[linux-2.6-block.git] / io_uring / io_uring.h

#ifndef IOU_CORE_H
#define IOU_CORE_H

#include <linux/errno.h>
#include <linux/lockdep.h>
#include <linux/resume_user_mode.h>
#include <linux/kasan.h>
#include <linux/poll.h>
#include <linux/io_uring_types.h>
#include <uapi/linux/eventpoll.h>
#include "io-wq.h"
#include "slist.h"
#include "filetable.h"

#ifndef CREATE_TRACE_POINTS
#include <trace/events/io_uring.h>
#endif

enum {
	IOU_OK			= 0,
	IOU_ISSUE_SKIP_COMPLETE	= -EIOCBQUEUED,

	/*
	 * Requeue the task_work to restart operations on this request. The
	 * actual value isn't important, should just be not an otherwise
	 * valid error code, yet less than -MAX_ERRNO and valid internally.
	 */
	IOU_REQUEUE		= -3072,

	/*
	 * Intended only when both IO_URING_F_MULTISHOT is passed
	 * to indicate to the poll runner that multishot should be
	 * removed and the result is set on req->cqe.res.
	 */
	IOU_STOP_MULTISHOT	= -ECANCELED,
};

struct io_wait_queue {
	struct wait_queue_entry wq;
	struct io_ring_ctx *ctx;
	unsigned cq_tail;
	unsigned cq_min_tail;
	unsigned nr_timeouts;
	int hit_timeout;
	ktime_t min_timeout;
	ktime_t timeout;
	struct hrtimer t;

#ifdef CONFIG_NET_RX_BUSY_POLL
	ktime_t napi_busy_poll_dt;
	bool napi_prefer_busy_poll;
#endif
};

static inline bool io_should_wake(struct io_wait_queue *iowq)
{
	struct io_ring_ctx *ctx = iowq->ctx;
	int dist = READ_ONCE(ctx->rings->cq.tail) - (int) iowq->cq_tail;

	/*
	 * Wake up if we have enough events, or if a timeout occurred since we
	 * started waiting. For timeouts, we always want to return to userspace,
	 * regardless of event count.
	 */
	return dist >= 0 || atomic_read(&ctx->cq_timeouts) != iowq->nr_timeouts;
}

bool io_cqe_cache_refill(struct io_ring_ctx *ctx, bool overflow);
int io_run_task_work_sig(struct io_ring_ctx *ctx);
void io_req_defer_failed(struct io_kiocb *req, s32 res);
bool io_post_aux_cqe(struct io_ring_ctx *ctx, u64 user_data, s32 res, u32 cflags);
void io_add_aux_cqe(struct io_ring_ctx *ctx, u64 user_data, s32 res, u32 cflags);
bool io_req_post_cqe(struct io_kiocb *req, s32 res, u32 cflags);
void __io_commit_cqring_flush(struct io_ring_ctx *ctx);

struct file *io_file_get_normal(struct io_kiocb *req, int fd);
struct file *io_file_get_fixed(struct io_kiocb *req, int fd,
			       unsigned issue_flags);

void __io_req_task_work_add(struct io_kiocb *req, unsigned flags);
void io_req_task_work_add_remote(struct io_kiocb *req, struct io_ring_ctx *ctx,
				 unsigned flags);
bool io_alloc_async_data(struct io_kiocb *req);
void io_req_task_queue(struct io_kiocb *req);
void io_req_task_complete(struct io_kiocb *req, struct io_tw_state *ts);
void io_req_task_queue_fail(struct io_kiocb *req, int ret);
void io_req_task_submit(struct io_kiocb *req, struct io_tw_state *ts);
struct llist_node *io_handle_tw_list(struct llist_node *node, unsigned int *count, unsigned int max_entries);
struct llist_node *tctx_task_work_run(struct io_uring_task *tctx, unsigned int max_entries, unsigned int *count);
void tctx_task_work(struct callback_head *cb);
__cold void io_uring_cancel_generic(bool cancel_all, struct io_sq_data *sqd);
int io_uring_alloc_task_context(struct task_struct *task,
				struct io_ring_ctx *ctx);

int io_ring_add_registered_file(struct io_uring_task *tctx, struct file *file,
				     int start, int end);
void io_req_queue_iowq(struct io_kiocb *req);

int io_poll_issue(struct io_kiocb *req, struct io_tw_state *ts);
int io_submit_sqes(struct io_ring_ctx *ctx, unsigned int nr);
int io_do_iopoll(struct io_ring_ctx *ctx, bool force_nonspin);
void __io_submit_flush_completions(struct io_ring_ctx *ctx);

struct io_wq_work *io_wq_free_work(struct io_wq_work *work);
void io_wq_submit_work(struct io_wq_work *work);

void io_free_req(struct io_kiocb *req);
void io_queue_next(struct io_kiocb *req);
void io_task_refs_refill(struct io_uring_task *tctx);
bool __io_alloc_req_refill(struct io_ring_ctx *ctx);

bool io_match_task_safe(struct io_kiocb *head, struct task_struct *task,
			bool cancel_all);

void io_activate_pollwq(struct io_ring_ctx *ctx);

static inline void io_lockdep_assert_cq_locked(struct io_ring_ctx *ctx)
{
#if defined(CONFIG_PROVE_LOCKING)
	lockdep_assert(in_task());

	if (ctx->flags & IORING_SETUP_IOPOLL) {
		lockdep_assert_held(&ctx->uring_lock);
	} else if (!ctx->task_complete) {
		lockdep_assert_held(&ctx->completion_lock);
	} else if (ctx->submitter_task) {
		/*
		 * ->submitter_task may be NULL and we can still post a CQE,
		 * if the ring has been setup with IORING_SETUP_R_DISABLED.
		 * Not from an SQE, as those cannot be submitted, but via
		 * updating tagged resources.
		 */
		if (ctx->submitter_task->flags & PF_EXITING)
			lockdep_assert(current_work());
		else
			lockdep_assert(current == ctx->submitter_task);
	}
#endif
}

static inline void io_req_task_work_add(struct io_kiocb *req)
{
	__io_req_task_work_add(req, 0);
}

static inline void io_submit_flush_completions(struct io_ring_ctx *ctx)
{
	if (!wq_list_empty(&ctx->submit_state.compl_reqs) ||
	    ctx->submit_state.cq_flush)
		__io_submit_flush_completions(ctx);
}

#define io_for_each_link(pos, head) \
	for (pos = (head); pos; pos = pos->link)

static inline bool io_get_cqe_overflow(struct io_ring_ctx *ctx,
					struct io_uring_cqe **ret,
					bool overflow)
{
	io_lockdep_assert_cq_locked(ctx);

	if (unlikely(ctx->cqe_cached >= ctx->cqe_sentinel)) {
		if (unlikely(!io_cqe_cache_refill(ctx, overflow)))
			return false;
	}
	*ret = ctx->cqe_cached;
	ctx->cached_cq_tail++;
	ctx->cqe_cached++;
	if (ctx->flags & IORING_SETUP_CQE32)
		ctx->cqe_cached++;
	return true;
}

static inline bool io_get_cqe(struct io_ring_ctx *ctx, struct io_uring_cqe **ret)
{
	return io_get_cqe_overflow(ctx, ret, false);
}

static __always_inline bool io_fill_cqe_req(struct io_ring_ctx *ctx,
					    struct io_kiocb *req)
{
	struct io_uring_cqe *cqe;

	/*
	 * If we can't get a cq entry, userspace overflowed the
	 * submission (by quite a lot). Increment the overflow count in
	 * the ring.
	 */
	if (unlikely(!io_get_cqe(ctx, &cqe)))
		return false;

	if (trace_io_uring_complete_enabled())
		trace_io_uring_complete(req->ctx, req, req->cqe.user_data,
					req->cqe.res, req->cqe.flags,
					req->big_cqe.extra1, req->big_cqe.extra2);

	memcpy(cqe, &req->cqe, sizeof(*cqe));
	if (ctx->flags & IORING_SETUP_CQE32) {
		memcpy(cqe->big_cqe, &req->big_cqe, sizeof(*cqe));
		memset(&req->big_cqe, 0, sizeof(req->big_cqe));
	}
	return true;
}

static inline void req_set_fail(struct io_kiocb *req)
{
	req->flags |= REQ_F_FAIL;
	if (req->flags & REQ_F_CQE_SKIP) {
		req->flags &= ~REQ_F_CQE_SKIP;
		req->flags |= REQ_F_SKIP_LINK_CQES;
	}
}

static inline void io_req_set_res(struct io_kiocb *req, s32 res, u32 cflags)
{
	req->cqe.res = res;
	req->cqe.flags = cflags;
}

static inline bool req_has_async_data(struct io_kiocb *req)
{
	return req->flags & REQ_F_ASYNC_DATA;
}

static inline void io_put_file(struct io_kiocb *req)
{
	if (!(req->flags & REQ_F_FIXED_FILE) && req->file)
		fput(req->file);
}

static inline void io_ring_submit_unlock(struct io_ring_ctx *ctx,
					 unsigned issue_flags)
{
	lockdep_assert_held(&ctx->uring_lock);
	if (unlikely(issue_flags & IO_URING_F_UNLOCKED))
		mutex_unlock(&ctx->uring_lock);
}

static inline void io_ring_submit_lock(struct io_ring_ctx *ctx,
				       unsigned issue_flags)
{
	/*
	 * "Normal" inline submissions always hold the uring_lock, since we
	 * grab it from the system call. Same is true for the SQPOLL offload.
	 * The only exception is when we've detached the request and issue it
	 * from an async worker thread, grab the lock for that case.
	 */
	if (unlikely(issue_flags & IO_URING_F_UNLOCKED))
		mutex_lock(&ctx->uring_lock);
	lockdep_assert_held(&ctx->uring_lock);
}

static inline void io_commit_cqring(struct io_ring_ctx *ctx)
{
	/* order cqe stores with ring update */
	smp_store_release(&ctx->rings->cq.tail, ctx->cached_cq_tail);
}

static inline void io_poll_wq_wake(struct io_ring_ctx *ctx)
{
	if (wq_has_sleeper(&ctx->poll_wq))
		__wake_up(&ctx->poll_wq, TASK_NORMAL, 0,
				poll_to_key(EPOLL_URING_WAKE | EPOLLIN));
}

static inline void io_cqring_wake(struct io_ring_ctx *ctx)
{
	/*
	 * Trigger waitqueue handler on all waiters on our waitqueue. This
	 * won't necessarily wake up all the tasks, io_should_wake() will make
	 * that decision.
	 *
	 * Pass in EPOLLIN|EPOLL_URING_WAKE as the poll wakeup key. The latter
	 * set in the mask so that if we recurse back into our own poll
	 * waitqueue handlers, we know we have a dependency between eventfd or
	 * epoll and should terminate multishot poll at that point.
	 */
	if (wq_has_sleeper(&ctx->cq_wait))
		__wake_up(&ctx->cq_wait, TASK_NORMAL, 0,
				poll_to_key(EPOLL_URING_WAKE | EPOLLIN));
}

static inline bool io_sqring_full(struct io_ring_ctx *ctx)
{
	struct io_rings *r = ctx->rings;

	return READ_ONCE(r->sq.tail) - ctx->cached_sq_head == ctx->sq_entries;
}

static inline unsigned int io_sqring_entries(struct io_ring_ctx *ctx)
{
	struct io_rings *rings = ctx->rings;
	unsigned int entries;

	/* make sure SQ entry isn't read before tail */
	entries = smp_load_acquire(&rings->sq.tail) - ctx->cached_sq_head;
	return min(entries, ctx->sq_entries);
}

static inline int io_run_task_work(void)
{
	bool ret = false;

	/*
	 * Always check-and-clear the task_work notification signal. With how
	 * signaling works for task_work, we can find it set with nothing to
	 * run. We need to clear it for that case, like get_signal() does.
	 */
	if (test_thread_flag(TIF_NOTIFY_SIGNAL))
		clear_notify_signal();
	/*
	 * PF_IO_WORKER never returns to userspace, so check here if we have
	 * notify work that needs processing.
	 */
	if (current->flags & PF_IO_WORKER) {
		if (test_thread_flag(TIF_NOTIFY_RESUME)) {
			__set_current_state(TASK_RUNNING);
			resume_user_mode_work(NULL);
		}
		if (current->io_uring) {
			unsigned int count = 0;

			tctx_task_work_run(current->io_uring, UINT_MAX, &count);
			if (count)
				ret = true;
		}
	}
	if (task_work_pending(current)) {
		__set_current_state(TASK_RUNNING);
		task_work_run();
		ret = true;
	}

	return ret;
}

static inline bool io_task_work_pending(struct io_ring_ctx *ctx)
{
	return task_work_pending(current) || !llist_empty(&ctx->work_llist);
}

static inline void io_tw_lock(struct io_ring_ctx *ctx, struct io_tw_state *ts)
{
	lockdep_assert_held(&ctx->uring_lock);
}

/*
 * Don't complete immediately but use deferred completion infrastructure.
 * Protected by ->uring_lock and can only be used either with
 * IO_URING_F_COMPLETE_DEFER or inside a tw handler holding the mutex.
 */
static inline void io_req_complete_defer(struct io_kiocb *req)
	__must_hold(&req->ctx->uring_lock)
{
	struct io_submit_state *state = &req->ctx->submit_state;

	lockdep_assert_held(&req->ctx->uring_lock);

	wq_list_add_tail(&req->comp_list, &state->compl_reqs);
}

static inline void io_commit_cqring_flush(struct io_ring_ctx *ctx)
{
	if (unlikely(ctx->off_timeout_used || ctx->drain_active ||
		     ctx->has_evfd || ctx->poll_activated))
		__io_commit_cqring_flush(ctx);
}

static inline void io_get_task_refs(int nr)
{
	struct io_uring_task *tctx = current->io_uring;

	tctx->cached_refs -= nr;
	if (unlikely(tctx->cached_refs < 0))
		io_task_refs_refill(tctx);
}

static inline bool io_req_cache_empty(struct io_ring_ctx *ctx)
{
	return !ctx->submit_state.free_list.next;
}

extern struct kmem_cache *req_cachep;
extern struct kmem_cache *io_buf_cachep;

static inline struct io_kiocb *io_extract_req(struct io_ring_ctx *ctx)
{
	struct io_kiocb *req;

	req = container_of(ctx->submit_state.free_list.next, struct io_kiocb, comp_list);
	wq_stack_extract(&ctx->submit_state.free_list);
	return req;
}

static inline bool io_alloc_req(struct io_ring_ctx *ctx, struct io_kiocb **req)
{
	if (unlikely(io_req_cache_empty(ctx))) {
		if (!__io_alloc_req_refill(ctx))
			return false;
	}
	*req = io_extract_req(ctx);
	return true;
}

static inline bool io_allowed_defer_tw_run(struct io_ring_ctx *ctx)
{
	return likely(ctx->submitter_task == current);
}

static inline bool io_allowed_run_tw(struct io_ring_ctx *ctx)
{
	return likely(!(ctx->flags & IORING_SETUP_DEFER_TASKRUN) ||
		      ctx->submitter_task == current);
}

static inline void io_req_queue_tw_complete(struct io_kiocb *req, s32 res)
{
	io_req_set_res(req, res, 0);
	req->io_task_work.func = io_req_task_complete;
	io_req_task_work_add(req);
}

/*
 * IORING_SETUP_SQE128 contexts allocate twice the normal SQE size for each
 * slot.
 */
static inline size_t uring_sqe_size(struct io_ring_ctx *ctx)
{
	if (ctx->flags & IORING_SETUP_SQE128)
		return 2 * sizeof(struct io_uring_sqe);
	return sizeof(struct io_uring_sqe);
}

static inline bool io_file_can_poll(struct io_kiocb *req)
{
	if (req->flags & REQ_F_CAN_POLL)
		return true;
	if (req->file && file_can_poll(req->file)) {
		req->flags |= REQ_F_CAN_POLL;
		return true;
	}
	return false;
}

static inline ktime_t io_get_time(struct io_ring_ctx *ctx)
{
	if (ctx->clockid == CLOCK_MONOTONIC)
		return ktime_get();

	return ktime_get_with_offset(ctx->clock_offset);
}

enum {
	IO_CHECK_CQ_OVERFLOW_BIT,
	IO_CHECK_CQ_DROPPED_BIT,
};

static inline bool io_has_work(struct io_ring_ctx *ctx)
{
	return test_bit(IO_CHECK_CQ_OVERFLOW_BIT, &ctx->check_cq) ||
	       !llist_empty(&ctx->work_llist);
}
#endif
Commit	Line	Data
de23077e JA	1	#ifndef IOU_CORE_H
	2	#define IOU_CORE_H
	3
	4	#include <linux/errno.h>
cd40cae2	5	#include <linux/lockdep.h>
b5d3ae20	6	#include <linux/resume_user_mode.h>
c1755c25	7	#include <linux/kasan.h>
95041b93	8	#include <linux/poll.h>
ab1c84d8	9	#include <linux/io_uring_types.h>
44648532	10	#include <uapi/linux/eventpoll.h>
ab1c84d8	11	#include "io-wq.h"
a6b21fbb	12	#include "slist.h"
ab1c84d8	13	#include "filetable.h"
de23077e	14
f3b44f92 JA	15	#ifndef CREATE_TRACE_POINTS
	16	#include <trace/events/io_uring.h>
	17	#endif
	18
97b388d7 JA	19	enum {
	20	IOU_OK = 0,
	21	IOU_ISSUE_SKIP_COMPLETE = -EIOCBQUEUED,
114eccdf	22
704ea888 JA	23	/*
	24	* Requeue the task_work to restart operations on this request. The
	25	* actual value isn't important, should just be not an otherwise
	26	* valid error code, yet less than -MAX_ERRNO and valid internally.
	27	*/
	28	IOU_REQUEUE = -3072,
	29
114eccdf	30	/*
91482864 PB	31	* Intended only when both IO_URING_F_MULTISHOT is passed
91482864 PB	32	* to indicate to the poll runner that multishot should be
114eccdf DY	33	* removed and the result is set on req->cqe.res.
	34	*/
	35	IOU_STOP_MULTISHOT = -ECANCELED,
97b388d7 JA	36	};
97b388d7 JA	37
405b4dc1 SR	38	struct io_wait_queue {
	39	struct wait_queue_entry wq;
	40	struct io_ring_ctx *ctx;
	41	unsigned cq_tail;
1100c4a2	42	unsigned cq_min_tail;
405b4dc1	43	unsigned nr_timeouts;
cebf123c	44	int hit_timeout;
1100c4a2	45	ktime_t min_timeout;
405b4dc1	46	ktime_t timeout;
cebf123c	47	struct hrtimer t;
405b4dc1	48
8d0c12a8	49	#ifdef CONFIG_NET_RX_BUSY_POLL
342b2e39	50	ktime_t napi_busy_poll_dt;
8d0c12a8 SR	51	bool napi_prefer_busy_poll;
8d0c12a8 SR	52	#endif
405b4dc1 SR	53	};
	54
	55	static inline bool io_should_wake(struct io_wait_queue *iowq)
	56	{
	57	struct io_ring_ctx *ctx = iowq->ctx;
	58	int dist = READ_ONCE(ctx->rings->cq.tail) - (int) iowq->cq_tail;
	59
	60	/*
	61	* Wake up if we have enough events, or if a timeout occurred since we
	62	* started waiting. For timeouts, we always want to return to userspace,
	63	* regardless of event count.
	64	*/
	65	return dist >= 0 \|\| atomic_read(&ctx->cq_timeouts) != iowq->nr_timeouts;
	66	}
	67
20d6b633	68	bool io_cqe_cache_refill(struct io_ring_ctx *ctx, bool overflow);
c0e0d6ba	69	int io_run_task_work_sig(struct io_ring_ctx *ctx);
973fc83f	70	void io_req_defer_failed(struct io_kiocb *req, s32 res);
b529c96a	71	bool io_post_aux_cqe(struct io_ring_ctx *ctx, u64 user_data, s32 res, u32 cflags);
f33096a3	72	void io_add_aux_cqe(struct io_ring_ctx *ctx, u64 user_data, s32 res, u32 cflags);
e5c12945	73	bool io_req_post_cqe(struct io_kiocb *req, s32 res, u32 cflags);
9046c641 PB	74	void __io_commit_cqring_flush(struct io_ring_ctx *ctx);
9046c641 PB	75
9046c641 PB	76	struct file io_file_get_normal(struct io_kiocb req, int fd);
	77	struct file io_file_get_fixed(struct io_kiocb req, int fd,
	78	unsigned issue_flags);
	79
8501fe70	80	void __io_req_task_work_add(struct io_kiocb *req, unsigned flags);
c3ac76f9 JA	81	void io_req_task_work_add_remote(struct io_kiocb req, struct io_ring_ctx ctx,
c3ac76f9 JA	82	unsigned flags);
9046c641	83	bool io_alloc_async_data(struct io_kiocb *req);
9046c641	84	void io_req_task_queue(struct io_kiocb *req);
a282967c	85	void io_req_task_complete(struct io_kiocb req, struct io_tw_state ts);
9046c641	86	void io_req_task_queue_fail(struct io_kiocb *req, int ret);
a282967c	87	void io_req_task_submit(struct io_kiocb req, struct io_tw_state ts);
af5d68f8 JA	88	struct llist_node io_handle_tw_list(struct llist_node node, unsigned int *count, unsigned int max_entries);
af5d68f8 JA	89	struct llist_node tctx_task_work_run(struct io_uring_task tctx, unsigned int max_entries, unsigned int *count);
9046c641 PB	90	void tctx_task_work(struct callback_head *cb);
	91	__cold void io_uring_cancel_generic(bool cancel_all, struct io_sq_data *sqd);
	92	int io_uring_alloc_task_context(struct task_struct *task,
	93	struct io_ring_ctx *ctx);
	94
6e76ac59 JT	95	int io_ring_add_registered_file(struct io_uring_task tctx, struct file file,
6e76ac59 JT	96	int start, int end);
6746ee4c	97	void io_req_queue_iowq(struct io_kiocb *req);
6e76ac59	98
a282967c	99	int io_poll_issue(struct io_kiocb req, struct io_tw_state ts);
9046c641 PB	100	int io_submit_sqes(struct io_ring_ctx *ctx, unsigned int nr);
9046c641 PB	101	int io_do_iopoll(struct io_ring_ctx *ctx, bool force_nonspin);
ec26c225	102	void __io_submit_flush_completions(struct io_ring_ctx *ctx);
9046c641 PB	103
	104	struct io_wq_work io_wq_free_work(struct io_wq_work work);
	105	void io_wq_submit_work(struct io_wq_work *work);
	106
	107	void io_free_req(struct io_kiocb *req);
	108	void io_queue_next(struct io_kiocb *req);
63809137	109	void io_task_refs_refill(struct io_uring_task *tctx);
bd1a3783	110	bool __io_alloc_req_refill(struct io_ring_ctx *ctx);
9046c641 PB	111
	112	bool io_match_task_safe(struct io_kiocb head, struct task_struct task,
	113	bool cancel_all);
	114
c4320315 JA	115	void io_activate_pollwq(struct io_ring_ctx *ctx);
c4320315 JA	116
1658633c JA	117	static inline void io_lockdep_assert_cq_locked(struct io_ring_ctx *ctx)
1658633c JA	118	{
c133b3b0	119	#if defined(CONFIG_PROVE_LOCKING)
1658633c JA	120	lockdep_assert(in_task());
	121
	122	if (ctx->flags & IORING_SETUP_IOPOLL) {
	123	lockdep_assert_held(&ctx->uring_lock);
	124	} else if (!ctx->task_complete) {
	125	lockdep_assert_held(&ctx->completion_lock);
	126	} else if (ctx->submitter_task) {
	127	/*
	128	* ->submitter_task may be NULL and we can still post a CQE,
	129	* if the ring has been setup with IORING_SETUP_R_DISABLED.
	130	* Not from an SQE, as those cannot be submitted, but via
	131	* updating tagged resources.
	132	*/
	133	if (ctx->submitter_task->flags & PF_EXITING)
	134	lockdep_assert(current_work());
	135	else
	136	lockdep_assert(current == ctx->submitter_task);
	137	}
1658633c	138	#endif
c133b3b0	139	}
f26cc959	140
e52d2e58 PB	141	static inline void io_req_task_work_add(struct io_kiocb *req)
e52d2e58 PB	142	{
8501fe70	143	__io_req_task_work_add(req, 0);
e52d2e58 PB	144	}
e52d2e58 PB	145
da12d9ab PB	146	static inline void io_submit_flush_completions(struct io_ring_ctx *ctx)
	147	{
	148	if (!wq_list_empty(&ctx->submit_state.compl_reqs) \|\|
902ce82c	149	ctx->submit_state.cq_flush)
da12d9ab PB	150	__io_submit_flush_completions(ctx);
	151	}
	152
9046c641 PB	153	#define io_for_each_link(pos, head) \
9046c641 PB	154	for (pos = (head); pos; pos = pos->link)
f3b44f92	155
59fbc409 PB	156	static inline bool io_get_cqe_overflow(struct io_ring_ctx *ctx,
	157	struct io_uring_cqe **ret,
	158	bool overflow)
f3b44f92	159	{
20d6b633	160	io_lockdep_assert_cq_locked(ctx);
f3b44f92	161
20d6b633 PB	162	if (unlikely(ctx->cqe_cached >= ctx->cqe_sentinel)) {
20d6b633 PB	163	if (unlikely(!io_cqe_cache_refill(ctx, overflow)))
59fbc409	164	return false;
f3b44f92	165	}
59fbc409	166	*ret = ctx->cqe_cached;
20d6b633 PB	167	ctx->cached_cq_tail++;
	168	ctx->cqe_cached++;
	169	if (ctx->flags & IORING_SETUP_CQE32)
	170	ctx->cqe_cached++;
59fbc409	171	return true;
aa1df3a3 PB	172	}
aa1df3a3 PB	173
59fbc409	174	static inline bool io_get_cqe(struct io_ring_ctx ctx, struct io_uring_cqe *ret)
aa1df3a3	175	{
59fbc409	176	return io_get_cqe_overflow(ctx, ret, false);
f3b44f92 JA	177	}
f3b44f92 JA	178
093a650b PB	179	static __always_inline bool io_fill_cqe_req(struct io_ring_ctx *ctx,
093a650b PB	180	struct io_kiocb *req)
f3b44f92 JA	181	{
	182	struct io_uring_cqe *cqe;
	183
e8c328c3 PB	184	/*
	185	* If we can't get a cq entry, userspace overflowed the
	186	* submission (by quite a lot). Increment the overflow count in
	187	* the ring.
	188	*/
59fbc409	189	if (unlikely(!io_get_cqe(ctx, &cqe)))
f66f7342	190	return false;
e0486f3f	191
a0727c73 PB	192	if (trace_io_uring_complete_enabled())
	193	trace_io_uring_complete(req->ctx, req, req->cqe.user_data,
	194	req->cqe.res, req->cqe.flags,
b24c5d75	195	req->big_cqe.extra1, req->big_cqe.extra2);
e0486f3f	196
e8c328c3	197	memcpy(cqe, &req->cqe, sizeof(*cqe));
e8c328c3	198	if (ctx->flags & IORING_SETUP_CQE32) {
b24c5d75 PB	199	memcpy(cqe->big_cqe, &req->big_cqe, sizeof(*cqe));
b24c5d75 PB	200	memset(&req->big_cqe, 0, sizeof(req->big_cqe));
f3b44f92	201	}
e8c328c3	202	return true;
f3b44f92 JA	203	}
f3b44f92 JA	204
531113bb JA	205	static inline void req_set_fail(struct io_kiocb *req)
	206	{
	207	req->flags \|= REQ_F_FAIL;
	208	if (req->flags & REQ_F_CQE_SKIP) {
	209	req->flags &= ~REQ_F_CQE_SKIP;
	210	req->flags \|= REQ_F_SKIP_LINK_CQES;
	211	}
	212	}
	213
de23077e JA	214	static inline void io_req_set_res(struct io_kiocb *req, s32 res, u32 cflags)
	215	{
	216	req->cqe.res = res;
	217	req->cqe.flags = cflags;
	218	}
	219
99f15d8d JA	220	static inline bool req_has_async_data(struct io_kiocb *req)
	221	{
	222	return req->flags & REQ_F_ASYNC_DATA;
	223	}
	224
17bc2837	225	static inline void io_put_file(struct io_kiocb *req)
531113bb	226	{
17bc2837 JA	227	if (!(req->flags & REQ_F_FIXED_FILE) && req->file)
17bc2837 JA	228	fput(req->file);
531113bb JA	229	}
531113bb JA	230
cd40cae2 JA	231	static inline void io_ring_submit_unlock(struct io_ring_ctx *ctx,
	232	unsigned issue_flags)
	233	{
	234	lockdep_assert_held(&ctx->uring_lock);
bfe30bfd	235	if (unlikely(issue_flags & IO_URING_F_UNLOCKED))
cd40cae2 JA	236	mutex_unlock(&ctx->uring_lock);
	237	}
	238
	239	static inline void io_ring_submit_lock(struct io_ring_ctx *ctx,
	240	unsigned issue_flags)
	241	{
	242	/*
	243	* "Normal" inline submissions always hold the uring_lock, since we
	244	* grab it from the system call. Same is true for the SQPOLL offload.
	245	* The only exception is when we've detached the request and issue it
	246	* from an async worker thread, grab the lock for that case.
	247	*/
bfe30bfd	248	if (unlikely(issue_flags & IO_URING_F_UNLOCKED))
cd40cae2 JA	249	mutex_lock(&ctx->uring_lock);
	250	lockdep_assert_held(&ctx->uring_lock);
	251	}
	252
f9ead18c JA	253	static inline void io_commit_cqring(struct io_ring_ctx *ctx)
	254	{
	255	/* order cqe stores with ring update */
	256	smp_store_release(&ctx->rings->cq.tail, ctx->cached_cq_tail);
	257	}
	258
7b235dd8 PB	259	static inline void io_poll_wq_wake(struct io_ring_ctx *ctx)
7b235dd8 PB	260	{
bca39f39	261	if (wq_has_sleeper(&ctx->poll_wq))
7b235dd8 PB	262	__wake_up(&ctx->poll_wq, TASK_NORMAL, 0,
	263	poll_to_key(EPOLL_URING_WAKE \| EPOLLIN));
	264	}
	265
6e7248ad	266	static inline void io_cqring_wake(struct io_ring_ctx *ctx)
f3b44f92 JA	267	{
f3b44f92 JA	268	/*
44648532 JA	269	* Trigger waitqueue handler on all waiters on our waitqueue. This
	270	* won't necessarily wake up all the tasks, io_should_wake() will make
	271	* that decision.
	272	*
	273	* Pass in EPOLLIN\|EPOLL_URING_WAKE as the poll wakeup key. The latter
	274	* set in the mask so that if we recurse back into our own poll
	275	* waitqueue handlers, we know we have a dependency between eventfd or
	276	* epoll and should terminate multishot poll at that point.
f3b44f92	277	*/
6e7248ad	278	if (wq_has_sleeper(&ctx->cq_wait))
44648532 JA	279	__wake_up(&ctx->cq_wait, TASK_NORMAL, 0,
44648532 JA	280	poll_to_key(EPOLL_URING_WAKE \| EPOLLIN));
f3b44f92 JA	281	}
f3b44f92 JA	282
17437f31 JA	283	static inline bool io_sqring_full(struct io_ring_ctx *ctx)
	284	{
	285	struct io_rings *r = ctx->rings;
	286
	287	return READ_ONCE(r->sq.tail) - ctx->cached_sq_head == ctx->sq_entries;
	288	}
	289
	290	static inline unsigned int io_sqring_entries(struct io_ring_ctx *ctx)
	291	{
	292	struct io_rings *rings = ctx->rings;
e3ef728f	293	unsigned int entries;
17437f31 JA	294
17437f31 JA	295	/* make sure SQ entry isn't read before tail */
e3ef728f JA	296	entries = smp_load_acquire(&rings->sq.tail) - ctx->cached_sq_head;
e3ef728f JA	297	return min(entries, ctx->sq_entries);
17437f31 JA	298	}
17437f31 JA	299
c0e0d6ba	300	static inline int io_run_task_work(void)
17437f31	301	{
af5d68f8 JA	302	bool ret = false;
af5d68f8 JA	303
7cfe7a09 JA	304	/*
	305	* Always check-and-clear the task_work notification signal. With how
	306	* signaling works for task_work, we can find it set with nothing to
	307	* run. We need to clear it for that case, like get_signal() does.
	308	*/
	309	if (test_thread_flag(TIF_NOTIFY_SIGNAL))
	310	clear_notify_signal();
b5d3ae20 JA	311	/*
	312	* PF_IO_WORKER never returns to userspace, so check here if we have
	313	* notify work that needs processing.
	314	*/
af5d68f8 JA	315	if (current->flags & PF_IO_WORKER) {
	316	if (test_thread_flag(TIF_NOTIFY_RESUME)) {
	317	__set_current_state(TASK_RUNNING);
	318	resume_user_mode_work(NULL);
	319	}
	320	if (current->io_uring) {
	321	unsigned int count = 0;
	322
	323	tctx_task_work_run(current->io_uring, UINT_MAX, &count);
	324	if (count)
	325	ret = true;
	326	}
2f2bb1ff	327	}
46a525e1	328	if (task_work_pending(current)) {
17437f31	329	__set_current_state(TASK_RUNNING);
46a525e1	330	task_work_run();
af5d68f8	331	ret = true;
17437f31 JA	332	}
17437f31 JA	333
af5d68f8	334	return ret;
c0e0d6ba DY	335	}
c0e0d6ba DY	336
dac6a0ea JA	337	static inline bool io_task_work_pending(struct io_ring_ctx *ctx)
dac6a0ea JA	338	{
22537c9f	339	return task_work_pending(current) \|\| !llist_empty(&ctx->work_llist);
dac6a0ea JA	340	}
dac6a0ea JA	341
a282967c	342	static inline void io_tw_lock(struct io_ring_ctx ctx, struct io_tw_state ts)
aa1e90f6	343	{
8e5b3b89	344	lockdep_assert_held(&ctx->uring_lock);
aa1e90f6 PB	345	}
aa1e90f6 PB	346
9da070b1 PB	347	/*
	348	* Don't complete immediately but use deferred completion infrastructure.
	349	* Protected by ->uring_lock and can only be used either with
	350	* IO_URING_F_COMPLETE_DEFER or inside a tw handler holding the mutex.
	351	*/
	352	static inline void io_req_complete_defer(struct io_kiocb *req)
	353	__must_hold(&req->ctx->uring_lock)
aa1e90f6 PB	354	{
	355	struct io_submit_state *state = &req->ctx->submit_state;
	356
9da070b1 PB	357	lockdep_assert_held(&req->ctx->uring_lock);
9da070b1 PB	358
aa1e90f6 PB	359	wq_list_add_tail(&req->comp_list, &state->compl_reqs);
	360	}
	361
46929b08 PB	362	static inline void io_commit_cqring_flush(struct io_ring_ctx *ctx)
46929b08 PB	363	{
bca39f39 PB	364	if (unlikely(ctx->off_timeout_used \|\| ctx->drain_active \|\|
bca39f39 PB	365	ctx->has_evfd \|\| ctx->poll_activated))
46929b08 PB	366	__io_commit_cqring_flush(ctx);
	367	}
	368
63809137 PB	369	static inline void io_get_task_refs(int nr)
	370	{
	371	struct io_uring_task *tctx = current->io_uring;
	372
	373	tctx->cached_refs -= nr;
	374	if (unlikely(tctx->cached_refs < 0))
	375	io_task_refs_refill(tctx);
	376	}
	377
bd1a3783 PB	378	static inline bool io_req_cache_empty(struct io_ring_ctx *ctx)
	379	{
	380	return !ctx->submit_state.free_list.next;
	381	}
	382
c1755c25	383	extern struct kmem_cache *req_cachep;
b3a4dbc8	384	extern struct kmem_cache *io_buf_cachep;
c1755c25	385
c8576f3e	386	static inline struct io_kiocb io_extract_req(struct io_ring_ctx ctx)
bd1a3783	387	{
c1755c25	388	struct io_kiocb *req;
bd1a3783	389
c1755c25	390	req = container_of(ctx->submit_state.free_list.next, struct io_kiocb, comp_list);
c1755c25 BL	391	wq_stack_extract(&ctx->submit_state.free_list);
c1755c25 BL	392	return req;
bd1a3783 PB	393	}
bd1a3783 PB	394
c8576f3e PB	395	static inline bool io_alloc_req(struct io_ring_ctx ctx, struct io_kiocb *req)
	396	{
	397	if (unlikely(io_req_cache_empty(ctx))) {
	398	if (!__io_alloc_req_refill(ctx))
	399	return false;
	400	}
	401	*req = io_extract_req(ctx);
	402	return true;
	403	}
	404
140102ae PB	405	static inline bool io_allowed_defer_tw_run(struct io_ring_ctx *ctx)
	406	{
	407	return likely(ctx->submitter_task == current);
	408	}
	409
76de6749 PB	410	static inline bool io_allowed_run_tw(struct io_ring_ctx *ctx)
76de6749 PB	411	{
6567506b PB	412	return likely(!(ctx->flags & IORING_SETUP_DEFER_TASKRUN) \|\|
6567506b PB	413	ctx->submitter_task == current);
76de6749 PB	414	}
76de6749 PB	415
833b5dff PB	416	static inline void io_req_queue_tw_complete(struct io_kiocb *req, s32 res)
	417	{
	418	io_req_set_res(req, res, 0);
	419	req->io_task_work.func = io_req_task_complete;
	420	io_req_task_work_add(req);
	421	}
	422
96c7d4f8 BL	423	/*
	424	* IORING_SETUP_SQE128 contexts allocate twice the normal SQE size for each
	425	* slot.
	426	*/
	427	static inline size_t uring_sqe_size(struct io_ring_ctx *ctx)
	428	{
	429	if (ctx->flags & IORING_SETUP_SQE128)
	430	return 2 * sizeof(struct io_uring_sqe);
	431	return sizeof(struct io_uring_sqe);
	432	}
95041b93 JA	433
	434	static inline bool io_file_can_poll(struct io_kiocb *req)
	435	{
	436	if (req->flags & REQ_F_CAN_POLL)
	437	return true;
5fc16fa5	438	if (req->file && file_can_poll(req->file)) {
95041b93 JA	439	req->flags \|= REQ_F_CAN_POLL;
	440	return true;
	441	}
	442	return false;
	443	}
428f1382	444
2b8e976b PB	445	static inline ktime_t io_get_time(struct io_ring_ctx *ctx)
	446	{
	447	if (ctx->clockid == CLOCK_MONOTONIC)
	448	return ktime_get();
	449
	450	return ktime_get_with_offset(ctx->clock_offset);
	451	}
	452
428f1382 JA	453	enum {
	454	IO_CHECK_CQ_OVERFLOW_BIT,
	455	IO_CHECK_CQ_DROPPED_BIT,
	456	};
	457
	458	static inline bool io_has_work(struct io_ring_ctx *ctx)
	459	{
	460	return test_bit(IO_CHECK_CQ_OVERFLOW_BIT, &ctx->check_cq) \|\|
	461	!llist_empty(&ctx->work_llist);
	462	}
de23077e	463	#endif