Re: [PATCH] xtime_lock vs update_process_times
On Wed, Feb 13, 2008 at 09:33:16PM +0100, Peter Zijlstra wrote: > Subject: xtime_lock vs update_process_times > From: Peter Zijlstra <[EMAIL PROTECTED]> > > ( repost from: http://lkml.org/lkml/2008/1/28/101 ) > > Commit: d3d74453c34f8fd87674a8cf5b8a327c68f22e99 > Subject: hrtimer: fixup the HRTIMER_CB_IRQSAFE_NO_SOFTIRQ fallback > > Broke several archs, since only Russel bothered to merge the fix, > and Greg to ACK his arch, I'm sending this for merger. > > I have confirmation that the Alpha bit results in a booting kernel. > That leaves: blackfin, frv, sh and sparc untested. > > The deadlock in question was found by Russell: > > IRQ handle > -> timer_tick() - xtime seqlock held for write > -> update_process_times() > -> run_local_timers() > -> hrtimer_run_queues() > -> hrtimer_get_softirq_time() - tries to get a read lock > > Now, Thomas assures me the fix is trivial, only do_timer() needs to be > done under the xtime_lock, and update_process_times() can savely be removed > from under it. > The SH bits also work fine. I've already merged that part in to my tree. Thanks, Peter. -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
[PATCH] xtime_lock vs update_process_times
This time with LKML CC'ed. Sorry for the duplication. Subject: xtime_lock vs update_process_times From: Peter Zijlstra <[EMAIL PROTECTED]> ( repost from: http://lkml.org/lkml/2008/1/28/101 ) Commit: d3d74453c34f8fd87674a8cf5b8a327c68f22e99 Subject: hrtimer: fixup the HRTIMER_CB_IRQSAFE_NO_SOFTIRQ fallback Broke several archs, since only Russel bothered to merge the fix, and Greg to ACK his arch, I'm sending this for merger. I have confirmation that the Alpha bit results in a booting kernel. That leaves: blackfin, frv, sh and sparc untested. The deadlock in question was found by Russell: IRQ handle -> timer_tick() - xtime seqlock held for write -> update_process_times() -> run_local_timers() -> hrtimer_run_queues() -> hrtimer_get_softirq_time() - tries to get a read lock Now, Thomas assures me the fix is trivial, only do_timer() needs to be done under the xtime_lock, and update_process_times() can savely be removed from under it. Signed-off-by: Peter Zijlstra <[EMAIL PROTECTED]> Acked-by: Greg Ungerer <[EMAIL PROTECTED]> CC: [EMAIL PROTECTED] CC: [EMAIL PROTECTED] CC: [EMAIL PROTECTED] CC: [EMAIL PROTECTED] CC: [EMAIL PROTECTED] --- arch/alpha/kernel/time.c | 15 --- arch/blackfin/kernel/time.c|8 +--- arch/frv/kernel/time.c |6 -- arch/m68knommu/kernel/time.c | 12 +++- arch/sh/kernel/timers/timer-cmt.c |9 - arch/sh/kernel/timers/timer-mtu2.c |2 -- arch/sparc/kernel/pcic.c |2 +- arch/sparc/kernel/time.c |7 +++ 8 files changed, 28 insertions(+), 33 deletions(-) Index: linux-2.6/arch/alpha/kernel/time.c === --- linux-2.6.orig/arch/alpha/kernel/time.c +++ linux-2.6/arch/alpha/kernel/time.c @@ -119,13 +119,8 @@ irqreturn_t timer_interrupt(int irq, voi state.partial_tick = delta & ((1UL << FIX_SHIFT) - 1); nticks = delta >> FIX_SHIFT; - while (nticks > 0) { - do_timer(1); -#ifndef CONFIG_SMP - update_process_times(user_mode(get_irq_regs())); -#endif - nticks--; - } + if (nticks) + do_timer(nticks); /* * If we have an externally synchronized Linux clock, then update @@ -141,6 +136,12 @@ irqreturn_t timer_interrupt(int irq, voi } write_sequnlock(_lock); + +#ifndef CONFIG_SMP + while (nticks--) + update_process_times(user_mode(get_irq_regs())); +#endif + return IRQ_HANDLED; } Index: linux-2.6/arch/blackfin/kernel/time.c === --- linux-2.6.orig/arch/blackfin/kernel/time.c +++ linux-2.6/arch/blackfin/kernel/time.c @@ -137,9 +137,6 @@ irqreturn_t timer_interrupt(int irq, voi do_timer(1); -#ifndef CONFIG_SMP - update_process_times(user_mode(get_irq_regs())); -#endif profile_tick(CPU_PROFILING); /* @@ -161,6 +158,11 @@ irqreturn_t timer_interrupt(int irq, voi last_rtc_update = xtime.tv_sec - 600; } write_sequnlock(_lock); + +#ifndef CONFIG_SMP + update_process_times(user_mode(get_irq_regs())); +#endif + return IRQ_HANDLED; } Index: linux-2.6/arch/frv/kernel/time.c === --- linux-2.6.orig/arch/frv/kernel/time.c +++ linux-2.6/arch/frv/kernel/time.c @@ -63,6 +63,7 @@ static irqreturn_t timer_interrupt(int i /* last time the cmos clock got updated */ static long last_rtc_update = 0; + profile_tick(CPU_PROFILING); /* * Here we are in the timer irq handler. We just have irqs locally * disabled but we don't know if the timer_bh is running on the other @@ -73,8 +74,6 @@ static irqreturn_t timer_interrupt(int i write_seqlock(_lock); do_timer(1); - update_process_times(user_mode(get_irq_regs())); - profile_tick(CPU_PROFILING); /* * If we have an externally synchronized Linux clock, then update @@ -99,6 +98,9 @@ static irqreturn_t timer_interrupt(int i #endif /* CONFIG_HEARTBEAT */ write_sequnlock(_lock); + + update_process_times(user_mode(get_irq_regs())); + return IRQ_HANDLED; } Index: linux-2.6/arch/m68knommu/kernel/time.c === --- linux-2.6.orig/arch/m68knommu/kernel/time.c +++ linux-2.6/arch/m68knommu/kernel/time.c @@ -42,14 +42,12 @@ irqreturn_t arch_timer_interrupt(int irq /* last time the cmos clock got updated */ static long last_rtc_update=0; + if (current->pid) + profile_tick(CPU_PROFILING); + write_seqlock(_lock); do_timer(1); -#ifndef CONFIG_SMP - update_process_times(user_mode(get_irq_regs())); -#endif - if (current->pid) -
[PATCH] xtime_lock vs update_process_times
This time with LKML CC'ed. Sorry for the duplication. Subject: xtime_lock vs update_process_times From: Peter Zijlstra [EMAIL PROTECTED] ( repost from: http://lkml.org/lkml/2008/1/28/101 ) Commit: d3d74453c34f8fd87674a8cf5b8a327c68f22e99 Subject: hrtimer: fixup the HRTIMER_CB_IRQSAFE_NO_SOFTIRQ fallback Broke several archs, since only Russel bothered to merge the fix, and Greg to ACK his arch, I'm sending this for merger. I have confirmation that the Alpha bit results in a booting kernel. That leaves: blackfin, frv, sh and sparc untested. The deadlock in question was found by Russell: IRQ handle - timer_tick() - xtime seqlock held for write - update_process_times() - run_local_timers() - hrtimer_run_queues() - hrtimer_get_softirq_time() - tries to get a read lock Now, Thomas assures me the fix is trivial, only do_timer() needs to be done under the xtime_lock, and update_process_times() can savely be removed from under it. Signed-off-by: Peter Zijlstra [EMAIL PROTECTED] Acked-by: Greg Ungerer [EMAIL PROTECTED] CC: [EMAIL PROTECTED] CC: [EMAIL PROTECTED] CC: [EMAIL PROTECTED] CC: [EMAIL PROTECTED] CC: [EMAIL PROTECTED] --- arch/alpha/kernel/time.c | 15 --- arch/blackfin/kernel/time.c|8 +--- arch/frv/kernel/time.c |6 -- arch/m68knommu/kernel/time.c | 12 +++- arch/sh/kernel/timers/timer-cmt.c |9 - arch/sh/kernel/timers/timer-mtu2.c |2 -- arch/sparc/kernel/pcic.c |2 +- arch/sparc/kernel/time.c |7 +++ 8 files changed, 28 insertions(+), 33 deletions(-) Index: linux-2.6/arch/alpha/kernel/time.c === --- linux-2.6.orig/arch/alpha/kernel/time.c +++ linux-2.6/arch/alpha/kernel/time.c @@ -119,13 +119,8 @@ irqreturn_t timer_interrupt(int irq, voi state.partial_tick = delta ((1UL FIX_SHIFT) - 1); nticks = delta FIX_SHIFT; - while (nticks 0) { - do_timer(1); -#ifndef CONFIG_SMP - update_process_times(user_mode(get_irq_regs())); -#endif - nticks--; - } + if (nticks) + do_timer(nticks); /* * If we have an externally synchronized Linux clock, then update @@ -141,6 +136,12 @@ irqreturn_t timer_interrupt(int irq, voi } write_sequnlock(xtime_lock); + +#ifndef CONFIG_SMP + while (nticks--) + update_process_times(user_mode(get_irq_regs())); +#endif + return IRQ_HANDLED; } Index: linux-2.6/arch/blackfin/kernel/time.c === --- linux-2.6.orig/arch/blackfin/kernel/time.c +++ linux-2.6/arch/blackfin/kernel/time.c @@ -137,9 +137,6 @@ irqreturn_t timer_interrupt(int irq, voi do_timer(1); -#ifndef CONFIG_SMP - update_process_times(user_mode(get_irq_regs())); -#endif profile_tick(CPU_PROFILING); /* @@ -161,6 +158,11 @@ irqreturn_t timer_interrupt(int irq, voi last_rtc_update = xtime.tv_sec - 600; } write_sequnlock(xtime_lock); + +#ifndef CONFIG_SMP + update_process_times(user_mode(get_irq_regs())); +#endif + return IRQ_HANDLED; } Index: linux-2.6/arch/frv/kernel/time.c === --- linux-2.6.orig/arch/frv/kernel/time.c +++ linux-2.6/arch/frv/kernel/time.c @@ -63,6 +63,7 @@ static irqreturn_t timer_interrupt(int i /* last time the cmos clock got updated */ static long last_rtc_update = 0; + profile_tick(CPU_PROFILING); /* * Here we are in the timer irq handler. We just have irqs locally * disabled but we don't know if the timer_bh is running on the other @@ -73,8 +74,6 @@ static irqreturn_t timer_interrupt(int i write_seqlock(xtime_lock); do_timer(1); - update_process_times(user_mode(get_irq_regs())); - profile_tick(CPU_PROFILING); /* * If we have an externally synchronized Linux clock, then update @@ -99,6 +98,9 @@ static irqreturn_t timer_interrupt(int i #endif /* CONFIG_HEARTBEAT */ write_sequnlock(xtime_lock); + + update_process_times(user_mode(get_irq_regs())); + return IRQ_HANDLED; } Index: linux-2.6/arch/m68knommu/kernel/time.c === --- linux-2.6.orig/arch/m68knommu/kernel/time.c +++ linux-2.6/arch/m68knommu/kernel/time.c @@ -42,14 +42,12 @@ irqreturn_t arch_timer_interrupt(int irq /* last time the cmos clock got updated */ static long last_rtc_update=0; + if (current-pid) + profile_tick(CPU_PROFILING); + write_seqlock(xtime_lock); do_timer(1); -#ifndef CONFIG_SMP - update_process_times(user_mode(get_irq_regs())); -#endif - if (current-pid) -
Re: [PATCH] xtime_lock vs update_process_times
On Wed, Feb 13, 2008 at 09:33:16PM +0100, Peter Zijlstra wrote: Subject: xtime_lock vs update_process_times From: Peter Zijlstra [EMAIL PROTECTED] ( repost from: http://lkml.org/lkml/2008/1/28/101 ) Commit: d3d74453c34f8fd87674a8cf5b8a327c68f22e99 Subject: hrtimer: fixup the HRTIMER_CB_IRQSAFE_NO_SOFTIRQ fallback Broke several archs, since only Russel bothered to merge the fix, and Greg to ACK his arch, I'm sending this for merger. I have confirmation that the Alpha bit results in a booting kernel. That leaves: blackfin, frv, sh and sparc untested. The deadlock in question was found by Russell: IRQ handle - timer_tick() - xtime seqlock held for write - update_process_times() - run_local_timers() - hrtimer_run_queues() - hrtimer_get_softirq_time() - tries to get a read lock Now, Thomas assures me the fix is trivial, only do_timer() needs to be done under the xtime_lock, and update_process_times() can savely be removed from under it. The SH bits also work fine. I've already merged that part in to my tree. Thanks, Peter. -- To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/