fs/super.c

/* */
This source file includes following definitions.
get_fs_type
__wait_on_super
sync_supers
get_super
put_super
read_super
get_unnamed_dev
put_unnamed_dev
do_umount
sys_umount
do_mount
do_remount_sb
do_remount
sys_mount
mount_root
   1 /*
   2  *  linux/fs/super.c
   3  *
   4  *  Copyright (C) 1991, 1992  Linus Torvalds
   5  */
   6 
   7 /*
   8  * super.c contains code to handle the super-block tables.
   9  */
  10 #include <linux/config.h>
  11 #include <linux/sched.h>
  12 #include <linux/kernel.h>
  13 #include <linux/major.h>
  14 #include <linux/stat.h>
  15 #include <linux/errno.h>
  16 #include <linux/string.h>
  17 #include <linux/locks.h>
  18 
  19 #include <asm/system.h>
  20 #include <asm/segment.h>
  21 
  22  
  23 /*
  24  * The definition of file_systems that used to be here is now in
  25  * filesystems.c.  Now super.c contains no fs specific code.  -- jrs
  26  */
  27 
  28 extern struct file_system_type file_systems[];
  29 extern struct file_operations * get_blkfops(unsigned int);
  30 extern struct file_operations * get_chrfops(unsigned int);
  31 
  32 extern void wait_for_keypress(void);
  33 extern void fcntl_init_locks(void);
  34 
  35 extern int root_mountflags;
  36 
  37 struct super_block super_blocks[NR_SUPER];
  38 
  39 static int do_remount_sb(struct super_block *sb, int flags);
  40 
  41 /* this is initialized in init/main.c */
  42 dev_t ROOT_DEV = 0;
  43 
  44 struct file_system_type *get_fs_type(char *name)
     /*  */
  45 {
  46         int a;
  47         
  48         if (!name)
  49                 return &file_systems[0];
  50         for(a = 0 ; file_systems[a].read_super ; a++)
  51                 if (!strcmp(name,file_systems[a].name))
  52                         return(&file_systems[a]);
  53         return NULL;
  54 }
  55 
  56 void __wait_on_super(struct super_block * sb)
     /*  */
  57 {
  58         struct wait_queue wait = { current, NULL };
  59 
  60         add_wait_queue(&sb->s_wait, &wait);
  61 repeat:
  62         current->state = TASK_UNINTERRUPTIBLE;
  63         if (sb->s_lock) {
  64                 schedule();
  65                 goto repeat;
  66         }
  67         remove_wait_queue(&sb->s_wait, &wait);
  68         current->state = TASK_RUNNING;
  69 }
  70 
  71 void sync_supers(dev_t dev)
     /*  */
  72 {
  73         struct super_block * sb;
  74 
  75         for (sb = super_blocks + 0 ; sb < super_blocks + NR_SUPER ; sb++) {
  76                 if (!sb->s_dev)
  77                         continue;
  78                 if (dev && sb->s_dev != dev)
  79                         continue;
  80                 wait_on_super(sb);
  81                 if (!sb->s_dev || !sb->s_dirt)
  82                         continue;
  83                 if (dev && (dev != sb->s_dev))
  84                         continue;
  85                 if (sb->s_op && sb->s_op->write_super)
  86                         sb->s_op->write_super(sb);
  87         }
  88 }
  89 
  90 static struct super_block * get_super(dev_t dev)
     /*  */
  91 {
  92         struct super_block * s;
  93 
  94         if (!dev)
  95                 return NULL;
  96         s = 0+super_blocks;
  97         while (s < NR_SUPER+super_blocks)
  98                 if (s->s_dev == dev) {
  99                         wait_on_super(s);
 100                         if (s->s_dev == dev)
 101                                 return s;
 102                         s = 0+super_blocks;
 103                 } else
 104                         s++;
 105         return NULL;
 106 }
 107 
 108 void put_super(dev_t dev)
     /*  */
 109 {
 110         struct super_block * sb;
 111 
 112         if (dev == ROOT_DEV) {
 113                 printk("VFS: Root device %d/%d: prepare for armageddon\n",
 114                                                         MAJOR(dev), MINOR(dev));
 115                 return;
 116         }
 117         if (!(sb = get_super(dev)))
 118                 return;
 119         if (sb->s_covered) {
 120                 printk("VFS: Mounted device %d/%d - tssk, tssk\n",
 121                                                 MAJOR(dev), MINOR(dev));
 122                 return;
 123         }
 124         if (sb->s_op && sb->s_op->put_super)
 125                 sb->s_op->put_super(sb);
 126 }
 127 
 128 static struct super_block * read_super(dev_t dev,char *name,int flags,
     /*  */
 129                                        void *data, int silent)
 130 {
 131         struct super_block * s;
 132         struct file_system_type *type;
 133 
 134         if (!dev)
 135                 return NULL;
 136         check_disk_change(dev);
 137         s = get_super(dev);
 138         if (s)
 139                 return s;
 140         if (!(type = get_fs_type(name))) {
 141                 printk("VFS: on device %d/%d: get_fs_type(%s) failed\n",
 142                                                 MAJOR(dev), MINOR(dev), name);
 143                 return NULL;
 144         }
 145         for (s = 0+super_blocks ;; s++) {
 146                 if (s >= NR_SUPER+super_blocks)
 147                         return NULL;
 148                 if (!s->s_dev)
 149                         break;
 150         }
 151         s->s_dev = dev;
 152         s->s_flags = flags;
 153         if (!type->read_super(s,data, silent)) {
 154                 s->s_dev = 0;
 155                 return NULL;
 156         }
 157         s->s_dev = dev;
 158         s->s_covered = NULL;
 159         s->s_rd_only = 0;
 160         s->s_dirt = 0;
 161         return s;
 162 }
 163 
 164 /*
 165  * Unnamed block devices are dummy devices used by virtual
 166  * filesystems which don't use real block-devices.  -- jrs
 167  */
 168 
 169 static char unnamed_dev_in_use[256];
 170 
 171 static dev_t get_unnamed_dev(void)
     /*  */
 172 {
 173         static int first_use = 0;
 174         int i;
 175 
 176         if (first_use == 0) {
 177                 first_use = 1;
 178                 memset(unnamed_dev_in_use, 0, sizeof(unnamed_dev_in_use));
 179                 unnamed_dev_in_use[0] = 1; /* minor 0 (nodev) is special */
 180         }
 181         for (i = 0; i < sizeof unnamed_dev_in_use/sizeof unnamed_dev_in_use[0]; i++) {
 182                 if (!unnamed_dev_in_use[i]) {
 183                         unnamed_dev_in_use[i] = 1;
 184                         return (UNNAMED_MAJOR << 8) | i;
 185                 }
 186         }
 187         return 0;
 188 }
 189 
 190 static void put_unnamed_dev(dev_t dev)
     /*  */
 191 {
 192         if (!dev)
 193                 return;
 194         if (!unnamed_dev_in_use[dev]) {
 195                 printk("VFS: put_unnamed_dev: freeing unused device %d/%d\n",
 196                                                         MAJOR(dev), MINOR(dev));
 197                 return;
 198         }
 199         unnamed_dev_in_use[dev] = 0;
 200 }
 201 
 202 static int do_umount(dev_t dev)
     /*  */
 203 {
 204         struct super_block * sb;
 205         int retval;
 206         
 207         if (dev==ROOT_DEV) {
 208                 /* Special case for "unmounting" root.  We just try to remount
 209                    it readonly, and sync() the device. */
 210                 if (!(sb=get_super(dev)))
 211                         return -ENOENT;
 212                 if (!(sb->s_flags & MS_RDONLY)) {
 213                         retval = do_remount_sb(sb, MS_RDONLY);
 214                         if (retval)
 215                                 return retval;
 216                 }
 217                 return 0;
 218         }
 219         if (!(sb=get_super(dev)) || !(sb->s_covered))
 220                 return -ENOENT;
 221         if (!sb->s_covered->i_mount)
 222                 printk("VFS: umount(%d/%d): mounted inode has i_mount=NULL\n",
 223                                                         MAJOR(dev), MINOR(dev));
 224         if (!fs_may_umount(dev, sb->s_mounted))
 225                 return -EBUSY;
 226         sb->s_covered->i_mount = NULL;
 227         iput(sb->s_covered);
 228         sb->s_covered = NULL;
 229         iput(sb->s_mounted);
 230         sb->s_mounted = NULL;
 231         if (sb->s_op && sb->s_op->write_super && sb->s_dirt)
 232                 sb->s_op->write_super(sb);
 233         put_super(dev);
 234         return 0;
 235 }
 236 
 237 /*
 238  * Now umount can handle mount points as well as block devices.
 239  * This is important for filesystems which use unnamed block devices.
 240  *
 241  * There is a little kludge here with the dummy_inode.  The current
 242  * vfs release functions only use the r_dev field in the inode so
 243  * we give them the info they need without using a real inode.
 244  * If any other fields are ever needed by any block device release
 245  * functions, they should be faked here.  -- jrs
 246  */
 247 
 248 asmlinkage int sys_umount(char * name)
     /*  */
 249 {
 250         struct inode * inode;
 251         dev_t dev;
 252         int retval;
 253         struct inode dummy_inode;
 254         struct file_operations * fops;
 255 
 256         if (!suser())
 257                 return -EPERM;
 258         retval = namei(name,&inode);
 259         if (retval) {
 260                 retval = lnamei(name,&inode);
 261                 if (retval)
 262                         return retval;
 263         }
 264         if (S_ISBLK(inode->i_mode)) {
 265                 dev = inode->i_rdev;
 266                 if (IS_NODEV(inode)) {
 267                         iput(inode);
 268                         return -EACCES;
 269                 }
 270         } else {
 271                 if (!inode || !inode->i_sb || inode != inode->i_sb->s_mounted) {
 272                         iput(inode);
 273                         return -EINVAL;
 274                 }
 275                 dev = inode->i_sb->s_dev;
 276                 iput(inode);
 277                 memset(&dummy_inode, 0, sizeof(dummy_inode));
 278                 dummy_inode.i_rdev = dev;
 279                 inode = &dummy_inode;
 280         }
 281         if (MAJOR(dev) >= MAX_BLKDEV) {
 282                 iput(inode);
 283                 return -ENXIO;
 284         }
 285         if (!(retval = do_umount(dev)) && dev != ROOT_DEV) {
 286                 fops = get_blkfops(MAJOR(dev));
 287                 if (fops && fops->release)
 288                         fops->release(inode,NULL);
 289                 if (MAJOR(dev) == UNNAMED_MAJOR)
 290                         put_unnamed_dev(dev);
 291         }
 292         if (inode != &dummy_inode)
 293                 iput(inode);
 294         if (retval)
 295                 return retval;
 296         fsync_dev(dev);
 297         return 0;
 298 }
 299 
 300 /*
 301  * do_mount() does the actual mounting after sys_mount has done the ugly
 302  * parameter parsing. When enough time has gone by, and everything uses the
 303  * new mount() parameters, sys_mount() can then be cleaned up.
 304  *
 305  * We cannot mount a filesystem if it has active, used, or dirty inodes.
 306  * We also have to flush all inode-data for this device, as the new mount
 307  * might need new info.
 308  */
 309 static int do_mount(dev_t dev, const char * dir, char * type, int flags, void * data)
     /*  */
 310 {
 311         struct inode * dir_i;
 312         struct super_block * sb;
 313         int error;
 314 
 315         error = namei(dir,&dir_i);
 316         if (error)
 317                 return error;
 318         if (dir_i->i_count != 1 || dir_i->i_mount) {
 319                 iput(dir_i);
 320                 return -EBUSY;
 321         }
 322         if (!S_ISDIR(dir_i->i_mode)) {
 323                 iput(dir_i);
 324                 return -EPERM;
 325         }
 326         if (!fs_may_mount(dev)) {
 327                 iput(dir_i);
 328                 return -EBUSY;
 329         }
 330         sb = read_super(dev,type,flags,data,0);
 331         if (!sb || sb->s_covered) {
 332                 iput(dir_i);
 333                 return -EBUSY;
 334         }
 335         sb->s_covered = dir_i;
 336         dir_i->i_mount = sb->s_mounted;
 337         return 0;               /* we don't iput(dir_i) - see umount */
 338 }
 339 
 340 
 341 /*
 342  * Alters the mount flags of a mounted file system. Only the mount point
 343  * is used as a reference - file system type and the device are ignored.
 344  * FS-specific mount options can't be altered by remounting.
 345  */
 346 
 347 static int do_remount_sb(struct super_block *sb, int flags)
     /*  */
 348 {
 349         int retval;
 350         
 351         /* If we are remounting RDONLY, make sure there are no rw files open */
 352         if ((flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY))
 353                 if (!fs_may_remount_ro(sb->s_dev))
 354                         return -EBUSY;
 355         if (sb->s_op && sb->s_op->remount_fs) {
 356                 retval = sb->s_op->remount_fs(sb, &flags);
 357                 if (retval)
 358                         return retval;
 359         }
 360         sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) |
 361                 (flags & MS_RMT_MASK);
 362         return 0;
 363 }
 364 
 365 static int do_remount(const char *dir,int flags)
     /*  */
 366 {
 367         struct inode *dir_i;
 368         int retval;
 369 
 370         retval = namei(dir,&dir_i);
 371         if (retval)
 372                 return retval;
 373         if (dir_i != dir_i->i_sb->s_mounted) {
 374                 iput(dir_i);
 375                 return -EINVAL;
 376         }
 377         retval = do_remount_sb(dir_i->i_sb, flags);
 378         iput(dir_i);
 379         return retval;
 380 }
 381 
 382 
 383 /*
 384  * Flags is a 16-bit value that allows up to 16 non-fs dependent flags to
 385  * be given to the mount() call (ie: read-only, no-dev, no-suid etc).
 386  *
 387  * data is a (void *) that can point to any structure up to PAGE_SIZE-1 bytes, which
 388  * can contain arbitrary fs-dependent information (or be NULL).
 389  *
 390  * NOTE! As old versions of mount() didn't use this setup, the flags has to have
 391  * a special 16-bit magic number in the hight word: 0xC0ED. If this magic word
 392  * isn't present, the flags and data info isn't used, as the syscall assumes we
 393  * are talking to an older version that didn't understand them.
 394  */
 395 asmlinkage int sys_mount(char * dev_name, char * dir_name, char * type,
     /*  */
 396         unsigned long new_flags, void * data)
 397 {
 398         struct file_system_type * fstype;
 399         struct inode * inode;
 400         struct file_operations * fops;
 401         dev_t dev;
 402         int retval;
 403         char tmp[100], * t;
 404         int i;
 405         unsigned long flags = 0;
 406         unsigned long page = 0;
 407 
 408         if (!suser())
 409                 return -EPERM;
 410         if ((new_flags & (MS_MGC_MSK | MS_REMOUNT)) == (MS_MGC_VAL | MS_REMOUNT)) {
 411                 return do_remount(dir_name,new_flags & ~MS_MGC_MSK & ~MS_REMOUNT);
 412         }
 413         if (type) {
 414                 for (i = 0 ; i < 100 ; i++) {
 415                         if (TASK_SIZE <= (unsigned long) type)
 416                                 return -EFAULT;
 417                         if (!(tmp[i] = get_fs_byte(type++)))
 418                                 break;
 419                 }
 420                 t = tmp;
 421         } else
 422                 t = NULL;
 423         if (!(fstype = get_fs_type(t)))
 424                 return -ENODEV;
 425         t = fstype->name;
 426         if (fstype->requires_dev) {
 427                 retval = namei(dev_name,&inode);
 428                 if (retval)
 429                         return retval;
 430                 if (!S_ISBLK(inode->i_mode)) {
 431                         iput(inode);
 432                         return -ENOTBLK;
 433                 }
 434                 if (IS_NODEV(inode)) {
 435                         iput(inode);
 436                         return -EACCES;
 437                 }
 438                 dev = inode->i_rdev;
 439                 if (MAJOR(dev) >= MAX_BLKDEV) {
 440                         iput(inode);
 441                         return -ENXIO;
 442                 }
 443         } else {
 444                 if (!(dev = get_unnamed_dev()))
 445                         return -EMFILE;
 446                 inode = NULL;
 447         }
 448         fops = get_blkfops(MAJOR(dev));
 449         if (fops && fops->open) {
 450                 retval = fops->open(inode,NULL);
 451                 if (retval) {
 452                         iput(inode);
 453                         return retval;
 454                 }
 455         }
 456         if ((new_flags & MS_MGC_MSK) == MS_MGC_VAL) {
 457                 flags = new_flags & ~MS_MGC_MSK;
 458                 if (data) {
 459                         struct vm_area_struct * vma;
 460 
 461                         for (vma = current->mmap ; ; ) {
 462                                 if (!vma || (unsigned long) data < vma->vm_start) {
 463                                         iput(inode);
 464                                         return -EFAULT;
 465                                 }
 466                                 if ((unsigned long) data < vma->vm_end)
 467                                         break;
 468                                 vma = vma->vm_next;
 469                         }
 470                         i = vma->vm_end - (unsigned long) data;
 471                         if (PAGE_SIZE <= (unsigned long) i)
 472                                 i = PAGE_SIZE-1;
 473                         if (!(page = __get_free_page(GFP_KERNEL))) {
 474                                 iput(inode);
 475                                 return -ENOMEM;
 476                         }
 477                         memcpy_fromfs((void *) page,data,i);
 478                 }
 479         }
 480         retval = do_mount(dev,dir_name,t,flags,(void *) page);
 481         free_page(page);
 482         if (retval && fops && fops->release)
 483                 fops->release(inode,NULL);
 484         iput(inode);
 485         return retval;
 486 }
 487 
 488 void mount_root(void)
     /*  */
 489 {
 490         struct file_system_type * fs_type;
 491         struct super_block * sb;
 492         struct inode * inode;
 493 
 494         memset(super_blocks, 0, sizeof(super_blocks));
 495         fcntl_init_locks();
 496         if (MAJOR(ROOT_DEV) == FLOPPY_MAJOR) {
 497                 printk(KERN_NOTICE "VFS: Insert root floppy and press ENTER\n");
 498                 wait_for_keypress();
 499         }
 500         for (fs_type = file_systems; fs_type->read_super; fs_type++) {
 501                 if (!fs_type->requires_dev)
 502                         continue;
 503                 sb = read_super(ROOT_DEV,fs_type->name,root_mountflags,NULL,1);
 504                 if (sb) {
 505                         inode = sb->s_mounted;
 506                         inode->i_count += 3 ;   /* NOTE! it is logically used 4 times, not 1 */
 507                         sb->s_covered = inode;
 508                         sb->s_flags = root_mountflags;
 509                         current->pwd = inode;
 510                         current->root = inode;
 511                         printk ("VFS: Mounted root (%s filesystem)%s.\n",
 512                                 fs_type->name,
 513                                 (sb->s_flags & MS_RDONLY) ? " readonly" : "");
 514                         return;
 515                 }
 516         }
 517         panic("VFS: Unable to mount root");
 518 }
/* */
root/fs/super.c

DEFINITIONS
