Messages in this thread | | | From | Lai Jiangshan <> | Date | Thu, 30 Nov 2017 22:43:16 +0800 | Subject | Re: [PATCH 6/7] smp/hotplug: Differentiate the AP-work lockdep class between up and down |
| |
On Thu, Sep 21, 2017 at 1:00 AM, Peter Zijlstra <peterz@infradead.org> wrote: > With lockdep-crossrelease we get deadlock reports that span cpu-up and > cpu-down chains. Such deadlocks cannot possibly happen because cpu-up > and cpu-down are globally serialized. > > CPU0 CPU1 CPU2 > cpuhp_up_callbacks: takedown_cpu: cpuhp_thread_fun: > > cpuhp_state > irq_lock_sparse() > irq_lock_sparse() > wait_for_completion() > cpuhp_state > complete() > > Now that we have consistent AP state, we can trivially separate the > AP-work class between up and down using st->bringup. > > Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> > --- > kernel/cpu.c | 41 ++++++++++++++++++++++++++++++++--------- > 1 file changed, 32 insertions(+), 9 deletions(-) > > --- a/kernel/cpu.c > +++ b/kernel/cpu.c > @@ -68,9 +68,26 @@ struct cpuhp_cpu_state { > static DEFINE_PER_CPU(struct cpuhp_cpu_state, cpuhp_state); > > #if defined(CONFIG_LOCKDEP) && defined(CONFIG_SMP) > -static struct lock_class_key cpuhp_state_key; > -static struct lockdep_map cpuhp_state_lock_map = > - STATIC_LOCKDEP_MAP_INIT("cpuhp_state", &cpuhp_state_key); > +static struct lockdep_map cpuhp_state_up_map = > + STATIC_LOCKDEP_MAP_INIT("cpuhp_state-up", &cpuhp_state_up_map); > +static struct lockdep_map cpuhp_state_down_map = > + STATIC_LOCKDEP_MAP_INIT("cpuhp_state-down", &cpuhp_state_down_map); > + > + > +static void inline cpuhp_lock_acquire(bool bringup) > +{ > + lock_map_acquire(bringup ? &cpuhp_state_up_map : &cpuhp_state_down_map); > +} > + > +static void inline cpuhp_lock_release(bool bringup) > +{ > + lock_map_release(bringup ? &cpuhp_state_up_map : &cpuhp_state_down_map); > +} > +#else > + > +static void inline cpuhp_lock_acquire(bool bringup) { } > +static void inline cpuhp_lock_release(bool bringup) { } > + > #endif > > /** > @@ -512,7 +529,7 @@ static void cpuhp_thread_fun(unsigned in > if (WARN_ON_ONCE(!st->should_run)) > return; > > - lock_map_acquire(&cpuhp_state_lock_map); > + cpuhp_lock_acquire(bringup); > > if (st->single) { > state = st->cb_state; > @@ -564,7 +581,7 @@ static void cpuhp_thread_fun(unsigned in > } > > next: > - lock_map_release(&cpuhp_state_lock_map); > + cpuhp_lock_release(bringup); > > if (!st->should_run) > complete(&st->done); > @@ -581,8 +598,11 @@ cpuhp_invoke_ap_callback(int cpu, enum c > if (!cpu_online(cpu)) > return 0; > > - lock_map_acquire(&cpuhp_state_lock_map); > - lock_map_release(&cpuhp_state_lock_map); > + cpuhp_lock_acquire(false); > + cpuhp_lock_release(false); > + > + cpuhp_lock_acquire(true); > + cpuhp_lock_release(true);
Hello, Peter,
I'm reading the code in kernel/cpu.c. I couldn't understand why both lockep_map are acquired here? Is the lockep_map matching for the argument @bringup enough here?
The log shows that the argument @bringup had been added when the time this commit was applied. But it was quite probably non-existed when you wrote the patch since the time was close.
thanks, Lai.
> > /* > * If we are up and running, use the hotplug thread. For early calls > @@ -620,8 +640,11 @@ static int cpuhp_kick_ap_work(unsigned i > enum cpuhp_state prev_state = st->state; > int ret; > > - lock_map_acquire(&cpuhp_state_lock_map); > - lock_map_release(&cpuhp_state_lock_map); > + cpuhp_lock_acquire(false); > + cpuhp_lock_release(false); > + > + cpuhp_lock_acquire(true); > + cpuhp_lock_release(true); > > trace_cpuhp_enter(cpu, st->target, prev_state, cpuhp_kick_ap_work); > ret = cpuhp_kick_ap(st, st->target); > >
| |