lh | 9ed821d | 2023-04-07 01:36:19 -0700 | [diff] [blame] | 1 | /* |
| 2 | * fs/timerfd.c |
| 3 | * |
| 4 | * Copyright (C) 2007 Davide Libenzi <davidel@xmailserver.org> |
| 5 | * |
| 6 | * |
| 7 | * Thanks to Thomas Gleixner for code reviews and useful comments. |
| 8 | * |
| 9 | */ |
| 10 | |
| 11 | #include <linux/file.h> |
| 12 | #include <linux/poll.h> |
| 13 | #include <linux/init.h> |
| 14 | #include <linux/fs.h> |
| 15 | #include <linux/sched.h> |
| 16 | #include <linux/kernel.h> |
| 17 | #include <linux/slab.h> |
| 18 | #include <linux/list.h> |
| 19 | #include <linux/spinlock.h> |
| 20 | #include <linux/time.h> |
| 21 | #include <linux/hrtimer.h> |
| 22 | #include <linux/anon_inodes.h> |
| 23 | #include <linux/timerfd.h> |
| 24 | #include <linux/syscalls.h> |
| 25 | #include <linux/rcupdate.h> |
| 26 | |
| 27 | struct timerfd_ctx { |
| 28 | struct hrtimer tmr; |
| 29 | ktime_t tintv; |
| 30 | ktime_t moffs; |
| 31 | wait_queue_head_t wqh; |
| 32 | u64 ticks; |
| 33 | int expired; |
| 34 | int clockid; |
| 35 | struct rcu_head rcu; |
| 36 | struct list_head clist; |
| 37 | spinlock_t cancel_lock; |
| 38 | bool might_cancel; |
| 39 | }; |
| 40 | |
| 41 | static LIST_HEAD(cancel_list); |
| 42 | static DEFINE_SPINLOCK(cancel_lock); |
| 43 | |
| 44 | /* |
| 45 | * This gets called when the timer event triggers. We set the "expired" |
| 46 | * flag, but we do not re-arm the timer (in case it's necessary, |
| 47 | * tintv.tv64 != 0) until the timer is accessed. |
| 48 | */ |
| 49 | static enum hrtimer_restart timerfd_tmrproc(struct hrtimer *htmr) |
| 50 | { |
| 51 | struct timerfd_ctx *ctx = container_of(htmr, struct timerfd_ctx, tmr); |
| 52 | unsigned long flags; |
| 53 | |
| 54 | spin_lock_irqsave(&ctx->wqh.lock, flags); |
| 55 | ctx->expired = 1; |
| 56 | ctx->ticks++; |
| 57 | wake_up_locked(&ctx->wqh); |
| 58 | spin_unlock_irqrestore(&ctx->wqh.lock, flags); |
| 59 | |
| 60 | return HRTIMER_NORESTART; |
| 61 | } |
| 62 | |
| 63 | /* |
| 64 | * Called when the clock was set to cancel the timers in the cancel |
| 65 | * list. This will wake up processes waiting on these timers. The |
| 66 | * wake-up requires ctx->ticks to be non zero, therefore we increment |
| 67 | * it before calling wake_up_locked(). |
| 68 | */ |
| 69 | void timerfd_clock_was_set(void) |
| 70 | { |
| 71 | ktime_t moffs = ktime_get_monotonic_offset(); |
| 72 | struct timerfd_ctx *ctx; |
| 73 | unsigned long flags; |
| 74 | |
| 75 | rcu_read_lock(); |
| 76 | list_for_each_entry_rcu(ctx, &cancel_list, clist) { |
| 77 | if (!ctx->might_cancel) |
| 78 | continue; |
| 79 | spin_lock_irqsave(&ctx->wqh.lock, flags); |
| 80 | if (ctx->moffs.tv64 != moffs.tv64) { |
| 81 | ctx->moffs.tv64 = KTIME_MAX; |
| 82 | ctx->ticks++; |
| 83 | wake_up_locked(&ctx->wqh); |
| 84 | } |
| 85 | spin_unlock_irqrestore(&ctx->wqh.lock, flags); |
| 86 | } |
| 87 | rcu_read_unlock(); |
| 88 | } |
| 89 | |
| 90 | static void __timerfd_remove_cancel(struct timerfd_ctx *ctx) |
| 91 | { |
| 92 | if (ctx->might_cancel) { |
| 93 | ctx->might_cancel = false; |
| 94 | spin_lock(&cancel_lock); |
| 95 | list_del_rcu(&ctx->clist); |
| 96 | spin_unlock(&cancel_lock); |
| 97 | } |
| 98 | } |
| 99 | static void timerfd_remove_cancel(struct timerfd_ctx *ctx) |
| 100 | { |
| 101 | spin_lock(&ctx->cancel_lock); |
| 102 | __timerfd_remove_cancel(ctx); |
| 103 | spin_unlock(&ctx->cancel_lock); |
| 104 | } |
| 105 | |
| 106 | static bool timerfd_canceled(struct timerfd_ctx *ctx) |
| 107 | { |
| 108 | if (!ctx->might_cancel || ctx->moffs.tv64 != KTIME_MAX) |
| 109 | return false; |
| 110 | ctx->moffs = ktime_get_monotonic_offset(); |
| 111 | return true; |
| 112 | } |
| 113 | |
| 114 | static void timerfd_setup_cancel(struct timerfd_ctx *ctx, int flags) |
| 115 | { |
| 116 | spin_lock(&ctx->cancel_lock); |
| 117 | if (ctx->clockid == CLOCK_REALTIME && (flags & TFD_TIMER_ABSTIME) && |
| 118 | (flags & TFD_TIMER_CANCEL_ON_SET)) { |
| 119 | if (!ctx->might_cancel) { |
| 120 | ctx->might_cancel = true; |
| 121 | spin_lock(&cancel_lock); |
| 122 | list_add_rcu(&ctx->clist, &cancel_list); |
| 123 | spin_unlock(&cancel_lock); |
| 124 | } |
| 125 | } else { |
| 126 | __timerfd_remove_cancel(ctx); |
| 127 | } |
| 128 | spin_unlock(&ctx->cancel_lock); |
| 129 | } |
| 130 | |
| 131 | static ktime_t timerfd_get_remaining(struct timerfd_ctx *ctx) |
| 132 | { |
| 133 | ktime_t remaining; |
| 134 | |
| 135 | remaining = hrtimer_expires_remaining(&ctx->tmr); |
| 136 | return remaining.tv64 < 0 ? ktime_set(0, 0): remaining; |
| 137 | } |
| 138 | |
| 139 | static int timerfd_setup(struct timerfd_ctx *ctx, int flags, |
| 140 | const struct itimerspec *ktmr) |
| 141 | { |
| 142 | enum hrtimer_mode htmode; |
| 143 | ktime_t texp; |
| 144 | int clockid = ctx->clockid; |
| 145 | |
| 146 | htmode = (flags & TFD_TIMER_ABSTIME) ? |
| 147 | HRTIMER_MODE_ABS: HRTIMER_MODE_REL; |
| 148 | |
| 149 | texp = timespec_to_ktime(ktmr->it_value); |
| 150 | ctx->expired = 0; |
| 151 | ctx->ticks = 0; |
| 152 | ctx->tintv = timespec_to_ktime(ktmr->it_interval); |
| 153 | hrtimer_init(&ctx->tmr, clockid, htmode); |
| 154 | hrtimer_set_expires(&ctx->tmr, texp); |
| 155 | ctx->tmr.function = timerfd_tmrproc; |
| 156 | if (texp.tv64 != 0) { |
| 157 | hrtimer_start(&ctx->tmr, texp, htmode); |
| 158 | if (timerfd_canceled(ctx)) |
| 159 | return -ECANCELED; |
| 160 | } |
| 161 | return 0; |
| 162 | } |
| 163 | |
| 164 | static int timerfd_release(struct inode *inode, struct file *file) |
| 165 | { |
| 166 | struct timerfd_ctx *ctx = file->private_data; |
| 167 | |
| 168 | timerfd_remove_cancel(ctx); |
| 169 | hrtimer_cancel(&ctx->tmr); |
| 170 | kfree_rcu(ctx, rcu); |
| 171 | return 0; |
| 172 | } |
| 173 | |
| 174 | static unsigned int timerfd_poll(struct file *file, poll_table *wait) |
| 175 | { |
| 176 | struct timerfd_ctx *ctx = file->private_data; |
| 177 | unsigned int events = 0; |
| 178 | unsigned long flags; |
| 179 | |
| 180 | poll_wait(file, &ctx->wqh, wait); |
| 181 | |
| 182 | spin_lock_irqsave(&ctx->wqh.lock, flags); |
| 183 | if (ctx->ticks) |
| 184 | events |= POLLIN; |
| 185 | spin_unlock_irqrestore(&ctx->wqh.lock, flags); |
| 186 | |
| 187 | return events; |
| 188 | } |
| 189 | |
| 190 | static ssize_t timerfd_read(struct file *file, char __user *buf, size_t count, |
| 191 | loff_t *ppos) |
| 192 | { |
| 193 | struct timerfd_ctx *ctx = file->private_data; |
| 194 | ssize_t res; |
| 195 | u64 ticks = 0; |
| 196 | |
| 197 | if (count < sizeof(ticks)) |
| 198 | return -EINVAL; |
| 199 | spin_lock_irq(&ctx->wqh.lock); |
| 200 | if (file->f_flags & O_NONBLOCK) |
| 201 | res = -EAGAIN; |
| 202 | else |
| 203 | res = wait_event_interruptible_locked_irq(ctx->wqh, ctx->ticks); |
| 204 | |
| 205 | /* |
| 206 | * If clock has changed, we do not care about the |
| 207 | * ticks and we do not rearm the timer. Userspace must |
| 208 | * reevaluate anyway. |
| 209 | */ |
| 210 | if (timerfd_canceled(ctx)) { |
| 211 | ctx->ticks = 0; |
| 212 | ctx->expired = 0; |
| 213 | res = -ECANCELED; |
| 214 | } |
| 215 | |
| 216 | if (ctx->ticks) { |
| 217 | ticks = ctx->ticks; |
| 218 | |
| 219 | if (ctx->expired && ctx->tintv.tv64) { |
| 220 | /* |
| 221 | * If tintv.tv64 != 0, this is a periodic timer that |
| 222 | * needs to be re-armed. We avoid doing it in the timer |
| 223 | * callback to avoid DoS attacks specifying a very |
| 224 | * short timer period. |
| 225 | */ |
| 226 | ticks += hrtimer_forward_now(&ctx->tmr, |
| 227 | ctx->tintv) - 1; |
| 228 | hrtimer_restart(&ctx->tmr); |
| 229 | } |
| 230 | ctx->expired = 0; |
| 231 | ctx->ticks = 0; |
| 232 | } |
| 233 | spin_unlock_irq(&ctx->wqh.lock); |
| 234 | if (ticks) |
| 235 | res = put_user(ticks, (u64 __user *) buf) ? -EFAULT: sizeof(ticks); |
| 236 | return res; |
| 237 | } |
| 238 | |
| 239 | static const struct file_operations timerfd_fops = { |
| 240 | .release = timerfd_release, |
| 241 | .poll = timerfd_poll, |
| 242 | .read = timerfd_read, |
| 243 | .llseek = noop_llseek, |
| 244 | }; |
| 245 | |
| 246 | static struct file *timerfd_fget(int fd) |
| 247 | { |
| 248 | struct file *file; |
| 249 | |
| 250 | file = fget(fd); |
| 251 | if (!file) |
| 252 | return ERR_PTR(-EBADF); |
| 253 | if (file->f_op != &timerfd_fops) { |
| 254 | fput(file); |
| 255 | return ERR_PTR(-EINVAL); |
| 256 | } |
| 257 | |
| 258 | return file; |
| 259 | } |
| 260 | |
| 261 | SYSCALL_DEFINE2(timerfd_create, int, clockid, int, flags) |
| 262 | { |
| 263 | int ufd; |
| 264 | struct timerfd_ctx *ctx; |
| 265 | |
| 266 | /* Check the TFD_* constants for consistency. */ |
| 267 | BUILD_BUG_ON(TFD_CLOEXEC != O_CLOEXEC); |
| 268 | BUILD_BUG_ON(TFD_NONBLOCK != O_NONBLOCK); |
| 269 | |
| 270 | if ((flags & ~TFD_CREATE_FLAGS) || |
| 271 | (clockid != CLOCK_MONOTONIC && |
| 272 | clockid != CLOCK_REALTIME)) |
| 273 | return -EINVAL; |
| 274 | |
| 275 | ctx = kzalloc(sizeof(*ctx), GFP_KERNEL); |
| 276 | if (!ctx) |
| 277 | return -ENOMEM; |
| 278 | |
| 279 | init_waitqueue_head(&ctx->wqh); |
| 280 | spin_lock_init(&ctx->cancel_lock); |
| 281 | ctx->clockid = clockid; |
| 282 | hrtimer_init(&ctx->tmr, clockid, HRTIMER_MODE_ABS); |
| 283 | ctx->moffs = ktime_get_monotonic_offset(); |
| 284 | |
| 285 | ufd = anon_inode_getfd("[timerfd]", &timerfd_fops, ctx, |
| 286 | O_RDWR | (flags & TFD_SHARED_FCNTL_FLAGS)); |
| 287 | if (ufd < 0) |
| 288 | kfree(ctx); |
| 289 | |
| 290 | return ufd; |
| 291 | } |
| 292 | |
| 293 | SYSCALL_DEFINE4(timerfd_settime, int, ufd, int, flags, |
| 294 | const struct itimerspec __user *, utmr, |
| 295 | struct itimerspec __user *, otmr) |
| 296 | { |
| 297 | struct file *file; |
| 298 | struct timerfd_ctx *ctx; |
| 299 | struct itimerspec ktmr, kotmr; |
| 300 | int ret; |
| 301 | |
| 302 | if (copy_from_user(&ktmr, utmr, sizeof(ktmr))) |
| 303 | return -EFAULT; |
| 304 | |
| 305 | if ((flags & ~TFD_SETTIME_FLAGS) || |
| 306 | !timespec_valid(&ktmr.it_value) || |
| 307 | !timespec_valid(&ktmr.it_interval)) |
| 308 | return -EINVAL; |
| 309 | |
| 310 | file = timerfd_fget(ufd); |
| 311 | if (IS_ERR(file)) |
| 312 | return PTR_ERR(file); |
| 313 | ctx = file->private_data; |
| 314 | |
| 315 | timerfd_setup_cancel(ctx, flags); |
| 316 | |
| 317 | /* |
| 318 | * We need to stop the existing timer before reprogramming |
| 319 | * it to the new values. |
| 320 | */ |
| 321 | for (;;) { |
| 322 | spin_lock_irq(&ctx->wqh.lock); |
| 323 | if (hrtimer_try_to_cancel(&ctx->tmr) >= 0) |
| 324 | break; |
| 325 | spin_unlock_irq(&ctx->wqh.lock); |
| 326 | hrtimer_wait_for_timer(&ctx->tmr); |
| 327 | } |
| 328 | |
| 329 | /* |
| 330 | * If the timer is expired and it's periodic, we need to advance it |
| 331 | * because the caller may want to know the previous expiration time. |
| 332 | * We do not update "ticks" and "expired" since the timer will be |
| 333 | * re-programmed again in the following timerfd_setup() call. |
| 334 | */ |
| 335 | if (ctx->expired && ctx->tintv.tv64) |
| 336 | hrtimer_forward_now(&ctx->tmr, ctx->tintv); |
| 337 | |
| 338 | kotmr.it_value = ktime_to_timespec(timerfd_get_remaining(ctx)); |
| 339 | kotmr.it_interval = ktime_to_timespec(ctx->tintv); |
| 340 | |
| 341 | /* |
| 342 | * Re-program the timer to the new value ... |
| 343 | */ |
| 344 | ret = timerfd_setup(ctx, flags, &ktmr); |
| 345 | |
| 346 | spin_unlock_irq(&ctx->wqh.lock); |
| 347 | fput(file); |
| 348 | if (otmr && copy_to_user(otmr, &kotmr, sizeof(kotmr))) |
| 349 | return -EFAULT; |
| 350 | |
| 351 | return ret; |
| 352 | } |
| 353 | |
| 354 | SYSCALL_DEFINE2(timerfd_gettime, int, ufd, struct itimerspec __user *, otmr) |
| 355 | { |
| 356 | struct file *file; |
| 357 | struct timerfd_ctx *ctx; |
| 358 | struct itimerspec kotmr; |
| 359 | |
| 360 | file = timerfd_fget(ufd); |
| 361 | if (IS_ERR(file)) |
| 362 | return PTR_ERR(file); |
| 363 | ctx = file->private_data; |
| 364 | |
| 365 | spin_lock_irq(&ctx->wqh.lock); |
| 366 | if (ctx->expired && ctx->tintv.tv64) { |
| 367 | ctx->expired = 0; |
| 368 | ctx->ticks += |
| 369 | hrtimer_forward_now(&ctx->tmr, ctx->tintv) - 1; |
| 370 | hrtimer_restart(&ctx->tmr); |
| 371 | } |
| 372 | kotmr.it_value = ktime_to_timespec(timerfd_get_remaining(ctx)); |
| 373 | kotmr.it_interval = ktime_to_timespec(ctx->tintv); |
| 374 | spin_unlock_irq(&ctx->wqh.lock); |
| 375 | fput(file); |
| 376 | |
| 377 | return copy_to_user(otmr, &kotmr, sizeof(kotmr)) ? -EFAULT: 0; |
| 378 | } |
| 379 | |