[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 <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_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;
};

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;
};

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_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 xarray		personalities;
		u32			pers_next;
	} ____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 {
		/*
		 * 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];
};

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