| Message ID | 54DE52BC.3020303@us.ibm.com (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
On Fri, Feb 13, 2015 at 01:38:36PM -0600, Paul Clarke wrote: > implement arch_irq_work_has_interrupt() for powerpc > > Commit 9b01f5bf3 introduced a dependency on "IRQ work self-IPIs" for > full dynamic ticks to be enabled, by expecting architectures to > implement a suitable arch_irq_work_has_interrupt() routine. > > Several arches have implemented this routine, including x86 > (3010279f) and arm (09f6edd4), but powerpc was omitted. > > This patch implements this routine for powerpc. > > The symptom, at boot (on powerpc arch systems) with "nohz_full=<CPU > list>" is displayed: > NO_HZ: Can't run full dynticks because arch doesn't support irq > work self-IPIs > > after this patch: > NO_HZ: Full dynticks CPUs: <CPU list>. > > Tested against 3.19. > > CC: Frederic Weisbecker <fweisbec@gmail.com> > CC: Paul E. McKenney <paulmck@linux.vnet.ibm.com> > Signed-off-by: Paul A. Clarke <pc@us.ibm.com> Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com> > diff --git a/arch/powerpc/include/asm/irq_work.h > b/arch/powerpc/include/asm/irq_work.h > new file mode 100644 > index 0000000..18365ec > --- /dev/null > +++ b/arch/powerpc/include/asm/irq_work.h > @@ -0,0 +1,11 @@ > +#ifndef _ASM_IRQ_WORK_H > +#define _ASM_IRQ_WORK_H > + > +#include <asm/processor.h> > + > +static inline bool arch_irq_work_has_interrupt(void) > +{ > + return 1; > +} > + > +#endif /* _ASM_IRQ_WORK_H */ > > -- > Regards, > Paul Clarke, IBM > > -- > 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/
On Fri, 2015-02-13 at 13:38 -0600, Paul Clarke wrote: > implement arch_irq_work_has_interrupt() for powerpc > > Commit 9b01f5bf3 introduced a dependency on "IRQ work self-IPIs" for > full dynamic ticks to be enabled, by expecting architectures to > implement a suitable arch_irq_work_has_interrupt() routine. > > Several arches have implemented this routine, including x86 (3010279f) > and arm (09f6edd4), but powerpc was omitted. > > This patch implements this routine for powerpc. > > The symptom, at boot (on powerpc arch systems) with "nohz_full=<CPU > list>" is displayed: > NO_HZ: Can't run full dynticks because arch doesn't support irq > work self-IPIs > > after this patch: > NO_HZ: Full dynticks CPUs: <CPU list>. > > Tested against 3.19. It makes the message change, but is that correct? ie. do we actually implement "IRQ work self-IPIs"? > diff --git a/arch/powerpc/include/asm/irq_work.h > b/arch/powerpc/include/asm/irq_work.h > new file mode 100644 > index 0000000..18365ec > --- /dev/null > +++ b/arch/powerpc/include/asm/irq_work.h > @@ -0,0 +1,11 @@ > +#ifndef _ASM_IRQ_WORK_H > +#define _ASM_IRQ_WORK_H > + > +#include <asm/processor.h> > + > +static inline bool arch_irq_work_has_interrupt(void) > +{ > + return 1; Should be "true"; > +} cheers
On Mon, 2015-02-16 at 11:08 +1100, Michael Ellerman wrote: > On Fri, 2015-02-13 at 13:38 -0600, Paul Clarke wrote: > > implement arch_irq_work_has_interrupt() for powerpc > > > > Commit 9b01f5bf3 introduced a dependency on "IRQ work self-IPIs" for > > full dynamic ticks to be enabled, by expecting architectures to > > implement a suitable arch_irq_work_has_interrupt() routine. > > > > Several arches have implemented this routine, including x86 (3010279f) > > and arm (09f6edd4), but powerpc was omitted. > > > > This patch implements this routine for powerpc. > > .../... > > It makes the message change, but is that correct? ie. do we actually implement > "IRQ work self-IPIs"? I think so... Fred, do you think what we do will work ? We hijack our decrementer (local timer) by making it shoot almost immediately (1 tick away) and run the irq work at the beginning of __timer_interrupt(). At that point we are on our irq stack and have done irq_enter but that's about it. Cheers, Ben. > > diff --git a/arch/powerpc/include/asm/irq_work.h > > b/arch/powerpc/include/asm/irq_work.h > > new file mode 100644 > > index 0000000..18365ec > > --- /dev/null > > +++ b/arch/powerpc/include/asm/irq_work.h > > @@ -0,0 +1,11 @@ > > +#ifndef _ASM_IRQ_WORK_H > > +#define _ASM_IRQ_WORK_H > > + > > +#include <asm/processor.h> > > + > > +static inline bool arch_irq_work_has_interrupt(void) > > +{ > > + return 1; > > Should be "true"; > > > +} > > cheers > > > _______________________________________________ > Linuxppc-dev mailing list > Linuxppc-dev@lists.ozlabs.org > https://lists.ozlabs.org/listinfo/linuxppc-dev
Hi Ben, 2015-02-16 5:06 GMT+01:00 Benjamin Herrenschmidt <benh@kernel.crashing.org>: > On Mon, 2015-02-16 at 11:08 +1100, Michael Ellerman wrote: >> On Fri, 2015-02-13 at 13:38 -0600, Paul Clarke wrote: >> > implement arch_irq_work_has_interrupt() for powerpc >> > >> > Commit 9b01f5bf3 introduced a dependency on "IRQ work self-IPIs" for >> > full dynamic ticks to be enabled, by expecting architectures to >> > implement a suitable arch_irq_work_has_interrupt() routine. >> > >> > Several arches have implemented this routine, including x86 (3010279f) >> > and arm (09f6edd4), but powerpc was omitted. >> > >> > This patch implements this routine for powerpc. >> > > .../... >> >> It makes the message change, but is that correct? ie. do we actually implement >> "IRQ work self-IPIs"? > > I think so... Fred, do you think what we do will work ? We hijack our > decrementer (local timer) by making it shoot almost immediately (1 tick > away) and run the irq work at the beginning of __timer_interrupt(). > > At that point we are on our irq stack and have done irq_enter but that's > about it. Yes that should work. After all "self-IPI" is an oxymoron. One would expect an IPI to be triggered by an irq controller but if such operation isn't supported with the current CPU being both source and destination, anything triggering the desired callback in an interrupt context in a reasonable amount of time ahead does the job here. I thought well that's what powerpc was doing for irq work but I wasn't sure I understood the code correctly. I should have pinged people about that, sorry. Thanks. > > Cheers, > Ben. > >> > diff --git a/arch/powerpc/include/asm/irq_work.h >> > b/arch/powerpc/include/asm/irq_work.h >> > new file mode 100644 >> > index 0000000..18365ec >> > --- /dev/null >> > +++ b/arch/powerpc/include/asm/irq_work.h >> > @@ -0,0 +1,11 @@ >> > +#ifndef _ASM_IRQ_WORK_H >> > +#define _ASM_IRQ_WORK_H >> > + >> > +#include <asm/processor.h> >> > + >> > +static inline bool arch_irq_work_has_interrupt(void) >> > +{ >> > + return 1; >> >> Should be "true"; >> >> > +} >> >> cheers >> >> >> _______________________________________________ >> Linuxppc-dev mailing list >> Linuxppc-dev@lists.ozlabs.org >> https://lists.ozlabs.org/listinfo/linuxppc-dev > >
On Sun, 2015-02-22 at 23:13 +0100, Frederic Weisbecker wrote: > Yes that should work. After all "self-IPI" is an oxymoron. One would > expect an IPI to be triggered by an irq controller but if such > operation isn't supported with the current CPU being both source and > destination, anything triggering the desired callback in an interrupt > context in a reasonable amount of time ahead does the job here. We could do self-IPI on platforms that have an SMP-capable interrupt controller too but it would probably have higher overhead and would require verifying that the code for each of our different interrupt controllers is safe to be called from NMIs (hint: ioremap space isn't safe to access from NMIs for us on some CPU families...). We might be able to do better than using the decrementer on some CPUs by using local doorbells, but for now this will do. > I thought well that's what powerpc was doing for irq work but I wasn't > sure I understood the code correctly. I should have pinged people > about that, sorry. No worries, Cheers, Ben.
diff --git a/arch/powerpc/include/asm/irq_work.h b/arch/powerpc/include/asm/irq_work.h new file mode 100644 index 0000000..18365ec --- /dev/null +++ b/arch/powerpc/include/asm/irq_work.h @@ -0,0 +1,11 @@ +#ifndef _ASM_IRQ_WORK_H +#define _ASM_IRQ_WORK_H + +#include <asm/processor.h> + +static inline bool arch_irq_work_has_interrupt(void) +{ + return 1; +} + +#endif /* _ASM_IRQ_WORK_H */
implement arch_irq_work_has_interrupt() for powerpc Commit 9b01f5bf3 introduced a dependency on "IRQ work self-IPIs" for full dynamic ticks to be enabled, by expecting architectures to implement a suitable arch_irq_work_has_interrupt() routine. Several arches have implemented this routine, including x86 (3010279f) and arm (09f6edd4), but powerpc was omitted. This patch implements this routine for powerpc. The symptom, at boot (on powerpc arch systems) with "nohz_full=<CPU list>" is displayed: NO_HZ: Can't run full dynticks because arch doesn't support irq work self-IPIs after this patch: NO_HZ: Full dynticks CPUs: <CPU list>. Tested against 3.19. CC: Frederic Weisbecker <fweisbec@gmail.com> CC: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Signed-off-by: Paul A. Clarke <pc@us.ibm.com> -- Regards, Paul Clarke, IBM