| Message ID | 1548836194-15264-1-git-send-email-vincent.guittot@linaro.org (mailing list archive) |
|---|---|
| State | Superseded |
| Delegated to: | Geert Uytterhoeven |
| Headers | show |
| Series | PM-runtime: fix deadlock with ktime | expand |
Hi Vincent, On Wed, Jan 30, 2019 at 9:16 AM Vincent Guittot <vincent.guittot@linaro.org> wrote: > A deadlock has been seen when swicthing clocksources which use PM runtime. > The call path is: > change_clocksource > ... > write_seqcount_begin > ... > timekeeping_update > ... > sh_cmt_clocksource_enable > ... > rpm_resume > pm_runtime_mark_last_busy > ktime_get > do > read_seqcount_begin > while read_seqcount_retry > .... > write_seqcount_end > > Although we should be safe because we haven't yet changed the clocksource > at that time, we can't because of seqcount protection. > > Use ktime_get_mono_fast_ns instead which is lock safe for such case > > Fixes: 8234f6734c5d ("PM-runtime: Switch autosuspend over to using hrtimers") > Reported-by: Biju Das <biju.das@bp.renesas.com> > Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org> Thanks for your patch! /** * ktime_get_mono_fast_ns - Fast NMI safe access to clock monotonic * * This timestamp is not guaranteed to be monotonic across an update. * The timestamp is calculated by: * * now = base_mono + clock_delta * slope * * So if the update lowers the slope, readers who are forced to the * not yet updated second array are still using the old steeper slope. * * tmono * ^ * | o n * | o n * | u * | o * |o * |12345678---> reader order * * o = old slope * u = update * n = new slope * * So reader 6 will observe time going backwards versus reader 5. * * While other CPUs are likely to be able observe that, the only way * for a CPU local observation is when an NMI hits in the middle of * the update. Timestamps taken from that NMI context might be ahead * of the following timestamps. Callers need to be aware of that and * deal with it. */ As this function is not guaranteed to be monotonic, have you checked how the Runtime PM code behaves if time goes backwards? Does it just make a suboptimal decision or does it crash? Thanks! Gr{oetje,eeting}s, Geert
Hi Geert, On Wed, 30 Jan 2019 at 09:21, Geert Uytterhoeven <geert@linux-m68k.org> wrote: > > Hi Vincent, > > On Wed, Jan 30, 2019 at 9:16 AM Vincent Guittot > <vincent.guittot@linaro.org> wrote: > > A deadlock has been seen when swicthing clocksources which use PM runtime. > > The call path is: > > change_clocksource > > ... > > write_seqcount_begin > > ... > > timekeeping_update > > ... > > sh_cmt_clocksource_enable > > ... > > rpm_resume > > pm_runtime_mark_last_busy > > ktime_get > > do > > read_seqcount_begin > > while read_seqcount_retry > > .... > > write_seqcount_end > > > > Although we should be safe because we haven't yet changed the clocksource > > at that time, we can't because of seqcount protection. > > > > Use ktime_get_mono_fast_ns instead which is lock safe for such case > > > > Fixes: 8234f6734c5d ("PM-runtime: Switch autosuspend over to using hrtimers") > > Reported-by: Biju Das <biju.das@bp.renesas.com> > > Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org> > > Thanks for your patch! > > /** > * ktime_get_mono_fast_ns - Fast NMI safe access to clock monotonic > * > * This timestamp is not guaranteed to be monotonic across an update. > * The timestamp is calculated by: > * > * now = base_mono + clock_delta * slope > * > * So if the update lowers the slope, readers who are forced to the > * not yet updated second array are still using the old steeper slope. > * > * tmono > * ^ > * | o n > * | o n > * | u > * | o > * |o > * |12345678---> reader order > * > * o = old slope > * u = update > * n = new slope > * > * So reader 6 will observe time going backwards versus reader 5. > * > * While other CPUs are likely to be able observe that, the only way > * for a CPU local observation is when an NMI hits in the middle of > * the update. Timestamps taken from that NMI context might be ahead > * of the following timestamps. Callers need to be aware of that and > * deal with it. > */ > > As this function is not guaranteed to be monotonic, have you checked how > the Runtime PM code behaves if time goes backwards? Does it just make > a suboptimal decision or does it crash? As a worst case this will generate a suboptimal decision around the update Regards, Vincent > > Thanks! > > Gr{oetje,eeting}s, > > Geert > > -- > Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org > > In personal conversations with technical people, I call myself a hacker. But > when I'm talking to journalists I just say "programmer" or something like that. > -- Linus Torvalds
On Wed, Jan 30, 2019 at 10:14 AM Vincent Guittot <vincent.guittot@linaro.org> wrote: > > Hi Geert, > > On Wed, 30 Jan 2019 at 09:21, Geert Uytterhoeven <geert@linux-m68k.org> wrote: > > > > Hi Vincent, > > > > On Wed, Jan 30, 2019 at 9:16 AM Vincent Guittot > > <vincent.guittot@linaro.org> wrote: > > > A deadlock has been seen when swicthing clocksources which use PM runtime. > > > The call path is: > > > change_clocksource > > > ... > > > write_seqcount_begin > > > ... > > > timekeeping_update > > > ... > > > sh_cmt_clocksource_enable > > > ... > > > rpm_resume > > > pm_runtime_mark_last_busy > > > ktime_get > > > do > > > read_seqcount_begin > > > while read_seqcount_retry > > > .... > > > write_seqcount_end > > > > > > Although we should be safe because we haven't yet changed the clocksource > > > at that time, we can't because of seqcount protection. > > > > > > Use ktime_get_mono_fast_ns instead which is lock safe for such case > > > > > > Fixes: 8234f6734c5d ("PM-runtime: Switch autosuspend over to using hrtimers") > > > Reported-by: Biju Das <biju.das@bp.renesas.com> > > > Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org> > > > > Thanks for your patch! > > > > /** > > * ktime_get_mono_fast_ns - Fast NMI safe access to clock monotonic > > * > > * This timestamp is not guaranteed to be monotonic across an update. > > * The timestamp is calculated by: > > * > > * now = base_mono + clock_delta * slope > > * > > * So if the update lowers the slope, readers who are forced to the > > * not yet updated second array are still using the old steeper slope. > > * > > * tmono > > * ^ > > * | o n > > * | o n > > * | u > > * | o > > * |o > > * |12345678---> reader order > > * > > * o = old slope > > * u = update > > * n = new slope > > * > > * So reader 6 will observe time going backwards versus reader 5. > > * > > * While other CPUs are likely to be able observe that, the only way > > * for a CPU local observation is when an NMI hits in the middle of > > * the update. Timestamps taken from that NMI context might be ahead > > * of the following timestamps. Callers need to be aware of that and > > * deal with it. > > */ > > > > As this function is not guaranteed to be monotonic, have you checked how > > the Runtime PM code behaves if time goes backwards? Does it just make > > a suboptimal decision or does it crash? > > As a worst case this will generate a suboptimal decision around the update So that should be explained in the changelog of the patch. In detail, if poss, please.
On Wed, 30 Jan 2019 at 10:39, Rafael J. Wysocki <rafael@kernel.org> wrote: > > On Wed, Jan 30, 2019 at 10:14 AM Vincent Guittot > <vincent.guittot@linaro.org> wrote: > > > > Hi Geert, > > > > On Wed, 30 Jan 2019 at 09:21, Geert Uytterhoeven <geert@linux-m68k.org> wrote: > > > > > > Hi Vincent, > > > > > > On Wed, Jan 30, 2019 at 9:16 AM Vincent Guittot > > > <vincent.guittot@linaro.org> wrote: > > > > A deadlock has been seen when swicthing clocksources which use PM runtime. > > > > The call path is: > > > > change_clocksource > > > > ... > > > > write_seqcount_begin > > > > ... > > > > timekeeping_update > > > > ... > > > > sh_cmt_clocksource_enable > > > > ... > > > > rpm_resume > > > > pm_runtime_mark_last_busy > > > > ktime_get > > > > do > > > > read_seqcount_begin > > > > while read_seqcount_retry > > > > .... > > > > write_seqcount_end > > > > > > > > Although we should be safe because we haven't yet changed the clocksource > > > > at that time, we can't because of seqcount protection. > > > > > > > > Use ktime_get_mono_fast_ns instead which is lock safe for such case > > > > > > > > Fixes: 8234f6734c5d ("PM-runtime: Switch autosuspend over to using hrtimers") > > > > Reported-by: Biju Das <biju.das@bp.renesas.com> > > > > Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org> > > > > > > Thanks for your patch! > > > > > > /** > > > * ktime_get_mono_fast_ns - Fast NMI safe access to clock monotonic > > > * > > > * This timestamp is not guaranteed to be monotonic across an update. > > > * The timestamp is calculated by: > > > * > > > * now = base_mono + clock_delta * slope > > > * > > > * So if the update lowers the slope, readers who are forced to the > > > * not yet updated second array are still using the old steeper slope. > > > * > > > * tmono > > > * ^ > > > * | o n > > > * | o n > > > * | u > > > * | o > > > * |o > > > * |12345678---> reader order > > > * > > > * o = old slope > > > * u = update > > > * n = new slope > > > * > > > * So reader 6 will observe time going backwards versus reader 5. > > > * > > > * While other CPUs are likely to be able observe that, the only way > > > * for a CPU local observation is when an NMI hits in the middle of > > > * the update. Timestamps taken from that NMI context might be ahead > > > * of the following timestamps. Callers need to be aware of that and > > > * deal with it. > > > */ > > > > > > As this function is not guaranteed to be monotonic, have you checked how > > > the Runtime PM code behaves if time goes backwards? Does it just make > > > a suboptimal decision or does it crash? > > > > As a worst case this will generate a suboptimal decision around the update > > So that should be explained in the changelog of the patch. In detail, > if poss, please. Ok, I'm going to update the commit message
diff --git a/drivers/base/power/runtime.c b/drivers/base/power/runtime.c index 457be03..708a13f 100644 --- a/drivers/base/power/runtime.c +++ b/drivers/base/power/runtime.c @@ -130,7 +130,7 @@ u64 pm_runtime_autosuspend_expiration(struct device *dev) { int autosuspend_delay; u64 last_busy, expires = 0; - u64 now = ktime_to_ns(ktime_get()); + u64 now = ktime_get_mono_fast_ns(); if (!dev->power.use_autosuspend) goto out; @@ -909,7 +909,7 @@ static enum hrtimer_restart pm_suspend_timer_fn(struct hrtimer *timer) * If 'expires' is after the current time, we've been called * too early. */ - if (expires > 0 && expires < ktime_to_ns(ktime_get())) { + if (expires > 0 && expires < ktime_get_mono_fast_ns()) { dev->power.timer_expires = 0; rpm_suspend(dev, dev->power.timer_autosuspends ? (RPM_ASYNC | RPM_AUTO) : RPM_ASYNC); @@ -928,7 +928,7 @@ static enum hrtimer_restart pm_suspend_timer_fn(struct hrtimer *timer) int pm_schedule_suspend(struct device *dev, unsigned int delay) { unsigned long flags; - ktime_t expires; + u64 expires; int retval; spin_lock_irqsave(&dev->power.lock, flags); @@ -945,8 +945,8 @@ int pm_schedule_suspend(struct device *dev, unsigned int delay) /* Other scheduled or pending requests need to be canceled. */ pm_runtime_cancel_pending(dev); - expires = ktime_add(ktime_get(), ms_to_ktime(delay)); - dev->power.timer_expires = ktime_to_ns(expires); + expires = ktime_get_mono_fast_ns() + (u64)delay * NSEC_PER_MSEC); + dev->power.timer_expires = expires; dev->power.timer_autosuspends = 0; hrtimer_start(&dev->power.suspend_timer, expires, HRTIMER_MODE_ABS); diff --git a/include/linux/pm_runtime.h b/include/linux/pm_runtime.h index 54af4ee..fed5be7 100644 --- a/include/linux/pm_runtime.h +++ b/include/linux/pm_runtime.h @@ -105,7 +105,7 @@ static inline bool pm_runtime_callbacks_present(struct device *dev) static inline void pm_runtime_mark_last_busy(struct device *dev) { - WRITE_ONCE(dev->power.last_busy, ktime_to_ns(ktime_get())); + WRITE_ONCE(dev->power.last_busy, ktime_get_mono_fast_ns()); } static inline bool pm_runtime_is_irq_safe(struct device *dev)
A deadlock has been seen when swicthing clocksources which use PM runtime. The call path is: change_clocksource ... write_seqcount_begin ... timekeeping_update ... sh_cmt_clocksource_enable ... rpm_resume pm_runtime_mark_last_busy ktime_get do read_seqcount_begin while read_seqcount_retry .... write_seqcount_end Although we should be safe because we haven't yet changed the clocksource at that time, we can't because of seqcount protection. Use ktime_get_mono_fast_ns instead which is lock safe for such case Fixes: 8234f6734c5d ("PM-runtime: Switch autosuspend over to using hrtimers") Reported-by: Biju Das <biju.das@bp.renesas.com> Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org> --- drivers/base/power/runtime.c | 10 +++++----- include/linux/pm_runtime.h | 2 +- 2 files changed, 6 insertions(+), 6 deletions(-)