2 * Copyright (C) 2005, 2006
3 * Avishay Traeger (avishay@gmail.com) (avishay@il.ibm.com)
4 * Copyright (C) 2005, 2006
5 * International Business Machines
6 * Copyright (C) 2008, 2009
7 * Boaz Harrosh <bharrosh@panasas.com>
9 * Copyrights for code taken from ext2:
10 * Copyright (C) 1992, 1993, 1994, 1995
11 * Remy Card (card@masi.ibp.fr)
12 * Laboratoire MASI - Institut Blaise Pascal
13 * Universite Pierre et Marie Curie (Paris VI)
15 * linux/fs/minix/inode.c
16 * Copyright (C) 1991, 1992 Linus Torvalds
18 * This file is part of exofs.
20 * exofs is free software; you can redistribute it and/or modify
21 * it under the terms of the GNU General Public License as published by
22 * the Free Software Foundation. Since it is based on ext2, and the only
23 * valid version of GPL for the Linux kernel is version 2, the only valid
24 * version of GPL for exofs is version 2.
26 * exofs is distributed in the hope that it will be useful,
27 * but WITHOUT ANY WARRANTY; without even the implied warranty of
28 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29 * GNU General Public License for more details.
31 * You should have received a copy of the GNU General Public License
32 * along with exofs; if not, write to the Free Software
33 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
36 #include <linux/string.h>
37 #include <linux/parser.h>
38 #include <linux/vfs.h>
39 #include <linux/random.h>
40 #include <linux/exportfs.h>
44 /******************************************************************************
46 *****************************************************************************/
49 * struct to hold what we get from mount options
51 struct exofs_mountopt {
58 * exofs-specific mount-time options.
60 enum { Opt_pid, Opt_to, Opt_mkfs, Opt_format, Opt_err };
63 * Our mount-time options. These should ideally be 64-bit unsigned, but the
64 * kernel's parsing functions do not currently support that. 32-bit should be
65 * sufficient for most applications now.
67 static match_table_t tokens = {
74 * The main option parsing method. Also makes sure that all of the mandatory
75 * mount options were set.
77 static int parse_options(char *options, struct exofs_mountopt *opts)
80 substring_t args[MAX_OPT_ARGS];
84 EXOFS_DBGMSG("parse_options %s\n", options);
86 memset(opts, 0, sizeof(*opts));
87 opts->timeout = BLK_DEFAULT_SG_TIMEOUT;
89 while ((p = strsep(&options, ",")) != NULL) {
96 token = match_token(p, tokens, args);
99 if (0 == match_strlcpy(str, &args[0], sizeof(str)))
101 opts->pid = simple_strtoull(str, NULL, 0);
102 if (opts->pid < EXOFS_MIN_PID) {
103 EXOFS_ERR("Partition ID must be >= %u",
110 if (match_int(&args[0], &option))
113 EXOFS_ERR("Timout must be > 0");
116 opts->timeout = option * HZ;
122 EXOFS_ERR("Need to specify the following options:\n");
123 EXOFS_ERR(" -o pid=pid_no_to_use\n");
130 /******************************************************************************
132 *****************************************************************************/
135 * Our inode cache. Isn't it pretty?
137 static struct kmem_cache *exofs_inode_cachep;
140 * Allocate an inode in the cache
142 static struct inode *exofs_alloc_inode(struct super_block *sb)
144 struct exofs_i_info *oi;
146 oi = kmem_cache_alloc(exofs_inode_cachep, GFP_KERNEL);
150 oi->vfs_inode.i_version = 1;
151 return &oi->vfs_inode;
155 * Remove an inode from the cache
157 static void exofs_destroy_inode(struct inode *inode)
159 kmem_cache_free(exofs_inode_cachep, exofs_i(inode));
163 * Initialize the inode
165 static void exofs_init_once(void *foo)
167 struct exofs_i_info *oi = foo;
169 inode_init_once(&oi->vfs_inode);
173 * Create and initialize the inode cache
175 static int init_inodecache(void)
177 exofs_inode_cachep = kmem_cache_create("exofs_inode_cache",
178 sizeof(struct exofs_i_info), 0,
179 SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD,
181 if (exofs_inode_cachep == NULL)
187 * Destroy the inode cache
189 static void destroy_inodecache(void)
191 kmem_cache_destroy(exofs_inode_cachep);
194 /******************************************************************************
195 * SUPERBLOCK FUNCTIONS
196 *****************************************************************************/
197 static const struct super_operations exofs_sops;
198 static const struct export_operations exofs_export_ops;
201 * Write the superblock to the OSD
203 static void exofs_write_super(struct super_block *sb)
205 struct exofs_sb_info *sbi;
206 struct exofs_fscb *fscb;
207 struct osd_request *or;
208 struct osd_obj_id obj;
211 fscb = kzalloc(sizeof(struct exofs_fscb), GFP_KERNEL);
213 EXOFS_ERR("exofs_write_super: memory allocation failed.\n");
219 fscb->s_nextid = cpu_to_le64(sbi->s_nextid);
220 fscb->s_numfiles = cpu_to_le32(sbi->s_numfiles);
221 fscb->s_magic = cpu_to_le16(sb->s_magic);
224 or = osd_start_request(sbi->s_dev, GFP_KERNEL);
226 EXOFS_ERR("exofs_write_super: osd_start_request failed.\n");
230 obj.partition = sbi->s_pid;
231 obj.id = EXOFS_SUPER_ID;
232 ret = osd_req_write_kern(or, &obj, 0, fscb, sizeof(*fscb));
234 EXOFS_ERR("exofs_write_super: osd_req_write_kern failed.\n");
238 ret = exofs_sync_op(or, sbi->s_timeout, sbi->s_cred);
240 EXOFS_ERR("exofs_write_super: exofs_sync_op failed.\n");
253 * This function is called when the vfs is freeing the superblock. We just
254 * need to free our own part.
256 static void exofs_put_super(struct super_block *sb)
259 struct exofs_sb_info *sbi = sb->s_fs_info;
264 exofs_write_super(sb);
266 /* make sure there are no pending commands */
267 for (num_pend = atomic_read(&sbi->s_curr_pending); num_pend > 0;
268 num_pend = atomic_read(&sbi->s_curr_pending)) {
269 wait_queue_head_t wq;
270 init_waitqueue_head(&wq);
271 wait_event_timeout(wq,
272 (atomic_read(&sbi->s_curr_pending) == 0),
273 msecs_to_jiffies(100));
276 osduld_put_device(sbi->s_dev);
277 kfree(sb->s_fs_info);
278 sb->s_fs_info = NULL;
284 * Read the superblock from the OSD and fill in the fields
286 static int exofs_fill_super(struct super_block *sb, void *data, int silent)
289 struct exofs_mountopt *opts = data;
290 struct exofs_sb_info *sbi; /*extended info */
291 struct exofs_fscb fscb; /*on-disk superblock info */
292 struct osd_request *or = NULL;
293 struct osd_obj_id obj;
296 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
301 /* use mount options to fill superblock */
302 sbi->s_dev = osduld_path_lookup(opts->dev_name);
303 if (IS_ERR(sbi->s_dev)) {
304 ret = PTR_ERR(sbi->s_dev);
309 sbi->s_pid = opts->pid;
310 sbi->s_timeout = opts->timeout;
312 /* fill in some other data by hand */
313 memset(sb->s_id, 0, sizeof(sb->s_id));
314 strcpy(sb->s_id, "exofs");
315 sb->s_blocksize = EXOFS_BLKSIZE;
316 sb->s_blocksize_bits = EXOFS_BLKSHIFT;
317 sb->s_maxbytes = MAX_LFS_FILESIZE;
318 atomic_set(&sbi->s_curr_pending, 0);
322 /* read data from on-disk superblock object */
323 obj.partition = sbi->s_pid;
324 obj.id = EXOFS_SUPER_ID;
325 exofs_make_credential(sbi->s_cred, &obj);
327 or = osd_start_request(sbi->s_dev, GFP_KERNEL);
331 "exofs_fill_super: osd_start_request failed.\n");
335 ret = osd_req_read_kern(or, &obj, 0, &fscb, sizeof(fscb));
339 "exofs_fill_super: osd_req_read_kern failed.\n");
344 ret = exofs_sync_op(or, sbi->s_timeout, sbi->s_cred);
347 EXOFS_ERR("exofs_fill_super: exofs_sync_op failed.\n");
352 sb->s_magic = le16_to_cpu(fscb.s_magic);
353 sbi->s_nextid = le64_to_cpu(fscb.s_nextid);
354 sbi->s_numfiles = le32_to_cpu(fscb.s_numfiles);
356 /* make sure what we read from the object store is correct */
357 if (sb->s_magic != EXOFS_SUPER_MAGIC) {
359 EXOFS_ERR("ERROR: Bad magic value\n");
364 /* start generation numbers from a random point */
365 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
366 spin_lock_init(&sbi->s_next_gen_lock);
368 /* set up operation vectors */
369 sb->s_op = &exofs_sops;
370 sb->s_export_op = &exofs_export_ops;
371 root = exofs_iget(sb, EXOFS_ROOT_ID - EXOFS_OBJ_OFF);
373 EXOFS_ERR("ERROR: exofs_iget failed\n");
377 sb->s_root = d_alloc_root(root);
380 EXOFS_ERR("ERROR: get root inode failed\n");
385 if (!S_ISDIR(root->i_mode)) {
388 EXOFS_ERR("ERROR: corrupt root inode (mode = %hd)\n",
401 osduld_put_device(sbi->s_dev); /* NULL safe */
407 * Set up the superblock (calls exofs_fill_super eventually)
409 static int exofs_get_sb(struct file_system_type *type,
410 int flags, const char *dev_name,
411 void *data, struct vfsmount *mnt)
413 struct exofs_mountopt opts;
416 ret = parse_options(data, &opts);
420 opts.dev_name = dev_name;
421 return get_sb_nodev(type, flags, &opts, exofs_fill_super, mnt);
425 * Return information about the file system state in the buffer. This is used
426 * by the 'df' command, for example.
428 static int exofs_statfs(struct dentry *dentry, struct kstatfs *buf)
430 struct super_block *sb = dentry->d_sb;
431 struct exofs_sb_info *sbi = sb->s_fs_info;
432 struct osd_obj_id obj = {sbi->s_pid, 0};
433 struct osd_attr attrs[] = {
434 ATTR_DEF(OSD_APAGE_PARTITION_QUOTAS,
435 OSD_ATTR_PQ_CAPACITY_QUOTA, sizeof(__be64)),
436 ATTR_DEF(OSD_APAGE_PARTITION_INFORMATION,
437 OSD_ATTR_PI_USED_CAPACITY, sizeof(__be64)),
439 uint64_t capacity = ULLONG_MAX;
440 uint64_t used = ULLONG_MAX;
441 struct osd_request *or;
442 uint8_t cred_a[OSD_CAP_LEN];
445 /* get used/capacity attributes */
446 exofs_make_credential(cred_a, &obj);
448 or = osd_start_request(sbi->s_dev, GFP_KERNEL);
450 EXOFS_DBGMSG("exofs_statfs: osd_start_request failed.\n");
454 osd_req_get_attributes(or, &obj);
455 osd_req_add_get_attr_list(or, attrs, ARRAY_SIZE(attrs));
456 ret = exofs_sync_op(or, sbi->s_timeout, cred_a);
460 ret = extract_attr_from_req(or, &attrs[0]);
462 capacity = get_unaligned_be64(attrs[0].val_ptr);
464 EXOFS_DBGMSG("exofs_statfs: get capacity failed.\n");
466 ret = extract_attr_from_req(or, &attrs[1]);
468 used = get_unaligned_be64(attrs[1].val_ptr);
470 EXOFS_DBGMSG("exofs_statfs: get used-space failed.\n");
472 /* fill in the stats buffer */
473 buf->f_type = EXOFS_SUPER_MAGIC;
474 buf->f_bsize = EXOFS_BLKSIZE;
475 buf->f_blocks = (capacity >> EXOFS_BLKSHIFT);
476 buf->f_bfree = ((capacity - used) >> EXOFS_BLKSHIFT);
477 buf->f_bavail = buf->f_bfree;
478 buf->f_files = sbi->s_numfiles;
479 buf->f_ffree = EXOFS_MAX_ID - sbi->s_numfiles;
480 buf->f_namelen = EXOFS_NAME_LEN;
487 static const struct super_operations exofs_sops = {
488 .alloc_inode = exofs_alloc_inode,
489 .destroy_inode = exofs_destroy_inode,
490 .write_inode = exofs_write_inode,
491 .delete_inode = exofs_delete_inode,
492 .put_super = exofs_put_super,
493 .write_super = exofs_write_super,
494 .statfs = exofs_statfs,
497 /******************************************************************************
499 *****************************************************************************/
501 struct dentry *exofs_get_parent(struct dentry *child)
503 unsigned long ino = exofs_parent_ino(child);
508 return d_obtain_alias(exofs_iget(child->d_inode->i_sb, ino));
511 static struct inode *exofs_nfs_get_inode(struct super_block *sb,
512 u64 ino, u32 generation)
516 inode = exofs_iget(sb, ino);
518 return ERR_CAST(inode);
519 if (generation && inode->i_generation != generation) {
520 /* we didn't find the right inode.. */
522 return ERR_PTR(-ESTALE);
527 static struct dentry *exofs_fh_to_dentry(struct super_block *sb,
528 struct fid *fid, int fh_len, int fh_type)
530 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
531 exofs_nfs_get_inode);
534 static struct dentry *exofs_fh_to_parent(struct super_block *sb,
535 struct fid *fid, int fh_len, int fh_type)
537 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
538 exofs_nfs_get_inode);
541 static const struct export_operations exofs_export_ops = {
542 .fh_to_dentry = exofs_fh_to_dentry,
543 .fh_to_parent = exofs_fh_to_parent,
544 .get_parent = exofs_get_parent,
547 /******************************************************************************
549 *****************************************************************************/
552 * struct that describes this file system
554 static struct file_system_type exofs_type = {
555 .owner = THIS_MODULE,
557 .get_sb = exofs_get_sb,
558 .kill_sb = generic_shutdown_super,
561 static int __init init_exofs(void)
565 err = init_inodecache();
569 err = register_filesystem(&exofs_type);
575 destroy_inodecache();
580 static void __exit exit_exofs(void)
582 unregister_filesystem(&exofs_type);
583 destroy_inodecache();
586 MODULE_AUTHOR("Avishay Traeger <avishay@gmail.com>");
587 MODULE_DESCRIPTION("exofs");
588 MODULE_LICENSE("GPL");
590 module_init(init_exofs)
591 module_exit(exit_exofs)