[linux-block.git] / io_uring / io_uring.h

#ifndef IOU_CORE_H
#define IOU_CORE_H

#include <linux/errno.h>
#include <linux/lockdep.h>
#include "io_uring_types.h"

#ifndef CREATE_TRACE_POINTS
#include <trace/events/io_uring.h>
#endif

enum {
	IOU_OK			= 0,
	IOU_ISSUE_SKIP_COMPLETE	= -EIOCBQUEUED,
};

struct io_uring_cqe *__io_get_cqe(struct io_ring_ctx *ctx);
bool io_req_cqe_overflow(struct io_kiocb *req);

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;

	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);
	/*
	 * 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);
	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) || task_work_pending(current)) {
		__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;
	}
}

static inline void io_req_add_compl_list(struct io_kiocb *req)
{
	struct io_submit_state *state = &req->ctx->submit_state;

	if (!(req->flags & REQ_F_CQE_SKIP))
		state->flush_cqes = true;
	wq_list_add_tail(&req->comp_list, &state->compl_reqs);
}

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);
void io_cqring_ev_posted(struct io_ring_ctx *ctx);
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);

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_task_prio_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);

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)

#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>
de23077e JA	6	#include "io_uring_types.h"
de23077e JA	7
f3b44f92 JA	8	#ifndef CREATE_TRACE_POINTS
	9	#include <trace/events/io_uring.h>
	10	#endif
	11
97b388d7 JA	12	enum {
	13	IOU_OK = 0,
	14	IOU_ISSUE_SKIP_COMPLETE = -EIOCBQUEUED,
	15	};
	16
faf88dde	17	struct io_uring_cqe __io_get_cqe(struct io_ring_ctx ctx);
68494a65	18	bool io_req_cqe_overflow(struct io_kiocb *req);
f3b44f92	19
f3b44f92 JA	20	static inline struct io_uring_cqe io_get_cqe(struct io_ring_ctx ctx)
	21	{
	22	if (likely(ctx->cqe_cached < ctx->cqe_sentinel)) {
	23	struct io_uring_cqe *cqe = ctx->cqe_cached;
	24
f3b44f92 JA	25	ctx->cached_cq_tail++;
f3b44f92 JA	26	ctx->cqe_cached++;
b3659a65 PB	27	if (ctx->flags & IORING_SETUP_CQE32)
b3659a65 PB	28	ctx->cqe_cached++;
f3b44f92 JA	29	return cqe;
	30	}
	31
	32	return __io_get_cqe(ctx);
	33	}
	34
	35	static inline bool __io_fill_cqe_req(struct io_ring_ctx *ctx,
	36	struct io_kiocb *req)
	37	{
	38	struct io_uring_cqe *cqe;
	39
ae5735c6 PB	40	trace_io_uring_complete(req->ctx, req, req->cqe.user_data,
	41	req->cqe.res, req->cqe.flags,
	42	(req->flags & REQ_F_CQE32_INIT) ? req->extra1 : 0,
	43	(req->flags & REQ_F_CQE32_INIT) ? req->extra2 : 0);
e8c328c3 PB	44	/*
	45	* If we can't get a cq entry, userspace overflowed the
	46	* submission (by quite a lot). Increment the overflow count in
	47	* the ring.
	48	*/
	49	cqe = io_get_cqe(ctx);
	50	if (unlikely(!cqe))
	51	return io_req_cqe_overflow(req);
	52	memcpy(cqe, &req->cqe, sizeof(*cqe));
	53
	54	if (ctx->flags & IORING_SETUP_CQE32) {
f3b44f92 JA	55	u64 extra1 = 0, extra2 = 0;
	56
	57	if (req->flags & REQ_F_CQE32_INIT) {
	58	extra1 = req->extra1;
	59	extra2 = req->extra2;
	60	}
	61
e8c328c3 PB	62	WRITE_ONCE(cqe->big_cqe[0], extra1);
e8c328c3 PB	63	WRITE_ONCE(cqe->big_cqe[1], extra2);
f3b44f92	64	}
e8c328c3	65	return true;
f3b44f92 JA	66	}
f3b44f92 JA	67
531113bb JA	68	static inline void req_set_fail(struct io_kiocb *req)
	69	{
	70	req->flags \|= REQ_F_FAIL;
	71	if (req->flags & REQ_F_CQE_SKIP) {
	72	req->flags &= ~REQ_F_CQE_SKIP;
	73	req->flags \|= REQ_F_SKIP_LINK_CQES;
	74	}
	75	}
	76
de23077e JA	77	static inline void io_req_set_res(struct io_kiocb *req, s32 res, u32 cflags)
	78	{
	79	req->cqe.res = res;
	80	req->cqe.flags = cflags;
	81	}
	82
99f15d8d JA	83	static inline bool req_has_async_data(struct io_kiocb *req)
	84	{
	85	return req->flags & REQ_F_ASYNC_DATA;
	86	}
	87
531113bb JA	88	static inline void io_put_file(struct file *file)
	89	{
	90	if (file)
	91	fput(file);
	92	}
	93
cd40cae2 JA	94	static inline void io_ring_submit_unlock(struct io_ring_ctx *ctx,
	95	unsigned issue_flags)
	96	{
	97	lockdep_assert_held(&ctx->uring_lock);
	98	if (issue_flags & IO_URING_F_UNLOCKED)
	99	mutex_unlock(&ctx->uring_lock);
	100	}
	101
	102	static inline void io_ring_submit_lock(struct io_ring_ctx *ctx,
	103	unsigned issue_flags)
	104	{
	105	/*
	106	* "Normal" inline submissions always hold the uring_lock, since we
	107	* grab it from the system call. Same is true for the SQPOLL offload.
	108	* The only exception is when we've detached the request and issue it
	109	* from an async worker thread, grab the lock for that case.
	110	*/
	111	if (issue_flags & IO_URING_F_UNLOCKED)
	112	mutex_lock(&ctx->uring_lock);
	113	lockdep_assert_held(&ctx->uring_lock);
	114	}
	115
f9ead18c JA	116	static inline void io_commit_cqring(struct io_ring_ctx *ctx)
	117	{
	118	/* order cqe stores with ring update */
	119	smp_store_release(&ctx->rings->cq.tail, ctx->cached_cq_tail);
	120	}
	121
f3b44f92 JA	122	static inline void io_cqring_wake(struct io_ring_ctx *ctx)
	123	{
	124	/*
	125	* wake_up_all() may seem excessive, but io_wake_function() and
	126	* io_should_wake() handle the termination of the loop and only
	127	* wake as many waiters as we need to.
	128	*/
	129	if (wq_has_sleeper(&ctx->cq_wait))
	130	wake_up_all(&ctx->cq_wait);
	131	}
	132
17437f31 JA	133	static inline bool io_sqring_full(struct io_ring_ctx *ctx)
	134	{
	135	struct io_rings *r = ctx->rings;
	136
	137	return READ_ONCE(r->sq.tail) - ctx->cached_sq_head == ctx->sq_entries;
	138	}
	139
	140	static inline unsigned int io_sqring_entries(struct io_ring_ctx *ctx)
	141	{
	142	struct io_rings *rings = ctx->rings;
	143
	144	/* make sure SQ entry isn't read before tail */
	145	return smp_load_acquire(&rings->sq.tail) - ctx->cached_sq_head;
	146	}
	147
	148	static inline bool io_run_task_work(void)
	149	{
	150	if (test_thread_flag(TIF_NOTIFY_SIGNAL) \|\| task_work_pending(current)) {
	151	__set_current_state(TASK_RUNNING);
	152	clear_notify_signal();
	153	if (task_work_pending(current))
	154	task_work_run();
	155	return true;
	156	}
	157
	158	return false;
	159	}
	160
aa1e90f6 PB	161	static inline void io_tw_lock(struct io_ring_ctx ctx, bool locked)
	162	{
	163	if (!*locked) {
	164	mutex_lock(&ctx->uring_lock);
	165	*locked = true;
	166	}
	167	}
	168
	169	static inline void io_req_add_compl_list(struct io_kiocb *req)
	170	{
	171	struct io_submit_state *state = &req->ctx->submit_state;
	172
	173	if (!(req->flags & REQ_F_CQE_SKIP))
	174	state->flush_cqes = true;
	175	wq_list_add_tail(&req->comp_list, &state->compl_reqs);
	176	}
	177
73572984	178	int io_run_task_work_sig(void);
329061d3	179	void io_req_complete_failed(struct io_kiocb *req, s32 res);
99f15d8d	180	void __io_req_complete(struct io_kiocb *req, unsigned issue_flags);
59915143 JA	181	void io_req_complete_post(struct io_kiocb *req);
59915143 JA	182	void __io_req_complete_post(struct io_kiocb *req);
d245bca6	183	bool io_post_aux_cqe(struct io_ring_ctx *ctx, u64 user_data, s32 res, u32 cflags);
f9ead18c	184	void io_cqring_ev_posted(struct io_ring_ctx *ctx);
f3b44f92	185	void __io_commit_cqring_flush(struct io_ring_ctx *ctx);
f9ead18c	186
3b77495a	187	struct page *io_pin_pages(unsigned long ubuf, unsigned long len, int npages);
329061d3	188
531113bb JA	189	struct file io_file_get_normal(struct io_kiocb req, int fd);
	190	struct file io_file_get_fixed(struct io_kiocb req, int fd,
	191	unsigned issue_flags);
cd40cae2	192
cd40cae2	193	bool io_is_uring_fops(struct file *file);
99f15d8d JA	194	bool io_alloc_async_data(struct io_kiocb *req);
99f15d8d JA	195	void io_req_task_work_add(struct io_kiocb *req);
f3b44f92	196	void io_req_task_prio_work_add(struct io_kiocb *req);
59915143	197	void io_req_tw_post_queue(struct io_kiocb *req, s32 res, u32 cflags);
f3b44f92 JA	198	void io_req_task_queue(struct io_kiocb *req);
f3b44f92 JA	199	void io_queue_iowq(struct io_kiocb req, bool dont_use);
59915143 JA	200	void io_req_task_complete(struct io_kiocb req, bool locked);
59915143 JA	201	void io_req_task_queue_fail(struct io_kiocb *req, int ret);
329061d3	202	void io_req_task_submit(struct io_kiocb req, bool locked);
c9f06aa7	203	void tctx_task_work(struct callback_head *cb);
17437f31 JA	204	__cold void io_uring_cancel_generic(bool cancel_all, struct io_sq_data *sqd);
	205	int io_uring_alloc_task_context(struct task_struct *task,
	206	struct io_ring_ctx *ctx);
	207
329061d3	208	int io_poll_issue(struct io_kiocb req, bool locked);
17437f31 JA	209	int io_submit_sqes(struct io_ring_ctx *ctx, unsigned int nr);
17437f31 JA	210	int io_do_iopoll(struct io_ring_ctx *ctx, bool force_nonspin);
f3b44f92 JA	211	void io_free_batch_list(struct io_ring_ctx ctx, struct io_wq_work_node node);
f3b44f92 JA	212	int io_req_prep_async(struct io_kiocb *req);
59915143	213
c9f06aa7 JA	214	struct io_wq_work io_wq_free_work(struct io_wq_work work);
	215	void io_wq_submit_work(struct io_wq_work *work);
	216
59915143 JA	217	void io_free_req(struct io_kiocb *req);
	218	void io_queue_next(struct io_kiocb *req);
	219
329061d3 JA	220	bool io_match_task_safe(struct io_kiocb head, struct task_struct task,
	221	bool cancel_all);
	222
59915143 JA	223	#define io_for_each_link(pos, head) \
59915143 JA	224	for (pos = (head); pos; pos = pos->link)
531113bb	225
de23077e	226	#endif