[PATCH v4 1/5] locking/mutex: implement mutex_trylock_nested

[Maxim Levitsky]

Despite the fact that several lockdep-related checks are skipped when
calling trylock* versions of the locking primitives, for example
mutex_trylock, each time the mutex is acquired, a held_lock is still
placed onto the lockdep stack by __lock_acquire() which is called
regardless of whether the trylock* or regular locking API was used.

This means that if the caller successfully acquires more than
MAX_LOCK_DEPTH locks of the same class, even when using mutex_trylock,
lockdep will still complain that the maximum depth of the held lock stack
has been reached and disable itself.

For example, the following error currently occurs in the ARM version
of KVM, once the code tries to lock all vCPUs of a VM configured with more
than MAX_LOCK_DEPTH vCPUs, a situation that can easily happen on modern
systems, where having more than 48 CPUs is common, and it's also common to
run VMs that have vCPU counts approaching that number:

[  328.171264] BUG: MAX_LOCK_DEPTH too low!
[  328.175227] turning off the locking correctness validator.
[  328.180726] Please attach the output of /proc/lock_stat to the bug report
[  328.187531] depth: 48  max: 48!
[  328.190678] 48 locks held by qemu-kvm/11664:
[  328.194957]  #0: ffff800086de5ba0 (&kvm->lock){+.+.}-{3:3}, at: kvm_ioctl_create_device+0x174/0x5b0
[  328.204048]  #1: ffff0800e78800b8 (&vcpu->mutex){+.+.}-{3:3}, at: lock_all_vcpus+0x16c/0x2a0
[  328.212521]  #2: ffff07ffeee51e98 (&vcpu->mutex){+.+.}-{3:3}, at: lock_all_vcpus+0x16c/0x2a0
[  328.220991]  #3: ffff0800dc7d80b8 (&vcpu->mutex){+.+.}-{3:3}, at: lock_all_vcpus+0x16c/0x2a0
[  328.229463]  #4: ffff07ffe0c980b8 (&vcpu->mutex){+.+.}-{3:3}, at: lock_all_vcpus+0x16c/0x2a0
[  328.237934]  #5: ffff0800a3883c78 (&vcpu->mutex){+.+.}-{3:3}, at: lock_all_vcpus+0x16c/0x2a0
[  328.246405]  #6: ffff07fffbe480b8 (&vcpu->mutex){+.+.}-{3:3}, at: lock_all_vcpus+0x16c/0x2a0

Luckily, in all instances that require locking all vCPUs, the
'kvm->lock' is taken a priori, and that fact makes it possible to use
the little known feature of lockdep, called a 'nest_lock', to avoid this
warning and subsequent lockdep self-disablement.

The action of 'nested lock' being provided to lockdep's lock_acquire(),
causes the lockdep to detect that the top of the held lock stack contains
a lock of the same class and then increment its reference counter instead
of pushing a new held_lock item onto that stack.

See __lock_acquire for more information.

Signed-off-by: Maxim Levitsky <mlevitsk at redhat.com>
---
 include/linux/mutex.h  | 15 +++++++++++++++
 kernel/locking/mutex.c | 14 +++++++++++---
 2 files changed, 26 insertions(+), 3 deletions(-)

diff --git a/include/linux/mutex.h b/include/linux/mutex.h
index 2143d05116be..da4518cfd59c 100644
--- a/include/linux/mutex.h
+++ b/include/linux/mutex.h
@@ -193,7 +193,22 @@ extern void mutex_lock_io(struct mutex *lock);
  *
  * Returns 1 if the mutex has been acquired successfully, and 0 on contention.
  */
+
+#ifdef CONFIG_DEBUG_LOCK_ALLOC
+extern int _mutex_trylock_nest_lock(struct mutex *lock, struct lockdep_map *nest_lock);
+
+#define mutex_trylock_nest_lock(lock, nest_lock)		\
+(								\
+	typecheck(struct lockdep_map *, &(nest_lock)->dep_map),	\
+	_mutex_trylock_nest_lock(lock, &(nest_lock)->dep_map)	\
+)
+
+#define mutex_trylock(lock) _mutex_trylock_nest_lock(lock, NULL)
+#else
 extern int mutex_trylock(struct mutex *lock);
+#define mutex_trylock_nest_lock(lock, nest_lock) mutex_trylock(lock)
+#endif
+
 extern void mutex_unlock(struct mutex *lock);
 
 extern int atomic_dec_and_mutex_lock(atomic_t *cnt, struct mutex *lock);
diff --git a/kernel/locking/mutex.c b/kernel/locking/mutex.c
index 555e2b3a665a..c75a838d3bae 100644
--- a/kernel/locking/mutex.c
+++ b/kernel/locking/mutex.c
@@ -1062,6 +1062,7 @@ __ww_mutex_lock_interruptible_slowpath(struct ww_mutex *lock,
 
 #endif
 
+#ifndef CONFIG_DEBUG_LOCK_ALLOC
 /**
  * mutex_trylock - try to acquire the mutex, without waiting
  * @lock: the mutex to be acquired
@@ -1077,18 +1078,25 @@ __ww_mutex_lock_interruptible_slowpath(struct ww_mutex *lock,
  * mutex must be released by the same task that acquired it.
  */
 int __sched mutex_trylock(struct mutex *lock)
+{
+	MUTEX_WARN_ON(lock->magic != lock);
+	return __mutex_trylock(lock);
+}
+EXPORT_SYMBOL(mutex_trylock);
+#else
+int __sched _mutex_trylock_nest_lock(struct mutex *lock, struct lockdep_map *nest_lock)
 {
 	bool locked;
 
 	MUTEX_WARN_ON(lock->magic != lock);
-
 	locked = __mutex_trylock(lock);
 	if (locked)
-		mutex_acquire(&lock->dep_map, 0, 1, _RET_IP_);
+		mutex_acquire_nest(&lock->dep_map, 0, 1, nest_lock, _RET_IP_);
 
 	return locked;
 }
-EXPORT_SYMBOL(mutex_trylock);
+EXPORT_SYMBOL(_mutex_trylock_nest_lock);
+#endif
 
 #ifndef CONFIG_DEBUG_LOCK_ALLOC
 int __sched
-- 
2.46.0