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         jfs_set_inode_flags(inode);
  59 }
  60
  61 /*
  62  * Workhorse of both fsync & write_inode
  63  */
  64 int jfs_commit_inode(struct inode *inode, int wait)
  65 {
  66         int rc = 0;
  67         tid_t tid;
  68         static int noisy = 5;
  69
  70         jfs_info("In jfs_commit_inode, inode = 0x%p", inode);
  71
  72         /*
  73          * Don't commit if inode has been committed since last being
  74          * marked dirty, or if it has been deleted.
  75          */
  76         if (inode->i_nlink == 0 || !test_cflag(COMMIT_Dirty, inode))
  77                 return 0;
  78
  79         if (isReadOnly(inode)) {
  80                 /* kernel allows writes to devices on read-only
  81                  * partitions and may think inode is dirty
  82                  */
  83                 if (!special_file(inode->i_mode) && noisy) {
  84                         jfs_err("jfs_commit_inode(0x%p) called on "
  85                                    "read-only volume", inode);
  86                         jfs_err("Is remount racy?");
  87                         noisy--;
  88                 }
  89                 return 0;
  90         }
  91
  92         tid = txBegin(inode->i_sb, COMMIT_INODE);
  93         mutex_lock(&JFS_IP(inode)->commit_mutex);
  94
  95         /*
  96          * Retest inode state after taking commit_mutex
  97          */
  98         if (inode->i_nlink && test_cflag(COMMIT_Dirty, inode))
  99                 rc = txCommit(tid, 1, &inode, wait ? COMMIT_SYNC : 0);
 100
 101         txEnd(tid);
 102         mutex_unlock(&JFS_IP(inode)->commit_mutex);
 103         return rc;
 104 }
 105
 106 int jfs_write_inode(struct inode *inode, int wait)
 107 {
 108         if (test_cflag(COMMIT_Nolink, inode))
 109                 return 0;
 110         /*
 111          * If COMMIT_DIRTY is not set, the inode isn't really dirty.
 112          * It has been committed since the last change, but was still
 113          * on the dirty inode list.
 114          */
 115          if (!test_cflag(COMMIT_Dirty, inode)) {
 116                 /* Make sure committed changes hit the disk */
 117                 jfs_flush_journal(JFS_SBI(inode->i_sb)->log, wait);
 118                 return 0;
 119          }
 120
 121         if (jfs_commit_inode(inode, wait)) {
 122                 jfs_err("jfs_write_inode: jfs_commit_inode failed!");
 123                 return -EIO;
 124         } else
 125                 return 0;
 126 }
 127
 128 void jfs_delete_inode(struct inode *inode)
 129 {
 130         jfs_info("In jfs_delete_inode, inode = 0x%p", inode);
 131
 132         if (!is_bad_inode(inode) &&
 133             (JFS_IP(inode)->fileset == FILESYSTEM_I)) {
 134                 truncate_inode_pages(&inode->i_data, 0);
 135
 136                 if (test_cflag(COMMIT_Freewmap, inode))
 137                         jfs_free_zero_link(inode);
 138
 139                 diFree(inode);
 140
 141                 /*
 142                  * Free the inode from the quota allocation.
 143                  */
 144                 DQUOT_INIT(inode);
 145                 DQUOT_FREE_INODE(inode);
 146                 DQUOT_DROP(inode);
 147         }
 148
 149         clear_inode(inode);
 150 }
 151
 152 void jfs_dirty_inode(struct inode *inode)
 153 {
 154         static int noisy = 5;
 155
 156         if (isReadOnly(inode)) {
 157                 if (!special_file(inode->i_mode) && noisy) {
 158                         /* kernel allows writes to devices on read-only
 159                          * partitions and may try to mark inode dirty
 160                          */
 161                         jfs_err("jfs_dirty_inode called on read-only volume");
 162                         jfs_err("Is remount racy?");
 163                         noisy--;
 164                 }
 165                 return;
 166         }
 167
 168         set_cflag(COMMIT_Dirty, inode);
 169 }
 170
 171 static int
 172 jfs_get_blocks(struct inode *ip, sector_t lblock, unsigned long max_blocks,
 173                         struct buffer_head *bh_result, int create)
 174 {
 175         s64 lblock64 = lblock;
 176         int rc = 0;
 177         xad_t xad;
 178         s64 xaddr;
 179         int xflag;
 180         s32 xlen = max_blocks;
 181
 182         /*
 183          * Take appropriate lock on inode
 184          */
 185         if (create)
 186                 IWRITE_LOCK(ip);
 187         else
 188                 IREAD_LOCK(ip);
 189
 190         if (((lblock64 << ip->i_sb->s_blocksize_bits) < ip->i_size) &&
 191             (!xtLookup(ip, lblock64, max_blocks, &xflag, &xaddr, &xlen, 0)) &&
 192             xaddr) {
 193                 if (xflag & XAD_NOTRECORDED) {
 194                         if (!create)
 195                                 /*
 196                                  * Allocated but not recorded, read treats
 197                                  * this as a hole
 198                                  */
 199                                 goto unlock;
 200 #ifdef _JFS_4K
 201                         XADoffset(&xad, lblock64);
 202                         XADlength(&xad, xlen);
 203                         XADaddress(&xad, xaddr);
 204 #else                           /* _JFS_4K */
 205                         /*
 206                          * As long as block size = 4K, this isn't a problem.
 207                          * We should mark the whole page not ABNR, but how
 208                          * will we know to mark the other blocks BH_New?
 209                          */
 210                         BUG();
 211 #endif                          /* _JFS_4K */
 212                         rc = extRecord(ip, &xad);
 213                         if (rc)
 214                                 goto unlock;
 215                         set_buffer_new(bh_result);
 216                 }
 217
 218                 map_bh(bh_result, ip->i_sb, xaddr);
 219                 bh_result->b_size = xlen << ip->i_blkbits;
 220                 goto unlock;
 221         }
 222         if (!create)
 223                 goto unlock;
 224
 225         /*
 226          * Allocate a new block
 227          */
 228 #ifdef _JFS_4K
 229         if ((rc = extHint(ip, lblock64 << ip->i_sb->s_blocksize_bits, &xad)))
 230                 goto unlock;
 231         rc = extAlloc(ip, xlen, lblock64, &xad, FALSE);
 232         if (rc)
 233                 goto unlock;
 234
 235         set_buffer_new(bh_result);
 236         map_bh(bh_result, ip->i_sb, addressXAD(&xad));
 237         bh_result->b_size = lengthXAD(&xad) << ip->i_blkbits;
 238
 239 #else                           /* _JFS_4K */
 240         /*
 241          * We need to do whatever it takes to keep all but the last buffers
 242          * in 4K pages - see jfs_write.c
 243          */
 244         BUG();
 245 #endif                          /* _JFS_4K */
 246
 247       unlock:
 248         /*
 249          * Release lock on inode
 250          */
 251         if (create)
 252                 IWRITE_UNLOCK(ip);
 253         else
 254                 IREAD_UNLOCK(ip);
 255         return rc;
 256 }
 257
 258 static int jfs_get_block(struct inode *ip, sector_t lblock,
 259                          struct buffer_head *bh_result, int create)
 260 {
 261         return jfs_get_blocks(ip, lblock, bh_result->b_size >> ip->i_blkbits,
 262                         bh_result, create);
 263 }
 264
 265 static int jfs_writepage(struct page *page, struct writeback_control *wbc)
 266 {
 267         return nobh_writepage(page, jfs_get_block, wbc);
 268 }
 269
 270 static int jfs_writepages(struct address_space *mapping,
 271                         struct writeback_control *wbc)
 272 {
 273         return mpage_writepages(mapping, wbc, jfs_get_block);
 274 }
 275
 276 static int jfs_readpage(struct file *file, struct page *page)
 277 {
 278         return mpage_readpage(page, jfs_get_block);
 279 }
 280
 281 static int jfs_readpages(struct file *file, struct address_space *mapping,
 282                 struct list_head *pages, unsigned nr_pages)
 283 {
 284         return mpage_readpages(mapping, pages, nr_pages, jfs_get_block);
 285 }
 286
 287 static int jfs_prepare_write(struct file *file,
 288                              struct page *page, unsigned from, unsigned to)
 289 {
 290         return nobh_prepare_write(page, from, to, jfs_get_block);
 291 }
 292
 293 static sector_t jfs_bmap(struct address_space *mapping, sector_t block)
 294 {
 295         return generic_block_bmap(mapping, block, jfs_get_block);
 296 }
 297
 298 static ssize_t jfs_direct_IO(int rw, struct kiocb *iocb,
 299         const struct iovec *iov, loff_t offset, unsigned long nr_segs)
 300 {
 301         struct file *file = iocb->ki_filp;
 302         struct inode *inode = file->f_mapping->host;
 303
 304         return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
 305                                 offset, nr_segs, jfs_get_block, NULL);
 306 }
 307
 308 struct address_space_operations jfs_aops = {
 309         .readpage       = jfs_readpage,
 310         .readpages      = jfs_readpages,
 311         .writepage      = jfs_writepage,
 312         .writepages     = jfs_writepages,
 313         .sync_page      = block_sync_page,
 314         .prepare_write  = jfs_prepare_write,
 315         .commit_write   = nobh_commit_write,
 316         .bmap           = jfs_bmap,
 317         .direct_IO      = jfs_direct_IO,
 318 };
 319
 320 /*
 321  * Guts of jfs_truncate.  Called with locks already held.  Can be called
 322  * with directory for truncating directory index table.
 323  */
 324 void jfs_truncate_nolock(struct inode *ip, loff_t length)
 325 {
 326         loff_t newsize;
 327         tid_t tid;
 328
 329         ASSERT(length >= 0);
 330
 331         if (test_cflag(COMMIT_Nolink, ip)) {
 332                 xtTruncate(0, ip, length, COMMIT_WMAP);
 333                 return;
 334         }
 335
 336         do {
 337                 tid = txBegin(ip->i_sb, 0);
 338
 339                 /*
 340                  * The commit_mutex cannot be taken before txBegin.
 341                  * txBegin may block and there is a chance the inode
 342                  * could be marked dirty and need to be committed
 343                  * before txBegin unblocks
 344                  */
 345                 mutex_lock(&JFS_IP(ip)->commit_mutex);
 346
 347                 newsize = xtTruncate(tid, ip, length,
 348                                      COMMIT_TRUNCATE | COMMIT_PWMAP);
 349                 if (newsize < 0) {
 350                         txEnd(tid);
 351                         mutex_unlock(&JFS_IP(ip)->commit_mutex);
 352                         break;
 353                 }
 354
 355                 ip->i_mtime = ip->i_ctime = CURRENT_TIME;
 356                 mark_inode_dirty(ip);
 357
 358                 txCommit(tid, 1, &ip, 0);
 359                 txEnd(tid);
 360                 mutex_unlock(&JFS_IP(ip)->commit_mutex);
 361         } while (newsize > length);     /* Truncate isn't always atomic */
 362 }
 363
 364 void jfs_truncate(struct inode *ip)
 365 {
 366         jfs_info("jfs_truncate: size = 0x%lx", (ulong) ip->i_size);
 367
 368         nobh_truncate_page(ip->i_mapping, ip->i_size);
 369
 370         IWRITE_LOCK(ip);
 371         jfs_truncate_nolock(ip, ip->i_size);
 372         IWRITE_UNLOCK(ip);
 373 }