[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/io_uring_types.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 REQ_F_POLLED and REQ_F_APOLL_MULTISHOT
	 * are set 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);
int __io_run_local_work(struct io_ring_ctx *ctx, bool *locked);
int io_run_local_work(struct io_ring_ctx *ctx);
void io_req_complete_failed(struct io_kiocb *req, s32 res);
void __io_req_complete(struct io_kiocb *req, unsigned issue_flags);
void io_req_complete_post(struct io_kiocb *req);
void __io_req_complete_post(struct io_kiocb *req);
bool io_post_aux_cqe(struct io_ring_ctx *ctx, u64 user_data, s32 res, u32 cflags,
		     bool allow_overflow);
bool io_fill_cqe_aux(struct io_ring_ctx *ctx, 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;
}

bool io_is_uring_fops(struct file *file);
bool io_alloc_async_data(struct io_kiocb *req);
void io_req_task_work_add(struct io_kiocb *req);
void io_req_tw_post_queue(struct io_kiocb *req, s32 res, u32 cflags);
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_put_task(struct task_struct *task, int nr);
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_for_each_link(pos, head) \
	for (pos = (head); pos; pos = pos->link)

static inline void io_cq_lock(struct io_ring_ctx *ctx)
	__acquires(ctx->completion_lock)
{
	spin_lock(&ctx->completion_lock);
}

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)
{
	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 io_req_cqe_overflow(req);

	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 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);
}

/* requires smb_mb() prior, see wq_has_sleeper() */
static inline void __io_cqring_wake(struct io_ring_ctx *ctx)
{
	/*
	 * wake_up_all() may seem excessive, but io_wake_function() and
	 * io_should_wake() handle the termination of the loop and only
	 * wake as many waiters as we need to.
	 */
	if (waitqueue_active(&ctx->cq_wait))
		wake_up_all(&ctx->cq_wait);
}

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)
{
	if (task_work_pending(current)) {
		if (test_thread_flag(TIF_NOTIFY_SIGNAL))
			clear_notify_signal();
		__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 test_thread_flag(TIF_NOTIFY_SIGNAL) ||
		!wq_list_empty(&ctx->work_llist);
}

static inline int io_run_task_work_ctx(struct io_ring_ctx *ctx)
{
	int ret = 0;
	int ret2;

	if (ctx->flags & IORING_SETUP_DEFER_TASKRUN)
		ret = io_run_local_work(ctx);

	/* want to run this after in case more is added */
	ret2 = io_run_task_work();

	/* Try propagate error in favour of if tasks were run,
	 * but still make sure to run them if requested
	 */
	if (ret >= 0)
		ret += ret2;

	return ret;
}

static inline int io_run_local_work_locked(struct io_ring_ctx *ctx)
{
	bool locked;
	int ret;

	if (llist_empty(&ctx->work_llist))
		return 0;

	locked = true;
	ret = __io_run_local_work(ctx, &locked);
	/* shouldn't happen! */
	if (WARN_ON_ONCE(!locked))
		mutex_lock(&ctx->uring_lock);
	return ret;
}

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))
		__io_commit_cqring_flush(ctx);
}

/* must to be called somewhat shortly after putting a request */
static inline void io_put_task(struct task_struct *task, int nr)
{
	if (likely(task == current))
		task->io_uring->cached_refs += nr;
	else
		__io_put_task(task, nr);
}

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;
}

static inline struct io_kiocb *io_alloc_req(struct io_ring_ctx *ctx)
{
	struct io_wq_work_node *node;

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

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

#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>
ab1c84d8 PB	6	#include <linux/io_uring_types.h>
ab1c84d8 PB	7	#include "io-wq.h"
a6b21fbb	8	#include "slist.h"
ab1c84d8	9	#include "filetable.h"
de23077e	10
f3b44f92 JA	11	#ifndef CREATE_TRACE_POINTS
	12	#include <trace/events/io_uring.h>
	13	#endif
	14
97b388d7 JA	15	enum {
	16	IOU_OK = 0,
	17	IOU_ISSUE_SKIP_COMPLETE = -EIOCBQUEUED,
114eccdf DY	18
	19	/*
	20	* Intended only when both REQ_F_POLLED and REQ_F_APOLL_MULTISHOT
	21	* are set to indicate to the poll runner that multishot should be
	22	* removed and the result is set on req->cqe.res.
	23	*/
	24	IOU_STOP_MULTISHOT = -ECANCELED,
97b388d7 JA	25	};
97b388d7 JA	26
aa1df3a3	27	struct io_uring_cqe __io_get_cqe(struct io_ring_ctx ctx, bool overflow);
68494a65	28	bool io_req_cqe_overflow(struct io_kiocb *req);
c0e0d6ba	29	int io_run_task_work_sig(struct io_ring_ctx *ctx);
b3026767	30	int __io_run_local_work(struct io_ring_ctx ctx, bool locked);
c0e0d6ba	31	int io_run_local_work(struct io_ring_ctx *ctx);
9046c641 PB	32	void io_req_complete_failed(struct io_kiocb *req, s32 res);
	33	void __io_req_complete(struct io_kiocb *req, unsigned issue_flags);
	34	void io_req_complete_post(struct io_kiocb *req);
	35	void __io_req_complete_post(struct io_kiocb *req);
52120f0f DY	36	bool io_post_aux_cqe(struct io_ring_ctx *ctx, u64 user_data, s32 res, u32 cflags,
52120f0f DY	37	bool allow_overflow);
eb42cebb PB	38	bool io_fill_cqe_aux(struct io_ring_ctx *ctx, u64 user_data, s32 res, u32 cflags,
eb42cebb PB	39	bool allow_overflow);
9046c641 PB	40	void __io_commit_cqring_flush(struct io_ring_ctx *ctx);
	41
	42	struct page *io_pin_pages(unsigned long ubuf, unsigned long len, int npages);
	43
	44	struct file io_file_get_normal(struct io_kiocb req, int fd);
	45	struct file io_file_get_fixed(struct io_kiocb req, int fd,
	46	unsigned issue_flags);
	47
f6b543fd JA	48	static inline bool io_req_ffs_set(struct io_kiocb *req)
	49	{
	50	return req->flags & REQ_F_FIXED_FILE;
	51	}
	52
9046c641 PB	53	bool io_is_uring_fops(struct file *file);
	54	bool io_alloc_async_data(struct io_kiocb *req);
	55	void io_req_task_work_add(struct io_kiocb *req);
9046c641 PB	56	void io_req_tw_post_queue(struct io_kiocb *req, s32 res, u32 cflags);
	57	void io_req_task_queue(struct io_kiocb *req);
	58	void io_queue_iowq(struct io_kiocb req, bool dont_use);
	59	void io_req_task_complete(struct io_kiocb req, bool locked);
	60	void io_req_task_queue_fail(struct io_kiocb *req, int ret);
	61	void io_req_task_submit(struct io_kiocb req, bool locked);
	62	void tctx_task_work(struct callback_head *cb);
	63	__cold void io_uring_cancel_generic(bool cancel_all, struct io_sq_data *sqd);
	64	int io_uring_alloc_task_context(struct task_struct *task,
	65	struct io_ring_ctx *ctx);
	66
	67	int io_poll_issue(struct io_kiocb req, bool locked);
	68	int io_submit_sqes(struct io_ring_ctx *ctx, unsigned int nr);
	69	int io_do_iopoll(struct io_ring_ctx *ctx, bool force_nonspin);
	70	void io_free_batch_list(struct io_ring_ctx ctx, struct io_wq_work_node node);
	71	int io_req_prep_async(struct io_kiocb *req);
	72
	73	struct io_wq_work io_wq_free_work(struct io_wq_work work);
	74	void io_wq_submit_work(struct io_wq_work *work);
	75
	76	void io_free_req(struct io_kiocb *req);
	77	void io_queue_next(struct io_kiocb *req);
e70cb608	78	void __io_put_task(struct task_struct *task, int nr);
63809137	79	void io_task_refs_refill(struct io_uring_task *tctx);
bd1a3783	80	bool __io_alloc_req_refill(struct io_ring_ctx *ctx);
9046c641 PB	81
	82	bool io_match_task_safe(struct io_kiocb head, struct task_struct task,
	83	bool cancel_all);
	84
	85	#define io_for_each_link(pos, head) \
	86	for (pos = (head); pos; pos = pos->link)
f3b44f92	87
25399321 PB	88	static inline void io_cq_lock(struct io_ring_ctx *ctx)
	89	__acquires(ctx->completion_lock)
	90	{
	91	spin_lock(&ctx->completion_lock);
	92	}
	93
	94	void io_cq_unlock_post(struct io_ring_ctx *ctx);
	95
aa1df3a3 PB	96	static inline struct io_uring_cqe io_get_cqe_overflow(struct io_ring_ctx ctx,
aa1df3a3 PB	97	bool overflow)
f3b44f92 JA	98	{
	99	if (likely(ctx->cqe_cached < ctx->cqe_sentinel)) {
	100	struct io_uring_cqe *cqe = ctx->cqe_cached;
	101
f3b44f92 JA	102	ctx->cached_cq_tail++;
f3b44f92 JA	103	ctx->cqe_cached++;
b3659a65 PB	104	if (ctx->flags & IORING_SETUP_CQE32)
b3659a65 PB	105	ctx->cqe_cached++;
f3b44f92 JA	106	return cqe;
	107	}
	108
aa1df3a3 PB	109	return __io_get_cqe(ctx, overflow);
	110	}
	111
	112	static inline struct io_uring_cqe io_get_cqe(struct io_ring_ctx ctx)
	113	{
	114	return io_get_cqe_overflow(ctx, false);
f3b44f92 JA	115	}
	116
	117	static inline bool __io_fill_cqe_req(struct io_ring_ctx *ctx,
	118	struct io_kiocb *req)
	119	{
	120	struct io_uring_cqe *cqe;
	121
e8c328c3 PB	122	/*
	123	* If we can't get a cq entry, userspace overflowed the
	124	* submission (by quite a lot). Increment the overflow count in
	125	* the ring.
	126	*/
	127	cqe = io_get_cqe(ctx);
	128	if (unlikely(!cqe))
	129	return io_req_cqe_overflow(req);
e0486f3f DY	130
	131	trace_io_uring_complete(req->ctx, req, req->cqe.user_data,
	132	req->cqe.res, req->cqe.flags,
	133	(req->flags & REQ_F_CQE32_INIT) ? req->extra1 : 0,
	134	(req->flags & REQ_F_CQE32_INIT) ? req->extra2 : 0);
	135
e8c328c3 PB	136	memcpy(cqe, &req->cqe, sizeof(*cqe));
	137
	138	if (ctx->flags & IORING_SETUP_CQE32) {
f3b44f92 JA	139	u64 extra1 = 0, extra2 = 0;
	140
	141	if (req->flags & REQ_F_CQE32_INIT) {
	142	extra1 = req->extra1;
	143	extra2 = req->extra2;
	144	}
	145
e8c328c3 PB	146	WRITE_ONCE(cqe->big_cqe[0], extra1);
e8c328c3 PB	147	WRITE_ONCE(cqe->big_cqe[1], extra2);
f3b44f92	148	}
e8c328c3	149	return true;
f3b44f92 JA	150	}
f3b44f92 JA	151
531113bb JA	152	static inline void req_set_fail(struct io_kiocb *req)
	153	{
	154	req->flags \|= REQ_F_FAIL;
	155	if (req->flags & REQ_F_CQE_SKIP) {
	156	req->flags &= ~REQ_F_CQE_SKIP;
	157	req->flags \|= REQ_F_SKIP_LINK_CQES;
	158	}
	159	}
	160
de23077e JA	161	static inline void io_req_set_res(struct io_kiocb *req, s32 res, u32 cflags)
	162	{
	163	req->cqe.res = res;
	164	req->cqe.flags = cflags;
	165	}
	166
99f15d8d JA	167	static inline bool req_has_async_data(struct io_kiocb *req)
	168	{
	169	return req->flags & REQ_F_ASYNC_DATA;
	170	}
	171
531113bb JA	172	static inline void io_put_file(struct file *file)
	173	{
	174	if (file)
	175	fput(file);
	176	}
	177
cd40cae2 JA	178	static inline void io_ring_submit_unlock(struct io_ring_ctx *ctx,
	179	unsigned issue_flags)
	180	{
	181	lockdep_assert_held(&ctx->uring_lock);
	182	if (issue_flags & IO_URING_F_UNLOCKED)
	183	mutex_unlock(&ctx->uring_lock);
	184	}
	185
	186	static inline void io_ring_submit_lock(struct io_ring_ctx *ctx,
	187	unsigned issue_flags)
	188	{
	189	/*
	190	* "Normal" inline submissions always hold the uring_lock, since we
	191	* grab it from the system call. Same is true for the SQPOLL offload.
	192	* The only exception is when we've detached the request and issue it
	193	* from an async worker thread, grab the lock for that case.
	194	*/
	195	if (issue_flags & IO_URING_F_UNLOCKED)
	196	mutex_lock(&ctx->uring_lock);
	197	lockdep_assert_held(&ctx->uring_lock);
	198	}
	199
f9ead18c JA	200	static inline void io_commit_cqring(struct io_ring_ctx *ctx)
	201	{
	202	/* order cqe stores with ring update */
	203	smp_store_release(&ctx->rings->cq.tail, ctx->cached_cq_tail);
	204	}
	205
fc86f9d3 PB	206	/* requires smb_mb() prior, see wq_has_sleeper() */
fc86f9d3 PB	207	static inline void __io_cqring_wake(struct io_ring_ctx *ctx)
f3b44f92 JA	208	{
	209	/*
	210	* wake_up_all() may seem excessive, but io_wake_function() and
	211	* io_should_wake() handle the termination of the loop and only
	212	* wake as many waiters as we need to.
	213	*/
fc86f9d3	214	if (waitqueue_active(&ctx->cq_wait))
f3b44f92 JA	215	wake_up_all(&ctx->cq_wait);
	216	}
	217
fc86f9d3 PB	218	static inline void io_cqring_wake(struct io_ring_ctx *ctx)
	219	{
	220	smp_mb();
	221	__io_cqring_wake(ctx);
	222	}
	223
17437f31 JA	224	static inline bool io_sqring_full(struct io_ring_ctx *ctx)
	225	{
	226	struct io_rings *r = ctx->rings;
	227
	228	return READ_ONCE(r->sq.tail) - ctx->cached_sq_head == ctx->sq_entries;
	229	}
	230
	231	static inline unsigned int io_sqring_entries(struct io_ring_ctx *ctx)
	232	{
	233	struct io_rings *rings = ctx->rings;
	234
	235	/* make sure SQ entry isn't read before tail */
	236	return smp_load_acquire(&rings->sq.tail) - ctx->cached_sq_head;
	237	}
	238
c0e0d6ba	239	static inline int io_run_task_work(void)
17437f31	240	{
46a525e1 JA	241	if (task_work_pending(current)) {
	242	if (test_thread_flag(TIF_NOTIFY_SIGNAL))
	243	clear_notify_signal();
17437f31	244	__set_current_state(TASK_RUNNING);
46a525e1	245	task_work_run();
c0e0d6ba	246	return 1;
17437f31 JA	247	}
17437f31 JA	248
c0e0d6ba DY	249	return 0;
	250	}
	251
dac6a0ea JA	252	static inline bool io_task_work_pending(struct io_ring_ctx *ctx)
	253	{
	254	return test_thread_flag(TIF_NOTIFY_SIGNAL) \|\|
	255	!wq_list_empty(&ctx->work_llist);
	256	}
	257
c0e0d6ba DY	258	static inline int io_run_task_work_ctx(struct io_ring_ctx *ctx)
	259	{
	260	int ret = 0;
	261	int ret2;
	262
	263	if (ctx->flags & IORING_SETUP_DEFER_TASKRUN)
	264	ret = io_run_local_work(ctx);
	265
	266	/* want to run this after in case more is added */
	267	ret2 = io_run_task_work();
	268
	269	/* Try propagate error in favour of if tasks were run,
	270	* but still make sure to run them if requested
	271	*/
	272	if (ret >= 0)
	273	ret += ret2;
	274
	275	return ret;
17437f31 JA	276	}
17437f31 JA	277
44f87745 PB	278	static inline int io_run_local_work_locked(struct io_ring_ctx *ctx)
44f87745 PB	279	{
b3026767 DY	280	bool locked;
	281	int ret;
	282
44f87745 PB	283	if (llist_empty(&ctx->work_llist))
44f87745 PB	284	return 0;
b3026767 DY	285
	286	locked = true;
	287	ret = __io_run_local_work(ctx, &locked);
	288	/* shouldn't happen! */
	289	if (WARN_ON_ONCE(!locked))
	290	mutex_lock(&ctx->uring_lock);
	291	return ret;
44f87745 PB	292	}
44f87745 PB	293
aa1e90f6 PB	294	static inline void io_tw_lock(struct io_ring_ctx ctx, bool locked)
	295	{
	296	if (!*locked) {
	297	mutex_lock(&ctx->uring_lock);
	298	*locked = true;
	299	}
	300	}
	301
9da070b1 PB	302	/*
	303	* Don't complete immediately but use deferred completion infrastructure.
	304	* Protected by ->uring_lock and can only be used either with
	305	* IO_URING_F_COMPLETE_DEFER or inside a tw handler holding the mutex.
	306	*/
	307	static inline void io_req_complete_defer(struct io_kiocb *req)
	308	__must_hold(&req->ctx->uring_lock)
aa1e90f6 PB	309	{
	310	struct io_submit_state *state = &req->ctx->submit_state;
	311
9da070b1 PB	312	lockdep_assert_held(&req->ctx->uring_lock);
9da070b1 PB	313
aa1e90f6 PB	314	wq_list_add_tail(&req->comp_list, &state->compl_reqs);
	315	}
	316
46929b08 PB	317	static inline void io_commit_cqring_flush(struct io_ring_ctx *ctx)
	318	{
	319	if (unlikely(ctx->off_timeout_used \|\| ctx->drain_active \|\| ctx->has_evfd))
	320	__io_commit_cqring_flush(ctx);
	321	}
	322
e70cb608 PB	323	/* must to be called somewhat shortly after putting a request */
	324	static inline void io_put_task(struct task_struct *task, int nr)
	325	{
	326	if (likely(task == current))
	327	task->io_uring->cached_refs += nr;
	328	else
	329	__io_put_task(task, nr);
	330	}
	331
63809137 PB	332	static inline void io_get_task_refs(int nr)
	333	{
	334	struct io_uring_task *tctx = current->io_uring;
	335
	336	tctx->cached_refs -= nr;
	337	if (unlikely(tctx->cached_refs < 0))
	338	io_task_refs_refill(tctx);
	339	}
	340
bd1a3783 PB	341	static inline bool io_req_cache_empty(struct io_ring_ctx *ctx)
	342	{
	343	return !ctx->submit_state.free_list.next;
	344	}
	345
	346	static inline bool io_alloc_req_refill(struct io_ring_ctx *ctx)
	347	{
	348	if (unlikely(io_req_cache_empty(ctx)))
	349	return __io_alloc_req_refill(ctx);
	350	return true;
	351	}
	352
	353	static inline struct io_kiocb io_alloc_req(struct io_ring_ctx ctx)
	354	{
	355	struct io_wq_work_node *node;
	356
	357	node = wq_stack_extract(&ctx->submit_state.free_list);
	358	return container_of(node, struct io_kiocb, comp_list);
	359	}
	360
76de6749 PB	361	static inline bool io_allowed_run_tw(struct io_ring_ctx *ctx)
76de6749 PB	362	{
6567506b PB	363	return likely(!(ctx->flags & IORING_SETUP_DEFER_TASKRUN) \|\|
6567506b PB	364	ctx->submitter_task == current);
76de6749 PB	365	}
76de6749 PB	366
de23077e	367	#endif