* supporting fast/efficient IO.
*
* A note on the read/write ordering memory barriers that are matched between
- * the application and kernel side. When the application reads the CQ ring
- * tail, it must use an appropriate smp_rmb() to order with the smp_wmb()
- * the kernel uses after writing the tail. Failure to do so could cause a
- * delay in when the application notices that completion events available.
- * This isn't a fatal condition. Likewise, the application must use an
- * appropriate smp_wmb() both before writing the SQ tail, and after writing
- * the SQ tail. The first one orders the sqe writes with the tail write, and
- * the latter is paired with the smp_rmb() the kernel will issue before
- * reading the SQ tail on submission.
+ * the application and kernel side.
+ *
+ * After the application reads the CQ ring tail, it must use an
+ * appropriate smp_rmb() to pair with the smp_wmb() the kernel uses
+ * before writing the tail (using smp_load_acquire to read the tail will
+ * do). It also needs a smp_mb() before updating CQ head (ordering the
+ * entry load(s) with the head store), pairing with an implicit barrier
+ * through a control-dependency in io_get_cqring (smp_store_release to
+ * store head will do). Failure to do so could lead to reading invalid
+ * CQ entries.
+ *
+ * Likewise, the application must use an appropriate smp_wmb() before
+ * writing the SQ tail (ordering SQ entry stores with the tail store),
+ * which pairs with smp_load_acquire in io_get_sqring (smp_store_release
+ * to store the tail will do). And it needs a barrier ordering the SQ
+ * head load before writing new SQ entries (smp_load_acquire to read
+ * head will do).
+ *
+ * When using the SQ poll thread (IORING_SETUP_SQPOLL), the application
+ * needs to check the SQ flags for IORING_SQ_NEED_WAKEUP *after*
+ * updating the SQ tail; a full memory barrier smp_mb() is needed
+ * between.
*
* Also see the examples in the liburing library:
*
u32 tail ____cacheline_aligned_in_smp;
};
+/*
+ * This data is shared with the application through the mmap at offset
+ * IORING_OFF_SQ_RING.
+ *
+ * The offsets to the member fields are published through struct
+ * io_sqring_offsets when calling io_uring_setup.
+ */
struct io_sq_ring {
+ /*
+ * Head and tail offsets into the ring; the offsets need to be
+ * masked to get valid indices.
+ *
+ * The kernel controls head and the application controls tail.
+ */
struct io_uring r;
+ /*
+ * Bitmask to apply to head and tail offsets (constant, equals
+ * ring_entries - 1)
+ */
u32 ring_mask;
+ /* Ring size (constant, power of 2) */
u32 ring_entries;
+ /*
+ * Number of invalid entries dropped by the kernel due to
+ * invalid index stored in array
+ *
+ * Written by the kernel, shouldn't be modified by the
+ * application (i.e. get number of "new events" by comparing to
+ * cached value).
+ *
+ * After a new SQ head value was read by the application this
+ * counter includes all submissions that were dropped reaching
+ * the new SQ head (and possibly more).
+ */
u32 dropped;
+ /*
+ * Runtime flags
+ *
+ * Written by the kernel, shouldn't be modified by the
+ * application.
+ *
+ * The application needs a full memory barrier before checking
+ * for IORING_SQ_NEED_WAKEUP after updating the sq tail.
+ */
u32 flags;
+ /*
+ * Ring buffer of indices into array of io_uring_sqe, which is
+ * mmapped by the application using the IORING_OFF_SQES offset.
+ *
+ * This indirection could e.g. be used to assign fixed
+ * io_uring_sqe entries to operations and only submit them to
+ * the queue when needed.
+ *
+ * The kernel modifies neither the indices array nor the entries
+ * array.
+ */
u32 array[];
};
+/*
+ * This data is shared with the application through the mmap at offset
+ * IORING_OFF_CQ_RING.
+ *
+ * The offsets to the member fields are published through struct
+ * io_cqring_offsets when calling io_uring_setup.
+ */
struct io_cq_ring {
+ /*
+ * Head and tail offsets into the ring; the offsets need to be
+ * masked to get valid indices.
+ *
+ * The application controls head and the kernel tail.
+ */
struct io_uring r;
+ /*
+ * Bitmask to apply to head and tail offsets (constant, equals
+ * ring_entries - 1)
+ */
u32 ring_mask;
+ /* Ring size (constant, power of 2) */
u32 ring_entries;
+ /*
+ * Number of completion events lost because the queue was full;
+ * this should be avoided by the application by making sure
+ * there are not more requests pending thatn there is space in
+ * the completion queue.
+ *
+ * Written by the kernel, shouldn't be modified by the
+ * application (i.e. get number of "new events" by comparing to
+ * cached value).
+ *
+ * As completion events come in out of order this counter is not
+ * ordered with any other data.
+ */
u32 overflow;
+ /*
+ * Ring buffer of completion events.
+ *
+ * The kernel writes completion events fresh every time they are
+ * produced, so the application is allowed to modify pending
+ * entries.
+ */
struct io_uring_cqe cqes[];
};
unsigned sq_mask;
unsigned sq_thread_idle;
struct io_uring_sqe *sq_sqes;
+
+ struct list_head defer_list;
} ____cacheline_aligned_in_smp;
/* IO offload */
struct task_struct *sqo_thread; /* if using sq thread polling */
struct mm_struct *sqo_mm;
wait_queue_head_t sqo_wait;
- unsigned sqo_stop;
struct {
/* CQ ring */
unsigned cq_mask;
struct wait_queue_head cq_wait;
struct fasync_struct *cq_fasync;
+ struct eventfd_ctx *cq_ev_fd;
} ____cacheline_aligned_in_smp;
/*
struct io_ring_ctx *ctx;
struct list_head list;
+ struct list_head link_list;
unsigned int flags;
refcount_t refs;
-#define REQ_F_FORCE_NONBLOCK 1 /* inline submission attempt */
+#define REQ_F_NOWAIT 1 /* must not punt to workers */
#define REQ_F_IOPOLL_COMPLETED 2 /* polled IO has completed */
#define REQ_F_FIXED_FILE 4 /* ctx owns file */
#define REQ_F_SEQ_PREV 8 /* sequential with previous */
-#define REQ_F_PREPPED 16 /* prep already done */
+#define REQ_F_IO_DRAIN 16 /* drain existing IO first */
+#define REQ_F_IO_DRAINED 32 /* drain done */
+#define REQ_F_LINK 64 /* linked sqes */
+#define REQ_F_FAIL_LINK 128 /* fail rest of links */
u64 user_data;
- u64 error;
+ u32 result;
+ u32 sequence;
struct work_struct work;
};
unsigned int ios_left;
};
+static void io_sq_wq_submit_work(struct work_struct *work);
+
static struct kmem_cache *req_cachep;
static const struct file_operations io_uring_fops;
spin_lock_init(&ctx->completion_lock);
INIT_LIST_HEAD(&ctx->poll_list);
INIT_LIST_HEAD(&ctx->cancel_list);
+ INIT_LIST_HEAD(&ctx->defer_list);
return ctx;
}
-static void io_commit_cqring(struct io_ring_ctx *ctx)
+static inline bool io_sequence_defer(struct io_ring_ctx *ctx,
+ struct io_kiocb *req)
+{
+ if ((req->flags & (REQ_F_IO_DRAIN|REQ_F_IO_DRAINED)) != REQ_F_IO_DRAIN)
+ return false;
+
+ return req->sequence > ctx->cached_cq_tail + ctx->sq_ring->dropped;
+}
+
+static struct io_kiocb *io_get_deferred_req(struct io_ring_ctx *ctx)
+{
+ struct io_kiocb *req;
+
+ if (list_empty(&ctx->defer_list))
+ return NULL;
+
+ req = list_first_entry(&ctx->defer_list, struct io_kiocb, list);
+ if (!io_sequence_defer(ctx, req)) {
+ list_del_init(&req->list);
+ return req;
+ }
+
+ return NULL;
+}
+
+static void __io_commit_cqring(struct io_ring_ctx *ctx)
{
struct io_cq_ring *ring = ctx->cq_ring;
/* order cqe stores with ring update */
smp_store_release(&ring->r.tail, ctx->cached_cq_tail);
- /*
- * Write sider barrier of tail update, app has read side. See
- * comment at the top of this file.
- */
- smp_wmb();
-
if (wq_has_sleeper(&ctx->cq_wait)) {
wake_up_interruptible(&ctx->cq_wait);
kill_fasync(&ctx->cq_fasync, SIGIO, POLL_IN);
}
}
+static void io_commit_cqring(struct io_ring_ctx *ctx)
+{
+ struct io_kiocb *req;
+
+ __io_commit_cqring(ctx);
+
+ while ((req = io_get_deferred_req(ctx)) != NULL) {
+ req->flags |= REQ_F_IO_DRAINED;
+ queue_work(ctx->sqo_wq, &req->work);
+ }
+}
+
static struct io_uring_cqe *io_get_cqring(struct io_ring_ctx *ctx)
{
struct io_cq_ring *ring = ctx->cq_ring;
unsigned tail;
tail = ctx->cached_cq_tail;
- /* See comment at the top of the file */
- smp_rmb();
+ /*
+ * writes to the cq entry need to come after reading head; the
+ * control dependency is enough as we're using WRITE_ONCE to
+ * fill the cq entry
+ */
if (tail - READ_ONCE(ring->r.head) == ring->ring_entries)
return NULL;
}
static void io_cqring_fill_event(struct io_ring_ctx *ctx, u64 ki_user_data,
- long res, unsigned ev_flags)
+ long res)
{
struct io_uring_cqe *cqe;
if (cqe) {
WRITE_ONCE(cqe->user_data, ki_user_data);
WRITE_ONCE(cqe->res, res);
- WRITE_ONCE(cqe->flags, ev_flags);
+ WRITE_ONCE(cqe->flags, 0);
} else {
unsigned overflow = READ_ONCE(ctx->cq_ring->overflow);
wake_up(&ctx->wait);
if (waitqueue_active(&ctx->sqo_wait))
wake_up(&ctx->sqo_wait);
+ if (ctx->cq_ev_fd)
+ eventfd_signal(ctx->cq_ev_fd, 1);
}
static void io_cqring_add_event(struct io_ring_ctx *ctx, u64 user_data,
- long res, unsigned ev_flags)
+ long res)
{
unsigned long flags;
spin_lock_irqsave(&ctx->completion_lock, flags);
- io_cqring_fill_event(ctx, user_data, res, ev_flags);
+ io_cqring_fill_event(ctx, user_data, res);
io_commit_cqring(ctx);
spin_unlock_irqrestore(&ctx->completion_lock, flags);
req->flags = 0;
/* one is dropped after submission, the other at completion */
refcount_set(&req->refs, 2);
+ req->result = 0;
return req;
out:
io_ring_drop_ctx_refs(ctx, 1);
}
}
-static void io_free_req(struct io_kiocb *req)
+static void __io_free_req(struct io_kiocb *req)
{
if (req->file && !(req->flags & REQ_F_FIXED_FILE))
fput(req->file);
kmem_cache_free(req_cachep, req);
}
+static void io_req_link_next(struct io_kiocb *req)
+{
+ struct io_kiocb *nxt;
+
+ /*
+ * The list should never be empty when we are called here. But could
+ * potentially happen if the chain is messed up, check to be on the
+ * safe side.
+ */
+ nxt = list_first_entry_or_null(&req->link_list, struct io_kiocb, list);
+ if (nxt) {
+ list_del(&nxt->list);
+ if (!list_empty(&req->link_list)) {
+ INIT_LIST_HEAD(&nxt->link_list);
+ list_splice(&req->link_list, &nxt->link_list);
+ nxt->flags |= REQ_F_LINK;
+ }
+
+ INIT_WORK(&nxt->work, io_sq_wq_submit_work);
+ queue_work(req->ctx->sqo_wq, &nxt->work);
+ }
+}
+
+/*
+ * Called if REQ_F_LINK is set, and we fail the head request
+ */
+static void io_fail_links(struct io_kiocb *req)
+{
+ struct io_kiocb *link;
+
+ while (!list_empty(&req->link_list)) {
+ link = list_first_entry(&req->link_list, struct io_kiocb, list);
+ list_del(&link->list);
+
+ io_cqring_add_event(req->ctx, link->user_data, -ECANCELED);
+ __io_free_req(link);
+ }
+}
+
+static void io_free_req(struct io_kiocb *req)
+{
+ /*
+ * If LINK is set, we have dependent requests in this chain. If we
+ * didn't fail this request, queue the first one up, moving any other
+ * dependencies to the next request. In case of failure, fail the rest
+ * of the chain.
+ */
+ if (req->flags & REQ_F_LINK) {
+ if (req->flags & REQ_F_FAIL_LINK)
+ io_fail_links(req);
+ else
+ io_req_link_next(req);
+ }
+
+ __io_free_req(req);
+}
+
static void io_put_req(struct io_kiocb *req)
{
if (refcount_dec_and_test(&req->refs))
req = list_first_entry(done, struct io_kiocb, list);
list_del(&req->list);
- io_cqring_fill_event(ctx, req->user_data, req->error, 0);
+ io_cqring_fill_event(ctx, req->user_data, req->result);
(*nr_events)++;
if (refcount_dec_and_test(&req->refs)) {
/* If we're not using fixed files, we have to pair the
* completion part with the file put. Use regular
* completions for those, only batch free for fixed
- * file.
+ * file and non-linked commands.
*/
- if (req->flags & REQ_F_FIXED_FILE) {
+ if ((req->flags & (REQ_F_FIXED_FILE|REQ_F_LINK)) ==
+ REQ_F_FIXED_FILE) {
reqs[to_free++] = req;
if (to_free == ARRAY_SIZE(reqs))
io_free_req_many(ctx, reqs, &to_free);
kiocb_end_write(kiocb);
- io_cqring_add_event(req->ctx, req->user_data, res, 0);
+ if ((req->flags & REQ_F_LINK) && res != req->result)
+ req->flags |= REQ_F_FAIL_LINK;
+ io_cqring_add_event(req->ctx, req->user_data, res);
io_put_req(req);
}
kiocb_end_write(kiocb);
- req->error = res;
+ if ((req->flags & REQ_F_LINK) && res != req->result)
+ req->flags |= REQ_F_FAIL_LINK;
+ req->result = res;
if (res != -EAGAIN)
req->flags |= REQ_F_IOPOLL_COMPLETED;
}
}
static int io_prep_rw(struct io_kiocb *req, const struct sqe_submit *s,
- bool force_nonblock, struct io_submit_state *state)
+ bool force_nonblock)
{
const struct io_uring_sqe *sqe = s->sqe;
struct io_ring_ctx *ctx = req->ctx;
if (!req->file)
return -EBADF;
- /* For -EAGAIN retry, everything is already prepped */
- if (req->flags & REQ_F_PREPPED)
- return 0;
if (force_nonblock && !io_file_supports_async(req->file))
force_nonblock = false;
ret = kiocb_set_rw_flags(kiocb, READ_ONCE(sqe->rw_flags));
if (unlikely(ret))
return ret;
- if (force_nonblock) {
+
+ /* don't allow async punt if RWF_NOWAIT was requested */
+ if (kiocb->ki_flags & IOCB_NOWAIT)
+ req->flags |= REQ_F_NOWAIT;
+
+ if (force_nonblock)
kiocb->ki_flags |= IOCB_NOWAIT;
- req->flags |= REQ_F_FORCE_NONBLOCK;
- }
+
if (ctx->flags & IORING_SETUP_IOPOLL) {
if (!(kiocb->ki_flags & IOCB_DIRECT) ||
!kiocb->ki_filp->f_op->iopoll)
return -EOPNOTSUPP;
- req->error = 0;
kiocb->ki_flags |= IOCB_HIPRI;
kiocb->ki_complete = io_complete_rw_iopoll;
} else {
return -EINVAL;
kiocb->ki_complete = io_complete_rw;
}
- req->flags |= REQ_F_PREPPED;
return 0;
}
return 0;
}
-static int io_import_iovec(struct io_ring_ctx *ctx, int rw,
- const struct sqe_submit *s, struct iovec **iovec,
- struct iov_iter *iter)
+static ssize_t io_import_iovec(struct io_ring_ctx *ctx, int rw,
+ const struct sqe_submit *s, struct iovec **iovec,
+ struct iov_iter *iter)
{
const struct io_uring_sqe *sqe = s->sqe;
void __user *buf = u64_to_user_ptr(READ_ONCE(sqe->addr));
opcode = READ_ONCE(sqe->opcode);
if (opcode == IORING_OP_READ_FIXED ||
opcode == IORING_OP_WRITE_FIXED) {
- int ret = io_import_fixed(ctx, rw, sqe, iter);
+ ssize_t ret = io_import_fixed(ctx, rw, sqe, iter);
*iovec = NULL;
return ret;
}
}
static int io_read(struct io_kiocb *req, const struct sqe_submit *s,
- bool force_nonblock, struct io_submit_state *state)
+ bool force_nonblock)
{
struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs;
struct kiocb *kiocb = &req->rw;
struct iov_iter iter;
struct file *file;
size_t iov_count;
- int ret;
+ ssize_t read_size, ret;
- ret = io_prep_rw(req, s, force_nonblock, state);
+ ret = io_prep_rw(req, s, force_nonblock);
if (ret)
return ret;
file = kiocb->ki_filp;
return -EINVAL;
ret = io_import_iovec(req->ctx, READ, s, &iovec, &iter);
- if (ret)
+ if (ret < 0)
return ret;
+ read_size = ret;
+ if (req->flags & REQ_F_LINK)
+ req->result = read_size;
+
iov_count = iov_iter_count(&iter);
ret = rw_verify_area(READ, file, &kiocb->ki_pos, iov_count);
if (!ret) {
ssize_t ret2;
- /* Catch -EAGAIN return for forced non-blocking submission */
ret2 = call_read_iter(file, kiocb, &iter);
+ /*
+ * In case of a short read, punt to async. This can happen
+ * if we have data partially cached. Alternatively we can
+ * return the short read, in which case the application will
+ * need to issue another SQE and wait for it. That SQE will
+ * need async punt anyway, so it's more efficient to do it
+ * here.
+ */
+ if (force_nonblock && ret2 > 0 && ret2 < read_size)
+ ret2 = -EAGAIN;
+ /* Catch -EAGAIN return for forced non-blocking submission */
if (!force_nonblock || ret2 != -EAGAIN) {
io_rw_done(kiocb, ret2);
} else {
}
static int io_write(struct io_kiocb *req, const struct sqe_submit *s,
- bool force_nonblock, struct io_submit_state *state)
+ bool force_nonblock)
{
struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs;
struct kiocb *kiocb = &req->rw;
struct iov_iter iter;
struct file *file;
size_t iov_count;
- int ret;
+ ssize_t ret;
- ret = io_prep_rw(req, s, force_nonblock, state);
+ ret = io_prep_rw(req, s, force_nonblock);
if (ret)
return ret;
return -EINVAL;
ret = io_import_iovec(req->ctx, WRITE, s, &iovec, &iter);
- if (ret)
+ if (ret < 0)
return ret;
+ if (req->flags & REQ_F_LINK)
+ req->result = ret;
+
iov_count = iov_iter_count(&iter);
ret = -EAGAIN;
if (unlikely(ctx->flags & IORING_SETUP_IOPOLL))
return -EINVAL;
- io_cqring_add_event(ctx, user_data, err, 0);
+ io_cqring_add_event(ctx, user_data, err);
io_put_req(req);
return 0;
}
if (!req->file)
return -EBADF;
- /* Prep already done (EAGAIN retry) */
- if (req->flags & REQ_F_PREPPED)
- return 0;
if (unlikely(ctx->flags & IORING_SETUP_IOPOLL))
return -EINVAL;
if (unlikely(sqe->addr || sqe->ioprio || sqe->buf_index))
return -EINVAL;
- req->flags |= REQ_F_PREPPED;
return 0;
}
end > 0 ? end : LLONG_MAX,
fsync_flags & IORING_FSYNC_DATASYNC);
- io_cqring_add_event(req->ctx, sqe->user_data, ret, 0);
+ if (ret < 0 && (req->flags & REQ_F_LINK))
+ req->flags |= REQ_F_FAIL_LINK;
+ io_cqring_add_event(req->ctx, sqe->user_data, ret);
io_put_req(req);
return 0;
}
+static int io_prep_sfr(struct io_kiocb *req, const struct io_uring_sqe *sqe)
+{
+ struct io_ring_ctx *ctx = req->ctx;
+ int ret = 0;
+
+ if (!req->file)
+ return -EBADF;
+
+ if (unlikely(ctx->flags & IORING_SETUP_IOPOLL))
+ return -EINVAL;
+ if (unlikely(sqe->addr || sqe->ioprio || sqe->buf_index))
+ return -EINVAL;
+
+ return ret;
+}
+
+static int io_sync_file_range(struct io_kiocb *req,
+ const struct io_uring_sqe *sqe,
+ bool force_nonblock)
+{
+ loff_t sqe_off;
+ loff_t sqe_len;
+ unsigned flags;
+ int ret;
+
+ ret = io_prep_sfr(req, sqe);
+ if (ret)
+ return ret;
+
+ /* sync_file_range always requires a blocking context */
+ if (force_nonblock)
+ return -EAGAIN;
+
+ sqe_off = READ_ONCE(sqe->off);
+ sqe_len = READ_ONCE(sqe->len);
+ flags = READ_ONCE(sqe->sync_range_flags);
+
+ ret = sync_file_range(req->rw.ki_filp, sqe_off, sqe_len, flags);
+
+ if (ret < 0 && (req->flags & REQ_F_LINK))
+ req->flags |= REQ_F_FAIL_LINK;
+ io_cqring_add_event(req->ctx, sqe->user_data, ret);
+ io_put_req(req);
+ return 0;
+}
+
+#if defined(CONFIG_NET)
+static int io_send_recvmsg(struct io_kiocb *req, const struct io_uring_sqe *sqe,
+ bool force_nonblock,
+ long (*fn)(struct socket *, struct user_msghdr __user *,
+ unsigned int))
+{
+ struct socket *sock;
+ int ret;
+
+ if (unlikely(req->ctx->flags & IORING_SETUP_IOPOLL))
+ return -EINVAL;
+
+ sock = sock_from_file(req->file, &ret);
+ if (sock) {
+ struct user_msghdr __user *msg;
+ unsigned flags;
+
+ flags = READ_ONCE(sqe->msg_flags);
+ if (flags & MSG_DONTWAIT)
+ req->flags |= REQ_F_NOWAIT;
+ else if (force_nonblock)
+ flags |= MSG_DONTWAIT;
+
+ msg = (struct user_msghdr __user *) (unsigned long)
+ READ_ONCE(sqe->addr);
+
+ ret = fn(sock, msg, flags);
+ if (force_nonblock && ret == -EAGAIN)
+ return ret;
+ }
+
+ io_cqring_add_event(req->ctx, sqe->user_data, ret);
+ io_put_req(req);
+ return 0;
+}
+#endif
+
+static int io_sendmsg(struct io_kiocb *req, const struct io_uring_sqe *sqe,
+ bool force_nonblock)
+{
+#if defined(CONFIG_NET)
+ return io_send_recvmsg(req, sqe, force_nonblock, __sys_sendmsg_sock);
+#else
+ return -EOPNOTSUPP;
+#endif
+}
+
+static int io_recvmsg(struct io_kiocb *req, const struct io_uring_sqe *sqe,
+ bool force_nonblock)
+{
+#if defined(CONFIG_NET)
+ return io_send_recvmsg(req, sqe, force_nonblock, __sys_recvmsg_sock);
+#else
+ return -EOPNOTSUPP;
+#endif
+}
+
static void io_poll_remove_one(struct io_kiocb *req)
{
struct io_poll_iocb *poll = &req->poll;
}
spin_unlock_irq(&ctx->completion_lock);
- io_cqring_add_event(req->ctx, sqe->user_data, ret, 0);
+ io_cqring_add_event(req->ctx, sqe->user_data, ret);
io_put_req(req);
return 0;
}
__poll_t mask)
{
req->poll.done = true;
- io_cqring_fill_event(ctx, req->user_data, mangle_poll(mask), 0);
+ io_cqring_fill_event(ctx, req->user_data, mangle_poll(mask));
io_commit_cqring(ctx);
}
spin_unlock(&poll->head->lock);
}
if (mask) { /* no async, we'd stolen it */
- req->error = mangle_poll(mask);
ipt.error = 0;
io_poll_complete(ctx, req, mask);
}
return ipt.error;
}
+static int io_req_defer(struct io_ring_ctx *ctx, struct io_kiocb *req,
+ const struct io_uring_sqe *sqe)
+{
+ struct io_uring_sqe *sqe_copy;
+
+ if (!io_sequence_defer(ctx, req) && list_empty(&ctx->defer_list))
+ return 0;
+
+ sqe_copy = kmalloc(sizeof(*sqe_copy), GFP_KERNEL);
+ if (!sqe_copy)
+ return -EAGAIN;
+
+ spin_lock_irq(&ctx->completion_lock);
+ if (!io_sequence_defer(ctx, req) && list_empty(&ctx->defer_list)) {
+ spin_unlock_irq(&ctx->completion_lock);
+ kfree(sqe_copy);
+ return 0;
+ }
+
+ memcpy(sqe_copy, sqe, sizeof(*sqe_copy));
+ req->submit.sqe = sqe_copy;
+
+ INIT_WORK(&req->work, io_sq_wq_submit_work);
+ list_add_tail(&req->list, &ctx->defer_list);
+ spin_unlock_irq(&ctx->completion_lock);
+ return -EIOCBQUEUED;
+}
+
static int __io_submit_sqe(struct io_ring_ctx *ctx, struct io_kiocb *req,
- const struct sqe_submit *s, bool force_nonblock,
- struct io_submit_state *state)
+ const struct sqe_submit *s, bool force_nonblock)
{
int ret, opcode;
+ req->user_data = READ_ONCE(s->sqe->user_data);
+
if (unlikely(s->index >= ctx->sq_entries))
return -EINVAL;
- req->user_data = READ_ONCE(s->sqe->user_data);
opcode = READ_ONCE(s->sqe->opcode);
switch (opcode) {
case IORING_OP_READV:
if (unlikely(s->sqe->buf_index))
return -EINVAL;
- ret = io_read(req, s, force_nonblock, state);
+ ret = io_read(req, s, force_nonblock);
break;
case IORING_OP_WRITEV:
if (unlikely(s->sqe->buf_index))
return -EINVAL;
- ret = io_write(req, s, force_nonblock, state);
+ ret = io_write(req, s, force_nonblock);
break;
case IORING_OP_READ_FIXED:
- ret = io_read(req, s, force_nonblock, state);
+ ret = io_read(req, s, force_nonblock);
break;
case IORING_OP_WRITE_FIXED:
- ret = io_write(req, s, force_nonblock, state);
+ ret = io_write(req, s, force_nonblock);
break;
case IORING_OP_FSYNC:
ret = io_fsync(req, s->sqe, force_nonblock);
case IORING_OP_POLL_REMOVE:
ret = io_poll_remove(req, s->sqe);
break;
+ case IORING_OP_SYNC_FILE_RANGE:
+ ret = io_sync_file_range(req, s->sqe, force_nonblock);
+ break;
+ case IORING_OP_SENDMSG:
+ ret = io_sendmsg(req, s->sqe, force_nonblock);
+ break;
+ case IORING_OP_RECVMSG:
+ ret = io_recvmsg(req, s->sqe, force_nonblock);
+ break;
default:
ret = -EINVAL;
break;
return ret;
if (ctx->flags & IORING_SETUP_IOPOLL) {
- if (req->error == -EAGAIN)
+ if (req->result == -EAGAIN)
return -EAGAIN;
/* workqueue context doesn't hold uring_lock, grab it now */
struct sqe_submit *s = &req->submit;
const struct io_uring_sqe *sqe = s->sqe;
- /* Ensure we clear previously set forced non-block flag */
- req->flags &= ~REQ_F_FORCE_NONBLOCK;
+ /* Ensure we clear previously set non-block flag */
req->rw.ki_flags &= ~IOCB_NOWAIT;
ret = 0;
s->has_user = cur_mm != NULL;
s->needs_lock = true;
do {
- ret = __io_submit_sqe(ctx, req, s, false, NULL);
+ ret = __io_submit_sqe(ctx, req, s, false);
/*
* We can get EAGAIN for polled IO even though
* we're forcing a sync submission from here,
break;
cond_resched();
} while (1);
-
- /* drop submission reference */
- io_put_req(req);
}
+
+ /* drop submission reference */
+ io_put_req(req);
+
if (ret) {
- io_cqring_add_event(ctx, sqe->user_data, ret, 0);
+ io_cqring_add_event(ctx, sqe->user_data, ret);
io_put_req(req);
}
flags = READ_ONCE(s->sqe->flags);
fd = READ_ONCE(s->sqe->fd);
+ if (flags & IOSQE_IO_DRAIN) {
+ req->flags |= REQ_F_IO_DRAIN;
+ req->sequence = ctx->cached_sq_head - 1;
+ }
+
if (!io_op_needs_file(s->sqe)) {
req->file = NULL;
return 0;
return 0;
}
-static int io_submit_sqe(struct io_ring_ctx *ctx, struct sqe_submit *s,
- struct io_submit_state *state)
+static int io_queue_sqe(struct io_ring_ctx *ctx, struct io_kiocb *req,
+ struct sqe_submit *s)
{
- struct io_kiocb *req;
int ret;
- /* enforce forwards compatibility on users */
- if (unlikely(s->sqe->flags & ~IOSQE_FIXED_FILE))
- return -EINVAL;
-
- req = io_get_req(ctx, state);
- if (unlikely(!req))
- return -EAGAIN;
-
- ret = io_req_set_file(ctx, s, state, req);
- if (unlikely(ret))
- goto out;
-
- ret = __io_submit_sqe(ctx, req, s, true, state);
- if (ret == -EAGAIN) {
+ ret = __io_submit_sqe(ctx, req, s, true);
+ if (ret == -EAGAIN && !(req->flags & REQ_F_NOWAIT)) {
struct io_uring_sqe *sqe_copy;
sqe_copy = kmalloc(sizeof(*sqe_copy), GFP_KERNEL);
/*
* Queued up for async execution, worker will release
- * submit reference when the iocb is actually
- * submitted.
+ * submit reference when the iocb is actually submitted.
*/
return 0;
}
}
-out:
/* drop submission reference */
io_put_req(req);
/* and drop final reference, if we failed */
- if (ret)
+ if (ret) {
+ io_cqring_add_event(ctx, req->user_data, ret);
+ if (req->flags & REQ_F_LINK)
+ req->flags |= REQ_F_FAIL_LINK;
io_put_req(req);
+ }
return ret;
}
+#define SQE_VALID_FLAGS (IOSQE_FIXED_FILE|IOSQE_IO_DRAIN|IOSQE_IO_LINK)
+
+static void io_submit_sqe(struct io_ring_ctx *ctx, struct sqe_submit *s,
+ struct io_submit_state *state, struct io_kiocb **link)
+{
+ struct io_uring_sqe *sqe_copy;
+ struct io_kiocb *req;
+ int ret;
+
+ /* enforce forwards compatibility on users */
+ if (unlikely(s->sqe->flags & ~SQE_VALID_FLAGS)) {
+ ret = -EINVAL;
+ goto err;
+ }
+
+ req = io_get_req(ctx, state);
+ if (unlikely(!req)) {
+ ret = -EAGAIN;
+ goto err;
+ }
+
+ ret = io_req_set_file(ctx, s, state, req);
+ if (unlikely(ret)) {
+err_req:
+ io_free_req(req);
+err:
+ io_cqring_add_event(ctx, s->sqe->user_data, ret);
+ return;
+ }
+
+ ret = io_req_defer(ctx, req, s->sqe);
+ if (ret) {
+ if (ret != -EIOCBQUEUED)
+ goto err_req;
+ return;
+ }
+
+ /*
+ * If we already have a head request, queue this one for async
+ * submittal once the head completes. If we don't have a head but
+ * IOSQE_IO_LINK is set in the sqe, start a new head. This one will be
+ * submitted sync once the chain is complete. If none of those
+ * conditions are true (normal request), then just queue it.
+ */
+ if (*link) {
+ struct io_kiocb *prev = *link;
+
+ sqe_copy = kmemdup(s->sqe, sizeof(*sqe_copy), GFP_KERNEL);
+ if (!sqe_copy) {
+ ret = -EAGAIN;
+ goto err_req;
+ }
+
+ s->sqe = sqe_copy;
+ memcpy(&req->submit, s, sizeof(*s));
+ list_add_tail(&req->list, &prev->link_list);
+ } else if (s->sqe->flags & IOSQE_IO_LINK) {
+ req->flags |= REQ_F_LINK;
+
+ memcpy(&req->submit, s, sizeof(*s));
+ INIT_LIST_HEAD(&req->link_list);
+ *link = req;
+ } else {
+ io_queue_sqe(ctx, req, s);
+ }
+}
+
/*
* Batched submission is done, ensure local IO is flushed out.
*/
* write new data to them.
*/
smp_store_release(&ring->r.head, ctx->cached_sq_head);
-
- /*
- * write side barrier of head update, app has read side. See
- * comment at the top of this file
- */
- smp_wmb();
}
}
/*
- * Undo last io_get_sqring()
- */
-static void io_drop_sqring(struct io_ring_ctx *ctx)
-{
- ctx->cached_sq_head--;
-}
-
-/*
* Fetch an sqe, if one is available. Note that s->sqe will point to memory
* that is mapped by userspace. This means that care needs to be taken to
* ensure that reads are stable, as we cannot rely on userspace always
* though the application is the one updating it.
*/
head = ctx->cached_sq_head;
- /* See comment at the top of this file */
- smp_rmb();
- if (head == READ_ONCE(ring->r.tail))
+ /* make sure SQ entry isn't read before tail */
+ if (head == smp_load_acquire(&ring->r.tail))
return false;
head = READ_ONCE(ring->array[head & ctx->sq_mask]);
/* drop invalid entries */
ctx->cached_sq_head++;
ring->dropped++;
- /* See comment at the top of this file */
- smp_wmb();
return false;
}
unsigned int nr, bool has_user, bool mm_fault)
{
struct io_submit_state state, *statep = NULL;
- int ret, i, submitted = 0;
+ struct io_kiocb *link = NULL;
+ bool prev_was_link = false;
+ int i, submitted = 0;
if (nr > IO_PLUG_THRESHOLD) {
io_submit_state_start(&state, ctx, nr);
}
for (i = 0; i < nr; i++) {
+ /*
+ * If previous wasn't linked and we have a linked command,
+ * that's the end of the chain. Submit the previous link.
+ */
+ if (!prev_was_link && link) {
+ io_queue_sqe(ctx, link, &link->submit);
+ link = NULL;
+ }
+ prev_was_link = (sqes[i].sqe->flags & IOSQE_IO_LINK) != 0;
+
if (unlikely(mm_fault)) {
- ret = -EFAULT;
+ io_cqring_add_event(ctx, sqes[i].sqe->user_data,
+ -EFAULT);
} else {
sqes[i].has_user = has_user;
sqes[i].needs_lock = true;
sqes[i].needs_fixed_file = true;
- ret = io_submit_sqe(ctx, &sqes[i], statep);
- }
- if (!ret) {
+ io_submit_sqe(ctx, &sqes[i], statep, &link);
submitted++;
- continue;
}
-
- io_cqring_add_event(ctx, sqes[i].sqe->user_data, ret, 0);
}
+ if (link)
+ io_queue_sqe(ctx, link, &link->submit);
if (statep)
io_submit_state_end(&state);
set_fs(USER_DS);
timeout = inflight = 0;
- while (!kthread_should_stop() && !ctx->sqo_stop) {
+ while (!kthread_should_park()) {
bool all_fixed, mm_fault = false;
int i;
/* Tell userspace we may need a wakeup call */
ctx->sq_ring->flags |= IORING_SQ_NEED_WAKEUP;
- smp_wmb();
+ /* make sure to read SQ tail after writing flags */
+ smp_mb();
if (!io_get_sqring(ctx, &sqes[0])) {
- if (kthread_should_stop()) {
+ if (kthread_should_park()) {
finish_wait(&ctx->sqo_wait, &wait);
break;
}
finish_wait(&ctx->sqo_wait, &wait);
ctx->sq_ring->flags &= ~IORING_SQ_NEED_WAKEUP;
- smp_wmb();
continue;
}
finish_wait(&ctx->sqo_wait, &wait);
ctx->sq_ring->flags &= ~IORING_SQ_NEED_WAKEUP;
- smp_wmb();
}
i = 0;
mmput(cur_mm);
}
- if (kthread_should_park())
- kthread_parkme();
+ kthread_parkme();
return 0;
}
static int io_ring_submit(struct io_ring_ctx *ctx, unsigned int to_submit)
{
struct io_submit_state state, *statep = NULL;
- int i, ret = 0, submit = 0;
+ struct io_kiocb *link = NULL;
+ bool prev_was_link = false;
+ int i, submit = 0;
if (to_submit > IO_PLUG_THRESHOLD) {
io_submit_state_start(&state, ctx, to_submit);
if (!io_get_sqring(ctx, &s))
break;
+ /*
+ * If previous wasn't linked and we have a linked command,
+ * that's the end of the chain. Submit the previous link.
+ */
+ if (!prev_was_link && link) {
+ io_queue_sqe(ctx, link, &link->submit);
+ link = NULL;
+ }
+ prev_was_link = (s.sqe->flags & IOSQE_IO_LINK) != 0;
+
s.has_user = true;
s.needs_lock = false;
s.needs_fixed_file = false;
-
- ret = io_submit_sqe(ctx, &s, statep);
- if (ret) {
- io_drop_sqring(ctx);
- break;
- }
-
submit++;
+ io_submit_sqe(ctx, &s, statep, &link);
}
io_commit_sqring(ctx);
+ if (link)
+ io_queue_sqe(ctx, link, &link->submit);
if (statep)
io_submit_state_end(statep);
- return submit ? submit : ret;
+ return submit;
}
static unsigned io_cqring_events(struct io_cq_ring *ring)
{
+ /* See comment at the top of this file */
+ smp_rmb();
return READ_ONCE(ring->r.tail) - READ_ONCE(ring->r.head);
}
{
struct io_cq_ring *ring = ctx->cq_ring;
sigset_t ksigmask, sigsaved;
- DEFINE_WAIT(wait);
int ret;
- /* See comment at the top of this file */
- smp_rmb();
if (io_cqring_events(ring) >= min_events)
return 0;
return ret;
}
- do {
- prepare_to_wait(&ctx->wait, &wait, TASK_INTERRUPTIBLE);
-
- ret = 0;
- /* See comment at the top of this file */
- smp_rmb();
- if (io_cqring_events(ring) >= min_events)
- break;
-
- schedule();
-
+ ret = wait_event_interruptible(ctx->wait, io_cqring_events(ring) >= min_events);
+ if (ret == -ERESTARTSYS)
ret = -EINTR;
- if (signal_pending(current))
- break;
- } while (1);
-
- finish_wait(&ctx->wait, &wait);
if (sig)
restore_user_sigmask(sig, &sigsaved);
static void io_sq_thread_stop(struct io_ring_ctx *ctx)
{
if (ctx->sqo_thread) {
- ctx->sqo_stop = 1;
- mb();
+ /*
+ * The park is a bit of a work-around, without it we get
+ * warning spews on shutdown with SQPOLL set and affinity
+ * set to a single CPU.
+ */
kthread_park(ctx->sqo_thread);
kthread_stop(ctx->sqo_thread);
ctx->sqo_thread = NULL;
left = ctx->nr_user_files;
while (left) {
unsigned this_files = min_t(unsigned, left, SCM_MAX_FD);
- int ret;
ret = __io_sqe_files_scm(ctx, this_files, total);
if (ret)
mmgrab(current->mm);
ctx->sqo_mm = current->mm;
- ret = -EINVAL;
- if (!cpu_possible(p->sq_thread_cpu))
- goto err;
-
if (ctx->flags & IORING_SETUP_SQPOLL) {
ret = -EPERM;
if (!capable(CAP_SYS_ADMIN))
ctx->sq_thread_idle = HZ;
if (p->flags & IORING_SETUP_SQ_AFF) {
- int cpu;
+ int cpu = p->sq_thread_cpu;
- cpu = array_index_nospec(p->sq_thread_cpu, NR_CPUS);
ret = -EINVAL;
- if (!cpu_possible(p->sq_thread_cpu))
+ if (cpu >= nr_cpu_ids)
+ goto err;
+ if (!cpu_online(cpu))
goto err;
ctx->sqo_thread = kthread_create_on_cpu(io_sq_thread,
static void io_mem_free(void *ptr)
{
- struct page *page = virt_to_head_page(ptr);
+ struct page *page;
+
+ if (!ptr)
+ return;
+ page = virt_to_head_page(ptr);
if (put_page_testzero(page))
free_compound_page(page);
}
if (ctx->account_mem)
io_unaccount_mem(ctx->user, imu->nr_bvecs);
- kfree(imu->bvec);
+ kvfree(imu->bvec);
imu->nr_bvecs = 0;
}
ret = io_copy_iov(ctx, &iov, arg, i);
if (ret)
- break;
+ goto err;
/*
* Don't impose further limits on the size and buffer
if (!pages || nr_pages > got_pages) {
kfree(vmas);
kfree(pages);
- pages = kmalloc_array(nr_pages, sizeof(struct page *),
+ pages = kvmalloc_array(nr_pages, sizeof(struct page *),
GFP_KERNEL);
- vmas = kmalloc_array(nr_pages,
+ vmas = kvmalloc_array(nr_pages,
sizeof(struct vm_area_struct *),
GFP_KERNEL);
if (!pages || !vmas) {
got_pages = nr_pages;
}
- imu->bvec = kmalloc_array(nr_pages, sizeof(struct bio_vec),
+ imu->bvec = kvmalloc_array(nr_pages, sizeof(struct bio_vec),
GFP_KERNEL);
ret = -ENOMEM;
if (!imu->bvec) {
ret = 0;
down_read(¤t->mm->mmap_sem);
- pret = get_user_pages_longterm(ubuf, nr_pages, FOLL_WRITE,
- pages, vmas);
+ pret = get_user_pages(ubuf, nr_pages,
+ FOLL_WRITE | FOLL_LONGTERM,
+ pages, vmas);
if (pret == nr_pages) {
/* don't support file backed memory */
for (j = 0; j < nr_pages; j++) {
}
if (ctx->account_mem)
io_unaccount_mem(ctx->user, nr_pages);
+ kvfree(imu->bvec);
goto err;
}
ctx->nr_user_bufs++;
}
- kfree(pages);
- kfree(vmas);
+ kvfree(pages);
+ kvfree(vmas);
return 0;
err:
- kfree(pages);
- kfree(vmas);
+ kvfree(pages);
+ kvfree(vmas);
io_sqe_buffer_unregister(ctx);
return ret;
}
+static int io_eventfd_register(struct io_ring_ctx *ctx, void __user *arg)
+{
+ __s32 __user *fds = arg;
+ int fd;
+
+ if (ctx->cq_ev_fd)
+ return -EBUSY;
+
+ if (copy_from_user(&fd, fds, sizeof(*fds)))
+ return -EFAULT;
+
+ ctx->cq_ev_fd = eventfd_ctx_fdget(fd);
+ if (IS_ERR(ctx->cq_ev_fd)) {
+ int ret = PTR_ERR(ctx->cq_ev_fd);
+ ctx->cq_ev_fd = NULL;
+ return ret;
+ }
+
+ return 0;
+}
+
+static int io_eventfd_unregister(struct io_ring_ctx *ctx)
+{
+ if (ctx->cq_ev_fd) {
+ eventfd_ctx_put(ctx->cq_ev_fd);
+ ctx->cq_ev_fd = NULL;
+ return 0;
+ }
+
+ return -ENXIO;
+}
+
static void io_ring_ctx_free(struct io_ring_ctx *ctx)
{
io_finish_async(ctx);
io_iopoll_reap_events(ctx);
io_sqe_buffer_unregister(ctx);
io_sqe_files_unregister(ctx);
+ io_eventfd_unregister(ctx);
#if defined(CONFIG_UNIX)
if (ctx->ring_sock)
__poll_t mask = 0;
poll_wait(file, &ctx->cq_wait, wait);
- /* See comment at the top of this file */
+ /*
+ * synchronizes with barrier from wq_has_sleeper call in
+ * io_commit_cqring
+ */
smp_rmb();
- if (READ_ONCE(ctx->sq_ring->r.tail) + 1 != ctx->cached_sq_head)
+ if (READ_ONCE(ctx->sq_ring->r.tail) - ctx->cached_sq_head !=
+ ctx->sq_ring->ring_entries)
mask |= EPOLLOUT | EPOLLWRNORM;
if (READ_ONCE(ctx->cq_ring->r.head) != ctx->cached_cq_tail)
mask |= EPOLLIN | EPOLLRDNORM;
mutex_lock(&ctx->uring_lock);
submitted = io_ring_submit(ctx, to_submit);
mutex_unlock(&ctx->uring_lock);
-
- if (submitted < 0)
- goto out_ctx;
}
if (flags & IORING_ENTER_GETEVENTS) {
unsigned nr_events = 0;
min_complete = min(min_complete, ctx->cq_entries);
- /*
- * The application could have included the 'to_submit' count
- * in how many events it wanted to wait for. If we failed to
- * submit the desired count, we may need to adjust the number
- * of events to poll/wait for.
- */
- if (submitted < to_submit)
- min_complete = min_t(unsigned, submitted, min_complete);
-
if (ctx->flags & IORING_SETUP_IOPOLL) {
mutex_lock(&ctx->uring_lock);
ret = io_iopoll_check(ctx, &nr_events, min_complete);
return -EOVERFLOW;
ctx->sq_sqes = io_mem_alloc(size);
- if (!ctx->sq_sqes) {
- io_mem_free(ctx->sq_ring);
+ if (!ctx->sq_sqes)
return -ENOMEM;
- }
cq_ring = io_mem_alloc(struct_size(cq_ring, cqes, p->cq_entries));
- if (!cq_ring) {
- io_mem_free(ctx->sq_ring);
- io_mem_free(ctx->sq_sqes);
+ if (!cq_ring)
return -ENOMEM;
- }
ctx->cq_ring = cq_ring;
cq_ring->ring_mask = p->cq_entries - 1;
{
int ret;
+ /*
+ * We're inside the ring mutex, if the ref is already dying, then
+ * someone else killed the ctx or is already going through
+ * io_uring_register().
+ */
+ if (percpu_ref_is_dying(&ctx->refs))
+ return -ENXIO;
+
percpu_ref_kill(&ctx->refs);
/*
break;
ret = io_sqe_files_unregister(ctx);
break;
+ case IORING_REGISTER_EVENTFD:
+ ret = -EINVAL;
+ if (nr_args != 1)
+ break;
+ ret = io_eventfd_register(ctx, arg);
+ break;
+ case IORING_UNREGISTER_EVENTFD:
+ ret = -EINVAL;
+ if (arg || nr_args)
+ break;
+ ret = io_eventfd_unregister(ctx);
+ break;
default:
ret = -EINVAL;
break;
OSDN Git Service
