Messages in this thread | | | Date | Tue, 2 Jun 2015 11:57:24 +0530 | From | Viresh Kumar <> | Subject | Re: [RFC PATCH] cpufreq/hotplug: Fix cpu-hotplug cpufreq race conditions |
| |
On 02-06-15, 11:50, Preeti U Murthy wrote: > CPU0 CPU1 > > store* store* > > lock(policy 1) lock(policy 2) > cpufreq_set_policy() cpufreq_set_policy() > EXIT() : > dbs-data->usage_count-- > > INIT() EXIT()
When will INIT() follow EXIT() in set_policy() for the same governor ? Maybe not, and so this sequence is hypothetical ?
> dbs_data exists dbs_data->usage_count -- = 0 > kfree(dbs_data) > dbs-data->usage_count++ > *NULL dereference*
But even if this happens, it should be handled with dbs_data->mutex_lock, which is used at other places already.
-- viresh
| |