[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/resume_user_mode.h>
#include <linux/kasan.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 {
	/* don't use deferred task_work */
	IOU_F_TWQ_FORCE_NORMAL			= 1,
};

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, unsigned flags);
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, struct io_tw_state *ts_dont_use);
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);
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, 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_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_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, 0);
}

#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));
}

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)
{
	/*
	 * 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 &&
	    test_thread_flag(TIF_NOTIFY_RESUME)) {
		__set_current_state(TASK_RUNNING);
		resume_user_mode_work(NULL);
	}
	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, struct io_tw_state *ts)
{
	if (!ts->locked) {
		mutex_lock(&ctx->uring_lock);
		ts->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);
}

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;

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);
	kasan_unpoison_object_data(req_cachep, req);
	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);
}

#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>
ab1c84d8	8	#include <linux/io_uring_types.h>
44648532	9	#include <uapi/linux/eventpoll.h>
ab1c84d8	10	#include "io-wq.h"
a6b21fbb	11	#include "slist.h"
ab1c84d8	12	#include "filetable.h"
de23077e	13
f3b44f92 JA	14	#ifndef CREATE_TRACE_POINTS
	15	#include <trace/events/io_uring.h>
	16	#endif
	17
8501fe70 PB	18	enum {
	19	/* don't use deferred task_work */
	20	IOU_F_TWQ_FORCE_NORMAL = 1,
	21	};
	22
97b388d7 JA	23	enum {
	24	IOU_OK = 0,
	25	IOU_ISSUE_SKIP_COMPLETE = -EIOCBQUEUED,
114eccdf DY	26
114eccdf DY	27	/*
91482864 PB	28	* Intended only when both IO_URING_F_MULTISHOT is passed
91482864 PB	29	* to indicate to the poll runner that multishot should be
114eccdf DY	30	* removed and the result is set on req->cqe.res.
	31	*/
	32	IOU_STOP_MULTISHOT = -ECANCELED,
97b388d7 JA	33	};
97b388d7 JA	34
aa1df3a3	35	struct io_uring_cqe __io_get_cqe(struct io_ring_ctx ctx, bool overflow);
68494a65	36	bool io_req_cqe_overflow(struct io_kiocb *req);
c0e0d6ba	37	int io_run_task_work_sig(struct io_ring_ctx *ctx);
973fc83f	38	void io_req_defer_failed(struct io_kiocb *req, s32 res);
1bec951c	39	void io_req_complete_post(struct io_kiocb *req, unsigned issue_flags);
b529c96a	40	bool io_post_aux_cqe(struct io_ring_ctx *ctx, u64 user_data, s32 res, u32 cflags);
9b8c5475 DY	41	bool io_aux_cqe(struct io_ring_ctx *ctx, bool defer, u64 user_data, s32 res, u32 cflags,
9b8c5475 DY	42	bool allow_overflow);
9046c641 PB	43	void __io_commit_cqring_flush(struct io_ring_ctx *ctx);
	44
	45	struct page *io_pin_pages(unsigned long ubuf, unsigned long len, int npages);
	46
	47	struct file io_file_get_normal(struct io_kiocb req, int fd);
	48	struct file io_file_get_fixed(struct io_kiocb req, int fd,
	49	unsigned issue_flags);
	50
f6b543fd JA	51	static inline bool io_req_ffs_set(struct io_kiocb *req)
	52	{
	53	return req->flags & REQ_F_FIXED_FILE;
	54	}
	55
8501fe70	56	void __io_req_task_work_add(struct io_kiocb *req, unsigned flags);
9046c641 PB	57	bool io_is_uring_fops(struct file *file);
9046c641 PB	58	bool io_alloc_async_data(struct io_kiocb *req);
9046c641	59	void io_req_task_queue(struct io_kiocb *req);
a282967c PB	60	void io_queue_iowq(struct io_kiocb req, struct io_tw_state ts_dont_use);
a282967c PB	61	void io_req_task_complete(struct io_kiocb req, struct io_tw_state ts);
9046c641	62	void io_req_task_queue_fail(struct io_kiocb *req, int ret);
a282967c	63	void io_req_task_submit(struct io_kiocb req, struct io_tw_state ts);
9046c641 PB	64	void tctx_task_work(struct callback_head *cb);
	65	__cold void io_uring_cancel_generic(bool cancel_all, struct io_sq_data *sqd);
	66	int io_uring_alloc_task_context(struct task_struct *task,
	67	struct io_ring_ctx *ctx);
	68
a282967c	69	int io_poll_issue(struct io_kiocb req, struct io_tw_state ts);
9046c641 PB	70	int io_submit_sqes(struct io_ring_ctx *ctx, unsigned int nr);
	71	int io_do_iopoll(struct io_ring_ctx *ctx, bool force_nonspin);
	72	void io_free_batch_list(struct io_ring_ctx ctx, struct io_wq_work_node node);
	73	int io_req_prep_async(struct io_kiocb *req);
	74
	75	struct io_wq_work io_wq_free_work(struct io_wq_work work);
	76	void io_wq_submit_work(struct io_wq_work *work);
	77
	78	void io_free_req(struct io_kiocb *req);
	79	void io_queue_next(struct io_kiocb *req);
63809137	80	void io_task_refs_refill(struct io_uring_task *tctx);
bd1a3783	81	bool __io_alloc_req_refill(struct io_ring_ctx *ctx);
9046c641 PB	82
	83	bool io_match_task_safe(struct io_kiocb head, struct task_struct task,
	84	bool cancel_all);
	85
f26cc959 PB	86	#define io_lockdep_assert_cq_locked(ctx) \
	87	do { \
	88	if (ctx->flags & IORING_SETUP_IOPOLL) { \
	89	lockdep_assert_held(&ctx->uring_lock); \
	90	} else if (!ctx->task_complete) { \
	91	lockdep_assert_held(&ctx->completion_lock); \
	92	} else if (ctx->submitter_task->flags & PF_EXITING) { \
	93	lockdep_assert(current_work()); \
	94	} else { \
	95	lockdep_assert(current == ctx->submitter_task); \
	96	} \
	97	} while (0)
	98
e52d2e58 PB	99	static inline void io_req_task_work_add(struct io_kiocb *req)
e52d2e58 PB	100	{
8501fe70	101	__io_req_task_work_add(req, 0);
e52d2e58 PB	102	}
e52d2e58 PB	103
9046c641 PB	104	#define io_for_each_link(pos, head) \
9046c641 PB	105	for (pos = (head); pos; pos = pos->link)
f3b44f92	106
25399321 PB	107	void io_cq_unlock_post(struct io_ring_ctx *ctx);
25399321 PB	108
aa1df3a3 PB	109	static inline struct io_uring_cqe io_get_cqe_overflow(struct io_ring_ctx ctx,
aa1df3a3 PB	110	bool overflow)
f3b44f92	111	{
f26cc959 PB	112	io_lockdep_assert_cq_locked(ctx);
f26cc959 PB	113
f3b44f92 JA	114	if (likely(ctx->cqe_cached < ctx->cqe_sentinel)) {
	115	struct io_uring_cqe *cqe = ctx->cqe_cached;
	116
f3b44f92 JA	117	ctx->cached_cq_tail++;
f3b44f92 JA	118	ctx->cqe_cached++;
b3659a65 PB	119	if (ctx->flags & IORING_SETUP_CQE32)
b3659a65 PB	120	ctx->cqe_cached++;
f3b44f92 JA	121	return cqe;
	122	}
	123
aa1df3a3 PB	124	return __io_get_cqe(ctx, overflow);
	125	}
	126
	127	static inline struct io_uring_cqe io_get_cqe(struct io_ring_ctx ctx)
	128	{
	129	return io_get_cqe_overflow(ctx, false);
f3b44f92 JA	130	}
	131
	132	static inline bool __io_fill_cqe_req(struct io_ring_ctx *ctx,
	133	struct io_kiocb *req)
	134	{
	135	struct io_uring_cqe *cqe;
	136
e8c328c3 PB	137	/*
	138	* If we can't get a cq entry, userspace overflowed the
	139	* submission (by quite a lot). Increment the overflow count in
	140	* the ring.
	141	*/
	142	cqe = io_get_cqe(ctx);
	143	if (unlikely(!cqe))
f66f7342	144	return false;
e0486f3f DY	145
	146	trace_io_uring_complete(req->ctx, req, req->cqe.user_data,
	147	req->cqe.res, req->cqe.flags,
	148	(req->flags & REQ_F_CQE32_INIT) ? req->extra1 : 0,
	149	(req->flags & REQ_F_CQE32_INIT) ? req->extra2 : 0);
	150
e8c328c3 PB	151	memcpy(cqe, &req->cqe, sizeof(*cqe));
	152
	153	if (ctx->flags & IORING_SETUP_CQE32) {
f3b44f92 JA	154	u64 extra1 = 0, extra2 = 0;
	155
	156	if (req->flags & REQ_F_CQE32_INIT) {
	157	extra1 = req->extra1;
	158	extra2 = req->extra2;
	159	}
	160
e8c328c3 PB	161	WRITE_ONCE(cqe->big_cqe[0], extra1);
e8c328c3 PB	162	WRITE_ONCE(cqe->big_cqe[1], extra2);
f3b44f92	163	}
e8c328c3	164	return true;
f3b44f92 JA	165	}
f3b44f92 JA	166
f66f7342 PB	167	static inline bool io_fill_cqe_req(struct io_ring_ctx *ctx,
	168	struct io_kiocb *req)
	169	{
	170	if (likely(__io_fill_cqe_req(ctx, req)))
	171	return true;
	172	return io_req_cqe_overflow(req);
	173	}
	174
531113bb JA	175	static inline void req_set_fail(struct io_kiocb *req)
	176	{
	177	req->flags \|= REQ_F_FAIL;
	178	if (req->flags & REQ_F_CQE_SKIP) {
	179	req->flags &= ~REQ_F_CQE_SKIP;
	180	req->flags \|= REQ_F_SKIP_LINK_CQES;
	181	}
	182	}
	183
de23077e JA	184	static inline void io_req_set_res(struct io_kiocb *req, s32 res, u32 cflags)
	185	{
	186	req->cqe.res = res;
	187	req->cqe.flags = cflags;
	188	}
	189
99f15d8d JA	190	static inline bool req_has_async_data(struct io_kiocb *req)
	191	{
	192	return req->flags & REQ_F_ASYNC_DATA;
	193	}
	194
531113bb JA	195	static inline void io_put_file(struct file *file)
	196	{
	197	if (file)
	198	fput(file);
	199	}
	200
cd40cae2 JA	201	static inline void io_ring_submit_unlock(struct io_ring_ctx *ctx,
	202	unsigned issue_flags)
	203	{
	204	lockdep_assert_held(&ctx->uring_lock);
	205	if (issue_flags & IO_URING_F_UNLOCKED)
	206	mutex_unlock(&ctx->uring_lock);
	207	}
	208
	209	static inline void io_ring_submit_lock(struct io_ring_ctx *ctx,
	210	unsigned issue_flags)
	211	{
	212	/*
	213	* "Normal" inline submissions always hold the uring_lock, since we
	214	* grab it from the system call. Same is true for the SQPOLL offload.
	215	* The only exception is when we've detached the request and issue it
	216	* from an async worker thread, grab the lock for that case.
	217	*/
	218	if (issue_flags & IO_URING_F_UNLOCKED)
	219	mutex_lock(&ctx->uring_lock);
	220	lockdep_assert_held(&ctx->uring_lock);
	221	}
	222
f9ead18c JA	223	static inline void io_commit_cqring(struct io_ring_ctx *ctx)
	224	{
	225	/* order cqe stores with ring update */
	226	smp_store_release(&ctx->rings->cq.tail, ctx->cached_cq_tail);
	227	}
	228
7b235dd8 PB	229	static inline void io_poll_wq_wake(struct io_ring_ctx *ctx)
7b235dd8 PB	230	{
bca39f39	231	if (wq_has_sleeper(&ctx->poll_wq))
7b235dd8 PB	232	__wake_up(&ctx->poll_wq, TASK_NORMAL, 0,
	233	poll_to_key(EPOLL_URING_WAKE \| EPOLLIN));
	234	}
	235
6e7248ad	236	static inline void io_cqring_wake(struct io_ring_ctx *ctx)
f3b44f92 JA	237	{
f3b44f92 JA	238	/*
44648532 JA	239	* Trigger waitqueue handler on all waiters on our waitqueue. This
	240	* won't necessarily wake up all the tasks, io_should_wake() will make
	241	* that decision.
	242	*
	243	* Pass in EPOLLIN\|EPOLL_URING_WAKE as the poll wakeup key. The latter
	244	* set in the mask so that if we recurse back into our own poll
	245	* waitqueue handlers, we know we have a dependency between eventfd or
	246	* epoll and should terminate multishot poll at that point.
f3b44f92	247	*/
6e7248ad	248	if (wq_has_sleeper(&ctx->cq_wait))
44648532 JA	249	__wake_up(&ctx->cq_wait, TASK_NORMAL, 0,
44648532 JA	250	poll_to_key(EPOLL_URING_WAKE \| EPOLLIN));
f3b44f92 JA	251	}
f3b44f92 JA	252
17437f31 JA	253	static inline bool io_sqring_full(struct io_ring_ctx *ctx)
	254	{
	255	struct io_rings *r = ctx->rings;
	256
	257	return READ_ONCE(r->sq.tail) - ctx->cached_sq_head == ctx->sq_entries;
	258	}
	259
	260	static inline unsigned int io_sqring_entries(struct io_ring_ctx *ctx)
	261	{
	262	struct io_rings *rings = ctx->rings;
e3ef728f	263	unsigned int entries;
17437f31 JA	264
17437f31 JA	265	/* make sure SQ entry isn't read before tail */
e3ef728f JA	266	entries = smp_load_acquire(&rings->sq.tail) - ctx->cached_sq_head;
e3ef728f JA	267	return min(entries, ctx->sq_entries);
17437f31 JA	268	}
17437f31 JA	269
c0e0d6ba	270	static inline int io_run_task_work(void)
17437f31	271	{
7cfe7a09 JA	272	/*
	273	* Always check-and-clear the task_work notification signal. With how
	274	* signaling works for task_work, we can find it set with nothing to
	275	* run. We need to clear it for that case, like get_signal() does.
	276	*/
	277	if (test_thread_flag(TIF_NOTIFY_SIGNAL))
	278	clear_notify_signal();
b5d3ae20 JA	279	/*
	280	* PF_IO_WORKER never returns to userspace, so check here if we have
	281	* notify work that needs processing.
	282	*/
	283	if (current->flags & PF_IO_WORKER &&
2f2bb1ff JA	284	test_thread_flag(TIF_NOTIFY_RESUME)) {
2f2bb1ff JA	285	__set_current_state(TASK_RUNNING);
b5d3ae20	286	resume_user_mode_work(NULL);
2f2bb1ff	287	}
46a525e1	288	if (task_work_pending(current)) {
17437f31	289	__set_current_state(TASK_RUNNING);
46a525e1	290	task_work_run();
c0e0d6ba	291	return 1;
17437f31 JA	292	}
17437f31 JA	293
c0e0d6ba DY	294	return 0;
	295	}
	296
dac6a0ea JA	297	static inline bool io_task_work_pending(struct io_ring_ctx *ctx)
dac6a0ea JA	298	{
6434ec01	299	return task_work_pending(current) \|\| !wq_list_empty(&ctx->work_llist);
dac6a0ea JA	300	}
dac6a0ea JA	301
a282967c	302	static inline void io_tw_lock(struct io_ring_ctx ctx, struct io_tw_state ts)
aa1e90f6	303	{
a282967c	304	if (!ts->locked) {
aa1e90f6	305	mutex_lock(&ctx->uring_lock);
a282967c	306	ts->locked = true;
aa1e90f6 PB	307	}
	308	}
	309
9da070b1 PB	310	/*
	311	* Don't complete immediately but use deferred completion infrastructure.
	312	* Protected by ->uring_lock and can only be used either with
	313	* IO_URING_F_COMPLETE_DEFER or inside a tw handler holding the mutex.
	314	*/
	315	static inline void io_req_complete_defer(struct io_kiocb *req)
	316	__must_hold(&req->ctx->uring_lock)
aa1e90f6 PB	317	{
	318	struct io_submit_state *state = &req->ctx->submit_state;
	319
9da070b1 PB	320	lockdep_assert_held(&req->ctx->uring_lock);
9da070b1 PB	321
aa1e90f6 PB	322	wq_list_add_tail(&req->comp_list, &state->compl_reqs);
	323	}
	324
46929b08 PB	325	static inline void io_commit_cqring_flush(struct io_ring_ctx *ctx)
46929b08 PB	326	{
bca39f39 PB	327	if (unlikely(ctx->off_timeout_used \|\| ctx->drain_active \|\|
bca39f39 PB	328	ctx->has_evfd \|\| ctx->poll_activated))
46929b08 PB	329	__io_commit_cqring_flush(ctx);
	330	}
	331
63809137 PB	332	static inline void io_get_task_refs(int nr)
	333	{
	334	struct io_uring_task *tctx = current->io_uring;
	335
	336	tctx->cached_refs -= nr;
	337	if (unlikely(tctx->cached_refs < 0))
	338	io_task_refs_refill(tctx);
	339	}
	340
bd1a3783 PB	341	static inline bool io_req_cache_empty(struct io_ring_ctx *ctx)
	342	{
	343	return !ctx->submit_state.free_list.next;
	344	}
	345
c1755c25 BL	346	extern struct kmem_cache *req_cachep;
c1755c25 BL	347
c8576f3e	348	static inline struct io_kiocb io_extract_req(struct io_ring_ctx ctx)
bd1a3783	349	{
c1755c25	350	struct io_kiocb *req;
bd1a3783	351
c1755c25 BL	352	req = container_of(ctx->submit_state.free_list.next, struct io_kiocb, comp_list);
	353	kasan_unpoison_object_data(req_cachep, req);
	354	wq_stack_extract(&ctx->submit_state.free_list);
	355	return req;
bd1a3783 PB	356	}
bd1a3783 PB	357
c8576f3e PB	358	static inline bool io_alloc_req(struct io_ring_ctx ctx, struct io_kiocb *req)
	359	{
	360	if (unlikely(io_req_cache_empty(ctx))) {
	361	if (!__io_alloc_req_refill(ctx))
	362	return false;
	363	}
	364	*req = io_extract_req(ctx);
	365	return true;
	366	}
	367
140102ae PB	368	static inline bool io_allowed_defer_tw_run(struct io_ring_ctx *ctx)
	369	{
	370	return likely(ctx->submitter_task == current);
	371	}
	372
76de6749 PB	373	static inline bool io_allowed_run_tw(struct io_ring_ctx *ctx)
76de6749 PB	374	{
6567506b PB	375	return likely(!(ctx->flags & IORING_SETUP_DEFER_TASKRUN) \|\|
6567506b PB	376	ctx->submitter_task == current);
76de6749 PB	377	}
76de6749 PB	378
833b5dff PB	379	static inline void io_req_queue_tw_complete(struct io_kiocb *req, s32 res)
	380	{
	381	io_req_set_res(req, res, 0);
	382	req->io_task_work.func = io_req_task_complete;
	383	io_req_task_work_add(req);
	384	}
	385
de23077e	386	#endif