[linux-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/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,

	/*
	 * 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_uring_cqe *__io_get_cqe(struct io_ring_ctx *ctx, bool overflow);
bool io_req_cqe_overflow(struct io_kiocb *req);
int io_run_task_work_sig(struct io_ring_ctx *ctx);
void io_req_defer_failed(struct io_kiocb *req, s32 res);
void io_req_complete_post(struct io_kiocb *req, unsigned issue_flags);
bool io_post_aux_cqe(struct io_ring_ctx *ctx, u64 user_data, s32 res, u32 cflags);
bool io_aux_cqe(struct io_ring_ctx *ctx, bool defer, u64 user_data, s32 res, u32 cflags,
		bool allow_overflow);
void __io_commit_cqring_flush(struct io_ring_ctx *ctx);

struct page **io_pin_pages(unsigned long ubuf, unsigned long len, int *npages);

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);

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

void __io_req_task_work_add(struct io_kiocb *req, bool allow_local);
bool io_is_uring_fops(struct file *file);
bool io_alloc_async_data(struct io_kiocb *req);
void io_req_task_queue(struct io_kiocb *req);
void io_queue_iowq(struct io_kiocb *req, bool *dont_use);
void io_req_task_complete(struct io_kiocb *req, bool *locked);
void io_req_task_queue_fail(struct io_kiocb *req, int ret);
void io_req_task_submit(struct io_kiocb *req, bool *locked);
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_poll_issue(struct io_kiocb *req, bool *locked);
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_free_batch_list(struct io_ring_ctx *ctx, struct io_wq_work_node *node);
int io_req_prep_async(struct io_kiocb *req);

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_put_task(struct task_struct *task, int nr);
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);

#define io_lockdep_assert_cq_locked(ctx)				\
	do {								\
		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->flags & PF_EXITING) {	\
			lockdep_assert(current_work());			\
		} else {						\
			lockdep_assert(current == ctx->submitter_task);	\
		}							\
	} while (0)

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

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

void io_cq_unlock_post(struct io_ring_ctx *ctx);

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

	if (likely(ctx->cqe_cached < ctx->cqe_sentinel)) {
		struct io_uring_cqe *cqe = ctx->cqe_cached;

		ctx->cached_cq_tail++;
		ctx->cqe_cached++;
		if (ctx->flags & IORING_SETUP_CQE32)
			ctx->cqe_cached++;
		return cqe;
	}

	return __io_get_cqe(ctx, overflow);
}

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

static 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.
	 */
	cqe = io_get_cqe(ctx);
	if (unlikely(!cqe))
		return false;

	trace_io_uring_complete(req->ctx, req, req->cqe.user_data,
				req->cqe.res, req->cqe.flags,
				(req->flags & REQ_F_CQE32_INIT) ? req->extra1 : 0,
				(req->flags & REQ_F_CQE32_INIT) ? req->extra2 : 0);

	memcpy(cqe, &req->cqe, sizeof(*cqe));

	if (ctx->flags & IORING_SETUP_CQE32) {
		u64 extra1 = 0, extra2 = 0;

		if (req->flags & REQ_F_CQE32_INIT) {
			extra1 = req->extra1;
			extra2 = req->extra2;
		}

		WRITE_ONCE(cqe->big_cqe[0], extra1);
		WRITE_ONCE(cqe->big_cqe[1], extra2);
	}
	return true;
}

static inline bool io_fill_cqe_req(struct io_ring_ctx *ctx,
				   struct io_kiocb *req)
{
	if (likely(__io_fill_cqe_req(ctx, req)))
		return true;
	return io_req_cqe_overflow(req);
}

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 file *file)
{
	if (file)
		fput(file);
}

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

/* requires smb_mb() prior, see wq_has_sleeper() */
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 (waitqueue_active(&ctx->cq_wait))
		__wake_up(&ctx->cq_wait, TASK_NORMAL, 0,
				poll_to_key(EPOLL_URING_WAKE | EPOLLIN));
}

static inline void io_cqring_wake(struct io_ring_ctx *ctx)
{
	smp_mb();
	__io_cqring_wake(ctx);
}

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;

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

static inline int io_run_task_work(void)
{
	/*
	 * 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();
	if (task_work_pending(current)) {
		__set_current_state(TASK_RUNNING);
		task_work_run();
		return 1;
	}

	return 0;
}

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

static inline void io_tw_lock(struct io_ring_ctx *ctx, bool *locked)
{
	if (!*locked) {
		mutex_lock(&ctx->uring_lock);
		*locked = true;
	}
}

/*
 * 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);
}

/* must to be called somewhat shortly after putting a request */
static inline void io_put_task(struct task_struct *task, int nr)
{
	if (likely(task == current))
		task->io_uring->cached_refs += nr;
	else
		__io_put_task(task, nr);
}

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;
}

static inline bool io_alloc_req_refill(struct io_ring_ctx *ctx)
{
	if (unlikely(io_req_cache_empty(ctx)))
		return __io_alloc_req_refill(ctx);
	return true;
}

static inline struct io_kiocb *io_alloc_req(struct io_ring_ctx *ctx)
{
	struct io_wq_work_node *node;

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

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);
}

#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>
ab1c84d8	6	#include <linux/io_uring_types.h>
44648532	7	#include <uapi/linux/eventpoll.h>
ab1c84d8	8	#include "io-wq.h"
a6b21fbb	9	#include "slist.h"
ab1c84d8	10	#include "filetable.h"
de23077e	11
f3b44f92 JA	12	#ifndef CREATE_TRACE_POINTS
	13	#include <trace/events/io_uring.h>
	14	#endif
	15
97b388d7 JA	16	enum {
	17	IOU_OK = 0,
	18	IOU_ISSUE_SKIP_COMPLETE = -EIOCBQUEUED,
114eccdf DY	19
114eccdf DY	20	/*
91482864 PB	21	* Intended only when both IO_URING_F_MULTISHOT is passed
91482864 PB	22	* to indicate to the poll runner that multishot should be
114eccdf DY	23	* removed and the result is set on req->cqe.res.
	24	*/
	25	IOU_STOP_MULTISHOT = -ECANCELED,
97b388d7 JA	26	};
97b388d7 JA	27
aa1df3a3	28	struct io_uring_cqe __io_get_cqe(struct io_ring_ctx ctx, bool overflow);
68494a65	29	bool io_req_cqe_overflow(struct io_kiocb *req);
c0e0d6ba	30	int io_run_task_work_sig(struct io_ring_ctx *ctx);
973fc83f	31	void io_req_defer_failed(struct io_kiocb *req, s32 res);
1bec951c	32	void io_req_complete_post(struct io_kiocb *req, unsigned issue_flags);
b529c96a	33	bool io_post_aux_cqe(struct io_ring_ctx *ctx, u64 user_data, s32 res, u32 cflags);
9b8c5475 DY	34	bool io_aux_cqe(struct io_ring_ctx *ctx, bool defer, u64 user_data, s32 res, u32 cflags,
9b8c5475 DY	35	bool allow_overflow);
9046c641 PB	36	void __io_commit_cqring_flush(struct io_ring_ctx *ctx);
	37
	38	struct page *io_pin_pages(unsigned long ubuf, unsigned long len, int npages);
	39
	40	struct file io_file_get_normal(struct io_kiocb req, int fd);
	41	struct file io_file_get_fixed(struct io_kiocb req, int fd,
	42	unsigned issue_flags);
	43
f6b543fd JA	44	static inline bool io_req_ffs_set(struct io_kiocb *req)
	45	{
	46	return req->flags & REQ_F_FIXED_FILE;
	47	}
	48
e52d2e58	49	void __io_req_task_work_add(struct io_kiocb *req, bool allow_local);
9046c641 PB	50	bool io_is_uring_fops(struct file *file);
9046c641 PB	51	bool io_alloc_async_data(struct io_kiocb *req);
9046c641 PB	52	void io_req_task_queue(struct io_kiocb *req);
	53	void io_queue_iowq(struct io_kiocb req, bool dont_use);
	54	void io_req_task_complete(struct io_kiocb req, bool locked);
	55	void io_req_task_queue_fail(struct io_kiocb *req, int ret);
	56	void io_req_task_submit(struct io_kiocb req, bool locked);
	57	void tctx_task_work(struct callback_head *cb);
	58	__cold void io_uring_cancel_generic(bool cancel_all, struct io_sq_data *sqd);
	59	int io_uring_alloc_task_context(struct task_struct *task,
	60	struct io_ring_ctx *ctx);
	61
	62	int io_poll_issue(struct io_kiocb req, bool locked);
	63	int io_submit_sqes(struct io_ring_ctx *ctx, unsigned int nr);
	64	int io_do_iopoll(struct io_ring_ctx *ctx, bool force_nonspin);
	65	void io_free_batch_list(struct io_ring_ctx ctx, struct io_wq_work_node node);
	66	int io_req_prep_async(struct io_kiocb *req);
	67
	68	struct io_wq_work io_wq_free_work(struct io_wq_work work);
	69	void io_wq_submit_work(struct io_wq_work *work);
	70
	71	void io_free_req(struct io_kiocb *req);
	72	void io_queue_next(struct io_kiocb *req);
e70cb608	73	void __io_put_task(struct task_struct *task, int nr);
63809137	74	void io_task_refs_refill(struct io_uring_task *tctx);
bd1a3783	75	bool __io_alloc_req_refill(struct io_ring_ctx *ctx);
9046c641 PB	76
	77	bool io_match_task_safe(struct io_kiocb head, struct task_struct task,
	78	bool cancel_all);
	79
f26cc959 PB	80	#define io_lockdep_assert_cq_locked(ctx) \
	81	do { \
	82	if (ctx->flags & IORING_SETUP_IOPOLL) { \
	83	lockdep_assert_held(&ctx->uring_lock); \
	84	} else if (!ctx->task_complete) { \
	85	lockdep_assert_held(&ctx->completion_lock); \
	86	} else if (ctx->submitter_task->flags & PF_EXITING) { \
	87	lockdep_assert(current_work()); \
	88	} else { \
	89	lockdep_assert(current == ctx->submitter_task); \
	90	} \
	91	} while (0)
	92
e52d2e58 PB	93	static inline void io_req_task_work_add(struct io_kiocb *req)
	94	{
	95	__io_req_task_work_add(req, true);
	96	}
	97
9046c641 PB	98	#define io_for_each_link(pos, head) \
9046c641 PB	99	for (pos = (head); pos; pos = pos->link)
f3b44f92	100
25399321 PB	101	void io_cq_unlock_post(struct io_ring_ctx *ctx);
25399321 PB	102
aa1df3a3 PB	103	static inline struct io_uring_cqe io_get_cqe_overflow(struct io_ring_ctx ctx,
aa1df3a3 PB	104	bool overflow)
f3b44f92	105	{
f26cc959 PB	106	io_lockdep_assert_cq_locked(ctx);
f26cc959 PB	107
f3b44f92 JA	108	if (likely(ctx->cqe_cached < ctx->cqe_sentinel)) {
	109	struct io_uring_cqe *cqe = ctx->cqe_cached;
	110
f3b44f92 JA	111	ctx->cached_cq_tail++;
f3b44f92 JA	112	ctx->cqe_cached++;
b3659a65 PB	113	if (ctx->flags & IORING_SETUP_CQE32)
b3659a65 PB	114	ctx->cqe_cached++;
f3b44f92 JA	115	return cqe;
	116	}
	117
aa1df3a3 PB	118	return __io_get_cqe(ctx, overflow);
	119	}
	120
	121	static inline struct io_uring_cqe io_get_cqe(struct io_ring_ctx ctx)
	122	{
	123	return io_get_cqe_overflow(ctx, false);
f3b44f92 JA	124	}
	125
	126	static inline bool __io_fill_cqe_req(struct io_ring_ctx *ctx,
	127	struct io_kiocb *req)
	128	{
	129	struct io_uring_cqe *cqe;
	130
e8c328c3 PB	131	/*
	132	* If we can't get a cq entry, userspace overflowed the
	133	* submission (by quite a lot). Increment the overflow count in
	134	* the ring.
	135	*/
	136	cqe = io_get_cqe(ctx);
	137	if (unlikely(!cqe))
f66f7342	138	return false;
e0486f3f DY	139
	140	trace_io_uring_complete(req->ctx, req, req->cqe.user_data,
	141	req->cqe.res, req->cqe.flags,
	142	(req->flags & REQ_F_CQE32_INIT) ? req->extra1 : 0,
	143	(req->flags & REQ_F_CQE32_INIT) ? req->extra2 : 0);
	144
e8c328c3 PB	145	memcpy(cqe, &req->cqe, sizeof(*cqe));
	146
	147	if (ctx->flags & IORING_SETUP_CQE32) {
f3b44f92 JA	148	u64 extra1 = 0, extra2 = 0;
	149
	150	if (req->flags & REQ_F_CQE32_INIT) {
	151	extra1 = req->extra1;
	152	extra2 = req->extra2;
	153	}
	154
e8c328c3 PB	155	WRITE_ONCE(cqe->big_cqe[0], extra1);
e8c328c3 PB	156	WRITE_ONCE(cqe->big_cqe[1], extra2);
f3b44f92	157	}
e8c328c3	158	return true;
f3b44f92 JA	159	}
f3b44f92 JA	160
f66f7342 PB	161	static inline bool io_fill_cqe_req(struct io_ring_ctx *ctx,
	162	struct io_kiocb *req)
	163	{
	164	if (likely(__io_fill_cqe_req(ctx, req)))
	165	return true;
	166	return io_req_cqe_overflow(req);
	167	}
	168
531113bb JA	169	static inline void req_set_fail(struct io_kiocb *req)
	170	{
	171	req->flags \|= REQ_F_FAIL;
	172	if (req->flags & REQ_F_CQE_SKIP) {
	173	req->flags &= ~REQ_F_CQE_SKIP;
	174	req->flags \|= REQ_F_SKIP_LINK_CQES;
	175	}
	176	}
	177
de23077e JA	178	static inline void io_req_set_res(struct io_kiocb *req, s32 res, u32 cflags)
	179	{
	180	req->cqe.res = res;
	181	req->cqe.flags = cflags;
	182	}
	183
99f15d8d JA	184	static inline bool req_has_async_data(struct io_kiocb *req)
	185	{
	186	return req->flags & REQ_F_ASYNC_DATA;
	187	}
	188
531113bb JA	189	static inline void io_put_file(struct file *file)
	190	{
	191	if (file)
	192	fput(file);
	193	}
	194
cd40cae2 JA	195	static inline void io_ring_submit_unlock(struct io_ring_ctx *ctx,
	196	unsigned issue_flags)
	197	{
	198	lockdep_assert_held(&ctx->uring_lock);
	199	if (issue_flags & IO_URING_F_UNLOCKED)
	200	mutex_unlock(&ctx->uring_lock);
	201	}
	202
	203	static inline void io_ring_submit_lock(struct io_ring_ctx *ctx,
	204	unsigned issue_flags)
	205	{
	206	/*
	207	* "Normal" inline submissions always hold the uring_lock, since we
	208	* grab it from the system call. Same is true for the SQPOLL offload.
	209	* The only exception is when we've detached the request and issue it
	210	* from an async worker thread, grab the lock for that case.
	211	*/
	212	if (issue_flags & IO_URING_F_UNLOCKED)
	213	mutex_lock(&ctx->uring_lock);
	214	lockdep_assert_held(&ctx->uring_lock);
	215	}
	216
f9ead18c JA	217	static inline void io_commit_cqring(struct io_ring_ctx *ctx)
	218	{
	219	/* order cqe stores with ring update */
	220	smp_store_release(&ctx->rings->cq.tail, ctx->cached_cq_tail);
	221	}
	222
7b235dd8 PB	223	static inline void io_poll_wq_wake(struct io_ring_ctx *ctx)
7b235dd8 PB	224	{
bca39f39	225	if (wq_has_sleeper(&ctx->poll_wq))
7b235dd8 PB	226	__wake_up(&ctx->poll_wq, TASK_NORMAL, 0,
	227	poll_to_key(EPOLL_URING_WAKE \| EPOLLIN));
	228	}
	229
fc86f9d3 PB	230	/* requires smb_mb() prior, see wq_has_sleeper() */
fc86f9d3 PB	231	static inline void __io_cqring_wake(struct io_ring_ctx *ctx)
f3b44f92 JA	232	{
f3b44f92 JA	233	/*
44648532 JA	234	* Trigger waitqueue handler on all waiters on our waitqueue. This
	235	* won't necessarily wake up all the tasks, io_should_wake() will make
	236	* that decision.
	237	*
	238	* Pass in EPOLLIN\|EPOLL_URING_WAKE as the poll wakeup key. The latter
	239	* set in the mask so that if we recurse back into our own poll
	240	* waitqueue handlers, we know we have a dependency between eventfd or
	241	* epoll and should terminate multishot poll at that point.
f3b44f92	242	*/
fc86f9d3	243	if (waitqueue_active(&ctx->cq_wait))
44648532 JA	244	__wake_up(&ctx->cq_wait, TASK_NORMAL, 0,
44648532 JA	245	poll_to_key(EPOLL_URING_WAKE \| EPOLLIN));
f3b44f92 JA	246	}
f3b44f92 JA	247
fc86f9d3 PB	248	static inline void io_cqring_wake(struct io_ring_ctx *ctx)
	249	{
	250	smp_mb();
	251	__io_cqring_wake(ctx);
	252	}
	253
17437f31 JA	254	static inline bool io_sqring_full(struct io_ring_ctx *ctx)
	255	{
	256	struct io_rings *r = ctx->rings;
	257
	258	return READ_ONCE(r->sq.tail) - ctx->cached_sq_head == ctx->sq_entries;
	259	}
	260
	261	static inline unsigned int io_sqring_entries(struct io_ring_ctx *ctx)
	262	{
	263	struct io_rings *rings = ctx->rings;
	264
	265	/* make sure SQ entry isn't read before tail */
	266	return smp_load_acquire(&rings->sq.tail) - ctx->cached_sq_head;
	267	}
	268
c0e0d6ba	269	static inline int io_run_task_work(void)
17437f31	270	{
7cfe7a09 JA	271	/*
	272	* Always check-and-clear the task_work notification signal. With how
	273	* signaling works for task_work, we can find it set with nothing to
	274	* run. We need to clear it for that case, like get_signal() does.
	275	*/
	276	if (test_thread_flag(TIF_NOTIFY_SIGNAL))
	277	clear_notify_signal();
46a525e1	278	if (task_work_pending(current)) {
17437f31	279	__set_current_state(TASK_RUNNING);
46a525e1	280	task_work_run();
c0e0d6ba	281	return 1;
17437f31 JA	282	}
17437f31 JA	283
c0e0d6ba DY	284	return 0;
	285	}
	286
dac6a0ea JA	287	static inline bool io_task_work_pending(struct io_ring_ctx *ctx)
dac6a0ea JA	288	{
6434ec01	289	return task_work_pending(current) \|\| !wq_list_empty(&ctx->work_llist);
dac6a0ea JA	290	}
dac6a0ea JA	291
aa1e90f6 PB	292	static inline void io_tw_lock(struct io_ring_ctx ctx, bool locked)
	293	{
	294	if (!*locked) {
	295	mutex_lock(&ctx->uring_lock);
	296	*locked = true;
	297	}
	298	}
	299
9da070b1 PB	300	/*
	301	* Don't complete immediately but use deferred completion infrastructure.
	302	* Protected by ->uring_lock and can only be used either with
	303	* IO_URING_F_COMPLETE_DEFER or inside a tw handler holding the mutex.
	304	*/
	305	static inline void io_req_complete_defer(struct io_kiocb *req)
	306	__must_hold(&req->ctx->uring_lock)
aa1e90f6 PB	307	{
	308	struct io_submit_state *state = &req->ctx->submit_state;
	309
9da070b1 PB	310	lockdep_assert_held(&req->ctx->uring_lock);
9da070b1 PB	311
aa1e90f6 PB	312	wq_list_add_tail(&req->comp_list, &state->compl_reqs);
	313	}
	314
46929b08 PB	315	static inline void io_commit_cqring_flush(struct io_ring_ctx *ctx)
46929b08 PB	316	{
bca39f39 PB	317	if (unlikely(ctx->off_timeout_used \|\| ctx->drain_active \|\|
bca39f39 PB	318	ctx->has_evfd \|\| ctx->poll_activated))
46929b08 PB	319	__io_commit_cqring_flush(ctx);
	320	}
	321
e70cb608 PB	322	/* must to be called somewhat shortly after putting a request */
	323	static inline void io_put_task(struct task_struct *task, int nr)
	324	{
	325	if (likely(task == current))
	326	task->io_uring->cached_refs += nr;
	327	else
	328	__io_put_task(task, nr);
	329	}
	330
63809137 PB	331	static inline void io_get_task_refs(int nr)
	332	{
	333	struct io_uring_task *tctx = current->io_uring;
	334
	335	tctx->cached_refs -= nr;
	336	if (unlikely(tctx->cached_refs < 0))
	337	io_task_refs_refill(tctx);
	338	}
	339
bd1a3783 PB	340	static inline bool io_req_cache_empty(struct io_ring_ctx *ctx)
	341	{
	342	return !ctx->submit_state.free_list.next;
	343	}
	344
	345	static inline bool io_alloc_req_refill(struct io_ring_ctx *ctx)
	346	{
	347	if (unlikely(io_req_cache_empty(ctx)))
	348	return __io_alloc_req_refill(ctx);
	349	return true;
	350	}
	351
	352	static inline struct io_kiocb io_alloc_req(struct io_ring_ctx ctx)
	353	{
	354	struct io_wq_work_node *node;
	355
	356	node = wq_stack_extract(&ctx->submit_state.free_list);
	357	return container_of(node, struct io_kiocb, comp_list);
	358	}
	359
140102ae PB	360	static inline bool io_allowed_defer_tw_run(struct io_ring_ctx *ctx)
	361	{
	362	return likely(ctx->submitter_task == current);
	363	}
	364
76de6749 PB	365	static inline bool io_allowed_run_tw(struct io_ring_ctx *ctx)
76de6749 PB	366	{
6567506b PB	367	return likely(!(ctx->flags & IORING_SETUP_DEFER_TASKRUN) \|\|
6567506b PB	368	ctx->submitter_task == current);
76de6749 PB	369	}
76de6749 PB	370
833b5dff PB	371	static inline void io_req_queue_tw_complete(struct io_kiocb *req, s32 res)
	372	{
	373	io_req_set_res(req, res, 0);
	374	req->io_task_work.func = io_req_task_complete;
	375	io_req_task_work_add(req);
	376	}
	377
de23077e	378	#endif