xref: /openbmc/linux/io_uring/sqpoll.c (revision 5ffd8c73)
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/io_uring.h>
14 
15 #include <uapi/linux/io_uring.h>
16 
17 #include "io_uring.h"
18 #include "sqpoll.h"
19 
20 #define IORING_SQPOLL_CAP_ENTRIES_VALUE 8
21 
22 enum {
23 	IO_SQ_THREAD_SHOULD_STOP = 0,
24 	IO_SQ_THREAD_SHOULD_PARK,
25 };
26 
27 void io_sq_thread_unpark(struct io_sq_data *sqd)
28 	__releases(&sqd->lock)
29 {
30 	WARN_ON_ONCE(sqd->thread == current);
31 
32 	/*
33 	 * Do the dance but not conditional clear_bit() because it'd race with
34 	 * other threads incrementing park_pending and setting the bit.
35 	 */
36 	clear_bit(IO_SQ_THREAD_SHOULD_PARK, &sqd->state);
37 	if (atomic_dec_return(&sqd->park_pending))
38 		set_bit(IO_SQ_THREAD_SHOULD_PARK, &sqd->state);
39 	mutex_unlock(&sqd->lock);
40 }
41 
42 void io_sq_thread_park(struct io_sq_data *sqd)
43 	__acquires(&sqd->lock)
44 {
45 	WARN_ON_ONCE(sqd->thread == current);
46 
47 	atomic_inc(&sqd->park_pending);
48 	set_bit(IO_SQ_THREAD_SHOULD_PARK, &sqd->state);
49 	mutex_lock(&sqd->lock);
50 	if (sqd->thread)
51 		wake_up_process(sqd->thread);
52 }
53 
54 void io_sq_thread_stop(struct io_sq_data *sqd)
55 {
56 	WARN_ON_ONCE(sqd->thread == current);
57 	WARN_ON_ONCE(test_bit(IO_SQ_THREAD_SHOULD_STOP, &sqd->state));
58 
59 	set_bit(IO_SQ_THREAD_SHOULD_STOP, &sqd->state);
60 	mutex_lock(&sqd->lock);
61 	if (sqd->thread)
62 		wake_up_process(sqd->thread);
63 	mutex_unlock(&sqd->lock);
64 	wait_for_completion(&sqd->exited);
65 }
66 
67 void io_put_sq_data(struct io_sq_data *sqd)
68 {
69 	if (refcount_dec_and_test(&sqd->refs)) {
70 		WARN_ON_ONCE(atomic_read(&sqd->park_pending));
71 
72 		io_sq_thread_stop(sqd);
73 		kfree(sqd);
74 	}
75 }
76 
77 static __cold void io_sqd_update_thread_idle(struct io_sq_data *sqd)
78 {
79 	struct io_ring_ctx *ctx;
80 	unsigned sq_thread_idle = 0;
81 
82 	list_for_each_entry(ctx, &sqd->ctx_list, sqd_list)
83 		sq_thread_idle = max(sq_thread_idle, ctx->sq_thread_idle);
84 	sqd->sq_thread_idle = sq_thread_idle;
85 }
86 
87 void io_sq_thread_finish(struct io_ring_ctx *ctx)
88 {
89 	struct io_sq_data *sqd = ctx->sq_data;
90 
91 	if (sqd) {
92 		io_sq_thread_park(sqd);
93 		list_del_init(&ctx->sqd_list);
94 		io_sqd_update_thread_idle(sqd);
95 		io_sq_thread_unpark(sqd);
96 
97 		io_put_sq_data(sqd);
98 		ctx->sq_data = NULL;
99 	}
100 }
101 
102 static struct io_sq_data *io_attach_sq_data(struct io_uring_params *p)
103 {
104 	struct io_ring_ctx *ctx_attach;
105 	struct io_sq_data *sqd;
106 	struct fd f;
107 
108 	f = fdget(p->wq_fd);
109 	if (!f.file)
110 		return ERR_PTR(-ENXIO);
111 	if (!io_is_uring_fops(f.file)) {
112 		fdput(f);
113 		return ERR_PTR(-EINVAL);
114 	}
115 
116 	ctx_attach = f.file->private_data;
117 	sqd = ctx_attach->sq_data;
118 	if (!sqd) {
119 		fdput(f);
120 		return ERR_PTR(-EINVAL);
121 	}
122 	if (sqd->task_tgid != current->tgid) {
123 		fdput(f);
124 		return ERR_PTR(-EPERM);
125 	}
126 
127 	refcount_inc(&sqd->refs);
128 	fdput(f);
129 	return sqd;
130 }
131 
132 static struct io_sq_data *io_get_sq_data(struct io_uring_params *p,
133 					 bool *attached)
134 {
135 	struct io_sq_data *sqd;
136 
137 	*attached = false;
138 	if (p->flags & IORING_SETUP_ATTACH_WQ) {
139 		sqd = io_attach_sq_data(p);
140 		if (!IS_ERR(sqd)) {
141 			*attached = true;
142 			return sqd;
143 		}
144 		/* fall through for EPERM case, setup new sqd/task */
145 		if (PTR_ERR(sqd) != -EPERM)
146 			return sqd;
147 	}
148 
149 	sqd = kzalloc(sizeof(*sqd), GFP_KERNEL);
150 	if (!sqd)
151 		return ERR_PTR(-ENOMEM);
152 
153 	atomic_set(&sqd->park_pending, 0);
154 	refcount_set(&sqd->refs, 1);
155 	INIT_LIST_HEAD(&sqd->ctx_list);
156 	mutex_init(&sqd->lock);
157 	init_waitqueue_head(&sqd->wait);
158 	init_completion(&sqd->exited);
159 	return sqd;
160 }
161 
162 static inline bool io_sqd_events_pending(struct io_sq_data *sqd)
163 {
164 	return READ_ONCE(sqd->state);
165 }
166 
167 static int __io_sq_thread(struct io_ring_ctx *ctx, bool cap_entries)
168 {
169 	unsigned int to_submit;
170 	int ret = 0;
171 
172 	to_submit = io_sqring_entries(ctx);
173 	/* if we're handling multiple rings, cap submit size for fairness */
174 	if (cap_entries && to_submit > IORING_SQPOLL_CAP_ENTRIES_VALUE)
175 		to_submit = IORING_SQPOLL_CAP_ENTRIES_VALUE;
176 
177 	if (!wq_list_empty(&ctx->iopoll_list) || to_submit) {
178 		const struct cred *creds = NULL;
179 
180 		if (ctx->sq_creds != current_cred())
181 			creds = override_creds(ctx->sq_creds);
182 
183 		mutex_lock(&ctx->uring_lock);
184 		if (!wq_list_empty(&ctx->iopoll_list))
185 			io_do_iopoll(ctx, true);
186 
187 		/*
188 		 * Don't submit if refs are dying, good for io_uring_register(),
189 		 * but also it is relied upon by io_ring_exit_work()
190 		 */
191 		if (to_submit && likely(!percpu_ref_is_dying(&ctx->refs)) &&
192 		    !(ctx->flags & IORING_SETUP_R_DISABLED))
193 			ret = io_submit_sqes(ctx, to_submit);
194 		mutex_unlock(&ctx->uring_lock);
195 
196 		if (to_submit && wq_has_sleeper(&ctx->sqo_sq_wait))
197 			wake_up(&ctx->sqo_sq_wait);
198 		if (creds)
199 			revert_creds(creds);
200 	}
201 
202 	return ret;
203 }
204 
205 static bool io_sqd_handle_event(struct io_sq_data *sqd)
206 {
207 	bool did_sig = false;
208 	struct ksignal ksig;
209 
210 	if (test_bit(IO_SQ_THREAD_SHOULD_PARK, &sqd->state) ||
211 	    signal_pending(current)) {
212 		mutex_unlock(&sqd->lock);
213 		if (signal_pending(current))
214 			did_sig = get_signal(&ksig);
215 		cond_resched();
216 		mutex_lock(&sqd->lock);
217 		sqd->sq_cpu = raw_smp_processor_id();
218 	}
219 	return did_sig || test_bit(IO_SQ_THREAD_SHOULD_STOP, &sqd->state);
220 }
221 
222 static int io_sq_thread(void *data)
223 {
224 	struct io_sq_data *sqd = data;
225 	struct io_ring_ctx *ctx;
226 	unsigned long timeout = 0;
227 	char buf[TASK_COMM_LEN];
228 	DEFINE_WAIT(wait);
229 
230 	snprintf(buf, sizeof(buf), "iou-sqp-%d", sqd->task_pid);
231 	set_task_comm(current, buf);
232 
233 	/* reset to our pid after we've set task_comm, for fdinfo */
234 	sqd->task_pid = current->pid;
235 
236 	if (sqd->sq_cpu != -1) {
237 		set_cpus_allowed_ptr(current, cpumask_of(sqd->sq_cpu));
238 	} else {
239 		set_cpus_allowed_ptr(current, cpu_online_mask);
240 		sqd->sq_cpu = raw_smp_processor_id();
241 	}
242 
243 	mutex_lock(&sqd->lock);
244 	while (1) {
245 		bool cap_entries, sqt_spin = false;
246 
247 		if (io_sqd_events_pending(sqd) || signal_pending(current)) {
248 			if (io_sqd_handle_event(sqd))
249 				break;
250 			timeout = jiffies + sqd->sq_thread_idle;
251 		}
252 
253 		cap_entries = !list_is_singular(&sqd->ctx_list);
254 		list_for_each_entry(ctx, &sqd->ctx_list, sqd_list) {
255 			int ret = __io_sq_thread(ctx, cap_entries);
256 
257 			if (!sqt_spin && (ret > 0 || !wq_list_empty(&ctx->iopoll_list)))
258 				sqt_spin = true;
259 		}
260 		if (io_run_task_work())
261 			sqt_spin = true;
262 
263 		if (sqt_spin || !time_after(jiffies, timeout)) {
264 			if (sqt_spin)
265 				timeout = jiffies + sqd->sq_thread_idle;
266 			if (unlikely(need_resched())) {
267 				mutex_unlock(&sqd->lock);
268 				cond_resched();
269 				mutex_lock(&sqd->lock);
270 				sqd->sq_cpu = raw_smp_processor_id();
271 			}
272 			continue;
273 		}
274 
275 		prepare_to_wait(&sqd->wait, &wait, TASK_INTERRUPTIBLE);
276 		if (!io_sqd_events_pending(sqd) && !task_work_pending(current)) {
277 			bool needs_sched = true;
278 
279 			list_for_each_entry(ctx, &sqd->ctx_list, sqd_list) {
280 				atomic_or(IORING_SQ_NEED_WAKEUP,
281 						&ctx->rings->sq_flags);
282 				if ((ctx->flags & IORING_SETUP_IOPOLL) &&
283 				    !wq_list_empty(&ctx->iopoll_list)) {
284 					needs_sched = false;
285 					break;
286 				}
287 
288 				/*
289 				 * Ensure the store of the wakeup flag is not
290 				 * reordered with the load of the SQ tail
291 				 */
292 				smp_mb__after_atomic();
293 
294 				if (io_sqring_entries(ctx)) {
295 					needs_sched = false;
296 					break;
297 				}
298 			}
299 
300 			if (needs_sched) {
301 				mutex_unlock(&sqd->lock);
302 				schedule();
303 				mutex_lock(&sqd->lock);
304 				sqd->sq_cpu = raw_smp_processor_id();
305 			}
306 			list_for_each_entry(ctx, &sqd->ctx_list, sqd_list)
307 				atomic_andnot(IORING_SQ_NEED_WAKEUP,
308 						&ctx->rings->sq_flags);
309 		}
310 
311 		finish_wait(&sqd->wait, &wait);
312 		timeout = jiffies + sqd->sq_thread_idle;
313 	}
314 
315 	io_uring_cancel_generic(true, sqd);
316 	sqd->thread = NULL;
317 	list_for_each_entry(ctx, &sqd->ctx_list, sqd_list)
318 		atomic_or(IORING_SQ_NEED_WAKEUP, &ctx->rings->sq_flags);
319 	io_run_task_work();
320 	mutex_unlock(&sqd->lock);
321 
322 	complete(&sqd->exited);
323 	do_exit(0);
324 }
325 
326 void io_sqpoll_wait_sq(struct io_ring_ctx *ctx)
327 {
328 	DEFINE_WAIT(wait);
329 
330 	do {
331 		if (!io_sqring_full(ctx))
332 			break;
333 		prepare_to_wait(&ctx->sqo_sq_wait, &wait, TASK_INTERRUPTIBLE);
334 
335 		if (!io_sqring_full(ctx))
336 			break;
337 		schedule();
338 	} while (!signal_pending(current));
339 
340 	finish_wait(&ctx->sqo_sq_wait, &wait);
341 }
342 
343 __cold int io_sq_offload_create(struct io_ring_ctx *ctx,
344 				struct io_uring_params *p)
345 {
346 	int ret;
347 
348 	/* Retain compatibility with failing for an invalid attach attempt */
349 	if ((ctx->flags & (IORING_SETUP_ATTACH_WQ | IORING_SETUP_SQPOLL)) ==
350 				IORING_SETUP_ATTACH_WQ) {
351 		struct fd f;
352 
353 		f = fdget(p->wq_fd);
354 		if (!f.file)
355 			return -ENXIO;
356 		if (!io_is_uring_fops(f.file)) {
357 			fdput(f);
358 			return -EINVAL;
359 		}
360 		fdput(f);
361 	}
362 	if (ctx->flags & IORING_SETUP_SQPOLL) {
363 		struct task_struct *tsk;
364 		struct io_sq_data *sqd;
365 		bool attached;
366 
367 		ret = security_uring_sqpoll();
368 		if (ret)
369 			return ret;
370 
371 		sqd = io_get_sq_data(p, &attached);
372 		if (IS_ERR(sqd)) {
373 			ret = PTR_ERR(sqd);
374 			goto err;
375 		}
376 
377 		ctx->sq_creds = get_current_cred();
378 		ctx->sq_data = sqd;
379 		ctx->sq_thread_idle = msecs_to_jiffies(p->sq_thread_idle);
380 		if (!ctx->sq_thread_idle)
381 			ctx->sq_thread_idle = HZ;
382 
383 		io_sq_thread_park(sqd);
384 		list_add(&ctx->sqd_list, &sqd->ctx_list);
385 		io_sqd_update_thread_idle(sqd);
386 		/* don't attach to a dying SQPOLL thread, would be racy */
387 		ret = (attached && !sqd->thread) ? -ENXIO : 0;
388 		io_sq_thread_unpark(sqd);
389 
390 		if (ret < 0)
391 			goto err;
392 		if (attached)
393 			return 0;
394 
395 		if (p->flags & IORING_SETUP_SQ_AFF) {
396 			int cpu = p->sq_thread_cpu;
397 
398 			ret = -EINVAL;
399 			if (cpu >= nr_cpu_ids || !cpu_online(cpu))
400 				goto err_sqpoll;
401 			sqd->sq_cpu = cpu;
402 		} else {
403 			sqd->sq_cpu = -1;
404 		}
405 
406 		sqd->task_pid = current->pid;
407 		sqd->task_tgid = current->tgid;
408 		tsk = create_io_thread(io_sq_thread, sqd, NUMA_NO_NODE);
409 		if (IS_ERR(tsk)) {
410 			ret = PTR_ERR(tsk);
411 			goto err_sqpoll;
412 		}
413 
414 		sqd->thread = tsk;
415 		ret = io_uring_alloc_task_context(tsk, ctx);
416 		wake_up_new_task(tsk);
417 		if (ret)
418 			goto err;
419 	} else if (p->flags & IORING_SETUP_SQ_AFF) {
420 		/* Can't have SQ_AFF without SQPOLL */
421 		ret = -EINVAL;
422 		goto err;
423 	}
424 
425 	return 0;
426 err_sqpoll:
427 	complete(&ctx->sq_data->exited);
428 err:
429 	io_sq_thread_finish(ctx);
430 	return ret;
431 }
432 
433 __cold int io_sqpoll_wq_cpu_affinity(struct io_ring_ctx *ctx,
434 				     cpumask_var_t mask)
435 {
436 	struct io_sq_data *sqd = ctx->sq_data;
437 	int ret = -EINVAL;
438 
439 	if (sqd) {
440 		io_sq_thread_park(sqd);
441 		/* Don't set affinity for a dying thread */
442 		if (sqd->thread)
443 			ret = io_wq_cpu_affinity(sqd->thread->io_uring, mask);
444 		io_sq_thread_unpark(sqd);
445 	}
446 
447 	return ret;
448 }
449