The patch titled
PI-futex: -V3
has been removed from the -mm tree. Its filename is
pi-futex-v3.patch
This patch was probably dropped from -mm because
it has now been merged into a subsystem tree or
into Linus's tree, or because it was folded into
its parent patch in the -mm tree.
------------------------------------------------------
Subject: PI-futex: -V3
From: Ingo Molnar <mingo@xxxxxxx>
this is version -V3 of the PI-futex patchset (ontop of current -mm plus
the -V2 delta patch). Changes since -V2:
- do the -> set_current_state() cleanups for real
- make plist_add() more readable
- whitespace cleanups in kernel/futex.c
- further cleanups in rtmutex_common.h
- fix cornercase race in the PI-locking code (it triggered in the -rt tree)
Signed-off-by: Ingo Molnar <mingo@xxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxx>
---
kernel/rtmutex.c | 245 ++++++++++++++++++++------------------
kernel/rtmutex_common.h | 6
lib/plist.c | 3
futex.c | 0
4 files changed, 134 insertions(+), 120 deletions(-)
diff -puN kernel/futex.c~pi-futex-v3 kernel/futex.c
diff -puN kernel/rtmutex.c~pi-futex-v3 kernel/rtmutex.c
--- devel/kernel/rtmutex.c~pi-futex-v3 2006-03-26 15:35:35.000000000 -0800
+++ devel-akpm/kernel/rtmutex.c 2006-03-26 15:35:35.000000000 -0800
@@ -173,11 +173,12 @@ static int lock_pi_chain(struct rt_mutex
struct task_struct *owner;
struct rt_mutex *nextlock, *lock = act_lock;
struct rt_mutex_waiter *nextwaiter;
+ int deadlock_detect;
/*
* Debugging might turn deadlock detection on, unconditionally:
*/
- detect_deadlock = debug_rt_mutex_detect_deadlock(detect_deadlock);
+ deadlock_detect = debug_rt_mutex_detect_deadlock(detect_deadlock);
for (;;) {
owner = rt_mutex_owner(lock);
@@ -185,7 +186,7 @@ static int lock_pi_chain(struct rt_mutex
/* Check for circular dependencies */
if (unlikely(owner->pi_locked_by == current)) {
debug_rt_mutex_deadlock(detect_deadlock, waiter, lock);
- return detect_deadlock ? -EDEADLK : 0;
+ return detect_deadlock ? -EDEADLK : 1;
}
while (!spin_trylock(&owner->pi_lock)) {
@@ -211,7 +212,7 @@ static int lock_pi_chain(struct rt_mutex
/* End of chain? */
if (!nextwaiter)
- return 0;
+ return 1;
nextlock = nextwaiter->lock;
@@ -223,7 +224,7 @@ static int lock_pi_chain(struct rt_mutex
list_del_init(&owner->pi_lock_chain);
owner->pi_locked_by = NULL;
spin_unlock(&owner->pi_lock);
- return detect_deadlock ? -EDEADLK : 0;
+ return detect_deadlock ? -EDEADLK : 1;
}
/* Try to get nextlock->wait_lock: */
@@ -242,7 +243,7 @@ static int lock_pi_chain(struct rt_mutex
* for userspace locks), we have to walk the full chain
* unconditionally.
*/
- if (detect_deadlock)
+ if (deadlock_detect)
continue;
/*
@@ -255,11 +256,11 @@ static int lock_pi_chain(struct rt_mutex
/* If the top waiter has higher priority, stop: */
if (rt_mutex_top_waiter(lock)->list_entry.prio <=
waiter->list_entry.prio)
- return 0;
+ return 1;
} else {
/* If nextwaiter is not the top waiter, stop: */
if (rt_mutex_top_waiter(lock) != nextwaiter)
- return 0;
+ return 1;
}
}
}
@@ -340,103 +341,6 @@ static void adjust_pi_chain(struct rt_mu
}
/*
- * Task blocks on lock.
- *
- * Prepare waiter and potentially propagate our priority into the pi chain.
- *
- * This must be called with lock->wait_lock held.
- */
-static int task_blocks_on_rt_mutex(struct rt_mutex *lock,
- struct rt_mutex_waiter *waiter,
- int detect_deadlock __IP_DECL__)
-{
- int res = 0;
- struct rt_mutex_waiter *top_waiter = waiter;
- LIST_HEAD(lock_chain);
-
- waiter->task = current;
- waiter->lock = lock;
- debug_rt_mutex_reset_waiter(waiter);
-
- spin_lock(¤t->pi_lock);
- current->pi_locked_by = current;
- plist_node_init(&waiter->list_entry, current->prio);
- plist_node_init(&waiter->pi_list_entry, current->prio);
-
- /* Get the top priority waiter of the lock: */
- if (rt_mutex_has_waiters(lock))
- top_waiter = rt_mutex_top_waiter(lock);
- plist_add(&waiter->list_entry, &lock->wait_list);
-
- current->pi_blocked_on = waiter;
-
- /*
- * Call adjust_prio_chain, when waiter is the new top waiter
- * or when deadlock detection is requested:
- */
- if (waiter != rt_mutex_top_waiter(lock) &&
- !debug_rt_mutex_detect_deadlock(detect_deadlock))
- goto out;
-
- /* Try to lock the full chain: */
- res = lock_pi_chain(lock, waiter, &lock_chain, 1, detect_deadlock);
-
- if (likely(!res))
- adjust_pi_chain(lock, waiter, top_waiter, &lock_chain);
-
- /* Common case: we managed to lock it: */
- if (res != -EBUSY)
- goto out_unlock;
-
- /* Rare case: we hit some other task running a pi chain operation: */
- unlock_pi_chain(&lock_chain);
-
- plist_del(&waiter->list_entry, &lock->wait_list);
- current->pi_blocked_on = NULL;
- current->pi_locked_by = NULL;
- spin_unlock(¤t->pi_lock);
- fixup_rt_mutex_waiters(lock);
-
- spin_unlock(&lock->wait_lock);
-
- spin_lock(&pi_conflicts_lock);
-
- spin_lock(¤t->pi_lock);
- current->pi_locked_by = current;
- spin_lock(&lock->wait_lock);
- if (!rt_mutex_owner(lock)) {
- waiter->task = NULL;
- spin_unlock(&pi_conflicts_lock);
- goto out;
- }
- plist_node_init(&waiter->list_entry, current->prio);
- plist_node_init(&waiter->pi_list_entry, current->prio);
-
- /* Get the top priority waiter of the lock: */
- if (rt_mutex_has_waiters(lock))
- top_waiter = rt_mutex_top_waiter(lock);
- plist_add(&waiter->list_entry, &lock->wait_list);
-
- current->pi_blocked_on = waiter;
-
- /* Lock the full chain: */
- res = lock_pi_chain(lock, waiter, &lock_chain, 0, detect_deadlock);
-
- /* Drop the conflicts lock before adjusting: */
- spin_unlock(&pi_conflicts_lock);
-
- if (likely(!res))
- adjust_pi_chain(lock, waiter, top_waiter, &lock_chain);
-
- out_unlock:
- unlock_pi_chain(&lock_chain);
- out:
- current->pi_locked_by = NULL;
- spin_unlock(¤t->pi_lock);
- return res;
-}
-
-/*
* Optimization: check if we can steal the lock from the
* assigned pending owner [which might not have taken the
* lock yet]:
@@ -542,6 +446,117 @@ static int try_to_take_rt_mutex(struct r
}
/*
+ * Task blocks on lock.
+ *
+ * Prepare waiter and potentially propagate our priority into the pi chain.
+ *
+ * This must be called with lock->wait_lock held.
+ * return values: 1: waiter queued, 0: got the lock,
+ * -EDEADLK: deadlock detected.
+ */
+static int task_blocks_on_rt_mutex(struct rt_mutex *lock,
+ struct rt_mutex_waiter *waiter,
+ int detect_deadlock __IP_DECL__)
+{
+ struct rt_mutex_waiter *top_waiter = waiter;
+ LIST_HEAD(lock_chain);
+ int res = 1;
+
+ waiter->task = current;
+ waiter->lock = lock;
+ debug_rt_mutex_reset_waiter(waiter);
+
+ spin_lock(¤t->pi_lock);
+ current->pi_locked_by = current;
+ plist_node_init(&waiter->list_entry, current->prio);
+ plist_node_init(&waiter->pi_list_entry, current->prio);
+
+ /* Get the top priority waiter of the lock: */
+ if (rt_mutex_has_waiters(lock))
+ top_waiter = rt_mutex_top_waiter(lock);
+ plist_add(&waiter->list_entry, &lock->wait_list);
+
+ current->pi_blocked_on = waiter;
+
+ /*
+ * Call adjust_prio_chain, when waiter is the new top waiter
+ * or when deadlock detection is requested:
+ */
+ if (waiter != rt_mutex_top_waiter(lock) &&
+ !debug_rt_mutex_detect_deadlock(detect_deadlock))
+ goto out_unlock_pi;
+
+ /* Try to lock the full chain: */
+ res = lock_pi_chain(lock, waiter, &lock_chain, 1, detect_deadlock);
+
+ if (likely(res == 1))
+ adjust_pi_chain(lock, waiter, top_waiter, &lock_chain);
+
+ /* Common case: we managed to lock it: */
+ if (res != -EBUSY)
+ goto out_unlock_chain_pi;
+
+ /* Rare case: we hit some other task running a pi chain operation: */
+ unlock_pi_chain(&lock_chain);
+
+ plist_del(&waiter->list_entry, &lock->wait_list);
+ current->pi_blocked_on = NULL;
+ current->pi_locked_by = NULL;
+ spin_unlock(¤t->pi_lock);
+ fixup_rt_mutex_waiters(lock);
+
+ spin_unlock(&lock->wait_lock);
+
+ /*
+ * Here we have dropped all locks, and take the global
+ * pi_conflicts_lock. We have to redo all the work, no
+ * previous information about the lock is valid anymore:
+ */
+ spin_lock(&pi_conflicts_lock);
+
+ spin_lock(&lock->wait_lock);
+ if (try_to_take_rt_mutex(lock __IP__)) {
+ /*
+ * Rare race: against all odds we got the lock.
+ */
+ res = 0;
+ goto out;
+ }
+
+ WARN_ON(!rt_mutex_owner(lock) || rt_mutex_owner(lock) == current);
+
+ spin_lock(¤t->pi_lock);
+ current->pi_locked_by = current;
+
+ plist_node_init(&waiter->list_entry, current->prio);
+ plist_node_init(&waiter->pi_list_entry, current->prio);
+
+ /* Get the top priority waiter of the lock: */
+ if (rt_mutex_has_waiters(lock))
+ top_waiter = rt_mutex_top_waiter(lock);
+ plist_add(&waiter->list_entry, &lock->wait_list);
+
+ current->pi_blocked_on = waiter;
+
+ /* Lock the full chain: */
+ res = lock_pi_chain(lock, waiter, &lock_chain, 0, detect_deadlock);
+
+ /* Drop the conflicts lock before adjusting: */
+ spin_unlock(&pi_conflicts_lock);
+
+ if (likely(res == 1))
+ adjust_pi_chain(lock, waiter, top_waiter, &lock_chain);
+
+ out_unlock_chain_pi:
+ unlock_pi_chain(&lock_chain);
+ out_unlock_pi:
+ current->pi_locked_by = NULL;
+ spin_unlock(¤t->pi_lock);
+ out:
+ return res;
+}
+
+/*
* Wake up the next waiter on the lock.
*
* Remove the top waiter from the current tasks waiter list and from
@@ -641,11 +656,11 @@ static int remove_waiter(struct rt_mutex
spin_lock(&pi_conflicts_lock);
+ spin_lock(&lock->wait_lock);
+
spin_lock(¤t->pi_lock);
current->pi_locked_by = current;
- spin_lock(&lock->wait_lock);
-
/* We might have been woken up: */
if (!waiter->task) {
spin_unlock(&pi_conflicts_lock);
@@ -709,7 +724,7 @@ rt_mutex_slowlock(struct rt_mutex *lock,
BUG_ON(rt_mutex_owner(lock) == current);
- set_task_state(current, state);
+ set_current_state(state);
/* Setup the timer, when timeout != NULL */
if (unlikely(timeout))
@@ -743,10 +758,10 @@ rt_mutex_slowlock(struct rt_mutex *lock,
if (!waiter.task) {
ret = task_blocks_on_rt_mutex(lock, &waiter,
detect_deadlock __IP__);
- if (ret == -EDEADLK)
+ /* got the lock or deadlock: */
+ if (ret == 0 || ret == -EDEADLK)
break;
- if (ret == -EBUSY)
- continue;
+ ret = 0;
}
spin_unlock_irqrestore(&lock->wait_lock, flags);
@@ -757,10 +772,10 @@ rt_mutex_slowlock(struct rt_mutex *lock,
schedule_rt_mutex(lock);
spin_lock_irqsave(&lock->wait_lock, flags);
- set_task_state(current, state);
+ set_current_state(state);
}
- set_task_state(current, TASK_RUNNING);
+ set_current_state(TASK_RUNNING);
if (unlikely(waiter.task))
remove_waiter(lock, &waiter __IP__);
@@ -806,11 +821,9 @@ rt_mutex_slowtrylock(struct rt_mutex *lo
ret = try_to_take_rt_mutex(lock __IP__);
/*
* try_to_take_rt_mutex() sets the lock waiters
- * bit. We might be the only waiter. Check if this
- * needs to be cleaned up.
+ * bit unconditionally. Clean this up.
*/
- if (!ret)
- fixup_rt_mutex_waiters(lock);
+ fixup_rt_mutex_waiters(lock);
}
spin_unlock_irqrestore(&lock->wait_lock, flags);
diff -puN kernel/rtmutex_common.h~pi-futex-v3 kernel/rtmutex_common.h
--- devel/kernel/rtmutex_common.h~pi-futex-v3 2006-03-26 15:35:35.000000000 -0800
+++ devel-akpm/kernel/rtmutex_common.h 2006-03-26 15:35:35.000000000 -0800
@@ -98,18 +98,18 @@ task_top_pi_waiter(struct task_struct *p
static inline struct task_struct *rt_mutex_owner(struct rt_mutex *lock)
{
return (struct task_struct *)
- ((unsigned long)((lock)->owner) & ~RT_MUTEX_OWNER_MASKALL);
+ ((unsigned long)lock->owner & ~RT_MUTEX_OWNER_MASKALL);
}
static inline struct task_struct *rt_mutex_real_owner(struct rt_mutex *lock)
{
return (struct task_struct *)
- ((unsigned long)((lock)->owner) & ~RT_MUTEX_HAS_WAITERS);
+ ((unsigned long)lock->owner & ~RT_MUTEX_HAS_WAITERS);
}
static inline unsigned long rt_mutex_owner_pending(struct rt_mutex *lock)
{
- return ((unsigned long)((lock)->owner) & RT_MUTEX_OWNER_PENDING);
+ return (unsigned long)lock->owner & RT_MUTEX_OWNER_PENDING;
}
/*
diff -puN lib/plist.c~pi-futex-v3 lib/plist.c
--- devel/lib/plist.c~pi-futex-v3 2006-03-26 15:35:35.000000000 -0800
+++ devel-akpm/lib/plist.c 2006-03-26 15:35:35.000000000 -0800
@@ -38,7 +38,7 @@ void plist_add(struct plist_node *node,
WARN_ON(!plist_node_empty(node));
- list_for_each_entry(iter, &head->prio_list, plist.prio_list)
+ list_for_each_entry(iter, &head->prio_list, plist.prio_list) {
if (node->prio < iter->prio)
goto lt_prio;
else if (node->prio == iter->prio) {
@@ -46,6 +46,7 @@ void plist_add(struct plist_node *node,
struct plist_node, plist.prio_list);
goto eq_prio;
}
+ }
lt_prio:
list_add_tail(&node->plist.prio_list, &iter->plist.prio_list);
_
Patches currently in -mm which might be from mingo@xxxxxxx are
git-acpi.patch
fix-drivers-mfd-ucb1x00-corec-irq-probing-bug.patch
git-infiniband.patch
git-netdev-all.patch
fix-for-serial-uart-lockup.patch
swapless-pm-add-r-w-migration-entries-fix.patch
i386-break-out-of-recursion-in-stackframe-walk.patch
x86-re-enable-generic-numa.patch
vdso-randomize-the-i386-vdso-by-moving-it-into-a-vma.patch
vdso-randomize-the-i386-vdso-by-moving-it-into-a-vma-tidy.patch
vdso-randomize-the-i386-vdso-by-moving-it-into-a-vma-arch_vma_name-fix.patch
work-around-ppc64-bootup-bug-by-making-mutex-debugging-save-restore-irqs.patch
kernel-kernel-cpuc-to-mutexes.patch
cond-resched-might-sleep-fix.patch
define-__raw_get_cpu_var-and-use-it.patch
ide-cd-end-of-media-error-fix.patch
spin-rwlock-init-cleanups.patch
time-clocksource-infrastructure.patch
sched-comment-bitmap-size-accounting.patch
sched-fix-interactive-ceiling-code.patch
sched-implement-smpnice.patch
sched-protect-calculation-of-max_pull-from-integer-wrap.patch
sched-store-weighted-load-on-up.patch
sched-add-discrete-weighted-cpu-load-function.patch
sched-prevent-high-load-weight-tasks-suppressing-balancing.patch
sched-improve-stability-of-smpnice-load-balancing.patch
sched-improve-smpnice-load-balancing-when-load-per-task.patch
smpnice-dont-consider-sched-groups-which-are-lightly-loaded-for-balancing.patch
smpnice-dont-consider-sched-groups-which-are-lightly-loaded-for-balancing-fix.patch
sched-modify-move_tasks-to-improve-load-balancing-outcomes.patch
sched-avoid-unnecessarily-moving-highest-priority-task-move_tasks.patch
sched-avoid-unnecessarily-moving-highest-priority-task-move_tasks-fix-2.patch
sched_domain-handle-kmalloc-failure.patch
sched_domain-handle-kmalloc-failure-fix.patch
sched_domain-dont-use-gfp_atomic.patch
sched_domain-use-kmalloc_node.patch
sched_domain-allocate-sched_group-structures-dynamically.patch
sched-add-above-background-load-function.patch
mm-implement-swap-prefetching-fix.patch
pi-futex-v3.patch
pi-futex-patchset-v4.patch
pi-futex-patchset-v4-update.patch
pi-futex-patchset-v4-fix.patch
rtmutex-remove-buggy-bug_on-in-pi-boosting-code.patch
futex-pi-enforce-waiter-bit-when-owner-died-is-detected.patch
rtmutex-debug-printk-correct-task-information.patch
futex-pi-make-use-of-restart_block-when-interrupted.patch
document-futex-pi-design.patch
futex_requeue-optimization.patch
reiser4.patch
reiser4-spin-rwlock-init-cleanups.patch
genirq-rename-desc-handler-to-desc-chip.patch
genirq-rename-desc-handler-to-desc-chip-power-fix.patch
genirq-rename-desc-handler-to-desc-chip-ia64-fix.patch
genirq-rename-desc-handler-to-desc-chip-ia64-fix-2.patch
genirq-sem2mutex-probe_sem-probing_active.patch
genirq-cleanup-merge-irq_affinity-into-irq_desc.patch
genirq-cleanup-remove-irq_descp.patch
genirq-cleanup-remove-fastcall.patch
genirq-cleanup-misc-code-cleanups.patch
genirq-cleanup-reduce-irq_desc_t-use-mark-it-obsolete.patch
genirq-cleanup-include-linux-irqh.patch
genirq-cleanup-merge-irq_dir-smp_affinity_entry-into-irq_desc.patch
genirq-cleanup-merge-pending_irq_cpumask-into-irq_desc.patch
genirq-cleanup-turn-arch_has_irq_per_cpu-into-config_irq_per_cpu.patch
genirq-debug-better-debug-printout-in-enable_irq.patch
genirq-add-retrigger-irq-op-to-consolidate-hw_irq_resend.patch
genirq-doc-comment-include-linux-irqh-structures.patch
genirq-doc-handle_irq_event-and-__do_irq-comments.patch
genirq-cleanup-no_irq_type-cleanups.patch
genirq-doc-add-design-documentation.patch
genirq-add-genirq-sw-irq-retrigger.patch
genirq-add-irq_noprobe-support.patch
genirq-add-irq_norequest-support.patch
genirq-add-irq_noautoen-support.patch
genirq-update-copyrights.patch
genirq-core.patch
genirq-add-irq-chip-support.patch
genirq-add-handle_bad_irq.patch
genirq-add-irq-wake-power-management-support.patch
genirq-add-sa_trigger-support.patch
genirq-cleanup-no_irq_type-no_irq_chip-rename.patch
genirq-convert-the-x86_64-architecture-to-irq-chips.patch
genirq-convert-the-i386-architecture-to-irq-chips.patch
genirq-convert-the-i386-architecture-to-irq-chips-fix-2.patch
genirq-more-verbose-debugging-on-unexpected-irq-vectors.patch
detect-atomic-counter-underflows.patch
debug-shared-irqs.patch
make-frame_pointer-default=y.patch
mutex-subsystem-synchro-test-module.patch
vdso-print-fatal-signals.patch
vdso-improve-print_fatal_signals-support-by-adding-memory-maps.patch
-
To unsubscribe from this list: send the line "unsubscribe mm-commits" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
