[thirdparty/linux.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 {
	/*
	 * A hint to not wake right away but delay until there are enough of
	 * tw's queued to match the number of CQEs the task is waiting for.
	 *
	 * Must not be used wirh requests generating more than one CQE.
	 * It's also ignored unless IORING_SETUP_DEFER_TASKRUN is set.
	 */
	IOU_F_TWQ_LAZY_WAKE			= 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(const struct io_kiocb *req, bool defer, 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);

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_ring_add_registered_file(struct io_uring_task *tctx, struct file *file,
				     int start, int end);

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 {								\
		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->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)

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);
}

/*
 * 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);
}
#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	18	enum {
8751d154 PB	19	/*
	20	* A hint to not wake right away but delay until there are enough of
	21	* tw's queued to match the number of CQEs the task is waiting for.
	22	*
	23	* Must not be used wirh requests generating more than one CQE.
	24	* It's also ignored unless IORING_SETUP_DEFER_TASKRUN is set.
	25	*/
91c7884a	26	IOU_F_TWQ_LAZY_WAKE = 1,
8501fe70 PB	27	};
8501fe70 PB	28
97b388d7 JA	29	enum {
	30	IOU_OK = 0,
	31	IOU_ISSUE_SKIP_COMPLETE = -EIOCBQUEUED,
114eccdf DY	32
114eccdf DY	33	/*
91482864 PB	34	* Intended only when both IO_URING_F_MULTISHOT is passed
91482864 PB	35	* to indicate to the poll runner that multishot should be
114eccdf DY	36	* removed and the result is set on req->cqe.res.
	37	*/
	38	IOU_STOP_MULTISHOT = -ECANCELED,
97b388d7 JA	39	};
97b388d7 JA	40
aa1df3a3	41	struct io_uring_cqe __io_get_cqe(struct io_ring_ctx ctx, bool overflow);
68494a65	42	bool io_req_cqe_overflow(struct io_kiocb *req);
c0e0d6ba	43	int io_run_task_work_sig(struct io_ring_ctx *ctx);
973fc83f	44	void io_req_defer_failed(struct io_kiocb *req, s32 res);
1bec951c	45	void io_req_complete_post(struct io_kiocb *req, unsigned issue_flags);
b529c96a	46	bool io_post_aux_cqe(struct io_ring_ctx *ctx, u64 user_data, s32 res, u32 cflags);
d86eaed1	47	bool io_aux_cqe(const struct io_kiocb *req, bool defer, s32 res, u32 cflags,
9b8c5475	48	bool allow_overflow);
9046c641 PB	49	void __io_commit_cqring_flush(struct io_ring_ctx *ctx);
	50
	51	struct page *io_pin_pages(unsigned long ubuf, unsigned long len, int npages);
	52
	53	struct file io_file_get_normal(struct io_kiocb req, int fd);
	54	struct file io_file_get_fixed(struct io_kiocb req, int fd,
	55	unsigned issue_flags);
	56
8501fe70	57	void __io_req_task_work_add(struct io_kiocb *req, unsigned flags);
9046c641 PB	58	bool io_is_uring_fops(struct file *file);
9046c641 PB	59	bool io_alloc_async_data(struct io_kiocb *req);
9046c641	60	void io_req_task_queue(struct io_kiocb *req);
a282967c PB	61	void io_queue_iowq(struct io_kiocb req, struct io_tw_state ts_dont_use);
a282967c PB	62	void io_req_task_complete(struct io_kiocb req, struct io_tw_state ts);
9046c641	63	void io_req_task_queue_fail(struct io_kiocb *req, int ret);
a282967c	64	void io_req_task_submit(struct io_kiocb req, struct io_tw_state ts);
9046c641 PB	65	void tctx_task_work(struct callback_head *cb);
	66	__cold void io_uring_cancel_generic(bool cancel_all, struct io_sq_data *sqd);
	67	int io_uring_alloc_task_context(struct task_struct *task,
	68	struct io_ring_ctx *ctx);
	69
6e76ac59 JT	70	int io_ring_add_registered_file(struct io_uring_task tctx, struct file file,
	71	int start, int end);
	72
a282967c	73	int io_poll_issue(struct io_kiocb req, struct io_tw_state ts);
9046c641 PB	74	int io_submit_sqes(struct io_ring_ctx *ctx, unsigned int nr);
	75	int io_do_iopoll(struct io_ring_ctx *ctx, bool force_nonspin);
	76	void io_free_batch_list(struct io_ring_ctx ctx, struct io_wq_work_node node);
	77	int io_req_prep_async(struct io_kiocb *req);
	78
	79	struct io_wq_work io_wq_free_work(struct io_wq_work work);
	80	void io_wq_submit_work(struct io_wq_work *work);
	81
	82	void io_free_req(struct io_kiocb *req);
	83	void io_queue_next(struct io_kiocb *req);
63809137	84	void io_task_refs_refill(struct io_uring_task *tctx);
bd1a3783	85	bool __io_alloc_req_refill(struct io_ring_ctx *ctx);
9046c641 PB	86
	87	bool io_match_task_safe(struct io_kiocb head, struct task_struct task,
	88	bool cancel_all);
	89
f26cc959 PB	90	#define io_lockdep_assert_cq_locked(ctx) \
f26cc959 PB	91	do { \
8ce4269e PB	92	lockdep_assert(in_task()); \
8ce4269e PB	93	\
f26cc959 PB	94	if (ctx->flags & IORING_SETUP_IOPOLL) { \
	95	lockdep_assert_held(&ctx->uring_lock); \
	96	} else if (!ctx->task_complete) { \
	97	lockdep_assert_held(&ctx->completion_lock); \
	98	} else if (ctx->submitter_task->flags & PF_EXITING) { \
	99	lockdep_assert(current_work()); \
	100	} else { \
	101	lockdep_assert(current == ctx->submitter_task); \
	102	} \
	103	} while (0)
	104
e52d2e58 PB	105	static inline void io_req_task_work_add(struct io_kiocb *req)
e52d2e58 PB	106	{
8501fe70	107	__io_req_task_work_add(req, 0);
e52d2e58 PB	108	}
e52d2e58 PB	109
9046c641 PB	110	#define io_for_each_link(pos, head) \
9046c641 PB	111	for (pos = (head); pos; pos = pos->link)
f3b44f92	112
aa1df3a3 PB	113	static inline struct io_uring_cqe io_get_cqe_overflow(struct io_ring_ctx ctx,
aa1df3a3 PB	114	bool overflow)
f3b44f92	115	{
f26cc959 PB	116	io_lockdep_assert_cq_locked(ctx);
f26cc959 PB	117
f3b44f92 JA	118	if (likely(ctx->cqe_cached < ctx->cqe_sentinel)) {
	119	struct io_uring_cqe *cqe = ctx->cqe_cached;
	120
f3b44f92 JA	121	ctx->cached_cq_tail++;
f3b44f92 JA	122	ctx->cqe_cached++;
b3659a65 PB	123	if (ctx->flags & IORING_SETUP_CQE32)
b3659a65 PB	124	ctx->cqe_cached++;
f3b44f92 JA	125	return cqe;
	126	}
	127
aa1df3a3 PB	128	return __io_get_cqe(ctx, overflow);
	129	}
	130
	131	static inline struct io_uring_cqe io_get_cqe(struct io_ring_ctx ctx)
	132	{
	133	return io_get_cqe_overflow(ctx, false);
f3b44f92 JA	134	}
	135
	136	static inline bool __io_fill_cqe_req(struct io_ring_ctx *ctx,
	137	struct io_kiocb *req)
	138	{
	139	struct io_uring_cqe *cqe;
	140
e8c328c3 PB	141	/*
	142	* If we can't get a cq entry, userspace overflowed the
	143	* submission (by quite a lot). Increment the overflow count in
	144	* the ring.
	145	*/
	146	cqe = io_get_cqe(ctx);
	147	if (unlikely(!cqe))
f66f7342	148	return false;
e0486f3f DY	149
	150	trace_io_uring_complete(req->ctx, req, req->cqe.user_data,
	151	req->cqe.res, req->cqe.flags,
	152	(req->flags & REQ_F_CQE32_INIT) ? req->extra1 : 0,
	153	(req->flags & REQ_F_CQE32_INIT) ? req->extra2 : 0);
	154
e8c328c3 PB	155	memcpy(cqe, &req->cqe, sizeof(*cqe));
	156
	157	if (ctx->flags & IORING_SETUP_CQE32) {
f3b44f92 JA	158	u64 extra1 = 0, extra2 = 0;
	159
	160	if (req->flags & REQ_F_CQE32_INIT) {
	161	extra1 = req->extra1;
	162	extra2 = req->extra2;
	163	}
	164
e8c328c3 PB	165	WRITE_ONCE(cqe->big_cqe[0], extra1);
e8c328c3 PB	166	WRITE_ONCE(cqe->big_cqe[1], extra2);
f3b44f92	167	}
e8c328c3	168	return true;
f3b44f92 JA	169	}
f3b44f92 JA	170
f66f7342 PB	171	static inline bool io_fill_cqe_req(struct io_ring_ctx *ctx,
	172	struct io_kiocb *req)
	173	{
	174	if (likely(__io_fill_cqe_req(ctx, req)))
	175	return true;
	176	return io_req_cqe_overflow(req);
	177	}
	178
531113bb JA	179	static inline void req_set_fail(struct io_kiocb *req)
	180	{
	181	req->flags \|= REQ_F_FAIL;
	182	if (req->flags & REQ_F_CQE_SKIP) {
	183	req->flags &= ~REQ_F_CQE_SKIP;
	184	req->flags \|= REQ_F_SKIP_LINK_CQES;
	185	}
	186	}
	187
de23077e JA	188	static inline void io_req_set_res(struct io_kiocb *req, s32 res, u32 cflags)
	189	{
	190	req->cqe.res = res;
	191	req->cqe.flags = cflags;
	192	}
	193
99f15d8d JA	194	static inline bool req_has_async_data(struct io_kiocb *req)
	195	{
	196	return req->flags & REQ_F_ASYNC_DATA;
	197	}
	198
531113bb JA	199	static inline void io_put_file(struct file *file)
	200	{
	201	if (file)
	202	fput(file);
	203	}
	204
cd40cae2 JA	205	static inline void io_ring_submit_unlock(struct io_ring_ctx *ctx,
	206	unsigned issue_flags)
	207	{
	208	lockdep_assert_held(&ctx->uring_lock);
	209	if (issue_flags & IO_URING_F_UNLOCKED)
	210	mutex_unlock(&ctx->uring_lock);
	211	}
	212
	213	static inline void io_ring_submit_lock(struct io_ring_ctx *ctx,
	214	unsigned issue_flags)
	215	{
	216	/*
	217	* "Normal" inline submissions always hold the uring_lock, since we
	218	* grab it from the system call. Same is true for the SQPOLL offload.
	219	* The only exception is when we've detached the request and issue it
	220	* from an async worker thread, grab the lock for that case.
	221	*/
	222	if (issue_flags & IO_URING_F_UNLOCKED)
	223	mutex_lock(&ctx->uring_lock);
	224	lockdep_assert_held(&ctx->uring_lock);
	225	}
	226
f9ead18c JA	227	static inline void io_commit_cqring(struct io_ring_ctx *ctx)
	228	{
	229	/* order cqe stores with ring update */
	230	smp_store_release(&ctx->rings->cq.tail, ctx->cached_cq_tail);
	231	}
	232
7b235dd8 PB	233	static inline void io_poll_wq_wake(struct io_ring_ctx *ctx)
7b235dd8 PB	234	{
bca39f39	235	if (wq_has_sleeper(&ctx->poll_wq))
7b235dd8 PB	236	__wake_up(&ctx->poll_wq, TASK_NORMAL, 0,
	237	poll_to_key(EPOLL_URING_WAKE \| EPOLLIN));
	238	}
	239
6e7248ad	240	static inline void io_cqring_wake(struct io_ring_ctx *ctx)
f3b44f92 JA	241	{
f3b44f92 JA	242	/*
44648532 JA	243	* Trigger waitqueue handler on all waiters on our waitqueue. This
	244	* won't necessarily wake up all the tasks, io_should_wake() will make
	245	* that decision.
	246	*
	247	* Pass in EPOLLIN\|EPOLL_URING_WAKE as the poll wakeup key. The latter
	248	* set in the mask so that if we recurse back into our own poll
	249	* waitqueue handlers, we know we have a dependency between eventfd or
	250	* epoll and should terminate multishot poll at that point.
f3b44f92	251	*/
6e7248ad	252	if (wq_has_sleeper(&ctx->cq_wait))
44648532 JA	253	__wake_up(&ctx->cq_wait, TASK_NORMAL, 0,
44648532 JA	254	poll_to_key(EPOLL_URING_WAKE \| EPOLLIN));
f3b44f92 JA	255	}
f3b44f92 JA	256
17437f31 JA	257	static inline bool io_sqring_full(struct io_ring_ctx *ctx)
	258	{
	259	struct io_rings *r = ctx->rings;
	260
	261	return READ_ONCE(r->sq.tail) - ctx->cached_sq_head == ctx->sq_entries;
	262	}
	263
	264	static inline unsigned int io_sqring_entries(struct io_ring_ctx *ctx)
	265	{
	266	struct io_rings *rings = ctx->rings;
e3ef728f	267	unsigned int entries;
17437f31 JA	268
17437f31 JA	269	/* make sure SQ entry isn't read before tail */
e3ef728f JA	270	entries = smp_load_acquire(&rings->sq.tail) - ctx->cached_sq_head;
e3ef728f JA	271	return min(entries, ctx->sq_entries);
17437f31 JA	272	}
17437f31 JA	273
c0e0d6ba	274	static inline int io_run_task_work(void)
17437f31	275	{
7cfe7a09 JA	276	/*
	277	* Always check-and-clear the task_work notification signal. With how
	278	* signaling works for task_work, we can find it set with nothing to
	279	* run. We need to clear it for that case, like get_signal() does.
	280	*/
	281	if (test_thread_flag(TIF_NOTIFY_SIGNAL))
	282	clear_notify_signal();
b5d3ae20 JA	283	/*
	284	* PF_IO_WORKER never returns to userspace, so check here if we have
	285	* notify work that needs processing.
	286	*/
	287	if (current->flags & PF_IO_WORKER &&
2f2bb1ff JA	288	test_thread_flag(TIF_NOTIFY_RESUME)) {
2f2bb1ff JA	289	__set_current_state(TASK_RUNNING);
b5d3ae20	290	resume_user_mode_work(NULL);
2f2bb1ff	291	}
46a525e1	292	if (task_work_pending(current)) {
17437f31	293	__set_current_state(TASK_RUNNING);
46a525e1	294	task_work_run();
c0e0d6ba	295	return 1;
17437f31 JA	296	}
17437f31 JA	297
c0e0d6ba DY	298	return 0;
	299	}
	300
dac6a0ea JA	301	static inline bool io_task_work_pending(struct io_ring_ctx *ctx)
dac6a0ea JA	302	{
6434ec01	303	return task_work_pending(current) \|\| !wq_list_empty(&ctx->work_llist);
dac6a0ea JA	304	}
dac6a0ea JA	305
a282967c	306	static inline void io_tw_lock(struct io_ring_ctx ctx, struct io_tw_state ts)
aa1e90f6	307	{
a282967c	308	if (!ts->locked) {
aa1e90f6	309	mutex_lock(&ctx->uring_lock);
a282967c	310	ts->locked = true;
aa1e90f6 PB	311	}
	312	}
	313
9da070b1 PB	314	/*
	315	* Don't complete immediately but use deferred completion infrastructure.
	316	* Protected by ->uring_lock and can only be used either with
	317	* IO_URING_F_COMPLETE_DEFER or inside a tw handler holding the mutex.
	318	*/
	319	static inline void io_req_complete_defer(struct io_kiocb *req)
	320	__must_hold(&req->ctx->uring_lock)
aa1e90f6 PB	321	{
	322	struct io_submit_state *state = &req->ctx->submit_state;
	323
9da070b1 PB	324	lockdep_assert_held(&req->ctx->uring_lock);
9da070b1 PB	325
aa1e90f6 PB	326	wq_list_add_tail(&req->comp_list, &state->compl_reqs);
	327	}
	328
46929b08 PB	329	static inline void io_commit_cqring_flush(struct io_ring_ctx *ctx)
46929b08 PB	330	{
bca39f39 PB	331	if (unlikely(ctx->off_timeout_used \|\| ctx->drain_active \|\|
bca39f39 PB	332	ctx->has_evfd \|\| ctx->poll_activated))
46929b08 PB	333	__io_commit_cqring_flush(ctx);
	334	}
	335
63809137 PB	336	static inline void io_get_task_refs(int nr)
	337	{
	338	struct io_uring_task *tctx = current->io_uring;
	339
	340	tctx->cached_refs -= nr;
	341	if (unlikely(tctx->cached_refs < 0))
	342	io_task_refs_refill(tctx);
	343	}
	344
bd1a3783 PB	345	static inline bool io_req_cache_empty(struct io_ring_ctx *ctx)
	346	{
	347	return !ctx->submit_state.free_list.next;
	348	}
	349
c1755c25 BL	350	extern struct kmem_cache *req_cachep;
c1755c25 BL	351
c8576f3e	352	static inline struct io_kiocb io_extract_req(struct io_ring_ctx ctx)
bd1a3783	353	{
c1755c25	354	struct io_kiocb *req;
bd1a3783	355
c1755c25 BL	356	req = container_of(ctx->submit_state.free_list.next, struct io_kiocb, comp_list);
	357	kasan_unpoison_object_data(req_cachep, req);
	358	wq_stack_extract(&ctx->submit_state.free_list);
	359	return req;
bd1a3783 PB	360	}
bd1a3783 PB	361
c8576f3e PB	362	static inline bool io_alloc_req(struct io_ring_ctx ctx, struct io_kiocb *req)
	363	{
	364	if (unlikely(io_req_cache_empty(ctx))) {
	365	if (!__io_alloc_req_refill(ctx))
	366	return false;
	367	}
	368	*req = io_extract_req(ctx);
	369	return true;
	370	}
	371
140102ae PB	372	static inline bool io_allowed_defer_tw_run(struct io_ring_ctx *ctx)
	373	{
	374	return likely(ctx->submitter_task == current);
	375	}
	376
76de6749 PB	377	static inline bool io_allowed_run_tw(struct io_ring_ctx *ctx)
76de6749 PB	378	{
6567506b PB	379	return likely(!(ctx->flags & IORING_SETUP_DEFER_TASKRUN) \|\|
6567506b PB	380	ctx->submitter_task == current);
76de6749 PB	381	}
76de6749 PB	382
833b5dff PB	383	static inline void io_req_queue_tw_complete(struct io_kiocb *req, s32 res)
	384	{
	385	io_req_set_res(req, res, 0);
	386	req->io_task_work.func = io_req_task_complete;
	387	io_req_task_work_add(req);
	388	}
	389
96c7d4f8 BL	390	/*
	391	* IORING_SETUP_SQE128 contexts allocate twice the normal SQE size for each
	392	* slot.
	393	*/
	394	static inline size_t uring_sqe_size(struct io_ring_ctx *ctx)
	395	{
	396	if (ctx->flags & IORING_SETUP_SQE128)
	397	return 2 * sizeof(struct io_uring_sqe);
	398	return sizeof(struct io_uring_sqe);
	399	}
de23077e	400	#endif