History

locking/lockdep: Untangle xhlock history save/restore from task independence

Where XHLOCK_{SOFT,HARD} are save/restore points in the xhlocks[] to ensure the temporal IRQ events don't interact with task state, the XHLOCK_PROC is a fundament different beast that just happens to share the interface. The purpose of XHLOCK_PROC is to annotate independent execution inside one task. For example workqueues, each work should appear to run in its own 'pristine' 'task'. Remove XHLOCK_PROC in favour of its own interface to avoid confusion.

Peter Zijlstra committed 9 years ago

Commit f52be571 parent 7b3d61c4 files · +48 −51Browse files

Showing the changes to kernel/workqueue.c (1 of 4 files). Show all

Unified · Split

kernel/workqueue.c+4−5
@@ -2094,8 +2094,8 @@ __acquires(&pool->lock) · 2093 unmodified lines
2094 lock_map_acquire(&pwq->wq->lockdep_map);2094 lock_map_acquire(&pwq->wq->lockdep_map);
2095 lock_map_acquire(&lockdep_map);2095 lock_map_acquire(&lockdep_map);
2096 /*2096 /*
2097 * Strictly speaking we should do start(PROC) without holding any2097 * Strictly speaking we should mark the invariant state without holding
2098 * locks, that is, before these two lock_map_acquire()'s.2098 * any locks, that is, before these two lock_map_acquire()'s.
2099 *2099 *
2100 * However, that would result in:2100 * However, that would result in:
2101 *2101 *
@@ -2107,14 +2107,14 @@ __acquires(&pool->lock) · 5 unmodified lines
2107 * Which would create W1->C->W1 dependencies, even though there is no2107 * Which would create W1->C->W1 dependencies, even though there is no
2108 * actual deadlock possible. There are two solutions, using a2108 * actual deadlock possible. There are two solutions, using a
2109 * read-recursive acquire on the work(queue) 'locks', but this will then2109 * read-recursive acquire on the work(queue) 'locks', but this will then
2110 * hit the lockdep limitation on recursive locks, or simly discard2110 * hit the lockdep limitation on recursive locks, or simply discard
2111 * these locks.2111 * these locks.
2112 *2112 *
2113 * AFAICT there is no possible deadlock scenario between the2113 * AFAICT there is no possible deadlock scenario between the
2114 * flush_work() and complete() primitives (except for single-threaded2114 * flush_work() and complete() primitives (except for single-threaded
2115 * workqueues), so hiding them isn't a problem.2115 * workqueues), so hiding them isn't a problem.
2116 */2116 */
2117 crossrelease_hist_start(XHLOCK_PROC, true);2117 lockdep_invariant_state(true);
2118 trace_workqueue_execute_start(work);2118 trace_workqueue_execute_start(work);
2119 worker->current_func(work);2119 worker->current_func(work);
2120 /*2120 /*
@@ -2122,7 +2122,6 @@ __acquires(&pool->lock) · 1 unmodified line
2122 * point will only record its address.2122 * point will only record its address.
2123 */2123 */
2124 trace_workqueue_execute_end(work);2124 trace_workqueue_execute_end(work);
2125 crossrelease_hist_end(XHLOCK_PROC);
2126 lock_map_release(&lockdep_map);2125 lock_map_release(&lockdep_map);
2127 lock_map_release(&pwq->wq->lockdep_map);2126 lock_map_release(&pwq->wq->lockdep_map);
2128 2127