[PATCH] perf: Protect pmu device removal with pmu_bus_running check CONFIG_DEBUG_TEST_DRIVER_REMOVE kernel panic

[Jiri Olsa]

On Thu, Oct 20, 2016 at 11:42:59AM +0200, Jiri Olsa wrote:
> On Thu, Oct 20, 2016 at 11:04:16AM +0200, Peter Zijlstra wrote:
> > On Thu, Oct 20, 2016 at 10:58:03AM +0200, Jiri Olsa wrote:
> > 
> > > @@ -8869,11 +8869,15 @@ void perf_pmu_unregister(struct pmu *pmu)
> > >   free_percpu(pmu->pmu_disable_count);
> > >   if (pmu->type >= PERF_TYPE_MAX)
> > >           idr_remove(&pmu_idr, pmu->type);
> > > - if (pmu->nr_addr_filters)
> > > -         device_remove_file(pmu->dev, &dev_attr_nr_addr_filters);
> > > - device_del(pmu->dev);
> > > - put_device(pmu->dev);
> > > + mutex_lock(&pmus_lock);
> > > + if (pmu_bus_running) {
> > > +         if (pmu->nr_addr_filters)
> > > +                 device_remove_file(pmu->dev, &dev_attr_nr_addr_filters);
> > > +         device_del(pmu->dev);
> > > +         put_device(pmu->dev);
> > > + }
> > >   free_pmu_context(pmu);
> > > + mutex_unlock(&pmus_lock);
> > >  }
> > >  EXPORT_SYMBOL_GPL(perf_pmu_unregister);
> > 
> > I think that is still racy..
> > 
> > 
> > unregister:         sysfs_init:
> > 
> > mutex_lock(&pmus_lock);
> > list_del_rcu(&pmu->entry);
> > mutex_unlock(&pmus_lock);
> > 
> > synchronize_*rcu();
> > 
> >                     mutex_lock(&pmus_lock);
> >                     list_for_each_entry(pmu, &pmus, entry) {
> >                             /* add device muck */
> 
> ah, I thought this part would add the device back.. but it's
> already out of the pmu list.. right :-\

attached fix, thanks

jirka


---
CAI Qian reported crash [1] in uncore device removal related
to CONFIG_DEBUG_TEST_DRIVER_REMOVE option.

The reason for crash is that  perf_pmu_unregister tries to remove
pmu device which is not added at this point. We add pmu devices
only after pmu_bus is registered which happens in perf_event_sysfs_init
init call and sets pmu_bus_running flag.

The fix is to get the pmu_bus_running flag state at the point
the pmu is taken  out of the pmus list and  remove the device
later only if it's set.

[1] https://marc.info/?l=linux-kernel&m=147688837328451

Reported-by: CAI Qian <caiq...@redhat.com>
Signed-off-by: Jiri Olsa <jo...@kernel.org>
---
 kernel/events/core.c | 13 +++++++++----
 1 file changed, 9 insertions(+), 4 deletions(-)

diff --git a/kernel/events/core.c b/kernel/events/core.c
index c6e47e97b33f..a5d2e62faf7e 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -8855,7 +8855,10 @@ EXPORT_SYMBOL_GPL(perf_pmu_register);
 
 void perf_pmu_unregister(struct pmu *pmu)
 {
+       int remove_device;
+
        mutex_lock(&pmus_lock);
+       remove_device = pmu_bus_running;
        list_del_rcu(&pmu->entry);
        mutex_unlock(&pmus_lock);
 
@@ -8869,10 +8872,12 @@ void perf_pmu_unregister(struct pmu *pmu)
        free_percpu(pmu->pmu_disable_count);
        if (pmu->type >= PERF_TYPE_MAX)
                idr_remove(&pmu_idr, pmu->type);
-       if (pmu->nr_addr_filters)
-               device_remove_file(pmu->dev, &dev_attr_nr_addr_filters);
-       device_del(pmu->dev);
-       put_device(pmu->dev);
+       if (remove_device) {
+               if (pmu->nr_addr_filters)
+                       device_remove_file(pmu->dev, &dev_attr_nr_addr_filters);
+               device_del(pmu->dev);
+               put_device(pmu->dev);
+       }
        free_pmu_context(pmu);
 }
 EXPORT_SYMBOL_GPL(perf_pmu_unregister);
-- 
2.7.4