1 // SPDX-License-Identifier: GPL-2.0
3 * Shared application/kernel submission and completion ring pairs, for
4 * supporting fast/efficient IO.
6 * A note on the read/write ordering memory barriers that are matched between
7 * the application and kernel side.
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
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
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
30 * Also see the examples in the liburing library:
32 * git://git.kernel.dk/liburing
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.
39 * Copyright (C) 2018-2019 Jens Axboe
40 * Copyright (c) 2018-2019 Christoph Hellwig
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>
50 #include <linux/sched/signal.h>
52 #include <linux/file.h>
53 #include <linux/fdtable.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>
60 #include <linux/kthread.h>
61 #include <linux/blkdev.h>
62 #include <linux/bvec.h>
63 #include <linux/net.h>
65 #include <net/af_unix.h>
67 #include <linux/anon_inodes.h>
68 #include <linux/sched/mm.h>
69 #include <linux/uaccess.h>
70 #include <linux/nospec.h>
71 #include <linux/sizes.h>
72 #include <linux/hugetlb.h>
74 #include <uapi/linux/io_uring.h>
78 #define IORING_MAX_ENTRIES 4096
79 #define IORING_MAX_FIXED_FILES 1024
82 u32 head ____cacheline_aligned_in_smp;
83 u32 tail ____cacheline_aligned_in_smp;
87 * This data is shared with the application through the mmap at offset
90 * The offsets to the member fields are published through struct
91 * io_sqring_offsets when calling io_uring_setup.
95 * Head and tail offsets into the ring; the offsets need to be
96 * masked to get valid indices.
98 * The kernel controls head and the application controls tail.
102 * Bitmask to apply to head and tail offsets (constant, equals
106 /* Ring size (constant, power of 2) */
109 * Number of invalid entries dropped by the kernel due to
110 * invalid index stored in array
112 * Written by the kernel, shouldn't be modified by the
113 * application (i.e. get number of "new events" by comparing to
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).
124 * Written by the kernel, shouldn't be modified by the
127 * The application needs a full memory barrier before checking
128 * for IORING_SQ_NEED_WAKEUP after updating the sq tail.
132 * Ring buffer of indices into array of io_uring_sqe, which is
133 * mmapped by the application using the IORING_OFF_SQES offset.
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.
139 * The kernel modifies neither the indices array nor the entries
146 * This data is shared with the application through the mmap at offset
147 * IORING_OFF_CQ_RING.
149 * The offsets to the member fields are published through struct
150 * io_cqring_offsets when calling io_uring_setup.
154 * Head and tail offsets into the ring; the offsets need to be
155 * masked to get valid indices.
157 * The application controls head and the kernel tail.
161 * Bitmask to apply to head and tail offsets (constant, equals
165 /* Ring size (constant, power of 2) */
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.
173 * Written by the kernel, shouldn't be modified by the
174 * application (i.e. get number of "new events" by comparing to
177 * As completion events come in out of order this counter is not
178 * ordered with any other data.
182 * Ring buffer of completion events.
184 * The kernel writes completion events fresh every time they are
185 * produced, so the application is allowed to modify pending
188 struct io_uring_cqe cqes[];
191 struct io_mapped_ubuf {
194 struct bio_vec *bvec;
195 unsigned int nr_bvecs;
201 struct list_head list;
210 struct percpu_ref refs;
211 } ____cacheline_aligned_in_smp;
219 struct io_sq_ring *sq_ring;
220 unsigned cached_sq_head;
223 unsigned sq_thread_idle;
224 struct io_uring_sqe *sq_sqes;
226 struct list_head defer_list;
227 } ____cacheline_aligned_in_smp;
230 struct workqueue_struct *sqo_wq;
231 struct task_struct *sqo_thread; /* if using sq thread polling */
232 struct mm_struct *sqo_mm;
233 wait_queue_head_t sqo_wait;
237 struct io_cq_ring *cq_ring;
238 unsigned cached_cq_tail;
241 struct wait_queue_head cq_wait;
242 struct fasync_struct *cq_fasync;
243 struct eventfd_ctx *cq_ev_fd;
244 } ____cacheline_aligned_in_smp;
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).
251 struct file **user_files;
252 unsigned nr_user_files;
254 /* if used, fixed mapped user buffers */
255 unsigned nr_user_bufs;
256 struct io_mapped_ubuf *user_bufs;
258 struct user_struct *user;
260 struct completion ctx_done;
263 struct mutex uring_lock;
264 wait_queue_head_t wait;
265 } ____cacheline_aligned_in_smp;
268 spinlock_t completion_lock;
269 bool poll_multi_file;
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.
276 struct list_head poll_list;
277 struct list_head cancel_list;
278 } ____cacheline_aligned_in_smp;
280 struct async_list pending_async[2];
282 #if defined(CONFIG_UNIX)
283 struct socket *ring_sock;
288 const struct io_uring_sqe *sqe;
289 unsigned short index;
292 bool needs_fixed_file;
296 * First field must be the file pointer in all the
297 * iocb unions! See also 'struct kiocb' in <linux/fs.h>
299 struct io_poll_iocb {
301 struct wait_queue_head *head;
305 struct wait_queue_entry wait;
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.
318 struct io_poll_iocb poll;
321 struct sqe_submit submit;
323 struct io_ring_ctx *ctx;
324 struct list_head list;
327 #define REQ_F_NOWAIT 1 /* must not punt to workers */
328 #define REQ_F_IOPOLL_COMPLETED 2 /* polled IO has completed */
329 #define REQ_F_FIXED_FILE 4 /* ctx owns file */
330 #define REQ_F_SEQ_PREV 8 /* sequential with previous */
331 #define REQ_F_IO_DRAIN 16 /* drain existing IO first */
332 #define REQ_F_IO_DRAINED 32 /* drain done */
334 u32 error; /* iopoll result from callback */
337 struct work_struct work;
340 #define IO_PLUG_THRESHOLD 2
341 #define IO_IOPOLL_BATCH 8
343 struct io_submit_state {
344 struct blk_plug plug;
347 * io_kiocb alloc cache
349 void *reqs[IO_IOPOLL_BATCH];
350 unsigned int free_reqs;
351 unsigned int cur_req;
354 * File reference cache
358 unsigned int has_refs;
359 unsigned int used_refs;
360 unsigned int ios_left;
363 static void io_sq_wq_submit_work(struct work_struct *work);
365 static struct kmem_cache *req_cachep;
367 static const struct file_operations io_uring_fops;
369 struct sock *io_uring_get_socket(struct file *file)
371 #if defined(CONFIG_UNIX)
372 if (file->f_op == &io_uring_fops) {
373 struct io_ring_ctx *ctx = file->private_data;
375 return ctx->ring_sock->sk;
380 EXPORT_SYMBOL(io_uring_get_socket);
382 static void io_ring_ctx_ref_free(struct percpu_ref *ref)
384 struct io_ring_ctx *ctx = container_of(ref, struct io_ring_ctx, refs);
386 complete(&ctx->ctx_done);
389 static struct io_ring_ctx *io_ring_ctx_alloc(struct io_uring_params *p)
391 struct io_ring_ctx *ctx;
394 ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
398 if (percpu_ref_init(&ctx->refs, io_ring_ctx_ref_free, 0, GFP_KERNEL)) {
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);
408 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);
413 spin_lock_init(&ctx->completion_lock);
414 INIT_LIST_HEAD(&ctx->poll_list);
415 INIT_LIST_HEAD(&ctx->cancel_list);
416 INIT_LIST_HEAD(&ctx->defer_list);
420 static inline bool io_sequence_defer(struct io_ring_ctx *ctx,
421 struct io_kiocb *req)
423 if ((req->flags & (REQ_F_IO_DRAIN|REQ_F_IO_DRAINED)) != REQ_F_IO_DRAIN)
426 return req->sequence > ctx->cached_cq_tail + ctx->sq_ring->dropped;
429 static struct io_kiocb *io_get_deferred_req(struct io_ring_ctx *ctx)
431 struct io_kiocb *req;
433 if (list_empty(&ctx->defer_list))
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);
445 static void __io_commit_cqring(struct io_ring_ctx *ctx)
447 struct io_cq_ring *ring = ctx->cq_ring;
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);
453 if (wq_has_sleeper(&ctx->cq_wait)) {
454 wake_up_interruptible(&ctx->cq_wait);
455 kill_fasync(&ctx->cq_fasync, SIGIO, POLL_IN);
460 static void io_commit_cqring(struct io_ring_ctx *ctx)
462 struct io_kiocb *req;
464 __io_commit_cqring(ctx);
466 while ((req = io_get_deferred_req(ctx)) != NULL) {
467 req->flags |= REQ_F_IO_DRAINED;
468 queue_work(ctx->sqo_wq, &req->work);
472 static struct io_uring_cqe *io_get_cqring(struct io_ring_ctx *ctx)
474 struct io_cq_ring *ring = ctx->cq_ring;
477 tail = ctx->cached_cq_tail;
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
483 if (tail - READ_ONCE(ring->r.head) == ring->ring_entries)
486 ctx->cached_cq_tail++;
487 return &ring->cqes[tail & ctx->cq_mask];
490 static void io_cqring_fill_event(struct io_ring_ctx *ctx, u64 ki_user_data,
493 struct io_uring_cqe *cqe;
496 * If we can't get a cq entry, userspace overflowed the
497 * submission (by quite a lot). Increment the overflow count in
500 cqe = io_get_cqring(ctx);
502 WRITE_ONCE(cqe->user_data, ki_user_data);
503 WRITE_ONCE(cqe->res, res);
504 WRITE_ONCE(cqe->flags, 0);
506 unsigned overflow = READ_ONCE(ctx->cq_ring->overflow);
508 WRITE_ONCE(ctx->cq_ring->overflow, overflow + 1);
512 static void io_cqring_ev_posted(struct io_ring_ctx *ctx)
514 if (waitqueue_active(&ctx->wait))
516 if (waitqueue_active(&ctx->sqo_wait))
517 wake_up(&ctx->sqo_wait);
519 eventfd_signal(ctx->cq_ev_fd, 1);
522 static void io_cqring_add_event(struct io_ring_ctx *ctx, u64 user_data,
527 spin_lock_irqsave(&ctx->completion_lock, flags);
528 io_cqring_fill_event(ctx, user_data, res);
529 io_commit_cqring(ctx);
530 spin_unlock_irqrestore(&ctx->completion_lock, flags);
532 io_cqring_ev_posted(ctx);
535 static void io_ring_drop_ctx_refs(struct io_ring_ctx *ctx, unsigned refs)
537 percpu_ref_put_many(&ctx->refs, refs);
539 if (waitqueue_active(&ctx->wait))
543 static struct io_kiocb *io_get_req(struct io_ring_ctx *ctx,
544 struct io_submit_state *state)
546 gfp_t gfp = GFP_KERNEL | __GFP_NOWARN;
547 struct io_kiocb *req;
549 if (!percpu_ref_tryget(&ctx->refs))
553 req = kmem_cache_alloc(req_cachep, gfp);
556 } else if (!state->free_reqs) {
560 sz = min_t(size_t, state->ios_left, ARRAY_SIZE(state->reqs));
561 ret = kmem_cache_alloc_bulk(req_cachep, gfp, sz, state->reqs);
564 * Bulk alloc is all-or-nothing. If we fail to get a batch,
565 * retry single alloc to be on the safe side.
567 if (unlikely(ret <= 0)) {
568 state->reqs[0] = kmem_cache_alloc(req_cachep, gfp);
573 state->free_reqs = ret - 1;
575 req = state->reqs[0];
577 req = state->reqs[state->cur_req];
584 /* one is dropped after submission, the other at completion */
585 refcount_set(&req->refs, 2);
588 io_ring_drop_ctx_refs(ctx, 1);
592 static void io_free_req_many(struct io_ring_ctx *ctx, void **reqs, int *nr)
595 kmem_cache_free_bulk(req_cachep, *nr, reqs);
596 io_ring_drop_ctx_refs(ctx, *nr);
601 static void io_free_req(struct io_kiocb *req)
603 if (req->file && !(req->flags & REQ_F_FIXED_FILE))
605 io_ring_drop_ctx_refs(req->ctx, 1);
606 kmem_cache_free(req_cachep, req);
609 static void io_put_req(struct io_kiocb *req)
611 if (refcount_dec_and_test(&req->refs))
616 * Find and free completed poll iocbs
618 static void io_iopoll_complete(struct io_ring_ctx *ctx, unsigned int *nr_events,
619 struct list_head *done)
621 void *reqs[IO_IOPOLL_BATCH];
622 struct io_kiocb *req;
626 while (!list_empty(done)) {
627 req = list_first_entry(done, struct io_kiocb, list);
628 list_del(&req->list);
630 io_cqring_fill_event(ctx, req->user_data, req->error);
633 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
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);
649 io_commit_cqring(ctx);
650 io_free_req_many(ctx, reqs, &to_free);
653 static int io_do_iopoll(struct io_ring_ctx *ctx, unsigned int *nr_events,
656 struct io_kiocb *req, *tmp;
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.
665 spin = !ctx->poll_multi_file && *nr_events < min;
668 list_for_each_entry_safe(req, tmp, &ctx->poll_list, list) {
669 struct kiocb *kiocb = &req->rw;
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.
676 if (req->flags & REQ_F_IOPOLL_COMPLETED) {
677 list_move_tail(&req->list, &done);
680 if (!list_empty(&done))
683 ret = kiocb->ki_filp->f_op->iopoll(kiocb, spin);
692 if (!list_empty(&done))
693 io_iopoll_complete(ctx, nr_events, &done);
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.
703 static int io_iopoll_getevents(struct io_ring_ctx *ctx, unsigned int *nr_events,
706 while (!list_empty(&ctx->poll_list)) {
709 ret = io_do_iopoll(ctx, nr_events, min);
712 if (!min || *nr_events >= min)
720 * We can't just wait for polled events to come to us, we have to actively
721 * find and complete them.
723 static void io_iopoll_reap_events(struct io_ring_ctx *ctx)
725 if (!(ctx->flags & IORING_SETUP_IOPOLL))
728 mutex_lock(&ctx->uring_lock);
729 while (!list_empty(&ctx->poll_list)) {
730 unsigned int nr_events = 0;
732 io_iopoll_getevents(ctx, &nr_events, 1);
734 mutex_unlock(&ctx->uring_lock);
737 static int io_iopoll_check(struct io_ring_ctx *ctx, unsigned *nr_events,
745 if (*nr_events < min)
746 tmin = min - *nr_events;
748 ret = io_iopoll_getevents(ctx, nr_events, tmin);
752 } while (min && !*nr_events && !need_resched());
757 static void kiocb_end_write(struct kiocb *kiocb)
759 if (kiocb->ki_flags & IOCB_WRITE) {
760 struct inode *inode = file_inode(kiocb->ki_filp);
763 * Tell lockdep we inherited freeze protection from submission
766 if (S_ISREG(inode->i_mode))
767 __sb_writers_acquired(inode->i_sb, SB_FREEZE_WRITE);
768 file_end_write(kiocb->ki_filp);
772 static void io_complete_rw(struct kiocb *kiocb, long res, long res2)
774 struct io_kiocb *req = container_of(kiocb, struct io_kiocb, rw);
776 kiocb_end_write(kiocb);
778 io_cqring_add_event(req->ctx, req->user_data, res);
782 static void io_complete_rw_iopoll(struct kiocb *kiocb, long res, long res2)
784 struct io_kiocb *req = container_of(kiocb, struct io_kiocb, rw);
786 kiocb_end_write(kiocb);
790 req->flags |= REQ_F_IOPOLL_COMPLETED;
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.
799 static void io_iopoll_req_issued(struct io_kiocb *req)
801 struct io_ring_ctx *ctx = req->ctx;
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
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;
813 list_req = list_first_entry(&ctx->poll_list, struct io_kiocb,
815 if (list_req->rw.ki_filp != req->rw.ki_filp)
816 ctx->poll_multi_file = true;
820 * For fast devices, IO may have already completed. If it has, add
821 * it to the front so we find it first.
823 if (req->flags & REQ_F_IOPOLL_COMPLETED)
824 list_add(&req->list, &ctx->poll_list);
826 list_add_tail(&req->list, &ctx->poll_list);
829 static void io_file_put(struct io_submit_state *state)
832 int diff = state->has_refs - state->used_refs;
835 fput_many(state->file, diff);
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.
845 static struct file *io_file_get(struct io_submit_state *state, int fd)
851 if (state->fd == fd) {
858 state->file = fget_many(fd, state->ios_left);
863 state->has_refs = state->ios_left;
864 state->used_refs = 1;
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
874 static bool io_file_supports_async(struct file *file)
876 umode_t mode = file_inode(file)->i_mode;
878 if (S_ISBLK(mode) || S_ISCHR(mode))
880 if (S_ISREG(mode) && file->f_op != &io_uring_fops)
886 static int io_prep_rw(struct io_kiocb *req, const struct sqe_submit *s,
889 const struct io_uring_sqe *sqe = s->sqe;
890 struct io_ring_ctx *ctx = req->ctx;
891 struct kiocb *kiocb = &req->rw;
898 if (force_nonblock && !io_file_supports_async(req->file))
899 force_nonblock = false;
901 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));
905 ioprio = READ_ONCE(sqe->ioprio);
907 ret = ioprio_check_cap(ioprio);
911 kiocb->ki_ioprio = ioprio;
913 kiocb->ki_ioprio = get_current_ioprio();
915 ret = kiocb_set_rw_flags(kiocb, READ_ONCE(sqe->rw_flags));
919 /* don't allow async punt if RWF_NOWAIT was requested */
920 if (kiocb->ki_flags & IOCB_NOWAIT)
921 req->flags |= REQ_F_NOWAIT;
924 kiocb->ki_flags |= IOCB_NOWAIT;
926 if (ctx->flags & IORING_SETUP_IOPOLL) {
927 if (!(kiocb->ki_flags & IOCB_DIRECT) ||
928 !kiocb->ki_filp->f_op->iopoll)
932 kiocb->ki_flags |= IOCB_HIPRI;
933 kiocb->ki_complete = io_complete_rw_iopoll;
935 if (kiocb->ki_flags & IOCB_HIPRI)
937 kiocb->ki_complete = io_complete_rw;
942 static inline void io_rw_done(struct kiocb *kiocb, ssize_t ret)
948 case -ERESTARTNOINTR:
949 case -ERESTARTNOHAND:
950 case -ERESTART_RESTARTBLOCK:
952 * We can't just restart the syscall, since previously
953 * submitted sqes may already be in progress. Just fail this
959 kiocb->ki_complete(kiocb, ret, 0);
963 static int io_import_fixed(struct io_ring_ctx *ctx, int rw,
964 const struct io_uring_sqe *sqe,
965 struct iov_iter *iter)
967 size_t len = READ_ONCE(sqe->len);
968 struct io_mapped_ubuf *imu;
969 unsigned index, buf_index;
973 /* attempt to use fixed buffers without having provided iovecs */
974 if (unlikely(!ctx->user_bufs))
977 buf_index = READ_ONCE(sqe->buf_index);
978 if (unlikely(buf_index >= ctx->nr_user_bufs))
981 index = array_index_nospec(buf_index, ctx->nr_user_bufs);
982 imu = &ctx->user_bufs[index];
983 buf_addr = READ_ONCE(sqe->addr);
986 if (buf_addr + len < buf_addr)
988 /* not inside the mapped region */
989 if (buf_addr < imu->ubuf || buf_addr + len > imu->ubuf + imu->len)
993 * May not be a start of buffer, set size appropriately
994 * and advance us to the beginning.
996 offset = buf_addr - imu->ubuf;
997 iov_iter_bvec(iter, rw, imu->bvec, imu->nr_bvecs, offset + len);
999 iov_iter_advance(iter, offset);
1001 /* don't drop a reference to these pages */
1002 iter->type |= ITER_BVEC_FLAG_NO_REF;
1006 static 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)
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);
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
1023 opcode = READ_ONCE(sqe->opcode);
1024 if (opcode == IORING_OP_READ_FIXED ||
1025 opcode == IORING_OP_WRITE_FIXED) {
1026 int ret = io_import_fixed(ctx, rw, sqe, iter);
1034 #ifdef CONFIG_COMPAT
1036 return compat_import_iovec(rw, buf, sqe_len, UIO_FASTIOV,
1040 return import_iovec(rw, buf, sqe_len, UIO_FASTIOV, iovec, iter);
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.
1049 static void io_async_list_note(int rw, struct io_kiocb *req, size_t len)
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;
1056 if (filp == async_list->file && kiocb->ki_pos == async_list->io_end) {
1057 unsigned long max_pages;
1059 /* Use 8x RA size as a decent limiter for both reads/writes */
1060 max_pages = filp->f_ra.ra_pages;
1062 max_pages = VM_READAHEAD_PAGES;
1065 /* If max pages are exceeded, reset the state */
1067 if (async_list->io_pages + len <= max_pages) {
1068 req->flags |= REQ_F_SEQ_PREV;
1069 async_list->io_pages += len;
1072 async_list->io_pages = 0;
1076 /* New file? Reset state. */
1077 if (async_list->file != filp) {
1078 async_list->io_pages = 0;
1079 async_list->file = filp;
1081 async_list->io_end = io_end;
1084 static int io_read(struct io_kiocb *req, const struct sqe_submit *s,
1085 bool force_nonblock)
1087 struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs;
1088 struct kiocb *kiocb = &req->rw;
1089 struct iov_iter iter;
1094 ret = io_prep_rw(req, s, force_nonblock);
1097 file = kiocb->ki_filp;
1099 if (unlikely(!(file->f_mode & FMODE_READ)))
1101 if (unlikely(!file->f_op->read_iter))
1104 ret = io_import_iovec(req->ctx, READ, s, &iovec, &iter);
1108 iov_count = iov_iter_count(&iter);
1109 ret = rw_verify_area(READ, file, &kiocb->ki_pos, iov_count);
1113 /* Catch -EAGAIN return for forced non-blocking submission */
1114 ret2 = call_read_iter(file, kiocb, &iter);
1115 if (!force_nonblock || ret2 != -EAGAIN) {
1116 io_rw_done(kiocb, ret2);
1119 * If ->needs_lock is true, we're already in async
1123 io_async_list_note(READ, req, iov_count);
1131 static int io_write(struct io_kiocb *req, const struct sqe_submit *s,
1132 bool force_nonblock)
1134 struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs;
1135 struct kiocb *kiocb = &req->rw;
1136 struct iov_iter iter;
1141 ret = io_prep_rw(req, s, force_nonblock);
1145 file = kiocb->ki_filp;
1146 if (unlikely(!(file->f_mode & FMODE_WRITE)))
1148 if (unlikely(!file->f_op->write_iter))
1151 ret = io_import_iovec(req->ctx, WRITE, s, &iovec, &iter);
1155 iov_count = iov_iter_count(&iter);
1158 if (force_nonblock && !(kiocb->ki_flags & IOCB_DIRECT)) {
1159 /* If ->needs_lock is true, we're already in async context. */
1161 io_async_list_note(WRITE, req, iov_count);
1165 ret = rw_verify_area(WRITE, file, &kiocb->ki_pos, iov_count);
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.
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,
1182 kiocb->ki_flags |= IOCB_WRITE;
1184 ret2 = call_write_iter(file, kiocb, &iter);
1185 if (!force_nonblock || ret2 != -EAGAIN) {
1186 io_rw_done(kiocb, ret2);
1189 * If ->needs_lock is true, we're already in async
1193 io_async_list_note(WRITE, req, iov_count);
1203 * IORING_OP_NOP just posts a completion event, nothing else.
1205 static int io_nop(struct io_kiocb *req, u64 user_data)
1207 struct io_ring_ctx *ctx = req->ctx;
1210 if (unlikely(ctx->flags & IORING_SETUP_IOPOLL))
1213 io_cqring_add_event(ctx, user_data, err);
1218 static int io_prep_fsync(struct io_kiocb *req, const struct io_uring_sqe *sqe)
1220 struct io_ring_ctx *ctx = req->ctx;
1225 if (unlikely(ctx->flags & IORING_SETUP_IOPOLL))
1227 if (unlikely(sqe->addr || sqe->ioprio || sqe->buf_index))
1233 static int io_fsync(struct io_kiocb *req, const struct io_uring_sqe *sqe,
1234 bool force_nonblock)
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;
1242 fsync_flags = READ_ONCE(sqe->fsync_flags);
1243 if (unlikely(fsync_flags & ~IORING_FSYNC_DATASYNC))
1246 ret = io_prep_fsync(req, sqe);
1250 /* fsync always requires a blocking context */
1254 ret = vfs_fsync_range(req->rw.ki_filp, sqe_off,
1255 end > 0 ? end : LLONG_MAX,
1256 fsync_flags & IORING_FSYNC_DATASYNC);
1258 io_cqring_add_event(req->ctx, sqe->user_data, ret);
1263 static int io_prep_sfr(struct io_kiocb *req, const struct io_uring_sqe *sqe)
1265 struct io_ring_ctx *ctx = req->ctx;
1271 if (unlikely(ctx->flags & IORING_SETUP_IOPOLL))
1273 if (unlikely(sqe->addr || sqe->ioprio || sqe->buf_index))
1279 static int io_sync_file_range(struct io_kiocb *req,
1280 const struct io_uring_sqe *sqe,
1281 bool force_nonblock)
1288 ret = io_prep_sfr(req, sqe);
1292 /* sync_file_range always requires a blocking context */
1296 sqe_off = READ_ONCE(sqe->off);
1297 sqe_len = READ_ONCE(sqe->len);
1298 flags = READ_ONCE(sqe->sync_range_flags);
1300 ret = sync_file_range(req->rw.ki_filp, sqe_off, sqe_len, flags);
1302 io_cqring_add_event(req->ctx, sqe->user_data, ret);
1307 static void io_poll_remove_one(struct io_kiocb *req)
1309 struct io_poll_iocb *poll = &req->poll;
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);
1317 spin_unlock(&poll->head->lock);
1319 list_del_init(&req->list);
1322 static void io_poll_remove_all(struct io_ring_ctx *ctx)
1324 struct io_kiocb *req;
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);
1331 spin_unlock_irq(&ctx->completion_lock);
1335 * Find a running poll command that matches one specified in sqe->addr,
1336 * and remove it if found.
1338 static int io_poll_remove(struct io_kiocb *req, const struct io_uring_sqe *sqe)
1340 struct io_ring_ctx *ctx = req->ctx;
1341 struct io_kiocb *poll_req, *next;
1344 if (unlikely(req->ctx->flags & IORING_SETUP_IOPOLL))
1346 if (sqe->ioprio || sqe->off || sqe->len || sqe->buf_index ||
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);
1358 spin_unlock_irq(&ctx->completion_lock);
1360 io_cqring_add_event(req->ctx, sqe->user_data, ret);
1365 static void io_poll_complete(struct io_ring_ctx *ctx, struct io_kiocb *req,
1368 req->poll.done = true;
1369 io_cqring_fill_event(ctx, req->user_data, mangle_poll(mask));
1370 io_commit_cqring(ctx);
1373 static void io_poll_complete_work(struct work_struct *work)
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;
1381 if (!READ_ONCE(poll->canceled))
1382 mask = vfs_poll(poll->file, &pt) & poll->events;
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.
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);
1397 list_del_init(&req->list);
1398 io_poll_complete(ctx, req, mask);
1399 spin_unlock_irq(&ctx->completion_lock);
1401 io_cqring_ev_posted(ctx);
1405 static int io_poll_wake(struct wait_queue_entry *wait, unsigned mode, int sync,
1408 struct io_poll_iocb *poll = container_of(wait, struct io_poll_iocb,
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);
1413 unsigned long flags;
1415 /* for instances that support it check for an event match first: */
1416 if (mask && !(mask & poll->events))
1419 list_del_init(&poll->wait.entry);
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);
1426 io_cqring_ev_posted(ctx);
1429 queue_work(ctx->sqo_wq, &req->work);
1435 struct io_poll_table {
1436 struct poll_table_struct pt;
1437 struct io_kiocb *req;
1441 static void io_poll_queue_proc(struct file *file, struct wait_queue_head *head,
1442 struct poll_table_struct *p)
1444 struct io_poll_table *pt = container_of(p, struct io_poll_table, pt);
1446 if (unlikely(pt->req->poll.head)) {
1447 pt->error = -EINVAL;
1452 pt->req->poll.head = head;
1453 add_wait_queue(head, &pt->req->poll.wait);
1456 static int io_poll_add(struct io_kiocb *req, const struct io_uring_sqe *sqe)
1458 struct io_poll_iocb *poll = &req->poll;
1459 struct io_ring_ctx *ctx = req->ctx;
1460 struct io_poll_table ipt;
1461 bool cancel = false;
1465 if (unlikely(req->ctx->flags & IORING_SETUP_IOPOLL))
1467 if (sqe->addr || sqe->ioprio || sqe->off || sqe->len || sqe->buf_index)
1472 INIT_WORK(&req->work, io_poll_complete_work);
1473 events = READ_ONCE(sqe->poll_events);
1474 poll->events = demangle_poll(events) | EPOLLERR | EPOLLHUP;
1478 poll->canceled = false;
1480 ipt.pt._qproc = io_poll_queue_proc;
1481 ipt.pt._key = poll->events;
1483 ipt.error = -EINVAL; /* same as no support for IOCB_CMD_POLL */
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);
1489 mask = vfs_poll(poll->file, &ipt.pt) & poll->events;
1491 spin_lock_irq(&ctx->completion_lock);
1492 if (likely(poll->head)) {
1493 spin_lock(&poll->head->lock);
1494 if (unlikely(list_empty(&poll->wait.entry))) {
1500 if (mask || ipt.error)
1501 list_del_init(&poll->wait.entry);
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);
1508 if (mask) { /* no async, we'd stolen it */
1510 io_poll_complete(ctx, req, mask);
1512 spin_unlock_irq(&ctx->completion_lock);
1515 io_cqring_ev_posted(ctx);
1521 static int io_req_defer(struct io_ring_ctx *ctx, struct io_kiocb *req,
1522 const struct io_uring_sqe *sqe)
1524 struct io_uring_sqe *sqe_copy;
1526 if (!io_sequence_defer(ctx, req) && list_empty(&ctx->defer_list))
1529 sqe_copy = kmalloc(sizeof(*sqe_copy), GFP_KERNEL);
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);
1540 memcpy(sqe_copy, sqe, sizeof(*sqe_copy));
1541 req->submit.sqe = sqe_copy;
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;
1549 static int __io_submit_sqe(struct io_ring_ctx *ctx, struct io_kiocb *req,
1550 const struct sqe_submit *s, bool force_nonblock)
1554 if (unlikely(s->index >= ctx->sq_entries))
1556 req->user_data = READ_ONCE(s->sqe->user_data);
1558 opcode = READ_ONCE(s->sqe->opcode);
1561 ret = io_nop(req, req->user_data);
1563 case IORING_OP_READV:
1564 if (unlikely(s->sqe->buf_index))
1566 ret = io_read(req, s, force_nonblock);
1568 case IORING_OP_WRITEV:
1569 if (unlikely(s->sqe->buf_index))
1571 ret = io_write(req, s, force_nonblock);
1573 case IORING_OP_READ_FIXED:
1574 ret = io_read(req, s, force_nonblock);
1576 case IORING_OP_WRITE_FIXED:
1577 ret = io_write(req, s, force_nonblock);
1579 case IORING_OP_FSYNC:
1580 ret = io_fsync(req, s->sqe, force_nonblock);
1582 case IORING_OP_POLL_ADD:
1583 ret = io_poll_add(req, s->sqe);
1585 case IORING_OP_POLL_REMOVE:
1586 ret = io_poll_remove(req, s->sqe);
1588 case IORING_OP_SYNC_FILE_RANGE:
1589 ret = io_sync_file_range(req, s->sqe, force_nonblock);
1599 if (ctx->flags & IORING_SETUP_IOPOLL) {
1600 if (req->error == -EAGAIN)
1603 /* workqueue context doesn't hold uring_lock, grab it now */
1605 mutex_lock(&ctx->uring_lock);
1606 io_iopoll_req_issued(req);
1608 mutex_unlock(&ctx->uring_lock);
1614 static struct async_list *io_async_list_from_sqe(struct io_ring_ctx *ctx,
1615 const struct io_uring_sqe *sqe)
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];
1629 static inline bool io_sqe_needs_user(const struct io_uring_sqe *sqe)
1631 u8 opcode = READ_ONCE(sqe->opcode);
1633 return !(opcode == IORING_OP_READ_FIXED ||
1634 opcode == IORING_OP_WRITE_FIXED);
1637 static void io_sq_wq_submit_work(struct work_struct *work)
1639 struct io_kiocb *req = container_of(work, struct io_kiocb, work);
1640 struct io_ring_ctx *ctx = req->ctx;
1641 struct mm_struct *cur_mm = NULL;
1642 struct async_list *async_list;
1643 LIST_HEAD(req_list);
1644 mm_segment_t old_fs;
1647 async_list = io_async_list_from_sqe(ctx, req->submit.sqe);
1650 struct sqe_submit *s = &req->submit;
1651 const struct io_uring_sqe *sqe = s->sqe;
1653 /* Ensure we clear previously set non-block flag */
1654 req->rw.ki_flags &= ~IOCB_NOWAIT;
1657 if (io_sqe_needs_user(sqe) && !cur_mm) {
1658 if (!mmget_not_zero(ctx->sqo_mm)) {
1661 cur_mm = ctx->sqo_mm;
1669 s->has_user = cur_mm != NULL;
1670 s->needs_lock = true;
1672 ret = __io_submit_sqe(ctx, req, s, false);
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
1685 /* drop submission reference */
1689 io_cqring_add_event(ctx, sqe->user_data, ret);
1693 /* async context always use a copy of the sqe */
1698 if (!list_empty(&req_list)) {
1699 req = list_first_entry(&req_list, struct io_kiocb,
1701 list_del(&req->list);
1704 if (list_empty(&async_list->list))
1708 spin_lock(&async_list->lock);
1709 if (list_empty(&async_list->list)) {
1710 spin_unlock(&async_list->lock);
1713 list_splice_init(&async_list->list, &req_list);
1714 spin_unlock(&async_list->lock);
1716 req = list_first_entry(&req_list, struct io_kiocb, list);
1717 list_del(&req->list);
1721 * 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.
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);
1733 if (!list_empty(&req_list)) {
1734 req = list_first_entry(&req_list,
1735 struct io_kiocb, list);
1736 list_del(&req->list);
1739 ret = atomic_dec_return(&async_list->cnt);
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.
1755 static bool io_add_to_prev_work(struct async_list *list, struct io_kiocb *req)
1761 if (!(req->flags & REQ_F_SEQ_PREV))
1763 if (!atomic_read(&list->cnt))
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);
1773 spin_unlock(&list->lock);
1777 static bool io_op_needs_file(const struct io_uring_sqe *sqe)
1779 int op = READ_ONCE(sqe->opcode);
1783 case IORING_OP_POLL_REMOVE:
1790 static 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)
1796 flags = READ_ONCE(s->sqe->flags);
1797 fd = READ_ONCE(s->sqe->fd);
1799 if (flags & IOSQE_IO_DRAIN) {
1800 req->flags |= REQ_F_IO_DRAIN;
1801 req->sequence = ctx->cached_sq_head - 1;
1804 if (!io_op_needs_file(s->sqe)) {
1809 if (flags & IOSQE_FIXED_FILE) {
1810 if (unlikely(!ctx->user_files ||
1811 (unsigned) fd >= ctx->nr_user_files))
1813 req->file = ctx->user_files[fd];
1814 req->flags |= REQ_F_FIXED_FILE;
1816 if (s->needs_fixed_file)
1818 req->file = io_file_get(state, fd);
1819 if (unlikely(!req->file))
1826 static int io_submit_sqe(struct io_ring_ctx *ctx, struct sqe_submit *s,
1827 struct io_submit_state *state)
1829 struct io_kiocb *req;
1832 /* enforce forwards compatibility on users */
1833 if (unlikely(s->sqe->flags & ~(IOSQE_FIXED_FILE | IOSQE_IO_DRAIN)))
1836 req = io_get_req(ctx, state);
1840 ret = io_req_set_file(ctx, s, state, req);
1844 ret = io_req_defer(ctx, req, s->sqe);
1846 if (ret == -EIOCBQUEUED)
1851 ret = __io_submit_sqe(ctx, req, s, true);
1852 if (ret == -EAGAIN && !(req->flags & REQ_F_NOWAIT)) {
1853 struct io_uring_sqe *sqe_copy;
1855 sqe_copy = kmalloc(sizeof(*sqe_copy), GFP_KERNEL);
1857 struct async_list *list;
1859 memcpy(sqe_copy, s->sqe, sizeof(*sqe_copy));
1862 memcpy(&req->submit, s, sizeof(*s));
1863 list = io_async_list_from_sqe(ctx, s->sqe);
1864 if (!io_add_to_prev_work(list, req)) {
1866 atomic_inc(&list->cnt);
1867 INIT_WORK(&req->work, io_sq_wq_submit_work);
1868 queue_work(ctx->sqo_wq, &req->work);
1872 * Queued up for async execution, worker will release
1873 * submit reference when the iocb is actually
1881 /* drop submission reference */
1884 /* and drop final reference, if we failed */
1892 * Batched submission is done, ensure local IO is flushed out.
1894 static void io_submit_state_end(struct io_submit_state *state)
1896 blk_finish_plug(&state->plug);
1898 if (state->free_reqs)
1899 kmem_cache_free_bulk(req_cachep, state->free_reqs,
1900 &state->reqs[state->cur_req]);
1904 * Start submission side cache.
1906 static void io_submit_state_start(struct io_submit_state *state,
1907 struct io_ring_ctx *ctx, unsigned max_ios)
1909 blk_start_plug(&state->plug);
1910 state->free_reqs = 0;
1912 state->ios_left = max_ios;
1915 static void io_commit_sqring(struct io_ring_ctx *ctx)
1917 struct io_sq_ring *ring = ctx->sq_ring;
1919 if (ctx->cached_sq_head != READ_ONCE(ring->r.head)) {
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.
1925 smp_store_release(&ring->r.head, ctx->cached_sq_head);
1930 * 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.
1937 static bool io_get_sqring(struct io_ring_ctx *ctx, struct sqe_submit *s)
1939 struct io_sq_ring *ring = ctx->sq_ring;
1943 * The cached sq head (or cq tail) serves two purposes:
1945 * 1) allows us to batch the cost of updating the user visible
1947 * 2) allows the kernel side to track the head on its own, even
1948 * though the application is the one updating it.
1950 head = ctx->cached_sq_head;
1951 /* make sure SQ entry isn't read before tail */
1952 if (head == smp_load_acquire(&ring->r.tail))
1955 head = READ_ONCE(ring->array[head & ctx->sq_mask]);
1956 if (head < ctx->sq_entries) {
1958 s->sqe = &ctx->sq_sqes[head];
1959 ctx->cached_sq_head++;
1963 /* drop invalid entries */
1964 ctx->cached_sq_head++;
1969 static int io_submit_sqes(struct io_ring_ctx *ctx, struct sqe_submit *sqes,
1970 unsigned int nr, bool has_user, bool mm_fault)
1972 struct io_submit_state state, *statep = NULL;
1973 int ret, i, submitted = 0;
1975 if (nr > IO_PLUG_THRESHOLD) {
1976 io_submit_state_start(&state, ctx, nr);
1980 for (i = 0; i < nr; i++) {
1981 if (unlikely(mm_fault)) {
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);
1994 io_cqring_add_event(ctx, sqes[i].sqe->user_data, ret);
1998 io_submit_state_end(&state);
2003 static int io_sq_thread(void *data)
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;
2011 unsigned long timeout;
2016 timeout = inflight = 0;
2017 while (!kthread_should_park()) {
2018 bool all_fixed, mm_fault = false;
2022 unsigned nr_events = 0;
2024 if (ctx->flags & IORING_SETUP_IOPOLL) {
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.
2031 mutex_lock(&ctx->uring_lock);
2032 io_iopoll_check(ctx, &nr_events, 0);
2033 mutex_unlock(&ctx->uring_lock);
2036 * Normal IO, just pretend everything completed.
2037 * We don't have to poll completions for that.
2039 nr_events = inflight;
2042 inflight -= nr_events;
2044 timeout = jiffies + ctx->sq_thread_idle;
2047 if (!io_get_sqring(ctx, &sqes[0])) {
2049 * We're polling. If we're within the defined idle
2050 * period, then let us spin without work before going
2053 if (inflight || !time_after(jiffies, timeout)) {
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
2070 prepare_to_wait(&ctx->sqo_wait, &wait,
2071 TASK_INTERRUPTIBLE);
2073 /* Tell userspace we may need a wakeup call */
2074 ctx->sq_ring->flags |= IORING_SQ_NEED_WAKEUP;
2075 /* make sure to read SQ tail after writing flags */
2078 if (!io_get_sqring(ctx, &sqes[0])) {
2079 if (kthread_should_park()) {
2080 finish_wait(&ctx->sqo_wait, &wait);
2083 if (signal_pending(current))
2084 flush_signals(current);
2086 finish_wait(&ctx->sqo_wait, &wait);
2088 ctx->sq_ring->flags &= ~IORING_SQ_NEED_WAKEUP;
2091 finish_wait(&ctx->sqo_wait, &wait);
2093 ctx->sq_ring->flags &= ~IORING_SQ_NEED_WAKEUP;
2099 if (all_fixed && io_sqe_needs_user(sqes[i].sqe))
2103 if (i == ARRAY_SIZE(sqes))
2105 } while (io_get_sqring(ctx, &sqes[i]));
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);
2111 use_mm(ctx->sqo_mm);
2112 cur_mm = ctx->sqo_mm;
2116 inflight += io_submit_sqes(ctx, sqes, i, cur_mm != NULL,
2119 /* Commit SQ ring head once we've consumed all SQEs */
2120 io_commit_sqring(ctx);
2134 static int io_ring_submit(struct io_ring_ctx *ctx, unsigned int to_submit)
2136 struct io_submit_state state, *statep = NULL;
2139 if (to_submit > IO_PLUG_THRESHOLD) {
2140 io_submit_state_start(&state, ctx, to_submit);
2144 for (i = 0; i < to_submit; i++) {
2145 struct sqe_submit s;
2148 if (!io_get_sqring(ctx, &s))
2152 s.needs_lock = false;
2153 s.needs_fixed_file = false;
2156 ret = io_submit_sqe(ctx, &s, statep);
2158 io_cqring_add_event(ctx, s.sqe->user_data, ret);
2160 io_commit_sqring(ctx);
2163 io_submit_state_end(statep);
2168 static unsigned io_cqring_events(struct io_cq_ring *ring)
2170 /* See comment at the top of this file */
2172 return READ_ONCE(ring->r.tail) - READ_ONCE(ring->r.head);
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.
2179 static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events,
2180 const sigset_t __user *sig, size_t sigsz)
2182 struct io_cq_ring *ring = ctx->cq_ring;
2183 sigset_t ksigmask, sigsaved;
2186 if (io_cqring_events(ring) >= min_events)
2190 #ifdef CONFIG_COMPAT
2191 if (in_compat_syscall())
2192 ret = set_compat_user_sigmask((const compat_sigset_t __user *)sig,
2193 &ksigmask, &sigsaved, sigsz);
2196 ret = set_user_sigmask(sig, &ksigmask,
2203 ret = wait_event_interruptible(ctx->wait, io_cqring_events(ring) >= min_events);
2204 if (ret == -ERESTARTSYS)
2208 restore_user_sigmask(sig, &sigsaved);
2210 return READ_ONCE(ring->r.head) == READ_ONCE(ring->r.tail) ? ret : 0;
2213 static void __io_sqe_files_unregister(struct io_ring_ctx *ctx)
2215 #if defined(CONFIG_UNIX)
2216 if (ctx->ring_sock) {
2217 struct sock *sock = ctx->ring_sock->sk;
2218 struct sk_buff *skb;
2220 while ((skb = skb_dequeue(&sock->sk_receive_queue)) != NULL)
2226 for (i = 0; i < ctx->nr_user_files; i++)
2227 fput(ctx->user_files[i]);
2231 static int io_sqe_files_unregister(struct io_ring_ctx *ctx)
2233 if (!ctx->user_files)
2236 __io_sqe_files_unregister(ctx);
2237 kfree(ctx->user_files);
2238 ctx->user_files = NULL;
2239 ctx->nr_user_files = 0;
2243 static void io_sq_thread_stop(struct io_ring_ctx *ctx)
2245 if (ctx->sqo_thread) {
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.
2251 kthread_park(ctx->sqo_thread);
2252 kthread_stop(ctx->sqo_thread);
2253 ctx->sqo_thread = NULL;
2257 static void io_finish_async(struct io_ring_ctx *ctx)
2259 io_sq_thread_stop(ctx);
2262 destroy_workqueue(ctx->sqo_wq);
2267 #if defined(CONFIG_UNIX)
2268 static void io_destruct_skb(struct sk_buff *skb)
2270 struct io_ring_ctx *ctx = skb->sk->sk_user_data;
2272 io_finish_async(ctx);
2273 unix_destruct_scm(skb);
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.
2281 static int __io_sqe_files_scm(struct io_ring_ctx *ctx, int nr, int offset)
2283 struct sock *sk = ctx->ring_sock->sk;
2284 struct scm_fp_list *fpl;
2285 struct sk_buff *skb;
2288 if (!capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN)) {
2289 unsigned long inflight = ctx->user->unix_inflight + nr;
2291 if (inflight > task_rlimit(current, RLIMIT_NOFILE))
2295 fpl = kzalloc(sizeof(*fpl), GFP_KERNEL);
2299 skb = alloc_skb(0, GFP_KERNEL);
2306 skb->destructor = io_destruct_skb;
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]);
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);
2319 for (i = 0; i < nr; i++)
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.
2330 static int io_sqe_files_scm(struct io_ring_ctx *ctx)
2332 unsigned left, total;
2336 left = ctx->nr_user_files;
2338 unsigned this_files = min_t(unsigned, left, SCM_MAX_FD);
2340 ret = __io_sqe_files_scm(ctx, this_files, total);
2344 total += this_files;
2350 while (total < ctx->nr_user_files) {
2351 fput(ctx->user_files[total]);
2358 static int io_sqe_files_scm(struct io_ring_ctx *ctx)
2364 static int io_sqe_files_register(struct io_ring_ctx *ctx, void __user *arg,
2367 __s32 __user *fds = (__s32 __user *) arg;
2371 if (ctx->user_files)
2375 if (nr_args > IORING_MAX_FIXED_FILES)
2378 ctx->user_files = kcalloc(nr_args, sizeof(struct file *), GFP_KERNEL);
2379 if (!ctx->user_files)
2382 for (i = 0; i < nr_args; i++) {
2384 if (copy_from_user(&fd, &fds[i], sizeof(fd)))
2387 ctx->user_files[i] = fget(fd);
2390 if (!ctx->user_files[i])
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.
2399 if (ctx->user_files[i]->f_op == &io_uring_fops) {
2400 fput(ctx->user_files[i]);
2403 ctx->nr_user_files++;
2408 for (i = 0; i < ctx->nr_user_files; i++)
2409 fput(ctx->user_files[i]);
2411 kfree(ctx->user_files);
2412 ctx->user_files = NULL;
2413 ctx->nr_user_files = 0;
2417 ret = io_sqe_files_scm(ctx);
2419 io_sqe_files_unregister(ctx);
2424 static int io_sq_offload_start(struct io_ring_ctx *ctx,
2425 struct io_uring_params *p)
2429 init_waitqueue_head(&ctx->sqo_wait);
2430 mmgrab(current->mm);
2431 ctx->sqo_mm = current->mm;
2433 if (ctx->flags & IORING_SETUP_SQPOLL) {
2435 if (!capable(CAP_SYS_ADMIN))
2438 ctx->sq_thread_idle = msecs_to_jiffies(p->sq_thread_idle);
2439 if (!ctx->sq_thread_idle)
2440 ctx->sq_thread_idle = HZ;
2442 if (p->flags & IORING_SETUP_SQ_AFF) {
2443 int cpu = p->sq_thread_cpu;
2446 if (cpu >= nr_cpu_ids)
2448 if (!cpu_online(cpu))
2451 ctx->sqo_thread = kthread_create_on_cpu(io_sq_thread,
2455 ctx->sqo_thread = kthread_create(io_sq_thread, ctx,
2458 if (IS_ERR(ctx->sqo_thread)) {
2459 ret = PTR_ERR(ctx->sqo_thread);
2460 ctx->sqo_thread = NULL;
2463 wake_up_process(ctx->sqo_thread);
2464 } else if (p->flags & IORING_SETUP_SQ_AFF) {
2465 /* Can't have SQ_AFF without SQPOLL */
2470 /* 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()));
2480 io_sq_thread_stop(ctx);
2481 mmdrop(ctx->sqo_mm);
2486 static void io_unaccount_mem(struct user_struct *user, unsigned long nr_pages)
2488 atomic_long_sub(nr_pages, &user->locked_vm);
2491 static int io_account_mem(struct user_struct *user, unsigned long nr_pages)
2493 unsigned long page_limit, cur_pages, new_pages;
2495 /* Don't allow more pages than we can safely lock */
2496 page_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT;
2499 cur_pages = atomic_long_read(&user->locked_vm);
2500 new_pages = cur_pages + nr_pages;
2501 if (new_pages > page_limit)
2503 } while (atomic_long_cmpxchg(&user->locked_vm, cur_pages,
2504 new_pages) != cur_pages);
2509 static void io_mem_free(void *ptr)
2516 page = virt_to_head_page(ptr);
2517 if (put_page_testzero(page))
2518 free_compound_page(page);
2521 static void *io_mem_alloc(size_t size)
2523 gfp_t gfp_flags = GFP_KERNEL | __GFP_ZERO | __GFP_NOWARN | __GFP_COMP |
2526 return (void *) __get_free_pages(gfp_flags, get_order(size));
2529 static unsigned long ring_pages(unsigned sq_entries, unsigned cq_entries)
2531 struct io_sq_ring *sq_ring;
2532 struct io_cq_ring *cq_ring;
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);
2539 return (bytes + PAGE_SIZE - 1) / PAGE_SIZE;
2542 static int io_sqe_buffer_unregister(struct io_ring_ctx *ctx)
2546 if (!ctx->user_bufs)
2549 for (i = 0; i < ctx->nr_user_bufs; i++) {
2550 struct io_mapped_ubuf *imu = &ctx->user_bufs[i];
2552 for (j = 0; j < imu->nr_bvecs; j++)
2553 put_page(imu->bvec[j].bv_page);
2555 if (ctx->account_mem)
2556 io_unaccount_mem(ctx->user, imu->nr_bvecs);
2561 kfree(ctx->user_bufs);
2562 ctx->user_bufs = NULL;
2563 ctx->nr_user_bufs = 0;
2567 static int io_copy_iov(struct io_ring_ctx *ctx, struct iovec *dst,
2568 void __user *arg, unsigned index)
2570 struct iovec __user *src;
2572 #ifdef CONFIG_COMPAT
2574 struct compat_iovec __user *ciovs;
2575 struct compat_iovec ciov;
2577 ciovs = (struct compat_iovec __user *) arg;
2578 if (copy_from_user(&ciov, &ciovs[index], sizeof(ciov)))
2581 dst->iov_base = (void __user *) (unsigned long) ciov.iov_base;
2582 dst->iov_len = ciov.iov_len;
2586 src = (struct iovec __user *) arg;
2587 if (copy_from_user(dst, &src[index], sizeof(*dst)))
2592 static int io_sqe_buffer_register(struct io_ring_ctx *ctx, void __user *arg,
2595 struct vm_area_struct **vmas = NULL;
2596 struct page **pages = NULL;
2597 int i, j, got_pages = 0;
2602 if (!nr_args || nr_args > UIO_MAXIOV)
2605 ctx->user_bufs = kcalloc(nr_args, sizeof(struct io_mapped_ubuf),
2607 if (!ctx->user_bufs)
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;
2617 ret = io_copy_iov(ctx, &iov, arg, i);
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.
2627 if (!iov.iov_base || !iov.iov_len)
2630 /* arbitrary limit, but we need something */
2631 if (iov.iov_len > SZ_1G)
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;
2639 if (ctx->account_mem) {
2640 ret = io_account_mem(ctx->user, nr_pages);
2646 if (!pages || nr_pages > got_pages) {
2649 pages = kvmalloc_array(nr_pages, sizeof(struct page *),
2651 vmas = kvmalloc_array(nr_pages,
2652 sizeof(struct vm_area_struct *),
2654 if (!pages || !vmas) {
2656 if (ctx->account_mem)
2657 io_unaccount_mem(ctx->user, nr_pages);
2660 got_pages = nr_pages;
2663 imu->bvec = kvmalloc_array(nr_pages, sizeof(struct bio_vec),
2667 if (ctx->account_mem)
2668 io_unaccount_mem(ctx->user, nr_pages);
2673 down_read(¤t->mm->mmap_sem);
2674 pret = get_user_pages(ubuf, nr_pages,
2675 FOLL_WRITE | FOLL_LONGTERM,
2677 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];
2683 !is_file_hugepages(vma->vm_file)) {
2689 ret = pret < 0 ? pret : -EFAULT;
2691 up_read(¤t->mm->mmap_sem);
2694 * if we did partial map, or found file backed vmas,
2695 * release any pages we did get
2698 for (j = 0; j < pret; j++)
2701 if (ctx->account_mem)
2702 io_unaccount_mem(ctx->user, nr_pages);
2707 off = ubuf & ~PAGE_MASK;
2709 for (j = 0; j < nr_pages; j++) {
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;
2719 /* store original address for later verification */
2721 imu->len = iov.iov_len;
2722 imu->nr_bvecs = nr_pages;
2724 ctx->nr_user_bufs++;
2732 io_sqe_buffer_unregister(ctx);
2736 static int io_eventfd_register(struct io_ring_ctx *ctx, void __user *arg)
2738 __s32 __user *fds = arg;
2744 if (copy_from_user(&fd, fds, sizeof(*fds)))
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;
2757 static int io_eventfd_unregister(struct io_ring_ctx *ctx)
2759 if (ctx->cq_ev_fd) {
2760 eventfd_ctx_put(ctx->cq_ev_fd);
2761 ctx->cq_ev_fd = NULL;
2768 static void io_ring_ctx_free(struct io_ring_ctx *ctx)
2770 io_finish_async(ctx);
2772 mmdrop(ctx->sqo_mm);
2774 io_iopoll_reap_events(ctx);
2775 io_sqe_buffer_unregister(ctx);
2776 io_sqe_files_unregister(ctx);
2777 io_eventfd_unregister(ctx);
2779 #if defined(CONFIG_UNIX)
2781 sock_release(ctx->ring_sock);
2784 io_mem_free(ctx->sq_ring);
2785 io_mem_free(ctx->sq_sqes);
2786 io_mem_free(ctx->cq_ring);
2788 percpu_ref_exit(&ctx->refs);
2789 if (ctx->account_mem)
2790 io_unaccount_mem(ctx->user,
2791 ring_pages(ctx->sq_entries, ctx->cq_entries));
2792 free_uid(ctx->user);
2796 static __poll_t io_uring_poll(struct file *file, poll_table *wait)
2798 struct io_ring_ctx *ctx = file->private_data;
2801 poll_wait(file, &ctx->cq_wait, wait);
2803 * synchronizes with barrier from wq_has_sleeper call in
2807 if (READ_ONCE(ctx->sq_ring->r.tail) - ctx->cached_sq_head !=
2808 ctx->sq_ring->ring_entries)
2809 mask |= EPOLLOUT | EPOLLWRNORM;
2810 if (READ_ONCE(ctx->cq_ring->r.head) != ctx->cached_cq_tail)
2811 mask |= EPOLLIN | EPOLLRDNORM;
2816 static int io_uring_fasync(int fd, struct file *file, int on)
2818 struct io_ring_ctx *ctx = file->private_data;
2820 return fasync_helper(fd, file, on, &ctx->cq_fasync);
2823 static void io_ring_ctx_wait_and_kill(struct io_ring_ctx *ctx)
2825 mutex_lock(&ctx->uring_lock);
2826 percpu_ref_kill(&ctx->refs);
2827 mutex_unlock(&ctx->uring_lock);
2829 io_poll_remove_all(ctx);
2830 io_iopoll_reap_events(ctx);
2831 wait_for_completion(&ctx->ctx_done);
2832 io_ring_ctx_free(ctx);
2835 static int io_uring_release(struct inode *inode, struct file *file)
2837 struct io_ring_ctx *ctx = file->private_data;
2839 file->private_data = NULL;
2840 io_ring_ctx_wait_and_kill(ctx);
2844 static int io_uring_mmap(struct file *file, struct vm_area_struct *vma)
2846 loff_t offset = (loff_t) vma->vm_pgoff << PAGE_SHIFT;
2847 unsigned long sz = vma->vm_end - vma->vm_start;
2848 struct io_ring_ctx *ctx = file->private_data;
2854 case IORING_OFF_SQ_RING:
2857 case IORING_OFF_SQES:
2860 case IORING_OFF_CQ_RING:
2867 page = virt_to_head_page(ptr);
2868 if (sz > (PAGE_SIZE << compound_order(page)))
2871 pfn = virt_to_phys(ptr) >> PAGE_SHIFT;
2872 return remap_pfn_range(vma, vma->vm_start, pfn, sz, vma->vm_page_prot);
2875 SYSCALL_DEFINE6(io_uring_enter, unsigned int, fd, u32, to_submit,
2876 u32, min_complete, u32, flags, const sigset_t __user *, sig,
2879 struct io_ring_ctx *ctx;
2884 if (flags & ~(IORING_ENTER_GETEVENTS | IORING_ENTER_SQ_WAKEUP))
2892 if (f.file->f_op != &io_uring_fops)
2896 ctx = f.file->private_data;
2897 if (!percpu_ref_tryget(&ctx->refs))
2901 * For SQ polling, the thread will do all submissions and completions.
2902 * Just return the requested submit count, and wake the thread if
2905 if (ctx->flags & IORING_SETUP_SQPOLL) {
2906 if (flags & IORING_ENTER_SQ_WAKEUP)
2907 wake_up(&ctx->sqo_wait);
2908 submitted = to_submit;
2914 to_submit = min(to_submit, ctx->sq_entries);
2916 mutex_lock(&ctx->uring_lock);
2917 submitted = io_ring_submit(ctx, to_submit);
2918 mutex_unlock(&ctx->uring_lock);
2920 if (flags & IORING_ENTER_GETEVENTS) {
2921 unsigned nr_events = 0;
2923 min_complete = min(min_complete, ctx->cq_entries);
2925 if (ctx->flags & IORING_SETUP_IOPOLL) {
2926 mutex_lock(&ctx->uring_lock);
2927 ret = io_iopoll_check(ctx, &nr_events, min_complete);
2928 mutex_unlock(&ctx->uring_lock);
2930 ret = io_cqring_wait(ctx, min_complete, sig, sigsz);
2935 io_ring_drop_ctx_refs(ctx, 1);
2938 return submitted ? submitted : ret;
2941 static const struct file_operations io_uring_fops = {
2942 .release = io_uring_release,
2943 .mmap = io_uring_mmap,
2944 .poll = io_uring_poll,
2945 .fasync = io_uring_fasync,
2948 static int io_allocate_scq_urings(struct io_ring_ctx *ctx,
2949 struct io_uring_params *p)
2951 struct io_sq_ring *sq_ring;
2952 struct io_cq_ring *cq_ring;
2955 sq_ring = io_mem_alloc(struct_size(sq_ring, array, p->sq_entries));
2959 ctx->sq_ring = sq_ring;
2960 sq_ring->ring_mask = p->sq_entries - 1;
2961 sq_ring->ring_entries = p->sq_entries;
2962 ctx->sq_mask = sq_ring->ring_mask;
2963 ctx->sq_entries = sq_ring->ring_entries;
2965 size = array_size(sizeof(struct io_uring_sqe), p->sq_entries);
2966 if (size == SIZE_MAX)
2969 ctx->sq_sqes = io_mem_alloc(size);
2973 cq_ring = io_mem_alloc(struct_size(cq_ring, cqes, p->cq_entries));
2977 ctx->cq_ring = cq_ring;
2978 cq_ring->ring_mask = p->cq_entries - 1;
2979 cq_ring->ring_entries = p->cq_entries;
2980 ctx->cq_mask = cq_ring->ring_mask;
2981 ctx->cq_entries = cq_ring->ring_entries;
2986 * Allocate an anonymous fd, this is what constitutes the application
2987 * visible backing of an io_uring instance. The application mmaps this
2988 * fd to gain access to the SQ/CQ ring details. If UNIX sockets are enabled,
2989 * we have to tie this fd to a socket for file garbage collection purposes.
2991 static int io_uring_get_fd(struct io_ring_ctx *ctx)
2996 #if defined(CONFIG_UNIX)
2997 ret = sock_create_kern(&init_net, PF_UNIX, SOCK_RAW, IPPROTO_IP,
3003 ret = get_unused_fd_flags(O_RDWR | O_CLOEXEC);
3007 file = anon_inode_getfile("[io_uring]", &io_uring_fops, ctx,
3008 O_RDWR | O_CLOEXEC);
3011 ret = PTR_ERR(file);
3015 #if defined(CONFIG_UNIX)
3016 ctx->ring_sock->file = file;
3017 ctx->ring_sock->sk->sk_user_data = ctx;
3019 fd_install(ret, file);
3022 #if defined(CONFIG_UNIX)
3023 sock_release(ctx->ring_sock);
3024 ctx->ring_sock = NULL;
3029 static int io_uring_create(unsigned entries, struct io_uring_params *p)
3031 struct user_struct *user = NULL;
3032 struct io_ring_ctx *ctx;
3036 if (!entries || entries > IORING_MAX_ENTRIES)
3040 * Use twice as many entries for the CQ ring. It's possible for the
3041 * application to drive a higher depth than the size of the SQ ring,
3042 * since the sqes are only used at submission time. This allows for
3043 * some flexibility in overcommitting a bit.
3045 p->sq_entries = roundup_pow_of_two(entries);
3046 p->cq_entries = 2 * p->sq_entries;
3048 user = get_uid(current_user());
3049 account_mem = !capable(CAP_IPC_LOCK);
3052 ret = io_account_mem(user,
3053 ring_pages(p->sq_entries, p->cq_entries));
3060 ctx = io_ring_ctx_alloc(p);
3063 io_unaccount_mem(user, ring_pages(p->sq_entries,
3068 ctx->compat = in_compat_syscall();
3069 ctx->account_mem = account_mem;
3072 ret = io_allocate_scq_urings(ctx, p);
3076 ret = io_sq_offload_start(ctx, p);
3080 ret = io_uring_get_fd(ctx);
3084 memset(&p->sq_off, 0, sizeof(p->sq_off));
3085 p->sq_off.head = offsetof(struct io_sq_ring, r.head);
3086 p->sq_off.tail = offsetof(struct io_sq_ring, r.tail);
3087 p->sq_off.ring_mask = offsetof(struct io_sq_ring, ring_mask);
3088 p->sq_off.ring_entries = offsetof(struct io_sq_ring, ring_entries);
3089 p->sq_off.flags = offsetof(struct io_sq_ring, flags);
3090 p->sq_off.dropped = offsetof(struct io_sq_ring, dropped);
3091 p->sq_off.array = offsetof(struct io_sq_ring, array);
3093 memset(&p->cq_off, 0, sizeof(p->cq_off));
3094 p->cq_off.head = offsetof(struct io_cq_ring, r.head);
3095 p->cq_off.tail = offsetof(struct io_cq_ring, r.tail);
3096 p->cq_off.ring_mask = offsetof(struct io_cq_ring, ring_mask);
3097 p->cq_off.ring_entries = offsetof(struct io_cq_ring, ring_entries);
3098 p->cq_off.overflow = offsetof(struct io_cq_ring, overflow);
3099 p->cq_off.cqes = offsetof(struct io_cq_ring, cqes);
3102 io_ring_ctx_wait_and_kill(ctx);
3107 * Sets up an aio uring context, and returns the fd. Applications asks for a
3108 * ring size, we return the actual sq/cq ring sizes (among other things) in the
3109 * params structure passed in.
3111 static long io_uring_setup(u32 entries, struct io_uring_params __user *params)
3113 struct io_uring_params p;
3117 if (copy_from_user(&p, params, sizeof(p)))
3119 for (i = 0; i < ARRAY_SIZE(p.resv); i++) {
3124 if (p.flags & ~(IORING_SETUP_IOPOLL | IORING_SETUP_SQPOLL |
3125 IORING_SETUP_SQ_AFF))
3128 ret = io_uring_create(entries, &p);
3132 if (copy_to_user(params, &p, sizeof(p)))
3138 SYSCALL_DEFINE2(io_uring_setup, u32, entries,
3139 struct io_uring_params __user *, params)
3141 return io_uring_setup(entries, params);
3144 static int __io_uring_register(struct io_ring_ctx *ctx, unsigned opcode,
3145 void __user *arg, unsigned nr_args)
3146 __releases(ctx->uring_lock)
3147 __acquires(ctx->uring_lock)
3152 * We're inside the ring mutex, if the ref is already dying, then
3153 * someone else killed the ctx or is already going through
3154 * io_uring_register().
3156 if (percpu_ref_is_dying(&ctx->refs))
3159 percpu_ref_kill(&ctx->refs);
3162 * Drop uring mutex before waiting for references to exit. If another
3163 * thread is currently inside io_uring_enter() it might need to grab
3164 * the uring_lock to make progress. If we hold it here across the drain
3165 * wait, then we can deadlock. It's safe to drop the mutex here, since
3166 * no new references will come in after we've killed the percpu ref.
3168 mutex_unlock(&ctx->uring_lock);
3169 wait_for_completion(&ctx->ctx_done);
3170 mutex_lock(&ctx->uring_lock);
3173 case IORING_REGISTER_BUFFERS:
3174 ret = io_sqe_buffer_register(ctx, arg, nr_args);
3176 case IORING_UNREGISTER_BUFFERS:
3180 ret = io_sqe_buffer_unregister(ctx);
3182 case IORING_REGISTER_FILES:
3183 ret = io_sqe_files_register(ctx, arg, nr_args);
3185 case IORING_UNREGISTER_FILES:
3189 ret = io_sqe_files_unregister(ctx);
3191 case IORING_REGISTER_EVENTFD:
3195 ret = io_eventfd_register(ctx, arg);
3197 case IORING_UNREGISTER_EVENTFD:
3201 ret = io_eventfd_unregister(ctx);
3208 /* bring the ctx back to life */
3209 reinit_completion(&ctx->ctx_done);
3210 percpu_ref_reinit(&ctx->refs);
3214 SYSCALL_DEFINE4(io_uring_register, unsigned int, fd, unsigned int, opcode,
3215 void __user *, arg, unsigned int, nr_args)
3217 struct io_ring_ctx *ctx;
3226 if (f.file->f_op != &io_uring_fops)
3229 ctx = f.file->private_data;
3231 mutex_lock(&ctx->uring_lock);
3232 ret = __io_uring_register(ctx, opcode, arg, nr_args);
3233 mutex_unlock(&ctx->uring_lock);
3239 static int __init io_uring_init(void)
3241 req_cachep = KMEM_CACHE(io_kiocb, SLAB_HWCACHE_ALIGN | SLAB_PANIC);
3244 __initcall(io_uring_init);