On 12/24/22 12:14, Michal Luczaj wrote:
This lock/unlock pair can cause a deadlock because it's inside the SRCU
read side critical section. Fortunately it's simpler to just use
mutex_lock for the whole function instead of using two small critical
sections, and then SRCU is not needed.
Ah, I didn't think using a single mutex_lock would be ok. And maybe that's
a silly question, but how can this lock/unlock pair cause a deadlock? My
assumption was it's a*sleepable* RCU after all.
This is a pretty simple AB-BA deadlock, just involving SRCU and a mutex
instead of two mutexes:
Thread 1 Thread 2
mutex_lock()
srcu_read_lock()
mutex_lock() // stops here
synchronize_srcu() // stops here
mutex_unlock()
mutex_unlock()
srcu_read_unlock()
Thread 2's synchronize_srcu() is waiting for srcu_read_unlock(), which
however won't happen until thread 1 can take the mutex. But the mutex
is taken by thread 2, hence the deadlock.
Paolo
