Commit | Line | Data |
---|---|---|
2b188cc1 JA |
1 | // 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 | |
1e84b97b SB |
7 | * 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. | |
2b188cc1 JA |
29 | * |
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 | |
c992fe29 | 40 | * Copyright (c) 2018-2019 Christoph Hellwig |
2b188cc1 JA |
41 | */ |
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> | |
6c271ce2 | 59 | #include <linux/kthread.h> |
2b188cc1 | 60 | #include <linux/blkdev.h> |
edafccee | 61 | #include <linux/bvec.h> |
2b188cc1 JA |
62 | #include <linux/net.h> |
63 | #include <net/sock.h> | |
64 | #include <net/af_unix.h> | |
6b06314c | 65 | #include <net/scm.h> |
2b188cc1 JA |
66 | #include <linux/anon_inodes.h> |
67 | #include <linux/sched/mm.h> | |
68 | #include <linux/uaccess.h> | |
69 | #include <linux/nospec.h> | |
edafccee JA |
70 | #include <linux/sizes.h> |
71 | #include <linux/hugetlb.h> | |
aa4c3967 | 72 | #include <linux/highmem.h> |
2b188cc1 | 73 | |
c826bd7a DD |
74 | #define CREATE_TRACE_POINTS |
75 | #include <trace/events/io_uring.h> | |
76 | ||
2b188cc1 JA |
77 | #include <uapi/linux/io_uring.h> |
78 | ||
79 | #include "internal.h" | |
561fb04a | 80 | #include "io-wq.h" |
2b188cc1 | 81 | |
5277deaa | 82 | #define IORING_MAX_ENTRIES 32768 |
33a107f0 | 83 | #define IORING_MAX_CQ_ENTRIES (2 * IORING_MAX_ENTRIES) |
65e19f54 JA |
84 | |
85 | /* | |
86 | * Shift of 9 is 512 entries, or exactly one page on 64-bit archs | |
87 | */ | |
88 | #define IORING_FILE_TABLE_SHIFT 9 | |
89 | #define IORING_MAX_FILES_TABLE (1U << IORING_FILE_TABLE_SHIFT) | |
90 | #define IORING_FILE_TABLE_MASK (IORING_MAX_FILES_TABLE - 1) | |
91 | #define IORING_MAX_FIXED_FILES (64 * IORING_MAX_FILES_TABLE) | |
2b188cc1 JA |
92 | |
93 | struct io_uring { | |
94 | u32 head ____cacheline_aligned_in_smp; | |
95 | u32 tail ____cacheline_aligned_in_smp; | |
96 | }; | |
97 | ||
1e84b97b | 98 | /* |
75b28aff HV |
99 | * This data is shared with the application through the mmap at offsets |
100 | * IORING_OFF_SQ_RING and IORING_OFF_CQ_RING. | |
1e84b97b SB |
101 | * |
102 | * The offsets to the member fields are published through struct | |
103 | * io_sqring_offsets when calling io_uring_setup. | |
104 | */ | |
75b28aff | 105 | struct io_rings { |
1e84b97b SB |
106 | /* |
107 | * Head and tail offsets into the ring; the offsets need to be | |
108 | * masked to get valid indices. | |
109 | * | |
75b28aff HV |
110 | * The kernel controls head of the sq ring and the tail of the cq ring, |
111 | * and the application controls tail of the sq ring and the head of the | |
112 | * cq ring. | |
1e84b97b | 113 | */ |
75b28aff | 114 | struct io_uring sq, cq; |
1e84b97b | 115 | /* |
75b28aff | 116 | * Bitmasks to apply to head and tail offsets (constant, equals |
1e84b97b SB |
117 | * ring_entries - 1) |
118 | */ | |
75b28aff HV |
119 | u32 sq_ring_mask, cq_ring_mask; |
120 | /* Ring sizes (constant, power of 2) */ | |
121 | u32 sq_ring_entries, cq_ring_entries; | |
1e84b97b SB |
122 | /* |
123 | * Number of invalid entries dropped by the kernel due to | |
124 | * invalid index stored in array | |
125 | * | |
126 | * Written by the kernel, shouldn't be modified by the | |
127 | * application (i.e. get number of "new events" by comparing to | |
128 | * cached value). | |
129 | * | |
130 | * After a new SQ head value was read by the application this | |
131 | * counter includes all submissions that were dropped reaching | |
132 | * the new SQ head (and possibly more). | |
133 | */ | |
75b28aff | 134 | u32 sq_dropped; |
1e84b97b SB |
135 | /* |
136 | * Runtime flags | |
137 | * | |
138 | * Written by the kernel, shouldn't be modified by the | |
139 | * application. | |
140 | * | |
141 | * The application needs a full memory barrier before checking | |
142 | * for IORING_SQ_NEED_WAKEUP after updating the sq tail. | |
143 | */ | |
75b28aff | 144 | u32 sq_flags; |
1e84b97b SB |
145 | /* |
146 | * Number of completion events lost because the queue was full; | |
147 | * this should be avoided by the application by making sure | |
0b4295b5 | 148 | * there are not more requests pending than there is space in |
1e84b97b SB |
149 | * the completion queue. |
150 | * | |
151 | * Written by the kernel, shouldn't be modified by the | |
152 | * application (i.e. get number of "new events" by comparing to | |
153 | * cached value). | |
154 | * | |
155 | * As completion events come in out of order this counter is not | |
156 | * ordered with any other data. | |
157 | */ | |
75b28aff | 158 | u32 cq_overflow; |
1e84b97b SB |
159 | /* |
160 | * Ring buffer of completion events. | |
161 | * | |
162 | * The kernel writes completion events fresh every time they are | |
163 | * produced, so the application is allowed to modify pending | |
164 | * entries. | |
165 | */ | |
75b28aff | 166 | struct io_uring_cqe cqes[] ____cacheline_aligned_in_smp; |
2b188cc1 JA |
167 | }; |
168 | ||
edafccee JA |
169 | struct io_mapped_ubuf { |
170 | u64 ubuf; | |
171 | size_t len; | |
172 | struct bio_vec *bvec; | |
173 | unsigned int nr_bvecs; | |
174 | }; | |
175 | ||
65e19f54 JA |
176 | struct fixed_file_table { |
177 | struct file **files; | |
31b51510 JA |
178 | }; |
179 | ||
2b188cc1 JA |
180 | struct io_ring_ctx { |
181 | struct { | |
182 | struct percpu_ref refs; | |
183 | } ____cacheline_aligned_in_smp; | |
184 | ||
185 | struct { | |
186 | unsigned int flags; | |
187 | bool compat; | |
188 | bool account_mem; | |
1d7bb1d5 | 189 | bool cq_overflow_flushed; |
1b4a51b6 | 190 | bool drain_next; |
2b188cc1 | 191 | |
75b28aff HV |
192 | /* |
193 | * Ring buffer of indices into array of io_uring_sqe, which is | |
194 | * mmapped by the application using the IORING_OFF_SQES offset. | |
195 | * | |
196 | * This indirection could e.g. be used to assign fixed | |
197 | * io_uring_sqe entries to operations and only submit them to | |
198 | * the queue when needed. | |
199 | * | |
200 | * The kernel modifies neither the indices array nor the entries | |
201 | * array. | |
202 | */ | |
203 | u32 *sq_array; | |
2b188cc1 JA |
204 | unsigned cached_sq_head; |
205 | unsigned sq_entries; | |
206 | unsigned sq_mask; | |
6c271ce2 | 207 | unsigned sq_thread_idle; |
498ccd9e | 208 | unsigned cached_sq_dropped; |
206aefde | 209 | atomic_t cached_cq_overflow; |
2b188cc1 | 210 | struct io_uring_sqe *sq_sqes; |
de0617e4 JA |
211 | |
212 | struct list_head defer_list; | |
5262f567 | 213 | struct list_head timeout_list; |
1d7bb1d5 | 214 | struct list_head cq_overflow_list; |
fcb323cc JA |
215 | |
216 | wait_queue_head_t inflight_wait; | |
2b188cc1 JA |
217 | } ____cacheline_aligned_in_smp; |
218 | ||
206aefde JA |
219 | struct io_rings *rings; |
220 | ||
2b188cc1 | 221 | /* IO offload */ |
561fb04a | 222 | struct io_wq *io_wq; |
6c271ce2 | 223 | struct task_struct *sqo_thread; /* if using sq thread polling */ |
2b188cc1 | 224 | struct mm_struct *sqo_mm; |
6c271ce2 | 225 | wait_queue_head_t sqo_wait; |
75b28aff | 226 | |
6b06314c JA |
227 | /* |
228 | * If used, fixed file set. Writers must ensure that ->refs is dead, | |
229 | * readers must ensure that ->refs is alive as long as the file* is | |
230 | * used. Only updated through io_uring_register(2). | |
231 | */ | |
65e19f54 | 232 | struct fixed_file_table *file_table; |
6b06314c JA |
233 | unsigned nr_user_files; |
234 | ||
edafccee JA |
235 | /* if used, fixed mapped user buffers */ |
236 | unsigned nr_user_bufs; | |
237 | struct io_mapped_ubuf *user_bufs; | |
238 | ||
2b188cc1 JA |
239 | struct user_struct *user; |
240 | ||
0b8c0ec7 | 241 | const struct cred *creds; |
181e448d | 242 | |
206aefde JA |
243 | /* 0 is for ctx quiesce/reinit/free, 1 is for sqo_thread started */ |
244 | struct completion *completions; | |
245 | ||
0ddf92e8 JA |
246 | /* if all else fails... */ |
247 | struct io_kiocb *fallback_req; | |
248 | ||
206aefde JA |
249 | #if defined(CONFIG_UNIX) |
250 | struct socket *ring_sock; | |
251 | #endif | |
252 | ||
253 | struct { | |
254 | unsigned cached_cq_tail; | |
255 | unsigned cq_entries; | |
256 | unsigned cq_mask; | |
257 | atomic_t cq_timeouts; | |
258 | struct wait_queue_head cq_wait; | |
259 | struct fasync_struct *cq_fasync; | |
260 | struct eventfd_ctx *cq_ev_fd; | |
261 | } ____cacheline_aligned_in_smp; | |
2b188cc1 JA |
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; | |
def596e9 JA |
270 | 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; | |
78076bb6 JA |
278 | struct hlist_head *cancel_hash; |
279 | unsigned cancel_hash_bits; | |
31b51510 | 280 | |
fcb323cc JA |
281 | spinlock_t inflight_lock; |
282 | struct list_head inflight_list; | |
2b188cc1 | 283 | } ____cacheline_aligned_in_smp; |
2b188cc1 JA |
284 | }; |
285 | ||
09bb8394 JA |
286 | /* |
287 | * First field must be the file pointer in all the | |
288 | * iocb unions! See also 'struct kiocb' in <linux/fs.h> | |
289 | */ | |
221c5eb2 JA |
290 | struct io_poll_iocb { |
291 | struct file *file; | |
0969e783 JA |
292 | union { |
293 | struct wait_queue_head *head; | |
294 | u64 addr; | |
295 | }; | |
221c5eb2 | 296 | __poll_t events; |
8c838788 | 297 | bool done; |
221c5eb2 | 298 | bool canceled; |
392edb45 | 299 | struct wait_queue_entry wait; |
221c5eb2 JA |
300 | }; |
301 | ||
ad8a48ac JA |
302 | struct io_timeout_data { |
303 | struct io_kiocb *req; | |
304 | struct hrtimer timer; | |
305 | struct timespec64 ts; | |
306 | enum hrtimer_mode mode; | |
cc42e0ac | 307 | u32 seq_offset; |
ad8a48ac JA |
308 | }; |
309 | ||
8ed8d3c3 JA |
310 | struct io_accept { |
311 | struct file *file; | |
312 | struct sockaddr __user *addr; | |
313 | int __user *addr_len; | |
314 | int flags; | |
315 | }; | |
316 | ||
317 | struct io_sync { | |
318 | struct file *file; | |
319 | loff_t len; | |
320 | loff_t off; | |
321 | int flags; | |
322 | }; | |
323 | ||
f499a021 JA |
324 | struct io_async_connect { |
325 | struct sockaddr_storage address; | |
326 | }; | |
327 | ||
03b1230c JA |
328 | struct io_async_msghdr { |
329 | struct iovec fast_iov[UIO_FASTIOV]; | |
330 | struct iovec *iov; | |
331 | struct sockaddr __user *uaddr; | |
332 | struct msghdr msg; | |
333 | }; | |
334 | ||
f67676d1 JA |
335 | struct io_async_rw { |
336 | struct iovec fast_iov[UIO_FASTIOV]; | |
337 | struct iovec *iov; | |
338 | ssize_t nr_segs; | |
339 | ssize_t size; | |
340 | }; | |
341 | ||
1a6b74fc JA |
342 | struct io_async_ctx { |
343 | struct io_uring_sqe sqe; | |
f67676d1 JA |
344 | union { |
345 | struct io_async_rw rw; | |
03b1230c | 346 | struct io_async_msghdr msg; |
f499a021 | 347 | struct io_async_connect connect; |
2d28390a | 348 | struct io_timeout_data timeout; |
f67676d1 | 349 | }; |
1a6b74fc JA |
350 | }; |
351 | ||
09bb8394 JA |
352 | /* |
353 | * NOTE! Each of the iocb union members has the file pointer | |
354 | * as the first entry in their struct definition. So you can | |
355 | * access the file pointer through any of the sub-structs, | |
356 | * or directly as just 'ki_filp' in this struct. | |
357 | */ | |
2b188cc1 | 358 | struct io_kiocb { |
221c5eb2 | 359 | union { |
09bb8394 | 360 | struct file *file; |
221c5eb2 JA |
361 | struct kiocb rw; |
362 | struct io_poll_iocb poll; | |
8ed8d3c3 JA |
363 | struct io_accept accept; |
364 | struct io_sync sync; | |
221c5eb2 | 365 | }; |
2b188cc1 | 366 | |
cf6fd4bd | 367 | const struct io_uring_sqe *sqe; |
1a6b74fc | 368 | struct io_async_ctx *io; |
cf6fd4bd PB |
369 | struct file *ring_file; |
370 | int ring_fd; | |
371 | bool has_user; | |
372 | bool in_async; | |
373 | bool needs_fixed_file; | |
2b188cc1 JA |
374 | |
375 | struct io_ring_ctx *ctx; | |
eac406c6 JA |
376 | union { |
377 | struct list_head list; | |
78076bb6 | 378 | struct hlist_node hash_node; |
eac406c6 | 379 | }; |
9e645e11 | 380 | struct list_head link_list; |
2b188cc1 | 381 | unsigned int flags; |
c16361c1 | 382 | refcount_t refs; |
8449eeda | 383 | #define REQ_F_NOWAIT 1 /* must not punt to workers */ |
def596e9 | 384 | #define REQ_F_IOPOLL_COMPLETED 2 /* polled IO has completed */ |
6b06314c | 385 | #define REQ_F_FIXED_FILE 4 /* ctx owns file */ |
4d7dd462 | 386 | #define REQ_F_LINK_NEXT 8 /* already grabbed next link */ |
e2033e33 SB |
387 | #define REQ_F_IO_DRAIN 16 /* drain existing IO first */ |
388 | #define REQ_F_IO_DRAINED 32 /* drain done */ | |
9e645e11 | 389 | #define REQ_F_LINK 64 /* linked sqes */ |
2665abfd | 390 | #define REQ_F_LINK_TIMEOUT 128 /* has linked timeout */ |
f7b76ac9 | 391 | #define REQ_F_FAIL_LINK 256 /* fail rest of links */ |
1b4a51b6 | 392 | #define REQ_F_DRAIN_LINK 512 /* link should be fully drained */ |
5262f567 | 393 | #define REQ_F_TIMEOUT 1024 /* timeout request */ |
491381ce JA |
394 | #define REQ_F_ISREG 2048 /* regular file */ |
395 | #define REQ_F_MUST_PUNT 4096 /* must be punted even for NONBLOCK */ | |
93bd25bb | 396 | #define REQ_F_TIMEOUT_NOSEQ 8192 /* no timeout sequence */ |
fb4b3d3f LT |
397 | #define REQ_F_INFLIGHT 16384 /* on inflight list */ |
398 | #define REQ_F_COMP_LOCKED 32768 /* completion under lock */ | |
4e88d6e7 | 399 | #define REQ_F_HARDLINK 65536 /* doesn't sever on completion < 0 */ |
8ed8d3c3 | 400 | #define REQ_F_PREPPED 131072 /* request already opcode prepared */ |
2b188cc1 | 401 | u64 user_data; |
9e645e11 | 402 | u32 result; |
de0617e4 | 403 | u32 sequence; |
2b188cc1 | 404 | |
fcb323cc JA |
405 | struct list_head inflight_entry; |
406 | ||
561fb04a | 407 | struct io_wq_work work; |
2b188cc1 JA |
408 | }; |
409 | ||
410 | #define IO_PLUG_THRESHOLD 2 | |
def596e9 | 411 | #define IO_IOPOLL_BATCH 8 |
2b188cc1 | 412 | |
9a56a232 JA |
413 | struct io_submit_state { |
414 | struct blk_plug plug; | |
415 | ||
2579f913 JA |
416 | /* |
417 | * io_kiocb alloc cache | |
418 | */ | |
419 | void *reqs[IO_IOPOLL_BATCH]; | |
420 | unsigned int free_reqs; | |
421 | unsigned int cur_req; | |
422 | ||
9a56a232 JA |
423 | /* |
424 | * File reference cache | |
425 | */ | |
426 | struct file *file; | |
427 | unsigned int fd; | |
428 | unsigned int has_refs; | |
429 | unsigned int used_refs; | |
430 | unsigned int ios_left; | |
431 | }; | |
432 | ||
561fb04a | 433 | static void io_wq_submit_work(struct io_wq_work **workptr); |
78e19bbe | 434 | static void io_cqring_fill_event(struct io_kiocb *req, long res); |
4fe2c963 | 435 | static void __io_free_req(struct io_kiocb *req); |
ec9c02ad | 436 | static void io_put_req(struct io_kiocb *req); |
78e19bbe | 437 | static void io_double_put_req(struct io_kiocb *req); |
978db57e | 438 | static void __io_double_put_req(struct io_kiocb *req); |
94ae5e77 JA |
439 | static struct io_kiocb *io_prep_linked_timeout(struct io_kiocb *req); |
440 | static void io_queue_linked_timeout(struct io_kiocb *req); | |
de0617e4 | 441 | |
2b188cc1 JA |
442 | static struct kmem_cache *req_cachep; |
443 | ||
444 | static const struct file_operations io_uring_fops; | |
445 | ||
446 | struct sock *io_uring_get_socket(struct file *file) | |
447 | { | |
448 | #if defined(CONFIG_UNIX) | |
449 | if (file->f_op == &io_uring_fops) { | |
450 | struct io_ring_ctx *ctx = file->private_data; | |
451 | ||
452 | return ctx->ring_sock->sk; | |
453 | } | |
454 | #endif | |
455 | return NULL; | |
456 | } | |
457 | EXPORT_SYMBOL(io_uring_get_socket); | |
458 | ||
459 | static void io_ring_ctx_ref_free(struct percpu_ref *ref) | |
460 | { | |
461 | struct io_ring_ctx *ctx = container_of(ref, struct io_ring_ctx, refs); | |
462 | ||
206aefde | 463 | complete(&ctx->completions[0]); |
2b188cc1 JA |
464 | } |
465 | ||
466 | static struct io_ring_ctx *io_ring_ctx_alloc(struct io_uring_params *p) | |
467 | { | |
468 | struct io_ring_ctx *ctx; | |
78076bb6 | 469 | int hash_bits; |
2b188cc1 JA |
470 | |
471 | ctx = kzalloc(sizeof(*ctx), GFP_KERNEL); | |
472 | if (!ctx) | |
473 | return NULL; | |
474 | ||
0ddf92e8 JA |
475 | ctx->fallback_req = kmem_cache_alloc(req_cachep, GFP_KERNEL); |
476 | if (!ctx->fallback_req) | |
477 | goto err; | |
478 | ||
206aefde JA |
479 | ctx->completions = kmalloc(2 * sizeof(struct completion), GFP_KERNEL); |
480 | if (!ctx->completions) | |
481 | goto err; | |
482 | ||
78076bb6 JA |
483 | /* |
484 | * Use 5 bits less than the max cq entries, that should give us around | |
485 | * 32 entries per hash list if totally full and uniformly spread. | |
486 | */ | |
487 | hash_bits = ilog2(p->cq_entries); | |
488 | hash_bits -= 5; | |
489 | if (hash_bits <= 0) | |
490 | hash_bits = 1; | |
491 | ctx->cancel_hash_bits = hash_bits; | |
492 | ctx->cancel_hash = kmalloc((1U << hash_bits) * sizeof(struct hlist_head), | |
493 | GFP_KERNEL); | |
494 | if (!ctx->cancel_hash) | |
495 | goto err; | |
496 | __hash_init(ctx->cancel_hash, 1U << hash_bits); | |
497 | ||
21482896 | 498 | if (percpu_ref_init(&ctx->refs, io_ring_ctx_ref_free, |
206aefde JA |
499 | PERCPU_REF_ALLOW_REINIT, GFP_KERNEL)) |
500 | goto err; | |
2b188cc1 JA |
501 | |
502 | ctx->flags = p->flags; | |
503 | init_waitqueue_head(&ctx->cq_wait); | |
1d7bb1d5 | 504 | INIT_LIST_HEAD(&ctx->cq_overflow_list); |
206aefde JA |
505 | init_completion(&ctx->completions[0]); |
506 | init_completion(&ctx->completions[1]); | |
2b188cc1 JA |
507 | mutex_init(&ctx->uring_lock); |
508 | init_waitqueue_head(&ctx->wait); | |
509 | spin_lock_init(&ctx->completion_lock); | |
def596e9 | 510 | INIT_LIST_HEAD(&ctx->poll_list); |
de0617e4 | 511 | INIT_LIST_HEAD(&ctx->defer_list); |
5262f567 | 512 | INIT_LIST_HEAD(&ctx->timeout_list); |
fcb323cc JA |
513 | init_waitqueue_head(&ctx->inflight_wait); |
514 | spin_lock_init(&ctx->inflight_lock); | |
515 | INIT_LIST_HEAD(&ctx->inflight_list); | |
2b188cc1 | 516 | return ctx; |
206aefde | 517 | err: |
0ddf92e8 JA |
518 | if (ctx->fallback_req) |
519 | kmem_cache_free(req_cachep, ctx->fallback_req); | |
206aefde | 520 | kfree(ctx->completions); |
78076bb6 | 521 | kfree(ctx->cancel_hash); |
206aefde JA |
522 | kfree(ctx); |
523 | return NULL; | |
2b188cc1 JA |
524 | } |
525 | ||
9d858b21 | 526 | static inline bool __req_need_defer(struct io_kiocb *req) |
7adf4eaf | 527 | { |
a197f664 JL |
528 | struct io_ring_ctx *ctx = req->ctx; |
529 | ||
498ccd9e JA |
530 | return req->sequence != ctx->cached_cq_tail + ctx->cached_sq_dropped |
531 | + atomic_read(&ctx->cached_cq_overflow); | |
7adf4eaf JA |
532 | } |
533 | ||
9d858b21 | 534 | static inline bool req_need_defer(struct io_kiocb *req) |
de0617e4 | 535 | { |
9d858b21 BL |
536 | if ((req->flags & (REQ_F_IO_DRAIN|REQ_F_IO_DRAINED)) == REQ_F_IO_DRAIN) |
537 | return __req_need_defer(req); | |
de0617e4 | 538 | |
9d858b21 | 539 | return false; |
de0617e4 JA |
540 | } |
541 | ||
7adf4eaf | 542 | static struct io_kiocb *io_get_deferred_req(struct io_ring_ctx *ctx) |
de0617e4 JA |
543 | { |
544 | struct io_kiocb *req; | |
545 | ||
7adf4eaf | 546 | req = list_first_entry_or_null(&ctx->defer_list, struct io_kiocb, list); |
9d858b21 | 547 | if (req && !req_need_defer(req)) { |
de0617e4 JA |
548 | list_del_init(&req->list); |
549 | return req; | |
550 | } | |
551 | ||
552 | return NULL; | |
553 | } | |
554 | ||
5262f567 JA |
555 | static struct io_kiocb *io_get_timeout_req(struct io_ring_ctx *ctx) |
556 | { | |
7adf4eaf JA |
557 | struct io_kiocb *req; |
558 | ||
559 | req = list_first_entry_or_null(&ctx->timeout_list, struct io_kiocb, list); | |
93bd25bb JA |
560 | if (req) { |
561 | if (req->flags & REQ_F_TIMEOUT_NOSEQ) | |
562 | return NULL; | |
fb4b3d3f | 563 | if (!__req_need_defer(req)) { |
93bd25bb JA |
564 | list_del_init(&req->list); |
565 | return req; | |
566 | } | |
7adf4eaf JA |
567 | } |
568 | ||
569 | return NULL; | |
5262f567 JA |
570 | } |
571 | ||
de0617e4 | 572 | static void __io_commit_cqring(struct io_ring_ctx *ctx) |
2b188cc1 | 573 | { |
75b28aff | 574 | struct io_rings *rings = ctx->rings; |
2b188cc1 | 575 | |
75b28aff | 576 | if (ctx->cached_cq_tail != READ_ONCE(rings->cq.tail)) { |
2b188cc1 | 577 | /* order cqe stores with ring update */ |
75b28aff | 578 | smp_store_release(&rings->cq.tail, ctx->cached_cq_tail); |
2b188cc1 | 579 | |
2b188cc1 JA |
580 | if (wq_has_sleeper(&ctx->cq_wait)) { |
581 | wake_up_interruptible(&ctx->cq_wait); | |
582 | kill_fasync(&ctx->cq_fasync, SIGIO, POLL_IN); | |
583 | } | |
584 | } | |
585 | } | |
586 | ||
561fb04a | 587 | static inline bool io_sqe_needs_user(const struct io_uring_sqe *sqe) |
18d9be1a | 588 | { |
561fb04a JA |
589 | u8 opcode = READ_ONCE(sqe->opcode); |
590 | ||
591 | return !(opcode == IORING_OP_READ_FIXED || | |
592 | opcode == IORING_OP_WRITE_FIXED); | |
593 | } | |
594 | ||
94ae5e77 JA |
595 | static inline bool io_prep_async_work(struct io_kiocb *req, |
596 | struct io_kiocb **link) | |
18d9be1a | 597 | { |
561fb04a | 598 | bool do_hashed = false; |
54a91f3b | 599 | |
cf6fd4bd PB |
600 | if (req->sqe) { |
601 | switch (req->sqe->opcode) { | |
6cc47d1d JA |
602 | case IORING_OP_WRITEV: |
603 | case IORING_OP_WRITE_FIXED: | |
53108d47 JA |
604 | /* only regular files should be hashed for writes */ |
605 | if (req->flags & REQ_F_ISREG) | |
606 | do_hashed = true; | |
5f8fd2d3 JA |
607 | /* fall-through */ |
608 | case IORING_OP_READV: | |
609 | case IORING_OP_READ_FIXED: | |
610 | case IORING_OP_SENDMSG: | |
611 | case IORING_OP_RECVMSG: | |
612 | case IORING_OP_ACCEPT: | |
613 | case IORING_OP_POLL_ADD: | |
f8e85cf2 | 614 | case IORING_OP_CONNECT: |
5f8fd2d3 JA |
615 | /* |
616 | * We know REQ_F_ISREG is not set on some of these | |
617 | * opcodes, but this enables us to keep the check in | |
618 | * just one place. | |
619 | */ | |
620 | if (!(req->flags & REQ_F_ISREG)) | |
621 | req->work.flags |= IO_WQ_WORK_UNBOUND; | |
6cc47d1d JA |
622 | break; |
623 | } | |
cf6fd4bd | 624 | if (io_sqe_needs_user(req->sqe)) |
561fb04a | 625 | req->work.flags |= IO_WQ_WORK_NEEDS_USER; |
54a91f3b JA |
626 | } |
627 | ||
94ae5e77 | 628 | *link = io_prep_linked_timeout(req); |
561fb04a JA |
629 | return do_hashed; |
630 | } | |
631 | ||
a197f664 | 632 | static inline void io_queue_async_work(struct io_kiocb *req) |
561fb04a | 633 | { |
a197f664 | 634 | struct io_ring_ctx *ctx = req->ctx; |
94ae5e77 JA |
635 | struct io_kiocb *link; |
636 | bool do_hashed; | |
637 | ||
638 | do_hashed = io_prep_async_work(req, &link); | |
561fb04a JA |
639 | |
640 | trace_io_uring_queue_async_work(ctx, do_hashed, req, &req->work, | |
641 | req->flags); | |
642 | if (!do_hashed) { | |
643 | io_wq_enqueue(ctx->io_wq, &req->work); | |
644 | } else { | |
645 | io_wq_enqueue_hashed(ctx->io_wq, &req->work, | |
646 | file_inode(req->file)); | |
647 | } | |
94ae5e77 JA |
648 | |
649 | if (link) | |
650 | io_queue_linked_timeout(link); | |
18d9be1a JA |
651 | } |
652 | ||
5262f567 JA |
653 | static void io_kill_timeout(struct io_kiocb *req) |
654 | { | |
655 | int ret; | |
656 | ||
2d28390a | 657 | ret = hrtimer_try_to_cancel(&req->io->timeout.timer); |
5262f567 JA |
658 | if (ret != -1) { |
659 | atomic_inc(&req->ctx->cq_timeouts); | |
842f9612 | 660 | list_del_init(&req->list); |
78e19bbe | 661 | io_cqring_fill_event(req, 0); |
ec9c02ad | 662 | io_put_req(req); |
5262f567 JA |
663 | } |
664 | } | |
665 | ||
666 | static void io_kill_timeouts(struct io_ring_ctx *ctx) | |
667 | { | |
668 | struct io_kiocb *req, *tmp; | |
669 | ||
670 | spin_lock_irq(&ctx->completion_lock); | |
671 | list_for_each_entry_safe(req, tmp, &ctx->timeout_list, list) | |
672 | io_kill_timeout(req); | |
673 | spin_unlock_irq(&ctx->completion_lock); | |
674 | } | |
675 | ||
de0617e4 JA |
676 | static void io_commit_cqring(struct io_ring_ctx *ctx) |
677 | { | |
678 | struct io_kiocb *req; | |
679 | ||
5262f567 JA |
680 | while ((req = io_get_timeout_req(ctx)) != NULL) |
681 | io_kill_timeout(req); | |
682 | ||
de0617e4 JA |
683 | __io_commit_cqring(ctx); |
684 | ||
685 | while ((req = io_get_deferred_req(ctx)) != NULL) { | |
686 | req->flags |= REQ_F_IO_DRAINED; | |
a197f664 | 687 | io_queue_async_work(req); |
de0617e4 JA |
688 | } |
689 | } | |
690 | ||
2b188cc1 JA |
691 | static struct io_uring_cqe *io_get_cqring(struct io_ring_ctx *ctx) |
692 | { | |
75b28aff | 693 | struct io_rings *rings = ctx->rings; |
2b188cc1 JA |
694 | unsigned tail; |
695 | ||
696 | tail = ctx->cached_cq_tail; | |
115e12e5 SB |
697 | /* |
698 | * writes to the cq entry need to come after reading head; the | |
699 | * control dependency is enough as we're using WRITE_ONCE to | |
700 | * fill the cq entry | |
701 | */ | |
75b28aff | 702 | if (tail - READ_ONCE(rings->cq.head) == rings->cq_ring_entries) |
2b188cc1 JA |
703 | return NULL; |
704 | ||
705 | ctx->cached_cq_tail++; | |
75b28aff | 706 | return &rings->cqes[tail & ctx->cq_mask]; |
2b188cc1 JA |
707 | } |
708 | ||
1d7bb1d5 JA |
709 | static void io_cqring_ev_posted(struct io_ring_ctx *ctx) |
710 | { | |
711 | if (waitqueue_active(&ctx->wait)) | |
712 | wake_up(&ctx->wait); | |
713 | if (waitqueue_active(&ctx->sqo_wait)) | |
714 | wake_up(&ctx->sqo_wait); | |
715 | if (ctx->cq_ev_fd) | |
716 | eventfd_signal(ctx->cq_ev_fd, 1); | |
717 | } | |
718 | ||
c4a2ed72 JA |
719 | /* Returns true if there are no backlogged entries after the flush */ |
720 | static bool io_cqring_overflow_flush(struct io_ring_ctx *ctx, bool force) | |
1d7bb1d5 JA |
721 | { |
722 | struct io_rings *rings = ctx->rings; | |
723 | struct io_uring_cqe *cqe; | |
724 | struct io_kiocb *req; | |
725 | unsigned long flags; | |
726 | LIST_HEAD(list); | |
727 | ||
728 | if (!force) { | |
729 | if (list_empty_careful(&ctx->cq_overflow_list)) | |
c4a2ed72 | 730 | return true; |
1d7bb1d5 JA |
731 | if ((ctx->cached_cq_tail - READ_ONCE(rings->cq.head) == |
732 | rings->cq_ring_entries)) | |
c4a2ed72 | 733 | return false; |
1d7bb1d5 JA |
734 | } |
735 | ||
736 | spin_lock_irqsave(&ctx->completion_lock, flags); | |
737 | ||
738 | /* if force is set, the ring is going away. always drop after that */ | |
739 | if (force) | |
740 | ctx->cq_overflow_flushed = true; | |
741 | ||
c4a2ed72 | 742 | cqe = NULL; |
1d7bb1d5 JA |
743 | while (!list_empty(&ctx->cq_overflow_list)) { |
744 | cqe = io_get_cqring(ctx); | |
745 | if (!cqe && !force) | |
746 | break; | |
747 | ||
748 | req = list_first_entry(&ctx->cq_overflow_list, struct io_kiocb, | |
749 | list); | |
750 | list_move(&req->list, &list); | |
751 | if (cqe) { | |
752 | WRITE_ONCE(cqe->user_data, req->user_data); | |
753 | WRITE_ONCE(cqe->res, req->result); | |
754 | WRITE_ONCE(cqe->flags, 0); | |
755 | } else { | |
756 | WRITE_ONCE(ctx->rings->cq_overflow, | |
757 | atomic_inc_return(&ctx->cached_cq_overflow)); | |
758 | } | |
759 | } | |
760 | ||
761 | io_commit_cqring(ctx); | |
762 | spin_unlock_irqrestore(&ctx->completion_lock, flags); | |
763 | io_cqring_ev_posted(ctx); | |
764 | ||
765 | while (!list_empty(&list)) { | |
766 | req = list_first_entry(&list, struct io_kiocb, list); | |
767 | list_del(&req->list); | |
ec9c02ad | 768 | io_put_req(req); |
1d7bb1d5 | 769 | } |
c4a2ed72 JA |
770 | |
771 | return cqe != NULL; | |
1d7bb1d5 JA |
772 | } |
773 | ||
78e19bbe | 774 | static void io_cqring_fill_event(struct io_kiocb *req, long res) |
2b188cc1 | 775 | { |
78e19bbe | 776 | struct io_ring_ctx *ctx = req->ctx; |
2b188cc1 JA |
777 | struct io_uring_cqe *cqe; |
778 | ||
78e19bbe | 779 | trace_io_uring_complete(ctx, req->user_data, res); |
51c3ff62 | 780 | |
2b188cc1 JA |
781 | /* |
782 | * If we can't get a cq entry, userspace overflowed the | |
783 | * submission (by quite a lot). Increment the overflow count in | |
784 | * the ring. | |
785 | */ | |
786 | cqe = io_get_cqring(ctx); | |
1d7bb1d5 | 787 | if (likely(cqe)) { |
78e19bbe | 788 | WRITE_ONCE(cqe->user_data, req->user_data); |
2b188cc1 | 789 | WRITE_ONCE(cqe->res, res); |
c71ffb67 | 790 | WRITE_ONCE(cqe->flags, 0); |
1d7bb1d5 | 791 | } else if (ctx->cq_overflow_flushed) { |
498ccd9e JA |
792 | WRITE_ONCE(ctx->rings->cq_overflow, |
793 | atomic_inc_return(&ctx->cached_cq_overflow)); | |
1d7bb1d5 JA |
794 | } else { |
795 | refcount_inc(&req->refs); | |
796 | req->result = res; | |
797 | list_add_tail(&req->list, &ctx->cq_overflow_list); | |
2b188cc1 JA |
798 | } |
799 | } | |
800 | ||
78e19bbe | 801 | static void io_cqring_add_event(struct io_kiocb *req, long res) |
2b188cc1 | 802 | { |
78e19bbe | 803 | struct io_ring_ctx *ctx = req->ctx; |
2b188cc1 JA |
804 | unsigned long flags; |
805 | ||
806 | spin_lock_irqsave(&ctx->completion_lock, flags); | |
78e19bbe | 807 | io_cqring_fill_event(req, res); |
2b188cc1 JA |
808 | io_commit_cqring(ctx); |
809 | spin_unlock_irqrestore(&ctx->completion_lock, flags); | |
810 | ||
8c838788 | 811 | io_cqring_ev_posted(ctx); |
2b188cc1 JA |
812 | } |
813 | ||
0ddf92e8 JA |
814 | static inline bool io_is_fallback_req(struct io_kiocb *req) |
815 | { | |
816 | return req == (struct io_kiocb *) | |
817 | ((unsigned long) req->ctx->fallback_req & ~1UL); | |
818 | } | |
819 | ||
820 | static struct io_kiocb *io_get_fallback_req(struct io_ring_ctx *ctx) | |
821 | { | |
822 | struct io_kiocb *req; | |
823 | ||
824 | req = ctx->fallback_req; | |
825 | if (!test_and_set_bit_lock(0, (unsigned long *) ctx->fallback_req)) | |
826 | return req; | |
827 | ||
828 | return NULL; | |
829 | } | |
830 | ||
2579f913 JA |
831 | static struct io_kiocb *io_get_req(struct io_ring_ctx *ctx, |
832 | struct io_submit_state *state) | |
2b188cc1 | 833 | { |
fd6fab2c | 834 | gfp_t gfp = GFP_KERNEL | __GFP_NOWARN; |
2b188cc1 JA |
835 | struct io_kiocb *req; |
836 | ||
837 | if (!percpu_ref_tryget(&ctx->refs)) | |
838 | return NULL; | |
839 | ||
2579f913 | 840 | if (!state) { |
fd6fab2c | 841 | req = kmem_cache_alloc(req_cachep, gfp); |
2579f913 | 842 | if (unlikely(!req)) |
0ddf92e8 | 843 | goto fallback; |
2579f913 JA |
844 | } else if (!state->free_reqs) { |
845 | size_t sz; | |
846 | int ret; | |
847 | ||
848 | sz = min_t(size_t, state->ios_left, ARRAY_SIZE(state->reqs)); | |
fd6fab2c JA |
849 | ret = kmem_cache_alloc_bulk(req_cachep, gfp, sz, state->reqs); |
850 | ||
851 | /* | |
852 | * Bulk alloc is all-or-nothing. If we fail to get a batch, | |
853 | * retry single alloc to be on the safe side. | |
854 | */ | |
855 | if (unlikely(ret <= 0)) { | |
856 | state->reqs[0] = kmem_cache_alloc(req_cachep, gfp); | |
857 | if (!state->reqs[0]) | |
0ddf92e8 | 858 | goto fallback; |
fd6fab2c JA |
859 | ret = 1; |
860 | } | |
2579f913 JA |
861 | state->free_reqs = ret - 1; |
862 | state->cur_req = 1; | |
863 | req = state->reqs[0]; | |
864 | } else { | |
865 | req = state->reqs[state->cur_req]; | |
866 | state->free_reqs--; | |
867 | state->cur_req++; | |
2b188cc1 JA |
868 | } |
869 | ||
0ddf92e8 | 870 | got_it: |
1a6b74fc | 871 | req->io = NULL; |
cf6fd4bd | 872 | req->ring_file = NULL; |
60c112b0 | 873 | req->file = NULL; |
2579f913 JA |
874 | req->ctx = ctx; |
875 | req->flags = 0; | |
e65ef56d JA |
876 | /* one is dropped after submission, the other at completion */ |
877 | refcount_set(&req->refs, 2); | |
9e645e11 | 878 | req->result = 0; |
561fb04a | 879 | INIT_IO_WORK(&req->work, io_wq_submit_work); |
2579f913 | 880 | return req; |
0ddf92e8 JA |
881 | fallback: |
882 | req = io_get_fallback_req(ctx); | |
883 | if (req) | |
884 | goto got_it; | |
6805b32e | 885 | percpu_ref_put(&ctx->refs); |
2b188cc1 JA |
886 | return NULL; |
887 | } | |
888 | ||
def596e9 JA |
889 | static void io_free_req_many(struct io_ring_ctx *ctx, void **reqs, int *nr) |
890 | { | |
891 | if (*nr) { | |
892 | kmem_cache_free_bulk(req_cachep, *nr, reqs); | |
6805b32e | 893 | percpu_ref_put_many(&ctx->refs, *nr); |
def596e9 JA |
894 | *nr = 0; |
895 | } | |
896 | } | |
897 | ||
9e645e11 | 898 | static void __io_free_req(struct io_kiocb *req) |
2b188cc1 | 899 | { |
fcb323cc JA |
900 | struct io_ring_ctx *ctx = req->ctx; |
901 | ||
1a6b74fc JA |
902 | if (req->io) |
903 | kfree(req->io); | |
09bb8394 JA |
904 | if (req->file && !(req->flags & REQ_F_FIXED_FILE)) |
905 | fput(req->file); | |
fcb323cc JA |
906 | if (req->flags & REQ_F_INFLIGHT) { |
907 | unsigned long flags; | |
908 | ||
909 | spin_lock_irqsave(&ctx->inflight_lock, flags); | |
910 | list_del(&req->inflight_entry); | |
911 | if (waitqueue_active(&ctx->inflight_wait)) | |
912 | wake_up(&ctx->inflight_wait); | |
913 | spin_unlock_irqrestore(&ctx->inflight_lock, flags); | |
914 | } | |
915 | percpu_ref_put(&ctx->refs); | |
0ddf92e8 JA |
916 | if (likely(!io_is_fallback_req(req))) |
917 | kmem_cache_free(req_cachep, req); | |
918 | else | |
919 | clear_bit_unlock(0, (unsigned long *) ctx->fallback_req); | |
e65ef56d JA |
920 | } |
921 | ||
a197f664 | 922 | static bool io_link_cancel_timeout(struct io_kiocb *req) |
2665abfd | 923 | { |
a197f664 | 924 | struct io_ring_ctx *ctx = req->ctx; |
2665abfd JA |
925 | int ret; |
926 | ||
2d28390a | 927 | ret = hrtimer_try_to_cancel(&req->io->timeout.timer); |
2665abfd | 928 | if (ret != -1) { |
78e19bbe | 929 | io_cqring_fill_event(req, -ECANCELED); |
2665abfd JA |
930 | io_commit_cqring(ctx); |
931 | req->flags &= ~REQ_F_LINK; | |
ec9c02ad | 932 | io_put_req(req); |
2665abfd JA |
933 | return true; |
934 | } | |
935 | ||
936 | return false; | |
e65ef56d JA |
937 | } |
938 | ||
ba816ad6 | 939 | static void io_req_link_next(struct io_kiocb *req, struct io_kiocb **nxtptr) |
9e645e11 | 940 | { |
2665abfd | 941 | struct io_ring_ctx *ctx = req->ctx; |
2665abfd | 942 | bool wake_ev = false; |
9e645e11 | 943 | |
4d7dd462 JA |
944 | /* Already got next link */ |
945 | if (req->flags & REQ_F_LINK_NEXT) | |
946 | return; | |
947 | ||
9e645e11 JA |
948 | /* |
949 | * The list should never be empty when we are called here. But could | |
950 | * potentially happen if the chain is messed up, check to be on the | |
951 | * safe side. | |
952 | */ | |
4493233e PB |
953 | while (!list_empty(&req->link_list)) { |
954 | struct io_kiocb *nxt = list_first_entry(&req->link_list, | |
955 | struct io_kiocb, link_list); | |
94ae5e77 | 956 | |
4493233e PB |
957 | if (unlikely((req->flags & REQ_F_LINK_TIMEOUT) && |
958 | (nxt->flags & REQ_F_TIMEOUT))) { | |
959 | list_del_init(&nxt->link_list); | |
94ae5e77 | 960 | wake_ev |= io_link_cancel_timeout(nxt); |
94ae5e77 JA |
961 | req->flags &= ~REQ_F_LINK_TIMEOUT; |
962 | continue; | |
963 | } | |
9e645e11 | 964 | |
4493233e PB |
965 | list_del_init(&req->link_list); |
966 | if (!list_empty(&nxt->link_list)) | |
967 | nxt->flags |= REQ_F_LINK; | |
b18fdf71 | 968 | *nxtptr = nxt; |
94ae5e77 | 969 | break; |
9e645e11 | 970 | } |
2665abfd | 971 | |
4d7dd462 | 972 | req->flags |= REQ_F_LINK_NEXT; |
2665abfd JA |
973 | if (wake_ev) |
974 | io_cqring_ev_posted(ctx); | |
9e645e11 JA |
975 | } |
976 | ||
977 | /* | |
978 | * Called if REQ_F_LINK is set, and we fail the head request | |
979 | */ | |
980 | static void io_fail_links(struct io_kiocb *req) | |
981 | { | |
2665abfd | 982 | struct io_ring_ctx *ctx = req->ctx; |
2665abfd JA |
983 | unsigned long flags; |
984 | ||
985 | spin_lock_irqsave(&ctx->completion_lock, flags); | |
9e645e11 JA |
986 | |
987 | while (!list_empty(&req->link_list)) { | |
4493233e PB |
988 | struct io_kiocb *link = list_first_entry(&req->link_list, |
989 | struct io_kiocb, link_list); | |
9e645e11 | 990 | |
4493233e | 991 | list_del_init(&link->link_list); |
c826bd7a | 992 | trace_io_uring_fail_link(req, link); |
2665abfd JA |
993 | |
994 | if ((req->flags & REQ_F_LINK_TIMEOUT) && | |
cf6fd4bd | 995 | link->sqe->opcode == IORING_OP_LINK_TIMEOUT) { |
a197f664 | 996 | io_link_cancel_timeout(link); |
2665abfd | 997 | } else { |
78e19bbe | 998 | io_cqring_fill_event(link, -ECANCELED); |
978db57e | 999 | __io_double_put_req(link); |
2665abfd | 1000 | } |
5d960724 | 1001 | req->flags &= ~REQ_F_LINK_TIMEOUT; |
9e645e11 | 1002 | } |
2665abfd JA |
1003 | |
1004 | io_commit_cqring(ctx); | |
1005 | spin_unlock_irqrestore(&ctx->completion_lock, flags); | |
1006 | io_cqring_ev_posted(ctx); | |
9e645e11 JA |
1007 | } |
1008 | ||
4d7dd462 | 1009 | static void io_req_find_next(struct io_kiocb *req, struct io_kiocb **nxt) |
9e645e11 | 1010 | { |
4d7dd462 | 1011 | if (likely(!(req->flags & REQ_F_LINK))) |
2665abfd | 1012 | return; |
2665abfd | 1013 | |
9e645e11 JA |
1014 | /* |
1015 | * If LINK is set, we have dependent requests in this chain. If we | |
1016 | * didn't fail this request, queue the first one up, moving any other | |
1017 | * dependencies to the next request. In case of failure, fail the rest | |
1018 | * of the chain. | |
1019 | */ | |
2665abfd JA |
1020 | if (req->flags & REQ_F_FAIL_LINK) { |
1021 | io_fail_links(req); | |
7c9e7f0f JA |
1022 | } else if ((req->flags & (REQ_F_LINK_TIMEOUT | REQ_F_COMP_LOCKED)) == |
1023 | REQ_F_LINK_TIMEOUT) { | |
2665abfd JA |
1024 | struct io_ring_ctx *ctx = req->ctx; |
1025 | unsigned long flags; | |
1026 | ||
1027 | /* | |
1028 | * If this is a timeout link, we could be racing with the | |
1029 | * timeout timer. Grab the completion lock for this case to | |
7c9e7f0f | 1030 | * protect against that. |
2665abfd JA |
1031 | */ |
1032 | spin_lock_irqsave(&ctx->completion_lock, flags); | |
1033 | io_req_link_next(req, nxt); | |
1034 | spin_unlock_irqrestore(&ctx->completion_lock, flags); | |
1035 | } else { | |
1036 | io_req_link_next(req, nxt); | |
9e645e11 | 1037 | } |
4d7dd462 | 1038 | } |
9e645e11 | 1039 | |
c69f8dbe JL |
1040 | static void io_free_req(struct io_kiocb *req) |
1041 | { | |
944e58bf PB |
1042 | struct io_kiocb *nxt = NULL; |
1043 | ||
1044 | io_req_find_next(req, &nxt); | |
70cf9f32 | 1045 | __io_free_req(req); |
944e58bf PB |
1046 | |
1047 | if (nxt) | |
1048 | io_queue_async_work(nxt); | |
c69f8dbe JL |
1049 | } |
1050 | ||
ba816ad6 JA |
1051 | /* |
1052 | * Drop reference to request, return next in chain (if there is one) if this | |
1053 | * was the last reference to this request. | |
1054 | */ | |
f9bd67f6 | 1055 | __attribute__((nonnull)) |
ec9c02ad | 1056 | static void io_put_req_find_next(struct io_kiocb *req, struct io_kiocb **nxtptr) |
e65ef56d | 1057 | { |
f9bd67f6 | 1058 | io_req_find_next(req, nxtptr); |
4d7dd462 | 1059 | |
e65ef56d | 1060 | if (refcount_dec_and_test(&req->refs)) |
4d7dd462 | 1061 | __io_free_req(req); |
2b188cc1 JA |
1062 | } |
1063 | ||
e65ef56d JA |
1064 | static void io_put_req(struct io_kiocb *req) |
1065 | { | |
1066 | if (refcount_dec_and_test(&req->refs)) | |
1067 | io_free_req(req); | |
2b188cc1 JA |
1068 | } |
1069 | ||
978db57e JA |
1070 | /* |
1071 | * Must only be used if we don't need to care about links, usually from | |
1072 | * within the completion handling itself. | |
1073 | */ | |
1074 | static void __io_double_put_req(struct io_kiocb *req) | |
78e19bbe JA |
1075 | { |
1076 | /* drop both submit and complete references */ | |
1077 | if (refcount_sub_and_test(2, &req->refs)) | |
1078 | __io_free_req(req); | |
1079 | } | |
1080 | ||
978db57e JA |
1081 | static void io_double_put_req(struct io_kiocb *req) |
1082 | { | |
1083 | /* drop both submit and complete references */ | |
1084 | if (refcount_sub_and_test(2, &req->refs)) | |
1085 | io_free_req(req); | |
1086 | } | |
1087 | ||
1d7bb1d5 | 1088 | static unsigned io_cqring_events(struct io_ring_ctx *ctx, bool noflush) |
a3a0e43f | 1089 | { |
84f97dc2 JA |
1090 | struct io_rings *rings = ctx->rings; |
1091 | ||
1d7bb1d5 JA |
1092 | /* |
1093 | * noflush == true is from the waitqueue handler, just ensure we wake | |
1094 | * up the task, and the next invocation will flush the entries. We | |
1095 | * cannot safely to it from here. | |
1096 | */ | |
1097 | if (noflush && !list_empty(&ctx->cq_overflow_list)) | |
1098 | return -1U; | |
1099 | ||
1100 | io_cqring_overflow_flush(ctx, false); | |
1101 | ||
a3a0e43f JA |
1102 | /* See comment at the top of this file */ |
1103 | smp_rmb(); | |
75b28aff | 1104 | return READ_ONCE(rings->cq.tail) - READ_ONCE(rings->cq.head); |
a3a0e43f JA |
1105 | } |
1106 | ||
fb5ccc98 PB |
1107 | static inline unsigned int io_sqring_entries(struct io_ring_ctx *ctx) |
1108 | { | |
1109 | struct io_rings *rings = ctx->rings; | |
1110 | ||
1111 | /* make sure SQ entry isn't read before tail */ | |
1112 | return smp_load_acquire(&rings->sq.tail) - ctx->cached_sq_head; | |
1113 | } | |
1114 | ||
def596e9 JA |
1115 | /* |
1116 | * Find and free completed poll iocbs | |
1117 | */ | |
1118 | static void io_iopoll_complete(struct io_ring_ctx *ctx, unsigned int *nr_events, | |
1119 | struct list_head *done) | |
1120 | { | |
1121 | void *reqs[IO_IOPOLL_BATCH]; | |
1122 | struct io_kiocb *req; | |
09bb8394 | 1123 | int to_free; |
def596e9 | 1124 | |
09bb8394 | 1125 | to_free = 0; |
def596e9 JA |
1126 | while (!list_empty(done)) { |
1127 | req = list_first_entry(done, struct io_kiocb, list); | |
1128 | list_del(&req->list); | |
1129 | ||
78e19bbe | 1130 | io_cqring_fill_event(req, req->result); |
def596e9 JA |
1131 | (*nr_events)++; |
1132 | ||
09bb8394 JA |
1133 | if (refcount_dec_and_test(&req->refs)) { |
1134 | /* If we're not using fixed files, we have to pair the | |
1135 | * completion part with the file put. Use regular | |
1136 | * completions for those, only batch free for fixed | |
9e645e11 | 1137 | * file and non-linked commands. |
09bb8394 | 1138 | */ |
1a6b74fc JA |
1139 | if (((req->flags & (REQ_F_FIXED_FILE|REQ_F_LINK)) == |
1140 | REQ_F_FIXED_FILE) && !io_is_fallback_req(req) && | |
1141 | !req->io) { | |
09bb8394 JA |
1142 | reqs[to_free++] = req; |
1143 | if (to_free == ARRAY_SIZE(reqs)) | |
1144 | io_free_req_many(ctx, reqs, &to_free); | |
6b06314c | 1145 | } else { |
09bb8394 | 1146 | io_free_req(req); |
6b06314c | 1147 | } |
9a56a232 | 1148 | } |
def596e9 | 1149 | } |
def596e9 | 1150 | |
09bb8394 | 1151 | io_commit_cqring(ctx); |
def596e9 JA |
1152 | io_free_req_many(ctx, reqs, &to_free); |
1153 | } | |
1154 | ||
1155 | static int io_do_iopoll(struct io_ring_ctx *ctx, unsigned int *nr_events, | |
1156 | long min) | |
1157 | { | |
1158 | struct io_kiocb *req, *tmp; | |
1159 | LIST_HEAD(done); | |
1160 | bool spin; | |
1161 | int ret; | |
1162 | ||
1163 | /* | |
1164 | * Only spin for completions if we don't have multiple devices hanging | |
1165 | * off our complete list, and we're under the requested amount. | |
1166 | */ | |
1167 | spin = !ctx->poll_multi_file && *nr_events < min; | |
1168 | ||
1169 | ret = 0; | |
1170 | list_for_each_entry_safe(req, tmp, &ctx->poll_list, list) { | |
1171 | struct kiocb *kiocb = &req->rw; | |
1172 | ||
1173 | /* | |
1174 | * Move completed entries to our local list. If we find a | |
1175 | * request that requires polling, break out and complete | |
1176 | * the done list first, if we have entries there. | |
1177 | */ | |
1178 | if (req->flags & REQ_F_IOPOLL_COMPLETED) { | |
1179 | list_move_tail(&req->list, &done); | |
1180 | continue; | |
1181 | } | |
1182 | if (!list_empty(&done)) | |
1183 | break; | |
1184 | ||
1185 | ret = kiocb->ki_filp->f_op->iopoll(kiocb, spin); | |
1186 | if (ret < 0) | |
1187 | break; | |
1188 | ||
1189 | if (ret && spin) | |
1190 | spin = false; | |
1191 | ret = 0; | |
1192 | } | |
1193 | ||
1194 | if (!list_empty(&done)) | |
1195 | io_iopoll_complete(ctx, nr_events, &done); | |
1196 | ||
1197 | return ret; | |
1198 | } | |
1199 | ||
1200 | /* | |
d195a66e | 1201 | * Poll for a minimum of 'min' events. Note that if min == 0 we consider that a |
def596e9 JA |
1202 | * non-spinning poll check - we'll still enter the driver poll loop, but only |
1203 | * as a non-spinning completion check. | |
1204 | */ | |
1205 | static int io_iopoll_getevents(struct io_ring_ctx *ctx, unsigned int *nr_events, | |
1206 | long min) | |
1207 | { | |
08f5439f | 1208 | while (!list_empty(&ctx->poll_list) && !need_resched()) { |
def596e9 JA |
1209 | int ret; |
1210 | ||
1211 | ret = io_do_iopoll(ctx, nr_events, min); | |
1212 | if (ret < 0) | |
1213 | return ret; | |
1214 | if (!min || *nr_events >= min) | |
1215 | return 0; | |
1216 | } | |
1217 | ||
1218 | return 1; | |
1219 | } | |
1220 | ||
1221 | /* | |
1222 | * We can't just wait for polled events to come to us, we have to actively | |
1223 | * find and complete them. | |
1224 | */ | |
1225 | static void io_iopoll_reap_events(struct io_ring_ctx *ctx) | |
1226 | { | |
1227 | if (!(ctx->flags & IORING_SETUP_IOPOLL)) | |
1228 | return; | |
1229 | ||
1230 | mutex_lock(&ctx->uring_lock); | |
1231 | while (!list_empty(&ctx->poll_list)) { | |
1232 | unsigned int nr_events = 0; | |
1233 | ||
1234 | io_iopoll_getevents(ctx, &nr_events, 1); | |
08f5439f JA |
1235 | |
1236 | /* | |
1237 | * Ensure we allow local-to-the-cpu processing to take place, | |
1238 | * in this case we need to ensure that we reap all events. | |
1239 | */ | |
1240 | cond_resched(); | |
def596e9 JA |
1241 | } |
1242 | mutex_unlock(&ctx->uring_lock); | |
1243 | } | |
1244 | ||
2b2ed975 JA |
1245 | static int __io_iopoll_check(struct io_ring_ctx *ctx, unsigned *nr_events, |
1246 | long min) | |
def596e9 | 1247 | { |
2b2ed975 | 1248 | int iters = 0, ret = 0; |
500f9fba | 1249 | |
def596e9 JA |
1250 | do { |
1251 | int tmin = 0; | |
1252 | ||
a3a0e43f JA |
1253 | /* |
1254 | * Don't enter poll loop if we already have events pending. | |
1255 | * If we do, we can potentially be spinning for commands that | |
1256 | * already triggered a CQE (eg in error). | |
1257 | */ | |
1d7bb1d5 | 1258 | if (io_cqring_events(ctx, false)) |
a3a0e43f JA |
1259 | break; |
1260 | ||
500f9fba JA |
1261 | /* |
1262 | * If a submit got punted to a workqueue, we can have the | |
1263 | * application entering polling for a command before it gets | |
1264 | * issued. That app will hold the uring_lock for the duration | |
1265 | * of the poll right here, so we need to take a breather every | |
1266 | * now and then to ensure that the issue has a chance to add | |
1267 | * the poll to the issued list. Otherwise we can spin here | |
1268 | * forever, while the workqueue is stuck trying to acquire the | |
1269 | * very same mutex. | |
1270 | */ | |
1271 | if (!(++iters & 7)) { | |
1272 | mutex_unlock(&ctx->uring_lock); | |
1273 | mutex_lock(&ctx->uring_lock); | |
1274 | } | |
1275 | ||
def596e9 JA |
1276 | if (*nr_events < min) |
1277 | tmin = min - *nr_events; | |
1278 | ||
1279 | ret = io_iopoll_getevents(ctx, nr_events, tmin); | |
1280 | if (ret <= 0) | |
1281 | break; | |
1282 | ret = 0; | |
1283 | } while (min && !*nr_events && !need_resched()); | |
1284 | ||
2b2ed975 JA |
1285 | return ret; |
1286 | } | |
1287 | ||
1288 | static int io_iopoll_check(struct io_ring_ctx *ctx, unsigned *nr_events, | |
1289 | long min) | |
1290 | { | |
1291 | int ret; | |
1292 | ||
1293 | /* | |
1294 | * We disallow the app entering submit/complete with polling, but we | |
1295 | * still need to lock the ring to prevent racing with polled issue | |
1296 | * that got punted to a workqueue. | |
1297 | */ | |
1298 | mutex_lock(&ctx->uring_lock); | |
1299 | ret = __io_iopoll_check(ctx, nr_events, min); | |
500f9fba | 1300 | mutex_unlock(&ctx->uring_lock); |
def596e9 JA |
1301 | return ret; |
1302 | } | |
1303 | ||
491381ce | 1304 | static void kiocb_end_write(struct io_kiocb *req) |
2b188cc1 | 1305 | { |
491381ce JA |
1306 | /* |
1307 | * Tell lockdep we inherited freeze protection from submission | |
1308 | * thread. | |
1309 | */ | |
1310 | if (req->flags & REQ_F_ISREG) { | |
1311 | struct inode *inode = file_inode(req->file); | |
2b188cc1 | 1312 | |
491381ce | 1313 | __sb_writers_acquired(inode->i_sb, SB_FREEZE_WRITE); |
2b188cc1 | 1314 | } |
491381ce | 1315 | file_end_write(req->file); |
2b188cc1 JA |
1316 | } |
1317 | ||
4e88d6e7 JA |
1318 | static inline void req_set_fail_links(struct io_kiocb *req) |
1319 | { | |
1320 | if ((req->flags & (REQ_F_LINK | REQ_F_HARDLINK)) == REQ_F_LINK) | |
1321 | req->flags |= REQ_F_FAIL_LINK; | |
1322 | } | |
1323 | ||
ba816ad6 | 1324 | static void io_complete_rw_common(struct kiocb *kiocb, long res) |
2b188cc1 JA |
1325 | { |
1326 | struct io_kiocb *req = container_of(kiocb, struct io_kiocb, rw); | |
1327 | ||
491381ce JA |
1328 | if (kiocb->ki_flags & IOCB_WRITE) |
1329 | kiocb_end_write(req); | |
2b188cc1 | 1330 | |
4e88d6e7 JA |
1331 | if (res != req->result) |
1332 | req_set_fail_links(req); | |
78e19bbe | 1333 | io_cqring_add_event(req, res); |
ba816ad6 JA |
1334 | } |
1335 | ||
1336 | static void io_complete_rw(struct kiocb *kiocb, long res, long res2) | |
1337 | { | |
1338 | struct io_kiocb *req = container_of(kiocb, struct io_kiocb, rw); | |
1339 | ||
1340 | io_complete_rw_common(kiocb, res); | |
e65ef56d | 1341 | io_put_req(req); |
2b188cc1 JA |
1342 | } |
1343 | ||
ba816ad6 JA |
1344 | static struct io_kiocb *__io_complete_rw(struct kiocb *kiocb, long res) |
1345 | { | |
1346 | struct io_kiocb *req = container_of(kiocb, struct io_kiocb, rw); | |
ec9c02ad | 1347 | struct io_kiocb *nxt = NULL; |
ba816ad6 JA |
1348 | |
1349 | io_complete_rw_common(kiocb, res); | |
ec9c02ad JL |
1350 | io_put_req_find_next(req, &nxt); |
1351 | ||
1352 | return nxt; | |
2b188cc1 JA |
1353 | } |
1354 | ||
def596e9 JA |
1355 | static void io_complete_rw_iopoll(struct kiocb *kiocb, long res, long res2) |
1356 | { | |
1357 | struct io_kiocb *req = container_of(kiocb, struct io_kiocb, rw); | |
1358 | ||
491381ce JA |
1359 | if (kiocb->ki_flags & IOCB_WRITE) |
1360 | kiocb_end_write(req); | |
def596e9 | 1361 | |
4e88d6e7 JA |
1362 | if (res != req->result) |
1363 | req_set_fail_links(req); | |
9e645e11 | 1364 | req->result = res; |
def596e9 JA |
1365 | if (res != -EAGAIN) |
1366 | req->flags |= REQ_F_IOPOLL_COMPLETED; | |
1367 | } | |
1368 | ||
1369 | /* | |
1370 | * After the iocb has been issued, it's safe to be found on the poll list. | |
1371 | * Adding the kiocb to the list AFTER submission ensures that we don't | |
1372 | * find it from a io_iopoll_getevents() thread before the issuer is done | |
1373 | * accessing the kiocb cookie. | |
1374 | */ | |
1375 | static void io_iopoll_req_issued(struct io_kiocb *req) | |
1376 | { | |
1377 | struct io_ring_ctx *ctx = req->ctx; | |
1378 | ||
1379 | /* | |
1380 | * Track whether we have multiple files in our lists. This will impact | |
1381 | * how we do polling eventually, not spinning if we're on potentially | |
1382 | * different devices. | |
1383 | */ | |
1384 | if (list_empty(&ctx->poll_list)) { | |
1385 | ctx->poll_multi_file = false; | |
1386 | } else if (!ctx->poll_multi_file) { | |
1387 | struct io_kiocb *list_req; | |
1388 | ||
1389 | list_req = list_first_entry(&ctx->poll_list, struct io_kiocb, | |
1390 | list); | |
1391 | if (list_req->rw.ki_filp != req->rw.ki_filp) | |
1392 | ctx->poll_multi_file = true; | |
1393 | } | |
1394 | ||
1395 | /* | |
1396 | * For fast devices, IO may have already completed. If it has, add | |
1397 | * it to the front so we find it first. | |
1398 | */ | |
1399 | if (req->flags & REQ_F_IOPOLL_COMPLETED) | |
1400 | list_add(&req->list, &ctx->poll_list); | |
1401 | else | |
1402 | list_add_tail(&req->list, &ctx->poll_list); | |
1403 | } | |
1404 | ||
3d6770fb | 1405 | static void io_file_put(struct io_submit_state *state) |
9a56a232 | 1406 | { |
3d6770fb | 1407 | if (state->file) { |
9a56a232 JA |
1408 | int diff = state->has_refs - state->used_refs; |
1409 | ||
1410 | if (diff) | |
1411 | fput_many(state->file, diff); | |
1412 | state->file = NULL; | |
1413 | } | |
1414 | } | |
1415 | ||
1416 | /* | |
1417 | * Get as many references to a file as we have IOs left in this submission, | |
1418 | * assuming most submissions are for one file, or at least that each file | |
1419 | * has more than one submission. | |
1420 | */ | |
1421 | static struct file *io_file_get(struct io_submit_state *state, int fd) | |
1422 | { | |
1423 | if (!state) | |
1424 | return fget(fd); | |
1425 | ||
1426 | if (state->file) { | |
1427 | if (state->fd == fd) { | |
1428 | state->used_refs++; | |
1429 | state->ios_left--; | |
1430 | return state->file; | |
1431 | } | |
3d6770fb | 1432 | io_file_put(state); |
9a56a232 JA |
1433 | } |
1434 | state->file = fget_many(fd, state->ios_left); | |
1435 | if (!state->file) | |
1436 | return NULL; | |
1437 | ||
1438 | state->fd = fd; | |
1439 | state->has_refs = state->ios_left; | |
1440 | state->used_refs = 1; | |
1441 | state->ios_left--; | |
1442 | return state->file; | |
1443 | } | |
1444 | ||
2b188cc1 JA |
1445 | /* |
1446 | * If we tracked the file through the SCM inflight mechanism, we could support | |
1447 | * any file. For now, just ensure that anything potentially problematic is done | |
1448 | * inline. | |
1449 | */ | |
1450 | static bool io_file_supports_async(struct file *file) | |
1451 | { | |
1452 | umode_t mode = file_inode(file)->i_mode; | |
1453 | ||
10d59345 | 1454 | if (S_ISBLK(mode) || S_ISCHR(mode) || S_ISSOCK(mode)) |
2b188cc1 JA |
1455 | return true; |
1456 | if (S_ISREG(mode) && file->f_op != &io_uring_fops) | |
1457 | return true; | |
1458 | ||
1459 | return false; | |
1460 | } | |
1461 | ||
267bc904 | 1462 | static int io_prep_rw(struct io_kiocb *req, bool force_nonblock) |
2b188cc1 | 1463 | { |
cf6fd4bd | 1464 | const struct io_uring_sqe *sqe = req->sqe; |
def596e9 | 1465 | struct io_ring_ctx *ctx = req->ctx; |
2b188cc1 | 1466 | struct kiocb *kiocb = &req->rw; |
09bb8394 JA |
1467 | unsigned ioprio; |
1468 | int ret; | |
2b188cc1 | 1469 | |
09bb8394 JA |
1470 | if (!req->file) |
1471 | return -EBADF; | |
2b188cc1 | 1472 | |
491381ce JA |
1473 | if (S_ISREG(file_inode(req->file)->i_mode)) |
1474 | req->flags |= REQ_F_ISREG; | |
1475 | ||
2b188cc1 JA |
1476 | kiocb->ki_pos = READ_ONCE(sqe->off); |
1477 | kiocb->ki_flags = iocb_flags(kiocb->ki_filp); | |
1478 | kiocb->ki_hint = ki_hint_validate(file_write_hint(kiocb->ki_filp)); | |
1479 | ||
1480 | ioprio = READ_ONCE(sqe->ioprio); | |
1481 | if (ioprio) { | |
1482 | ret = ioprio_check_cap(ioprio); | |
1483 | if (ret) | |
09bb8394 | 1484 | return ret; |
2b188cc1 JA |
1485 | |
1486 | kiocb->ki_ioprio = ioprio; | |
1487 | } else | |
1488 | kiocb->ki_ioprio = get_current_ioprio(); | |
1489 | ||
1490 | ret = kiocb_set_rw_flags(kiocb, READ_ONCE(sqe->rw_flags)); | |
1491 | if (unlikely(ret)) | |
09bb8394 | 1492 | return ret; |
8449eeda SB |
1493 | |
1494 | /* don't allow async punt if RWF_NOWAIT was requested */ | |
491381ce JA |
1495 | if ((kiocb->ki_flags & IOCB_NOWAIT) || |
1496 | (req->file->f_flags & O_NONBLOCK)) | |
8449eeda SB |
1497 | req->flags |= REQ_F_NOWAIT; |
1498 | ||
1499 | if (force_nonblock) | |
2b188cc1 | 1500 | kiocb->ki_flags |= IOCB_NOWAIT; |
8449eeda | 1501 | |
def596e9 | 1502 | if (ctx->flags & IORING_SETUP_IOPOLL) { |
def596e9 JA |
1503 | if (!(kiocb->ki_flags & IOCB_DIRECT) || |
1504 | !kiocb->ki_filp->f_op->iopoll) | |
09bb8394 | 1505 | return -EOPNOTSUPP; |
2b188cc1 | 1506 | |
def596e9 JA |
1507 | kiocb->ki_flags |= IOCB_HIPRI; |
1508 | kiocb->ki_complete = io_complete_rw_iopoll; | |
6873e0bd | 1509 | req->result = 0; |
def596e9 | 1510 | } else { |
09bb8394 JA |
1511 | if (kiocb->ki_flags & IOCB_HIPRI) |
1512 | return -EINVAL; | |
def596e9 JA |
1513 | kiocb->ki_complete = io_complete_rw; |
1514 | } | |
2b188cc1 | 1515 | return 0; |
2b188cc1 JA |
1516 | } |
1517 | ||
1518 | static inline void io_rw_done(struct kiocb *kiocb, ssize_t ret) | |
1519 | { | |
1520 | switch (ret) { | |
1521 | case -EIOCBQUEUED: | |
1522 | break; | |
1523 | case -ERESTARTSYS: | |
1524 | case -ERESTARTNOINTR: | |
1525 | case -ERESTARTNOHAND: | |
1526 | case -ERESTART_RESTARTBLOCK: | |
1527 | /* | |
1528 | * We can't just restart the syscall, since previously | |
1529 | * submitted sqes may already be in progress. Just fail this | |
1530 | * IO with EINTR. | |
1531 | */ | |
1532 | ret = -EINTR; | |
1533 | /* fall through */ | |
1534 | default: | |
1535 | kiocb->ki_complete(kiocb, ret, 0); | |
1536 | } | |
1537 | } | |
1538 | ||
ba816ad6 JA |
1539 | static void kiocb_done(struct kiocb *kiocb, ssize_t ret, struct io_kiocb **nxt, |
1540 | bool in_async) | |
1541 | { | |
f9bd67f6 | 1542 | if (in_async && ret >= 0 && kiocb->ki_complete == io_complete_rw) |
ba816ad6 JA |
1543 | *nxt = __io_complete_rw(kiocb, ret); |
1544 | else | |
1545 | io_rw_done(kiocb, ret); | |
1546 | } | |
1547 | ||
7d009165 PB |
1548 | static ssize_t io_import_fixed(struct io_ring_ctx *ctx, int rw, |
1549 | const struct io_uring_sqe *sqe, | |
1550 | struct iov_iter *iter) | |
edafccee JA |
1551 | { |
1552 | size_t len = READ_ONCE(sqe->len); | |
1553 | struct io_mapped_ubuf *imu; | |
1554 | unsigned index, buf_index; | |
1555 | size_t offset; | |
1556 | u64 buf_addr; | |
1557 | ||
1558 | /* attempt to use fixed buffers without having provided iovecs */ | |
1559 | if (unlikely(!ctx->user_bufs)) | |
1560 | return -EFAULT; | |
1561 | ||
1562 | buf_index = READ_ONCE(sqe->buf_index); | |
1563 | if (unlikely(buf_index >= ctx->nr_user_bufs)) | |
1564 | return -EFAULT; | |
1565 | ||
1566 | index = array_index_nospec(buf_index, ctx->nr_user_bufs); | |
1567 | imu = &ctx->user_bufs[index]; | |
1568 | buf_addr = READ_ONCE(sqe->addr); | |
1569 | ||
1570 | /* overflow */ | |
1571 | if (buf_addr + len < buf_addr) | |
1572 | return -EFAULT; | |
1573 | /* not inside the mapped region */ | |
1574 | if (buf_addr < imu->ubuf || buf_addr + len > imu->ubuf + imu->len) | |
1575 | return -EFAULT; | |
1576 | ||
1577 | /* | |
1578 | * May not be a start of buffer, set size appropriately | |
1579 | * and advance us to the beginning. | |
1580 | */ | |
1581 | offset = buf_addr - imu->ubuf; | |
1582 | iov_iter_bvec(iter, rw, imu->bvec, imu->nr_bvecs, offset + len); | |
bd11b3a3 JA |
1583 | |
1584 | if (offset) { | |
1585 | /* | |
1586 | * Don't use iov_iter_advance() here, as it's really slow for | |
1587 | * using the latter parts of a big fixed buffer - it iterates | |
1588 | * over each segment manually. We can cheat a bit here, because | |
1589 | * we know that: | |
1590 | * | |
1591 | * 1) it's a BVEC iter, we set it up | |
1592 | * 2) all bvecs are PAGE_SIZE in size, except potentially the | |
1593 | * first and last bvec | |
1594 | * | |
1595 | * So just find our index, and adjust the iterator afterwards. | |
1596 | * If the offset is within the first bvec (or the whole first | |
1597 | * bvec, just use iov_iter_advance(). This makes it easier | |
1598 | * since we can just skip the first segment, which may not | |
1599 | * be PAGE_SIZE aligned. | |
1600 | */ | |
1601 | const struct bio_vec *bvec = imu->bvec; | |
1602 | ||
1603 | if (offset <= bvec->bv_len) { | |
1604 | iov_iter_advance(iter, offset); | |
1605 | } else { | |
1606 | unsigned long seg_skip; | |
1607 | ||
1608 | /* skip first vec */ | |
1609 | offset -= bvec->bv_len; | |
1610 | seg_skip = 1 + (offset >> PAGE_SHIFT); | |
1611 | ||
1612 | iter->bvec = bvec + seg_skip; | |
1613 | iter->nr_segs -= seg_skip; | |
99c79f66 | 1614 | iter->count -= bvec->bv_len + offset; |
bd11b3a3 | 1615 | iter->iov_offset = offset & ~PAGE_MASK; |
bd11b3a3 JA |
1616 | } |
1617 | } | |
1618 | ||
5e559561 | 1619 | return len; |
edafccee JA |
1620 | } |
1621 | ||
cf6fd4bd PB |
1622 | static ssize_t io_import_iovec(int rw, struct io_kiocb *req, |
1623 | struct iovec **iovec, struct iov_iter *iter) | |
2b188cc1 | 1624 | { |
cf6fd4bd | 1625 | const struct io_uring_sqe *sqe = req->sqe; |
2b188cc1 JA |
1626 | void __user *buf = u64_to_user_ptr(READ_ONCE(sqe->addr)); |
1627 | size_t sqe_len = READ_ONCE(sqe->len); | |
edafccee JA |
1628 | u8 opcode; |
1629 | ||
1630 | /* | |
1631 | * We're reading ->opcode for the second time, but the first read | |
1632 | * doesn't care whether it's _FIXED or not, so it doesn't matter | |
1633 | * whether ->opcode changes concurrently. The first read does care | |
1634 | * about whether it is a READ or a WRITE, so we don't trust this read | |
1635 | * for that purpose and instead let the caller pass in the read/write | |
1636 | * flag. | |
1637 | */ | |
1638 | opcode = READ_ONCE(sqe->opcode); | |
7d009165 | 1639 | if (opcode == IORING_OP_READ_FIXED || opcode == IORING_OP_WRITE_FIXED) { |
edafccee | 1640 | *iovec = NULL; |
7d009165 | 1641 | return io_import_fixed(req->ctx, rw, sqe, iter); |
edafccee | 1642 | } |
2b188cc1 | 1643 | |
f67676d1 JA |
1644 | if (req->io) { |
1645 | struct io_async_rw *iorw = &req->io->rw; | |
1646 | ||
1647 | *iovec = iorw->iov; | |
1648 | iov_iter_init(iter, rw, *iovec, iorw->nr_segs, iorw->size); | |
1649 | if (iorw->iov == iorw->fast_iov) | |
1650 | *iovec = NULL; | |
1651 | return iorw->size; | |
1652 | } | |
1653 | ||
cf6fd4bd | 1654 | if (!req->has_user) |
2b188cc1 JA |
1655 | return -EFAULT; |
1656 | ||
1657 | #ifdef CONFIG_COMPAT | |
cf6fd4bd | 1658 | if (req->ctx->compat) |
2b188cc1 JA |
1659 | return compat_import_iovec(rw, buf, sqe_len, UIO_FASTIOV, |
1660 | iovec, iter); | |
1661 | #endif | |
1662 | ||
1663 | return import_iovec(rw, buf, sqe_len, UIO_FASTIOV, iovec, iter); | |
1664 | } | |
1665 | ||
31b51510 | 1666 | /* |
32960613 JA |
1667 | * For files that don't have ->read_iter() and ->write_iter(), handle them |
1668 | * by looping over ->read() or ->write() manually. | |
31b51510 | 1669 | */ |
32960613 JA |
1670 | static ssize_t loop_rw_iter(int rw, struct file *file, struct kiocb *kiocb, |
1671 | struct iov_iter *iter) | |
1672 | { | |
1673 | ssize_t ret = 0; | |
1674 | ||
1675 | /* | |
1676 | * Don't support polled IO through this interface, and we can't | |
1677 | * support non-blocking either. For the latter, this just causes | |
1678 | * the kiocb to be handled from an async context. | |
1679 | */ | |
1680 | if (kiocb->ki_flags & IOCB_HIPRI) | |
1681 | return -EOPNOTSUPP; | |
1682 | if (kiocb->ki_flags & IOCB_NOWAIT) | |
1683 | return -EAGAIN; | |
1684 | ||
1685 | while (iov_iter_count(iter)) { | |
311ae9e1 | 1686 | struct iovec iovec; |
32960613 JA |
1687 | ssize_t nr; |
1688 | ||
311ae9e1 PB |
1689 | if (!iov_iter_is_bvec(iter)) { |
1690 | iovec = iov_iter_iovec(iter); | |
1691 | } else { | |
1692 | /* fixed buffers import bvec */ | |
1693 | iovec.iov_base = kmap(iter->bvec->bv_page) | |
1694 | + iter->iov_offset; | |
1695 | iovec.iov_len = min(iter->count, | |
1696 | iter->bvec->bv_len - iter->iov_offset); | |
1697 | } | |
1698 | ||
32960613 JA |
1699 | if (rw == READ) { |
1700 | nr = file->f_op->read(file, iovec.iov_base, | |
1701 | iovec.iov_len, &kiocb->ki_pos); | |
1702 | } else { | |
1703 | nr = file->f_op->write(file, iovec.iov_base, | |
1704 | iovec.iov_len, &kiocb->ki_pos); | |
1705 | } | |
1706 | ||
311ae9e1 PB |
1707 | if (iov_iter_is_bvec(iter)) |
1708 | kunmap(iter->bvec->bv_page); | |
1709 | ||
32960613 JA |
1710 | if (nr < 0) { |
1711 | if (!ret) | |
1712 | ret = nr; | |
1713 | break; | |
1714 | } | |
1715 | ret += nr; | |
1716 | if (nr != iovec.iov_len) | |
1717 | break; | |
1718 | iov_iter_advance(iter, nr); | |
1719 | } | |
1720 | ||
1721 | return ret; | |
1722 | } | |
1723 | ||
b7bb4f7d | 1724 | static void io_req_map_rw(struct io_kiocb *req, ssize_t io_size, |
f67676d1 JA |
1725 | struct iovec *iovec, struct iovec *fast_iov, |
1726 | struct iov_iter *iter) | |
1727 | { | |
1728 | req->io->rw.nr_segs = iter->nr_segs; | |
1729 | req->io->rw.size = io_size; | |
1730 | req->io->rw.iov = iovec; | |
1731 | if (!req->io->rw.iov) { | |
1732 | req->io->rw.iov = req->io->rw.fast_iov; | |
1733 | memcpy(req->io->rw.iov, fast_iov, | |
1734 | sizeof(struct iovec) * iter->nr_segs); | |
1735 | } | |
1736 | } | |
1737 | ||
b7bb4f7d | 1738 | static int io_alloc_async_ctx(struct io_kiocb *req) |
f67676d1 JA |
1739 | { |
1740 | req->io = kmalloc(sizeof(*req->io), GFP_KERNEL); | |
1741 | if (req->io) { | |
f67676d1 JA |
1742 | memcpy(&req->io->sqe, req->sqe, sizeof(req->io->sqe)); |
1743 | req->sqe = &req->io->sqe; | |
1744 | return 0; | |
1745 | } | |
1746 | ||
b7bb4f7d JA |
1747 | return 1; |
1748 | } | |
1749 | ||
1750 | static void io_rw_async(struct io_wq_work **workptr) | |
1751 | { | |
1752 | struct io_kiocb *req = container_of(*workptr, struct io_kiocb, work); | |
1753 | struct iovec *iov = NULL; | |
1754 | ||
1755 | if (req->io->rw.iov != req->io->rw.fast_iov) | |
1756 | iov = req->io->rw.iov; | |
1757 | io_wq_submit_work(workptr); | |
1758 | kfree(iov); | |
1759 | } | |
1760 | ||
1761 | static int io_setup_async_rw(struct io_kiocb *req, ssize_t io_size, | |
1762 | struct iovec *iovec, struct iovec *fast_iov, | |
1763 | struct iov_iter *iter) | |
1764 | { | |
1765 | if (!req->io && io_alloc_async_ctx(req)) | |
1766 | return -ENOMEM; | |
1767 | ||
1768 | io_req_map_rw(req, io_size, iovec, fast_iov, iter); | |
1769 | req->work.func = io_rw_async; | |
1770 | return 0; | |
f67676d1 JA |
1771 | } |
1772 | ||
1773 | static int io_read_prep(struct io_kiocb *req, struct iovec **iovec, | |
1774 | struct iov_iter *iter, bool force_nonblock) | |
1775 | { | |
1776 | ssize_t ret; | |
1777 | ||
1778 | ret = io_prep_rw(req, force_nonblock); | |
1779 | if (ret) | |
1780 | return ret; | |
1781 | ||
1782 | if (unlikely(!(req->file->f_mode & FMODE_READ))) | |
1783 | return -EBADF; | |
1784 | ||
1785 | return io_import_iovec(READ, req, iovec, iter); | |
1786 | } | |
1787 | ||
267bc904 | 1788 | static int io_read(struct io_kiocb *req, struct io_kiocb **nxt, |
8358e3a8 | 1789 | bool force_nonblock) |
2b188cc1 JA |
1790 | { |
1791 | struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs; | |
1792 | struct kiocb *kiocb = &req->rw; | |
1793 | struct iov_iter iter; | |
1794 | struct file *file; | |
31b51510 | 1795 | size_t iov_count; |
f67676d1 | 1796 | ssize_t io_size, ret; |
2b188cc1 | 1797 | |
f67676d1 JA |
1798 | if (!req->io) { |
1799 | ret = io_read_prep(req, &iovec, &iter, force_nonblock); | |
1800 | if (ret < 0) | |
1801 | return ret; | |
1802 | } else { | |
1803 | ret = io_import_iovec(READ, req, &iovec, &iter); | |
1804 | if (ret < 0) | |
1805 | return ret; | |
1806 | } | |
2b188cc1 | 1807 | |
f67676d1 JA |
1808 | file = req->file; |
1809 | io_size = ret; | |
9e645e11 | 1810 | if (req->flags & REQ_F_LINK) |
f67676d1 JA |
1811 | req->result = io_size; |
1812 | ||
1813 | /* | |
1814 | * If the file doesn't support async, mark it as REQ_F_MUST_PUNT so | |
1815 | * we know to async punt it even if it was opened O_NONBLOCK | |
1816 | */ | |
1817 | if (force_nonblock && !io_file_supports_async(file)) { | |
1818 | req->flags |= REQ_F_MUST_PUNT; | |
1819 | goto copy_iov; | |
1820 | } | |
9e645e11 | 1821 | |
31b51510 JA |
1822 | iov_count = iov_iter_count(&iter); |
1823 | ret = rw_verify_area(READ, file, &kiocb->ki_pos, iov_count); | |
2b188cc1 JA |
1824 | if (!ret) { |
1825 | ssize_t ret2; | |
1826 | ||
32960613 JA |
1827 | if (file->f_op->read_iter) |
1828 | ret2 = call_read_iter(file, kiocb, &iter); | |
1829 | else | |
1830 | ret2 = loop_rw_iter(READ, file, kiocb, &iter); | |
1831 | ||
9d93a3f5 JA |
1832 | /* |
1833 | * In case of a short read, punt to async. This can happen | |
1834 | * if we have data partially cached. Alternatively we can | |
1835 | * return the short read, in which case the application will | |
1836 | * need to issue another SQE and wait for it. That SQE will | |
1837 | * need async punt anyway, so it's more efficient to do it | |
1838 | * here. | |
1839 | */ | |
491381ce JA |
1840 | if (force_nonblock && !(req->flags & REQ_F_NOWAIT) && |
1841 | (req->flags & REQ_F_ISREG) && | |
f67676d1 | 1842 | ret2 > 0 && ret2 < io_size) |
9d93a3f5 JA |
1843 | ret2 = -EAGAIN; |
1844 | /* Catch -EAGAIN return for forced non-blocking submission */ | |
f67676d1 | 1845 | if (!force_nonblock || ret2 != -EAGAIN) { |
cf6fd4bd | 1846 | kiocb_done(kiocb, ret2, nxt, req->in_async); |
f67676d1 JA |
1847 | } else { |
1848 | copy_iov: | |
b7bb4f7d | 1849 | ret = io_setup_async_rw(req, io_size, iovec, |
f67676d1 JA |
1850 | inline_vecs, &iter); |
1851 | if (ret) | |
1852 | goto out_free; | |
1853 | return -EAGAIN; | |
1854 | } | |
2b188cc1 | 1855 | } |
f67676d1 | 1856 | out_free: |
b7bb4f7d JA |
1857 | if (!io_wq_current_is_worker()) |
1858 | kfree(iovec); | |
2b188cc1 JA |
1859 | return ret; |
1860 | } | |
1861 | ||
f67676d1 JA |
1862 | static int io_write_prep(struct io_kiocb *req, struct iovec **iovec, |
1863 | struct iov_iter *iter, bool force_nonblock) | |
1864 | { | |
1865 | ssize_t ret; | |
1866 | ||
1867 | ret = io_prep_rw(req, force_nonblock); | |
1868 | if (ret) | |
1869 | return ret; | |
1870 | ||
1871 | if (unlikely(!(req->file->f_mode & FMODE_WRITE))) | |
1872 | return -EBADF; | |
1873 | ||
1874 | return io_import_iovec(WRITE, req, iovec, iter); | |
1875 | } | |
1876 | ||
267bc904 | 1877 | static int io_write(struct io_kiocb *req, struct io_kiocb **nxt, |
8358e3a8 | 1878 | bool force_nonblock) |
2b188cc1 JA |
1879 | { |
1880 | struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs; | |
1881 | struct kiocb *kiocb = &req->rw; | |
1882 | struct iov_iter iter; | |
1883 | struct file *file; | |
31b51510 | 1884 | size_t iov_count; |
f67676d1 | 1885 | ssize_t ret, io_size; |
2b188cc1 | 1886 | |
f67676d1 JA |
1887 | if (!req->io) { |
1888 | ret = io_write_prep(req, &iovec, &iter, force_nonblock); | |
1889 | if (ret < 0) | |
1890 | return ret; | |
1891 | } else { | |
1892 | ret = io_import_iovec(WRITE, req, &iovec, &iter); | |
1893 | if (ret < 0) | |
1894 | return ret; | |
1895 | } | |
2b188cc1 | 1896 | |
2b188cc1 | 1897 | file = kiocb->ki_filp; |
f67676d1 | 1898 | io_size = ret; |
9e645e11 | 1899 | if (req->flags & REQ_F_LINK) |
f67676d1 | 1900 | req->result = io_size; |
9e645e11 | 1901 | |
f67676d1 JA |
1902 | /* |
1903 | * If the file doesn't support async, mark it as REQ_F_MUST_PUNT so | |
1904 | * we know to async punt it even if it was opened O_NONBLOCK | |
1905 | */ | |
1906 | if (force_nonblock && !io_file_supports_async(req->file)) { | |
1907 | req->flags |= REQ_F_MUST_PUNT; | |
1908 | goto copy_iov; | |
1909 | } | |
31b51510 | 1910 | |
10d59345 JA |
1911 | /* file path doesn't support NOWAIT for non-direct_IO */ |
1912 | if (force_nonblock && !(kiocb->ki_flags & IOCB_DIRECT) && | |
1913 | (req->flags & REQ_F_ISREG)) | |
f67676d1 | 1914 | goto copy_iov; |
31b51510 | 1915 | |
f67676d1 | 1916 | iov_count = iov_iter_count(&iter); |
31b51510 | 1917 | ret = rw_verify_area(WRITE, file, &kiocb->ki_pos, iov_count); |
2b188cc1 | 1918 | if (!ret) { |
9bf7933f RP |
1919 | ssize_t ret2; |
1920 | ||
2b188cc1 JA |
1921 | /* |
1922 | * Open-code file_start_write here to grab freeze protection, | |
1923 | * which will be released by another thread in | |
1924 | * io_complete_rw(). Fool lockdep by telling it the lock got | |
1925 | * released so that it doesn't complain about the held lock when | |
1926 | * we return to userspace. | |
1927 | */ | |
491381ce | 1928 | if (req->flags & REQ_F_ISREG) { |
2b188cc1 JA |
1929 | __sb_start_write(file_inode(file)->i_sb, |
1930 | SB_FREEZE_WRITE, true); | |
1931 | __sb_writers_release(file_inode(file)->i_sb, | |
1932 | SB_FREEZE_WRITE); | |
1933 | } | |
1934 | kiocb->ki_flags |= IOCB_WRITE; | |
9bf7933f | 1935 | |
32960613 JA |
1936 | if (file->f_op->write_iter) |
1937 | ret2 = call_write_iter(file, kiocb, &iter); | |
1938 | else | |
1939 | ret2 = loop_rw_iter(WRITE, file, kiocb, &iter); | |
f67676d1 | 1940 | if (!force_nonblock || ret2 != -EAGAIN) { |
cf6fd4bd | 1941 | kiocb_done(kiocb, ret2, nxt, req->in_async); |
f67676d1 JA |
1942 | } else { |
1943 | copy_iov: | |
b7bb4f7d | 1944 | ret = io_setup_async_rw(req, io_size, iovec, |
f67676d1 JA |
1945 | inline_vecs, &iter); |
1946 | if (ret) | |
1947 | goto out_free; | |
1948 | return -EAGAIN; | |
1949 | } | |
2b188cc1 | 1950 | } |
31b51510 | 1951 | out_free: |
b7bb4f7d JA |
1952 | if (!io_wq_current_is_worker()) |
1953 | kfree(iovec); | |
2b188cc1 JA |
1954 | return ret; |
1955 | } | |
1956 | ||
1957 | /* | |
1958 | * IORING_OP_NOP just posts a completion event, nothing else. | |
1959 | */ | |
78e19bbe | 1960 | static int io_nop(struct io_kiocb *req) |
2b188cc1 JA |
1961 | { |
1962 | struct io_ring_ctx *ctx = req->ctx; | |
2b188cc1 | 1963 | |
def596e9 JA |
1964 | if (unlikely(ctx->flags & IORING_SETUP_IOPOLL)) |
1965 | return -EINVAL; | |
1966 | ||
78e19bbe | 1967 | io_cqring_add_event(req, 0); |
e65ef56d | 1968 | io_put_req(req); |
2b188cc1 JA |
1969 | return 0; |
1970 | } | |
1971 | ||
fc4df999 | 1972 | static int io_prep_fsync(struct io_kiocb *req) |
c992fe29 | 1973 | { |
fc4df999 | 1974 | const struct io_uring_sqe *sqe = req->sqe; |
6b06314c | 1975 | struct io_ring_ctx *ctx = req->ctx; |
c992fe29 | 1976 | |
8ed8d3c3 JA |
1977 | if (req->flags & REQ_F_PREPPED) |
1978 | return 0; | |
09bb8394 JA |
1979 | if (!req->file) |
1980 | return -EBADF; | |
c992fe29 | 1981 | |
6b06314c | 1982 | if (unlikely(ctx->flags & IORING_SETUP_IOPOLL)) |
def596e9 | 1983 | return -EINVAL; |
edafccee | 1984 | if (unlikely(sqe->addr || sqe->ioprio || sqe->buf_index)) |
c992fe29 CH |
1985 | return -EINVAL; |
1986 | ||
8ed8d3c3 JA |
1987 | req->sync.flags = READ_ONCE(sqe->fsync_flags); |
1988 | if (unlikely(req->sync.flags & ~IORING_FSYNC_DATASYNC)) | |
1989 | return -EINVAL; | |
1990 | ||
1991 | req->sync.off = READ_ONCE(sqe->off); | |
1992 | req->sync.len = READ_ONCE(sqe->len); | |
1993 | req->flags |= REQ_F_PREPPED; | |
c992fe29 CH |
1994 | return 0; |
1995 | } | |
1996 | ||
8ed8d3c3 JA |
1997 | static bool io_req_cancelled(struct io_kiocb *req) |
1998 | { | |
1999 | if (req->work.flags & IO_WQ_WORK_CANCEL) { | |
2000 | req_set_fail_links(req); | |
2001 | io_cqring_add_event(req, -ECANCELED); | |
2002 | io_put_req(req); | |
2003 | return true; | |
2004 | } | |
2005 | ||
2006 | return false; | |
2007 | } | |
2008 | ||
2009 | static void io_fsync_finish(struct io_wq_work **workptr) | |
2010 | { | |
2011 | struct io_kiocb *req = container_of(*workptr, struct io_kiocb, work); | |
2012 | loff_t end = req->sync.off + req->sync.len; | |
2013 | struct io_kiocb *nxt = NULL; | |
2014 | int ret; | |
2015 | ||
2016 | if (io_req_cancelled(req)) | |
2017 | return; | |
2018 | ||
2019 | ret = vfs_fsync_range(req->rw.ki_filp, req->sync.off, | |
2020 | end > 0 ? end : LLONG_MAX, | |
2021 | req->sync.flags & IORING_FSYNC_DATASYNC); | |
2022 | if (ret < 0) | |
2023 | req_set_fail_links(req); | |
2024 | io_cqring_add_event(req, ret); | |
2025 | io_put_req_find_next(req, &nxt); | |
2026 | if (nxt) | |
2027 | *workptr = &nxt->work; | |
2028 | } | |
2029 | ||
fc4df999 JA |
2030 | static int io_fsync(struct io_kiocb *req, struct io_kiocb **nxt, |
2031 | bool force_nonblock) | |
c992fe29 | 2032 | { |
8ed8d3c3 | 2033 | struct io_wq_work *work, *old_work; |
c992fe29 CH |
2034 | int ret; |
2035 | ||
fc4df999 | 2036 | ret = io_prep_fsync(req); |
c992fe29 CH |
2037 | if (ret) |
2038 | return ret; | |
2039 | ||
2040 | /* fsync always requires a blocking context */ | |
8ed8d3c3 JA |
2041 | if (force_nonblock) { |
2042 | io_put_req(req); | |
2043 | req->work.func = io_fsync_finish; | |
c992fe29 | 2044 | return -EAGAIN; |
8ed8d3c3 | 2045 | } |
c992fe29 | 2046 | |
8ed8d3c3 JA |
2047 | work = old_work = &req->work; |
2048 | io_fsync_finish(&work); | |
2049 | if (work && work != old_work) | |
2050 | *nxt = container_of(work, struct io_kiocb, work); | |
c992fe29 CH |
2051 | return 0; |
2052 | } | |
2053 | ||
fc4df999 | 2054 | static int io_prep_sfr(struct io_kiocb *req) |
5d17b4a4 | 2055 | { |
fc4df999 | 2056 | const struct io_uring_sqe *sqe = req->sqe; |
5d17b4a4 | 2057 | struct io_ring_ctx *ctx = req->ctx; |
5d17b4a4 | 2058 | |
8ed8d3c3 JA |
2059 | if (req->flags & REQ_F_PREPPED) |
2060 | return 0; | |
5d17b4a4 JA |
2061 | if (!req->file) |
2062 | return -EBADF; | |
5d17b4a4 JA |
2063 | |
2064 | if (unlikely(ctx->flags & IORING_SETUP_IOPOLL)) | |
2065 | return -EINVAL; | |
2066 | if (unlikely(sqe->addr || sqe->ioprio || sqe->buf_index)) | |
2067 | return -EINVAL; | |
2068 | ||
8ed8d3c3 JA |
2069 | req->sync.off = READ_ONCE(sqe->off); |
2070 | req->sync.len = READ_ONCE(sqe->len); | |
2071 | req->sync.flags = READ_ONCE(sqe->sync_range_flags); | |
2072 | req->flags |= REQ_F_PREPPED; | |
2073 | return 0; | |
2074 | } | |
2075 | ||
2076 | static void io_sync_file_range_finish(struct io_wq_work **workptr) | |
2077 | { | |
2078 | struct io_kiocb *req = container_of(*workptr, struct io_kiocb, work); | |
2079 | struct io_kiocb *nxt = NULL; | |
2080 | int ret; | |
2081 | ||
2082 | if (io_req_cancelled(req)) | |
2083 | return; | |
2084 | ||
2085 | ret = sync_file_range(req->rw.ki_filp, req->sync.off, req->sync.len, | |
2086 | req->sync.flags); | |
2087 | if (ret < 0) | |
2088 | req_set_fail_links(req); | |
2089 | io_cqring_add_event(req, ret); | |
2090 | io_put_req_find_next(req, &nxt); | |
2091 | if (nxt) | |
2092 | *workptr = &nxt->work; | |
5d17b4a4 JA |
2093 | } |
2094 | ||
fc4df999 | 2095 | static int io_sync_file_range(struct io_kiocb *req, struct io_kiocb **nxt, |
5d17b4a4 JA |
2096 | bool force_nonblock) |
2097 | { | |
8ed8d3c3 | 2098 | struct io_wq_work *work, *old_work; |
5d17b4a4 JA |
2099 | int ret; |
2100 | ||
fc4df999 | 2101 | ret = io_prep_sfr(req); |
5d17b4a4 JA |
2102 | if (ret) |
2103 | return ret; | |
2104 | ||
2105 | /* sync_file_range always requires a blocking context */ | |
8ed8d3c3 JA |
2106 | if (force_nonblock) { |
2107 | io_put_req(req); | |
2108 | req->work.func = io_sync_file_range_finish; | |
5d17b4a4 | 2109 | return -EAGAIN; |
8ed8d3c3 | 2110 | } |
5d17b4a4 | 2111 | |
8ed8d3c3 JA |
2112 | work = old_work = &req->work; |
2113 | io_sync_file_range_finish(&work); | |
2114 | if (work && work != old_work) | |
2115 | *nxt = container_of(work, struct io_kiocb, work); | |
5d17b4a4 JA |
2116 | return 0; |
2117 | } | |
2118 | ||
b7bb4f7d JA |
2119 | #if defined(CONFIG_NET) |
2120 | static void io_sendrecv_async(struct io_wq_work **workptr) | |
2121 | { | |
2122 | struct io_kiocb *req = container_of(*workptr, struct io_kiocb, work); | |
2123 | struct iovec *iov = NULL; | |
2124 | ||
2125 | if (req->io->rw.iov != req->io->rw.fast_iov) | |
2126 | iov = req->io->msg.iov; | |
2127 | io_wq_submit_work(workptr); | |
2128 | kfree(iov); | |
2129 | } | |
2130 | #endif | |
2131 | ||
03b1230c JA |
2132 | static int io_sendmsg_prep(struct io_kiocb *req, struct io_async_ctx *io) |
2133 | { | |
0fa03c62 | 2134 | #if defined(CONFIG_NET) |
03b1230c JA |
2135 | const struct io_uring_sqe *sqe = req->sqe; |
2136 | struct user_msghdr __user *msg; | |
2137 | unsigned flags; | |
2138 | ||
2139 | flags = READ_ONCE(sqe->msg_flags); | |
2140 | msg = (struct user_msghdr __user *)(unsigned long) READ_ONCE(sqe->addr); | |
d9688565 | 2141 | io->msg.iov = io->msg.fast_iov; |
03b1230c JA |
2142 | return sendmsg_copy_msghdr(&io->msg.msg, msg, flags, &io->msg.iov); |
2143 | #else | |
2144 | return 0; | |
2145 | #endif | |
2146 | } | |
2147 | ||
fc4df999 JA |
2148 | static int io_sendmsg(struct io_kiocb *req, struct io_kiocb **nxt, |
2149 | bool force_nonblock) | |
aa1fa28f | 2150 | { |
03b1230c | 2151 | #if defined(CONFIG_NET) |
fc4df999 | 2152 | const struct io_uring_sqe *sqe = req->sqe; |
0b416c3e | 2153 | struct io_async_msghdr *kmsg = NULL; |
0fa03c62 JA |
2154 | struct socket *sock; |
2155 | int ret; | |
2156 | ||
2157 | if (unlikely(req->ctx->flags & IORING_SETUP_IOPOLL)) | |
2158 | return -EINVAL; | |
2159 | ||
2160 | sock = sock_from_file(req->file, &ret); | |
2161 | if (sock) { | |
b7bb4f7d | 2162 | struct io_async_ctx io; |
03b1230c | 2163 | struct sockaddr_storage addr; |
0fa03c62 JA |
2164 | unsigned flags; |
2165 | ||
2166 | flags = READ_ONCE(sqe->msg_flags); | |
2167 | if (flags & MSG_DONTWAIT) | |
2168 | req->flags |= REQ_F_NOWAIT; | |
2169 | else if (force_nonblock) | |
2170 | flags |= MSG_DONTWAIT; | |
2171 | ||
03b1230c | 2172 | if (req->io) { |
0b416c3e JA |
2173 | kmsg = &req->io->msg; |
2174 | kmsg->msg.msg_name = &addr; | |
2175 | /* if iov is set, it's allocated already */ | |
2176 | if (!kmsg->iov) | |
2177 | kmsg->iov = kmsg->fast_iov; | |
2178 | kmsg->msg.msg_iter.iov = kmsg->iov; | |
03b1230c | 2179 | } else { |
0b416c3e JA |
2180 | kmsg = &io.msg; |
2181 | kmsg->msg.msg_name = &addr; | |
03b1230c JA |
2182 | ret = io_sendmsg_prep(req, &io); |
2183 | if (ret) | |
2184 | goto out; | |
2185 | } | |
0fa03c62 | 2186 | |
0b416c3e | 2187 | ret = __sys_sendmsg_sock(sock, &kmsg->msg, flags); |
03b1230c | 2188 | if (force_nonblock && ret == -EAGAIN) { |
b7bb4f7d JA |
2189 | if (req->io) |
2190 | return -EAGAIN; | |
2191 | if (io_alloc_async_ctx(req)) | |
2192 | return -ENOMEM; | |
2193 | memcpy(&req->io->msg, &io.msg, sizeof(io.msg)); | |
2194 | req->work.func = io_sendrecv_async; | |
0b416c3e | 2195 | return -EAGAIN; |
03b1230c | 2196 | } |
441cdbd5 JA |
2197 | if (ret == -ERESTARTSYS) |
2198 | ret = -EINTR; | |
0fa03c62 JA |
2199 | } |
2200 | ||
03b1230c | 2201 | out: |
b7bb4f7d | 2202 | if (!io_wq_current_is_worker() && kmsg && kmsg->iov != kmsg->fast_iov) |
0b416c3e | 2203 | kfree(kmsg->iov); |
78e19bbe | 2204 | io_cqring_add_event(req, ret); |
4e88d6e7 JA |
2205 | if (ret < 0) |
2206 | req_set_fail_links(req); | |
ec9c02ad | 2207 | io_put_req_find_next(req, nxt); |
5d17b4a4 | 2208 | return 0; |
03b1230c JA |
2209 | #else |
2210 | return -EOPNOTSUPP; | |
aa1fa28f | 2211 | #endif |
03b1230c | 2212 | } |
aa1fa28f | 2213 | |
03b1230c | 2214 | static int io_recvmsg_prep(struct io_kiocb *req, struct io_async_ctx *io) |
aa1fa28f JA |
2215 | { |
2216 | #if defined(CONFIG_NET) | |
03b1230c JA |
2217 | const struct io_uring_sqe *sqe = req->sqe; |
2218 | struct user_msghdr __user *msg; | |
2219 | unsigned flags; | |
2220 | ||
2221 | flags = READ_ONCE(sqe->msg_flags); | |
2222 | msg = (struct user_msghdr __user *)(unsigned long) READ_ONCE(sqe->addr); | |
d9688565 | 2223 | io->msg.iov = io->msg.fast_iov; |
03b1230c JA |
2224 | return recvmsg_copy_msghdr(&io->msg.msg, msg, flags, &io->msg.uaddr, |
2225 | &io->msg.iov); | |
aa1fa28f | 2226 | #else |
03b1230c | 2227 | return 0; |
aa1fa28f JA |
2228 | #endif |
2229 | } | |
2230 | ||
fc4df999 JA |
2231 | static int io_recvmsg(struct io_kiocb *req, struct io_kiocb **nxt, |
2232 | bool force_nonblock) | |
aa1fa28f JA |
2233 | { |
2234 | #if defined(CONFIG_NET) | |
fc4df999 | 2235 | const struct io_uring_sqe *sqe = req->sqe; |
0b416c3e | 2236 | struct io_async_msghdr *kmsg = NULL; |
03b1230c JA |
2237 | struct socket *sock; |
2238 | int ret; | |
2239 | ||
2240 | if (unlikely(req->ctx->flags & IORING_SETUP_IOPOLL)) | |
2241 | return -EINVAL; | |
2242 | ||
2243 | sock = sock_from_file(req->file, &ret); | |
2244 | if (sock) { | |
2245 | struct user_msghdr __user *msg; | |
b7bb4f7d | 2246 | struct io_async_ctx io; |
03b1230c | 2247 | struct sockaddr_storage addr; |
03b1230c JA |
2248 | unsigned flags; |
2249 | ||
2250 | flags = READ_ONCE(sqe->msg_flags); | |
2251 | if (flags & MSG_DONTWAIT) | |
2252 | req->flags |= REQ_F_NOWAIT; | |
2253 | else if (force_nonblock) | |
2254 | flags |= MSG_DONTWAIT; | |
2255 | ||
2256 | msg = (struct user_msghdr __user *) (unsigned long) | |
2257 | READ_ONCE(sqe->addr); | |
2258 | if (req->io) { | |
0b416c3e JA |
2259 | kmsg = &req->io->msg; |
2260 | kmsg->msg.msg_name = &addr; | |
2261 | /* if iov is set, it's allocated already */ | |
2262 | if (!kmsg->iov) | |
2263 | kmsg->iov = kmsg->fast_iov; | |
2264 | kmsg->msg.msg_iter.iov = kmsg->iov; | |
03b1230c | 2265 | } else { |
0b416c3e JA |
2266 | kmsg = &io.msg; |
2267 | kmsg->msg.msg_name = &addr; | |
03b1230c JA |
2268 | ret = io_recvmsg_prep(req, &io); |
2269 | if (ret) | |
2270 | goto out; | |
2271 | } | |
2272 | ||
0b416c3e | 2273 | ret = __sys_recvmsg_sock(sock, &kmsg->msg, msg, kmsg->uaddr, flags); |
03b1230c | 2274 | if (force_nonblock && ret == -EAGAIN) { |
b7bb4f7d JA |
2275 | if (req->io) |
2276 | return -EAGAIN; | |
2277 | if (io_alloc_async_ctx(req)) | |
2278 | return -ENOMEM; | |
2279 | memcpy(&req->io->msg, &io.msg, sizeof(io.msg)); | |
2280 | req->work.func = io_sendrecv_async; | |
0b416c3e | 2281 | return -EAGAIN; |
03b1230c JA |
2282 | } |
2283 | if (ret == -ERESTARTSYS) | |
2284 | ret = -EINTR; | |
2285 | } | |
2286 | ||
2287 | out: | |
b7bb4f7d | 2288 | if (!io_wq_current_is_worker() && kmsg && kmsg->iov != kmsg->fast_iov) |
0b416c3e | 2289 | kfree(kmsg->iov); |
03b1230c | 2290 | io_cqring_add_event(req, ret); |
4e88d6e7 JA |
2291 | if (ret < 0) |
2292 | req_set_fail_links(req); | |
03b1230c JA |
2293 | io_put_req_find_next(req, nxt); |
2294 | return 0; | |
0fa03c62 JA |
2295 | #else |
2296 | return -EOPNOTSUPP; | |
2297 | #endif | |
2298 | } | |
5d17b4a4 | 2299 | |
8ed8d3c3 | 2300 | static int io_accept_prep(struct io_kiocb *req) |
17f2fe35 JA |
2301 | { |
2302 | #if defined(CONFIG_NET) | |
fc4df999 | 2303 | const struct io_uring_sqe *sqe = req->sqe; |
8ed8d3c3 JA |
2304 | struct io_accept *accept = &req->accept; |
2305 | ||
2306 | if (req->flags & REQ_F_PREPPED) | |
2307 | return 0; | |
17f2fe35 JA |
2308 | |
2309 | if (unlikely(req->ctx->flags & (IORING_SETUP_IOPOLL|IORING_SETUP_SQPOLL))) | |
2310 | return -EINVAL; | |
8042d6ce | 2311 | if (sqe->ioprio || sqe->len || sqe->buf_index) |
17f2fe35 JA |
2312 | return -EINVAL; |
2313 | ||
8ed8d3c3 JA |
2314 | accept->addr = (struct sockaddr __user *) |
2315 | (unsigned long) READ_ONCE(sqe->addr); | |
2316 | accept->addr_len = (int __user *) (unsigned long) READ_ONCE(sqe->addr2); | |
2317 | accept->flags = READ_ONCE(sqe->accept_flags); | |
2318 | req->flags |= REQ_F_PREPPED; | |
2319 | return 0; | |
2320 | #else | |
2321 | return -EOPNOTSUPP; | |
2322 | #endif | |
2323 | } | |
17f2fe35 | 2324 | |
8ed8d3c3 JA |
2325 | #if defined(CONFIG_NET) |
2326 | static int __io_accept(struct io_kiocb *req, struct io_kiocb **nxt, | |
2327 | bool force_nonblock) | |
2328 | { | |
2329 | struct io_accept *accept = &req->accept; | |
2330 | unsigned file_flags; | |
2331 | int ret; | |
2332 | ||
2333 | file_flags = force_nonblock ? O_NONBLOCK : 0; | |
2334 | ret = __sys_accept4_file(req->file, file_flags, accept->addr, | |
2335 | accept->addr_len, accept->flags); | |
2336 | if (ret == -EAGAIN && force_nonblock) | |
17f2fe35 | 2337 | return -EAGAIN; |
8e3cca12 JA |
2338 | if (ret == -ERESTARTSYS) |
2339 | ret = -EINTR; | |
4e88d6e7 JA |
2340 | if (ret < 0) |
2341 | req_set_fail_links(req); | |
78e19bbe | 2342 | io_cqring_add_event(req, ret); |
ec9c02ad | 2343 | io_put_req_find_next(req, nxt); |
17f2fe35 | 2344 | return 0; |
8ed8d3c3 JA |
2345 | } |
2346 | ||
2347 | static void io_accept_finish(struct io_wq_work **workptr) | |
2348 | { | |
2349 | struct io_kiocb *req = container_of(*workptr, struct io_kiocb, work); | |
2350 | struct io_kiocb *nxt = NULL; | |
2351 | ||
2352 | if (io_req_cancelled(req)) | |
2353 | return; | |
2354 | __io_accept(req, &nxt, false); | |
2355 | if (nxt) | |
2356 | *workptr = &nxt->work; | |
2357 | } | |
2358 | #endif | |
2359 | ||
2360 | static int io_accept(struct io_kiocb *req, struct io_kiocb **nxt, | |
2361 | bool force_nonblock) | |
2362 | { | |
2363 | #if defined(CONFIG_NET) | |
2364 | int ret; | |
2365 | ||
2366 | ret = io_accept_prep(req); | |
2367 | if (ret) | |
2368 | return ret; | |
2369 | ||
2370 | ret = __io_accept(req, nxt, force_nonblock); | |
2371 | if (ret == -EAGAIN && force_nonblock) { | |
2372 | req->work.func = io_accept_finish; | |
2373 | req->work.flags |= IO_WQ_WORK_NEEDS_FILES; | |
2374 | io_put_req(req); | |
2375 | return -EAGAIN; | |
2376 | } | |
2377 | return 0; | |
0fa03c62 JA |
2378 | #else |
2379 | return -EOPNOTSUPP; | |
2380 | #endif | |
2381 | } | |
5d17b4a4 | 2382 | |
f499a021 JA |
2383 | static int io_connect_prep(struct io_kiocb *req, struct io_async_ctx *io) |
2384 | { | |
2385 | #if defined(CONFIG_NET) | |
2386 | const struct io_uring_sqe *sqe = req->sqe; | |
2387 | struct sockaddr __user *addr; | |
2388 | int addr_len; | |
2389 | ||
2390 | addr = (struct sockaddr __user *) (unsigned long) READ_ONCE(sqe->addr); | |
2391 | addr_len = READ_ONCE(sqe->addr2); | |
2392 | return move_addr_to_kernel(addr, addr_len, &io->connect.address); | |
2393 | #else | |
2394 | return 0; | |
2395 | #endif | |
2396 | } | |
2397 | ||
fc4df999 JA |
2398 | static int io_connect(struct io_kiocb *req, struct io_kiocb **nxt, |
2399 | bool force_nonblock) | |
f8e85cf2 JA |
2400 | { |
2401 | #if defined(CONFIG_NET) | |
fc4df999 | 2402 | const struct io_uring_sqe *sqe = req->sqe; |
f499a021 | 2403 | struct io_async_ctx __io, *io; |
f8e85cf2 JA |
2404 | unsigned file_flags; |
2405 | int addr_len, ret; | |
2406 | ||
2407 | if (unlikely(req->ctx->flags & (IORING_SETUP_IOPOLL|IORING_SETUP_SQPOLL))) | |
2408 | return -EINVAL; | |
2409 | if (sqe->ioprio || sqe->len || sqe->buf_index || sqe->rw_flags) | |
2410 | return -EINVAL; | |
2411 | ||
f8e85cf2 JA |
2412 | addr_len = READ_ONCE(sqe->addr2); |
2413 | file_flags = force_nonblock ? O_NONBLOCK : 0; | |
2414 | ||
f499a021 JA |
2415 | if (req->io) { |
2416 | io = req->io; | |
2417 | } else { | |
2418 | ret = io_connect_prep(req, &__io); | |
2419 | if (ret) | |
2420 | goto out; | |
2421 | io = &__io; | |
2422 | } | |
2423 | ||
2424 | ret = __sys_connect_file(req->file, &io->connect.address, addr_len, | |
2425 | file_flags); | |
87f80d62 | 2426 | if ((ret == -EAGAIN || ret == -EINPROGRESS) && force_nonblock) { |
b7bb4f7d JA |
2427 | if (req->io) |
2428 | return -EAGAIN; | |
2429 | if (io_alloc_async_ctx(req)) { | |
f499a021 JA |
2430 | ret = -ENOMEM; |
2431 | goto out; | |
2432 | } | |
b7bb4f7d | 2433 | memcpy(&req->io->connect, &__io.connect, sizeof(__io.connect)); |
f8e85cf2 | 2434 | return -EAGAIN; |
f499a021 | 2435 | } |
f8e85cf2 JA |
2436 | if (ret == -ERESTARTSYS) |
2437 | ret = -EINTR; | |
f499a021 | 2438 | out: |
4e88d6e7 JA |
2439 | if (ret < 0) |
2440 | req_set_fail_links(req); | |
f8e85cf2 JA |
2441 | io_cqring_add_event(req, ret); |
2442 | io_put_req_find_next(req, nxt); | |
2443 | return 0; | |
2444 | #else | |
2445 | return -EOPNOTSUPP; | |
2446 | #endif | |
2447 | } | |
2448 | ||
221c5eb2 JA |
2449 | static void io_poll_remove_one(struct io_kiocb *req) |
2450 | { | |
2451 | struct io_poll_iocb *poll = &req->poll; | |
2452 | ||
2453 | spin_lock(&poll->head->lock); | |
2454 | WRITE_ONCE(poll->canceled, true); | |
392edb45 JA |
2455 | if (!list_empty(&poll->wait.entry)) { |
2456 | list_del_init(&poll->wait.entry); | |
a197f664 | 2457 | io_queue_async_work(req); |
221c5eb2 JA |
2458 | } |
2459 | spin_unlock(&poll->head->lock); | |
78076bb6 | 2460 | hash_del(&req->hash_node); |
221c5eb2 JA |
2461 | } |
2462 | ||
2463 | static void io_poll_remove_all(struct io_ring_ctx *ctx) | |
2464 | { | |
78076bb6 | 2465 | struct hlist_node *tmp; |
221c5eb2 | 2466 | struct io_kiocb *req; |
78076bb6 | 2467 | int i; |
221c5eb2 JA |
2468 | |
2469 | spin_lock_irq(&ctx->completion_lock); | |
78076bb6 JA |
2470 | for (i = 0; i < (1U << ctx->cancel_hash_bits); i++) { |
2471 | struct hlist_head *list; | |
2472 | ||
2473 | list = &ctx->cancel_hash[i]; | |
2474 | hlist_for_each_entry_safe(req, tmp, list, hash_node) | |
2475 | io_poll_remove_one(req); | |
221c5eb2 JA |
2476 | } |
2477 | spin_unlock_irq(&ctx->completion_lock); | |
2478 | } | |
2479 | ||
47f46768 JA |
2480 | static int io_poll_cancel(struct io_ring_ctx *ctx, __u64 sqe_addr) |
2481 | { | |
78076bb6 | 2482 | struct hlist_head *list; |
47f46768 JA |
2483 | struct io_kiocb *req; |
2484 | ||
78076bb6 JA |
2485 | list = &ctx->cancel_hash[hash_long(sqe_addr, ctx->cancel_hash_bits)]; |
2486 | hlist_for_each_entry(req, list, hash_node) { | |
2487 | if (sqe_addr == req->user_data) { | |
eac406c6 JA |
2488 | io_poll_remove_one(req); |
2489 | return 0; | |
2490 | } | |
47f46768 JA |
2491 | } |
2492 | ||
2493 | return -ENOENT; | |
2494 | } | |
2495 | ||
0969e783 JA |
2496 | static int io_poll_remove_prep(struct io_kiocb *req) |
2497 | { | |
2498 | const struct io_uring_sqe *sqe = req->sqe; | |
2499 | ||
2500 | if (req->flags & REQ_F_PREPPED) | |
2501 | return 0; | |
2502 | if (unlikely(req->ctx->flags & IORING_SETUP_IOPOLL)) | |
2503 | return -EINVAL; | |
2504 | if (sqe->ioprio || sqe->off || sqe->len || sqe->buf_index || | |
2505 | sqe->poll_events) | |
2506 | return -EINVAL; | |
2507 | ||
2508 | req->poll.addr = READ_ONCE(sqe->addr); | |
2509 | req->flags |= REQ_F_PREPPED; | |
2510 | return 0; | |
2511 | } | |
2512 | ||
221c5eb2 JA |
2513 | /* |
2514 | * Find a running poll command that matches one specified in sqe->addr, | |
2515 | * and remove it if found. | |
2516 | */ | |
fc4df999 | 2517 | static int io_poll_remove(struct io_kiocb *req) |
221c5eb2 JA |
2518 | { |
2519 | struct io_ring_ctx *ctx = req->ctx; | |
0969e783 | 2520 | u64 addr; |
47f46768 | 2521 | int ret; |
221c5eb2 | 2522 | |
0969e783 JA |
2523 | ret = io_poll_remove_prep(req); |
2524 | if (ret) | |
2525 | return ret; | |
221c5eb2 | 2526 | |
0969e783 | 2527 | addr = req->poll.addr; |
221c5eb2 | 2528 | spin_lock_irq(&ctx->completion_lock); |
0969e783 | 2529 | ret = io_poll_cancel(ctx, addr); |
221c5eb2 JA |
2530 | spin_unlock_irq(&ctx->completion_lock); |
2531 | ||
78e19bbe | 2532 | io_cqring_add_event(req, ret); |
4e88d6e7 JA |
2533 | if (ret < 0) |
2534 | req_set_fail_links(req); | |
e65ef56d | 2535 | io_put_req(req); |
221c5eb2 JA |
2536 | return 0; |
2537 | } | |
2538 | ||
b0dd8a41 | 2539 | static void io_poll_complete(struct io_kiocb *req, __poll_t mask, int error) |
221c5eb2 | 2540 | { |
a197f664 JL |
2541 | struct io_ring_ctx *ctx = req->ctx; |
2542 | ||
8c838788 | 2543 | req->poll.done = true; |
b0dd8a41 JA |
2544 | if (error) |
2545 | io_cqring_fill_event(req, error); | |
2546 | else | |
2547 | io_cqring_fill_event(req, mangle_poll(mask)); | |
8c838788 | 2548 | io_commit_cqring(ctx); |
221c5eb2 JA |
2549 | } |
2550 | ||
561fb04a | 2551 | static void io_poll_complete_work(struct io_wq_work **workptr) |
221c5eb2 | 2552 | { |
561fb04a | 2553 | struct io_wq_work *work = *workptr; |
221c5eb2 JA |
2554 | struct io_kiocb *req = container_of(work, struct io_kiocb, work); |
2555 | struct io_poll_iocb *poll = &req->poll; | |
2556 | struct poll_table_struct pt = { ._key = poll->events }; | |
2557 | struct io_ring_ctx *ctx = req->ctx; | |
89723d0b | 2558 | struct io_kiocb *nxt = NULL; |
221c5eb2 | 2559 | __poll_t mask = 0; |
b0dd8a41 | 2560 | int ret = 0; |
221c5eb2 | 2561 | |
b0dd8a41 | 2562 | if (work->flags & IO_WQ_WORK_CANCEL) { |
561fb04a | 2563 | WRITE_ONCE(poll->canceled, true); |
b0dd8a41 JA |
2564 | ret = -ECANCELED; |
2565 | } else if (READ_ONCE(poll->canceled)) { | |
2566 | ret = -ECANCELED; | |
2567 | } | |
561fb04a | 2568 | |
b0dd8a41 | 2569 | if (ret != -ECANCELED) |
221c5eb2 JA |
2570 | mask = vfs_poll(poll->file, &pt) & poll->events; |
2571 | ||
2572 | /* | |
2573 | * Note that ->ki_cancel callers also delete iocb from active_reqs after | |
2574 | * calling ->ki_cancel. We need the ctx_lock roundtrip here to | |
2575 | * synchronize with them. In the cancellation case the list_del_init | |
2576 | * itself is not actually needed, but harmless so we keep it in to | |
2577 | * avoid further branches in the fast path. | |
2578 | */ | |
2579 | spin_lock_irq(&ctx->completion_lock); | |
b0dd8a41 | 2580 | if (!mask && ret != -ECANCELED) { |
392edb45 | 2581 | add_wait_queue(poll->head, &poll->wait); |
221c5eb2 JA |
2582 | spin_unlock_irq(&ctx->completion_lock); |
2583 | return; | |
2584 | } | |
78076bb6 | 2585 | hash_del(&req->hash_node); |
b0dd8a41 | 2586 | io_poll_complete(req, mask, ret); |
221c5eb2 JA |
2587 | spin_unlock_irq(&ctx->completion_lock); |
2588 | ||
8c838788 | 2589 | io_cqring_ev_posted(ctx); |
89723d0b | 2590 | |
4e88d6e7 JA |
2591 | if (ret < 0) |
2592 | req_set_fail_links(req); | |
ec9c02ad | 2593 | io_put_req_find_next(req, &nxt); |
89723d0b JA |
2594 | if (nxt) |
2595 | *workptr = &nxt->work; | |
221c5eb2 JA |
2596 | } |
2597 | ||
2598 | static int io_poll_wake(struct wait_queue_entry *wait, unsigned mode, int sync, | |
2599 | void *key) | |
2600 | { | |
e944475e | 2601 | struct io_poll_iocb *poll = wait->private; |
221c5eb2 JA |
2602 | struct io_kiocb *req = container_of(poll, struct io_kiocb, poll); |
2603 | struct io_ring_ctx *ctx = req->ctx; | |
2604 | __poll_t mask = key_to_poll(key); | |
8c838788 | 2605 | unsigned long flags; |
221c5eb2 JA |
2606 | |
2607 | /* for instances that support it check for an event match first: */ | |
8c838788 JA |
2608 | if (mask && !(mask & poll->events)) |
2609 | return 0; | |
221c5eb2 | 2610 | |
392edb45 | 2611 | list_del_init(&poll->wait.entry); |
221c5eb2 | 2612 | |
7c9e7f0f JA |
2613 | /* |
2614 | * Run completion inline if we can. We're using trylock here because | |
2615 | * we are violating the completion_lock -> poll wq lock ordering. | |
2616 | * If we have a link timeout we're going to need the completion_lock | |
2617 | * for finalizing the request, mark us as having grabbed that already. | |
2618 | */ | |
8c838788 | 2619 | if (mask && spin_trylock_irqsave(&ctx->completion_lock, flags)) { |
78076bb6 | 2620 | hash_del(&req->hash_node); |
b0dd8a41 | 2621 | io_poll_complete(req, mask, 0); |
7c9e7f0f JA |
2622 | req->flags |= REQ_F_COMP_LOCKED; |
2623 | io_put_req(req); | |
8c838788 | 2624 | spin_unlock_irqrestore(&ctx->completion_lock, flags); |
221c5eb2 | 2625 | |
8c838788 | 2626 | io_cqring_ev_posted(ctx); |
8c838788 | 2627 | } else { |
a197f664 | 2628 | io_queue_async_work(req); |
221c5eb2 JA |
2629 | } |
2630 | ||
221c5eb2 JA |
2631 | return 1; |
2632 | } | |
2633 | ||
2634 | struct io_poll_table { | |
2635 | struct poll_table_struct pt; | |
2636 | struct io_kiocb *req; | |
2637 | int error; | |
2638 | }; | |
2639 | ||
2640 | static void io_poll_queue_proc(struct file *file, struct wait_queue_head *head, | |
2641 | struct poll_table_struct *p) | |
2642 | { | |
2643 | struct io_poll_table *pt = container_of(p, struct io_poll_table, pt); | |
2644 | ||
2645 | if (unlikely(pt->req->poll.head)) { | |
2646 | pt->error = -EINVAL; | |
2647 | return; | |
2648 | } | |
2649 | ||
2650 | pt->error = 0; | |
2651 | pt->req->poll.head = head; | |
392edb45 | 2652 | add_wait_queue(head, &pt->req->poll.wait); |
221c5eb2 JA |
2653 | } |
2654 | ||
eac406c6 JA |
2655 | static void io_poll_req_insert(struct io_kiocb *req) |
2656 | { | |
2657 | struct io_ring_ctx *ctx = req->ctx; | |
78076bb6 JA |
2658 | struct hlist_head *list; |
2659 | ||
2660 | list = &ctx->cancel_hash[hash_long(req->user_data, ctx->cancel_hash_bits)]; | |
2661 | hlist_add_head(&req->hash_node, list); | |
eac406c6 JA |
2662 | } |
2663 | ||
0969e783 | 2664 | static int io_poll_add_prep(struct io_kiocb *req) |
221c5eb2 | 2665 | { |
fc4df999 | 2666 | const struct io_uring_sqe *sqe = req->sqe; |
221c5eb2 | 2667 | struct io_poll_iocb *poll = &req->poll; |
221c5eb2 | 2668 | u16 events; |
221c5eb2 | 2669 | |
0969e783 JA |
2670 | if (req->flags & REQ_F_PREPPED) |
2671 | return 0; | |
221c5eb2 JA |
2672 | if (unlikely(req->ctx->flags & IORING_SETUP_IOPOLL)) |
2673 | return -EINVAL; | |
2674 | if (sqe->addr || sqe->ioprio || sqe->off || sqe->len || sqe->buf_index) | |
2675 | return -EINVAL; | |
09bb8394 JA |
2676 | if (!poll->file) |
2677 | return -EBADF; | |
221c5eb2 | 2678 | |
0969e783 | 2679 | req->flags |= REQ_F_PREPPED; |
221c5eb2 JA |
2680 | events = READ_ONCE(sqe->poll_events); |
2681 | poll->events = demangle_poll(events) | EPOLLERR | EPOLLHUP; | |
0969e783 JA |
2682 | return 0; |
2683 | } | |
2684 | ||
2685 | static int io_poll_add(struct io_kiocb *req, struct io_kiocb **nxt) | |
2686 | { | |
2687 | struct io_poll_iocb *poll = &req->poll; | |
2688 | struct io_ring_ctx *ctx = req->ctx; | |
2689 | struct io_poll_table ipt; | |
2690 | bool cancel = false; | |
2691 | __poll_t mask; | |
2692 | int ret; | |
2693 | ||
2694 | ret = io_poll_add_prep(req); | |
2695 | if (ret) | |
2696 | return ret; | |
2697 | ||
2698 | INIT_IO_WORK(&req->work, io_poll_complete_work); | |
78076bb6 | 2699 | INIT_HLIST_NODE(&req->hash_node); |
221c5eb2 | 2700 | |
221c5eb2 | 2701 | poll->head = NULL; |
8c838788 | 2702 | poll->done = false; |
221c5eb2 JA |
2703 | poll->canceled = false; |
2704 | ||
2705 | ipt.pt._qproc = io_poll_queue_proc; | |
2706 | ipt.pt._key = poll->events; | |
2707 | ipt.req = req; | |
2708 | ipt.error = -EINVAL; /* same as no support for IOCB_CMD_POLL */ | |
2709 | ||
2710 | /* initialized the list so that we can do list_empty checks */ | |
392edb45 JA |
2711 | INIT_LIST_HEAD(&poll->wait.entry); |
2712 | init_waitqueue_func_entry(&poll->wait, io_poll_wake); | |
2713 | poll->wait.private = poll; | |
221c5eb2 | 2714 | |
36703247 JA |
2715 | INIT_LIST_HEAD(&req->list); |
2716 | ||
221c5eb2 | 2717 | mask = vfs_poll(poll->file, &ipt.pt) & poll->events; |
221c5eb2 JA |
2718 | |
2719 | spin_lock_irq(&ctx->completion_lock); | |
8c838788 JA |
2720 | if (likely(poll->head)) { |
2721 | spin_lock(&poll->head->lock); | |
392edb45 | 2722 | if (unlikely(list_empty(&poll->wait.entry))) { |
8c838788 JA |
2723 | if (ipt.error) |
2724 | cancel = true; | |
2725 | ipt.error = 0; | |
2726 | mask = 0; | |
2727 | } | |
2728 | if (mask || ipt.error) | |
392edb45 | 2729 | list_del_init(&poll->wait.entry); |
8c838788 JA |
2730 | else if (cancel) |
2731 | WRITE_ONCE(poll->canceled, true); | |
2732 | else if (!poll->done) /* actually waiting for an event */ | |
eac406c6 | 2733 | io_poll_req_insert(req); |
8c838788 JA |
2734 | spin_unlock(&poll->head->lock); |
2735 | } | |
2736 | if (mask) { /* no async, we'd stolen it */ | |
221c5eb2 | 2737 | ipt.error = 0; |
b0dd8a41 | 2738 | io_poll_complete(req, mask, 0); |
221c5eb2 | 2739 | } |
221c5eb2 JA |
2740 | spin_unlock_irq(&ctx->completion_lock); |
2741 | ||
8c838788 JA |
2742 | if (mask) { |
2743 | io_cqring_ev_posted(ctx); | |
ec9c02ad | 2744 | io_put_req_find_next(req, nxt); |
221c5eb2 | 2745 | } |
8c838788 | 2746 | return ipt.error; |
221c5eb2 JA |
2747 | } |
2748 | ||
5262f567 JA |
2749 | static enum hrtimer_restart io_timeout_fn(struct hrtimer *timer) |
2750 | { | |
ad8a48ac JA |
2751 | struct io_timeout_data *data = container_of(timer, |
2752 | struct io_timeout_data, timer); | |
2753 | struct io_kiocb *req = data->req; | |
2754 | struct io_ring_ctx *ctx = req->ctx; | |
5262f567 JA |
2755 | unsigned long flags; |
2756 | ||
5262f567 JA |
2757 | atomic_inc(&ctx->cq_timeouts); |
2758 | ||
2759 | spin_lock_irqsave(&ctx->completion_lock, flags); | |
ef03681a | 2760 | /* |
11365043 JA |
2761 | * We could be racing with timeout deletion. If the list is empty, |
2762 | * then timeout lookup already found it and will be handling it. | |
ef03681a | 2763 | */ |
842f9612 | 2764 | if (!list_empty(&req->list)) { |
11365043 | 2765 | struct io_kiocb *prev; |
5262f567 | 2766 | |
11365043 JA |
2767 | /* |
2768 | * Adjust the reqs sequence before the current one because it | |
d195a66e | 2769 | * will consume a slot in the cq_ring and the cq_tail |
11365043 JA |
2770 | * pointer will be increased, otherwise other timeout reqs may |
2771 | * return in advance without waiting for enough wait_nr. | |
2772 | */ | |
2773 | prev = req; | |
2774 | list_for_each_entry_continue_reverse(prev, &ctx->timeout_list, list) | |
2775 | prev->sequence++; | |
11365043 | 2776 | list_del_init(&req->list); |
11365043 | 2777 | } |
5262f567 | 2778 | |
78e19bbe | 2779 | io_cqring_fill_event(req, -ETIME); |
5262f567 JA |
2780 | io_commit_cqring(ctx); |
2781 | spin_unlock_irqrestore(&ctx->completion_lock, flags); | |
2782 | ||
2783 | io_cqring_ev_posted(ctx); | |
4e88d6e7 | 2784 | req_set_fail_links(req); |
5262f567 JA |
2785 | io_put_req(req); |
2786 | return HRTIMER_NORESTART; | |
2787 | } | |
2788 | ||
47f46768 JA |
2789 | static int io_timeout_cancel(struct io_ring_ctx *ctx, __u64 user_data) |
2790 | { | |
2791 | struct io_kiocb *req; | |
2792 | int ret = -ENOENT; | |
2793 | ||
2794 | list_for_each_entry(req, &ctx->timeout_list, list) { | |
2795 | if (user_data == req->user_data) { | |
2796 | list_del_init(&req->list); | |
2797 | ret = 0; | |
2798 | break; | |
2799 | } | |
2800 | } | |
2801 | ||
2802 | if (ret == -ENOENT) | |
2803 | return ret; | |
2804 | ||
2d28390a | 2805 | ret = hrtimer_try_to_cancel(&req->io->timeout.timer); |
47f46768 JA |
2806 | if (ret == -1) |
2807 | return -EALREADY; | |
2808 | ||
4e88d6e7 | 2809 | req_set_fail_links(req); |
47f46768 JA |
2810 | io_cqring_fill_event(req, -ECANCELED); |
2811 | io_put_req(req); | |
2812 | return 0; | |
2813 | } | |
2814 | ||
11365043 JA |
2815 | /* |
2816 | * Remove or update an existing timeout command | |
2817 | */ | |
fc4df999 | 2818 | static int io_timeout_remove(struct io_kiocb *req) |
11365043 | 2819 | { |
fc4df999 | 2820 | const struct io_uring_sqe *sqe = req->sqe; |
11365043 | 2821 | struct io_ring_ctx *ctx = req->ctx; |
11365043 | 2822 | unsigned flags; |
47f46768 | 2823 | int ret; |
11365043 JA |
2824 | |
2825 | if (unlikely(ctx->flags & IORING_SETUP_IOPOLL)) | |
2826 | return -EINVAL; | |
2827 | if (sqe->flags || sqe->ioprio || sqe->buf_index || sqe->len) | |
2828 | return -EINVAL; | |
2829 | flags = READ_ONCE(sqe->timeout_flags); | |
2830 | if (flags) | |
2831 | return -EINVAL; | |
2832 | ||
11365043 | 2833 | spin_lock_irq(&ctx->completion_lock); |
47f46768 | 2834 | ret = io_timeout_cancel(ctx, READ_ONCE(sqe->addr)); |
11365043 | 2835 | |
47f46768 | 2836 | io_cqring_fill_event(req, ret); |
11365043 JA |
2837 | io_commit_cqring(ctx); |
2838 | spin_unlock_irq(&ctx->completion_lock); | |
5262f567 | 2839 | io_cqring_ev_posted(ctx); |
4e88d6e7 JA |
2840 | if (ret < 0) |
2841 | req_set_fail_links(req); | |
ec9c02ad | 2842 | io_put_req(req); |
11365043 | 2843 | return 0; |
5262f567 JA |
2844 | } |
2845 | ||
2d28390a JA |
2846 | static int io_timeout_prep(struct io_kiocb *req, struct io_async_ctx *io, |
2847 | bool is_timeout_link) | |
5262f567 | 2848 | { |
cf6fd4bd | 2849 | const struct io_uring_sqe *sqe = req->sqe; |
ad8a48ac | 2850 | struct io_timeout_data *data; |
a41525ab | 2851 | unsigned flags; |
5262f567 | 2852 | |
ad8a48ac | 2853 | if (unlikely(req->ctx->flags & IORING_SETUP_IOPOLL)) |
5262f567 | 2854 | return -EINVAL; |
ad8a48ac | 2855 | if (sqe->ioprio || sqe->buf_index || sqe->len != 1) |
a41525ab | 2856 | return -EINVAL; |
2d28390a JA |
2857 | if (sqe->off && is_timeout_link) |
2858 | return -EINVAL; | |
a41525ab JA |
2859 | flags = READ_ONCE(sqe->timeout_flags); |
2860 | if (flags & ~IORING_TIMEOUT_ABS) | |
5262f567 | 2861 | return -EINVAL; |
bdf20073 | 2862 | |
2d28390a | 2863 | data = &io->timeout; |
ad8a48ac | 2864 | data->req = req; |
ad8a48ac JA |
2865 | req->flags |= REQ_F_TIMEOUT; |
2866 | ||
2867 | if (get_timespec64(&data->ts, u64_to_user_ptr(sqe->addr))) | |
5262f567 JA |
2868 | return -EFAULT; |
2869 | ||
11365043 | 2870 | if (flags & IORING_TIMEOUT_ABS) |
ad8a48ac | 2871 | data->mode = HRTIMER_MODE_ABS; |
11365043 | 2872 | else |
ad8a48ac | 2873 | data->mode = HRTIMER_MODE_REL; |
11365043 | 2874 | |
ad8a48ac JA |
2875 | hrtimer_init(&data->timer, CLOCK_MONOTONIC, data->mode); |
2876 | return 0; | |
2877 | } | |
2878 | ||
fc4df999 | 2879 | static int io_timeout(struct io_kiocb *req) |
ad8a48ac | 2880 | { |
fc4df999 | 2881 | const struct io_uring_sqe *sqe = req->sqe; |
ad8a48ac JA |
2882 | unsigned count; |
2883 | struct io_ring_ctx *ctx = req->ctx; | |
2884 | struct io_timeout_data *data; | |
2885 | struct list_head *entry; | |
2886 | unsigned span = 0; | |
b7bb4f7d | 2887 | int ret; |
ad8a48ac | 2888 | |
b7bb4f7d JA |
2889 | if (!req->io) { |
2890 | if (io_alloc_async_ctx(req)) | |
2d28390a | 2891 | return -ENOMEM; |
b7bb4f7d JA |
2892 | ret = io_timeout_prep(req, req->io, false); |
2893 | if (ret) | |
2d28390a | 2894 | return ret; |
2d28390a JA |
2895 | } |
2896 | data = &req->io->timeout; | |
93bd25bb | 2897 | |
5262f567 JA |
2898 | /* |
2899 | * sqe->off holds how many events that need to occur for this | |
93bd25bb JA |
2900 | * timeout event to be satisfied. If it isn't set, then this is |
2901 | * a pure timeout request, sequence isn't used. | |
5262f567 JA |
2902 | */ |
2903 | count = READ_ONCE(sqe->off); | |
93bd25bb JA |
2904 | if (!count) { |
2905 | req->flags |= REQ_F_TIMEOUT_NOSEQ; | |
2906 | spin_lock_irq(&ctx->completion_lock); | |
2907 | entry = ctx->timeout_list.prev; | |
2908 | goto add; | |
2909 | } | |
5262f567 JA |
2910 | |
2911 | req->sequence = ctx->cached_sq_head + count - 1; | |
2d28390a | 2912 | data->seq_offset = count; |
5262f567 JA |
2913 | |
2914 | /* | |
2915 | * Insertion sort, ensuring the first entry in the list is always | |
2916 | * the one we need first. | |
2917 | */ | |
5262f567 JA |
2918 | spin_lock_irq(&ctx->completion_lock); |
2919 | list_for_each_prev(entry, &ctx->timeout_list) { | |
2920 | struct io_kiocb *nxt = list_entry(entry, struct io_kiocb, list); | |
5da0fb1a | 2921 | unsigned nxt_sq_head; |
2922 | long long tmp, tmp_nxt; | |
2d28390a | 2923 | u32 nxt_offset = nxt->io->timeout.seq_offset; |
5262f567 | 2924 | |
93bd25bb JA |
2925 | if (nxt->flags & REQ_F_TIMEOUT_NOSEQ) |
2926 | continue; | |
2927 | ||
5da0fb1a | 2928 | /* |
2929 | * Since cached_sq_head + count - 1 can overflow, use type long | |
2930 | * long to store it. | |
2931 | */ | |
2932 | tmp = (long long)ctx->cached_sq_head + count - 1; | |
cc42e0ac PB |
2933 | nxt_sq_head = nxt->sequence - nxt_offset + 1; |
2934 | tmp_nxt = (long long)nxt_sq_head + nxt_offset - 1; | |
5da0fb1a | 2935 | |
2936 | /* | |
2937 | * cached_sq_head may overflow, and it will never overflow twice | |
2938 | * once there is some timeout req still be valid. | |
2939 | */ | |
2940 | if (ctx->cached_sq_head < nxt_sq_head) | |
8b07a65a | 2941 | tmp += UINT_MAX; |
5da0fb1a | 2942 | |
a1f58ba4 | 2943 | if (tmp > tmp_nxt) |
5262f567 | 2944 | break; |
a1f58ba4 | 2945 | |
2946 | /* | |
2947 | * Sequence of reqs after the insert one and itself should | |
2948 | * be adjusted because each timeout req consumes a slot. | |
2949 | */ | |
2950 | span++; | |
2951 | nxt->sequence++; | |
5262f567 | 2952 | } |
a1f58ba4 | 2953 | req->sequence -= span; |
93bd25bb | 2954 | add: |
5262f567 | 2955 | list_add(&req->list, entry); |
ad8a48ac JA |
2956 | data->timer.function = io_timeout_fn; |
2957 | hrtimer_start(&data->timer, timespec64_to_ktime(data->ts), data->mode); | |
5262f567 | 2958 | spin_unlock_irq(&ctx->completion_lock); |
5262f567 JA |
2959 | return 0; |
2960 | } | |
5262f567 | 2961 | |
62755e35 JA |
2962 | static bool io_cancel_cb(struct io_wq_work *work, void *data) |
2963 | { | |
2964 | struct io_kiocb *req = container_of(work, struct io_kiocb, work); | |
2965 | ||
2966 | return req->user_data == (unsigned long) data; | |
2967 | } | |
2968 | ||
e977d6d3 | 2969 | static int io_async_cancel_one(struct io_ring_ctx *ctx, void *sqe_addr) |
62755e35 | 2970 | { |
62755e35 | 2971 | enum io_wq_cancel cancel_ret; |
62755e35 JA |
2972 | int ret = 0; |
2973 | ||
62755e35 JA |
2974 | cancel_ret = io_wq_cancel_cb(ctx->io_wq, io_cancel_cb, sqe_addr); |
2975 | switch (cancel_ret) { | |
2976 | case IO_WQ_CANCEL_OK: | |
2977 | ret = 0; | |
2978 | break; | |
2979 | case IO_WQ_CANCEL_RUNNING: | |
2980 | ret = -EALREADY; | |
2981 | break; | |
2982 | case IO_WQ_CANCEL_NOTFOUND: | |
2983 | ret = -ENOENT; | |
2984 | break; | |
2985 | } | |
2986 | ||
e977d6d3 JA |
2987 | return ret; |
2988 | } | |
2989 | ||
47f46768 JA |
2990 | static void io_async_find_and_cancel(struct io_ring_ctx *ctx, |
2991 | struct io_kiocb *req, __u64 sqe_addr, | |
b0dd8a41 | 2992 | struct io_kiocb **nxt, int success_ret) |
47f46768 JA |
2993 | { |
2994 | unsigned long flags; | |
2995 | int ret; | |
2996 | ||
2997 | ret = io_async_cancel_one(ctx, (void *) (unsigned long) sqe_addr); | |
2998 | if (ret != -ENOENT) { | |
2999 | spin_lock_irqsave(&ctx->completion_lock, flags); | |
3000 | goto done; | |
3001 | } | |
3002 | ||
3003 | spin_lock_irqsave(&ctx->completion_lock, flags); | |
3004 | ret = io_timeout_cancel(ctx, sqe_addr); | |
3005 | if (ret != -ENOENT) | |
3006 | goto done; | |
3007 | ret = io_poll_cancel(ctx, sqe_addr); | |
3008 | done: | |
b0dd8a41 JA |
3009 | if (!ret) |
3010 | ret = success_ret; | |
47f46768 JA |
3011 | io_cqring_fill_event(req, ret); |
3012 | io_commit_cqring(ctx); | |
3013 | spin_unlock_irqrestore(&ctx->completion_lock, flags); | |
3014 | io_cqring_ev_posted(ctx); | |
3015 | ||
4e88d6e7 JA |
3016 | if (ret < 0) |
3017 | req_set_fail_links(req); | |
47f46768 JA |
3018 | io_put_req_find_next(req, nxt); |
3019 | } | |
3020 | ||
fc4df999 | 3021 | static int io_async_cancel(struct io_kiocb *req, struct io_kiocb **nxt) |
e977d6d3 | 3022 | { |
fc4df999 | 3023 | const struct io_uring_sqe *sqe = req->sqe; |
e977d6d3 | 3024 | struct io_ring_ctx *ctx = req->ctx; |
e977d6d3 JA |
3025 | |
3026 | if (unlikely(ctx->flags & IORING_SETUP_IOPOLL)) | |
3027 | return -EINVAL; | |
3028 | if (sqe->flags || sqe->ioprio || sqe->off || sqe->len || | |
3029 | sqe->cancel_flags) | |
3030 | return -EINVAL; | |
3031 | ||
b0dd8a41 | 3032 | io_async_find_and_cancel(ctx, req, READ_ONCE(sqe->addr), nxt, 0); |
5262f567 JA |
3033 | return 0; |
3034 | } | |
3035 | ||
b7bb4f7d | 3036 | static int io_req_defer_prep(struct io_kiocb *req) |
f67676d1 JA |
3037 | { |
3038 | struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs; | |
b7bb4f7d | 3039 | struct io_async_ctx *io = req->io; |
f67676d1 JA |
3040 | struct iov_iter iter; |
3041 | ssize_t ret; | |
3042 | ||
f67676d1 JA |
3043 | switch (io->sqe.opcode) { |
3044 | case IORING_OP_READV: | |
3045 | case IORING_OP_READ_FIXED: | |
b7bb4f7d JA |
3046 | /* ensure prep does right import */ |
3047 | req->io = NULL; | |
f67676d1 | 3048 | ret = io_read_prep(req, &iovec, &iter, true); |
b7bb4f7d JA |
3049 | req->io = io; |
3050 | if (ret < 0) | |
3051 | break; | |
3052 | io_req_map_rw(req, ret, iovec, inline_vecs, &iter); | |
3053 | ret = 0; | |
f67676d1 JA |
3054 | break; |
3055 | case IORING_OP_WRITEV: | |
3056 | case IORING_OP_WRITE_FIXED: | |
b7bb4f7d JA |
3057 | /* ensure prep does right import */ |
3058 | req->io = NULL; | |
f67676d1 | 3059 | ret = io_write_prep(req, &iovec, &iter, true); |
b7bb4f7d JA |
3060 | req->io = io; |
3061 | if (ret < 0) | |
3062 | break; | |
3063 | io_req_map_rw(req, ret, iovec, inline_vecs, &iter); | |
3064 | ret = 0; | |
f67676d1 | 3065 | break; |
0969e783 JA |
3066 | case IORING_OP_POLL_ADD: |
3067 | ret = io_poll_add_prep(req); | |
3068 | break; | |
3069 | case IORING_OP_POLL_REMOVE: | |
3070 | ret = io_poll_remove_prep(req); | |
3071 | break; | |
8ed8d3c3 JA |
3072 | case IORING_OP_FSYNC: |
3073 | ret = io_prep_fsync(req); | |
3074 | break; | |
3075 | case IORING_OP_SYNC_FILE_RANGE: | |
3076 | ret = io_prep_sfr(req); | |
3077 | break; | |
03b1230c JA |
3078 | case IORING_OP_SENDMSG: |
3079 | ret = io_sendmsg_prep(req, io); | |
3080 | break; | |
3081 | case IORING_OP_RECVMSG: | |
3082 | ret = io_recvmsg_prep(req, io); | |
3083 | break; | |
f499a021 JA |
3084 | case IORING_OP_CONNECT: |
3085 | ret = io_connect_prep(req, io); | |
3086 | break; | |
2d28390a | 3087 | case IORING_OP_TIMEOUT: |
b7bb4f7d JA |
3088 | ret = io_timeout_prep(req, io, false); |
3089 | break; | |
2d28390a | 3090 | case IORING_OP_LINK_TIMEOUT: |
b7bb4f7d JA |
3091 | ret = io_timeout_prep(req, io, true); |
3092 | break; | |
8ed8d3c3 JA |
3093 | case IORING_OP_ACCEPT: |
3094 | ret = io_accept_prep(req); | |
3095 | break; | |
f67676d1 | 3096 | default: |
b7bb4f7d JA |
3097 | ret = 0; |
3098 | break; | |
f67676d1 JA |
3099 | } |
3100 | ||
b7bb4f7d | 3101 | return ret; |
f67676d1 JA |
3102 | } |
3103 | ||
a197f664 | 3104 | static int io_req_defer(struct io_kiocb *req) |
de0617e4 | 3105 | { |
a197f664 | 3106 | struct io_ring_ctx *ctx = req->ctx; |
f67676d1 | 3107 | int ret; |
de0617e4 | 3108 | |
9d858b21 BL |
3109 | /* Still need defer if there is pending req in defer list. */ |
3110 | if (!req_need_defer(req) && list_empty(&ctx->defer_list)) | |
de0617e4 JA |
3111 | return 0; |
3112 | ||
b7bb4f7d | 3113 | if (io_alloc_async_ctx(req)) |
de0617e4 JA |
3114 | return -EAGAIN; |
3115 | ||
b7bb4f7d JA |
3116 | ret = io_req_defer_prep(req); |
3117 | if (ret < 0) | |
2d28390a | 3118 | return ret; |
2d28390a | 3119 | |
de0617e4 | 3120 | spin_lock_irq(&ctx->completion_lock); |
9d858b21 | 3121 | if (!req_need_defer(req) && list_empty(&ctx->defer_list)) { |
de0617e4 | 3122 | spin_unlock_irq(&ctx->completion_lock); |
de0617e4 JA |
3123 | return 0; |
3124 | } | |
3125 | ||
915967f6 | 3126 | trace_io_uring_defer(ctx, req, req->user_data); |
de0617e4 JA |
3127 | list_add_tail(&req->list, &ctx->defer_list); |
3128 | spin_unlock_irq(&ctx->completion_lock); | |
3129 | return -EIOCBQUEUED; | |
3130 | } | |
3131 | ||
f9bd67f6 | 3132 | __attribute__((nonnull)) |
d732447f PB |
3133 | static int io_issue_sqe(struct io_kiocb *req, struct io_kiocb **nxt, |
3134 | bool force_nonblock) | |
2b188cc1 | 3135 | { |
e0c5c576 | 3136 | int ret, opcode; |
a197f664 | 3137 | struct io_ring_ctx *ctx = req->ctx; |
2b188cc1 | 3138 | |
cf6fd4bd | 3139 | opcode = READ_ONCE(req->sqe->opcode); |
2b188cc1 JA |
3140 | switch (opcode) { |
3141 | case IORING_OP_NOP: | |
78e19bbe | 3142 | ret = io_nop(req); |
2b188cc1 JA |
3143 | break; |
3144 | case IORING_OP_READV: | |
cf6fd4bd | 3145 | if (unlikely(req->sqe->buf_index)) |
edafccee | 3146 | return -EINVAL; |
267bc904 | 3147 | ret = io_read(req, nxt, force_nonblock); |
2b188cc1 JA |
3148 | break; |
3149 | case IORING_OP_WRITEV: | |
cf6fd4bd | 3150 | if (unlikely(req->sqe->buf_index)) |
edafccee | 3151 | return -EINVAL; |
267bc904 | 3152 | ret = io_write(req, nxt, force_nonblock); |
edafccee JA |
3153 | break; |
3154 | case IORING_OP_READ_FIXED: | |
267bc904 | 3155 | ret = io_read(req, nxt, force_nonblock); |
edafccee JA |
3156 | break; |
3157 | case IORING_OP_WRITE_FIXED: | |
267bc904 | 3158 | ret = io_write(req, nxt, force_nonblock); |
2b188cc1 | 3159 | break; |
c992fe29 | 3160 | case IORING_OP_FSYNC: |
fc4df999 | 3161 | ret = io_fsync(req, nxt, force_nonblock); |
c992fe29 | 3162 | break; |
221c5eb2 | 3163 | case IORING_OP_POLL_ADD: |
fc4df999 | 3164 | ret = io_poll_add(req, nxt); |
221c5eb2 JA |
3165 | break; |
3166 | case IORING_OP_POLL_REMOVE: | |
fc4df999 | 3167 | ret = io_poll_remove(req); |
221c5eb2 | 3168 | break; |
5d17b4a4 | 3169 | case IORING_OP_SYNC_FILE_RANGE: |
fc4df999 | 3170 | ret = io_sync_file_range(req, nxt, force_nonblock); |
5d17b4a4 | 3171 | break; |
0fa03c62 | 3172 | case IORING_OP_SENDMSG: |
fc4df999 | 3173 | ret = io_sendmsg(req, nxt, force_nonblock); |
0fa03c62 | 3174 | break; |
aa1fa28f | 3175 | case IORING_OP_RECVMSG: |
fc4df999 | 3176 | ret = io_recvmsg(req, nxt, force_nonblock); |
aa1fa28f | 3177 | break; |
5262f567 | 3178 | case IORING_OP_TIMEOUT: |
fc4df999 | 3179 | ret = io_timeout(req); |
5262f567 | 3180 | break; |
11365043 | 3181 | case IORING_OP_TIMEOUT_REMOVE: |
fc4df999 | 3182 | ret = io_timeout_remove(req); |
11365043 | 3183 | break; |
17f2fe35 | 3184 | case IORING_OP_ACCEPT: |
fc4df999 | 3185 | ret = io_accept(req, nxt, force_nonblock); |
17f2fe35 | 3186 | break; |
f8e85cf2 | 3187 | case IORING_OP_CONNECT: |
fc4df999 | 3188 | ret = io_connect(req, nxt, force_nonblock); |
f8e85cf2 | 3189 | break; |
62755e35 | 3190 | case IORING_OP_ASYNC_CANCEL: |
fc4df999 | 3191 | ret = io_async_cancel(req, nxt); |
62755e35 | 3192 | break; |
2b188cc1 JA |
3193 | default: |
3194 | ret = -EINVAL; | |
3195 | break; | |
3196 | } | |
3197 | ||
def596e9 JA |
3198 | if (ret) |
3199 | return ret; | |
3200 | ||
3201 | if (ctx->flags & IORING_SETUP_IOPOLL) { | |
9e645e11 | 3202 | if (req->result == -EAGAIN) |
def596e9 JA |
3203 | return -EAGAIN; |
3204 | ||
def596e9 | 3205 | io_iopoll_req_issued(req); |
def596e9 JA |
3206 | } |
3207 | ||
3208 | return 0; | |
2b188cc1 JA |
3209 | } |
3210 | ||
b76da70f JA |
3211 | static void io_link_work_cb(struct io_wq_work **workptr) |
3212 | { | |
3213 | struct io_wq_work *work = *workptr; | |
3214 | struct io_kiocb *link = work->data; | |
3215 | ||
3216 | io_queue_linked_timeout(link); | |
3217 | work->func = io_wq_submit_work; | |
3218 | } | |
3219 | ||
561fb04a | 3220 | static void io_wq_submit_work(struct io_wq_work **workptr) |
2b188cc1 | 3221 | { |
561fb04a | 3222 | struct io_wq_work *work = *workptr; |
2b188cc1 | 3223 | struct io_kiocb *req = container_of(work, struct io_kiocb, work); |
561fb04a JA |
3224 | struct io_kiocb *nxt = NULL; |
3225 | int ret = 0; | |
2b188cc1 | 3226 | |
561fb04a JA |
3227 | /* Ensure we clear previously set non-block flag */ |
3228 | req->rw.ki_flags &= ~IOCB_NOWAIT; | |
2b188cc1 | 3229 | |
561fb04a JA |
3230 | if (work->flags & IO_WQ_WORK_CANCEL) |
3231 | ret = -ECANCELED; | |
31b51510 | 3232 | |
561fb04a | 3233 | if (!ret) { |
cf6fd4bd PB |
3234 | req->has_user = (work->flags & IO_WQ_WORK_HAS_MM) != 0; |
3235 | req->in_async = true; | |
561fb04a | 3236 | do { |
d732447f | 3237 | ret = io_issue_sqe(req, &nxt, false); |
561fb04a JA |
3238 | /* |
3239 | * We can get EAGAIN for polled IO even though we're | |
3240 | * forcing a sync submission from here, since we can't | |
3241 | * wait for request slots on the block side. | |
3242 | */ | |
3243 | if (ret != -EAGAIN) | |
3244 | break; | |
3245 | cond_resched(); | |
3246 | } while (1); | |
3247 | } | |
31b51510 | 3248 | |
561fb04a | 3249 | /* drop submission reference */ |
ec9c02ad | 3250 | io_put_req(req); |
817869d2 | 3251 | |
561fb04a | 3252 | if (ret) { |
4e88d6e7 | 3253 | req_set_fail_links(req); |
78e19bbe | 3254 | io_cqring_add_event(req, ret); |
817869d2 | 3255 | io_put_req(req); |
edafccee | 3256 | } |
2b188cc1 | 3257 | |
561fb04a JA |
3258 | /* if a dependent link is ready, pass it back */ |
3259 | if (!ret && nxt) { | |
94ae5e77 JA |
3260 | struct io_kiocb *link; |
3261 | ||
3262 | io_prep_async_work(nxt, &link); | |
561fb04a | 3263 | *workptr = &nxt->work; |
b76da70f JA |
3264 | if (link) { |
3265 | nxt->work.flags |= IO_WQ_WORK_CB; | |
3266 | nxt->work.func = io_link_work_cb; | |
3267 | nxt->work.data = link; | |
3268 | } | |
31b51510 | 3269 | } |
2b188cc1 JA |
3270 | } |
3271 | ||
9e3aa61a JA |
3272 | static bool io_req_op_valid(int op) |
3273 | { | |
3274 | return op >= IORING_OP_NOP && op < IORING_OP_LAST; | |
3275 | } | |
3276 | ||
3277 | static int io_op_needs_file(const struct io_uring_sqe *sqe) | |
09bb8394 JA |
3278 | { |
3279 | int op = READ_ONCE(sqe->opcode); | |
3280 | ||
3281 | switch (op) { | |
3282 | case IORING_OP_NOP: | |
3283 | case IORING_OP_POLL_REMOVE: | |
5683e540 | 3284 | case IORING_OP_TIMEOUT: |
a320e9fa PB |
3285 | case IORING_OP_TIMEOUT_REMOVE: |
3286 | case IORING_OP_ASYNC_CANCEL: | |
3287 | case IORING_OP_LINK_TIMEOUT: | |
9e3aa61a | 3288 | return 0; |
09bb8394 | 3289 | default: |
9e3aa61a JA |
3290 | if (io_req_op_valid(op)) |
3291 | return 1; | |
3292 | return -EINVAL; | |
09bb8394 JA |
3293 | } |
3294 | } | |
3295 | ||
65e19f54 JA |
3296 | static inline struct file *io_file_from_index(struct io_ring_ctx *ctx, |
3297 | int index) | |
3298 | { | |
3299 | struct fixed_file_table *table; | |
3300 | ||
3301 | table = &ctx->file_table[index >> IORING_FILE_TABLE_SHIFT]; | |
3302 | return table->files[index & IORING_FILE_TABLE_MASK]; | |
3303 | } | |
3304 | ||
a197f664 | 3305 | static int io_req_set_file(struct io_submit_state *state, struct io_kiocb *req) |
09bb8394 | 3306 | { |
a197f664 | 3307 | struct io_ring_ctx *ctx = req->ctx; |
09bb8394 | 3308 | unsigned flags; |
9e3aa61a | 3309 | int fd, ret; |
09bb8394 | 3310 | |
cf6fd4bd PB |
3311 | flags = READ_ONCE(req->sqe->flags); |
3312 | fd = READ_ONCE(req->sqe->fd); | |
09bb8394 | 3313 | |
4fe2c963 | 3314 | if (flags & IOSQE_IO_DRAIN) |
de0617e4 | 3315 | req->flags |= REQ_F_IO_DRAIN; |
de0617e4 | 3316 | |
9e3aa61a JA |
3317 | ret = io_op_needs_file(req->sqe); |
3318 | if (ret <= 0) | |
3319 | return ret; | |
09bb8394 JA |
3320 | |
3321 | if (flags & IOSQE_FIXED_FILE) { | |
65e19f54 | 3322 | if (unlikely(!ctx->file_table || |
09bb8394 JA |
3323 | (unsigned) fd >= ctx->nr_user_files)) |
3324 | return -EBADF; | |
b7620121 | 3325 | fd = array_index_nospec(fd, ctx->nr_user_files); |
65e19f54 JA |
3326 | req->file = io_file_from_index(ctx, fd); |
3327 | if (!req->file) | |
08a45173 | 3328 | return -EBADF; |
09bb8394 JA |
3329 | req->flags |= REQ_F_FIXED_FILE; |
3330 | } else { | |
cf6fd4bd | 3331 | if (req->needs_fixed_file) |
09bb8394 | 3332 | return -EBADF; |
c826bd7a | 3333 | trace_io_uring_file_get(ctx, fd); |
09bb8394 JA |
3334 | req->file = io_file_get(state, fd); |
3335 | if (unlikely(!req->file)) | |
3336 | return -EBADF; | |
3337 | } | |
3338 | ||
3339 | return 0; | |
3340 | } | |
3341 | ||
a197f664 | 3342 | static int io_grab_files(struct io_kiocb *req) |
fcb323cc JA |
3343 | { |
3344 | int ret = -EBADF; | |
a197f664 | 3345 | struct io_ring_ctx *ctx = req->ctx; |
fcb323cc JA |
3346 | |
3347 | rcu_read_lock(); | |
3348 | spin_lock_irq(&ctx->inflight_lock); | |
3349 | /* | |
3350 | * We use the f_ops->flush() handler to ensure that we can flush | |
3351 | * out work accessing these files if the fd is closed. Check if | |
3352 | * the fd has changed since we started down this path, and disallow | |
3353 | * this operation if it has. | |
3354 | */ | |
cf6fd4bd | 3355 | if (fcheck(req->ring_fd) == req->ring_file) { |
fcb323cc JA |
3356 | list_add(&req->inflight_entry, &ctx->inflight_list); |
3357 | req->flags |= REQ_F_INFLIGHT; | |
3358 | req->work.files = current->files; | |
3359 | ret = 0; | |
3360 | } | |
3361 | spin_unlock_irq(&ctx->inflight_lock); | |
3362 | rcu_read_unlock(); | |
3363 | ||
3364 | return ret; | |
3365 | } | |
3366 | ||
2665abfd | 3367 | static enum hrtimer_restart io_link_timeout_fn(struct hrtimer *timer) |
2b188cc1 | 3368 | { |
ad8a48ac JA |
3369 | struct io_timeout_data *data = container_of(timer, |
3370 | struct io_timeout_data, timer); | |
3371 | struct io_kiocb *req = data->req; | |
2665abfd JA |
3372 | struct io_ring_ctx *ctx = req->ctx; |
3373 | struct io_kiocb *prev = NULL; | |
3374 | unsigned long flags; | |
2665abfd JA |
3375 | |
3376 | spin_lock_irqsave(&ctx->completion_lock, flags); | |
3377 | ||
3378 | /* | |
3379 | * We don't expect the list to be empty, that will only happen if we | |
3380 | * race with the completion of the linked work. | |
3381 | */ | |
4493233e PB |
3382 | if (!list_empty(&req->link_list)) { |
3383 | prev = list_entry(req->link_list.prev, struct io_kiocb, | |
3384 | link_list); | |
5d960724 | 3385 | if (refcount_inc_not_zero(&prev->refs)) { |
4493233e | 3386 | list_del_init(&req->link_list); |
5d960724 JA |
3387 | prev->flags &= ~REQ_F_LINK_TIMEOUT; |
3388 | } else | |
76a46e06 | 3389 | prev = NULL; |
2665abfd JA |
3390 | } |
3391 | ||
3392 | spin_unlock_irqrestore(&ctx->completion_lock, flags); | |
3393 | ||
3394 | if (prev) { | |
4e88d6e7 | 3395 | req_set_fail_links(prev); |
b0dd8a41 JA |
3396 | io_async_find_and_cancel(ctx, req, prev->user_data, NULL, |
3397 | -ETIME); | |
76a46e06 | 3398 | io_put_req(prev); |
47f46768 JA |
3399 | } else { |
3400 | io_cqring_add_event(req, -ETIME); | |
3401 | io_put_req(req); | |
2665abfd | 3402 | } |
2665abfd JA |
3403 | return HRTIMER_NORESTART; |
3404 | } | |
3405 | ||
ad8a48ac | 3406 | static void io_queue_linked_timeout(struct io_kiocb *req) |
2665abfd | 3407 | { |
76a46e06 | 3408 | struct io_ring_ctx *ctx = req->ctx; |
2665abfd | 3409 | |
76a46e06 JA |
3410 | /* |
3411 | * If the list is now empty, then our linked request finished before | |
3412 | * we got a chance to setup the timer | |
3413 | */ | |
3414 | spin_lock_irq(&ctx->completion_lock); | |
4493233e | 3415 | if (!list_empty(&req->link_list)) { |
2d28390a | 3416 | struct io_timeout_data *data = &req->io->timeout; |
94ae5e77 | 3417 | |
ad8a48ac JA |
3418 | data->timer.function = io_link_timeout_fn; |
3419 | hrtimer_start(&data->timer, timespec64_to_ktime(data->ts), | |
3420 | data->mode); | |
2665abfd | 3421 | } |
76a46e06 | 3422 | spin_unlock_irq(&ctx->completion_lock); |
2665abfd | 3423 | |
2665abfd | 3424 | /* drop submission reference */ |
76a46e06 JA |
3425 | io_put_req(req); |
3426 | } | |
2665abfd | 3427 | |
ad8a48ac | 3428 | static struct io_kiocb *io_prep_linked_timeout(struct io_kiocb *req) |
2665abfd JA |
3429 | { |
3430 | struct io_kiocb *nxt; | |
3431 | ||
3432 | if (!(req->flags & REQ_F_LINK)) | |
3433 | return NULL; | |
3434 | ||
4493233e PB |
3435 | nxt = list_first_entry_or_null(&req->link_list, struct io_kiocb, |
3436 | link_list); | |
cf6fd4bd | 3437 | if (!nxt || nxt->sqe->opcode != IORING_OP_LINK_TIMEOUT) |
76a46e06 | 3438 | return NULL; |
2665abfd | 3439 | |
76a46e06 | 3440 | req->flags |= REQ_F_LINK_TIMEOUT; |
76a46e06 | 3441 | return nxt; |
2665abfd JA |
3442 | } |
3443 | ||
0e0702da | 3444 | static void __io_queue_sqe(struct io_kiocb *req) |
2b188cc1 | 3445 | { |
4a0a7a18 | 3446 | struct io_kiocb *linked_timeout; |
f9bd67f6 | 3447 | struct io_kiocb *nxt = NULL; |
e0c5c576 | 3448 | int ret; |
2b188cc1 | 3449 | |
4a0a7a18 JA |
3450 | again: |
3451 | linked_timeout = io_prep_linked_timeout(req); | |
3452 | ||
f9bd67f6 | 3453 | ret = io_issue_sqe(req, &nxt, true); |
491381ce JA |
3454 | |
3455 | /* | |
3456 | * We async punt it if the file wasn't marked NOWAIT, or if the file | |
3457 | * doesn't support non-blocking read/write attempts | |
3458 | */ | |
3459 | if (ret == -EAGAIN && (!(req->flags & REQ_F_NOWAIT) || | |
3460 | (req->flags & REQ_F_MUST_PUNT))) { | |
bbad27b2 PB |
3461 | if (req->work.flags & IO_WQ_WORK_NEEDS_FILES) { |
3462 | ret = io_grab_files(req); | |
3463 | if (ret) | |
3464 | goto err; | |
2b188cc1 | 3465 | } |
bbad27b2 PB |
3466 | |
3467 | /* | |
3468 | * Queued up for async execution, worker will release | |
3469 | * submit reference when the iocb is actually submitted. | |
3470 | */ | |
3471 | io_queue_async_work(req); | |
4a0a7a18 | 3472 | goto done_req; |
2b188cc1 | 3473 | } |
e65ef56d | 3474 | |
fcb323cc | 3475 | err: |
76a46e06 | 3476 | /* drop submission reference */ |
ec9c02ad | 3477 | io_put_req(req); |
e65ef56d | 3478 | |
f9bd67f6 | 3479 | if (linked_timeout) { |
76a46e06 | 3480 | if (!ret) |
f9bd67f6 | 3481 | io_queue_linked_timeout(linked_timeout); |
76a46e06 | 3482 | else |
f9bd67f6 | 3483 | io_put_req(linked_timeout); |
76a46e06 JA |
3484 | } |
3485 | ||
e65ef56d | 3486 | /* and drop final reference, if we failed */ |
9e645e11 | 3487 | if (ret) { |
78e19bbe | 3488 | io_cqring_add_event(req, ret); |
4e88d6e7 | 3489 | req_set_fail_links(req); |
e65ef56d | 3490 | io_put_req(req); |
9e645e11 | 3491 | } |
4a0a7a18 JA |
3492 | done_req: |
3493 | if (nxt) { | |
3494 | req = nxt; | |
3495 | nxt = NULL; | |
3496 | goto again; | |
3497 | } | |
2b188cc1 JA |
3498 | } |
3499 | ||
0e0702da | 3500 | static void io_queue_sqe(struct io_kiocb *req) |
4fe2c963 JL |
3501 | { |
3502 | int ret; | |
3503 | ||
1b4a51b6 PB |
3504 | if (unlikely(req->ctx->drain_next)) { |
3505 | req->flags |= REQ_F_IO_DRAIN; | |
3506 | req->ctx->drain_next = false; | |
3507 | } | |
3508 | req->ctx->drain_next = (req->flags & REQ_F_DRAIN_LINK); | |
3509 | ||
a197f664 | 3510 | ret = io_req_defer(req); |
4fe2c963 JL |
3511 | if (ret) { |
3512 | if (ret != -EIOCBQUEUED) { | |
78e19bbe | 3513 | io_cqring_add_event(req, ret); |
4e88d6e7 | 3514 | req_set_fail_links(req); |
78e19bbe | 3515 | io_double_put_req(req); |
4fe2c963 | 3516 | } |
0e0702da JA |
3517 | } else |
3518 | __io_queue_sqe(req); | |
4fe2c963 JL |
3519 | } |
3520 | ||
1b4a51b6 | 3521 | static inline void io_queue_link_head(struct io_kiocb *req) |
4fe2c963 | 3522 | { |
94ae5e77 | 3523 | if (unlikely(req->flags & REQ_F_FAIL_LINK)) { |
1b4a51b6 PB |
3524 | io_cqring_add_event(req, -ECANCELED); |
3525 | io_double_put_req(req); | |
3526 | } else | |
0e0702da | 3527 | io_queue_sqe(req); |
4fe2c963 JL |
3528 | } |
3529 | ||
4e88d6e7 JA |
3530 | #define SQE_VALID_FLAGS (IOSQE_FIXED_FILE|IOSQE_IO_DRAIN|IOSQE_IO_LINK| \ |
3531 | IOSQE_IO_HARDLINK) | |
9e645e11 | 3532 | |
2e6e1fde | 3533 | static bool io_submit_sqe(struct io_kiocb *req, struct io_submit_state *state, |
a197f664 | 3534 | struct io_kiocb **link) |
9e645e11 | 3535 | { |
a197f664 | 3536 | struct io_ring_ctx *ctx = req->ctx; |
9e645e11 JA |
3537 | int ret; |
3538 | ||
cf6fd4bd | 3539 | req->user_data = req->sqe->user_data; |
78e19bbe | 3540 | |
9e645e11 | 3541 | /* enforce forwards compatibility on users */ |
cf6fd4bd | 3542 | if (unlikely(req->sqe->flags & ~SQE_VALID_FLAGS)) { |
9e645e11 | 3543 | ret = -EINVAL; |
196be95c | 3544 | goto err_req; |
9e645e11 JA |
3545 | } |
3546 | ||
a197f664 | 3547 | ret = io_req_set_file(state, req); |
9e645e11 JA |
3548 | if (unlikely(ret)) { |
3549 | err_req: | |
78e19bbe JA |
3550 | io_cqring_add_event(req, ret); |
3551 | io_double_put_req(req); | |
2e6e1fde | 3552 | return false; |
9e645e11 JA |
3553 | } |
3554 | ||
9e645e11 JA |
3555 | /* |
3556 | * If we already have a head request, queue this one for async | |
3557 | * submittal once the head completes. If we don't have a head but | |
3558 | * IOSQE_IO_LINK is set in the sqe, start a new head. This one will be | |
3559 | * submitted sync once the chain is complete. If none of those | |
3560 | * conditions are true (normal request), then just queue it. | |
3561 | */ | |
3562 | if (*link) { | |
3563 | struct io_kiocb *prev = *link; | |
3564 | ||
cf6fd4bd | 3565 | if (req->sqe->flags & IOSQE_IO_DRAIN) |
1b4a51b6 PB |
3566 | (*link)->flags |= REQ_F_DRAIN_LINK | REQ_F_IO_DRAIN; |
3567 | ||
4e88d6e7 JA |
3568 | if (req->sqe->flags & IOSQE_IO_HARDLINK) |
3569 | req->flags |= REQ_F_HARDLINK; | |
3570 | ||
b7bb4f7d | 3571 | if (io_alloc_async_ctx(req)) { |
9e645e11 JA |
3572 | ret = -EAGAIN; |
3573 | goto err_req; | |
3574 | } | |
3575 | ||
b7bb4f7d | 3576 | ret = io_req_defer_prep(req); |
2d28390a | 3577 | if (ret) { |
4e88d6e7 | 3578 | /* fail even hard links since we don't submit */ |
2d28390a | 3579 | prev->flags |= REQ_F_FAIL_LINK; |
f67676d1 | 3580 | goto err_req; |
2d28390a | 3581 | } |
c826bd7a | 3582 | trace_io_uring_link(ctx, req, prev); |
4493233e | 3583 | list_add_tail(&req->link_list, &prev->link_list); |
4e88d6e7 | 3584 | } else if (req->sqe->flags & (IOSQE_IO_LINK|IOSQE_IO_HARDLINK)) { |
9e645e11 | 3585 | req->flags |= REQ_F_LINK; |
4e88d6e7 JA |
3586 | if (req->sqe->flags & IOSQE_IO_HARDLINK) |
3587 | req->flags |= REQ_F_HARDLINK; | |
9e645e11 | 3588 | |
9e645e11 JA |
3589 | INIT_LIST_HEAD(&req->link_list); |
3590 | *link = req; | |
3591 | } else { | |
a197f664 | 3592 | io_queue_sqe(req); |
9e645e11 | 3593 | } |
2e6e1fde PB |
3594 | |
3595 | return true; | |
9e645e11 JA |
3596 | } |
3597 | ||
9a56a232 JA |
3598 | /* |
3599 | * Batched submission is done, ensure local IO is flushed out. | |
3600 | */ | |
3601 | static void io_submit_state_end(struct io_submit_state *state) | |
3602 | { | |
3603 | blk_finish_plug(&state->plug); | |
3d6770fb | 3604 | io_file_put(state); |
2579f913 JA |
3605 | if (state->free_reqs) |
3606 | kmem_cache_free_bulk(req_cachep, state->free_reqs, | |
3607 | &state->reqs[state->cur_req]); | |
9a56a232 JA |
3608 | } |
3609 | ||
3610 | /* | |
3611 | * Start submission side cache. | |
3612 | */ | |
3613 | static void io_submit_state_start(struct io_submit_state *state, | |
22efde59 | 3614 | unsigned int max_ios) |
9a56a232 JA |
3615 | { |
3616 | blk_start_plug(&state->plug); | |
2579f913 | 3617 | state->free_reqs = 0; |
9a56a232 JA |
3618 | state->file = NULL; |
3619 | state->ios_left = max_ios; | |
3620 | } | |
3621 | ||
2b188cc1 JA |
3622 | static void io_commit_sqring(struct io_ring_ctx *ctx) |
3623 | { | |
75b28aff | 3624 | struct io_rings *rings = ctx->rings; |
2b188cc1 | 3625 | |
75b28aff | 3626 | if (ctx->cached_sq_head != READ_ONCE(rings->sq.head)) { |
2b188cc1 JA |
3627 | /* |
3628 | * Ensure any loads from the SQEs are done at this point, | |
3629 | * since once we write the new head, the application could | |
3630 | * write new data to them. | |
3631 | */ | |
75b28aff | 3632 | smp_store_release(&rings->sq.head, ctx->cached_sq_head); |
2b188cc1 JA |
3633 | } |
3634 | } | |
3635 | ||
2b188cc1 | 3636 | /* |
d195a66e | 3637 | * Fetch an sqe, if one is available. Note that req->sqe will point to memory |
2b188cc1 JA |
3638 | * that is mapped by userspace. This means that care needs to be taken to |
3639 | * ensure that reads are stable, as we cannot rely on userspace always | |
3640 | * being a good citizen. If members of the sqe are validated and then later | |
3641 | * used, it's important that those reads are done through READ_ONCE() to | |
3642 | * prevent a re-load down the line. | |
3643 | */ | |
cf6fd4bd | 3644 | static bool io_get_sqring(struct io_ring_ctx *ctx, struct io_kiocb *req) |
2b188cc1 | 3645 | { |
75b28aff HV |
3646 | struct io_rings *rings = ctx->rings; |
3647 | u32 *sq_array = ctx->sq_array; | |
2b188cc1 JA |
3648 | unsigned head; |
3649 | ||
3650 | /* | |
3651 | * The cached sq head (or cq tail) serves two purposes: | |
3652 | * | |
3653 | * 1) allows us to batch the cost of updating the user visible | |
3654 | * head updates. | |
3655 | * 2) allows the kernel side to track the head on its own, even | |
3656 | * though the application is the one updating it. | |
3657 | */ | |
3658 | head = ctx->cached_sq_head; | |
e523a29c | 3659 | /* make sure SQ entry isn't read before tail */ |
9835d6fa | 3660 | if (unlikely(head == smp_load_acquire(&rings->sq.tail))) |
2b188cc1 JA |
3661 | return false; |
3662 | ||
75b28aff | 3663 | head = READ_ONCE(sq_array[head & ctx->sq_mask]); |
9835d6fa | 3664 | if (likely(head < ctx->sq_entries)) { |
cf6fd4bd PB |
3665 | /* |
3666 | * All io need record the previous position, if LINK vs DARIN, | |
3667 | * it can be used to mark the position of the first IO in the | |
3668 | * link list. | |
3669 | */ | |
3670 | req->sequence = ctx->cached_sq_head; | |
3671 | req->sqe = &ctx->sq_sqes[head]; | |
2b188cc1 JA |
3672 | ctx->cached_sq_head++; |
3673 | return true; | |
3674 | } | |
3675 | ||
3676 | /* drop invalid entries */ | |
3677 | ctx->cached_sq_head++; | |
498ccd9e JA |
3678 | ctx->cached_sq_dropped++; |
3679 | WRITE_ONCE(rings->sq_dropped, ctx->cached_sq_dropped); | |
2b188cc1 JA |
3680 | return false; |
3681 | } | |
3682 | ||
fb5ccc98 | 3683 | static int io_submit_sqes(struct io_ring_ctx *ctx, unsigned int nr, |
ae9428ca PB |
3684 | struct file *ring_file, int ring_fd, |
3685 | struct mm_struct **mm, bool async) | |
6c271ce2 JA |
3686 | { |
3687 | struct io_submit_state state, *statep = NULL; | |
9e645e11 | 3688 | struct io_kiocb *link = NULL; |
9e645e11 | 3689 | int i, submitted = 0; |
95a1b3ff | 3690 | bool mm_fault = false; |
6c271ce2 | 3691 | |
c4a2ed72 JA |
3692 | /* if we have a backlog and couldn't flush it all, return BUSY */ |
3693 | if (!list_empty(&ctx->cq_overflow_list) && | |
3694 | !io_cqring_overflow_flush(ctx, false)) | |
1d7bb1d5 | 3695 | return -EBUSY; |
6c271ce2 JA |
3696 | |
3697 | if (nr > IO_PLUG_THRESHOLD) { | |
22efde59 | 3698 | io_submit_state_start(&state, nr); |
6c271ce2 JA |
3699 | statep = &state; |
3700 | } | |
3701 | ||
3702 | for (i = 0; i < nr; i++) { | |
196be95c | 3703 | struct io_kiocb *req; |
50585b9a | 3704 | unsigned int sqe_flags; |
fb5ccc98 | 3705 | |
196be95c PB |
3706 | req = io_get_req(ctx, statep); |
3707 | if (unlikely(!req)) { | |
3708 | if (!submitted) | |
3709 | submitted = -EAGAIN; | |
fb5ccc98 | 3710 | break; |
196be95c | 3711 | } |
cf6fd4bd | 3712 | if (!io_get_sqring(ctx, req)) { |
196be95c PB |
3713 | __io_free_req(req); |
3714 | break; | |
3715 | } | |
fb5ccc98 | 3716 | |
cf6fd4bd | 3717 | if (io_sqe_needs_user(req->sqe) && !*mm) { |
95a1b3ff PB |
3718 | mm_fault = mm_fault || !mmget_not_zero(ctx->sqo_mm); |
3719 | if (!mm_fault) { | |
3720 | use_mm(ctx->sqo_mm); | |
3721 | *mm = ctx->sqo_mm; | |
3722 | } | |
9e645e11 | 3723 | } |
9e645e11 | 3724 | |
2e6e1fde | 3725 | submitted++; |
cf6fd4bd | 3726 | sqe_flags = req->sqe->flags; |
50585b9a | 3727 | |
cf6fd4bd PB |
3728 | req->ring_file = ring_file; |
3729 | req->ring_fd = ring_fd; | |
3730 | req->has_user = *mm != NULL; | |
3731 | req->in_async = async; | |
3732 | req->needs_fixed_file = async; | |
3733 | trace_io_uring_submit_sqe(ctx, req->sqe->user_data, | |
50585b9a | 3734 | true, async); |
2e6e1fde PB |
3735 | if (!io_submit_sqe(req, statep, &link)) |
3736 | break; | |
e5eb6366 PB |
3737 | /* |
3738 | * If previous wasn't linked and we have a linked command, | |
3739 | * that's the end of the chain. Submit the previous link. | |
3740 | */ | |
ffbb8d6b | 3741 | if (!(sqe_flags & (IOSQE_IO_LINK|IOSQE_IO_HARDLINK)) && link) { |
1b4a51b6 | 3742 | io_queue_link_head(link); |
e5eb6366 | 3743 | link = NULL; |
6c271ce2 | 3744 | } |
6c271ce2 JA |
3745 | } |
3746 | ||
9e645e11 | 3747 | if (link) |
1b4a51b6 | 3748 | io_queue_link_head(link); |
6c271ce2 JA |
3749 | if (statep) |
3750 | io_submit_state_end(&state); | |
3751 | ||
ae9428ca PB |
3752 | /* Commit SQ ring head once we've consumed and submitted all SQEs */ |
3753 | io_commit_sqring(ctx); | |
3754 | ||
6c271ce2 JA |
3755 | return submitted; |
3756 | } | |
3757 | ||
3758 | static int io_sq_thread(void *data) | |
3759 | { | |
6c271ce2 JA |
3760 | struct io_ring_ctx *ctx = data; |
3761 | struct mm_struct *cur_mm = NULL; | |
181e448d | 3762 | const struct cred *old_cred; |
6c271ce2 JA |
3763 | mm_segment_t old_fs; |
3764 | DEFINE_WAIT(wait); | |
3765 | unsigned inflight; | |
3766 | unsigned long timeout; | |
c1edbf5f | 3767 | int ret; |
6c271ce2 | 3768 | |
206aefde | 3769 | complete(&ctx->completions[1]); |
a4c0b3de | 3770 | |
6c271ce2 JA |
3771 | old_fs = get_fs(); |
3772 | set_fs(USER_DS); | |
181e448d | 3773 | old_cred = override_creds(ctx->creds); |
6c271ce2 | 3774 | |
c1edbf5f | 3775 | ret = timeout = inflight = 0; |
2bbcd6d3 | 3776 | while (!kthread_should_park()) { |
fb5ccc98 | 3777 | unsigned int to_submit; |
6c271ce2 JA |
3778 | |
3779 | if (inflight) { | |
3780 | unsigned nr_events = 0; | |
3781 | ||
3782 | if (ctx->flags & IORING_SETUP_IOPOLL) { | |
2b2ed975 JA |
3783 | /* |
3784 | * inflight is the count of the maximum possible | |
3785 | * entries we submitted, but it can be smaller | |
3786 | * if we dropped some of them. If we don't have | |
3787 | * poll entries available, then we know that we | |
3788 | * have nothing left to poll for. Reset the | |
3789 | * inflight count to zero in that case. | |
3790 | */ | |
3791 | mutex_lock(&ctx->uring_lock); | |
3792 | if (!list_empty(&ctx->poll_list)) | |
3793 | __io_iopoll_check(ctx, &nr_events, 0); | |
3794 | else | |
3795 | inflight = 0; | |
3796 | mutex_unlock(&ctx->uring_lock); | |
6c271ce2 JA |
3797 | } else { |
3798 | /* | |
3799 | * Normal IO, just pretend everything completed. | |
3800 | * We don't have to poll completions for that. | |
3801 | */ | |
3802 | nr_events = inflight; | |
3803 | } | |
3804 | ||
3805 | inflight -= nr_events; | |
3806 | if (!inflight) | |
3807 | timeout = jiffies + ctx->sq_thread_idle; | |
3808 | } | |
3809 | ||
fb5ccc98 | 3810 | to_submit = io_sqring_entries(ctx); |
c1edbf5f JA |
3811 | |
3812 | /* | |
3813 | * If submit got -EBUSY, flag us as needing the application | |
3814 | * to enter the kernel to reap and flush events. | |
3815 | */ | |
3816 | if (!to_submit || ret == -EBUSY) { | |
6c271ce2 JA |
3817 | /* |
3818 | * We're polling. If we're within the defined idle | |
3819 | * period, then let us spin without work before going | |
c1edbf5f JA |
3820 | * to sleep. The exception is if we got EBUSY doing |
3821 | * more IO, we should wait for the application to | |
3822 | * reap events and wake us up. | |
6c271ce2 | 3823 | */ |
c1edbf5f JA |
3824 | if (inflight || |
3825 | (!time_after(jiffies, timeout) && ret != -EBUSY)) { | |
9831a90c | 3826 | cond_resched(); |
6c271ce2 JA |
3827 | continue; |
3828 | } | |
3829 | ||
3830 | /* | |
3831 | * Drop cur_mm before scheduling, we can't hold it for | |
3832 | * long periods (or over schedule()). Do this before | |
3833 | * adding ourselves to the waitqueue, as the unuse/drop | |
3834 | * may sleep. | |
3835 | */ | |
3836 | if (cur_mm) { | |
3837 | unuse_mm(cur_mm); | |
3838 | mmput(cur_mm); | |
3839 | cur_mm = NULL; | |
3840 | } | |
3841 | ||
3842 | prepare_to_wait(&ctx->sqo_wait, &wait, | |
3843 | TASK_INTERRUPTIBLE); | |
3844 | ||
3845 | /* Tell userspace we may need a wakeup call */ | |
75b28aff | 3846 | ctx->rings->sq_flags |= IORING_SQ_NEED_WAKEUP; |
0d7bae69 SB |
3847 | /* make sure to read SQ tail after writing flags */ |
3848 | smp_mb(); | |
6c271ce2 | 3849 | |
fb5ccc98 | 3850 | to_submit = io_sqring_entries(ctx); |
c1edbf5f | 3851 | if (!to_submit || ret == -EBUSY) { |
2bbcd6d3 | 3852 | if (kthread_should_park()) { |
6c271ce2 JA |
3853 | finish_wait(&ctx->sqo_wait, &wait); |
3854 | break; | |
3855 | } | |
3856 | if (signal_pending(current)) | |
3857 | flush_signals(current); | |
3858 | schedule(); | |
3859 | finish_wait(&ctx->sqo_wait, &wait); | |
3860 | ||
75b28aff | 3861 | ctx->rings->sq_flags &= ~IORING_SQ_NEED_WAKEUP; |
6c271ce2 JA |
3862 | continue; |
3863 | } | |
3864 | finish_wait(&ctx->sqo_wait, &wait); | |
3865 | ||
75b28aff | 3866 | ctx->rings->sq_flags &= ~IORING_SQ_NEED_WAKEUP; |
6c271ce2 JA |
3867 | } |
3868 | ||
fb5ccc98 | 3869 | to_submit = min(to_submit, ctx->sq_entries); |
8a4955ff | 3870 | mutex_lock(&ctx->uring_lock); |
1d7bb1d5 | 3871 | ret = io_submit_sqes(ctx, to_submit, NULL, -1, &cur_mm, true); |
8a4955ff | 3872 | mutex_unlock(&ctx->uring_lock); |
1d7bb1d5 JA |
3873 | if (ret > 0) |
3874 | inflight += ret; | |
6c271ce2 JA |
3875 | } |
3876 | ||
3877 | set_fs(old_fs); | |
3878 | if (cur_mm) { | |
3879 | unuse_mm(cur_mm); | |
3880 | mmput(cur_mm); | |
3881 | } | |
181e448d | 3882 | revert_creds(old_cred); |
06058632 | 3883 | |
2bbcd6d3 | 3884 | kthread_parkme(); |
06058632 | 3885 | |
6c271ce2 JA |
3886 | return 0; |
3887 | } | |
3888 | ||
bda52162 JA |
3889 | struct io_wait_queue { |
3890 | struct wait_queue_entry wq; | |
3891 | struct io_ring_ctx *ctx; | |
3892 | unsigned to_wait; | |
3893 | unsigned nr_timeouts; | |
3894 | }; | |
3895 | ||
1d7bb1d5 | 3896 | static inline bool io_should_wake(struct io_wait_queue *iowq, bool noflush) |
bda52162 JA |
3897 | { |
3898 | struct io_ring_ctx *ctx = iowq->ctx; | |
3899 | ||
3900 | /* | |
d195a66e | 3901 | * Wake up if we have enough events, or if a timeout occurred since we |
bda52162 JA |
3902 | * started waiting. For timeouts, we always want to return to userspace, |
3903 | * regardless of event count. | |
3904 | */ | |
1d7bb1d5 | 3905 | return io_cqring_events(ctx, noflush) >= iowq->to_wait || |
bda52162 JA |
3906 | atomic_read(&ctx->cq_timeouts) != iowq->nr_timeouts; |
3907 | } | |
3908 | ||
3909 | static int io_wake_function(struct wait_queue_entry *curr, unsigned int mode, | |
3910 | int wake_flags, void *key) | |
3911 | { | |
3912 | struct io_wait_queue *iowq = container_of(curr, struct io_wait_queue, | |
3913 | wq); | |
3914 | ||
1d7bb1d5 JA |
3915 | /* use noflush == true, as we can't safely rely on locking context */ |
3916 | if (!io_should_wake(iowq, true)) | |
bda52162 JA |
3917 | return -1; |
3918 | ||
3919 | return autoremove_wake_function(curr, mode, wake_flags, key); | |
3920 | } | |
3921 | ||
2b188cc1 JA |
3922 | /* |
3923 | * Wait until events become available, if we don't already have some. The | |
3924 | * application must reap them itself, as they reside on the shared cq ring. | |
3925 | */ | |
3926 | static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events, | |
3927 | const sigset_t __user *sig, size_t sigsz) | |
3928 | { | |
bda52162 JA |
3929 | struct io_wait_queue iowq = { |
3930 | .wq = { | |
3931 | .private = current, | |
3932 | .func = io_wake_function, | |
3933 | .entry = LIST_HEAD_INIT(iowq.wq.entry), | |
3934 | }, | |
3935 | .ctx = ctx, | |
3936 | .to_wait = min_events, | |
3937 | }; | |
75b28aff | 3938 | struct io_rings *rings = ctx->rings; |
e9ffa5c2 | 3939 | int ret = 0; |
2b188cc1 | 3940 | |
1d7bb1d5 | 3941 | if (io_cqring_events(ctx, false) >= min_events) |
2b188cc1 JA |
3942 | return 0; |
3943 | ||
3944 | if (sig) { | |
9e75ad5d AB |
3945 | #ifdef CONFIG_COMPAT |
3946 | if (in_compat_syscall()) | |
3947 | ret = set_compat_user_sigmask((const compat_sigset_t __user *)sig, | |
b772434b | 3948 | sigsz); |
9e75ad5d AB |
3949 | else |
3950 | #endif | |
b772434b | 3951 | ret = set_user_sigmask(sig, sigsz); |
9e75ad5d | 3952 | |
2b188cc1 JA |
3953 | if (ret) |
3954 | return ret; | |
3955 | } | |
3956 | ||
bda52162 | 3957 | iowq.nr_timeouts = atomic_read(&ctx->cq_timeouts); |
c826bd7a | 3958 | trace_io_uring_cqring_wait(ctx, min_events); |
bda52162 JA |
3959 | do { |
3960 | prepare_to_wait_exclusive(&ctx->wait, &iowq.wq, | |
3961 | TASK_INTERRUPTIBLE); | |
1d7bb1d5 | 3962 | if (io_should_wake(&iowq, false)) |
bda52162 JA |
3963 | break; |
3964 | schedule(); | |
3965 | if (signal_pending(current)) { | |
e9ffa5c2 | 3966 | ret = -EINTR; |
bda52162 JA |
3967 | break; |
3968 | } | |
3969 | } while (1); | |
3970 | finish_wait(&ctx->wait, &iowq.wq); | |
3971 | ||
e9ffa5c2 | 3972 | restore_saved_sigmask_unless(ret == -EINTR); |
2b188cc1 | 3973 | |
75b28aff | 3974 | return READ_ONCE(rings->cq.head) == READ_ONCE(rings->cq.tail) ? ret : 0; |
2b188cc1 JA |
3975 | } |
3976 | ||
6b06314c JA |
3977 | static void __io_sqe_files_unregister(struct io_ring_ctx *ctx) |
3978 | { | |
3979 | #if defined(CONFIG_UNIX) | |
3980 | if (ctx->ring_sock) { | |
3981 | struct sock *sock = ctx->ring_sock->sk; | |
3982 | struct sk_buff *skb; | |
3983 | ||
3984 | while ((skb = skb_dequeue(&sock->sk_receive_queue)) != NULL) | |
3985 | kfree_skb(skb); | |
3986 | } | |
3987 | #else | |
3988 | int i; | |
3989 | ||
65e19f54 JA |
3990 | for (i = 0; i < ctx->nr_user_files; i++) { |
3991 | struct file *file; | |
3992 | ||
3993 | file = io_file_from_index(ctx, i); | |
3994 | if (file) | |
3995 | fput(file); | |
3996 | } | |
6b06314c JA |
3997 | #endif |
3998 | } | |
3999 | ||
4000 | static int io_sqe_files_unregister(struct io_ring_ctx *ctx) | |
4001 | { | |
65e19f54 JA |
4002 | unsigned nr_tables, i; |
4003 | ||
4004 | if (!ctx->file_table) | |
6b06314c JA |
4005 | return -ENXIO; |
4006 | ||
4007 | __io_sqe_files_unregister(ctx); | |
65e19f54 JA |
4008 | nr_tables = DIV_ROUND_UP(ctx->nr_user_files, IORING_MAX_FILES_TABLE); |
4009 | for (i = 0; i < nr_tables; i++) | |
4010 | kfree(ctx->file_table[i].files); | |
4011 | kfree(ctx->file_table); | |
4012 | ctx->file_table = NULL; | |
6b06314c JA |
4013 | ctx->nr_user_files = 0; |
4014 | return 0; | |
4015 | } | |
4016 | ||
6c271ce2 JA |
4017 | static void io_sq_thread_stop(struct io_ring_ctx *ctx) |
4018 | { | |
4019 | if (ctx->sqo_thread) { | |
206aefde | 4020 | wait_for_completion(&ctx->completions[1]); |
2bbcd6d3 RP |
4021 | /* |
4022 | * The park is a bit of a work-around, without it we get | |
4023 | * warning spews on shutdown with SQPOLL set and affinity | |
4024 | * set to a single CPU. | |
4025 | */ | |
06058632 | 4026 | kthread_park(ctx->sqo_thread); |
6c271ce2 JA |
4027 | kthread_stop(ctx->sqo_thread); |
4028 | ctx->sqo_thread = NULL; | |
4029 | } | |
4030 | } | |
4031 | ||
6b06314c JA |
4032 | static void io_finish_async(struct io_ring_ctx *ctx) |
4033 | { | |
6c271ce2 JA |
4034 | io_sq_thread_stop(ctx); |
4035 | ||
561fb04a JA |
4036 | if (ctx->io_wq) { |
4037 | io_wq_destroy(ctx->io_wq); | |
4038 | ctx->io_wq = NULL; | |
6b06314c JA |
4039 | } |
4040 | } | |
4041 | ||
4042 | #if defined(CONFIG_UNIX) | |
4043 | static void io_destruct_skb(struct sk_buff *skb) | |
4044 | { | |
4045 | struct io_ring_ctx *ctx = skb->sk->sk_user_data; | |
8a997340 | 4046 | |
561fb04a JA |
4047 | if (ctx->io_wq) |
4048 | io_wq_flush(ctx->io_wq); | |
6b06314c | 4049 | |
6b06314c JA |
4050 | unix_destruct_scm(skb); |
4051 | } | |
4052 | ||
4053 | /* | |
4054 | * Ensure the UNIX gc is aware of our file set, so we are certain that | |
4055 | * the io_uring can be safely unregistered on process exit, even if we have | |
4056 | * loops in the file referencing. | |
4057 | */ | |
4058 | static int __io_sqe_files_scm(struct io_ring_ctx *ctx, int nr, int offset) | |
4059 | { | |
4060 | struct sock *sk = ctx->ring_sock->sk; | |
4061 | struct scm_fp_list *fpl; | |
4062 | struct sk_buff *skb; | |
08a45173 | 4063 | int i, nr_files; |
6b06314c JA |
4064 | |
4065 | if (!capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN)) { | |
4066 | unsigned long inflight = ctx->user->unix_inflight + nr; | |
4067 | ||
4068 | if (inflight > task_rlimit(current, RLIMIT_NOFILE)) | |
4069 | return -EMFILE; | |
4070 | } | |
4071 | ||
4072 | fpl = kzalloc(sizeof(*fpl), GFP_KERNEL); | |
4073 | if (!fpl) | |
4074 | return -ENOMEM; | |
4075 | ||
4076 | skb = alloc_skb(0, GFP_KERNEL); | |
4077 | if (!skb) { | |
4078 | kfree(fpl); | |
4079 | return -ENOMEM; | |
4080 | } | |
4081 | ||
4082 | skb->sk = sk; | |
6b06314c | 4083 | |
08a45173 | 4084 | nr_files = 0; |
6b06314c JA |
4085 | fpl->user = get_uid(ctx->user); |
4086 | for (i = 0; i < nr; i++) { | |
65e19f54 JA |
4087 | struct file *file = io_file_from_index(ctx, i + offset); |
4088 | ||
4089 | if (!file) | |
08a45173 | 4090 | continue; |
65e19f54 | 4091 | fpl->fp[nr_files] = get_file(file); |
08a45173 JA |
4092 | unix_inflight(fpl->user, fpl->fp[nr_files]); |
4093 | nr_files++; | |
6b06314c JA |
4094 | } |
4095 | ||
08a45173 JA |
4096 | if (nr_files) { |
4097 | fpl->max = SCM_MAX_FD; | |
4098 | fpl->count = nr_files; | |
4099 | UNIXCB(skb).fp = fpl; | |
4100 | skb->destructor = io_destruct_skb; | |
4101 | refcount_add(skb->truesize, &sk->sk_wmem_alloc); | |
4102 | skb_queue_head(&sk->sk_receive_queue, skb); | |
6b06314c | 4103 | |
08a45173 JA |
4104 | for (i = 0; i < nr_files; i++) |
4105 | fput(fpl->fp[i]); | |
4106 | } else { | |
4107 | kfree_skb(skb); | |
4108 | kfree(fpl); | |
4109 | } | |
6b06314c JA |
4110 | |
4111 | return 0; | |
4112 | } | |
4113 | ||
4114 | /* | |
4115 | * If UNIX sockets are enabled, fd passing can cause a reference cycle which | |
4116 | * causes regular reference counting to break down. We rely on the UNIX | |
4117 | * garbage collection to take care of this problem for us. | |
4118 | */ | |
4119 | static int io_sqe_files_scm(struct io_ring_ctx *ctx) | |
4120 | { | |
4121 | unsigned left, total; | |
4122 | int ret = 0; | |
4123 | ||
4124 | total = 0; | |
4125 | left = ctx->nr_user_files; | |
4126 | while (left) { | |
4127 | unsigned this_files = min_t(unsigned, left, SCM_MAX_FD); | |
6b06314c JA |
4128 | |
4129 | ret = __io_sqe_files_scm(ctx, this_files, total); | |
4130 | if (ret) | |
4131 | break; | |
4132 | left -= this_files; | |
4133 | total += this_files; | |
4134 | } | |
4135 | ||
4136 | if (!ret) | |
4137 | return 0; | |
4138 | ||
4139 | while (total < ctx->nr_user_files) { | |
65e19f54 JA |
4140 | struct file *file = io_file_from_index(ctx, total); |
4141 | ||
4142 | if (file) | |
4143 | fput(file); | |
6b06314c JA |
4144 | total++; |
4145 | } | |
4146 | ||
4147 | return ret; | |
4148 | } | |
4149 | #else | |
4150 | static int io_sqe_files_scm(struct io_ring_ctx *ctx) | |
4151 | { | |
4152 | return 0; | |
4153 | } | |
4154 | #endif | |
4155 | ||
65e19f54 JA |
4156 | static int io_sqe_alloc_file_tables(struct io_ring_ctx *ctx, unsigned nr_tables, |
4157 | unsigned nr_files) | |
4158 | { | |
4159 | int i; | |
4160 | ||
4161 | for (i = 0; i < nr_tables; i++) { | |
4162 | struct fixed_file_table *table = &ctx->file_table[i]; | |
4163 | unsigned this_files; | |
4164 | ||
4165 | this_files = min(nr_files, IORING_MAX_FILES_TABLE); | |
4166 | table->files = kcalloc(this_files, sizeof(struct file *), | |
4167 | GFP_KERNEL); | |
4168 | if (!table->files) | |
4169 | break; | |
4170 | nr_files -= this_files; | |
4171 | } | |
4172 | ||
4173 | if (i == nr_tables) | |
4174 | return 0; | |
4175 | ||
4176 | for (i = 0; i < nr_tables; i++) { | |
4177 | struct fixed_file_table *table = &ctx->file_table[i]; | |
4178 | kfree(table->files); | |
4179 | } | |
4180 | return 1; | |
4181 | } | |
4182 | ||
6b06314c JA |
4183 | static int io_sqe_files_register(struct io_ring_ctx *ctx, void __user *arg, |
4184 | unsigned nr_args) | |
4185 | { | |
4186 | __s32 __user *fds = (__s32 __user *) arg; | |
65e19f54 | 4187 | unsigned nr_tables; |
6b06314c JA |
4188 | int fd, ret = 0; |
4189 | unsigned i; | |
4190 | ||
65e19f54 | 4191 | if (ctx->file_table) |
6b06314c JA |
4192 | return -EBUSY; |
4193 | if (!nr_args) | |
4194 | return -EINVAL; | |
4195 | if (nr_args > IORING_MAX_FIXED_FILES) | |
4196 | return -EMFILE; | |
4197 | ||
65e19f54 JA |
4198 | nr_tables = DIV_ROUND_UP(nr_args, IORING_MAX_FILES_TABLE); |
4199 | ctx->file_table = kcalloc(nr_tables, sizeof(struct fixed_file_table), | |
4200 | GFP_KERNEL); | |
4201 | if (!ctx->file_table) | |
6b06314c JA |
4202 | return -ENOMEM; |
4203 | ||
65e19f54 JA |
4204 | if (io_sqe_alloc_file_tables(ctx, nr_tables, nr_args)) { |
4205 | kfree(ctx->file_table); | |
46568e9b | 4206 | ctx->file_table = NULL; |
65e19f54 JA |
4207 | return -ENOMEM; |
4208 | } | |
4209 | ||
08a45173 | 4210 | for (i = 0; i < nr_args; i++, ctx->nr_user_files++) { |
65e19f54 JA |
4211 | struct fixed_file_table *table; |
4212 | unsigned index; | |
4213 | ||
6b06314c JA |
4214 | ret = -EFAULT; |
4215 | if (copy_from_user(&fd, &fds[i], sizeof(fd))) | |
4216 | break; | |
08a45173 JA |
4217 | /* allow sparse sets */ |
4218 | if (fd == -1) { | |
4219 | ret = 0; | |
4220 | continue; | |
4221 | } | |
6b06314c | 4222 | |
65e19f54 JA |
4223 | table = &ctx->file_table[i >> IORING_FILE_TABLE_SHIFT]; |
4224 | index = i & IORING_FILE_TABLE_MASK; | |
4225 | table->files[index] = fget(fd); | |
6b06314c JA |
4226 | |
4227 | ret = -EBADF; | |
65e19f54 | 4228 | if (!table->files[index]) |
6b06314c JA |
4229 | break; |
4230 | /* | |
4231 | * Don't allow io_uring instances to be registered. If UNIX | |
4232 | * isn't enabled, then this causes a reference cycle and this | |
4233 | * instance can never get freed. If UNIX is enabled we'll | |
4234 | * handle it just fine, but there's still no point in allowing | |
4235 | * a ring fd as it doesn't support regular read/write anyway. | |
4236 | */ | |
65e19f54 JA |
4237 | if (table->files[index]->f_op == &io_uring_fops) { |
4238 | fput(table->files[index]); | |
6b06314c JA |
4239 | break; |
4240 | } | |
6b06314c JA |
4241 | ret = 0; |
4242 | } | |
4243 | ||
4244 | if (ret) { | |
65e19f54 JA |
4245 | for (i = 0; i < ctx->nr_user_files; i++) { |
4246 | struct file *file; | |
6b06314c | 4247 | |
65e19f54 JA |
4248 | file = io_file_from_index(ctx, i); |
4249 | if (file) | |
4250 | fput(file); | |
4251 | } | |
4252 | for (i = 0; i < nr_tables; i++) | |
4253 | kfree(ctx->file_table[i].files); | |
6b06314c | 4254 | |
65e19f54 JA |
4255 | kfree(ctx->file_table); |
4256 | ctx->file_table = NULL; | |
6b06314c JA |
4257 | ctx->nr_user_files = 0; |
4258 | return ret; | |
4259 | } | |
4260 | ||
4261 | ret = io_sqe_files_scm(ctx); | |
4262 | if (ret) | |
4263 | io_sqe_files_unregister(ctx); | |
4264 | ||
4265 | return ret; | |
4266 | } | |
4267 | ||
c3a31e60 JA |
4268 | static void io_sqe_file_unregister(struct io_ring_ctx *ctx, int index) |
4269 | { | |
4270 | #if defined(CONFIG_UNIX) | |
65e19f54 | 4271 | struct file *file = io_file_from_index(ctx, index); |
c3a31e60 JA |
4272 | struct sock *sock = ctx->ring_sock->sk; |
4273 | struct sk_buff_head list, *head = &sock->sk_receive_queue; | |
4274 | struct sk_buff *skb; | |
4275 | int i; | |
4276 | ||
4277 | __skb_queue_head_init(&list); | |
4278 | ||
4279 | /* | |
4280 | * Find the skb that holds this file in its SCM_RIGHTS. When found, | |
4281 | * remove this entry and rearrange the file array. | |
4282 | */ | |
4283 | skb = skb_dequeue(head); | |
4284 | while (skb) { | |
4285 | struct scm_fp_list *fp; | |
4286 | ||
4287 | fp = UNIXCB(skb).fp; | |
4288 | for (i = 0; i < fp->count; i++) { | |
4289 | int left; | |
4290 | ||
4291 | if (fp->fp[i] != file) | |
4292 | continue; | |
4293 | ||
4294 | unix_notinflight(fp->user, fp->fp[i]); | |
4295 | left = fp->count - 1 - i; | |
4296 | if (left) { | |
4297 | memmove(&fp->fp[i], &fp->fp[i + 1], | |
4298 | left * sizeof(struct file *)); | |
4299 | } | |
4300 | fp->count--; | |
4301 | if (!fp->count) { | |
4302 | kfree_skb(skb); | |
4303 | skb = NULL; | |
4304 | } else { | |
4305 | __skb_queue_tail(&list, skb); | |
4306 | } | |
4307 | fput(file); | |
4308 | file = NULL; | |
4309 | break; | |
4310 | } | |
4311 | ||
4312 | if (!file) | |
4313 | break; | |
4314 | ||
4315 | __skb_queue_tail(&list, skb); | |
4316 | ||
4317 | skb = skb_dequeue(head); | |
4318 | } | |
4319 | ||
4320 | if (skb_peek(&list)) { | |
4321 | spin_lock_irq(&head->lock); | |
4322 | while ((skb = __skb_dequeue(&list)) != NULL) | |
4323 | __skb_queue_tail(head, skb); | |
4324 | spin_unlock_irq(&head->lock); | |
4325 | } | |
4326 | #else | |
65e19f54 | 4327 | fput(io_file_from_index(ctx, index)); |
c3a31e60 JA |
4328 | #endif |
4329 | } | |
4330 | ||
4331 | static int io_sqe_file_register(struct io_ring_ctx *ctx, struct file *file, | |
4332 | int index) | |
4333 | { | |
4334 | #if defined(CONFIG_UNIX) | |
4335 | struct sock *sock = ctx->ring_sock->sk; | |
4336 | struct sk_buff_head *head = &sock->sk_receive_queue; | |
4337 | struct sk_buff *skb; | |
4338 | ||
4339 | /* | |
4340 | * See if we can merge this file into an existing skb SCM_RIGHTS | |
4341 | * file set. If there's no room, fall back to allocating a new skb | |
4342 | * and filling it in. | |
4343 | */ | |
4344 | spin_lock_irq(&head->lock); | |
4345 | skb = skb_peek(head); | |
4346 | if (skb) { | |
4347 | struct scm_fp_list *fpl = UNIXCB(skb).fp; | |
4348 | ||
4349 | if (fpl->count < SCM_MAX_FD) { | |
4350 | __skb_unlink(skb, head); | |
4351 | spin_unlock_irq(&head->lock); | |
4352 | fpl->fp[fpl->count] = get_file(file); | |
4353 | unix_inflight(fpl->user, fpl->fp[fpl->count]); | |
4354 | fpl->count++; | |
4355 | spin_lock_irq(&head->lock); | |
4356 | __skb_queue_head(head, skb); | |
4357 | } else { | |
4358 | skb = NULL; | |
4359 | } | |
4360 | } | |
4361 | spin_unlock_irq(&head->lock); | |
4362 | ||
4363 | if (skb) { | |
4364 | fput(file); | |
4365 | return 0; | |
4366 | } | |
4367 | ||
4368 | return __io_sqe_files_scm(ctx, 1, index); | |
4369 | #else | |
4370 | return 0; | |
4371 | #endif | |
4372 | } | |
4373 | ||
4374 | static int io_sqe_files_update(struct io_ring_ctx *ctx, void __user *arg, | |
4375 | unsigned nr_args) | |
4376 | { | |
4377 | struct io_uring_files_update up; | |
4378 | __s32 __user *fds; | |
4379 | int fd, i, err; | |
4380 | __u32 done; | |
4381 | ||
65e19f54 | 4382 | if (!ctx->file_table) |
c3a31e60 JA |
4383 | return -ENXIO; |
4384 | if (!nr_args) | |
4385 | return -EINVAL; | |
4386 | if (copy_from_user(&up, arg, sizeof(up))) | |
4387 | return -EFAULT; | |
4388 | if (check_add_overflow(up.offset, nr_args, &done)) | |
4389 | return -EOVERFLOW; | |
4390 | if (done > ctx->nr_user_files) | |
4391 | return -EINVAL; | |
4392 | ||
4393 | done = 0; | |
4394 | fds = (__s32 __user *) up.fds; | |
4395 | while (nr_args) { | |
65e19f54 JA |
4396 | struct fixed_file_table *table; |
4397 | unsigned index; | |
4398 | ||
c3a31e60 JA |
4399 | err = 0; |
4400 | if (copy_from_user(&fd, &fds[done], sizeof(fd))) { | |
4401 | err = -EFAULT; | |
4402 | break; | |
4403 | } | |
4404 | i = array_index_nospec(up.offset, ctx->nr_user_files); | |
65e19f54 JA |
4405 | table = &ctx->file_table[i >> IORING_FILE_TABLE_SHIFT]; |
4406 | index = i & IORING_FILE_TABLE_MASK; | |
4407 | if (table->files[index]) { | |
c3a31e60 | 4408 | io_sqe_file_unregister(ctx, i); |
65e19f54 | 4409 | table->files[index] = NULL; |
c3a31e60 JA |
4410 | } |
4411 | if (fd != -1) { | |
4412 | struct file *file; | |
4413 | ||
4414 | file = fget(fd); | |
4415 | if (!file) { | |
4416 | err = -EBADF; | |
4417 | break; | |
4418 | } | |
4419 | /* | |
4420 | * Don't allow io_uring instances to be registered. If | |
4421 | * UNIX isn't enabled, then this causes a reference | |
4422 | * cycle and this instance can never get freed. If UNIX | |
4423 | * is enabled we'll handle it just fine, but there's | |
4424 | * still no point in allowing a ring fd as it doesn't | |
4425 | * support regular read/write anyway. | |
4426 | */ | |
4427 | if (file->f_op == &io_uring_fops) { | |
4428 | fput(file); | |
4429 | err = -EBADF; | |
4430 | break; | |
4431 | } | |
65e19f54 | 4432 | table->files[index] = file; |
c3a31e60 JA |
4433 | err = io_sqe_file_register(ctx, file, i); |
4434 | if (err) | |
4435 | break; | |
4436 | } | |
4437 | nr_args--; | |
4438 | done++; | |
4439 | up.offset++; | |
4440 | } | |
4441 | ||
4442 | return done ? done : err; | |
4443 | } | |
4444 | ||
7d723065 JA |
4445 | static void io_put_work(struct io_wq_work *work) |
4446 | { | |
4447 | struct io_kiocb *req = container_of(work, struct io_kiocb, work); | |
4448 | ||
4449 | io_put_req(req); | |
4450 | } | |
4451 | ||
4452 | static void io_get_work(struct io_wq_work *work) | |
4453 | { | |
4454 | struct io_kiocb *req = container_of(work, struct io_kiocb, work); | |
4455 | ||
4456 | refcount_inc(&req->refs); | |
4457 | } | |
4458 | ||
6c271ce2 JA |
4459 | static int io_sq_offload_start(struct io_ring_ctx *ctx, |
4460 | struct io_uring_params *p) | |
2b188cc1 | 4461 | { |
576a347b | 4462 | struct io_wq_data data; |
561fb04a | 4463 | unsigned concurrency; |
2b188cc1 JA |
4464 | int ret; |
4465 | ||
6c271ce2 | 4466 | init_waitqueue_head(&ctx->sqo_wait); |
2b188cc1 JA |
4467 | mmgrab(current->mm); |
4468 | ctx->sqo_mm = current->mm; | |
4469 | ||
6c271ce2 | 4470 | if (ctx->flags & IORING_SETUP_SQPOLL) { |
3ec482d1 JA |
4471 | ret = -EPERM; |
4472 | if (!capable(CAP_SYS_ADMIN)) | |
4473 | goto err; | |
4474 | ||
917257da JA |
4475 | ctx->sq_thread_idle = msecs_to_jiffies(p->sq_thread_idle); |
4476 | if (!ctx->sq_thread_idle) | |
4477 | ctx->sq_thread_idle = HZ; | |
4478 | ||
6c271ce2 | 4479 | if (p->flags & IORING_SETUP_SQ_AFF) { |
44a9bd18 | 4480 | int cpu = p->sq_thread_cpu; |
6c271ce2 | 4481 | |
917257da | 4482 | ret = -EINVAL; |
44a9bd18 JA |
4483 | if (cpu >= nr_cpu_ids) |
4484 | goto err; | |
7889f44d | 4485 | if (!cpu_online(cpu)) |
917257da JA |
4486 | goto err; |
4487 | ||
6c271ce2 JA |
4488 | ctx->sqo_thread = kthread_create_on_cpu(io_sq_thread, |
4489 | ctx, cpu, | |
4490 | "io_uring-sq"); | |
4491 | } else { | |
4492 | ctx->sqo_thread = kthread_create(io_sq_thread, ctx, | |
4493 | "io_uring-sq"); | |
4494 | } | |
4495 | if (IS_ERR(ctx->sqo_thread)) { | |
4496 | ret = PTR_ERR(ctx->sqo_thread); | |
4497 | ctx->sqo_thread = NULL; | |
4498 | goto err; | |
4499 | } | |
4500 | wake_up_process(ctx->sqo_thread); | |
4501 | } else if (p->flags & IORING_SETUP_SQ_AFF) { | |
4502 | /* Can't have SQ_AFF without SQPOLL */ | |
4503 | ret = -EINVAL; | |
4504 | goto err; | |
4505 | } | |
4506 | ||
576a347b JA |
4507 | data.mm = ctx->sqo_mm; |
4508 | data.user = ctx->user; | |
181e448d | 4509 | data.creds = ctx->creds; |
576a347b JA |
4510 | data.get_work = io_get_work; |
4511 | data.put_work = io_put_work; | |
4512 | ||
561fb04a JA |
4513 | /* Do QD, or 4 * CPUS, whatever is smallest */ |
4514 | concurrency = min(ctx->sq_entries, 4 * num_online_cpus()); | |
576a347b | 4515 | ctx->io_wq = io_wq_create(concurrency, &data); |
975c99a5 JA |
4516 | if (IS_ERR(ctx->io_wq)) { |
4517 | ret = PTR_ERR(ctx->io_wq); | |
4518 | ctx->io_wq = NULL; | |
2b188cc1 JA |
4519 | goto err; |
4520 | } | |
4521 | ||
4522 | return 0; | |
4523 | err: | |
54a91f3b | 4524 | io_finish_async(ctx); |
2b188cc1 JA |
4525 | mmdrop(ctx->sqo_mm); |
4526 | ctx->sqo_mm = NULL; | |
4527 | return ret; | |
4528 | } | |
4529 | ||
4530 | static void io_unaccount_mem(struct user_struct *user, unsigned long nr_pages) | |
4531 | { | |
4532 | atomic_long_sub(nr_pages, &user->locked_vm); | |
4533 | } | |
4534 | ||
4535 | static int io_account_mem(struct user_struct *user, unsigned long nr_pages) | |
4536 | { | |
4537 | unsigned long page_limit, cur_pages, new_pages; | |
4538 | ||
4539 | /* Don't allow more pages than we can safely lock */ | |
4540 | page_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT; | |
4541 | ||
4542 | do { | |
4543 | cur_pages = atomic_long_read(&user->locked_vm); | |
4544 | new_pages = cur_pages + nr_pages; | |
4545 | if (new_pages > page_limit) | |
4546 | return -ENOMEM; | |
4547 | } while (atomic_long_cmpxchg(&user->locked_vm, cur_pages, | |
4548 | new_pages) != cur_pages); | |
4549 | ||
4550 | return 0; | |
4551 | } | |
4552 | ||
4553 | static void io_mem_free(void *ptr) | |
4554 | { | |
52e04ef4 MR |
4555 | struct page *page; |
4556 | ||
4557 | if (!ptr) | |
4558 | return; | |
2b188cc1 | 4559 | |
52e04ef4 | 4560 | page = virt_to_head_page(ptr); |
2b188cc1 JA |
4561 | if (put_page_testzero(page)) |
4562 | free_compound_page(page); | |
4563 | } | |
4564 | ||
4565 | static void *io_mem_alloc(size_t size) | |
4566 | { | |
4567 | gfp_t gfp_flags = GFP_KERNEL | __GFP_ZERO | __GFP_NOWARN | __GFP_COMP | | |
4568 | __GFP_NORETRY; | |
4569 | ||
4570 | return (void *) __get_free_pages(gfp_flags, get_order(size)); | |
4571 | } | |
4572 | ||
75b28aff HV |
4573 | static unsigned long rings_size(unsigned sq_entries, unsigned cq_entries, |
4574 | size_t *sq_offset) | |
4575 | { | |
4576 | struct io_rings *rings; | |
4577 | size_t off, sq_array_size; | |
4578 | ||
4579 | off = struct_size(rings, cqes, cq_entries); | |
4580 | if (off == SIZE_MAX) | |
4581 | return SIZE_MAX; | |
4582 | ||
4583 | #ifdef CONFIG_SMP | |
4584 | off = ALIGN(off, SMP_CACHE_BYTES); | |
4585 | if (off == 0) | |
4586 | return SIZE_MAX; | |
4587 | #endif | |
4588 | ||
4589 | sq_array_size = array_size(sizeof(u32), sq_entries); | |
4590 | if (sq_array_size == SIZE_MAX) | |
4591 | return SIZE_MAX; | |
4592 | ||
4593 | if (check_add_overflow(off, sq_array_size, &off)) | |
4594 | return SIZE_MAX; | |
4595 | ||
4596 | if (sq_offset) | |
4597 | *sq_offset = off; | |
4598 | ||
4599 | return off; | |
4600 | } | |
4601 | ||
2b188cc1 JA |
4602 | static unsigned long ring_pages(unsigned sq_entries, unsigned cq_entries) |
4603 | { | |
75b28aff | 4604 | size_t pages; |
2b188cc1 | 4605 | |
75b28aff HV |
4606 | pages = (size_t)1 << get_order( |
4607 | rings_size(sq_entries, cq_entries, NULL)); | |
4608 | pages += (size_t)1 << get_order( | |
4609 | array_size(sizeof(struct io_uring_sqe), sq_entries)); | |
2b188cc1 | 4610 | |
75b28aff | 4611 | return pages; |
2b188cc1 JA |
4612 | } |
4613 | ||
edafccee JA |
4614 | static int io_sqe_buffer_unregister(struct io_ring_ctx *ctx) |
4615 | { | |
4616 | int i, j; | |
4617 | ||
4618 | if (!ctx->user_bufs) | |
4619 | return -ENXIO; | |
4620 | ||
4621 | for (i = 0; i < ctx->nr_user_bufs; i++) { | |
4622 | struct io_mapped_ubuf *imu = &ctx->user_bufs[i]; | |
4623 | ||
4624 | for (j = 0; j < imu->nr_bvecs; j++) | |
27c4d3a3 | 4625 | put_user_page(imu->bvec[j].bv_page); |
edafccee JA |
4626 | |
4627 | if (ctx->account_mem) | |
4628 | io_unaccount_mem(ctx->user, imu->nr_bvecs); | |
d4ef6475 | 4629 | kvfree(imu->bvec); |
edafccee JA |
4630 | imu->nr_bvecs = 0; |
4631 | } | |
4632 | ||
4633 | kfree(ctx->user_bufs); | |
4634 | ctx->user_bufs = NULL; | |
4635 | ctx->nr_user_bufs = 0; | |
4636 | return 0; | |
4637 | } | |
4638 | ||
4639 | static int io_copy_iov(struct io_ring_ctx *ctx, struct iovec *dst, | |
4640 | void __user *arg, unsigned index) | |
4641 | { | |
4642 | struct iovec __user *src; | |
4643 | ||
4644 | #ifdef CONFIG_COMPAT | |
4645 | if (ctx->compat) { | |
4646 | struct compat_iovec __user *ciovs; | |
4647 | struct compat_iovec ciov; | |
4648 | ||
4649 | ciovs = (struct compat_iovec __user *) arg; | |
4650 | if (copy_from_user(&ciov, &ciovs[index], sizeof(ciov))) | |
4651 | return -EFAULT; | |
4652 | ||
4653 | dst->iov_base = (void __user *) (unsigned long) ciov.iov_base; | |
4654 | dst->iov_len = ciov.iov_len; | |
4655 | return 0; | |
4656 | } | |
4657 | #endif | |
4658 | src = (struct iovec __user *) arg; | |
4659 | if (copy_from_user(dst, &src[index], sizeof(*dst))) | |
4660 | return -EFAULT; | |
4661 | return 0; | |
4662 | } | |
4663 | ||
4664 | static int io_sqe_buffer_register(struct io_ring_ctx *ctx, void __user *arg, | |
4665 | unsigned nr_args) | |
4666 | { | |
4667 | struct vm_area_struct **vmas = NULL; | |
4668 | struct page **pages = NULL; | |
4669 | int i, j, got_pages = 0; | |
4670 | int ret = -EINVAL; | |
4671 | ||
4672 | if (ctx->user_bufs) | |
4673 | return -EBUSY; | |
4674 | if (!nr_args || nr_args > UIO_MAXIOV) | |
4675 | return -EINVAL; | |
4676 | ||
4677 | ctx->user_bufs = kcalloc(nr_args, sizeof(struct io_mapped_ubuf), | |
4678 | GFP_KERNEL); | |
4679 | if (!ctx->user_bufs) | |
4680 | return -ENOMEM; | |
4681 | ||
4682 | for (i = 0; i < nr_args; i++) { | |
4683 | struct io_mapped_ubuf *imu = &ctx->user_bufs[i]; | |
4684 | unsigned long off, start, end, ubuf; | |
4685 | int pret, nr_pages; | |
4686 | struct iovec iov; | |
4687 | size_t size; | |
4688 | ||
4689 | ret = io_copy_iov(ctx, &iov, arg, i); | |
4690 | if (ret) | |
a278682d | 4691 | goto err; |
edafccee JA |
4692 | |
4693 | /* | |
4694 | * Don't impose further limits on the size and buffer | |
4695 | * constraints here, we'll -EINVAL later when IO is | |
4696 | * submitted if they are wrong. | |
4697 | */ | |
4698 | ret = -EFAULT; | |
4699 | if (!iov.iov_base || !iov.iov_len) | |
4700 | goto err; | |
4701 | ||
4702 | /* arbitrary limit, but we need something */ | |
4703 | if (iov.iov_len > SZ_1G) | |
4704 | goto err; | |
4705 | ||
4706 | ubuf = (unsigned long) iov.iov_base; | |
4707 | end = (ubuf + iov.iov_len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
4708 | start = ubuf >> PAGE_SHIFT; | |
4709 | nr_pages = end - start; | |
4710 | ||
4711 | if (ctx->account_mem) { | |
4712 | ret = io_account_mem(ctx->user, nr_pages); | |
4713 | if (ret) | |
4714 | goto err; | |
4715 | } | |
4716 | ||
4717 | ret = 0; | |
4718 | if (!pages || nr_pages > got_pages) { | |
4719 | kfree(vmas); | |
4720 | kfree(pages); | |
d4ef6475 | 4721 | pages = kvmalloc_array(nr_pages, sizeof(struct page *), |
edafccee | 4722 | GFP_KERNEL); |
d4ef6475 | 4723 | vmas = kvmalloc_array(nr_pages, |
edafccee JA |
4724 | sizeof(struct vm_area_struct *), |
4725 | GFP_KERNEL); | |
4726 | if (!pages || !vmas) { | |
4727 | ret = -ENOMEM; | |
4728 | if (ctx->account_mem) | |
4729 | io_unaccount_mem(ctx->user, nr_pages); | |
4730 | goto err; | |
4731 | } | |
4732 | got_pages = nr_pages; | |
4733 | } | |
4734 | ||
d4ef6475 | 4735 | imu->bvec = kvmalloc_array(nr_pages, sizeof(struct bio_vec), |
edafccee JA |
4736 | GFP_KERNEL); |
4737 | ret = -ENOMEM; | |
4738 | if (!imu->bvec) { | |
4739 | if (ctx->account_mem) | |
4740 | io_unaccount_mem(ctx->user, nr_pages); | |
4741 | goto err; | |
4742 | } | |
4743 | ||
4744 | ret = 0; | |
4745 | down_read(¤t->mm->mmap_sem); | |
932f4a63 IW |
4746 | pret = get_user_pages(ubuf, nr_pages, |
4747 | FOLL_WRITE | FOLL_LONGTERM, | |
4748 | pages, vmas); | |
edafccee JA |
4749 | if (pret == nr_pages) { |
4750 | /* don't support file backed memory */ | |
4751 | for (j = 0; j < nr_pages; j++) { | |
4752 | struct vm_area_struct *vma = vmas[j]; | |
4753 | ||
4754 | if (vma->vm_file && | |
4755 | !is_file_hugepages(vma->vm_file)) { | |
4756 | ret = -EOPNOTSUPP; | |
4757 | break; | |
4758 | } | |
4759 | } | |
4760 | } else { | |
4761 | ret = pret < 0 ? pret : -EFAULT; | |
4762 | } | |
4763 | up_read(¤t->mm->mmap_sem); | |
4764 | if (ret) { | |
4765 | /* | |
4766 | * if we did partial map, or found file backed vmas, | |
4767 | * release any pages we did get | |
4768 | */ | |
27c4d3a3 JH |
4769 | if (pret > 0) |
4770 | put_user_pages(pages, pret); | |
edafccee JA |
4771 | if (ctx->account_mem) |
4772 | io_unaccount_mem(ctx->user, nr_pages); | |
d4ef6475 | 4773 | kvfree(imu->bvec); |
edafccee JA |
4774 | goto err; |
4775 | } | |
4776 | ||
4777 | off = ubuf & ~PAGE_MASK; | |
4778 | size = iov.iov_len; | |
4779 | for (j = 0; j < nr_pages; j++) { | |
4780 | size_t vec_len; | |
4781 | ||
4782 | vec_len = min_t(size_t, size, PAGE_SIZE - off); | |
4783 | imu->bvec[j].bv_page = pages[j]; | |
4784 | imu->bvec[j].bv_len = vec_len; | |
4785 | imu->bvec[j].bv_offset = off; | |
4786 | off = 0; | |
4787 | size -= vec_len; | |
4788 | } | |
4789 | /* store original address for later verification */ | |
4790 | imu->ubuf = ubuf; | |
4791 | imu->len = iov.iov_len; | |
4792 | imu->nr_bvecs = nr_pages; | |
4793 | ||
4794 | ctx->nr_user_bufs++; | |
4795 | } | |
d4ef6475 MR |
4796 | kvfree(pages); |
4797 | kvfree(vmas); | |
edafccee JA |
4798 | return 0; |
4799 | err: | |
d4ef6475 MR |
4800 | kvfree(pages); |
4801 | kvfree(vmas); | |
edafccee JA |
4802 | io_sqe_buffer_unregister(ctx); |
4803 | return ret; | |
4804 | } | |
4805 | ||
9b402849 JA |
4806 | static int io_eventfd_register(struct io_ring_ctx *ctx, void __user *arg) |
4807 | { | |
4808 | __s32 __user *fds = arg; | |
4809 | int fd; | |
4810 | ||
4811 | if (ctx->cq_ev_fd) | |
4812 | return -EBUSY; | |
4813 | ||
4814 | if (copy_from_user(&fd, fds, sizeof(*fds))) | |
4815 | return -EFAULT; | |
4816 | ||
4817 | ctx->cq_ev_fd = eventfd_ctx_fdget(fd); | |
4818 | if (IS_ERR(ctx->cq_ev_fd)) { | |
4819 | int ret = PTR_ERR(ctx->cq_ev_fd); | |
4820 | ctx->cq_ev_fd = NULL; | |
4821 | return ret; | |
4822 | } | |
4823 | ||
4824 | return 0; | |
4825 | } | |
4826 | ||
4827 | static int io_eventfd_unregister(struct io_ring_ctx *ctx) | |
4828 | { | |
4829 | if (ctx->cq_ev_fd) { | |
4830 | eventfd_ctx_put(ctx->cq_ev_fd); | |
4831 | ctx->cq_ev_fd = NULL; | |
4832 | return 0; | |
4833 | } | |
4834 | ||
4835 | return -ENXIO; | |
4836 | } | |
4837 | ||
2b188cc1 JA |
4838 | static void io_ring_ctx_free(struct io_ring_ctx *ctx) |
4839 | { | |
6b06314c | 4840 | io_finish_async(ctx); |
2b188cc1 JA |
4841 | if (ctx->sqo_mm) |
4842 | mmdrop(ctx->sqo_mm); | |
def596e9 JA |
4843 | |
4844 | io_iopoll_reap_events(ctx); | |
edafccee | 4845 | io_sqe_buffer_unregister(ctx); |
6b06314c | 4846 | io_sqe_files_unregister(ctx); |
9b402849 | 4847 | io_eventfd_unregister(ctx); |
def596e9 | 4848 | |
2b188cc1 | 4849 | #if defined(CONFIG_UNIX) |
355e8d26 EB |
4850 | if (ctx->ring_sock) { |
4851 | ctx->ring_sock->file = NULL; /* so that iput() is called */ | |
2b188cc1 | 4852 | sock_release(ctx->ring_sock); |
355e8d26 | 4853 | } |
2b188cc1 JA |
4854 | #endif |
4855 | ||
75b28aff | 4856 | io_mem_free(ctx->rings); |
2b188cc1 | 4857 | io_mem_free(ctx->sq_sqes); |
2b188cc1 JA |
4858 | |
4859 | percpu_ref_exit(&ctx->refs); | |
4860 | if (ctx->account_mem) | |
4861 | io_unaccount_mem(ctx->user, | |
4862 | ring_pages(ctx->sq_entries, ctx->cq_entries)); | |
4863 | free_uid(ctx->user); | |
181e448d | 4864 | put_cred(ctx->creds); |
206aefde | 4865 | kfree(ctx->completions); |
78076bb6 | 4866 | kfree(ctx->cancel_hash); |
0ddf92e8 | 4867 | kmem_cache_free(req_cachep, ctx->fallback_req); |
2b188cc1 JA |
4868 | kfree(ctx); |
4869 | } | |
4870 | ||
4871 | static __poll_t io_uring_poll(struct file *file, poll_table *wait) | |
4872 | { | |
4873 | struct io_ring_ctx *ctx = file->private_data; | |
4874 | __poll_t mask = 0; | |
4875 | ||
4876 | poll_wait(file, &ctx->cq_wait, wait); | |
4f7067c3 SB |
4877 | /* |
4878 | * synchronizes with barrier from wq_has_sleeper call in | |
4879 | * io_commit_cqring | |
4880 | */ | |
2b188cc1 | 4881 | smp_rmb(); |
75b28aff HV |
4882 | if (READ_ONCE(ctx->rings->sq.tail) - ctx->cached_sq_head != |
4883 | ctx->rings->sq_ring_entries) | |
2b188cc1 | 4884 | mask |= EPOLLOUT | EPOLLWRNORM; |
daa5de54 | 4885 | if (READ_ONCE(ctx->rings->cq.head) != ctx->cached_cq_tail) |
2b188cc1 JA |
4886 | mask |= EPOLLIN | EPOLLRDNORM; |
4887 | ||
4888 | return mask; | |
4889 | } | |
4890 | ||
4891 | static int io_uring_fasync(int fd, struct file *file, int on) | |
4892 | { | |
4893 | struct io_ring_ctx *ctx = file->private_data; | |
4894 | ||
4895 | return fasync_helper(fd, file, on, &ctx->cq_fasync); | |
4896 | } | |
4897 | ||
4898 | static void io_ring_ctx_wait_and_kill(struct io_ring_ctx *ctx) | |
4899 | { | |
4900 | mutex_lock(&ctx->uring_lock); | |
4901 | percpu_ref_kill(&ctx->refs); | |
4902 | mutex_unlock(&ctx->uring_lock); | |
4903 | ||
5262f567 | 4904 | io_kill_timeouts(ctx); |
221c5eb2 | 4905 | io_poll_remove_all(ctx); |
561fb04a JA |
4906 | |
4907 | if (ctx->io_wq) | |
4908 | io_wq_cancel_all(ctx->io_wq); | |
4909 | ||
def596e9 | 4910 | io_iopoll_reap_events(ctx); |
15dff286 JA |
4911 | /* if we failed setting up the ctx, we might not have any rings */ |
4912 | if (ctx->rings) | |
4913 | io_cqring_overflow_flush(ctx, true); | |
206aefde | 4914 | wait_for_completion(&ctx->completions[0]); |
2b188cc1 JA |
4915 | io_ring_ctx_free(ctx); |
4916 | } | |
4917 | ||
4918 | static int io_uring_release(struct inode *inode, struct file *file) | |
4919 | { | |
4920 | struct io_ring_ctx *ctx = file->private_data; | |
4921 | ||
4922 | file->private_data = NULL; | |
4923 | io_ring_ctx_wait_and_kill(ctx); | |
4924 | return 0; | |
4925 | } | |
4926 | ||
fcb323cc JA |
4927 | static void io_uring_cancel_files(struct io_ring_ctx *ctx, |
4928 | struct files_struct *files) | |
4929 | { | |
4930 | struct io_kiocb *req; | |
4931 | DEFINE_WAIT(wait); | |
4932 | ||
4933 | while (!list_empty_careful(&ctx->inflight_list)) { | |
768134d4 | 4934 | struct io_kiocb *cancel_req = NULL; |
fcb323cc JA |
4935 | |
4936 | spin_lock_irq(&ctx->inflight_lock); | |
4937 | list_for_each_entry(req, &ctx->inflight_list, inflight_entry) { | |
768134d4 JA |
4938 | if (req->work.files != files) |
4939 | continue; | |
4940 | /* req is being completed, ignore */ | |
4941 | if (!refcount_inc_not_zero(&req->refs)) | |
4942 | continue; | |
4943 | cancel_req = req; | |
4944 | break; | |
fcb323cc | 4945 | } |
768134d4 | 4946 | if (cancel_req) |
fcb323cc | 4947 | prepare_to_wait(&ctx->inflight_wait, &wait, |
768134d4 | 4948 | TASK_UNINTERRUPTIBLE); |
fcb323cc JA |
4949 | spin_unlock_irq(&ctx->inflight_lock); |
4950 | ||
768134d4 JA |
4951 | /* We need to keep going until we don't find a matching req */ |
4952 | if (!cancel_req) | |
fcb323cc | 4953 | break; |
2f6d9b9d BL |
4954 | |
4955 | io_wq_cancel_work(ctx->io_wq, &cancel_req->work); | |
4956 | io_put_req(cancel_req); | |
fcb323cc JA |
4957 | schedule(); |
4958 | } | |
768134d4 | 4959 | finish_wait(&ctx->inflight_wait, &wait); |
fcb323cc JA |
4960 | } |
4961 | ||
4962 | static int io_uring_flush(struct file *file, void *data) | |
4963 | { | |
4964 | struct io_ring_ctx *ctx = file->private_data; | |
4965 | ||
4966 | io_uring_cancel_files(ctx, data); | |
1d7bb1d5 JA |
4967 | if (fatal_signal_pending(current) || (current->flags & PF_EXITING)) { |
4968 | io_cqring_overflow_flush(ctx, true); | |
fcb323cc | 4969 | io_wq_cancel_all(ctx->io_wq); |
1d7bb1d5 | 4970 | } |
fcb323cc JA |
4971 | return 0; |
4972 | } | |
4973 | ||
6c5c240e RP |
4974 | static void *io_uring_validate_mmap_request(struct file *file, |
4975 | loff_t pgoff, size_t sz) | |
2b188cc1 | 4976 | { |
2b188cc1 | 4977 | struct io_ring_ctx *ctx = file->private_data; |
6c5c240e | 4978 | loff_t offset = pgoff << PAGE_SHIFT; |
2b188cc1 JA |
4979 | struct page *page; |
4980 | void *ptr; | |
4981 | ||
4982 | switch (offset) { | |
4983 | case IORING_OFF_SQ_RING: | |
75b28aff HV |
4984 | case IORING_OFF_CQ_RING: |
4985 | ptr = ctx->rings; | |
2b188cc1 JA |
4986 | break; |
4987 | case IORING_OFF_SQES: | |
4988 | ptr = ctx->sq_sqes; | |
4989 | break; | |
2b188cc1 | 4990 | default: |
6c5c240e | 4991 | return ERR_PTR(-EINVAL); |
2b188cc1 JA |
4992 | } |
4993 | ||
4994 | page = virt_to_head_page(ptr); | |
a50b854e | 4995 | if (sz > page_size(page)) |
6c5c240e RP |
4996 | return ERR_PTR(-EINVAL); |
4997 | ||
4998 | return ptr; | |
4999 | } | |
5000 | ||
5001 | #ifdef CONFIG_MMU | |
5002 | ||
5003 | static int io_uring_mmap(struct file *file, struct vm_area_struct *vma) | |
5004 | { | |
5005 | size_t sz = vma->vm_end - vma->vm_start; | |
5006 | unsigned long pfn; | |
5007 | void *ptr; | |
5008 | ||
5009 | ptr = io_uring_validate_mmap_request(file, vma->vm_pgoff, sz); | |
5010 | if (IS_ERR(ptr)) | |
5011 | return PTR_ERR(ptr); | |
2b188cc1 JA |
5012 | |
5013 | pfn = virt_to_phys(ptr) >> PAGE_SHIFT; | |
5014 | return remap_pfn_range(vma, vma->vm_start, pfn, sz, vma->vm_page_prot); | |
5015 | } | |
5016 | ||
6c5c240e RP |
5017 | #else /* !CONFIG_MMU */ |
5018 | ||
5019 | static int io_uring_mmap(struct file *file, struct vm_area_struct *vma) | |
5020 | { | |
5021 | return vma->vm_flags & (VM_SHARED | VM_MAYSHARE) ? 0 : -EINVAL; | |
5022 | } | |
5023 | ||
5024 | static unsigned int io_uring_nommu_mmap_capabilities(struct file *file) | |
5025 | { | |
5026 | return NOMMU_MAP_DIRECT | NOMMU_MAP_READ | NOMMU_MAP_WRITE; | |
5027 | } | |
5028 | ||
5029 | static unsigned long io_uring_nommu_get_unmapped_area(struct file *file, | |
5030 | unsigned long addr, unsigned long len, | |
5031 | unsigned long pgoff, unsigned long flags) | |
5032 | { | |
5033 | void *ptr; | |
5034 | ||
5035 | ptr = io_uring_validate_mmap_request(file, pgoff, len); | |
5036 | if (IS_ERR(ptr)) | |
5037 | return PTR_ERR(ptr); | |
5038 | ||
5039 | return (unsigned long) ptr; | |
5040 | } | |
5041 | ||
5042 | #endif /* !CONFIG_MMU */ | |
5043 | ||
2b188cc1 JA |
5044 | SYSCALL_DEFINE6(io_uring_enter, unsigned int, fd, u32, to_submit, |
5045 | u32, min_complete, u32, flags, const sigset_t __user *, sig, | |
5046 | size_t, sigsz) | |
5047 | { | |
5048 | struct io_ring_ctx *ctx; | |
5049 | long ret = -EBADF; | |
5050 | int submitted = 0; | |
5051 | struct fd f; | |
5052 | ||
6c271ce2 | 5053 | if (flags & ~(IORING_ENTER_GETEVENTS | IORING_ENTER_SQ_WAKEUP)) |
2b188cc1 JA |
5054 | return -EINVAL; |
5055 | ||
5056 | f = fdget(fd); | |
5057 | if (!f.file) | |
5058 | return -EBADF; | |
5059 | ||
5060 | ret = -EOPNOTSUPP; | |
5061 | if (f.file->f_op != &io_uring_fops) | |
5062 | goto out_fput; | |
5063 | ||
5064 | ret = -ENXIO; | |
5065 | ctx = f.file->private_data; | |
5066 | if (!percpu_ref_tryget(&ctx->refs)) | |
5067 | goto out_fput; | |
5068 | ||
6c271ce2 JA |
5069 | /* |
5070 | * For SQ polling, the thread will do all submissions and completions. | |
5071 | * Just return the requested submit count, and wake the thread if | |
5072 | * we were asked to. | |
5073 | */ | |
b2a9eada | 5074 | ret = 0; |
6c271ce2 | 5075 | if (ctx->flags & IORING_SETUP_SQPOLL) { |
c1edbf5f JA |
5076 | if (!list_empty_careful(&ctx->cq_overflow_list)) |
5077 | io_cqring_overflow_flush(ctx, false); | |
6c271ce2 JA |
5078 | if (flags & IORING_ENTER_SQ_WAKEUP) |
5079 | wake_up(&ctx->sqo_wait); | |
5080 | submitted = to_submit; | |
b2a9eada | 5081 | } else if (to_submit) { |
ae9428ca | 5082 | struct mm_struct *cur_mm; |
2b188cc1 | 5083 | |
ae9428ca | 5084 | to_submit = min(to_submit, ctx->sq_entries); |
2b188cc1 | 5085 | mutex_lock(&ctx->uring_lock); |
ae9428ca PB |
5086 | /* already have mm, so io_submit_sqes() won't try to grab it */ |
5087 | cur_mm = ctx->sqo_mm; | |
5088 | submitted = io_submit_sqes(ctx, to_submit, f.file, fd, | |
5089 | &cur_mm, false); | |
2b188cc1 | 5090 | mutex_unlock(&ctx->uring_lock); |
2b188cc1 JA |
5091 | } |
5092 | if (flags & IORING_ENTER_GETEVENTS) { | |
def596e9 JA |
5093 | unsigned nr_events = 0; |
5094 | ||
2b188cc1 JA |
5095 | min_complete = min(min_complete, ctx->cq_entries); |
5096 | ||
def596e9 | 5097 | if (ctx->flags & IORING_SETUP_IOPOLL) { |
def596e9 | 5098 | ret = io_iopoll_check(ctx, &nr_events, min_complete); |
def596e9 JA |
5099 | } else { |
5100 | ret = io_cqring_wait(ctx, min_complete, sig, sigsz); | |
5101 | } | |
2b188cc1 JA |
5102 | } |
5103 | ||
6805b32e | 5104 | percpu_ref_put(&ctx->refs); |
2b188cc1 JA |
5105 | out_fput: |
5106 | fdput(f); | |
5107 | return submitted ? submitted : ret; | |
5108 | } | |
5109 | ||
5110 | static const struct file_operations io_uring_fops = { | |
5111 | .release = io_uring_release, | |
fcb323cc | 5112 | .flush = io_uring_flush, |
2b188cc1 | 5113 | .mmap = io_uring_mmap, |
6c5c240e RP |
5114 | #ifndef CONFIG_MMU |
5115 | .get_unmapped_area = io_uring_nommu_get_unmapped_area, | |
5116 | .mmap_capabilities = io_uring_nommu_mmap_capabilities, | |
5117 | #endif | |
2b188cc1 JA |
5118 | .poll = io_uring_poll, |
5119 | .fasync = io_uring_fasync, | |
5120 | }; | |
5121 | ||
5122 | static int io_allocate_scq_urings(struct io_ring_ctx *ctx, | |
5123 | struct io_uring_params *p) | |
5124 | { | |
75b28aff HV |
5125 | struct io_rings *rings; |
5126 | size_t size, sq_array_offset; | |
2b188cc1 | 5127 | |
75b28aff HV |
5128 | size = rings_size(p->sq_entries, p->cq_entries, &sq_array_offset); |
5129 | if (size == SIZE_MAX) | |
5130 | return -EOVERFLOW; | |
5131 | ||
5132 | rings = io_mem_alloc(size); | |
5133 | if (!rings) | |
2b188cc1 JA |
5134 | return -ENOMEM; |
5135 | ||
75b28aff HV |
5136 | ctx->rings = rings; |
5137 | ctx->sq_array = (u32 *)((char *)rings + sq_array_offset); | |
5138 | rings->sq_ring_mask = p->sq_entries - 1; | |
5139 | rings->cq_ring_mask = p->cq_entries - 1; | |
5140 | rings->sq_ring_entries = p->sq_entries; | |
5141 | rings->cq_ring_entries = p->cq_entries; | |
5142 | ctx->sq_mask = rings->sq_ring_mask; | |
5143 | ctx->cq_mask = rings->cq_ring_mask; | |
5144 | ctx->sq_entries = rings->sq_ring_entries; | |
5145 | ctx->cq_entries = rings->cq_ring_entries; | |
2b188cc1 JA |
5146 | |
5147 | size = array_size(sizeof(struct io_uring_sqe), p->sq_entries); | |
eb065d30 JA |
5148 | if (size == SIZE_MAX) { |
5149 | io_mem_free(ctx->rings); | |
5150 | ctx->rings = NULL; | |
2b188cc1 | 5151 | return -EOVERFLOW; |
eb065d30 | 5152 | } |
2b188cc1 JA |
5153 | |
5154 | ctx->sq_sqes = io_mem_alloc(size); | |
eb065d30 JA |
5155 | if (!ctx->sq_sqes) { |
5156 | io_mem_free(ctx->rings); | |
5157 | ctx->rings = NULL; | |
2b188cc1 | 5158 | return -ENOMEM; |
eb065d30 | 5159 | } |
2b188cc1 | 5160 | |
2b188cc1 JA |
5161 | return 0; |
5162 | } | |
5163 | ||
5164 | /* | |
5165 | * Allocate an anonymous fd, this is what constitutes the application | |
5166 | * visible backing of an io_uring instance. The application mmaps this | |
5167 | * fd to gain access to the SQ/CQ ring details. If UNIX sockets are enabled, | |
5168 | * we have to tie this fd to a socket for file garbage collection purposes. | |
5169 | */ | |
5170 | static int io_uring_get_fd(struct io_ring_ctx *ctx) | |
5171 | { | |
5172 | struct file *file; | |
5173 | int ret; | |
5174 | ||
5175 | #if defined(CONFIG_UNIX) | |
5176 | ret = sock_create_kern(&init_net, PF_UNIX, SOCK_RAW, IPPROTO_IP, | |
5177 | &ctx->ring_sock); | |
5178 | if (ret) | |
5179 | return ret; | |
5180 | #endif | |
5181 | ||
5182 | ret = get_unused_fd_flags(O_RDWR | O_CLOEXEC); | |
5183 | if (ret < 0) | |
5184 | goto err; | |
5185 | ||
5186 | file = anon_inode_getfile("[io_uring]", &io_uring_fops, ctx, | |
5187 | O_RDWR | O_CLOEXEC); | |
5188 | if (IS_ERR(file)) { | |
5189 | put_unused_fd(ret); | |
5190 | ret = PTR_ERR(file); | |
5191 | goto err; | |
5192 | } | |
5193 | ||
5194 | #if defined(CONFIG_UNIX) | |
5195 | ctx->ring_sock->file = file; | |
6b06314c | 5196 | ctx->ring_sock->sk->sk_user_data = ctx; |
2b188cc1 JA |
5197 | #endif |
5198 | fd_install(ret, file); | |
5199 | return ret; | |
5200 | err: | |
5201 | #if defined(CONFIG_UNIX) | |
5202 | sock_release(ctx->ring_sock); | |
5203 | ctx->ring_sock = NULL; | |
5204 | #endif | |
5205 | return ret; | |
5206 | } | |
5207 | ||
5208 | static int io_uring_create(unsigned entries, struct io_uring_params *p) | |
5209 | { | |
5210 | struct user_struct *user = NULL; | |
5211 | struct io_ring_ctx *ctx; | |
5212 | bool account_mem; | |
5213 | int ret; | |
5214 | ||
5215 | if (!entries || entries > IORING_MAX_ENTRIES) | |
5216 | return -EINVAL; | |
5217 | ||
5218 | /* | |
5219 | * Use twice as many entries for the CQ ring. It's possible for the | |
5220 | * application to drive a higher depth than the size of the SQ ring, | |
5221 | * since the sqes are only used at submission time. This allows for | |
33a107f0 JA |
5222 | * some flexibility in overcommitting a bit. If the application has |
5223 | * set IORING_SETUP_CQSIZE, it will have passed in the desired number | |
5224 | * of CQ ring entries manually. | |
2b188cc1 JA |
5225 | */ |
5226 | p->sq_entries = roundup_pow_of_two(entries); | |
33a107f0 JA |
5227 | if (p->flags & IORING_SETUP_CQSIZE) { |
5228 | /* | |
5229 | * If IORING_SETUP_CQSIZE is set, we do the same roundup | |
5230 | * to a power-of-two, if it isn't already. We do NOT impose | |
5231 | * any cq vs sq ring sizing. | |
5232 | */ | |
5233 | if (p->cq_entries < p->sq_entries || p->cq_entries > IORING_MAX_CQ_ENTRIES) | |
5234 | return -EINVAL; | |
5235 | p->cq_entries = roundup_pow_of_two(p->cq_entries); | |
5236 | } else { | |
5237 | p->cq_entries = 2 * p->sq_entries; | |
5238 | } | |
2b188cc1 JA |
5239 | |
5240 | user = get_uid(current_user()); | |
5241 | account_mem = !capable(CAP_IPC_LOCK); | |
5242 | ||
5243 | if (account_mem) { | |
5244 | ret = io_account_mem(user, | |
5245 | ring_pages(p->sq_entries, p->cq_entries)); | |
5246 | if (ret) { | |
5247 | free_uid(user); | |
5248 | return ret; | |
5249 | } | |
5250 | } | |
5251 | ||
5252 | ctx = io_ring_ctx_alloc(p); | |
5253 | if (!ctx) { | |
5254 | if (account_mem) | |
5255 | io_unaccount_mem(user, ring_pages(p->sq_entries, | |
5256 | p->cq_entries)); | |
5257 | free_uid(user); | |
5258 | return -ENOMEM; | |
5259 | } | |
5260 | ctx->compat = in_compat_syscall(); | |
5261 | ctx->account_mem = account_mem; | |
5262 | ctx->user = user; | |
0b8c0ec7 | 5263 | ctx->creds = get_current_cred(); |
2b188cc1 JA |
5264 | |
5265 | ret = io_allocate_scq_urings(ctx, p); | |
5266 | if (ret) | |
5267 | goto err; | |
5268 | ||
6c271ce2 | 5269 | ret = io_sq_offload_start(ctx, p); |
2b188cc1 JA |
5270 | if (ret) |
5271 | goto err; | |
5272 | ||
2b188cc1 | 5273 | memset(&p->sq_off, 0, sizeof(p->sq_off)); |
75b28aff HV |
5274 | p->sq_off.head = offsetof(struct io_rings, sq.head); |
5275 | p->sq_off.tail = offsetof(struct io_rings, sq.tail); | |
5276 | p->sq_off.ring_mask = offsetof(struct io_rings, sq_ring_mask); | |
5277 | p->sq_off.ring_entries = offsetof(struct io_rings, sq_ring_entries); | |
5278 | p->sq_off.flags = offsetof(struct io_rings, sq_flags); | |
5279 | p->sq_off.dropped = offsetof(struct io_rings, sq_dropped); | |
5280 | p->sq_off.array = (char *)ctx->sq_array - (char *)ctx->rings; | |
2b188cc1 JA |
5281 | |
5282 | memset(&p->cq_off, 0, sizeof(p->cq_off)); | |
75b28aff HV |
5283 | p->cq_off.head = offsetof(struct io_rings, cq.head); |
5284 | p->cq_off.tail = offsetof(struct io_rings, cq.tail); | |
5285 | p->cq_off.ring_mask = offsetof(struct io_rings, cq_ring_mask); | |
5286 | p->cq_off.ring_entries = offsetof(struct io_rings, cq_ring_entries); | |
5287 | p->cq_off.overflow = offsetof(struct io_rings, cq_overflow); | |
5288 | p->cq_off.cqes = offsetof(struct io_rings, cqes); | |
ac90f249 | 5289 | |
044c1ab3 JA |
5290 | /* |
5291 | * Install ring fd as the very last thing, so we don't risk someone | |
5292 | * having closed it before we finish setup | |
5293 | */ | |
5294 | ret = io_uring_get_fd(ctx); | |
5295 | if (ret < 0) | |
5296 | goto err; | |
5297 | ||
da8c9690 JA |
5298 | p->features = IORING_FEAT_SINGLE_MMAP | IORING_FEAT_NODROP | |
5299 | IORING_FEAT_SUBMIT_STABLE; | |
c826bd7a | 5300 | trace_io_uring_create(ret, ctx, p->sq_entries, p->cq_entries, p->flags); |
2b188cc1 JA |
5301 | return ret; |
5302 | err: | |
5303 | io_ring_ctx_wait_and_kill(ctx); | |
5304 | return ret; | |
5305 | } | |
5306 | ||
5307 | /* | |
5308 | * Sets up an aio uring context, and returns the fd. Applications asks for a | |
5309 | * ring size, we return the actual sq/cq ring sizes (among other things) in the | |
5310 | * params structure passed in. | |
5311 | */ | |
5312 | static long io_uring_setup(u32 entries, struct io_uring_params __user *params) | |
5313 | { | |
5314 | struct io_uring_params p; | |
5315 | long ret; | |
5316 | int i; | |
5317 | ||
5318 | if (copy_from_user(&p, params, sizeof(p))) | |
5319 | return -EFAULT; | |
5320 | for (i = 0; i < ARRAY_SIZE(p.resv); i++) { | |
5321 | if (p.resv[i]) | |
5322 | return -EINVAL; | |
5323 | } | |
5324 | ||
6c271ce2 | 5325 | if (p.flags & ~(IORING_SETUP_IOPOLL | IORING_SETUP_SQPOLL | |
33a107f0 | 5326 | IORING_SETUP_SQ_AFF | IORING_SETUP_CQSIZE)) |
2b188cc1 JA |
5327 | return -EINVAL; |
5328 | ||
5329 | ret = io_uring_create(entries, &p); | |
5330 | if (ret < 0) | |
5331 | return ret; | |
5332 | ||
5333 | if (copy_to_user(params, &p, sizeof(p))) | |
5334 | return -EFAULT; | |
5335 | ||
5336 | return ret; | |
5337 | } | |
5338 | ||
5339 | SYSCALL_DEFINE2(io_uring_setup, u32, entries, | |
5340 | struct io_uring_params __user *, params) | |
5341 | { | |
5342 | return io_uring_setup(entries, params); | |
5343 | } | |
5344 | ||
edafccee JA |
5345 | static int __io_uring_register(struct io_ring_ctx *ctx, unsigned opcode, |
5346 | void __user *arg, unsigned nr_args) | |
b19062a5 JA |
5347 | __releases(ctx->uring_lock) |
5348 | __acquires(ctx->uring_lock) | |
edafccee JA |
5349 | { |
5350 | int ret; | |
5351 | ||
35fa71a0 JA |
5352 | /* |
5353 | * We're inside the ring mutex, if the ref is already dying, then | |
5354 | * someone else killed the ctx or is already going through | |
5355 | * io_uring_register(). | |
5356 | */ | |
5357 | if (percpu_ref_is_dying(&ctx->refs)) | |
5358 | return -ENXIO; | |
5359 | ||
edafccee | 5360 | percpu_ref_kill(&ctx->refs); |
b19062a5 JA |
5361 | |
5362 | /* | |
5363 | * Drop uring mutex before waiting for references to exit. If another | |
5364 | * thread is currently inside io_uring_enter() it might need to grab | |
5365 | * the uring_lock to make progress. If we hold it here across the drain | |
5366 | * wait, then we can deadlock. It's safe to drop the mutex here, since | |
5367 | * no new references will come in after we've killed the percpu ref. | |
5368 | */ | |
5369 | mutex_unlock(&ctx->uring_lock); | |
206aefde | 5370 | wait_for_completion(&ctx->completions[0]); |
b19062a5 | 5371 | mutex_lock(&ctx->uring_lock); |
edafccee JA |
5372 | |
5373 | switch (opcode) { | |
5374 | case IORING_REGISTER_BUFFERS: | |
5375 | ret = io_sqe_buffer_register(ctx, arg, nr_args); | |
5376 | break; | |
5377 | case IORING_UNREGISTER_BUFFERS: | |
5378 | ret = -EINVAL; | |
5379 | if (arg || nr_args) | |
5380 | break; | |
5381 | ret = io_sqe_buffer_unregister(ctx); | |
5382 | break; | |
6b06314c JA |
5383 | case IORING_REGISTER_FILES: |
5384 | ret = io_sqe_files_register(ctx, arg, nr_args); | |
5385 | break; | |
5386 | case IORING_UNREGISTER_FILES: | |
5387 | ret = -EINVAL; | |
5388 | if (arg || nr_args) | |
5389 | break; | |
5390 | ret = io_sqe_files_unregister(ctx); | |
5391 | break; | |
c3a31e60 JA |
5392 | case IORING_REGISTER_FILES_UPDATE: |
5393 | ret = io_sqe_files_update(ctx, arg, nr_args); | |
5394 | break; | |
9b402849 JA |
5395 | case IORING_REGISTER_EVENTFD: |
5396 | ret = -EINVAL; | |
5397 | if (nr_args != 1) | |
5398 | break; | |
5399 | ret = io_eventfd_register(ctx, arg); | |
5400 | break; | |
5401 | case IORING_UNREGISTER_EVENTFD: | |
5402 | ret = -EINVAL; | |
5403 | if (arg || nr_args) | |
5404 | break; | |
5405 | ret = io_eventfd_unregister(ctx); | |
5406 | break; | |
edafccee JA |
5407 | default: |
5408 | ret = -EINVAL; | |
5409 | break; | |
5410 | } | |
5411 | ||
5412 | /* bring the ctx back to life */ | |
206aefde | 5413 | reinit_completion(&ctx->completions[0]); |
edafccee JA |
5414 | percpu_ref_reinit(&ctx->refs); |
5415 | return ret; | |
5416 | } | |
5417 | ||
5418 | SYSCALL_DEFINE4(io_uring_register, unsigned int, fd, unsigned int, opcode, | |
5419 | void __user *, arg, unsigned int, nr_args) | |
5420 | { | |
5421 | struct io_ring_ctx *ctx; | |
5422 | long ret = -EBADF; | |
5423 | struct fd f; | |
5424 | ||
5425 | f = fdget(fd); | |
5426 | if (!f.file) | |
5427 | return -EBADF; | |
5428 | ||
5429 | ret = -EOPNOTSUPP; | |
5430 | if (f.file->f_op != &io_uring_fops) | |
5431 | goto out_fput; | |
5432 | ||
5433 | ctx = f.file->private_data; | |
5434 | ||
5435 | mutex_lock(&ctx->uring_lock); | |
5436 | ret = __io_uring_register(ctx, opcode, arg, nr_args); | |
5437 | mutex_unlock(&ctx->uring_lock); | |
c826bd7a DD |
5438 | trace_io_uring_register(ctx, opcode, ctx->nr_user_files, ctx->nr_user_bufs, |
5439 | ctx->cq_ev_fd != NULL, ret); | |
edafccee JA |
5440 | out_fput: |
5441 | fdput(f); | |
5442 | return ret; | |
5443 | } | |
5444 | ||
2b188cc1 JA |
5445 | static int __init io_uring_init(void) |
5446 | { | |
5447 | req_cachep = KMEM_CACHE(io_kiocb, SLAB_HWCACHE_ALIGN | SLAB_PANIC); | |
5448 | return 0; | |
5449 | }; | |
5450 | __initcall(io_uring_init); |