[linux-2.6-block.git] / fs / orangefs / super.c

/*
 * (C) 2001 Clemson University and The University of Chicago
 *
 * See COPYING in top-level directory.
 */

#include "protocol.h"
#include "pvfs2-kernel.h"
#include "pvfs2-bufmap.h"

#include <linux/parser.h>

/* a cache for orangefs-inode objects (i.e. orangefs inode private data) */
static struct kmem_cache *orangefs_inode_cache;

/* list for storing orangefs specific superblocks in use */
LIST_HEAD(orangefs_superblocks);

DEFINE_SPINLOCK(orangefs_superblocks_lock);

enum {
	Opt_intr,
	Opt_acl,
	Opt_local_lock,

	Opt_err
};

static const match_table_t tokens = {
	{ Opt_acl,		"acl" },
	{ Opt_intr,		"intr" },
	{ Opt_local_lock,	"local_lock" },
	{ Opt_err,	NULL }
};


static int parse_mount_options(struct super_block *sb, char *options,
		int silent)
{
	struct orangefs_sb_info_s *orangefs_sb = ORANGEFS_SB(sb);
	substring_t args[MAX_OPT_ARGS];
	char *p;

	/*
	 * Force any potential flags that might be set from the mount
	 * to zero, ie, initialize to unset.
	 */
	sb->s_flags &= ~MS_POSIXACL;
	orangefs_sb->flags &= ~ORANGEFS_OPT_INTR;
	orangefs_sb->flags &= ~ORANGEFS_OPT_LOCAL_LOCK;

	while ((p = strsep(&options, ",")) != NULL) {
		int token;

		if (!*p)
			continue;

		token = match_token(p, tokens, args);
		switch (token) {
		case Opt_acl:
			sb->s_flags |= MS_POSIXACL;
			break;
		case Opt_intr:
			orangefs_sb->flags |= ORANGEFS_OPT_INTR;
			break;
		case Opt_local_lock:
			orangefs_sb->flags |= ORANGEFS_OPT_LOCAL_LOCK;
			break;
		default:
			goto fail;
		}
	}

	return 0;
fail:
	if (!silent)
		gossip_err("Error: mount option [%s] is not supported.\n", p);
	return -EINVAL;
}

static void orangefs_inode_cache_ctor(void *req)
{
	struct orangefs_inode_s *orangefs_inode = req;

	inode_init_once(&orangefs_inode->vfs_inode);
	init_rwsem(&orangefs_inode->xattr_sem);

	orangefs_inode->vfs_inode.i_version = 1;
}

static struct inode *orangefs_alloc_inode(struct super_block *sb)
{
	struct orangefs_inode_s *orangefs_inode;

	orangefs_inode = kmem_cache_alloc(orangefs_inode_cache,
				       ORANGEFS_CACHE_ALLOC_FLAGS);
	if (orangefs_inode == NULL) {
		gossip_err("Failed to allocate orangefs_inode\n");
		return NULL;
	}

	/*
	 * We want to clear everything except for rw_semaphore and the
	 * vfs_inode.
	 */
	memset(&orangefs_inode->refn.khandle, 0, 16);
	orangefs_inode->refn.fs_id = ORANGEFS_FS_ID_NULL;
	orangefs_inode->last_failed_block_index_read = 0;
	memset(orangefs_inode->link_target, 0, sizeof(orangefs_inode->link_target));
	orangefs_inode->pinode_flags = 0;

	gossip_debug(GOSSIP_SUPER_DEBUG,
		     "orangefs_alloc_inode: allocated %p\n",
		     &orangefs_inode->vfs_inode);
	return &orangefs_inode->vfs_inode;
}

static void orangefs_destroy_inode(struct inode *inode)
{
	struct orangefs_inode_s *orangefs_inode = ORANGEFS_I(inode);

	gossip_debug(GOSSIP_SUPER_DEBUG,
			"%s: deallocated %p destroying inode %pU\n",
			__func__, orangefs_inode, get_khandle_from_ino(inode));

	kmem_cache_free(orangefs_inode_cache, orangefs_inode);
}

/*
 * NOTE: information filled in here is typically reflected in the
 * output of the system command 'df'
*/
static int orangefs_statfs(struct dentry *dentry, struct kstatfs *buf)
{
	int ret = -ENOMEM;
	struct orangefs_kernel_op_s *new_op = NULL;
	int flags = 0;
	struct super_block *sb = NULL;

	sb = dentry->d_sb;

	gossip_debug(GOSSIP_SUPER_DEBUG,
		     "orangefs_statfs: called on sb %p (fs_id is %d)\n",
		     sb,
		     (int)(ORANGEFS_SB(sb)->fs_id));

	new_op = op_alloc(ORANGEFS_VFS_OP_STATFS);
	if (!new_op)
		return ret;
	new_op->upcall.req.statfs.fs_id = ORANGEFS_SB(sb)->fs_id;

	if (ORANGEFS_SB(sb)->flags & ORANGEFS_OPT_INTR)
		flags = ORANGEFS_OP_INTERRUPTIBLE;

	ret = service_operation(new_op, "orangefs_statfs", flags);

	if (new_op->downcall.status < 0)
		goto out_op_release;

	gossip_debug(GOSSIP_SUPER_DEBUG,
		     "orangefs_statfs: got %ld blocks available | "
		     "%ld blocks total | %ld block size\n",
		     (long)new_op->downcall.resp.statfs.blocks_avail,
		     (long)new_op->downcall.resp.statfs.blocks_total,
		     (long)new_op->downcall.resp.statfs.block_size);

	buf->f_type = sb->s_magic;
	memcpy(&buf->f_fsid, &ORANGEFS_SB(sb)->fs_id, sizeof(buf->f_fsid));
	buf->f_bsize = new_op->downcall.resp.statfs.block_size;
	buf->f_namelen = ORANGEFS_NAME_LEN;

	buf->f_blocks = (sector_t) new_op->downcall.resp.statfs.blocks_total;
	buf->f_bfree = (sector_t) new_op->downcall.resp.statfs.blocks_avail;
	buf->f_bavail = (sector_t) new_op->downcall.resp.statfs.blocks_avail;
	buf->f_files = (sector_t) new_op->downcall.resp.statfs.files_total;
	buf->f_ffree = (sector_t) new_op->downcall.resp.statfs.files_avail;
	buf->f_frsize = sb->s_blocksize;

out_op_release:
	op_release(new_op);
	gossip_debug(GOSSIP_SUPER_DEBUG, "orangefs_statfs: returning %d\n", ret);
	return ret;
}

/*
 * Remount as initiated by VFS layer.  We just need to reparse the mount
 * options, no need to signal pvfs2-client-core about it.
 */
static int orangefs_remount_fs(struct super_block *sb, int *flags, char *data)
{
	gossip_debug(GOSSIP_SUPER_DEBUG, "orangefs_remount_fs: called\n");
	return parse_mount_options(sb, data, 1);
}

/*
 * Remount as initiated by pvfs2-client-core on restart.  This is used to
 * repopulate mount information left from previous pvfs2-client-core.
 *
 * the idea here is that given a valid superblock, we're
 * re-initializing the user space client with the initial mount
 * information specified when the super block was first initialized.
 * this is very different than the first initialization/creation of a
 * superblock.  we use the special service_priority_operation to make
 * sure that the mount gets ahead of any other pending operation that
 * is waiting for servicing.  this means that the pvfs2-client won't
 * fail to start several times for all other pending operations before
 * the client regains all of the mount information from us.
 * NOTE: this function assumes that the request_mutex is already acquired!
 */
int orangefs_remount(struct super_block *sb)
{
	struct orangefs_kernel_op_s *new_op;
	int ret = -EINVAL;

	gossip_debug(GOSSIP_SUPER_DEBUG, "orangefs_remount: called\n");

	new_op = op_alloc(ORANGEFS_VFS_OP_FS_MOUNT);
	if (!new_op)
		return -ENOMEM;
	strncpy(new_op->upcall.req.fs_mount.orangefs_config_server,
		ORANGEFS_SB(sb)->devname,
		ORANGEFS_MAX_SERVER_ADDR_LEN);

	gossip_debug(GOSSIP_SUPER_DEBUG,
		     "Attempting ORANGEFS Remount via host %s\n",
		     new_op->upcall.req.fs_mount.orangefs_config_server);

	/*
	 * we assume that the calling function has already acquire the
	 * request_mutex to prevent other operations from bypassing
	 * this one
	 */
	ret = service_operation(new_op, "orangefs_remount",
		ORANGEFS_OP_PRIORITY | ORANGEFS_OP_NO_SEMAPHORE);
	gossip_debug(GOSSIP_SUPER_DEBUG,
		     "orangefs_remount: mount got return value of %d\n",
		     ret);
	if (ret == 0) {
		/*
		 * store the id assigned to this sb -- it's just a
		 * short-lived mapping that the system interface uses
		 * to map this superblock to a particular mount entry
		 */
		ORANGEFS_SB(sb)->id = new_op->downcall.resp.fs_mount.id;
		ORANGEFS_SB(sb)->mount_pending = 0;
	}

	op_release(new_op);
	return ret;
}

int fsid_key_table_initialize(void)
{
	return 0;
}

void fsid_key_table_finalize(void)
{
}

/* Called whenever the VFS dirties the inode in response to atime updates */
static void orangefs_dirty_inode(struct inode *inode, int flags)
{
	struct orangefs_inode_s *orangefs_inode = ORANGEFS_I(inode);

	gossip_debug(GOSSIP_SUPER_DEBUG,
		     "orangefs_dirty_inode: %pU\n",
		     get_khandle_from_ino(inode));
	SetAtimeFlag(orangefs_inode);
}

static const struct super_operations orangefs_s_ops = {
	.alloc_inode = orangefs_alloc_inode,
	.destroy_inode = orangefs_destroy_inode,
	.dirty_inode = orangefs_dirty_inode,
	.drop_inode = generic_delete_inode,
	.statfs = orangefs_statfs,
	.remount_fs = orangefs_remount_fs,
	.show_options = generic_show_options,
};

static struct dentry *orangefs_fh_to_dentry(struct super_block *sb,
				  struct fid *fid,
				  int fh_len,
				  int fh_type)
{
	struct orangefs_object_kref refn;

	if (fh_len < 5 || fh_type > 2)
		return NULL;

	ORANGEFS_khandle_from(&(refn.khandle), fid->raw, 16);
	refn.fs_id = (u32) fid->raw[4];
	gossip_debug(GOSSIP_SUPER_DEBUG,
		     "fh_to_dentry: handle %pU, fs_id %d\n",
		     &refn.khandle,
		     refn.fs_id);

	return d_obtain_alias(orangefs_iget(sb, &refn));
}

static int orangefs_encode_fh(struct inode *inode,
		    __u32 *fh,
		    int *max_len,
		    struct inode *parent)
{
	int len = parent ? 10 : 5;
	int type = 1;
	struct orangefs_object_kref refn;

	if (*max_len < len) {
		gossip_lerr("fh buffer is too small for encoding\n");
		*max_len = len;
		type = 255;
		goto out;
	}

	refn = ORANGEFS_I(inode)->refn;
	ORANGEFS_khandle_to(&refn.khandle, fh, 16);
	fh[4] = refn.fs_id;

	gossip_debug(GOSSIP_SUPER_DEBUG,
		     "Encoding fh: handle %pU, fsid %u\n",
		     &refn.khandle,
		     refn.fs_id);


	if (parent) {
		refn = ORANGEFS_I(parent)->refn;
		ORANGEFS_khandle_to(&refn.khandle, (char *) fh + 20, 16);
		fh[9] = refn.fs_id;

		type = 2;
		gossip_debug(GOSSIP_SUPER_DEBUG,
			     "Encoding parent: handle %pU, fsid %u\n",
			     &refn.khandle,
			     refn.fs_id);
	}
	*max_len = len;

out:
	return type;
}

static struct export_operations orangefs_export_ops = {
	.encode_fh = orangefs_encode_fh,
	.fh_to_dentry = orangefs_fh_to_dentry,
};

static int orangefs_fill_sb(struct super_block *sb,
		struct orangefs_fs_mount_response *fs_mount,
		void *data, int silent)
{
	int ret = -EINVAL;
	struct inode *root = NULL;
	struct dentry *root_dentry = NULL;
	struct orangefs_object_kref root_object;

	/* alloc and init our private orangefs sb info */
	sb->s_fs_info =
		kzalloc(sizeof(struct orangefs_sb_info_s), ORANGEFS_GFP_FLAGS);
	if (!ORANGEFS_SB(sb))
		return -ENOMEM;
	ORANGEFS_SB(sb)->sb = sb;

	ORANGEFS_SB(sb)->root_khandle = fs_mount->root_khandle;
	ORANGEFS_SB(sb)->fs_id = fs_mount->fs_id;
	ORANGEFS_SB(sb)->id = fs_mount->id;

	if (data) {
		ret = parse_mount_options(sb, data, silent);
		if (ret)
			return ret;
	}

	/* Hang the xattr handlers off the superblock */
	sb->s_xattr = orangefs_xattr_handlers;
	sb->s_magic = ORANGEFS_SUPER_MAGIC;
	sb->s_op = &orangefs_s_ops;
	sb->s_d_op = &orangefs_dentry_operations;

	sb->s_blocksize = orangefs_bufmap_size_query();
	sb->s_blocksize_bits = orangefs_bufmap_shift_query();
	sb->s_maxbytes = MAX_LFS_FILESIZE;

	root_object.khandle = ORANGEFS_SB(sb)->root_khandle;
	root_object.fs_id = ORANGEFS_SB(sb)->fs_id;
	gossip_debug(GOSSIP_SUPER_DEBUG,
		     "get inode %pU, fsid %d\n",
		     &root_object.khandle,
		     root_object.fs_id);

	root = orangefs_iget(sb, &root_object);
	if (IS_ERR(root))
		return PTR_ERR(root);

	gossip_debug(GOSSIP_SUPER_DEBUG,
		     "Allocated root inode [%p] with mode %x\n",
		     root,
		     root->i_mode);

	/* allocates and places root dentry in dcache */
	root_dentry = d_make_root(root);
	if (!root_dentry)
		return -ENOMEM;

	sb->s_export_op = &orangefs_export_ops;
	sb->s_root = root_dentry;
	return 0;
}

struct dentry *orangefs_mount(struct file_system_type *fst,
			   int flags,
			   const char *devname,
			   void *data)
{
	int ret = -EINVAL;
	struct super_block *sb = ERR_PTR(-EINVAL);
	struct orangefs_kernel_op_s *new_op;
	struct dentry *d = ERR_PTR(-EINVAL);

	gossip_debug(GOSSIP_SUPER_DEBUG,
		     "orangefs_mount: called with devname %s\n",
		     devname);

	if (!devname) {
		gossip_err("ERROR: device name not specified.\n");
		return ERR_PTR(-EINVAL);
	}

	new_op = op_alloc(ORANGEFS_VFS_OP_FS_MOUNT);
	if (!new_op)
		return ERR_PTR(-ENOMEM);

	strncpy(new_op->upcall.req.fs_mount.orangefs_config_server,
		devname,
		ORANGEFS_MAX_SERVER_ADDR_LEN);

	gossip_debug(GOSSIP_SUPER_DEBUG,
		     "Attempting ORANGEFS Mount via host %s\n",
		     new_op->upcall.req.fs_mount.orangefs_config_server);

	ret = service_operation(new_op, "orangefs_mount", 0);
	gossip_debug(GOSSIP_SUPER_DEBUG,
		     "orangefs_mount: mount got return value of %d\n", ret);
	if (ret)
		goto free_op;

	if (new_op->downcall.resp.fs_mount.fs_id == ORANGEFS_FS_ID_NULL) {
		gossip_err("ERROR: Retrieved null fs_id\n");
		ret = -EINVAL;
		goto free_op;
	}

	sb = sget(fst, NULL, set_anon_super, flags, NULL);

	if (IS_ERR(sb)) {
		d = ERR_CAST(sb);
		goto free_op;
	}

	ret = orangefs_fill_sb(sb,
	      &new_op->downcall.resp.fs_mount, data,
	      flags & MS_SILENT ? 1 : 0);

	if (ret) {
		d = ERR_PTR(ret);
		goto free_op;
	}

	/*
	 * on successful mount, store the devname and data
	 * used
	 */
	strncpy(ORANGEFS_SB(sb)->devname,
		devname,
		ORANGEFS_MAX_SERVER_ADDR_LEN);

	/* mount_pending must be cleared */
	ORANGEFS_SB(sb)->mount_pending = 0;

	/*
	 * finally, add this sb to our list of known orangefs
	 * sb's
	 */
	add_orangefs_sb(sb);
	op_release(new_op);
	return dget(sb->s_root);

free_op:
	gossip_err("orangefs_mount: mount request failed with %d\n", ret);
	if (ret == -EINVAL) {
		gossip_err("Ensure that all orangefs-servers have the same FS configuration files\n");
		gossip_err("Look at pvfs2-client-core log file (typically /tmp/pvfs2-client.log) for more details\n");
	}

	op_release(new_op);

	return d;
}

void orangefs_kill_sb(struct super_block *sb)
{
	gossip_debug(GOSSIP_SUPER_DEBUG, "orangefs_kill_sb: called\n");

	/*
	 * issue the unmount to userspace to tell it to remove the
	 * dynamic mount info it has for this superblock
	 */
	orangefs_unmount_sb(sb);

	/* remove the sb from our list of orangefs specific sb's */
	remove_orangefs_sb(sb);

	/* provided sb cleanup */
	kill_anon_super(sb);

	/* free the orangefs superblock private data */
	kfree(ORANGEFS_SB(sb));
}

int orangefs_inode_cache_initialize(void)
{
	orangefs_inode_cache = kmem_cache_create("orangefs_inode_cache",
					      sizeof(struct orangefs_inode_s),
					      0,
					      ORANGEFS_CACHE_CREATE_FLAGS,
					      orangefs_inode_cache_ctor);

	if (!orangefs_inode_cache) {
		gossip_err("Cannot create orangefs_inode_cache\n");
		return -ENOMEM;
	}
	return 0;
}

int orangefs_inode_cache_finalize(void)
{
	kmem_cache_destroy(orangefs_inode_cache);
	return 0;
}
Commit	Line	Data
1182fca3 MM	1	/*
	2	* (C) 2001 Clemson University and The University of Chicago
	3	*
	4	* See COPYING in top-level directory.
	5	*/
	6
	7	#include "protocol.h"
	8	#include "pvfs2-kernel.h"
	9	#include "pvfs2-bufmap.h"
	10
	11	#include <linux/parser.h>
	12
8bb8aefd YL	13	/* a cache for orangefs-inode objects (i.e. orangefs inode private data) */
8bb8aefd YL	14	static struct kmem_cache *orangefs_inode_cache;
1182fca3	15
8bb8aefd YL	16	/* list for storing orangefs specific superblocks in use */
8bb8aefd YL	17	LIST_HEAD(orangefs_superblocks);
1182fca3	18
8bb8aefd	19	DEFINE_SPINLOCK(orangefs_superblocks_lock);
1182fca3 MM	20
	21	enum {
	22	Opt_intr,
	23	Opt_acl,
	24	Opt_local_lock,
	25
	26	Opt_err
	27	};
	28
	29	static const match_table_t tokens = {
	30	{ Opt_acl, "acl" },
	31	{ Opt_intr, "intr" },
	32	{ Opt_local_lock, "local_lock" },
	33	{ Opt_err, NULL }
	34	};
	35
	36
	37	static int parse_mount_options(struct super_block sb, char options,
	38	int silent)
	39	{
8bb8aefd	40	struct orangefs_sb_info_s *orangefs_sb = ORANGEFS_SB(sb);
1182fca3 MM	41	substring_t args[MAX_OPT_ARGS];
	42	char *p;
	43
	44	/*
	45	* Force any potential flags that might be set from the mount
	46	* to zero, ie, initialize to unset.
	47	*/
	48	sb->s_flags &= ~MS_POSIXACL;
8bb8aefd YL	49	orangefs_sb->flags &= ~ORANGEFS_OPT_INTR;
8bb8aefd YL	50	orangefs_sb->flags &= ~ORANGEFS_OPT_LOCAL_LOCK;
1182fca3 MM	51
	52	while ((p = strsep(&options, ",")) != NULL) {
	53	int token;
	54
	55	if (!*p)
	56	continue;
	57
	58	token = match_token(p, tokens, args);
	59	switch (token) {
	60	case Opt_acl:
	61	sb->s_flags \|= MS_POSIXACL;
	62	break;
	63	case Opt_intr:
8bb8aefd	64	orangefs_sb->flags \|= ORANGEFS_OPT_INTR;
1182fca3 MM	65	break;
1182fca3 MM	66	case Opt_local_lock:
8bb8aefd	67	orangefs_sb->flags \|= ORANGEFS_OPT_LOCAL_LOCK;
1182fca3 MM	68	break;
	69	default:
	70	goto fail;
	71	}
	72	}
	73
	74	return 0;
	75	fail:
	76	if (!silent)
	77	gossip_err("Error: mount option [%s] is not supported.\n", p);
	78	return -EINVAL;
	79	}
	80
8bb8aefd	81	static void orangefs_inode_cache_ctor(void *req)
1182fca3	82	{
8bb8aefd	83	struct orangefs_inode_s *orangefs_inode = req;
1182fca3	84
8bb8aefd YL	85	inode_init_once(&orangefs_inode->vfs_inode);
8bb8aefd YL	86	init_rwsem(&orangefs_inode->xattr_sem);
1182fca3	87
8bb8aefd	88	orangefs_inode->vfs_inode.i_version = 1;
1182fca3 MM	89	}
1182fca3 MM	90
8bb8aefd	91	static struct inode orangefs_alloc_inode(struct super_block sb)
1182fca3	92	{
8bb8aefd	93	struct orangefs_inode_s *orangefs_inode;
1182fca3	94
8bb8aefd YL	95	orangefs_inode = kmem_cache_alloc(orangefs_inode_cache,
	96	ORANGEFS_CACHE_ALLOC_FLAGS);
	97	if (orangefs_inode == NULL) {
	98	gossip_err("Failed to allocate orangefs_inode\n");
1182fca3 MM	99	return NULL;
	100	}
	101
	102	/*
	103	* We want to clear everything except for rw_semaphore and the
	104	* vfs_inode.
	105	*/
8bb8aefd YL	106	memset(&orangefs_inode->refn.khandle, 0, 16);
	107	orangefs_inode->refn.fs_id = ORANGEFS_FS_ID_NULL;
	108	orangefs_inode->last_failed_block_index_read = 0;
	109	memset(orangefs_inode->link_target, 0, sizeof(orangefs_inode->link_target));
	110	orangefs_inode->pinode_flags = 0;
1182fca3 MM	111
1182fca3 MM	112	gossip_debug(GOSSIP_SUPER_DEBUG,
8bb8aefd YL	113	"orangefs_alloc_inode: allocated %p\n",
	114	&orangefs_inode->vfs_inode);
	115	return &orangefs_inode->vfs_inode;
1182fca3 MM	116	}
1182fca3 MM	117
8bb8aefd	118	static void orangefs_destroy_inode(struct inode *inode)
1182fca3	119	{
8bb8aefd	120	struct orangefs_inode_s *orangefs_inode = ORANGEFS_I(inode);
1182fca3 MM	121
	122	gossip_debug(GOSSIP_SUPER_DEBUG,
	123	"%s: deallocated %p destroying inode %pU\n",
8bb8aefd	124	__func__, orangefs_inode, get_khandle_from_ino(inode));
1182fca3	125
8bb8aefd	126	kmem_cache_free(orangefs_inode_cache, orangefs_inode);
1182fca3 MM	127	}
	128
	129	/*
	130	* NOTE: information filled in here is typically reflected in the
	131	* output of the system command 'df'
	132	*/
8bb8aefd	133	static int orangefs_statfs(struct dentry dentry, struct kstatfs buf)
1182fca3 MM	134	{
1182fca3 MM	135	int ret = -ENOMEM;
8bb8aefd	136	struct orangefs_kernel_op_s *new_op = NULL;
1182fca3 MM	137	int flags = 0;
	138	struct super_block *sb = NULL;
	139
	140	sb = dentry->d_sb;
	141
	142	gossip_debug(GOSSIP_SUPER_DEBUG,
8bb8aefd	143	"orangefs_statfs: called on sb %p (fs_id is %d)\n",
1182fca3	144	sb,
8bb8aefd	145	(int)(ORANGEFS_SB(sb)->fs_id));
1182fca3	146
8bb8aefd	147	new_op = op_alloc(ORANGEFS_VFS_OP_STATFS);
1182fca3 MM	148	if (!new_op)
1182fca3 MM	149	return ret;
8bb8aefd	150	new_op->upcall.req.statfs.fs_id = ORANGEFS_SB(sb)->fs_id;
1182fca3	151
8bb8aefd YL	152	if (ORANGEFS_SB(sb)->flags & ORANGEFS_OPT_INTR)
8bb8aefd YL	153	flags = ORANGEFS_OP_INTERRUPTIBLE;
1182fca3	154
8bb8aefd	155	ret = service_operation(new_op, "orangefs_statfs", flags);
1182fca3 MM	156
	157	if (new_op->downcall.status < 0)
	158	goto out_op_release;
	159
	160	gossip_debug(GOSSIP_SUPER_DEBUG,
8bb8aefd	161	"orangefs_statfs: got %ld blocks available \| "
1182fca3 MM	162	"%ld blocks total \| %ld block size\n",
	163	(long)new_op->downcall.resp.statfs.blocks_avail,
	164	(long)new_op->downcall.resp.statfs.blocks_total,
	165	(long)new_op->downcall.resp.statfs.block_size);
	166
	167	buf->f_type = sb->s_magic;
8bb8aefd	168	memcpy(&buf->f_fsid, &ORANGEFS_SB(sb)->fs_id, sizeof(buf->f_fsid));
1182fca3	169	buf->f_bsize = new_op->downcall.resp.statfs.block_size;
8bb8aefd	170	buf->f_namelen = ORANGEFS_NAME_LEN;
1182fca3 MM	171
	172	buf->f_blocks = (sector_t) new_op->downcall.resp.statfs.blocks_total;
	173	buf->f_bfree = (sector_t) new_op->downcall.resp.statfs.blocks_avail;
	174	buf->f_bavail = (sector_t) new_op->downcall.resp.statfs.blocks_avail;
	175	buf->f_files = (sector_t) new_op->downcall.resp.statfs.files_total;
	176	buf->f_ffree = (sector_t) new_op->downcall.resp.statfs.files_avail;
	177	buf->f_frsize = sb->s_blocksize;
	178
	179	out_op_release:
	180	op_release(new_op);
8bb8aefd	181	gossip_debug(GOSSIP_SUPER_DEBUG, "orangefs_statfs: returning %d\n", ret);
1182fca3 MM	182	return ret;
	183	}
	184
	185	/*
	186	* Remount as initiated by VFS layer. We just need to reparse the mount
	187	* options, no need to signal pvfs2-client-core about it.
	188	*/
8bb8aefd	189	static int orangefs_remount_fs(struct super_block sb, int flags, char *data)
1182fca3	190	{
8bb8aefd	191	gossip_debug(GOSSIP_SUPER_DEBUG, "orangefs_remount_fs: called\n");
1182fca3 MM	192	return parse_mount_options(sb, data, 1);
	193	}
	194
	195	/*
	196	* Remount as initiated by pvfs2-client-core on restart. This is used to
	197	* repopulate mount information left from previous pvfs2-client-core.
	198	*
	199	* the idea here is that given a valid superblock, we're
	200	* re-initializing the user space client with the initial mount
	201	* information specified when the super block was first initialized.
	202	* this is very different than the first initialization/creation of a
	203	* superblock. we use the special service_priority_operation to make
	204	* sure that the mount gets ahead of any other pending operation that
	205	* is waiting for servicing. this means that the pvfs2-client won't
	206	* fail to start several times for all other pending operations before
	207	* the client regains all of the mount information from us.
	208	* NOTE: this function assumes that the request_mutex is already acquired!
	209	*/
8bb8aefd	210	int orangefs_remount(struct super_block *sb)
1182fca3	211	{
8bb8aefd	212	struct orangefs_kernel_op_s *new_op;
1182fca3 MM	213	int ret = -EINVAL;
1182fca3 MM	214
8bb8aefd	215	gossip_debug(GOSSIP_SUPER_DEBUG, "orangefs_remount: called\n");
1182fca3	216
8bb8aefd	217	new_op = op_alloc(ORANGEFS_VFS_OP_FS_MOUNT);
1182fca3 MM	218	if (!new_op)
1182fca3 MM	219	return -ENOMEM;
8bb8aefd YL	220	strncpy(new_op->upcall.req.fs_mount.orangefs_config_server,
	221	ORANGEFS_SB(sb)->devname,
	222	ORANGEFS_MAX_SERVER_ADDR_LEN);
1182fca3 MM	223
1182fca3 MM	224	gossip_debug(GOSSIP_SUPER_DEBUG,
8bb8aefd YL	225	"Attempting ORANGEFS Remount via host %s\n",
8bb8aefd YL	226	new_op->upcall.req.fs_mount.orangefs_config_server);
1182fca3 MM	227
	228	/*
	229	* we assume that the calling function has already acquire the
	230	* request_mutex to prevent other operations from bypassing
	231	* this one
	232	*/
8bb8aefd YL	233	ret = service_operation(new_op, "orangefs_remount",
8bb8aefd YL	234	ORANGEFS_OP_PRIORITY \| ORANGEFS_OP_NO_SEMAPHORE);
1182fca3	235	gossip_debug(GOSSIP_SUPER_DEBUG,
8bb8aefd	236	"orangefs_remount: mount got return value of %d\n",
1182fca3 MM	237	ret);
	238	if (ret == 0) {
	239	/*
	240	* store the id assigned to this sb -- it's just a
	241	* short-lived mapping that the system interface uses
	242	* to map this superblock to a particular mount entry
	243	*/
8bb8aefd YL	244	ORANGEFS_SB(sb)->id = new_op->downcall.resp.fs_mount.id;
8bb8aefd YL	245	ORANGEFS_SB(sb)->mount_pending = 0;
1182fca3 MM	246	}
	247
	248	op_release(new_op);
	249	return ret;
	250	}
	251
	252	int fsid_key_table_initialize(void)
	253	{
	254	return 0;
	255	}
	256
	257	void fsid_key_table_finalize(void)
	258	{
	259	}
	260
	261	/* Called whenever the VFS dirties the inode in response to atime updates */
8bb8aefd	262	static void orangefs_dirty_inode(struct inode *inode, int flags)
1182fca3	263	{
8bb8aefd	264	struct orangefs_inode_s *orangefs_inode = ORANGEFS_I(inode);
1182fca3 MM	265
1182fca3 MM	266	gossip_debug(GOSSIP_SUPER_DEBUG,
8bb8aefd	267	"orangefs_dirty_inode: %pU\n",
1182fca3	268	get_khandle_from_ino(inode));
8bb8aefd	269	SetAtimeFlag(orangefs_inode);
1182fca3 MM	270	}
1182fca3 MM	271
8bb8aefd YL	272	static const struct super_operations orangefs_s_ops = {
	273	.alloc_inode = orangefs_alloc_inode,
	274	.destroy_inode = orangefs_destroy_inode,
	275	.dirty_inode = orangefs_dirty_inode,
1182fca3	276	.drop_inode = generic_delete_inode,
8bb8aefd YL	277	.statfs = orangefs_statfs,
8bb8aefd YL	278	.remount_fs = orangefs_remount_fs,
1182fca3 MM	279	.show_options = generic_show_options,
	280	};
	281
8bb8aefd	282	static struct dentry orangefs_fh_to_dentry(struct super_block sb,
1182fca3 MM	283	struct fid *fid,
	284	int fh_len,
	285	int fh_type)
	286	{
8bb8aefd	287	struct orangefs_object_kref refn;
1182fca3 MM	288
	289	if (fh_len < 5 \|\| fh_type > 2)
	290	return NULL;
	291
8bb8aefd	292	ORANGEFS_khandle_from(&(refn.khandle), fid->raw, 16);
1182fca3 MM	293	refn.fs_id = (u32) fid->raw[4];
	294	gossip_debug(GOSSIP_SUPER_DEBUG,
	295	"fh_to_dentry: handle %pU, fs_id %d\n",
	296	&refn.khandle,
	297	refn.fs_id);
	298
8bb8aefd	299	return d_obtain_alias(orangefs_iget(sb, &refn));
1182fca3 MM	300	}
1182fca3 MM	301
8bb8aefd	302	static int orangefs_encode_fh(struct inode *inode,
1182fca3 MM	303	__u32 *fh,
	304	int *max_len,
	305	struct inode *parent)
	306	{
	307	int len = parent ? 10 : 5;
	308	int type = 1;
8bb8aefd	309	struct orangefs_object_kref refn;
1182fca3 MM	310
	311	if (*max_len < len) {
	312	gossip_lerr("fh buffer is too small for encoding\n");
	313	*max_len = len;
	314	type = 255;
	315	goto out;
	316	}
	317
8bb8aefd YL	318	refn = ORANGEFS_I(inode)->refn;
8bb8aefd YL	319	ORANGEFS_khandle_to(&refn.khandle, fh, 16);
1182fca3 MM	320	fh[4] = refn.fs_id;
	321
	322	gossip_debug(GOSSIP_SUPER_DEBUG,
	323	"Encoding fh: handle %pU, fsid %u\n",
	324	&refn.khandle,
	325	refn.fs_id);
	326
	327
	328	if (parent) {
8bb8aefd YL	329	refn = ORANGEFS_I(parent)->refn;
8bb8aefd YL	330	ORANGEFS_khandle_to(&refn.khandle, (char *) fh + 20, 16);
1182fca3 MM	331	fh[9] = refn.fs_id;
	332
	333	type = 2;
	334	gossip_debug(GOSSIP_SUPER_DEBUG,
	335	"Encoding parent: handle %pU, fsid %u\n",
	336	&refn.khandle,
	337	refn.fs_id);
	338	}
	339	*max_len = len;
	340
	341	out:
	342	return type;
	343	}
	344
8bb8aefd YL	345	static struct export_operations orangefs_export_ops = {
	346	.encode_fh = orangefs_encode_fh,
	347	.fh_to_dentry = orangefs_fh_to_dentry,
1182fca3 MM	348	};
1182fca3 MM	349
8bb8aefd YL	350	static int orangefs_fill_sb(struct super_block *sb,
8bb8aefd YL	351	struct orangefs_fs_mount_response *fs_mount,
5c0dbbc6	352	void *data, int silent)
1182fca3 MM	353	{
	354	int ret = -EINVAL;
	355	struct inode *root = NULL;
	356	struct dentry *root_dentry = NULL;
8bb8aefd	357	struct orangefs_object_kref root_object;
1182fca3	358
8bb8aefd	359	/* alloc and init our private orangefs sb info */
1182fca3	360	sb->s_fs_info =
8bb8aefd YL	361	kzalloc(sizeof(struct orangefs_sb_info_s), ORANGEFS_GFP_FLAGS);
8bb8aefd YL	362	if (!ORANGEFS_SB(sb))
1182fca3	363	return -ENOMEM;
8bb8aefd	364	ORANGEFS_SB(sb)->sb = sb;
1182fca3	365
8bb8aefd YL	366	ORANGEFS_SB(sb)->root_khandle = fs_mount->root_khandle;
	367	ORANGEFS_SB(sb)->fs_id = fs_mount->fs_id;
	368	ORANGEFS_SB(sb)->id = fs_mount->id;
1182fca3	369
5c0dbbc6 AV	370	if (data) {
5c0dbbc6 AV	371	ret = parse_mount_options(sb, data, silent);
1182fca3 MM	372	if (ret)
	373	return ret;
	374	}
	375
	376	/* Hang the xattr handlers off the superblock */
8bb8aefd YL	377	sb->s_xattr = orangefs_xattr_handlers;
	378	sb->s_magic = ORANGEFS_SUPER_MAGIC;
	379	sb->s_op = &orangefs_s_ops;
	380	sb->s_d_op = &orangefs_dentry_operations;
1182fca3	381
8bb8aefd YL	382	sb->s_blocksize = orangefs_bufmap_size_query();
8bb8aefd YL	383	sb->s_blocksize_bits = orangefs_bufmap_shift_query();
1182fca3 MM	384	sb->s_maxbytes = MAX_LFS_FILESIZE;
1182fca3 MM	385
8bb8aefd YL	386	root_object.khandle = ORANGEFS_SB(sb)->root_khandle;
8bb8aefd YL	387	root_object.fs_id = ORANGEFS_SB(sb)->fs_id;
1182fca3 MM	388	gossip_debug(GOSSIP_SUPER_DEBUG,
	389	"get inode %pU, fsid %d\n",
	390	&root_object.khandle,
	391	root_object.fs_id);
	392
8bb8aefd	393	root = orangefs_iget(sb, &root_object);
1182fca3 MM	394	if (IS_ERR(root))
	395	return PTR_ERR(root);
	396
	397	gossip_debug(GOSSIP_SUPER_DEBUG,
	398	"Allocated root inode [%p] with mode %x\n",
	399	root,
	400	root->i_mode);
	401
	402	/* allocates and places root dentry in dcache */
	403	root_dentry = d_make_root(root);
b05a7851	404	if (!root_dentry)
1182fca3	405	return -ENOMEM;
1182fca3	406
8bb8aefd	407	sb->s_export_op = &orangefs_export_ops;
1182fca3 MM	408	sb->s_root = root_dentry;
	409	return 0;
	410	}
	411
8bb8aefd	412	struct dentry orangefs_mount(struct file_system_type fst,
1182fca3 MM	413	int flags,
	414	const char *devname,
	415	void *data)
	416	{
	417	int ret = -EINVAL;
	418	struct super_block *sb = ERR_PTR(-EINVAL);
8bb8aefd	419	struct orangefs_kernel_op_s *new_op;
1be21f86	420	struct dentry *d = ERR_PTR(-EINVAL);
1182fca3 MM	421
1182fca3 MM	422	gossip_debug(GOSSIP_SUPER_DEBUG,
8bb8aefd	423	"orangefs_mount: called with devname %s\n",
1182fca3 MM	424	devname);
	425
	426	if (!devname) {
	427	gossip_err("ERROR: device name not specified.\n");
	428	return ERR_PTR(-EINVAL);
	429	}
	430
8bb8aefd	431	new_op = op_alloc(ORANGEFS_VFS_OP_FS_MOUNT);
1182fca3 MM	432	if (!new_op)
	433	return ERR_PTR(-ENOMEM);
	434
8bb8aefd	435	strncpy(new_op->upcall.req.fs_mount.orangefs_config_server,
1182fca3	436	devname,
8bb8aefd	437	ORANGEFS_MAX_SERVER_ADDR_LEN);
1182fca3 MM	438
1182fca3 MM	439	gossip_debug(GOSSIP_SUPER_DEBUG,
8bb8aefd YL	440	"Attempting ORANGEFS Mount via host %s\n",
8bb8aefd YL	441	new_op->upcall.req.fs_mount.orangefs_config_server);
1182fca3	442
8bb8aefd	443	ret = service_operation(new_op, "orangefs_mount", 0);
1182fca3	444	gossip_debug(GOSSIP_SUPER_DEBUG,
8bb8aefd	445	"orangefs_mount: mount got return value of %d\n", ret);
1182fca3 MM	446	if (ret)
	447	goto free_op;
	448
8bb8aefd	449	if (new_op->downcall.resp.fs_mount.fs_id == ORANGEFS_FS_ID_NULL) {
1182fca3 MM	450	gossip_err("ERROR: Retrieved null fs_id\n");
	451	ret = -EINVAL;
	452	goto free_op;
	453	}
	454
1be21f86 MM	455	sb = sget(fst, NULL, set_anon_super, flags, NULL);
	456
	457	if (IS_ERR(sb)) {
	458	d = ERR_CAST(sb);
1182fca3 MM	459	goto free_op;
	460	}
	461
8bb8aefd	462	ret = orangefs_fill_sb(sb,
5c0dbbc6	463	&new_op->downcall.resp.fs_mount, data,
1be21f86 MM	464	flags & MS_SILENT ? 1 : 0);
	465
	466	if (ret) {
	467	d = ERR_PTR(ret);
	468	goto free_op;
	469	}
1182fca3 MM	470
	471	/*
	472	* on successful mount, store the devname and data
	473	* used
	474	*/
8bb8aefd	475	strncpy(ORANGEFS_SB(sb)->devname,
1182fca3	476	devname,
8bb8aefd	477	ORANGEFS_MAX_SERVER_ADDR_LEN);
1182fca3 MM	478
1182fca3 MM	479	/* mount_pending must be cleared */
8bb8aefd	480	ORANGEFS_SB(sb)->mount_pending = 0;
1182fca3 MM	481
1182fca3 MM	482	/*
8bb8aefd	483	* finally, add this sb to our list of known orangefs
1182fca3 MM	484	* sb's
1182fca3 MM	485	*/
8bb8aefd	486	add_orangefs_sb(sb);
1182fca3	487	op_release(new_op);
1be21f86	488	return dget(sb->s_root);
1182fca3 MM	489
1182fca3 MM	490	free_op:
8bb8aefd	491	gossip_err("orangefs_mount: mount request failed with %d\n", ret);
1182fca3	492	if (ret == -EINVAL) {
8bb8aefd	493	gossip_err("Ensure that all orangefs-servers have the same FS configuration files\n");
1182fca3 MM	494	gossip_err("Look at pvfs2-client-core log file (typically /tmp/pvfs2-client.log) for more details\n");
	495	}
	496
	497	op_release(new_op);
	498
1be21f86	499	return d;
1182fca3 MM	500	}
1182fca3 MM	501
8bb8aefd	502	void orangefs_kill_sb(struct super_block *sb)
1182fca3	503	{
8bb8aefd	504	gossip_debug(GOSSIP_SUPER_DEBUG, "orangefs_kill_sb: called\n");
1182fca3 MM	505
	506	/*
	507	* issue the unmount to userspace to tell it to remove the
	508	* dynamic mount info it has for this superblock
	509	*/
8bb8aefd	510	orangefs_unmount_sb(sb);
1182fca3	511
8bb8aefd YL	512	/* remove the sb from our list of orangefs specific sb's */
8bb8aefd YL	513	remove_orangefs_sb(sb);
1182fca3 MM	514
	515	/* provided sb cleanup */
	516	kill_anon_super(sb);
	517
8bb8aefd YL	518	/* free the orangefs superblock private data */
8bb8aefd YL	519	kfree(ORANGEFS_SB(sb));
1182fca3 MM	520	}
1182fca3 MM	521
8bb8aefd	522	int orangefs_inode_cache_initialize(void)
1182fca3	523	{
8bb8aefd YL	524	orangefs_inode_cache = kmem_cache_create("orangefs_inode_cache",
8bb8aefd YL	525	sizeof(struct orangefs_inode_s),
1182fca3	526	0,
8bb8aefd YL	527	ORANGEFS_CACHE_CREATE_FLAGS,
8bb8aefd YL	528	orangefs_inode_cache_ctor);
1182fca3	529
8bb8aefd YL	530	if (!orangefs_inode_cache) {
8bb8aefd YL	531	gossip_err("Cannot create orangefs_inode_cache\n");
1182fca3 MM	532	return -ENOMEM;
	533	}
	534	return 0;
	535	}
	536
8bb8aefd	537	int orangefs_inode_cache_finalize(void)
1182fca3	538	{
8bb8aefd	539	kmem_cache_destroy(orangefs_inode_cache);
1182fca3 MM	540	return 0;
1182fca3 MM	541	}