[linux-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 "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 }
};

uint64_t orangefs_features;

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);

	if (orangefs_userspace_version >= 20907) {
		new_op = op_alloc(ORANGEFS_VFS_OP_FEATURES);
		if (!new_op)
			return -ENOMEM;
		new_op->upcall.req.features.features = 0;
		ret = service_operation(new_op, "orangefs_features", 0);
		orangefs_features = new_op->downcall.resp.features.features;
		op_release(new_op);
	} else {
		orangefs_features = 0;
	}

	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);

	if (orangefs_userspace_version >= 20907) {
		new_op = op_alloc(ORANGEFS_VFS_OP_FEATURES);
		if (!new_op)
			return ERR_PTR(-ENOMEM);
		new_op->upcall.req.features.features = 0;
		ret = service_operation(new_op, "orangefs_features", 0);
		orangefs_features = new_op->downcall.resp.features.features;
		op_release(new_op);
	} else {
		orangefs_features = 0;
	}

	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(&orangefs_request_mutex);
	mutex_unlock(&orangefs_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
482664dd	36	uint64_t orangefs_features;
1182fca3 MM	37
	38	static int parse_mount_options(struct super_block sb, char options,
	39	int silent)
	40	{
8bb8aefd	41	struct orangefs_sb_info_s *orangefs_sb = ORANGEFS_SB(sb);
1182fca3 MM	42	substring_t args[MAX_OPT_ARGS];
	43	char *p;
	44
	45	/*
	46	* Force any potential flags that might be set from the mount
	47	* to zero, ie, initialize to unset.
	48	*/
	49	sb->s_flags &= ~MS_POSIXACL;
8bb8aefd YL	50	orangefs_sb->flags &= ~ORANGEFS_OPT_INTR;
8bb8aefd YL	51	orangefs_sb->flags &= ~ORANGEFS_OPT_LOCAL_LOCK;
1182fca3 MM	52
	53	while ((p = strsep(&options, ",")) != NULL) {
	54	int token;
	55
	56	if (!*p)
	57	continue;
	58
	59	token = match_token(p, tokens, args);
	60	switch (token) {
	61	case Opt_acl:
	62	sb->s_flags \|= MS_POSIXACL;
	63	break;
	64	case Opt_intr:
8bb8aefd	65	orangefs_sb->flags \|= ORANGEFS_OPT_INTR;
1182fca3 MM	66	break;
1182fca3 MM	67	case Opt_local_lock:
8bb8aefd	68	orangefs_sb->flags \|= ORANGEFS_OPT_LOCAL_LOCK;
1182fca3 MM	69	break;
	70	default:
	71	goto fail;
	72	}
	73	}
	74
	75	return 0;
	76	fail:
	77	if (!silent)
	78	gossip_err("Error: mount option [%s] is not supported.\n", p);
	79	return -EINVAL;
	80	}
	81
8bb8aefd	82	static void orangefs_inode_cache_ctor(void *req)
1182fca3	83	{
8bb8aefd	84	struct orangefs_inode_s *orangefs_inode = req;
1182fca3	85
8bb8aefd YL	86	inode_init_once(&orangefs_inode->vfs_inode);
8bb8aefd YL	87	init_rwsem(&orangefs_inode->xattr_sem);
1182fca3	88
8bb8aefd	89	orangefs_inode->vfs_inode.i_version = 1;
1182fca3 MM	90	}
1182fca3 MM	91
8bb8aefd	92	static struct inode orangefs_alloc_inode(struct super_block sb)
1182fca3	93	{
8bb8aefd	94	struct orangefs_inode_s *orangefs_inode;
1182fca3	95
2d4cae0d	96	orangefs_inode = kmem_cache_alloc(orangefs_inode_cache, GFP_KERNEL);
8bb8aefd YL	97	if (orangefs_inode == NULL) {
8bb8aefd YL	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,
be57366e MM	161	"%s: got %ld blocks available \| "
	162	"%ld blocks total \| %ld block size \| "
	163	"%ld files total \| %ld files avail\n",
	164	__func__,
1182fca3 MM	165	(long)new_op->downcall.resp.statfs.blocks_avail,
1182fca3 MM	166	(long)new_op->downcall.resp.statfs.blocks_total,
be57366e MM	167	(long)new_op->downcall.resp.statfs.block_size,
	168	(long)new_op->downcall.resp.statfs.files_total,
	169	(long)new_op->downcall.resp.statfs.files_avail);
1182fca3 MM	170
1182fca3 MM	171	buf->f_type = sb->s_magic;
8bb8aefd	172	memcpy(&buf->f_fsid, &ORANGEFS_SB(sb)->fs_id, sizeof(buf->f_fsid));
1182fca3	173	buf->f_bsize = new_op->downcall.resp.statfs.block_size;
47b4948f	174	buf->f_namelen = ORANGEFS_NAME_MAX;
1182fca3 MM	175
	176	buf->f_blocks = (sector_t) new_op->downcall.resp.statfs.blocks_total;
	177	buf->f_bfree = (sector_t) new_op->downcall.resp.statfs.blocks_avail;
	178	buf->f_bavail = (sector_t) new_op->downcall.resp.statfs.blocks_avail;
	179	buf->f_files = (sector_t) new_op->downcall.resp.statfs.files_total;
	180	buf->f_ffree = (sector_t) new_op->downcall.resp.statfs.files_avail;
	181	buf->f_frsize = sb->s_blocksize;
	182
	183	out_op_release:
	184	op_release(new_op);
8bb8aefd	185	gossip_debug(GOSSIP_SUPER_DEBUG, "orangefs_statfs: returning %d\n", ret);
1182fca3 MM	186	return ret;
	187	}
	188
	189	/*
	190	* Remount as initiated by VFS layer. We just need to reparse the mount
	191	* options, no need to signal pvfs2-client-core about it.
	192	*/
8bb8aefd	193	static int orangefs_remount_fs(struct super_block sb, int flags, char *data)
1182fca3	194	{
8bb8aefd	195	gossip_debug(GOSSIP_SUPER_DEBUG, "orangefs_remount_fs: called\n");
1182fca3 MM	196	return parse_mount_options(sb, data, 1);
	197	}
	198
	199	/*
	200	* Remount as initiated by pvfs2-client-core on restart. This is used to
	201	* repopulate mount information left from previous pvfs2-client-core.
	202	*
	203	* the idea here is that given a valid superblock, we're
	204	* re-initializing the user space client with the initial mount
	205	* information specified when the super block was first initialized.
	206	* this is very different than the first initialization/creation of a
	207	* superblock. we use the special service_priority_operation to make
	208	* sure that the mount gets ahead of any other pending operation that
	209	* is waiting for servicing. this means that the pvfs2-client won't
	210	* fail to start several times for all other pending operations before
	211	* the client regains all of the mount information from us.
	212	* NOTE: this function assumes that the request_mutex is already acquired!
	213	*/
45996492	214	int orangefs_remount(struct orangefs_sb_info_s *orangefs_sb)
1182fca3	215	{
8bb8aefd	216	struct orangefs_kernel_op_s *new_op;
1182fca3 MM	217	int ret = -EINVAL;
1182fca3 MM	218
8bb8aefd	219	gossip_debug(GOSSIP_SUPER_DEBUG, "orangefs_remount: called\n");
1182fca3	220
8bb8aefd	221	new_op = op_alloc(ORANGEFS_VFS_OP_FS_MOUNT);
1182fca3 MM	222	if (!new_op)
1182fca3 MM	223	return -ENOMEM;
8bb8aefd	224	strncpy(new_op->upcall.req.fs_mount.orangefs_config_server,
45996492	225	orangefs_sb->devname,
8bb8aefd	226	ORANGEFS_MAX_SERVER_ADDR_LEN);
1182fca3 MM	227
1182fca3 MM	228	gossip_debug(GOSSIP_SUPER_DEBUG,
8bb8aefd YL	229	"Attempting ORANGEFS Remount via host %s\n",
8bb8aefd YL	230	new_op->upcall.req.fs_mount.orangefs_config_server);
1182fca3 MM	231
1182fca3 MM	232	/*
adcf34a2	233	* we assume that the calling function has already acquired the
1182fca3 MM	234	* request_mutex to prevent other operations from bypassing
	235	* this one
	236	*/
8bb8aefd	237	ret = service_operation(new_op, "orangefs_remount",
adcf34a2	238	ORANGEFS_OP_PRIORITY \| ORANGEFS_OP_NO_MUTEX);
1182fca3	239	gossip_debug(GOSSIP_SUPER_DEBUG,
8bb8aefd	240	"orangefs_remount: mount got return value of %d\n",
1182fca3 MM	241	ret);
	242	if (ret == 0) {
	243	/*
	244	* store the id assigned to this sb -- it's just a
	245	* short-lived mapping that the system interface uses
	246	* to map this superblock to a particular mount entry
	247	*/
45996492 AV	248	orangefs_sb->id = new_op->downcall.resp.fs_mount.id;
45996492 AV	249	orangefs_sb->mount_pending = 0;
1182fca3 MM	250	}
	251
	252	op_release(new_op);
482664dd	253
0c95ad76	254	if (orangefs_userspace_version >= 20907) {
482664dd MB	255	new_op = op_alloc(ORANGEFS_VFS_OP_FEATURES);
	256	if (!new_op)
	257	return -ENOMEM;
	258	new_op->upcall.req.features.features = 0;
	259	ret = service_operation(new_op, "orangefs_features", 0);
	260	orangefs_features = new_op->downcall.resp.features.features;
	261	op_release(new_op);
	262	} else {
	263	orangefs_features = 0;
	264	}
	265
1182fca3 MM	266	return ret;
	267	}
	268
	269	int fsid_key_table_initialize(void)
	270	{
	271	return 0;
	272	}
	273
	274	void fsid_key_table_finalize(void)
	275	{
	276	}
	277
	278	/* Called whenever the VFS dirties the inode in response to atime updates */
8bb8aefd	279	static void orangefs_dirty_inode(struct inode *inode, int flags)
1182fca3	280	{
8bb8aefd	281	struct orangefs_inode_s *orangefs_inode = ORANGEFS_I(inode);
1182fca3 MM	282
1182fca3 MM	283	gossip_debug(GOSSIP_SUPER_DEBUG,
8bb8aefd	284	"orangefs_dirty_inode: %pU\n",
1182fca3	285	get_khandle_from_ino(inode));
8bb8aefd	286	SetAtimeFlag(orangefs_inode);
1182fca3 MM	287	}
1182fca3 MM	288
8bb8aefd YL	289	static const struct super_operations orangefs_s_ops = {
	290	.alloc_inode = orangefs_alloc_inode,
	291	.destroy_inode = orangefs_destroy_inode,
	292	.dirty_inode = orangefs_dirty_inode,
1182fca3	293	.drop_inode = generic_delete_inode,
8bb8aefd YL	294	.statfs = orangefs_statfs,
8bb8aefd YL	295	.remount_fs = orangefs_remount_fs,
1182fca3 MM	296	.show_options = generic_show_options,
	297	};
	298
8bb8aefd	299	static struct dentry orangefs_fh_to_dentry(struct super_block sb,
1182fca3 MM	300	struct fid *fid,
	301	int fh_len,
	302	int fh_type)
	303	{
8bb8aefd	304	struct orangefs_object_kref refn;
1182fca3 MM	305
	306	if (fh_len < 5 \|\| fh_type > 2)
	307	return NULL;
	308
8bb8aefd	309	ORANGEFS_khandle_from(&(refn.khandle), fid->raw, 16);
1182fca3 MM	310	refn.fs_id = (u32) fid->raw[4];
	311	gossip_debug(GOSSIP_SUPER_DEBUG,
	312	"fh_to_dentry: handle %pU, fs_id %d\n",
	313	&refn.khandle,
	314	refn.fs_id);
	315
8bb8aefd	316	return d_obtain_alias(orangefs_iget(sb, &refn));
1182fca3 MM	317	}
1182fca3 MM	318
8bb8aefd	319	static int orangefs_encode_fh(struct inode *inode,
1182fca3 MM	320	__u32 *fh,
	321	int *max_len,
	322	struct inode *parent)
	323	{
	324	int len = parent ? 10 : 5;
	325	int type = 1;
8bb8aefd	326	struct orangefs_object_kref refn;
1182fca3 MM	327
	328	if (*max_len < len) {
	329	gossip_lerr("fh buffer is too small for encoding\n");
	330	*max_len = len;
	331	type = 255;
	332	goto out;
	333	}
	334
8bb8aefd YL	335	refn = ORANGEFS_I(inode)->refn;
8bb8aefd YL	336	ORANGEFS_khandle_to(&refn.khandle, fh, 16);
1182fca3 MM	337	fh[4] = refn.fs_id;
	338
	339	gossip_debug(GOSSIP_SUPER_DEBUG,
	340	"Encoding fh: handle %pU, fsid %u\n",
	341	&refn.khandle,
	342	refn.fs_id);
	343
	344
	345	if (parent) {
8bb8aefd YL	346	refn = ORANGEFS_I(parent)->refn;
8bb8aefd YL	347	ORANGEFS_khandle_to(&refn.khandle, (char *) fh + 20, 16);
1182fca3 MM	348	fh[9] = refn.fs_id;
	349
	350	type = 2;
	351	gossip_debug(GOSSIP_SUPER_DEBUG,
	352	"Encoding parent: handle %pU, fsid %u\n",
	353	&refn.khandle,
	354	refn.fs_id);
	355	}
	356	*max_len = len;
	357
	358	out:
	359	return type;
	360	}
	361
acaca36d	362	static const struct export_operations orangefs_export_ops = {
8bb8aefd YL	363	.encode_fh = orangefs_encode_fh,
8bb8aefd YL	364	.fh_to_dentry = orangefs_fh_to_dentry,
1182fca3 MM	365	};
1182fca3 MM	366
8bb8aefd YL	367	static int orangefs_fill_sb(struct super_block *sb,
8bb8aefd YL	368	struct orangefs_fs_mount_response *fs_mount,
5c0dbbc6	369	void *data, int silent)
1182fca3 MM	370	{
	371	int ret = -EINVAL;
	372	struct inode *root = NULL;
	373	struct dentry *root_dentry = NULL;
8bb8aefd	374	struct orangefs_object_kref root_object;
1182fca3	375
8bb8aefd	376	/* alloc and init our private orangefs sb info */
05d31c5c	377	sb->s_fs_info = kzalloc(sizeof(struct orangefs_sb_info_s), GFP_KERNEL);
8bb8aefd	378	if (!ORANGEFS_SB(sb))
1182fca3	379	return -ENOMEM;
8bb8aefd	380	ORANGEFS_SB(sb)->sb = sb;
1182fca3	381
8bb8aefd YL	382	ORANGEFS_SB(sb)->root_khandle = fs_mount->root_khandle;
	383	ORANGEFS_SB(sb)->fs_id = fs_mount->fs_id;
	384	ORANGEFS_SB(sb)->id = fs_mount->id;
1182fca3	385
5c0dbbc6 AV	386	if (data) {
5c0dbbc6 AV	387	ret = parse_mount_options(sb, data, silent);
1182fca3 MM	388	if (ret)
	389	return ret;
	390	}
	391
	392	/* Hang the xattr handlers off the superblock */
8bb8aefd YL	393	sb->s_xattr = orangefs_xattr_handlers;
	394	sb->s_magic = ORANGEFS_SUPER_MAGIC;
	395	sb->s_op = &orangefs_s_ops;
	396	sb->s_d_op = &orangefs_dentry_operations;
1182fca3	397
8bb8aefd YL	398	sb->s_blocksize = orangefs_bufmap_size_query();
8bb8aefd YL	399	sb->s_blocksize_bits = orangefs_bufmap_shift_query();
1182fca3 MM	400	sb->s_maxbytes = MAX_LFS_FILESIZE;
1182fca3 MM	401
8bb8aefd YL	402	root_object.khandle = ORANGEFS_SB(sb)->root_khandle;
8bb8aefd YL	403	root_object.fs_id = ORANGEFS_SB(sb)->fs_id;
1182fca3 MM	404	gossip_debug(GOSSIP_SUPER_DEBUG,
	405	"get inode %pU, fsid %d\n",
	406	&root_object.khandle,
	407	root_object.fs_id);
	408
8bb8aefd	409	root = orangefs_iget(sb, &root_object);
1182fca3 MM	410	if (IS_ERR(root))
	411	return PTR_ERR(root);
	412
	413	gossip_debug(GOSSIP_SUPER_DEBUG,
	414	"Allocated root inode [%p] with mode %x\n",
	415	root,
	416	root->i_mode);
	417
	418	/* allocates and places root dentry in dcache */
	419	root_dentry = d_make_root(root);
b05a7851	420	if (!root_dentry)
1182fca3	421	return -ENOMEM;
1182fca3	422
8bb8aefd	423	sb->s_export_op = &orangefs_export_ops;
1182fca3 MM	424	sb->s_root = root_dentry;
	425	return 0;
	426	}
	427
8bb8aefd	428	struct dentry orangefs_mount(struct file_system_type fst,
1182fca3 MM	429	int flags,
	430	const char *devname,
	431	void *data)
	432	{
	433	int ret = -EINVAL;
	434	struct super_block *sb = ERR_PTR(-EINVAL);
8bb8aefd	435	struct orangefs_kernel_op_s *new_op;
1be21f86	436	struct dentry *d = ERR_PTR(-EINVAL);
1182fca3 MM	437
1182fca3 MM	438	gossip_debug(GOSSIP_SUPER_DEBUG,
8bb8aefd	439	"orangefs_mount: called with devname %s\n",
1182fca3 MM	440	devname);
	441
	442	if (!devname) {
	443	gossip_err("ERROR: device name not specified.\n");
	444	return ERR_PTR(-EINVAL);
	445	}
	446
8bb8aefd	447	new_op = op_alloc(ORANGEFS_VFS_OP_FS_MOUNT);
1182fca3 MM	448	if (!new_op)
	449	return ERR_PTR(-ENOMEM);
	450
8bb8aefd	451	strncpy(new_op->upcall.req.fs_mount.orangefs_config_server,
1182fca3	452	devname,
8bb8aefd	453	ORANGEFS_MAX_SERVER_ADDR_LEN);
1182fca3 MM	454
1182fca3 MM	455	gossip_debug(GOSSIP_SUPER_DEBUG,
8bb8aefd YL	456	"Attempting ORANGEFS Mount via host %s\n",
8bb8aefd YL	457	new_op->upcall.req.fs_mount.orangefs_config_server);
1182fca3	458
8bb8aefd	459	ret = service_operation(new_op, "orangefs_mount", 0);
1182fca3	460	gossip_debug(GOSSIP_SUPER_DEBUG,
8bb8aefd	461	"orangefs_mount: mount got return value of %d\n", ret);
1182fca3 MM	462	if (ret)
	463	goto free_op;
	464
8bb8aefd	465	if (new_op->downcall.resp.fs_mount.fs_id == ORANGEFS_FS_ID_NULL) {
1182fca3 MM	466	gossip_err("ERROR: Retrieved null fs_id\n");
	467	ret = -EINVAL;
	468	goto free_op;
	469	}
	470
1be21f86 MM	471	sb = sget(fst, NULL, set_anon_super, flags, NULL);
	472
	473	if (IS_ERR(sb)) {
	474	d = ERR_CAST(sb);
1182fca3 MM	475	goto free_op;
	476	}
	477
8bb8aefd	478	ret = orangefs_fill_sb(sb,
5c0dbbc6	479	&new_op->downcall.resp.fs_mount, data,
1be21f86 MM	480	flags & MS_SILENT ? 1 : 0);
	481
	482	if (ret) {
	483	d = ERR_PTR(ret);
	484	goto free_op;
	485	}
1182fca3 MM	486
	487	/*
	488	* on successful mount, store the devname and data
	489	* used
	490	*/
8bb8aefd	491	strncpy(ORANGEFS_SB(sb)->devname,
1182fca3	492	devname,
8bb8aefd	493	ORANGEFS_MAX_SERVER_ADDR_LEN);
1182fca3 MM	494
1182fca3 MM	495	/* mount_pending must be cleared */
8bb8aefd	496	ORANGEFS_SB(sb)->mount_pending = 0;
1182fca3 MM	497
1182fca3 MM	498	/*
8bb8aefd	499	* finally, add this sb to our list of known orangefs
1182fca3 MM	500	* sb's
1182fca3 MM	501	*/
45996492 AV	502	gossip_debug(GOSSIP_SUPER_DEBUG,
	503	"Adding SB %p to orangefs superblocks\n",
	504	ORANGEFS_SB(sb));
	505	spin_lock(&orangefs_superblocks_lock);
	506	list_add_tail(&ORANGEFS_SB(sb)->list, &orangefs_superblocks);
	507	spin_unlock(&orangefs_superblocks_lock);
1182fca3	508	op_release(new_op);
482664dd	509
0c95ad76	510	if (orangefs_userspace_version >= 20907) {
482664dd MB	511	new_op = op_alloc(ORANGEFS_VFS_OP_FEATURES);
	512	if (!new_op)
	513	return ERR_PTR(-ENOMEM);
	514	new_op->upcall.req.features.features = 0;
	515	ret = service_operation(new_op, "orangefs_features", 0);
	516	orangefs_features = new_op->downcall.resp.features.features;
	517	op_release(new_op);
	518	} else {
	519	orangefs_features = 0;
	520	}
	521
1be21f86	522	return dget(sb->s_root);
1182fca3 MM	523
1182fca3 MM	524	free_op:
8bb8aefd	525	gossip_err("orangefs_mount: mount request failed with %d\n", ret);
1182fca3	526	if (ret == -EINVAL) {
8bb8aefd	527	gossip_err("Ensure that all orangefs-servers have the same FS configuration files\n");
1182fca3 MM	528	gossip_err("Look at pvfs2-client-core log file (typically /tmp/pvfs2-client.log) for more details\n");
	529	}
	530
	531	op_release(new_op);
	532
1be21f86	533	return d;
1182fca3 MM	534	}
1182fca3 MM	535
8bb8aefd	536	void orangefs_kill_sb(struct super_block *sb)
1182fca3	537	{
8bb8aefd	538	gossip_debug(GOSSIP_SUPER_DEBUG, "orangefs_kill_sb: called\n");
1182fca3	539
524b1d30 AV	540	/* provided sb cleanup */
	541	kill_anon_super(sb);
	542
1182fca3 MM	543	/*
	544	* issue the unmount to userspace to tell it to remove the
	545	* dynamic mount info it has for this superblock
	546	*/
45996492	547	orangefs_unmount_sb(sb);
1182fca3	548
8bb8aefd	549	/* remove the sb from our list of orangefs specific sb's */
45996492 AV	550
	551	spin_lock(&orangefs_superblocks_lock);
	552	__list_del_entry(&ORANGEFS_SB(sb)->list); /* not list_del_init */
	553	ORANGEFS_SB(sb)->list.prev = NULL;
	554	spin_unlock(&orangefs_superblocks_lock);
	555
	556	/*
	557	* make sure that ORANGEFS_DEV_REMOUNT_ALL loop that might've seen us
	558	* gets completed before we free the dang thing.
	559	*/
1d503617 MB	560	mutex_lock(&orangefs_request_mutex);
1d503617 MB	561	mutex_unlock(&orangefs_request_mutex);
1182fca3	562
8bb8aefd YL	563	/* free the orangefs superblock private data */
8bb8aefd YL	564	kfree(ORANGEFS_SB(sb));
1182fca3 MM	565	}
1182fca3 MM	566
8bb8aefd	567	int orangefs_inode_cache_initialize(void)
1182fca3	568	{
8bb8aefd YL	569	orangefs_inode_cache = kmem_cache_create("orangefs_inode_cache",
8bb8aefd YL	570	sizeof(struct orangefs_inode_s),
1182fca3	571	0,
8bb8aefd YL	572	ORANGEFS_CACHE_CREATE_FLAGS,
8bb8aefd YL	573	orangefs_inode_cache_ctor);
1182fca3	574
8bb8aefd YL	575	if (!orangefs_inode_cache) {
8bb8aefd YL	576	gossip_err("Cannot create orangefs_inode_cache\n");
1182fca3 MM	577	return -ENOMEM;
	578	}
	579	return 0;
	580	}
	581
8bb8aefd	582	int orangefs_inode_cache_finalize(void)
1182fca3	583	{
8bb8aefd	584	kmem_cache_destroy(orangefs_inode_cache);
1182fca3 MM	585	return 0;
1182fca3 MM	586	}