[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/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);
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_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);
}

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);
}

#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);
eb42cebb PB	36	bool io_fill_cqe_aux(struct io_ring_ctx *ctx, u64 user_data, s32 res, u32 cflags,
eb42cebb PB	37	bool allow_overflow);
9046c641 PB	38	void __io_commit_cqring_flush(struct io_ring_ctx *ctx);
	39
	40	struct page *io_pin_pages(unsigned long ubuf, unsigned long len, int npages);
	41
	42	struct file io_file_get_normal(struct io_kiocb req, int fd);
	43	struct file io_file_get_fixed(struct io_kiocb req, int fd,
	44	unsigned issue_flags);
	45
f6b543fd JA	46	static inline bool io_req_ffs_set(struct io_kiocb *req)
	47	{
	48	return req->flags & REQ_F_FIXED_FILE;
	49	}
	50
9046c641 PB	51	bool io_is_uring_fops(struct file *file);
	52	bool io_alloc_async_data(struct io_kiocb *req);
	53	void io_req_task_work_add(struct io_kiocb *req);
9046c641 PB	54	void io_req_tw_post_queue(struct io_kiocb *req, s32 res, u32 cflags);
	55	void io_req_task_queue(struct io_kiocb *req);
	56	void io_queue_iowq(struct io_kiocb req, bool dont_use);
	57	void io_req_task_complete(struct io_kiocb req, bool locked);
	58	void io_req_task_queue_fail(struct io_kiocb *req, int ret);
	59	void io_req_task_submit(struct io_kiocb req, bool locked);
	60	void tctx_task_work(struct callback_head *cb);
	61	__cold void io_uring_cancel_generic(bool cancel_all, struct io_sq_data *sqd);
	62	int io_uring_alloc_task_context(struct task_struct *task,
	63	struct io_ring_ctx *ctx);
	64
	65	int io_poll_issue(struct io_kiocb req, bool locked);
	66	int io_submit_sqes(struct io_ring_ctx *ctx, unsigned int nr);
	67	int io_do_iopoll(struct io_ring_ctx *ctx, bool force_nonspin);
	68	void io_free_batch_list(struct io_ring_ctx ctx, struct io_wq_work_node node);
	69	int io_req_prep_async(struct io_kiocb *req);
	70
	71	struct io_wq_work io_wq_free_work(struct io_wq_work work);
	72	void io_wq_submit_work(struct io_wq_work *work);
	73
	74	void io_free_req(struct io_kiocb *req);
	75	void io_queue_next(struct io_kiocb *req);
e70cb608	76	void __io_put_task(struct task_struct *task, int nr);
63809137	77	void io_task_refs_refill(struct io_uring_task *tctx);
9046c641 PB	78
	79	bool io_match_task_safe(struct io_kiocb head, struct task_struct task,
	80	bool cancel_all);
	81
	82	#define io_for_each_link(pos, head) \
	83	for (pos = (head); pos; pos = pos->link)
f3b44f92	84
25399321 PB	85	static inline void io_cq_lock(struct io_ring_ctx *ctx)
	86	__acquires(ctx->completion_lock)
	87	{
	88	spin_lock(&ctx->completion_lock);
	89	}
	90
	91	void io_cq_unlock_post(struct io_ring_ctx *ctx);
	92
f3b44f92 JA	93	static inline struct io_uring_cqe io_get_cqe(struct io_ring_ctx ctx)
	94	{
	95	if (likely(ctx->cqe_cached < ctx->cqe_sentinel)) {
	96	struct io_uring_cqe *cqe = ctx->cqe_cached;
	97
f3b44f92 JA	98	ctx->cached_cq_tail++;
f3b44f92 JA	99	ctx->cqe_cached++;
b3659a65 PB	100	if (ctx->flags & IORING_SETUP_CQE32)
b3659a65 PB	101	ctx->cqe_cached++;
f3b44f92 JA	102	return cqe;
	103	}
	104
	105	return __io_get_cqe(ctx);
	106	}
	107
	108	static inline bool __io_fill_cqe_req(struct io_ring_ctx *ctx,
	109	struct io_kiocb *req)
	110	{
	111	struct io_uring_cqe *cqe;
	112
e8c328c3 PB	113	/*
	114	* If we can't get a cq entry, userspace overflowed the
	115	* submission (by quite a lot). Increment the overflow count in
	116	* the ring.
	117	*/
	118	cqe = io_get_cqe(ctx);
	119	if (unlikely(!cqe))
	120	return io_req_cqe_overflow(req);
e0486f3f DY	121
	122	trace_io_uring_complete(req->ctx, req, req->cqe.user_data,
	123	req->cqe.res, req->cqe.flags,
	124	(req->flags & REQ_F_CQE32_INIT) ? req->extra1 : 0,
	125	(req->flags & REQ_F_CQE32_INIT) ? req->extra2 : 0);
	126
e8c328c3 PB	127	memcpy(cqe, &req->cqe, sizeof(*cqe));
	128
	129	if (ctx->flags & IORING_SETUP_CQE32) {
f3b44f92 JA	130	u64 extra1 = 0, extra2 = 0;
	131
	132	if (req->flags & REQ_F_CQE32_INIT) {
	133	extra1 = req->extra1;
	134	extra2 = req->extra2;
	135	}
	136
e8c328c3 PB	137	WRITE_ONCE(cqe->big_cqe[0], extra1);
e8c328c3 PB	138	WRITE_ONCE(cqe->big_cqe[1], extra2);
f3b44f92	139	}
e8c328c3	140	return true;
f3b44f92 JA	141	}
f3b44f92 JA	142
531113bb JA	143	static inline void req_set_fail(struct io_kiocb *req)
	144	{
	145	req->flags \|= REQ_F_FAIL;
	146	if (req->flags & REQ_F_CQE_SKIP) {
	147	req->flags &= ~REQ_F_CQE_SKIP;
	148	req->flags \|= REQ_F_SKIP_LINK_CQES;
	149	}
	150	}
	151
de23077e JA	152	static inline void io_req_set_res(struct io_kiocb *req, s32 res, u32 cflags)
	153	{
	154	req->cqe.res = res;
	155	req->cqe.flags = cflags;
	156	}
	157
99f15d8d JA	158	static inline bool req_has_async_data(struct io_kiocb *req)
	159	{
	160	return req->flags & REQ_F_ASYNC_DATA;
	161	}
	162
531113bb JA	163	static inline void io_put_file(struct file *file)
	164	{
	165	if (file)
	166	fput(file);
	167	}
	168
cd40cae2 JA	169	static inline void io_ring_submit_unlock(struct io_ring_ctx *ctx,
	170	unsigned issue_flags)
	171	{
	172	lockdep_assert_held(&ctx->uring_lock);
	173	if (issue_flags & IO_URING_F_UNLOCKED)
	174	mutex_unlock(&ctx->uring_lock);
	175	}
	176
	177	static inline void io_ring_submit_lock(struct io_ring_ctx *ctx,
	178	unsigned issue_flags)
	179	{
	180	/*
	181	* "Normal" inline submissions always hold the uring_lock, since we
	182	* grab it from the system call. Same is true for the SQPOLL offload.
	183	* The only exception is when we've detached the request and issue it
	184	* from an async worker thread, grab the lock for that case.
	185	*/
	186	if (issue_flags & IO_URING_F_UNLOCKED)
	187	mutex_lock(&ctx->uring_lock);
	188	lockdep_assert_held(&ctx->uring_lock);
	189	}
	190
f9ead18c JA	191	static inline void io_commit_cqring(struct io_ring_ctx *ctx)
	192	{
	193	/* order cqe stores with ring update */
	194	smp_store_release(&ctx->rings->cq.tail, ctx->cached_cq_tail);
	195	}
	196
f3b44f92 JA	197	static inline void io_cqring_wake(struct io_ring_ctx *ctx)
	198	{
	199	/*
	200	* wake_up_all() may seem excessive, but io_wake_function() and
	201	* io_should_wake() handle the termination of the loop and only
	202	* wake as many waiters as we need to.
	203	*/
	204	if (wq_has_sleeper(&ctx->cq_wait))
	205	wake_up_all(&ctx->cq_wait);
	206	}
	207
17437f31 JA	208	static inline bool io_sqring_full(struct io_ring_ctx *ctx)
	209	{
	210	struct io_rings *r = ctx->rings;
	211
	212	return READ_ONCE(r->sq.tail) - ctx->cached_sq_head == ctx->sq_entries;
	213	}
	214
	215	static inline unsigned int io_sqring_entries(struct io_ring_ctx *ctx)
	216	{
	217	struct io_rings *rings = ctx->rings;
	218
	219	/* make sure SQ entry isn't read before tail */
	220	return smp_load_acquire(&rings->sq.tail) - ctx->cached_sq_head;
	221	}
	222
	223	static inline bool io_run_task_work(void)
	224	{
625d38b3	225	if (test_thread_flag(TIF_NOTIFY_SIGNAL)) {
17437f31 JA	226	__set_current_state(TASK_RUNNING);
	227	clear_notify_signal();
	228	if (task_work_pending(current))
	229	task_work_run();
	230	return true;
	231	}
	232
	233	return false;
	234	}
	235
aa1e90f6 PB	236	static inline void io_tw_lock(struct io_ring_ctx ctx, bool locked)
	237	{
	238	if (!*locked) {
	239	mutex_lock(&ctx->uring_lock);
	240	*locked = true;
	241	}
	242	}
	243
9da070b1 PB	244	/*
	245	* Don't complete immediately but use deferred completion infrastructure.
	246	* Protected by ->uring_lock and can only be used either with
	247	* IO_URING_F_COMPLETE_DEFER or inside a tw handler holding the mutex.
	248	*/
	249	static inline void io_req_complete_defer(struct io_kiocb *req)
	250	__must_hold(&req->ctx->uring_lock)
aa1e90f6 PB	251	{
	252	struct io_submit_state *state = &req->ctx->submit_state;
	253
9da070b1 PB	254	lockdep_assert_held(&req->ctx->uring_lock);
9da070b1 PB	255
aa1e90f6 PB	256	wq_list_add_tail(&req->comp_list, &state->compl_reqs);
	257	}
	258
46929b08 PB	259	static inline void io_commit_cqring_flush(struct io_ring_ctx *ctx)
	260	{
	261	if (unlikely(ctx->off_timeout_used \|\| ctx->drain_active \|\| ctx->has_evfd))
	262	__io_commit_cqring_flush(ctx);
	263	}
	264
e70cb608 PB	265	/* must to be called somewhat shortly after putting a request */
	266	static inline void io_put_task(struct task_struct *task, int nr)
	267	{
	268	if (likely(task == current))
	269	task->io_uring->cached_refs += nr;
	270	else
	271	__io_put_task(task, nr);
	272	}
	273
63809137 PB	274	static inline void io_get_task_refs(int nr)
	275	{
	276	struct io_uring_task *tctx = current->io_uring;
	277
	278	tctx->cached_refs -= nr;
	279	if (unlikely(tctx->cached_refs < 0))
	280	io_task_refs_refill(tctx);
	281	}
	282
de23077e	283	#endif