[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);

void *io_mem_alloc(size_t size);
void io_mem_free(void *ptr);

#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
edecf168 JA	89	void *io_mem_alloc(size_t size);
	90	void io_mem_free(void *ptr);
	91
1658633c JA	92	#if defined(CONFIG_PROVE_LOCKING)
	93	static inline void io_lockdep_assert_cq_locked(struct io_ring_ctx *ctx)
	94	{
	95	lockdep_assert(in_task());
	96
	97	if (ctx->flags & IORING_SETUP_IOPOLL) {
	98	lockdep_assert_held(&ctx->uring_lock);
	99	} else if (!ctx->task_complete) {
	100	lockdep_assert_held(&ctx->completion_lock);
	101	} else if (ctx->submitter_task) {
	102	/*
	103	* ->submitter_task may be NULL and we can still post a CQE,
	104	* if the ring has been setup with IORING_SETUP_R_DISABLED.
	105	* Not from an SQE, as those cannot be submitted, but via
	106	* updating tagged resources.
	107	*/
	108	if (ctx->submitter_task->flags & PF_EXITING)
	109	lockdep_assert(current_work());
	110	else
	111	lockdep_assert(current == ctx->submitter_task);
	112	}
	113	}
	114	#else
	115	static inline void io_lockdep_assert_cq_locked(struct io_ring_ctx *ctx)
	116	{
	117	}
	118	#endif
f26cc959	119
e52d2e58 PB	120	static inline void io_req_task_work_add(struct io_kiocb *req)
e52d2e58 PB	121	{
8501fe70	122	__io_req_task_work_add(req, 0);
e52d2e58 PB	123	}
e52d2e58 PB	124
9046c641 PB	125	#define io_for_each_link(pos, head) \
9046c641 PB	126	for (pos = (head); pos; pos = pos->link)
f3b44f92	127
59fbc409 PB	128	static inline bool io_get_cqe_overflow(struct io_ring_ctx *ctx,
	129	struct io_uring_cqe **ret,
	130	bool overflow)
f3b44f92	131	{
20d6b633	132	io_lockdep_assert_cq_locked(ctx);
f3b44f92	133
20d6b633 PB	134	if (unlikely(ctx->cqe_cached >= ctx->cqe_sentinel)) {
20d6b633 PB	135	if (unlikely(!io_cqe_cache_refill(ctx, overflow)))
59fbc409	136	return false;
f3b44f92	137	}
59fbc409	138	*ret = ctx->cqe_cached;
20d6b633 PB	139	ctx->cached_cq_tail++;
	140	ctx->cqe_cached++;
	141	if (ctx->flags & IORING_SETUP_CQE32)
	142	ctx->cqe_cached++;
59fbc409	143	return true;
aa1df3a3 PB	144	}
aa1df3a3 PB	145
59fbc409	146	static inline bool io_get_cqe(struct io_ring_ctx ctx, struct io_uring_cqe *ret)
aa1df3a3	147	{
59fbc409	148	return io_get_cqe_overflow(ctx, ret, false);
f3b44f92 JA	149	}
f3b44f92 JA	150
093a650b PB	151	static __always_inline bool io_fill_cqe_req(struct io_ring_ctx *ctx,
093a650b PB	152	struct io_kiocb *req)
f3b44f92 JA	153	{
	154	struct io_uring_cqe *cqe;
	155
e8c328c3 PB	156	/*
	157	* If we can't get a cq entry, userspace overflowed the
	158	* submission (by quite a lot). Increment the overflow count in
	159	* the ring.
	160	*/
59fbc409	161	if (unlikely(!io_get_cqe(ctx, &cqe)))
f66f7342	162	return false;
e0486f3f	163
a0727c73 PB	164	if (trace_io_uring_complete_enabled())
	165	trace_io_uring_complete(req->ctx, req, req->cqe.user_data,
	166	req->cqe.res, req->cqe.flags,
b24c5d75	167	req->big_cqe.extra1, req->big_cqe.extra2);
e0486f3f	168
e8c328c3	169	memcpy(cqe, &req->cqe, sizeof(*cqe));
e8c328c3	170	if (ctx->flags & IORING_SETUP_CQE32) {
b24c5d75 PB	171	memcpy(cqe->big_cqe, &req->big_cqe, sizeof(*cqe));
b24c5d75 PB	172	memset(&req->big_cqe, 0, sizeof(req->big_cqe));
f3b44f92	173	}
e8c328c3	174	return true;
f3b44f92 JA	175	}
f3b44f92 JA	176
531113bb JA	177	static inline void req_set_fail(struct io_kiocb *req)
	178	{
	179	req->flags \|= REQ_F_FAIL;
	180	if (req->flags & REQ_F_CQE_SKIP) {
	181	req->flags &= ~REQ_F_CQE_SKIP;
	182	req->flags \|= REQ_F_SKIP_LINK_CQES;
	183	}
	184	}
	185
de23077e JA	186	static inline void io_req_set_res(struct io_kiocb *req, s32 res, u32 cflags)
	187	{
	188	req->cqe.res = res;
	189	req->cqe.flags = cflags;
	190	}
	191
99f15d8d JA	192	static inline bool req_has_async_data(struct io_kiocb *req)
	193	{
	194	return req->flags & REQ_F_ASYNC_DATA;
	195	}
	196
17bc2837	197	static inline void io_put_file(struct io_kiocb *req)
531113bb	198	{
17bc2837 JA	199	if (!(req->flags & REQ_F_FIXED_FILE) && req->file)
17bc2837 JA	200	fput(req->file);
531113bb JA	201	}
531113bb JA	202
cd40cae2 JA	203	static inline void io_ring_submit_unlock(struct io_ring_ctx *ctx,
	204	unsigned issue_flags)
	205	{
	206	lockdep_assert_held(&ctx->uring_lock);
	207	if (issue_flags & IO_URING_F_UNLOCKED)
	208	mutex_unlock(&ctx->uring_lock);
	209	}
	210
	211	static inline void io_ring_submit_lock(struct io_ring_ctx *ctx,
	212	unsigned issue_flags)
	213	{
	214	/*
	215	* "Normal" inline submissions always hold the uring_lock, since we
	216	* grab it from the system call. Same is true for the SQPOLL offload.
	217	* The only exception is when we've detached the request and issue it
	218	* from an async worker thread, grab the lock for that case.
	219	*/
	220	if (issue_flags & IO_URING_F_UNLOCKED)
	221	mutex_lock(&ctx->uring_lock);
	222	lockdep_assert_held(&ctx->uring_lock);
	223	}
	224
f9ead18c JA	225	static inline void io_commit_cqring(struct io_ring_ctx *ctx)
	226	{
	227	/* order cqe stores with ring update */
	228	smp_store_release(&ctx->rings->cq.tail, ctx->cached_cq_tail);
	229	}
	230
7b235dd8 PB	231	static inline void io_poll_wq_wake(struct io_ring_ctx *ctx)
7b235dd8 PB	232	{
bca39f39	233	if (wq_has_sleeper(&ctx->poll_wq))
7b235dd8 PB	234	__wake_up(&ctx->poll_wq, TASK_NORMAL, 0,
	235	poll_to_key(EPOLL_URING_WAKE \| EPOLLIN));
	236	}
	237
6e7248ad	238	static inline void io_cqring_wake(struct io_ring_ctx *ctx)
f3b44f92 JA	239	{
f3b44f92 JA	240	/*
44648532 JA	241	* Trigger waitqueue handler on all waiters on our waitqueue. This
	242	* won't necessarily wake up all the tasks, io_should_wake() will make
	243	* that decision.
	244	*
	245	* Pass in EPOLLIN\|EPOLL_URING_WAKE as the poll wakeup key. The latter
	246	* set in the mask so that if we recurse back into our own poll
	247	* waitqueue handlers, we know we have a dependency between eventfd or
	248	* epoll and should terminate multishot poll at that point.
f3b44f92	249	*/
6e7248ad	250	if (wq_has_sleeper(&ctx->cq_wait))
44648532 JA	251	__wake_up(&ctx->cq_wait, TASK_NORMAL, 0,
44648532 JA	252	poll_to_key(EPOLL_URING_WAKE \| EPOLLIN));
f3b44f92 JA	253	}
f3b44f92 JA	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;
e3ef728f	265	unsigned int entries;
17437f31 JA	266
17437f31 JA	267	/* make sure SQ entry isn't read before tail */
e3ef728f JA	268	entries = smp_load_acquire(&rings->sq.tail) - ctx->cached_sq_head;
e3ef728f JA	269	return min(entries, ctx->sq_entries);
17437f31 JA	270	}
17437f31 JA	271
c0e0d6ba	272	static inline int io_run_task_work(void)
17437f31	273	{
7cfe7a09 JA	274	/*
	275	* Always check-and-clear the task_work notification signal. With how
	276	* signaling works for task_work, we can find it set with nothing to
	277	* run. We need to clear it for that case, like get_signal() does.
	278	*/
	279	if (test_thread_flag(TIF_NOTIFY_SIGNAL))
	280	clear_notify_signal();
b5d3ae20 JA	281	/*
	282	* PF_IO_WORKER never returns to userspace, so check here if we have
	283	* notify work that needs processing.
	284	*/
	285	if (current->flags & PF_IO_WORKER &&
2f2bb1ff JA	286	test_thread_flag(TIF_NOTIFY_RESUME)) {
2f2bb1ff JA	287	__set_current_state(TASK_RUNNING);
b5d3ae20	288	resume_user_mode_work(NULL);
2f2bb1ff	289	}
46a525e1	290	if (task_work_pending(current)) {
17437f31	291	__set_current_state(TASK_RUNNING);
46a525e1	292	task_work_run();
c0e0d6ba	293	return 1;
17437f31 JA	294	}
17437f31 JA	295
c0e0d6ba DY	296	return 0;
	297	}
	298
dac6a0ea JA	299	static inline bool io_task_work_pending(struct io_ring_ctx *ctx)
dac6a0ea JA	300	{
6434ec01	301	return task_work_pending(current) \|\| !wq_list_empty(&ctx->work_llist);
dac6a0ea JA	302	}
dac6a0ea JA	303
a282967c	304	static inline void io_tw_lock(struct io_ring_ctx ctx, struct io_tw_state ts)
aa1e90f6	305	{
a282967c	306	if (!ts->locked) {
aa1e90f6	307	mutex_lock(&ctx->uring_lock);
a282967c	308	ts->locked = true;
aa1e90f6 PB	309	}
	310	}
	311
9da070b1 PB	312	/*
	313	* Don't complete immediately but use deferred completion infrastructure.
	314	* Protected by ->uring_lock and can only be used either with
	315	* IO_URING_F_COMPLETE_DEFER or inside a tw handler holding the mutex.
	316	*/
	317	static inline void io_req_complete_defer(struct io_kiocb *req)
	318	__must_hold(&req->ctx->uring_lock)
aa1e90f6 PB	319	{
	320	struct io_submit_state *state = &req->ctx->submit_state;
	321
9da070b1 PB	322	lockdep_assert_held(&req->ctx->uring_lock);
9da070b1 PB	323
aa1e90f6 PB	324	wq_list_add_tail(&req->comp_list, &state->compl_reqs);
	325	}
	326
46929b08 PB	327	static inline void io_commit_cqring_flush(struct io_ring_ctx *ctx)
46929b08 PB	328	{
bca39f39 PB	329	if (unlikely(ctx->off_timeout_used \|\| ctx->drain_active \|\|
bca39f39 PB	330	ctx->has_evfd \|\| ctx->poll_activated))
46929b08 PB	331	__io_commit_cqring_flush(ctx);
	332	}
	333
63809137 PB	334	static inline void io_get_task_refs(int nr)
	335	{
	336	struct io_uring_task *tctx = current->io_uring;
	337
	338	tctx->cached_refs -= nr;
	339	if (unlikely(tctx->cached_refs < 0))
	340	io_task_refs_refill(tctx);
	341	}
	342
bd1a3783 PB	343	static inline bool io_req_cache_empty(struct io_ring_ctx *ctx)
	344	{
	345	return !ctx->submit_state.free_list.next;
	346	}
	347
c1755c25	348	extern struct kmem_cache *req_cachep;
b3a4dbc8	349	extern struct kmem_cache *io_buf_cachep;
c1755c25	350
c8576f3e	351	static inline struct io_kiocb io_extract_req(struct io_ring_ctx ctx)
bd1a3783	352	{
c1755c25	353	struct io_kiocb *req;
bd1a3783	354
c1755c25	355	req = container_of(ctx->submit_state.free_list.next, struct io_kiocb, comp_list);
c1755c25 BL	356	wq_stack_extract(&ctx->submit_state.free_list);
c1755c25 BL	357	return req;
bd1a3783 PB	358	}
bd1a3783 PB	359
c8576f3e PB	360	static inline bool io_alloc_req(struct io_ring_ctx ctx, struct io_kiocb *req)
	361	{
	362	if (unlikely(io_req_cache_empty(ctx))) {
	363	if (!__io_alloc_req_refill(ctx))
	364	return false;
	365	}
	366	*req = io_extract_req(ctx);
	367	return true;
	368	}
	369
140102ae PB	370	static inline bool io_allowed_defer_tw_run(struct io_ring_ctx *ctx)
	371	{
	372	return likely(ctx->submitter_task == current);
	373	}
	374
76de6749 PB	375	static inline bool io_allowed_run_tw(struct io_ring_ctx *ctx)
76de6749 PB	376	{
6567506b PB	377	return likely(!(ctx->flags & IORING_SETUP_DEFER_TASKRUN) \|\|
6567506b PB	378	ctx->submitter_task == current);
76de6749 PB	379	}
76de6749 PB	380
833b5dff PB	381	static inline void io_req_queue_tw_complete(struct io_kiocb *req, s32 res)
	382	{
	383	io_req_set_res(req, res, 0);
	384	req->io_task_work.func = io_req_task_complete;
	385	io_req_task_work_add(req);
	386	}
	387
96c7d4f8 BL	388	/*
	389	* IORING_SETUP_SQE128 contexts allocate twice the normal SQE size for each
	390	* slot.
	391	*/
	392	static inline size_t uring_sqe_size(struct io_ring_ctx *ctx)
	393	{
	394	if (ctx->flags & IORING_SETUP_SQE128)
	395	return 2 * sizeof(struct io_uring_sqe);
	396	return sizeof(struct io_uring_sqe);
	397	}
de23077e	398	#endif