[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 io_req_cqe_overflow(struct io_kiocb *req);
int io_run_task_work_sig(void);
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);
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);

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(struct io_ring_ctx *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);
}

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);
}

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 (wq_has_sleeper(&ctx->cq_wait))
		wake_up_all(&ctx->cq_wait);
}

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 bool io_run_task_work(void)
{
	if (test_thread_flag(TIF_NOTIFY_SIGNAL)) {
		__set_current_state(TASK_RUNNING);
		clear_notify_signal();
		if (task_work_pending(current))
			task_work_run();
		return true;
	}

	return false;
}

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);
}

#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
faf88dde	27	struct io_uring_cqe __io_get_cqe(struct io_ring_ctx ctx);
68494a65	28	bool io_req_cqe_overflow(struct io_kiocb *req);
9046c641 PB	29	int io_run_task_work_sig(void);
	30	void io_req_complete_failed(struct io_kiocb *req, s32 res);
	31	void __io_req_complete(struct io_kiocb *req, unsigned issue_flags);
	32	void io_req_complete_post(struct io_kiocb *req);
	33	void __io_req_complete_post(struct io_kiocb *req);
52120f0f DY	34	bool io_post_aux_cqe(struct io_ring_ctx *ctx, u64 user_data, s32 res, u32 cflags,
52120f0f DY	35	bool allow_overflow);
9046c641 PB	36	void __io_commit_cqring_flush(struct io_ring_ctx *ctx);
	37
	38	struct page *io_pin_pages(unsigned long ubuf, unsigned long len, int npages);
	39
	40	struct file io_file_get_normal(struct io_kiocb req, int fd);
	41	struct file io_file_get_fixed(struct io_kiocb req, int fd,
	42	unsigned issue_flags);
	43
f6b543fd JA	44	static inline bool io_req_ffs_set(struct io_kiocb *req)
	45	{
	46	return req->flags & REQ_F_FIXED_FILE;
	47	}
	48
9046c641 PB	49	bool io_is_uring_fops(struct file *file);
	50	bool io_alloc_async_data(struct io_kiocb *req);
	51	void io_req_task_work_add(struct io_kiocb *req);
9046c641 PB	52	void io_req_tw_post_queue(struct io_kiocb *req, s32 res, u32 cflags);
	53	void io_req_task_queue(struct io_kiocb *req);
	54	void io_queue_iowq(struct io_kiocb req, bool dont_use);
	55	void io_req_task_complete(struct io_kiocb req, bool locked);
	56	void io_req_task_queue_fail(struct io_kiocb *req, int ret);
	57	void io_req_task_submit(struct io_kiocb req, bool locked);
	58	void tctx_task_work(struct callback_head *cb);
	59	__cold void io_uring_cancel_generic(bool cancel_all, struct io_sq_data *sqd);
	60	int io_uring_alloc_task_context(struct task_struct *task,
	61	struct io_ring_ctx *ctx);
	62
	63	int io_poll_issue(struct io_kiocb req, bool locked);
	64	int io_submit_sqes(struct io_ring_ctx *ctx, unsigned int nr);
	65	int io_do_iopoll(struct io_ring_ctx *ctx, bool force_nonspin);
	66	void io_free_batch_list(struct io_ring_ctx ctx, struct io_wq_work_node node);
	67	int io_req_prep_async(struct io_kiocb *req);
	68
	69	struct io_wq_work io_wq_free_work(struct io_wq_work work);
	70	void io_wq_submit_work(struct io_wq_work *work);
	71
	72	void io_free_req(struct io_kiocb *req);
	73	void io_queue_next(struct io_kiocb *req);
e70cb608	74	void __io_put_task(struct task_struct *task, int nr);
9046c641 PB	75
	76	bool io_match_task_safe(struct io_kiocb head, struct task_struct task,
	77	bool cancel_all);
	78
	79	#define io_for_each_link(pos, head) \
	80	for (pos = (head); pos; pos = pos->link)
f3b44f92	81
25399321 PB	82	static inline void io_cq_lock(struct io_ring_ctx *ctx)
	83	__acquires(ctx->completion_lock)
	84	{
	85	spin_lock(&ctx->completion_lock);
	86	}
	87
	88	void io_cq_unlock_post(struct io_ring_ctx *ctx);
	89
f3b44f92 JA	90	static inline struct io_uring_cqe io_get_cqe(struct io_ring_ctx ctx)
	91	{
	92	if (likely(ctx->cqe_cached < ctx->cqe_sentinel)) {
	93	struct io_uring_cqe *cqe = ctx->cqe_cached;
	94
f3b44f92 JA	95	ctx->cached_cq_tail++;
f3b44f92 JA	96	ctx->cqe_cached++;
b3659a65 PB	97	if (ctx->flags & IORING_SETUP_CQE32)
b3659a65 PB	98	ctx->cqe_cached++;
f3b44f92 JA	99	return cqe;
	100	}
	101
	102	return __io_get_cqe(ctx);
	103	}
	104
	105	static inline bool __io_fill_cqe_req(struct io_ring_ctx *ctx,
	106	struct io_kiocb *req)
	107	{
	108	struct io_uring_cqe *cqe;
	109
e8c328c3 PB	110	/*
	111	* If we can't get a cq entry, userspace overflowed the
	112	* submission (by quite a lot). Increment the overflow count in
	113	* the ring.
	114	*/
	115	cqe = io_get_cqe(ctx);
	116	if (unlikely(!cqe))
	117	return io_req_cqe_overflow(req);
e0486f3f DY	118
	119	trace_io_uring_complete(req->ctx, req, req->cqe.user_data,
	120	req->cqe.res, req->cqe.flags,
	121	(req->flags & REQ_F_CQE32_INIT) ? req->extra1 : 0,
	122	(req->flags & REQ_F_CQE32_INIT) ? req->extra2 : 0);
	123
e8c328c3 PB	124	memcpy(cqe, &req->cqe, sizeof(*cqe));
	125
	126	if (ctx->flags & IORING_SETUP_CQE32) {
f3b44f92 JA	127	u64 extra1 = 0, extra2 = 0;
	128
	129	if (req->flags & REQ_F_CQE32_INIT) {
	130	extra1 = req->extra1;
	131	extra2 = req->extra2;
	132	}
	133
e8c328c3 PB	134	WRITE_ONCE(cqe->big_cqe[0], extra1);
e8c328c3 PB	135	WRITE_ONCE(cqe->big_cqe[1], extra2);
f3b44f92	136	}
e8c328c3	137	return true;
f3b44f92 JA	138	}
f3b44f92 JA	139
531113bb JA	140	static inline void req_set_fail(struct io_kiocb *req)
	141	{
	142	req->flags \|= REQ_F_FAIL;
	143	if (req->flags & REQ_F_CQE_SKIP) {
	144	req->flags &= ~REQ_F_CQE_SKIP;
	145	req->flags \|= REQ_F_SKIP_LINK_CQES;
	146	}
	147	}
	148
de23077e JA	149	static inline void io_req_set_res(struct io_kiocb *req, s32 res, u32 cflags)
	150	{
	151	req->cqe.res = res;
	152	req->cqe.flags = cflags;
	153	}
	154
99f15d8d JA	155	static inline bool req_has_async_data(struct io_kiocb *req)
	156	{
	157	return req->flags & REQ_F_ASYNC_DATA;
	158	}
	159
531113bb JA	160	static inline void io_put_file(struct file *file)
	161	{
	162	if (file)
	163	fput(file);
	164	}
	165
cd40cae2 JA	166	static inline void io_ring_submit_unlock(struct io_ring_ctx *ctx,
	167	unsigned issue_flags)
	168	{
	169	lockdep_assert_held(&ctx->uring_lock);
	170	if (issue_flags & IO_URING_F_UNLOCKED)
	171	mutex_unlock(&ctx->uring_lock);
	172	}
	173
	174	static inline void io_ring_submit_lock(struct io_ring_ctx *ctx,
	175	unsigned issue_flags)
	176	{
	177	/*
	178	* "Normal" inline submissions always hold the uring_lock, since we
	179	* grab it from the system call. Same is true for the SQPOLL offload.
	180	* The only exception is when we've detached the request and issue it
	181	* from an async worker thread, grab the lock for that case.
	182	*/
	183	if (issue_flags & IO_URING_F_UNLOCKED)
	184	mutex_lock(&ctx->uring_lock);
	185	lockdep_assert_held(&ctx->uring_lock);
	186	}
	187
f9ead18c JA	188	static inline void io_commit_cqring(struct io_ring_ctx *ctx)
	189	{
	190	/* order cqe stores with ring update */
	191	smp_store_release(&ctx->rings->cq.tail, ctx->cached_cq_tail);
	192	}
	193
f3b44f92 JA	194	static inline void io_cqring_wake(struct io_ring_ctx *ctx)
	195	{
	196	/*
	197	* wake_up_all() may seem excessive, but io_wake_function() and
	198	* io_should_wake() handle the termination of the loop and only
	199	* wake as many waiters as we need to.
	200	*/
	201	if (wq_has_sleeper(&ctx->cq_wait))
	202	wake_up_all(&ctx->cq_wait);
	203	}
	204
17437f31 JA	205	static inline bool io_sqring_full(struct io_ring_ctx *ctx)
	206	{
	207	struct io_rings *r = ctx->rings;
	208
	209	return READ_ONCE(r->sq.tail) - ctx->cached_sq_head == ctx->sq_entries;
	210	}
	211
	212	static inline unsigned int io_sqring_entries(struct io_ring_ctx *ctx)
	213	{
	214	struct io_rings *rings = ctx->rings;
	215
	216	/* make sure SQ entry isn't read before tail */
	217	return smp_load_acquire(&rings->sq.tail) - ctx->cached_sq_head;
	218	}
	219
	220	static inline bool io_run_task_work(void)
	221	{
625d38b3	222	if (test_thread_flag(TIF_NOTIFY_SIGNAL)) {
17437f31 JA	223	__set_current_state(TASK_RUNNING);
	224	clear_notify_signal();
	225	if (task_work_pending(current))
	226	task_work_run();
	227	return true;
	228	}
	229
	230	return false;
	231	}
	232
aa1e90f6 PB	233	static inline void io_tw_lock(struct io_ring_ctx ctx, bool locked)
	234	{
	235	if (!*locked) {
	236	mutex_lock(&ctx->uring_lock);
	237	*locked = true;
	238	}
	239	}
	240
9da070b1 PB	241	/*
	242	* Don't complete immediately but use deferred completion infrastructure.
	243	* Protected by ->uring_lock and can only be used either with
	244	* IO_URING_F_COMPLETE_DEFER or inside a tw handler holding the mutex.
	245	*/
	246	static inline void io_req_complete_defer(struct io_kiocb *req)
	247	__must_hold(&req->ctx->uring_lock)
aa1e90f6 PB	248	{
	249	struct io_submit_state *state = &req->ctx->submit_state;
	250
9da070b1 PB	251	lockdep_assert_held(&req->ctx->uring_lock);
9da070b1 PB	252
aa1e90f6 PB	253	wq_list_add_tail(&req->comp_list, &state->compl_reqs);
	254	}
	255
46929b08 PB	256	static inline void io_commit_cqring_flush(struct io_ring_ctx *ctx)
	257	{
	258	if (unlikely(ctx->off_timeout_used \|\| ctx->drain_active \|\| ctx->has_evfd))
	259	__io_commit_cqring_flush(ctx);
	260	}
	261
e70cb608 PB	262	/* must to be called somewhat shortly after putting a request */
	263	static inline void io_put_task(struct task_struct *task, int nr)
	264	{
	265	if (likely(task == current))
	266	task->io_uring->cached_refs += nr;
	267	else
	268	__io_put_task(task, nr);
	269	}
	270
de23077e	271	#endif