? I'm a little >> bit lost there. >> >> Regards, >> Mohammed > > I have not looked into the details of CONFIG_PREEMPT before. > > When it is enabled, does it mean every interrupt that invokes the > scheduler can preempt a task, even if it is in kernel space. > > Or does it mean there is a API that allows a task to say "I'm high > priority, run me at the next interrupt even if the kernel is what was > interrupted?" > > If the latter, I believe the question is "What's the API to do that?" > > If the former, the question makes no sense. > > Greg > Hi All, Basically, my questions was to know at which points the kernel would normally call preempt_schedule() in order to trigger kernel preemption, a first look at LXR didn't help me find where that is performed. However after a second look at preempt.h, I found that preempt_schedule() gets called only in the preempt_check_resched() macro, which is in turn called by preempt_enable(). Once we enable preemption via preempt_enable(), we check if rescheduling is needed via preempt_check_resched() which calls preempt_schedule() if scheduling is needed. Take a look at preempt.h (lines 42-53) Regards, Mohammed -- To unsubscribe from this list: send an email with "unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx Please read the FAQ at http://kernelnewbies.org/FAQ
