blob: cfb48bd088e12a0bbd8847e2b78e7ba14cb26fd8 [file] [log] [blame]
Jens Axboe2b188cc2019-01-07 10:46:33 -07001// SPDX-License-Identifier: GPL-2.0
2/*
3 * Shared application/kernel submission and completion ring pairs, for
4 * supporting fast/efficient IO.
5 *
6 * A note on the read/write ordering memory barriers that are matched between
Stefan Bühler1e84b972019-04-24 23:54:16 +02007 * the application and kernel side.
8 *
9 * After the application reads the CQ ring tail, it must use an
10 * appropriate smp_rmb() to pair with the smp_wmb() the kernel uses
11 * before writing the tail (using smp_load_acquire to read the tail will
12 * do). It also needs a smp_mb() before updating CQ head (ordering the
13 * entry load(s) with the head store), pairing with an implicit barrier
14 * through a control-dependency in io_get_cqring (smp_store_release to
15 * store head will do). Failure to do so could lead to reading invalid
16 * CQ entries.
17 *
18 * Likewise, the application must use an appropriate smp_wmb() before
19 * writing the SQ tail (ordering SQ entry stores with the tail store),
20 * which pairs with smp_load_acquire in io_get_sqring (smp_store_release
21 * to store the tail will do). And it needs a barrier ordering the SQ
22 * head load before writing new SQ entries (smp_load_acquire to read
23 * head will do).
24 *
25 * When using the SQ poll thread (IORING_SETUP_SQPOLL), the application
26 * needs to check the SQ flags for IORING_SQ_NEED_WAKEUP *after*
27 * updating the SQ tail; a full memory barrier smp_mb() is needed
28 * between.
Jens Axboe2b188cc2019-01-07 10:46:33 -070029 *
30 * Also see the examples in the liburing library:
31 *
32 * git://git.kernel.dk/liburing
33 *
34 * io_uring also uses READ/WRITE_ONCE() for _any_ store or load that happens
35 * from data shared between the kernel and application. This is done both
36 * for ordering purposes, but also to ensure that once a value is loaded from
37 * data that the application could potentially modify, it remains stable.
38 *
39 * Copyright (C) 2018-2019 Jens Axboe
Christoph Hellwigc992fe22019-01-11 09:43:02 -070040 * Copyright (c) 2018-2019 Christoph Hellwig
Jens Axboe2b188cc2019-01-07 10:46:33 -070041 */
42#include <linux/kernel.h>
43#include <linux/init.h>
44#include <linux/errno.h>
45#include <linux/syscalls.h>
46#include <linux/compat.h>
47#include <linux/refcount.h>
48#include <linux/uio.h>
49
50#include <linux/sched/signal.h>
51#include <linux/fs.h>
52#include <linux/file.h>
53#include <linux/fdtable.h>
54#include <linux/mm.h>
55#include <linux/mman.h>
56#include <linux/mmu_context.h>
57#include <linux/percpu.h>
58#include <linux/slab.h>
59#include <linux/workqueue.h>
Jens Axboe6c271ce2019-01-10 11:22:30 -070060#include <linux/kthread.h>
Jens Axboe2b188cc2019-01-07 10:46:33 -070061#include <linux/blkdev.h>
Jens Axboeedafcce2019-01-09 09:16:05 -070062#include <linux/bvec.h>
Jens Axboe2b188cc2019-01-07 10:46:33 -070063#include <linux/net.h>
64#include <net/sock.h>
65#include <net/af_unix.h>
Jens Axboe6b063142019-01-10 22:13:58 -070066#include <net/scm.h>
Jens Axboe2b188cc2019-01-07 10:46:33 -070067#include <linux/anon_inodes.h>
68#include <linux/sched/mm.h>
69#include <linux/uaccess.h>
70#include <linux/nospec.h>
Jens Axboeedafcce2019-01-09 09:16:05 -070071#include <linux/sizes.h>
72#include <linux/hugetlb.h>
Jens Axboe2b188cc2019-01-07 10:46:33 -070073
74#include <uapi/linux/io_uring.h>
75
76#include "internal.h"
77
78#define IORING_MAX_ENTRIES 4096
Jens Axboe6b063142019-01-10 22:13:58 -070079#define IORING_MAX_FIXED_FILES 1024
Jens Axboe2b188cc2019-01-07 10:46:33 -070080
81struct io_uring {
82 u32 head ____cacheline_aligned_in_smp;
83 u32 tail ____cacheline_aligned_in_smp;
84};
85
Stefan Bühler1e84b972019-04-24 23:54:16 +020086/*
87 * This data is shared with the application through the mmap at offset
88 * IORING_OFF_SQ_RING.
89 *
90 * The offsets to the member fields are published through struct
91 * io_sqring_offsets when calling io_uring_setup.
92 */
Jens Axboe2b188cc2019-01-07 10:46:33 -070093struct io_sq_ring {
Stefan Bühler1e84b972019-04-24 23:54:16 +020094 /*
95 * Head and tail offsets into the ring; the offsets need to be
96 * masked to get valid indices.
97 *
98 * The kernel controls head and the application controls tail.
99 */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700100 struct io_uring r;
Stefan Bühler1e84b972019-04-24 23:54:16 +0200101 /*
102 * Bitmask to apply to head and tail offsets (constant, equals
103 * ring_entries - 1)
104 */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700105 u32 ring_mask;
Stefan Bühler1e84b972019-04-24 23:54:16 +0200106 /* Ring size (constant, power of 2) */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700107 u32 ring_entries;
Stefan Bühler1e84b972019-04-24 23:54:16 +0200108 /*
109 * Number of invalid entries dropped by the kernel due to
110 * invalid index stored in array
111 *
112 * Written by the kernel, shouldn't be modified by the
113 * application (i.e. get number of "new events" by comparing to
114 * cached value).
115 *
116 * After a new SQ head value was read by the application this
117 * counter includes all submissions that were dropped reaching
118 * the new SQ head (and possibly more).
119 */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700120 u32 dropped;
Stefan Bühler1e84b972019-04-24 23:54:16 +0200121 /*
122 * Runtime flags
123 *
124 * Written by the kernel, shouldn't be modified by the
125 * application.
126 *
127 * The application needs a full memory barrier before checking
128 * for IORING_SQ_NEED_WAKEUP after updating the sq tail.
129 */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700130 u32 flags;
Stefan Bühler1e84b972019-04-24 23:54:16 +0200131 /*
132 * Ring buffer of indices into array of io_uring_sqe, which is
133 * mmapped by the application using the IORING_OFF_SQES offset.
134 *
135 * This indirection could e.g. be used to assign fixed
136 * io_uring_sqe entries to operations and only submit them to
137 * the queue when needed.
138 *
139 * The kernel modifies neither the indices array nor the entries
140 * array.
141 */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700142 u32 array[];
143};
144
Stefan Bühler1e84b972019-04-24 23:54:16 +0200145/*
146 * This data is shared with the application through the mmap at offset
147 * IORING_OFF_CQ_RING.
148 *
149 * The offsets to the member fields are published through struct
150 * io_cqring_offsets when calling io_uring_setup.
151 */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700152struct io_cq_ring {
Stefan Bühler1e84b972019-04-24 23:54:16 +0200153 /*
154 * Head and tail offsets into the ring; the offsets need to be
155 * masked to get valid indices.
156 *
157 * The application controls head and the kernel tail.
158 */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700159 struct io_uring r;
Stefan Bühler1e84b972019-04-24 23:54:16 +0200160 /*
161 * Bitmask to apply to head and tail offsets (constant, equals
162 * ring_entries - 1)
163 */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700164 u32 ring_mask;
Stefan Bühler1e84b972019-04-24 23:54:16 +0200165 /* Ring size (constant, power of 2) */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700166 u32 ring_entries;
Stefan Bühler1e84b972019-04-24 23:54:16 +0200167 /*
168 * Number of completion events lost because the queue was full;
169 * this should be avoided by the application by making sure
170 * there are not more requests pending thatn there is space in
171 * the completion queue.
172 *
173 * Written by the kernel, shouldn't be modified by the
174 * application (i.e. get number of "new events" by comparing to
175 * cached value).
176 *
177 * As completion events come in out of order this counter is not
178 * ordered with any other data.
179 */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700180 u32 overflow;
Stefan Bühler1e84b972019-04-24 23:54:16 +0200181 /*
182 * Ring buffer of completion events.
183 *
184 * The kernel writes completion events fresh every time they are
185 * produced, so the application is allowed to modify pending
186 * entries.
187 */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700188 struct io_uring_cqe cqes[];
189};
190
Jens Axboeedafcce2019-01-09 09:16:05 -0700191struct io_mapped_ubuf {
192 u64 ubuf;
193 size_t len;
194 struct bio_vec *bvec;
195 unsigned int nr_bvecs;
196};
197
Jens Axboe31b51512019-01-18 22:56:34 -0700198struct async_list {
199 spinlock_t lock;
200 atomic_t cnt;
201 struct list_head list;
202
203 struct file *file;
204 off_t io_end;
Zhengyuan Liu9310a7ba2019-07-22 10:23:27 +0800205 size_t io_len;
Jens Axboe31b51512019-01-18 22:56:34 -0700206};
207
Jens Axboe2b188cc2019-01-07 10:46:33 -0700208struct io_ring_ctx {
209 struct {
210 struct percpu_ref refs;
211 } ____cacheline_aligned_in_smp;
212
213 struct {
214 unsigned int flags;
215 bool compat;
216 bool account_mem;
217
218 /* SQ ring */
219 struct io_sq_ring *sq_ring;
220 unsigned cached_sq_head;
221 unsigned sq_entries;
222 unsigned sq_mask;
Jens Axboe6c271ce2019-01-10 11:22:30 -0700223 unsigned sq_thread_idle;
Jens Axboe2b188cc2019-01-07 10:46:33 -0700224 struct io_uring_sqe *sq_sqes;
Jens Axboede0617e2019-04-06 21:51:27 -0600225
226 struct list_head defer_list;
Jens Axboe2b188cc2019-01-07 10:46:33 -0700227 } ____cacheline_aligned_in_smp;
228
229 /* IO offload */
230 struct workqueue_struct *sqo_wq;
Jens Axboe6c271ce2019-01-10 11:22:30 -0700231 struct task_struct *sqo_thread; /* if using sq thread polling */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700232 struct mm_struct *sqo_mm;
Jens Axboe6c271ce2019-01-10 11:22:30 -0700233 wait_queue_head_t sqo_wait;
Jackie Liua4c0b3d2019-07-08 13:41:12 +0800234 struct completion sqo_thread_started;
Jens Axboe2b188cc2019-01-07 10:46:33 -0700235
236 struct {
237 /* CQ ring */
238 struct io_cq_ring *cq_ring;
239 unsigned cached_cq_tail;
240 unsigned cq_entries;
241 unsigned cq_mask;
242 struct wait_queue_head cq_wait;
243 struct fasync_struct *cq_fasync;
Jens Axboe9b402842019-04-11 11:45:41 -0600244 struct eventfd_ctx *cq_ev_fd;
Jens Axboe2b188cc2019-01-07 10:46:33 -0700245 } ____cacheline_aligned_in_smp;
246
Jens Axboe6b063142019-01-10 22:13:58 -0700247 /*
248 * If used, fixed file set. Writers must ensure that ->refs is dead,
249 * readers must ensure that ->refs is alive as long as the file* is
250 * used. Only updated through io_uring_register(2).
251 */
252 struct file **user_files;
253 unsigned nr_user_files;
254
Jens Axboeedafcce2019-01-09 09:16:05 -0700255 /* if used, fixed mapped user buffers */
256 unsigned nr_user_bufs;
257 struct io_mapped_ubuf *user_bufs;
258
Jens Axboe2b188cc2019-01-07 10:46:33 -0700259 struct user_struct *user;
260
261 struct completion ctx_done;
262
263 struct {
264 struct mutex uring_lock;
265 wait_queue_head_t wait;
266 } ____cacheline_aligned_in_smp;
267
268 struct {
269 spinlock_t completion_lock;
Jens Axboedef596e2019-01-09 08:59:42 -0700270 bool poll_multi_file;
271 /*
272 * ->poll_list is protected by the ctx->uring_lock for
273 * io_uring instances that don't use IORING_SETUP_SQPOLL.
274 * For SQPOLL, only the single threaded io_sq_thread() will
275 * manipulate the list, hence no extra locking is needed there.
276 */
277 struct list_head poll_list;
Jens Axboe221c5eb2019-01-17 09:41:58 -0700278 struct list_head cancel_list;
Jens Axboe2b188cc2019-01-07 10:46:33 -0700279 } ____cacheline_aligned_in_smp;
280
Jens Axboe31b51512019-01-18 22:56:34 -0700281 struct async_list pending_async[2];
282
Jens Axboe2b188cc2019-01-07 10:46:33 -0700283#if defined(CONFIG_UNIX)
284 struct socket *ring_sock;
285#endif
286};
287
288struct sqe_submit {
289 const struct io_uring_sqe *sqe;
290 unsigned short index;
291 bool has_user;
Jens Axboedef596e2019-01-09 08:59:42 -0700292 bool needs_lock;
Jens Axboe6c271ce2019-01-10 11:22:30 -0700293 bool needs_fixed_file;
Jens Axboe2b188cc2019-01-07 10:46:33 -0700294};
295
Jens Axboe09bb8392019-03-13 12:39:28 -0600296/*
297 * First field must be the file pointer in all the
298 * iocb unions! See also 'struct kiocb' in <linux/fs.h>
299 */
Jens Axboe221c5eb2019-01-17 09:41:58 -0700300struct io_poll_iocb {
301 struct file *file;
302 struct wait_queue_head *head;
303 __poll_t events;
Jens Axboe8c838782019-03-12 15:48:16 -0600304 bool done;
Jens Axboe221c5eb2019-01-17 09:41:58 -0700305 bool canceled;
306 struct wait_queue_entry wait;
307};
308
Jens Axboe09bb8392019-03-13 12:39:28 -0600309/*
310 * NOTE! Each of the iocb union members has the file pointer
311 * as the first entry in their struct definition. So you can
312 * access the file pointer through any of the sub-structs,
313 * or directly as just 'ki_filp' in this struct.
314 */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700315struct io_kiocb {
Jens Axboe221c5eb2019-01-17 09:41:58 -0700316 union {
Jens Axboe09bb8392019-03-13 12:39:28 -0600317 struct file *file;
Jens Axboe221c5eb2019-01-17 09:41:58 -0700318 struct kiocb rw;
319 struct io_poll_iocb poll;
320 };
Jens Axboe2b188cc2019-01-07 10:46:33 -0700321
322 struct sqe_submit submit;
323
324 struct io_ring_ctx *ctx;
325 struct list_head list;
Jens Axboe9e645e112019-05-10 16:07:28 -0600326 struct list_head link_list;
Jens Axboe2b188cc2019-01-07 10:46:33 -0700327 unsigned int flags;
Jens Axboec16361c2019-01-17 08:39:48 -0700328 refcount_t refs;
Stefan Bühler8449eed2019-04-27 20:34:19 +0200329#define REQ_F_NOWAIT 1 /* must not punt to workers */
Jens Axboedef596e2019-01-09 08:59:42 -0700330#define REQ_F_IOPOLL_COMPLETED 2 /* polled IO has completed */
Jens Axboe6b063142019-01-10 22:13:58 -0700331#define REQ_F_FIXED_FILE 4 /* ctx owns file */
Jens Axboe31b51512019-01-18 22:56:34 -0700332#define REQ_F_SEQ_PREV 8 /* sequential with previous */
Stefan Bühlere2033e32019-05-11 19:08:01 +0200333#define REQ_F_IO_DRAIN 16 /* drain existing IO first */
334#define REQ_F_IO_DRAINED 32 /* drain done */
Jens Axboe9e645e112019-05-10 16:07:28 -0600335#define REQ_F_LINK 64 /* linked sqes */
Zhengyuan Liuf7b76ac2019-07-16 23:26:14 +0800336#define REQ_F_LINK_DONE 128 /* linked sqes done */
337#define REQ_F_FAIL_LINK 256 /* fail rest of links */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700338 u64 user_data;
Jens Axboe9e645e112019-05-10 16:07:28 -0600339 u32 result;
Jens Axboede0617e2019-04-06 21:51:27 -0600340 u32 sequence;
Jens Axboe2b188cc2019-01-07 10:46:33 -0700341
342 struct work_struct work;
343};
344
345#define IO_PLUG_THRESHOLD 2
Jens Axboedef596e2019-01-09 08:59:42 -0700346#define IO_IOPOLL_BATCH 8
Jens Axboe2b188cc2019-01-07 10:46:33 -0700347
Jens Axboe9a56a232019-01-09 09:06:50 -0700348struct io_submit_state {
349 struct blk_plug plug;
350
351 /*
Jens Axboe2579f912019-01-09 09:10:43 -0700352 * io_kiocb alloc cache
353 */
354 void *reqs[IO_IOPOLL_BATCH];
355 unsigned int free_reqs;
356 unsigned int cur_req;
357
358 /*
Jens Axboe9a56a232019-01-09 09:06:50 -0700359 * File reference cache
360 */
361 struct file *file;
362 unsigned int fd;
363 unsigned int has_refs;
364 unsigned int used_refs;
365 unsigned int ios_left;
366};
367
Jens Axboede0617e2019-04-06 21:51:27 -0600368static void io_sq_wq_submit_work(struct work_struct *work);
369
Jens Axboe2b188cc2019-01-07 10:46:33 -0700370static struct kmem_cache *req_cachep;
371
372static const struct file_operations io_uring_fops;
373
374struct sock *io_uring_get_socket(struct file *file)
375{
376#if defined(CONFIG_UNIX)
377 if (file->f_op == &io_uring_fops) {
378 struct io_ring_ctx *ctx = file->private_data;
379
380 return ctx->ring_sock->sk;
381 }
382#endif
383 return NULL;
384}
385EXPORT_SYMBOL(io_uring_get_socket);
386
387static void io_ring_ctx_ref_free(struct percpu_ref *ref)
388{
389 struct io_ring_ctx *ctx = container_of(ref, struct io_ring_ctx, refs);
390
391 complete(&ctx->ctx_done);
392}
393
394static struct io_ring_ctx *io_ring_ctx_alloc(struct io_uring_params *p)
395{
396 struct io_ring_ctx *ctx;
Jens Axboe31b51512019-01-18 22:56:34 -0700397 int i;
Jens Axboe2b188cc2019-01-07 10:46:33 -0700398
399 ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
400 if (!ctx)
401 return NULL;
402
Roman Gushchin21482892019-05-07 10:01:48 -0700403 if (percpu_ref_init(&ctx->refs, io_ring_ctx_ref_free,
404 PERCPU_REF_ALLOW_REINIT, GFP_KERNEL)) {
Jens Axboe2b188cc2019-01-07 10:46:33 -0700405 kfree(ctx);
406 return NULL;
407 }
408
409 ctx->flags = p->flags;
410 init_waitqueue_head(&ctx->cq_wait);
411 init_completion(&ctx->ctx_done);
Jackie Liua4c0b3d2019-07-08 13:41:12 +0800412 init_completion(&ctx->sqo_thread_started);
Jens Axboe2b188cc2019-01-07 10:46:33 -0700413 mutex_init(&ctx->uring_lock);
414 init_waitqueue_head(&ctx->wait);
Jens Axboe31b51512019-01-18 22:56:34 -0700415 for (i = 0; i < ARRAY_SIZE(ctx->pending_async); i++) {
416 spin_lock_init(&ctx->pending_async[i].lock);
417 INIT_LIST_HEAD(&ctx->pending_async[i].list);
418 atomic_set(&ctx->pending_async[i].cnt, 0);
419 }
Jens Axboe2b188cc2019-01-07 10:46:33 -0700420 spin_lock_init(&ctx->completion_lock);
Jens Axboedef596e2019-01-09 08:59:42 -0700421 INIT_LIST_HEAD(&ctx->poll_list);
Jens Axboe221c5eb2019-01-17 09:41:58 -0700422 INIT_LIST_HEAD(&ctx->cancel_list);
Jens Axboede0617e2019-04-06 21:51:27 -0600423 INIT_LIST_HEAD(&ctx->defer_list);
Jens Axboe2b188cc2019-01-07 10:46:33 -0700424 return ctx;
425}
426
Jens Axboede0617e2019-04-06 21:51:27 -0600427static inline bool io_sequence_defer(struct io_ring_ctx *ctx,
428 struct io_kiocb *req)
429{
430 if ((req->flags & (REQ_F_IO_DRAIN|REQ_F_IO_DRAINED)) != REQ_F_IO_DRAIN)
431 return false;
432
Zhengyuan Liudbd0f6d2019-07-13 11:58:26 +0800433 return req->sequence != ctx->cached_cq_tail + ctx->sq_ring->dropped;
Jens Axboede0617e2019-04-06 21:51:27 -0600434}
435
436static struct io_kiocb *io_get_deferred_req(struct io_ring_ctx *ctx)
437{
438 struct io_kiocb *req;
439
440 if (list_empty(&ctx->defer_list))
441 return NULL;
442
443 req = list_first_entry(&ctx->defer_list, struct io_kiocb, list);
444 if (!io_sequence_defer(ctx, req)) {
445 list_del_init(&req->list);
446 return req;
447 }
448
449 return NULL;
450}
451
452static void __io_commit_cqring(struct io_ring_ctx *ctx)
Jens Axboe2b188cc2019-01-07 10:46:33 -0700453{
454 struct io_cq_ring *ring = ctx->cq_ring;
455
456 if (ctx->cached_cq_tail != READ_ONCE(ring->r.tail)) {
457 /* order cqe stores with ring update */
458 smp_store_release(&ring->r.tail, ctx->cached_cq_tail);
459
Jens Axboe2b188cc2019-01-07 10:46:33 -0700460 if (wq_has_sleeper(&ctx->cq_wait)) {
461 wake_up_interruptible(&ctx->cq_wait);
462 kill_fasync(&ctx->cq_fasync, SIGIO, POLL_IN);
463 }
464 }
465}
466
Jens Axboede0617e2019-04-06 21:51:27 -0600467static void io_commit_cqring(struct io_ring_ctx *ctx)
468{
469 struct io_kiocb *req;
470
471 __io_commit_cqring(ctx);
472
473 while ((req = io_get_deferred_req(ctx)) != NULL) {
474 req->flags |= REQ_F_IO_DRAINED;
475 queue_work(ctx->sqo_wq, &req->work);
476 }
477}
478
Jens Axboe2b188cc2019-01-07 10:46:33 -0700479static struct io_uring_cqe *io_get_cqring(struct io_ring_ctx *ctx)
480{
481 struct io_cq_ring *ring = ctx->cq_ring;
482 unsigned tail;
483
484 tail = ctx->cached_cq_tail;
Stefan Bühler115e12e2019-04-24 23:54:18 +0200485 /*
486 * writes to the cq entry need to come after reading head; the
487 * control dependency is enough as we're using WRITE_ONCE to
488 * fill the cq entry
489 */
Jens Axboe74f464e2019-04-17 08:57:48 -0600490 if (tail - READ_ONCE(ring->r.head) == ring->ring_entries)
Jens Axboe2b188cc2019-01-07 10:46:33 -0700491 return NULL;
492
493 ctx->cached_cq_tail++;
494 return &ring->cqes[tail & ctx->cq_mask];
495}
496
497static void io_cqring_fill_event(struct io_ring_ctx *ctx, u64 ki_user_data,
Jens Axboec71ffb62019-05-13 20:58:29 -0600498 long res)
Jens Axboe2b188cc2019-01-07 10:46:33 -0700499{
500 struct io_uring_cqe *cqe;
501
502 /*
503 * If we can't get a cq entry, userspace overflowed the
504 * submission (by quite a lot). Increment the overflow count in
505 * the ring.
506 */
507 cqe = io_get_cqring(ctx);
508 if (cqe) {
509 WRITE_ONCE(cqe->user_data, ki_user_data);
510 WRITE_ONCE(cqe->res, res);
Jens Axboec71ffb62019-05-13 20:58:29 -0600511 WRITE_ONCE(cqe->flags, 0);
Jens Axboe2b188cc2019-01-07 10:46:33 -0700512 } else {
513 unsigned overflow = READ_ONCE(ctx->cq_ring->overflow);
514
515 WRITE_ONCE(ctx->cq_ring->overflow, overflow + 1);
516 }
517}
518
Jens Axboe8c838782019-03-12 15:48:16 -0600519static void io_cqring_ev_posted(struct io_ring_ctx *ctx)
520{
521 if (waitqueue_active(&ctx->wait))
522 wake_up(&ctx->wait);
523 if (waitqueue_active(&ctx->sqo_wait))
524 wake_up(&ctx->sqo_wait);
Jens Axboe9b402842019-04-11 11:45:41 -0600525 if (ctx->cq_ev_fd)
526 eventfd_signal(ctx->cq_ev_fd, 1);
Jens Axboe8c838782019-03-12 15:48:16 -0600527}
528
529static void io_cqring_add_event(struct io_ring_ctx *ctx, u64 user_data,
Jens Axboec71ffb62019-05-13 20:58:29 -0600530 long res)
Jens Axboe2b188cc2019-01-07 10:46:33 -0700531{
532 unsigned long flags;
533
534 spin_lock_irqsave(&ctx->completion_lock, flags);
Jens Axboec71ffb62019-05-13 20:58:29 -0600535 io_cqring_fill_event(ctx, user_data, res);
Jens Axboe2b188cc2019-01-07 10:46:33 -0700536 io_commit_cqring(ctx);
537 spin_unlock_irqrestore(&ctx->completion_lock, flags);
538
Jens Axboe8c838782019-03-12 15:48:16 -0600539 io_cqring_ev_posted(ctx);
Jens Axboe2b188cc2019-01-07 10:46:33 -0700540}
541
542static void io_ring_drop_ctx_refs(struct io_ring_ctx *ctx, unsigned refs)
543{
544 percpu_ref_put_many(&ctx->refs, refs);
545
546 if (waitqueue_active(&ctx->wait))
547 wake_up(&ctx->wait);
548}
549
Jens Axboe2579f912019-01-09 09:10:43 -0700550static struct io_kiocb *io_get_req(struct io_ring_ctx *ctx,
551 struct io_submit_state *state)
Jens Axboe2b188cc2019-01-07 10:46:33 -0700552{
Jens Axboefd6fab22019-03-14 16:30:06 -0600553 gfp_t gfp = GFP_KERNEL | __GFP_NOWARN;
Jens Axboe2b188cc2019-01-07 10:46:33 -0700554 struct io_kiocb *req;
555
556 if (!percpu_ref_tryget(&ctx->refs))
557 return NULL;
558
Jens Axboe2579f912019-01-09 09:10:43 -0700559 if (!state) {
Jens Axboefd6fab22019-03-14 16:30:06 -0600560 req = kmem_cache_alloc(req_cachep, gfp);
Jens Axboe2579f912019-01-09 09:10:43 -0700561 if (unlikely(!req))
562 goto out;
563 } else if (!state->free_reqs) {
564 size_t sz;
565 int ret;
566
567 sz = min_t(size_t, state->ios_left, ARRAY_SIZE(state->reqs));
Jens Axboefd6fab22019-03-14 16:30:06 -0600568 ret = kmem_cache_alloc_bulk(req_cachep, gfp, sz, state->reqs);
569
570 /*
571 * Bulk alloc is all-or-nothing. If we fail to get a batch,
572 * retry single alloc to be on the safe side.
573 */
574 if (unlikely(ret <= 0)) {
575 state->reqs[0] = kmem_cache_alloc(req_cachep, gfp);
576 if (!state->reqs[0])
577 goto out;
578 ret = 1;
579 }
Jens Axboe2579f912019-01-09 09:10:43 -0700580 state->free_reqs = ret - 1;
581 state->cur_req = 1;
582 req = state->reqs[0];
583 } else {
584 req = state->reqs[state->cur_req];
585 state->free_reqs--;
586 state->cur_req++;
Jens Axboe2b188cc2019-01-07 10:46:33 -0700587 }
588
Jens Axboe60c112b2019-06-21 10:20:18 -0600589 req->file = NULL;
Jens Axboe2579f912019-01-09 09:10:43 -0700590 req->ctx = ctx;
591 req->flags = 0;
Jens Axboee65ef562019-03-12 10:16:44 -0600592 /* one is dropped after submission, the other at completion */
593 refcount_set(&req->refs, 2);
Jens Axboe9e645e112019-05-10 16:07:28 -0600594 req->result = 0;
Jens Axboe2579f912019-01-09 09:10:43 -0700595 return req;
596out:
Jens Axboe2b188cc2019-01-07 10:46:33 -0700597 io_ring_drop_ctx_refs(ctx, 1);
598 return NULL;
599}
600
Jens Axboedef596e2019-01-09 08:59:42 -0700601static void io_free_req_many(struct io_ring_ctx *ctx, void **reqs, int *nr)
602{
603 if (*nr) {
604 kmem_cache_free_bulk(req_cachep, *nr, reqs);
605 io_ring_drop_ctx_refs(ctx, *nr);
606 *nr = 0;
607 }
608}
609
Jens Axboe9e645e112019-05-10 16:07:28 -0600610static void __io_free_req(struct io_kiocb *req)
Jens Axboe2b188cc2019-01-07 10:46:33 -0700611{
Jens Axboe09bb8392019-03-13 12:39:28 -0600612 if (req->file && !(req->flags & REQ_F_FIXED_FILE))
613 fput(req->file);
Jens Axboee65ef562019-03-12 10:16:44 -0600614 io_ring_drop_ctx_refs(req->ctx, 1);
615 kmem_cache_free(req_cachep, req);
616}
617
Jens Axboe9e645e112019-05-10 16:07:28 -0600618static void io_req_link_next(struct io_kiocb *req)
619{
620 struct io_kiocb *nxt;
621
622 /*
623 * The list should never be empty when we are called here. But could
624 * potentially happen if the chain is messed up, check to be on the
625 * safe side.
626 */
627 nxt = list_first_entry_or_null(&req->link_list, struct io_kiocb, list);
628 if (nxt) {
629 list_del(&nxt->list);
630 if (!list_empty(&req->link_list)) {
631 INIT_LIST_HEAD(&nxt->link_list);
632 list_splice(&req->link_list, &nxt->link_list);
633 nxt->flags |= REQ_F_LINK;
634 }
635
Zhengyuan Liuf7b76ac2019-07-16 23:26:14 +0800636 nxt->flags |= REQ_F_LINK_DONE;
Jens Axboe9e645e112019-05-10 16:07:28 -0600637 INIT_WORK(&nxt->work, io_sq_wq_submit_work);
638 queue_work(req->ctx->sqo_wq, &nxt->work);
639 }
640}
641
642/*
643 * Called if REQ_F_LINK is set, and we fail the head request
644 */
645static void io_fail_links(struct io_kiocb *req)
646{
647 struct io_kiocb *link;
648
649 while (!list_empty(&req->link_list)) {
650 link = list_first_entry(&req->link_list, struct io_kiocb, list);
651 list_del(&link->list);
652
653 io_cqring_add_event(req->ctx, link->user_data, -ECANCELED);
654 __io_free_req(link);
655 }
656}
657
658static void io_free_req(struct io_kiocb *req)
659{
660 /*
661 * If LINK is set, we have dependent requests in this chain. If we
662 * didn't fail this request, queue the first one up, moving any other
663 * dependencies to the next request. In case of failure, fail the rest
664 * of the chain.
665 */
666 if (req->flags & REQ_F_LINK) {
667 if (req->flags & REQ_F_FAIL_LINK)
668 io_fail_links(req);
669 else
670 io_req_link_next(req);
671 }
672
673 __io_free_req(req);
674}
675
Jens Axboee65ef562019-03-12 10:16:44 -0600676static void io_put_req(struct io_kiocb *req)
677{
678 if (refcount_dec_and_test(&req->refs))
679 io_free_req(req);
Jens Axboe2b188cc2019-01-07 10:46:33 -0700680}
681
Jens Axboea3a0e432019-08-20 11:03:11 -0600682static unsigned io_cqring_events(struct io_cq_ring *ring)
683{
684 /* See comment at the top of this file */
685 smp_rmb();
686 return READ_ONCE(ring->r.tail) - READ_ONCE(ring->r.head);
687}
688
Jens Axboedef596e2019-01-09 08:59:42 -0700689/*
690 * Find and free completed poll iocbs
691 */
692static void io_iopoll_complete(struct io_ring_ctx *ctx, unsigned int *nr_events,
693 struct list_head *done)
694{
695 void *reqs[IO_IOPOLL_BATCH];
696 struct io_kiocb *req;
Jens Axboe09bb8392019-03-13 12:39:28 -0600697 int to_free;
Jens Axboedef596e2019-01-09 08:59:42 -0700698
Jens Axboe09bb8392019-03-13 12:39:28 -0600699 to_free = 0;
Jens Axboedef596e2019-01-09 08:59:42 -0700700 while (!list_empty(done)) {
701 req = list_first_entry(done, struct io_kiocb, list);
702 list_del(&req->list);
703
Jens Axboe9e645e112019-05-10 16:07:28 -0600704 io_cqring_fill_event(ctx, req->user_data, req->result);
Jens Axboedef596e2019-01-09 08:59:42 -0700705 (*nr_events)++;
706
Jens Axboe09bb8392019-03-13 12:39:28 -0600707 if (refcount_dec_and_test(&req->refs)) {
708 /* If we're not using fixed files, we have to pair the
709 * completion part with the file put. Use regular
710 * completions for those, only batch free for fixed
Jens Axboe9e645e112019-05-10 16:07:28 -0600711 * file and non-linked commands.
Jens Axboe09bb8392019-03-13 12:39:28 -0600712 */
Jens Axboe9e645e112019-05-10 16:07:28 -0600713 if ((req->flags & (REQ_F_FIXED_FILE|REQ_F_LINK)) ==
714 REQ_F_FIXED_FILE) {
Jens Axboe09bb8392019-03-13 12:39:28 -0600715 reqs[to_free++] = req;
716 if (to_free == ARRAY_SIZE(reqs))
717 io_free_req_many(ctx, reqs, &to_free);
Jens Axboe6b063142019-01-10 22:13:58 -0700718 } else {
Jens Axboe09bb8392019-03-13 12:39:28 -0600719 io_free_req(req);
Jens Axboe6b063142019-01-10 22:13:58 -0700720 }
Jens Axboe9a56a232019-01-09 09:06:50 -0700721 }
Jens Axboedef596e2019-01-09 08:59:42 -0700722 }
Jens Axboedef596e2019-01-09 08:59:42 -0700723
Jens Axboe09bb8392019-03-13 12:39:28 -0600724 io_commit_cqring(ctx);
Jens Axboedef596e2019-01-09 08:59:42 -0700725 io_free_req_many(ctx, reqs, &to_free);
726}
727
728static int io_do_iopoll(struct io_ring_ctx *ctx, unsigned int *nr_events,
729 long min)
730{
731 struct io_kiocb *req, *tmp;
732 LIST_HEAD(done);
733 bool spin;
734 int ret;
735
736 /*
737 * Only spin for completions if we don't have multiple devices hanging
738 * off our complete list, and we're under the requested amount.
739 */
740 spin = !ctx->poll_multi_file && *nr_events < min;
741
742 ret = 0;
743 list_for_each_entry_safe(req, tmp, &ctx->poll_list, list) {
744 struct kiocb *kiocb = &req->rw;
745
746 /*
747 * Move completed entries to our local list. If we find a
748 * request that requires polling, break out and complete
749 * the done list first, if we have entries there.
750 */
751 if (req->flags & REQ_F_IOPOLL_COMPLETED) {
752 list_move_tail(&req->list, &done);
753 continue;
754 }
755 if (!list_empty(&done))
756 break;
757
758 ret = kiocb->ki_filp->f_op->iopoll(kiocb, spin);
759 if (ret < 0)
760 break;
761
762 if (ret && spin)
763 spin = false;
764 ret = 0;
765 }
766
767 if (!list_empty(&done))
768 io_iopoll_complete(ctx, nr_events, &done);
769
770 return ret;
771}
772
773/*
774 * Poll for a mininum of 'min' events. Note that if min == 0 we consider that a
775 * non-spinning poll check - we'll still enter the driver poll loop, but only
776 * as a non-spinning completion check.
777 */
778static int io_iopoll_getevents(struct io_ring_ctx *ctx, unsigned int *nr_events,
779 long min)
780{
Jens Axboe08f54392019-08-21 22:19:11 -0600781 while (!list_empty(&ctx->poll_list) && !need_resched()) {
Jens Axboedef596e2019-01-09 08:59:42 -0700782 int ret;
783
784 ret = io_do_iopoll(ctx, nr_events, min);
785 if (ret < 0)
786 return ret;
787 if (!min || *nr_events >= min)
788 return 0;
789 }
790
791 return 1;
792}
793
794/*
795 * We can't just wait for polled events to come to us, we have to actively
796 * find and complete them.
797 */
798static void io_iopoll_reap_events(struct io_ring_ctx *ctx)
799{
800 if (!(ctx->flags & IORING_SETUP_IOPOLL))
801 return;
802
803 mutex_lock(&ctx->uring_lock);
804 while (!list_empty(&ctx->poll_list)) {
805 unsigned int nr_events = 0;
806
807 io_iopoll_getevents(ctx, &nr_events, 1);
Jens Axboe08f54392019-08-21 22:19:11 -0600808
809 /*
810 * Ensure we allow local-to-the-cpu processing to take place,
811 * in this case we need to ensure that we reap all events.
812 */
813 cond_resched();
Jens Axboedef596e2019-01-09 08:59:42 -0700814 }
815 mutex_unlock(&ctx->uring_lock);
816}
817
818static int io_iopoll_check(struct io_ring_ctx *ctx, unsigned *nr_events,
819 long min)
820{
Jens Axboe500f9fb2019-08-19 12:15:59 -0600821 int iters, ret = 0;
Jens Axboedef596e2019-01-09 08:59:42 -0700822
Jens Axboe500f9fb2019-08-19 12:15:59 -0600823 /*
824 * We disallow the app entering submit/complete with polling, but we
825 * still need to lock the ring to prevent racing with polled issue
826 * that got punted to a workqueue.
827 */
828 mutex_lock(&ctx->uring_lock);
829
830 iters = 0;
Jens Axboedef596e2019-01-09 08:59:42 -0700831 do {
832 int tmin = 0;
833
Jens Axboe500f9fb2019-08-19 12:15:59 -0600834 /*
Jens Axboea3a0e432019-08-20 11:03:11 -0600835 * Don't enter poll loop if we already have events pending.
836 * If we do, we can potentially be spinning for commands that
837 * already triggered a CQE (eg in error).
838 */
839 if (io_cqring_events(ctx->cq_ring))
840 break;
841
842 /*
Jens Axboe500f9fb2019-08-19 12:15:59 -0600843 * If a submit got punted to a workqueue, we can have the
844 * application entering polling for a command before it gets
845 * issued. That app will hold the uring_lock for the duration
846 * of the poll right here, so we need to take a breather every
847 * now and then to ensure that the issue has a chance to add
848 * the poll to the issued list. Otherwise we can spin here
849 * forever, while the workqueue is stuck trying to acquire the
850 * very same mutex.
851 */
852 if (!(++iters & 7)) {
853 mutex_unlock(&ctx->uring_lock);
854 mutex_lock(&ctx->uring_lock);
855 }
856
Jens Axboedef596e2019-01-09 08:59:42 -0700857 if (*nr_events < min)
858 tmin = min - *nr_events;
859
860 ret = io_iopoll_getevents(ctx, nr_events, tmin);
861 if (ret <= 0)
862 break;
863 ret = 0;
864 } while (min && !*nr_events && !need_resched());
865
Jens Axboe500f9fb2019-08-19 12:15:59 -0600866 mutex_unlock(&ctx->uring_lock);
Jens Axboedef596e2019-01-09 08:59:42 -0700867 return ret;
868}
869
Jens Axboe2b188cc2019-01-07 10:46:33 -0700870static void kiocb_end_write(struct kiocb *kiocb)
871{
872 if (kiocb->ki_flags & IOCB_WRITE) {
873 struct inode *inode = file_inode(kiocb->ki_filp);
874
875 /*
876 * Tell lockdep we inherited freeze protection from submission
877 * thread.
878 */
879 if (S_ISREG(inode->i_mode))
880 __sb_writers_acquired(inode->i_sb, SB_FREEZE_WRITE);
881 file_end_write(kiocb->ki_filp);
882 }
883}
884
885static void io_complete_rw(struct kiocb *kiocb, long res, long res2)
886{
887 struct io_kiocb *req = container_of(kiocb, struct io_kiocb, rw);
888
889 kiocb_end_write(kiocb);
890
Jens Axboe9e645e112019-05-10 16:07:28 -0600891 if ((req->flags & REQ_F_LINK) && res != req->result)
892 req->flags |= REQ_F_FAIL_LINK;
Jens Axboec71ffb62019-05-13 20:58:29 -0600893 io_cqring_add_event(req->ctx, req->user_data, res);
Jens Axboee65ef562019-03-12 10:16:44 -0600894 io_put_req(req);
Jens Axboe2b188cc2019-01-07 10:46:33 -0700895}
896
Jens Axboedef596e2019-01-09 08:59:42 -0700897static void io_complete_rw_iopoll(struct kiocb *kiocb, long res, long res2)
898{
899 struct io_kiocb *req = container_of(kiocb, struct io_kiocb, rw);
900
901 kiocb_end_write(kiocb);
902
Jens Axboe9e645e112019-05-10 16:07:28 -0600903 if ((req->flags & REQ_F_LINK) && res != req->result)
904 req->flags |= REQ_F_FAIL_LINK;
905 req->result = res;
Jens Axboedef596e2019-01-09 08:59:42 -0700906 if (res != -EAGAIN)
907 req->flags |= REQ_F_IOPOLL_COMPLETED;
908}
909
910/*
911 * After the iocb has been issued, it's safe to be found on the poll list.
912 * Adding the kiocb to the list AFTER submission ensures that we don't
913 * find it from a io_iopoll_getevents() thread before the issuer is done
914 * accessing the kiocb cookie.
915 */
916static void io_iopoll_req_issued(struct io_kiocb *req)
917{
918 struct io_ring_ctx *ctx = req->ctx;
919
920 /*
921 * Track whether we have multiple files in our lists. This will impact
922 * how we do polling eventually, not spinning if we're on potentially
923 * different devices.
924 */
925 if (list_empty(&ctx->poll_list)) {
926 ctx->poll_multi_file = false;
927 } else if (!ctx->poll_multi_file) {
928 struct io_kiocb *list_req;
929
930 list_req = list_first_entry(&ctx->poll_list, struct io_kiocb,
931 list);
932 if (list_req->rw.ki_filp != req->rw.ki_filp)
933 ctx->poll_multi_file = true;
934 }
935
936 /*
937 * For fast devices, IO may have already completed. If it has, add
938 * it to the front so we find it first.
939 */
940 if (req->flags & REQ_F_IOPOLL_COMPLETED)
941 list_add(&req->list, &ctx->poll_list);
942 else
943 list_add_tail(&req->list, &ctx->poll_list);
944}
945
Jens Axboe3d6770f2019-04-13 11:50:54 -0600946static void io_file_put(struct io_submit_state *state)
Jens Axboe9a56a232019-01-09 09:06:50 -0700947{
Jens Axboe3d6770f2019-04-13 11:50:54 -0600948 if (state->file) {
Jens Axboe9a56a232019-01-09 09:06:50 -0700949 int diff = state->has_refs - state->used_refs;
950
951 if (diff)
952 fput_many(state->file, diff);
953 state->file = NULL;
954 }
955}
956
957/*
958 * Get as many references to a file as we have IOs left in this submission,
959 * assuming most submissions are for one file, or at least that each file
960 * has more than one submission.
961 */
962static struct file *io_file_get(struct io_submit_state *state, int fd)
963{
964 if (!state)
965 return fget(fd);
966
967 if (state->file) {
968 if (state->fd == fd) {
969 state->used_refs++;
970 state->ios_left--;
971 return state->file;
972 }
Jens Axboe3d6770f2019-04-13 11:50:54 -0600973 io_file_put(state);
Jens Axboe9a56a232019-01-09 09:06:50 -0700974 }
975 state->file = fget_many(fd, state->ios_left);
976 if (!state->file)
977 return NULL;
978
979 state->fd = fd;
980 state->has_refs = state->ios_left;
981 state->used_refs = 1;
982 state->ios_left--;
983 return state->file;
984}
985
Jens Axboe2b188cc2019-01-07 10:46:33 -0700986/*
987 * If we tracked the file through the SCM inflight mechanism, we could support
988 * any file. For now, just ensure that anything potentially problematic is done
989 * inline.
990 */
991static bool io_file_supports_async(struct file *file)
992{
993 umode_t mode = file_inode(file)->i_mode;
994
995 if (S_ISBLK(mode) || S_ISCHR(mode))
996 return true;
997 if (S_ISREG(mode) && file->f_op != &io_uring_fops)
998 return true;
999
1000 return false;
1001}
1002
Jens Axboe6c271ce2019-01-10 11:22:30 -07001003static int io_prep_rw(struct io_kiocb *req, const struct sqe_submit *s,
Jens Axboe8358e3a2019-04-23 08:17:58 -06001004 bool force_nonblock)
Jens Axboe2b188cc2019-01-07 10:46:33 -07001005{
Jens Axboe6c271ce2019-01-10 11:22:30 -07001006 const struct io_uring_sqe *sqe = s->sqe;
Jens Axboedef596e2019-01-09 08:59:42 -07001007 struct io_ring_ctx *ctx = req->ctx;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001008 struct kiocb *kiocb = &req->rw;
Jens Axboe09bb8392019-03-13 12:39:28 -06001009 unsigned ioprio;
1010 int ret;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001011
Jens Axboe09bb8392019-03-13 12:39:28 -06001012 if (!req->file)
1013 return -EBADF;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001014
Jens Axboe09bb8392019-03-13 12:39:28 -06001015 if (force_nonblock && !io_file_supports_async(req->file))
1016 force_nonblock = false;
Jens Axboe6b063142019-01-10 22:13:58 -07001017
Jens Axboe2b188cc2019-01-07 10:46:33 -07001018 kiocb->ki_pos = READ_ONCE(sqe->off);
1019 kiocb->ki_flags = iocb_flags(kiocb->ki_filp);
1020 kiocb->ki_hint = ki_hint_validate(file_write_hint(kiocb->ki_filp));
1021
1022 ioprio = READ_ONCE(sqe->ioprio);
1023 if (ioprio) {
1024 ret = ioprio_check_cap(ioprio);
1025 if (ret)
Jens Axboe09bb8392019-03-13 12:39:28 -06001026 return ret;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001027
1028 kiocb->ki_ioprio = ioprio;
1029 } else
1030 kiocb->ki_ioprio = get_current_ioprio();
1031
1032 ret = kiocb_set_rw_flags(kiocb, READ_ONCE(sqe->rw_flags));
1033 if (unlikely(ret))
Jens Axboe09bb8392019-03-13 12:39:28 -06001034 return ret;
Stefan Bühler8449eed2019-04-27 20:34:19 +02001035
1036 /* don't allow async punt if RWF_NOWAIT was requested */
1037 if (kiocb->ki_flags & IOCB_NOWAIT)
1038 req->flags |= REQ_F_NOWAIT;
1039
1040 if (force_nonblock)
Jens Axboe2b188cc2019-01-07 10:46:33 -07001041 kiocb->ki_flags |= IOCB_NOWAIT;
Stefan Bühler8449eed2019-04-27 20:34:19 +02001042
Jens Axboedef596e2019-01-09 08:59:42 -07001043 if (ctx->flags & IORING_SETUP_IOPOLL) {
Jens Axboedef596e2019-01-09 08:59:42 -07001044 if (!(kiocb->ki_flags & IOCB_DIRECT) ||
1045 !kiocb->ki_filp->f_op->iopoll)
Jens Axboe09bb8392019-03-13 12:39:28 -06001046 return -EOPNOTSUPP;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001047
Jens Axboedef596e2019-01-09 08:59:42 -07001048 kiocb->ki_flags |= IOCB_HIPRI;
1049 kiocb->ki_complete = io_complete_rw_iopoll;
1050 } else {
Jens Axboe09bb8392019-03-13 12:39:28 -06001051 if (kiocb->ki_flags & IOCB_HIPRI)
1052 return -EINVAL;
Jens Axboedef596e2019-01-09 08:59:42 -07001053 kiocb->ki_complete = io_complete_rw;
1054 }
Jens Axboe2b188cc2019-01-07 10:46:33 -07001055 return 0;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001056}
1057
1058static inline void io_rw_done(struct kiocb *kiocb, ssize_t ret)
1059{
1060 switch (ret) {
1061 case -EIOCBQUEUED:
1062 break;
1063 case -ERESTARTSYS:
1064 case -ERESTARTNOINTR:
1065 case -ERESTARTNOHAND:
1066 case -ERESTART_RESTARTBLOCK:
1067 /*
1068 * We can't just restart the syscall, since previously
1069 * submitted sqes may already be in progress. Just fail this
1070 * IO with EINTR.
1071 */
1072 ret = -EINTR;
1073 /* fall through */
1074 default:
1075 kiocb->ki_complete(kiocb, ret, 0);
1076 }
1077}
1078
Jens Axboeedafcce2019-01-09 09:16:05 -07001079static int io_import_fixed(struct io_ring_ctx *ctx, int rw,
1080 const struct io_uring_sqe *sqe,
1081 struct iov_iter *iter)
1082{
1083 size_t len = READ_ONCE(sqe->len);
1084 struct io_mapped_ubuf *imu;
1085 unsigned index, buf_index;
1086 size_t offset;
1087 u64 buf_addr;
1088
1089 /* attempt to use fixed buffers without having provided iovecs */
1090 if (unlikely(!ctx->user_bufs))
1091 return -EFAULT;
1092
1093 buf_index = READ_ONCE(sqe->buf_index);
1094 if (unlikely(buf_index >= ctx->nr_user_bufs))
1095 return -EFAULT;
1096
1097 index = array_index_nospec(buf_index, ctx->nr_user_bufs);
1098 imu = &ctx->user_bufs[index];
1099 buf_addr = READ_ONCE(sqe->addr);
1100
1101 /* overflow */
1102 if (buf_addr + len < buf_addr)
1103 return -EFAULT;
1104 /* not inside the mapped region */
1105 if (buf_addr < imu->ubuf || buf_addr + len > imu->ubuf + imu->len)
1106 return -EFAULT;
1107
1108 /*
1109 * May not be a start of buffer, set size appropriately
1110 * and advance us to the beginning.
1111 */
1112 offset = buf_addr - imu->ubuf;
1113 iov_iter_bvec(iter, rw, imu->bvec, imu->nr_bvecs, offset + len);
Jens Axboebd11b3a2019-07-20 08:37:31 -06001114
1115 if (offset) {
1116 /*
1117 * Don't use iov_iter_advance() here, as it's really slow for
1118 * using the latter parts of a big fixed buffer - it iterates
1119 * over each segment manually. We can cheat a bit here, because
1120 * we know that:
1121 *
1122 * 1) it's a BVEC iter, we set it up
1123 * 2) all bvecs are PAGE_SIZE in size, except potentially the
1124 * first and last bvec
1125 *
1126 * So just find our index, and adjust the iterator afterwards.
1127 * If the offset is within the first bvec (or the whole first
1128 * bvec, just use iov_iter_advance(). This makes it easier
1129 * since we can just skip the first segment, which may not
1130 * be PAGE_SIZE aligned.
1131 */
1132 const struct bio_vec *bvec = imu->bvec;
1133
1134 if (offset <= bvec->bv_len) {
1135 iov_iter_advance(iter, offset);
1136 } else {
1137 unsigned long seg_skip;
1138
1139 /* skip first vec */
1140 offset -= bvec->bv_len;
1141 seg_skip = 1 + (offset >> PAGE_SHIFT);
1142
1143 iter->bvec = bvec + seg_skip;
1144 iter->nr_segs -= seg_skip;
Aleix Roca Nonell99c79f62019-08-15 14:03:22 +02001145 iter->count -= bvec->bv_len + offset;
Jens Axboebd11b3a2019-07-20 08:37:31 -06001146 iter->iov_offset = offset & ~PAGE_MASK;
Jens Axboebd11b3a2019-07-20 08:37:31 -06001147 }
1148 }
1149
Jens Axboeedafcce2019-01-09 09:16:05 -07001150 return 0;
1151}
1152
Jens Axboe87e5e6d2019-05-14 16:02:22 -06001153static ssize_t io_import_iovec(struct io_ring_ctx *ctx, int rw,
1154 const struct sqe_submit *s, struct iovec **iovec,
1155 struct iov_iter *iter)
Jens Axboe2b188cc2019-01-07 10:46:33 -07001156{
1157 const struct io_uring_sqe *sqe = s->sqe;
1158 void __user *buf = u64_to_user_ptr(READ_ONCE(sqe->addr));
1159 size_t sqe_len = READ_ONCE(sqe->len);
Jens Axboeedafcce2019-01-09 09:16:05 -07001160 u8 opcode;
1161
1162 /*
1163 * We're reading ->opcode for the second time, but the first read
1164 * doesn't care whether it's _FIXED or not, so it doesn't matter
1165 * whether ->opcode changes concurrently. The first read does care
1166 * about whether it is a READ or a WRITE, so we don't trust this read
1167 * for that purpose and instead let the caller pass in the read/write
1168 * flag.
1169 */
1170 opcode = READ_ONCE(sqe->opcode);
1171 if (opcode == IORING_OP_READ_FIXED ||
1172 opcode == IORING_OP_WRITE_FIXED) {
Jens Axboe87e5e6d2019-05-14 16:02:22 -06001173 ssize_t ret = io_import_fixed(ctx, rw, sqe, iter);
Jens Axboeedafcce2019-01-09 09:16:05 -07001174 *iovec = NULL;
1175 return ret;
1176 }
Jens Axboe2b188cc2019-01-07 10:46:33 -07001177
1178 if (!s->has_user)
1179 return -EFAULT;
1180
1181#ifdef CONFIG_COMPAT
1182 if (ctx->compat)
1183 return compat_import_iovec(rw, buf, sqe_len, UIO_FASTIOV,
1184 iovec, iter);
1185#endif
1186
1187 return import_iovec(rw, buf, sqe_len, UIO_FASTIOV, iovec, iter);
1188}
1189
Jens Axboe31b51512019-01-18 22:56:34 -07001190/*
1191 * Make a note of the last file/offset/direction we punted to async
1192 * context. We'll use this information to see if we can piggy back a
1193 * sequential request onto the previous one, if it's still hasn't been
1194 * completed by the async worker.
1195 */
1196static void io_async_list_note(int rw, struct io_kiocb *req, size_t len)
1197{
1198 struct async_list *async_list = &req->ctx->pending_async[rw];
1199 struct kiocb *kiocb = &req->rw;
1200 struct file *filp = kiocb->ki_filp;
1201 off_t io_end = kiocb->ki_pos + len;
1202
1203 if (filp == async_list->file && kiocb->ki_pos == async_list->io_end) {
Zhengyuan Liu9310a7ba2019-07-22 10:23:27 +08001204 unsigned long max_bytes;
Jens Axboe31b51512019-01-18 22:56:34 -07001205
1206 /* Use 8x RA size as a decent limiter for both reads/writes */
Zhengyuan Liu9310a7ba2019-07-22 10:23:27 +08001207 max_bytes = filp->f_ra.ra_pages << (PAGE_SHIFT + 3);
1208 if (!max_bytes)
1209 max_bytes = VM_READAHEAD_PAGES << (PAGE_SHIFT + 3);
Jens Axboe31b51512019-01-18 22:56:34 -07001210
Zhengyuan Liu9310a7ba2019-07-22 10:23:27 +08001211 /* If max len are exceeded, reset the state */
1212 if (async_list->io_len + len <= max_bytes) {
Jens Axboe31b51512019-01-18 22:56:34 -07001213 req->flags |= REQ_F_SEQ_PREV;
Zhengyuan Liu9310a7ba2019-07-22 10:23:27 +08001214 async_list->io_len += len;
Jens Axboe31b51512019-01-18 22:56:34 -07001215 } else {
1216 io_end = 0;
Zhengyuan Liu9310a7ba2019-07-22 10:23:27 +08001217 async_list->io_len = 0;
Jens Axboe31b51512019-01-18 22:56:34 -07001218 }
1219 }
1220
1221 /* New file? Reset state. */
1222 if (async_list->file != filp) {
Zhengyuan Liu9310a7ba2019-07-22 10:23:27 +08001223 async_list->io_len = 0;
Jens Axboe31b51512019-01-18 22:56:34 -07001224 async_list->file = filp;
1225 }
1226 async_list->io_end = io_end;
1227}
1228
Jens Axboee0c5c572019-03-12 10:18:47 -06001229static int io_read(struct io_kiocb *req, const struct sqe_submit *s,
Jens Axboe8358e3a2019-04-23 08:17:58 -06001230 bool force_nonblock)
Jens Axboe2b188cc2019-01-07 10:46:33 -07001231{
1232 struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs;
1233 struct kiocb *kiocb = &req->rw;
1234 struct iov_iter iter;
1235 struct file *file;
Jens Axboe31b51512019-01-18 22:56:34 -07001236 size_t iov_count;
Jens Axboe9d93a3f2019-05-15 13:53:07 -06001237 ssize_t read_size, ret;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001238
Jens Axboe8358e3a2019-04-23 08:17:58 -06001239 ret = io_prep_rw(req, s, force_nonblock);
Jens Axboe2b188cc2019-01-07 10:46:33 -07001240 if (ret)
1241 return ret;
1242 file = kiocb->ki_filp;
1243
Jens Axboe2b188cc2019-01-07 10:46:33 -07001244 if (unlikely(!(file->f_mode & FMODE_READ)))
Jens Axboe09bb8392019-03-13 12:39:28 -06001245 return -EBADF;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001246 if (unlikely(!file->f_op->read_iter))
Jens Axboe09bb8392019-03-13 12:39:28 -06001247 return -EINVAL;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001248
1249 ret = io_import_iovec(req->ctx, READ, s, &iovec, &iter);
Jens Axboe87e5e6d2019-05-14 16:02:22 -06001250 if (ret < 0)
Jens Axboe09bb8392019-03-13 12:39:28 -06001251 return ret;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001252
Jens Axboe9d93a3f2019-05-15 13:53:07 -06001253 read_size = ret;
Jens Axboe9e645e112019-05-10 16:07:28 -06001254 if (req->flags & REQ_F_LINK)
1255 req->result = read_size;
1256
Jens Axboe31b51512019-01-18 22:56:34 -07001257 iov_count = iov_iter_count(&iter);
1258 ret = rw_verify_area(READ, file, &kiocb->ki_pos, iov_count);
Jens Axboe2b188cc2019-01-07 10:46:33 -07001259 if (!ret) {
1260 ssize_t ret2;
1261
Jens Axboe2b188cc2019-01-07 10:46:33 -07001262 ret2 = call_read_iter(file, kiocb, &iter);
Jens Axboe9d93a3f2019-05-15 13:53:07 -06001263 /*
1264 * In case of a short read, punt to async. This can happen
1265 * if we have data partially cached. Alternatively we can
1266 * return the short read, in which case the application will
1267 * need to issue another SQE and wait for it. That SQE will
1268 * need async punt anyway, so it's more efficient to do it
1269 * here.
1270 */
1271 if (force_nonblock && ret2 > 0 && ret2 < read_size)
1272 ret2 = -EAGAIN;
1273 /* Catch -EAGAIN return for forced non-blocking submission */
Jens Axboe31b51512019-01-18 22:56:34 -07001274 if (!force_nonblock || ret2 != -EAGAIN) {
Jens Axboe2b188cc2019-01-07 10:46:33 -07001275 io_rw_done(kiocb, ret2);
Jens Axboe31b51512019-01-18 22:56:34 -07001276 } else {
1277 /*
1278 * If ->needs_lock is true, we're already in async
1279 * context.
1280 */
1281 if (!s->needs_lock)
1282 io_async_list_note(READ, req, iov_count);
Jens Axboe2b188cc2019-01-07 10:46:33 -07001283 ret = -EAGAIN;
Jens Axboe31b51512019-01-18 22:56:34 -07001284 }
Jens Axboe2b188cc2019-01-07 10:46:33 -07001285 }
1286 kfree(iovec);
Jens Axboe2b188cc2019-01-07 10:46:33 -07001287 return ret;
1288}
1289
Jens Axboee0c5c572019-03-12 10:18:47 -06001290static int io_write(struct io_kiocb *req, const struct sqe_submit *s,
Jens Axboe8358e3a2019-04-23 08:17:58 -06001291 bool force_nonblock)
Jens Axboe2b188cc2019-01-07 10:46:33 -07001292{
1293 struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs;
1294 struct kiocb *kiocb = &req->rw;
1295 struct iov_iter iter;
1296 struct file *file;
Jens Axboe31b51512019-01-18 22:56:34 -07001297 size_t iov_count;
Jens Axboe87e5e6d2019-05-14 16:02:22 -06001298 ssize_t ret;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001299
Jens Axboe8358e3a2019-04-23 08:17:58 -06001300 ret = io_prep_rw(req, s, force_nonblock);
Jens Axboe2b188cc2019-01-07 10:46:33 -07001301 if (ret)
1302 return ret;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001303
Jens Axboe2b188cc2019-01-07 10:46:33 -07001304 file = kiocb->ki_filp;
1305 if (unlikely(!(file->f_mode & FMODE_WRITE)))
Jens Axboe09bb8392019-03-13 12:39:28 -06001306 return -EBADF;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001307 if (unlikely(!file->f_op->write_iter))
Jens Axboe09bb8392019-03-13 12:39:28 -06001308 return -EINVAL;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001309
1310 ret = io_import_iovec(req->ctx, WRITE, s, &iovec, &iter);
Jens Axboe87e5e6d2019-05-14 16:02:22 -06001311 if (ret < 0)
Jens Axboe09bb8392019-03-13 12:39:28 -06001312 return ret;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001313
Jens Axboe9e645e112019-05-10 16:07:28 -06001314 if (req->flags & REQ_F_LINK)
1315 req->result = ret;
1316
Jens Axboe31b51512019-01-18 22:56:34 -07001317 iov_count = iov_iter_count(&iter);
1318
1319 ret = -EAGAIN;
1320 if (force_nonblock && !(kiocb->ki_flags & IOCB_DIRECT)) {
1321 /* If ->needs_lock is true, we're already in async context. */
1322 if (!s->needs_lock)
1323 io_async_list_note(WRITE, req, iov_count);
1324 goto out_free;
1325 }
1326
1327 ret = rw_verify_area(WRITE, file, &kiocb->ki_pos, iov_count);
Jens Axboe2b188cc2019-01-07 10:46:33 -07001328 if (!ret) {
Roman Penyaev9bf79332019-03-25 20:09:24 +01001329 ssize_t ret2;
1330
Jens Axboe2b188cc2019-01-07 10:46:33 -07001331 /*
1332 * Open-code file_start_write here to grab freeze protection,
1333 * which will be released by another thread in
1334 * io_complete_rw(). Fool lockdep by telling it the lock got
1335 * released so that it doesn't complain about the held lock when
1336 * we return to userspace.
1337 */
1338 if (S_ISREG(file_inode(file)->i_mode)) {
1339 __sb_start_write(file_inode(file)->i_sb,
1340 SB_FREEZE_WRITE, true);
1341 __sb_writers_release(file_inode(file)->i_sb,
1342 SB_FREEZE_WRITE);
1343 }
1344 kiocb->ki_flags |= IOCB_WRITE;
Roman Penyaev9bf79332019-03-25 20:09:24 +01001345
1346 ret2 = call_write_iter(file, kiocb, &iter);
1347 if (!force_nonblock || ret2 != -EAGAIN) {
1348 io_rw_done(kiocb, ret2);
1349 } else {
1350 /*
1351 * If ->needs_lock is true, we're already in async
1352 * context.
1353 */
1354 if (!s->needs_lock)
1355 io_async_list_note(WRITE, req, iov_count);
1356 ret = -EAGAIN;
1357 }
Jens Axboe2b188cc2019-01-07 10:46:33 -07001358 }
Jens Axboe31b51512019-01-18 22:56:34 -07001359out_free:
Jens Axboe2b188cc2019-01-07 10:46:33 -07001360 kfree(iovec);
Jens Axboe2b188cc2019-01-07 10:46:33 -07001361 return ret;
1362}
1363
1364/*
1365 * IORING_OP_NOP just posts a completion event, nothing else.
1366 */
1367static int io_nop(struct io_kiocb *req, u64 user_data)
1368{
1369 struct io_ring_ctx *ctx = req->ctx;
1370 long err = 0;
1371
Jens Axboedef596e2019-01-09 08:59:42 -07001372 if (unlikely(ctx->flags & IORING_SETUP_IOPOLL))
1373 return -EINVAL;
1374
Jens Axboec71ffb62019-05-13 20:58:29 -06001375 io_cqring_add_event(ctx, user_data, err);
Jens Axboee65ef562019-03-12 10:16:44 -06001376 io_put_req(req);
Jens Axboe2b188cc2019-01-07 10:46:33 -07001377 return 0;
1378}
1379
Christoph Hellwigc992fe22019-01-11 09:43:02 -07001380static int io_prep_fsync(struct io_kiocb *req, const struct io_uring_sqe *sqe)
1381{
Jens Axboe6b063142019-01-10 22:13:58 -07001382 struct io_ring_ctx *ctx = req->ctx;
Christoph Hellwigc992fe22019-01-11 09:43:02 -07001383
Jens Axboe09bb8392019-03-13 12:39:28 -06001384 if (!req->file)
1385 return -EBADF;
Christoph Hellwigc992fe22019-01-11 09:43:02 -07001386
Jens Axboe6b063142019-01-10 22:13:58 -07001387 if (unlikely(ctx->flags & IORING_SETUP_IOPOLL))
Jens Axboedef596e2019-01-09 08:59:42 -07001388 return -EINVAL;
Jens Axboeedafcce2019-01-09 09:16:05 -07001389 if (unlikely(sqe->addr || sqe->ioprio || sqe->buf_index))
Christoph Hellwigc992fe22019-01-11 09:43:02 -07001390 return -EINVAL;
1391
Christoph Hellwigc992fe22019-01-11 09:43:02 -07001392 return 0;
1393}
1394
1395static int io_fsync(struct io_kiocb *req, const struct io_uring_sqe *sqe,
1396 bool force_nonblock)
1397{
1398 loff_t sqe_off = READ_ONCE(sqe->off);
1399 loff_t sqe_len = READ_ONCE(sqe->len);
1400 loff_t end = sqe_off + sqe_len;
1401 unsigned fsync_flags;
1402 int ret;
1403
1404 fsync_flags = READ_ONCE(sqe->fsync_flags);
1405 if (unlikely(fsync_flags & ~IORING_FSYNC_DATASYNC))
1406 return -EINVAL;
1407
1408 ret = io_prep_fsync(req, sqe);
1409 if (ret)
1410 return ret;
1411
1412 /* fsync always requires a blocking context */
1413 if (force_nonblock)
1414 return -EAGAIN;
1415
1416 ret = vfs_fsync_range(req->rw.ki_filp, sqe_off,
1417 end > 0 ? end : LLONG_MAX,
1418 fsync_flags & IORING_FSYNC_DATASYNC);
1419
Jens Axboe9e645e112019-05-10 16:07:28 -06001420 if (ret < 0 && (req->flags & REQ_F_LINK))
1421 req->flags |= REQ_F_FAIL_LINK;
Jens Axboec71ffb62019-05-13 20:58:29 -06001422 io_cqring_add_event(req->ctx, sqe->user_data, ret);
Jens Axboee65ef562019-03-12 10:16:44 -06001423 io_put_req(req);
Christoph Hellwigc992fe22019-01-11 09:43:02 -07001424 return 0;
1425}
1426
Jens Axboe5d17b4a2019-04-09 14:56:44 -06001427static int io_prep_sfr(struct io_kiocb *req, const struct io_uring_sqe *sqe)
1428{
1429 struct io_ring_ctx *ctx = req->ctx;
1430 int ret = 0;
1431
1432 if (!req->file)
1433 return -EBADF;
Jens Axboe5d17b4a2019-04-09 14:56:44 -06001434
1435 if (unlikely(ctx->flags & IORING_SETUP_IOPOLL))
1436 return -EINVAL;
1437 if (unlikely(sqe->addr || sqe->ioprio || sqe->buf_index))
1438 return -EINVAL;
1439
Jens Axboe5d17b4a2019-04-09 14:56:44 -06001440 return ret;
1441}
1442
1443static int io_sync_file_range(struct io_kiocb *req,
1444 const struct io_uring_sqe *sqe,
1445 bool force_nonblock)
1446{
1447 loff_t sqe_off;
1448 loff_t sqe_len;
1449 unsigned flags;
1450 int ret;
1451
1452 ret = io_prep_sfr(req, sqe);
1453 if (ret)
1454 return ret;
1455
1456 /* sync_file_range always requires a blocking context */
1457 if (force_nonblock)
1458 return -EAGAIN;
1459
1460 sqe_off = READ_ONCE(sqe->off);
1461 sqe_len = READ_ONCE(sqe->len);
1462 flags = READ_ONCE(sqe->sync_range_flags);
1463
1464 ret = sync_file_range(req->rw.ki_filp, sqe_off, sqe_len, flags);
1465
Jens Axboe9e645e112019-05-10 16:07:28 -06001466 if (ret < 0 && (req->flags & REQ_F_LINK))
1467 req->flags |= REQ_F_FAIL_LINK;
Jens Axboec71ffb62019-05-13 20:58:29 -06001468 io_cqring_add_event(req->ctx, sqe->user_data, ret);
Jens Axboe5d17b4a2019-04-09 14:56:44 -06001469 io_put_req(req);
1470 return 0;
1471}
1472
Jens Axboe0fa03c62019-04-19 13:34:07 -06001473#if defined(CONFIG_NET)
Jens Axboeaa1fa282019-04-19 13:38:09 -06001474static int io_send_recvmsg(struct io_kiocb *req, const struct io_uring_sqe *sqe,
1475 bool force_nonblock,
1476 long (*fn)(struct socket *, struct user_msghdr __user *,
1477 unsigned int))
1478{
Jens Axboe0fa03c62019-04-19 13:34:07 -06001479 struct socket *sock;
1480 int ret;
1481
1482 if (unlikely(req->ctx->flags & IORING_SETUP_IOPOLL))
1483 return -EINVAL;
1484
1485 sock = sock_from_file(req->file, &ret);
1486 if (sock) {
1487 struct user_msghdr __user *msg;
1488 unsigned flags;
1489
1490 flags = READ_ONCE(sqe->msg_flags);
1491 if (flags & MSG_DONTWAIT)
1492 req->flags |= REQ_F_NOWAIT;
1493 else if (force_nonblock)
1494 flags |= MSG_DONTWAIT;
1495
1496 msg = (struct user_msghdr __user *) (unsigned long)
1497 READ_ONCE(sqe->addr);
1498
Jens Axboeaa1fa282019-04-19 13:38:09 -06001499 ret = fn(sock, msg, flags);
Jens Axboe0fa03c62019-04-19 13:34:07 -06001500 if (force_nonblock && ret == -EAGAIN)
1501 return ret;
1502 }
1503
1504 io_cqring_add_event(req->ctx, sqe->user_data, ret);
1505 io_put_req(req);
1506 return 0;
Jens Axboeaa1fa282019-04-19 13:38:09 -06001507}
1508#endif
1509
1510static int io_sendmsg(struct io_kiocb *req, const struct io_uring_sqe *sqe,
1511 bool force_nonblock)
1512{
1513#if defined(CONFIG_NET)
1514 return io_send_recvmsg(req, sqe, force_nonblock, __sys_sendmsg_sock);
1515#else
1516 return -EOPNOTSUPP;
1517#endif
1518}
1519
1520static int io_recvmsg(struct io_kiocb *req, const struct io_uring_sqe *sqe,
1521 bool force_nonblock)
1522{
1523#if defined(CONFIG_NET)
1524 return io_send_recvmsg(req, sqe, force_nonblock, __sys_recvmsg_sock);
Jens Axboe0fa03c62019-04-19 13:34:07 -06001525#else
1526 return -EOPNOTSUPP;
1527#endif
1528}
1529
Jens Axboe221c5eb2019-01-17 09:41:58 -07001530static void io_poll_remove_one(struct io_kiocb *req)
1531{
1532 struct io_poll_iocb *poll = &req->poll;
1533
1534 spin_lock(&poll->head->lock);
1535 WRITE_ONCE(poll->canceled, true);
1536 if (!list_empty(&poll->wait.entry)) {
1537 list_del_init(&poll->wait.entry);
1538 queue_work(req->ctx->sqo_wq, &req->work);
1539 }
1540 spin_unlock(&poll->head->lock);
1541
1542 list_del_init(&req->list);
1543}
1544
1545static void io_poll_remove_all(struct io_ring_ctx *ctx)
1546{
1547 struct io_kiocb *req;
1548
1549 spin_lock_irq(&ctx->completion_lock);
1550 while (!list_empty(&ctx->cancel_list)) {
1551 req = list_first_entry(&ctx->cancel_list, struct io_kiocb,list);
1552 io_poll_remove_one(req);
1553 }
1554 spin_unlock_irq(&ctx->completion_lock);
1555}
1556
1557/*
1558 * Find a running poll command that matches one specified in sqe->addr,
1559 * and remove it if found.
1560 */
1561static int io_poll_remove(struct io_kiocb *req, const struct io_uring_sqe *sqe)
1562{
1563 struct io_ring_ctx *ctx = req->ctx;
1564 struct io_kiocb *poll_req, *next;
1565 int ret = -ENOENT;
1566
1567 if (unlikely(req->ctx->flags & IORING_SETUP_IOPOLL))
1568 return -EINVAL;
1569 if (sqe->ioprio || sqe->off || sqe->len || sqe->buf_index ||
1570 sqe->poll_events)
1571 return -EINVAL;
1572
1573 spin_lock_irq(&ctx->completion_lock);
1574 list_for_each_entry_safe(poll_req, next, &ctx->cancel_list, list) {
1575 if (READ_ONCE(sqe->addr) == poll_req->user_data) {
1576 io_poll_remove_one(poll_req);
1577 ret = 0;
1578 break;
1579 }
1580 }
1581 spin_unlock_irq(&ctx->completion_lock);
1582
Jens Axboec71ffb62019-05-13 20:58:29 -06001583 io_cqring_add_event(req->ctx, sqe->user_data, ret);
Jens Axboee65ef562019-03-12 10:16:44 -06001584 io_put_req(req);
Jens Axboe221c5eb2019-01-17 09:41:58 -07001585 return 0;
1586}
1587
Jens Axboe8c838782019-03-12 15:48:16 -06001588static void io_poll_complete(struct io_ring_ctx *ctx, struct io_kiocb *req,
1589 __poll_t mask)
Jens Axboe221c5eb2019-01-17 09:41:58 -07001590{
Jens Axboe8c838782019-03-12 15:48:16 -06001591 req->poll.done = true;
Jens Axboec71ffb62019-05-13 20:58:29 -06001592 io_cqring_fill_event(ctx, req->user_data, mangle_poll(mask));
Jens Axboe8c838782019-03-12 15:48:16 -06001593 io_commit_cqring(ctx);
Jens Axboe221c5eb2019-01-17 09:41:58 -07001594}
1595
1596static void io_poll_complete_work(struct work_struct *work)
1597{
1598 struct io_kiocb *req = container_of(work, struct io_kiocb, work);
1599 struct io_poll_iocb *poll = &req->poll;
1600 struct poll_table_struct pt = { ._key = poll->events };
1601 struct io_ring_ctx *ctx = req->ctx;
1602 __poll_t mask = 0;
1603
1604 if (!READ_ONCE(poll->canceled))
1605 mask = vfs_poll(poll->file, &pt) & poll->events;
1606
1607 /*
1608 * Note that ->ki_cancel callers also delete iocb from active_reqs after
1609 * calling ->ki_cancel. We need the ctx_lock roundtrip here to
1610 * synchronize with them. In the cancellation case the list_del_init
1611 * itself is not actually needed, but harmless so we keep it in to
1612 * avoid further branches in the fast path.
1613 */
1614 spin_lock_irq(&ctx->completion_lock);
1615 if (!mask && !READ_ONCE(poll->canceled)) {
1616 add_wait_queue(poll->head, &poll->wait);
1617 spin_unlock_irq(&ctx->completion_lock);
1618 return;
1619 }
1620 list_del_init(&req->list);
Jens Axboe8c838782019-03-12 15:48:16 -06001621 io_poll_complete(ctx, req, mask);
Jens Axboe221c5eb2019-01-17 09:41:58 -07001622 spin_unlock_irq(&ctx->completion_lock);
1623
Jens Axboe8c838782019-03-12 15:48:16 -06001624 io_cqring_ev_posted(ctx);
1625 io_put_req(req);
Jens Axboe221c5eb2019-01-17 09:41:58 -07001626}
1627
1628static int io_poll_wake(struct wait_queue_entry *wait, unsigned mode, int sync,
1629 void *key)
1630{
1631 struct io_poll_iocb *poll = container_of(wait, struct io_poll_iocb,
1632 wait);
1633 struct io_kiocb *req = container_of(poll, struct io_kiocb, poll);
1634 struct io_ring_ctx *ctx = req->ctx;
1635 __poll_t mask = key_to_poll(key);
Jens Axboe8c838782019-03-12 15:48:16 -06001636 unsigned long flags;
Jens Axboe221c5eb2019-01-17 09:41:58 -07001637
1638 /* for instances that support it check for an event match first: */
Jens Axboe8c838782019-03-12 15:48:16 -06001639 if (mask && !(mask & poll->events))
1640 return 0;
Jens Axboe221c5eb2019-01-17 09:41:58 -07001641
1642 list_del_init(&poll->wait.entry);
Jens Axboe8c838782019-03-12 15:48:16 -06001643
1644 if (mask && spin_trylock_irqsave(&ctx->completion_lock, flags)) {
1645 list_del(&req->list);
1646 io_poll_complete(ctx, req, mask);
1647 spin_unlock_irqrestore(&ctx->completion_lock, flags);
1648
1649 io_cqring_ev_posted(ctx);
1650 io_put_req(req);
1651 } else {
1652 queue_work(ctx->sqo_wq, &req->work);
1653 }
1654
Jens Axboe221c5eb2019-01-17 09:41:58 -07001655 return 1;
1656}
1657
1658struct io_poll_table {
1659 struct poll_table_struct pt;
1660 struct io_kiocb *req;
1661 int error;
1662};
1663
1664static void io_poll_queue_proc(struct file *file, struct wait_queue_head *head,
1665 struct poll_table_struct *p)
1666{
1667 struct io_poll_table *pt = container_of(p, struct io_poll_table, pt);
1668
1669 if (unlikely(pt->req->poll.head)) {
1670 pt->error = -EINVAL;
1671 return;
1672 }
1673
1674 pt->error = 0;
1675 pt->req->poll.head = head;
1676 add_wait_queue(head, &pt->req->poll.wait);
1677}
1678
1679static int io_poll_add(struct io_kiocb *req, const struct io_uring_sqe *sqe)
1680{
1681 struct io_poll_iocb *poll = &req->poll;
1682 struct io_ring_ctx *ctx = req->ctx;
1683 struct io_poll_table ipt;
Jens Axboe8c838782019-03-12 15:48:16 -06001684 bool cancel = false;
Jens Axboe221c5eb2019-01-17 09:41:58 -07001685 __poll_t mask;
1686 u16 events;
Jens Axboe221c5eb2019-01-17 09:41:58 -07001687
1688 if (unlikely(req->ctx->flags & IORING_SETUP_IOPOLL))
1689 return -EINVAL;
1690 if (sqe->addr || sqe->ioprio || sqe->off || sqe->len || sqe->buf_index)
1691 return -EINVAL;
Jens Axboe09bb8392019-03-13 12:39:28 -06001692 if (!poll->file)
1693 return -EBADF;
Jens Axboe221c5eb2019-01-17 09:41:58 -07001694
1695 INIT_WORK(&req->work, io_poll_complete_work);
1696 events = READ_ONCE(sqe->poll_events);
1697 poll->events = demangle_poll(events) | EPOLLERR | EPOLLHUP;
1698
Jens Axboe221c5eb2019-01-17 09:41:58 -07001699 poll->head = NULL;
Jens Axboe8c838782019-03-12 15:48:16 -06001700 poll->done = false;
Jens Axboe221c5eb2019-01-17 09:41:58 -07001701 poll->canceled = false;
1702
1703 ipt.pt._qproc = io_poll_queue_proc;
1704 ipt.pt._key = poll->events;
1705 ipt.req = req;
1706 ipt.error = -EINVAL; /* same as no support for IOCB_CMD_POLL */
1707
1708 /* initialized the list so that we can do list_empty checks */
1709 INIT_LIST_HEAD(&poll->wait.entry);
1710 init_waitqueue_func_entry(&poll->wait, io_poll_wake);
1711
Jens Axboe36703242019-07-25 10:20:18 -06001712 INIT_LIST_HEAD(&req->list);
1713
Jens Axboe221c5eb2019-01-17 09:41:58 -07001714 mask = vfs_poll(poll->file, &ipt.pt) & poll->events;
Jens Axboe221c5eb2019-01-17 09:41:58 -07001715
1716 spin_lock_irq(&ctx->completion_lock);
Jens Axboe8c838782019-03-12 15:48:16 -06001717 if (likely(poll->head)) {
1718 spin_lock(&poll->head->lock);
1719 if (unlikely(list_empty(&poll->wait.entry))) {
1720 if (ipt.error)
1721 cancel = true;
1722 ipt.error = 0;
1723 mask = 0;
1724 }
1725 if (mask || ipt.error)
1726 list_del_init(&poll->wait.entry);
1727 else if (cancel)
1728 WRITE_ONCE(poll->canceled, true);
1729 else if (!poll->done) /* actually waiting for an event */
1730 list_add_tail(&req->list, &ctx->cancel_list);
1731 spin_unlock(&poll->head->lock);
Jens Axboe221c5eb2019-01-17 09:41:58 -07001732 }
Jens Axboe8c838782019-03-12 15:48:16 -06001733 if (mask) { /* no async, we'd stolen it */
Jens Axboe8c838782019-03-12 15:48:16 -06001734 ipt.error = 0;
1735 io_poll_complete(ctx, req, mask);
1736 }
Jens Axboe221c5eb2019-01-17 09:41:58 -07001737 spin_unlock_irq(&ctx->completion_lock);
1738
Jens Axboe8c838782019-03-12 15:48:16 -06001739 if (mask) {
1740 io_cqring_ev_posted(ctx);
Jens Axboee65ef562019-03-12 10:16:44 -06001741 io_put_req(req);
Jens Axboe221c5eb2019-01-17 09:41:58 -07001742 }
Jens Axboe8c838782019-03-12 15:48:16 -06001743 return ipt.error;
Jens Axboe221c5eb2019-01-17 09:41:58 -07001744}
1745
Jens Axboede0617e2019-04-06 21:51:27 -06001746static int io_req_defer(struct io_ring_ctx *ctx, struct io_kiocb *req,
1747 const struct io_uring_sqe *sqe)
1748{
1749 struct io_uring_sqe *sqe_copy;
1750
1751 if (!io_sequence_defer(ctx, req) && list_empty(&ctx->defer_list))
1752 return 0;
1753
1754 sqe_copy = kmalloc(sizeof(*sqe_copy), GFP_KERNEL);
1755 if (!sqe_copy)
1756 return -EAGAIN;
1757
1758 spin_lock_irq(&ctx->completion_lock);
1759 if (!io_sequence_defer(ctx, req) && list_empty(&ctx->defer_list)) {
1760 spin_unlock_irq(&ctx->completion_lock);
1761 kfree(sqe_copy);
1762 return 0;
1763 }
1764
1765 memcpy(sqe_copy, sqe, sizeof(*sqe_copy));
1766 req->submit.sqe = sqe_copy;
1767
1768 INIT_WORK(&req->work, io_sq_wq_submit_work);
1769 list_add_tail(&req->list, &ctx->defer_list);
1770 spin_unlock_irq(&ctx->completion_lock);
1771 return -EIOCBQUEUED;
1772}
1773
Jens Axboe2b188cc2019-01-07 10:46:33 -07001774static int __io_submit_sqe(struct io_ring_ctx *ctx, struct io_kiocb *req,
Jens Axboe8358e3a2019-04-23 08:17:58 -06001775 const struct sqe_submit *s, bool force_nonblock)
Jens Axboe2b188cc2019-01-07 10:46:33 -07001776{
Jens Axboee0c5c572019-03-12 10:18:47 -06001777 int ret, opcode;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001778
Jens Axboe9e645e112019-05-10 16:07:28 -06001779 req->user_data = READ_ONCE(s->sqe->user_data);
1780
Jens Axboe2b188cc2019-01-07 10:46:33 -07001781 if (unlikely(s->index >= ctx->sq_entries))
1782 return -EINVAL;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001783
1784 opcode = READ_ONCE(s->sqe->opcode);
1785 switch (opcode) {
1786 case IORING_OP_NOP:
1787 ret = io_nop(req, req->user_data);
1788 break;
1789 case IORING_OP_READV:
Jens Axboeedafcce2019-01-09 09:16:05 -07001790 if (unlikely(s->sqe->buf_index))
1791 return -EINVAL;
Jens Axboe8358e3a2019-04-23 08:17:58 -06001792 ret = io_read(req, s, force_nonblock);
Jens Axboe2b188cc2019-01-07 10:46:33 -07001793 break;
1794 case IORING_OP_WRITEV:
Jens Axboeedafcce2019-01-09 09:16:05 -07001795 if (unlikely(s->sqe->buf_index))
1796 return -EINVAL;
Jens Axboe8358e3a2019-04-23 08:17:58 -06001797 ret = io_write(req, s, force_nonblock);
Jens Axboeedafcce2019-01-09 09:16:05 -07001798 break;
1799 case IORING_OP_READ_FIXED:
Jens Axboe8358e3a2019-04-23 08:17:58 -06001800 ret = io_read(req, s, force_nonblock);
Jens Axboeedafcce2019-01-09 09:16:05 -07001801 break;
1802 case IORING_OP_WRITE_FIXED:
Jens Axboe8358e3a2019-04-23 08:17:58 -06001803 ret = io_write(req, s, force_nonblock);
Jens Axboe2b188cc2019-01-07 10:46:33 -07001804 break;
Christoph Hellwigc992fe22019-01-11 09:43:02 -07001805 case IORING_OP_FSYNC:
1806 ret = io_fsync(req, s->sqe, force_nonblock);
1807 break;
Jens Axboe221c5eb2019-01-17 09:41:58 -07001808 case IORING_OP_POLL_ADD:
1809 ret = io_poll_add(req, s->sqe);
1810 break;
1811 case IORING_OP_POLL_REMOVE:
1812 ret = io_poll_remove(req, s->sqe);
1813 break;
Jens Axboe5d17b4a2019-04-09 14:56:44 -06001814 case IORING_OP_SYNC_FILE_RANGE:
1815 ret = io_sync_file_range(req, s->sqe, force_nonblock);
1816 break;
Jens Axboe0fa03c62019-04-19 13:34:07 -06001817 case IORING_OP_SENDMSG:
1818 ret = io_sendmsg(req, s->sqe, force_nonblock);
1819 break;
Jens Axboeaa1fa282019-04-19 13:38:09 -06001820 case IORING_OP_RECVMSG:
1821 ret = io_recvmsg(req, s->sqe, force_nonblock);
1822 break;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001823 default:
1824 ret = -EINVAL;
1825 break;
1826 }
1827
Jens Axboedef596e2019-01-09 08:59:42 -07001828 if (ret)
1829 return ret;
1830
1831 if (ctx->flags & IORING_SETUP_IOPOLL) {
Jens Axboe9e645e112019-05-10 16:07:28 -06001832 if (req->result == -EAGAIN)
Jens Axboedef596e2019-01-09 08:59:42 -07001833 return -EAGAIN;
1834
1835 /* workqueue context doesn't hold uring_lock, grab it now */
1836 if (s->needs_lock)
1837 mutex_lock(&ctx->uring_lock);
1838 io_iopoll_req_issued(req);
1839 if (s->needs_lock)
1840 mutex_unlock(&ctx->uring_lock);
1841 }
1842
1843 return 0;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001844}
1845
Jens Axboe31b51512019-01-18 22:56:34 -07001846static struct async_list *io_async_list_from_sqe(struct io_ring_ctx *ctx,
1847 const struct io_uring_sqe *sqe)
1848{
1849 switch (sqe->opcode) {
1850 case IORING_OP_READV:
1851 case IORING_OP_READ_FIXED:
1852 return &ctx->pending_async[READ];
1853 case IORING_OP_WRITEV:
1854 case IORING_OP_WRITE_FIXED:
1855 return &ctx->pending_async[WRITE];
1856 default:
1857 return NULL;
1858 }
1859}
1860
Jens Axboeedafcce2019-01-09 09:16:05 -07001861static inline bool io_sqe_needs_user(const struct io_uring_sqe *sqe)
1862{
1863 u8 opcode = READ_ONCE(sqe->opcode);
1864
1865 return !(opcode == IORING_OP_READ_FIXED ||
1866 opcode == IORING_OP_WRITE_FIXED);
1867}
1868
Jens Axboe2b188cc2019-01-07 10:46:33 -07001869static void io_sq_wq_submit_work(struct work_struct *work)
1870{
1871 struct io_kiocb *req = container_of(work, struct io_kiocb, work);
Jens Axboe2b188cc2019-01-07 10:46:33 -07001872 struct io_ring_ctx *ctx = req->ctx;
Jens Axboe31b51512019-01-18 22:56:34 -07001873 struct mm_struct *cur_mm = NULL;
1874 struct async_list *async_list;
1875 LIST_HEAD(req_list);
Jens Axboeedafcce2019-01-09 09:16:05 -07001876 mm_segment_t old_fs;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001877 int ret;
1878
Jens Axboe31b51512019-01-18 22:56:34 -07001879 async_list = io_async_list_from_sqe(ctx, req->submit.sqe);
1880restart:
1881 do {
1882 struct sqe_submit *s = &req->submit;
1883 const struct io_uring_sqe *sqe = s->sqe;
Jackie Liud0ee8792019-07-31 14:39:33 +08001884 unsigned int flags = req->flags;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001885
Stefan Bühler8449eed2019-04-27 20:34:19 +02001886 /* Ensure we clear previously set non-block flag */
Jens Axboe31b51512019-01-18 22:56:34 -07001887 req->rw.ki_flags &= ~IOCB_NOWAIT;
1888
1889 ret = 0;
1890 if (io_sqe_needs_user(sqe) && !cur_mm) {
1891 if (!mmget_not_zero(ctx->sqo_mm)) {
1892 ret = -EFAULT;
1893 } else {
1894 cur_mm = ctx->sqo_mm;
1895 use_mm(cur_mm);
1896 old_fs = get_fs();
1897 set_fs(USER_DS);
1898 }
1899 }
1900
1901 if (!ret) {
1902 s->has_user = cur_mm != NULL;
1903 s->needs_lock = true;
1904 do {
Jens Axboe8358e3a2019-04-23 08:17:58 -06001905 ret = __io_submit_sqe(ctx, req, s, false);
Jens Axboe31b51512019-01-18 22:56:34 -07001906 /*
1907 * We can get EAGAIN for polled IO even though
1908 * we're forcing a sync submission from here,
1909 * since we can't wait for request slots on the
1910 * block side.
1911 */
1912 if (ret != -EAGAIN)
1913 break;
1914 cond_resched();
1915 } while (1);
1916 }
Jens Axboe817869d2019-04-30 14:44:05 -06001917
1918 /* drop submission reference */
1919 io_put_req(req);
1920
Jens Axboe31b51512019-01-18 22:56:34 -07001921 if (ret) {
Jens Axboec71ffb62019-05-13 20:58:29 -06001922 io_cqring_add_event(ctx, sqe->user_data, ret);
Jens Axboee65ef562019-03-12 10:16:44 -06001923 io_put_req(req);
Jens Axboe31b51512019-01-18 22:56:34 -07001924 }
1925
1926 /* async context always use a copy of the sqe */
1927 kfree(sqe);
1928
Zhengyuan Liuf7b76ac2019-07-16 23:26:14 +08001929 /* req from defer and link list needn't decrease async cnt */
Jackie Liud0ee8792019-07-31 14:39:33 +08001930 if (flags & (REQ_F_IO_DRAINED | REQ_F_LINK_DONE))
Zhengyuan Liuf7b76ac2019-07-16 23:26:14 +08001931 goto out;
1932
Jens Axboe31b51512019-01-18 22:56:34 -07001933 if (!async_list)
1934 break;
1935 if (!list_empty(&req_list)) {
1936 req = list_first_entry(&req_list, struct io_kiocb,
1937 list);
1938 list_del(&req->list);
1939 continue;
1940 }
1941 if (list_empty(&async_list->list))
1942 break;
1943
1944 req = NULL;
1945 spin_lock(&async_list->lock);
1946 if (list_empty(&async_list->list)) {
1947 spin_unlock(&async_list->lock);
1948 break;
1949 }
1950 list_splice_init(&async_list->list, &req_list);
1951 spin_unlock(&async_list->lock);
1952
1953 req = list_first_entry(&req_list, struct io_kiocb, list);
1954 list_del(&req->list);
1955 } while (req);
Jens Axboeedafcce2019-01-09 09:16:05 -07001956
1957 /*
Jens Axboe31b51512019-01-18 22:56:34 -07001958 * Rare case of racing with a submitter. If we find the count has
1959 * dropped to zero AND we have pending work items, then restart
1960 * the processing. This is a tiny race window.
Jens Axboeedafcce2019-01-09 09:16:05 -07001961 */
Jens Axboe31b51512019-01-18 22:56:34 -07001962 if (async_list) {
1963 ret = atomic_dec_return(&async_list->cnt);
1964 while (!ret && !list_empty(&async_list->list)) {
1965 spin_lock(&async_list->lock);
1966 atomic_inc(&async_list->cnt);
1967 list_splice_init(&async_list->list, &req_list);
1968 spin_unlock(&async_list->lock);
1969
1970 if (!list_empty(&req_list)) {
1971 req = list_first_entry(&req_list,
1972 struct io_kiocb, list);
1973 list_del(&req->list);
1974 goto restart;
1975 }
1976 ret = atomic_dec_return(&async_list->cnt);
Jens Axboeedafcce2019-01-09 09:16:05 -07001977 }
Jens Axboeedafcce2019-01-09 09:16:05 -07001978 }
Jens Axboe2b188cc2019-01-07 10:46:33 -07001979
Zhengyuan Liuf7b76ac2019-07-16 23:26:14 +08001980out:
Jens Axboe31b51512019-01-18 22:56:34 -07001981 if (cur_mm) {
Jens Axboeedafcce2019-01-09 09:16:05 -07001982 set_fs(old_fs);
Jens Axboe31b51512019-01-18 22:56:34 -07001983 unuse_mm(cur_mm);
1984 mmput(cur_mm);
Jens Axboeedafcce2019-01-09 09:16:05 -07001985 }
Jens Axboe31b51512019-01-18 22:56:34 -07001986}
Jens Axboe2b188cc2019-01-07 10:46:33 -07001987
Jens Axboe31b51512019-01-18 22:56:34 -07001988/*
1989 * See if we can piggy back onto previously submitted work, that is still
1990 * running. We currently only allow this if the new request is sequential
1991 * to the previous one we punted.
1992 */
1993static bool io_add_to_prev_work(struct async_list *list, struct io_kiocb *req)
1994{
1995 bool ret = false;
1996
1997 if (!list)
1998 return false;
1999 if (!(req->flags & REQ_F_SEQ_PREV))
2000 return false;
2001 if (!atomic_read(&list->cnt))
2002 return false;
2003
2004 ret = true;
2005 spin_lock(&list->lock);
2006 list_add_tail(&req->list, &list->list);
Zhengyuan Liuc0e48f92019-07-18 20:44:00 +08002007 /*
2008 * Ensure we see a simultaneous modification from io_sq_wq_submit_work()
2009 */
2010 smp_mb();
Jens Axboe31b51512019-01-18 22:56:34 -07002011 if (!atomic_read(&list->cnt)) {
2012 list_del_init(&req->list);
2013 ret = false;
2014 }
2015 spin_unlock(&list->lock);
2016 return ret;
Jens Axboe2b188cc2019-01-07 10:46:33 -07002017}
2018
Jens Axboe09bb8392019-03-13 12:39:28 -06002019static bool io_op_needs_file(const struct io_uring_sqe *sqe)
2020{
2021 int op = READ_ONCE(sqe->opcode);
2022
2023 switch (op) {
2024 case IORING_OP_NOP:
2025 case IORING_OP_POLL_REMOVE:
2026 return false;
2027 default:
2028 return true;
2029 }
2030}
2031
2032static int io_req_set_file(struct io_ring_ctx *ctx, const struct sqe_submit *s,
2033 struct io_submit_state *state, struct io_kiocb *req)
2034{
2035 unsigned flags;
2036 int fd;
2037
2038 flags = READ_ONCE(s->sqe->flags);
2039 fd = READ_ONCE(s->sqe->fd);
2040
Jens Axboede0617e2019-04-06 21:51:27 -06002041 if (flags & IOSQE_IO_DRAIN) {
2042 req->flags |= REQ_F_IO_DRAIN;
2043 req->sequence = ctx->cached_sq_head - 1;
2044 }
2045
Jens Axboe60c112b2019-06-21 10:20:18 -06002046 if (!io_op_needs_file(s->sqe))
Jens Axboe09bb8392019-03-13 12:39:28 -06002047 return 0;
Jens Axboe09bb8392019-03-13 12:39:28 -06002048
2049 if (flags & IOSQE_FIXED_FILE) {
2050 if (unlikely(!ctx->user_files ||
2051 (unsigned) fd >= ctx->nr_user_files))
2052 return -EBADF;
2053 req->file = ctx->user_files[fd];
2054 req->flags |= REQ_F_FIXED_FILE;
2055 } else {
2056 if (s->needs_fixed_file)
2057 return -EBADF;
2058 req->file = io_file_get(state, fd);
2059 if (unlikely(!req->file))
2060 return -EBADF;
2061 }
2062
2063 return 0;
2064}
2065
Jens Axboe9e645e112019-05-10 16:07:28 -06002066static int io_queue_sqe(struct io_ring_ctx *ctx, struct io_kiocb *req,
2067 struct sqe_submit *s)
Jens Axboe2b188cc2019-01-07 10:46:33 -07002068{
Jens Axboee0c5c572019-03-12 10:18:47 -06002069 int ret;
Jens Axboe2b188cc2019-01-07 10:46:33 -07002070
Jackie Liua982eeb2019-08-14 17:35:22 +08002071 ret = io_req_defer(ctx, req, s->sqe);
2072 if (ret) {
2073 if (ret != -EIOCBQUEUED) {
2074 io_free_req(req);
2075 io_cqring_add_event(ctx, s->sqe->user_data, ret);
2076 }
2077 return 0;
2078 }
2079
Jens Axboe8358e3a2019-04-23 08:17:58 -06002080 ret = __io_submit_sqe(ctx, req, s, true);
Stefan Bühler8449eed2019-04-27 20:34:19 +02002081 if (ret == -EAGAIN && !(req->flags & REQ_F_NOWAIT)) {
Jens Axboe2b188cc2019-01-07 10:46:33 -07002082 struct io_uring_sqe *sqe_copy;
2083
2084 sqe_copy = kmalloc(sizeof(*sqe_copy), GFP_KERNEL);
2085 if (sqe_copy) {
Jens Axboe31b51512019-01-18 22:56:34 -07002086 struct async_list *list;
2087
Jens Axboe2b188cc2019-01-07 10:46:33 -07002088 memcpy(sqe_copy, s->sqe, sizeof(*sqe_copy));
2089 s->sqe = sqe_copy;
2090
2091 memcpy(&req->submit, s, sizeof(*s));
Jens Axboe31b51512019-01-18 22:56:34 -07002092 list = io_async_list_from_sqe(ctx, s->sqe);
2093 if (!io_add_to_prev_work(list, req)) {
2094 if (list)
2095 atomic_inc(&list->cnt);
2096 INIT_WORK(&req->work, io_sq_wq_submit_work);
2097 queue_work(ctx->sqo_wq, &req->work);
2098 }
Jens Axboee65ef562019-03-12 10:16:44 -06002099
2100 /*
2101 * Queued up for async execution, worker will release
Jens Axboe9e645e112019-05-10 16:07:28 -06002102 * submit reference when the iocb is actually submitted.
Jens Axboee65ef562019-03-12 10:16:44 -06002103 */
2104 return 0;
Jens Axboe2b188cc2019-01-07 10:46:33 -07002105 }
2106 }
Jens Axboee65ef562019-03-12 10:16:44 -06002107
2108 /* drop submission reference */
2109 io_put_req(req);
2110
2111 /* and drop final reference, if we failed */
Jens Axboe9e645e112019-05-10 16:07:28 -06002112 if (ret) {
2113 io_cqring_add_event(ctx, req->user_data, ret);
2114 if (req->flags & REQ_F_LINK)
2115 req->flags |= REQ_F_FAIL_LINK;
Jens Axboee65ef562019-03-12 10:16:44 -06002116 io_put_req(req);
Jens Axboe9e645e112019-05-10 16:07:28 -06002117 }
Jens Axboe2b188cc2019-01-07 10:46:33 -07002118
2119 return ret;
2120}
2121
Jens Axboe9e645e112019-05-10 16:07:28 -06002122#define SQE_VALID_FLAGS (IOSQE_FIXED_FILE|IOSQE_IO_DRAIN|IOSQE_IO_LINK)
2123
2124static void io_submit_sqe(struct io_ring_ctx *ctx, struct sqe_submit *s,
2125 struct io_submit_state *state, struct io_kiocb **link)
2126{
2127 struct io_uring_sqe *sqe_copy;
2128 struct io_kiocb *req;
2129 int ret;
2130
2131 /* enforce forwards compatibility on users */
2132 if (unlikely(s->sqe->flags & ~SQE_VALID_FLAGS)) {
2133 ret = -EINVAL;
2134 goto err;
2135 }
2136
2137 req = io_get_req(ctx, state);
2138 if (unlikely(!req)) {
2139 ret = -EAGAIN;
2140 goto err;
2141 }
2142
2143 ret = io_req_set_file(ctx, s, state, req);
2144 if (unlikely(ret)) {
2145err_req:
2146 io_free_req(req);
2147err:
2148 io_cqring_add_event(ctx, s->sqe->user_data, ret);
2149 return;
2150 }
2151
Jens Axboe9e645e112019-05-10 16:07:28 -06002152 /*
2153 * If we already have a head request, queue this one for async
2154 * submittal once the head completes. If we don't have a head but
2155 * IOSQE_IO_LINK is set in the sqe, start a new head. This one will be
2156 * submitted sync once the chain is complete. If none of those
2157 * conditions are true (normal request), then just queue it.
2158 */
2159 if (*link) {
2160 struct io_kiocb *prev = *link;
2161
2162 sqe_copy = kmemdup(s->sqe, sizeof(*sqe_copy), GFP_KERNEL);
2163 if (!sqe_copy) {
2164 ret = -EAGAIN;
2165 goto err_req;
2166 }
2167
2168 s->sqe = sqe_copy;
2169 memcpy(&req->submit, s, sizeof(*s));
2170 list_add_tail(&req->list, &prev->link_list);
2171 } else if (s->sqe->flags & IOSQE_IO_LINK) {
2172 req->flags |= REQ_F_LINK;
2173
2174 memcpy(&req->submit, s, sizeof(*s));
2175 INIT_LIST_HEAD(&req->link_list);
2176 *link = req;
2177 } else {
2178 io_queue_sqe(ctx, req, s);
2179 }
2180}
2181
Jens Axboe9a56a232019-01-09 09:06:50 -07002182/*
2183 * Batched submission is done, ensure local IO is flushed out.
2184 */
2185static void io_submit_state_end(struct io_submit_state *state)
2186{
2187 blk_finish_plug(&state->plug);
Jens Axboe3d6770f2019-04-13 11:50:54 -06002188 io_file_put(state);
Jens Axboe2579f912019-01-09 09:10:43 -07002189 if (state->free_reqs)
2190 kmem_cache_free_bulk(req_cachep, state->free_reqs,
2191 &state->reqs[state->cur_req]);
Jens Axboe9a56a232019-01-09 09:06:50 -07002192}
2193
2194/*
2195 * Start submission side cache.
2196 */
2197static void io_submit_state_start(struct io_submit_state *state,
2198 struct io_ring_ctx *ctx, unsigned max_ios)
2199{
2200 blk_start_plug(&state->plug);
Jens Axboe2579f912019-01-09 09:10:43 -07002201 state->free_reqs = 0;
Jens Axboe9a56a232019-01-09 09:06:50 -07002202 state->file = NULL;
2203 state->ios_left = max_ios;
2204}
2205
Jens Axboe2b188cc2019-01-07 10:46:33 -07002206static void io_commit_sqring(struct io_ring_ctx *ctx)
2207{
2208 struct io_sq_ring *ring = ctx->sq_ring;
2209
2210 if (ctx->cached_sq_head != READ_ONCE(ring->r.head)) {
2211 /*
2212 * Ensure any loads from the SQEs are done at this point,
2213 * since once we write the new head, the application could
2214 * write new data to them.
2215 */
2216 smp_store_release(&ring->r.head, ctx->cached_sq_head);
Jens Axboe2b188cc2019-01-07 10:46:33 -07002217 }
2218}
2219
2220/*
Jens Axboe2b188cc2019-01-07 10:46:33 -07002221 * Fetch an sqe, if one is available. Note that s->sqe will point to memory
2222 * that is mapped by userspace. This means that care needs to be taken to
2223 * ensure that reads are stable, as we cannot rely on userspace always
2224 * being a good citizen. If members of the sqe are validated and then later
2225 * used, it's important that those reads are done through READ_ONCE() to
2226 * prevent a re-load down the line.
2227 */
2228static bool io_get_sqring(struct io_ring_ctx *ctx, struct sqe_submit *s)
2229{
2230 struct io_sq_ring *ring = ctx->sq_ring;
2231 unsigned head;
2232
2233 /*
2234 * The cached sq head (or cq tail) serves two purposes:
2235 *
2236 * 1) allows us to batch the cost of updating the user visible
2237 * head updates.
2238 * 2) allows the kernel side to track the head on its own, even
2239 * though the application is the one updating it.
2240 */
2241 head = ctx->cached_sq_head;
Stefan Bühlere523a292019-04-19 11:57:44 +02002242 /* make sure SQ entry isn't read before tail */
2243 if (head == smp_load_acquire(&ring->r.tail))
Jens Axboe2b188cc2019-01-07 10:46:33 -07002244 return false;
2245
2246 head = READ_ONCE(ring->array[head & ctx->sq_mask]);
2247 if (head < ctx->sq_entries) {
2248 s->index = head;
2249 s->sqe = &ctx->sq_sqes[head];
2250 ctx->cached_sq_head++;
2251 return true;
2252 }
2253
2254 /* drop invalid entries */
2255 ctx->cached_sq_head++;
2256 ring->dropped++;
Jens Axboe2b188cc2019-01-07 10:46:33 -07002257 return false;
2258}
2259
Jens Axboe6c271ce2019-01-10 11:22:30 -07002260static int io_submit_sqes(struct io_ring_ctx *ctx, struct sqe_submit *sqes,
2261 unsigned int nr, bool has_user, bool mm_fault)
2262{
2263 struct io_submit_state state, *statep = NULL;
Jens Axboe9e645e112019-05-10 16:07:28 -06002264 struct io_kiocb *link = NULL;
2265 bool prev_was_link = false;
2266 int i, submitted = 0;
Jens Axboe6c271ce2019-01-10 11:22:30 -07002267
2268 if (nr > IO_PLUG_THRESHOLD) {
2269 io_submit_state_start(&state, ctx, nr);
2270 statep = &state;
2271 }
2272
2273 for (i = 0; i < nr; i++) {
Jens Axboe9e645e112019-05-10 16:07:28 -06002274 /*
2275 * If previous wasn't linked and we have a linked command,
2276 * that's the end of the chain. Submit the previous link.
2277 */
2278 if (!prev_was_link && link) {
2279 io_queue_sqe(ctx, link, &link->submit);
2280 link = NULL;
2281 }
2282 prev_was_link = (sqes[i].sqe->flags & IOSQE_IO_LINK) != 0;
2283
Jens Axboe6c271ce2019-01-10 11:22:30 -07002284 if (unlikely(mm_fault)) {
Jens Axboe9e645e112019-05-10 16:07:28 -06002285 io_cqring_add_event(ctx, sqes[i].sqe->user_data,
2286 -EFAULT);
Jens Axboe6c271ce2019-01-10 11:22:30 -07002287 } else {
2288 sqes[i].has_user = has_user;
2289 sqes[i].needs_lock = true;
2290 sqes[i].needs_fixed_file = true;
Jens Axboe9e645e112019-05-10 16:07:28 -06002291 io_submit_sqe(ctx, &sqes[i], statep, &link);
Jens Axboe6c271ce2019-01-10 11:22:30 -07002292 submitted++;
Jens Axboe6c271ce2019-01-10 11:22:30 -07002293 }
Jens Axboe6c271ce2019-01-10 11:22:30 -07002294 }
2295
Jens Axboe9e645e112019-05-10 16:07:28 -06002296 if (link)
2297 io_queue_sqe(ctx, link, &link->submit);
Jens Axboe6c271ce2019-01-10 11:22:30 -07002298 if (statep)
2299 io_submit_state_end(&state);
2300
2301 return submitted;
2302}
2303
2304static int io_sq_thread(void *data)
2305{
2306 struct sqe_submit sqes[IO_IOPOLL_BATCH];
2307 struct io_ring_ctx *ctx = data;
2308 struct mm_struct *cur_mm = NULL;
2309 mm_segment_t old_fs;
2310 DEFINE_WAIT(wait);
2311 unsigned inflight;
2312 unsigned long timeout;
2313
Jackie Liua4c0b3d2019-07-08 13:41:12 +08002314 complete(&ctx->sqo_thread_started);
2315
Jens Axboe6c271ce2019-01-10 11:22:30 -07002316 old_fs = get_fs();
2317 set_fs(USER_DS);
2318
2319 timeout = inflight = 0;
Roman Penyaev2bbcd6d2019-05-16 10:53:57 +02002320 while (!kthread_should_park()) {
Jens Axboe6c271ce2019-01-10 11:22:30 -07002321 bool all_fixed, mm_fault = false;
2322 int i;
2323
2324 if (inflight) {
2325 unsigned nr_events = 0;
2326
2327 if (ctx->flags & IORING_SETUP_IOPOLL) {
Jens Axboe6c271ce2019-01-10 11:22:30 -07002328 io_iopoll_check(ctx, &nr_events, 0);
Jens Axboe6c271ce2019-01-10 11:22:30 -07002329 } else {
2330 /*
2331 * Normal IO, just pretend everything completed.
2332 * We don't have to poll completions for that.
2333 */
2334 nr_events = inflight;
2335 }
2336
2337 inflight -= nr_events;
2338 if (!inflight)
2339 timeout = jiffies + ctx->sq_thread_idle;
2340 }
2341
2342 if (!io_get_sqring(ctx, &sqes[0])) {
2343 /*
2344 * We're polling. If we're within the defined idle
2345 * period, then let us spin without work before going
2346 * to sleep.
2347 */
2348 if (inflight || !time_after(jiffies, timeout)) {
2349 cpu_relax();
2350 continue;
2351 }
2352
2353 /*
2354 * Drop cur_mm before scheduling, we can't hold it for
2355 * long periods (or over schedule()). Do this before
2356 * adding ourselves to the waitqueue, as the unuse/drop
2357 * may sleep.
2358 */
2359 if (cur_mm) {
2360 unuse_mm(cur_mm);
2361 mmput(cur_mm);
2362 cur_mm = NULL;
2363 }
2364
2365 prepare_to_wait(&ctx->sqo_wait, &wait,
2366 TASK_INTERRUPTIBLE);
2367
2368 /* Tell userspace we may need a wakeup call */
2369 ctx->sq_ring->flags |= IORING_SQ_NEED_WAKEUP;
Stefan Bühler0d7bae62019-04-19 11:57:45 +02002370 /* make sure to read SQ tail after writing flags */
2371 smp_mb();
Jens Axboe6c271ce2019-01-10 11:22:30 -07002372
2373 if (!io_get_sqring(ctx, &sqes[0])) {
Roman Penyaev2bbcd6d2019-05-16 10:53:57 +02002374 if (kthread_should_park()) {
Jens Axboe6c271ce2019-01-10 11:22:30 -07002375 finish_wait(&ctx->sqo_wait, &wait);
2376 break;
2377 }
2378 if (signal_pending(current))
2379 flush_signals(current);
2380 schedule();
2381 finish_wait(&ctx->sqo_wait, &wait);
2382
2383 ctx->sq_ring->flags &= ~IORING_SQ_NEED_WAKEUP;
Jens Axboe6c271ce2019-01-10 11:22:30 -07002384 continue;
2385 }
2386 finish_wait(&ctx->sqo_wait, &wait);
2387
2388 ctx->sq_ring->flags &= ~IORING_SQ_NEED_WAKEUP;
Jens Axboe6c271ce2019-01-10 11:22:30 -07002389 }
2390
2391 i = 0;
2392 all_fixed = true;
2393 do {
2394 if (all_fixed && io_sqe_needs_user(sqes[i].sqe))
2395 all_fixed = false;
2396
2397 i++;
2398 if (i == ARRAY_SIZE(sqes))
2399 break;
2400 } while (io_get_sqring(ctx, &sqes[i]));
2401
2402 /* Unless all new commands are FIXED regions, grab mm */
2403 if (!all_fixed && !cur_mm) {
2404 mm_fault = !mmget_not_zero(ctx->sqo_mm);
2405 if (!mm_fault) {
2406 use_mm(ctx->sqo_mm);
2407 cur_mm = ctx->sqo_mm;
2408 }
2409 }
2410
2411 inflight += io_submit_sqes(ctx, sqes, i, cur_mm != NULL,
2412 mm_fault);
2413
2414 /* Commit SQ ring head once we've consumed all SQEs */
2415 io_commit_sqring(ctx);
2416 }
2417
2418 set_fs(old_fs);
2419 if (cur_mm) {
2420 unuse_mm(cur_mm);
2421 mmput(cur_mm);
2422 }
Jens Axboe06058632019-04-13 09:26:03 -06002423
Roman Penyaev2bbcd6d2019-05-16 10:53:57 +02002424 kthread_parkme();
Jens Axboe06058632019-04-13 09:26:03 -06002425
Jens Axboe6c271ce2019-01-10 11:22:30 -07002426 return 0;
2427}
2428
Jens Axboe2b188cc2019-01-07 10:46:33 -07002429static int io_ring_submit(struct io_ring_ctx *ctx, unsigned int to_submit)
2430{
Jens Axboe9a56a232019-01-09 09:06:50 -07002431 struct io_submit_state state, *statep = NULL;
Jens Axboe9e645e112019-05-10 16:07:28 -06002432 struct io_kiocb *link = NULL;
2433 bool prev_was_link = false;
Jens Axboe5c8b0b52019-04-30 10:16:07 -06002434 int i, submit = 0;
Jens Axboe2b188cc2019-01-07 10:46:33 -07002435
Jens Axboe9a56a232019-01-09 09:06:50 -07002436 if (to_submit > IO_PLUG_THRESHOLD) {
2437 io_submit_state_start(&state, ctx, to_submit);
2438 statep = &state;
2439 }
Jens Axboe2b188cc2019-01-07 10:46:33 -07002440
2441 for (i = 0; i < to_submit; i++) {
2442 struct sqe_submit s;
2443
2444 if (!io_get_sqring(ctx, &s))
2445 break;
2446
Jens Axboe9e645e112019-05-10 16:07:28 -06002447 /*
2448 * If previous wasn't linked and we have a linked command,
2449 * that's the end of the chain. Submit the previous link.
2450 */
2451 if (!prev_was_link && link) {
2452 io_queue_sqe(ctx, link, &link->submit);
2453 link = NULL;
2454 }
2455 prev_was_link = (s.sqe->flags & IOSQE_IO_LINK) != 0;
2456
Jens Axboe2b188cc2019-01-07 10:46:33 -07002457 s.has_user = true;
Jens Axboedef596e2019-01-09 08:59:42 -07002458 s.needs_lock = false;
Jens Axboe6c271ce2019-01-10 11:22:30 -07002459 s.needs_fixed_file = false;
Jens Axboe5c8b0b52019-04-30 10:16:07 -06002460 submit++;
Jens Axboe9e645e112019-05-10 16:07:28 -06002461 io_submit_sqe(ctx, &s, statep, &link);
Jens Axboe2b188cc2019-01-07 10:46:33 -07002462 }
2463 io_commit_sqring(ctx);
2464
Jens Axboe9e645e112019-05-10 16:07:28 -06002465 if (link)
2466 io_queue_sqe(ctx, link, &link->submit);
Jens Axboe9a56a232019-01-09 09:06:50 -07002467 if (statep)
2468 io_submit_state_end(statep);
Jens Axboe2b188cc2019-01-07 10:46:33 -07002469
Jens Axboe5c8b0b52019-04-30 10:16:07 -06002470 return submit;
Jens Axboe2b188cc2019-01-07 10:46:33 -07002471}
2472
Jens Axboe2b188cc2019-01-07 10:46:33 -07002473/*
2474 * Wait until events become available, if we don't already have some. The
2475 * application must reap them itself, as they reside on the shared cq ring.
2476 */
2477static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events,
2478 const sigset_t __user *sig, size_t sigsz)
2479{
2480 struct io_cq_ring *ring = ctx->cq_ring;
Jens Axboe2b188cc2019-01-07 10:46:33 -07002481 int ret;
2482
Jens Axboe2b188cc2019-01-07 10:46:33 -07002483 if (io_cqring_events(ring) >= min_events)
2484 return 0;
2485
2486 if (sig) {
Arnd Bergmann9e75ad52019-03-25 15:34:53 +01002487#ifdef CONFIG_COMPAT
2488 if (in_compat_syscall())
2489 ret = set_compat_user_sigmask((const compat_sigset_t __user *)sig,
Oleg Nesterovb7724342019-07-16 16:29:53 -07002490 sigsz);
Arnd Bergmann9e75ad52019-03-25 15:34:53 +01002491 else
2492#endif
Oleg Nesterovb7724342019-07-16 16:29:53 -07002493 ret = set_user_sigmask(sig, sigsz);
Arnd Bergmann9e75ad52019-03-25 15:34:53 +01002494
Jens Axboe2b188cc2019-01-07 10:46:33 -07002495 if (ret)
2496 return ret;
2497 }
2498
Jackie Liufdb288a2019-05-16 11:46:31 +08002499 ret = wait_event_interruptible(ctx->wait, io_cqring_events(ring) >= min_events);
Oleg Nesterovb7724342019-07-16 16:29:53 -07002500 restore_saved_sigmask_unless(ret == -ERESTARTSYS);
Oleg Nesterov97abc882019-06-28 12:06:50 -07002501 if (ret == -ERESTARTSYS)
2502 ret = -EINTR;
Jens Axboe2b188cc2019-01-07 10:46:33 -07002503
2504 return READ_ONCE(ring->r.head) == READ_ONCE(ring->r.tail) ? ret : 0;
2505}
2506
Jens Axboe6b063142019-01-10 22:13:58 -07002507static void __io_sqe_files_unregister(struct io_ring_ctx *ctx)
2508{
2509#if defined(CONFIG_UNIX)
2510 if (ctx->ring_sock) {
2511 struct sock *sock = ctx->ring_sock->sk;
2512 struct sk_buff *skb;
2513
2514 while ((skb = skb_dequeue(&sock->sk_receive_queue)) != NULL)
2515 kfree_skb(skb);
2516 }
2517#else
2518 int i;
2519
2520 for (i = 0; i < ctx->nr_user_files; i++)
2521 fput(ctx->user_files[i]);
2522#endif
2523}
2524
2525static int io_sqe_files_unregister(struct io_ring_ctx *ctx)
2526{
2527 if (!ctx->user_files)
2528 return -ENXIO;
2529
2530 __io_sqe_files_unregister(ctx);
2531 kfree(ctx->user_files);
2532 ctx->user_files = NULL;
2533 ctx->nr_user_files = 0;
2534 return 0;
2535}
2536
Jens Axboe6c271ce2019-01-10 11:22:30 -07002537static void io_sq_thread_stop(struct io_ring_ctx *ctx)
2538{
2539 if (ctx->sqo_thread) {
Jackie Liua4c0b3d2019-07-08 13:41:12 +08002540 wait_for_completion(&ctx->sqo_thread_started);
Roman Penyaev2bbcd6d2019-05-16 10:53:57 +02002541 /*
2542 * The park is a bit of a work-around, without it we get
2543 * warning spews on shutdown with SQPOLL set and affinity
2544 * set to a single CPU.
2545 */
Jens Axboe06058632019-04-13 09:26:03 -06002546 kthread_park(ctx->sqo_thread);
Jens Axboe6c271ce2019-01-10 11:22:30 -07002547 kthread_stop(ctx->sqo_thread);
2548 ctx->sqo_thread = NULL;
2549 }
2550}
2551
Jens Axboe6b063142019-01-10 22:13:58 -07002552static void io_finish_async(struct io_ring_ctx *ctx)
2553{
Jens Axboe6c271ce2019-01-10 11:22:30 -07002554 io_sq_thread_stop(ctx);
2555
Jens Axboe6b063142019-01-10 22:13:58 -07002556 if (ctx->sqo_wq) {
2557 destroy_workqueue(ctx->sqo_wq);
2558 ctx->sqo_wq = NULL;
2559 }
2560}
2561
2562#if defined(CONFIG_UNIX)
2563static void io_destruct_skb(struct sk_buff *skb)
2564{
2565 struct io_ring_ctx *ctx = skb->sk->sk_user_data;
2566
2567 io_finish_async(ctx);
2568 unix_destruct_scm(skb);
2569}
2570
2571/*
2572 * Ensure the UNIX gc is aware of our file set, so we are certain that
2573 * the io_uring can be safely unregistered on process exit, even if we have
2574 * loops in the file referencing.
2575 */
2576static int __io_sqe_files_scm(struct io_ring_ctx *ctx, int nr, int offset)
2577{
2578 struct sock *sk = ctx->ring_sock->sk;
2579 struct scm_fp_list *fpl;
2580 struct sk_buff *skb;
2581 int i;
2582
2583 if (!capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN)) {
2584 unsigned long inflight = ctx->user->unix_inflight + nr;
2585
2586 if (inflight > task_rlimit(current, RLIMIT_NOFILE))
2587 return -EMFILE;
2588 }
2589
2590 fpl = kzalloc(sizeof(*fpl), GFP_KERNEL);
2591 if (!fpl)
2592 return -ENOMEM;
2593
2594 skb = alloc_skb(0, GFP_KERNEL);
2595 if (!skb) {
2596 kfree(fpl);
2597 return -ENOMEM;
2598 }
2599
2600 skb->sk = sk;
2601 skb->destructor = io_destruct_skb;
2602
2603 fpl->user = get_uid(ctx->user);
2604 for (i = 0; i < nr; i++) {
2605 fpl->fp[i] = get_file(ctx->user_files[i + offset]);
2606 unix_inflight(fpl->user, fpl->fp[i]);
2607 }
2608
2609 fpl->max = fpl->count = nr;
2610 UNIXCB(skb).fp = fpl;
2611 refcount_add(skb->truesize, &sk->sk_wmem_alloc);
2612 skb_queue_head(&sk->sk_receive_queue, skb);
2613
2614 for (i = 0; i < nr; i++)
2615 fput(fpl->fp[i]);
2616
2617 return 0;
2618}
2619
2620/*
2621 * If UNIX sockets are enabled, fd passing can cause a reference cycle which
2622 * causes regular reference counting to break down. We rely on the UNIX
2623 * garbage collection to take care of this problem for us.
2624 */
2625static int io_sqe_files_scm(struct io_ring_ctx *ctx)
2626{
2627 unsigned left, total;
2628 int ret = 0;
2629
2630 total = 0;
2631 left = ctx->nr_user_files;
2632 while (left) {
2633 unsigned this_files = min_t(unsigned, left, SCM_MAX_FD);
Jens Axboe6b063142019-01-10 22:13:58 -07002634
2635 ret = __io_sqe_files_scm(ctx, this_files, total);
2636 if (ret)
2637 break;
2638 left -= this_files;
2639 total += this_files;
2640 }
2641
2642 if (!ret)
2643 return 0;
2644
2645 while (total < ctx->nr_user_files) {
2646 fput(ctx->user_files[total]);
2647 total++;
2648 }
2649
2650 return ret;
2651}
2652#else
2653static int io_sqe_files_scm(struct io_ring_ctx *ctx)
2654{
2655 return 0;
2656}
2657#endif
2658
2659static int io_sqe_files_register(struct io_ring_ctx *ctx, void __user *arg,
2660 unsigned nr_args)
2661{
2662 __s32 __user *fds = (__s32 __user *) arg;
2663 int fd, ret = 0;
2664 unsigned i;
2665
2666 if (ctx->user_files)
2667 return -EBUSY;
2668 if (!nr_args)
2669 return -EINVAL;
2670 if (nr_args > IORING_MAX_FIXED_FILES)
2671 return -EMFILE;
2672
2673 ctx->user_files = kcalloc(nr_args, sizeof(struct file *), GFP_KERNEL);
2674 if (!ctx->user_files)
2675 return -ENOMEM;
2676
2677 for (i = 0; i < nr_args; i++) {
2678 ret = -EFAULT;
2679 if (copy_from_user(&fd, &fds[i], sizeof(fd)))
2680 break;
2681
2682 ctx->user_files[i] = fget(fd);
2683
2684 ret = -EBADF;
2685 if (!ctx->user_files[i])
2686 break;
2687 /*
2688 * Don't allow io_uring instances to be registered. If UNIX
2689 * isn't enabled, then this causes a reference cycle and this
2690 * instance can never get freed. If UNIX is enabled we'll
2691 * handle it just fine, but there's still no point in allowing
2692 * a ring fd as it doesn't support regular read/write anyway.
2693 */
2694 if (ctx->user_files[i]->f_op == &io_uring_fops) {
2695 fput(ctx->user_files[i]);
2696 break;
2697 }
2698 ctx->nr_user_files++;
2699 ret = 0;
2700 }
2701
2702 if (ret) {
2703 for (i = 0; i < ctx->nr_user_files; i++)
2704 fput(ctx->user_files[i]);
2705
2706 kfree(ctx->user_files);
Jens Axboe25adf502019-04-03 09:52:40 -06002707 ctx->user_files = NULL;
Jens Axboe6b063142019-01-10 22:13:58 -07002708 ctx->nr_user_files = 0;
2709 return ret;
2710 }
2711
2712 ret = io_sqe_files_scm(ctx);
2713 if (ret)
2714 io_sqe_files_unregister(ctx);
2715
2716 return ret;
2717}
2718
Jens Axboe6c271ce2019-01-10 11:22:30 -07002719static int io_sq_offload_start(struct io_ring_ctx *ctx,
2720 struct io_uring_params *p)
Jens Axboe2b188cc2019-01-07 10:46:33 -07002721{
2722 int ret;
2723
Jens Axboe6c271ce2019-01-10 11:22:30 -07002724 init_waitqueue_head(&ctx->sqo_wait);
Jens Axboe2b188cc2019-01-07 10:46:33 -07002725 mmgrab(current->mm);
2726 ctx->sqo_mm = current->mm;
2727
Jens Axboe6c271ce2019-01-10 11:22:30 -07002728 if (ctx->flags & IORING_SETUP_SQPOLL) {
Jens Axboe3ec482d2019-04-08 10:51:01 -06002729 ret = -EPERM;
2730 if (!capable(CAP_SYS_ADMIN))
2731 goto err;
2732
Jens Axboe917257d2019-04-13 09:28:55 -06002733 ctx->sq_thread_idle = msecs_to_jiffies(p->sq_thread_idle);
2734 if (!ctx->sq_thread_idle)
2735 ctx->sq_thread_idle = HZ;
2736
Jens Axboe6c271ce2019-01-10 11:22:30 -07002737 if (p->flags & IORING_SETUP_SQ_AFF) {
Jens Axboe44a9bd12019-05-14 20:00:30 -06002738 int cpu = p->sq_thread_cpu;
Jens Axboe6c271ce2019-01-10 11:22:30 -07002739
Jens Axboe917257d2019-04-13 09:28:55 -06002740 ret = -EINVAL;
Jens Axboe44a9bd12019-05-14 20:00:30 -06002741 if (cpu >= nr_cpu_ids)
2742 goto err;
Shenghui Wang7889f442019-05-07 16:03:19 +08002743 if (!cpu_online(cpu))
Jens Axboe917257d2019-04-13 09:28:55 -06002744 goto err;
2745
Jens Axboe6c271ce2019-01-10 11:22:30 -07002746 ctx->sqo_thread = kthread_create_on_cpu(io_sq_thread,
2747 ctx, cpu,
2748 "io_uring-sq");
2749 } else {
2750 ctx->sqo_thread = kthread_create(io_sq_thread, ctx,
2751 "io_uring-sq");
2752 }
2753 if (IS_ERR(ctx->sqo_thread)) {
2754 ret = PTR_ERR(ctx->sqo_thread);
2755 ctx->sqo_thread = NULL;
2756 goto err;
2757 }
2758 wake_up_process(ctx->sqo_thread);
2759 } else if (p->flags & IORING_SETUP_SQ_AFF) {
2760 /* Can't have SQ_AFF without SQPOLL */
2761 ret = -EINVAL;
2762 goto err;
2763 }
2764
Jens Axboe2b188cc2019-01-07 10:46:33 -07002765 /* Do QD, or 2 * CPUS, whatever is smallest */
2766 ctx->sqo_wq = alloc_workqueue("io_ring-wq", WQ_UNBOUND | WQ_FREEZABLE,
2767 min(ctx->sq_entries - 1, 2 * num_online_cpus()));
2768 if (!ctx->sqo_wq) {
2769 ret = -ENOMEM;
2770 goto err;
2771 }
2772
2773 return 0;
2774err:
Jens Axboe6c271ce2019-01-10 11:22:30 -07002775 io_sq_thread_stop(ctx);
Jens Axboe2b188cc2019-01-07 10:46:33 -07002776 mmdrop(ctx->sqo_mm);
2777 ctx->sqo_mm = NULL;
2778 return ret;
2779}
2780
2781static void io_unaccount_mem(struct user_struct *user, unsigned long nr_pages)
2782{
2783 atomic_long_sub(nr_pages, &user->locked_vm);
2784}
2785
2786static int io_account_mem(struct user_struct *user, unsigned long nr_pages)
2787{
2788 unsigned long page_limit, cur_pages, new_pages;
2789
2790 /* Don't allow more pages than we can safely lock */
2791 page_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT;
2792
2793 do {
2794 cur_pages = atomic_long_read(&user->locked_vm);
2795 new_pages = cur_pages + nr_pages;
2796 if (new_pages > page_limit)
2797 return -ENOMEM;
2798 } while (atomic_long_cmpxchg(&user->locked_vm, cur_pages,
2799 new_pages) != cur_pages);
2800
2801 return 0;
2802}
2803
2804static void io_mem_free(void *ptr)
2805{
Mark Rutland52e04ef2019-04-30 17:30:21 +01002806 struct page *page;
Jens Axboe2b188cc2019-01-07 10:46:33 -07002807
Mark Rutland52e04ef2019-04-30 17:30:21 +01002808 if (!ptr)
2809 return;
2810
2811 page = virt_to_head_page(ptr);
Jens Axboe2b188cc2019-01-07 10:46:33 -07002812 if (put_page_testzero(page))
2813 free_compound_page(page);
2814}
2815
2816static void *io_mem_alloc(size_t size)
2817{
2818 gfp_t gfp_flags = GFP_KERNEL | __GFP_ZERO | __GFP_NOWARN | __GFP_COMP |
2819 __GFP_NORETRY;
2820
2821 return (void *) __get_free_pages(gfp_flags, get_order(size));
2822}
2823
2824static unsigned long ring_pages(unsigned sq_entries, unsigned cq_entries)
2825{
2826 struct io_sq_ring *sq_ring;
2827 struct io_cq_ring *cq_ring;
2828 size_t bytes;
2829
2830 bytes = struct_size(sq_ring, array, sq_entries);
2831 bytes += array_size(sizeof(struct io_uring_sqe), sq_entries);
2832 bytes += struct_size(cq_ring, cqes, cq_entries);
2833
2834 return (bytes + PAGE_SIZE - 1) / PAGE_SIZE;
2835}
2836
Jens Axboeedafcce2019-01-09 09:16:05 -07002837static int io_sqe_buffer_unregister(struct io_ring_ctx *ctx)
2838{
2839 int i, j;
2840
2841 if (!ctx->user_bufs)
2842 return -ENXIO;
2843
2844 for (i = 0; i < ctx->nr_user_bufs; i++) {
2845 struct io_mapped_ubuf *imu = &ctx->user_bufs[i];
2846
2847 for (j = 0; j < imu->nr_bvecs; j++)
2848 put_page(imu->bvec[j].bv_page);
2849
2850 if (ctx->account_mem)
2851 io_unaccount_mem(ctx->user, imu->nr_bvecs);
Mark Rutlandd4ef6472019-05-01 16:59:16 +01002852 kvfree(imu->bvec);
Jens Axboeedafcce2019-01-09 09:16:05 -07002853 imu->nr_bvecs = 0;
2854 }
2855
2856 kfree(ctx->user_bufs);
2857 ctx->user_bufs = NULL;
2858 ctx->nr_user_bufs = 0;
2859 return 0;
2860}
2861
2862static int io_copy_iov(struct io_ring_ctx *ctx, struct iovec *dst,
2863 void __user *arg, unsigned index)
2864{
2865 struct iovec __user *src;
2866
2867#ifdef CONFIG_COMPAT
2868 if (ctx->compat) {
2869 struct compat_iovec __user *ciovs;
2870 struct compat_iovec ciov;
2871
2872 ciovs = (struct compat_iovec __user *) arg;
2873 if (copy_from_user(&ciov, &ciovs[index], sizeof(ciov)))
2874 return -EFAULT;
2875
2876 dst->iov_base = (void __user *) (unsigned long) ciov.iov_base;
2877 dst->iov_len = ciov.iov_len;
2878 return 0;
2879 }
2880#endif
2881 src = (struct iovec __user *) arg;
2882 if (copy_from_user(dst, &src[index], sizeof(*dst)))
2883 return -EFAULT;
2884 return 0;
2885}
2886
2887static int io_sqe_buffer_register(struct io_ring_ctx *ctx, void __user *arg,
2888 unsigned nr_args)
2889{
2890 struct vm_area_struct **vmas = NULL;
2891 struct page **pages = NULL;
2892 int i, j, got_pages = 0;
2893 int ret = -EINVAL;
2894
2895 if (ctx->user_bufs)
2896 return -EBUSY;
2897 if (!nr_args || nr_args > UIO_MAXIOV)
2898 return -EINVAL;
2899
2900 ctx->user_bufs = kcalloc(nr_args, sizeof(struct io_mapped_ubuf),
2901 GFP_KERNEL);
2902 if (!ctx->user_bufs)
2903 return -ENOMEM;
2904
2905 for (i = 0; i < nr_args; i++) {
2906 struct io_mapped_ubuf *imu = &ctx->user_bufs[i];
2907 unsigned long off, start, end, ubuf;
2908 int pret, nr_pages;
2909 struct iovec iov;
2910 size_t size;
2911
2912 ret = io_copy_iov(ctx, &iov, arg, i);
2913 if (ret)
Pavel Begunkova2786822019-05-26 12:35:47 +03002914 goto err;
Jens Axboeedafcce2019-01-09 09:16:05 -07002915
2916 /*
2917 * Don't impose further limits on the size and buffer
2918 * constraints here, we'll -EINVAL later when IO is
2919 * submitted if they are wrong.
2920 */
2921 ret = -EFAULT;
2922 if (!iov.iov_base || !iov.iov_len)
2923 goto err;
2924
2925 /* arbitrary limit, but we need something */
2926 if (iov.iov_len > SZ_1G)
2927 goto err;
2928
2929 ubuf = (unsigned long) iov.iov_base;
2930 end = (ubuf + iov.iov_len + PAGE_SIZE - 1) >> PAGE_SHIFT;
2931 start = ubuf >> PAGE_SHIFT;
2932 nr_pages = end - start;
2933
2934 if (ctx->account_mem) {
2935 ret = io_account_mem(ctx->user, nr_pages);
2936 if (ret)
2937 goto err;
2938 }
2939
2940 ret = 0;
2941 if (!pages || nr_pages > got_pages) {
2942 kfree(vmas);
2943 kfree(pages);
Mark Rutlandd4ef6472019-05-01 16:59:16 +01002944 pages = kvmalloc_array(nr_pages, sizeof(struct page *),
Jens Axboeedafcce2019-01-09 09:16:05 -07002945 GFP_KERNEL);
Mark Rutlandd4ef6472019-05-01 16:59:16 +01002946 vmas = kvmalloc_array(nr_pages,
Jens Axboeedafcce2019-01-09 09:16:05 -07002947 sizeof(struct vm_area_struct *),
2948 GFP_KERNEL);
2949 if (!pages || !vmas) {
2950 ret = -ENOMEM;
2951 if (ctx->account_mem)
2952 io_unaccount_mem(ctx->user, nr_pages);
2953 goto err;
2954 }
2955 got_pages = nr_pages;
2956 }
2957
Mark Rutlandd4ef6472019-05-01 16:59:16 +01002958 imu->bvec = kvmalloc_array(nr_pages, sizeof(struct bio_vec),
Jens Axboeedafcce2019-01-09 09:16:05 -07002959 GFP_KERNEL);
2960 ret = -ENOMEM;
2961 if (!imu->bvec) {
2962 if (ctx->account_mem)
2963 io_unaccount_mem(ctx->user, nr_pages);
2964 goto err;
2965 }
2966
2967 ret = 0;
2968 down_read(&current->mm->mmap_sem);
Ira Weiny932f4a62019-05-13 17:17:03 -07002969 pret = get_user_pages(ubuf, nr_pages,
2970 FOLL_WRITE | FOLL_LONGTERM,
2971 pages, vmas);
Jens Axboeedafcce2019-01-09 09:16:05 -07002972 if (pret == nr_pages) {
2973 /* don't support file backed memory */
2974 for (j = 0; j < nr_pages; j++) {
2975 struct vm_area_struct *vma = vmas[j];
2976
2977 if (vma->vm_file &&
2978 !is_file_hugepages(vma->vm_file)) {
2979 ret = -EOPNOTSUPP;
2980 break;
2981 }
2982 }
2983 } else {
2984 ret = pret < 0 ? pret : -EFAULT;
2985 }
2986 up_read(&current->mm->mmap_sem);
2987 if (ret) {
2988 /*
2989 * if we did partial map, or found file backed vmas,
2990 * release any pages we did get
2991 */
2992 if (pret > 0) {
2993 for (j = 0; j < pret; j++)
2994 put_page(pages[j]);
2995 }
2996 if (ctx->account_mem)
2997 io_unaccount_mem(ctx->user, nr_pages);
Mark Rutlandd4ef6472019-05-01 16:59:16 +01002998 kvfree(imu->bvec);
Jens Axboeedafcce2019-01-09 09:16:05 -07002999 goto err;
3000 }
3001
3002 off = ubuf & ~PAGE_MASK;
3003 size = iov.iov_len;
3004 for (j = 0; j < nr_pages; j++) {
3005 size_t vec_len;
3006
3007 vec_len = min_t(size_t, size, PAGE_SIZE - off);
3008 imu->bvec[j].bv_page = pages[j];
3009 imu->bvec[j].bv_len = vec_len;
3010 imu->bvec[j].bv_offset = off;
3011 off = 0;
3012 size -= vec_len;
3013 }
3014 /* store original address for later verification */
3015 imu->ubuf = ubuf;
3016 imu->len = iov.iov_len;
3017 imu->nr_bvecs = nr_pages;
3018
3019 ctx->nr_user_bufs++;
3020 }
Mark Rutlandd4ef6472019-05-01 16:59:16 +01003021 kvfree(pages);
3022 kvfree(vmas);
Jens Axboeedafcce2019-01-09 09:16:05 -07003023 return 0;
3024err:
Mark Rutlandd4ef6472019-05-01 16:59:16 +01003025 kvfree(pages);
3026 kvfree(vmas);
Jens Axboeedafcce2019-01-09 09:16:05 -07003027 io_sqe_buffer_unregister(ctx);
3028 return ret;
3029}
3030
Jens Axboe9b402842019-04-11 11:45:41 -06003031static int io_eventfd_register(struct io_ring_ctx *ctx, void __user *arg)
3032{
3033 __s32 __user *fds = arg;
3034 int fd;
3035
3036 if (ctx->cq_ev_fd)
3037 return -EBUSY;
3038
3039 if (copy_from_user(&fd, fds, sizeof(*fds)))
3040 return -EFAULT;
3041
3042 ctx->cq_ev_fd = eventfd_ctx_fdget(fd);
3043 if (IS_ERR(ctx->cq_ev_fd)) {
3044 int ret = PTR_ERR(ctx->cq_ev_fd);
3045 ctx->cq_ev_fd = NULL;
3046 return ret;
3047 }
3048
3049 return 0;
3050}
3051
3052static int io_eventfd_unregister(struct io_ring_ctx *ctx)
3053{
3054 if (ctx->cq_ev_fd) {
3055 eventfd_ctx_put(ctx->cq_ev_fd);
3056 ctx->cq_ev_fd = NULL;
3057 return 0;
3058 }
3059
3060 return -ENXIO;
3061}
3062
Jens Axboe2b188cc2019-01-07 10:46:33 -07003063static void io_ring_ctx_free(struct io_ring_ctx *ctx)
3064{
Jens Axboe6b063142019-01-10 22:13:58 -07003065 io_finish_async(ctx);
Jens Axboe2b188cc2019-01-07 10:46:33 -07003066 if (ctx->sqo_mm)
3067 mmdrop(ctx->sqo_mm);
Jens Axboedef596e2019-01-09 08:59:42 -07003068
3069 io_iopoll_reap_events(ctx);
Jens Axboeedafcce2019-01-09 09:16:05 -07003070 io_sqe_buffer_unregister(ctx);
Jens Axboe6b063142019-01-10 22:13:58 -07003071 io_sqe_files_unregister(ctx);
Jens Axboe9b402842019-04-11 11:45:41 -06003072 io_eventfd_unregister(ctx);
Jens Axboedef596e2019-01-09 08:59:42 -07003073
Jens Axboe2b188cc2019-01-07 10:46:33 -07003074#if defined(CONFIG_UNIX)
Eric Biggers355e8d22019-06-12 14:58:43 -07003075 if (ctx->ring_sock) {
3076 ctx->ring_sock->file = NULL; /* so that iput() is called */
Jens Axboe2b188cc2019-01-07 10:46:33 -07003077 sock_release(ctx->ring_sock);
Eric Biggers355e8d22019-06-12 14:58:43 -07003078 }
Jens Axboe2b188cc2019-01-07 10:46:33 -07003079#endif
3080
3081 io_mem_free(ctx->sq_ring);
3082 io_mem_free(ctx->sq_sqes);
3083 io_mem_free(ctx->cq_ring);
3084
3085 percpu_ref_exit(&ctx->refs);
3086 if (ctx->account_mem)
3087 io_unaccount_mem(ctx->user,
3088 ring_pages(ctx->sq_entries, ctx->cq_entries));
3089 free_uid(ctx->user);
3090 kfree(ctx);
3091}
3092
3093static __poll_t io_uring_poll(struct file *file, poll_table *wait)
3094{
3095 struct io_ring_ctx *ctx = file->private_data;
3096 __poll_t mask = 0;
3097
3098 poll_wait(file, &ctx->cq_wait, wait);
Stefan Bühler4f7067c2019-04-24 23:54:17 +02003099 /*
3100 * synchronizes with barrier from wq_has_sleeper call in
3101 * io_commit_cqring
3102 */
Jens Axboe2b188cc2019-01-07 10:46:33 -07003103 smp_rmb();
Stefan Bühlerfb775fa2019-04-19 11:57:46 +02003104 if (READ_ONCE(ctx->sq_ring->r.tail) - ctx->cached_sq_head !=
3105 ctx->sq_ring->ring_entries)
Jens Axboe2b188cc2019-01-07 10:46:33 -07003106 mask |= EPOLLOUT | EPOLLWRNORM;
3107 if (READ_ONCE(ctx->cq_ring->r.head) != ctx->cached_cq_tail)
3108 mask |= EPOLLIN | EPOLLRDNORM;
3109
3110 return mask;
3111}
3112
3113static int io_uring_fasync(int fd, struct file *file, int on)
3114{
3115 struct io_ring_ctx *ctx = file->private_data;
3116
3117 return fasync_helper(fd, file, on, &ctx->cq_fasync);
3118}
3119
3120static void io_ring_ctx_wait_and_kill(struct io_ring_ctx *ctx)
3121{
3122 mutex_lock(&ctx->uring_lock);
3123 percpu_ref_kill(&ctx->refs);
3124 mutex_unlock(&ctx->uring_lock);
3125
Jens Axboe221c5eb2019-01-17 09:41:58 -07003126 io_poll_remove_all(ctx);
Jens Axboedef596e2019-01-09 08:59:42 -07003127 io_iopoll_reap_events(ctx);
Jens Axboe2b188cc2019-01-07 10:46:33 -07003128 wait_for_completion(&ctx->ctx_done);
3129 io_ring_ctx_free(ctx);
3130}
3131
3132static int io_uring_release(struct inode *inode, struct file *file)
3133{
3134 struct io_ring_ctx *ctx = file->private_data;
3135
3136 file->private_data = NULL;
3137 io_ring_ctx_wait_and_kill(ctx);
3138 return 0;
3139}
3140
3141static int io_uring_mmap(struct file *file, struct vm_area_struct *vma)
3142{
3143 loff_t offset = (loff_t) vma->vm_pgoff << PAGE_SHIFT;
3144 unsigned long sz = vma->vm_end - vma->vm_start;
3145 struct io_ring_ctx *ctx = file->private_data;
3146 unsigned long pfn;
3147 struct page *page;
3148 void *ptr;
3149
3150 switch (offset) {
3151 case IORING_OFF_SQ_RING:
3152 ptr = ctx->sq_ring;
3153 break;
3154 case IORING_OFF_SQES:
3155 ptr = ctx->sq_sqes;
3156 break;
3157 case IORING_OFF_CQ_RING:
3158 ptr = ctx->cq_ring;
3159 break;
3160 default:
3161 return -EINVAL;
3162 }
3163
3164 page = virt_to_head_page(ptr);
3165 if (sz > (PAGE_SIZE << compound_order(page)))
3166 return -EINVAL;
3167
3168 pfn = virt_to_phys(ptr) >> PAGE_SHIFT;
3169 return remap_pfn_range(vma, vma->vm_start, pfn, sz, vma->vm_page_prot);
3170}
3171
3172SYSCALL_DEFINE6(io_uring_enter, unsigned int, fd, u32, to_submit,
3173 u32, min_complete, u32, flags, const sigset_t __user *, sig,
3174 size_t, sigsz)
3175{
3176 struct io_ring_ctx *ctx;
3177 long ret = -EBADF;
3178 int submitted = 0;
3179 struct fd f;
3180
Jens Axboe6c271ce2019-01-10 11:22:30 -07003181 if (flags & ~(IORING_ENTER_GETEVENTS | IORING_ENTER_SQ_WAKEUP))
Jens Axboe2b188cc2019-01-07 10:46:33 -07003182 return -EINVAL;
3183
3184 f = fdget(fd);
3185 if (!f.file)
3186 return -EBADF;
3187
3188 ret = -EOPNOTSUPP;
3189 if (f.file->f_op != &io_uring_fops)
3190 goto out_fput;
3191
3192 ret = -ENXIO;
3193 ctx = f.file->private_data;
3194 if (!percpu_ref_tryget(&ctx->refs))
3195 goto out_fput;
3196
Jens Axboe6c271ce2019-01-10 11:22:30 -07003197 /*
3198 * For SQ polling, the thread will do all submissions and completions.
3199 * Just return the requested submit count, and wake the thread if
3200 * we were asked to.
3201 */
3202 if (ctx->flags & IORING_SETUP_SQPOLL) {
3203 if (flags & IORING_ENTER_SQ_WAKEUP)
3204 wake_up(&ctx->sqo_wait);
3205 submitted = to_submit;
3206 goto out_ctx;
3207 }
3208
Jens Axboe2b188cc2019-01-07 10:46:33 -07003209 ret = 0;
3210 if (to_submit) {
3211 to_submit = min(to_submit, ctx->sq_entries);
3212
3213 mutex_lock(&ctx->uring_lock);
3214 submitted = io_ring_submit(ctx, to_submit);
3215 mutex_unlock(&ctx->uring_lock);
Jens Axboe2b188cc2019-01-07 10:46:33 -07003216 }
3217 if (flags & IORING_ENTER_GETEVENTS) {
Jens Axboedef596e2019-01-09 08:59:42 -07003218 unsigned nr_events = 0;
3219
Jens Axboe2b188cc2019-01-07 10:46:33 -07003220 min_complete = min(min_complete, ctx->cq_entries);
3221
Jens Axboedef596e2019-01-09 08:59:42 -07003222 if (ctx->flags & IORING_SETUP_IOPOLL) {
Jens Axboedef596e2019-01-09 08:59:42 -07003223 ret = io_iopoll_check(ctx, &nr_events, min_complete);
Jens Axboedef596e2019-01-09 08:59:42 -07003224 } else {
3225 ret = io_cqring_wait(ctx, min_complete, sig, sigsz);
3226 }
Jens Axboe2b188cc2019-01-07 10:46:33 -07003227 }
3228
3229out_ctx:
3230 io_ring_drop_ctx_refs(ctx, 1);
3231out_fput:
3232 fdput(f);
3233 return submitted ? submitted : ret;
3234}
3235
3236static const struct file_operations io_uring_fops = {
3237 .release = io_uring_release,
3238 .mmap = io_uring_mmap,
3239 .poll = io_uring_poll,
3240 .fasync = io_uring_fasync,
3241};
3242
3243static int io_allocate_scq_urings(struct io_ring_ctx *ctx,
3244 struct io_uring_params *p)
3245{
3246 struct io_sq_ring *sq_ring;
3247 struct io_cq_ring *cq_ring;
3248 size_t size;
3249
3250 sq_ring = io_mem_alloc(struct_size(sq_ring, array, p->sq_entries));
3251 if (!sq_ring)
3252 return -ENOMEM;
3253
3254 ctx->sq_ring = sq_ring;
3255 sq_ring->ring_mask = p->sq_entries - 1;
3256 sq_ring->ring_entries = p->sq_entries;
3257 ctx->sq_mask = sq_ring->ring_mask;
3258 ctx->sq_entries = sq_ring->ring_entries;
3259
3260 size = array_size(sizeof(struct io_uring_sqe), p->sq_entries);
3261 if (size == SIZE_MAX)
3262 return -EOVERFLOW;
3263
3264 ctx->sq_sqes = io_mem_alloc(size);
Mark Rutland52e04ef2019-04-30 17:30:21 +01003265 if (!ctx->sq_sqes)
Jens Axboe2b188cc2019-01-07 10:46:33 -07003266 return -ENOMEM;
Jens Axboe2b188cc2019-01-07 10:46:33 -07003267
3268 cq_ring = io_mem_alloc(struct_size(cq_ring, cqes, p->cq_entries));
Mark Rutland52e04ef2019-04-30 17:30:21 +01003269 if (!cq_ring)
Jens Axboe2b188cc2019-01-07 10:46:33 -07003270 return -ENOMEM;
Jens Axboe2b188cc2019-01-07 10:46:33 -07003271
3272 ctx->cq_ring = cq_ring;
3273 cq_ring->ring_mask = p->cq_entries - 1;
3274 cq_ring->ring_entries = p->cq_entries;
3275 ctx->cq_mask = cq_ring->ring_mask;
3276 ctx->cq_entries = cq_ring->ring_entries;
3277 return 0;
3278}
3279
3280/*
3281 * Allocate an anonymous fd, this is what constitutes the application
3282 * visible backing of an io_uring instance. The application mmaps this
3283 * fd to gain access to the SQ/CQ ring details. If UNIX sockets are enabled,
3284 * we have to tie this fd to a socket for file garbage collection purposes.
3285 */
3286static int io_uring_get_fd(struct io_ring_ctx *ctx)
3287{
3288 struct file *file;
3289 int ret;
3290
3291#if defined(CONFIG_UNIX)
3292 ret = sock_create_kern(&init_net, PF_UNIX, SOCK_RAW, IPPROTO_IP,
3293 &ctx->ring_sock);
3294 if (ret)
3295 return ret;
3296#endif
3297
3298 ret = get_unused_fd_flags(O_RDWR | O_CLOEXEC);
3299 if (ret < 0)
3300 goto err;
3301
3302 file = anon_inode_getfile("[io_uring]", &io_uring_fops, ctx,
3303 O_RDWR | O_CLOEXEC);
3304 if (IS_ERR(file)) {
3305 put_unused_fd(ret);
3306 ret = PTR_ERR(file);
3307 goto err;
3308 }
3309
3310#if defined(CONFIG_UNIX)
3311 ctx->ring_sock->file = file;
Jens Axboe6b063142019-01-10 22:13:58 -07003312 ctx->ring_sock->sk->sk_user_data = ctx;
Jens Axboe2b188cc2019-01-07 10:46:33 -07003313#endif
3314 fd_install(ret, file);
3315 return ret;
3316err:
3317#if defined(CONFIG_UNIX)
3318 sock_release(ctx->ring_sock);
3319 ctx->ring_sock = NULL;
3320#endif
3321 return ret;
3322}
3323
3324static int io_uring_create(unsigned entries, struct io_uring_params *p)
3325{
3326 struct user_struct *user = NULL;
3327 struct io_ring_ctx *ctx;
3328 bool account_mem;
3329 int ret;
3330
3331 if (!entries || entries > IORING_MAX_ENTRIES)
3332 return -EINVAL;
3333
3334 /*
3335 * Use twice as many entries for the CQ ring. It's possible for the
3336 * application to drive a higher depth than the size of the SQ ring,
3337 * since the sqes are only used at submission time. This allows for
3338 * some flexibility in overcommitting a bit.
3339 */
3340 p->sq_entries = roundup_pow_of_two(entries);
3341 p->cq_entries = 2 * p->sq_entries;
3342
3343 user = get_uid(current_user());
3344 account_mem = !capable(CAP_IPC_LOCK);
3345
3346 if (account_mem) {
3347 ret = io_account_mem(user,
3348 ring_pages(p->sq_entries, p->cq_entries));
3349 if (ret) {
3350 free_uid(user);
3351 return ret;
3352 }
3353 }
3354
3355 ctx = io_ring_ctx_alloc(p);
3356 if (!ctx) {
3357 if (account_mem)
3358 io_unaccount_mem(user, ring_pages(p->sq_entries,
3359 p->cq_entries));
3360 free_uid(user);
3361 return -ENOMEM;
3362 }
3363 ctx->compat = in_compat_syscall();
3364 ctx->account_mem = account_mem;
3365 ctx->user = user;
3366
3367 ret = io_allocate_scq_urings(ctx, p);
3368 if (ret)
3369 goto err;
3370
Jens Axboe6c271ce2019-01-10 11:22:30 -07003371 ret = io_sq_offload_start(ctx, p);
Jens Axboe2b188cc2019-01-07 10:46:33 -07003372 if (ret)
3373 goto err;
3374
3375 ret = io_uring_get_fd(ctx);
3376 if (ret < 0)
3377 goto err;
3378
3379 memset(&p->sq_off, 0, sizeof(p->sq_off));
3380 p->sq_off.head = offsetof(struct io_sq_ring, r.head);
3381 p->sq_off.tail = offsetof(struct io_sq_ring, r.tail);
3382 p->sq_off.ring_mask = offsetof(struct io_sq_ring, ring_mask);
3383 p->sq_off.ring_entries = offsetof(struct io_sq_ring, ring_entries);
3384 p->sq_off.flags = offsetof(struct io_sq_ring, flags);
3385 p->sq_off.dropped = offsetof(struct io_sq_ring, dropped);
3386 p->sq_off.array = offsetof(struct io_sq_ring, array);
3387
3388 memset(&p->cq_off, 0, sizeof(p->cq_off));
3389 p->cq_off.head = offsetof(struct io_cq_ring, r.head);
3390 p->cq_off.tail = offsetof(struct io_cq_ring, r.tail);
3391 p->cq_off.ring_mask = offsetof(struct io_cq_ring, ring_mask);
3392 p->cq_off.ring_entries = offsetof(struct io_cq_ring, ring_entries);
3393 p->cq_off.overflow = offsetof(struct io_cq_ring, overflow);
3394 p->cq_off.cqes = offsetof(struct io_cq_ring, cqes);
3395 return ret;
3396err:
3397 io_ring_ctx_wait_and_kill(ctx);
3398 return ret;
3399}
3400
3401/*
3402 * Sets up an aio uring context, and returns the fd. Applications asks for a
3403 * ring size, we return the actual sq/cq ring sizes (among other things) in the
3404 * params structure passed in.
3405 */
3406static long io_uring_setup(u32 entries, struct io_uring_params __user *params)
3407{
3408 struct io_uring_params p;
3409 long ret;
3410 int i;
3411
3412 if (copy_from_user(&p, params, sizeof(p)))
3413 return -EFAULT;
3414 for (i = 0; i < ARRAY_SIZE(p.resv); i++) {
3415 if (p.resv[i])
3416 return -EINVAL;
3417 }
3418
Jens Axboe6c271ce2019-01-10 11:22:30 -07003419 if (p.flags & ~(IORING_SETUP_IOPOLL | IORING_SETUP_SQPOLL |
3420 IORING_SETUP_SQ_AFF))
Jens Axboe2b188cc2019-01-07 10:46:33 -07003421 return -EINVAL;
3422
3423 ret = io_uring_create(entries, &p);
3424 if (ret < 0)
3425 return ret;
3426
3427 if (copy_to_user(params, &p, sizeof(p)))
3428 return -EFAULT;
3429
3430 return ret;
3431}
3432
3433SYSCALL_DEFINE2(io_uring_setup, u32, entries,
3434 struct io_uring_params __user *, params)
3435{
3436 return io_uring_setup(entries, params);
3437}
3438
Jens Axboeedafcce2019-01-09 09:16:05 -07003439static int __io_uring_register(struct io_ring_ctx *ctx, unsigned opcode,
3440 void __user *arg, unsigned nr_args)
Jens Axboeb19062a2019-04-15 10:49:38 -06003441 __releases(ctx->uring_lock)
3442 __acquires(ctx->uring_lock)
Jens Axboeedafcce2019-01-09 09:16:05 -07003443{
3444 int ret;
3445
Jens Axboe35fa71a2019-04-22 10:23:23 -06003446 /*
3447 * We're inside the ring mutex, if the ref is already dying, then
3448 * someone else killed the ctx or is already going through
3449 * io_uring_register().
3450 */
3451 if (percpu_ref_is_dying(&ctx->refs))
3452 return -ENXIO;
3453
Jens Axboeedafcce2019-01-09 09:16:05 -07003454 percpu_ref_kill(&ctx->refs);
Jens Axboeb19062a2019-04-15 10:49:38 -06003455
3456 /*
3457 * Drop uring mutex before waiting for references to exit. If another
3458 * thread is currently inside io_uring_enter() it might need to grab
3459 * the uring_lock to make progress. If we hold it here across the drain
3460 * wait, then we can deadlock. It's safe to drop the mutex here, since
3461 * no new references will come in after we've killed the percpu ref.
3462 */
3463 mutex_unlock(&ctx->uring_lock);
Jens Axboeedafcce2019-01-09 09:16:05 -07003464 wait_for_completion(&ctx->ctx_done);
Jens Axboeb19062a2019-04-15 10:49:38 -06003465 mutex_lock(&ctx->uring_lock);
Jens Axboeedafcce2019-01-09 09:16:05 -07003466
3467 switch (opcode) {
3468 case IORING_REGISTER_BUFFERS:
3469 ret = io_sqe_buffer_register(ctx, arg, nr_args);
3470 break;
3471 case IORING_UNREGISTER_BUFFERS:
3472 ret = -EINVAL;
3473 if (arg || nr_args)
3474 break;
3475 ret = io_sqe_buffer_unregister(ctx);
3476 break;
Jens Axboe6b063142019-01-10 22:13:58 -07003477 case IORING_REGISTER_FILES:
3478 ret = io_sqe_files_register(ctx, arg, nr_args);
3479 break;
3480 case IORING_UNREGISTER_FILES:
3481 ret = -EINVAL;
3482 if (arg || nr_args)
3483 break;
3484 ret = io_sqe_files_unregister(ctx);
3485 break;
Jens Axboe9b402842019-04-11 11:45:41 -06003486 case IORING_REGISTER_EVENTFD:
3487 ret = -EINVAL;
3488 if (nr_args != 1)
3489 break;
3490 ret = io_eventfd_register(ctx, arg);
3491 break;
3492 case IORING_UNREGISTER_EVENTFD:
3493 ret = -EINVAL;
3494 if (arg || nr_args)
3495 break;
3496 ret = io_eventfd_unregister(ctx);
3497 break;
Jens Axboeedafcce2019-01-09 09:16:05 -07003498 default:
3499 ret = -EINVAL;
3500 break;
3501 }
3502
3503 /* bring the ctx back to life */
3504 reinit_completion(&ctx->ctx_done);
3505 percpu_ref_reinit(&ctx->refs);
3506 return ret;
3507}
3508
3509SYSCALL_DEFINE4(io_uring_register, unsigned int, fd, unsigned int, opcode,
3510 void __user *, arg, unsigned int, nr_args)
3511{
3512 struct io_ring_ctx *ctx;
3513 long ret = -EBADF;
3514 struct fd f;
3515
3516 f = fdget(fd);
3517 if (!f.file)
3518 return -EBADF;
3519
3520 ret = -EOPNOTSUPP;
3521 if (f.file->f_op != &io_uring_fops)
3522 goto out_fput;
3523
3524 ctx = f.file->private_data;
3525
3526 mutex_lock(&ctx->uring_lock);
3527 ret = __io_uring_register(ctx, opcode, arg, nr_args);
3528 mutex_unlock(&ctx->uring_lock);
3529out_fput:
3530 fdput(f);
3531 return ret;
3532}
3533
Jens Axboe2b188cc2019-01-07 10:46:33 -07003534static int __init io_uring_init(void)
3535{
3536 req_cachep = KMEM_CACHE(io_kiocb, SLAB_HWCACHE_ALIGN | SLAB_PANIC);
3537 return 0;
3538};
3539__initcall(io_uring_init);