On 1/13/23 18:57, Boqun Feng wrote:
Lock scenario print is always a weak spot of lockdep splats. Improvement
can be made if we rework the dependency search and the error printing.
However without touching the graph search, we can improve a little for
the circular deadlock case, since we have the to-be-added lock
dependency, and know whether these two locks are read/write/sync.
In order to know whether a held_lock is sync or not, a bit was
"stolen" from ->references, which reduce our limit for the same lock
class nesting from 2^12 to 2^11, and it should still be good enough.
Besides, since we now have bit in held_lock for sync, we don't need the
"hardirqoffs being 1" trick, and also we can avoid the __lock_release()
if we jump out of __lock_acquire() before the held_lock stored.
With these changes, a deadlock case evolved with read lock and sync gets
a better print-out from:
[...] Possible unsafe locking scenario:
[...]
[...] CPU0 CPU1
[...] ---- ----
[...] lock(srcuA);
[...] lock(srcuB);
[...] lock(srcuA);
[...] lock(srcuB);
to
[...] Possible unsafe locking scenario:
[...]
[...] CPU0 CPU1
[...] ---- ----
[...] rlock(srcuA);
[...] lock(srcuB);
[...] lock(srcuA);
[...] sync(srcuB);
Signed-off-by: Boqun Feng <boqun.feng@xxxxxxxxx>
---
include/linux/lockdep.h | 3 ++-
kernel/locking/lockdep.c | 48 ++++++++++++++++++++++++++--------------
2 files changed, 34 insertions(+), 17 deletions(-)
diff --git a/include/linux/lockdep.h b/include/linux/lockdep.h
index ba09df6a0872..febd7ecc225c 100644
--- a/include/linux/lockdep.h
+++ b/include/linux/lockdep.h
@@ -134,7 +134,8 @@ struct held_lock {
unsigned int read:2; /* see lock_acquire() comment */
unsigned int check:1; /* see lock_acquire() comment */
unsigned int hardirqs_off:1;
- unsigned int references:12; /* 32 bits */
+ unsigned int sync:1;
+ unsigned int references:11; /* 32 bits */
unsigned int pin_count;
};
diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
index cffa026a765f..4031d87f6829 100644
--- a/kernel/locking/lockdep.c
+++ b/kernel/locking/lockdep.c
@@ -1880,6 +1880,8 @@ print_circular_lock_scenario(struct held_lock *src,
struct lock_class *source = hlock_class(src);
struct lock_class *target = hlock_class(tgt);
struct lock_class *parent = prt->class;
+ int src_read = src->read;
+ int tgt_read = tgt->read;
/*
* A direct locking problem where unsafe_class lock is taken
@@ -1907,7 +1909,10 @@ print_circular_lock_scenario(struct held_lock *src,
printk(" Possible unsafe locking scenario:\n\n");
printk(" CPU0 CPU1\n");
printk(" ---- ----\n");
- printk(" lock(");
+ if (tgt_read != 0)
+ printk(" rlock(");
+ else
+ printk(" lock(");
__print_lock_name(target);
printk(KERN_CONT ");\n");
printk(" lock(");
@@ -1916,7 +1921,12 @@ print_circular_lock_scenario(struct held_lock *src,
printk(" lock(");
__print_lock_name(target);
printk(KERN_CONT ");\n");
- printk(" lock(");
+ if (src_read != 0)
+ printk(" rlock(");
+ else if (src->sync)
+ printk(" sync(");
+ else
+ printk(" lock(");
__print_lock_name(source);
printk(KERN_CONT ");\n");
printk("\n *** DEADLOCK ***\n\n");
src can be sync() but not the target. Is there a reason why that is the
case?
@@ -4530,7 +4540,13 @@ mark_usage(struct task_struct *curr, struct held_lock *hlock, int check)
return 0;
}
}
- if (!hlock->hardirqs_off) {
+
+ /*
+ * For lock_sync(), don't mark the ENABLED usage, since lock_sync()
+ * creates no critical section and no extra dependency can be introduced
+ * by interrupts
+ */
+ if (!hlock->hardirqs_off && !hlock->sync) {
if (hlock->read) {
if (!mark_lock(curr, hlock,
LOCK_ENABLED_HARDIRQ_READ))
@@ -4909,7 +4925,7 @@ static int __lock_is_held(const struct lockdep_map *lock, int read);
static int __lock_acquire(struct lockdep_map *lock, unsigned int subclass,
int trylock, int read, int check, int hardirqs_off,
struct lockdep_map *nest_lock, unsigned long ip,
- int references, int pin_count)
+ int references, int pin_count, int sync)
{
struct task_struct *curr = current;
struct lock_class *class = NULL;
@@ -4960,7 +4976,8 @@ static int __lock_acquire(struct lockdep_map *lock, unsigned int subclass,
class_idx = class - lock_classes;
- if (depth) { /* we're holding locks */
+ if (depth && !sync) {
+ /* we're holding locks and the new held lock is not a sync */
hlock = curr->held_locks + depth - 1;
if (hlock->class_idx == class_idx && nest_lock) {
if (!references)
@@ -4994,6 +5011,7 @@ static int __lock_acquire(struct lockdep_map *lock, unsigned int subclass,
hlock->trylock = trylock;
hlock->read = read;
hlock->check = check;
+ hlock->sync = !!sync;
hlock->hardirqs_off = !!hardirqs_off;
hlock->references = references;
#ifdef CONFIG_LOCK_STAT
@@ -5055,6 +5073,10 @@ static int __lock_acquire(struct lockdep_map *lock, unsigned int subclass,
if (!validate_chain(curr, hlock, chain_head, chain_key))
return 0;
+ /* For lock_sync(), we are done here since no actual critical section */
+ if (hlock->sync)
+ return 1;
+
curr->curr_chain_key = chain_key;
curr->lockdep_depth++;
check_chain_key(curr);
Even with sync, there is still a corresponding lock_acquire() and
lock_release(), you can't exit here without increasing lockdep_depth.
That can cause underflow.
Cheers,
Longman
