[mirror_ubuntu-artful-kernel.git] / fs / orangefs / orangefs-kernel.h

/*
 * (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;
};

#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;
	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 __s32 get_fsid_from_ino(struct inode *inode)
{
	return ORANGEFS_I(inode)->refn.fs_id;
}

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(struct vfsmount *mnt,
		  struct dentry *dentry,
		  struct kstat *kstat);

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);

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
f7ab093f MM	1	/*
	2	* (C) 2001 Clemson University and The University of Chicago
	3	*
	4	* See COPYING in top-level directory.
	5	*/
	6
	7	/*
8bb8aefd	8	* The ORANGEFS Linux kernel support allows ORANGEFS volumes to be mounted and
f7ab093f MM	9	* accessed through the Linux VFS (i.e. using standard I/O system calls).
	10	* This support is only needed on clients that wish to mount the file system.
	11	*
	12	*/
	13
	14	/*
8bb8aefd	15	* Declarations and macros for the ORANGEFS Linux kernel support.
f7ab093f MM	16	*/
f7ab093f MM	17
8bb8aefd YL	18	#ifndef __ORANGEFSKERNEL_H
8bb8aefd YL	19	#define __ORANGEFSKERNEL_H
f7ab093f MM	20
	21	#include <linux/kernel.h>
	22	#include <linux/moduleparam.h>
	23	#include <linux/statfs.h>
	24	#include <linux/backing-dev.h>
	25	#include <linux/device.h>
	26	#include <linux/mpage.h>
	27	#include <linux/namei.h>
	28	#include <linux/errno.h>
	29	#include <linux/init.h>
	30	#include <linux/module.h>
	31	#include <linux/slab.h>
	32	#include <linux/types.h>
	33	#include <linux/fs.h>
	34	#include <linux/vmalloc.h>
	35
	36	#include <linux/aio.h>
	37	#include <linux/posix_acl.h>
	38	#include <linux/posix_acl_xattr.h>
	39	#include <linux/compat.h>
	40	#include <linux/mount.h>
	41	#include <linux/uaccess.h>
	42	#include <linux/atomic.h>
	43	#include <linux/uio.h>
174cd4b1	44	#include <linux/sched/signal.h>
f7ab093f MM	45	#include <linux/mm.h>
	46	#include <linux/wait.h>
	47	#include <linux/dcache.h>
	48	#include <linux/pagemap.h>
	49	#include <linux/poll.h>
	50	#include <linux/rwsem.h>
	51	#include <linux/xattr.h>
	52	#include <linux/exportfs.h>
	53
	54	#include <asm/unaligned.h>
	55
575e9461	56	#include "orangefs-dev-proto.h"
f7ab093f	57
8bb8aefd YL	58	#ifdef ORANGEFS_KERNEL_DEBUG
8bb8aefd YL	59	#define ORANGEFS_DEFAULT_OP_TIMEOUT_SECS 10
f7ab093f	60	#else
8bb8aefd	61	#define ORANGEFS_DEFAULT_OP_TIMEOUT_SECS 20
f7ab093f MM	62	#endif
f7ab093f MM	63
8bb8aefd	64	#define ORANGEFS_BUFMAP_WAIT_TIMEOUT_SECS 30
f7ab093f	65
8bb8aefd	66	#define ORANGEFS_DEFAULT_SLOT_TIMEOUT_SECS 900 /* 15 minutes */
f7ab093f	67
8bb8aefd	68	#define ORANGEFS_REQDEVICE_NAME "pvfs2-req"
f7ab093f	69
8bb8aefd YL	70	#define ORANGEFS_DEVREQ_MAGIC 0x20030529
	71	#define ORANGEFS_LINK_MAX 0x000000FF
	72	#define ORANGEFS_PURGE_RETRY_COUNT 0x00000005
8bb8aefd YL	73	#define ORANGEFS_MAX_NUM_OPTIONS 0x00000004
	74	#define ORANGEFS_MAX_MOUNT_OPT_LEN 0x00000080
	75	#define ORANGEFS_MAX_FSKEY_LEN 64
f7ab093f	76
cf0c2771	77	#define MAX_DEV_REQ_UPSIZE (2 * sizeof(__s32) + \
8bb8aefd	78	sizeof(__u64) + sizeof(struct orangefs_upcall_s))
cf0c2771	79	#define MAX_DEV_REQ_DOWNSIZE (2 * sizeof(__s32) + \
8bb8aefd	80	sizeof(__u64) + sizeof(struct orangefs_downcall_s))
f7ab093f	81
f7ab093f	82	/*
8bb8aefd	83	* valid orangefs kernel operation states
f7ab093f MM	84	*
	85	* unknown - op was just initialized
	86	* waiting - op is on request_list (upward bound)
	87	* inprogr - op is in progress (waiting for downcall)
	88	* serviced - op has matching downcall; ok
	89	* purged - op has to start a timer since client-core
	90	* exited uncleanly before servicing op
ed42fe05	91	* given up - submitter has given up waiting for it
f7ab093f	92	*/
8bb8aefd	93	enum orangefs_vfs_op_states {
f7ab093f MM	94	OP_VFS_STATE_UNKNOWN = 0,
	95	OP_VFS_STATE_WAITING = 1,
	96	OP_VFS_STATE_INPROGR = 2,
	97	OP_VFS_STATE_SERVICED = 4,
	98	OP_VFS_STATE_PURGED = 8,
ed42fe05	99	OP_VFS_STATE_GIVEN_UP = 16,
f7ab093f MM	100	};
f7ab093f MM	101
f7ab093f	102	/*
8bb8aefd	103	* orangefs kernel memory related flags
f7ab093f MM	104	*/
f7ab093f MM	105
8bb8aefd YL	106	#if ((defined ORANGEFS_KERNEL_DEBUG) && (defined CONFIG_DEBUG_SLAB))
8bb8aefd YL	107	#define ORANGEFS_CACHE_CREATE_FLAGS SLAB_RED_ZONE
f7ab093f	108	#else
8bb8aefd YL	109	#define ORANGEFS_CACHE_CREATE_FLAGS 0
8bb8aefd YL	110	#endif /* ((defined ORANGEFS_KERNEL_DEBUG) && (defined CONFIG_DEBUG_SLAB)) */
f7ab093f	111
8bb8aefd YL	112	extern int orangefs_init_acl(struct inode inode, struct inode dir);
8bb8aefd YL	113	extern const struct xattr_handler *orangefs_xattr_handlers[];
f7ab093f	114
8bb8aefd YL	115	extern struct posix_acl orangefs_get_acl(struct inode inode, int type);
8bb8aefd YL	116	extern int orangefs_set_acl(struct inode inode, struct posix_acl acl, int type);
f7ab093f	117
f7ab093f MM	118	/*
	119	* Redefine xtvec structure so that we could move helper functions out of
	120	* the define
	121	*/
	122	struct xtvec {
	123	__kernel_off_t xtv_off; /* must be off_t */
	124	__kernel_size_t xtv_len; /* must be size_t */
	125	};
	126
	127	/*
8bb8aefd	128	* orangefs data structures
f7ab093f	129	*/
8bb8aefd YL	130	struct orangefs_kernel_op_s {
8bb8aefd YL	131	enum orangefs_vfs_op_states op_state;
f7ab093f MM	132	__u64 tag;
	133
	134	/*
d37c0f30 MM	135	* Set uses_shared_memory to non zero if this operation uses
	136	* shared memory. If true, then a retry on the op must also
	137	* get a new shared memory buffer and re-populate it.
	138	* Cancels don't care - it only matters for service_operation()
	139	* retry logics and cancels don't go through it anymore. It
	140	* safely stays non-zero when we use it as slot_to_free.
f7ab093f	141	*/
78699e29 AV	142	union {
	143	int uses_shared_memory;
	144	int slot_to_free;
	145	};
f7ab093f	146
8bb8aefd YL	147	struct orangefs_upcall_s upcall;
8bb8aefd YL	148	struct orangefs_downcall_s downcall;
f7ab093f	149
d2d87a3b	150	struct completion waitq;
f7ab093f MM	151	spinlock_t lock;
f7ab093f MM	152
f7ab093f MM	153	int attempts;
	154
	155	struct list_head list;
	156	};
	157
60831949 AV	158	#define set_op_state_waiting(op) ((op)->op_state = OP_VFS_STATE_WAITING)
60831949 AV	159	#define set_op_state_inprogress(op) ((op)->op_state = OP_VFS_STATE_INPROGR)
2d4cae0d	160	#define set_op_state_given_up(op) ((op)->op_state = OP_VFS_STATE_GIVEN_UP)
60831949 AV	161	static inline void set_op_state_serviced(struct orangefs_kernel_op_s *op)
	162	{
	163	op->op_state = OP_VFS_STATE_SERVICED;
d2d87a3b	164	complete(&op->waitq);
60831949	165	}
60831949 AV	166
	167	#define op_state_waiting(op) ((op)->op_state & OP_VFS_STATE_WAITING)
	168	#define op_state_in_progress(op) ((op)->op_state & OP_VFS_STATE_INPROGR)
	169	#define op_state_serviced(op) ((op)->op_state & OP_VFS_STATE_SERVICED)
	170	#define op_state_purged(op) ((op)->op_state & OP_VFS_STATE_PURGED)
ed42fe05	171	#define op_state_given_up(op) ((op)->op_state & OP_VFS_STATE_GIVEN_UP)
78699e29	172	#define op_is_cancel(op) ((op)->upcall.type == ORANGEFS_VFS_OP_CANCEL)
ed42fe05	173
c1223ca4	174	void op_release(struct orangefs_kernel_op_s *op);
60831949	175
78699e29 AV	176	extern void orangefs_bufmap_put(int);
	177	static inline void put_cancel(struct orangefs_kernel_op_s *op)
	178	{
	179	orangefs_bufmap_put(op->slot_to_free);
	180	op_release(op);
	181	}
	182
	183	static inline void set_op_state_purged(struct orangefs_kernel_op_s *op)
	184	{
	185	spin_lock(&op->lock);
	186	if (unlikely(op_is_cancel(op))) {
05a50a5b	187	list_del_init(&op->list);
78699e29 AV	188	spin_unlock(&op->lock);
	189	put_cancel(op);
	190	} else {
	191	op->op_state \|= OP_VFS_STATE_PURGED;
d2d87a3b	192	complete(&op->waitq);
78699e29 AV	193	spin_unlock(&op->lock);
	194	}
	195	}
	196
8bb8aefd YL	197	/* per inode private orangefs info */
	198	struct orangefs_inode_s {
	199	struct orangefs_object_kref refn;
	200	char link_target[ORANGEFS_NAME_MAX];
f7ab093f MM	201	__s64 blksize;
	202	/*
	203	* Reading/Writing Extended attributes need to acquire the appropriate
8bb8aefd	204	* reader/writer semaphore on the orangefs_inode_s structure.
f7ab093f MM	205	*/
	206	struct rw_semaphore xattr_sem;
	207
	208	struct inode vfs_inode;
	209	sector_t last_failed_block_index_read;
	210
	211	/*
	212	* State of in-memory attributes not yet flushed to disk associated
	213	* with this object
	214	*/
	215	unsigned long pinode_flags;
71680c18 MB	216
71680c18 MB	217	unsigned long getattr_time;
f7ab093f MM	218	};
	219
	220	#define P_ATIME_FLAG 0
	221	#define P_MTIME_FLAG 1
	222	#define P_CTIME_FLAG 2
	223	#define P_MODE_FLAG 3
	224
	225	#define ClearAtimeFlag(pinode) clear_bit(P_ATIME_FLAG, &(pinode)->pinode_flags)
	226	#define SetAtimeFlag(pinode) set_bit(P_ATIME_FLAG, &(pinode)->pinode_flags)
	227	#define AtimeFlag(pinode) test_bit(P_ATIME_FLAG, &(pinode)->pinode_flags)
	228
	229	#define ClearMtimeFlag(pinode) clear_bit(P_MTIME_FLAG, &(pinode)->pinode_flags)
	230	#define SetMtimeFlag(pinode) set_bit(P_MTIME_FLAG, &(pinode)->pinode_flags)
	231	#define MtimeFlag(pinode) test_bit(P_MTIME_FLAG, &(pinode)->pinode_flags)
	232
	233	#define ClearCtimeFlag(pinode) clear_bit(P_CTIME_FLAG, &(pinode)->pinode_flags)
	234	#define SetCtimeFlag(pinode) set_bit(P_CTIME_FLAG, &(pinode)->pinode_flags)
	235	#define CtimeFlag(pinode) test_bit(P_CTIME_FLAG, &(pinode)->pinode_flags)
	236
	237	#define ClearModeFlag(pinode) clear_bit(P_MODE_FLAG, &(pinode)->pinode_flags)
	238	#define SetModeFlag(pinode) set_bit(P_MODE_FLAG, &(pinode)->pinode_flags)
	239	#define ModeFlag(pinode) test_bit(P_MODE_FLAG, &(pinode)->pinode_flags)
	240
8bb8aefd YL	241	/* per superblock private orangefs info */
	242	struct orangefs_sb_info_s {
	243	struct orangefs_khandle root_khandle;
f7ab093f MM	244	__s32 fs_id;
	245	int id;
	246	int flags;
8bb8aefd YL	247	#define ORANGEFS_OPT_INTR 0x01
	248	#define ORANGEFS_OPT_LOCAL_LOCK 0x02
	249	char devname[ORANGEFS_MAX_SERVER_ADDR_LEN];
f7ab093f MM	250	struct super_block *sb;
	251	int mount_pending;
	252	struct list_head list;
	253	};
	254
f7ab093f MM	255	/*
	256	* structure that holds the state of any async I/O operation issued
	257	* through the VFS. Needed especially to handle cancellation requests
	258	* or even completion notification so that the VFS client-side daemon
	259	* can free up its vfs_request slots.
	260	*/
8bb8aefd	261	struct orangefs_kiocb_s {
f7ab093f MM	262	/* the pointer to the task that initiated the AIO */
	263	struct task_struct *tsk;
	264
	265	/* pointer to the kiocb that kicked this operation */
	266	struct kiocb *kiocb;
	267
	268	/* buffer index that was used for the I/O */
8bb8aefd	269	struct orangefs_bufmap *bufmap;
f7ab093f MM	270	int buffer_index;
f7ab093f MM	271
8bb8aefd YL	272	/* orangefs kernel operation type */
8bb8aefd YL	273	struct orangefs_kernel_op_s *op;
f7ab093f MM	274
	275	/* The user space buffers from/to which I/O is being staged */
	276	struct iovec *iov;
	277
	278	/* number of elements in the iovector */
	279	unsigned long nr_segs;
	280
	281	/* set to indicate the type of the operation */
	282	int rw;
	283
	284	/* file offset */
	285	loff_t offset;
	286
	287	/* and the count in bytes */
	288	size_t bytes_to_be_copied;
	289
	290	ssize_t bytes_copied;
	291	int needs_cleanup;
	292	};
	293
8bb8aefd	294	struct orangefs_stats {
f7ab093f MM	295	unsigned long cache_hits;
	296	unsigned long cache_misses;
	297	unsigned long reads;
	298	unsigned long writes;
	299	};
	300
889d5f1b	301	extern struct orangefs_stats orangefs_stats;
f7ab093f MM	302
f7ab093f MM	303	/*
54804949 MM	304	* NOTE: See Documentation/filesystems/porting for information
	305	* on implementing FOO_I and properly accessing fs private data
	306	*/
8bb8aefd	307	static inline struct orangefs_inode_s ORANGEFS_I(struct inode inode)
f7ab093f	308	{
8bb8aefd	309	return container_of(inode, struct orangefs_inode_s, vfs_inode);
f7ab093f MM	310	}
f7ab093f MM	311
8bb8aefd	312	static inline struct orangefs_sb_info_s ORANGEFS_SB(struct super_block sb)
f7ab093f	313	{
8bb8aefd	314	return (struct orangefs_sb_info_s *) sb->s_fs_info;
f7ab093f MM	315	}
	316
	317	/* ino_t descends from "unsigned long", 8 bytes, 64 bits. */
8bb8aefd	318	static inline ino_t orangefs_khandle_to_ino(struct orangefs_khandle *khandle)
f7ab093f MM	319	{
	320	union {
	321	unsigned char u[8];
	322	__u64 ino;
	323	} ihandle;
	324
	325	ihandle.u[0] = khandle->u[0] ^ khandle->u[4];
	326	ihandle.u[1] = khandle->u[1] ^ khandle->u[5];
	327	ihandle.u[2] = khandle->u[2] ^ khandle->u[6];
	328	ihandle.u[3] = khandle->u[3] ^ khandle->u[7];
	329	ihandle.u[4] = khandle->u[12] ^ khandle->u[8];
	330	ihandle.u[5] = khandle->u[13] ^ khandle->u[9];
	331	ihandle.u[6] = khandle->u[14] ^ khandle->u[10];
	332	ihandle.u[7] = khandle->u[15] ^ khandle->u[11];
	333
	334	return ihandle.ino;
	335	}
	336
8bb8aefd	337	static inline struct orangefs_khandle get_khandle_from_ino(struct inode inode)
f7ab093f	338	{
8bb8aefd	339	return &(ORANGEFS_I(inode)->refn.khandle);
f7ab093f MM	340	}
	341
	342	static inline __s32 get_fsid_from_ino(struct inode *inode)
	343	{
8bb8aefd	344	return ORANGEFS_I(inode)->refn.fs_id;
f7ab093f MM	345	}
	346
	347	static inline ino_t get_ino_from_khandle(struct inode *inode)
	348	{
8bb8aefd	349	struct orangefs_khandle *khandle;
f7ab093f MM	350	ino_t ino;
	351
	352	khandle = get_khandle_from_ino(inode);
8bb8aefd	353	ino = orangefs_khandle_to_ino(khandle);
f7ab093f MM	354	return ino;
	355	}
	356
	357	static inline ino_t get_parent_ino_from_dentry(struct dentry *dentry)
	358	{
	359	return get_ino_from_khandle(dentry->d_parent->d_inode);
	360	}
	361
	362	static inline int is_root_handle(struct inode *inode)
	363	{
	364	gossip_debug(GOSSIP_DCACHE_DEBUG,
	365	"%s: root handle: %pU, this handle: %pU:\n",
	366	__func__,
8bb8aefd	367	&ORANGEFS_SB(inode->i_sb)->root_khandle,
f7ab093f MM	368	get_khandle_from_ino(inode));
f7ab093f MM	369
8bb8aefd	370	if (ORANGEFS_khandle_cmp(&(ORANGEFS_SB(inode->i_sb)->root_khandle),
f7ab093f MM	371	get_khandle_from_ino(inode)))
	372	return 0;
	373	else
	374	return 1;
	375	}
	376
8bb8aefd	377	static inline int match_handle(struct orangefs_khandle resp_handle,
f7ab093f MM	378	struct inode *inode)
	379	{
	380	gossip_debug(GOSSIP_DCACHE_DEBUG,
	381	"%s: one handle: %pU, another handle:%pU:\n",
	382	__func__,
	383	&resp_handle,
	384	get_khandle_from_ino(inode));
	385
8bb8aefd	386	if (ORANGEFS_khandle_cmp(&resp_handle, get_khandle_from_ino(inode)))
f7ab093f MM	387	return 0;
	388	else
	389	return 1;
	390	}
	391
	392	/*
8bb8aefd	393	* defined in orangefs-cache.c
f7ab093f MM	394	*/
	395	int op_cache_initialize(void);
	396	int op_cache_finalize(void);
8bb8aefd	397	struct orangefs_kernel_op_s *op_alloc(__s32 type);
78699e29	398	void orangefs_new_tag(struct orangefs_kernel_op_s *op);
8bb8aefd	399	char get_opname_string(struct orangefs_kernel_op_s new_op);
f7ab093f	400
8bb8aefd YL	401	int orangefs_inode_cache_initialize(void);
8bb8aefd YL	402	int orangefs_inode_cache_finalize(void);
f7ab093f	403
f7ab093f	404	/*
8bb8aefd	405	* defined in orangefs-mod.c
f7ab093f MM	406	*/
	407	void purge_inprogress_ops(void);
	408
	409	/*
	410	* defined in waitqueue.c
	411	*/
f7ab093f MM	412	void purge_waiting_ops(void);
	413
	414	/*
	415	* defined in super.c
	416	*/
482664dd MB	417	extern uint64_t orangefs_features;
482664dd MB	418
8bb8aefd	419	struct dentry orangefs_mount(struct file_system_type fst,
f7ab093f MM	420	int flags,
	421	const char *devname,
	422	void *data);
	423
8bb8aefd	424	void orangefs_kill_sb(struct super_block *sb);
45996492	425	int orangefs_remount(struct orangefs_sb_info_s *);
f7ab093f MM	426
	427	int fsid_key_table_initialize(void);
	428	void fsid_key_table_finalize(void);
	429
	430	/*
	431	* defined in inode.c
	432	*/
8bb8aefd YL	433	__u32 convert_to_orangefs_mask(unsigned long lite_mask);
8bb8aefd YL	434	struct inode orangefs_new_inode(struct super_block sb,
f7ab093f MM	435	struct inode *dir,
	436	int mode,
	437	dev_t dev,
8bb8aefd	438	struct orangefs_object_kref *ref);
f7ab093f	439
8bb8aefd	440	int orangefs_setattr(struct dentry dentry, struct iattr iattr);
f7ab093f	441
8bb8aefd	442	int orangefs_getattr(struct vfsmount *mnt,
f7ab093f MM	443	struct dentry *dentry,
	444	struct kstat *kstat);
	445
933287da MB	446	int orangefs_permission(struct inode *inode, int mask);
933287da MB	447
f7ab093f MM	448	/*
	449	* defined in xattr.c
	450	*/
8bb8aefd	451	int orangefs_setxattr(struct dentry *dentry,
f7ab093f MM	452	const char *name,
	453	const void *value,
	454	size_t size,
	455	int flags);
	456
8bb8aefd	457	ssize_t orangefs_getxattr(struct dentry *dentry,
f7ab093f MM	458	const char *name,
	459	void *buffer,
	460	size_t size);
	461
8bb8aefd	462	ssize_t orangefs_listxattr(struct dentry dentry, char buffer, size_t size);
f7ab093f MM	463
	464	/*
	465	* defined in namei.c
	466	*/
8bb8aefd YL	467	struct inode orangefs_iget(struct super_block sb,
8bb8aefd YL	468	struct orangefs_object_kref *ref);
f7ab093f	469
8bb8aefd YL	470	ssize_t orangefs_inode_read(struct inode *inode,
	471	struct iov_iter *iter,
	472	loff_t *offset,
	473	loff_t readahead_size);
f7ab093f MM	474
f7ab093f MM	475	/*
8bb8aefd	476	* defined in devorangefs-req.c
f7ab093f	477	*/
482664dd	478	extern uint32_t orangefs_userspace_version;
f2ee3b75	479
8bb8aefd YL	480	int orangefs_dev_init(void);
8bb8aefd YL	481	void orangefs_dev_cleanup(void);
f7ab093f	482	int is_daemon_in_service(void);
78699e29	483	bool __is_daemon_in_service(void);
f7ab093f MM	484
f7ab093f MM	485	/*
8bb8aefd	486	* defined in orangefs-utils.c
f7ab093f	487	*/
8bb8aefd	488	__s32 fsid_of_op(struct orangefs_kernel_op_s *op);
f7ab093f	489
8bb8aefd	490	int orangefs_flush_inode(struct inode *inode);
f7ab093f	491
8bb8aefd	492	ssize_t orangefs_inode_getxattr(struct inode *inode,
f7ab093f MM	493	const char *name,
	494	void *buffer,
	495	size_t size);
	496
8bb8aefd	497	int orangefs_inode_setxattr(struct inode *inode,
f7ab093f MM	498	const char *name,
	499	const void *value,
	500	size_t size,
	501	int flags);
	502
71680c18	503	int orangefs_inode_getattr(struct inode *inode, int new, int bypass);
f7ab093f	504
5859d77e MB	505	int orangefs_inode_check_changed(struct inode *inode);
5859d77e MB	506
8bb8aefd	507	int orangefs_inode_setattr(struct inode inode, struct iattr iattr);
f7ab093f	508
8bb8aefd	509	void orangefs_make_bad_inode(struct inode *inode);
f7ab093f	510
8bb8aefd	511	int orangefs_unmount_sb(struct super_block *sb);
f7ab093f	512
78699e29	513	bool orangefs_cancel_op_in_progress(struct orangefs_kernel_op_s *op);
f7ab093f	514
8bb8aefd	515	int orangefs_normalize_to_errno(__s32 error_code);
f7ab093f	516
1d503617	517	extern struct mutex orangefs_request_mutex;
f7ab093f MM	518	extern int op_timeout_secs;
f7ab093f MM	519	extern int slot_timeout_secs;
1d503617 MB	520	extern int orangefs_dcache_timeout_msecs;
1d503617 MB	521	extern int orangefs_getattr_timeout_msecs;
8bb8aefd YL	522	extern struct list_head orangefs_superblocks;
	523	extern spinlock_t orangefs_superblocks_lock;
	524	extern struct list_head orangefs_request_list;
	525	extern spinlock_t orangefs_request_list_lock;
	526	extern wait_queue_head_t orangefs_request_list_waitq;
1d503617 MB	527	extern struct list_head *orangefs_htable_ops_in_progress;
1d503617 MB	528	extern spinlock_t orangefs_htable_ops_in_progress_lock;
f7ab093f MM	529	extern int hash_table_size;
f7ab093f MM	530
8bb8aefd	531	extern const struct address_space_operations orangefs_address_operations;
6f3fc107	532	extern const struct inode_operations orangefs_file_inode_operations;
8bb8aefd	533	extern const struct file_operations orangefs_file_operations;
6f3fc107 AV	534	extern const struct inode_operations orangefs_symlink_inode_operations;
6f3fc107 AV	535	extern const struct inode_operations orangefs_dir_inode_operations;
8bb8aefd YL	536	extern const struct file_operations orangefs_dir_operations;
	537	extern const struct dentry_operations orangefs_dentry_operations;
	538	extern const struct file_operations orangefs_devreq_file_operations;
f7ab093f	539
8bb8aefd	540	extern wait_queue_head_t orangefs_bufmap_init_waitq;
f7ab093f MM	541
	542	/*
	543	* misc convenience macros
	544	*/
f7ab093f	545
8bb8aefd YL	546	#define ORANGEFS_OP_INTERRUPTIBLE 1 /* service_operation() is interruptible */
	547	#define ORANGEFS_OP_PRIORITY 2 /* service_operation() is high priority */
	548	#define ORANGEFS_OP_CANCELLATION 4 /* this is a cancellation */
adcf34a2	549	#define ORANGEFS_OP_NO_MUTEX 8 /* don't acquire request_mutex */
8bb8aefd	550	#define ORANGEFS_OP_ASYNC 16 /* Queue it, but don't wait */
f7ab093f	551
8bb8aefd	552	int service_operation(struct orangefs_kernel_op_s *op,
f7ab093f MM	553	const char *op_name,
	554	int flags);
	555
f7ab093f	556	#define get_interruptible_flag(inode) \
8bb8aefd YL	557	((ORANGEFS_SB(inode->i_sb)->flags & ORANGEFS_OPT_INTR) ? \
8bb8aefd YL	558	ORANGEFS_OP_INTERRUPTIBLE : 0)
f7ab093f	559
f7ab093f MM	560	#define fill_default_sys_attrs(sys_attr, type, mode) \
f7ab093f MM	561	do { \
78fee0b6 JH	562	sys_attr.owner = from_kuid(&init_user_ns, current_fsuid()); \
78fee0b6 JH	563	sys_attr.group = from_kgid(&init_user_ns, current_fsgid()); \
8bb8aefd	564	sys_attr.perms = ORANGEFS_util_translate_mode(mode); \
69a23de2 MB	565	sys_attr.mtime = 0; \
	566	sys_attr.atime = 0; \
	567	sys_attr.ctime = 0; \
8bb8aefd	568	sys_attr.mask = ORANGEFS_ATTR_SYS_ALL_SETABLE; \
f7ab093f MM	569	} while (0)
f7ab093f MM	570
8bb8aefd	571	static inline void orangefs_i_size_write(struct inode *inode, loff_t i_size)
f7ab093f MM	572	{
f7ab093f MM	573	#if BITS_PER_LONG == 32 && defined(CONFIG_SMP)
5ecfcb26	574	inode_lock(inode);
f7ab093f MM	575	#endif
	576	i_size_write(inode, i_size);
	577	#if BITS_PER_LONG == 32 && defined(CONFIG_SMP)
5ecfcb26	578	inode_unlock(inode);
f7ab093f MM	579	#endif
	580	}
	581
804b1737 MS	582	static inline void orangefs_set_timeout(struct dentry *dentry)
	583	{
	584	unsigned long time = jiffies + orangefs_dcache_timeout_msecs*HZ/1000;
	585
	586	dentry->d_fsdata = (void *) time;
	587	}
	588
8bb8aefd	589	#endif /* __ORANGEFSKERNEL_H */