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