Magliano Andrea <[email protected]> added the comment: To revise my comment:
TAILQ_EMPTY was already in camisr_runqueue(), so 0001-Optimize-isr-routine-from-FreeBSD-proved-to-work.patch has really nothing to do with locking issues, just a drop in optimization from FreeBSD Nevertheless 0002-TAILQ_EMPTY-now-done-in-camisr_runqueue-so-useless-h.patch still makes sense to me: in camisr_runqueue() 'while ((ccb_h = TAILQ_FIRST(&sim->sim_doneq)) != NULL)' lets TAILQ_FIRST != NULL make the same job as TAILQ_EMPTY, and here the queue macro is do protected by a spinlock ---------- status: unread -> chatting _____________________________________________________ DragonFly issue tracker <[email protected]> <http://bugs.dragonflybsd.org/issue2092> _____________________________________________________
