On 20.03.2019 21:52, Andrew Morton wrote: > On Wed, 20 Mar 2019 14:19:27 +0300 Kirill Tkhai <ktkhai@xxxxxxxxxxxxx> wrote: > >> 1)Spinlock must be locked in any case, so assert_spin_locked() >> are moved above the switch; > > This isn't true. When the ->isolate() handler xfs_buftarg_wait_rele() > (at least) returns LRU_SKIP, the lock is not held. Oh, I missed that :( Thanks for pointing. Kirill
