fs/jfs/inode.c

   1 /*
   2  *   Copyright (C) International Business Machines Corp., 2000-2004
   3  *   Portions Copyright (C) Christoph Hellwig, 2001-2002
   4  *
   5  *   This program is free software;  you can redistribute it and/or modify
   6  *   it under the terms of the GNU General Public License as published by
   7  *   the Free Software Foundation; either version 2 of the License, or
   8  *   (at your option) any later version.
   9  *
  10  *   This program is distributed in the hope that it will be useful,
  11  *   but WITHOUT ANY WARRANTY;  without even the implied warranty of
  12  *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See
  13  *   the GNU General Public License for more details.
  14  *
  15  *   You should have received a copy of the GNU General Public License
  16  *   along with this program;  if not, write to the Free Software
  17  *   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
  18  */
  19
  20 #include <linux/fs.h>
  21 #include <linux/mpage.h>
  22 #include <linux/buffer_head.h>
  23 #include <linux/pagemap.h>
  24 #include <linux/quotaops.h>
  25 #include "jfs_incore.h"
  26 #include "jfs_filsys.h"
  27 #include "jfs_imap.h"
  28 #include "jfs_extent.h"
  29 #include "jfs_unicode.h"
  30 #include "jfs_debug.h"
  31
  32
  33 extern struct inode_operations jfs_dir_inode_operations;
  34 extern struct inode_operations jfs_file_inode_operations;
  35 extern struct inode_operations jfs_symlink_inode_operations;
  36 extern struct file_operations jfs_dir_operations;
  37 extern struct file_operations jfs_file_operations;
  38 struct address_space_operations jfs_aops;
  39 extern int freeZeroLink(struct inode *);
  40
  41 void jfs_read_inode(struct inode *inode)
  42 {
  43         if (diRead(inode)) {
  44                 make_bad_inode(inode);
  45                 return;
  46         }
  47
  48         if (S_ISREG(inode->i_mode)) {
  49                 inode->i_op = &jfs_file_inode_operations;
  50                 inode->i_fop = &jfs_file_operations;
  51                 inode->i_mapping->a_ops = &jfs_aops;
  52         } else if (S_ISDIR(inode->i_mode)) {
  53                 inode->i_op = &jfs_dir_inode_operations;
  54                 inode->i_fop = &jfs_dir_operations;
  55                 inode->i_mapping->a_ops = &jfs_aops;
  56                 mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
  57         } else if (S_ISLNK(inode->i_mode)) {
  58                 if (inode->i_size >= IDATASIZE) {
  59                         inode->i_op = &page_symlink_inode_operations;
  60                         inode->i_mapping->a_ops = &jfs_aops;
  61                 } else
  62                         inode->i_op = &jfs_symlink_inode_operations;
  63         } else {
  64                 inode->i_op = &jfs_file_inode_operations;
  65                 init_special_inode(inode, inode->i_mode, inode->i_rdev);
  66         }
  67 }
  68
  69 /*
  70  * Workhorse of both fsync & write_inode
  71  */
  72 int jfs_commit_inode(struct inode *inode, int wait)
  73 {
  74         int rc = 0;
  75         tid_t tid;
  76         static int noisy = 5;
  77
  78         jfs_info("In jfs_commit_inode, inode = 0x%p", inode);
  79
  80         /*
  81          * Don't commit if inode has been committed since last being
  82          * marked dirty, or if it has been deleted.
  83          */
  84         if (test_cflag(COMMIT_Nolink, inode) ||
  85             !test_cflag(COMMIT_Dirty, inode))
  86                 return 0;
  87
  88         if (isReadOnly(inode)) {
  89                 /* kernel allows writes to devices on read-only
  90                  * partitions and may think inode is dirty
  91                  */
  92                 if (!special_file(inode->i_mode) && noisy) {
  93                         jfs_err("jfs_commit_inode(0x%p) called on "
  94                                    "read-only volume", inode);
  95                         jfs_err("Is remount racy?");
  96                         noisy--;
  97                 }
  98                 return 0;
  99         }
 100
 101         tid = txBegin(inode->i_sb, COMMIT_INODE);
 102         down(&JFS_IP(inode)->commit_sem);
 103         rc = txCommit(tid, 1, &inode, wait ? COMMIT_SYNC : 0);
 104         txEnd(tid);
 105         up(&JFS_IP(inode)->commit_sem);
 106         return rc;
 107 }
 108
 109 int jfs_write_inode(struct inode *inode, int wait)
 110 {
 111         if (test_cflag(COMMIT_Nolink, inode))
 112                 return 0;
 113         /*
 114          * If COMMIT_DIRTY is not set, the inode isn't really dirty.
 115          * It has been committed since the last change, but was still
 116          * on the dirty inode list.
 117          */
 118          if (!test_cflag(COMMIT_Dirty, inode)) {
 119                 /* Make sure committed changes hit the disk */
 120                 jfs_flush_journal(JFS_SBI(inode->i_sb)->log, wait);
 121                 return 0;
 122          }
 123
 124         if (jfs_commit_inode(inode, wait)) {
 125                 jfs_err("jfs_write_inode: jfs_commit_inode failed!");
 126                 return -EIO;
 127         } else
 128                 return 0;
 129 }
 130
 131 void jfs_delete_inode(struct inode *inode)
 132 {
 133         jfs_info("In jfs_delete_inode, inode = 0x%p", inode);
 134
 135         if (test_cflag(COMMIT_Freewmap, inode))
 136                 freeZeroLink(inode);
 137
 138         diFree(inode);
 139
 140         /*
 141          * Free the inode from the quota allocation.
 142          */
 143         DQUOT_INIT(inode);
 144         DQUOT_FREE_INODE(inode);
 145         DQUOT_DROP(inode);
 146
 147         clear_inode(inode);
 148 }
 149
 150 void jfs_dirty_inode(struct inode *inode)
 151 {
 152         static int noisy = 5;
 153
 154         if (isReadOnly(inode)) {
 155                 if (!special_file(inode->i_mode) && noisy) {
 156                         /* kernel allows writes to devices on read-only
 157                          * partitions and may try to mark inode dirty
 158                          */
 159                         jfs_err("jfs_dirty_inode called on read-only volume");
 160                         jfs_err("Is remount racy?");
 161                         noisy--;
 162                 }
 163                 return;
 164         }
 165
 166         set_cflag(COMMIT_Dirty, inode);
 167 }
 168
 169 static int
 170 jfs_get_blocks(struct inode *ip, sector_t lblock, unsigned long max_blocks,
 171                         struct buffer_head *bh_result, int create)
 172 {
 173         s64 lblock64 = lblock;
 174         int no_size_check = 0;
 175         int rc = 0;
 176         int take_locks;
 177         xad_t xad;
 178         s64 xaddr;
 179         int xflag;
 180         s32 xlen;
 181
 182         /*
 183          * If this is a special inode (imap, dmap) or directory,
 184          * the lock should already be taken
 185          */
 186         take_locks = ((JFS_IP(ip)->fileset != AGGREGATE_I) &&
 187                       !S_ISDIR(ip->i_mode));
 188         /*
 189          * Take appropriate lock on inode
 190          */
 191         if (take_locks) {
 192                 if (create)
 193                         IWRITE_LOCK(ip);
 194                 else
 195                         IREAD_LOCK(ip);
 196         }
 197
 198         /*
 199          * A directory's "data" is the inode index table, but i_size is the
 200          * size of the d-tree, so don't check the offset against i_size
 201          */
 202         if (S_ISDIR(ip->i_mode))
 203                 no_size_check = 1;
 204
 205         if ((no_size_check ||
 206              ((lblock64 << ip->i_sb->s_blocksize_bits) < ip->i_size)) &&
 207             (xtLookup(ip, lblock64, max_blocks, &xflag, &xaddr, &xlen, no_size_check)
 208              == 0) && xlen) {
 209                 if (xflag & XAD_NOTRECORDED) {
 210                         if (!create)
 211                                 /*
 212                                  * Allocated but not recorded, read treats
 213                                  * this as a hole
 214                                  */
 215                                 goto unlock;
 216 #ifdef _JFS_4K
 217                         XADoffset(&xad, lblock64);
 218                         XADlength(&xad, xlen);
 219                         XADaddress(&xad, xaddr);
 220 #else                           /* _JFS_4K */
 221                         /*
 222                          * As long as block size = 4K, this isn't a problem.
 223                          * We should mark the whole page not ABNR, but how
 224                          * will we know to mark the other blocks BH_New?
 225                          */
 226                         BUG();
 227 #endif                          /* _JFS_4K */
 228                         rc = extRecord(ip, &xad);
 229                         if (rc)
 230                                 goto unlock;
 231                         set_buffer_new(bh_result);
 232                 }
 233
 234                 map_bh(bh_result, ip->i_sb, xaddr);
 235                 bh_result->b_size = xlen << ip->i_blkbits;
 236                 goto unlock;
 237         }
 238         if (!create)
 239                 goto unlock;
 240
 241         /*
 242          * Allocate a new block
 243          */
 244 #ifdef _JFS_4K
 245         if ((rc = extHint(ip, lblock64 << ip->i_sb->s_blocksize_bits, &xad)))
 246                 goto unlock;
 247         rc = extAlloc(ip, max_blocks, lblock64, &xad, FALSE);
 248         if (rc)
 249                 goto unlock;
 250
 251         set_buffer_new(bh_result);
 252         map_bh(bh_result, ip->i_sb, addressXAD(&xad));
 253         bh_result->b_size = lengthXAD(&xad) << ip->i_blkbits;
 254
 255 #else                           /* _JFS_4K */
 256         /*
 257          * We need to do whatever it takes to keep all but the last buffers
 258          * in 4K pages - see jfs_write.c
 259          */
 260         BUG();
 261 #endif                          /* _JFS_4K */
 262
 263       unlock:
 264         /*
 265          * Release lock on inode
 266          */
 267         if (take_locks) {
 268                 if (create)
 269                         IWRITE_UNLOCK(ip);
 270                 else
 271                         IREAD_UNLOCK(ip);
 272         }
 273         return rc;
 274 }
 275
 276 static int jfs_get_block(struct inode *ip, sector_t lblock,
 277                          struct buffer_head *bh_result, int create)
 278 {
 279         return jfs_get_blocks(ip, lblock, 1, bh_result, create);
 280 }
 281
 282 static int jfs_writepage(struct page *page, struct writeback_control *wbc)
 283 {
 284         return block_write_full_page(page, jfs_get_block, wbc);
 285 }
 286
 287 static int jfs_writepages(struct address_space *mapping,
 288                         struct writeback_control *wbc)
 289 {
 290         return mpage_writepages(mapping, wbc, jfs_get_block);
 291 }
 292
 293 static int jfs_readpage(struct file *file, struct page *page)
 294 {
 295         return mpage_readpage(page, jfs_get_block);
 296 }
 297
 298 static int jfs_readpages(struct file *file, struct address_space *mapping,
 299                 struct list_head *pages, unsigned nr_pages)
 300 {
 301         return mpage_readpages(mapping, pages, nr_pages, jfs_get_block);
 302 }
 303
 304 static int jfs_prepare_write(struct file *file,
 305                              struct page *page, unsigned from, unsigned to)
 306 {
 307         return nobh_prepare_write(page, from, to, jfs_get_block);
 308 }
 309
 310 static sector_t jfs_bmap(struct address_space *mapping, sector_t block)
 311 {
 312         return generic_block_bmap(mapping, block, jfs_get_block);
 313 }
 314
 315 static ssize_t jfs_direct_IO(int rw, struct kiocb *iocb,
 316         const struct iovec *iov, loff_t offset, unsigned long nr_segs)
 317 {
 318         struct file *file = iocb->ki_filp;
 319         struct inode *inode = file->f_mapping->host;
 320
 321         return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
 322                                 offset, nr_segs, jfs_get_blocks, NULL);
 323 }
 324
 325 struct address_space_operations jfs_aops = {
 326         .readpage       = jfs_readpage,
 327         .readpages      = jfs_readpages,
 328         .writepage      = jfs_writepage,
 329         .writepages     = jfs_writepages,
 330         .sync_page      = block_sync_page,
 331         .prepare_write  = jfs_prepare_write,
 332         .commit_write   = nobh_commit_write,
 333         .bmap           = jfs_bmap,
 334         .direct_IO      = jfs_direct_IO,
 335 };
 336
 337 /*
 338  * Guts of jfs_truncate.  Called with locks already held.  Can be called
 339  * with directory for truncating directory index table.
 340  */
 341 void jfs_truncate_nolock(struct inode *ip, loff_t length)
 342 {
 343         loff_t newsize;
 344         tid_t tid;
 345
 346         ASSERT(length >= 0);
 347
 348         if (test_cflag(COMMIT_Nolink, ip)) {
 349                 xtTruncate(0, ip, length, COMMIT_WMAP);
 350                 return;
 351         }
 352
 353         do {
 354                 tid = txBegin(ip->i_sb, 0);
 355
 356                 /*
 357                  * The commit_sem cannot be taken before txBegin.
 358                  * txBegin may block and there is a chance the inode
 359                  * could be marked dirty and need to be committed
 360                  * before txBegin unblocks
 361                  */
 362                 down(&JFS_IP(ip)->commit_sem);
 363
 364                 newsize = xtTruncate(tid, ip, length,
 365                                      COMMIT_TRUNCATE | COMMIT_PWMAP);
 366                 if (newsize < 0) {
 367                         txEnd(tid);
 368                         up(&JFS_IP(ip)->commit_sem);
 369                         break;
 370                 }
 371
 372                 ip->i_mtime = ip->i_ctime = CURRENT_TIME;
 373                 mark_inode_dirty(ip);
 374
 375                 txCommit(tid, 1, &ip, 0);
 376                 txEnd(tid);
 377                 up(&JFS_IP(ip)->commit_sem);
 378         } while (newsize > length);     /* Truncate isn't always atomic */
 379 }
 380
 381 void jfs_truncate(struct inode *ip)
 382 {
 383         jfs_info("jfs_truncate: size = 0x%lx", (ulong) ip->i_size);
 384
 385         nobh_truncate_page(ip->i_mapping, ip->i_size);
 386
 387         IWRITE_LOCK(ip);
 388         jfs_truncate_nolock(ip, ip->i_size);
 389         IWRITE_UNLOCK(ip);
 390 }