blob: 86a2bd7219005b149f76b811117d5312ef8a63b8 [file] [log] [blame]
Jens Axboe2b188cc2019-01-07 10:46:33 -07001// SPDX-License-Identifier: GPL-2.0
2/*
3 * Shared application/kernel submission and completion ring pairs, for
4 * supporting fast/efficient IO.
5 *
6 * A note on the read/write ordering memory barriers that are matched between
Stefan Bühler1e84b972019-04-24 23:54:16 +02007 * the application and kernel side.
8 *
9 * After the application reads the CQ ring tail, it must use an
10 * appropriate smp_rmb() to pair with the smp_wmb() the kernel uses
11 * before writing the tail (using smp_load_acquire to read the tail will
12 * do). It also needs a smp_mb() before updating CQ head (ordering the
13 * entry load(s) with the head store), pairing with an implicit barrier
14 * through a control-dependency in io_get_cqring (smp_store_release to
15 * store head will do). Failure to do so could lead to reading invalid
16 * CQ entries.
17 *
18 * Likewise, the application must use an appropriate smp_wmb() before
19 * writing the SQ tail (ordering SQ entry stores with the tail store),
20 * which pairs with smp_load_acquire in io_get_sqring (smp_store_release
21 * to store the tail will do). And it needs a barrier ordering the SQ
22 * head load before writing new SQ entries (smp_load_acquire to read
23 * head will do).
24 *
25 * When using the SQ poll thread (IORING_SETUP_SQPOLL), the application
26 * needs to check the SQ flags for IORING_SQ_NEED_WAKEUP *after*
27 * updating the SQ tail; a full memory barrier smp_mb() is needed
28 * between.
Jens Axboe2b188cc2019-01-07 10:46:33 -070029 *
30 * Also see the examples in the liburing library:
31 *
32 * git://git.kernel.dk/liburing
33 *
34 * io_uring also uses READ/WRITE_ONCE() for _any_ store or load that happens
35 * from data shared between the kernel and application. This is done both
36 * for ordering purposes, but also to ensure that once a value is loaded from
37 * data that the application could potentially modify, it remains stable.
38 *
39 * Copyright (C) 2018-2019 Jens Axboe
Christoph Hellwigc992fe22019-01-11 09:43:02 -070040 * Copyright (c) 2018-2019 Christoph Hellwig
Jens Axboe2b188cc2019-01-07 10:46:33 -070041 */
42#include <linux/kernel.h>
43#include <linux/init.h>
44#include <linux/errno.h>
45#include <linux/syscalls.h>
46#include <linux/compat.h>
47#include <linux/refcount.h>
48#include <linux/uio.h>
49
50#include <linux/sched/signal.h>
51#include <linux/fs.h>
52#include <linux/file.h>
53#include <linux/fdtable.h>
54#include <linux/mm.h>
55#include <linux/mman.h>
56#include <linux/mmu_context.h>
57#include <linux/percpu.h>
58#include <linux/slab.h>
59#include <linux/workqueue.h>
Jens Axboe6c271ce2019-01-10 11:22:30 -070060#include <linux/kthread.h>
Jens Axboe2b188cc2019-01-07 10:46:33 -070061#include <linux/blkdev.h>
Jens Axboeedafcce2019-01-09 09:16:05 -070062#include <linux/bvec.h>
Jens Axboe2b188cc2019-01-07 10:46:33 -070063#include <linux/net.h>
64#include <net/sock.h>
65#include <net/af_unix.h>
Jens Axboe6b063142019-01-10 22:13:58 -070066#include <net/scm.h>
Jens Axboe2b188cc2019-01-07 10:46:33 -070067#include <linux/anon_inodes.h>
68#include <linux/sched/mm.h>
69#include <linux/uaccess.h>
70#include <linux/nospec.h>
Jens Axboeedafcce2019-01-09 09:16:05 -070071#include <linux/sizes.h>
72#include <linux/hugetlb.h>
Jens Axboe2b188cc2019-01-07 10:46:33 -070073
74#include <uapi/linux/io_uring.h>
75
76#include "internal.h"
77
78#define IORING_MAX_ENTRIES 4096
Jens Axboe6b063142019-01-10 22:13:58 -070079#define IORING_MAX_FIXED_FILES 1024
Jens Axboe2b188cc2019-01-07 10:46:33 -070080
81struct io_uring {
82 u32 head ____cacheline_aligned_in_smp;
83 u32 tail ____cacheline_aligned_in_smp;
84};
85
Stefan Bühler1e84b972019-04-24 23:54:16 +020086/*
87 * This data is shared with the application through the mmap at offset
88 * IORING_OFF_SQ_RING.
89 *
90 * The offsets to the member fields are published through struct
91 * io_sqring_offsets when calling io_uring_setup.
92 */
Jens Axboe2b188cc2019-01-07 10:46:33 -070093struct io_sq_ring {
Stefan Bühler1e84b972019-04-24 23:54:16 +020094 /*
95 * Head and tail offsets into the ring; the offsets need to be
96 * masked to get valid indices.
97 *
98 * The kernel controls head and the application controls tail.
99 */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700100 struct io_uring r;
Stefan Bühler1e84b972019-04-24 23:54:16 +0200101 /*
102 * Bitmask to apply to head and tail offsets (constant, equals
103 * ring_entries - 1)
104 */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700105 u32 ring_mask;
Stefan Bühler1e84b972019-04-24 23:54:16 +0200106 /* Ring size (constant, power of 2) */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700107 u32 ring_entries;
Stefan Bühler1e84b972019-04-24 23:54:16 +0200108 /*
109 * Number of invalid entries dropped by the kernel due to
110 * invalid index stored in array
111 *
112 * Written by the kernel, shouldn't be modified by the
113 * application (i.e. get number of "new events" by comparing to
114 * cached value).
115 *
116 * After a new SQ head value was read by the application this
117 * counter includes all submissions that were dropped reaching
118 * the new SQ head (and possibly more).
119 */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700120 u32 dropped;
Stefan Bühler1e84b972019-04-24 23:54:16 +0200121 /*
122 * Runtime flags
123 *
124 * Written by the kernel, shouldn't be modified by the
125 * application.
126 *
127 * The application needs a full memory barrier before checking
128 * for IORING_SQ_NEED_WAKEUP after updating the sq tail.
129 */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700130 u32 flags;
Stefan Bühler1e84b972019-04-24 23:54:16 +0200131 /*
132 * Ring buffer of indices into array of io_uring_sqe, which is
133 * mmapped by the application using the IORING_OFF_SQES offset.
134 *
135 * This indirection could e.g. be used to assign fixed
136 * io_uring_sqe entries to operations and only submit them to
137 * the queue when needed.
138 *
139 * The kernel modifies neither the indices array nor the entries
140 * array.
141 */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700142 u32 array[];
143};
144
Stefan Bühler1e84b972019-04-24 23:54:16 +0200145/*
146 * This data is shared with the application through the mmap at offset
147 * IORING_OFF_CQ_RING.
148 *
149 * The offsets to the member fields are published through struct
150 * io_cqring_offsets when calling io_uring_setup.
151 */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700152struct io_cq_ring {
Stefan Bühler1e84b972019-04-24 23:54:16 +0200153 /*
154 * Head and tail offsets into the ring; the offsets need to be
155 * masked to get valid indices.
156 *
157 * The application controls head and the kernel tail.
158 */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700159 struct io_uring r;
Stefan Bühler1e84b972019-04-24 23:54:16 +0200160 /*
161 * Bitmask to apply to head and tail offsets (constant, equals
162 * ring_entries - 1)
163 */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700164 u32 ring_mask;
Stefan Bühler1e84b972019-04-24 23:54:16 +0200165 /* Ring size (constant, power of 2) */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700166 u32 ring_entries;
Stefan Bühler1e84b972019-04-24 23:54:16 +0200167 /*
168 * Number of completion events lost because the queue was full;
169 * this should be avoided by the application by making sure
170 * there are not more requests pending thatn there is space in
171 * the completion queue.
172 *
173 * Written by the kernel, shouldn't be modified by the
174 * application (i.e. get number of "new events" by comparing to
175 * cached value).
176 *
177 * As completion events come in out of order this counter is not
178 * ordered with any other data.
179 */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700180 u32 overflow;
Stefan Bühler1e84b972019-04-24 23:54:16 +0200181 /*
182 * Ring buffer of completion events.
183 *
184 * The kernel writes completion events fresh every time they are
185 * produced, so the application is allowed to modify pending
186 * entries.
187 */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700188 struct io_uring_cqe cqes[];
189};
190
Jens Axboeedafcce2019-01-09 09:16:05 -0700191struct io_mapped_ubuf {
192 u64 ubuf;
193 size_t len;
194 struct bio_vec *bvec;
195 unsigned int nr_bvecs;
196};
197
Jens Axboe31b51512019-01-18 22:56:34 -0700198struct async_list {
199 spinlock_t lock;
200 atomic_t cnt;
201 struct list_head list;
202
203 struct file *file;
204 off_t io_end;
205 size_t io_pages;
206};
207
Jens Axboe2b188cc2019-01-07 10:46:33 -0700208struct io_ring_ctx {
209 struct {
210 struct percpu_ref refs;
211 } ____cacheline_aligned_in_smp;
212
213 struct {
214 unsigned int flags;
215 bool compat;
216 bool account_mem;
217
218 /* SQ ring */
219 struct io_sq_ring *sq_ring;
220 unsigned cached_sq_head;
221 unsigned sq_entries;
222 unsigned sq_mask;
Jens Axboe6c271ce2019-01-10 11:22:30 -0700223 unsigned sq_thread_idle;
Jens Axboe2b188cc2019-01-07 10:46:33 -0700224 struct io_uring_sqe *sq_sqes;
Jens Axboede0617e2019-04-06 21:51:27 -0600225
226 struct list_head defer_list;
Jens Axboe2b188cc2019-01-07 10:46:33 -0700227 } ____cacheline_aligned_in_smp;
228
229 /* IO offload */
230 struct workqueue_struct *sqo_wq;
Jens Axboe6c271ce2019-01-10 11:22:30 -0700231 struct task_struct *sqo_thread; /* if using sq thread polling */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700232 struct mm_struct *sqo_mm;
Jens Axboe6c271ce2019-01-10 11:22:30 -0700233 wait_queue_head_t sqo_wait;
Jens Axboe2b188cc2019-01-07 10:46:33 -0700234
235 struct {
236 /* CQ ring */
237 struct io_cq_ring *cq_ring;
238 unsigned cached_cq_tail;
239 unsigned cq_entries;
240 unsigned cq_mask;
241 struct wait_queue_head cq_wait;
242 struct fasync_struct *cq_fasync;
Jens Axboe9b402842019-04-11 11:45:41 -0600243 struct eventfd_ctx *cq_ev_fd;
Jens Axboe2b188cc2019-01-07 10:46:33 -0700244 } ____cacheline_aligned_in_smp;
245
Jens Axboe6b063142019-01-10 22:13:58 -0700246 /*
247 * If used, fixed file set. Writers must ensure that ->refs is dead,
248 * readers must ensure that ->refs is alive as long as the file* is
249 * used. Only updated through io_uring_register(2).
250 */
251 struct file **user_files;
252 unsigned nr_user_files;
253
Jens Axboeedafcce2019-01-09 09:16:05 -0700254 /* if used, fixed mapped user buffers */
255 unsigned nr_user_bufs;
256 struct io_mapped_ubuf *user_bufs;
257
Jens Axboe2b188cc2019-01-07 10:46:33 -0700258 struct user_struct *user;
259
260 struct completion ctx_done;
261
262 struct {
263 struct mutex uring_lock;
264 wait_queue_head_t wait;
265 } ____cacheline_aligned_in_smp;
266
267 struct {
268 spinlock_t completion_lock;
Jens Axboedef596e2019-01-09 08:59:42 -0700269 bool poll_multi_file;
270 /*
271 * ->poll_list is protected by the ctx->uring_lock for
272 * io_uring instances that don't use IORING_SETUP_SQPOLL.
273 * For SQPOLL, only the single threaded io_sq_thread() will
274 * manipulate the list, hence no extra locking is needed there.
275 */
276 struct list_head poll_list;
Jens Axboe221c5eb2019-01-17 09:41:58 -0700277 struct list_head cancel_list;
Jens Axboe2b188cc2019-01-07 10:46:33 -0700278 } ____cacheline_aligned_in_smp;
279
Jens Axboe31b51512019-01-18 22:56:34 -0700280 struct async_list pending_async[2];
281
Jens Axboe2b188cc2019-01-07 10:46:33 -0700282#if defined(CONFIG_UNIX)
283 struct socket *ring_sock;
284#endif
285};
286
287struct sqe_submit {
288 const struct io_uring_sqe *sqe;
289 unsigned short index;
290 bool has_user;
Jens Axboedef596e2019-01-09 08:59:42 -0700291 bool needs_lock;
Jens Axboe6c271ce2019-01-10 11:22:30 -0700292 bool needs_fixed_file;
Jens Axboe2b188cc2019-01-07 10:46:33 -0700293};
294
Jens Axboe09bb8392019-03-13 12:39:28 -0600295/*
296 * First field must be the file pointer in all the
297 * iocb unions! See also 'struct kiocb' in <linux/fs.h>
298 */
Jens Axboe221c5eb2019-01-17 09:41:58 -0700299struct io_poll_iocb {
300 struct file *file;
301 struct wait_queue_head *head;
302 __poll_t events;
Jens Axboe8c838782019-03-12 15:48:16 -0600303 bool done;
Jens Axboe221c5eb2019-01-17 09:41:58 -0700304 bool canceled;
305 struct wait_queue_entry wait;
306};
307
Jens Axboe09bb8392019-03-13 12:39:28 -0600308/*
309 * NOTE! Each of the iocb union members has the file pointer
310 * as the first entry in their struct definition. So you can
311 * access the file pointer through any of the sub-structs,
312 * or directly as just 'ki_filp' in this struct.
313 */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700314struct io_kiocb {
Jens Axboe221c5eb2019-01-17 09:41:58 -0700315 union {
Jens Axboe09bb8392019-03-13 12:39:28 -0600316 struct file *file;
Jens Axboe221c5eb2019-01-17 09:41:58 -0700317 struct kiocb rw;
318 struct io_poll_iocb poll;
319 };
Jens Axboe2b188cc2019-01-07 10:46:33 -0700320
321 struct sqe_submit submit;
322
323 struct io_ring_ctx *ctx;
324 struct list_head list;
325 unsigned int flags;
Jens Axboec16361c2019-01-17 08:39:48 -0700326 refcount_t refs;
Stefan Bühler8449eed2019-04-27 20:34:19 +0200327#define REQ_F_NOWAIT 1 /* must not punt to workers */
Jens Axboedef596e2019-01-09 08:59:42 -0700328#define REQ_F_IOPOLL_COMPLETED 2 /* polled IO has completed */
Jens Axboe6b063142019-01-10 22:13:58 -0700329#define REQ_F_FIXED_FILE 4 /* ctx owns file */
Jens Axboe31b51512019-01-18 22:56:34 -0700330#define REQ_F_SEQ_PREV 8 /* sequential with previous */
Stefan Bühlere2033e32019-05-11 19:08:01 +0200331#define REQ_F_IO_DRAIN 16 /* drain existing IO first */
332#define REQ_F_IO_DRAINED 32 /* drain done */
Jens Axboe2b188cc2019-01-07 10:46:33 -0700333 u64 user_data;
Stefan Bühler5dcf8772019-05-01 13:53:36 +0200334 u32 error; /* iopoll result from callback */
Jens Axboede0617e2019-04-06 21:51:27 -0600335 u32 sequence;
Jens Axboe2b188cc2019-01-07 10:46:33 -0700336
337 struct work_struct work;
338};
339
340#define IO_PLUG_THRESHOLD 2
Jens Axboedef596e2019-01-09 08:59:42 -0700341#define IO_IOPOLL_BATCH 8
Jens Axboe2b188cc2019-01-07 10:46:33 -0700342
Jens Axboe9a56a232019-01-09 09:06:50 -0700343struct io_submit_state {
344 struct blk_plug plug;
345
346 /*
Jens Axboe2579f912019-01-09 09:10:43 -0700347 * io_kiocb alloc cache
348 */
349 void *reqs[IO_IOPOLL_BATCH];
350 unsigned int free_reqs;
351 unsigned int cur_req;
352
353 /*
Jens Axboe9a56a232019-01-09 09:06:50 -0700354 * File reference cache
355 */
356 struct file *file;
357 unsigned int fd;
358 unsigned int has_refs;
359 unsigned int used_refs;
360 unsigned int ios_left;
361};
362
Jens Axboede0617e2019-04-06 21:51:27 -0600363static void io_sq_wq_submit_work(struct work_struct *work);
364
Jens Axboe2b188cc2019-01-07 10:46:33 -0700365static struct kmem_cache *req_cachep;
366
367static const struct file_operations io_uring_fops;
368
369struct sock *io_uring_get_socket(struct file *file)
370{
371#if defined(CONFIG_UNIX)
372 if (file->f_op == &io_uring_fops) {
373 struct io_ring_ctx *ctx = file->private_data;
374
375 return ctx->ring_sock->sk;
376 }
377#endif
378 return NULL;
379}
380EXPORT_SYMBOL(io_uring_get_socket);
381
382static void io_ring_ctx_ref_free(struct percpu_ref *ref)
383{
384 struct io_ring_ctx *ctx = container_of(ref, struct io_ring_ctx, refs);
385
386 complete(&ctx->ctx_done);
387}
388
389static struct io_ring_ctx *io_ring_ctx_alloc(struct io_uring_params *p)
390{
391 struct io_ring_ctx *ctx;
Jens Axboe31b51512019-01-18 22:56:34 -0700392 int i;
Jens Axboe2b188cc2019-01-07 10:46:33 -0700393
394 ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
395 if (!ctx)
396 return NULL;
397
398 if (percpu_ref_init(&ctx->refs, io_ring_ctx_ref_free, 0, GFP_KERNEL)) {
399 kfree(ctx);
400 return NULL;
401 }
402
403 ctx->flags = p->flags;
404 init_waitqueue_head(&ctx->cq_wait);
405 init_completion(&ctx->ctx_done);
406 mutex_init(&ctx->uring_lock);
407 init_waitqueue_head(&ctx->wait);
Jens Axboe31b51512019-01-18 22:56:34 -0700408 for (i = 0; i < ARRAY_SIZE(ctx->pending_async); i++) {
409 spin_lock_init(&ctx->pending_async[i].lock);
410 INIT_LIST_HEAD(&ctx->pending_async[i].list);
411 atomic_set(&ctx->pending_async[i].cnt, 0);
412 }
Jens Axboe2b188cc2019-01-07 10:46:33 -0700413 spin_lock_init(&ctx->completion_lock);
Jens Axboedef596e2019-01-09 08:59:42 -0700414 INIT_LIST_HEAD(&ctx->poll_list);
Jens Axboe221c5eb2019-01-17 09:41:58 -0700415 INIT_LIST_HEAD(&ctx->cancel_list);
Jens Axboede0617e2019-04-06 21:51:27 -0600416 INIT_LIST_HEAD(&ctx->defer_list);
Jens Axboe2b188cc2019-01-07 10:46:33 -0700417 return ctx;
418}
419
Jens Axboede0617e2019-04-06 21:51:27 -0600420static inline bool io_sequence_defer(struct io_ring_ctx *ctx,
421 struct io_kiocb *req)
422{
423 if ((req->flags & (REQ_F_IO_DRAIN|REQ_F_IO_DRAINED)) != REQ_F_IO_DRAIN)
424 return false;
425
426 return req->sequence > ctx->cached_cq_tail + ctx->sq_ring->dropped;
427}
428
429static struct io_kiocb *io_get_deferred_req(struct io_ring_ctx *ctx)
430{
431 struct io_kiocb *req;
432
433 if (list_empty(&ctx->defer_list))
434 return NULL;
435
436 req = list_first_entry(&ctx->defer_list, struct io_kiocb, list);
437 if (!io_sequence_defer(ctx, req)) {
438 list_del_init(&req->list);
439 return req;
440 }
441
442 return NULL;
443}
444
445static void __io_commit_cqring(struct io_ring_ctx *ctx)
Jens Axboe2b188cc2019-01-07 10:46:33 -0700446{
447 struct io_cq_ring *ring = ctx->cq_ring;
448
449 if (ctx->cached_cq_tail != READ_ONCE(ring->r.tail)) {
450 /* order cqe stores with ring update */
451 smp_store_release(&ring->r.tail, ctx->cached_cq_tail);
452
Jens Axboe2b188cc2019-01-07 10:46:33 -0700453 if (wq_has_sleeper(&ctx->cq_wait)) {
454 wake_up_interruptible(&ctx->cq_wait);
455 kill_fasync(&ctx->cq_fasync, SIGIO, POLL_IN);
456 }
457 }
458}
459
Jens Axboede0617e2019-04-06 21:51:27 -0600460static void io_commit_cqring(struct io_ring_ctx *ctx)
461{
462 struct io_kiocb *req;
463
464 __io_commit_cqring(ctx);
465
466 while ((req = io_get_deferred_req(ctx)) != NULL) {
467 req->flags |= REQ_F_IO_DRAINED;
468 queue_work(ctx->sqo_wq, &req->work);
469 }
470}
471
Jens Axboe2b188cc2019-01-07 10:46:33 -0700472static struct io_uring_cqe *io_get_cqring(struct io_ring_ctx *ctx)
473{
474 struct io_cq_ring *ring = ctx->cq_ring;
475 unsigned tail;
476
477 tail = ctx->cached_cq_tail;
Stefan Bühler115e12e2019-04-24 23:54:18 +0200478 /*
479 * writes to the cq entry need to come after reading head; the
480 * control dependency is enough as we're using WRITE_ONCE to
481 * fill the cq entry
482 */
Jens Axboe74f464e2019-04-17 08:57:48 -0600483 if (tail - READ_ONCE(ring->r.head) == ring->ring_entries)
Jens Axboe2b188cc2019-01-07 10:46:33 -0700484 return NULL;
485
486 ctx->cached_cq_tail++;
487 return &ring->cqes[tail & ctx->cq_mask];
488}
489
490static void io_cqring_fill_event(struct io_ring_ctx *ctx, u64 ki_user_data,
Jens Axboec71ffb62019-05-13 20:58:29 -0600491 long res)
Jens Axboe2b188cc2019-01-07 10:46:33 -0700492{
493 struct io_uring_cqe *cqe;
494
495 /*
496 * If we can't get a cq entry, userspace overflowed the
497 * submission (by quite a lot). Increment the overflow count in
498 * the ring.
499 */
500 cqe = io_get_cqring(ctx);
501 if (cqe) {
502 WRITE_ONCE(cqe->user_data, ki_user_data);
503 WRITE_ONCE(cqe->res, res);
Jens Axboec71ffb62019-05-13 20:58:29 -0600504 WRITE_ONCE(cqe->flags, 0);
Jens Axboe2b188cc2019-01-07 10:46:33 -0700505 } else {
506 unsigned overflow = READ_ONCE(ctx->cq_ring->overflow);
507
508 WRITE_ONCE(ctx->cq_ring->overflow, overflow + 1);
509 }
510}
511
Jens Axboe8c838782019-03-12 15:48:16 -0600512static void io_cqring_ev_posted(struct io_ring_ctx *ctx)
513{
514 if (waitqueue_active(&ctx->wait))
515 wake_up(&ctx->wait);
516 if (waitqueue_active(&ctx->sqo_wait))
517 wake_up(&ctx->sqo_wait);
Jens Axboe9b402842019-04-11 11:45:41 -0600518 if (ctx->cq_ev_fd)
519 eventfd_signal(ctx->cq_ev_fd, 1);
Jens Axboe8c838782019-03-12 15:48:16 -0600520}
521
522static void io_cqring_add_event(struct io_ring_ctx *ctx, u64 user_data,
Jens Axboec71ffb62019-05-13 20:58:29 -0600523 long res)
Jens Axboe2b188cc2019-01-07 10:46:33 -0700524{
525 unsigned long flags;
526
527 spin_lock_irqsave(&ctx->completion_lock, flags);
Jens Axboec71ffb62019-05-13 20:58:29 -0600528 io_cqring_fill_event(ctx, user_data, res);
Jens Axboe2b188cc2019-01-07 10:46:33 -0700529 io_commit_cqring(ctx);
530 spin_unlock_irqrestore(&ctx->completion_lock, flags);
531
Jens Axboe8c838782019-03-12 15:48:16 -0600532 io_cqring_ev_posted(ctx);
Jens Axboe2b188cc2019-01-07 10:46:33 -0700533}
534
535static void io_ring_drop_ctx_refs(struct io_ring_ctx *ctx, unsigned refs)
536{
537 percpu_ref_put_many(&ctx->refs, refs);
538
539 if (waitqueue_active(&ctx->wait))
540 wake_up(&ctx->wait);
541}
542
Jens Axboe2579f912019-01-09 09:10:43 -0700543static struct io_kiocb *io_get_req(struct io_ring_ctx *ctx,
544 struct io_submit_state *state)
Jens Axboe2b188cc2019-01-07 10:46:33 -0700545{
Jens Axboefd6fab22019-03-14 16:30:06 -0600546 gfp_t gfp = GFP_KERNEL | __GFP_NOWARN;
Jens Axboe2b188cc2019-01-07 10:46:33 -0700547 struct io_kiocb *req;
548
549 if (!percpu_ref_tryget(&ctx->refs))
550 return NULL;
551
Jens Axboe2579f912019-01-09 09:10:43 -0700552 if (!state) {
Jens Axboefd6fab22019-03-14 16:30:06 -0600553 req = kmem_cache_alloc(req_cachep, gfp);
Jens Axboe2579f912019-01-09 09:10:43 -0700554 if (unlikely(!req))
555 goto out;
556 } else if (!state->free_reqs) {
557 size_t sz;
558 int ret;
559
560 sz = min_t(size_t, state->ios_left, ARRAY_SIZE(state->reqs));
Jens Axboefd6fab22019-03-14 16:30:06 -0600561 ret = kmem_cache_alloc_bulk(req_cachep, gfp, sz, state->reqs);
562
563 /*
564 * Bulk alloc is all-or-nothing. If we fail to get a batch,
565 * retry single alloc to be on the safe side.
566 */
567 if (unlikely(ret <= 0)) {
568 state->reqs[0] = kmem_cache_alloc(req_cachep, gfp);
569 if (!state->reqs[0])
570 goto out;
571 ret = 1;
572 }
Jens Axboe2579f912019-01-09 09:10:43 -0700573 state->free_reqs = ret - 1;
574 state->cur_req = 1;
575 req = state->reqs[0];
576 } else {
577 req = state->reqs[state->cur_req];
578 state->free_reqs--;
579 state->cur_req++;
Jens Axboe2b188cc2019-01-07 10:46:33 -0700580 }
581
Jens Axboe2579f912019-01-09 09:10:43 -0700582 req->ctx = ctx;
583 req->flags = 0;
Jens Axboee65ef562019-03-12 10:16:44 -0600584 /* one is dropped after submission, the other at completion */
585 refcount_set(&req->refs, 2);
Jens Axboe2579f912019-01-09 09:10:43 -0700586 return req;
587out:
Jens Axboe2b188cc2019-01-07 10:46:33 -0700588 io_ring_drop_ctx_refs(ctx, 1);
589 return NULL;
590}
591
Jens Axboedef596e2019-01-09 08:59:42 -0700592static void io_free_req_many(struct io_ring_ctx *ctx, void **reqs, int *nr)
593{
594 if (*nr) {
595 kmem_cache_free_bulk(req_cachep, *nr, reqs);
596 io_ring_drop_ctx_refs(ctx, *nr);
597 *nr = 0;
598 }
599}
600
Jens Axboe2b188cc2019-01-07 10:46:33 -0700601static void io_free_req(struct io_kiocb *req)
602{
Jens Axboe09bb8392019-03-13 12:39:28 -0600603 if (req->file && !(req->flags & REQ_F_FIXED_FILE))
604 fput(req->file);
Jens Axboee65ef562019-03-12 10:16:44 -0600605 io_ring_drop_ctx_refs(req->ctx, 1);
606 kmem_cache_free(req_cachep, req);
607}
608
609static void io_put_req(struct io_kiocb *req)
610{
611 if (refcount_dec_and_test(&req->refs))
612 io_free_req(req);
Jens Axboe2b188cc2019-01-07 10:46:33 -0700613}
614
Jens Axboedef596e2019-01-09 08:59:42 -0700615/*
616 * Find and free completed poll iocbs
617 */
618static void io_iopoll_complete(struct io_ring_ctx *ctx, unsigned int *nr_events,
619 struct list_head *done)
620{
621 void *reqs[IO_IOPOLL_BATCH];
622 struct io_kiocb *req;
Jens Axboe09bb8392019-03-13 12:39:28 -0600623 int to_free;
Jens Axboedef596e2019-01-09 08:59:42 -0700624
Jens Axboe09bb8392019-03-13 12:39:28 -0600625 to_free = 0;
Jens Axboedef596e2019-01-09 08:59:42 -0700626 while (!list_empty(done)) {
627 req = list_first_entry(done, struct io_kiocb, list);
628 list_del(&req->list);
629
Jens Axboec71ffb62019-05-13 20:58:29 -0600630 io_cqring_fill_event(ctx, req->user_data, req->error);
Jens Axboedef596e2019-01-09 08:59:42 -0700631 (*nr_events)++;
632
Jens Axboe09bb8392019-03-13 12:39:28 -0600633 if (refcount_dec_and_test(&req->refs)) {
634 /* If we're not using fixed files, we have to pair the
635 * completion part with the file put. Use regular
636 * completions for those, only batch free for fixed
637 * file.
638 */
639 if (req->flags & REQ_F_FIXED_FILE) {
640 reqs[to_free++] = req;
641 if (to_free == ARRAY_SIZE(reqs))
642 io_free_req_many(ctx, reqs, &to_free);
Jens Axboe6b063142019-01-10 22:13:58 -0700643 } else {
Jens Axboe09bb8392019-03-13 12:39:28 -0600644 io_free_req(req);
Jens Axboe6b063142019-01-10 22:13:58 -0700645 }
Jens Axboe9a56a232019-01-09 09:06:50 -0700646 }
Jens Axboedef596e2019-01-09 08:59:42 -0700647 }
Jens Axboedef596e2019-01-09 08:59:42 -0700648
Jens Axboe09bb8392019-03-13 12:39:28 -0600649 io_commit_cqring(ctx);
Jens Axboedef596e2019-01-09 08:59:42 -0700650 io_free_req_many(ctx, reqs, &to_free);
651}
652
653static int io_do_iopoll(struct io_ring_ctx *ctx, unsigned int *nr_events,
654 long min)
655{
656 struct io_kiocb *req, *tmp;
657 LIST_HEAD(done);
658 bool spin;
659 int ret;
660
661 /*
662 * Only spin for completions if we don't have multiple devices hanging
663 * off our complete list, and we're under the requested amount.
664 */
665 spin = !ctx->poll_multi_file && *nr_events < min;
666
667 ret = 0;
668 list_for_each_entry_safe(req, tmp, &ctx->poll_list, list) {
669 struct kiocb *kiocb = &req->rw;
670
671 /*
672 * Move completed entries to our local list. If we find a
673 * request that requires polling, break out and complete
674 * the done list first, if we have entries there.
675 */
676 if (req->flags & REQ_F_IOPOLL_COMPLETED) {
677 list_move_tail(&req->list, &done);
678 continue;
679 }
680 if (!list_empty(&done))
681 break;
682
683 ret = kiocb->ki_filp->f_op->iopoll(kiocb, spin);
684 if (ret < 0)
685 break;
686
687 if (ret && spin)
688 spin = false;
689 ret = 0;
690 }
691
692 if (!list_empty(&done))
693 io_iopoll_complete(ctx, nr_events, &done);
694
695 return ret;
696}
697
698/*
699 * Poll for a mininum of 'min' events. Note that if min == 0 we consider that a
700 * non-spinning poll check - we'll still enter the driver poll loop, but only
701 * as a non-spinning completion check.
702 */
703static int io_iopoll_getevents(struct io_ring_ctx *ctx, unsigned int *nr_events,
704 long min)
705{
706 while (!list_empty(&ctx->poll_list)) {
707 int ret;
708
709 ret = io_do_iopoll(ctx, nr_events, min);
710 if (ret < 0)
711 return ret;
712 if (!min || *nr_events >= min)
713 return 0;
714 }
715
716 return 1;
717}
718
719/*
720 * We can't just wait for polled events to come to us, we have to actively
721 * find and complete them.
722 */
723static void io_iopoll_reap_events(struct io_ring_ctx *ctx)
724{
725 if (!(ctx->flags & IORING_SETUP_IOPOLL))
726 return;
727
728 mutex_lock(&ctx->uring_lock);
729 while (!list_empty(&ctx->poll_list)) {
730 unsigned int nr_events = 0;
731
732 io_iopoll_getevents(ctx, &nr_events, 1);
733 }
734 mutex_unlock(&ctx->uring_lock);
735}
736
737static int io_iopoll_check(struct io_ring_ctx *ctx, unsigned *nr_events,
738 long min)
739{
740 int ret = 0;
741
742 do {
743 int tmin = 0;
744
745 if (*nr_events < min)
746 tmin = min - *nr_events;
747
748 ret = io_iopoll_getevents(ctx, nr_events, tmin);
749 if (ret <= 0)
750 break;
751 ret = 0;
752 } while (min && !*nr_events && !need_resched());
753
754 return ret;
755}
756
Jens Axboe2b188cc2019-01-07 10:46:33 -0700757static void kiocb_end_write(struct kiocb *kiocb)
758{
759 if (kiocb->ki_flags & IOCB_WRITE) {
760 struct inode *inode = file_inode(kiocb->ki_filp);
761
762 /*
763 * Tell lockdep we inherited freeze protection from submission
764 * thread.
765 */
766 if (S_ISREG(inode->i_mode))
767 __sb_writers_acquired(inode->i_sb, SB_FREEZE_WRITE);
768 file_end_write(kiocb->ki_filp);
769 }
770}
771
772static void io_complete_rw(struct kiocb *kiocb, long res, long res2)
773{
774 struct io_kiocb *req = container_of(kiocb, struct io_kiocb, rw);
775
776 kiocb_end_write(kiocb);
777
Jens Axboec71ffb62019-05-13 20:58:29 -0600778 io_cqring_add_event(req->ctx, req->user_data, res);
Jens Axboee65ef562019-03-12 10:16:44 -0600779 io_put_req(req);
Jens Axboe2b188cc2019-01-07 10:46:33 -0700780}
781
Jens Axboedef596e2019-01-09 08:59:42 -0700782static void io_complete_rw_iopoll(struct kiocb *kiocb, long res, long res2)
783{
784 struct io_kiocb *req = container_of(kiocb, struct io_kiocb, rw);
785
786 kiocb_end_write(kiocb);
787
788 req->error = res;
789 if (res != -EAGAIN)
790 req->flags |= REQ_F_IOPOLL_COMPLETED;
791}
792
793/*
794 * After the iocb has been issued, it's safe to be found on the poll list.
795 * Adding the kiocb to the list AFTER submission ensures that we don't
796 * find it from a io_iopoll_getevents() thread before the issuer is done
797 * accessing the kiocb cookie.
798 */
799static void io_iopoll_req_issued(struct io_kiocb *req)
800{
801 struct io_ring_ctx *ctx = req->ctx;
802
803 /*
804 * Track whether we have multiple files in our lists. This will impact
805 * how we do polling eventually, not spinning if we're on potentially
806 * different devices.
807 */
808 if (list_empty(&ctx->poll_list)) {
809 ctx->poll_multi_file = false;
810 } else if (!ctx->poll_multi_file) {
811 struct io_kiocb *list_req;
812
813 list_req = list_first_entry(&ctx->poll_list, struct io_kiocb,
814 list);
815 if (list_req->rw.ki_filp != req->rw.ki_filp)
816 ctx->poll_multi_file = true;
817 }
818
819 /*
820 * For fast devices, IO may have already completed. If it has, add
821 * it to the front so we find it first.
822 */
823 if (req->flags & REQ_F_IOPOLL_COMPLETED)
824 list_add(&req->list, &ctx->poll_list);
825 else
826 list_add_tail(&req->list, &ctx->poll_list);
827}
828
Jens Axboe3d6770f2019-04-13 11:50:54 -0600829static void io_file_put(struct io_submit_state *state)
Jens Axboe9a56a232019-01-09 09:06:50 -0700830{
Jens Axboe3d6770f2019-04-13 11:50:54 -0600831 if (state->file) {
Jens Axboe9a56a232019-01-09 09:06:50 -0700832 int diff = state->has_refs - state->used_refs;
833
834 if (diff)
835 fput_many(state->file, diff);
836 state->file = NULL;
837 }
838}
839
840/*
841 * Get as many references to a file as we have IOs left in this submission,
842 * assuming most submissions are for one file, or at least that each file
843 * has more than one submission.
844 */
845static struct file *io_file_get(struct io_submit_state *state, int fd)
846{
847 if (!state)
848 return fget(fd);
849
850 if (state->file) {
851 if (state->fd == fd) {
852 state->used_refs++;
853 state->ios_left--;
854 return state->file;
855 }
Jens Axboe3d6770f2019-04-13 11:50:54 -0600856 io_file_put(state);
Jens Axboe9a56a232019-01-09 09:06:50 -0700857 }
858 state->file = fget_many(fd, state->ios_left);
859 if (!state->file)
860 return NULL;
861
862 state->fd = fd;
863 state->has_refs = state->ios_left;
864 state->used_refs = 1;
865 state->ios_left--;
866 return state->file;
867}
868
Jens Axboe2b188cc2019-01-07 10:46:33 -0700869/*
870 * If we tracked the file through the SCM inflight mechanism, we could support
871 * any file. For now, just ensure that anything potentially problematic is done
872 * inline.
873 */
874static bool io_file_supports_async(struct file *file)
875{
876 umode_t mode = file_inode(file)->i_mode;
877
878 if (S_ISBLK(mode) || S_ISCHR(mode))
879 return true;
880 if (S_ISREG(mode) && file->f_op != &io_uring_fops)
881 return true;
882
883 return false;
884}
885
Jens Axboe6c271ce2019-01-10 11:22:30 -0700886static int io_prep_rw(struct io_kiocb *req, const struct sqe_submit *s,
Jens Axboe8358e3a2019-04-23 08:17:58 -0600887 bool force_nonblock)
Jens Axboe2b188cc2019-01-07 10:46:33 -0700888{
Jens Axboe6c271ce2019-01-10 11:22:30 -0700889 const struct io_uring_sqe *sqe = s->sqe;
Jens Axboedef596e2019-01-09 08:59:42 -0700890 struct io_ring_ctx *ctx = req->ctx;
Jens Axboe2b188cc2019-01-07 10:46:33 -0700891 struct kiocb *kiocb = &req->rw;
Jens Axboe09bb8392019-03-13 12:39:28 -0600892 unsigned ioprio;
893 int ret;
Jens Axboe2b188cc2019-01-07 10:46:33 -0700894
Jens Axboe09bb8392019-03-13 12:39:28 -0600895 if (!req->file)
896 return -EBADF;
Jens Axboe2b188cc2019-01-07 10:46:33 -0700897
Jens Axboe09bb8392019-03-13 12:39:28 -0600898 if (force_nonblock && !io_file_supports_async(req->file))
899 force_nonblock = false;
Jens Axboe6b063142019-01-10 22:13:58 -0700900
Jens Axboe2b188cc2019-01-07 10:46:33 -0700901 kiocb->ki_pos = READ_ONCE(sqe->off);
902 kiocb->ki_flags = iocb_flags(kiocb->ki_filp);
903 kiocb->ki_hint = ki_hint_validate(file_write_hint(kiocb->ki_filp));
904
905 ioprio = READ_ONCE(sqe->ioprio);
906 if (ioprio) {
907 ret = ioprio_check_cap(ioprio);
908 if (ret)
Jens Axboe09bb8392019-03-13 12:39:28 -0600909 return ret;
Jens Axboe2b188cc2019-01-07 10:46:33 -0700910
911 kiocb->ki_ioprio = ioprio;
912 } else
913 kiocb->ki_ioprio = get_current_ioprio();
914
915 ret = kiocb_set_rw_flags(kiocb, READ_ONCE(sqe->rw_flags));
916 if (unlikely(ret))
Jens Axboe09bb8392019-03-13 12:39:28 -0600917 return ret;
Stefan Bühler8449eed2019-04-27 20:34:19 +0200918
919 /* don't allow async punt if RWF_NOWAIT was requested */
920 if (kiocb->ki_flags & IOCB_NOWAIT)
921 req->flags |= REQ_F_NOWAIT;
922
923 if (force_nonblock)
Jens Axboe2b188cc2019-01-07 10:46:33 -0700924 kiocb->ki_flags |= IOCB_NOWAIT;
Stefan Bühler8449eed2019-04-27 20:34:19 +0200925
Jens Axboedef596e2019-01-09 08:59:42 -0700926 if (ctx->flags & IORING_SETUP_IOPOLL) {
Jens Axboedef596e2019-01-09 08:59:42 -0700927 if (!(kiocb->ki_flags & IOCB_DIRECT) ||
928 !kiocb->ki_filp->f_op->iopoll)
Jens Axboe09bb8392019-03-13 12:39:28 -0600929 return -EOPNOTSUPP;
Jens Axboe2b188cc2019-01-07 10:46:33 -0700930
Jens Axboedef596e2019-01-09 08:59:42 -0700931 req->error = 0;
932 kiocb->ki_flags |= IOCB_HIPRI;
933 kiocb->ki_complete = io_complete_rw_iopoll;
934 } else {
Jens Axboe09bb8392019-03-13 12:39:28 -0600935 if (kiocb->ki_flags & IOCB_HIPRI)
936 return -EINVAL;
Jens Axboedef596e2019-01-09 08:59:42 -0700937 kiocb->ki_complete = io_complete_rw;
938 }
Jens Axboe2b188cc2019-01-07 10:46:33 -0700939 return 0;
Jens Axboe2b188cc2019-01-07 10:46:33 -0700940}
941
942static inline void io_rw_done(struct kiocb *kiocb, ssize_t ret)
943{
944 switch (ret) {
945 case -EIOCBQUEUED:
946 break;
947 case -ERESTARTSYS:
948 case -ERESTARTNOINTR:
949 case -ERESTARTNOHAND:
950 case -ERESTART_RESTARTBLOCK:
951 /*
952 * We can't just restart the syscall, since previously
953 * submitted sqes may already be in progress. Just fail this
954 * IO with EINTR.
955 */
956 ret = -EINTR;
957 /* fall through */
958 default:
959 kiocb->ki_complete(kiocb, ret, 0);
960 }
961}
962
Jens Axboeedafcce2019-01-09 09:16:05 -0700963static int io_import_fixed(struct io_ring_ctx *ctx, int rw,
964 const struct io_uring_sqe *sqe,
965 struct iov_iter *iter)
966{
967 size_t len = READ_ONCE(sqe->len);
968 struct io_mapped_ubuf *imu;
969 unsigned index, buf_index;
970 size_t offset;
971 u64 buf_addr;
972
973 /* attempt to use fixed buffers without having provided iovecs */
974 if (unlikely(!ctx->user_bufs))
975 return -EFAULT;
976
977 buf_index = READ_ONCE(sqe->buf_index);
978 if (unlikely(buf_index >= ctx->nr_user_bufs))
979 return -EFAULT;
980
981 index = array_index_nospec(buf_index, ctx->nr_user_bufs);
982 imu = &ctx->user_bufs[index];
983 buf_addr = READ_ONCE(sqe->addr);
984
985 /* overflow */
986 if (buf_addr + len < buf_addr)
987 return -EFAULT;
988 /* not inside the mapped region */
989 if (buf_addr < imu->ubuf || buf_addr + len > imu->ubuf + imu->len)
990 return -EFAULT;
991
992 /*
993 * May not be a start of buffer, set size appropriately
994 * and advance us to the beginning.
995 */
996 offset = buf_addr - imu->ubuf;
997 iov_iter_bvec(iter, rw, imu->bvec, imu->nr_bvecs, offset + len);
998 if (offset)
999 iov_iter_advance(iter, offset);
Jens Axboe875f1d02019-02-27 13:05:25 -07001000
1001 /* don't drop a reference to these pages */
1002 iter->type |= ITER_BVEC_FLAG_NO_REF;
Jens Axboeedafcce2019-01-09 09:16:05 -07001003 return 0;
1004}
1005
Jens Axboe2b188cc2019-01-07 10:46:33 -07001006static int io_import_iovec(struct io_ring_ctx *ctx, int rw,
1007 const struct sqe_submit *s, struct iovec **iovec,
1008 struct iov_iter *iter)
1009{
1010 const struct io_uring_sqe *sqe = s->sqe;
1011 void __user *buf = u64_to_user_ptr(READ_ONCE(sqe->addr));
1012 size_t sqe_len = READ_ONCE(sqe->len);
Jens Axboeedafcce2019-01-09 09:16:05 -07001013 u8 opcode;
1014
1015 /*
1016 * We're reading ->opcode for the second time, but the first read
1017 * doesn't care whether it's _FIXED or not, so it doesn't matter
1018 * whether ->opcode changes concurrently. The first read does care
1019 * about whether it is a READ or a WRITE, so we don't trust this read
1020 * for that purpose and instead let the caller pass in the read/write
1021 * flag.
1022 */
1023 opcode = READ_ONCE(sqe->opcode);
1024 if (opcode == IORING_OP_READ_FIXED ||
1025 opcode == IORING_OP_WRITE_FIXED) {
Jens Axboee0c5c572019-03-12 10:18:47 -06001026 int ret = io_import_fixed(ctx, rw, sqe, iter);
Jens Axboeedafcce2019-01-09 09:16:05 -07001027 *iovec = NULL;
1028 return ret;
1029 }
Jens Axboe2b188cc2019-01-07 10:46:33 -07001030
1031 if (!s->has_user)
1032 return -EFAULT;
1033
1034#ifdef CONFIG_COMPAT
1035 if (ctx->compat)
1036 return compat_import_iovec(rw, buf, sqe_len, UIO_FASTIOV,
1037 iovec, iter);
1038#endif
1039
1040 return import_iovec(rw, buf, sqe_len, UIO_FASTIOV, iovec, iter);
1041}
1042
Jens Axboe31b51512019-01-18 22:56:34 -07001043/*
1044 * Make a note of the last file/offset/direction we punted to async
1045 * context. We'll use this information to see if we can piggy back a
1046 * sequential request onto the previous one, if it's still hasn't been
1047 * completed by the async worker.
1048 */
1049static void io_async_list_note(int rw, struct io_kiocb *req, size_t len)
1050{
1051 struct async_list *async_list = &req->ctx->pending_async[rw];
1052 struct kiocb *kiocb = &req->rw;
1053 struct file *filp = kiocb->ki_filp;
1054 off_t io_end = kiocb->ki_pos + len;
1055
1056 if (filp == async_list->file && kiocb->ki_pos == async_list->io_end) {
1057 unsigned long max_pages;
1058
1059 /* Use 8x RA size as a decent limiter for both reads/writes */
1060 max_pages = filp->f_ra.ra_pages;
1061 if (!max_pages)
Nikolay Borisovb5420232019-03-11 23:28:13 -07001062 max_pages = VM_READAHEAD_PAGES;
Jens Axboe31b51512019-01-18 22:56:34 -07001063 max_pages *= 8;
1064
1065 /* If max pages are exceeded, reset the state */
1066 len >>= PAGE_SHIFT;
1067 if (async_list->io_pages + len <= max_pages) {
1068 req->flags |= REQ_F_SEQ_PREV;
1069 async_list->io_pages += len;
1070 } else {
1071 io_end = 0;
1072 async_list->io_pages = 0;
1073 }
1074 }
1075
1076 /* New file? Reset state. */
1077 if (async_list->file != filp) {
1078 async_list->io_pages = 0;
1079 async_list->file = filp;
1080 }
1081 async_list->io_end = io_end;
1082}
1083
Jens Axboee0c5c572019-03-12 10:18:47 -06001084static int io_read(struct io_kiocb *req, const struct sqe_submit *s,
Jens Axboe8358e3a2019-04-23 08:17:58 -06001085 bool force_nonblock)
Jens Axboe2b188cc2019-01-07 10:46:33 -07001086{
1087 struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs;
1088 struct kiocb *kiocb = &req->rw;
1089 struct iov_iter iter;
1090 struct file *file;
Jens Axboe31b51512019-01-18 22:56:34 -07001091 size_t iov_count;
Jens Axboee0c5c572019-03-12 10:18:47 -06001092 int ret;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001093
Jens Axboe8358e3a2019-04-23 08:17:58 -06001094 ret = io_prep_rw(req, s, force_nonblock);
Jens Axboe2b188cc2019-01-07 10:46:33 -07001095 if (ret)
1096 return ret;
1097 file = kiocb->ki_filp;
1098
Jens Axboe2b188cc2019-01-07 10:46:33 -07001099 if (unlikely(!(file->f_mode & FMODE_READ)))
Jens Axboe09bb8392019-03-13 12:39:28 -06001100 return -EBADF;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001101 if (unlikely(!file->f_op->read_iter))
Jens Axboe09bb8392019-03-13 12:39:28 -06001102 return -EINVAL;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001103
1104 ret = io_import_iovec(req->ctx, READ, s, &iovec, &iter);
1105 if (ret)
Jens Axboe09bb8392019-03-13 12:39:28 -06001106 return ret;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001107
Jens Axboe31b51512019-01-18 22:56:34 -07001108 iov_count = iov_iter_count(&iter);
1109 ret = rw_verify_area(READ, file, &kiocb->ki_pos, iov_count);
Jens Axboe2b188cc2019-01-07 10:46:33 -07001110 if (!ret) {
1111 ssize_t ret2;
1112
1113 /* Catch -EAGAIN return for forced non-blocking submission */
1114 ret2 = call_read_iter(file, kiocb, &iter);
Jens Axboe31b51512019-01-18 22:56:34 -07001115 if (!force_nonblock || ret2 != -EAGAIN) {
Jens Axboe2b188cc2019-01-07 10:46:33 -07001116 io_rw_done(kiocb, ret2);
Jens Axboe31b51512019-01-18 22:56:34 -07001117 } else {
1118 /*
1119 * If ->needs_lock is true, we're already in async
1120 * context.
1121 */
1122 if (!s->needs_lock)
1123 io_async_list_note(READ, req, iov_count);
Jens Axboe2b188cc2019-01-07 10:46:33 -07001124 ret = -EAGAIN;
Jens Axboe31b51512019-01-18 22:56:34 -07001125 }
Jens Axboe2b188cc2019-01-07 10:46:33 -07001126 }
1127 kfree(iovec);
Jens Axboe2b188cc2019-01-07 10:46:33 -07001128 return ret;
1129}
1130
Jens Axboee0c5c572019-03-12 10:18:47 -06001131static int io_write(struct io_kiocb *req, const struct sqe_submit *s,
Jens Axboe8358e3a2019-04-23 08:17:58 -06001132 bool force_nonblock)
Jens Axboe2b188cc2019-01-07 10:46:33 -07001133{
1134 struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs;
1135 struct kiocb *kiocb = &req->rw;
1136 struct iov_iter iter;
1137 struct file *file;
Jens Axboe31b51512019-01-18 22:56:34 -07001138 size_t iov_count;
Jens Axboee0c5c572019-03-12 10:18:47 -06001139 int ret;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001140
Jens Axboe8358e3a2019-04-23 08:17:58 -06001141 ret = io_prep_rw(req, s, force_nonblock);
Jens Axboe2b188cc2019-01-07 10:46:33 -07001142 if (ret)
1143 return ret;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001144
Jens Axboe2b188cc2019-01-07 10:46:33 -07001145 file = kiocb->ki_filp;
1146 if (unlikely(!(file->f_mode & FMODE_WRITE)))
Jens Axboe09bb8392019-03-13 12:39:28 -06001147 return -EBADF;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001148 if (unlikely(!file->f_op->write_iter))
Jens Axboe09bb8392019-03-13 12:39:28 -06001149 return -EINVAL;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001150
1151 ret = io_import_iovec(req->ctx, WRITE, s, &iovec, &iter);
1152 if (ret)
Jens Axboe09bb8392019-03-13 12:39:28 -06001153 return ret;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001154
Jens Axboe31b51512019-01-18 22:56:34 -07001155 iov_count = iov_iter_count(&iter);
1156
1157 ret = -EAGAIN;
1158 if (force_nonblock && !(kiocb->ki_flags & IOCB_DIRECT)) {
1159 /* If ->needs_lock is true, we're already in async context. */
1160 if (!s->needs_lock)
1161 io_async_list_note(WRITE, req, iov_count);
1162 goto out_free;
1163 }
1164
1165 ret = rw_verify_area(WRITE, file, &kiocb->ki_pos, iov_count);
Jens Axboe2b188cc2019-01-07 10:46:33 -07001166 if (!ret) {
Roman Penyaev9bf79332019-03-25 20:09:24 +01001167 ssize_t ret2;
1168
Jens Axboe2b188cc2019-01-07 10:46:33 -07001169 /*
1170 * Open-code file_start_write here to grab freeze protection,
1171 * which will be released by another thread in
1172 * io_complete_rw(). Fool lockdep by telling it the lock got
1173 * released so that it doesn't complain about the held lock when
1174 * we return to userspace.
1175 */
1176 if (S_ISREG(file_inode(file)->i_mode)) {
1177 __sb_start_write(file_inode(file)->i_sb,
1178 SB_FREEZE_WRITE, true);
1179 __sb_writers_release(file_inode(file)->i_sb,
1180 SB_FREEZE_WRITE);
1181 }
1182 kiocb->ki_flags |= IOCB_WRITE;
Roman Penyaev9bf79332019-03-25 20:09:24 +01001183
1184 ret2 = call_write_iter(file, kiocb, &iter);
1185 if (!force_nonblock || ret2 != -EAGAIN) {
1186 io_rw_done(kiocb, ret2);
1187 } else {
1188 /*
1189 * If ->needs_lock is true, we're already in async
1190 * context.
1191 */
1192 if (!s->needs_lock)
1193 io_async_list_note(WRITE, req, iov_count);
1194 ret = -EAGAIN;
1195 }
Jens Axboe2b188cc2019-01-07 10:46:33 -07001196 }
Jens Axboe31b51512019-01-18 22:56:34 -07001197out_free:
Jens Axboe2b188cc2019-01-07 10:46:33 -07001198 kfree(iovec);
Jens Axboe2b188cc2019-01-07 10:46:33 -07001199 return ret;
1200}
1201
1202/*
1203 * IORING_OP_NOP just posts a completion event, nothing else.
1204 */
1205static int io_nop(struct io_kiocb *req, u64 user_data)
1206{
1207 struct io_ring_ctx *ctx = req->ctx;
1208 long err = 0;
1209
Jens Axboedef596e2019-01-09 08:59:42 -07001210 if (unlikely(ctx->flags & IORING_SETUP_IOPOLL))
1211 return -EINVAL;
1212
Jens Axboec71ffb62019-05-13 20:58:29 -06001213 io_cqring_add_event(ctx, user_data, err);
Jens Axboee65ef562019-03-12 10:16:44 -06001214 io_put_req(req);
Jens Axboe2b188cc2019-01-07 10:46:33 -07001215 return 0;
1216}
1217
Christoph Hellwigc992fe22019-01-11 09:43:02 -07001218static int io_prep_fsync(struct io_kiocb *req, const struct io_uring_sqe *sqe)
1219{
Jens Axboe6b063142019-01-10 22:13:58 -07001220 struct io_ring_ctx *ctx = req->ctx;
Christoph Hellwigc992fe22019-01-11 09:43:02 -07001221
Jens Axboe09bb8392019-03-13 12:39:28 -06001222 if (!req->file)
1223 return -EBADF;
Christoph Hellwigc992fe22019-01-11 09:43:02 -07001224
Jens Axboe6b063142019-01-10 22:13:58 -07001225 if (unlikely(ctx->flags & IORING_SETUP_IOPOLL))
Jens Axboedef596e2019-01-09 08:59:42 -07001226 return -EINVAL;
Jens Axboeedafcce2019-01-09 09:16:05 -07001227 if (unlikely(sqe->addr || sqe->ioprio || sqe->buf_index))
Christoph Hellwigc992fe22019-01-11 09:43:02 -07001228 return -EINVAL;
1229
Christoph Hellwigc992fe22019-01-11 09:43:02 -07001230 return 0;
1231}
1232
1233static int io_fsync(struct io_kiocb *req, const struct io_uring_sqe *sqe,
1234 bool force_nonblock)
1235{
1236 loff_t sqe_off = READ_ONCE(sqe->off);
1237 loff_t sqe_len = READ_ONCE(sqe->len);
1238 loff_t end = sqe_off + sqe_len;
1239 unsigned fsync_flags;
1240 int ret;
1241
1242 fsync_flags = READ_ONCE(sqe->fsync_flags);
1243 if (unlikely(fsync_flags & ~IORING_FSYNC_DATASYNC))
1244 return -EINVAL;
1245
1246 ret = io_prep_fsync(req, sqe);
1247 if (ret)
1248 return ret;
1249
1250 /* fsync always requires a blocking context */
1251 if (force_nonblock)
1252 return -EAGAIN;
1253
1254 ret = vfs_fsync_range(req->rw.ki_filp, sqe_off,
1255 end > 0 ? end : LLONG_MAX,
1256 fsync_flags & IORING_FSYNC_DATASYNC);
1257
Jens Axboec71ffb62019-05-13 20:58:29 -06001258 io_cqring_add_event(req->ctx, sqe->user_data, ret);
Jens Axboee65ef562019-03-12 10:16:44 -06001259 io_put_req(req);
Christoph Hellwigc992fe22019-01-11 09:43:02 -07001260 return 0;
1261}
1262
Jens Axboe5d17b4a2019-04-09 14:56:44 -06001263static int io_prep_sfr(struct io_kiocb *req, const struct io_uring_sqe *sqe)
1264{
1265 struct io_ring_ctx *ctx = req->ctx;
1266 int ret = 0;
1267
1268 if (!req->file)
1269 return -EBADF;
Jens Axboe5d17b4a2019-04-09 14:56:44 -06001270
1271 if (unlikely(ctx->flags & IORING_SETUP_IOPOLL))
1272 return -EINVAL;
1273 if (unlikely(sqe->addr || sqe->ioprio || sqe->buf_index))
1274 return -EINVAL;
1275
Jens Axboe5d17b4a2019-04-09 14:56:44 -06001276 return ret;
1277}
1278
1279static int io_sync_file_range(struct io_kiocb *req,
1280 const struct io_uring_sqe *sqe,
1281 bool force_nonblock)
1282{
1283 loff_t sqe_off;
1284 loff_t sqe_len;
1285 unsigned flags;
1286 int ret;
1287
1288 ret = io_prep_sfr(req, sqe);
1289 if (ret)
1290 return ret;
1291
1292 /* sync_file_range always requires a blocking context */
1293 if (force_nonblock)
1294 return -EAGAIN;
1295
1296 sqe_off = READ_ONCE(sqe->off);
1297 sqe_len = READ_ONCE(sqe->len);
1298 flags = READ_ONCE(sqe->sync_range_flags);
1299
1300 ret = sync_file_range(req->rw.ki_filp, sqe_off, sqe_len, flags);
1301
Jens Axboec71ffb62019-05-13 20:58:29 -06001302 io_cqring_add_event(req->ctx, sqe->user_data, ret);
Jens Axboe5d17b4a2019-04-09 14:56:44 -06001303 io_put_req(req);
1304 return 0;
1305}
1306
Jens Axboe221c5eb2019-01-17 09:41:58 -07001307static void io_poll_remove_one(struct io_kiocb *req)
1308{
1309 struct io_poll_iocb *poll = &req->poll;
1310
1311 spin_lock(&poll->head->lock);
1312 WRITE_ONCE(poll->canceled, true);
1313 if (!list_empty(&poll->wait.entry)) {
1314 list_del_init(&poll->wait.entry);
1315 queue_work(req->ctx->sqo_wq, &req->work);
1316 }
1317 spin_unlock(&poll->head->lock);
1318
1319 list_del_init(&req->list);
1320}
1321
1322static void io_poll_remove_all(struct io_ring_ctx *ctx)
1323{
1324 struct io_kiocb *req;
1325
1326 spin_lock_irq(&ctx->completion_lock);
1327 while (!list_empty(&ctx->cancel_list)) {
1328 req = list_first_entry(&ctx->cancel_list, struct io_kiocb,list);
1329 io_poll_remove_one(req);
1330 }
1331 spin_unlock_irq(&ctx->completion_lock);
1332}
1333
1334/*
1335 * Find a running poll command that matches one specified in sqe->addr,
1336 * and remove it if found.
1337 */
1338static int io_poll_remove(struct io_kiocb *req, const struct io_uring_sqe *sqe)
1339{
1340 struct io_ring_ctx *ctx = req->ctx;
1341 struct io_kiocb *poll_req, *next;
1342 int ret = -ENOENT;
1343
1344 if (unlikely(req->ctx->flags & IORING_SETUP_IOPOLL))
1345 return -EINVAL;
1346 if (sqe->ioprio || sqe->off || sqe->len || sqe->buf_index ||
1347 sqe->poll_events)
1348 return -EINVAL;
1349
1350 spin_lock_irq(&ctx->completion_lock);
1351 list_for_each_entry_safe(poll_req, next, &ctx->cancel_list, list) {
1352 if (READ_ONCE(sqe->addr) == poll_req->user_data) {
1353 io_poll_remove_one(poll_req);
1354 ret = 0;
1355 break;
1356 }
1357 }
1358 spin_unlock_irq(&ctx->completion_lock);
1359
Jens Axboec71ffb62019-05-13 20:58:29 -06001360 io_cqring_add_event(req->ctx, sqe->user_data, ret);
Jens Axboee65ef562019-03-12 10:16:44 -06001361 io_put_req(req);
Jens Axboe221c5eb2019-01-17 09:41:58 -07001362 return 0;
1363}
1364
Jens Axboe8c838782019-03-12 15:48:16 -06001365static void io_poll_complete(struct io_ring_ctx *ctx, struct io_kiocb *req,
1366 __poll_t mask)
Jens Axboe221c5eb2019-01-17 09:41:58 -07001367{
Jens Axboe8c838782019-03-12 15:48:16 -06001368 req->poll.done = true;
Jens Axboec71ffb62019-05-13 20:58:29 -06001369 io_cqring_fill_event(ctx, req->user_data, mangle_poll(mask));
Jens Axboe8c838782019-03-12 15:48:16 -06001370 io_commit_cqring(ctx);
Jens Axboe221c5eb2019-01-17 09:41:58 -07001371}
1372
1373static void io_poll_complete_work(struct work_struct *work)
1374{
1375 struct io_kiocb *req = container_of(work, struct io_kiocb, work);
1376 struct io_poll_iocb *poll = &req->poll;
1377 struct poll_table_struct pt = { ._key = poll->events };
1378 struct io_ring_ctx *ctx = req->ctx;
1379 __poll_t mask = 0;
1380
1381 if (!READ_ONCE(poll->canceled))
1382 mask = vfs_poll(poll->file, &pt) & poll->events;
1383
1384 /*
1385 * Note that ->ki_cancel callers also delete iocb from active_reqs after
1386 * calling ->ki_cancel. We need the ctx_lock roundtrip here to
1387 * synchronize with them. In the cancellation case the list_del_init
1388 * itself is not actually needed, but harmless so we keep it in to
1389 * avoid further branches in the fast path.
1390 */
1391 spin_lock_irq(&ctx->completion_lock);
1392 if (!mask && !READ_ONCE(poll->canceled)) {
1393 add_wait_queue(poll->head, &poll->wait);
1394 spin_unlock_irq(&ctx->completion_lock);
1395 return;
1396 }
1397 list_del_init(&req->list);
Jens Axboe8c838782019-03-12 15:48:16 -06001398 io_poll_complete(ctx, req, mask);
Jens Axboe221c5eb2019-01-17 09:41:58 -07001399 spin_unlock_irq(&ctx->completion_lock);
1400
Jens Axboe8c838782019-03-12 15:48:16 -06001401 io_cqring_ev_posted(ctx);
1402 io_put_req(req);
Jens Axboe221c5eb2019-01-17 09:41:58 -07001403}
1404
1405static int io_poll_wake(struct wait_queue_entry *wait, unsigned mode, int sync,
1406 void *key)
1407{
1408 struct io_poll_iocb *poll = container_of(wait, struct io_poll_iocb,
1409 wait);
1410 struct io_kiocb *req = container_of(poll, struct io_kiocb, poll);
1411 struct io_ring_ctx *ctx = req->ctx;
1412 __poll_t mask = key_to_poll(key);
Jens Axboe8c838782019-03-12 15:48:16 -06001413 unsigned long flags;
Jens Axboe221c5eb2019-01-17 09:41:58 -07001414
1415 /* for instances that support it check for an event match first: */
Jens Axboe8c838782019-03-12 15:48:16 -06001416 if (mask && !(mask & poll->events))
1417 return 0;
Jens Axboe221c5eb2019-01-17 09:41:58 -07001418
1419 list_del_init(&poll->wait.entry);
Jens Axboe8c838782019-03-12 15:48:16 -06001420
1421 if (mask && spin_trylock_irqsave(&ctx->completion_lock, flags)) {
1422 list_del(&req->list);
1423 io_poll_complete(ctx, req, mask);
1424 spin_unlock_irqrestore(&ctx->completion_lock, flags);
1425
1426 io_cqring_ev_posted(ctx);
1427 io_put_req(req);
1428 } else {
1429 queue_work(ctx->sqo_wq, &req->work);
1430 }
1431
Jens Axboe221c5eb2019-01-17 09:41:58 -07001432 return 1;
1433}
1434
1435struct io_poll_table {
1436 struct poll_table_struct pt;
1437 struct io_kiocb *req;
1438 int error;
1439};
1440
1441static void io_poll_queue_proc(struct file *file, struct wait_queue_head *head,
1442 struct poll_table_struct *p)
1443{
1444 struct io_poll_table *pt = container_of(p, struct io_poll_table, pt);
1445
1446 if (unlikely(pt->req->poll.head)) {
1447 pt->error = -EINVAL;
1448 return;
1449 }
1450
1451 pt->error = 0;
1452 pt->req->poll.head = head;
1453 add_wait_queue(head, &pt->req->poll.wait);
1454}
1455
1456static int io_poll_add(struct io_kiocb *req, const struct io_uring_sqe *sqe)
1457{
1458 struct io_poll_iocb *poll = &req->poll;
1459 struct io_ring_ctx *ctx = req->ctx;
1460 struct io_poll_table ipt;
Jens Axboe8c838782019-03-12 15:48:16 -06001461 bool cancel = false;
Jens Axboe221c5eb2019-01-17 09:41:58 -07001462 __poll_t mask;
1463 u16 events;
Jens Axboe221c5eb2019-01-17 09:41:58 -07001464
1465 if (unlikely(req->ctx->flags & IORING_SETUP_IOPOLL))
1466 return -EINVAL;
1467 if (sqe->addr || sqe->ioprio || sqe->off || sqe->len || sqe->buf_index)
1468 return -EINVAL;
Jens Axboe09bb8392019-03-13 12:39:28 -06001469 if (!poll->file)
1470 return -EBADF;
Jens Axboe221c5eb2019-01-17 09:41:58 -07001471
1472 INIT_WORK(&req->work, io_poll_complete_work);
1473 events = READ_ONCE(sqe->poll_events);
1474 poll->events = demangle_poll(events) | EPOLLERR | EPOLLHUP;
1475
Jens Axboe221c5eb2019-01-17 09:41:58 -07001476 poll->head = NULL;
Jens Axboe8c838782019-03-12 15:48:16 -06001477 poll->done = false;
Jens Axboe221c5eb2019-01-17 09:41:58 -07001478 poll->canceled = false;
1479
1480 ipt.pt._qproc = io_poll_queue_proc;
1481 ipt.pt._key = poll->events;
1482 ipt.req = req;
1483 ipt.error = -EINVAL; /* same as no support for IOCB_CMD_POLL */
1484
1485 /* initialized the list so that we can do list_empty checks */
1486 INIT_LIST_HEAD(&poll->wait.entry);
1487 init_waitqueue_func_entry(&poll->wait, io_poll_wake);
1488
Jens Axboe221c5eb2019-01-17 09:41:58 -07001489 mask = vfs_poll(poll->file, &ipt.pt) & poll->events;
Jens Axboe221c5eb2019-01-17 09:41:58 -07001490
1491 spin_lock_irq(&ctx->completion_lock);
Jens Axboe8c838782019-03-12 15:48:16 -06001492 if (likely(poll->head)) {
1493 spin_lock(&poll->head->lock);
1494 if (unlikely(list_empty(&poll->wait.entry))) {
1495 if (ipt.error)
1496 cancel = true;
1497 ipt.error = 0;
1498 mask = 0;
1499 }
1500 if (mask || ipt.error)
1501 list_del_init(&poll->wait.entry);
1502 else if (cancel)
1503 WRITE_ONCE(poll->canceled, true);
1504 else if (!poll->done) /* actually waiting for an event */
1505 list_add_tail(&req->list, &ctx->cancel_list);
1506 spin_unlock(&poll->head->lock);
Jens Axboe221c5eb2019-01-17 09:41:58 -07001507 }
Jens Axboe8c838782019-03-12 15:48:16 -06001508 if (mask) { /* no async, we'd stolen it */
Jens Axboe8c838782019-03-12 15:48:16 -06001509 ipt.error = 0;
1510 io_poll_complete(ctx, req, mask);
1511 }
Jens Axboe221c5eb2019-01-17 09:41:58 -07001512 spin_unlock_irq(&ctx->completion_lock);
1513
Jens Axboe8c838782019-03-12 15:48:16 -06001514 if (mask) {
1515 io_cqring_ev_posted(ctx);
Jens Axboee65ef562019-03-12 10:16:44 -06001516 io_put_req(req);
Jens Axboe221c5eb2019-01-17 09:41:58 -07001517 }
Jens Axboe8c838782019-03-12 15:48:16 -06001518 return ipt.error;
Jens Axboe221c5eb2019-01-17 09:41:58 -07001519}
1520
Jens Axboede0617e2019-04-06 21:51:27 -06001521static int io_req_defer(struct io_ring_ctx *ctx, struct io_kiocb *req,
1522 const struct io_uring_sqe *sqe)
1523{
1524 struct io_uring_sqe *sqe_copy;
1525
1526 if (!io_sequence_defer(ctx, req) && list_empty(&ctx->defer_list))
1527 return 0;
1528
1529 sqe_copy = kmalloc(sizeof(*sqe_copy), GFP_KERNEL);
1530 if (!sqe_copy)
1531 return -EAGAIN;
1532
1533 spin_lock_irq(&ctx->completion_lock);
1534 if (!io_sequence_defer(ctx, req) && list_empty(&ctx->defer_list)) {
1535 spin_unlock_irq(&ctx->completion_lock);
1536 kfree(sqe_copy);
1537 return 0;
1538 }
1539
1540 memcpy(sqe_copy, sqe, sizeof(*sqe_copy));
1541 req->submit.sqe = sqe_copy;
1542
1543 INIT_WORK(&req->work, io_sq_wq_submit_work);
1544 list_add_tail(&req->list, &ctx->defer_list);
1545 spin_unlock_irq(&ctx->completion_lock);
1546 return -EIOCBQUEUED;
1547}
1548
Jens Axboe2b188cc2019-01-07 10:46:33 -07001549static int __io_submit_sqe(struct io_ring_ctx *ctx, struct io_kiocb *req,
Jens Axboe8358e3a2019-04-23 08:17:58 -06001550 const struct sqe_submit *s, bool force_nonblock)
Jens Axboe2b188cc2019-01-07 10:46:33 -07001551{
Jens Axboee0c5c572019-03-12 10:18:47 -06001552 int ret, opcode;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001553
1554 if (unlikely(s->index >= ctx->sq_entries))
1555 return -EINVAL;
1556 req->user_data = READ_ONCE(s->sqe->user_data);
1557
1558 opcode = READ_ONCE(s->sqe->opcode);
1559 switch (opcode) {
1560 case IORING_OP_NOP:
1561 ret = io_nop(req, req->user_data);
1562 break;
1563 case IORING_OP_READV:
Jens Axboeedafcce2019-01-09 09:16:05 -07001564 if (unlikely(s->sqe->buf_index))
1565 return -EINVAL;
Jens Axboe8358e3a2019-04-23 08:17:58 -06001566 ret = io_read(req, s, force_nonblock);
Jens Axboe2b188cc2019-01-07 10:46:33 -07001567 break;
1568 case IORING_OP_WRITEV:
Jens Axboeedafcce2019-01-09 09:16:05 -07001569 if (unlikely(s->sqe->buf_index))
1570 return -EINVAL;
Jens Axboe8358e3a2019-04-23 08:17:58 -06001571 ret = io_write(req, s, force_nonblock);
Jens Axboeedafcce2019-01-09 09:16:05 -07001572 break;
1573 case IORING_OP_READ_FIXED:
Jens Axboe8358e3a2019-04-23 08:17:58 -06001574 ret = io_read(req, s, force_nonblock);
Jens Axboeedafcce2019-01-09 09:16:05 -07001575 break;
1576 case IORING_OP_WRITE_FIXED:
Jens Axboe8358e3a2019-04-23 08:17:58 -06001577 ret = io_write(req, s, force_nonblock);
Jens Axboe2b188cc2019-01-07 10:46:33 -07001578 break;
Christoph Hellwigc992fe22019-01-11 09:43:02 -07001579 case IORING_OP_FSYNC:
1580 ret = io_fsync(req, s->sqe, force_nonblock);
1581 break;
Jens Axboe221c5eb2019-01-17 09:41:58 -07001582 case IORING_OP_POLL_ADD:
1583 ret = io_poll_add(req, s->sqe);
1584 break;
1585 case IORING_OP_POLL_REMOVE:
1586 ret = io_poll_remove(req, s->sqe);
1587 break;
Jens Axboe5d17b4a2019-04-09 14:56:44 -06001588 case IORING_OP_SYNC_FILE_RANGE:
1589 ret = io_sync_file_range(req, s->sqe, force_nonblock);
1590 break;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001591 default:
1592 ret = -EINVAL;
1593 break;
1594 }
1595
Jens Axboedef596e2019-01-09 08:59:42 -07001596 if (ret)
1597 return ret;
1598
1599 if (ctx->flags & IORING_SETUP_IOPOLL) {
1600 if (req->error == -EAGAIN)
1601 return -EAGAIN;
1602
1603 /* workqueue context doesn't hold uring_lock, grab it now */
1604 if (s->needs_lock)
1605 mutex_lock(&ctx->uring_lock);
1606 io_iopoll_req_issued(req);
1607 if (s->needs_lock)
1608 mutex_unlock(&ctx->uring_lock);
1609 }
1610
1611 return 0;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001612}
1613
Jens Axboe31b51512019-01-18 22:56:34 -07001614static struct async_list *io_async_list_from_sqe(struct io_ring_ctx *ctx,
1615 const struct io_uring_sqe *sqe)
1616{
1617 switch (sqe->opcode) {
1618 case IORING_OP_READV:
1619 case IORING_OP_READ_FIXED:
1620 return &ctx->pending_async[READ];
1621 case IORING_OP_WRITEV:
1622 case IORING_OP_WRITE_FIXED:
1623 return &ctx->pending_async[WRITE];
1624 default:
1625 return NULL;
1626 }
1627}
1628
Jens Axboeedafcce2019-01-09 09:16:05 -07001629static inline bool io_sqe_needs_user(const struct io_uring_sqe *sqe)
1630{
1631 u8 opcode = READ_ONCE(sqe->opcode);
1632
1633 return !(opcode == IORING_OP_READ_FIXED ||
1634 opcode == IORING_OP_WRITE_FIXED);
1635}
1636
Jens Axboe2b188cc2019-01-07 10:46:33 -07001637static void io_sq_wq_submit_work(struct work_struct *work)
1638{
1639 struct io_kiocb *req = container_of(work, struct io_kiocb, work);
Jens Axboe2b188cc2019-01-07 10:46:33 -07001640 struct io_ring_ctx *ctx = req->ctx;
Jens Axboe31b51512019-01-18 22:56:34 -07001641 struct mm_struct *cur_mm = NULL;
1642 struct async_list *async_list;
1643 LIST_HEAD(req_list);
Jens Axboeedafcce2019-01-09 09:16:05 -07001644 mm_segment_t old_fs;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001645 int ret;
1646
Jens Axboe31b51512019-01-18 22:56:34 -07001647 async_list = io_async_list_from_sqe(ctx, req->submit.sqe);
1648restart:
1649 do {
1650 struct sqe_submit *s = &req->submit;
1651 const struct io_uring_sqe *sqe = s->sqe;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001652
Stefan Bühler8449eed2019-04-27 20:34:19 +02001653 /* Ensure we clear previously set non-block flag */
Jens Axboe31b51512019-01-18 22:56:34 -07001654 req->rw.ki_flags &= ~IOCB_NOWAIT;
1655
1656 ret = 0;
1657 if (io_sqe_needs_user(sqe) && !cur_mm) {
1658 if (!mmget_not_zero(ctx->sqo_mm)) {
1659 ret = -EFAULT;
1660 } else {
1661 cur_mm = ctx->sqo_mm;
1662 use_mm(cur_mm);
1663 old_fs = get_fs();
1664 set_fs(USER_DS);
1665 }
1666 }
1667
1668 if (!ret) {
1669 s->has_user = cur_mm != NULL;
1670 s->needs_lock = true;
1671 do {
Jens Axboe8358e3a2019-04-23 08:17:58 -06001672 ret = __io_submit_sqe(ctx, req, s, false);
Jens Axboe31b51512019-01-18 22:56:34 -07001673 /*
1674 * We can get EAGAIN for polled IO even though
1675 * we're forcing a sync submission from here,
1676 * since we can't wait for request slots on the
1677 * block side.
1678 */
1679 if (ret != -EAGAIN)
1680 break;
1681 cond_resched();
1682 } while (1);
1683 }
Jens Axboe817869d2019-04-30 14:44:05 -06001684
1685 /* drop submission reference */
1686 io_put_req(req);
1687
Jens Axboe31b51512019-01-18 22:56:34 -07001688 if (ret) {
Jens Axboec71ffb62019-05-13 20:58:29 -06001689 io_cqring_add_event(ctx, sqe->user_data, ret);
Jens Axboee65ef562019-03-12 10:16:44 -06001690 io_put_req(req);
Jens Axboe31b51512019-01-18 22:56:34 -07001691 }
1692
1693 /* async context always use a copy of the sqe */
1694 kfree(sqe);
1695
1696 if (!async_list)
1697 break;
1698 if (!list_empty(&req_list)) {
1699 req = list_first_entry(&req_list, struct io_kiocb,
1700 list);
1701 list_del(&req->list);
1702 continue;
1703 }
1704 if (list_empty(&async_list->list))
1705 break;
1706
1707 req = NULL;
1708 spin_lock(&async_list->lock);
1709 if (list_empty(&async_list->list)) {
1710 spin_unlock(&async_list->lock);
1711 break;
1712 }
1713 list_splice_init(&async_list->list, &req_list);
1714 spin_unlock(&async_list->lock);
1715
1716 req = list_first_entry(&req_list, struct io_kiocb, list);
1717 list_del(&req->list);
1718 } while (req);
Jens Axboeedafcce2019-01-09 09:16:05 -07001719
1720 /*
Jens Axboe31b51512019-01-18 22:56:34 -07001721 * Rare case of racing with a submitter. If we find the count has
1722 * dropped to zero AND we have pending work items, then restart
1723 * the processing. This is a tiny race window.
Jens Axboeedafcce2019-01-09 09:16:05 -07001724 */
Jens Axboe31b51512019-01-18 22:56:34 -07001725 if (async_list) {
1726 ret = atomic_dec_return(&async_list->cnt);
1727 while (!ret && !list_empty(&async_list->list)) {
1728 spin_lock(&async_list->lock);
1729 atomic_inc(&async_list->cnt);
1730 list_splice_init(&async_list->list, &req_list);
1731 spin_unlock(&async_list->lock);
1732
1733 if (!list_empty(&req_list)) {
1734 req = list_first_entry(&req_list,
1735 struct io_kiocb, list);
1736 list_del(&req->list);
1737 goto restart;
1738 }
1739 ret = atomic_dec_return(&async_list->cnt);
Jens Axboeedafcce2019-01-09 09:16:05 -07001740 }
Jens Axboeedafcce2019-01-09 09:16:05 -07001741 }
Jens Axboe2b188cc2019-01-07 10:46:33 -07001742
Jens Axboe31b51512019-01-18 22:56:34 -07001743 if (cur_mm) {
Jens Axboeedafcce2019-01-09 09:16:05 -07001744 set_fs(old_fs);
Jens Axboe31b51512019-01-18 22:56:34 -07001745 unuse_mm(cur_mm);
1746 mmput(cur_mm);
Jens Axboeedafcce2019-01-09 09:16:05 -07001747 }
Jens Axboe31b51512019-01-18 22:56:34 -07001748}
Jens Axboe2b188cc2019-01-07 10:46:33 -07001749
Jens Axboe31b51512019-01-18 22:56:34 -07001750/*
1751 * See if we can piggy back onto previously submitted work, that is still
1752 * running. We currently only allow this if the new request is sequential
1753 * to the previous one we punted.
1754 */
1755static bool io_add_to_prev_work(struct async_list *list, struct io_kiocb *req)
1756{
1757 bool ret = false;
1758
1759 if (!list)
1760 return false;
1761 if (!(req->flags & REQ_F_SEQ_PREV))
1762 return false;
1763 if (!atomic_read(&list->cnt))
1764 return false;
1765
1766 ret = true;
1767 spin_lock(&list->lock);
1768 list_add_tail(&req->list, &list->list);
1769 if (!atomic_read(&list->cnt)) {
1770 list_del_init(&req->list);
1771 ret = false;
1772 }
1773 spin_unlock(&list->lock);
1774 return ret;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001775}
1776
Jens Axboe09bb8392019-03-13 12:39:28 -06001777static bool io_op_needs_file(const struct io_uring_sqe *sqe)
1778{
1779 int op = READ_ONCE(sqe->opcode);
1780
1781 switch (op) {
1782 case IORING_OP_NOP:
1783 case IORING_OP_POLL_REMOVE:
1784 return false;
1785 default:
1786 return true;
1787 }
1788}
1789
1790static int io_req_set_file(struct io_ring_ctx *ctx, const struct sqe_submit *s,
1791 struct io_submit_state *state, struct io_kiocb *req)
1792{
1793 unsigned flags;
1794 int fd;
1795
1796 flags = READ_ONCE(s->sqe->flags);
1797 fd = READ_ONCE(s->sqe->fd);
1798
Jens Axboede0617e2019-04-06 21:51:27 -06001799 if (flags & IOSQE_IO_DRAIN) {
1800 req->flags |= REQ_F_IO_DRAIN;
1801 req->sequence = ctx->cached_sq_head - 1;
1802 }
1803
Jens Axboe09bb8392019-03-13 12:39:28 -06001804 if (!io_op_needs_file(s->sqe)) {
1805 req->file = NULL;
1806 return 0;
1807 }
1808
1809 if (flags & IOSQE_FIXED_FILE) {
1810 if (unlikely(!ctx->user_files ||
1811 (unsigned) fd >= ctx->nr_user_files))
1812 return -EBADF;
1813 req->file = ctx->user_files[fd];
1814 req->flags |= REQ_F_FIXED_FILE;
1815 } else {
1816 if (s->needs_fixed_file)
1817 return -EBADF;
1818 req->file = io_file_get(state, fd);
1819 if (unlikely(!req->file))
1820 return -EBADF;
1821 }
1822
1823 return 0;
1824}
1825
Jens Axboe9a56a232019-01-09 09:06:50 -07001826static int io_submit_sqe(struct io_ring_ctx *ctx, struct sqe_submit *s,
1827 struct io_submit_state *state)
Jens Axboe2b188cc2019-01-07 10:46:33 -07001828{
1829 struct io_kiocb *req;
Jens Axboee0c5c572019-03-12 10:18:47 -06001830 int ret;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001831
1832 /* enforce forwards compatibility on users */
Jens Axboede0617e2019-04-06 21:51:27 -06001833 if (unlikely(s->sqe->flags & ~(IOSQE_FIXED_FILE | IOSQE_IO_DRAIN)))
Jens Axboe2b188cc2019-01-07 10:46:33 -07001834 return -EINVAL;
1835
Jens Axboe2579f912019-01-09 09:10:43 -07001836 req = io_get_req(ctx, state);
Jens Axboe2b188cc2019-01-07 10:46:33 -07001837 if (unlikely(!req))
1838 return -EAGAIN;
1839
Jens Axboe09bb8392019-03-13 12:39:28 -06001840 ret = io_req_set_file(ctx, s, state, req);
1841 if (unlikely(ret))
1842 goto out;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001843
Jens Axboede0617e2019-04-06 21:51:27 -06001844 ret = io_req_defer(ctx, req, s->sqe);
1845 if (ret) {
1846 if (ret == -EIOCBQUEUED)
1847 ret = 0;
1848 return ret;
1849 }
1850
Jens Axboe8358e3a2019-04-23 08:17:58 -06001851 ret = __io_submit_sqe(ctx, req, s, true);
Stefan Bühler8449eed2019-04-27 20:34:19 +02001852 if (ret == -EAGAIN && !(req->flags & REQ_F_NOWAIT)) {
Jens Axboe2b188cc2019-01-07 10:46:33 -07001853 struct io_uring_sqe *sqe_copy;
1854
1855 sqe_copy = kmalloc(sizeof(*sqe_copy), GFP_KERNEL);
1856 if (sqe_copy) {
Jens Axboe31b51512019-01-18 22:56:34 -07001857 struct async_list *list;
1858
Jens Axboe2b188cc2019-01-07 10:46:33 -07001859 memcpy(sqe_copy, s->sqe, sizeof(*sqe_copy));
1860 s->sqe = sqe_copy;
1861
1862 memcpy(&req->submit, s, sizeof(*s));
Jens Axboe31b51512019-01-18 22:56:34 -07001863 list = io_async_list_from_sqe(ctx, s->sqe);
1864 if (!io_add_to_prev_work(list, req)) {
1865 if (list)
1866 atomic_inc(&list->cnt);
1867 INIT_WORK(&req->work, io_sq_wq_submit_work);
1868 queue_work(ctx->sqo_wq, &req->work);
1869 }
Jens Axboee65ef562019-03-12 10:16:44 -06001870
1871 /*
1872 * Queued up for async execution, worker will release
1873 * submit reference when the iocb is actually
1874 * submitted.
1875 */
1876 return 0;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001877 }
1878 }
Jens Axboee65ef562019-03-12 10:16:44 -06001879
Jens Axboe09bb8392019-03-13 12:39:28 -06001880out:
Jens Axboee65ef562019-03-12 10:16:44 -06001881 /* drop submission reference */
1882 io_put_req(req);
1883
1884 /* and drop final reference, if we failed */
Jens Axboe2b188cc2019-01-07 10:46:33 -07001885 if (ret)
Jens Axboee65ef562019-03-12 10:16:44 -06001886 io_put_req(req);
Jens Axboe2b188cc2019-01-07 10:46:33 -07001887
1888 return ret;
1889}
1890
Jens Axboe9a56a232019-01-09 09:06:50 -07001891/*
1892 * Batched submission is done, ensure local IO is flushed out.
1893 */
1894static void io_submit_state_end(struct io_submit_state *state)
1895{
1896 blk_finish_plug(&state->plug);
Jens Axboe3d6770f2019-04-13 11:50:54 -06001897 io_file_put(state);
Jens Axboe2579f912019-01-09 09:10:43 -07001898 if (state->free_reqs)
1899 kmem_cache_free_bulk(req_cachep, state->free_reqs,
1900 &state->reqs[state->cur_req]);
Jens Axboe9a56a232019-01-09 09:06:50 -07001901}
1902
1903/*
1904 * Start submission side cache.
1905 */
1906static void io_submit_state_start(struct io_submit_state *state,
1907 struct io_ring_ctx *ctx, unsigned max_ios)
1908{
1909 blk_start_plug(&state->plug);
Jens Axboe2579f912019-01-09 09:10:43 -07001910 state->free_reqs = 0;
Jens Axboe9a56a232019-01-09 09:06:50 -07001911 state->file = NULL;
1912 state->ios_left = max_ios;
1913}
1914
Jens Axboe2b188cc2019-01-07 10:46:33 -07001915static void io_commit_sqring(struct io_ring_ctx *ctx)
1916{
1917 struct io_sq_ring *ring = ctx->sq_ring;
1918
1919 if (ctx->cached_sq_head != READ_ONCE(ring->r.head)) {
1920 /*
1921 * Ensure any loads from the SQEs are done at this point,
1922 * since once we write the new head, the application could
1923 * write new data to them.
1924 */
1925 smp_store_release(&ring->r.head, ctx->cached_sq_head);
Jens Axboe2b188cc2019-01-07 10:46:33 -07001926 }
1927}
1928
1929/*
Jens Axboe2b188cc2019-01-07 10:46:33 -07001930 * Fetch an sqe, if one is available. Note that s->sqe will point to memory
1931 * that is mapped by userspace. This means that care needs to be taken to
1932 * ensure that reads are stable, as we cannot rely on userspace always
1933 * being a good citizen. If members of the sqe are validated and then later
1934 * used, it's important that those reads are done through READ_ONCE() to
1935 * prevent a re-load down the line.
1936 */
1937static bool io_get_sqring(struct io_ring_ctx *ctx, struct sqe_submit *s)
1938{
1939 struct io_sq_ring *ring = ctx->sq_ring;
1940 unsigned head;
1941
1942 /*
1943 * The cached sq head (or cq tail) serves two purposes:
1944 *
1945 * 1) allows us to batch the cost of updating the user visible
1946 * head updates.
1947 * 2) allows the kernel side to track the head on its own, even
1948 * though the application is the one updating it.
1949 */
1950 head = ctx->cached_sq_head;
Stefan Bühlere523a292019-04-19 11:57:44 +02001951 /* make sure SQ entry isn't read before tail */
1952 if (head == smp_load_acquire(&ring->r.tail))
Jens Axboe2b188cc2019-01-07 10:46:33 -07001953 return false;
1954
1955 head = READ_ONCE(ring->array[head & ctx->sq_mask]);
1956 if (head < ctx->sq_entries) {
1957 s->index = head;
1958 s->sqe = &ctx->sq_sqes[head];
1959 ctx->cached_sq_head++;
1960 return true;
1961 }
1962
1963 /* drop invalid entries */
1964 ctx->cached_sq_head++;
1965 ring->dropped++;
Jens Axboe2b188cc2019-01-07 10:46:33 -07001966 return false;
1967}
1968
Jens Axboe6c271ce2019-01-10 11:22:30 -07001969static int io_submit_sqes(struct io_ring_ctx *ctx, struct sqe_submit *sqes,
1970 unsigned int nr, bool has_user, bool mm_fault)
1971{
1972 struct io_submit_state state, *statep = NULL;
1973 int ret, i, submitted = 0;
1974
1975 if (nr > IO_PLUG_THRESHOLD) {
1976 io_submit_state_start(&state, ctx, nr);
1977 statep = &state;
1978 }
1979
1980 for (i = 0; i < nr; i++) {
1981 if (unlikely(mm_fault)) {
1982 ret = -EFAULT;
1983 } else {
1984 sqes[i].has_user = has_user;
1985 sqes[i].needs_lock = true;
1986 sqes[i].needs_fixed_file = true;
1987 ret = io_submit_sqe(ctx, &sqes[i], statep);
1988 }
1989 if (!ret) {
1990 submitted++;
1991 continue;
1992 }
1993
Jens Axboec71ffb62019-05-13 20:58:29 -06001994 io_cqring_add_event(ctx, sqes[i].sqe->user_data, ret);
Jens Axboe6c271ce2019-01-10 11:22:30 -07001995 }
1996
1997 if (statep)
1998 io_submit_state_end(&state);
1999
2000 return submitted;
2001}
2002
2003static int io_sq_thread(void *data)
2004{
2005 struct sqe_submit sqes[IO_IOPOLL_BATCH];
2006 struct io_ring_ctx *ctx = data;
2007 struct mm_struct *cur_mm = NULL;
2008 mm_segment_t old_fs;
2009 DEFINE_WAIT(wait);
2010 unsigned inflight;
2011 unsigned long timeout;
2012
2013 old_fs = get_fs();
2014 set_fs(USER_DS);
2015
2016 timeout = inflight = 0;
Roman Penyaev2bbcd6d2019-05-16 10:53:57 +02002017 while (!kthread_should_park()) {
Jens Axboe6c271ce2019-01-10 11:22:30 -07002018 bool all_fixed, mm_fault = false;
2019 int i;
2020
2021 if (inflight) {
2022 unsigned nr_events = 0;
2023
2024 if (ctx->flags & IORING_SETUP_IOPOLL) {
2025 /*
2026 * We disallow the app entering submit/complete
2027 * with polling, but we still need to lock the
2028 * ring to prevent racing with polled issue
2029 * that got punted to a workqueue.
2030 */
2031 mutex_lock(&ctx->uring_lock);
2032 io_iopoll_check(ctx, &nr_events, 0);
2033 mutex_unlock(&ctx->uring_lock);
2034 } else {
2035 /*
2036 * Normal IO, just pretend everything completed.
2037 * We don't have to poll completions for that.
2038 */
2039 nr_events = inflight;
2040 }
2041
2042 inflight -= nr_events;
2043 if (!inflight)
2044 timeout = jiffies + ctx->sq_thread_idle;
2045 }
2046
2047 if (!io_get_sqring(ctx, &sqes[0])) {
2048 /*
2049 * We're polling. If we're within the defined idle
2050 * period, then let us spin without work before going
2051 * to sleep.
2052 */
2053 if (inflight || !time_after(jiffies, timeout)) {
2054 cpu_relax();
2055 continue;
2056 }
2057
2058 /*
2059 * Drop cur_mm before scheduling, we can't hold it for
2060 * long periods (or over schedule()). Do this before
2061 * adding ourselves to the waitqueue, as the unuse/drop
2062 * may sleep.
2063 */
2064 if (cur_mm) {
2065 unuse_mm(cur_mm);
2066 mmput(cur_mm);
2067 cur_mm = NULL;
2068 }
2069
2070 prepare_to_wait(&ctx->sqo_wait, &wait,
2071 TASK_INTERRUPTIBLE);
2072
2073 /* Tell userspace we may need a wakeup call */
2074 ctx->sq_ring->flags |= IORING_SQ_NEED_WAKEUP;
Stefan Bühler0d7bae62019-04-19 11:57:45 +02002075 /* make sure to read SQ tail after writing flags */
2076 smp_mb();
Jens Axboe6c271ce2019-01-10 11:22:30 -07002077
2078 if (!io_get_sqring(ctx, &sqes[0])) {
Roman Penyaev2bbcd6d2019-05-16 10:53:57 +02002079 if (kthread_should_park()) {
Jens Axboe6c271ce2019-01-10 11:22:30 -07002080 finish_wait(&ctx->sqo_wait, &wait);
2081 break;
2082 }
2083 if (signal_pending(current))
2084 flush_signals(current);
2085 schedule();
2086 finish_wait(&ctx->sqo_wait, &wait);
2087
2088 ctx->sq_ring->flags &= ~IORING_SQ_NEED_WAKEUP;
Jens Axboe6c271ce2019-01-10 11:22:30 -07002089 continue;
2090 }
2091 finish_wait(&ctx->sqo_wait, &wait);
2092
2093 ctx->sq_ring->flags &= ~IORING_SQ_NEED_WAKEUP;
Jens Axboe6c271ce2019-01-10 11:22:30 -07002094 }
2095
2096 i = 0;
2097 all_fixed = true;
2098 do {
2099 if (all_fixed && io_sqe_needs_user(sqes[i].sqe))
2100 all_fixed = false;
2101
2102 i++;
2103 if (i == ARRAY_SIZE(sqes))
2104 break;
2105 } while (io_get_sqring(ctx, &sqes[i]));
2106
2107 /* Unless all new commands are FIXED regions, grab mm */
2108 if (!all_fixed && !cur_mm) {
2109 mm_fault = !mmget_not_zero(ctx->sqo_mm);
2110 if (!mm_fault) {
2111 use_mm(ctx->sqo_mm);
2112 cur_mm = ctx->sqo_mm;
2113 }
2114 }
2115
2116 inflight += io_submit_sqes(ctx, sqes, i, cur_mm != NULL,
2117 mm_fault);
2118
2119 /* Commit SQ ring head once we've consumed all SQEs */
2120 io_commit_sqring(ctx);
2121 }
2122
2123 set_fs(old_fs);
2124 if (cur_mm) {
2125 unuse_mm(cur_mm);
2126 mmput(cur_mm);
2127 }
Jens Axboe06058632019-04-13 09:26:03 -06002128
Roman Penyaev2bbcd6d2019-05-16 10:53:57 +02002129 kthread_parkme();
Jens Axboe06058632019-04-13 09:26:03 -06002130
Jens Axboe6c271ce2019-01-10 11:22:30 -07002131 return 0;
2132}
2133
Jens Axboe2b188cc2019-01-07 10:46:33 -07002134static int io_ring_submit(struct io_ring_ctx *ctx, unsigned int to_submit)
2135{
Jens Axboe9a56a232019-01-09 09:06:50 -07002136 struct io_submit_state state, *statep = NULL;
Jens Axboe5c8b0b52019-04-30 10:16:07 -06002137 int i, submit = 0;
Jens Axboe2b188cc2019-01-07 10:46:33 -07002138
Jens Axboe9a56a232019-01-09 09:06:50 -07002139 if (to_submit > IO_PLUG_THRESHOLD) {
2140 io_submit_state_start(&state, ctx, to_submit);
2141 statep = &state;
2142 }
Jens Axboe2b188cc2019-01-07 10:46:33 -07002143
2144 for (i = 0; i < to_submit; i++) {
2145 struct sqe_submit s;
Jens Axboe5c8b0b52019-04-30 10:16:07 -06002146 int ret;
Jens Axboe2b188cc2019-01-07 10:46:33 -07002147
2148 if (!io_get_sqring(ctx, &s))
2149 break;
2150
2151 s.has_user = true;
Jens Axboedef596e2019-01-09 08:59:42 -07002152 s.needs_lock = false;
Jens Axboe6c271ce2019-01-10 11:22:30 -07002153 s.needs_fixed_file = false;
Jens Axboe5c8b0b52019-04-30 10:16:07 -06002154 submit++;
Jens Axboedef596e2019-01-09 08:59:42 -07002155
Jens Axboe9a56a232019-01-09 09:06:50 -07002156 ret = io_submit_sqe(ctx, &s, statep);
Jens Axboe5c8b0b52019-04-30 10:16:07 -06002157 if (ret)
Jens Axboec71ffb62019-05-13 20:58:29 -06002158 io_cqring_add_event(ctx, s.sqe->user_data, ret);
Jens Axboe2b188cc2019-01-07 10:46:33 -07002159 }
2160 io_commit_sqring(ctx);
2161
Jens Axboe9a56a232019-01-09 09:06:50 -07002162 if (statep)
2163 io_submit_state_end(statep);
Jens Axboe2b188cc2019-01-07 10:46:33 -07002164
Jens Axboe5c8b0b52019-04-30 10:16:07 -06002165 return submit;
Jens Axboe2b188cc2019-01-07 10:46:33 -07002166}
2167
2168static unsigned io_cqring_events(struct io_cq_ring *ring)
2169{
Jackie Liudc6ce4b2019-05-16 11:46:30 +08002170 /* See comment at the top of this file */
2171 smp_rmb();
Jens Axboe2b188cc2019-01-07 10:46:33 -07002172 return READ_ONCE(ring->r.tail) - READ_ONCE(ring->r.head);
2173}
2174
2175/*
2176 * Wait until events become available, if we don't already have some. The
2177 * application must reap them itself, as they reside on the shared cq ring.
2178 */
2179static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events,
2180 const sigset_t __user *sig, size_t sigsz)
2181{
2182 struct io_cq_ring *ring = ctx->cq_ring;
2183 sigset_t ksigmask, sigsaved;
Jens Axboe2b188cc2019-01-07 10:46:33 -07002184 int ret;
2185
Jens Axboe2b188cc2019-01-07 10:46:33 -07002186 if (io_cqring_events(ring) >= min_events)
2187 return 0;
2188
2189 if (sig) {
Arnd Bergmann9e75ad52019-03-25 15:34:53 +01002190#ifdef CONFIG_COMPAT
2191 if (in_compat_syscall())
2192 ret = set_compat_user_sigmask((const compat_sigset_t __user *)sig,
2193 &ksigmask, &sigsaved, sigsz);
2194 else
2195#endif
2196 ret = set_user_sigmask(sig, &ksigmask,
2197 &sigsaved, sigsz);
2198
Jens Axboe2b188cc2019-01-07 10:46:33 -07002199 if (ret)
2200 return ret;
2201 }
2202
Jackie Liufdb288a2019-05-16 11:46:31 +08002203 ret = wait_event_interruptible(ctx->wait, io_cqring_events(ring) >= min_events);
2204 if (ret == -ERESTARTSYS)
Jens Axboe2b188cc2019-01-07 10:46:33 -07002205 ret = -EINTR;
Jens Axboe2b188cc2019-01-07 10:46:33 -07002206
2207 if (sig)
2208 restore_user_sigmask(sig, &sigsaved);
2209
2210 return READ_ONCE(ring->r.head) == READ_ONCE(ring->r.tail) ? ret : 0;
2211}
2212
Jens Axboe6b063142019-01-10 22:13:58 -07002213static void __io_sqe_files_unregister(struct io_ring_ctx *ctx)
2214{
2215#if defined(CONFIG_UNIX)
2216 if (ctx->ring_sock) {
2217 struct sock *sock = ctx->ring_sock->sk;
2218 struct sk_buff *skb;
2219
2220 while ((skb = skb_dequeue(&sock->sk_receive_queue)) != NULL)
2221 kfree_skb(skb);
2222 }
2223#else
2224 int i;
2225
2226 for (i = 0; i < ctx->nr_user_files; i++)
2227 fput(ctx->user_files[i]);
2228#endif
2229}
2230
2231static int io_sqe_files_unregister(struct io_ring_ctx *ctx)
2232{
2233 if (!ctx->user_files)
2234 return -ENXIO;
2235
2236 __io_sqe_files_unregister(ctx);
2237 kfree(ctx->user_files);
2238 ctx->user_files = NULL;
2239 ctx->nr_user_files = 0;
2240 return 0;
2241}
2242
Jens Axboe6c271ce2019-01-10 11:22:30 -07002243static void io_sq_thread_stop(struct io_ring_ctx *ctx)
2244{
2245 if (ctx->sqo_thread) {
Roman Penyaev2bbcd6d2019-05-16 10:53:57 +02002246 /*
2247 * The park is a bit of a work-around, without it we get
2248 * warning spews on shutdown with SQPOLL set and affinity
2249 * set to a single CPU.
2250 */
Jens Axboe06058632019-04-13 09:26:03 -06002251 kthread_park(ctx->sqo_thread);
Jens Axboe6c271ce2019-01-10 11:22:30 -07002252 kthread_stop(ctx->sqo_thread);
2253 ctx->sqo_thread = NULL;
2254 }
2255}
2256
Jens Axboe6b063142019-01-10 22:13:58 -07002257static void io_finish_async(struct io_ring_ctx *ctx)
2258{
Jens Axboe6c271ce2019-01-10 11:22:30 -07002259 io_sq_thread_stop(ctx);
2260
Jens Axboe6b063142019-01-10 22:13:58 -07002261 if (ctx->sqo_wq) {
2262 destroy_workqueue(ctx->sqo_wq);
2263 ctx->sqo_wq = NULL;
2264 }
2265}
2266
2267#if defined(CONFIG_UNIX)
2268static void io_destruct_skb(struct sk_buff *skb)
2269{
2270 struct io_ring_ctx *ctx = skb->sk->sk_user_data;
2271
2272 io_finish_async(ctx);
2273 unix_destruct_scm(skb);
2274}
2275
2276/*
2277 * Ensure the UNIX gc is aware of our file set, so we are certain that
2278 * the io_uring can be safely unregistered on process exit, even if we have
2279 * loops in the file referencing.
2280 */
2281static int __io_sqe_files_scm(struct io_ring_ctx *ctx, int nr, int offset)
2282{
2283 struct sock *sk = ctx->ring_sock->sk;
2284 struct scm_fp_list *fpl;
2285 struct sk_buff *skb;
2286 int i;
2287
2288 if (!capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN)) {
2289 unsigned long inflight = ctx->user->unix_inflight + nr;
2290
2291 if (inflight > task_rlimit(current, RLIMIT_NOFILE))
2292 return -EMFILE;
2293 }
2294
2295 fpl = kzalloc(sizeof(*fpl), GFP_KERNEL);
2296 if (!fpl)
2297 return -ENOMEM;
2298
2299 skb = alloc_skb(0, GFP_KERNEL);
2300 if (!skb) {
2301 kfree(fpl);
2302 return -ENOMEM;
2303 }
2304
2305 skb->sk = sk;
2306 skb->destructor = io_destruct_skb;
2307
2308 fpl->user = get_uid(ctx->user);
2309 for (i = 0; i < nr; i++) {
2310 fpl->fp[i] = get_file(ctx->user_files[i + offset]);
2311 unix_inflight(fpl->user, fpl->fp[i]);
2312 }
2313
2314 fpl->max = fpl->count = nr;
2315 UNIXCB(skb).fp = fpl;
2316 refcount_add(skb->truesize, &sk->sk_wmem_alloc);
2317 skb_queue_head(&sk->sk_receive_queue, skb);
2318
2319 for (i = 0; i < nr; i++)
2320 fput(fpl->fp[i]);
2321
2322 return 0;
2323}
2324
2325/*
2326 * If UNIX sockets are enabled, fd passing can cause a reference cycle which
2327 * causes regular reference counting to break down. We rely on the UNIX
2328 * garbage collection to take care of this problem for us.
2329 */
2330static int io_sqe_files_scm(struct io_ring_ctx *ctx)
2331{
2332 unsigned left, total;
2333 int ret = 0;
2334
2335 total = 0;
2336 left = ctx->nr_user_files;
2337 while (left) {
2338 unsigned this_files = min_t(unsigned, left, SCM_MAX_FD);
Jens Axboe6b063142019-01-10 22:13:58 -07002339
2340 ret = __io_sqe_files_scm(ctx, this_files, total);
2341 if (ret)
2342 break;
2343 left -= this_files;
2344 total += this_files;
2345 }
2346
2347 if (!ret)
2348 return 0;
2349
2350 while (total < ctx->nr_user_files) {
2351 fput(ctx->user_files[total]);
2352 total++;
2353 }
2354
2355 return ret;
2356}
2357#else
2358static int io_sqe_files_scm(struct io_ring_ctx *ctx)
2359{
2360 return 0;
2361}
2362#endif
2363
2364static int io_sqe_files_register(struct io_ring_ctx *ctx, void __user *arg,
2365 unsigned nr_args)
2366{
2367 __s32 __user *fds = (__s32 __user *) arg;
2368 int fd, ret = 0;
2369 unsigned i;
2370
2371 if (ctx->user_files)
2372 return -EBUSY;
2373 if (!nr_args)
2374 return -EINVAL;
2375 if (nr_args > IORING_MAX_FIXED_FILES)
2376 return -EMFILE;
2377
2378 ctx->user_files = kcalloc(nr_args, sizeof(struct file *), GFP_KERNEL);
2379 if (!ctx->user_files)
2380 return -ENOMEM;
2381
2382 for (i = 0; i < nr_args; i++) {
2383 ret = -EFAULT;
2384 if (copy_from_user(&fd, &fds[i], sizeof(fd)))
2385 break;
2386
2387 ctx->user_files[i] = fget(fd);
2388
2389 ret = -EBADF;
2390 if (!ctx->user_files[i])
2391 break;
2392 /*
2393 * Don't allow io_uring instances to be registered. If UNIX
2394 * isn't enabled, then this causes a reference cycle and this
2395 * instance can never get freed. If UNIX is enabled we'll
2396 * handle it just fine, but there's still no point in allowing
2397 * a ring fd as it doesn't support regular read/write anyway.
2398 */
2399 if (ctx->user_files[i]->f_op == &io_uring_fops) {
2400 fput(ctx->user_files[i]);
2401 break;
2402 }
2403 ctx->nr_user_files++;
2404 ret = 0;
2405 }
2406
2407 if (ret) {
2408 for (i = 0; i < ctx->nr_user_files; i++)
2409 fput(ctx->user_files[i]);
2410
2411 kfree(ctx->user_files);
Jens Axboe25adf502019-04-03 09:52:40 -06002412 ctx->user_files = NULL;
Jens Axboe6b063142019-01-10 22:13:58 -07002413 ctx->nr_user_files = 0;
2414 return ret;
2415 }
2416
2417 ret = io_sqe_files_scm(ctx);
2418 if (ret)
2419 io_sqe_files_unregister(ctx);
2420
2421 return ret;
2422}
2423
Jens Axboe6c271ce2019-01-10 11:22:30 -07002424static int io_sq_offload_start(struct io_ring_ctx *ctx,
2425 struct io_uring_params *p)
Jens Axboe2b188cc2019-01-07 10:46:33 -07002426{
2427 int ret;
2428
Jens Axboe6c271ce2019-01-10 11:22:30 -07002429 init_waitqueue_head(&ctx->sqo_wait);
Jens Axboe2b188cc2019-01-07 10:46:33 -07002430 mmgrab(current->mm);
2431 ctx->sqo_mm = current->mm;
2432
Jens Axboe6c271ce2019-01-10 11:22:30 -07002433 if (ctx->flags & IORING_SETUP_SQPOLL) {
Jens Axboe3ec482d2019-04-08 10:51:01 -06002434 ret = -EPERM;
2435 if (!capable(CAP_SYS_ADMIN))
2436 goto err;
2437
Jens Axboe917257d2019-04-13 09:28:55 -06002438 ctx->sq_thread_idle = msecs_to_jiffies(p->sq_thread_idle);
2439 if (!ctx->sq_thread_idle)
2440 ctx->sq_thread_idle = HZ;
2441
Jens Axboe6c271ce2019-01-10 11:22:30 -07002442 if (p->flags & IORING_SETUP_SQ_AFF) {
Jens Axboe44a9bd12019-05-14 20:00:30 -06002443 int cpu = p->sq_thread_cpu;
Jens Axboe6c271ce2019-01-10 11:22:30 -07002444
Jens Axboe917257d2019-04-13 09:28:55 -06002445 ret = -EINVAL;
Jens Axboe44a9bd12019-05-14 20:00:30 -06002446 if (cpu >= nr_cpu_ids)
2447 goto err;
Shenghui Wang7889f442019-05-07 16:03:19 +08002448 if (!cpu_online(cpu))
Jens Axboe917257d2019-04-13 09:28:55 -06002449 goto err;
2450
Jens Axboe6c271ce2019-01-10 11:22:30 -07002451 ctx->sqo_thread = kthread_create_on_cpu(io_sq_thread,
2452 ctx, cpu,
2453 "io_uring-sq");
2454 } else {
2455 ctx->sqo_thread = kthread_create(io_sq_thread, ctx,
2456 "io_uring-sq");
2457 }
2458 if (IS_ERR(ctx->sqo_thread)) {
2459 ret = PTR_ERR(ctx->sqo_thread);
2460 ctx->sqo_thread = NULL;
2461 goto err;
2462 }
2463 wake_up_process(ctx->sqo_thread);
2464 } else if (p->flags & IORING_SETUP_SQ_AFF) {
2465 /* Can't have SQ_AFF without SQPOLL */
2466 ret = -EINVAL;
2467 goto err;
2468 }
2469
Jens Axboe2b188cc2019-01-07 10:46:33 -07002470 /* Do QD, or 2 * CPUS, whatever is smallest */
2471 ctx->sqo_wq = alloc_workqueue("io_ring-wq", WQ_UNBOUND | WQ_FREEZABLE,
2472 min(ctx->sq_entries - 1, 2 * num_online_cpus()));
2473 if (!ctx->sqo_wq) {
2474 ret = -ENOMEM;
2475 goto err;
2476 }
2477
2478 return 0;
2479err:
Jens Axboe6c271ce2019-01-10 11:22:30 -07002480 io_sq_thread_stop(ctx);
Jens Axboe2b188cc2019-01-07 10:46:33 -07002481 mmdrop(ctx->sqo_mm);
2482 ctx->sqo_mm = NULL;
2483 return ret;
2484}
2485
2486static void io_unaccount_mem(struct user_struct *user, unsigned long nr_pages)
2487{
2488 atomic_long_sub(nr_pages, &user->locked_vm);
2489}
2490
2491static int io_account_mem(struct user_struct *user, unsigned long nr_pages)
2492{
2493 unsigned long page_limit, cur_pages, new_pages;
2494
2495 /* Don't allow more pages than we can safely lock */
2496 page_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT;
2497
2498 do {
2499 cur_pages = atomic_long_read(&user->locked_vm);
2500 new_pages = cur_pages + nr_pages;
2501 if (new_pages > page_limit)
2502 return -ENOMEM;
2503 } while (atomic_long_cmpxchg(&user->locked_vm, cur_pages,
2504 new_pages) != cur_pages);
2505
2506 return 0;
2507}
2508
2509static void io_mem_free(void *ptr)
2510{
Mark Rutland52e04ef2019-04-30 17:30:21 +01002511 struct page *page;
Jens Axboe2b188cc2019-01-07 10:46:33 -07002512
Mark Rutland52e04ef2019-04-30 17:30:21 +01002513 if (!ptr)
2514 return;
2515
2516 page = virt_to_head_page(ptr);
Jens Axboe2b188cc2019-01-07 10:46:33 -07002517 if (put_page_testzero(page))
2518 free_compound_page(page);
2519}
2520
2521static void *io_mem_alloc(size_t size)
2522{
2523 gfp_t gfp_flags = GFP_KERNEL | __GFP_ZERO | __GFP_NOWARN | __GFP_COMP |
2524 __GFP_NORETRY;
2525
2526 return (void *) __get_free_pages(gfp_flags, get_order(size));
2527}
2528
2529static unsigned long ring_pages(unsigned sq_entries, unsigned cq_entries)
2530{
2531 struct io_sq_ring *sq_ring;
2532 struct io_cq_ring *cq_ring;
2533 size_t bytes;
2534
2535 bytes = struct_size(sq_ring, array, sq_entries);
2536 bytes += array_size(sizeof(struct io_uring_sqe), sq_entries);
2537 bytes += struct_size(cq_ring, cqes, cq_entries);
2538
2539 return (bytes + PAGE_SIZE - 1) / PAGE_SIZE;
2540}
2541
Jens Axboeedafcce2019-01-09 09:16:05 -07002542static int io_sqe_buffer_unregister(struct io_ring_ctx *ctx)
2543{
2544 int i, j;
2545
2546 if (!ctx->user_bufs)
2547 return -ENXIO;
2548
2549 for (i = 0; i < ctx->nr_user_bufs; i++) {
2550 struct io_mapped_ubuf *imu = &ctx->user_bufs[i];
2551
2552 for (j = 0; j < imu->nr_bvecs; j++)
2553 put_page(imu->bvec[j].bv_page);
2554
2555 if (ctx->account_mem)
2556 io_unaccount_mem(ctx->user, imu->nr_bvecs);
Mark Rutlandd4ef6472019-05-01 16:59:16 +01002557 kvfree(imu->bvec);
Jens Axboeedafcce2019-01-09 09:16:05 -07002558 imu->nr_bvecs = 0;
2559 }
2560
2561 kfree(ctx->user_bufs);
2562 ctx->user_bufs = NULL;
2563 ctx->nr_user_bufs = 0;
2564 return 0;
2565}
2566
2567static int io_copy_iov(struct io_ring_ctx *ctx, struct iovec *dst,
2568 void __user *arg, unsigned index)
2569{
2570 struct iovec __user *src;
2571
2572#ifdef CONFIG_COMPAT
2573 if (ctx->compat) {
2574 struct compat_iovec __user *ciovs;
2575 struct compat_iovec ciov;
2576
2577 ciovs = (struct compat_iovec __user *) arg;
2578 if (copy_from_user(&ciov, &ciovs[index], sizeof(ciov)))
2579 return -EFAULT;
2580
2581 dst->iov_base = (void __user *) (unsigned long) ciov.iov_base;
2582 dst->iov_len = ciov.iov_len;
2583 return 0;
2584 }
2585#endif
2586 src = (struct iovec __user *) arg;
2587 if (copy_from_user(dst, &src[index], sizeof(*dst)))
2588 return -EFAULT;
2589 return 0;
2590}
2591
2592static int io_sqe_buffer_register(struct io_ring_ctx *ctx, void __user *arg,
2593 unsigned nr_args)
2594{
2595 struct vm_area_struct **vmas = NULL;
2596 struct page **pages = NULL;
2597 int i, j, got_pages = 0;
2598 int ret = -EINVAL;
2599
2600 if (ctx->user_bufs)
2601 return -EBUSY;
2602 if (!nr_args || nr_args > UIO_MAXIOV)
2603 return -EINVAL;
2604
2605 ctx->user_bufs = kcalloc(nr_args, sizeof(struct io_mapped_ubuf),
2606 GFP_KERNEL);
2607 if (!ctx->user_bufs)
2608 return -ENOMEM;
2609
2610 for (i = 0; i < nr_args; i++) {
2611 struct io_mapped_ubuf *imu = &ctx->user_bufs[i];
2612 unsigned long off, start, end, ubuf;
2613 int pret, nr_pages;
2614 struct iovec iov;
2615 size_t size;
2616
2617 ret = io_copy_iov(ctx, &iov, arg, i);
2618 if (ret)
Pavel Begunkova2786822019-05-26 12:35:47 +03002619 goto err;
Jens Axboeedafcce2019-01-09 09:16:05 -07002620
2621 /*
2622 * Don't impose further limits on the size and buffer
2623 * constraints here, we'll -EINVAL later when IO is
2624 * submitted if they are wrong.
2625 */
2626 ret = -EFAULT;
2627 if (!iov.iov_base || !iov.iov_len)
2628 goto err;
2629
2630 /* arbitrary limit, but we need something */
2631 if (iov.iov_len > SZ_1G)
2632 goto err;
2633
2634 ubuf = (unsigned long) iov.iov_base;
2635 end = (ubuf + iov.iov_len + PAGE_SIZE - 1) >> PAGE_SHIFT;
2636 start = ubuf >> PAGE_SHIFT;
2637 nr_pages = end - start;
2638
2639 if (ctx->account_mem) {
2640 ret = io_account_mem(ctx->user, nr_pages);
2641 if (ret)
2642 goto err;
2643 }
2644
2645 ret = 0;
2646 if (!pages || nr_pages > got_pages) {
2647 kfree(vmas);
2648 kfree(pages);
Mark Rutlandd4ef6472019-05-01 16:59:16 +01002649 pages = kvmalloc_array(nr_pages, sizeof(struct page *),
Jens Axboeedafcce2019-01-09 09:16:05 -07002650 GFP_KERNEL);
Mark Rutlandd4ef6472019-05-01 16:59:16 +01002651 vmas = kvmalloc_array(nr_pages,
Jens Axboeedafcce2019-01-09 09:16:05 -07002652 sizeof(struct vm_area_struct *),
2653 GFP_KERNEL);
2654 if (!pages || !vmas) {
2655 ret = -ENOMEM;
2656 if (ctx->account_mem)
2657 io_unaccount_mem(ctx->user, nr_pages);
2658 goto err;
2659 }
2660 got_pages = nr_pages;
2661 }
2662
Mark Rutlandd4ef6472019-05-01 16:59:16 +01002663 imu->bvec = kvmalloc_array(nr_pages, sizeof(struct bio_vec),
Jens Axboeedafcce2019-01-09 09:16:05 -07002664 GFP_KERNEL);
2665 ret = -ENOMEM;
2666 if (!imu->bvec) {
2667 if (ctx->account_mem)
2668 io_unaccount_mem(ctx->user, nr_pages);
2669 goto err;
2670 }
2671
2672 ret = 0;
2673 down_read(&current->mm->mmap_sem);
Ira Weiny932f4a62019-05-13 17:17:03 -07002674 pret = get_user_pages(ubuf, nr_pages,
2675 FOLL_WRITE | FOLL_LONGTERM,
2676 pages, vmas);
Jens Axboeedafcce2019-01-09 09:16:05 -07002677 if (pret == nr_pages) {
2678 /* don't support file backed memory */
2679 for (j = 0; j < nr_pages; j++) {
2680 struct vm_area_struct *vma = vmas[j];
2681
2682 if (vma->vm_file &&
2683 !is_file_hugepages(vma->vm_file)) {
2684 ret = -EOPNOTSUPP;
2685 break;
2686 }
2687 }
2688 } else {
2689 ret = pret < 0 ? pret : -EFAULT;
2690 }
2691 up_read(&current->mm->mmap_sem);
2692 if (ret) {
2693 /*
2694 * if we did partial map, or found file backed vmas,
2695 * release any pages we did get
2696 */
2697 if (pret > 0) {
2698 for (j = 0; j < pret; j++)
2699 put_page(pages[j]);
2700 }
2701 if (ctx->account_mem)
2702 io_unaccount_mem(ctx->user, nr_pages);
Mark Rutlandd4ef6472019-05-01 16:59:16 +01002703 kvfree(imu->bvec);
Jens Axboeedafcce2019-01-09 09:16:05 -07002704 goto err;
2705 }
2706
2707 off = ubuf & ~PAGE_MASK;
2708 size = iov.iov_len;
2709 for (j = 0; j < nr_pages; j++) {
2710 size_t vec_len;
2711
2712 vec_len = min_t(size_t, size, PAGE_SIZE - off);
2713 imu->bvec[j].bv_page = pages[j];
2714 imu->bvec[j].bv_len = vec_len;
2715 imu->bvec[j].bv_offset = off;
2716 off = 0;
2717 size -= vec_len;
2718 }
2719 /* store original address for later verification */
2720 imu->ubuf = ubuf;
2721 imu->len = iov.iov_len;
2722 imu->nr_bvecs = nr_pages;
2723
2724 ctx->nr_user_bufs++;
2725 }
Mark Rutlandd4ef6472019-05-01 16:59:16 +01002726 kvfree(pages);
2727 kvfree(vmas);
Jens Axboeedafcce2019-01-09 09:16:05 -07002728 return 0;
2729err:
Mark Rutlandd4ef6472019-05-01 16:59:16 +01002730 kvfree(pages);
2731 kvfree(vmas);
Jens Axboeedafcce2019-01-09 09:16:05 -07002732 io_sqe_buffer_unregister(ctx);
2733 return ret;
2734}
2735
Jens Axboe9b402842019-04-11 11:45:41 -06002736static int io_eventfd_register(struct io_ring_ctx *ctx, void __user *arg)
2737{
2738 __s32 __user *fds = arg;
2739 int fd;
2740
2741 if (ctx->cq_ev_fd)
2742 return -EBUSY;
2743
2744 if (copy_from_user(&fd, fds, sizeof(*fds)))
2745 return -EFAULT;
2746
2747 ctx->cq_ev_fd = eventfd_ctx_fdget(fd);
2748 if (IS_ERR(ctx->cq_ev_fd)) {
2749 int ret = PTR_ERR(ctx->cq_ev_fd);
2750 ctx->cq_ev_fd = NULL;
2751 return ret;
2752 }
2753
2754 return 0;
2755}
2756
2757static int io_eventfd_unregister(struct io_ring_ctx *ctx)
2758{
2759 if (ctx->cq_ev_fd) {
2760 eventfd_ctx_put(ctx->cq_ev_fd);
2761 ctx->cq_ev_fd = NULL;
2762 return 0;
2763 }
2764
2765 return -ENXIO;
2766}
2767
Jens Axboe2b188cc2019-01-07 10:46:33 -07002768static void io_ring_ctx_free(struct io_ring_ctx *ctx)
2769{
Jens Axboe6b063142019-01-10 22:13:58 -07002770 io_finish_async(ctx);
Jens Axboe2b188cc2019-01-07 10:46:33 -07002771 if (ctx->sqo_mm)
2772 mmdrop(ctx->sqo_mm);
Jens Axboedef596e2019-01-09 08:59:42 -07002773
2774 io_iopoll_reap_events(ctx);
Jens Axboeedafcce2019-01-09 09:16:05 -07002775 io_sqe_buffer_unregister(ctx);
Jens Axboe6b063142019-01-10 22:13:58 -07002776 io_sqe_files_unregister(ctx);
Jens Axboe9b402842019-04-11 11:45:41 -06002777 io_eventfd_unregister(ctx);
Jens Axboedef596e2019-01-09 08:59:42 -07002778
Jens Axboe2b188cc2019-01-07 10:46:33 -07002779#if defined(CONFIG_UNIX)
Eric Biggers355e8d22019-06-12 14:58:43 -07002780 if (ctx->ring_sock) {
2781 ctx->ring_sock->file = NULL; /* so that iput() is called */
Jens Axboe2b188cc2019-01-07 10:46:33 -07002782 sock_release(ctx->ring_sock);
Eric Biggers355e8d22019-06-12 14:58:43 -07002783 }
Jens Axboe2b188cc2019-01-07 10:46:33 -07002784#endif
2785
2786 io_mem_free(ctx->sq_ring);
2787 io_mem_free(ctx->sq_sqes);
2788 io_mem_free(ctx->cq_ring);
2789
2790 percpu_ref_exit(&ctx->refs);
2791 if (ctx->account_mem)
2792 io_unaccount_mem(ctx->user,
2793 ring_pages(ctx->sq_entries, ctx->cq_entries));
2794 free_uid(ctx->user);
2795 kfree(ctx);
2796}
2797
2798static __poll_t io_uring_poll(struct file *file, poll_table *wait)
2799{
2800 struct io_ring_ctx *ctx = file->private_data;
2801 __poll_t mask = 0;
2802
2803 poll_wait(file, &ctx->cq_wait, wait);
Stefan Bühler4f7067c2019-04-24 23:54:17 +02002804 /*
2805 * synchronizes with barrier from wq_has_sleeper call in
2806 * io_commit_cqring
2807 */
Jens Axboe2b188cc2019-01-07 10:46:33 -07002808 smp_rmb();
Stefan Bühlerfb775fa2019-04-19 11:57:46 +02002809 if (READ_ONCE(ctx->sq_ring->r.tail) - ctx->cached_sq_head !=
2810 ctx->sq_ring->ring_entries)
Jens Axboe2b188cc2019-01-07 10:46:33 -07002811 mask |= EPOLLOUT | EPOLLWRNORM;
2812 if (READ_ONCE(ctx->cq_ring->r.head) != ctx->cached_cq_tail)
2813 mask |= EPOLLIN | EPOLLRDNORM;
2814
2815 return mask;
2816}
2817
2818static int io_uring_fasync(int fd, struct file *file, int on)
2819{
2820 struct io_ring_ctx *ctx = file->private_data;
2821
2822 return fasync_helper(fd, file, on, &ctx->cq_fasync);
2823}
2824
2825static void io_ring_ctx_wait_and_kill(struct io_ring_ctx *ctx)
2826{
2827 mutex_lock(&ctx->uring_lock);
2828 percpu_ref_kill(&ctx->refs);
2829 mutex_unlock(&ctx->uring_lock);
2830
Jens Axboe221c5eb2019-01-17 09:41:58 -07002831 io_poll_remove_all(ctx);
Jens Axboedef596e2019-01-09 08:59:42 -07002832 io_iopoll_reap_events(ctx);
Jens Axboe2b188cc2019-01-07 10:46:33 -07002833 wait_for_completion(&ctx->ctx_done);
2834 io_ring_ctx_free(ctx);
2835}
2836
2837static int io_uring_release(struct inode *inode, struct file *file)
2838{
2839 struct io_ring_ctx *ctx = file->private_data;
2840
2841 file->private_data = NULL;
2842 io_ring_ctx_wait_and_kill(ctx);
2843 return 0;
2844}
2845
2846static int io_uring_mmap(struct file *file, struct vm_area_struct *vma)
2847{
2848 loff_t offset = (loff_t) vma->vm_pgoff << PAGE_SHIFT;
2849 unsigned long sz = vma->vm_end - vma->vm_start;
2850 struct io_ring_ctx *ctx = file->private_data;
2851 unsigned long pfn;
2852 struct page *page;
2853 void *ptr;
2854
2855 switch (offset) {
2856 case IORING_OFF_SQ_RING:
2857 ptr = ctx->sq_ring;
2858 break;
2859 case IORING_OFF_SQES:
2860 ptr = ctx->sq_sqes;
2861 break;
2862 case IORING_OFF_CQ_RING:
2863 ptr = ctx->cq_ring;
2864 break;
2865 default:
2866 return -EINVAL;
2867 }
2868
2869 page = virt_to_head_page(ptr);
2870 if (sz > (PAGE_SIZE << compound_order(page)))
2871 return -EINVAL;
2872
2873 pfn = virt_to_phys(ptr) >> PAGE_SHIFT;
2874 return remap_pfn_range(vma, vma->vm_start, pfn, sz, vma->vm_page_prot);
2875}
2876
2877SYSCALL_DEFINE6(io_uring_enter, unsigned int, fd, u32, to_submit,
2878 u32, min_complete, u32, flags, const sigset_t __user *, sig,
2879 size_t, sigsz)
2880{
2881 struct io_ring_ctx *ctx;
2882 long ret = -EBADF;
2883 int submitted = 0;
2884 struct fd f;
2885
Jens Axboe6c271ce2019-01-10 11:22:30 -07002886 if (flags & ~(IORING_ENTER_GETEVENTS | IORING_ENTER_SQ_WAKEUP))
Jens Axboe2b188cc2019-01-07 10:46:33 -07002887 return -EINVAL;
2888
2889 f = fdget(fd);
2890 if (!f.file)
2891 return -EBADF;
2892
2893 ret = -EOPNOTSUPP;
2894 if (f.file->f_op != &io_uring_fops)
2895 goto out_fput;
2896
2897 ret = -ENXIO;
2898 ctx = f.file->private_data;
2899 if (!percpu_ref_tryget(&ctx->refs))
2900 goto out_fput;
2901
Jens Axboe6c271ce2019-01-10 11:22:30 -07002902 /*
2903 * For SQ polling, the thread will do all submissions and completions.
2904 * Just return the requested submit count, and wake the thread if
2905 * we were asked to.
2906 */
2907 if (ctx->flags & IORING_SETUP_SQPOLL) {
2908 if (flags & IORING_ENTER_SQ_WAKEUP)
2909 wake_up(&ctx->sqo_wait);
2910 submitted = to_submit;
2911 goto out_ctx;
2912 }
2913
Jens Axboe2b188cc2019-01-07 10:46:33 -07002914 ret = 0;
2915 if (to_submit) {
2916 to_submit = min(to_submit, ctx->sq_entries);
2917
2918 mutex_lock(&ctx->uring_lock);
2919 submitted = io_ring_submit(ctx, to_submit);
2920 mutex_unlock(&ctx->uring_lock);
Jens Axboe2b188cc2019-01-07 10:46:33 -07002921 }
2922 if (flags & IORING_ENTER_GETEVENTS) {
Jens Axboedef596e2019-01-09 08:59:42 -07002923 unsigned nr_events = 0;
2924
Jens Axboe2b188cc2019-01-07 10:46:33 -07002925 min_complete = min(min_complete, ctx->cq_entries);
2926
Jens Axboedef596e2019-01-09 08:59:42 -07002927 if (ctx->flags & IORING_SETUP_IOPOLL) {
2928 mutex_lock(&ctx->uring_lock);
2929 ret = io_iopoll_check(ctx, &nr_events, min_complete);
2930 mutex_unlock(&ctx->uring_lock);
2931 } else {
2932 ret = io_cqring_wait(ctx, min_complete, sig, sigsz);
2933 }
Jens Axboe2b188cc2019-01-07 10:46:33 -07002934 }
2935
2936out_ctx:
2937 io_ring_drop_ctx_refs(ctx, 1);
2938out_fput:
2939 fdput(f);
2940 return submitted ? submitted : ret;
2941}
2942
2943static const struct file_operations io_uring_fops = {
2944 .release = io_uring_release,
2945 .mmap = io_uring_mmap,
2946 .poll = io_uring_poll,
2947 .fasync = io_uring_fasync,
2948};
2949
2950static int io_allocate_scq_urings(struct io_ring_ctx *ctx,
2951 struct io_uring_params *p)
2952{
2953 struct io_sq_ring *sq_ring;
2954 struct io_cq_ring *cq_ring;
2955 size_t size;
2956
2957 sq_ring = io_mem_alloc(struct_size(sq_ring, array, p->sq_entries));
2958 if (!sq_ring)
2959 return -ENOMEM;
2960
2961 ctx->sq_ring = sq_ring;
2962 sq_ring->ring_mask = p->sq_entries - 1;
2963 sq_ring->ring_entries = p->sq_entries;
2964 ctx->sq_mask = sq_ring->ring_mask;
2965 ctx->sq_entries = sq_ring->ring_entries;
2966
2967 size = array_size(sizeof(struct io_uring_sqe), p->sq_entries);
2968 if (size == SIZE_MAX)
2969 return -EOVERFLOW;
2970
2971 ctx->sq_sqes = io_mem_alloc(size);
Mark Rutland52e04ef2019-04-30 17:30:21 +01002972 if (!ctx->sq_sqes)
Jens Axboe2b188cc2019-01-07 10:46:33 -07002973 return -ENOMEM;
Jens Axboe2b188cc2019-01-07 10:46:33 -07002974
2975 cq_ring = io_mem_alloc(struct_size(cq_ring, cqes, p->cq_entries));
Mark Rutland52e04ef2019-04-30 17:30:21 +01002976 if (!cq_ring)
Jens Axboe2b188cc2019-01-07 10:46:33 -07002977 return -ENOMEM;
Jens Axboe2b188cc2019-01-07 10:46:33 -07002978
2979 ctx->cq_ring = cq_ring;
2980 cq_ring->ring_mask = p->cq_entries - 1;
2981 cq_ring->ring_entries = p->cq_entries;
2982 ctx->cq_mask = cq_ring->ring_mask;
2983 ctx->cq_entries = cq_ring->ring_entries;
2984 return 0;
2985}
2986
2987/*
2988 * Allocate an anonymous fd, this is what constitutes the application
2989 * visible backing of an io_uring instance. The application mmaps this
2990 * fd to gain access to the SQ/CQ ring details. If UNIX sockets are enabled,
2991 * we have to tie this fd to a socket for file garbage collection purposes.
2992 */
2993static int io_uring_get_fd(struct io_ring_ctx *ctx)
2994{
2995 struct file *file;
2996 int ret;
2997
2998#if defined(CONFIG_UNIX)
2999 ret = sock_create_kern(&init_net, PF_UNIX, SOCK_RAW, IPPROTO_IP,
3000 &ctx->ring_sock);
3001 if (ret)
3002 return ret;
3003#endif
3004
3005 ret = get_unused_fd_flags(O_RDWR | O_CLOEXEC);
3006 if (ret < 0)
3007 goto err;
3008
3009 file = anon_inode_getfile("[io_uring]", &io_uring_fops, ctx,
3010 O_RDWR | O_CLOEXEC);
3011 if (IS_ERR(file)) {
3012 put_unused_fd(ret);
3013 ret = PTR_ERR(file);
3014 goto err;
3015 }
3016
3017#if defined(CONFIG_UNIX)
3018 ctx->ring_sock->file = file;
Jens Axboe6b063142019-01-10 22:13:58 -07003019 ctx->ring_sock->sk->sk_user_data = ctx;
Jens Axboe2b188cc2019-01-07 10:46:33 -07003020#endif
3021 fd_install(ret, file);
3022 return ret;
3023err:
3024#if defined(CONFIG_UNIX)
3025 sock_release(ctx->ring_sock);
3026 ctx->ring_sock = NULL;
3027#endif
3028 return ret;
3029}
3030
3031static int io_uring_create(unsigned entries, struct io_uring_params *p)
3032{
3033 struct user_struct *user = NULL;
3034 struct io_ring_ctx *ctx;
3035 bool account_mem;
3036 int ret;
3037
3038 if (!entries || entries > IORING_MAX_ENTRIES)
3039 return -EINVAL;
3040
3041 /*
3042 * Use twice as many entries for the CQ ring. It's possible for the
3043 * application to drive a higher depth than the size of the SQ ring,
3044 * since the sqes are only used at submission time. This allows for
3045 * some flexibility in overcommitting a bit.
3046 */
3047 p->sq_entries = roundup_pow_of_two(entries);
3048 p->cq_entries = 2 * p->sq_entries;
3049
3050 user = get_uid(current_user());
3051 account_mem = !capable(CAP_IPC_LOCK);
3052
3053 if (account_mem) {
3054 ret = io_account_mem(user,
3055 ring_pages(p->sq_entries, p->cq_entries));
3056 if (ret) {
3057 free_uid(user);
3058 return ret;
3059 }
3060 }
3061
3062 ctx = io_ring_ctx_alloc(p);
3063 if (!ctx) {
3064 if (account_mem)
3065 io_unaccount_mem(user, ring_pages(p->sq_entries,
3066 p->cq_entries));
3067 free_uid(user);
3068 return -ENOMEM;
3069 }
3070 ctx->compat = in_compat_syscall();
3071 ctx->account_mem = account_mem;
3072 ctx->user = user;
3073
3074 ret = io_allocate_scq_urings(ctx, p);
3075 if (ret)
3076 goto err;
3077
Jens Axboe6c271ce2019-01-10 11:22:30 -07003078 ret = io_sq_offload_start(ctx, p);
Jens Axboe2b188cc2019-01-07 10:46:33 -07003079 if (ret)
3080 goto err;
3081
3082 ret = io_uring_get_fd(ctx);
3083 if (ret < 0)
3084 goto err;
3085
3086 memset(&p->sq_off, 0, sizeof(p->sq_off));
3087 p->sq_off.head = offsetof(struct io_sq_ring, r.head);
3088 p->sq_off.tail = offsetof(struct io_sq_ring, r.tail);
3089 p->sq_off.ring_mask = offsetof(struct io_sq_ring, ring_mask);
3090 p->sq_off.ring_entries = offsetof(struct io_sq_ring, ring_entries);
3091 p->sq_off.flags = offsetof(struct io_sq_ring, flags);
3092 p->sq_off.dropped = offsetof(struct io_sq_ring, dropped);
3093 p->sq_off.array = offsetof(struct io_sq_ring, array);
3094
3095 memset(&p->cq_off, 0, sizeof(p->cq_off));
3096 p->cq_off.head = offsetof(struct io_cq_ring, r.head);
3097 p->cq_off.tail = offsetof(struct io_cq_ring, r.tail);
3098 p->cq_off.ring_mask = offsetof(struct io_cq_ring, ring_mask);
3099 p->cq_off.ring_entries = offsetof(struct io_cq_ring, ring_entries);
3100 p->cq_off.overflow = offsetof(struct io_cq_ring, overflow);
3101 p->cq_off.cqes = offsetof(struct io_cq_ring, cqes);
3102 return ret;
3103err:
3104 io_ring_ctx_wait_and_kill(ctx);
3105 return ret;
3106}
3107
3108/*
3109 * Sets up an aio uring context, and returns the fd. Applications asks for a
3110 * ring size, we return the actual sq/cq ring sizes (among other things) in the
3111 * params structure passed in.
3112 */
3113static long io_uring_setup(u32 entries, struct io_uring_params __user *params)
3114{
3115 struct io_uring_params p;
3116 long ret;
3117 int i;
3118
3119 if (copy_from_user(&p, params, sizeof(p)))
3120 return -EFAULT;
3121 for (i = 0; i < ARRAY_SIZE(p.resv); i++) {
3122 if (p.resv[i])
3123 return -EINVAL;
3124 }
3125
Jens Axboe6c271ce2019-01-10 11:22:30 -07003126 if (p.flags & ~(IORING_SETUP_IOPOLL | IORING_SETUP_SQPOLL |
3127 IORING_SETUP_SQ_AFF))
Jens Axboe2b188cc2019-01-07 10:46:33 -07003128 return -EINVAL;
3129
3130 ret = io_uring_create(entries, &p);
3131 if (ret < 0)
3132 return ret;
3133
3134 if (copy_to_user(params, &p, sizeof(p)))
3135 return -EFAULT;
3136
3137 return ret;
3138}
3139
3140SYSCALL_DEFINE2(io_uring_setup, u32, entries,
3141 struct io_uring_params __user *, params)
3142{
3143 return io_uring_setup(entries, params);
3144}
3145
Jens Axboeedafcce2019-01-09 09:16:05 -07003146static int __io_uring_register(struct io_ring_ctx *ctx, unsigned opcode,
3147 void __user *arg, unsigned nr_args)
Jens Axboeb19062a2019-04-15 10:49:38 -06003148 __releases(ctx->uring_lock)
3149 __acquires(ctx->uring_lock)
Jens Axboeedafcce2019-01-09 09:16:05 -07003150{
3151 int ret;
3152
Jens Axboe35fa71a2019-04-22 10:23:23 -06003153 /*
3154 * We're inside the ring mutex, if the ref is already dying, then
3155 * someone else killed the ctx or is already going through
3156 * io_uring_register().
3157 */
3158 if (percpu_ref_is_dying(&ctx->refs))
3159 return -ENXIO;
3160
Jens Axboeedafcce2019-01-09 09:16:05 -07003161 percpu_ref_kill(&ctx->refs);
Jens Axboeb19062a2019-04-15 10:49:38 -06003162
3163 /*
3164 * Drop uring mutex before waiting for references to exit. If another
3165 * thread is currently inside io_uring_enter() it might need to grab
3166 * the uring_lock to make progress. If we hold it here across the drain
3167 * wait, then we can deadlock. It's safe to drop the mutex here, since
3168 * no new references will come in after we've killed the percpu ref.
3169 */
3170 mutex_unlock(&ctx->uring_lock);
Jens Axboeedafcce2019-01-09 09:16:05 -07003171 wait_for_completion(&ctx->ctx_done);
Jens Axboeb19062a2019-04-15 10:49:38 -06003172 mutex_lock(&ctx->uring_lock);
Jens Axboeedafcce2019-01-09 09:16:05 -07003173
3174 switch (opcode) {
3175 case IORING_REGISTER_BUFFERS:
3176 ret = io_sqe_buffer_register(ctx, arg, nr_args);
3177 break;
3178 case IORING_UNREGISTER_BUFFERS:
3179 ret = -EINVAL;
3180 if (arg || nr_args)
3181 break;
3182 ret = io_sqe_buffer_unregister(ctx);
3183 break;
Jens Axboe6b063142019-01-10 22:13:58 -07003184 case IORING_REGISTER_FILES:
3185 ret = io_sqe_files_register(ctx, arg, nr_args);
3186 break;
3187 case IORING_UNREGISTER_FILES:
3188 ret = -EINVAL;
3189 if (arg || nr_args)
3190 break;
3191 ret = io_sqe_files_unregister(ctx);
3192 break;
Jens Axboe9b402842019-04-11 11:45:41 -06003193 case IORING_REGISTER_EVENTFD:
3194 ret = -EINVAL;
3195 if (nr_args != 1)
3196 break;
3197 ret = io_eventfd_register(ctx, arg);
3198 break;
3199 case IORING_UNREGISTER_EVENTFD:
3200 ret = -EINVAL;
3201 if (arg || nr_args)
3202 break;
3203 ret = io_eventfd_unregister(ctx);
3204 break;
Jens Axboeedafcce2019-01-09 09:16:05 -07003205 default:
3206 ret = -EINVAL;
3207 break;
3208 }
3209
3210 /* bring the ctx back to life */
3211 reinit_completion(&ctx->ctx_done);
3212 percpu_ref_reinit(&ctx->refs);
3213 return ret;
3214}
3215
3216SYSCALL_DEFINE4(io_uring_register, unsigned int, fd, unsigned int, opcode,
3217 void __user *, arg, unsigned int, nr_args)
3218{
3219 struct io_ring_ctx *ctx;
3220 long ret = -EBADF;
3221 struct fd f;
3222
3223 f = fdget(fd);
3224 if (!f.file)
3225 return -EBADF;
3226
3227 ret = -EOPNOTSUPP;
3228 if (f.file->f_op != &io_uring_fops)
3229 goto out_fput;
3230
3231 ctx = f.file->private_data;
3232
3233 mutex_lock(&ctx->uring_lock);
3234 ret = __io_uring_register(ctx, opcode, arg, nr_args);
3235 mutex_unlock(&ctx->uring_lock);
3236out_fput:
3237 fdput(f);
3238 return ret;
3239}
3240
Jens Axboe2b188cc2019-01-07 10:46:33 -07003241static int __init io_uring_init(void)
3242{
3243 req_cachep = KMEM_CACHE(io_kiocb, SLAB_HWCACHE_ALIGN | SLAB_PANIC);
3244 return 0;
3245};
3246__initcall(io_uring_init);