[thirdparty/kernel/linux.git] / fs / orangefs / orangefs-kernel.h

/* SPDX-License-Identifier: GPL-2.0 */
/*
 * (C) 2001 Clemson University and The University of Chicago
 *
 * See COPYING in top-level directory.
 */

/*
 *  The ORANGEFS Linux kernel support allows ORANGEFS volumes to be mounted and
 *  accessed through the Linux VFS (i.e. using standard I/O system calls).
 *  This support is only needed on clients that wish to mount the file system.
 *
 */

/*
 *  Declarations and macros for the ORANGEFS Linux kernel support.
 */

#ifndef __ORANGEFSKERNEL_H
#define __ORANGEFSKERNEL_H

#include <linux/kernel.h>
#include <linux/moduleparam.h>
#include <linux/statfs.h>
#include <linux/backing-dev.h>
#include <linux/device.h>
#include <linux/mpage.h>
#include <linux/namei.h>
#include <linux/errno.h>
#include <linux/init.h>
#include <linux/module.h>
#include <linux/slab.h>
#include <linux/types.h>
#include <linux/fs.h>
#include <linux/vmalloc.h>

#include <linux/aio.h>
#include <linux/posix_acl.h>
#include <linux/posix_acl_xattr.h>
#include <linux/compat.h>
#include <linux/mount.h>
#include <linux/uaccess.h>
#include <linux/atomic.h>
#include <linux/uio.h>
#include <linux/sched/signal.h>
#include <linux/mm.h>
#include <linux/wait.h>
#include <linux/dcache.h>
#include <linux/pagemap.h>
#include <linux/poll.h>
#include <linux/rwsem.h>
#include <linux/xattr.h>
#include <linux/exportfs.h>

#include <asm/unaligned.h>

#include "orangefs-dev-proto.h"

#ifdef ORANGEFS_KERNEL_DEBUG
#define ORANGEFS_DEFAULT_OP_TIMEOUT_SECS       10
#else
#define ORANGEFS_DEFAULT_OP_TIMEOUT_SECS       20
#endif

#define ORANGEFS_BUFMAP_WAIT_TIMEOUT_SECS   30

#define ORANGEFS_DEFAULT_SLOT_TIMEOUT_SECS     900	/* 15 minutes */

#define ORANGEFS_REQDEVICE_NAME          "pvfs2-req"

#define ORANGEFS_DEVREQ_MAGIC             0x20030529
#define ORANGEFS_LINK_MAX                 0x000000FF
#define ORANGEFS_PURGE_RETRY_COUNT     0x00000005
#define ORANGEFS_MAX_NUM_OPTIONS          0x00000004
#define ORANGEFS_MAX_MOUNT_OPT_LEN        0x00000080
#define ORANGEFS_MAX_FSKEY_LEN            64

#define MAX_DEV_REQ_UPSIZE (2 * sizeof(__s32) +   \
sizeof(__u64) + sizeof(struct orangefs_upcall_s))
#define MAX_DEV_REQ_DOWNSIZE (2 * sizeof(__s32) + \
sizeof(__u64) + sizeof(struct orangefs_downcall_s))

/*
 * valid orangefs kernel operation states
 *
 * unknown  - op was just initialized
 * waiting  - op is on request_list (upward bound)
 * inprogr  - op is in progress (waiting for downcall)
 * serviced - op has matching downcall; ok
 * purged   - op has to start a timer since client-core
 *            exited uncleanly before servicing op
 * given up - submitter has given up waiting for it
 */
enum orangefs_vfs_op_states {
	OP_VFS_STATE_UNKNOWN = 0,
	OP_VFS_STATE_WAITING = 1,
	OP_VFS_STATE_INPROGR = 2,
	OP_VFS_STATE_SERVICED = 4,
	OP_VFS_STATE_PURGED = 8,
	OP_VFS_STATE_GIVEN_UP = 16,
};

/*
 * orangefs kernel memory related flags
 */

#if ((defined ORANGEFS_KERNEL_DEBUG) && (defined CONFIG_DEBUG_SLAB))
#define ORANGEFS_CACHE_CREATE_FLAGS SLAB_RED_ZONE
#else
#define ORANGEFS_CACHE_CREATE_FLAGS 0
#endif /* ((defined ORANGEFS_KERNEL_DEBUG) && (defined CONFIG_DEBUG_SLAB)) */

extern int orangefs_init_acl(struct inode *inode, struct inode *dir);
extern const struct xattr_handler *orangefs_xattr_handlers[];

extern struct posix_acl *orangefs_get_acl(struct inode *inode, int type);
extern int orangefs_set_acl(struct inode *inode, struct posix_acl *acl, int type);

/*
 * Redefine xtvec structure so that we could move helper functions out of
 * the define
 */
struct xtvec {
	__kernel_off_t xtv_off;		/* must be off_t */
	__kernel_size_t xtv_len;	/* must be size_t */
};

/*
 * orangefs data structures
 */
struct orangefs_kernel_op_s {
	enum orangefs_vfs_op_states op_state;
	__u64 tag;

	/*
	 * Set uses_shared_memory to non zero if this operation uses
	 * shared memory. If true, then a retry on the op must also
	 * get a new shared memory buffer and re-populate it.
	 * Cancels don't care - it only matters for service_operation()
	 * retry logics and cancels don't go through it anymore. It
	 * safely stays non-zero when we use it as slot_to_free.
	 */
	union {
		int uses_shared_memory;
		int slot_to_free;
	};

	struct orangefs_upcall_s upcall;
	struct orangefs_downcall_s downcall;

	struct completion waitq;
	spinlock_t lock;

	int attempts;

	struct list_head list;
};

#define set_op_state_waiting(op)     ((op)->op_state = OP_VFS_STATE_WAITING)
#define set_op_state_inprogress(op)  ((op)->op_state = OP_VFS_STATE_INPROGR)
#define set_op_state_given_up(op)  ((op)->op_state = OP_VFS_STATE_GIVEN_UP)
static inline void set_op_state_serviced(struct orangefs_kernel_op_s *op)
{
	op->op_state = OP_VFS_STATE_SERVICED;
	complete(&op->waitq);
}

#define op_state_waiting(op)     ((op)->op_state & OP_VFS_STATE_WAITING)
#define op_state_in_progress(op) ((op)->op_state & OP_VFS_STATE_INPROGR)
#define op_state_serviced(op)    ((op)->op_state & OP_VFS_STATE_SERVICED)
#define op_state_purged(op)      ((op)->op_state & OP_VFS_STATE_PURGED)
#define op_state_given_up(op)    ((op)->op_state & OP_VFS_STATE_GIVEN_UP)
#define op_is_cancel(op)         ((op)->upcall.type == ORANGEFS_VFS_OP_CANCEL)

void op_release(struct orangefs_kernel_op_s *op);

extern void orangefs_bufmap_put(int);
static inline void put_cancel(struct orangefs_kernel_op_s *op)
{
	orangefs_bufmap_put(op->slot_to_free);
	op_release(op);
}

static inline void set_op_state_purged(struct orangefs_kernel_op_s *op)
{
	spin_lock(&op->lock);
	if (unlikely(op_is_cancel(op))) {
		list_del_init(&op->list);
		spin_unlock(&op->lock);
		put_cancel(op);
	} else {
		op->op_state |= OP_VFS_STATE_PURGED;
		complete(&op->waitq);
		spin_unlock(&op->lock);
	}
}

/* per inode private orangefs info */
struct orangefs_inode_s {
	struct orangefs_object_kref refn;
	char link_target[ORANGEFS_NAME_MAX];
	__s64 blksize;
	/*
	 * Reading/Writing Extended attributes need to acquire the appropriate
	 * reader/writer semaphore on the orangefs_inode_s structure.
	 */
	struct rw_semaphore xattr_sem;

	struct inode vfs_inode;
	sector_t last_failed_block_index_read;

	/*
	 * State of in-memory attributes not yet flushed to disk associated
	 * with this object
	 */
	unsigned long pinode_flags;

	unsigned long getattr_time;
	u32 getattr_mask;
};

#define P_ATIME_FLAG 0
#define P_MTIME_FLAG 1
#define P_CTIME_FLAG 2
#define P_MODE_FLAG  3

#define ClearAtimeFlag(pinode) clear_bit(P_ATIME_FLAG, &(pinode)->pinode_flags)
#define SetAtimeFlag(pinode)   set_bit(P_ATIME_FLAG, &(pinode)->pinode_flags)
#define AtimeFlag(pinode)      test_bit(P_ATIME_FLAG, &(pinode)->pinode_flags)

#define ClearMtimeFlag(pinode) clear_bit(P_MTIME_FLAG, &(pinode)->pinode_flags)
#define SetMtimeFlag(pinode)   set_bit(P_MTIME_FLAG, &(pinode)->pinode_flags)
#define MtimeFlag(pinode)      test_bit(P_MTIME_FLAG, &(pinode)->pinode_flags)

#define ClearCtimeFlag(pinode) clear_bit(P_CTIME_FLAG, &(pinode)->pinode_flags)
#define SetCtimeFlag(pinode)   set_bit(P_CTIME_FLAG, &(pinode)->pinode_flags)
#define CtimeFlag(pinode)      test_bit(P_CTIME_FLAG, &(pinode)->pinode_flags)

#define ClearModeFlag(pinode) clear_bit(P_MODE_FLAG, &(pinode)->pinode_flags)
#define SetModeFlag(pinode)   set_bit(P_MODE_FLAG, &(pinode)->pinode_flags)
#define ModeFlag(pinode)      test_bit(P_MODE_FLAG, &(pinode)->pinode_flags)

/* per superblock private orangefs info */
struct orangefs_sb_info_s {
	struct orangefs_khandle root_khandle;
	__s32 fs_id;
	int id;
	int flags;
#define ORANGEFS_OPT_INTR	0x01
#define ORANGEFS_OPT_LOCAL_LOCK	0x02
	char devname[ORANGEFS_MAX_SERVER_ADDR_LEN];
	struct super_block *sb;
	int mount_pending;
	int no_list;
	struct list_head list;
};

/*
 * structure that holds the state of any async I/O operation issued
 * through the VFS. Needed especially to handle cancellation requests
 * or even completion notification so that the VFS client-side daemon
 * can free up its vfs_request slots.
 */
struct orangefs_kiocb_s {
	/* the pointer to the task that initiated the AIO */
	struct task_struct *tsk;

	/* pointer to the kiocb that kicked this operation */
	struct kiocb *kiocb;

	/* buffer index that was used for the I/O */
	struct orangefs_bufmap *bufmap;
	int buffer_index;

	/* orangefs kernel operation type */
	struct orangefs_kernel_op_s *op;

	/* The user space buffers from/to which I/O is being staged */
	struct iovec *iov;

	/* number of elements in the iovector */
	unsigned long nr_segs;

	/* set to indicate the type of the operation */
	int rw;

	/* file offset */
	loff_t offset;

	/* and the count in bytes */
	size_t bytes_to_be_copied;

	ssize_t bytes_copied;
	int needs_cleanup;
};

struct orangefs_stats {
	unsigned long cache_hits;
	unsigned long cache_misses;
	unsigned long reads;
	unsigned long writes;
};

extern struct orangefs_stats orangefs_stats;

/*
 * NOTE: See Documentation/filesystems/porting for information
 * on implementing FOO_I and properly accessing fs private data
 */
static inline struct orangefs_inode_s *ORANGEFS_I(struct inode *inode)
{
	return container_of(inode, struct orangefs_inode_s, vfs_inode);
}

static inline struct orangefs_sb_info_s *ORANGEFS_SB(struct super_block *sb)
{
	return (struct orangefs_sb_info_s *) sb->s_fs_info;
}

/* ino_t descends from "unsigned long", 8 bytes, 64 bits. */
static inline ino_t orangefs_khandle_to_ino(struct orangefs_khandle *khandle)
{
	union {
		unsigned char u[8];
		__u64 ino;
	} ihandle;

	ihandle.u[0] = khandle->u[0] ^ khandle->u[4];
	ihandle.u[1] = khandle->u[1] ^ khandle->u[5];
	ihandle.u[2] = khandle->u[2] ^ khandle->u[6];
	ihandle.u[3] = khandle->u[3] ^ khandle->u[7];
	ihandle.u[4] = khandle->u[12] ^ khandle->u[8];
	ihandle.u[5] = khandle->u[13] ^ khandle->u[9];
	ihandle.u[6] = khandle->u[14] ^ khandle->u[10];
	ihandle.u[7] = khandle->u[15] ^ khandle->u[11];

	return ihandle.ino;
}

static inline struct orangefs_khandle *get_khandle_from_ino(struct inode *inode)
{
	return &(ORANGEFS_I(inode)->refn.khandle);
}

static inline ino_t get_ino_from_khandle(struct inode *inode)
{
	struct orangefs_khandle *khandle;
	ino_t ino;

	khandle = get_khandle_from_ino(inode);
	ino = orangefs_khandle_to_ino(khandle);
	return ino;
}

static inline ino_t get_parent_ino_from_dentry(struct dentry *dentry)
{
	return get_ino_from_khandle(dentry->d_parent->d_inode);
}

static inline int is_root_handle(struct inode *inode)
{
	gossip_debug(GOSSIP_DCACHE_DEBUG,
		     "%s: root handle: %pU, this handle: %pU:\n",
		     __func__,
		     &ORANGEFS_SB(inode->i_sb)->root_khandle,
		     get_khandle_from_ino(inode));

	if (ORANGEFS_khandle_cmp(&(ORANGEFS_SB(inode->i_sb)->root_khandle),
			     get_khandle_from_ino(inode)))
		return 0;
	else
		return 1;
}

static inline int match_handle(struct orangefs_khandle resp_handle,
			       struct inode *inode)
{
	gossip_debug(GOSSIP_DCACHE_DEBUG,
		     "%s: one handle: %pU, another handle:%pU:\n",
		     __func__,
		     &resp_handle,
		     get_khandle_from_ino(inode));

	if (ORANGEFS_khandle_cmp(&resp_handle, get_khandle_from_ino(inode)))
		return 0;
	else
		return 1;
}

/*
 * defined in orangefs-cache.c
 */
int op_cache_initialize(void);
int op_cache_finalize(void);
struct orangefs_kernel_op_s *op_alloc(__s32 type);
void orangefs_new_tag(struct orangefs_kernel_op_s *op);
char *get_opname_string(struct orangefs_kernel_op_s *new_op);

int orangefs_inode_cache_initialize(void);
int orangefs_inode_cache_finalize(void);

/*
 * defined in orangefs-mod.c
 */
void purge_inprogress_ops(void);

/*
 * defined in waitqueue.c
 */
void purge_waiting_ops(void);

/*
 * defined in super.c
 */
extern uint64_t orangefs_features;

struct dentry *orangefs_mount(struct file_system_type *fst,
			   int flags,
			   const char *devname,
			   void *data);

void orangefs_kill_sb(struct super_block *sb);
int orangefs_remount(struct orangefs_sb_info_s *);

int fsid_key_table_initialize(void);
void fsid_key_table_finalize(void);

/*
 * defined in inode.c
 */
__u32 convert_to_orangefs_mask(unsigned long lite_mask);
struct inode *orangefs_new_inode(struct super_block *sb,
			      struct inode *dir,
			      int mode,
			      dev_t dev,
			      struct orangefs_object_kref *ref);

int orangefs_setattr(struct dentry *dentry, struct iattr *iattr);

int orangefs_getattr(const struct path *path, struct kstat *stat,
		     u32 request_mask, unsigned int flags);

int orangefs_permission(struct inode *inode, int mask);

/*
 * defined in xattr.c
 */
int orangefs_setxattr(struct dentry *dentry,
		   const char *name,
		   const void *value,
		   size_t size,
		   int flags);

ssize_t orangefs_getxattr(struct dentry *dentry,
		       const char *name,
		       void *buffer,
		       size_t size);

ssize_t orangefs_listxattr(struct dentry *dentry, char *buffer, size_t size);

/*
 * defined in namei.c
 */
struct inode *orangefs_iget(struct super_block *sb,
			 struct orangefs_object_kref *ref);

ssize_t orangefs_inode_read(struct inode *inode,
			    struct iov_iter *iter,
			    loff_t *offset,
			    loff_t readahead_size);

/*
 * defined in devorangefs-req.c
 */
extern uint32_t orangefs_userspace_version;

int orangefs_dev_init(void);
void orangefs_dev_cleanup(void);
int is_daemon_in_service(void);
bool __is_daemon_in_service(void);

/*
 * defined in orangefs-utils.c
 */
__s32 fsid_of_op(struct orangefs_kernel_op_s *op);

int orangefs_flush_inode(struct inode *inode);

ssize_t orangefs_inode_getxattr(struct inode *inode,
			     const char *name,
			     void *buffer,
			     size_t size);

int orangefs_inode_setxattr(struct inode *inode,
			 const char *name,
			 const void *value,
			 size_t size,
			 int flags);

int orangefs_inode_getattr(struct inode *inode, int new, int bypass,
    u32 request_mask);

int orangefs_inode_check_changed(struct inode *inode);

int orangefs_inode_setattr(struct inode *inode, struct iattr *iattr);

void orangefs_make_bad_inode(struct inode *inode);

int orangefs_unmount_sb(struct super_block *sb);

bool orangefs_cancel_op_in_progress(struct orangefs_kernel_op_s *op);

int orangefs_normalize_to_errno(__s32 error_code);

extern struct mutex orangefs_request_mutex;
extern int op_timeout_secs;
extern int slot_timeout_secs;
extern int orangefs_dcache_timeout_msecs;
extern int orangefs_getattr_timeout_msecs;
extern struct list_head orangefs_superblocks;
extern spinlock_t orangefs_superblocks_lock;
extern struct list_head orangefs_request_list;
extern spinlock_t orangefs_request_list_lock;
extern wait_queue_head_t orangefs_request_list_waitq;
extern struct list_head *orangefs_htable_ops_in_progress;
extern spinlock_t orangefs_htable_ops_in_progress_lock;
extern int hash_table_size;

extern const struct address_space_operations orangefs_address_operations;
extern const struct inode_operations orangefs_file_inode_operations;
extern const struct file_operations orangefs_file_operations;
extern const struct inode_operations orangefs_symlink_inode_operations;
extern const struct inode_operations orangefs_dir_inode_operations;
extern const struct file_operations orangefs_dir_operations;
extern const struct dentry_operations orangefs_dentry_operations;
extern const struct file_operations orangefs_devreq_file_operations;

extern wait_queue_head_t orangefs_bufmap_init_waitq;

/*
 * misc convenience macros
 */

#define ORANGEFS_OP_INTERRUPTIBLE 1   /* service_operation() is interruptible */
#define ORANGEFS_OP_PRIORITY      2   /* service_operation() is high priority */
#define ORANGEFS_OP_CANCELLATION  4   /* this is a cancellation */
#define ORANGEFS_OP_NO_MUTEX      8   /* don't acquire request_mutex */
#define ORANGEFS_OP_ASYNC         16  /* Queue it, but don't wait */

int service_operation(struct orangefs_kernel_op_s *op,
		      const char *op_name,
		      int flags);

#define get_interruptible_flag(inode) \
	((ORANGEFS_SB(inode->i_sb)->flags & ORANGEFS_OPT_INTR) ? \
		ORANGEFS_OP_INTERRUPTIBLE : 0)

#define fill_default_sys_attrs(sys_attr, type, mode)			\
do {									\
	sys_attr.owner = from_kuid(&init_user_ns, current_fsuid()); \
	sys_attr.group = from_kgid(&init_user_ns, current_fsgid()); \
	sys_attr.perms = ORANGEFS_util_translate_mode(mode);		\
	sys_attr.mtime = 0;						\
	sys_attr.atime = 0;						\
	sys_attr.ctime = 0;						\
	sys_attr.mask = ORANGEFS_ATTR_SYS_ALL_SETABLE;			\
} while (0)

static inline void orangefs_i_size_write(struct inode *inode, loff_t i_size)
{
#if BITS_PER_LONG == 32 && defined(CONFIG_SMP)
	inode_lock(inode);
#endif
	i_size_write(inode, i_size);
#if BITS_PER_LONG == 32 && defined(CONFIG_SMP)
	inode_unlock(inode);
#endif
}

static inline void orangefs_set_timeout(struct dentry *dentry)
{
	unsigned long time = jiffies + orangefs_dcache_timeout_msecs*HZ/1000;

	dentry->d_fsdata = (void *) time;
}

#endif /* __ORANGEFSKERNEL_H */
Commit	Line	Data
b2441318	1	/* SPDX-License-Identifier: GPL-2.0 */
f7ab093f MM	2	/*
	3	* (C) 2001 Clemson University and The University of Chicago
	4	*
	5	* See COPYING in top-level directory.
	6	*/
	7
	8	/*
8bb8aefd	9	* The ORANGEFS Linux kernel support allows ORANGEFS volumes to be mounted and
f7ab093f MM	10	* accessed through the Linux VFS (i.e. using standard I/O system calls).
	11	* This support is only needed on clients that wish to mount the file system.
	12	*
	13	*/
	14
	15	/*
8bb8aefd	16	* Declarations and macros for the ORANGEFS Linux kernel support.
f7ab093f MM	17	*/
f7ab093f MM	18
8bb8aefd YL	19	#ifndef __ORANGEFSKERNEL_H
8bb8aefd YL	20	#define __ORANGEFSKERNEL_H
f7ab093f MM	21
	22	#include <linux/kernel.h>
	23	#include <linux/moduleparam.h>
	24	#include <linux/statfs.h>
	25	#include <linux/backing-dev.h>
	26	#include <linux/device.h>
	27	#include <linux/mpage.h>
	28	#include <linux/namei.h>
	29	#include <linux/errno.h>
	30	#include <linux/init.h>
	31	#include <linux/module.h>
	32	#include <linux/slab.h>
	33	#include <linux/types.h>
	34	#include <linux/fs.h>
	35	#include <linux/vmalloc.h>
	36
	37	#include <linux/aio.h>
	38	#include <linux/posix_acl.h>
	39	#include <linux/posix_acl_xattr.h>
	40	#include <linux/compat.h>
	41	#include <linux/mount.h>
	42	#include <linux/uaccess.h>
	43	#include <linux/atomic.h>
	44	#include <linux/uio.h>
174cd4b1	45	#include <linux/sched/signal.h>
f7ab093f MM	46	#include <linux/mm.h>
	47	#include <linux/wait.h>
	48	#include <linux/dcache.h>
	49	#include <linux/pagemap.h>
	50	#include <linux/poll.h>
	51	#include <linux/rwsem.h>
	52	#include <linux/xattr.h>
	53	#include <linux/exportfs.h>
	54
	55	#include <asm/unaligned.h>
	56
575e9461	57	#include "orangefs-dev-proto.h"
f7ab093f	58
8bb8aefd YL	59	#ifdef ORANGEFS_KERNEL_DEBUG
8bb8aefd YL	60	#define ORANGEFS_DEFAULT_OP_TIMEOUT_SECS 10
f7ab093f	61	#else
8bb8aefd	62	#define ORANGEFS_DEFAULT_OP_TIMEOUT_SECS 20
f7ab093f MM	63	#endif
f7ab093f MM	64
8bb8aefd	65	#define ORANGEFS_BUFMAP_WAIT_TIMEOUT_SECS 30
f7ab093f	66
8bb8aefd	67	#define ORANGEFS_DEFAULT_SLOT_TIMEOUT_SECS 900 /* 15 minutes */
f7ab093f	68
8bb8aefd	69	#define ORANGEFS_REQDEVICE_NAME "pvfs2-req"
f7ab093f	70
8bb8aefd YL	71	#define ORANGEFS_DEVREQ_MAGIC 0x20030529
	72	#define ORANGEFS_LINK_MAX 0x000000FF
	73	#define ORANGEFS_PURGE_RETRY_COUNT 0x00000005
8bb8aefd YL	74	#define ORANGEFS_MAX_NUM_OPTIONS 0x00000004
	75	#define ORANGEFS_MAX_MOUNT_OPT_LEN 0x00000080
	76	#define ORANGEFS_MAX_FSKEY_LEN 64
f7ab093f	77
cf0c2771	78	#define MAX_DEV_REQ_UPSIZE (2 * sizeof(__s32) + \
8bb8aefd	79	sizeof(__u64) + sizeof(struct orangefs_upcall_s))
cf0c2771	80	#define MAX_DEV_REQ_DOWNSIZE (2 * sizeof(__s32) + \
8bb8aefd	81	sizeof(__u64) + sizeof(struct orangefs_downcall_s))
f7ab093f	82
f7ab093f	83	/*
8bb8aefd	84	* valid orangefs kernel operation states
f7ab093f MM	85	*
	86	* unknown - op was just initialized
	87	* waiting - op is on request_list (upward bound)
	88	* inprogr - op is in progress (waiting for downcall)
	89	* serviced - op has matching downcall; ok
	90	* purged - op has to start a timer since client-core
	91	* exited uncleanly before servicing op
ed42fe05	92	* given up - submitter has given up waiting for it
f7ab093f	93	*/
8bb8aefd	94	enum orangefs_vfs_op_states {
f7ab093f MM	95	OP_VFS_STATE_UNKNOWN = 0,
	96	OP_VFS_STATE_WAITING = 1,
	97	OP_VFS_STATE_INPROGR = 2,
	98	OP_VFS_STATE_SERVICED = 4,
	99	OP_VFS_STATE_PURGED = 8,
ed42fe05	100	OP_VFS_STATE_GIVEN_UP = 16,
f7ab093f MM	101	};
f7ab093f MM	102
f7ab093f	103	/*
8bb8aefd	104	* orangefs kernel memory related flags
f7ab093f MM	105	*/
f7ab093f MM	106
8bb8aefd YL	107	#if ((defined ORANGEFS_KERNEL_DEBUG) && (defined CONFIG_DEBUG_SLAB))
8bb8aefd YL	108	#define ORANGEFS_CACHE_CREATE_FLAGS SLAB_RED_ZONE
f7ab093f	109	#else
8bb8aefd YL	110	#define ORANGEFS_CACHE_CREATE_FLAGS 0
8bb8aefd YL	111	#endif /* ((defined ORANGEFS_KERNEL_DEBUG) && (defined CONFIG_DEBUG_SLAB)) */
f7ab093f	112
8bb8aefd YL	113	extern int orangefs_init_acl(struct inode inode, struct inode dir);
8bb8aefd YL	114	extern const struct xattr_handler *orangefs_xattr_handlers[];
f7ab093f	115
8bb8aefd YL	116	extern struct posix_acl orangefs_get_acl(struct inode inode, int type);
8bb8aefd YL	117	extern int orangefs_set_acl(struct inode inode, struct posix_acl acl, int type);
f7ab093f	118
f7ab093f MM	119	/*
	120	* Redefine xtvec structure so that we could move helper functions out of
	121	* the define
	122	*/
	123	struct xtvec {
	124	__kernel_off_t xtv_off; /* must be off_t */
	125	__kernel_size_t xtv_len; /* must be size_t */
	126	};
	127
	128	/*
8bb8aefd	129	* orangefs data structures
f7ab093f	130	*/
8bb8aefd YL	131	struct orangefs_kernel_op_s {
8bb8aefd YL	132	enum orangefs_vfs_op_states op_state;
f7ab093f MM	133	__u64 tag;
	134
	135	/*
d37c0f30 MM	136	* Set uses_shared_memory to non zero if this operation uses
	137	* shared memory. If true, then a retry on the op must also
	138	* get a new shared memory buffer and re-populate it.
	139	* Cancels don't care - it only matters for service_operation()
	140	* retry logics and cancels don't go through it anymore. It
	141	* safely stays non-zero when we use it as slot_to_free.
f7ab093f	142	*/
78699e29 AV	143	union {
	144	int uses_shared_memory;
	145	int slot_to_free;
	146	};
f7ab093f	147
8bb8aefd YL	148	struct orangefs_upcall_s upcall;
8bb8aefd YL	149	struct orangefs_downcall_s downcall;
f7ab093f	150
d2d87a3b	151	struct completion waitq;
f7ab093f MM	152	spinlock_t lock;
f7ab093f MM	153
f7ab093f MM	154	int attempts;
	155
	156	struct list_head list;
	157	};
	158
60831949 AV	159	#define set_op_state_waiting(op) ((op)->op_state = OP_VFS_STATE_WAITING)
60831949 AV	160	#define set_op_state_inprogress(op) ((op)->op_state = OP_VFS_STATE_INPROGR)
2d4cae0d	161	#define set_op_state_given_up(op) ((op)->op_state = OP_VFS_STATE_GIVEN_UP)
60831949 AV	162	static inline void set_op_state_serviced(struct orangefs_kernel_op_s *op)
	163	{
	164	op->op_state = OP_VFS_STATE_SERVICED;
d2d87a3b	165	complete(&op->waitq);
60831949	166	}
60831949 AV	167
	168	#define op_state_waiting(op) ((op)->op_state & OP_VFS_STATE_WAITING)
	169	#define op_state_in_progress(op) ((op)->op_state & OP_VFS_STATE_INPROGR)
	170	#define op_state_serviced(op) ((op)->op_state & OP_VFS_STATE_SERVICED)
	171	#define op_state_purged(op) ((op)->op_state & OP_VFS_STATE_PURGED)
ed42fe05	172	#define op_state_given_up(op) ((op)->op_state & OP_VFS_STATE_GIVEN_UP)
78699e29	173	#define op_is_cancel(op) ((op)->upcall.type == ORANGEFS_VFS_OP_CANCEL)
ed42fe05	174
c1223ca4	175	void op_release(struct orangefs_kernel_op_s *op);
60831949	176
78699e29 AV	177	extern void orangefs_bufmap_put(int);
	178	static inline void put_cancel(struct orangefs_kernel_op_s *op)
	179	{
	180	orangefs_bufmap_put(op->slot_to_free);
	181	op_release(op);
	182	}
	183
	184	static inline void set_op_state_purged(struct orangefs_kernel_op_s *op)
	185	{
	186	spin_lock(&op->lock);
	187	if (unlikely(op_is_cancel(op))) {
05a50a5b	188	list_del_init(&op->list);
78699e29 AV	189	spin_unlock(&op->lock);
	190	put_cancel(op);
	191	} else {
	192	op->op_state \|= OP_VFS_STATE_PURGED;
d2d87a3b	193	complete(&op->waitq);
78699e29 AV	194	spin_unlock(&op->lock);
	195	}
	196	}
	197
8bb8aefd YL	198	/* per inode private orangefs info */
	199	struct orangefs_inode_s {
	200	struct orangefs_object_kref refn;
	201	char link_target[ORANGEFS_NAME_MAX];
f7ab093f MM	202	__s64 blksize;
	203	/*
	204	* Reading/Writing Extended attributes need to acquire the appropriate
8bb8aefd	205	* reader/writer semaphore on the orangefs_inode_s structure.
f7ab093f MM	206	*/
	207	struct rw_semaphore xattr_sem;
	208
	209	struct inode vfs_inode;
	210	sector_t last_failed_block_index_read;
	211
	212	/*
	213	* State of in-memory attributes not yet flushed to disk associated
	214	* with this object
	215	*/
	216	unsigned long pinode_flags;
71680c18 MB	217
71680c18 MB	218	unsigned long getattr_time;
68a24a6c	219	u32 getattr_mask;
f7ab093f MM	220	};
	221
	222	#define P_ATIME_FLAG 0
	223	#define P_MTIME_FLAG 1
	224	#define P_CTIME_FLAG 2
	225	#define P_MODE_FLAG 3
	226
	227	#define ClearAtimeFlag(pinode) clear_bit(P_ATIME_FLAG, &(pinode)->pinode_flags)
	228	#define SetAtimeFlag(pinode) set_bit(P_ATIME_FLAG, &(pinode)->pinode_flags)
	229	#define AtimeFlag(pinode) test_bit(P_ATIME_FLAG, &(pinode)->pinode_flags)
	230
	231	#define ClearMtimeFlag(pinode) clear_bit(P_MTIME_FLAG, &(pinode)->pinode_flags)
	232	#define SetMtimeFlag(pinode) set_bit(P_MTIME_FLAG, &(pinode)->pinode_flags)
	233	#define MtimeFlag(pinode) test_bit(P_MTIME_FLAG, &(pinode)->pinode_flags)
	234
	235	#define ClearCtimeFlag(pinode) clear_bit(P_CTIME_FLAG, &(pinode)->pinode_flags)
	236	#define SetCtimeFlag(pinode) set_bit(P_CTIME_FLAG, &(pinode)->pinode_flags)
	237	#define CtimeFlag(pinode) test_bit(P_CTIME_FLAG, &(pinode)->pinode_flags)
	238
	239	#define ClearModeFlag(pinode) clear_bit(P_MODE_FLAG, &(pinode)->pinode_flags)
	240	#define SetModeFlag(pinode) set_bit(P_MODE_FLAG, &(pinode)->pinode_flags)
	241	#define ModeFlag(pinode) test_bit(P_MODE_FLAG, &(pinode)->pinode_flags)
	242
8bb8aefd YL	243	/* per superblock private orangefs info */
	244	struct orangefs_sb_info_s {
	245	struct orangefs_khandle root_khandle;
f7ab093f MM	246	__s32 fs_id;
	247	int id;
	248	int flags;
8bb8aefd YL	249	#define ORANGEFS_OPT_INTR 0x01
	250	#define ORANGEFS_OPT_LOCAL_LOCK 0x02
	251	char devname[ORANGEFS_MAX_SERVER_ADDR_LEN];
f7ab093f MM	252	struct super_block *sb;
f7ab093f MM	253	int mount_pending;
1ec1688c	254	int no_list;
f7ab093f MM	255	struct list_head list;
	256	};
	257
f7ab093f MM	258	/*
	259	* structure that holds the state of any async I/O operation issued
	260	* through the VFS. Needed especially to handle cancellation requests
	261	* or even completion notification so that the VFS client-side daemon
	262	* can free up its vfs_request slots.
	263	*/
8bb8aefd	264	struct orangefs_kiocb_s {
f7ab093f MM	265	/* the pointer to the task that initiated the AIO */
	266	struct task_struct *tsk;
	267
	268	/* pointer to the kiocb that kicked this operation */
	269	struct kiocb *kiocb;
	270
	271	/* buffer index that was used for the I/O */
8bb8aefd	272	struct orangefs_bufmap *bufmap;
f7ab093f MM	273	int buffer_index;
f7ab093f MM	274
8bb8aefd YL	275	/* orangefs kernel operation type */
8bb8aefd YL	276	struct orangefs_kernel_op_s *op;
f7ab093f MM	277
	278	/* The user space buffers from/to which I/O is being staged */
	279	struct iovec *iov;
	280
	281	/* number of elements in the iovector */
	282	unsigned long nr_segs;
	283
	284	/* set to indicate the type of the operation */
	285	int rw;
	286
	287	/* file offset */
	288	loff_t offset;
	289
	290	/* and the count in bytes */
	291	size_t bytes_to_be_copied;
	292
	293	ssize_t bytes_copied;
	294	int needs_cleanup;
	295	};
	296
8bb8aefd	297	struct orangefs_stats {
f7ab093f MM	298	unsigned long cache_hits;
	299	unsigned long cache_misses;
	300	unsigned long reads;
	301	unsigned long writes;
	302	};
	303
889d5f1b	304	extern struct orangefs_stats orangefs_stats;
f7ab093f MM	305
f7ab093f MM	306	/*
54804949 MM	307	* NOTE: See Documentation/filesystems/porting for information
	308	* on implementing FOO_I and properly accessing fs private data
	309	*/
8bb8aefd	310	static inline struct orangefs_inode_s ORANGEFS_I(struct inode inode)
f7ab093f	311	{
8bb8aefd	312	return container_of(inode, struct orangefs_inode_s, vfs_inode);
f7ab093f MM	313	}
f7ab093f MM	314
8bb8aefd	315	static inline struct orangefs_sb_info_s ORANGEFS_SB(struct super_block sb)
f7ab093f	316	{
8bb8aefd	317	return (struct orangefs_sb_info_s *) sb->s_fs_info;
f7ab093f MM	318	}
	319
	320	/* ino_t descends from "unsigned long", 8 bytes, 64 bits. */
8bb8aefd	321	static inline ino_t orangefs_khandle_to_ino(struct orangefs_khandle *khandle)
f7ab093f MM	322	{
	323	union {
	324	unsigned char u[8];
	325	__u64 ino;
	326	} ihandle;
	327
	328	ihandle.u[0] = khandle->u[0] ^ khandle->u[4];
	329	ihandle.u[1] = khandle->u[1] ^ khandle->u[5];
	330	ihandle.u[2] = khandle->u[2] ^ khandle->u[6];
	331	ihandle.u[3] = khandle->u[3] ^ khandle->u[7];
	332	ihandle.u[4] = khandle->u[12] ^ khandle->u[8];
	333	ihandle.u[5] = khandle->u[13] ^ khandle->u[9];
	334	ihandle.u[6] = khandle->u[14] ^ khandle->u[10];
	335	ihandle.u[7] = khandle->u[15] ^ khandle->u[11];
	336
	337	return ihandle.ino;
	338	}
	339
8bb8aefd	340	static inline struct orangefs_khandle get_khandle_from_ino(struct inode inode)
f7ab093f	341	{
8bb8aefd	342	return &(ORANGEFS_I(inode)->refn.khandle);
f7ab093f MM	343	}
f7ab093f MM	344
f7ab093f MM	345	static inline ino_t get_ino_from_khandle(struct inode *inode)
f7ab093f MM	346	{
8bb8aefd	347	struct orangefs_khandle *khandle;
f7ab093f MM	348	ino_t ino;
	349
	350	khandle = get_khandle_from_ino(inode);
8bb8aefd	351	ino = orangefs_khandle_to_ino(khandle);
f7ab093f MM	352	return ino;
	353	}
	354
	355	static inline ino_t get_parent_ino_from_dentry(struct dentry *dentry)
	356	{
	357	return get_ino_from_khandle(dentry->d_parent->d_inode);
	358	}
	359
	360	static inline int is_root_handle(struct inode *inode)
	361	{
	362	gossip_debug(GOSSIP_DCACHE_DEBUG,
	363	"%s: root handle: %pU, this handle: %pU:\n",
	364	__func__,
8bb8aefd	365	&ORANGEFS_SB(inode->i_sb)->root_khandle,
f7ab093f MM	366	get_khandle_from_ino(inode));
f7ab093f MM	367
8bb8aefd	368	if (ORANGEFS_khandle_cmp(&(ORANGEFS_SB(inode->i_sb)->root_khandle),
f7ab093f MM	369	get_khandle_from_ino(inode)))
	370	return 0;
	371	else
	372	return 1;
	373	}
	374
8bb8aefd	375	static inline int match_handle(struct orangefs_khandle resp_handle,
f7ab093f MM	376	struct inode *inode)
	377	{
	378	gossip_debug(GOSSIP_DCACHE_DEBUG,
	379	"%s: one handle: %pU, another handle:%pU:\n",
	380	__func__,
	381	&resp_handle,
	382	get_khandle_from_ino(inode));
	383
8bb8aefd	384	if (ORANGEFS_khandle_cmp(&resp_handle, get_khandle_from_ino(inode)))
f7ab093f MM	385	return 0;
	386	else
	387	return 1;
	388	}
	389
	390	/*
8bb8aefd	391	* defined in orangefs-cache.c
f7ab093f MM	392	*/
	393	int op_cache_initialize(void);
	394	int op_cache_finalize(void);
8bb8aefd	395	struct orangefs_kernel_op_s *op_alloc(__s32 type);
78699e29	396	void orangefs_new_tag(struct orangefs_kernel_op_s *op);
8bb8aefd	397	char get_opname_string(struct orangefs_kernel_op_s new_op);
f7ab093f	398
8bb8aefd YL	399	int orangefs_inode_cache_initialize(void);
8bb8aefd YL	400	int orangefs_inode_cache_finalize(void);
f7ab093f	401
f7ab093f	402	/*
8bb8aefd	403	* defined in orangefs-mod.c
f7ab093f MM	404	*/
	405	void purge_inprogress_ops(void);
	406
	407	/*
	408	* defined in waitqueue.c
	409	*/
f7ab093f MM	410	void purge_waiting_ops(void);
	411
	412	/*
	413	* defined in super.c
	414	*/
482664dd MB	415	extern uint64_t orangefs_features;
482664dd MB	416
8bb8aefd	417	struct dentry orangefs_mount(struct file_system_type fst,
f7ab093f MM	418	int flags,
	419	const char *devname,
	420	void *data);
	421
8bb8aefd	422	void orangefs_kill_sb(struct super_block *sb);
45996492	423	int orangefs_remount(struct orangefs_sb_info_s *);
f7ab093f MM	424
	425	int fsid_key_table_initialize(void);
	426	void fsid_key_table_finalize(void);
	427
	428	/*
	429	* defined in inode.c
	430	*/
8bb8aefd YL	431	__u32 convert_to_orangefs_mask(unsigned long lite_mask);
8bb8aefd YL	432	struct inode orangefs_new_inode(struct super_block sb,
f7ab093f MM	433	struct inode *dir,
	434	int mode,
	435	dev_t dev,
8bb8aefd	436	struct orangefs_object_kref *ref);
f7ab093f	437
8bb8aefd	438	int orangefs_setattr(struct dentry dentry, struct iattr iattr);
f7ab093f	439
a528d35e DH	440	int orangefs_getattr(const struct path path, struct kstat stat,
a528d35e DH	441	u32 request_mask, unsigned int flags);
f7ab093f	442
933287da MB	443	int orangefs_permission(struct inode *inode, int mask);
933287da MB	444
f7ab093f MM	445	/*
	446	* defined in xattr.c
	447	*/
8bb8aefd	448	int orangefs_setxattr(struct dentry *dentry,
f7ab093f MM	449	const char *name,
	450	const void *value,
	451	size_t size,
	452	int flags);
	453
8bb8aefd	454	ssize_t orangefs_getxattr(struct dentry *dentry,
f7ab093f MM	455	const char *name,
	456	void *buffer,
	457	size_t size);
	458
8bb8aefd	459	ssize_t orangefs_listxattr(struct dentry dentry, char buffer, size_t size);
f7ab093f MM	460
	461	/*
	462	* defined in namei.c
	463	*/
8bb8aefd YL	464	struct inode orangefs_iget(struct super_block sb,
8bb8aefd YL	465	struct orangefs_object_kref *ref);
f7ab093f	466
8bb8aefd YL	467	ssize_t orangefs_inode_read(struct inode *inode,
	468	struct iov_iter *iter,
	469	loff_t *offset,
	470	loff_t readahead_size);
f7ab093f MM	471
f7ab093f MM	472	/*
8bb8aefd	473	* defined in devorangefs-req.c
f7ab093f	474	*/
482664dd	475	extern uint32_t orangefs_userspace_version;
f2ee3b75	476
8bb8aefd YL	477	int orangefs_dev_init(void);
8bb8aefd YL	478	void orangefs_dev_cleanup(void);
f7ab093f	479	int is_daemon_in_service(void);
78699e29	480	bool __is_daemon_in_service(void);
f7ab093f MM	481
f7ab093f MM	482	/*
8bb8aefd	483	* defined in orangefs-utils.c
f7ab093f	484	*/
8bb8aefd	485	__s32 fsid_of_op(struct orangefs_kernel_op_s *op);
f7ab093f	486
8bb8aefd	487	int orangefs_flush_inode(struct inode *inode);
f7ab093f	488
8bb8aefd	489	ssize_t orangefs_inode_getxattr(struct inode *inode,
f7ab093f MM	490	const char *name,
	491	void *buffer,
	492	size_t size);
	493
8bb8aefd	494	int orangefs_inode_setxattr(struct inode *inode,
f7ab093f MM	495	const char *name,
	496	const void *value,
	497	size_t size,
	498	int flags);
	499
68a24a6c MB	500	int orangefs_inode_getattr(struct inode *inode, int new, int bypass,
68a24a6c MB	501	u32 request_mask);
f7ab093f	502
5859d77e MB	503	int orangefs_inode_check_changed(struct inode *inode);
5859d77e MB	504
8bb8aefd	505	int orangefs_inode_setattr(struct inode inode, struct iattr iattr);
f7ab093f	506
8bb8aefd	507	void orangefs_make_bad_inode(struct inode *inode);
f7ab093f	508
8bb8aefd	509	int orangefs_unmount_sb(struct super_block *sb);
f7ab093f	510
78699e29	511	bool orangefs_cancel_op_in_progress(struct orangefs_kernel_op_s *op);
f7ab093f	512
8bb8aefd	513	int orangefs_normalize_to_errno(__s32 error_code);
f7ab093f	514
1d503617	515	extern struct mutex orangefs_request_mutex;
f7ab093f MM	516	extern int op_timeout_secs;
f7ab093f MM	517	extern int slot_timeout_secs;
1d503617 MB	518	extern int orangefs_dcache_timeout_msecs;
1d503617 MB	519	extern int orangefs_getattr_timeout_msecs;
8bb8aefd YL	520	extern struct list_head orangefs_superblocks;
	521	extern spinlock_t orangefs_superblocks_lock;
	522	extern struct list_head orangefs_request_list;
	523	extern spinlock_t orangefs_request_list_lock;
	524	extern wait_queue_head_t orangefs_request_list_waitq;
1d503617 MB	525	extern struct list_head *orangefs_htable_ops_in_progress;
1d503617 MB	526	extern spinlock_t orangefs_htable_ops_in_progress_lock;
f7ab093f MM	527	extern int hash_table_size;
f7ab093f MM	528
8bb8aefd	529	extern const struct address_space_operations orangefs_address_operations;
6f3fc107	530	extern const struct inode_operations orangefs_file_inode_operations;
8bb8aefd	531	extern const struct file_operations orangefs_file_operations;
6f3fc107 AV	532	extern const struct inode_operations orangefs_symlink_inode_operations;
6f3fc107 AV	533	extern const struct inode_operations orangefs_dir_inode_operations;
8bb8aefd YL	534	extern const struct file_operations orangefs_dir_operations;
	535	extern const struct dentry_operations orangefs_dentry_operations;
	536	extern const struct file_operations orangefs_devreq_file_operations;
f7ab093f	537
8bb8aefd	538	extern wait_queue_head_t orangefs_bufmap_init_waitq;
f7ab093f MM	539
	540	/*
	541	* misc convenience macros
	542	*/
f7ab093f	543
8bb8aefd YL	544	#define ORANGEFS_OP_INTERRUPTIBLE 1 /* service_operation() is interruptible */
	545	#define ORANGEFS_OP_PRIORITY 2 /* service_operation() is high priority */
	546	#define ORANGEFS_OP_CANCELLATION 4 /* this is a cancellation */
adcf34a2	547	#define ORANGEFS_OP_NO_MUTEX 8 /* don't acquire request_mutex */
8bb8aefd	548	#define ORANGEFS_OP_ASYNC 16 /* Queue it, but don't wait */
f7ab093f	549
8bb8aefd	550	int service_operation(struct orangefs_kernel_op_s *op,
f7ab093f MM	551	const char *op_name,
	552	int flags);
	553
f7ab093f	554	#define get_interruptible_flag(inode) \
8bb8aefd YL	555	((ORANGEFS_SB(inode->i_sb)->flags & ORANGEFS_OPT_INTR) ? \
8bb8aefd YL	556	ORANGEFS_OP_INTERRUPTIBLE : 0)
f7ab093f	557
f7ab093f MM	558	#define fill_default_sys_attrs(sys_attr, type, mode) \
f7ab093f MM	559	do { \
78fee0b6 JH	560	sys_attr.owner = from_kuid(&init_user_ns, current_fsuid()); \
78fee0b6 JH	561	sys_attr.group = from_kgid(&init_user_ns, current_fsgid()); \
8bb8aefd	562	sys_attr.perms = ORANGEFS_util_translate_mode(mode); \
69a23de2 MB	563	sys_attr.mtime = 0; \
	564	sys_attr.atime = 0; \
	565	sys_attr.ctime = 0; \
8bb8aefd	566	sys_attr.mask = ORANGEFS_ATTR_SYS_ALL_SETABLE; \
f7ab093f MM	567	} while (0)
f7ab093f MM	568
8bb8aefd	569	static inline void orangefs_i_size_write(struct inode *inode, loff_t i_size)
f7ab093f MM	570	{
f7ab093f MM	571	#if BITS_PER_LONG == 32 && defined(CONFIG_SMP)
5ecfcb26	572	inode_lock(inode);
f7ab093f MM	573	#endif
	574	i_size_write(inode, i_size);
	575	#if BITS_PER_LONG == 32 && defined(CONFIG_SMP)
5ecfcb26	576	inode_unlock(inode);
f7ab093f MM	577	#endif
	578	}
	579
804b1737 MS	580	static inline void orangefs_set_timeout(struct dentry *dentry)
	581	{
	582	unsigned long time = jiffies + orangefs_dcache_timeout_msecs*HZ/1000;
	583
	584	dentry->d_fsdata = (void *) time;
	585	}
	586
8bb8aefd	587	#endif /* __ORANGEFSKERNEL_H */