[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,
};

bool io_cqe_cache_refill(struct io_ring_ctx *ctx, bool overflow);
void 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_fill_cqe_req_aux(struct io_kiocb *req, bool defer, s32 res, u32 cflags);
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_submit_flush_completions(struct io_ring_ctx *ctx);
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);

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

/*
 * IORING_SETUP_SQE128 contexts allocate twice the normal SQE size for each
 * slot.
 */
static inline size_t uring_sqe_size(struct io_ring_ctx *ctx)
{
	if (ctx->flags & IORING_SETUP_SQE128)
		return 2 * sizeof(struct io_uring_sqe);
	return sizeof(struct io_uring_sqe);
}
#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
20d6b633	41	bool io_cqe_cache_refill(struct io_ring_ctx *ctx, bool overflow);
056695bf	42	void 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);
b6b2bb58	47	bool io_fill_cqe_req_aux(struct io_kiocb *req, bool defer, s32 res, u32 cflags);
9046c641 PB	48	void __io_commit_cqring_flush(struct io_ring_ctx *ctx);
	49
	50	struct page *io_pin_pages(unsigned long ubuf, unsigned long len, int npages);
	51
	52	struct file io_file_get_normal(struct io_kiocb req, int fd);
	53	struct file io_file_get_fixed(struct io_kiocb req, int fd,
	54	unsigned issue_flags);
	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
6e76ac59 JT	69	int io_ring_add_registered_file(struct io_uring_task tctx, struct file file,
	70	int start, int end);
	71
a282967c	72	int io_poll_issue(struct io_kiocb req, struct io_tw_state ts);
9046c641 PB	73	int io_submit_sqes(struct io_ring_ctx *ctx, unsigned int nr);
9046c641 PB	74	int io_do_iopoll(struct io_ring_ctx *ctx, bool force_nonspin);
ec26c225	75	void __io_submit_flush_completions(struct io_ring_ctx *ctx);
9046c641 PB	76	int io_req_prep_async(struct io_kiocb *req);
	77
	78	struct io_wq_work io_wq_free_work(struct io_wq_work work);
	79	void io_wq_submit_work(struct io_wq_work *work);
	80
	81	void io_free_req(struct io_kiocb *req);
	82	void io_queue_next(struct io_kiocb *req);
63809137	83	void io_task_refs_refill(struct io_uring_task *tctx);
bd1a3783	84	bool __io_alloc_req_refill(struct io_ring_ctx *ctx);
9046c641 PB	85
	86	bool io_match_task_safe(struct io_kiocb head, struct task_struct task,
	87	bool cancel_all);
	88
1658633c JA	89	#if defined(CONFIG_PROVE_LOCKING)
	90	static inline void io_lockdep_assert_cq_locked(struct io_ring_ctx *ctx)
	91	{
	92	lockdep_assert(in_task());
	93
	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) {
	99	/*
	100	* ->submitter_task may be NULL and we can still post a CQE,
	101	* if the ring has been setup with IORING_SETUP_R_DISABLED.
	102	* Not from an SQE, as those cannot be submitted, but via
	103	* updating tagged resources.
	104	*/
	105	if (ctx->submitter_task->flags & PF_EXITING)
	106	lockdep_assert(current_work());
	107	else
	108	lockdep_assert(current == ctx->submitter_task);
	109	}
	110	}
	111	#else
	112	static inline void io_lockdep_assert_cq_locked(struct io_ring_ctx *ctx)
	113	{
	114	}
	115	#endif
f26cc959	116
e52d2e58 PB	117	static inline void io_req_task_work_add(struct io_kiocb *req)
e52d2e58 PB	118	{
8501fe70	119	__io_req_task_work_add(req, 0);
e52d2e58 PB	120	}
e52d2e58 PB	121
9046c641 PB	122	#define io_for_each_link(pos, head) \
9046c641 PB	123	for (pos = (head); pos; pos = pos->link)
f3b44f92	124
59fbc409 PB	125	static inline bool io_get_cqe_overflow(struct io_ring_ctx *ctx,
	126	struct io_uring_cqe **ret,
	127	bool overflow)
f3b44f92	128	{
20d6b633	129	io_lockdep_assert_cq_locked(ctx);
f3b44f92	130
20d6b633 PB	131	if (unlikely(ctx->cqe_cached >= ctx->cqe_sentinel)) {
20d6b633 PB	132	if (unlikely(!io_cqe_cache_refill(ctx, overflow)))
59fbc409	133	return false;
f3b44f92	134	}
59fbc409	135	*ret = ctx->cqe_cached;
20d6b633 PB	136	ctx->cached_cq_tail++;
	137	ctx->cqe_cached++;
	138	if (ctx->flags & IORING_SETUP_CQE32)
	139	ctx->cqe_cached++;
59fbc409	140	return true;
aa1df3a3 PB	141	}
aa1df3a3 PB	142
59fbc409	143	static inline bool io_get_cqe(struct io_ring_ctx ctx, struct io_uring_cqe *ret)
aa1df3a3	144	{
59fbc409	145	return io_get_cqe_overflow(ctx, ret, false);
f3b44f92 JA	146	}
f3b44f92 JA	147
093a650b PB	148	static __always_inline bool io_fill_cqe_req(struct io_ring_ctx *ctx,
093a650b PB	149	struct io_kiocb *req)
f3b44f92 JA	150	{
	151	struct io_uring_cqe *cqe;
	152
e8c328c3 PB	153	/*
	154	* If we can't get a cq entry, userspace overflowed the
	155	* submission (by quite a lot). Increment the overflow count in
	156	* the ring.
	157	*/
59fbc409	158	if (unlikely(!io_get_cqe(ctx, &cqe)))
f66f7342	159	return false;
e0486f3f	160
a0727c73 PB	161	if (trace_io_uring_complete_enabled())
	162	trace_io_uring_complete(req->ctx, req, req->cqe.user_data,
	163	req->cqe.res, req->cqe.flags,
b24c5d75	164	req->big_cqe.extra1, req->big_cqe.extra2);
e0486f3f	165
e8c328c3	166	memcpy(cqe, &req->cqe, sizeof(*cqe));
e8c328c3	167	if (ctx->flags & IORING_SETUP_CQE32) {
b24c5d75 PB	168	memcpy(cqe->big_cqe, &req->big_cqe, sizeof(*cqe));
b24c5d75 PB	169	memset(&req->big_cqe, 0, sizeof(req->big_cqe));
f3b44f92	170	}
e8c328c3	171	return true;
f3b44f92 JA	172	}
f3b44f92 JA	173
531113bb JA	174	static inline void req_set_fail(struct io_kiocb *req)
	175	{
	176	req->flags \|= REQ_F_FAIL;
	177	if (req->flags & REQ_F_CQE_SKIP) {
	178	req->flags &= ~REQ_F_CQE_SKIP;
	179	req->flags \|= REQ_F_SKIP_LINK_CQES;
	180	}
	181	}
	182
de23077e JA	183	static inline void io_req_set_res(struct io_kiocb *req, s32 res, u32 cflags)
	184	{
	185	req->cqe.res = res;
	186	req->cqe.flags = cflags;
	187	}
	188
99f15d8d JA	189	static inline bool req_has_async_data(struct io_kiocb *req)
	190	{
	191	return req->flags & REQ_F_ASYNC_DATA;
	192	}
	193
17bc2837	194	static inline void io_put_file(struct io_kiocb *req)
531113bb	195	{
17bc2837 JA	196	if (!(req->flags & REQ_F_FIXED_FILE) && req->file)
17bc2837 JA	197	fput(req->file);
531113bb JA	198	}
531113bb JA	199
cd40cae2 JA	200	static inline void io_ring_submit_unlock(struct io_ring_ctx *ctx,
	201	unsigned issue_flags)
	202	{
	203	lockdep_assert_held(&ctx->uring_lock);
	204	if (issue_flags & IO_URING_F_UNLOCKED)
	205	mutex_unlock(&ctx->uring_lock);
	206	}
	207
	208	static inline void io_ring_submit_lock(struct io_ring_ctx *ctx,
	209	unsigned issue_flags)
	210	{
	211	/*
	212	* "Normal" inline submissions always hold the uring_lock, since we
	213	* grab it from the system call. Same is true for the SQPOLL offload.
	214	* The only exception is when we've detached the request and issue it
	215	* from an async worker thread, grab the lock for that case.
	216	*/
	217	if (issue_flags & IO_URING_F_UNLOCKED)
	218	mutex_lock(&ctx->uring_lock);
	219	lockdep_assert_held(&ctx->uring_lock);
	220	}
	221
f9ead18c JA	222	static inline void io_commit_cqring(struct io_ring_ctx *ctx)
	223	{
	224	/* order cqe stores with ring update */
	225	smp_store_release(&ctx->rings->cq.tail, ctx->cached_cq_tail);
	226	}
	227
7b235dd8 PB	228	static inline void io_poll_wq_wake(struct io_ring_ctx *ctx)
7b235dd8 PB	229	{
bca39f39	230	if (wq_has_sleeper(&ctx->poll_wq))
7b235dd8 PB	231	__wake_up(&ctx->poll_wq, TASK_NORMAL, 0,
	232	poll_to_key(EPOLL_URING_WAKE \| EPOLLIN));
	233	}
	234
6e7248ad	235	static inline void io_cqring_wake(struct io_ring_ctx *ctx)
f3b44f92 JA	236	{
f3b44f92 JA	237	/*
44648532 JA	238	* Trigger waitqueue handler on all waiters on our waitqueue. This
	239	* won't necessarily wake up all the tasks, io_should_wake() will make
	240	* that decision.
	241	*
	242	* Pass in EPOLLIN\|EPOLL_URING_WAKE as the poll wakeup key. The latter
	243	* set in the mask so that if we recurse back into our own poll
	244	* waitqueue handlers, we know we have a dependency between eventfd or
	245	* epoll and should terminate multishot poll at that point.
f3b44f92	246	*/
6e7248ad	247	if (wq_has_sleeper(&ctx->cq_wait))
44648532 JA	248	__wake_up(&ctx->cq_wait, TASK_NORMAL, 0,
44648532 JA	249	poll_to_key(EPOLL_URING_WAKE \| EPOLLIN));
f3b44f92 JA	250	}
f3b44f92 JA	251
17437f31 JA	252	static inline bool io_sqring_full(struct io_ring_ctx *ctx)
	253	{
	254	struct io_rings *r = ctx->rings;
	255
	256	return READ_ONCE(r->sq.tail) - ctx->cached_sq_head == ctx->sq_entries;
	257	}
	258
	259	static inline unsigned int io_sqring_entries(struct io_ring_ctx *ctx)
	260	{
	261	struct io_rings *rings = ctx->rings;
e3ef728f	262	unsigned int entries;
17437f31 JA	263
17437f31 JA	264	/* make sure SQ entry isn't read before tail */
e3ef728f JA	265	entries = smp_load_acquire(&rings->sq.tail) - ctx->cached_sq_head;
e3ef728f JA	266	return min(entries, ctx->sq_entries);
17437f31 JA	267	}
17437f31 JA	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();
b5d3ae20 JA	278	/*
	279	* PF_IO_WORKER never returns to userspace, so check here if we have
	280	* notify work that needs processing.
	281	*/
	282	if (current->flags & PF_IO_WORKER &&
2f2bb1ff JA	283	test_thread_flag(TIF_NOTIFY_RESUME)) {
2f2bb1ff JA	284	__set_current_state(TASK_RUNNING);
b5d3ae20	285	resume_user_mode_work(NULL);
2f2bb1ff	286	}
46a525e1	287	if (task_work_pending(current)) {
17437f31	288	__set_current_state(TASK_RUNNING);
46a525e1	289	task_work_run();
c0e0d6ba	290	return 1;
17437f31 JA	291	}
17437f31 JA	292
c0e0d6ba DY	293	return 0;
	294	}
	295
dac6a0ea JA	296	static inline bool io_task_work_pending(struct io_ring_ctx *ctx)
dac6a0ea JA	297	{
6434ec01	298	return task_work_pending(current) \|\| !wq_list_empty(&ctx->work_llist);
dac6a0ea JA	299	}
dac6a0ea JA	300
a282967c	301	static inline void io_tw_lock(struct io_ring_ctx ctx, struct io_tw_state ts)
aa1e90f6	302	{
a282967c	303	if (!ts->locked) {
aa1e90f6	304	mutex_lock(&ctx->uring_lock);
a282967c	305	ts->locked = true;
aa1e90f6 PB	306	}
	307	}
	308
9da070b1 PB	309	/*
	310	* Don't complete immediately but use deferred completion infrastructure.
	311	* Protected by ->uring_lock and can only be used either with
	312	* IO_URING_F_COMPLETE_DEFER or inside a tw handler holding the mutex.
	313	*/
	314	static inline void io_req_complete_defer(struct io_kiocb *req)
	315	__must_hold(&req->ctx->uring_lock)
aa1e90f6 PB	316	{
	317	struct io_submit_state *state = &req->ctx->submit_state;
	318
9da070b1 PB	319	lockdep_assert_held(&req->ctx->uring_lock);
9da070b1 PB	320
aa1e90f6 PB	321	wq_list_add_tail(&req->comp_list, &state->compl_reqs);
	322	}
	323
46929b08 PB	324	static inline void io_commit_cqring_flush(struct io_ring_ctx *ctx)
46929b08 PB	325	{
bca39f39 PB	326	if (unlikely(ctx->off_timeout_used \|\| ctx->drain_active \|\|
bca39f39 PB	327	ctx->has_evfd \|\| ctx->poll_activated))
46929b08 PB	328	__io_commit_cqring_flush(ctx);
	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
c1755c25	345	extern struct kmem_cache *req_cachep;
b3a4dbc8	346	extern struct kmem_cache *io_buf_cachep;
c1755c25	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	352	req = container_of(ctx->submit_state.free_list.next, struct io_kiocb, comp_list);
c1755c25 BL	353	wq_stack_extract(&ctx->submit_state.free_list);
c1755c25 BL	354	return req;
bd1a3783 PB	355	}
bd1a3783 PB	356
c8576f3e PB	357	static inline bool io_alloc_req(struct io_ring_ctx ctx, struct io_kiocb *req)
	358	{
	359	if (unlikely(io_req_cache_empty(ctx))) {
	360	if (!__io_alloc_req_refill(ctx))
	361	return false;
	362	}
	363	*req = io_extract_req(ctx);
	364	return true;
	365	}
	366
140102ae PB	367	static inline bool io_allowed_defer_tw_run(struct io_ring_ctx *ctx)
	368	{
	369	return likely(ctx->submitter_task == current);
	370	}
	371
76de6749 PB	372	static inline bool io_allowed_run_tw(struct io_ring_ctx *ctx)
76de6749 PB	373	{
6567506b PB	374	return likely(!(ctx->flags & IORING_SETUP_DEFER_TASKRUN) \|\|
6567506b PB	375	ctx->submitter_task == current);
76de6749 PB	376	}
76de6749 PB	377
833b5dff PB	378	static inline void io_req_queue_tw_complete(struct io_kiocb *req, s32 res)
	379	{
	380	io_req_set_res(req, res, 0);
	381	req->io_task_work.func = io_req_task_complete;
	382	io_req_task_work_add(req);
	383	}
	384
96c7d4f8 BL	385	/*
	386	* IORING_SETUP_SQE128 contexts allocate twice the normal SQE size for each
	387	* slot.
	388	*/
	389	static inline size_t uring_sqe_size(struct io_ring_ctx *ctx)
	390	{
	391	if (ctx->flags & IORING_SETUP_SQE128)
	392	return 2 * sizeof(struct io_uring_sqe);
	393	return sizeof(struct io_uring_sqe);
	394	}
de23077e	395	#endif