[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 <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,
};

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);
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_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)

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;

	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)) {
		__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);
}

#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,
	18	};
	19
faf88dde	20	struct io_uring_cqe __io_get_cqe(struct io_ring_ctx ctx);
68494a65	21	bool io_req_cqe_overflow(struct io_kiocb *req);
9046c641 PB	22	int io_run_task_work_sig(void);
	23	void io_req_complete_failed(struct io_kiocb *req, s32 res);
	24	void __io_req_complete(struct io_kiocb *req, unsigned issue_flags);
	25	void io_req_complete_post(struct io_kiocb *req);
	26	void __io_req_complete_post(struct io_kiocb *req);
	27	bool io_post_aux_cqe(struct io_ring_ctx *ctx, u64 user_data, s32 res, u32 cflags);
9046c641 PB	28	void __io_commit_cqring_flush(struct io_ring_ctx *ctx);
	29
	30	struct page *io_pin_pages(unsigned long ubuf, unsigned long len, int npages);
	31
	32	struct file io_file_get_normal(struct io_kiocb req, int fd);
	33	struct file io_file_get_fixed(struct io_kiocb req, int fd,
	34	unsigned issue_flags);
	35
	36	bool io_is_uring_fops(struct file *file);
	37	bool io_alloc_async_data(struct io_kiocb *req);
	38	void io_req_task_work_add(struct io_kiocb *req);
9046c641 PB	39	void io_req_tw_post_queue(struct io_kiocb *req, s32 res, u32 cflags);
	40	void io_req_task_queue(struct io_kiocb *req);
	41	void io_queue_iowq(struct io_kiocb req, bool dont_use);
	42	void io_req_task_complete(struct io_kiocb req, bool locked);
	43	void io_req_task_queue_fail(struct io_kiocb *req, int ret);
	44	void io_req_task_submit(struct io_kiocb req, bool locked);
	45	void tctx_task_work(struct callback_head *cb);
	46	__cold void io_uring_cancel_generic(bool cancel_all, struct io_sq_data *sqd);
	47	int io_uring_alloc_task_context(struct task_struct *task,
	48	struct io_ring_ctx *ctx);
	49
	50	int io_poll_issue(struct io_kiocb req, bool locked);
	51	int io_submit_sqes(struct io_ring_ctx *ctx, unsigned int nr);
	52	int io_do_iopoll(struct io_ring_ctx *ctx, bool force_nonspin);
	53	void io_free_batch_list(struct io_ring_ctx ctx, struct io_wq_work_node node);
	54	int io_req_prep_async(struct io_kiocb *req);
	55
	56	struct io_wq_work io_wq_free_work(struct io_wq_work work);
	57	void io_wq_submit_work(struct io_wq_work *work);
	58
	59	void io_free_req(struct io_kiocb *req);
	60	void io_queue_next(struct io_kiocb *req);
	61
	62	bool io_match_task_safe(struct io_kiocb head, struct task_struct task,
	63	bool cancel_all);
	64
	65	#define io_for_each_link(pos, head) \
	66	for (pos = (head); pos; pos = pos->link)
f3b44f92	67
25399321 PB	68	static inline void io_cq_lock(struct io_ring_ctx *ctx)
	69	__acquires(ctx->completion_lock)
	70	{
	71	spin_lock(&ctx->completion_lock);
	72	}
	73
	74	void io_cq_unlock_post(struct io_ring_ctx *ctx);
	75
f3b44f92 JA	76	static inline struct io_uring_cqe io_get_cqe(struct io_ring_ctx ctx)
	77	{
	78	if (likely(ctx->cqe_cached < ctx->cqe_sentinel)) {
	79	struct io_uring_cqe *cqe = ctx->cqe_cached;
	80
f3b44f92 JA	81	ctx->cached_cq_tail++;
f3b44f92 JA	82	ctx->cqe_cached++;
b3659a65 PB	83	if (ctx->flags & IORING_SETUP_CQE32)
b3659a65 PB	84	ctx->cqe_cached++;
f3b44f92 JA	85	return cqe;
	86	}
	87
	88	return __io_get_cqe(ctx);
	89	}
	90
	91	static inline bool __io_fill_cqe_req(struct io_ring_ctx *ctx,
	92	struct io_kiocb *req)
	93	{
	94	struct io_uring_cqe *cqe;
	95
ae5735c6 PB	96	trace_io_uring_complete(req->ctx, req, req->cqe.user_data,
	97	req->cqe.res, req->cqe.flags,
	98	(req->flags & REQ_F_CQE32_INIT) ? req->extra1 : 0,
	99	(req->flags & REQ_F_CQE32_INIT) ? req->extra2 : 0);
e8c328c3 PB	100	/*
	101	* If we can't get a cq entry, userspace overflowed the
	102	* submission (by quite a lot). Increment the overflow count in
	103	* the ring.
	104	*/
	105	cqe = io_get_cqe(ctx);
	106	if (unlikely(!cqe))
	107	return io_req_cqe_overflow(req);
	108	memcpy(cqe, &req->cqe, sizeof(*cqe));
	109
	110	if (ctx->flags & IORING_SETUP_CQE32) {
f3b44f92 JA	111	u64 extra1 = 0, extra2 = 0;
	112
	113	if (req->flags & REQ_F_CQE32_INIT) {
	114	extra1 = req->extra1;
	115	extra2 = req->extra2;
	116	}
	117
e8c328c3 PB	118	WRITE_ONCE(cqe->big_cqe[0], extra1);
e8c328c3 PB	119	WRITE_ONCE(cqe->big_cqe[1], extra2);
f3b44f92	120	}
e8c328c3	121	return true;
f3b44f92 JA	122	}
f3b44f92 JA	123
531113bb JA	124	static inline void req_set_fail(struct io_kiocb *req)
	125	{
	126	req->flags \|= REQ_F_FAIL;
	127	if (req->flags & REQ_F_CQE_SKIP) {
	128	req->flags &= ~REQ_F_CQE_SKIP;
	129	req->flags \|= REQ_F_SKIP_LINK_CQES;
	130	}
	131	}
	132
de23077e JA	133	static inline void io_req_set_res(struct io_kiocb *req, s32 res, u32 cflags)
	134	{
	135	req->cqe.res = res;
	136	req->cqe.flags = cflags;
	137	}
	138
99f15d8d JA	139	static inline bool req_has_async_data(struct io_kiocb *req)
	140	{
	141	return req->flags & REQ_F_ASYNC_DATA;
	142	}
	143
531113bb JA	144	static inline void io_put_file(struct file *file)
	145	{
	146	if (file)
	147	fput(file);
	148	}
	149
cd40cae2 JA	150	static inline void io_ring_submit_unlock(struct io_ring_ctx *ctx,
	151	unsigned issue_flags)
	152	{
	153	lockdep_assert_held(&ctx->uring_lock);
	154	if (issue_flags & IO_URING_F_UNLOCKED)
	155	mutex_unlock(&ctx->uring_lock);
	156	}
	157
	158	static inline void io_ring_submit_lock(struct io_ring_ctx *ctx,
	159	unsigned issue_flags)
	160	{
	161	/*
	162	* "Normal" inline submissions always hold the uring_lock, since we
	163	* grab it from the system call. Same is true for the SQPOLL offload.
	164	* The only exception is when we've detached the request and issue it
	165	* from an async worker thread, grab the lock for that case.
	166	*/
	167	if (issue_flags & IO_URING_F_UNLOCKED)
	168	mutex_lock(&ctx->uring_lock);
	169	lockdep_assert_held(&ctx->uring_lock);
	170	}
	171
f9ead18c JA	172	static inline void io_commit_cqring(struct io_ring_ctx *ctx)
	173	{
	174	/* order cqe stores with ring update */
	175	smp_store_release(&ctx->rings->cq.tail, ctx->cached_cq_tail);
	176	}
	177
f3b44f92 JA	178	static inline void io_cqring_wake(struct io_ring_ctx *ctx)
	179	{
	180	/*
	181	* wake_up_all() may seem excessive, but io_wake_function() and
	182	* io_should_wake() handle the termination of the loop and only
	183	* wake as many waiters as we need to.
	184	*/
	185	if (wq_has_sleeper(&ctx->cq_wait))
	186	wake_up_all(&ctx->cq_wait);
	187	}
	188
17437f31 JA	189	static inline bool io_sqring_full(struct io_ring_ctx *ctx)
	190	{
	191	struct io_rings *r = ctx->rings;
	192
	193	return READ_ONCE(r->sq.tail) - ctx->cached_sq_head == ctx->sq_entries;
	194	}
	195
	196	static inline unsigned int io_sqring_entries(struct io_ring_ctx *ctx)
	197	{
	198	struct io_rings *rings = ctx->rings;
	199
	200	/* make sure SQ entry isn't read before tail */
	201	return smp_load_acquire(&rings->sq.tail) - ctx->cached_sq_head;
	202	}
	203
	204	static inline bool io_run_task_work(void)
	205	{
625d38b3	206	if (test_thread_flag(TIF_NOTIFY_SIGNAL)) {
17437f31 JA	207	__set_current_state(TASK_RUNNING);
	208	clear_notify_signal();
	209	if (task_work_pending(current))
	210	task_work_run();
	211	return true;
	212	}
	213
	214	return false;
	215	}
	216
aa1e90f6 PB	217	static inline void io_tw_lock(struct io_ring_ctx ctx, bool locked)
	218	{
	219	if (!*locked) {
	220	mutex_lock(&ctx->uring_lock);
	221	*locked = true;
	222	}
	223	}
	224
9da070b1 PB	225	/*
	226	* Don't complete immediately but use deferred completion infrastructure.
	227	* Protected by ->uring_lock and can only be used either with
	228	* IO_URING_F_COMPLETE_DEFER or inside a tw handler holding the mutex.
	229	*/
	230	static inline void io_req_complete_defer(struct io_kiocb *req)
	231	__must_hold(&req->ctx->uring_lock)
aa1e90f6 PB	232	{
	233	struct io_submit_state *state = &req->ctx->submit_state;
	234
9da070b1 PB	235	lockdep_assert_held(&req->ctx->uring_lock);
9da070b1 PB	236
aa1e90f6 PB	237	wq_list_add_tail(&req->comp_list, &state->compl_reqs);
	238	}
	239
46929b08 PB	240	static inline void io_commit_cqring_flush(struct io_ring_ctx *ctx)
	241	{
	242	if (unlikely(ctx->off_timeout_used \|\| ctx->drain_active \|\| ctx->has_evfd))
	243	__io_commit_cqring_flush(ctx);
	244	}
	245
de23077e	246	#endif