tjh.dev / kernel
Linux kernel mirror (for testing) git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel os linux
1
fork

Configure Feed

Select the types of activity you want to include in your feed.

lockding/lockdep: Avoid to find wrong lock dep path in check_irq_usage()

In the step #3 of check_irq_usage(), we seach backwards to find a lock
whose usage conflicts the usage of @target_entry1 on safe/unsafe.
However, we should only keep the irq-unsafe usage of @target_entry1 into
consideration, because it could be a case where a lock is hardirq-unsafe
but soft-safe, and in check_irq_usage() we find it because its
hardirq-unsafe could result into a hardirq-safe-unsafe deadlock, but
currently since we don't filter out the other usage bits, so we may find
a lock dependency path softirq-unsafe -> softirq-safe, which in fact
doesn't cause a deadlock. And this may cause misleading lockdep splats.

Fix this by only keeping LOCKF_ENABLED_IRQ_ALL bits when we try the
backwards search.

Reported-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20210618170110.3699115-4-boqun.feng@gmail.com

authored by

Boqun Feng and committed by
Peter Zijlstra 7b1f8c61 d4c157c7

+11 -1
+11 -1
kernel/locking/lockdep.c
··· 2768 2768 * Step 3: we found a bad match! Now retrieve a lock from the backward 2769 2769 * list whose usage mask matches the exclusive usage mask from the 2770 2770 * lock found on the forward list. 2771 + * 2772 + * Note, we should only keep the LOCKF_ENABLED_IRQ_ALL bits, considering 2773 + * the follow case: 2774 + * 2775 + * When trying to add A -> B to the graph, we find that there is a 2776 + * hardirq-safe L, that L -> ... -> A, and another hardirq-unsafe M, 2777 + * that B -> ... -> M. However M is **softirq-safe**, if we use exact 2778 + * invert bits of M's usage_mask, we will find another lock N that is 2779 + * **softirq-unsafe** and N -> ... -> A, however N -> .. -> M will not 2780 + * cause a inversion deadlock. 2771 2781 */ 2772 - backward_mask = original_mask(target_entry1->class->usage_mask); 2782 + backward_mask = original_mask(target_entry1->class->usage_mask & LOCKF_ENABLED_IRQ_ALL); 2773 2783 2774 2784 ret = find_usage_backwards(&this, backward_mask, &target_entry); 2775 2785 if (bfs_error(ret)) {