aboutsummaryrefslogtreecommitdiffstats
path: root/kernel
diff options
context:
space:
mode:
authorRussell King <rmk+kernel@arm.linux.org.uk>2014-07-12 08:59:24 -0400
committerRussell King <rmk+kernel@arm.linux.org.uk>2014-07-12 08:59:24 -0400
commitcda390bb8f0fc751194ef5f4c691ca8d3ff91009 (patch)
tree9739758690c5107b73a09f632ae2f8fc2507db09 /kernel
parent9a2c33a422d60e8e55c5aff6752522dc39993b16 (diff)
parent449fd15fbcde239875031d70fd06abb4a0e1ed09 (diff)
Merge branch 'kprobes-test-fixes' of git://git.linaro.org/people/tixy/kernel into fixes
Diffstat (limited to 'kernel')
-rw-r--r--kernel/context_tracking.c3
-rw-r--r--kernel/events/core.c37
-rw-r--r--kernel/locking/rtmutex-debug.h5
-rw-r--r--kernel/locking/rtmutex.c243
-rw-r--r--kernel/locking/rtmutex.h5
-rw-r--r--kernel/power/hibernate.c37
-rw-r--r--kernel/power/main.c6
-rw-r--r--kernel/power/user.c3
-rw-r--r--kernel/sysctl.c4
9 files changed, 295 insertions, 48 deletions
diff --git a/kernel/context_tracking.c b/kernel/context_tracking.c
index 019d45008448..5664985c46a0 100644
--- a/kernel/context_tracking.c
+++ b/kernel/context_tracking.c
@@ -19,6 +19,7 @@
19#include <linux/sched.h> 19#include <linux/sched.h>
20#include <linux/hardirq.h> 20#include <linux/hardirq.h>
21#include <linux/export.h> 21#include <linux/export.h>
22#include <linux/kprobes.h>
22 23
23#define CREATE_TRACE_POINTS 24#define CREATE_TRACE_POINTS
24#include <trace/events/context_tracking.h> 25#include <trace/events/context_tracking.h>
@@ -104,6 +105,7 @@ void context_tracking_user_enter(void)
104 } 105 }
105 local_irq_restore(flags); 106 local_irq_restore(flags);
106} 107}
108NOKPROBE_SYMBOL(context_tracking_user_enter);
107 109
108#ifdef CONFIG_PREEMPT 110#ifdef CONFIG_PREEMPT
109/** 111/**
@@ -181,6 +183,7 @@ void context_tracking_user_exit(void)
181 } 183 }
182 local_irq_restore(flags); 184 local_irq_restore(flags);
183} 185}
186NOKPROBE_SYMBOL(context_tracking_user_exit);
184 187
185/** 188/**
186 * __context_tracking_task_switch - context switch the syscall callbacks 189 * __context_tracking_task_switch - context switch the syscall callbacks
diff --git a/kernel/events/core.c b/kernel/events/core.c
index 5fa58e4cffac..a33d9a2bcbd7 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -40,6 +40,7 @@
40#include <linux/mm_types.h> 40#include <linux/mm_types.h>
41#include <linux/cgroup.h> 41#include <linux/cgroup.h>
42#include <linux/module.h> 42#include <linux/module.h>
43#include <linux/mman.h>
43 44
44#include "internal.h" 45#include "internal.h"
45 46
@@ -5128,6 +5129,7 @@ struct perf_mmap_event {
5128 int maj, min; 5129 int maj, min;
5129 u64 ino; 5130 u64 ino;
5130 u64 ino_generation; 5131 u64 ino_generation;
5132 u32 prot, flags;
5131 5133
5132 struct { 5134 struct {
5133 struct perf_event_header header; 5135 struct perf_event_header header;
@@ -5169,6 +5171,8 @@ static void perf_event_mmap_output(struct perf_event *event,
5169 mmap_event->event_id.header.size += sizeof(mmap_event->min); 5171 mmap_event->event_id.header.size += sizeof(mmap_event->min);
5170 mmap_event->event_id.header.size += sizeof(mmap_event->ino); 5172 mmap_event->event_id.header.size += sizeof(mmap_event->ino);
5171 mmap_event->event_id.header.size += sizeof(mmap_event->ino_generation); 5173 mmap_event->event_id.header.size += sizeof(mmap_event->ino_generation);
5174 mmap_event->event_id.header.size += sizeof(mmap_event->prot);
5175 mmap_event->event_id.header.size += sizeof(mmap_event->flags);
5172 } 5176 }
5173 5177
5174 perf_event_header__init_id(&mmap_event->event_id.header, &sample, event); 5178 perf_event_header__init_id(&mmap_event->event_id.header, &sample, event);
@@ -5187,6 +5191,8 @@ static void perf_event_mmap_output(struct perf_event *event,
5187 perf_output_put(&handle, mmap_event->min); 5191 perf_output_put(&handle, mmap_event->min);
5188 perf_output_put(&handle, mmap_event->ino); 5192 perf_output_put(&handle, mmap_event->ino);
5189 perf_output_put(&handle, mmap_event->ino_generation); 5193 perf_output_put(&handle, mmap_event->ino_generation);
5194 perf_output_put(&handle, mmap_event->prot);
5195 perf_output_put(&handle, mmap_event->flags);
5190 } 5196 }
5191 5197
5192 __output_copy(&handle, mmap_event->file_name, 5198 __output_copy(&handle, mmap_event->file_name,
@@ -5205,6 +5211,7 @@ static void perf_event_mmap_event(struct perf_mmap_event *mmap_event)
5205 struct file *file = vma->vm_file; 5211 struct file *file = vma->vm_file;
5206 int maj = 0, min = 0; 5212 int maj = 0, min = 0;
5207 u64 ino = 0, gen = 0; 5213 u64 ino = 0, gen = 0;
5214 u32 prot = 0, flags = 0;
5208 unsigned int size; 5215 unsigned int size;
5209 char tmp[16]; 5216 char tmp[16];
5210 char *buf = NULL; 5217 char *buf = NULL;
@@ -5235,6 +5242,28 @@ static void perf_event_mmap_event(struct perf_mmap_event *mmap_event)
5235 gen = inode->i_generation; 5242 gen = inode->i_generation;
5236 maj = MAJOR(dev); 5243 maj = MAJOR(dev);
5237 min = MINOR(dev); 5244 min = MINOR(dev);
5245
5246 if (vma->vm_flags & VM_READ)
5247 prot |= PROT_READ;
5248 if (vma->vm_flags & VM_WRITE)
5249 prot |= PROT_WRITE;
5250 if (vma->vm_flags & VM_EXEC)
5251 prot |= PROT_EXEC;
5252
5253 if (vma->vm_flags & VM_MAYSHARE)
5254 flags = MAP_SHARED;
5255 else
5256 flags = MAP_PRIVATE;
5257
5258 if (vma->vm_flags & VM_DENYWRITE)
5259 flags |= MAP_DENYWRITE;
5260 if (vma->vm_flags & VM_MAYEXEC)
5261 flags |= MAP_EXECUTABLE;
5262 if (vma->vm_flags & VM_LOCKED)
5263 flags |= MAP_LOCKED;
5264 if (vma->vm_flags & VM_HUGETLB)
5265 flags |= MAP_HUGETLB;
5266
5238 goto got_name; 5267 goto got_name;
5239 } else { 5268 } else {
5240 name = (char *)arch_vma_name(vma); 5269 name = (char *)arch_vma_name(vma);
@@ -5275,6 +5304,8 @@ got_name:
5275 mmap_event->min = min; 5304 mmap_event->min = min;
5276 mmap_event->ino = ino; 5305 mmap_event->ino = ino;
5277 mmap_event->ino_generation = gen; 5306 mmap_event->ino_generation = gen;
5307 mmap_event->prot = prot;
5308 mmap_event->flags = flags;
5278 5309
5279 if (!(vma->vm_flags & VM_EXEC)) 5310 if (!(vma->vm_flags & VM_EXEC))
5280 mmap_event->event_id.header.misc |= PERF_RECORD_MISC_MMAP_DATA; 5311 mmap_event->event_id.header.misc |= PERF_RECORD_MISC_MMAP_DATA;
@@ -5315,6 +5346,8 @@ void perf_event_mmap(struct vm_area_struct *vma)
5315 /* .min (attr_mmap2 only) */ 5346 /* .min (attr_mmap2 only) */
5316 /* .ino (attr_mmap2 only) */ 5347 /* .ino (attr_mmap2 only) */
5317 /* .ino_generation (attr_mmap2 only) */ 5348 /* .ino_generation (attr_mmap2 only) */
5349 /* .prot (attr_mmap2 only) */
5350 /* .flags (attr_mmap2 only) */
5318 }; 5351 };
5319 5352
5320 perf_event_mmap_event(&mmap_event); 5353 perf_event_mmap_event(&mmap_event);
@@ -6897,10 +6930,6 @@ static int perf_copy_attr(struct perf_event_attr __user *uattr,
6897 if (ret) 6930 if (ret)
6898 return -EFAULT; 6931 return -EFAULT;
6899 6932
6900 /* disabled for now */
6901 if (attr->mmap2)
6902 return -EINVAL;
6903
6904 if (attr->__reserved_1) 6933 if (attr->__reserved_1)
6905 return -EINVAL; 6934 return -EINVAL;
6906 6935
diff --git a/kernel/locking/rtmutex-debug.h b/kernel/locking/rtmutex-debug.h
index 14193d596d78..ab29b6a22669 100644
--- a/kernel/locking/rtmutex-debug.h
+++ b/kernel/locking/rtmutex-debug.h
@@ -31,3 +31,8 @@ static inline int debug_rt_mutex_detect_deadlock(struct rt_mutex_waiter *waiter,
31{ 31{
32 return (waiter != NULL); 32 return (waiter != NULL);
33} 33}
34
35static inline void rt_mutex_print_deadlock(struct rt_mutex_waiter *w)
36{
37 debug_rt_mutex_print_deadlock(w);
38}
diff --git a/kernel/locking/rtmutex.c b/kernel/locking/rtmutex.c
index a620d4d08ca6..fc605941b9b8 100644
--- a/kernel/locking/rtmutex.c
+++ b/kernel/locking/rtmutex.c
@@ -83,6 +83,47 @@ static inline void mark_rt_mutex_waiters(struct rt_mutex *lock)
83 owner = *p; 83 owner = *p;
84 } while (cmpxchg(p, owner, owner | RT_MUTEX_HAS_WAITERS) != owner); 84 } while (cmpxchg(p, owner, owner | RT_MUTEX_HAS_WAITERS) != owner);
85} 85}
86
87/*
88 * Safe fastpath aware unlock:
89 * 1) Clear the waiters bit
90 * 2) Drop lock->wait_lock
91 * 3) Try to unlock the lock with cmpxchg
92 */
93static inline bool unlock_rt_mutex_safe(struct rt_mutex *lock)
94 __releases(lock->wait_lock)
95{
96 struct task_struct *owner = rt_mutex_owner(lock);
97
98 clear_rt_mutex_waiters(lock);
99 raw_spin_unlock(&lock->wait_lock);
100 /*
101 * If a new waiter comes in between the unlock and the cmpxchg
102 * we have two situations:
103 *
104 * unlock(wait_lock);
105 * lock(wait_lock);
106 * cmpxchg(p, owner, 0) == owner
107 * mark_rt_mutex_waiters(lock);
108 * acquire(lock);
109 * or:
110 *
111 * unlock(wait_lock);
112 * lock(wait_lock);
113 * mark_rt_mutex_waiters(lock);
114 *
115 * cmpxchg(p, owner, 0) != owner
116 * enqueue_waiter();
117 * unlock(wait_lock);
118 * lock(wait_lock);
119 * wake waiter();
120 * unlock(wait_lock);
121 * lock(wait_lock);
122 * acquire(lock);
123 */
124 return rt_mutex_cmpxchg(lock, owner, NULL);
125}
126
86#else 127#else
87# define rt_mutex_cmpxchg(l,c,n) (0) 128# define rt_mutex_cmpxchg(l,c,n) (0)
88static inline void mark_rt_mutex_waiters(struct rt_mutex *lock) 129static inline void mark_rt_mutex_waiters(struct rt_mutex *lock)
@@ -90,6 +131,17 @@ static inline void mark_rt_mutex_waiters(struct rt_mutex *lock)
90 lock->owner = (struct task_struct *) 131 lock->owner = (struct task_struct *)
91 ((unsigned long)lock->owner | RT_MUTEX_HAS_WAITERS); 132 ((unsigned long)lock->owner | RT_MUTEX_HAS_WAITERS);
92} 133}
134
135/*
136 * Simple slow path only version: lock->owner is protected by lock->wait_lock.
137 */
138static inline bool unlock_rt_mutex_safe(struct rt_mutex *lock)
139 __releases(lock->wait_lock)
140{
141 lock->owner = NULL;
142 raw_spin_unlock(&lock->wait_lock);
143 return true;
144}
93#endif 145#endif
94 146
95static inline int 147static inline int
@@ -260,27 +312,36 @@ static void rt_mutex_adjust_prio(struct task_struct *task)
260 */ 312 */
261int max_lock_depth = 1024; 313int max_lock_depth = 1024;
262 314
315static inline struct rt_mutex *task_blocked_on_lock(struct task_struct *p)
316{
317 return p->pi_blocked_on ? p->pi_blocked_on->lock : NULL;
318}
319
263/* 320/*
264 * Adjust the priority chain. Also used for deadlock detection. 321 * Adjust the priority chain. Also used for deadlock detection.
265 * Decreases task's usage by one - may thus free the task. 322 * Decreases task's usage by one - may thus free the task.
266 * 323 *
267 * @task: the task owning the mutex (owner) for which a chain walk is probably 324 * @task: the task owning the mutex (owner) for which a chain walk is
268 * needed 325 * probably needed
269 * @deadlock_detect: do we have to carry out deadlock detection? 326 * @deadlock_detect: do we have to carry out deadlock detection?
270 * @orig_lock: the mutex (can be NULL if we are walking the chain to recheck 327 * @orig_lock: the mutex (can be NULL if we are walking the chain to recheck
271 * things for a task that has just got its priority adjusted, and 328 * things for a task that has just got its priority adjusted, and
272 * is waiting on a mutex) 329 * is waiting on a mutex)
330 * @next_lock: the mutex on which the owner of @orig_lock was blocked before
331 * we dropped its pi_lock. Is never dereferenced, only used for
332 * comparison to detect lock chain changes.
273 * @orig_waiter: rt_mutex_waiter struct for the task that has just donated 333 * @orig_waiter: rt_mutex_waiter struct for the task that has just donated
274 * its priority to the mutex owner (can be NULL in the case 334 * its priority to the mutex owner (can be NULL in the case
275 * depicted above or if the top waiter is gone away and we are 335 * depicted above or if the top waiter is gone away and we are
276 * actually deboosting the owner) 336 * actually deboosting the owner)
277 * @top_task: the current top waiter 337 * @top_task: the current top waiter
278 * 338 *
279 * Returns 0 or -EDEADLK. 339 * Returns 0 or -EDEADLK.
280 */ 340 */
281static int rt_mutex_adjust_prio_chain(struct task_struct *task, 341static int rt_mutex_adjust_prio_chain(struct task_struct *task,
282 int deadlock_detect, 342 int deadlock_detect,
283 struct rt_mutex *orig_lock, 343 struct rt_mutex *orig_lock,
344 struct rt_mutex *next_lock,
284 struct rt_mutex_waiter *orig_waiter, 345 struct rt_mutex_waiter *orig_waiter,
285 struct task_struct *top_task) 346 struct task_struct *top_task)
286{ 347{
@@ -314,7 +375,7 @@ static int rt_mutex_adjust_prio_chain(struct task_struct *task,
314 } 375 }
315 put_task_struct(task); 376 put_task_struct(task);
316 377
317 return deadlock_detect ? -EDEADLK : 0; 378 return -EDEADLK;
318 } 379 }
319 retry: 380 retry:
320 /* 381 /*
@@ -339,6 +400,18 @@ static int rt_mutex_adjust_prio_chain(struct task_struct *task,
339 goto out_unlock_pi; 400 goto out_unlock_pi;
340 401
341 /* 402 /*
403 * We dropped all locks after taking a refcount on @task, so
404 * the task might have moved on in the lock chain or even left
405 * the chain completely and blocks now on an unrelated lock or
406 * on @orig_lock.
407 *
408 * We stored the lock on which @task was blocked in @next_lock,
409 * so we can detect the chain change.
410 */
411 if (next_lock != waiter->lock)
412 goto out_unlock_pi;
413
414 /*
342 * Drop out, when the task has no waiters. Note, 415 * Drop out, when the task has no waiters. Note,
343 * top_waiter can be NULL, when we are in the deboosting 416 * top_waiter can be NULL, when we are in the deboosting
344 * mode! 417 * mode!
@@ -377,7 +450,7 @@ static int rt_mutex_adjust_prio_chain(struct task_struct *task,
377 if (lock == orig_lock || rt_mutex_owner(lock) == top_task) { 450 if (lock == orig_lock || rt_mutex_owner(lock) == top_task) {
378 debug_rt_mutex_deadlock(deadlock_detect, orig_waiter, lock); 451 debug_rt_mutex_deadlock(deadlock_detect, orig_waiter, lock);
379 raw_spin_unlock(&lock->wait_lock); 452 raw_spin_unlock(&lock->wait_lock);
380 ret = deadlock_detect ? -EDEADLK : 0; 453 ret = -EDEADLK;
381 goto out_unlock_pi; 454 goto out_unlock_pi;
382 } 455 }
383 456
@@ -422,11 +495,26 @@ static int rt_mutex_adjust_prio_chain(struct task_struct *task,
422 __rt_mutex_adjust_prio(task); 495 __rt_mutex_adjust_prio(task);
423 } 496 }
424 497
498 /*
499 * Check whether the task which owns the current lock is pi
500 * blocked itself. If yes we store a pointer to the lock for
501 * the lock chain change detection above. After we dropped
502 * task->pi_lock next_lock cannot be dereferenced anymore.
503 */
504 next_lock = task_blocked_on_lock(task);
505
425 raw_spin_unlock_irqrestore(&task->pi_lock, flags); 506 raw_spin_unlock_irqrestore(&task->pi_lock, flags);
426 507
427 top_waiter = rt_mutex_top_waiter(lock); 508 top_waiter = rt_mutex_top_waiter(lock);
428 raw_spin_unlock(&lock->wait_lock); 509 raw_spin_unlock(&lock->wait_lock);
429 510
511 /*
512 * We reached the end of the lock chain. Stop right here. No
513 * point to go back just to figure that out.
514 */
515 if (!next_lock)
516 goto out_put_task;
517
430 if (!detect_deadlock && waiter != top_waiter) 518 if (!detect_deadlock && waiter != top_waiter)
431 goto out_put_task; 519 goto out_put_task;
432 520
@@ -536,8 +624,9 @@ static int task_blocks_on_rt_mutex(struct rt_mutex *lock,
536{ 624{
537 struct task_struct *owner = rt_mutex_owner(lock); 625 struct task_struct *owner = rt_mutex_owner(lock);
538 struct rt_mutex_waiter *top_waiter = waiter; 626 struct rt_mutex_waiter *top_waiter = waiter;
539 unsigned long flags; 627 struct rt_mutex *next_lock;
540 int chain_walk = 0, res; 628 int chain_walk = 0, res;
629 unsigned long flags;
541 630
542 /* 631 /*
543 * Early deadlock detection. We really don't want the task to 632 * Early deadlock detection. We really don't want the task to
@@ -548,7 +637,7 @@ static int task_blocks_on_rt_mutex(struct rt_mutex *lock,
548 * which is wrong, as the other waiter is not in a deadlock 637 * which is wrong, as the other waiter is not in a deadlock
549 * situation. 638 * situation.
550 */ 639 */
551 if (detect_deadlock && owner == task) 640 if (owner == task)
552 return -EDEADLK; 641 return -EDEADLK;
553 642
554 raw_spin_lock_irqsave(&task->pi_lock, flags); 643 raw_spin_lock_irqsave(&task->pi_lock, flags);
@@ -569,20 +658,28 @@ static int task_blocks_on_rt_mutex(struct rt_mutex *lock,
569 if (!owner) 658 if (!owner)
570 return 0; 659 return 0;
571 660
661 raw_spin_lock_irqsave(&owner->pi_lock, flags);
572 if (waiter == rt_mutex_top_waiter(lock)) { 662 if (waiter == rt_mutex_top_waiter(lock)) {
573 raw_spin_lock_irqsave(&owner->pi_lock, flags);
574 rt_mutex_dequeue_pi(owner, top_waiter); 663 rt_mutex_dequeue_pi(owner, top_waiter);
575 rt_mutex_enqueue_pi(owner, waiter); 664 rt_mutex_enqueue_pi(owner, waiter);
576 665
577 __rt_mutex_adjust_prio(owner); 666 __rt_mutex_adjust_prio(owner);
578 if (owner->pi_blocked_on) 667 if (owner->pi_blocked_on)
579 chain_walk = 1; 668 chain_walk = 1;
580 raw_spin_unlock_irqrestore(&owner->pi_lock, flags); 669 } else if (debug_rt_mutex_detect_deadlock(waiter, detect_deadlock)) {
581 }
582 else if (debug_rt_mutex_detect_deadlock(waiter, detect_deadlock))
583 chain_walk = 1; 670 chain_walk = 1;
671 }
584 672
585 if (!chain_walk) 673 /* Store the lock on which owner is blocked or NULL */
674 next_lock = task_blocked_on_lock(owner);
675
676 raw_spin_unlock_irqrestore(&owner->pi_lock, flags);
677 /*
678 * Even if full deadlock detection is on, if the owner is not
679 * blocked itself, we can avoid finding this out in the chain
680 * walk.
681 */
682 if (!chain_walk || !next_lock)
586 return 0; 683 return 0;
587 684
588 /* 685 /*
@@ -594,8 +691,8 @@ static int task_blocks_on_rt_mutex(struct rt_mutex *lock,
594 691
595 raw_spin_unlock(&lock->wait_lock); 692 raw_spin_unlock(&lock->wait_lock);
596 693
597 res = rt_mutex_adjust_prio_chain(owner, detect_deadlock, lock, waiter, 694 res = rt_mutex_adjust_prio_chain(owner, detect_deadlock, lock,
598 task); 695 next_lock, waiter, task);
599 696
600 raw_spin_lock(&lock->wait_lock); 697 raw_spin_lock(&lock->wait_lock);
601 698
@@ -605,7 +702,8 @@ static int task_blocks_on_rt_mutex(struct rt_mutex *lock,
605/* 702/*
606 * Wake up the next waiter on the lock. 703 * Wake up the next waiter on the lock.
607 * 704 *
608 * Remove the top waiter from the current tasks waiter list and wake it up. 705 * Remove the top waiter from the current tasks pi waiter list and
706 * wake it up.
609 * 707 *
610 * Called with lock->wait_lock held. 708 * Called with lock->wait_lock held.
611 */ 709 */
@@ -626,10 +724,23 @@ static void wakeup_next_waiter(struct rt_mutex *lock)
626 */ 724 */
627 rt_mutex_dequeue_pi(current, waiter); 725 rt_mutex_dequeue_pi(current, waiter);
628 726
629 rt_mutex_set_owner(lock, NULL); 727 /*
728 * As we are waking up the top waiter, and the waiter stays
729 * queued on the lock until it gets the lock, this lock
730 * obviously has waiters. Just set the bit here and this has
731 * the added benefit of forcing all new tasks into the
732 * slow path making sure no task of lower priority than
733 * the top waiter can steal this lock.
734 */
735 lock->owner = (void *) RT_MUTEX_HAS_WAITERS;
630 736
631 raw_spin_unlock_irqrestore(&current->pi_lock, flags); 737 raw_spin_unlock_irqrestore(&current->pi_lock, flags);
632 738
739 /*
740 * It's safe to dereference waiter as it cannot go away as
741 * long as we hold lock->wait_lock. The waiter task needs to
742 * acquire it in order to dequeue the waiter.
743 */
633 wake_up_process(waiter->task); 744 wake_up_process(waiter->task);
634} 745}
635 746
@@ -644,8 +755,8 @@ static void remove_waiter(struct rt_mutex *lock,
644{ 755{
645 int first = (waiter == rt_mutex_top_waiter(lock)); 756 int first = (waiter == rt_mutex_top_waiter(lock));
646 struct task_struct *owner = rt_mutex_owner(lock); 757 struct task_struct *owner = rt_mutex_owner(lock);
758 struct rt_mutex *next_lock = NULL;
647 unsigned long flags; 759 unsigned long flags;
648 int chain_walk = 0;
649 760
650 raw_spin_lock_irqsave(&current->pi_lock, flags); 761 raw_spin_lock_irqsave(&current->pi_lock, flags);
651 rt_mutex_dequeue(lock, waiter); 762 rt_mutex_dequeue(lock, waiter);
@@ -669,13 +780,13 @@ static void remove_waiter(struct rt_mutex *lock,
669 } 780 }
670 __rt_mutex_adjust_prio(owner); 781 __rt_mutex_adjust_prio(owner);
671 782
672 if (owner->pi_blocked_on) 783 /* Store the lock on which owner is blocked or NULL */
673 chain_walk = 1; 784 next_lock = task_blocked_on_lock(owner);
674 785
675 raw_spin_unlock_irqrestore(&owner->pi_lock, flags); 786 raw_spin_unlock_irqrestore(&owner->pi_lock, flags);
676 } 787 }
677 788
678 if (!chain_walk) 789 if (!next_lock)
679 return; 790 return;
680 791
681 /* gets dropped in rt_mutex_adjust_prio_chain()! */ 792 /* gets dropped in rt_mutex_adjust_prio_chain()! */
@@ -683,7 +794,7 @@ static void remove_waiter(struct rt_mutex *lock,
683 794
684 raw_spin_unlock(&lock->wait_lock); 795 raw_spin_unlock(&lock->wait_lock);
685 796
686 rt_mutex_adjust_prio_chain(owner, 0, lock, NULL, current); 797 rt_mutex_adjust_prio_chain(owner, 0, lock, next_lock, NULL, current);
687 798
688 raw_spin_lock(&lock->wait_lock); 799 raw_spin_lock(&lock->wait_lock);
689} 800}
@@ -696,6 +807,7 @@ static void remove_waiter(struct rt_mutex *lock,
696void rt_mutex_adjust_pi(struct task_struct *task) 807void rt_mutex_adjust_pi(struct task_struct *task)
697{ 808{
698 struct rt_mutex_waiter *waiter; 809 struct rt_mutex_waiter *waiter;
810 struct rt_mutex *next_lock;
699 unsigned long flags; 811 unsigned long flags;
700 812
701 raw_spin_lock_irqsave(&task->pi_lock, flags); 813 raw_spin_lock_irqsave(&task->pi_lock, flags);
@@ -706,12 +818,13 @@ void rt_mutex_adjust_pi(struct task_struct *task)
706 raw_spin_unlock_irqrestore(&task->pi_lock, flags); 818 raw_spin_unlock_irqrestore(&task->pi_lock, flags);
707 return; 819 return;
708 } 820 }
709 821 next_lock = waiter->lock;
710 raw_spin_unlock_irqrestore(&task->pi_lock, flags); 822 raw_spin_unlock_irqrestore(&task->pi_lock, flags);
711 823
712 /* gets dropped in rt_mutex_adjust_prio_chain()! */ 824 /* gets dropped in rt_mutex_adjust_prio_chain()! */
713 get_task_struct(task); 825 get_task_struct(task);
714 rt_mutex_adjust_prio_chain(task, 0, NULL, NULL, task); 826
827 rt_mutex_adjust_prio_chain(task, 0, NULL, next_lock, NULL, task);
715} 828}
716 829
717/** 830/**
@@ -763,6 +876,26 @@ __rt_mutex_slowlock(struct rt_mutex *lock, int state,
763 return ret; 876 return ret;
764} 877}
765 878
879static void rt_mutex_handle_deadlock(int res, int detect_deadlock,
880 struct rt_mutex_waiter *w)
881{
882 /*
883 * If the result is not -EDEADLOCK or the caller requested
884 * deadlock detection, nothing to do here.
885 */
886 if (res != -EDEADLOCK || detect_deadlock)
887 return;
888
889 /*
890 * Yell lowdly and stop the task right here.
891 */
892 rt_mutex_print_deadlock(w);
893 while (1) {
894 set_current_state(TASK_INTERRUPTIBLE);
895 schedule();
896 }
897}
898
766/* 899/*
767 * Slow path lock function: 900 * Slow path lock function:
768 */ 901 */
@@ -802,8 +935,10 @@ rt_mutex_slowlock(struct rt_mutex *lock, int state,
802 935
803 set_current_state(TASK_RUNNING); 936 set_current_state(TASK_RUNNING);
804 937
805 if (unlikely(ret)) 938 if (unlikely(ret)) {
806 remove_waiter(lock, &waiter); 939 remove_waiter(lock, &waiter);
940 rt_mutex_handle_deadlock(ret, detect_deadlock, &waiter);
941 }
807 942
808 /* 943 /*
809 * try_to_take_rt_mutex() sets the waiter bit 944 * try_to_take_rt_mutex() sets the waiter bit
@@ -859,12 +994,49 @@ rt_mutex_slowunlock(struct rt_mutex *lock)
859 994
860 rt_mutex_deadlock_account_unlock(current); 995 rt_mutex_deadlock_account_unlock(current);
861 996
862 if (!rt_mutex_has_waiters(lock)) { 997 /*
863 lock->owner = NULL; 998 * We must be careful here if the fast path is enabled. If we
864 raw_spin_unlock(&lock->wait_lock); 999 * have no waiters queued we cannot set owner to NULL here
865 return; 1000 * because of:
1001 *
1002 * foo->lock->owner = NULL;
1003 * rtmutex_lock(foo->lock); <- fast path
1004 * free = atomic_dec_and_test(foo->refcnt);
1005 * rtmutex_unlock(foo->lock); <- fast path
1006 * if (free)
1007 * kfree(foo);
1008 * raw_spin_unlock(foo->lock->wait_lock);
1009 *
1010 * So for the fastpath enabled kernel:
1011 *
1012 * Nothing can set the waiters bit as long as we hold
1013 * lock->wait_lock. So we do the following sequence:
1014 *
1015 * owner = rt_mutex_owner(lock);
1016 * clear_rt_mutex_waiters(lock);
1017 * raw_spin_unlock(&lock->wait_lock);
1018 * if (cmpxchg(&lock->owner, owner, 0) == owner)
1019 * return;
1020 * goto retry;
1021 *
1022 * The fastpath disabled variant is simple as all access to
1023 * lock->owner is serialized by lock->wait_lock:
1024 *
1025 * lock->owner = NULL;
1026 * raw_spin_unlock(&lock->wait_lock);
1027 */
1028 while (!rt_mutex_has_waiters(lock)) {
1029 /* Drops lock->wait_lock ! */
1030 if (unlock_rt_mutex_safe(lock) == true)
1031 return;
1032 /* Relock the rtmutex and try again */
1033 raw_spin_lock(&lock->wait_lock);
866 } 1034 }
867 1035
1036 /*
1037 * The wakeup next waiter path does not suffer from the above
1038 * race. See the comments there.
1039 */
868 wakeup_next_waiter(lock); 1040 wakeup_next_waiter(lock);
869 1041
870 raw_spin_unlock(&lock->wait_lock); 1042 raw_spin_unlock(&lock->wait_lock);
@@ -1112,7 +1284,8 @@ int rt_mutex_start_proxy_lock(struct rt_mutex *lock,
1112 return 1; 1284 return 1;
1113 } 1285 }
1114 1286
1115 ret = task_blocks_on_rt_mutex(lock, waiter, task, detect_deadlock); 1287 /* We enforce deadlock detection for futexes */
1288 ret = task_blocks_on_rt_mutex(lock, waiter, task, 1);
1116 1289
1117 if (ret && !rt_mutex_owner(lock)) { 1290 if (ret && !rt_mutex_owner(lock)) {
1118 /* 1291 /*
diff --git a/kernel/locking/rtmutex.h b/kernel/locking/rtmutex.h
index a1a1dd06421d..f6a1f3c133b1 100644
--- a/kernel/locking/rtmutex.h
+++ b/kernel/locking/rtmutex.h
@@ -24,3 +24,8 @@
24#define debug_rt_mutex_print_deadlock(w) do { } while (0) 24#define debug_rt_mutex_print_deadlock(w) do { } while (0)
25#define debug_rt_mutex_detect_deadlock(w,d) (d) 25#define debug_rt_mutex_detect_deadlock(w,d) (d)
26#define debug_rt_mutex_reset_waiter(w) do { } while (0) 26#define debug_rt_mutex_reset_waiter(w) do { } while (0)
27
28static inline void rt_mutex_print_deadlock(struct rt_mutex_waiter *w)
29{
30 WARN(1, "rtmutex deadlock detected\n");
31}
diff --git a/kernel/power/hibernate.c b/kernel/power/hibernate.c
index 49e0a20fd010..fcc2611d3f14 100644
--- a/kernel/power/hibernate.c
+++ b/kernel/power/hibernate.c
@@ -35,6 +35,7 @@
35 35
36static int nocompress; 36static int nocompress;
37static int noresume; 37static int noresume;
38static int nohibernate;
38static int resume_wait; 39static int resume_wait;
39static unsigned int resume_delay; 40static unsigned int resume_delay;
40static char resume_file[256] = CONFIG_PM_STD_PARTITION; 41static char resume_file[256] = CONFIG_PM_STD_PARTITION;
@@ -62,6 +63,11 @@ bool freezer_test_done;
62 63
63static const struct platform_hibernation_ops *hibernation_ops; 64static const struct platform_hibernation_ops *hibernation_ops;
64 65
66bool hibernation_available(void)
67{
68 return (nohibernate == 0);
69}
70
65/** 71/**
66 * hibernation_set_ops - Set the global hibernate operations. 72 * hibernation_set_ops - Set the global hibernate operations.
67 * @ops: Hibernation operations to use in subsequent hibernation transitions. 73 * @ops: Hibernation operations to use in subsequent hibernation transitions.
@@ -642,6 +648,11 @@ int hibernate(void)
642{ 648{
643 int error; 649 int error;
644 650
651 if (!hibernation_available()) {
652 pr_debug("PM: Hibernation not available.\n");
653 return -EPERM;
654 }
655
645 lock_system_sleep(); 656 lock_system_sleep();
646 /* The snapshot device should not be opened while we're running */ 657 /* The snapshot device should not be opened while we're running */
647 if (!atomic_add_unless(&snapshot_device_available, -1, 0)) { 658 if (!atomic_add_unless(&snapshot_device_available, -1, 0)) {
@@ -734,7 +745,7 @@ static int software_resume(void)
734 /* 745 /*
735 * If the user said "noresume".. bail out early. 746 * If the user said "noresume".. bail out early.
736 */ 747 */
737 if (noresume) 748 if (noresume || !hibernation_available())
738 return 0; 749 return 0;
739 750
740 /* 751 /*
@@ -900,6 +911,9 @@ static ssize_t disk_show(struct kobject *kobj, struct kobj_attribute *attr,
900 int i; 911 int i;
901 char *start = buf; 912 char *start = buf;
902 913
914 if (!hibernation_available())
915 return sprintf(buf, "[disabled]\n");
916
903 for (i = HIBERNATION_FIRST; i <= HIBERNATION_MAX; i++) { 917 for (i = HIBERNATION_FIRST; i <= HIBERNATION_MAX; i++) {
904 if (!hibernation_modes[i]) 918 if (!hibernation_modes[i])
905 continue; 919 continue;
@@ -934,6 +948,9 @@ static ssize_t disk_store(struct kobject *kobj, struct kobj_attribute *attr,
934 char *p; 948 char *p;
935 int mode = HIBERNATION_INVALID; 949 int mode = HIBERNATION_INVALID;
936 950
951 if (!hibernation_available())
952 return -EPERM;
953
937 p = memchr(buf, '\n', n); 954 p = memchr(buf, '\n', n);
938 len = p ? p - buf : n; 955 len = p ? p - buf : n;
939 956
@@ -1101,6 +1118,10 @@ static int __init hibernate_setup(char *str)
1101 noresume = 1; 1118 noresume = 1;
1102 else if (!strncmp(str, "nocompress", 10)) 1119 else if (!strncmp(str, "nocompress", 10))
1103 nocompress = 1; 1120 nocompress = 1;
1121 else if (!strncmp(str, "no", 2)) {
1122 noresume = 1;
1123 nohibernate = 1;
1124 }
1104 return 1; 1125 return 1;
1105} 1126}
1106 1127
@@ -1125,9 +1146,23 @@ static int __init resumedelay_setup(char *str)
1125 return 1; 1146 return 1;
1126} 1147}
1127 1148
1149static int __init nohibernate_setup(char *str)
1150{
1151 noresume = 1;
1152 nohibernate = 1;
1153 return 1;
1154}
1155
1156static int __init kaslr_nohibernate_setup(char *str)
1157{
1158 return nohibernate_setup(str);
1159}
1160
1128__setup("noresume", noresume_setup); 1161__setup("noresume", noresume_setup);
1129__setup("resume_offset=", resume_offset_setup); 1162__setup("resume_offset=", resume_offset_setup);
1130__setup("resume=", resume_setup); 1163__setup("resume=", resume_setup);
1131__setup("hibernate=", hibernate_setup); 1164__setup("hibernate=", hibernate_setup);
1132__setup("resumewait", resumewait_setup); 1165__setup("resumewait", resumewait_setup);
1133__setup("resumedelay=", resumedelay_setup); 1166__setup("resumedelay=", resumedelay_setup);
1167__setup("nohibernate", nohibernate_setup);
1168__setup("kaslr", kaslr_nohibernate_setup);
diff --git a/kernel/power/main.c b/kernel/power/main.c
index 573410d6647e..8e90f330f139 100644
--- a/kernel/power/main.c
+++ b/kernel/power/main.c
@@ -300,13 +300,11 @@ static ssize_t state_show(struct kobject *kobj, struct kobj_attribute *attr,
300 s += sprintf(s,"%s ", pm_states[i].label); 300 s += sprintf(s,"%s ", pm_states[i].label);
301 301
302#endif 302#endif
303#ifdef CONFIG_HIBERNATION 303 if (hibernation_available())
304 s += sprintf(s, "%s\n", "disk"); 304 s += sprintf(s, "disk ");
305#else
306 if (s != buf) 305 if (s != buf)
307 /* convert the last space to a newline */ 306 /* convert the last space to a newline */
308 *(s-1) = '\n'; 307 *(s-1) = '\n';
309#endif
310 return (s - buf); 308 return (s - buf);
311} 309}
312 310
diff --git a/kernel/power/user.c b/kernel/power/user.c
index 98d357584cd6..526e8911460a 100644
--- a/kernel/power/user.c
+++ b/kernel/power/user.c
@@ -49,6 +49,9 @@ static int snapshot_open(struct inode *inode, struct file *filp)
49 struct snapshot_data *data; 49 struct snapshot_data *data;
50 int error; 50 int error;
51 51
52 if (!hibernation_available())
53 return -EPERM;
54
52 lock_system_sleep(); 55 lock_system_sleep();
53 56
54 if (!atomic_add_unless(&snapshot_device_available, -1, 0)) { 57 if (!atomic_add_unless(&snapshot_device_available, -1, 0)) {
diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index ba9ed453c4ed..7de6555cfea0 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -152,10 +152,6 @@ static unsigned long hung_task_timeout_max = (LONG_MAX/HZ);
152#ifdef CONFIG_SPARC 152#ifdef CONFIG_SPARC
153#endif 153#endif
154 154
155#ifdef CONFIG_SPARC64
156extern int sysctl_tsb_ratio;
157#endif
158
159#ifdef __hppa__ 155#ifdef __hppa__
160extern int pwrsw_enabled; 156extern int pwrsw_enabled;
161#endif 157#endif