1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * Contains the core associated with submission side polling of the SQ 4 * ring, offloading submissions from the application to a kernel thread. 5 */ 6 #include <linux/kernel.h> 7 #include <linux/errno.h> 8 #include <linux/file.h> 9 #include <linux/mm.h> 10 #include <linux/slab.h> 11 #include <linux/audit.h> 12 #include <linux/security.h> 13 #include <linux/cpuset.h> 14 #include <linux/io_uring.h> 15 16 #include <uapi/linux/io_uring.h> 17 18 #include "io_uring.h" 19 #include "napi.h" 20 #include "sqpoll.h" 21 22 #define IORING_SQPOLL_CAP_ENTRIES_VALUE 8 23 #define IORING_TW_CAP_ENTRIES_VALUE 8 24 25 enum { 26 IO_SQ_THREAD_SHOULD_STOP = 0, 27 IO_SQ_THREAD_SHOULD_PARK, 28 }; 29 30 void io_sq_thread_unpark(struct io_sq_data *sqd) 31 __releases(&sqd->lock) 32 { 33 WARN_ON_ONCE(sqd->thread == current); 34 35 /* 36 * Do the dance but not conditional clear_bit() because it'd race with 37 * other threads incrementing park_pending and setting the bit. 38 */ 39 clear_bit(IO_SQ_THREAD_SHOULD_PARK, &sqd->state); 40 if (atomic_dec_return(&sqd->park_pending)) 41 set_bit(IO_SQ_THREAD_SHOULD_PARK, &sqd->state); 42 mutex_unlock(&sqd->lock); 43 } 44 45 void io_sq_thread_park(struct io_sq_data *sqd) 46 __acquires(&sqd->lock) 47 { 48 WARN_ON_ONCE(data_race(sqd->thread) == current); 49 50 atomic_inc(&sqd->park_pending); 51 set_bit(IO_SQ_THREAD_SHOULD_PARK, &sqd->state); 52 mutex_lock(&sqd->lock); 53 if (sqd->thread) 54 wake_up_process(sqd->thread); 55 } 56 57 void io_sq_thread_stop(struct io_sq_data *sqd) 58 { 59 WARN_ON_ONCE(sqd->thread == current); 60 WARN_ON_ONCE(test_bit(IO_SQ_THREAD_SHOULD_STOP, &sqd->state)); 61 62 set_bit(IO_SQ_THREAD_SHOULD_STOP, &sqd->state); 63 mutex_lock(&sqd->lock); 64 if (sqd->thread) 65 wake_up_process(sqd->thread); 66 mutex_unlock(&sqd->lock); 67 wait_for_completion(&sqd->exited); 68 } 69 70 void io_put_sq_data(struct io_sq_data *sqd) 71 { 72 if (refcount_dec_and_test(&sqd->refs)) { 73 WARN_ON_ONCE(atomic_read(&sqd->park_pending)); 74 75 io_sq_thread_stop(sqd); 76 kfree(sqd); 77 } 78 } 79 80 static __cold void io_sqd_update_thread_idle(struct io_sq_data *sqd) 81 { 82 struct io_ring_ctx *ctx; 83 unsigned sq_thread_idle = 0; 84 85 list_for_each_entry(ctx, &sqd->ctx_list, sqd_list) 86 sq_thread_idle = max(sq_thread_idle, ctx->sq_thread_idle); 87 sqd->sq_thread_idle = sq_thread_idle; 88 } 89 90 void io_sq_thread_finish(struct io_ring_ctx *ctx) 91 { 92 struct io_sq_data *sqd = ctx->sq_data; 93 94 if (sqd) { 95 io_sq_thread_park(sqd); 96 list_del_init(&ctx->sqd_list); 97 io_sqd_update_thread_idle(sqd); 98 io_sq_thread_unpark(sqd); 99 100 io_put_sq_data(sqd); 101 ctx->sq_data = NULL; 102 } 103 } 104 105 static struct io_sq_data *io_attach_sq_data(struct io_uring_params *p) 106 { 107 struct io_ring_ctx *ctx_attach; 108 struct io_sq_data *sqd; 109 struct fd f; 110 111 f = fdget(p->wq_fd); 112 if (!f.file) 113 return ERR_PTR(-ENXIO); 114 if (!io_is_uring_fops(f.file)) { 115 fdput(f); 116 return ERR_PTR(-EINVAL); 117 } 118 119 ctx_attach = f.file->private_data; 120 sqd = ctx_attach->sq_data; 121 if (!sqd) { 122 fdput(f); 123 return ERR_PTR(-EINVAL); 124 } 125 if (sqd->task_tgid != current->tgid) { 126 fdput(f); 127 return ERR_PTR(-EPERM); 128 } 129 130 refcount_inc(&sqd->refs); 131 fdput(f); 132 return sqd; 133 } 134 135 static struct io_sq_data *io_get_sq_data(struct io_uring_params *p, 136 bool *attached) 137 { 138 struct io_sq_data *sqd; 139 140 *attached = false; 141 if (p->flags & IORING_SETUP_ATTACH_WQ) { 142 sqd = io_attach_sq_data(p); 143 if (!IS_ERR(sqd)) { 144 *attached = true; 145 return sqd; 146 } 147 /* fall through for EPERM case, setup new sqd/task */ 148 if (PTR_ERR(sqd) != -EPERM) 149 return sqd; 150 } 151 152 sqd = kzalloc(sizeof(*sqd), GFP_KERNEL); 153 if (!sqd) 154 return ERR_PTR(-ENOMEM); 155 156 atomic_set(&sqd->park_pending, 0); 157 refcount_set(&sqd->refs, 1); 158 INIT_LIST_HEAD(&sqd->ctx_list); 159 mutex_init(&sqd->lock); 160 init_waitqueue_head(&sqd->wait); 161 init_completion(&sqd->exited); 162 return sqd; 163 } 164 165 static inline bool io_sqd_events_pending(struct io_sq_data *sqd) 166 { 167 return READ_ONCE(sqd->state); 168 } 169 170 static int __io_sq_thread(struct io_ring_ctx *ctx, bool cap_entries) 171 { 172 unsigned int to_submit; 173 int ret = 0; 174 175 to_submit = io_sqring_entries(ctx); 176 /* if we're handling multiple rings, cap submit size for fairness */ 177 if (cap_entries && to_submit > IORING_SQPOLL_CAP_ENTRIES_VALUE) 178 to_submit = IORING_SQPOLL_CAP_ENTRIES_VALUE; 179 180 if (to_submit || !wq_list_empty(&ctx->iopoll_list)) { 181 const struct cred *creds = NULL; 182 183 if (ctx->sq_creds != current_cred()) 184 creds = override_creds(ctx->sq_creds); 185 186 mutex_lock(&ctx->uring_lock); 187 if (!wq_list_empty(&ctx->iopoll_list)) 188 io_do_iopoll(ctx, true); 189 190 /* 191 * Don't submit if refs are dying, good for io_uring_register(), 192 * but also it is relied upon by io_ring_exit_work() 193 */ 194 if (to_submit && likely(!percpu_ref_is_dying(&ctx->refs)) && 195 !(ctx->flags & IORING_SETUP_R_DISABLED)) 196 ret = io_submit_sqes(ctx, to_submit); 197 mutex_unlock(&ctx->uring_lock); 198 199 if (io_napi(ctx)) 200 ret += io_napi_sqpoll_busy_poll(ctx); 201 202 if (to_submit && wq_has_sleeper(&ctx->sqo_sq_wait)) 203 wake_up(&ctx->sqo_sq_wait); 204 if (creds) 205 revert_creds(creds); 206 } 207 208 return ret; 209 } 210 211 static bool io_sqd_handle_event(struct io_sq_data *sqd) 212 { 213 bool did_sig = false; 214 struct ksignal ksig; 215 216 if (test_bit(IO_SQ_THREAD_SHOULD_PARK, &sqd->state) || 217 signal_pending(current)) { 218 mutex_unlock(&sqd->lock); 219 if (signal_pending(current)) 220 did_sig = get_signal(&ksig); 221 cond_resched(); 222 mutex_lock(&sqd->lock); 223 sqd->sq_cpu = raw_smp_processor_id(); 224 } 225 return did_sig || test_bit(IO_SQ_THREAD_SHOULD_STOP, &sqd->state); 226 } 227 228 /* 229 * Run task_work, processing the retry_list first. The retry_list holds 230 * entries that we passed on in the previous run, if we had more task_work 231 * than we were asked to process. Newly queued task_work isn't run until the 232 * retry list has been fully processed. 233 */ 234 static unsigned int io_sq_tw(struct llist_node **retry_list, int max_entries) 235 { 236 struct io_uring_task *tctx = current->io_uring; 237 unsigned int count = 0; 238 239 if (*retry_list) { 240 *retry_list = io_handle_tw_list(*retry_list, &count, max_entries); 241 if (count >= max_entries) 242 goto out; 243 max_entries -= count; 244 } 245 *retry_list = tctx_task_work_run(tctx, max_entries, &count); 246 out: 247 if (task_work_pending(current)) 248 task_work_run(); 249 return count; 250 } 251 252 static bool io_sq_tw_pending(struct llist_node *retry_list) 253 { 254 struct io_uring_task *tctx = current->io_uring; 255 256 return retry_list || !llist_empty(&tctx->task_list); 257 } 258 259 static void io_sq_update_worktime(struct io_sq_data *sqd, struct rusage *start) 260 { 261 struct rusage end; 262 263 getrusage(current, RUSAGE_SELF, &end); 264 end.ru_stime.tv_sec -= start->ru_stime.tv_sec; 265 end.ru_stime.tv_usec -= start->ru_stime.tv_usec; 266 267 sqd->work_time += end.ru_stime.tv_usec + end.ru_stime.tv_sec * 1000000; 268 } 269 270 static int io_sq_thread(void *data) 271 { 272 struct llist_node *retry_list = NULL; 273 struct io_sq_data *sqd = data; 274 struct io_ring_ctx *ctx; 275 struct rusage start; 276 unsigned long timeout = 0; 277 char buf[TASK_COMM_LEN]; 278 DEFINE_WAIT(wait); 279 280 /* offload context creation failed, just exit */ 281 if (!current->io_uring) 282 goto err_out; 283 284 snprintf(buf, sizeof(buf), "iou-sqp-%d", sqd->task_pid); 285 set_task_comm(current, buf); 286 287 /* reset to our pid after we've set task_comm, for fdinfo */ 288 sqd->task_pid = current->pid; 289 290 if (sqd->sq_cpu != -1) { 291 set_cpus_allowed_ptr(current, cpumask_of(sqd->sq_cpu)); 292 } else { 293 set_cpus_allowed_ptr(current, cpu_online_mask); 294 sqd->sq_cpu = raw_smp_processor_id(); 295 } 296 297 /* 298 * Force audit context to get setup, in case we do prep side async 299 * operations that would trigger an audit call before any issue side 300 * audit has been done. 301 */ 302 audit_uring_entry(IORING_OP_NOP); 303 audit_uring_exit(true, 0); 304 305 mutex_lock(&sqd->lock); 306 while (1) { 307 bool cap_entries, sqt_spin = false; 308 309 if (io_sqd_events_pending(sqd) || signal_pending(current)) { 310 if (io_sqd_handle_event(sqd)) 311 break; 312 timeout = jiffies + sqd->sq_thread_idle; 313 } 314 315 cap_entries = !list_is_singular(&sqd->ctx_list); 316 getrusage(current, RUSAGE_SELF, &start); 317 list_for_each_entry(ctx, &sqd->ctx_list, sqd_list) { 318 int ret = __io_sq_thread(ctx, cap_entries); 319 320 if (!sqt_spin && (ret > 0 || !wq_list_empty(&ctx->iopoll_list))) 321 sqt_spin = true; 322 } 323 if (io_sq_tw(&retry_list, IORING_TW_CAP_ENTRIES_VALUE)) 324 sqt_spin = true; 325 326 if (sqt_spin || !time_after(jiffies, timeout)) { 327 if (sqt_spin) { 328 io_sq_update_worktime(sqd, &start); 329 timeout = jiffies + sqd->sq_thread_idle; 330 } 331 if (unlikely(need_resched())) { 332 mutex_unlock(&sqd->lock); 333 cond_resched(); 334 mutex_lock(&sqd->lock); 335 sqd->sq_cpu = raw_smp_processor_id(); 336 } 337 continue; 338 } 339 340 prepare_to_wait(&sqd->wait, &wait, TASK_INTERRUPTIBLE); 341 if (!io_sqd_events_pending(sqd) && !io_sq_tw_pending(retry_list)) { 342 bool needs_sched = true; 343 344 list_for_each_entry(ctx, &sqd->ctx_list, sqd_list) { 345 atomic_or(IORING_SQ_NEED_WAKEUP, 346 &ctx->rings->sq_flags); 347 if ((ctx->flags & IORING_SETUP_IOPOLL) && 348 !wq_list_empty(&ctx->iopoll_list)) { 349 needs_sched = false; 350 break; 351 } 352 353 /* 354 * Ensure the store of the wakeup flag is not 355 * reordered with the load of the SQ tail 356 */ 357 smp_mb__after_atomic(); 358 359 if (io_sqring_entries(ctx)) { 360 needs_sched = false; 361 break; 362 } 363 } 364 365 if (needs_sched) { 366 mutex_unlock(&sqd->lock); 367 schedule(); 368 mutex_lock(&sqd->lock); 369 sqd->sq_cpu = raw_smp_processor_id(); 370 } 371 list_for_each_entry(ctx, &sqd->ctx_list, sqd_list) 372 atomic_andnot(IORING_SQ_NEED_WAKEUP, 373 &ctx->rings->sq_flags); 374 } 375 376 finish_wait(&sqd->wait, &wait); 377 timeout = jiffies + sqd->sq_thread_idle; 378 } 379 380 if (retry_list) 381 io_sq_tw(&retry_list, UINT_MAX); 382 383 io_uring_cancel_generic(true, sqd); 384 sqd->thread = NULL; 385 list_for_each_entry(ctx, &sqd->ctx_list, sqd_list) 386 atomic_or(IORING_SQ_NEED_WAKEUP, &ctx->rings->sq_flags); 387 io_run_task_work(); 388 mutex_unlock(&sqd->lock); 389 err_out: 390 complete(&sqd->exited); 391 do_exit(0); 392 } 393 394 void io_sqpoll_wait_sq(struct io_ring_ctx *ctx) 395 { 396 DEFINE_WAIT(wait); 397 398 do { 399 if (!io_sqring_full(ctx)) 400 break; 401 prepare_to_wait(&ctx->sqo_sq_wait, &wait, TASK_INTERRUPTIBLE); 402 403 if (!io_sqring_full(ctx)) 404 break; 405 schedule(); 406 } while (!signal_pending(current)); 407 408 finish_wait(&ctx->sqo_sq_wait, &wait); 409 } 410 411 __cold int io_sq_offload_create(struct io_ring_ctx *ctx, 412 struct io_uring_params *p) 413 { 414 int ret; 415 416 /* Retain compatibility with failing for an invalid attach attempt */ 417 if ((ctx->flags & (IORING_SETUP_ATTACH_WQ | IORING_SETUP_SQPOLL)) == 418 IORING_SETUP_ATTACH_WQ) { 419 struct fd f; 420 421 f = fdget(p->wq_fd); 422 if (!f.file) 423 return -ENXIO; 424 if (!io_is_uring_fops(f.file)) { 425 fdput(f); 426 return -EINVAL; 427 } 428 fdput(f); 429 } 430 if (ctx->flags & IORING_SETUP_SQPOLL) { 431 struct task_struct *tsk; 432 struct io_sq_data *sqd; 433 bool attached; 434 435 ret = security_uring_sqpoll(); 436 if (ret) 437 return ret; 438 439 sqd = io_get_sq_data(p, &attached); 440 if (IS_ERR(sqd)) { 441 ret = PTR_ERR(sqd); 442 goto err; 443 } 444 445 ctx->sq_creds = get_current_cred(); 446 ctx->sq_data = sqd; 447 ctx->sq_thread_idle = msecs_to_jiffies(p->sq_thread_idle); 448 if (!ctx->sq_thread_idle) 449 ctx->sq_thread_idle = HZ; 450 451 io_sq_thread_park(sqd); 452 list_add(&ctx->sqd_list, &sqd->ctx_list); 453 io_sqd_update_thread_idle(sqd); 454 /* don't attach to a dying SQPOLL thread, would be racy */ 455 ret = (attached && !sqd->thread) ? -ENXIO : 0; 456 io_sq_thread_unpark(sqd); 457 458 if (ret < 0) 459 goto err; 460 if (attached) 461 return 0; 462 463 if (p->flags & IORING_SETUP_SQ_AFF) { 464 struct cpumask allowed_mask; 465 int cpu = p->sq_thread_cpu; 466 467 ret = -EINVAL; 468 cpuset_cpus_allowed(current, &allowed_mask); 469 if (!cpumask_test_cpu(cpu, &allowed_mask)) 470 goto err_sqpoll; 471 sqd->sq_cpu = cpu; 472 } else { 473 sqd->sq_cpu = -1; 474 } 475 476 sqd->task_pid = current->pid; 477 sqd->task_tgid = current->tgid; 478 tsk = create_io_thread(io_sq_thread, sqd, NUMA_NO_NODE); 479 if (IS_ERR(tsk)) { 480 ret = PTR_ERR(tsk); 481 goto err_sqpoll; 482 } 483 484 sqd->thread = tsk; 485 ret = io_uring_alloc_task_context(tsk, ctx); 486 wake_up_new_task(tsk); 487 if (ret) 488 goto err; 489 } else if (p->flags & IORING_SETUP_SQ_AFF) { 490 /* Can't have SQ_AFF without SQPOLL */ 491 ret = -EINVAL; 492 goto err; 493 } 494 495 return 0; 496 err_sqpoll: 497 complete(&ctx->sq_data->exited); 498 err: 499 io_sq_thread_finish(ctx); 500 return ret; 501 } 502 503 __cold int io_sqpoll_wq_cpu_affinity(struct io_ring_ctx *ctx, 504 cpumask_var_t mask) 505 { 506 struct io_sq_data *sqd = ctx->sq_data; 507 int ret = -EINVAL; 508 509 if (sqd) { 510 io_sq_thread_park(sqd); 511 /* Don't set affinity for a dying thread */ 512 if (sqd->thread) 513 ret = io_wq_cpu_affinity(sqd->thread->io_uring, mask); 514 io_sq_thread_unpark(sqd); 515 } 516 517 return ret; 518 } 519