fs/gfs2/inode.c

   1 /*
   2  * Copyright (C) Sistina Software, Inc.  1997-2003 All rights reserved.
   3  * Copyright (C) 2004-2008 Red Hat, Inc.  All rights reserved.
   4  *
   5  * This copyrighted material is made available to anyone wishing to use,
   6  * modify, copy, or redistribute it subject to the terms and conditions
   7  * of the GNU General Public License version 2.
   8  */
   9
  10 #include <linux/sched.h>
  11 #include <linux/slab.h>
  12 #include <linux/spinlock.h>
  13 #include <linux/completion.h>
  14 #include <linux/buffer_head.h>
  15 #include <linux/posix_acl.h>
  16 #include <linux/sort.h>
  17 #include <linux/gfs2_ondisk.h>
  18 #include <linux/crc32.h>
  19 #include <linux/security.h>
  20 #include <linux/time.h>
  21
  22 #include "gfs2.h"
  23 #include "incore.h"
  24 #include "acl.h"
  25 #include "bmap.h"
  26 #include "dir.h"
  27 #include "xattr.h"
  28 #include "glock.h"
  29 #include "glops.h"
  30 #include "inode.h"
  31 #include "log.h"
  32 #include "meta_io.h"
  33 #include "quota.h"
  34 #include "rgrp.h"
  35 #include "trans.h"
  36 #include "util.h"
  37
  38 struct gfs2_inum_range_host {
  39         u64 ir_start;
  40         u64 ir_length;
  41 };
  42
  43 struct gfs2_skip_data {
  44         u64 no_addr;
  45         int skipped;
  46         int non_block;
  47 };
  48
  49 static int iget_test(struct inode *inode, void *opaque)
  50 {
  51         struct gfs2_inode *ip = GFS2_I(inode);
  52         struct gfs2_skip_data *data = opaque;
  53
  54         if (ip->i_no_addr == data->no_addr) {
  55                 if (data->non_block &&
  56                     inode->i_state & (I_FREEING|I_CLEAR|I_WILL_FREE)) {
  57                         data->skipped = 1;
  58                         return 0;
  59                 }
  60                 return 1;
  61         }
  62         return 0;
  63 }
  64
  65 static int iget_set(struct inode *inode, void *opaque)
  66 {
  67         struct gfs2_inode *ip = GFS2_I(inode);
  68         struct gfs2_skip_data *data = opaque;
  69
  70         if (data->skipped)
  71                 return -ENOENT;
  72         inode->i_ino = (unsigned long)(data->no_addr);
  73         ip->i_no_addr = data->no_addr;
  74         return 0;
  75 }
  76
  77 struct inode *gfs2_ilookup(struct super_block *sb, u64 no_addr)
  78 {
  79         unsigned long hash = (unsigned long)no_addr;
  80         struct gfs2_skip_data data;
  81
  82         data.no_addr = no_addr;
  83         data.skipped = 0;
  84         data.non_block = 0;
  85         return ilookup5(sb, hash, iget_test, &data);
  86 }
  87
  88 static struct inode *gfs2_iget(struct super_block *sb, u64 no_addr,
  89                                int non_block)
  90 {
  91         struct gfs2_skip_data data;
  92         unsigned long hash = (unsigned long)no_addr;
  93
  94         data.no_addr = no_addr;
  95         data.skipped = 0;
  96         data.non_block = non_block;
  97         return iget5_locked(sb, hash, iget_test, iget_set, &data);
  98 }
  99
 100 /**
 101  * gfs2_set_iop - Sets inode operations
 102  * @inode: The inode with correct i_mode filled in
 103  *
 104  * GFS2 lookup code fills in vfs inode contents based on info obtained
 105  * from directory entry inside gfs2_inode_lookup().
 106  */
 107
 108 static void gfs2_set_iop(struct inode *inode)
 109 {
 110         struct gfs2_sbd *sdp = GFS2_SB(inode);
 111         umode_t mode = inode->i_mode;
 112
 113         if (S_ISREG(mode)) {
 114                 inode->i_op = &gfs2_file_iops;
 115                 if (gfs2_localflocks(sdp))
 116                         inode->i_fop = &gfs2_file_fops_nolock;
 117                 else
 118                         inode->i_fop = &gfs2_file_fops;
 119         } else if (S_ISDIR(mode)) {
 120                 inode->i_op = &gfs2_dir_iops;
 121                 if (gfs2_localflocks(sdp))
 122                         inode->i_fop = &gfs2_dir_fops_nolock;
 123                 else
 124                         inode->i_fop = &gfs2_dir_fops;
 125         } else if (S_ISLNK(mode)) {
 126                 inode->i_op = &gfs2_symlink_iops;
 127         } else {
 128                 inode->i_op = &gfs2_file_iops;
 129                 init_special_inode(inode, inode->i_mode, inode->i_rdev);
 130         }
 131 }
 132
 133 /**
 134  * gfs2_inode_lookup - Lookup an inode
 135  * @sb: The super block
 136  * @no_addr: The inode number
 137  * @type: The type of the inode
 138  * non_block: Can we block on inodes that are being freed?
 139  *
 140  * Returns: A VFS inode, or an error
 141  */
 142
 143 struct inode *gfs2_inode_lookup(struct super_block *sb, unsigned int type,
 144                                 u64 no_addr, u64 no_formal_ino, int non_block)
 145 {
 146         struct inode *inode;
 147         struct gfs2_inode *ip;
 148         struct gfs2_glock *io_gl = NULL;
 149         int error;
 150
 151         inode = gfs2_iget(sb, no_addr, non_block);
 152         ip = GFS2_I(inode);
 153
 154         if (!inode)
 155                 return ERR_PTR(-ENOBUFS);
 156
 157         if (inode->i_state & I_NEW) {
 158                 struct gfs2_sbd *sdp = GFS2_SB(inode);
 159                 ip->i_no_formal_ino = no_formal_ino;
 160
 161                 error = gfs2_glock_get(sdp, no_addr, &gfs2_inode_glops, CREATE, &ip->i_gl);
 162                 if (unlikely(error))
 163                         goto fail;
 164                 ip->i_gl->gl_object = ip;
 165
 166                 error = gfs2_glock_get(sdp, no_addr, &gfs2_iopen_glops, CREATE, &io_gl);
 167                 if (unlikely(error))
 168                         goto fail_put;
 169
 170                 set_bit(GIF_INVALID, &ip->i_flags);
 171                 error = gfs2_glock_nq_init(io_gl, LM_ST_SHARED, GL_EXACT, &ip->i_iopen_gh);
 172                 if (unlikely(error))
 173                         goto fail_iopen;
 174
 175                 ip->i_iopen_gh.gh_gl->gl_object = ip;
 176                 gfs2_glock_put(io_gl);
 177                 io_gl = NULL;
 178
 179                 if (type == DT_UNKNOWN) {
 180                         /* Inode glock must be locked already */
 181                         error = gfs2_inode_refresh(GFS2_I(inode));
 182                         if (error)
 183                                 goto fail_refresh;
 184                 } else {
 185                         inode->i_mode = DT2IF(type);
 186                 }
 187
 188                 gfs2_set_iop(inode);
 189                 unlock_new_inode(inode);
 190         }
 191
 192         return inode;
 193
 194 fail_refresh:
 195         ip->i_iopen_gh.gh_gl->gl_object = NULL;
 196         gfs2_glock_dq_uninit(&ip->i_iopen_gh);
 197 fail_iopen:
 198         if (io_gl)
 199                 gfs2_glock_put(io_gl);
 200 fail_put:
 201         ip->i_gl->gl_object = NULL;
 202         gfs2_glock_put(ip->i_gl);
 203 fail:
 204         iget_failed(inode);
 205         return ERR_PTR(error);
 206 }
 207
 208 struct inode *gfs2_lookup_by_inum(struct gfs2_sbd *sdp, u64 no_addr,
 209                                   u64 *no_formal_ino, unsigned int blktype)
 210 {
 211         struct super_block *sb = sdp->sd_vfs;
 212         struct gfs2_holder i_gh;
 213         struct inode *inode = NULL;
 214         int error;
 215
 216         /* Must not read in block until block type is verified */
 217         error = gfs2_glock_nq_num(sdp, no_addr, &gfs2_inode_glops,
 218                                   LM_ST_EXCLUSIVE, GL_SKIP, &i_gh);
 219         if (error)
 220                 return ERR_PTR(error);
 221
 222         error = gfs2_check_blk_type(sdp, no_addr, blktype);
 223         if (error)
 224                 goto fail;
 225
 226         inode = gfs2_inode_lookup(sb, DT_UNKNOWN, no_addr, 0, 1);
 227         if (IS_ERR(inode))
 228                 goto fail;
 229
 230         /* Two extra checks for NFS only */
 231         if (no_formal_ino) {
 232                 error = -ESTALE;
 233                 if (GFS2_I(inode)->i_no_formal_ino != *no_formal_ino)
 234                         goto fail_iput;
 235
 236                 error = -EIO;
 237                 if (GFS2_I(inode)->i_diskflags & GFS2_DIF_SYSTEM)
 238                         goto fail_iput;
 239
 240                 error = 0;
 241         }
 242
 243 fail:
 244         gfs2_glock_dq_uninit(&i_gh);
 245         return error ? ERR_PTR(error) : inode;
 246 fail_iput:
 247         iput(inode);
 248         goto fail;
 249 }
 250
 251 static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf)
 252 {
 253         const struct gfs2_dinode *str = buf;
 254         struct timespec atime;
 255         u16 height, depth;
 256
 257         if (unlikely(ip->i_no_addr != be64_to_cpu(str->di_num.no_addr)))
 258                 goto corrupt;
 259         ip->i_no_formal_ino = be64_to_cpu(str->di_num.no_formal_ino);
 260         ip->i_inode.i_mode = be32_to_cpu(str->di_mode);
 261         ip->i_inode.i_rdev = 0;
 262         switch (ip->i_inode.i_mode & S_IFMT) {
 263         case S_IFBLK:
 264         case S_IFCHR:
 265                 ip->i_inode.i_rdev = MKDEV(be32_to_cpu(str->di_major),
 266                                            be32_to_cpu(str->di_minor));
 267                 break;
 268         };
 269
 270         ip->i_inode.i_uid = be32_to_cpu(str->di_uid);
 271         ip->i_inode.i_gid = be32_to_cpu(str->di_gid);
 272         /*
 273          * We will need to review setting the nlink count here in the
 274          * light of the forthcoming ro bind mount work. This is a reminder
 275          * to do that.
 276          */
 277         ip->i_inode.i_nlink = be32_to_cpu(str->di_nlink);
 278         i_size_write(&ip->i_inode, be64_to_cpu(str->di_size));
 279         gfs2_set_inode_blocks(&ip->i_inode, be64_to_cpu(str->di_blocks));
 280         atime.tv_sec = be64_to_cpu(str->di_atime);
 281         atime.tv_nsec = be32_to_cpu(str->di_atime_nsec);
 282         if (timespec_compare(&ip->i_inode.i_atime, &atime) < 0)
 283                 ip->i_inode.i_atime = atime;
 284         ip->i_inode.i_mtime.tv_sec = be64_to_cpu(str->di_mtime);
 285         ip->i_inode.i_mtime.tv_nsec = be32_to_cpu(str->di_mtime_nsec);
 286         ip->i_inode.i_ctime.tv_sec = be64_to_cpu(str->di_ctime);
 287         ip->i_inode.i_ctime.tv_nsec = be32_to_cpu(str->di_ctime_nsec);
 288
 289         ip->i_goal = be64_to_cpu(str->di_goal_meta);
 290         ip->i_generation = be64_to_cpu(str->di_generation);
 291
 292         ip->i_diskflags = be32_to_cpu(str->di_flags);
 293         gfs2_set_inode_flags(&ip->i_inode);
 294         height = be16_to_cpu(str->di_height);
 295         if (unlikely(height > GFS2_MAX_META_HEIGHT))
 296                 goto corrupt;
 297         ip->i_height = (u8)height;
 298
 299         depth = be16_to_cpu(str->di_depth);
 300         if (unlikely(depth > GFS2_DIR_MAX_DEPTH))
 301                 goto corrupt;
 302         ip->i_depth = (u8)depth;
 303         ip->i_entries = be32_to_cpu(str->di_entries);
 304
 305         ip->i_eattr = be64_to_cpu(str->di_eattr);
 306         if (S_ISREG(ip->i_inode.i_mode))
 307                 gfs2_set_aops(&ip->i_inode);
 308
 309         return 0;
 310 corrupt:
 311         if (gfs2_consist_inode(ip))
 312                 gfs2_dinode_print(ip);
 313         return -EIO;
 314 }
 315
 316 /**
 317  * gfs2_inode_refresh - Refresh the incore copy of the dinode
 318  * @ip: The GFS2 inode
 319  *
 320  * Returns: errno
 321  */
 322
 323 int gfs2_inode_refresh(struct gfs2_inode *ip)
 324 {
 325         struct buffer_head *dibh;
 326         int error;
 327
 328         error = gfs2_meta_inode_buffer(ip, &dibh);
 329         if (error)
 330                 return error;
 331
 332         if (gfs2_metatype_check(GFS2_SB(&ip->i_inode), dibh, GFS2_METATYPE_DI)) {
 333                 brelse(dibh);
 334                 return -EIO;
 335         }
 336
 337         error = gfs2_dinode_in(ip, dibh->b_data);
 338         brelse(dibh);
 339         clear_bit(GIF_INVALID, &ip->i_flags);
 340
 341         return error;
 342 }
 343
 344 /**
 345  * gfs2_change_nlink - Change nlink count on inode
 346  * @ip: The GFS2 inode
 347  * @diff: The change in the nlink count required
 348  *
 349  * Returns: errno
 350  */
 351 int gfs2_change_nlink(struct gfs2_inode *ip, int diff)
 352 {
 353         struct buffer_head *dibh;
 354         u32 nlink;
 355         int error;
 356
 357         BUG_ON(diff != 1 && diff != -1);
 358         nlink = ip->i_inode.i_nlink + diff;
 359
 360         /* If we are reducing the nlink count, but the new value ends up being
 361            bigger than the old one, we must have underflowed. */
 362         if (diff < 0 && nlink > ip->i_inode.i_nlink) {
 363                 if (gfs2_consist_inode(ip))
 364                         gfs2_dinode_print(ip);
 365                 return -EIO;
 366         }
 367
 368         error = gfs2_meta_inode_buffer(ip, &dibh);
 369         if (error)
 370                 return error;
 371
 372         if (diff > 0)
 373                 inc_nlink(&ip->i_inode);
 374         else
 375                 drop_nlink(&ip->i_inode);
 376
 377         ip->i_inode.i_ctime = CURRENT_TIME;
 378
 379         gfs2_trans_add_bh(ip->i_gl, dibh, 1);
 380         gfs2_dinode_out(ip, dibh->b_data);
 381         brelse(dibh);
 382         mark_inode_dirty(&ip->i_inode);
 383
 384         if (ip->i_inode.i_nlink == 0)
 385                 gfs2_unlink_di(&ip->i_inode); /* mark inode unlinked */
 386
 387         return error;
 388 }
 389
 390 struct inode *gfs2_lookup_simple(struct inode *dip, const char *name)
 391 {
 392         struct qstr qstr;
 393         struct inode *inode;
 394         gfs2_str2qstr(&qstr, name);
 395         inode = gfs2_lookupi(dip, &qstr, 1);
 396         /* gfs2_lookupi has inconsistent callers: vfs
 397          * related routines expect NULL for no entry found,
 398          * gfs2_lookup_simple callers expect ENOENT
 399          * and do not check for NULL.
 400          */
 401         if (inode == NULL)
 402                 return ERR_PTR(-ENOENT);
 403         else
 404                 return inode;
 405 }
 406
 407
 408 /**
 409  * gfs2_lookupi - Look up a filename in a directory and return its inode
 410  * @d_gh: An initialized holder for the directory glock
 411  * @name: The name of the inode to look for
 412  * @is_root: If 1, ignore the caller's permissions
 413  * @i_gh: An uninitialized holder for the new inode glock
 414  *
 415  * This can be called via the VFS filldir function when NFS is doing
 416  * a readdirplus and the inode which its intending to stat isn't
 417  * already in cache. In this case we must not take the directory glock
 418  * again, since the readdir call will have already taken that lock.
 419  *
 420  * Returns: errno
 421  */
 422
 423 struct inode *gfs2_lookupi(struct inode *dir, const struct qstr *name,
 424                            int is_root)
 425 {
 426         struct super_block *sb = dir->i_sb;
 427         struct gfs2_inode *dip = GFS2_I(dir);
 428         struct gfs2_holder d_gh;
 429         int error = 0;
 430         struct inode *inode = NULL;
 431         int unlock = 0;
 432
 433         if (!name->len || name->len > GFS2_FNAMESIZE)
 434                 return ERR_PTR(-ENAMETOOLONG);
 435
 436         if ((name->len == 1 && memcmp(name->name, ".", 1) == 0) ||
 437             (name->len == 2 && memcmp(name->name, "..", 2) == 0 &&
 438              dir == sb->s_root->d_inode)) {
 439                 igrab(dir);
 440                 return dir;
 441         }
 442
 443         if (gfs2_glock_is_locked_by_me(dip->i_gl) == NULL) {
 444                 error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, &d_gh);
 445                 if (error)
 446                         return ERR_PTR(error);
 447                 unlock = 1;
 448         }
 449
 450         if (!is_root) {
 451                 error = gfs2_permission(dir, MAY_EXEC, 0);
 452                 if (error)
 453                         goto out;
 454         }
 455
 456         inode = gfs2_dir_search(dir, name);
 457         if (IS_ERR(inode))
 458                 error = PTR_ERR(inode);
 459 out:
 460         if (unlock)
 461                 gfs2_glock_dq_uninit(&d_gh);
 462         if (error == -ENOENT)
 463                 return NULL;
 464         return inode ? inode : ERR_PTR(error);
 465 }
 466
 467 /**
 468  * create_ok - OK to create a new on-disk inode here?
 469  * @dip:  Directory in which dinode is to be created
 470  * @name:  Name of new dinode
 471  * @mode:
 472  *
 473  * Returns: errno
 474  */
 475
 476 static int create_ok(struct gfs2_inode *dip, const struct qstr *name,
 477                      unsigned int mode)
 478 {
 479         int error;
 480
 481         error = gfs2_permission(&dip->i_inode, MAY_WRITE | MAY_EXEC, 0);
 482         if (error)
 483                 return error;
 484
 485         /*  Don't create entries in an unlinked directory  */
 486         if (!dip->i_inode.i_nlink)
 487                 return -EPERM;
 488
 489         error = gfs2_dir_check(&dip->i_inode, name, NULL);
 490         switch (error) {
 491         case -ENOENT:
 492                 error = 0;
 493                 break;
 494         case 0:
 495                 return -EEXIST;
 496         default:
 497                 return error;
 498         }
 499
 500         if (dip->i_entries == (u32)-1)
 501                 return -EFBIG;
 502         if (S_ISDIR(mode) && dip->i_inode.i_nlink == (u32)-1)
 503                 return -EMLINK;
 504
 505         return 0;
 506 }
 507
 508 static void munge_mode_uid_gid(struct gfs2_inode *dip, unsigned int *mode,
 509                                unsigned int *uid, unsigned int *gid)
 510 {
 511         if (GFS2_SB(&dip->i_inode)->sd_args.ar_suiddir &&
 512             (dip->i_inode.i_mode & S_ISUID) && dip->i_inode.i_uid) {
 513                 if (S_ISDIR(*mode))
 514                         *mode |= S_ISUID;
 515                 else if (dip->i_inode.i_uid != current_fsuid())
 516                         *mode &= ~07111;
 517                 *uid = dip->i_inode.i_uid;
 518         } else
 519                 *uid = current_fsuid();
 520
 521         if (dip->i_inode.i_mode & S_ISGID) {
 522                 if (S_ISDIR(*mode))
 523                         *mode |= S_ISGID;
 524                 *gid = dip->i_inode.i_gid;
 525         } else
 526                 *gid = current_fsgid();
 527 }
 528
 529 static int alloc_dinode(struct gfs2_inode *dip, u64 *no_addr, u64 *generation)
 530 {
 531         struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
 532         int error;
 533
 534         if (gfs2_alloc_get(dip) == NULL)
 535                 return -ENOMEM;
 536
 537         dip->i_alloc->al_requested = RES_DINODE;
 538         error = gfs2_inplace_reserve(dip);
 539         if (error)
 540                 goto out;
 541
 542         error = gfs2_trans_begin(sdp, RES_RG_BIT + RES_STATFS, 0);
 543         if (error)
 544                 goto out_ipreserv;
 545
 546         error = gfs2_alloc_di(dip, no_addr, generation);
 547
 548         gfs2_trans_end(sdp);
 549
 550 out_ipreserv:
 551         gfs2_inplace_release(dip);
 552 out:
 553         gfs2_alloc_put(dip);
 554         return error;
 555 }
 556
 557 /**
 558  * init_dinode - Fill in a new dinode structure
 559  * @dip: the directory this inode is being created in
 560  * @gl: The glock covering the new inode
 561  * @inum: the inode number
 562  * @mode: the file permissions
 563  * @uid:
 564  * @gid:
 565  *
 566  */
 567
 568 static void init_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl,
 569                         const struct gfs2_inum_host *inum, unsigned int mode,
 570                         unsigned int uid, unsigned int gid,
 571                         const u64 *generation, dev_t dev, struct buffer_head **bhp)
 572 {
 573         struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
 574         struct gfs2_dinode *di;
 575         struct buffer_head *dibh;
 576         struct timespec tv = CURRENT_TIME;
 577
 578         dibh = gfs2_meta_new(gl, inum->no_addr);
 579         gfs2_trans_add_bh(gl, dibh, 1);
 580         gfs2_metatype_set(dibh, GFS2_METATYPE_DI, GFS2_FORMAT_DI);
 581         gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode));
 582         di = (struct gfs2_dinode *)dibh->b_data;
 583
 584         di->di_num.no_formal_ino = cpu_to_be64(inum->no_formal_ino);
 585         di->di_num.no_addr = cpu_to_be64(inum->no_addr);
 586         di->di_mode = cpu_to_be32(mode);
 587         di->di_uid = cpu_to_be32(uid);
 588         di->di_gid = cpu_to_be32(gid);
 589         di->di_nlink = 0;
 590         di->di_size = 0;
 591         di->di_blocks = cpu_to_be64(1);
 592         di->di_atime = di->di_mtime = di->di_ctime = cpu_to_be64(tv.tv_sec);
 593         di->di_major = cpu_to_be32(MAJOR(dev));
 594         di->di_minor = cpu_to_be32(MINOR(dev));
 595         di->di_goal_meta = di->di_goal_data = cpu_to_be64(inum->no_addr);
 596         di->di_generation = cpu_to_be64(*generation);
 597         di->di_flags = 0;
 598
 599         if (S_ISREG(mode)) {
 600                 if ((dip->i_diskflags & GFS2_DIF_INHERIT_JDATA) ||
 601                     gfs2_tune_get(sdp, gt_new_files_jdata))
 602                         di->di_flags |= cpu_to_be32(GFS2_DIF_JDATA);
 603         } else if (S_ISDIR(mode)) {
 604                 di->di_flags |= cpu_to_be32(dip->i_diskflags &
 605                                             GFS2_DIF_INHERIT_JDATA);
 606         }
 607
 608         di->__pad1 = 0;
 609         di->di_payload_format = cpu_to_be32(S_ISDIR(mode) ? GFS2_FORMAT_DE : 0);
 610         di->di_height = 0;
 611         di->__pad2 = 0;
 612         di->__pad3 = 0;
 613         di->di_depth = 0;
 614         di->di_entries = 0;
 615         memset(&di->__pad4, 0, sizeof(di->__pad4));
 616         di->di_eattr = 0;
 617         di->di_atime_nsec = cpu_to_be32(tv.tv_nsec);
 618         di->di_mtime_nsec = cpu_to_be32(tv.tv_nsec);
 619         di->di_ctime_nsec = cpu_to_be32(tv.tv_nsec);
 620         memset(&di->di_reserved, 0, sizeof(di->di_reserved));
 621
 622         set_buffer_uptodate(dibh);
 623
 624         *bhp = dibh;
 625 }
 626
 627 static int make_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl,
 628                        unsigned int mode, const struct gfs2_inum_host *inum,
 629                        const u64 *generation, dev_t dev, struct buffer_head **bhp)
 630 {
 631         struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
 632         unsigned int uid, gid;
 633         int error;
 634
 635         munge_mode_uid_gid(dip, &mode, &uid, &gid);
 636         if (!gfs2_alloc_get(dip))
 637                 return -ENOMEM;
 638
 639         error = gfs2_quota_lock(dip, uid, gid);
 640         if (error)
 641                 goto out;
 642
 643         error = gfs2_quota_check(dip, uid, gid);
 644         if (error)
 645                 goto out_quota;
 646
 647         error = gfs2_trans_begin(sdp, RES_DINODE + RES_QUOTA, 0);
 648         if (error)
 649                 goto out_quota;
 650
 651         init_dinode(dip, gl, inum, mode, uid, gid, generation, dev, bhp);
 652         gfs2_quota_change(dip, +1, uid, gid);
 653         gfs2_trans_end(sdp);
 654
 655 out_quota:
 656         gfs2_quota_unlock(dip);
 657 out:
 658         gfs2_alloc_put(dip);
 659         return error;
 660 }
 661
 662 static int link_dinode(struct gfs2_inode *dip, const struct qstr *name,
 663                        struct gfs2_inode *ip)
 664 {
 665         struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
 666         struct gfs2_alloc *al;
 667         int alloc_required;
 668         struct buffer_head *dibh;
 669         int error;
 670
 671         al = gfs2_alloc_get(dip);
 672         if (!al)
 673                 return -ENOMEM;
 674
 675         error = gfs2_quota_lock(dip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
 676         if (error)
 677                 goto fail;
 678
 679         error = alloc_required = gfs2_diradd_alloc_required(&dip->i_inode, name);
 680         if (alloc_required < 0)
 681                 goto fail_quota_locks;
 682         if (alloc_required) {
 683                 error = gfs2_quota_check(dip, dip->i_inode.i_uid, dip->i_inode.i_gid);
 684                 if (error)
 685                         goto fail_quota_locks;
 686
 687                 al->al_requested = sdp->sd_max_dirres;
 688
 689                 error = gfs2_inplace_reserve(dip);
 690                 if (error)
 691                         goto fail_quota_locks;
 692
 693                 error = gfs2_trans_begin(sdp, sdp->sd_max_dirres +
 694                                          al->al_rgd->rd_length +
 695                                          2 * RES_DINODE +
 696                                          RES_STATFS + RES_QUOTA, 0);
 697                 if (error)
 698                         goto fail_ipreserv;
 699         } else {
 700                 error = gfs2_trans_begin(sdp, RES_LEAF + 2 * RES_DINODE, 0);
 701                 if (error)
 702                         goto fail_quota_locks;
 703         }
 704
 705         error = gfs2_dir_add(&dip->i_inode, name, ip, IF2DT(ip->i_inode.i_mode));
 706         if (error)
 707                 goto fail_end_trans;
 708
 709         error = gfs2_meta_inode_buffer(ip, &dibh);
 710         if (error)
 711                 goto fail_end_trans;
 712         ip->i_inode.i_nlink = 1;
 713         gfs2_trans_add_bh(ip->i_gl, dibh, 1);
 714         gfs2_dinode_out(ip, dibh->b_data);
 715         brelse(dibh);
 716         return 0;
 717
 718 fail_end_trans:
 719         gfs2_trans_end(sdp);
 720
 721 fail_ipreserv:
 722         if (dip->i_alloc->al_rgd)
 723                 gfs2_inplace_release(dip);
 724
 725 fail_quota_locks:
 726         gfs2_quota_unlock(dip);
 727
 728 fail:
 729         gfs2_alloc_put(dip);
 730         return error;
 731 }
 732
 733 static int gfs2_security_init(struct gfs2_inode *dip, struct gfs2_inode *ip,
 734                               const struct qstr *qstr)
 735 {
 736         int err;
 737         size_t len;
 738         void *value;
 739         char *name;
 740
 741         err = security_inode_init_security(&ip->i_inode, &dip->i_inode, qstr,
 742                                            &name, &value, &len);
 743
 744         if (err) {
 745                 if (err == -EOPNOTSUPP)
 746                         return 0;
 747                 return err;
 748         }
 749
 750         err = __gfs2_xattr_set(&ip->i_inode, name, value, len, 0,
 751                                GFS2_EATYPE_SECURITY);
 752         kfree(value);
 753         kfree(name);
 754
 755         return err;
 756 }
 757
 758 /**
 759  * gfs2_createi - Create a new inode
 760  * @ghs: An array of two holders
 761  * @name: The name of the new file
 762  * @mode: the permissions on the new inode
 763  *
 764  * @ghs[0] is an initialized holder for the directory
 765  * @ghs[1] is the holder for the inode lock
 766  *
 767  * If the return value is not NULL, the glocks on both the directory and the new
 768  * file are held.  A transaction has been started and an inplace reservation
 769  * is held, as well.
 770  *
 771  * Returns: An inode
 772  */
 773
 774 struct inode *gfs2_createi(struct gfs2_holder *ghs, const struct qstr *name,
 775                            unsigned int mode, dev_t dev)
 776 {
 777         struct inode *inode = NULL;
 778         struct gfs2_inode *dip = ghs->gh_gl->gl_object;
 779         struct inode *dir = &dip->i_inode;
 780         struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
 781         struct gfs2_inum_host inum = { .no_addr = 0, .no_formal_ino = 0 };
 782         int error;
 783         u64 generation;
 784         struct buffer_head *bh = NULL;
 785
 786         if (!name->len || name->len > GFS2_FNAMESIZE)
 787                 return ERR_PTR(-ENAMETOOLONG);
 788
 789         gfs2_holder_reinit(LM_ST_EXCLUSIVE, 0, ghs);
 790         error = gfs2_glock_nq(ghs);
 791         if (error)
 792                 goto fail;
 793
 794         error = create_ok(dip, name, mode);
 795         if (error)
 796                 goto fail_gunlock;
 797
 798         error = alloc_dinode(dip, &inum.no_addr, &generation);
 799         if (error)
 800                 goto fail_gunlock;
 801         inum.no_formal_ino = generation;
 802
 803         error = gfs2_glock_nq_num(sdp, inum.no_addr, &gfs2_inode_glops,
 804                                   LM_ST_EXCLUSIVE, GL_SKIP, ghs + 1);
 805         if (error)
 806                 goto fail_gunlock;
 807
 808         error = make_dinode(dip, ghs[1].gh_gl, mode, &inum, &generation, dev, &bh);
 809         if (error)
 810                 goto fail_gunlock2;
 811
 812         inode = gfs2_inode_lookup(dir->i_sb, IF2DT(mode), inum.no_addr,
 813                                   inum.no_formal_ino, 0);
 814         if (IS_ERR(inode))
 815                 goto fail_gunlock2;
 816
 817         error = gfs2_inode_refresh(GFS2_I(inode));
 818         if (error)
 819                 goto fail_gunlock2;
 820
 821         error = gfs2_acl_create(dip, inode);
 822         if (error)
 823                 goto fail_gunlock2;
 824
 825         error = gfs2_security_init(dip, GFS2_I(inode), name);
 826         if (error)
 827                 goto fail_gunlock2;
 828
 829         error = link_dinode(dip, name, GFS2_I(inode));
 830         if (error)
 831                 goto fail_gunlock2;
 832
 833         if (bh)
 834                 brelse(bh);
 835         return inode;
 836
 837 fail_gunlock2:
 838         gfs2_glock_dq_uninit(ghs + 1);
 839         if (inode && !IS_ERR(inode))
 840                 iput(inode);
 841 fail_gunlock:
 842         gfs2_glock_dq(ghs);
 843 fail:
 844         if (bh)
 845                 brelse(bh);
 846         return ERR_PTR(error);
 847 }
 848
 849 static int __gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
 850 {
 851         struct inode *inode = &ip->i_inode;
 852         struct buffer_head *dibh;
 853         int error;
 854
 855         error = gfs2_meta_inode_buffer(ip, &dibh);
 856         if (error)
 857                 return error;
 858
 859         setattr_copy(inode, attr);
 860         mark_inode_dirty(inode);
 861         gfs2_trans_add_bh(ip->i_gl, dibh, 1);
 862         gfs2_dinode_out(ip, dibh->b_data);
 863         brelse(dibh);
 864         return 0;
 865 }
 866
 867 /**
 868  * gfs2_setattr_simple -
 869  * @ip:
 870  * @attr:
 871  *
 872  * Called with a reference on the vnode.
 873  *
 874  * Returns: errno
 875  */
 876
 877 int gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
 878 {
 879         int error;
 880
 881         if (current->journal_info)
 882                 return __gfs2_setattr_simple(ip, attr);
 883
 884         error = gfs2_trans_begin(GFS2_SB(&ip->i_inode), RES_DINODE, 0);
 885         if (error)
 886                 return error;
 887
 888         error = __gfs2_setattr_simple(ip, attr);
 889         gfs2_trans_end(GFS2_SB(&ip->i_inode));
 890         return error;
 891 }
 892
 893 void gfs2_dinode_out(const struct gfs2_inode *ip, void *buf)
 894 {
 895         struct gfs2_dinode *str = buf;
 896
 897         str->di_header.mh_magic = cpu_to_be32(GFS2_MAGIC);
 898         str->di_header.mh_type = cpu_to_be32(GFS2_METATYPE_DI);
 899         str->di_header.mh_format = cpu_to_be32(GFS2_FORMAT_DI);
 900         str->di_num.no_addr = cpu_to_be64(ip->i_no_addr);
 901         str->di_num.no_formal_ino = cpu_to_be64(ip->i_no_formal_ino);
 902         str->di_mode = cpu_to_be32(ip->i_inode.i_mode);
 903         str->di_uid = cpu_to_be32(ip->i_inode.i_uid);
 904         str->di_gid = cpu_to_be32(ip->i_inode.i_gid);
 905         str->di_nlink = cpu_to_be32(ip->i_inode.i_nlink);
 906         str->di_size = cpu_to_be64(i_size_read(&ip->i_inode));
 907         str->di_blocks = cpu_to_be64(gfs2_get_inode_blocks(&ip->i_inode));
 908         str->di_atime = cpu_to_be64(ip->i_inode.i_atime.tv_sec);
 909         str->di_mtime = cpu_to_be64(ip->i_inode.i_mtime.tv_sec);
 910         str->di_ctime = cpu_to_be64(ip->i_inode.i_ctime.tv_sec);
 911
 912         str->di_goal_meta = cpu_to_be64(ip->i_goal);
 913         str->di_goal_data = cpu_to_be64(ip->i_goal);
 914         str->di_generation = cpu_to_be64(ip->i_generation);
 915
 916         str->di_flags = cpu_to_be32(ip->i_diskflags);
 917         str->di_height = cpu_to_be16(ip->i_height);
 918         str->di_payload_format = cpu_to_be32(S_ISDIR(ip->i_inode.i_mode) &&
 919                                              !(ip->i_diskflags & GFS2_DIF_EXHASH) ?
 920                                              GFS2_FORMAT_DE : 0);
 921         str->di_depth = cpu_to_be16(ip->i_depth);
 922         str->di_entries = cpu_to_be32(ip->i_entries);
 923
 924         str->di_eattr = cpu_to_be64(ip->i_eattr);
 925         str->di_atime_nsec = cpu_to_be32(ip->i_inode.i_atime.tv_nsec);
 926         str->di_mtime_nsec = cpu_to_be32(ip->i_inode.i_mtime.tv_nsec);
 927         str->di_ctime_nsec = cpu_to_be32(ip->i_inode.i_ctime.tv_nsec);
 928 }
 929
 930 void gfs2_dinode_print(const struct gfs2_inode *ip)
 931 {
 932         printk(KERN_INFO "  no_formal_ino = %llu\n",
 933                (unsigned long long)ip->i_no_formal_ino);
 934         printk(KERN_INFO "  no_addr = %llu\n",
 935                (unsigned long long)ip->i_no_addr);
 936         printk(KERN_INFO "  i_size = %llu\n",
 937                (unsigned long long)i_size_read(&ip->i_inode));
 938         printk(KERN_INFO "  blocks = %llu\n",
 939                (unsigned long long)gfs2_get_inode_blocks(&ip->i_inode));
 940         printk(KERN_INFO "  i_goal = %llu\n",
 941                (unsigned long long)ip->i_goal);
 942         printk(KERN_INFO "  i_diskflags = 0x%.8X\n", ip->i_diskflags);
 943         printk(KERN_INFO "  i_height = %u\n", ip->i_height);
 944         printk(KERN_INFO "  i_depth = %u\n", ip->i_depth);
 945         printk(KERN_INFO "  i_entries = %u\n", ip->i_entries);
 946         printk(KERN_INFO "  i_eattr = %llu\n",
 947                (unsigned long long)ip->i_eattr);
 948 }
 949