[linux.git] / fs / orangefs / super.c

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

#include "protocol.h"
#include "orangefs-kernel.h"
#include "orangefs-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, GFP_KERNEL);
	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,
		     "%s: got %ld blocks available | "
		     "%ld blocks total | %ld block size | "
		     "%ld files total | %ld files avail\n",
		     __func__,
		     (long)new_op->downcall.resp.statfs.blocks_avail,
		     (long)new_op->downcall.resp.statfs.blocks_total,
		     (long)new_op->downcall.resp.statfs.block_size,
		     (long)new_op->downcall.resp.statfs.files_total,
		     (long)new_op->downcall.resp.statfs.files_avail);

	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_MAX;

	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 orangefs_sb_info_s *orangefs_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->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 acquired the
	 * request_mutex to prevent other operations from bypassing
	 * this one
	 */
	ret = service_operation(new_op, "orangefs_remount",
		ORANGEFS_OP_PRIORITY | ORANGEFS_OP_NO_MUTEX);
	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->id = new_op->downcall.resp.fs_mount.id;
		orangefs_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 const 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), GFP_KERNEL);
	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
	 */
	gossip_debug(GOSSIP_SUPER_DEBUG,
		     "Adding SB %p to orangefs superblocks\n",
		     ORANGEFS_SB(sb));
	spin_lock(&orangefs_superblocks_lock);
	list_add_tail(&ORANGEFS_SB(sb)->list, &orangefs_superblocks);
	spin_unlock(&orangefs_superblocks_lock);
	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");

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

	/*
	 * 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 */

	spin_lock(&orangefs_superblocks_lock);
	__list_del_entry(&ORANGEFS_SB(sb)->list);	/* not list_del_init */
	ORANGEFS_SB(sb)->list.prev = NULL;
	spin_unlock(&orangefs_superblocks_lock);

	/*
	 * make sure that ORANGEFS_DEV_REMOUNT_ALL loop that might've seen us
	 * gets completed before we free the dang thing.
	 */
	mutex_lock(&request_mutex);
	mutex_unlock(&request_mutex);

	/* 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"
575e9461 MM	8	#include "orangefs-kernel.h"
575e9461 MM	9	#include "orangefs-bufmap.h"
1182fca3 MM	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
2d4cae0d	95	orangefs_inode = kmem_cache_alloc(orangefs_inode_cache, GFP_KERNEL);
8bb8aefd YL	96	if (orangefs_inode == NULL) {
8bb8aefd YL	97	gossip_err("Failed to allocate orangefs_inode\n");
1182fca3 MM	98	return NULL;
	99	}
	100
	101	/*
	102	* We want to clear everything except for rw_semaphore and the
	103	* vfs_inode.
	104	*/
8bb8aefd YL	105	memset(&orangefs_inode->refn.khandle, 0, 16);
	106	orangefs_inode->refn.fs_id = ORANGEFS_FS_ID_NULL;
	107	orangefs_inode->last_failed_block_index_read = 0;
	108	memset(orangefs_inode->link_target, 0, sizeof(orangefs_inode->link_target));
	109	orangefs_inode->pinode_flags = 0;
1182fca3 MM	110
1182fca3 MM	111	gossip_debug(GOSSIP_SUPER_DEBUG,
8bb8aefd YL	112	"orangefs_alloc_inode: allocated %p\n",
	113	&orangefs_inode->vfs_inode);
	114	return &orangefs_inode->vfs_inode;
1182fca3 MM	115	}
1182fca3 MM	116
8bb8aefd	117	static void orangefs_destroy_inode(struct inode *inode)
1182fca3	118	{
8bb8aefd	119	struct orangefs_inode_s *orangefs_inode = ORANGEFS_I(inode);
1182fca3 MM	120
	121	gossip_debug(GOSSIP_SUPER_DEBUG,
	122	"%s: deallocated %p destroying inode %pU\n",
8bb8aefd	123	__func__, orangefs_inode, get_khandle_from_ino(inode));
1182fca3	124
8bb8aefd	125	kmem_cache_free(orangefs_inode_cache, orangefs_inode);
1182fca3 MM	126	}
	127
	128	/*
	129	* NOTE: information filled in here is typically reflected in the
	130	* output of the system command 'df'
	131	*/
8bb8aefd	132	static int orangefs_statfs(struct dentry dentry, struct kstatfs buf)
1182fca3 MM	133	{
1182fca3 MM	134	int ret = -ENOMEM;
8bb8aefd	135	struct orangefs_kernel_op_s *new_op = NULL;
1182fca3 MM	136	int flags = 0;
	137	struct super_block *sb = NULL;
	138
	139	sb = dentry->d_sb;
	140
	141	gossip_debug(GOSSIP_SUPER_DEBUG,
8bb8aefd	142	"orangefs_statfs: called on sb %p (fs_id is %d)\n",
1182fca3	143	sb,
8bb8aefd	144	(int)(ORANGEFS_SB(sb)->fs_id));
1182fca3	145
8bb8aefd	146	new_op = op_alloc(ORANGEFS_VFS_OP_STATFS);
1182fca3 MM	147	if (!new_op)
1182fca3 MM	148	return ret;
8bb8aefd	149	new_op->upcall.req.statfs.fs_id = ORANGEFS_SB(sb)->fs_id;
1182fca3	150
8bb8aefd YL	151	if (ORANGEFS_SB(sb)->flags & ORANGEFS_OPT_INTR)
8bb8aefd YL	152	flags = ORANGEFS_OP_INTERRUPTIBLE;
1182fca3	153
8bb8aefd	154	ret = service_operation(new_op, "orangefs_statfs", flags);
1182fca3 MM	155
	156	if (new_op->downcall.status < 0)
	157	goto out_op_release;
	158
	159	gossip_debug(GOSSIP_SUPER_DEBUG,
be57366e MM	160	"%s: got %ld blocks available \| "
	161	"%ld blocks total \| %ld block size \| "
	162	"%ld files total \| %ld files avail\n",
	163	__func__,
1182fca3 MM	164	(long)new_op->downcall.resp.statfs.blocks_avail,
1182fca3 MM	165	(long)new_op->downcall.resp.statfs.blocks_total,
be57366e MM	166	(long)new_op->downcall.resp.statfs.block_size,
	167	(long)new_op->downcall.resp.statfs.files_total,
	168	(long)new_op->downcall.resp.statfs.files_avail);
1182fca3 MM	169
1182fca3 MM	170	buf->f_type = sb->s_magic;
8bb8aefd	171	memcpy(&buf->f_fsid, &ORANGEFS_SB(sb)->fs_id, sizeof(buf->f_fsid));
1182fca3	172	buf->f_bsize = new_op->downcall.resp.statfs.block_size;
47b4948f	173	buf->f_namelen = ORANGEFS_NAME_MAX;
1182fca3 MM	174
	175	buf->f_blocks = (sector_t) new_op->downcall.resp.statfs.blocks_total;
	176	buf->f_bfree = (sector_t) new_op->downcall.resp.statfs.blocks_avail;
	177	buf->f_bavail = (sector_t) new_op->downcall.resp.statfs.blocks_avail;
	178	buf->f_files = (sector_t) new_op->downcall.resp.statfs.files_total;
	179	buf->f_ffree = (sector_t) new_op->downcall.resp.statfs.files_avail;
	180	buf->f_frsize = sb->s_blocksize;
	181
	182	out_op_release:
	183	op_release(new_op);
8bb8aefd	184	gossip_debug(GOSSIP_SUPER_DEBUG, "orangefs_statfs: returning %d\n", ret);
1182fca3 MM	185	return ret;
	186	}
	187
	188	/*
	189	* Remount as initiated by VFS layer. We just need to reparse the mount
	190	* options, no need to signal pvfs2-client-core about it.
	191	*/
8bb8aefd	192	static int orangefs_remount_fs(struct super_block sb, int flags, char *data)
1182fca3	193	{
8bb8aefd	194	gossip_debug(GOSSIP_SUPER_DEBUG, "orangefs_remount_fs: called\n");
1182fca3 MM	195	return parse_mount_options(sb, data, 1);
	196	}
	197
	198	/*
	199	* Remount as initiated by pvfs2-client-core on restart. This is used to
	200	* repopulate mount information left from previous pvfs2-client-core.
	201	*
	202	* the idea here is that given a valid superblock, we're
	203	* re-initializing the user space client with the initial mount
	204	* information specified when the super block was first initialized.
	205	* this is very different than the first initialization/creation of a
	206	* superblock. we use the special service_priority_operation to make
	207	* sure that the mount gets ahead of any other pending operation that
	208	* is waiting for servicing. this means that the pvfs2-client won't
	209	* fail to start several times for all other pending operations before
	210	* the client regains all of the mount information from us.
	211	* NOTE: this function assumes that the request_mutex is already acquired!
	212	*/
45996492	213	int orangefs_remount(struct orangefs_sb_info_s *orangefs_sb)
1182fca3	214	{
8bb8aefd	215	struct orangefs_kernel_op_s *new_op;
1182fca3 MM	216	int ret = -EINVAL;
1182fca3 MM	217
8bb8aefd	218	gossip_debug(GOSSIP_SUPER_DEBUG, "orangefs_remount: called\n");
1182fca3	219
8bb8aefd	220	new_op = op_alloc(ORANGEFS_VFS_OP_FS_MOUNT);
1182fca3 MM	221	if (!new_op)
1182fca3 MM	222	return -ENOMEM;
8bb8aefd	223	strncpy(new_op->upcall.req.fs_mount.orangefs_config_server,
45996492	224	orangefs_sb->devname,
8bb8aefd	225	ORANGEFS_MAX_SERVER_ADDR_LEN);
1182fca3 MM	226
1182fca3 MM	227	gossip_debug(GOSSIP_SUPER_DEBUG,
8bb8aefd YL	228	"Attempting ORANGEFS Remount via host %s\n",
8bb8aefd YL	229	new_op->upcall.req.fs_mount.orangefs_config_server);
1182fca3 MM	230
1182fca3 MM	231	/*
adcf34a2	232	* we assume that the calling function has already acquired the
1182fca3 MM	233	* request_mutex to prevent other operations from bypassing
	234	* this one
	235	*/
8bb8aefd	236	ret = service_operation(new_op, "orangefs_remount",
adcf34a2	237	ORANGEFS_OP_PRIORITY \| ORANGEFS_OP_NO_MUTEX);
1182fca3	238	gossip_debug(GOSSIP_SUPER_DEBUG,
8bb8aefd	239	"orangefs_remount: mount got return value of %d\n",
1182fca3 MM	240	ret);
	241	if (ret == 0) {
	242	/*
	243	* store the id assigned to this sb -- it's just a
	244	* short-lived mapping that the system interface uses
	245	* to map this superblock to a particular mount entry
	246	*/
45996492 AV	247	orangefs_sb->id = new_op->downcall.resp.fs_mount.id;
45996492 AV	248	orangefs_sb->mount_pending = 0;
1182fca3 MM	249	}
	250
	251	op_release(new_op);
	252	return ret;
	253	}
	254
	255	int fsid_key_table_initialize(void)
	256	{
	257	return 0;
	258	}
	259
	260	void fsid_key_table_finalize(void)
	261	{
	262	}
	263
	264	/* Called whenever the VFS dirties the inode in response to atime updates */
8bb8aefd	265	static void orangefs_dirty_inode(struct inode *inode, int flags)
1182fca3	266	{
8bb8aefd	267	struct orangefs_inode_s *orangefs_inode = ORANGEFS_I(inode);
1182fca3 MM	268
1182fca3 MM	269	gossip_debug(GOSSIP_SUPER_DEBUG,
8bb8aefd	270	"orangefs_dirty_inode: %pU\n",
1182fca3	271	get_khandle_from_ino(inode));
8bb8aefd	272	SetAtimeFlag(orangefs_inode);
1182fca3 MM	273	}
1182fca3 MM	274
8bb8aefd YL	275	static const struct super_operations orangefs_s_ops = {
	276	.alloc_inode = orangefs_alloc_inode,
	277	.destroy_inode = orangefs_destroy_inode,
	278	.dirty_inode = orangefs_dirty_inode,
1182fca3	279	.drop_inode = generic_delete_inode,
8bb8aefd YL	280	.statfs = orangefs_statfs,
8bb8aefd YL	281	.remount_fs = orangefs_remount_fs,
1182fca3 MM	282	.show_options = generic_show_options,
	283	};
	284
8bb8aefd	285	static struct dentry orangefs_fh_to_dentry(struct super_block sb,
1182fca3 MM	286	struct fid *fid,
	287	int fh_len,
	288	int fh_type)
	289	{
8bb8aefd	290	struct orangefs_object_kref refn;
1182fca3 MM	291
	292	if (fh_len < 5 \|\| fh_type > 2)
	293	return NULL;
	294
8bb8aefd	295	ORANGEFS_khandle_from(&(refn.khandle), fid->raw, 16);
1182fca3 MM	296	refn.fs_id = (u32) fid->raw[4];
	297	gossip_debug(GOSSIP_SUPER_DEBUG,
	298	"fh_to_dentry: handle %pU, fs_id %d\n",
	299	&refn.khandle,
	300	refn.fs_id);
	301
8bb8aefd	302	return d_obtain_alias(orangefs_iget(sb, &refn));
1182fca3 MM	303	}
1182fca3 MM	304
8bb8aefd	305	static int orangefs_encode_fh(struct inode *inode,
1182fca3 MM	306	__u32 *fh,
	307	int *max_len,
	308	struct inode *parent)
	309	{
	310	int len = parent ? 10 : 5;
	311	int type = 1;
8bb8aefd	312	struct orangefs_object_kref refn;
1182fca3 MM	313
	314	if (*max_len < len) {
	315	gossip_lerr("fh buffer is too small for encoding\n");
	316	*max_len = len;
	317	type = 255;
	318	goto out;
	319	}
	320
8bb8aefd YL	321	refn = ORANGEFS_I(inode)->refn;
8bb8aefd YL	322	ORANGEFS_khandle_to(&refn.khandle, fh, 16);
1182fca3 MM	323	fh[4] = refn.fs_id;
	324
	325	gossip_debug(GOSSIP_SUPER_DEBUG,
	326	"Encoding fh: handle %pU, fsid %u\n",
	327	&refn.khandle,
	328	refn.fs_id);
	329
	330
	331	if (parent) {
8bb8aefd YL	332	refn = ORANGEFS_I(parent)->refn;
8bb8aefd YL	333	ORANGEFS_khandle_to(&refn.khandle, (char *) fh + 20, 16);
1182fca3 MM	334	fh[9] = refn.fs_id;
	335
	336	type = 2;
	337	gossip_debug(GOSSIP_SUPER_DEBUG,
	338	"Encoding parent: handle %pU, fsid %u\n",
	339	&refn.khandle,
	340	refn.fs_id);
	341	}
	342	*max_len = len;
	343
	344	out:
	345	return type;
	346	}
	347
acaca36d	348	static const struct export_operations orangefs_export_ops = {
8bb8aefd YL	349	.encode_fh = orangefs_encode_fh,
8bb8aefd YL	350	.fh_to_dentry = orangefs_fh_to_dentry,
1182fca3 MM	351	};
1182fca3 MM	352
8bb8aefd YL	353	static int orangefs_fill_sb(struct super_block *sb,
8bb8aefd YL	354	struct orangefs_fs_mount_response *fs_mount,
5c0dbbc6	355	void *data, int silent)
1182fca3 MM	356	{
	357	int ret = -EINVAL;
	358	struct inode *root = NULL;
	359	struct dentry *root_dentry = NULL;
8bb8aefd	360	struct orangefs_object_kref root_object;
1182fca3	361
8bb8aefd	362	/* alloc and init our private orangefs sb info */
05d31c5c	363	sb->s_fs_info = kzalloc(sizeof(struct orangefs_sb_info_s), GFP_KERNEL);
8bb8aefd	364	if (!ORANGEFS_SB(sb))
1182fca3	365	return -ENOMEM;
8bb8aefd	366	ORANGEFS_SB(sb)->sb = sb;
1182fca3	367
8bb8aefd YL	368	ORANGEFS_SB(sb)->root_khandle = fs_mount->root_khandle;
	369	ORANGEFS_SB(sb)->fs_id = fs_mount->fs_id;
	370	ORANGEFS_SB(sb)->id = fs_mount->id;
1182fca3	371
5c0dbbc6 AV	372	if (data) {
5c0dbbc6 AV	373	ret = parse_mount_options(sb, data, silent);
1182fca3 MM	374	if (ret)
	375	return ret;
	376	}
	377
	378	/* Hang the xattr handlers off the superblock */
8bb8aefd YL	379	sb->s_xattr = orangefs_xattr_handlers;
	380	sb->s_magic = ORANGEFS_SUPER_MAGIC;
	381	sb->s_op = &orangefs_s_ops;
	382	sb->s_d_op = &orangefs_dentry_operations;
1182fca3	383
8bb8aefd YL	384	sb->s_blocksize = orangefs_bufmap_size_query();
8bb8aefd YL	385	sb->s_blocksize_bits = orangefs_bufmap_shift_query();
1182fca3 MM	386	sb->s_maxbytes = MAX_LFS_FILESIZE;
1182fca3 MM	387
8bb8aefd YL	388	root_object.khandle = ORANGEFS_SB(sb)->root_khandle;
8bb8aefd YL	389	root_object.fs_id = ORANGEFS_SB(sb)->fs_id;
1182fca3 MM	390	gossip_debug(GOSSIP_SUPER_DEBUG,
	391	"get inode %pU, fsid %d\n",
	392	&root_object.khandle,
	393	root_object.fs_id);
	394
8bb8aefd	395	root = orangefs_iget(sb, &root_object);
1182fca3 MM	396	if (IS_ERR(root))
	397	return PTR_ERR(root);
	398
	399	gossip_debug(GOSSIP_SUPER_DEBUG,
	400	"Allocated root inode [%p] with mode %x\n",
	401	root,
	402	root->i_mode);
	403
	404	/* allocates and places root dentry in dcache */
	405	root_dentry = d_make_root(root);
b05a7851	406	if (!root_dentry)
1182fca3	407	return -ENOMEM;
1182fca3	408
8bb8aefd	409	sb->s_export_op = &orangefs_export_ops;
1182fca3 MM	410	sb->s_root = root_dentry;
	411	return 0;
	412	}
	413
8bb8aefd	414	struct dentry orangefs_mount(struct file_system_type fst,
1182fca3 MM	415	int flags,
	416	const char *devname,
	417	void *data)
	418	{
	419	int ret = -EINVAL;
	420	struct super_block *sb = ERR_PTR(-EINVAL);
8bb8aefd	421	struct orangefs_kernel_op_s *new_op;
1be21f86	422	struct dentry *d = ERR_PTR(-EINVAL);
1182fca3 MM	423
1182fca3 MM	424	gossip_debug(GOSSIP_SUPER_DEBUG,
8bb8aefd	425	"orangefs_mount: called with devname %s\n",
1182fca3 MM	426	devname);
	427
	428	if (!devname) {
	429	gossip_err("ERROR: device name not specified.\n");
	430	return ERR_PTR(-EINVAL);
	431	}
	432
8bb8aefd	433	new_op = op_alloc(ORANGEFS_VFS_OP_FS_MOUNT);
1182fca3 MM	434	if (!new_op)
	435	return ERR_PTR(-ENOMEM);
	436
8bb8aefd	437	strncpy(new_op->upcall.req.fs_mount.orangefs_config_server,
1182fca3	438	devname,
8bb8aefd	439	ORANGEFS_MAX_SERVER_ADDR_LEN);
1182fca3 MM	440
1182fca3 MM	441	gossip_debug(GOSSIP_SUPER_DEBUG,
8bb8aefd YL	442	"Attempting ORANGEFS Mount via host %s\n",
8bb8aefd YL	443	new_op->upcall.req.fs_mount.orangefs_config_server);
1182fca3	444
8bb8aefd	445	ret = service_operation(new_op, "orangefs_mount", 0);
1182fca3	446	gossip_debug(GOSSIP_SUPER_DEBUG,
8bb8aefd	447	"orangefs_mount: mount got return value of %d\n", ret);
1182fca3 MM	448	if (ret)
	449	goto free_op;
	450
8bb8aefd	451	if (new_op->downcall.resp.fs_mount.fs_id == ORANGEFS_FS_ID_NULL) {
1182fca3 MM	452	gossip_err("ERROR: Retrieved null fs_id\n");
	453	ret = -EINVAL;
	454	goto free_op;
	455	}
	456
1be21f86 MM	457	sb = sget(fst, NULL, set_anon_super, flags, NULL);
	458
	459	if (IS_ERR(sb)) {
	460	d = ERR_CAST(sb);
1182fca3 MM	461	goto free_op;
	462	}
	463
8bb8aefd	464	ret = orangefs_fill_sb(sb,
5c0dbbc6	465	&new_op->downcall.resp.fs_mount, data,
1be21f86 MM	466	flags & MS_SILENT ? 1 : 0);
	467
	468	if (ret) {
	469	d = ERR_PTR(ret);
	470	goto free_op;
	471	}
1182fca3 MM	472
	473	/*
	474	* on successful mount, store the devname and data
	475	* used
	476	*/
8bb8aefd	477	strncpy(ORANGEFS_SB(sb)->devname,
1182fca3	478	devname,
8bb8aefd	479	ORANGEFS_MAX_SERVER_ADDR_LEN);
1182fca3 MM	480
1182fca3 MM	481	/* mount_pending must be cleared */
8bb8aefd	482	ORANGEFS_SB(sb)->mount_pending = 0;
1182fca3 MM	483
1182fca3 MM	484	/*
8bb8aefd	485	* finally, add this sb to our list of known orangefs
1182fca3 MM	486	* sb's
1182fca3 MM	487	*/
45996492 AV	488	gossip_debug(GOSSIP_SUPER_DEBUG,
	489	"Adding SB %p to orangefs superblocks\n",
	490	ORANGEFS_SB(sb));
	491	spin_lock(&orangefs_superblocks_lock);
	492	list_add_tail(&ORANGEFS_SB(sb)->list, &orangefs_superblocks);
	493	spin_unlock(&orangefs_superblocks_lock);
1182fca3	494	op_release(new_op);
1be21f86	495	return dget(sb->s_root);
1182fca3 MM	496
1182fca3 MM	497	free_op:
8bb8aefd	498	gossip_err("orangefs_mount: mount request failed with %d\n", ret);
1182fca3	499	if (ret == -EINVAL) {
8bb8aefd	500	gossip_err("Ensure that all orangefs-servers have the same FS configuration files\n");
1182fca3 MM	501	gossip_err("Look at pvfs2-client-core log file (typically /tmp/pvfs2-client.log) for more details\n");
	502	}
	503
	504	op_release(new_op);
	505
1be21f86	506	return d;
1182fca3 MM	507	}
1182fca3 MM	508
8bb8aefd	509	void orangefs_kill_sb(struct super_block *sb)
1182fca3	510	{
8bb8aefd	511	gossip_debug(GOSSIP_SUPER_DEBUG, "orangefs_kill_sb: called\n");
1182fca3	512
524b1d30 AV	513	/* provided sb cleanup */
	514	kill_anon_super(sb);
	515
1182fca3 MM	516	/*
	517	* issue the unmount to userspace to tell it to remove the
	518	* dynamic mount info it has for this superblock
	519	*/
45996492	520	orangefs_unmount_sb(sb);
1182fca3	521
8bb8aefd	522	/* remove the sb from our list of orangefs specific sb's */
45996492 AV	523
	524	spin_lock(&orangefs_superblocks_lock);
	525	__list_del_entry(&ORANGEFS_SB(sb)->list); /* not list_del_init */
	526	ORANGEFS_SB(sb)->list.prev = NULL;
	527	spin_unlock(&orangefs_superblocks_lock);
	528
	529	/*
	530	* make sure that ORANGEFS_DEV_REMOUNT_ALL loop that might've seen us
	531	* gets completed before we free the dang thing.
	532	*/
	533	mutex_lock(&request_mutex);
	534	mutex_unlock(&request_mutex);
1182fca3	535
8bb8aefd YL	536	/* free the orangefs superblock private data */
8bb8aefd YL	537	kfree(ORANGEFS_SB(sb));
1182fca3 MM	538	}
1182fca3 MM	539
8bb8aefd	540	int orangefs_inode_cache_initialize(void)
1182fca3	541	{
8bb8aefd YL	542	orangefs_inode_cache = kmem_cache_create("orangefs_inode_cache",
8bb8aefd YL	543	sizeof(struct orangefs_inode_s),
1182fca3	544	0,
8bb8aefd YL	545	ORANGEFS_CACHE_CREATE_FLAGS,
8bb8aefd YL	546	orangefs_inode_cache_ctor);
1182fca3	547
8bb8aefd YL	548	if (!orangefs_inode_cache) {
8bb8aefd YL	549	gossip_err("Cannot create orangefs_inode_cache\n");
1182fca3 MM	550	return -ENOMEM;
	551	}
	552	return 0;
	553	}
	554
8bb8aefd	555	int orangefs_inode_cache_finalize(void)
1182fca3	556	{
8bb8aefd	557	kmem_cache_destroy(orangefs_inode_cache);
1182fca3 MM	558	return 0;
1182fca3 MM	559	}