Commit 23870f12 authored by peterz@infradead.org's avatar peterz@infradead.org Committed by Ingo Molnar

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: f6f48e18 ("lockdep: Teach lockdep about "USED" <- "IN-NMI" inversions")
Signed-off-by: default avatarPeter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: default avatarIngo Molnar <mingo@kernel.org>
Tested-by: default avatarMasami Hiramatsu <mhiramat@kernel.org>
Acked-by: default avatarPaul E. McKenney <paulmck@kernel.org>
Link: https://lore.kernel.org/r/20200902160323.GK1362448@hirez.programming.kicks-ass.net
parent fc3abb53
......@@ -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;
......
......@@ -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 |
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment