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 static void _exofs_print_device(const char *msg, const char *dev_path,
 261                                 struct osd_dev *od, u64 pid)
 262 {
 263         const struct osd_dev_info *odi = osduld_device_info(od);
 264
 265         printk(KERN_NOTICE "exofs: %s %s osd_name-%s pid-0x%llx\n",
 266                 msg, dev_path ?: "", odi->osdname, _LLU(pid));
 267 }
 268
 269 /*
 270  * This function is called when the vfs is freeing the superblock.  We just
 271  * need to free our own part.
 272  */
 273 static void exofs_put_super(struct super_block *sb)
 274 {
 275         int num_pend;
 276         struct exofs_sb_info *sbi = sb->s_fs_info;
 277
 278         if (sb->s_dirt)
 279                 exofs_write_super(sb);
 280
 281         /* make sure there are no pending commands */
 282         for (num_pend = atomic_read(&sbi->s_curr_pending); num_pend > 0;
 283              num_pend = atomic_read(&sbi->s_curr_pending)) {
 284                 wait_queue_head_t wq;
 285                 init_waitqueue_head(&wq);
 286                 wait_event_timeout(wq,
 287                                   (atomic_read(&sbi->s_curr_pending) == 0),
 288                                   msecs_to_jiffies(100));
 289         }
 290
 291         _exofs_print_device("Unmounting", NULL, sbi->s_dev, sbi->s_pid);
 292         osduld_put_device(sbi->s_dev);
 293         kfree(sb->s_fs_info);
 294         sb->s_fs_info = NULL;
 295 }
 296
 297 /*
 298  * Read the superblock from the OSD and fill in the fields
 299  */
 300 static int exofs_fill_super(struct super_block *sb, void *data, int silent)
 301 {
 302         struct inode *root;
 303         struct exofs_mountopt *opts = data;
 304         struct exofs_sb_info *sbi;      /*extended info                  */
 305         struct exofs_fscb fscb;         /*on-disk superblock info        */
 306         struct osd_request *or = NULL;
 307         struct osd_obj_id obj;
 308         int ret;
 309
 310         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 311         if (!sbi)
 312                 return -ENOMEM;
 313         sb->s_fs_info = sbi;
 314
 315         /* use mount options to fill superblock */
 316         sbi->s_dev = osduld_path_lookup(opts->dev_name);
 317         if (IS_ERR(sbi->s_dev)) {
 318                 ret = PTR_ERR(sbi->s_dev);
 319                 sbi->s_dev = NULL;
 320                 goto free_sbi;
 321         }
 322
 323         sbi->s_pid = opts->pid;
 324         sbi->s_timeout = opts->timeout;
 325
 326         /* fill in some other data by hand */
 327         memset(sb->s_id, 0, sizeof(sb->s_id));
 328         strcpy(sb->s_id, "exofs");
 329         sb->s_blocksize = EXOFS_BLKSIZE;
 330         sb->s_blocksize_bits = EXOFS_BLKSHIFT;
 331         sb->s_maxbytes = MAX_LFS_FILESIZE;
 332         atomic_set(&sbi->s_curr_pending, 0);
 333         sb->s_bdev = NULL;
 334         sb->s_dev = 0;
 335
 336         /* read data from on-disk superblock object */
 337         obj.partition = sbi->s_pid;
 338         obj.id = EXOFS_SUPER_ID;
 339         exofs_make_credential(sbi->s_cred, &obj);
 340
 341         or = osd_start_request(sbi->s_dev, GFP_KERNEL);
 342         if (unlikely(!or)) {
 343                 if (!silent)
 344                         EXOFS_ERR(
 345                                "exofs_fill_super: osd_start_request failed.\n");
 346                 ret = -ENOMEM;
 347                 goto free_sbi;
 348         }
 349         ret = osd_req_read_kern(or, &obj, 0, &fscb, sizeof(fscb));
 350         if (unlikely(ret)) {
 351                 if (!silent)
 352                         EXOFS_ERR(
 353                                "exofs_fill_super: osd_req_read_kern failed.\n");
 354                 ret = -ENOMEM;
 355                 goto free_sbi;
 356         }
 357
 358         ret = exofs_sync_op(or, sbi->s_timeout, sbi->s_cred);
 359         if (unlikely(ret)) {
 360                 if (!silent)
 361                         EXOFS_ERR("exofs_fill_super: exofs_sync_op failed.\n");
 362                 ret = -EIO;
 363                 goto free_sbi;
 364         }
 365
 366         sb->s_magic = le16_to_cpu(fscb.s_magic);
 367         sbi->s_nextid = le64_to_cpu(fscb.s_nextid);
 368         sbi->s_numfiles = le32_to_cpu(fscb.s_numfiles);
 369
 370         /* make sure what we read from the object store is correct */
 371         if (sb->s_magic != EXOFS_SUPER_MAGIC) {
 372                 if (!silent)
 373                         EXOFS_ERR("ERROR: Bad magic value\n");
 374                 ret = -EINVAL;
 375                 goto free_sbi;
 376         }
 377
 378         /* start generation numbers from a random point */
 379         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
 380         spin_lock_init(&sbi->s_next_gen_lock);
 381
 382         /* set up operation vectors */
 383         sb->s_op = &exofs_sops;
 384         sb->s_export_op = &exofs_export_ops;
 385         root = exofs_iget(sb, EXOFS_ROOT_ID - EXOFS_OBJ_OFF);
 386         if (IS_ERR(root)) {
 387                 EXOFS_ERR("ERROR: exofs_iget failed\n");
 388                 ret = PTR_ERR(root);
 389                 goto free_sbi;
 390         }
 391         sb->s_root = d_alloc_root(root);
 392         if (!sb->s_root) {
 393                 iput(root);
 394                 EXOFS_ERR("ERROR: get root inode failed\n");
 395                 ret = -ENOMEM;
 396                 goto free_sbi;
 397         }
 398
 399         if (!S_ISDIR(root->i_mode)) {
 400                 dput(sb->s_root);
 401                 sb->s_root = NULL;
 402                 EXOFS_ERR("ERROR: corrupt root inode (mode = %hd)\n",
 403                        root->i_mode);
 404                 ret = -EINVAL;
 405                 goto free_sbi;
 406         }
 407
 408         _exofs_print_device("Mounting", opts->dev_name, sbi->s_dev, sbi->s_pid);
 409         ret = 0;
 410 out:
 411         if (or)
 412                 osd_end_request(or);
 413         return ret;
 414
 415 free_sbi:
 416         osduld_put_device(sbi->s_dev); /* NULL safe */
 417         kfree(sbi);
 418         goto out;
 419 }
 420
 421 /*
 422  * Set up the superblock (calls exofs_fill_super eventually)
 423  */
 424 static int exofs_get_sb(struct file_system_type *type,
 425                           int flags, const char *dev_name,
 426                           void *data, struct vfsmount *mnt)
 427 {
 428         struct exofs_mountopt opts;
 429         int ret;
 430
 431         ret = parse_options(data, &opts);
 432         if (ret)
 433                 return ret;
 434
 435         opts.dev_name = dev_name;
 436         return get_sb_nodev(type, flags, &opts, exofs_fill_super, mnt);
 437 }
 438
 439 /*
 440  * Return information about the file system state in the buffer.  This is used
 441  * by the 'df' command, for example.
 442  */
 443 static int exofs_statfs(struct dentry *dentry, struct kstatfs *buf)
 444 {
 445         struct super_block *sb = dentry->d_sb;
 446         struct exofs_sb_info *sbi = sb->s_fs_info;
 447         struct osd_obj_id obj = {sbi->s_pid, 0};
 448         struct osd_attr attrs[] = {
 449                 ATTR_DEF(OSD_APAGE_PARTITION_QUOTAS,
 450                         OSD_ATTR_PQ_CAPACITY_QUOTA, sizeof(__be64)),
 451                 ATTR_DEF(OSD_APAGE_PARTITION_INFORMATION,
 452                         OSD_ATTR_PI_USED_CAPACITY, sizeof(__be64)),
 453         };
 454         uint64_t capacity = ULLONG_MAX;
 455         uint64_t used = ULLONG_MAX;
 456         struct osd_request *or;
 457         uint8_t cred_a[OSD_CAP_LEN];
 458         int ret;
 459
 460         /* get used/capacity attributes */
 461         exofs_make_credential(cred_a, &obj);
 462
 463         or = osd_start_request(sbi->s_dev, GFP_KERNEL);
 464         if (unlikely(!or)) {
 465                 EXOFS_DBGMSG("exofs_statfs: osd_start_request failed.\n");
 466                 return -ENOMEM;
 467         }
 468
 469         osd_req_get_attributes(or, &obj);
 470         osd_req_add_get_attr_list(or, attrs, ARRAY_SIZE(attrs));
 471         ret = exofs_sync_op(or, sbi->s_timeout, cred_a);
 472         if (unlikely(ret))
 473                 goto out;
 474
 475         ret = extract_attr_from_req(or, &attrs[0]);
 476         if (likely(!ret))
 477                 capacity = get_unaligned_be64(attrs[0].val_ptr);
 478         else
 479                 EXOFS_DBGMSG("exofs_statfs: get capacity failed.\n");
 480
 481         ret = extract_attr_from_req(or, &attrs[1]);
 482         if (likely(!ret))
 483                 used = get_unaligned_be64(attrs[1].val_ptr);
 484         else
 485                 EXOFS_DBGMSG("exofs_statfs: get used-space failed.\n");
 486
 487         /* fill in the stats buffer */
 488         buf->f_type = EXOFS_SUPER_MAGIC;
 489         buf->f_bsize = EXOFS_BLKSIZE;
 490         buf->f_blocks = (capacity >> EXOFS_BLKSHIFT);
 491         buf->f_bfree = ((capacity - used) >> EXOFS_BLKSHIFT);
 492         buf->f_bavail = buf->f_bfree;
 493         buf->f_files = sbi->s_numfiles;
 494         buf->f_ffree = EXOFS_MAX_ID - sbi->s_numfiles;
 495         buf->f_namelen = EXOFS_NAME_LEN;
 496
 497 out:
 498         osd_end_request(or);
 499         return ret;
 500 }
 501
 502 static const struct super_operations exofs_sops = {
 503         .alloc_inode    = exofs_alloc_inode,
 504         .destroy_inode  = exofs_destroy_inode,
 505         .write_inode    = exofs_write_inode,
 506         .delete_inode   = exofs_delete_inode,
 507         .put_super      = exofs_put_super,
 508         .write_super    = exofs_write_super,
 509         .sync_fs        = exofs_sync_fs,
 510         .statfs         = exofs_statfs,
 511 };
 512
 513 /******************************************************************************
 514  * EXPORT OPERATIONS
 515  *****************************************************************************/
 516
 517 struct dentry *exofs_get_parent(struct dentry *child)
 518 {
 519         unsigned long ino = exofs_parent_ino(child);
 520
 521         if (!ino)
 522                 return NULL;
 523
 524         return d_obtain_alias(exofs_iget(child->d_inode->i_sb, ino));
 525 }
 526
 527 static struct inode *exofs_nfs_get_inode(struct super_block *sb,
 528                 u64 ino, u32 generation)
 529 {
 530         struct inode *inode;
 531
 532         inode = exofs_iget(sb, ino);
 533         if (IS_ERR(inode))
 534                 return ERR_CAST(inode);
 535         if (generation && inode->i_generation != generation) {
 536                 /* we didn't find the right inode.. */
 537                 iput(inode);
 538                 return ERR_PTR(-ESTALE);
 539         }
 540         return inode;
 541 }
 542
 543 static struct dentry *exofs_fh_to_dentry(struct super_block *sb,
 544                                 struct fid *fid, int fh_len, int fh_type)
 545 {
 546         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 547                                     exofs_nfs_get_inode);
 548 }
 549
 550 static struct dentry *exofs_fh_to_parent(struct super_block *sb,
 551                                 struct fid *fid, int fh_len, int fh_type)
 552 {
 553         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 554                                     exofs_nfs_get_inode);
 555 }
 556
 557 static const struct export_operations exofs_export_ops = {
 558         .fh_to_dentry = exofs_fh_to_dentry,
 559         .fh_to_parent = exofs_fh_to_parent,
 560         .get_parent = exofs_get_parent,
 561 };
 562
 563 /******************************************************************************
 564  * INSMOD/RMMOD
 565  *****************************************************************************/
 566
 567 /*
 568  * struct that describes this file system
 569  */
 570 static struct file_system_type exofs_type = {
 571         .owner          = THIS_MODULE,
 572         .name           = "exofs",
 573         .get_sb         = exofs_get_sb,
 574         .kill_sb        = generic_shutdown_super,
 575 };
 576
 577 static int __init init_exofs(void)
 578 {
 579         int err;
 580
 581         err = init_inodecache();
 582         if (err)
 583                 goto out;
 584
 585         err = register_filesystem(&exofs_type);
 586         if (err)
 587                 goto out_d;
 588
 589         return 0;
 590 out_d:
 591         destroy_inodecache();
 592 out:
 593         return err;
 594 }
 595
 596 static void __exit exit_exofs(void)
 597 {
 598         unregister_filesystem(&exofs_type);
 599         destroy_inodecache();
 600 }
 601
 602 MODULE_AUTHOR("Avishay Traeger <avishay@gmail.com>");
 603 MODULE_DESCRIPTION("exofs");
 604 MODULE_LICENSE("GPL");
 605
 606 module_init(init_exofs)
 607 module_exit(exit_exofs)