From 23870f1227680d2aacff6f79c3ab2222bd04e86e Mon Sep 17 00:00:00 2001 From: "peterz@infradead.org" Date: Wed, 2 Sep 2020 18:03:23 +0200 Subject: locking/lockdep: Fix "USED" <- "IN-NMI" inversions During the LPC RCU BoF Paul asked how come the "USED" <- "IN-NMI" detector doesn't trip over rcu_read_lock()'s lockdep annotation. Looking into this I found a very embarrasing typo in verify_lock_unused(): - if (!(class->usage_mask & LOCK_USED)) + if (!(class->usage_mask & LOCKF_USED)) fixing that will indeed cause rcu_read_lock() to insta-splat :/ The above typo means that instead of testing for: 0x100 (1 << LOCK_USED), we test for 8 (LOCK_USED), which corresponds to (1 << LOCK_ENABLED_HARDIRQ). So instead of testing for _any_ used lock, it will only match any lock used with interrupts enabled. The rcu_read_lock() annotation uses .check=0, which means it will not set any of the interrupt bits and will thus never match. In order to properly fix the situation and allow rcu_read_lock() to correctly work, split LOCK_USED into LOCK_USED and LOCK_USED_READ and by having .read users set USED_READ and test USED, pure read-recursive locks are permitted. Fixes: f6f48e180404 ("lockdep: Teach lockdep about "USED" <- "IN-NMI" inversions") Signed-off-by: Peter Zijlstra (Intel) Signed-off-by: Ingo Molnar Tested-by: Masami Hiramatsu Acked-by: Paul E. McKenney Link: https://lore.kernel.org/r/20200902160323.GK1362448@hirez.programming.kicks-ass.net --- kernel/locking/lockdep.c | 35 +++++++++++++++++++++++++++++------ kernel/locking/lockdep_internals.h | 2 ++ 2 files changed, 31 insertions(+), 6 deletions(-) (limited to 'kernel') diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c index 54b74fabf40c..2facbbd146ec 100644 --- a/kernel/locking/lockdep.c +++ b/kernel/locking/lockdep.c @@ -3969,13 +3969,18 @@ static int separate_irq_context(struct task_struct *curr, static int mark_lock(struct task_struct *curr, struct held_lock *this, enum lock_usage_bit new_bit) { - unsigned int new_mask = 1 << new_bit, ret = 1; + unsigned int old_mask, new_mask, ret = 1; if (new_bit >= LOCK_USAGE_STATES) { DEBUG_LOCKS_WARN_ON(1); return 0; } + if (new_bit == LOCK_USED && this->read) + new_bit = LOCK_USED_READ; + + new_mask = 1 << new_bit; + /* * If already set then do not dirty the cacheline, * nor do any checks: @@ -3988,13 +3993,22 @@ static int mark_lock(struct task_struct *curr, struct held_lock *this, /* * Make sure we didn't race: */ - if (unlikely(hlock_class(this)->usage_mask & new_mask)) { - graph_unlock(); - return 1; - } + if (unlikely(hlock_class(this)->usage_mask & new_mask)) + goto unlock; + old_mask = hlock_class(this)->usage_mask; hlock_class(this)->usage_mask |= new_mask; + /* + * Save one usage_traces[] entry and map both LOCK_USED and + * LOCK_USED_READ onto the same entry. + */ + if (new_bit == LOCK_USED || new_bit == LOCK_USED_READ) { + if (old_mask & (LOCKF_USED | LOCKF_USED_READ)) + goto unlock; + new_bit = LOCK_USED; + } + if (!(hlock_class(this)->usage_traces[new_bit] = save_trace())) return 0; @@ -4008,6 +4022,7 @@ static int mark_lock(struct task_struct *curr, struct held_lock *this, return 0; } +unlock: graph_unlock(); /* @@ -4942,12 +4957,20 @@ static void verify_lock_unused(struct lockdep_map *lock, struct held_lock *hlock { #ifdef CONFIG_PROVE_LOCKING struct lock_class *class = look_up_lock_class(lock, subclass); + unsigned long mask = LOCKF_USED; /* if it doesn't have a class (yet), it certainly hasn't been used yet */ if (!class) return; - if (!(class->usage_mask & LOCK_USED)) + /* + * READ locks only conflict with USED, such that if we only ever use + * READ locks, there is no deadlock possible -- RCU. + */ + if (!hlock->read) + mask |= LOCKF_USED_READ; + + if (!(class->usage_mask & mask)) return; hlock->class_idx = class - lock_classes; diff --git a/kernel/locking/lockdep_internals.h b/kernel/locking/lockdep_internals.h index baca699b94e9..b0be1560ed17 100644 --- a/kernel/locking/lockdep_internals.h +++ b/kernel/locking/lockdep_internals.h @@ -19,6 +19,7 @@ enum lock_usage_bit { #include "lockdep_states.h" #undef LOCKDEP_STATE LOCK_USED, + LOCK_USED_READ, LOCK_USAGE_STATES }; @@ -40,6 +41,7 @@ enum { #include "lockdep_states.h" #undef LOCKDEP_STATE __LOCKF(USED) + __LOCKF(USED_READ) }; #define LOCKDEP_STATE(__STATE) LOCKF_ENABLED_##__STATE | -- cgit From e6b1a44eccfcab5e5e280be376f65478c3b2c7a2 Mon Sep 17 00:00:00 2001 From: Hou Tao Date: Tue, 15 Sep 2020 22:07:50 +0800 Subject: locking/percpu-rwsem: Use this_cpu_{inc,dec}() for read_count The __this_cpu*() accessors are (in general) IRQ-unsafe which, given that percpu-rwsem is a blocking primitive, should be just fine. However, file_end_write() is used from IRQ context and will cause load-store issues on architectures where the per-cpu accessors are not natively irq-safe. Fix it by using the IRQ-safe this_cpu_*() for operations on read_count. This will generate more expensive code on a number of platforms, which might cause a performance regression for some of the other percpu-rwsem users. If any such is reported, we can consider alternative solutions. Fixes: 70fe2f48152e ("aio: fix freeze protection of aio writes") Signed-off-by: Hou Tao Signed-off-by: Peter Zijlstra (Intel) Acked-by: Will Deacon Acked-by: Oleg Nesterov Link: https://lkml.kernel.org/r/20200915140750.137881-1-houtao1@huawei.com --- kernel/locking/percpu-rwsem.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/locking/percpu-rwsem.c b/kernel/locking/percpu-rwsem.c index 8bbafe3e5203..70a32a576f3f 100644 --- a/kernel/locking/percpu-rwsem.c +++ b/kernel/locking/percpu-rwsem.c @@ -45,7 +45,7 @@ EXPORT_SYMBOL_GPL(percpu_free_rwsem); static bool __percpu_down_read_trylock(struct percpu_rw_semaphore *sem) { - __this_cpu_inc(*sem->read_count); + this_cpu_inc(*sem->read_count); /* * Due to having preemption disabled the decrement happens on @@ -71,7 +71,7 @@ static bool __percpu_down_read_trylock(struct percpu_rw_semaphore *sem) if (likely(!atomic_read_acquire(&sem->block))) return true; - __this_cpu_dec(*sem->read_count); + this_cpu_dec(*sem->read_count); /* Prod writer to re-evaluate readers_active_check() */ rcuwait_wake_up(&sem->writer); -- cgit