Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* rwsem.c: R/W semaphores: contention handling functions |
| 2 | * |
| 3 | * Written by David Howells (dhowells@redhat.com). |
| 4 | * Derived from arch/i386/kernel/semaphore.c |
| 5 | */ |
| 6 | #include <linux/rwsem.h> |
| 7 | #include <linux/sched.h> |
| 8 | #include <linux/init.h> |
| 9 | #include <linux/module.h> |
| 10 | |
| 11 | struct rwsem_waiter { |
| 12 | struct list_head list; |
| 13 | struct task_struct *task; |
| 14 | unsigned int flags; |
| 15 | #define RWSEM_WAITING_FOR_READ 0x00000001 |
| 16 | #define RWSEM_WAITING_FOR_WRITE 0x00000002 |
| 17 | }; |
| 18 | |
| 19 | #if RWSEM_DEBUG |
| 20 | #undef rwsemtrace |
| 21 | void rwsemtrace(struct rw_semaphore *sem, const char *str) |
| 22 | { |
| 23 | printk("sem=%p\n", sem); |
| 24 | printk("(sem)=%08lx\n", sem->count); |
| 25 | if (sem->debug) |
| 26 | printk("[%d] %s({%08lx})\n", current->pid, str, sem->count); |
| 27 | } |
| 28 | #endif |
| 29 | |
| 30 | /* |
| 31 | * handle the lock release when processes blocked on it that can now run |
| 32 | * - if we come here from up_xxxx(), then: |
| 33 | * - the 'active part' of count (&0x0000ffff) reached 0 (but may have changed) |
| 34 | * - the 'waiting part' of count (&0xffff0000) is -ve (and will still be so) |
| 35 | * - there must be someone on the queue |
| 36 | * - the spinlock must be held by the caller |
| 37 | * - woken process blocks are discarded from the list after having task zeroed |
| 38 | * - writers are only woken if downgrading is false |
| 39 | */ |
| 40 | static inline struct rw_semaphore * |
| 41 | __rwsem_do_wake(struct rw_semaphore *sem, int downgrading) |
| 42 | { |
| 43 | struct rwsem_waiter *waiter; |
| 44 | struct task_struct *tsk; |
| 45 | struct list_head *next; |
| 46 | signed long oldcount, woken, loop; |
| 47 | |
| 48 | rwsemtrace(sem, "Entering __rwsem_do_wake"); |
| 49 | |
| 50 | if (downgrading) |
| 51 | goto dont_wake_writers; |
| 52 | |
| 53 | /* if we came through an up_xxxx() call, we only only wake someone up |
| 54 | * if we can transition the active part of the count from 0 -> 1 |
| 55 | */ |
| 56 | try_again: |
| 57 | oldcount = rwsem_atomic_update(RWSEM_ACTIVE_BIAS, sem) |
| 58 | - RWSEM_ACTIVE_BIAS; |
| 59 | if (oldcount & RWSEM_ACTIVE_MASK) |
| 60 | goto undo; |
| 61 | |
| 62 | waiter = list_entry(sem->wait_list.next, struct rwsem_waiter, list); |
| 63 | |
| 64 | /* try to grant a single write lock if there's a writer at the front |
| 65 | * of the queue - note we leave the 'active part' of the count |
| 66 | * incremented by 1 and the waiting part incremented by 0x00010000 |
| 67 | */ |
| 68 | if (!(waiter->flags & RWSEM_WAITING_FOR_WRITE)) |
| 69 | goto readers_only; |
| 70 | |
| 71 | /* We must be careful not to touch 'waiter' after we set ->task = NULL. |
| 72 | * It is an allocated on the waiter's stack and may become invalid at |
| 73 | * any time after that point (due to a wakeup from another source). |
| 74 | */ |
| 75 | list_del(&waiter->list); |
| 76 | tsk = waiter->task; |
akpm@osdl.org | d59dd46 | 2005-05-01 08:58:47 -0700 | [diff] [blame] | 77 | smp_mb(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 78 | waiter->task = NULL; |
| 79 | wake_up_process(tsk); |
| 80 | put_task_struct(tsk); |
| 81 | goto out; |
| 82 | |
| 83 | /* don't want to wake any writers */ |
| 84 | dont_wake_writers: |
| 85 | waiter = list_entry(sem->wait_list.next, struct rwsem_waiter, list); |
| 86 | if (waiter->flags & RWSEM_WAITING_FOR_WRITE) |
| 87 | goto out; |
| 88 | |
| 89 | /* grant an infinite number of read locks to the readers at the front |
| 90 | * of the queue |
| 91 | * - note we increment the 'active part' of the count by the number of |
| 92 | * readers before waking any processes up |
| 93 | */ |
| 94 | readers_only: |
| 95 | woken = 0; |
| 96 | do { |
| 97 | woken++; |
| 98 | |
| 99 | if (waiter->list.next == &sem->wait_list) |
| 100 | break; |
| 101 | |
| 102 | waiter = list_entry(waiter->list.next, |
| 103 | struct rwsem_waiter, list); |
| 104 | |
| 105 | } while (waiter->flags & RWSEM_WAITING_FOR_READ); |
| 106 | |
| 107 | loop = woken; |
| 108 | woken *= RWSEM_ACTIVE_BIAS - RWSEM_WAITING_BIAS; |
| 109 | if (!downgrading) |
| 110 | /* we'd already done one increment earlier */ |
| 111 | woken -= RWSEM_ACTIVE_BIAS; |
| 112 | |
| 113 | rwsem_atomic_add(woken, sem); |
| 114 | |
| 115 | next = sem->wait_list.next; |
| 116 | for (; loop > 0; loop--) { |
| 117 | waiter = list_entry(next, struct rwsem_waiter, list); |
| 118 | next = waiter->list.next; |
| 119 | tsk = waiter->task; |
akpm@osdl.org | d59dd46 | 2005-05-01 08:58:47 -0700 | [diff] [blame] | 120 | smp_mb(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 121 | waiter->task = NULL; |
| 122 | wake_up_process(tsk); |
| 123 | put_task_struct(tsk); |
| 124 | } |
| 125 | |
| 126 | sem->wait_list.next = next; |
| 127 | next->prev = &sem->wait_list; |
| 128 | |
| 129 | out: |
| 130 | rwsemtrace(sem, "Leaving __rwsem_do_wake"); |
| 131 | return sem; |
| 132 | |
| 133 | /* undo the change to count, but check for a transition 1->0 */ |
| 134 | undo: |
| 135 | if (rwsem_atomic_update(-RWSEM_ACTIVE_BIAS, sem) != 0) |
| 136 | goto out; |
| 137 | goto try_again; |
| 138 | } |
| 139 | |
| 140 | /* |
| 141 | * wait for a lock to be granted |
| 142 | */ |
| 143 | static inline struct rw_semaphore * |
| 144 | rwsem_down_failed_common(struct rw_semaphore *sem, |
| 145 | struct rwsem_waiter *waiter, signed long adjustment) |
| 146 | { |
| 147 | struct task_struct *tsk = current; |
| 148 | signed long count; |
| 149 | |
| 150 | set_task_state(tsk, TASK_UNINTERRUPTIBLE); |
| 151 | |
| 152 | /* set up my own style of waitqueue */ |
| 153 | spin_lock_irq(&sem->wait_lock); |
| 154 | waiter->task = tsk; |
| 155 | get_task_struct(tsk); |
| 156 | |
| 157 | list_add_tail(&waiter->list, &sem->wait_list); |
| 158 | |
| 159 | /* we're now waiting on the lock, but no longer actively read-locking */ |
| 160 | count = rwsem_atomic_update(adjustment, sem); |
| 161 | |
| 162 | /* if there are no active locks, wake the front queued process(es) up */ |
| 163 | if (!(count & RWSEM_ACTIVE_MASK)) |
| 164 | sem = __rwsem_do_wake(sem, 0); |
| 165 | |
| 166 | spin_unlock_irq(&sem->wait_lock); |
| 167 | |
| 168 | /* wait to be given the lock */ |
| 169 | for (;;) { |
| 170 | if (!waiter->task) |
| 171 | break; |
| 172 | schedule(); |
| 173 | set_task_state(tsk, TASK_UNINTERRUPTIBLE); |
| 174 | } |
| 175 | |
| 176 | tsk->state = TASK_RUNNING; |
| 177 | |
| 178 | return sem; |
| 179 | } |
| 180 | |
| 181 | /* |
| 182 | * wait for the read lock to be granted |
| 183 | */ |
| 184 | struct rw_semaphore fastcall __sched * |
| 185 | rwsem_down_read_failed(struct rw_semaphore *sem) |
| 186 | { |
| 187 | struct rwsem_waiter waiter; |
| 188 | |
| 189 | rwsemtrace(sem, "Entering rwsem_down_read_failed"); |
| 190 | |
| 191 | waiter.flags = RWSEM_WAITING_FOR_READ; |
| 192 | rwsem_down_failed_common(sem, &waiter, |
| 193 | RWSEM_WAITING_BIAS - RWSEM_ACTIVE_BIAS); |
| 194 | |
| 195 | rwsemtrace(sem, "Leaving rwsem_down_read_failed"); |
| 196 | return sem; |
| 197 | } |
| 198 | |
| 199 | /* |
| 200 | * wait for the write lock to be granted |
| 201 | */ |
| 202 | struct rw_semaphore fastcall __sched * |
| 203 | rwsem_down_write_failed(struct rw_semaphore *sem) |
| 204 | { |
| 205 | struct rwsem_waiter waiter; |
| 206 | |
| 207 | rwsemtrace(sem, "Entering rwsem_down_write_failed"); |
| 208 | |
| 209 | waiter.flags = RWSEM_WAITING_FOR_WRITE; |
| 210 | rwsem_down_failed_common(sem, &waiter, -RWSEM_ACTIVE_BIAS); |
| 211 | |
| 212 | rwsemtrace(sem, "Leaving rwsem_down_write_failed"); |
| 213 | return sem; |
| 214 | } |
| 215 | |
| 216 | /* |
| 217 | * handle waking up a waiter on the semaphore |
| 218 | * - up_read/up_write has decremented the active part of count if we come here |
| 219 | */ |
| 220 | struct rw_semaphore fastcall *rwsem_wake(struct rw_semaphore *sem) |
| 221 | { |
| 222 | unsigned long flags; |
| 223 | |
| 224 | rwsemtrace(sem, "Entering rwsem_wake"); |
| 225 | |
| 226 | spin_lock_irqsave(&sem->wait_lock, flags); |
| 227 | |
| 228 | /* do nothing if list empty */ |
| 229 | if (!list_empty(&sem->wait_list)) |
| 230 | sem = __rwsem_do_wake(sem, 0); |
| 231 | |
| 232 | spin_unlock_irqrestore(&sem->wait_lock, flags); |
| 233 | |
| 234 | rwsemtrace(sem, "Leaving rwsem_wake"); |
| 235 | |
| 236 | return sem; |
| 237 | } |
| 238 | |
| 239 | /* |
| 240 | * downgrade a write lock into a read lock |
| 241 | * - caller incremented waiting part of count and discovered it still negative |
| 242 | * - just wake up any readers at the front of the queue |
| 243 | */ |
| 244 | struct rw_semaphore fastcall *rwsem_downgrade_wake(struct rw_semaphore *sem) |
| 245 | { |
| 246 | unsigned long flags; |
| 247 | |
| 248 | rwsemtrace(sem, "Entering rwsem_downgrade_wake"); |
| 249 | |
| 250 | spin_lock_irqsave(&sem->wait_lock, flags); |
| 251 | |
| 252 | /* do nothing if list empty */ |
| 253 | if (!list_empty(&sem->wait_list)) |
| 254 | sem = __rwsem_do_wake(sem, 1); |
| 255 | |
| 256 | spin_unlock_irqrestore(&sem->wait_lock, flags); |
| 257 | |
| 258 | rwsemtrace(sem, "Leaving rwsem_downgrade_wake"); |
| 259 | return sem; |
| 260 | } |
| 261 | |
| 262 | EXPORT_SYMBOL(rwsem_down_read_failed); |
| 263 | EXPORT_SYMBOL(rwsem_down_write_failed); |
| 264 | EXPORT_SYMBOL(rwsem_wake); |
| 265 | EXPORT_SYMBOL(rwsem_downgrade_wake); |
| 266 | #if RWSEM_DEBUG |
| 267 | EXPORT_SYMBOL(rwsemtrace); |
| 268 | #endif |