[linux-block.git] / include / linux / io_uring_types.h

#ifndef IO_URING_TYPES_H
#define IO_URING_TYPES_H

#include <linux/blkdev.h>
#include <linux/task_work.h>
#include <linux/bitmap.h>
#include <linux/llist.h>
#include <uapi/linux/io_uring.h>

struct io_wq_work_node {
	struct io_wq_work_node *next;
};

struct io_wq_work_list {
	struct io_wq_work_node *first;
	struct io_wq_work_node *last;
};

struct io_wq_work {
	struct io_wq_work_node list;
	unsigned flags;
	/* place it here instead of io_kiocb as it fills padding and saves 4B */
	int cancel_seq;
};

struct io_fixed_file {
	/* file * with additional FFS_* flags */
	unsigned long file_ptr;
};

struct io_file_table {
	struct io_fixed_file *files;
	unsigned long *bitmap;
	unsigned int alloc_hint;
};

struct io_notif;
struct io_notif_slot;

struct io_hash_bucket {
	spinlock_t		lock;
	struct hlist_head	list;
} ____cacheline_aligned_in_smp;

struct io_hash_table {
	struct io_hash_bucket	*hbs;
	unsigned		hash_bits;
};

/*
 * Arbitrary limit, can be raised if need be
 */
#define IO_RINGFD_REG_MAX 16

struct io_uring_task {
	/* submission side */
	int				cached_refs;
	const struct io_ring_ctx 	*last;
	struct io_wq			*io_wq;
	struct file			*registered_rings[IO_RINGFD_REG_MAX];

	struct xarray			xa;
	struct wait_queue_head		wait;
	atomic_t			in_idle;
	atomic_t			inflight_tracked;
	struct percpu_counter		inflight;

	struct { /* task_work */
		struct llist_head	task_list;
		struct callback_head	task_work;
	} ____cacheline_aligned_in_smp;
};

struct io_uring {
	u32 head ____cacheline_aligned_in_smp;
	u32 tail ____cacheline_aligned_in_smp;
};

/*
 * This data is shared with the application through the mmap at offsets
 * IORING_OFF_SQ_RING and IORING_OFF_CQ_RING.
 *
 * The offsets to the member fields are published through struct
 * io_sqring_offsets when calling io_uring_setup.
 */
struct io_rings {
	/*
	 * Head and tail offsets into the ring; the offsets need to be
	 * masked to get valid indices.
	 *
	 * The kernel controls head of the sq ring and the tail of the cq ring,
	 * and the application controls tail of the sq ring and the head of the
	 * cq ring.
	 */
	struct io_uring		sq, cq;
	/*
	 * Bitmasks to apply to head and tail offsets (constant, equals
	 * ring_entries - 1)
	 */
	u32			sq_ring_mask, cq_ring_mask;
	/* Ring sizes (constant, power of 2) */
	u32			sq_ring_entries, cq_ring_entries;
	/*
	 * Number of invalid entries dropped by the kernel due to
	 * invalid index stored in array
	 *
	 * Written by the kernel, shouldn't be modified by the
	 * application (i.e. get number of "new events" by comparing to
	 * cached value).
	 *
	 * After a new SQ head value was read by the application this
	 * counter includes all submissions that were dropped reaching
	 * the new SQ head (and possibly more).
	 */
	u32			sq_dropped;
	/*
	 * Runtime SQ flags
	 *
	 * Written by the kernel, shouldn't be modified by the
	 * application.
	 *
	 * The application needs a full memory barrier before checking
	 * for IORING_SQ_NEED_WAKEUP after updating the sq tail.
	 */
	atomic_t		sq_flags;
	/*
	 * Runtime CQ flags
	 *
	 * Written by the application, shouldn't be modified by the
	 * kernel.
	 */
	u32			cq_flags;
	/*
	 * Number of completion events lost because the queue was full;
	 * this should be avoided by the application by making sure
	 * there are not more requests pending than there is space in
	 * the completion queue.
	 *
	 * Written by the kernel, shouldn't be modified by the
	 * application (i.e. get number of "new events" by comparing to
	 * cached value).
	 *
	 * As completion events come in out of order this counter is not
	 * ordered with any other data.
	 */
	u32			cq_overflow;
	/*
	 * Ring buffer of completion events.
	 *
	 * The kernel writes completion events fresh every time they are
	 * produced, so the application is allowed to modify pending
	 * entries.
	 */
	struct io_uring_cqe	cqes[] ____cacheline_aligned_in_smp;
};

struct io_restriction {
	DECLARE_BITMAP(register_op, IORING_REGISTER_LAST);
	DECLARE_BITMAP(sqe_op, IORING_OP_LAST);
	u8 sqe_flags_allowed;
	u8 sqe_flags_required;
	bool registered;
};

struct io_submit_link {
	struct io_kiocb		*head;
	struct io_kiocb		*last;
};

struct io_submit_state {
	/* inline/task_work completion list, under ->uring_lock */
	struct io_wq_work_node	free_list;
	/* batch completion logic */
	struct io_wq_work_list	compl_reqs;
	struct io_submit_link	link;

	bool			plug_started;
	bool			need_plug;
	unsigned short		submit_nr;
	struct blk_plug		plug;
};

struct io_ev_fd {
	struct eventfd_ctx	*cq_ev_fd;
	unsigned int		eventfd_async: 1;
	struct rcu_head		rcu;
};

struct io_alloc_cache {
	struct hlist_head	list;
	unsigned int		nr_cached;
};

struct io_ring_ctx {
	/* const or read-mostly hot data */
	struct {
		struct percpu_ref	refs;

		struct io_rings		*rings;
		unsigned int		flags;
		enum task_work_notify_mode	notify_method;
		unsigned int		compat: 1;
		unsigned int		drain_next: 1;
		unsigned int		restricted: 1;
		unsigned int		off_timeout_used: 1;
		unsigned int		drain_active: 1;
		unsigned int		drain_disabled: 1;
		unsigned int		has_evfd: 1;
		unsigned int		syscall_iopoll: 1;
	} ____cacheline_aligned_in_smp;

	/* submission data */
	struct {
		struct mutex		uring_lock;

		/*
		 * Ring buffer of indices into array of io_uring_sqe, which is
		 * mmapped by the application using the IORING_OFF_SQES offset.
		 *
		 * This indirection could e.g. be used to assign fixed
		 * io_uring_sqe entries to operations and only submit them to
		 * the queue when needed.
		 *
		 * The kernel modifies neither the indices array nor the entries
		 * array.
		 */
		u32			*sq_array;
		struct io_uring_sqe	*sq_sqes;
		unsigned		cached_sq_head;
		unsigned		sq_entries;

		/*
		 * Fixed resources fast path, should be accessed only under
		 * uring_lock, and updated through io_uring_register(2)
		 */
		struct io_rsrc_node	*rsrc_node;
		int			rsrc_cached_refs;
		atomic_t		cancel_seq;
		struct io_file_table	file_table;
		unsigned		nr_user_files;
		unsigned		nr_user_bufs;
		struct io_mapped_ubuf	**user_bufs;
		struct io_notif_slot	*notif_slots;
		unsigned		nr_notif_slots;

		struct io_submit_state	submit_state;

		struct io_buffer_list	*io_bl;
		struct xarray		io_bl_xa;
		struct list_head	io_buffers_cache;

		struct io_hash_table	cancel_table_locked;
		struct list_head	cq_overflow_list;
		struct io_alloc_cache	apoll_cache;
		struct io_alloc_cache	netmsg_cache;
	} ____cacheline_aligned_in_smp;

	/* IRQ completion list, under ->completion_lock */
	struct io_wq_work_list	locked_free_list;
	unsigned int		locked_free_nr;

	const struct cred	*sq_creds;	/* cred used for __io_sq_thread() */
	struct io_sq_data	*sq_data;	/* if using sq thread polling */

	struct wait_queue_head	sqo_sq_wait;
	struct list_head	sqd_list;

	unsigned long		check_cq;

	unsigned int		file_alloc_start;
	unsigned int		file_alloc_end;

	struct xarray		personalities;
	u32			pers_next;

	struct {
		/*
		 * We cache a range of free CQEs we can use, once exhausted it
		 * should go through a slower range setup, see __io_get_cqe()
		 */
		struct io_uring_cqe	*cqe_cached;
		struct io_uring_cqe	*cqe_sentinel;

		unsigned		cached_cq_tail;
		unsigned		cq_entries;
		struct io_ev_fd	__rcu	*io_ev_fd;
		struct wait_queue_head	cq_wait;
		unsigned		cq_extra;
	} ____cacheline_aligned_in_smp;

	struct {
		spinlock_t		completion_lock;

		/*
		 * ->iopoll_list is protected by the ctx->uring_lock for
		 * io_uring instances that don't use IORING_SETUP_SQPOLL.
		 * For SQPOLL, only the single threaded io_sq_thread() will
		 * manipulate the list, hence no extra locking is needed there.
		 */
		struct io_wq_work_list	iopoll_list;
		struct io_hash_table	cancel_table;
		bool			poll_multi_queue;

		struct list_head	io_buffers_comp;
	} ____cacheline_aligned_in_smp;

	/* timeouts */
	struct {
		spinlock_t		timeout_lock;
		atomic_t		cq_timeouts;
		struct list_head	timeout_list;
		struct list_head	ltimeout_list;
		unsigned		cq_last_tm_flush;
	} ____cacheline_aligned_in_smp;

	/* Keep this last, we don't need it for the fast path */

	struct io_restriction		restrictions;
	struct task_struct		*submitter_task;

	/* slow path rsrc auxilary data, used by update/register */
	struct io_rsrc_node		*rsrc_backup_node;
	struct io_mapped_ubuf		*dummy_ubuf;
	struct io_rsrc_data		*file_data;
	struct io_rsrc_data		*buf_data;

	struct delayed_work		rsrc_put_work;
	struct llist_head		rsrc_put_llist;
	struct list_head		rsrc_ref_list;
	spinlock_t			rsrc_ref_lock;

	struct list_head		io_buffers_pages;

	#if defined(CONFIG_UNIX)
		struct socket		*ring_sock;
	#endif
	/* hashed buffered write serialization */
	struct io_wq_hash		*hash_map;

	/* Only used for accounting purposes */
	struct user_struct		*user;
	struct mm_struct		*mm_account;

	/* ctx exit and cancelation */
	struct llist_head		fallback_llist;
	struct delayed_work		fallback_work;
	struct work_struct		exit_work;
	struct list_head		tctx_list;
	struct completion		ref_comp;

	/* io-wq management, e.g. thread count */
	u32				iowq_limits[2];
	bool				iowq_limits_set;

	struct list_head		defer_list;
	unsigned			sq_thread_idle;
	/* protected by ->completion_lock */
	unsigned			evfd_last_cq_tail;
};

enum {
	REQ_F_FIXED_FILE_BIT	= IOSQE_FIXED_FILE_BIT,
	REQ_F_IO_DRAIN_BIT	= IOSQE_IO_DRAIN_BIT,
	REQ_F_LINK_BIT		= IOSQE_IO_LINK_BIT,
	REQ_F_HARDLINK_BIT	= IOSQE_IO_HARDLINK_BIT,
	REQ_F_FORCE_ASYNC_BIT	= IOSQE_ASYNC_BIT,
	REQ_F_BUFFER_SELECT_BIT	= IOSQE_BUFFER_SELECT_BIT,
	REQ_F_CQE_SKIP_BIT	= IOSQE_CQE_SKIP_SUCCESS_BIT,

	/* first byte is taken by user flags, shift it to not overlap */
	REQ_F_FAIL_BIT		= 8,
	REQ_F_INFLIGHT_BIT,
	REQ_F_CUR_POS_BIT,
	REQ_F_NOWAIT_BIT,
	REQ_F_LINK_TIMEOUT_BIT,
	REQ_F_NEED_CLEANUP_BIT,
	REQ_F_POLLED_BIT,
	REQ_F_BUFFER_SELECTED_BIT,
	REQ_F_BUFFER_RING_BIT,
	REQ_F_REISSUE_BIT,
	REQ_F_CREDS_BIT,
	REQ_F_REFCOUNT_BIT,
	REQ_F_ARM_LTIMEOUT_BIT,
	REQ_F_ASYNC_DATA_BIT,
	REQ_F_SKIP_LINK_CQES_BIT,
	REQ_F_SINGLE_POLL_BIT,
	REQ_F_DOUBLE_POLL_BIT,
	REQ_F_PARTIAL_IO_BIT,
	REQ_F_CQE32_INIT_BIT,
	REQ_F_APOLL_MULTISHOT_BIT,
	REQ_F_CLEAR_POLLIN_BIT,
	REQ_F_HASH_LOCKED_BIT,
	/* keep async read/write and isreg together and in order */
	REQ_F_SUPPORT_NOWAIT_BIT,
	REQ_F_ISREG_BIT,

	/* not a real bit, just to check we're not overflowing the space */
	__REQ_F_LAST_BIT,
};

enum {
	/* ctx owns file */
	REQ_F_FIXED_FILE	= BIT(REQ_F_FIXED_FILE_BIT),
	/* drain existing IO first */
	REQ_F_IO_DRAIN		= BIT(REQ_F_IO_DRAIN_BIT),
	/* linked sqes */
	REQ_F_LINK		= BIT(REQ_F_LINK_BIT),
	/* doesn't sever on completion < 0 */
	REQ_F_HARDLINK		= BIT(REQ_F_HARDLINK_BIT),
	/* IOSQE_ASYNC */
	REQ_F_FORCE_ASYNC	= BIT(REQ_F_FORCE_ASYNC_BIT),
	/* IOSQE_BUFFER_SELECT */
	REQ_F_BUFFER_SELECT	= BIT(REQ_F_BUFFER_SELECT_BIT),
	/* IOSQE_CQE_SKIP_SUCCESS */
	REQ_F_CQE_SKIP		= BIT(REQ_F_CQE_SKIP_BIT),

	/* fail rest of links */
	REQ_F_FAIL		= BIT(REQ_F_FAIL_BIT),
	/* on inflight list, should be cancelled and waited on exit reliably */
	REQ_F_INFLIGHT		= BIT(REQ_F_INFLIGHT_BIT),
	/* read/write uses file position */
	REQ_F_CUR_POS		= BIT(REQ_F_CUR_POS_BIT),
	/* must not punt to workers */
	REQ_F_NOWAIT		= BIT(REQ_F_NOWAIT_BIT),
	/* has or had linked timeout */
	REQ_F_LINK_TIMEOUT	= BIT(REQ_F_LINK_TIMEOUT_BIT),
	/* needs cleanup */
	REQ_F_NEED_CLEANUP	= BIT(REQ_F_NEED_CLEANUP_BIT),
	/* already went through poll handler */
	REQ_F_POLLED		= BIT(REQ_F_POLLED_BIT),
	/* buffer already selected */
	REQ_F_BUFFER_SELECTED	= BIT(REQ_F_BUFFER_SELECTED_BIT),
	/* buffer selected from ring, needs commit */
	REQ_F_BUFFER_RING	= BIT(REQ_F_BUFFER_RING_BIT),
	/* caller should reissue async */
	REQ_F_REISSUE		= BIT(REQ_F_REISSUE_BIT),
	/* supports async reads/writes */
	REQ_F_SUPPORT_NOWAIT	= BIT(REQ_F_SUPPORT_NOWAIT_BIT),
	/* regular file */
	REQ_F_ISREG		= BIT(REQ_F_ISREG_BIT),
	/* has creds assigned */
	REQ_F_CREDS		= BIT(REQ_F_CREDS_BIT),
	/* skip refcounting if not set */
	REQ_F_REFCOUNT		= BIT(REQ_F_REFCOUNT_BIT),
	/* there is a linked timeout that has to be armed */
	REQ_F_ARM_LTIMEOUT	= BIT(REQ_F_ARM_LTIMEOUT_BIT),
	/* ->async_data allocated */
	REQ_F_ASYNC_DATA	= BIT(REQ_F_ASYNC_DATA_BIT),
	/* don't post CQEs while failing linked requests */
	REQ_F_SKIP_LINK_CQES	= BIT(REQ_F_SKIP_LINK_CQES_BIT),
	/* single poll may be active */
	REQ_F_SINGLE_POLL	= BIT(REQ_F_SINGLE_POLL_BIT),
	/* double poll may active */
	REQ_F_DOUBLE_POLL	= BIT(REQ_F_DOUBLE_POLL_BIT),
	/* request has already done partial IO */
	REQ_F_PARTIAL_IO	= BIT(REQ_F_PARTIAL_IO_BIT),
	/* fast poll multishot mode */
	REQ_F_APOLL_MULTISHOT	= BIT(REQ_F_APOLL_MULTISHOT_BIT),
	/* ->extra1 and ->extra2 are initialised */
	REQ_F_CQE32_INIT	= BIT(REQ_F_CQE32_INIT_BIT),
	/* recvmsg special flag, clear EPOLLIN */
	REQ_F_CLEAR_POLLIN	= BIT(REQ_F_CLEAR_POLLIN_BIT),
	/* hashed into ->cancel_hash_locked, protected by ->uring_lock */
	REQ_F_HASH_LOCKED	= BIT(REQ_F_HASH_LOCKED_BIT),
};

typedef void (*io_req_tw_func_t)(struct io_kiocb *req, bool *locked);

struct io_task_work {
	struct llist_node		node;
	io_req_tw_func_t		func;
};

struct io_cqe {
	__u64	user_data;
	__s32	res;
	/* fd initially, then cflags for completion */
	union {
		__u32	flags;
		int	fd;
	};
};

/*
 * Each request type overlays its private data structure on top of this one.
 * They must not exceed this one in size.
 */
struct io_cmd_data {
	struct file		*file;
	/* each command gets 56 bytes of data */
	__u8			data[56];
};

static inline void io_kiocb_cmd_sz_check(size_t cmd_sz)
{
	BUILD_BUG_ON(cmd_sz > sizeof(struct io_cmd_data));
}
#define io_kiocb_to_cmd(req, cmd_type) ( \
	io_kiocb_cmd_sz_check(sizeof(cmd_type)) , \
	((cmd_type *)&(req)->cmd) \
)
#define cmd_to_io_kiocb(ptr)	((struct io_kiocb *) ptr)

struct io_kiocb {
	union {
		/*
		 * NOTE! Each of the io_kiocb union members has the file pointer
		 * as the first entry in their struct definition. So you can
		 * access the file pointer through any of the sub-structs,
		 * or directly as just 'file' in this struct.
		 */
		struct file		*file;
		struct io_cmd_data	cmd;
	};

	u8				opcode;
	/* polled IO has completed */
	u8				iopoll_completed;
	/*
	 * Can be either a fixed buffer index, or used with provided buffers.
	 * For the latter, before issue it points to the buffer group ID,
	 * and after selection it points to the buffer ID itself.
	 */
	u16				buf_index;
	unsigned int			flags;

	struct io_cqe			cqe;

	struct io_ring_ctx		*ctx;
	struct task_struct		*task;

	struct io_rsrc_node		*rsrc_node;

	union {
		/* store used ubuf, so we can prevent reloading */
		struct io_mapped_ubuf	*imu;

		/* stores selected buf, valid IFF REQ_F_BUFFER_SELECTED is set */
		struct io_buffer	*kbuf;

		/*
		 * stores buffer ID for ring provided buffers, valid IFF
		 * REQ_F_BUFFER_RING is set.
		 */
		struct io_buffer_list	*buf_list;
	};

	union {
		/* used by request caches, completion batching and iopoll */
		struct io_wq_work_node	comp_list;
		/* cache ->apoll->events */
		__poll_t apoll_events;
	};
	atomic_t			refs;
	atomic_t			poll_refs;
	struct io_task_work		io_task_work;
	/* for polled requests, i.e. IORING_OP_POLL_ADD and async armed poll */
	union {
		struct hlist_node	hash_node;
		struct {
			u64		extra1;
			u64		extra2;
		};
	};
	/* internal polling, see IORING_FEAT_FAST_POLL */
	struct async_poll		*apoll;
	/* opcode allocated if it needs to store data for async defer */
	void				*async_data;
	/* linked requests, IFF REQ_F_HARDLINK or REQ_F_LINK are set */
	struct io_kiocb			*link;
	/* custom credentials, valid IFF REQ_F_CREDS is set */
	const struct cred		*creds;
	struct io_wq_work		work;
};

struct io_overflow_cqe {
	struct list_head list;
	struct io_uring_cqe cqe;
};

#endif
Commit	Line	Data
e27f928e JA	1	#ifndef IO_URING_TYPES_H
	2	#define IO_URING_TYPES_H
	3
	4	#include <linux/blkdev.h>
	5	#include <linux/task_work.h>
d9b57aa3	6	#include <linux/bitmap.h>
e70cb608	7	#include <linux/llist.h>
d9b57aa3	8	#include <uapi/linux/io_uring.h>
e27f928e	9
ab1c84d8 PB	10	struct io_wq_work_node {
	11	struct io_wq_work_node *next;
	12	};
	13
	14	struct io_wq_work_list {
	15	struct io_wq_work_node *first;
	16	struct io_wq_work_node *last;
	17	};
	18
	19	struct io_wq_work {
	20	struct io_wq_work_node list;
	21	unsigned flags;
	22	/* place it here instead of io_kiocb as it fills padding and saves 4B */
	23	int cancel_seq;
	24	};
	25
	26	struct io_fixed_file {
	27	/* file * with additional FFS_* flags */
	28	unsigned long file_ptr;
	29	};
	30
	31	struct io_file_table {
	32	struct io_fixed_file *files;
	33	unsigned long *bitmap;
	34	unsigned int alloc_hint;
	35	};
e27f928e	36
eb42cebb PB	37	struct io_notif;
	38	struct io_notif_slot;
	39
e6f89be6 PB	40	struct io_hash_bucket {
	41	spinlock_t lock;
	42	struct hlist_head list;
	43	} ____cacheline_aligned_in_smp;
	44
	45	struct io_hash_table {
	46	struct io_hash_bucket *hbs;
	47	unsigned hash_bits;
	48	};
	49
e70cb608 PB	50	/*
	51	* Arbitrary limit, can be raised if need be
	52	*/
	53	#define IO_RINGFD_REG_MAX 16
	54
	55	struct io_uring_task {
	56	/* submission side */
	57	int cached_refs;
	58	const struct io_ring_ctx *last;
	59	struct io_wq *io_wq;
	60	struct file *registered_rings[IO_RINGFD_REG_MAX];
	61
	62	struct xarray xa;
	63	struct wait_queue_head wait;
	64	atomic_t in_idle;
	65	atomic_t inflight_tracked;
	66	struct percpu_counter inflight;
	67
	68	struct { /* task_work */
	69	struct llist_head task_list;
	70	struct callback_head task_work;
	71	} ____cacheline_aligned_in_smp;
	72	};
	73
e27f928e JA	74	struct io_uring {
	75	u32 head ____cacheline_aligned_in_smp;
	76	u32 tail ____cacheline_aligned_in_smp;
	77	};
	78
	79	/*
	80	* This data is shared with the application through the mmap at offsets
	81	* IORING_OFF_SQ_RING and IORING_OFF_CQ_RING.
	82	*
	83	* The offsets to the member fields are published through struct
	84	* io_sqring_offsets when calling io_uring_setup.
	85	*/
	86	struct io_rings {
	87	/*
	88	* Head and tail offsets into the ring; the offsets need to be
	89	* masked to get valid indices.
	90	*
	91	* The kernel controls head of the sq ring and the tail of the cq ring,
	92	* and the application controls tail of the sq ring and the head of the
	93	* cq ring.
	94	*/
	95	struct io_uring sq, cq;
	96	/*
	97	* Bitmasks to apply to head and tail offsets (constant, equals
	98	* ring_entries - 1)
	99	*/
	100	u32 sq_ring_mask, cq_ring_mask;
	101	/* Ring sizes (constant, power of 2) */
	102	u32 sq_ring_entries, cq_ring_entries;
	103	/*
	104	* Number of invalid entries dropped by the kernel due to
	105	* invalid index stored in array
	106	*
	107	* Written by the kernel, shouldn't be modified by the
	108	* application (i.e. get number of "new events" by comparing to
	109	* cached value).
	110	*
	111	* After a new SQ head value was read by the application this
	112	* counter includes all submissions that were dropped reaching
	113	* the new SQ head (and possibly more).
	114	*/
	115	u32 sq_dropped;
	116	/*
	117	* Runtime SQ flags
	118	*
	119	* Written by the kernel, shouldn't be modified by the
	120	* application.
	121	*
	122	* The application needs a full memory barrier before checking
	123	* for IORING_SQ_NEED_WAKEUP after updating the sq tail.
	124	*/
	125	atomic_t sq_flags;
	126	/*
	127	* Runtime CQ flags
	128	*
	129	* Written by the application, shouldn't be modified by the
	130	* kernel.
	131	*/
	132	u32 cq_flags;
	133	/*
	134	* Number of completion events lost because the queue was full;
	135	* this should be avoided by the application by making sure
	136	* there are not more requests pending than there is space in
	137	* the completion queue.
138	*
139	* Written by the kernel, shouldn't be modified by the
140	* application (i.e. get number of "new events" by comparing to
141	* cached value).
142	*
143	* As completion events come in out of order this counter is not
144	* ordered with any other data.
145	*/
146	u32 cq_overflow;
147	/*
148	* Ring buffer of completion events.
149	*
150	* The kernel writes completion events fresh every time they are
151	* produced, so the application is allowed to modify pending
152	* entries.
153	*/
154	struct io_uring_cqe cqes[] ____cacheline_aligned_in_smp;
155	};
156
157	struct io_restriction {
158	DECLARE_BITMAP(register_op, IORING_REGISTER_LAST);
159	DECLARE_BITMAP(sqe_op, IORING_OP_LAST);
160	u8 sqe_flags_allowed;
161	u8 sqe_flags_required;
162	bool registered;
163	};
164
165	struct io_submit_link {
166	struct io_kiocb *head;
167	struct io_kiocb *last;
168	};
169
170	struct io_submit_state {
171	/* inline/task_work completion list, under ->uring_lock */
172	struct io_wq_work_node free_list;
173	/* batch completion logic */
174	struct io_wq_work_list compl_reqs;
175	struct io_submit_link link;
176
177	bool plug_started;
178	bool need_plug;
e27f928e JA	179	unsigned short submit_nr;
	180	struct blk_plug plug;
	181	};
	182
	183	struct io_ev_fd {
	184	struct eventfd_ctx *cq_ev_fd;
	185	unsigned int eventfd_async: 1;
	186	struct rcu_head rcu;
	187	};
	188
9b797a37 JA	189	struct io_alloc_cache {
9b797a37 JA	190	struct hlist_head list;
9731bc98	191	unsigned int nr_cached;
9b797a37 JA	192	};
9b797a37 JA	193
e27f928e JA	194	struct io_ring_ctx {
	195	/* const or read-mostly hot data */
	196	struct {
	197	struct percpu_ref refs;
	198
	199	struct io_rings *rings;
	200	unsigned int flags;
	201	enum task_work_notify_mode notify_method;
	202	unsigned int compat: 1;
	203	unsigned int drain_next: 1;
	204	unsigned int restricted: 1;
	205	unsigned int off_timeout_used: 1;
	206	unsigned int drain_active: 1;
	207	unsigned int drain_disabled: 1;
	208	unsigned int has_evfd: 1;
	209	unsigned int syscall_iopoll: 1;
	210	} ____cacheline_aligned_in_smp;
	211
	212	/* submission data */
	213	struct {
	214	struct mutex uring_lock;
	215
	216	/*
	217	* Ring buffer of indices into array of io_uring_sqe, which is
	218	* mmapped by the application using the IORING_OFF_SQES offset.
	219	*
	220	* This indirection could e.g. be used to assign fixed
	221	* io_uring_sqe entries to operations and only submit them to
	222	* the queue when needed.
	223	*
	224	* The kernel modifies neither the indices array nor the entries
	225	* array.
	226	*/
	227	u32 *sq_array;
	228	struct io_uring_sqe *sq_sqes;
	229	unsigned cached_sq_head;
	230	unsigned sq_entries;
e27f928e JA	231
	232	/*
	233	* Fixed resources fast path, should be accessed only under
	234	* uring_lock, and updated through io_uring_register(2)
	235	*/
	236	struct io_rsrc_node *rsrc_node;
	237	int rsrc_cached_refs;
	238	atomic_t cancel_seq;
	239	struct io_file_table file_table;
	240	unsigned nr_user_files;
	241	unsigned nr_user_bufs;
	242	struct io_mapped_ubuf **user_bufs;
eb42cebb PB	243	struct io_notif_slot *notif_slots;
eb42cebb PB	244	unsigned nr_notif_slots;
e27f928e JA	245
	246	struct io_submit_state submit_state;
	247
	248	struct io_buffer_list *io_bl;
	249	struct xarray io_bl_xa;
	250	struct list_head io_buffers_cache;
	251
9ca9fb24	252	struct io_hash_table cancel_table_locked;
e27f928e	253	struct list_head cq_overflow_list;
9b797a37	254	struct io_alloc_cache apoll_cache;
43e0bbbd	255	struct io_alloc_cache netmsg_cache;
e27f928e JA	256	} ____cacheline_aligned_in_smp;
	257
	258	/* IRQ completion list, under ->completion_lock */
	259	struct io_wq_work_list locked_free_list;
	260	unsigned int locked_free_nr;
	261
	262	const struct cred sq_creds; / cred used for __io_sq_thread() */
	263	struct io_sq_data sq_data; / if using sq thread polling */
	264
	265	struct wait_queue_head sqo_sq_wait;
	266	struct list_head sqd_list;
	267
	268	unsigned long check_cq;
	269
6e73dffb PB	270	unsigned int file_alloc_start;
	271	unsigned int file_alloc_end;
	272
43e0bbbd JA	273	struct xarray personalities;
	274	u32 pers_next;
	275
e27f928e JA	276	struct {
	277	/*
	278	* We cache a range of free CQEs we can use, once exhausted it
	279	* should go through a slower range setup, see __io_get_cqe()
	280	*/
	281	struct io_uring_cqe *cqe_cached;
	282	struct io_uring_cqe *cqe_sentinel;
	283
	284	unsigned cached_cq_tail;
	285	unsigned cq_entries;
	286	struct io_ev_fd __rcu *io_ev_fd;
	287	struct wait_queue_head cq_wait;
	288	unsigned cq_extra;
e27f928e JA	289	} ____cacheline_aligned_in_smp;
	290
	291	struct {
	292	spinlock_t completion_lock;
	293
e27f928e JA	294	/*
	295	* ->iopoll_list is protected by the ctx->uring_lock for
	296	* io_uring instances that don't use IORING_SETUP_SQPOLL.
	297	* For SQPOLL, only the single threaded io_sq_thread() will
	298	* manipulate the list, hence no extra locking is needed there.
	299	*/
	300	struct io_wq_work_list iopoll_list;
e6f89be6	301	struct io_hash_table cancel_table;
e27f928e JA	302	bool poll_multi_queue;
	303
	304	struct list_head io_buffers_comp;
	305	} ____cacheline_aligned_in_smp;
	306
aff5b2df PB	307	/* timeouts */
	308	struct {
	309	spinlock_t timeout_lock;
	310	atomic_t cq_timeouts;
	311	struct list_head timeout_list;
	312	struct list_head ltimeout_list;
	313	unsigned cq_last_tm_flush;
	314	} ____cacheline_aligned_in_smp;
	315
e27f928e	316	/* Keep this last, we don't need it for the fast path */
22eb2a3f PB	317
22eb2a3f PB	318	struct io_restriction restrictions;
97bbdc06	319	struct task_struct *submitter_task;
22eb2a3f PB	320
	321	/* slow path rsrc auxilary data, used by update/register */
	322	struct io_rsrc_node *rsrc_backup_node;
	323	struct io_mapped_ubuf *dummy_ubuf;
	324	struct io_rsrc_data *file_data;
	325	struct io_rsrc_data *buf_data;
	326
	327	struct delayed_work rsrc_put_work;
	328	struct llist_head rsrc_put_llist;
	329	struct list_head rsrc_ref_list;
	330	spinlock_t rsrc_ref_lock;
	331
	332	struct list_head io_buffers_pages;
	333
	334	#if defined(CONFIG_UNIX)
	335	struct socket *ring_sock;
	336	#endif
	337	/* hashed buffered write serialization */
	338	struct io_wq_hash *hash_map;
	339
	340	/* Only used for accounting purposes */
	341	struct user_struct *user;
	342	struct mm_struct *mm_account;
	343
	344	/* ctx exit and cancelation */
	345	struct llist_head fallback_llist;
	346	struct delayed_work fallback_work;
	347	struct work_struct exit_work;
	348	struct list_head tctx_list;
	349	struct completion ref_comp;
	350
	351	/* io-wq management, e.g. thread count */
	352	u32 iowq_limits[2];
	353	bool iowq_limits_set;
	354
	355	struct list_head defer_list;
	356	unsigned sq_thread_idle;
305bef98 PB	357	/* protected by ->completion_lock */
305bef98 PB	358	unsigned evfd_last_cq_tail;
e27f928e JA	359	};
	360
	361	enum {
	362	REQ_F_FIXED_FILE_BIT = IOSQE_FIXED_FILE_BIT,
	363	REQ_F_IO_DRAIN_BIT = IOSQE_IO_DRAIN_BIT,
	364	REQ_F_LINK_BIT = IOSQE_IO_LINK_BIT,
	365	REQ_F_HARDLINK_BIT = IOSQE_IO_HARDLINK_BIT,
	366	REQ_F_FORCE_ASYNC_BIT = IOSQE_ASYNC_BIT,
	367	REQ_F_BUFFER_SELECT_BIT = IOSQE_BUFFER_SELECT_BIT,
	368	REQ_F_CQE_SKIP_BIT = IOSQE_CQE_SKIP_SUCCESS_BIT,
	369
	370	/* first byte is taken by user flags, shift it to not overlap */
	371	REQ_F_FAIL_BIT = 8,
	372	REQ_F_INFLIGHT_BIT,
	373	REQ_F_CUR_POS_BIT,
	374	REQ_F_NOWAIT_BIT,
	375	REQ_F_LINK_TIMEOUT_BIT,
	376	REQ_F_NEED_CLEANUP_BIT,
	377	REQ_F_POLLED_BIT,
	378	REQ_F_BUFFER_SELECTED_BIT,
	379	REQ_F_BUFFER_RING_BIT,
e27f928e JA	380	REQ_F_REISSUE_BIT,
	381	REQ_F_CREDS_BIT,
	382	REQ_F_REFCOUNT_BIT,
	383	REQ_F_ARM_LTIMEOUT_BIT,
	384	REQ_F_ASYNC_DATA_BIT,
	385	REQ_F_SKIP_LINK_CQES_BIT,
	386	REQ_F_SINGLE_POLL_BIT,
	387	REQ_F_DOUBLE_POLL_BIT,
	388	REQ_F_PARTIAL_IO_BIT,
	389	REQ_F_CQE32_INIT_BIT,
	390	REQ_F_APOLL_MULTISHOT_BIT,
	391	REQ_F_CLEAR_POLLIN_BIT,
9ca9fb24	392	REQ_F_HASH_LOCKED_BIT,
e27f928e JA	393	/* keep async read/write and isreg together and in order */
	394	REQ_F_SUPPORT_NOWAIT_BIT,
	395	REQ_F_ISREG_BIT,
	396
	397	/* not a real bit, just to check we're not overflowing the space */
	398	__REQ_F_LAST_BIT,
	399	};
	400
	401	enum {
	402	/* ctx owns file */
	403	REQ_F_FIXED_FILE = BIT(REQ_F_FIXED_FILE_BIT),
	404	/* drain existing IO first */
	405	REQ_F_IO_DRAIN = BIT(REQ_F_IO_DRAIN_BIT),
	406	/* linked sqes */
	407	REQ_F_LINK = BIT(REQ_F_LINK_BIT),
	408	/* doesn't sever on completion < 0 */
	409	REQ_F_HARDLINK = BIT(REQ_F_HARDLINK_BIT),
	410	/* IOSQE_ASYNC */
	411	REQ_F_FORCE_ASYNC = BIT(REQ_F_FORCE_ASYNC_BIT),
	412	/* IOSQE_BUFFER_SELECT */
	413	REQ_F_BUFFER_SELECT = BIT(REQ_F_BUFFER_SELECT_BIT),
	414	/* IOSQE_CQE_SKIP_SUCCESS */
	415	REQ_F_CQE_SKIP = BIT(REQ_F_CQE_SKIP_BIT),
	416
	417	/* fail rest of links */
	418	REQ_F_FAIL = BIT(REQ_F_FAIL_BIT),
	419	/* on inflight list, should be cancelled and waited on exit reliably */
	420	REQ_F_INFLIGHT = BIT(REQ_F_INFLIGHT_BIT),
	421	/* read/write uses file position */
	422	REQ_F_CUR_POS = BIT(REQ_F_CUR_POS_BIT),
	423	/* must not punt to workers */
	424	REQ_F_NOWAIT = BIT(REQ_F_NOWAIT_BIT),
	425	/* has or had linked timeout */
	426	REQ_F_LINK_TIMEOUT = BIT(REQ_F_LINK_TIMEOUT_BIT),
	427	/* needs cleanup */
	428	REQ_F_NEED_CLEANUP = BIT(REQ_F_NEED_CLEANUP_BIT),
	429	/* already went through poll handler */
	430	REQ_F_POLLED = BIT(REQ_F_POLLED_BIT),
	431	/* buffer already selected */
	432	REQ_F_BUFFER_SELECTED = BIT(REQ_F_BUFFER_SELECTED_BIT),
	433	/* buffer selected from ring, needs commit */
	434	REQ_F_BUFFER_RING = BIT(REQ_F_BUFFER_RING_BIT),
e27f928e JA	435	/* caller should reissue async */
	436	REQ_F_REISSUE = BIT(REQ_F_REISSUE_BIT),
	437	/* supports async reads/writes */
	438	REQ_F_SUPPORT_NOWAIT = BIT(REQ_F_SUPPORT_NOWAIT_BIT),
	439	/* regular file */
	440	REQ_F_ISREG = BIT(REQ_F_ISREG_BIT),
	441	/* has creds assigned */
	442	REQ_F_CREDS = BIT(REQ_F_CREDS_BIT),
	443	/* skip refcounting if not set */
	444	REQ_F_REFCOUNT = BIT(REQ_F_REFCOUNT_BIT),
	445	/* there is a linked timeout that has to be armed */
	446	REQ_F_ARM_LTIMEOUT = BIT(REQ_F_ARM_LTIMEOUT_BIT),
	447	/* ->async_data allocated */
	448	REQ_F_ASYNC_DATA = BIT(REQ_F_ASYNC_DATA_BIT),
	449	/* don't post CQEs while failing linked requests */
	450	REQ_F_SKIP_LINK_CQES = BIT(REQ_F_SKIP_LINK_CQES_BIT),
	451	/* single poll may be active */
	452	REQ_F_SINGLE_POLL = BIT(REQ_F_SINGLE_POLL_BIT),
	453	/* double poll may active */
	454	REQ_F_DOUBLE_POLL = BIT(REQ_F_DOUBLE_POLL_BIT),
	455	/* request has already done partial IO */
	456	REQ_F_PARTIAL_IO = BIT(REQ_F_PARTIAL_IO_BIT),
	457	/* fast poll multishot mode */
	458	REQ_F_APOLL_MULTISHOT = BIT(REQ_F_APOLL_MULTISHOT_BIT),
	459	/* ->extra1 and ->extra2 are initialised */
	460	REQ_F_CQE32_INIT = BIT(REQ_F_CQE32_INIT_BIT),
	461	/* recvmsg special flag, clear EPOLLIN */
	462	REQ_F_CLEAR_POLLIN = BIT(REQ_F_CLEAR_POLLIN_BIT),
9ca9fb24 PB	463	/* hashed into ->cancel_hash_locked, protected by ->uring_lock */
9ca9fb24 PB	464	REQ_F_HASH_LOCKED = BIT(REQ_F_HASH_LOCKED_BIT),
e27f928e JA	465	};
	466
	467	typedef void (io_req_tw_func_t)(struct io_kiocb req, bool *locked);
	468
	469	struct io_task_work {
3218e5d3	470	struct llist_node node;
e27f928e JA	471	io_req_tw_func_t func;
	472	};
	473
	474	struct io_cqe {
	475	__u64 user_data;
	476	__s32 res;
	477	/* fd initially, then cflags for completion */
	478	union {
	479	__u32 flags;
	480	int fd;
	481	};
	482	};
	483
	484	/*
	485	* Each request type overlays its private data structure on top of this one.
	486	* They must not exceed this one in size.
	487	*/
	488	struct io_cmd_data {
	489	struct file *file;
	490	/* each command gets 56 bytes of data */
	491	__u8 data[56];
	492	};
	493
f2ccb5ae SM	494	static inline void io_kiocb_cmd_sz_check(size_t cmd_sz)
	495	{
	496	BUILD_BUG_ON(cmd_sz > sizeof(struct io_cmd_data));
	497	}
	498	#define io_kiocb_to_cmd(req, cmd_type) ( \
	499	io_kiocb_cmd_sz_check(sizeof(cmd_type)) , \
	500	((cmd_type *)&(req)->cmd) \
	501	)
e27f928e JA	502	#define cmd_to_io_kiocb(ptr) ((struct io_kiocb *) ptr)
	503
	504	struct io_kiocb {
	505	union {
	506	/*
	507	* NOTE! Each of the io_kiocb union members has the file pointer
	508	* as the first entry in their struct definition. So you can
	509	* access the file pointer through any of the sub-structs,
	510	* or directly as just 'file' in this struct.
	511	*/
	512	struct file *file;
	513	struct io_cmd_data cmd;
	514	};
	515
	516	u8 opcode;
	517	/* polled IO has completed */
	518	u8 iopoll_completed;
	519	/*
	520	* Can be either a fixed buffer index, or used with provided buffers.
	521	* For the latter, before issue it points to the buffer group ID,
	522	* and after selection it points to the buffer ID itself.
	523	*/
	524	u16 buf_index;
	525	unsigned int flags;
	526
	527	struct io_cqe cqe;
	528
	529	struct io_ring_ctx *ctx;
	530	struct task_struct *task;
	531
	532	struct io_rsrc_node *rsrc_node;
	533
	534	union {
	535	/* store used ubuf, so we can prevent reloading */
	536	struct io_mapped_ubuf *imu;
	537
	538	/* stores selected buf, valid IFF REQ_F_BUFFER_SELECTED is set */
	539	struct io_buffer *kbuf;
	540
	541	/*
	542	* stores buffer ID for ring provided buffers, valid IFF
	543	* REQ_F_BUFFER_RING is set.
	544	*/
	545	struct io_buffer_list *buf_list;
	546	};
	547
	548	union {
	549	/* used by request caches, completion batching and iopoll */
	550	struct io_wq_work_node comp_list;
	551	/* cache ->apoll->events */
	552	__poll_t apoll_events;
	553	};
	554	atomic_t refs;
	555	atomic_t poll_refs;
	556	struct io_task_work io_task_work;
	557	/* for polled requests, i.e. IORING_OP_POLL_ADD and async armed poll */
	558	union {
	559	struct hlist_node hash_node;
	560	struct {
	561	u64 extra1;
	562	u64 extra2;
	563	};
	564	};
	565	/* internal polling, see IORING_FEAT_FAST_POLL */
566	struct async_poll *apoll;
567	/* opcode allocated if it needs to store data for async defer */
568	void *async_data;
569	/* linked requests, IFF REQ_F_HARDLINK or REQ_F_LINK are set */
570	struct io_kiocb *link;
571	/* custom credentials, valid IFF REQ_F_CREDS is set */
572	const struct cred *creds;
573	struct io_wq_work work;
574	};
575
a4ad4f74 JA	576	struct io_overflow_cqe {
	577	struct list_head list;
	578	struct io_uring_cqe cqe;
	579	};
	580
e27f928e	581	#endif