[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_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_cancel;
	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;
	u32 tail;
};

/*
 * 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;
	unsigned int		cqes_count;
	struct blk_plug		plug;
};

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

struct io_alloc_cache {
	struct io_wq_work_node	list;
	unsigned int		nr_cached;
	unsigned int		max_cached;
	size_t			elem_size;
};

struct io_ring_ctx {
	/* const or read-mostly hot data */
	struct {
		unsigned int		flags;
		unsigned int		drain_next: 1;
		unsigned int		restricted: 1;
		unsigned int		off_timeout_used: 1;
		unsigned int		drain_active: 1;
		unsigned int		has_evfd: 1;
		/* all CQEs should be posted only by the submitter task */
		unsigned int		task_complete: 1;
		unsigned int		lockless_cq: 1;
		unsigned int		syscall_iopoll: 1;
		unsigned int		poll_activated: 1;
		unsigned int		drain_disabled: 1;
		unsigned int		compat: 1;

		struct task_struct	*submitter_task;
		struct io_rings		*rings;
		struct percpu_ref	refs;

		enum task_work_notify_mode	notify_method;
	} ____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;
		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 io_hash_table	cancel_table_locked;
		struct io_alloc_cache	apoll_cache;
		struct io_alloc_cache	netmsg_cache;
	} ____cacheline_aligned_in_smp;

	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;
		unsigned		cq_extra;
	} ____cacheline_aligned_in_smp;

	/*
	 * task_work and async notification delivery cacheline. Expected to
	 * regularly bounce b/w CPUs.
	 */
	struct {
		struct llist_head	work_llist;
		unsigned long		check_cq;
		atomic_t		cq_wait_nr;
		atomic_t		cq_timeouts;
		struct wait_queue_head	cq_wait;
	} ____cacheline_aligned_in_smp;

	struct {
		spinlock_t		completion_lock;

		bool			poll_multi_queue;

		/*
		 * ->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;
	} ____cacheline_aligned_in_smp;

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

	struct io_uring_cqe	completion_cqes[16];

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

	struct list_head	io_buffers_comp;
	struct list_head	cq_overflow_list;
	struct io_hash_table	cancel_table;

	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 int		file_alloc_start;
	unsigned int		file_alloc_end;

	struct xarray		personalities;
	u32			pers_next;

	struct list_head	io_buffers_cache;

	/* Keep this last, we don't need it for the fast path */
	struct wait_queue_head		poll_wq;
	struct io_restriction		restrictions;

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

	/* protected by ->uring_lock */
	struct list_head		rsrc_ref_list;
	struct io_alloc_cache		rsrc_node_cache;
	struct wait_queue_head		rsrc_quiesce_wq;
	unsigned			rsrc_quiesce;

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

	/*
	 * If IORING_SETUP_NO_MMAP is used, then the below holds
	 * the gup'ed pages for the two rings, and the sqes.
	 */
	unsigned short			n_ring_pages;
	unsigned short			n_sqe_pages;
	struct page			**ring_pages;
	struct page			**sqe_pages;
};

struct io_tw_state {
	/* ->uring_lock is taken, callbacks can use io_tw_lock to lock it */
	bool locked;
};

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_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),
	/* 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, struct io_tw_state *ts);

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;
	unsigned			nr_tw;
	/* for polled requests, i.e. IORING_OP_POLL_ADD and async armed poll */
	struct hlist_node		hash_node;
	/* 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 {
		u64			extra1;
		u64			extra2;
	} big_cqe;
};

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
e6f89be6 PB	37	struct io_hash_bucket {
	38	spinlock_t lock;
	39	struct hlist_head list;
	40	} ____cacheline_aligned_in_smp;
	41
	42	struct io_hash_table {
	43	struct io_hash_bucket *hbs;
	44	unsigned hash_bits;
	45	};
	46
e70cb608 PB	47	/*
	48	* Arbitrary limit, can be raised if need be
	49	*/
	50	#define IO_RINGFD_REG_MAX 16
	51
	52	struct io_uring_task {
	53	/* submission side */
	54	int cached_refs;
	55	const struct io_ring_ctx *last;
	56	struct io_wq *io_wq;
	57	struct file *registered_rings[IO_RINGFD_REG_MAX];
	58
	59	struct xarray xa;
	60	struct wait_queue_head wait;
8d664282	61	atomic_t in_cancel;
e70cb608 PB	62	atomic_t inflight_tracked;
	63	struct percpu_counter inflight;
	64
	65	struct { /* task_work */
	66	struct llist_head task_list;
	67	struct callback_head task_work;
	68	} ____cacheline_aligned_in_smp;
	69	};
	70
e27f928e	71	struct io_uring {
e5598d6a PB	72	u32 head;
e5598d6a PB	73	u32 tail;
e27f928e JA	74	};
	75
	76	/*
	77	* This data is shared with the application through the mmap at offsets
	78	* IORING_OFF_SQ_RING and IORING_OFF_CQ_RING.
	79	*
	80	* The offsets to the member fields are published through struct
	81	* io_sqring_offsets when calling io_uring_setup.
	82	*/
	83	struct io_rings {
	84	/*
	85	* Head and tail offsets into the ring; the offsets need to be
	86	* masked to get valid indices.
	87	*
	88	* The kernel controls head of the sq ring and the tail of the cq ring,
	89	* and the application controls tail of the sq ring and the head of the
	90	* cq ring.
	91	*/
	92	struct io_uring sq, cq;
	93	/*
	94	* Bitmasks to apply to head and tail offsets (constant, equals
	95	* ring_entries - 1)
	96	*/
	97	u32 sq_ring_mask, cq_ring_mask;
	98	/* Ring sizes (constant, power of 2) */
	99	u32 sq_ring_entries, cq_ring_entries;
	100	/*
	101	* Number of invalid entries dropped by the kernel due to
	102	* invalid index stored in array
	103	*
	104	* Written by the kernel, shouldn't be modified by the
	105	* application (i.e. get number of "new events" by comparing to
	106	* cached value).
	107	*
	108	* After a new SQ head value was read by the application this
	109	* counter includes all submissions that were dropped reaching
	110	* the new SQ head (and possibly more).
	111	*/
	112	u32 sq_dropped;
	113	/*
	114	* Runtime SQ flags
	115	*
	116	* Written by the kernel, shouldn't be modified by the
	117	* application.
	118	*
	119	* The application needs a full memory barrier before checking
	120	* for IORING_SQ_NEED_WAKEUP after updating the sq tail.
	121	*/
	122	atomic_t sq_flags;
	123	/*
	124	* Runtime CQ flags
	125	*
	126	* Written by the application, shouldn't be modified by the
	127	* kernel.
	128	*/
	129	u32 cq_flags;
	130	/*
	131	* Number of completion events lost because the queue was full;
	132	* this should be avoided by the application by making sure
	133	* there are not more requests pending than there is space in
	134	* the completion queue.
	135	*
	136	* Written by the kernel, shouldn't be modified by the
	137	* application (i.e. get number of "new events" by comparing to
138	* cached value).
139	*
140	* As completion events come in out of order this counter is not
141	* ordered with any other data.
142	*/
143	u32 cq_overflow;
144	/*
145	* Ring buffer of completion events.
146	*
147	* The kernel writes completion events fresh every time they are
148	* produced, so the application is allowed to modify pending
149	* entries.
150	*/
151	struct io_uring_cqe cqes[] ____cacheline_aligned_in_smp;
152	};
153
154	struct io_restriction {
155	DECLARE_BITMAP(register_op, IORING_REGISTER_LAST);
156	DECLARE_BITMAP(sqe_op, IORING_OP_LAST);
157	u8 sqe_flags_allowed;
158	u8 sqe_flags_required;
159	bool registered;
160	};
161
162	struct io_submit_link {
163	struct io_kiocb *head;
164	struct io_kiocb *last;
165	};
166
167	struct io_submit_state {
168	/* inline/task_work completion list, under ->uring_lock */
169	struct io_wq_work_node free_list;
170	/* batch completion logic */
171	struct io_wq_work_list compl_reqs;
172	struct io_submit_link link;
173
174	bool plug_started;
175	bool need_plug;
e27f928e	176	unsigned short submit_nr;
931147dd	177	unsigned int cqes_count;
e27f928e JA	178	struct blk_plug plug;
	179	};
	180
	181	struct io_ev_fd {
	182	struct eventfd_ctx *cq_ev_fd;
	183	unsigned int eventfd_async: 1;
	184	struct rcu_head rcu;
21a091b9 DY	185	atomic_t refs;
21a091b9 DY	186	atomic_t ops;
e27f928e JA	187	};
e27f928e JA	188
9b797a37	189	struct io_alloc_cache {
efba1a9e	190	struct io_wq_work_node list;
9731bc98	191	unsigned int nr_cached;
69bbc6ad	192	unsigned int max_cached;
e1fe7ee8	193	size_t elem_size;
9b797a37 JA	194	};
9b797a37 JA	195
e27f928e JA	196	struct io_ring_ctx {
	197	/* const or read-mostly hot data */
	198	struct {
e27f928e	199	unsigned int flags;
e27f928e JA	200	unsigned int drain_next: 1;
	201	unsigned int restricted: 1;
	202	unsigned int off_timeout_used: 1;
	203	unsigned int drain_active: 1;
e27f928e	204	unsigned int has_evfd: 1;
e6aeb272 PB	205	/* all CQEs should be posted only by the submitter task */
e6aeb272 PB	206	unsigned int task_complete: 1;
ec26c225	207	unsigned int lockless_cq: 1;
632ffe09	208	unsigned int syscall_iopoll: 1;
bca39f39	209	unsigned int poll_activated: 1;
632ffe09 PB	210	unsigned int drain_disabled: 1;
632ffe09 PB	211	unsigned int compat: 1;
dde40322	212
18df385f PB	213	struct task_struct *submitter_task;
	214	struct io_rings *rings;
	215	struct percpu_ref refs;
03d89a2d	216
18df385f	217	enum task_work_notify_mode notify_method;
e27f928e JA	218	} ____cacheline_aligned_in_smp;
	219
	220	/* submission data */
	221	struct {
	222	struct mutex uring_lock;
	223
	224	/*
	225	* Ring buffer of indices into array of io_uring_sqe, which is
	226	* mmapped by the application using the IORING_OFF_SQES offset.
	227	*
	228	* This indirection could e.g. be used to assign fixed
	229	* io_uring_sqe entries to operations and only submit them to
	230	* the queue when needed.
	231	*
	232	* The kernel modifies neither the indices array nor the entries
	233	* array.
	234	*/
	235	u32 *sq_array;
	236	struct io_uring_sqe *sq_sqes;
	237	unsigned cached_sq_head;
	238	unsigned sq_entries;
e27f928e JA	239
	240	/*
	241	* Fixed resources fast path, should be accessed only under
	242	* uring_lock, and updated through io_uring_register(2)
	243	*/
	244	struct io_rsrc_node *rsrc_node;
e27f928e JA	245	atomic_t cancel_seq;
	246	struct io_file_table file_table;
	247	unsigned nr_user_files;
	248	unsigned nr_user_bufs;
	249	struct io_mapped_ubuf **user_bufs;
	250
	251	struct io_submit_state submit_state;
	252
	253	struct io_buffer_list *io_bl;
	254	struct xarray io_bl_xa;
e27f928e	255
9ca9fb24	256	struct io_hash_table cancel_table_locked;
9b797a37	257	struct io_alloc_cache apoll_cache;
43e0bbbd	258	struct io_alloc_cache netmsg_cache;
e27f928e JA	259	} ____cacheline_aligned_in_smp;
e27f928e JA	260
e27f928e JA	261	struct {
	262	/*
	263	* We cache a range of free CQEs we can use, once exhausted it
	264	* should go through a slower range setup, see __io_get_cqe()
	265	*/
	266	struct io_uring_cqe *cqe_cached;
	267	struct io_uring_cqe *cqe_sentinel;
	268
	269	unsigned cached_cq_tail;
	270	unsigned cq_entries;
	271	struct io_ev_fd __rcu *io_ev_fd;
e27f928e	272	unsigned cq_extra;
e27f928e JA	273	} ____cacheline_aligned_in_smp;
e27f928e JA	274
c9def23d PB	275	/*
	276	* task_work and async notification delivery cacheline. Expected to
	277	* regularly bounce b/w CPUs.
	278	*/
	279	struct {
	280	struct llist_head work_llist;
	281	unsigned long check_cq;
	282	atomic_t cq_wait_nr;
	283	atomic_t cq_timeouts;
	284	struct wait_queue_head cq_wait;
	285	} ____cacheline_aligned_in_smp;
	286
e27f928e JA	287	struct {
	288	spinlock_t completion_lock;
	289
59b745bb JA	290	bool poll_multi_queue;
59b745bb JA	291
e27f928e JA	292	/*
	293	* ->iopoll_list is protected by the ctx->uring_lock for
	294	* io_uring instances that don't use IORING_SETUP_SQPOLL.
	295	* For SQPOLL, only the single threaded io_sq_thread() will
	296	* manipulate the list, hence no extra locking is needed there.
	297	*/
	298	struct io_wq_work_list iopoll_list;
e27f928e JA	299	} ____cacheline_aligned_in_smp;
e27f928e JA	300
aff5b2df PB	301	/* timeouts */
	302	struct {
	303	spinlock_t timeout_lock;
aff5b2df PB	304	struct list_head timeout_list;
	305	struct list_head ltimeout_list;
	306	unsigned cq_last_tm_flush;
	307	} ____cacheline_aligned_in_smp;
	308
0aa7aa5f PB	309	struct io_uring_cqe completion_cqes[16];
0aa7aa5f PB	310
d7f06fea PB	311	/* IRQ completion list, under ->completion_lock */
	312	struct io_wq_work_list locked_free_list;
	313	unsigned int locked_free_nr;
	314
18df385f PB	315	struct list_head io_buffers_comp;
	316	struct list_head cq_overflow_list;
	317	struct io_hash_table cancel_table;
	318
d7f06fea PB	319	const struct cred sq_creds; / cred used for __io_sq_thread() */
	320	struct io_sq_data sq_data; / if using sq thread polling */
	321
	322	struct wait_queue_head sqo_sq_wait;
	323	struct list_head sqd_list;
	324
d7f06fea PB	325	unsigned int file_alloc_start;
	326	unsigned int file_alloc_end;
	327
	328	struct xarray personalities;
	329	u32 pers_next;
	330
18df385f PB	331	struct list_head io_buffers_cache;
18df385f PB	332
e27f928e	333	/* Keep this last, we don't need it for the fast path */
7b235dd8	334	struct wait_queue_head poll_wq;
22eb2a3f PB	335	struct io_restriction restrictions;
	336
	337	/* slow path rsrc auxilary data, used by update/register */
22eb2a3f PB	338	struct io_mapped_ubuf *dummy_ubuf;
	339	struct io_rsrc_data *file_data;
	340	struct io_rsrc_data *buf_data;
	341
0a4813b1	342	/* protected by ->uring_lock */
22eb2a3f	343	struct list_head rsrc_ref_list;
9eae8655	344	struct io_alloc_cache rsrc_node_cache;
4ea15b56	345	struct wait_queue_head rsrc_quiesce_wq;
0b222eeb	346	unsigned rsrc_quiesce;
22eb2a3f PB	347
	348	struct list_head io_buffers_pages;
	349
	350	#if defined(CONFIG_UNIX)
	351	struct socket *ring_sock;
	352	#endif
	353	/* hashed buffered write serialization */
	354	struct io_wq_hash *hash_map;
	355
	356	/* Only used for accounting purposes */
	357	struct user_struct *user;
	358	struct mm_struct *mm_account;
	359
	360	/* ctx exit and cancelation */
	361	struct llist_head fallback_llist;
	362	struct delayed_work fallback_work;
	363	struct work_struct exit_work;
	364	struct list_head tctx_list;
	365	struct completion ref_comp;
	366
	367	/* io-wq management, e.g. thread count */
	368	u32 iowq_limits[2];
	369	bool iowq_limits_set;
	370
bca39f39	371	struct callback_head poll_wq_task_work;
22eb2a3f PB	372	struct list_head defer_list;
22eb2a3f PB	373	unsigned sq_thread_idle;
305bef98 PB	374	/* protected by ->completion_lock */
305bef98 PB	375	unsigned evfd_last_cq_tail;
18df385f PB	376
	377	/*
	378	* If IORING_SETUP_NO_MMAP is used, then the below holds
	379	* the gup'ed pages for the two rings, and the sqes.
	380	*/
	381	unsigned short n_ring_pages;
	382	unsigned short n_sqe_pages;
	383	struct page **ring_pages;
	384	struct page **sqe_pages;
e27f928e JA	385	};
e27f928e JA	386
a282967c PB	387	struct io_tw_state {
	388	/* ->uring_lock is taken, callbacks can use io_tw_lock to lock it */
	389	bool locked;
	390	};
	391
e27f928e JA	392	enum {
	393	REQ_F_FIXED_FILE_BIT = IOSQE_FIXED_FILE_BIT,
	394	REQ_F_IO_DRAIN_BIT = IOSQE_IO_DRAIN_BIT,
	395	REQ_F_LINK_BIT = IOSQE_IO_LINK_BIT,
	396	REQ_F_HARDLINK_BIT = IOSQE_IO_HARDLINK_BIT,
	397	REQ_F_FORCE_ASYNC_BIT = IOSQE_ASYNC_BIT,
	398	REQ_F_BUFFER_SELECT_BIT = IOSQE_BUFFER_SELECT_BIT,
	399	REQ_F_CQE_SKIP_BIT = IOSQE_CQE_SKIP_SUCCESS_BIT,
	400
	401	/* first byte is taken by user flags, shift it to not overlap */
	402	REQ_F_FAIL_BIT = 8,
	403	REQ_F_INFLIGHT_BIT,
	404	REQ_F_CUR_POS_BIT,
	405	REQ_F_NOWAIT_BIT,
	406	REQ_F_LINK_TIMEOUT_BIT,
	407	REQ_F_NEED_CLEANUP_BIT,
	408	REQ_F_POLLED_BIT,
	409	REQ_F_BUFFER_SELECTED_BIT,
	410	REQ_F_BUFFER_RING_BIT,
e27f928e JA	411	REQ_F_REISSUE_BIT,
	412	REQ_F_CREDS_BIT,
	413	REQ_F_REFCOUNT_BIT,
	414	REQ_F_ARM_LTIMEOUT_BIT,
	415	REQ_F_ASYNC_DATA_BIT,
	416	REQ_F_SKIP_LINK_CQES_BIT,
	417	REQ_F_SINGLE_POLL_BIT,
	418	REQ_F_DOUBLE_POLL_BIT,
	419	REQ_F_PARTIAL_IO_BIT,
e27f928e JA	420	REQ_F_APOLL_MULTISHOT_BIT,
e27f928e JA	421	REQ_F_CLEAR_POLLIN_BIT,
9ca9fb24	422	REQ_F_HASH_LOCKED_BIT,
e27f928e JA	423	/* keep async read/write and isreg together and in order */
	424	REQ_F_SUPPORT_NOWAIT_BIT,
	425	REQ_F_ISREG_BIT,
	426
	427	/* not a real bit, just to check we're not overflowing the space */
	428	__REQ_F_LAST_BIT,
	429	};
	430
	431	enum {
	432	/* ctx owns file */
	433	REQ_F_FIXED_FILE = BIT(REQ_F_FIXED_FILE_BIT),
	434	/* drain existing IO first */
	435	REQ_F_IO_DRAIN = BIT(REQ_F_IO_DRAIN_BIT),
	436	/* linked sqes */
	437	REQ_F_LINK = BIT(REQ_F_LINK_BIT),
	438	/* doesn't sever on completion < 0 */
	439	REQ_F_HARDLINK = BIT(REQ_F_HARDLINK_BIT),
	440	/* IOSQE_ASYNC */
	441	REQ_F_FORCE_ASYNC = BIT(REQ_F_FORCE_ASYNC_BIT),
	442	/* IOSQE_BUFFER_SELECT */
	443	REQ_F_BUFFER_SELECT = BIT(REQ_F_BUFFER_SELECT_BIT),
	444	/* IOSQE_CQE_SKIP_SUCCESS */
	445	REQ_F_CQE_SKIP = BIT(REQ_F_CQE_SKIP_BIT),
	446
	447	/* fail rest of links */
	448	REQ_F_FAIL = BIT(REQ_F_FAIL_BIT),
	449	/* on inflight list, should be cancelled and waited on exit reliably */
	450	REQ_F_INFLIGHT = BIT(REQ_F_INFLIGHT_BIT),
	451	/* read/write uses file position */
	452	REQ_F_CUR_POS = BIT(REQ_F_CUR_POS_BIT),
	453	/* must not punt to workers */
	454	REQ_F_NOWAIT = BIT(REQ_F_NOWAIT_BIT),
	455	/* has or had linked timeout */
	456	REQ_F_LINK_TIMEOUT = BIT(REQ_F_LINK_TIMEOUT_BIT),
	457	/* needs cleanup */
	458	REQ_F_NEED_CLEANUP = BIT(REQ_F_NEED_CLEANUP_BIT),
	459	/* already went through poll handler */
	460	REQ_F_POLLED = BIT(REQ_F_POLLED_BIT),
	461	/* buffer already selected */
	462	REQ_F_BUFFER_SELECTED = BIT(REQ_F_BUFFER_SELECTED_BIT),
	463	/* buffer selected from ring, needs commit */
	464	REQ_F_BUFFER_RING = BIT(REQ_F_BUFFER_RING_BIT),
e27f928e JA	465	/* caller should reissue async */
	466	REQ_F_REISSUE = BIT(REQ_F_REISSUE_BIT),
	467	/* supports async reads/writes */
	468	REQ_F_SUPPORT_NOWAIT = BIT(REQ_F_SUPPORT_NOWAIT_BIT),
	469	/* regular file */
	470	REQ_F_ISREG = BIT(REQ_F_ISREG_BIT),
	471	/* has creds assigned */
	472	REQ_F_CREDS = BIT(REQ_F_CREDS_BIT),
	473	/* skip refcounting if not set */
	474	REQ_F_REFCOUNT = BIT(REQ_F_REFCOUNT_BIT),
	475	/* there is a linked timeout that has to be armed */
	476	REQ_F_ARM_LTIMEOUT = BIT(REQ_F_ARM_LTIMEOUT_BIT),
	477	/* ->async_data allocated */
	478	REQ_F_ASYNC_DATA = BIT(REQ_F_ASYNC_DATA_BIT),
	479	/* don't post CQEs while failing linked requests */
	480	REQ_F_SKIP_LINK_CQES = BIT(REQ_F_SKIP_LINK_CQES_BIT),
	481	/* single poll may be active */
	482	REQ_F_SINGLE_POLL = BIT(REQ_F_SINGLE_POLL_BIT),
	483	/* double poll may active */
	484	REQ_F_DOUBLE_POLL = BIT(REQ_F_DOUBLE_POLL_BIT),
	485	/* request has already done partial IO */
	486	REQ_F_PARTIAL_IO = BIT(REQ_F_PARTIAL_IO_BIT),
	487	/* fast poll multishot mode */
	488	REQ_F_APOLL_MULTISHOT = BIT(REQ_F_APOLL_MULTISHOT_BIT),
e27f928e JA	489	/* recvmsg special flag, clear EPOLLIN */
e27f928e JA	490	REQ_F_CLEAR_POLLIN = BIT(REQ_F_CLEAR_POLLIN_BIT),
9ca9fb24 PB	491	/* hashed into ->cancel_hash_locked, protected by ->uring_lock */
9ca9fb24 PB	492	REQ_F_HASH_LOCKED = BIT(REQ_F_HASH_LOCKED_BIT),
e27f928e JA	493	};
e27f928e JA	494
a282967c	495	typedef void (io_req_tw_func_t)(struct io_kiocb req, struct io_tw_state *ts);
e27f928e JA	496
e27f928e JA	497	struct io_task_work {
3218e5d3	498	struct llist_node node;
e27f928e JA	499	io_req_tw_func_t func;
	500	};
	501
	502	struct io_cqe {
	503	__u64 user_data;
	504	__s32 res;
	505	/* fd initially, then cflags for completion */
	506	union {
	507	__u32 flags;
	508	int fd;
	509	};
	510	};
	511
	512	/*
	513	* Each request type overlays its private data structure on top of this one.
	514	* They must not exceed this one in size.
	515	*/
	516	struct io_cmd_data {
	517	struct file *file;
	518	/* each command gets 56 bytes of data */
	519	__u8 data[56];
	520	};
	521
f2ccb5ae SM	522	static inline void io_kiocb_cmd_sz_check(size_t cmd_sz)
	523	{
	524	BUILD_BUG_ON(cmd_sz > sizeof(struct io_cmd_data));
	525	}
	526	#define io_kiocb_to_cmd(req, cmd_type) ( \
	527	io_kiocb_cmd_sz_check(sizeof(cmd_type)) , \
	528	((cmd_type *)&(req)->cmd) \
	529	)
e27f928e JA	530	#define cmd_to_io_kiocb(ptr) ((struct io_kiocb *) ptr)
	531
	532	struct io_kiocb {
	533	union {
	534	/*
	535	* NOTE! Each of the io_kiocb union members has the file pointer
	536	* as the first entry in their struct definition. So you can
	537	* access the file pointer through any of the sub-structs,
	538	* or directly as just 'file' in this struct.
	539	*/
	540	struct file *file;
	541	struct io_cmd_data cmd;
	542	};
	543
	544	u8 opcode;
	545	/* polled IO has completed */
	546	u8 iopoll_completed;
	547	/*
	548	* Can be either a fixed buffer index, or used with provided buffers.
	549	* For the latter, before issue it points to the buffer group ID,
	550	* and after selection it points to the buffer ID itself.
	551	*/
	552	u16 buf_index;
	553	unsigned int flags;
	554
	555	struct io_cqe cqe;
	556
	557	struct io_ring_ctx *ctx;
	558	struct task_struct *task;
	559
	560	struct io_rsrc_node *rsrc_node;
	561
	562	union {
	563	/* store used ubuf, so we can prevent reloading */
	564	struct io_mapped_ubuf *imu;
	565
	566	/* stores selected buf, valid IFF REQ_F_BUFFER_SELECTED is set */
	567	struct io_buffer *kbuf;
	568
	569	/*
	570	* stores buffer ID for ring provided buffers, valid IFF
	571	* REQ_F_BUFFER_RING is set.
	572	*/
	573	struct io_buffer_list *buf_list;
	574	};
	575
	576	union {
	577	/* used by request caches, completion batching and iopoll */
	578	struct io_wq_work_node comp_list;
	579	/* cache ->apoll->events */
	580	__poll_t apoll_events;
	581	};
	582	atomic_t refs;
	583	atomic_t poll_refs;
	584	struct io_task_work io_task_work;
8751d154	585	unsigned nr_tw;
e27f928e	586	/* for polled requests, i.e. IORING_OP_POLL_ADD and async armed poll */
b24c5d75	587	struct hlist_node hash_node;
e27f928e JA	588	/* internal polling, see IORING_FEAT_FAST_POLL */
	589	struct async_poll *apoll;
	590	/* opcode allocated if it needs to store data for async defer */
	591	void *async_data;
	592	/* linked requests, IFF REQ_F_HARDLINK or REQ_F_LINK are set */
	593	struct io_kiocb *link;
	594	/* custom credentials, valid IFF REQ_F_CREDS is set */
	595	const struct cred *creds;
	596	struct io_wq_work work;
b24c5d75 PB	597
	598	struct {
	599	u64 extra1;
	600	u64 extra2;
	601	} big_cqe;
e27f928e JA	602	};
e27f928e JA	603
a4ad4f74 JA	604	struct io_overflow_cqe {
	605	struct list_head list;
	606	struct io_uring_cqe cqe;
	607	};
	608
e27f928e	609	#endif