[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 {
	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_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();
	/*
	 * 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))
		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, 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);
}

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;
}

extern struct kmem_cache *req_cachep;

static inline struct io_kiocb *io_alloc_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_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
97b388d7 JA	18	enum {
	19	IOU_OK = 0,
	20	IOU_ISSUE_SKIP_COMPLETE = -EIOCBQUEUED,
114eccdf DY	21
114eccdf DY	22	/*
91482864 PB	23	* Intended only when both IO_URING_F_MULTISHOT is passed
91482864 PB	24	* to indicate to the poll runner that multishot should be
114eccdf DY	25	* removed and the result is set on req->cqe.res.
	26	*/
	27	IOU_STOP_MULTISHOT = -ECANCELED,
97b388d7 JA	28	};
97b388d7 JA	29
aa1df3a3	30	struct io_uring_cqe __io_get_cqe(struct io_ring_ctx ctx, bool overflow);
68494a65	31	bool io_req_cqe_overflow(struct io_kiocb *req);
c0e0d6ba	32	int io_run_task_work_sig(struct io_ring_ctx *ctx);
973fc83f	33	void io_req_defer_failed(struct io_kiocb *req, s32 res);
1bec951c	34	void io_req_complete_post(struct io_kiocb *req, unsigned issue_flags);
b529c96a	35	bool io_post_aux_cqe(struct io_ring_ctx *ctx, u64 user_data, s32 res, u32 cflags);
9b8c5475 DY	36	bool io_aux_cqe(struct io_ring_ctx *ctx, bool defer, u64 user_data, s32 res, u32 cflags,
9b8c5475 DY	37	bool allow_overflow);
9046c641 PB	38	void __io_commit_cqring_flush(struct io_ring_ctx *ctx);
	39
	40	struct page *io_pin_pages(unsigned long ubuf, unsigned long len, int npages);
	41
	42	struct file io_file_get_normal(struct io_kiocb req, int fd);
	43	struct file io_file_get_fixed(struct io_kiocb req, int fd,
	44	unsigned issue_flags);
	45
f6b543fd JA	46	static inline bool io_req_ffs_set(struct io_kiocb *req)
	47	{
	48	return req->flags & REQ_F_FIXED_FILE;
	49	}
	50
e52d2e58	51	void __io_req_task_work_add(struct io_kiocb *req, bool allow_local);
9046c641 PB	52	bool io_is_uring_fops(struct file *file);
9046c641 PB	53	bool io_alloc_async_data(struct io_kiocb *req);
9046c641 PB	54	void io_req_task_queue(struct io_kiocb *req);
	55	void io_queue_iowq(struct io_kiocb req, bool dont_use);
	56	void io_req_task_complete(struct io_kiocb req, bool locked);
	57	void io_req_task_queue_fail(struct io_kiocb *req, int ret);
	58	void io_req_task_submit(struct io_kiocb req, bool locked);
	59	void tctx_task_work(struct callback_head *cb);
	60	__cold void io_uring_cancel_generic(bool cancel_all, struct io_sq_data *sqd);
	61	int io_uring_alloc_task_context(struct task_struct *task,
	62	struct io_ring_ctx *ctx);
	63
	64	int io_poll_issue(struct io_kiocb req, bool locked);
	65	int io_submit_sqes(struct io_ring_ctx *ctx, unsigned int nr);
	66	int io_do_iopoll(struct io_ring_ctx *ctx, bool force_nonspin);
	67	void io_free_batch_list(struct io_ring_ctx ctx, struct io_wq_work_node node);
	68	int io_req_prep_async(struct io_kiocb *req);
	69
	70	struct io_wq_work io_wq_free_work(struct io_wq_work work);
	71	void io_wq_submit_work(struct io_wq_work *work);
	72
	73	void io_free_req(struct io_kiocb *req);
	74	void io_queue_next(struct io_kiocb *req);
63809137	75	void io_task_refs_refill(struct io_uring_task *tctx);
bd1a3783	76	bool __io_alloc_req_refill(struct io_ring_ctx *ctx);
9046c641 PB	77
	78	bool io_match_task_safe(struct io_kiocb head, struct task_struct task,
	79	bool cancel_all);
	80
f26cc959 PB	81	#define io_lockdep_assert_cq_locked(ctx) \
	82	do { \
	83	if (ctx->flags & IORING_SETUP_IOPOLL) { \
	84	lockdep_assert_held(&ctx->uring_lock); \
	85	} else if (!ctx->task_complete) { \
	86	lockdep_assert_held(&ctx->completion_lock); \
	87	} else if (ctx->submitter_task->flags & PF_EXITING) { \
	88	lockdep_assert(current_work()); \
	89	} else { \
	90	lockdep_assert(current == ctx->submitter_task); \
	91	} \
	92	} while (0)
	93
e52d2e58 PB	94	static inline void io_req_task_work_add(struct io_kiocb *req)
	95	{
	96	__io_req_task_work_add(req, true);
	97	}
	98
9046c641 PB	99	#define io_for_each_link(pos, head) \
9046c641 PB	100	for (pos = (head); pos; pos = pos->link)
f3b44f92	101
25399321 PB	102	void io_cq_unlock_post(struct io_ring_ctx *ctx);
25399321 PB	103
aa1df3a3 PB	104	static inline struct io_uring_cqe io_get_cqe_overflow(struct io_ring_ctx ctx,
aa1df3a3 PB	105	bool overflow)
f3b44f92	106	{
f26cc959 PB	107	io_lockdep_assert_cq_locked(ctx);
f26cc959 PB	108
f3b44f92 JA	109	if (likely(ctx->cqe_cached < ctx->cqe_sentinel)) {
	110	struct io_uring_cqe *cqe = ctx->cqe_cached;
	111
f3b44f92 JA	112	ctx->cached_cq_tail++;
f3b44f92 JA	113	ctx->cqe_cached++;
b3659a65 PB	114	if (ctx->flags & IORING_SETUP_CQE32)
b3659a65 PB	115	ctx->cqe_cached++;
f3b44f92 JA	116	return cqe;
	117	}
	118
aa1df3a3 PB	119	return __io_get_cqe(ctx, overflow);
	120	}
	121
	122	static inline struct io_uring_cqe io_get_cqe(struct io_ring_ctx ctx)
	123	{
	124	return io_get_cqe_overflow(ctx, false);
f3b44f92 JA	125	}
	126
	127	static inline bool __io_fill_cqe_req(struct io_ring_ctx *ctx,
	128	struct io_kiocb *req)
	129	{
	130	struct io_uring_cqe *cqe;
	131
e8c328c3 PB	132	/*
	133	* If we can't get a cq entry, userspace overflowed the
	134	* submission (by quite a lot). Increment the overflow count in
	135	* the ring.
	136	*/
	137	cqe = io_get_cqe(ctx);
	138	if (unlikely(!cqe))
f66f7342	139	return false;
e0486f3f DY	140
	141	trace_io_uring_complete(req->ctx, req, req->cqe.user_data,
	142	req->cqe.res, req->cqe.flags,
	143	(req->flags & REQ_F_CQE32_INIT) ? req->extra1 : 0,
	144	(req->flags & REQ_F_CQE32_INIT) ? req->extra2 : 0);
	145
e8c328c3 PB	146	memcpy(cqe, &req->cqe, sizeof(*cqe));
	147
	148	if (ctx->flags & IORING_SETUP_CQE32) {
f3b44f92 JA	149	u64 extra1 = 0, extra2 = 0;
	150
	151	if (req->flags & REQ_F_CQE32_INIT) {
	152	extra1 = req->extra1;
	153	extra2 = req->extra2;
	154	}
	155
e8c328c3 PB	156	WRITE_ONCE(cqe->big_cqe[0], extra1);
e8c328c3 PB	157	WRITE_ONCE(cqe->big_cqe[1], extra2);
f3b44f92	158	}
e8c328c3	159	return true;
f3b44f92 JA	160	}
f3b44f92 JA	161
f66f7342 PB	162	static inline bool io_fill_cqe_req(struct io_ring_ctx *ctx,
	163	struct io_kiocb *req)
	164	{
	165	if (likely(__io_fill_cqe_req(ctx, req)))
	166	return true;
	167	return io_req_cqe_overflow(req);
	168	}
	169
531113bb JA	170	static inline void req_set_fail(struct io_kiocb *req)
	171	{
	172	req->flags \|= REQ_F_FAIL;
	173	if (req->flags & REQ_F_CQE_SKIP) {
	174	req->flags &= ~REQ_F_CQE_SKIP;
	175	req->flags \|= REQ_F_SKIP_LINK_CQES;
	176	}
	177	}
	178
de23077e JA	179	static inline void io_req_set_res(struct io_kiocb *req, s32 res, u32 cflags)
	180	{
	181	req->cqe.res = res;
	182	req->cqe.flags = cflags;
	183	}
	184
99f15d8d JA	185	static inline bool req_has_async_data(struct io_kiocb *req)
	186	{
	187	return req->flags & REQ_F_ASYNC_DATA;
	188	}
	189
531113bb JA	190	static inline void io_put_file(struct file *file)
	191	{
	192	if (file)
	193	fput(file);
	194	}
	195
cd40cae2 JA	196	static inline void io_ring_submit_unlock(struct io_ring_ctx *ctx,
	197	unsigned issue_flags)
	198	{
	199	lockdep_assert_held(&ctx->uring_lock);
	200	if (issue_flags & IO_URING_F_UNLOCKED)
	201	mutex_unlock(&ctx->uring_lock);
	202	}
	203
	204	static inline void io_ring_submit_lock(struct io_ring_ctx *ctx,
	205	unsigned issue_flags)
	206	{
	207	/*
	208	* "Normal" inline submissions always hold the uring_lock, since we
	209	* grab it from the system call. Same is true for the SQPOLL offload.
	210	* The only exception is when we've detached the request and issue it
	211	* from an async worker thread, grab the lock for that case.
	212	*/
	213	if (issue_flags & IO_URING_F_UNLOCKED)
	214	mutex_lock(&ctx->uring_lock);
	215	lockdep_assert_held(&ctx->uring_lock);
	216	}
	217
f9ead18c JA	218	static inline void io_commit_cqring(struct io_ring_ctx *ctx)
	219	{
	220	/* order cqe stores with ring update */
	221	smp_store_release(&ctx->rings->cq.tail, ctx->cached_cq_tail);
	222	}
	223
7b235dd8 PB	224	static inline void io_poll_wq_wake(struct io_ring_ctx *ctx)
7b235dd8 PB	225	{
bca39f39	226	if (wq_has_sleeper(&ctx->poll_wq))
7b235dd8 PB	227	__wake_up(&ctx->poll_wq, TASK_NORMAL, 0,
	228	poll_to_key(EPOLL_URING_WAKE \| EPOLLIN));
	229	}
	230
fc86f9d3 PB	231	/* requires smb_mb() prior, see wq_has_sleeper() */
fc86f9d3 PB	232	static inline void __io_cqring_wake(struct io_ring_ctx *ctx)
f3b44f92 JA	233	{
f3b44f92 JA	234	/*
44648532 JA	235	* Trigger waitqueue handler on all waiters on our waitqueue. This
	236	* won't necessarily wake up all the tasks, io_should_wake() will make
	237	* that decision.
	238	*
	239	* Pass in EPOLLIN\|EPOLL_URING_WAKE as the poll wakeup key. The latter
	240	* set in the mask so that if we recurse back into our own poll
	241	* waitqueue handlers, we know we have a dependency between eventfd or
	242	* epoll and should terminate multishot poll at that point.
f3b44f92	243	*/
fc86f9d3	244	if (waitqueue_active(&ctx->cq_wait))
44648532 JA	245	__wake_up(&ctx->cq_wait, TASK_NORMAL, 0,
44648532 JA	246	poll_to_key(EPOLL_URING_WAKE \| EPOLLIN));
f3b44f92 JA	247	}
f3b44f92 JA	248
fc86f9d3 PB	249	static inline void io_cqring_wake(struct io_ring_ctx *ctx)
	250	{
	251	smp_mb();
	252	__io_cqring_wake(ctx);
	253	}
	254
17437f31 JA	255	static inline bool io_sqring_full(struct io_ring_ctx *ctx)
	256	{
	257	struct io_rings *r = ctx->rings;
	258
	259	return READ_ONCE(r->sq.tail) - ctx->cached_sq_head == ctx->sq_entries;
	260	}
	261
	262	static inline unsigned int io_sqring_entries(struct io_ring_ctx *ctx)
	263	{
	264	struct io_rings *rings = ctx->rings;
	265
	266	/* make sure SQ entry isn't read before tail */
	267	return smp_load_acquire(&rings->sq.tail) - ctx->cached_sq_head;
	268	}
	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 &&
	284	test_thread_flag(TIF_NOTIFY_RESUME))
	285	resume_user_mode_work(NULL);
46a525e1	286	if (task_work_pending(current)) {
17437f31	287	__set_current_state(TASK_RUNNING);
46a525e1	288	task_work_run();
c0e0d6ba	289	return 1;
17437f31 JA	290	}
17437f31 JA	291
c0e0d6ba DY	292	return 0;
	293	}
	294
dac6a0ea JA	295	static inline bool io_task_work_pending(struct io_ring_ctx *ctx)
dac6a0ea JA	296	{
6434ec01	297	return task_work_pending(current) \|\| !wq_list_empty(&ctx->work_llist);
dac6a0ea JA	298	}
dac6a0ea JA	299
aa1e90f6 PB	300	static inline void io_tw_lock(struct io_ring_ctx ctx, bool locked)
	301	{
	302	if (!*locked) {
	303	mutex_lock(&ctx->uring_lock);
	304	*locked = true;
	305	}
	306	}
	307
9da070b1 PB	308	/*
	309	* Don't complete immediately but use deferred completion infrastructure.
	310	* Protected by ->uring_lock and can only be used either with
	311	* IO_URING_F_COMPLETE_DEFER or inside a tw handler holding the mutex.
	312	*/
	313	static inline void io_req_complete_defer(struct io_kiocb *req)
	314	__must_hold(&req->ctx->uring_lock)
aa1e90f6 PB	315	{
	316	struct io_submit_state *state = &req->ctx->submit_state;
	317
9da070b1 PB	318	lockdep_assert_held(&req->ctx->uring_lock);
9da070b1 PB	319
aa1e90f6 PB	320	wq_list_add_tail(&req->comp_list, &state->compl_reqs);
	321	}
	322
46929b08 PB	323	static inline void io_commit_cqring_flush(struct io_ring_ctx *ctx)
46929b08 PB	324	{
bca39f39 PB	325	if (unlikely(ctx->off_timeout_used \|\| ctx->drain_active \|\|
bca39f39 PB	326	ctx->has_evfd \|\| ctx->poll_activated))
46929b08 PB	327	__io_commit_cqring_flush(ctx);
	328	}
	329
63809137 PB	330	static inline void io_get_task_refs(int nr)
	331	{
	332	struct io_uring_task *tctx = current->io_uring;
	333
	334	tctx->cached_refs -= nr;
	335	if (unlikely(tctx->cached_refs < 0))
	336	io_task_refs_refill(tctx);
	337	}
	338
bd1a3783 PB	339	static inline bool io_req_cache_empty(struct io_ring_ctx *ctx)
	340	{
	341	return !ctx->submit_state.free_list.next;
	342	}
	343
	344	static inline bool io_alloc_req_refill(struct io_ring_ctx *ctx)
	345	{
	346	if (unlikely(io_req_cache_empty(ctx)))
	347	return __io_alloc_req_refill(ctx);
	348	return true;
	349	}
	350
c1755c25 BL	351	extern struct kmem_cache *req_cachep;
c1755c25 BL	352
bd1a3783 PB	353	static inline struct io_kiocb io_alloc_req(struct io_ring_ctx ctx)
bd1a3783 PB	354	{
c1755c25	355	struct io_kiocb *req;
bd1a3783	356
c1755c25 BL	357	req = container_of(ctx->submit_state.free_list.next, struct io_kiocb, comp_list);
	358	kasan_unpoison_object_data(req_cachep, req);
	359	wq_stack_extract(&ctx->submit_state.free_list);
	360	return req;
bd1a3783 PB	361	}
bd1a3783 PB	362
140102ae PB	363	static inline bool io_allowed_defer_tw_run(struct io_ring_ctx *ctx)
	364	{
	365	return likely(ctx->submitter_task == current);
	366	}
	367
76de6749 PB	368	static inline bool io_allowed_run_tw(struct io_ring_ctx *ctx)
76de6749 PB	369	{
6567506b PB	370	return likely(!(ctx->flags & IORING_SETUP_DEFER_TASKRUN) \|\|
6567506b PB	371	ctx->submitter_task == current);
76de6749 PB	372	}
76de6749 PB	373
833b5dff PB	374	static inline void io_req_queue_tw_complete(struct io_kiocb *req, s32 res)
	375	{
	376	io_req_set_res(req, res, 0);
	377	req->io_task_work.func = io_req_task_complete;
	378	io_req_task_work_add(req);
	379	}
	380
de23077e	381	#endif