fs/ext2/file.c

   1 /*
   2  *  linux/fs/ext2/file.c
   3  *
   4  * Copyright (C) 1992, 1993, 1994, 1995
   5  * Remy Card (card@masi.ibp.fr)
   6  * Laboratoire MASI - Institut Blaise Pascal
   7  * Universite Pierre et Marie Curie (Paris VI)
   8  *
   9  *  from
  10  *
  11  *  linux/fs/minix/file.c
  12  *
  13  *  Copyright (C) 1991, 1992  Linus Torvalds
  14  *
  15  *  ext2 fs regular file handling primitives
  16  *
  17  *  64-bit file support on 64-bit platforms by Jakub Jelinek
  18  *      (jj@sunsite.ms.mff.cuni.cz)
  19  */
  20
  21 #include <asm/uaccess.h>
  22 #include <asm/system.h>
  23
  24 #include <linux/errno.h>
  25 #include <linux/fs.h>
  26 #include <linux/ext2_fs.h>
  27 #include <linux/fcntl.h>
  28 #include <linux/sched.h>
  29 #include <linux/stat.h>
  30 #include <linux/locks.h>
  31 #include <linux/mm.h>
  32 #include <linux/pagemap.h>
  33
  34 #define NBUF    32
  35
  36 #define MIN(a,b) (((a)<(b))?(a):(b))
  37 #define MAX(a,b) (((a)>(b))?(a):(b))
  38
  39 static long long ext2_file_lseek(struct file *, long long, int);
  40 static ssize_t ext2_file_write (struct file *, const char *, size_t, loff_t *);
  41 static int ext2_release_file (struct inode *, struct file *);
  42 #if BITS_PER_LONG < 64
  43 static int ext2_open_file (struct inode *, struct file *);
  44
  45 #else
  46
  47 #define EXT2_MAX_SIZE(bits)                                                     \
  48         (((EXT2_NDIR_BLOCKS + (1LL << (bits - 2)) +                             \
  49            (1LL << (bits - 2)) * (1LL << (bits - 2)) +                          \
  50            (1LL << (bits - 2)) * (1LL << (bits - 2)) * (1LL << (bits - 2))) *   \
  51           (1LL << bits)) - 1)
  52
  53 static long long ext2_max_sizes[] = {
  54 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  55 EXT2_MAX_SIZE(10), EXT2_MAX_SIZE(11), EXT2_MAX_SIZE(12), EXT2_MAX_SIZE(13)
  56 };
  57
  58 #endif
  59
  60 /*
  61  * We have mostly NULL's here: the current defaults are ok for
  62  * the ext2 filesystem.
  63  */
  64 static struct file_operations ext2_file_operations = {
  65         ext2_file_lseek,        /* lseek */
  66         generic_file_read,      /* read */
  67         ext2_file_write,        /* write */
  68         NULL,                   /* readdir - bad */
  69         NULL,                   /* poll - default */
  70         ext2_ioctl,             /* ioctl */
  71         generic_file_mmap,      /* mmap */
  72 #if BITS_PER_LONG == 64
  73         NULL,                   /* no special open is needed */
  74 #else
  75         ext2_open_file,
  76 #endif
  77         NULL,                   /* flush */
  78         ext2_release_file,      /* release */
  79         ext2_sync_file,         /* fsync */
  80         NULL,                   /* fasync */
  81         NULL,                   /* check_media_change */
  82         NULL                    /* revalidate */
  83 };
  84
  85 struct inode_operations ext2_file_inode_operations = {
  86         &ext2_file_operations,/* default file operations */
  87         NULL,                   /* create */
  88         NULL,                   /* lookup */
  89         NULL,                   /* link */
  90         NULL,                   /* unlink */
  91         NULL,                   /* symlink */
  92         NULL,                   /* mkdir */
  93         NULL,                   /* rmdir */
  94         NULL,                   /* mknod */
  95         NULL,                   /* rename */
  96         NULL,                   /* readlink */
  97         NULL,                   /* follow_link */
  98         generic_readpage,       /* readpage */
  99         NULL,                   /* writepage */
 100         ext2_bmap,              /* bmap */
 101         ext2_truncate,          /* truncate */
 102         ext2_permission,        /* permission */
 103         NULL                    /* smap */
 104 };
 105
 106 /*
 107  * Make sure the offset never goes beyond the 32-bit mark..
 108  */
 109 static long long ext2_file_lseek(
 110         struct file *file,
 111         long long offset,
 112         int origin)
 113 {
 114         struct inode *inode = file->f_dentry->d_inode;
 115
 116         switch (origin) {
 117                 case 2:
 118                         offset += inode->i_size;
 119                         break;
 120                 case 1:
 121                         offset += file->f_pos;
 122         }
 123         if (((unsigned long long) offset >> 32) != 0) {
 124 #if BITS_PER_LONG < 64
 125                 return -EINVAL;
 126 #else
 127                 if (offset > ext2_max_sizes[EXT2_BLOCK_SIZE_BITS(inode->i_sb)])
 128                         return -EINVAL;
 129 #endif
 130         }
 131         if (offset != file->f_pos) {
 132                 file->f_pos = offset;
 133                 file->f_reada = 0;
 134                 file->f_version = ++event;
 135         }
 136         return offset;
 137 }
 138
 139 static inline void remove_suid(struct inode *inode)
 140 {
 141         unsigned int mode;
 142
 143         /* set S_IGID if S_IXGRP is set, and always set S_ISUID */
 144         mode = (inode->i_mode & S_IXGRP)*(S_ISGID/S_IXGRP) | S_ISUID;
 145
 146         /* was any of the uid bits set? */
 147         mode &= inode->i_mode;
 148         if (mode && !capable(CAP_FSETID)) {
 149                 inode->i_mode &= ~mode;
 150                 mark_inode_dirty(inode);
 151         }
 152 }
 153
 154 static ssize_t ext2_file_write (struct file * filp, const char * buf,
 155                                 size_t count, loff_t *ppos)
 156 {
 157         struct inode * inode = filp->f_dentry->d_inode;
 158         off_t pos;
 159         long block;
 160         int offset;
 161         int written, c;
 162         struct buffer_head * bh, *bufferlist[NBUF];
 163         struct super_block * sb;
 164         int err;
 165         int i,buffercount,write_error;
 166
 167         /* POSIX: mtime/ctime may not change for 0 count */
 168         if (!count)
 169                 return 0;
 170         write_error = buffercount = 0;
 171         if (!inode) {
 172                 printk("ext2_file_write: inode = NULL\n");
 173                 return -EINVAL;
 174         }
 175         sb = inode->i_sb;
 176         if (sb->s_flags & MS_RDONLY)
 177                 /*
 178                  * This fs has been automatically remounted ro because of errors
 179                  */
 180                 return -ENOSPC;
 181
 182         if (!S_ISREG(inode->i_mode)) {
 183                 ext2_warning (sb, "ext2_file_write", "mode = %07o",
 184                               inode->i_mode);
 185                 return -EINVAL;
 186         }
 187         remove_suid(inode);
 188
 189         if (filp->f_flags & O_APPEND)
 190                 pos = inode->i_size;
 191         else {
 192                 pos = *ppos;
 193                 if (pos != *ppos)
 194                         return -EINVAL;
 195 #if BITS_PER_LONG >= 64
 196                 if (pos > ext2_max_sizes[EXT2_BLOCK_SIZE_BITS(sb)])
 197                         return -EINVAL;
 198 #endif
 199         }
 200
 201         /* Check for overflow.. */
 202 #if BITS_PER_LONG < 64
 203         if (pos > (__u32) (pos + count)) {
 204                 count = ~pos; /* == 0xFFFFFFFF - pos */
 205                 if (!count)
 206                         return -EFBIG;
 207         }
 208 #else
 209         {
 210                 off_t max = ext2_max_sizes[EXT2_BLOCK_SIZE_BITS(sb)];
 211
 212                 if (pos + count > max) {
 213                         count = max - pos;
 214                         if (!count)
 215                                 return -EFBIG;
 216                 }
 217                 if (((pos + count) >> 32) &&
 218                     !(sb->u.ext2_sb.s_es->s_feature_ro_compat &
 219                       cpu_to_le32(EXT2_FEATURE_RO_COMPAT_LARGE_FILE))) {
 220                         /* If this is the first large file created, add a flag
 221                            to the superblock */
 222                         sb->u.ext2_sb.s_es->s_feature_ro_compat |=
 223                                 cpu_to_le32(EXT2_FEATURE_RO_COMPAT_LARGE_FILE);
 224                         mark_buffer_dirty(sb->u.ext2_sb.s_sbh, 1);
 225                 }
 226         }
 227 #endif
 228
 229         /*
 230          * If a file has been opened in synchronous mode, we have to ensure
 231          * that meta-data will also be written synchronously.  Thus, we
 232          * set the i_osync field.  This field is tested by the allocation
 233          * routines.
 234          */
 235         if (filp->f_flags & O_SYNC)
 236                 inode->u.ext2_i.i_osync++;
 237         block = pos >> EXT2_BLOCK_SIZE_BITS(sb);
 238         offset = pos & (sb->s_blocksize - 1);
 239         c = sb->s_blocksize - offset;
 240         written = 0;
 241         do {
 242                 bh = ext2_getblk (inode, block, 1, &err);
 243                 if (!bh) {
 244                         if (!written)
 245                                 written = err;
 246                         break;
 247                 }
 248                 if (c > count)
 249                         c = count;
 250                 if (c != sb->s_blocksize && !buffer_uptodate(bh)) {
 251                         ll_rw_block (READ, 1, &bh);
 252                         wait_on_buffer (bh);
 253                         if (!buffer_uptodate(bh)) {
 254                                 brelse (bh);
 255                                 if (!written)
 256                                         written = -EIO;
 257                                 break;
 258                         }
 259                 }
 260                 c -= copy_from_user (bh->b_data + offset, buf, c);
 261                 if (!c) {
 262                         brelse(bh);
 263                         if (!written)
 264                                 written = -EFAULT;
 265                         break;
 266                 }
 267                 update_vm_cache(inode, pos, bh->b_data + offset, c);
 268                 pos += c;
 269                 written += c;
 270                 buf += c;
 271                 count -= c;
 272                 mark_buffer_uptodate(bh, 1);
 273                 mark_buffer_dirty(bh, 0);
 274
 275                 /* Mark the buffer untouched if we'll move on to the next one.. */
 276                 if (!(pos & (sb->s_blocksize-1)))
 277                         clear_bit(BH_Touched, &bh->b_state);
 278
 279                 if (filp->f_flags & O_SYNC)
 280                         bufferlist[buffercount++] = bh;
 281                 else
 282                         brelse(bh);
 283                 if (buffercount == NBUF){
 284                         ll_rw_block(WRITE, buffercount, bufferlist);
 285                         for(i=0; i<buffercount; i++){
 286                                 wait_on_buffer(bufferlist[i]);
 287                                 if (!buffer_uptodate(bufferlist[i]))
 288                                         write_error=1;
 289                                 brelse(bufferlist[i]);
 290                         }
 291                         buffercount=0;
 292                 }
 293                 if(write_error)
 294                         break;
 295                 block++;
 296                 offset = 0;
 297                 c = sb->s_blocksize;
 298         } while (count);
 299         if ( buffercount ){
 300                 ll_rw_block(WRITE, buffercount, bufferlist);
 301                 for(i=0; i<buffercount; i++){
 302                         wait_on_buffer(bufferlist[i]);
 303                         if (!buffer_uptodate(bufferlist[i]))
 304                                 write_error=1;
 305                         brelse(bufferlist[i]);
 306                 }
 307         }
 308         if (pos > inode->i_size)
 309                 inode->i_size = pos;
 310         if (filp->f_flags & O_SYNC)
 311                 inode->u.ext2_i.i_osync--;
 312         inode->i_ctime = inode->i_mtime = CURRENT_TIME;
 313         *ppos = pos;
 314         mark_inode_dirty(inode);
 315         return written;
 316 }
 317
 318 /*
 319  * Called when an inode is released. Note that this is different
 320  * from ext2_file_open: open gets called at every open, but release
 321  * gets called only when /all/ the files are closed.
 322  */
 323 static int ext2_release_file (struct inode * inode, struct file * filp)
 324 {
 325         if (filp->f_mode & FMODE_WRITE)
 326                 ext2_discard_prealloc (inode);
 327         return 0;
 328 }
 329
 330 #if BITS_PER_LONG < 64
 331 /*
 332  * Called when an inode is about to be open.
 333  * We use this to disallow opening RW large files on 32bit systems.
 334  */
 335 static int ext2_open_file (struct inode * inode, struct file * filp)
 336 {
 337         if (inode->u.ext2_i.i_high_size && (filp->f_mode & FMODE_WRITE))
 338                 return -EFBIG;
 339         return 0;
 340 }
 341 #endif