fs/exofs/super.c

   1 /*
   2  * Copyright (C) 2005, 2006
   3  * Avishay Traeger (avishay@gmail.com)
   4  * Copyright (C) 2008, 2009
   5  * Boaz Harrosh <bharrosh@panasas.com>
   6  *
   7  * Copyrights for code taken from ext2:
   8  *     Copyright (C) 1992, 1993, 1994, 1995
   9  *     Remy Card (card@masi.ibp.fr)
  10  *     Laboratoire MASI - Institut Blaise Pascal
  11  *     Universite Pierre et Marie Curie (Paris VI)
  12  *     from
  13  *     linux/fs/minix/inode.c
  14  *     Copyright (C) 1991, 1992  Linus Torvalds
  15  *
  16  * This file is part of exofs.
  17  *
  18  * exofs is free software; you can redistribute it and/or modify
  19  * it under the terms of the GNU General Public License as published by
  20  * the Free Software Foundation.  Since it is based on ext2, and the only
  21  * valid version of GPL for the Linux kernel is version 2, the only valid
  22  * version of GPL for exofs is version 2.
  23  *
  24  * exofs is distributed in the hope that it will be useful,
  25  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  26  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  27  * GNU General Public License for more details.
  28  *
  29  * You should have received a copy of the GNU General Public License
  30  * along with exofs; if not, write to the Free Software
  31  * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
  32  */
  33
  34 #include <linux/smp_lock.h>
  35 #include <linux/string.h>
  36 #include <linux/parser.h>
  37 #include <linux/vfs.h>
  38 #include <linux/random.h>
  39 #include <linux/exportfs.h>
  40
  41 #include "exofs.h"
  42
  43 /******************************************************************************
  44  * MOUNT OPTIONS
  45  *****************************************************************************/
  46
  47 /*
  48  * struct to hold what we get from mount options
  49  */
  50 struct exofs_mountopt {
  51         const char *dev_name;
  52         uint64_t pid;
  53         int timeout;
  54 };
  55
  56 /*
  57  * exofs-specific mount-time options.
  58  */
  59 enum { Opt_pid, Opt_to, Opt_mkfs, Opt_format, Opt_err };
  60
  61 /*
  62  * Our mount-time options.  These should ideally be 64-bit unsigned, but the
  63  * kernel's parsing functions do not currently support that.  32-bit should be
  64  * sufficient for most applications now.
  65  */
  66 static match_table_t tokens = {
  67         {Opt_pid, "pid=%u"},
  68         {Opt_to, "to=%u"},
  69         {Opt_err, NULL}
  70 };
  71
  72 /*
  73  * The main option parsing method.  Also makes sure that all of the mandatory
  74  * mount options were set.
  75  */
  76 static int parse_options(char *options, struct exofs_mountopt *opts)
  77 {
  78         char *p;
  79         substring_t args[MAX_OPT_ARGS];
  80         int option;
  81         bool s_pid = false;
  82
  83         EXOFS_DBGMSG("parse_options %s\n", options);
  84         /* defaults */
  85         memset(opts, 0, sizeof(*opts));
  86         opts->timeout = BLK_DEFAULT_SG_TIMEOUT;
  87
  88         while ((p = strsep(&options, ",")) != NULL) {
  89                 int token;
  90                 char str[32];
  91
  92                 if (!*p)
  93                         continue;
  94
  95                 token = match_token(p, tokens, args);
  96                 switch (token) {
  97                 case Opt_pid:
  98                         if (0 == match_strlcpy(str, &args[0], sizeof(str)))
  99                                 return -EINVAL;
 100                         opts->pid = simple_strtoull(str, NULL, 0);
 101                         if (opts->pid < EXOFS_MIN_PID) {
 102                                 EXOFS_ERR("Partition ID must be >= %u",
 103                                           EXOFS_MIN_PID);
 104                                 return -EINVAL;
 105                         }
 106                         s_pid = 1;
 107                         break;
 108                 case Opt_to:
 109                         if (match_int(&args[0], &option))
 110                                 return -EINVAL;
 111                         if (option <= 0) {
 112                                 EXOFS_ERR("Timout must be > 0");
 113                                 return -EINVAL;
 114                         }
 115                         opts->timeout = option * HZ;
 116                         break;
 117                 }
 118         }
 119
 120         if (!s_pid) {
 121                 EXOFS_ERR("Need to specify the following options:\n");
 122                 EXOFS_ERR("    -o pid=pid_no_to_use\n");
 123                 return -EINVAL;
 124         }
 125
 126         return 0;
 127 }
 128
 129 /******************************************************************************
 130  * INODE CACHE
 131  *****************************************************************************/
 132
 133 /*
 134  * Our inode cache.  Isn't it pretty?
 135  */
 136 static struct kmem_cache *exofs_inode_cachep;
 137
 138 /*
 139  * Allocate an inode in the cache
 140  */
 141 static struct inode *exofs_alloc_inode(struct super_block *sb)
 142 {
 143         struct exofs_i_info *oi;
 144
 145         oi = kmem_cache_alloc(exofs_inode_cachep, GFP_KERNEL);
 146         if (!oi)
 147                 return NULL;
 148
 149         oi->vfs_inode.i_version = 1;
 150         return &oi->vfs_inode;
 151 }
 152
 153 /*
 154  * Remove an inode from the cache
 155  */
 156 static void exofs_destroy_inode(struct inode *inode)
 157 {
 158         kmem_cache_free(exofs_inode_cachep, exofs_i(inode));
 159 }
 160
 161 /*
 162  * Initialize the inode
 163  */
 164 static void exofs_init_once(void *foo)
 165 {
 166         struct exofs_i_info *oi = foo;
 167
 168         inode_init_once(&oi->vfs_inode);
 169 }
 170
 171 /*
 172  * Create and initialize the inode cache
 173  */
 174 static int init_inodecache(void)
 175 {
 176         exofs_inode_cachep = kmem_cache_create("exofs_inode_cache",
 177                                 sizeof(struct exofs_i_info), 0,
 178                                 SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD,
 179                                 exofs_init_once);
 180         if (exofs_inode_cachep == NULL)
 181                 return -ENOMEM;
 182         return 0;
 183 }
 184
 185 /*
 186  * Destroy the inode cache
 187  */
 188 static void destroy_inodecache(void)
 189 {
 190         kmem_cache_destroy(exofs_inode_cachep);
 191 }
 192
 193 /******************************************************************************
 194  * SUPERBLOCK FUNCTIONS
 195  *****************************************************************************/
 196 static const struct super_operations exofs_sops;
 197 static const struct export_operations exofs_export_ops;
 198
 199 /*
 200  * Write the superblock to the OSD
 201  */
 202 int exofs_sync_fs(struct super_block *sb, int wait)
 203 {
 204         struct exofs_sb_info *sbi;
 205         struct exofs_fscb *fscb;
 206         struct osd_request *or;
 207         struct osd_obj_id obj;
 208         int ret = -ENOMEM;
 209
 210         fscb = kzalloc(sizeof(struct exofs_fscb), GFP_KERNEL);
 211         if (!fscb) {
 212                 EXOFS_ERR("exofs_write_super: memory allocation failed.\n");
 213                 return -ENOMEM;
 214         }
 215
 216         lock_super(sb);
 217         sbi = sb->s_fs_info;
 218         fscb->s_nextid = cpu_to_le64(sbi->s_nextid);
 219         fscb->s_numfiles = cpu_to_le32(sbi->s_numfiles);
 220         fscb->s_magic = cpu_to_le16(sb->s_magic);
 221         fscb->s_newfs = 0;
 222
 223         or = osd_start_request(sbi->s_dev, GFP_KERNEL);
 224         if (unlikely(!or)) {
 225                 EXOFS_ERR("exofs_write_super: osd_start_request failed.\n");
 226                 goto out;
 227         }
 228
 229         obj.partition = sbi->s_pid;
 230         obj.id = EXOFS_SUPER_ID;
 231         ret = osd_req_write_kern(or, &obj, 0, fscb, sizeof(*fscb));
 232         if (unlikely(ret)) {
 233                 EXOFS_ERR("exofs_write_super: osd_req_write_kern failed.\n");
 234                 goto out;
 235         }
 236
 237         ret = exofs_sync_op(or, sbi->s_timeout, sbi->s_cred);
 238         if (unlikely(ret)) {
 239                 EXOFS_ERR("exofs_write_super: exofs_sync_op failed.\n");
 240                 goto out;
 241         }
 242         sb->s_dirt = 0;
 243
 244 out:
 245         if (or)
 246                 osd_end_request(or);
 247         unlock_super(sb);
 248         kfree(fscb);
 249         return ret;
 250 }
 251
 252 static void exofs_write_super(struct super_block *sb)
 253 {
 254         if (!(sb->s_flags & MS_RDONLY))
 255                 exofs_sync_fs(sb, 1);
 256         else
 257                 sb->s_dirt = 0;
 258 }
 259
 260 /*
 261  * This function is called when the vfs is freeing the superblock.  We just
 262  * need to free our own part.
 263  */
 264 static void exofs_put_super(struct super_block *sb)
 265 {
 266         int num_pend;
 267         struct exofs_sb_info *sbi = sb->s_fs_info;
 268
 269         if (sb->s_dirt)
 270                 exofs_write_super(sb);
 271
 272         /* make sure there are no pending commands */
 273         for (num_pend = atomic_read(&sbi->s_curr_pending); num_pend > 0;
 274              num_pend = atomic_read(&sbi->s_curr_pending)) {
 275                 wait_queue_head_t wq;
 276                 init_waitqueue_head(&wq);
 277                 wait_event_timeout(wq,
 278                                   (atomic_read(&sbi->s_curr_pending) == 0),
 279                                   msecs_to_jiffies(100));
 280         }
 281
 282         osduld_put_device(sbi->s_dev);
 283         kfree(sb->s_fs_info);
 284         sb->s_fs_info = NULL;
 285 }
 286
 287 /*
 288  * Read the superblock from the OSD and fill in the fields
 289  */
 290 static int exofs_fill_super(struct super_block *sb, void *data, int silent)
 291 {
 292         struct inode *root;
 293         struct exofs_mountopt *opts = data;
 294         struct exofs_sb_info *sbi;      /*extended info                  */
 295         struct exofs_fscb fscb;         /*on-disk superblock info        */
 296         struct osd_request *or = NULL;
 297         struct osd_obj_id obj;
 298         int ret;
 299
 300         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 301         if (!sbi)
 302                 return -ENOMEM;
 303         sb->s_fs_info = sbi;
 304
 305         /* use mount options to fill superblock */
 306         sbi->s_dev = osduld_path_lookup(opts->dev_name);
 307         if (IS_ERR(sbi->s_dev)) {
 308                 ret = PTR_ERR(sbi->s_dev);
 309                 sbi->s_dev = NULL;
 310                 goto free_sbi;
 311         }
 312
 313         sbi->s_pid = opts->pid;
 314         sbi->s_timeout = opts->timeout;
 315
 316         /* fill in some other data by hand */
 317         memset(sb->s_id, 0, sizeof(sb->s_id));
 318         strcpy(sb->s_id, "exofs");
 319         sb->s_blocksize = EXOFS_BLKSIZE;
 320         sb->s_blocksize_bits = EXOFS_BLKSHIFT;
 321         sb->s_maxbytes = MAX_LFS_FILESIZE;
 322         atomic_set(&sbi->s_curr_pending, 0);
 323         sb->s_bdev = NULL;
 324         sb->s_dev = 0;
 325
 326         /* read data from on-disk superblock object */
 327         obj.partition = sbi->s_pid;
 328         obj.id = EXOFS_SUPER_ID;
 329         exofs_make_credential(sbi->s_cred, &obj);
 330
 331         or = osd_start_request(sbi->s_dev, GFP_KERNEL);
 332         if (unlikely(!or)) {
 333                 if (!silent)
 334                         EXOFS_ERR(
 335                                "exofs_fill_super: osd_start_request failed.\n");
 336                 ret = -ENOMEM;
 337                 goto free_sbi;
 338         }
 339         ret = osd_req_read_kern(or, &obj, 0, &fscb, sizeof(fscb));
 340         if (unlikely(ret)) {
 341                 if (!silent)
 342                         EXOFS_ERR(
 343                                "exofs_fill_super: osd_req_read_kern failed.\n");
 344                 ret = -ENOMEM;
 345                 goto free_sbi;
 346         }
 347
 348         ret = exofs_sync_op(or, sbi->s_timeout, sbi->s_cred);
 349         if (unlikely(ret)) {
 350                 if (!silent)
 351                         EXOFS_ERR("exofs_fill_super: exofs_sync_op failed.\n");
 352                 ret = -EIO;
 353                 goto free_sbi;
 354         }
 355
 356         sb->s_magic = le16_to_cpu(fscb.s_magic);
 357         sbi->s_nextid = le64_to_cpu(fscb.s_nextid);
 358         sbi->s_numfiles = le32_to_cpu(fscb.s_numfiles);
 359
 360         /* make sure what we read from the object store is correct */
 361         if (sb->s_magic != EXOFS_SUPER_MAGIC) {
 362                 if (!silent)
 363                         EXOFS_ERR("ERROR: Bad magic value\n");
 364                 ret = -EINVAL;
 365                 goto free_sbi;
 366         }
 367
 368         /* start generation numbers from a random point */
 369         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
 370         spin_lock_init(&sbi->s_next_gen_lock);
 371
 372         /* set up operation vectors */
 373         sb->s_op = &exofs_sops;
 374         sb->s_export_op = &exofs_export_ops;
 375         root = exofs_iget(sb, EXOFS_ROOT_ID - EXOFS_OBJ_OFF);
 376         if (IS_ERR(root)) {
 377                 EXOFS_ERR("ERROR: exofs_iget failed\n");
 378                 ret = PTR_ERR(root);
 379                 goto free_sbi;
 380         }
 381         sb->s_root = d_alloc_root(root);
 382         if (!sb->s_root) {
 383                 iput(root);
 384                 EXOFS_ERR("ERROR: get root inode failed\n");
 385                 ret = -ENOMEM;
 386                 goto free_sbi;
 387         }
 388
 389         if (!S_ISDIR(root->i_mode)) {
 390                 dput(sb->s_root);
 391                 sb->s_root = NULL;
 392                 EXOFS_ERR("ERROR: corrupt root inode (mode = %hd)\n",
 393                        root->i_mode);
 394                 ret = -EINVAL;
 395                 goto free_sbi;
 396         }
 397
 398         ret = 0;
 399 out:
 400         if (or)
 401                 osd_end_request(or);
 402         return ret;
 403
 404 free_sbi:
 405         osduld_put_device(sbi->s_dev); /* NULL safe */
 406         kfree(sbi);
 407         goto out;
 408 }
 409
 410 /*
 411  * Set up the superblock (calls exofs_fill_super eventually)
 412  */
 413 static int exofs_get_sb(struct file_system_type *type,
 414                           int flags, const char *dev_name,
 415                           void *data, struct vfsmount *mnt)
 416 {
 417         struct exofs_mountopt opts;
 418         int ret;
 419
 420         ret = parse_options(data, &opts);
 421         if (ret)
 422                 return ret;
 423
 424         opts.dev_name = dev_name;
 425         return get_sb_nodev(type, flags, &opts, exofs_fill_super, mnt);
 426 }
 427
 428 /*
 429  * Return information about the file system state in the buffer.  This is used
 430  * by the 'df' command, for example.
 431  */
 432 static int exofs_statfs(struct dentry *dentry, struct kstatfs *buf)
 433 {
 434         struct super_block *sb = dentry->d_sb;
 435         struct exofs_sb_info *sbi = sb->s_fs_info;
 436         struct osd_obj_id obj = {sbi->s_pid, 0};
 437         struct osd_attr attrs[] = {
 438                 ATTR_DEF(OSD_APAGE_PARTITION_QUOTAS,
 439                         OSD_ATTR_PQ_CAPACITY_QUOTA, sizeof(__be64)),
 440                 ATTR_DEF(OSD_APAGE_PARTITION_INFORMATION,
 441                         OSD_ATTR_PI_USED_CAPACITY, sizeof(__be64)),
 442         };
 443         uint64_t capacity = ULLONG_MAX;
 444         uint64_t used = ULLONG_MAX;
 445         struct osd_request *or;
 446         uint8_t cred_a[OSD_CAP_LEN];
 447         int ret;
 448
 449         /* get used/capacity attributes */
 450         exofs_make_credential(cred_a, &obj);
 451
 452         or = osd_start_request(sbi->s_dev, GFP_KERNEL);
 453         if (unlikely(!or)) {
 454                 EXOFS_DBGMSG("exofs_statfs: osd_start_request failed.\n");
 455                 return -ENOMEM;
 456         }
 457
 458         osd_req_get_attributes(or, &obj);
 459         osd_req_add_get_attr_list(or, attrs, ARRAY_SIZE(attrs));
 460         ret = exofs_sync_op(or, sbi->s_timeout, cred_a);
 461         if (unlikely(ret))
 462                 goto out;
 463
 464         ret = extract_attr_from_req(or, &attrs[0]);
 465         if (likely(!ret))
 466                 capacity = get_unaligned_be64(attrs[0].val_ptr);
 467         else
 468                 EXOFS_DBGMSG("exofs_statfs: get capacity failed.\n");
 469
 470         ret = extract_attr_from_req(or, &attrs[1]);
 471         if (likely(!ret))
 472                 used = get_unaligned_be64(attrs[1].val_ptr);
 473         else
 474                 EXOFS_DBGMSG("exofs_statfs: get used-space failed.\n");
 475
 476         /* fill in the stats buffer */
 477         buf->f_type = EXOFS_SUPER_MAGIC;
 478         buf->f_bsize = EXOFS_BLKSIZE;
 479         buf->f_blocks = (capacity >> EXOFS_BLKSHIFT);
 480         buf->f_bfree = ((capacity - used) >> EXOFS_BLKSHIFT);
 481         buf->f_bavail = buf->f_bfree;
 482         buf->f_files = sbi->s_numfiles;
 483         buf->f_ffree = EXOFS_MAX_ID - sbi->s_numfiles;
 484         buf->f_namelen = EXOFS_NAME_LEN;
 485
 486 out:
 487         osd_end_request(or);
 488         return ret;
 489 }
 490
 491 static const struct super_operations exofs_sops = {
 492         .alloc_inode    = exofs_alloc_inode,
 493         .destroy_inode  = exofs_destroy_inode,
 494         .write_inode    = exofs_write_inode,
 495         .delete_inode   = exofs_delete_inode,
 496         .put_super      = exofs_put_super,
 497         .write_super    = exofs_write_super,
 498         .sync_fs        = exofs_sync_fs,
 499         .statfs         = exofs_statfs,
 500 };
 501
 502 /******************************************************************************
 503  * EXPORT OPERATIONS
 504  *****************************************************************************/
 505
 506 struct dentry *exofs_get_parent(struct dentry *child)
 507 {
 508         unsigned long ino = exofs_parent_ino(child);
 509
 510         if (!ino)
 511                 return NULL;
 512
 513         return d_obtain_alias(exofs_iget(child->d_inode->i_sb, ino));
 514 }
 515
 516 static struct inode *exofs_nfs_get_inode(struct super_block *sb,
 517                 u64 ino, u32 generation)
 518 {
 519         struct inode *inode;
 520
 521         inode = exofs_iget(sb, ino);
 522         if (IS_ERR(inode))
 523                 return ERR_CAST(inode);
 524         if (generation && inode->i_generation != generation) {
 525                 /* we didn't find the right inode.. */
 526                 iput(inode);
 527                 return ERR_PTR(-ESTALE);
 528         }
 529         return inode;
 530 }
 531
 532 static struct dentry *exofs_fh_to_dentry(struct super_block *sb,
 533                                 struct fid *fid, int fh_len, int fh_type)
 534 {
 535         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 536                                     exofs_nfs_get_inode);
 537 }
 538
 539 static struct dentry *exofs_fh_to_parent(struct super_block *sb,
 540                                 struct fid *fid, int fh_len, int fh_type)
 541 {
 542         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 543                                     exofs_nfs_get_inode);
 544 }
 545
 546 static const struct export_operations exofs_export_ops = {
 547         .fh_to_dentry = exofs_fh_to_dentry,
 548         .fh_to_parent = exofs_fh_to_parent,
 549         .get_parent = exofs_get_parent,
 550 };
 551
 552 /******************************************************************************
 553  * INSMOD/RMMOD
 554  *****************************************************************************/
 555
 556 /*
 557  * struct that describes this file system
 558  */
 559 static struct file_system_type exofs_type = {
 560         .owner          = THIS_MODULE,
 561         .name           = "exofs",
 562         .get_sb         = exofs_get_sb,
 563         .kill_sb        = generic_shutdown_super,
 564 };
 565
 566 static int __init init_exofs(void)
 567 {
 568         int err;
 569
 570         err = init_inodecache();
 571         if (err)
 572                 goto out;
 573
 574         err = register_filesystem(&exofs_type);
 575         if (err)
 576                 goto out_d;
 577
 578         return 0;
 579 out_d:
 580         destroy_inodecache();
 581 out:
 582         return err;
 583 }
 584
 585 static void __exit exit_exofs(void)
 586 {
 587         unregister_filesystem(&exofs_type);
 588         destroy_inodecache();
 589 }
 590
 591 MODULE_AUTHOR("Avishay Traeger <avishay@gmail.com>");
 592 MODULE_DESCRIPTION("exofs");
 593 MODULE_LICENSE("GPL");
 594
 595 module_init(init_exofs)
 596 module_exit(exit_exofs)