diff options
| author | Kent Overstreet <kent.overstreet@linux.dev> | 2023-10-24 20:46:58 +0200 |
|---|---|---|
| committer | Kent Overstreet <kent.overstreet@linux.dev> | 2023-10-31 02:48:22 +0100 |
| commit | ee526b88caaa4b4182144bf2576af2c3b1e9c759 (patch) | |
| tree | d25c1a6b69d86c8a7ed518334dbcc13a842b37e4 /lib/closure.c | |
| parent | closures: Better memory barriers (diff) | |
| download | linux-ee526b88caaa4b4182144bf2576af2c3b1e9c759.tar.xz linux-ee526b88caaa4b4182144bf2576af2c3b1e9c759.zip | |
closures: Fix race in closure_sync()
As pointed out by Linus, closure_sync() was racy; we could skip blocking
immediately after a get() and a put(), but then that would skip any
barrier corresponding to the other thread's put() barrier.
To fix this, always do the full __closure_sync() sequence whenever any
get() has happened and the closure might have been used by other
threads.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Diffstat (limited to 'lib/closure.c')
| -rw-r--r-- | lib/closure.c | 3 |
1 files changed, 3 insertions, 0 deletions
diff --git a/lib/closure.c b/lib/closure.c index 501dfa277b59..f86c9eeafb35 100644 --- a/lib/closure.c +++ b/lib/closure.c @@ -23,6 +23,8 @@ static inline void closure_put_after_sub(struct closure *cl, int flags) if (!r) { smp_acquire__after_ctrl_dep(); + cl->closure_get_happened = false; + if (cl->fn && !(flags & CLOSURE_DESTRUCTOR)) { atomic_set(&cl->remaining, CLOSURE_REMAINING_INITIALIZER); @@ -92,6 +94,7 @@ bool closure_wait(struct closure_waitlist *waitlist, struct closure *cl) if (atomic_read(&cl->remaining) & CLOSURE_WAITING) return false; + cl->closure_get_happened = true; closure_set_waiting(cl, _RET_IP_); atomic_add(CLOSURE_WAITING + 1, &cl->remaining); llist_add(&cl->list, &waitlist->list); |