Messages in this thread | | | Date | Fri, 1 Aug 2003 08:27:35 +0200 (CEST) | From | Ingo Molnar <> | Subject | Re: PATCH: Race in 2.6.0-test2 timer code |
| |
On Thu, 31 Jul 2003 linas@austin.ibm.com wrote:
> OK, I looked at removing run_all_timers, it doesn't seem too hard. > > I would need to: > -- add TIMER_SOFTIRQ to interrupts.h, > -- add open_softirq (run_timer_softirq) to timer.c init_timer() > -- move guts of run_local_timers() to run_timer_softirq() > -- remove bh locks in above, not yet sure about other locks > -- remove TIMER_BH everywhere. Or rather, remove it for those > arches that support cpu-local timer interupts (curently x86 & freinds, > soon hopefully ppc64, I attach it below, in case other arches want to > play with this). > > Is that right?
no. In 2.4 there are (and/or can be) all sorts of assumptions about TIMER_BH being serialized with other bh contexts (eg. the serial bh), that's why i added the TIMER_BH logic to the 2.4 timer-scalability patch. You cannot just remove TIMER_BH. The way we did it in 2.5 was to remove _all_ bhs and thus all assumptions about serialization. This is not an option for 2.4 in any case.
Ingo
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |