diff options
| author | Peter Zijlstra <peterz@infradead.org> | 2013-11-07 08:43:43 -0500 |
|---|---|---|
| committer | Ingo Molnar <mingo@kernel.org> | 2014-01-13 07:41:50 -0500 |
| commit | fb00aca474405f4fa8a8519c3179fed722eabd83 (patch) | |
| tree | 8a779629a771dd3340d5a3ba0ba16b732b8de1c8 /kernel/locking | |
| parent | af6ace764d03900524e9b1ac621a1c520ee49fc6 (diff) | |
rtmutex: Turn the plist into an rb-tree
Turn the pi-chains from plist to rb-tree, in the rt_mutex code,
and provide a proper comparison function for -deadline and
-priority tasks.
This is done mainly because:
- classical prio field of the plist is just an int, which might
not be enough for representing a deadline;
- manipulating such a list would become O(nr_deadline_tasks),
which might be to much, as the number of -deadline task increases.
Therefore, an rb-tree is used, and tasks are queued in it according
to the following logic:
- among two -priority (i.e., SCHED_BATCH/OTHER/RR/FIFO) tasks, the
one with the higher (lower, actually!) prio wins;
- among a -priority and a -deadline task, the latter always wins;
- among two -deadline tasks, the one with the earliest deadline
wins.
Queueing and dequeueing functions are changed accordingly, for both
the list of a task's pi-waiters and the list of tasks blocked on
a pi-lock.
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Dario Faggioli <raistlin@linux.it>
Signed-off-by: Juri Lelli <juri.lelli@gmail.com>
Signed-off-again-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1383831828-15501-10-git-send-email-juri.lelli@gmail.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Diffstat (limited to 'kernel/locking')
| -rw-r--r-- | kernel/locking/rtmutex-debug.c | 8 | ||||
| -rw-r--r-- | kernel/locking/rtmutex.c | 151 | ||||
| -rw-r--r-- | kernel/locking/rtmutex_common.h | 22 |
3 files changed, 134 insertions, 47 deletions
diff --git a/kernel/locking/rtmutex-debug.c b/kernel/locking/rtmutex-debug.c index 13b243a323fa..49b2ed3dced8 100644 --- a/kernel/locking/rtmutex-debug.c +++ b/kernel/locking/rtmutex-debug.c | |||
| @@ -24,7 +24,7 @@ | |||
| 24 | #include <linux/kallsyms.h> | 24 | #include <linux/kallsyms.h> |
| 25 | #include <linux/syscalls.h> | 25 | #include <linux/syscalls.h> |
| 26 | #include <linux/interrupt.h> | 26 | #include <linux/interrupt.h> |
| 27 | #include <linux/plist.h> | 27 | #include <linux/rbtree.h> |
| 28 | #include <linux/fs.h> | 28 | #include <linux/fs.h> |
| 29 | #include <linux/debug_locks.h> | 29 | #include <linux/debug_locks.h> |
| 30 | 30 | ||
| @@ -57,7 +57,7 @@ static void printk_lock(struct rt_mutex *lock, int print_owner) | |||
| 57 | 57 | ||
| 58 | void rt_mutex_debug_task_free(struct task_struct *task) | 58 | void rt_mutex_debug_task_free(struct task_struct *task) |
| 59 | { | 59 | { |
| 60 | DEBUG_LOCKS_WARN_ON(!plist_head_empty(&task->pi_waiters)); | 60 | DEBUG_LOCKS_WARN_ON(!RB_EMPTY_ROOT(&task->pi_waiters)); |
| 61 | DEBUG_LOCKS_WARN_ON(task->pi_blocked_on); | 61 | DEBUG_LOCKS_WARN_ON(task->pi_blocked_on); |
| 62 | } | 62 | } |
| 63 | 63 | ||
| @@ -154,16 +154,12 @@ void debug_rt_mutex_proxy_unlock(struct rt_mutex *lock) | |||
| 154 | void debug_rt_mutex_init_waiter(struct rt_mutex_waiter *waiter) | 154 | void debug_rt_mutex_init_waiter(struct rt_mutex_waiter *waiter) |
| 155 | { | 155 | { |
| 156 | memset(waiter, 0x11, sizeof(*waiter)); | 156 | memset(waiter, 0x11, sizeof(*waiter)); |
| 157 | plist_node_init(&waiter->list_entry, MAX_PRIO); | ||
| 158 | plist_node_init(&waiter->pi_list_entry, MAX_PRIO); | ||
| 159 | waiter->deadlock_task_pid = NULL; | 157 | waiter->deadlock_task_pid = NULL; |
| 160 | } | 158 | } |
| 161 | 159 | ||
| 162 | void debug_rt_mutex_free_waiter(struct rt_mutex_waiter *waiter) | 160 | void debug_rt_mutex_free_waiter(struct rt_mutex_waiter *waiter) |
| 163 | { | 161 | { |
| 164 | put_pid(waiter->deadlock_task_pid); | 162 | put_pid(waiter->deadlock_task_pid); |
| 165 | DEBUG_LOCKS_WARN_ON(!plist_node_empty(&waiter->list_entry)); | ||
| 166 | DEBUG_LOCKS_WARN_ON(!plist_node_empty(&waiter->pi_list_entry)); | ||
| 167 | memset(waiter, 0x22, sizeof(*waiter)); | 163 | memset(waiter, 0x22, sizeof(*waiter)); |
| 168 | } | 164 | } |
| 169 | 165 | ||
diff --git a/kernel/locking/rtmutex.c b/kernel/locking/rtmutex.c index 0dd6aec1cb6a..3bf0aa68dd3f 100644 --- a/kernel/locking/rtmutex.c +++ b/kernel/locking/rtmutex.c | |||
| @@ -14,6 +14,7 @@ | |||
| 14 | #include <linux/export.h> | 14 | #include <linux/export.h> |
| 15 | #include <linux/sched.h> | 15 | #include <linux/sched.h> |
| 16 | #include <linux/sched/rt.h> | 16 | #include <linux/sched/rt.h> |
| 17 | #include <linux/sched/deadline.h> | ||
| 17 | #include <linux/timer.h> | 18 | #include <linux/timer.h> |
| 18 | 19 | ||
| 19 | #include "rtmutex_common.h" | 20 | #include "rtmutex_common.h" |
| @@ -91,10 +92,104 @@ static inline void mark_rt_mutex_waiters(struct rt_mutex *lock) | |||
| 91 | } | 92 | } |
| 92 | #endif | 93 | #endif |
| 93 | 94 | ||
| 95 | static inline int | ||
| 96 | rt_mutex_waiter_less(struct rt_mutex_waiter *left, | ||
| 97 | struct rt_mutex_waiter *right) | ||
| 98 | { | ||
| 99 | if (left->task->prio < right->task->prio) | ||
| 100 | return 1; | ||
| 101 | |||
| 102 | /* | ||
| 103 | * If both tasks are dl_task(), we check their deadlines. | ||
| 104 | */ | ||
| 105 | if (dl_prio(left->task->prio) && dl_prio(right->task->prio)) | ||
| 106 | return (left->task->dl.deadline < right->task->dl.deadline); | ||
| 107 | |||
| 108 | return 0; | ||
| 109 | } | ||
| 110 | |||
| 111 | static void | ||
| 112 | rt_mutex_enqueue(struct rt_mutex *lock, struct rt_mutex_waiter *waiter) | ||
| 113 | { | ||
| 114 | struct rb_node **link = &lock->waiters.rb_node; | ||
| 115 | struct rb_node *parent = NULL; | ||
| 116 | struct rt_mutex_waiter *entry; | ||
| 117 | int leftmost = 1; | ||
| 118 | |||
| 119 | while (*link) { | ||
| 120 | parent = *link; | ||
| 121 | entry = rb_entry(parent, struct rt_mutex_waiter, tree_entry); | ||
| 122 | if (rt_mutex_waiter_less(waiter, entry)) { | ||
| 123 | link = &parent->rb_left; | ||
| 124 | } else { | ||
| 125 | link = &parent->rb_right; | ||
| 126 | leftmost = 0; | ||
| 127 | } | ||
| 128 | } | ||
| 129 | |||
| 130 | if (leftmost) | ||
| 131 | lock->waiters_leftmost = &waiter->tree_entry; | ||
| 132 | |||
| 133 | rb_link_node(&waiter->tree_entry, parent, link); | ||
| 134 | rb_insert_color(&waiter->tree_entry, &lock->waiters); | ||
| 135 | } | ||
| 136 | |||
| 137 | static void | ||
| 138 | rt_mutex_dequeue(struct rt_mutex *lock, struct rt_mutex_waiter *waiter) | ||
| 139 | { | ||
| 140 | if (RB_EMPTY_NODE(&waiter->tree_entry)) | ||
| 141 | return; | ||
| 142 | |||
| 143 | if (lock->waiters_leftmost == &waiter->tree_entry) | ||
| 144 | lock->waiters_leftmost = rb_next(&waiter->tree_entry); | ||
| 145 | |||
| 146 | rb_erase(&waiter->tree_entry, &lock->waiters); | ||
| 147 | RB_CLEAR_NODE(&waiter->tree_entry); | ||
| 148 | } | ||
| 149 | |||
| 150 | static void | ||
| 151 | rt_mutex_enqueue_pi(struct task_struct *task, struct rt_mutex_waiter *waiter) | ||
| 152 | { | ||
| 153 | struct rb_node **link = &task->pi_waiters.rb_node; | ||
| 154 | struct rb_node *parent = NULL; | ||
| 155 | struct rt_mutex_waiter *entry; | ||
| 156 | int leftmost = 1; | ||
| 157 | |||
| 158 | while (*link) { | ||
| 159 | parent = *link; | ||
| 160 | entry = rb_entry(parent, struct rt_mutex_waiter, pi_tree_entry); | ||
| 161 | if (rt_mutex_waiter_less(waiter, entry)) { | ||
| 162 | link = &parent->rb_left; | ||
| 163 | } else { | ||
| 164 | link = &parent->rb_right; | ||
| 165 | leftmost = 0; | ||
| 166 | } | ||
| 167 | } | ||
| 168 | |||
| 169 | if (leftmost) | ||
| 170 | task->pi_waiters_leftmost = &waiter->pi_tree_entry; | ||
| 171 | |||
| 172 | rb_link_node(&waiter->pi_tree_entry, parent, link); | ||
| 173 | rb_insert_color(&waiter->pi_tree_entry, &task->pi_waiters); | ||
| 174 | } | ||
| 175 | |||
| 176 | static void | ||
| 177 | rt_mutex_dequeue_pi(struct task_struct *task, struct rt_mutex_waiter *waiter) | ||
| 178 | { | ||
| 179 | if (RB_EMPTY_NODE(&waiter->pi_tree_entry)) | ||
| 180 | return; | ||
| 181 | |||
| 182 | if (task->pi_waiters_leftmost == &waiter->pi_tree_entry) | ||
| 183 | task->pi_waiters_leftmost = rb_next(&waiter->pi_tree_entry); | ||
| 184 | |||
| 185 | rb_erase(&waiter->pi_tree_entry, &task->pi_waiters); | ||
| 186 | RB_CLEAR_NODE(&waiter->pi_tree_entry); | ||
| 187 | } | ||
| 188 | |||
| 94 | /* | 189 | /* |
| 95 | * Calculate task priority from the waiter list priority | 190 | * Calculate task priority from the waiter tree priority |
| 96 | * | 191 | * |
| 97 | * Return task->normal_prio when the waiter list is empty or when | 192 | * Return task->normal_prio when the waiter tree is empty or when |
| 98 | * the waiter is not allowed to do priority boosting | 193 | * the waiter is not allowed to do priority boosting |
| 99 | */ | 194 | */ |
| 100 | int rt_mutex_getprio(struct task_struct *task) | 195 | int rt_mutex_getprio(struct task_struct *task) |
| @@ -102,7 +197,7 @@ int rt_mutex_getprio(struct task_struct *task) | |||
| 102 | if (likely(!task_has_pi_waiters(task))) | 197 | if (likely(!task_has_pi_waiters(task))) |
| 103 | return task->normal_prio; | 198 | return task->normal_prio; |
| 104 | 199 | ||
| 105 | return min(task_top_pi_waiter(task)->pi_list_entry.prio, | 200 | return min(task_top_pi_waiter(task)->task->prio, |
| 106 | task->normal_prio); | 201 | task->normal_prio); |
| 107 | } | 202 | } |
| 108 | 203 | ||
| @@ -233,7 +328,7 @@ static int rt_mutex_adjust_prio_chain(struct task_struct *task, | |||
| 233 | * When deadlock detection is off then we check, if further | 328 | * When deadlock detection is off then we check, if further |
| 234 | * priority adjustment is necessary. | 329 | * priority adjustment is necessary. |
| 235 | */ | 330 | */ |
| 236 | if (!detect_deadlock && waiter->list_entry.prio == task->prio) | 331 | if (!detect_deadlock && waiter->task->prio == task->prio) |
