NFS: Ensure that mmapped pages remain stable during writeback
[linux-2.6.git] / init / do_mounts.c
blob2974c8b3b351b4a15b38acca4478e4032fece9fa
1 #include <linux/module.h>
2 #include <linux/sched.h>
3 #include <linux/ctype.h>
4 #include <linux/fd.h>
5 #include <linux/tty.h>
6 #include <linux/suspend.h>
7 #include <linux/root_dev.h>
8 #include <linux/security.h>
9 #include <linux/delay.h>
10 #include <linux/genhd.h>
11 #include <linux/mount.h>
12 #include <linux/device.h>
13 #include <linux/init.h>
14 #include <linux/fs.h>
15 #include <linux/initrd.h>
16 #include <linux/async.h>
17 #include <linux/fs_struct.h>
18 #include <linux/slab.h>
20 #include <linux/nfs_fs.h>
21 #include <linux/nfs_fs_sb.h>
22 #include <linux/nfs_mount.h>
24 #include "do_mounts.h"
26 int __initdata rd_doload; /* 1 = load RAM disk, 0 = don't load */
28 int root_mountflags = MS_RDONLY | MS_SILENT;
29 static char * __initdata root_device_name;
30 static char __initdata saved_root_name[64];
31 static int root_wait;
33 dev_t ROOT_DEV;
35 static int __init load_ramdisk(char *str)
37 rd_doload = simple_strtol(str,NULL,0) & 3;
38 return 1;
40 __setup("load_ramdisk=", load_ramdisk);
42 static int __init readonly(char *str)
44 if (*str)
45 return 0;
46 root_mountflags |= MS_RDONLY;
47 return 1;
50 static int __init readwrite(char *str)
52 if (*str)
53 return 0;
54 root_mountflags &= ~MS_RDONLY;
55 return 1;
58 __setup("ro", readonly);
59 __setup("rw", readwrite);
61 #ifdef CONFIG_BLOCK
62 /**
63 * match_dev_by_uuid - callback for finding a partition using its uuid
64 * @dev: device passed in by the caller
65 * @data: opaque pointer to a 36 byte char array with a UUID
67 * Returns 1 if the device matches, and 0 otherwise.
69 static int match_dev_by_uuid(struct device *dev, void *data)
71 u8 *uuid = data;
72 struct hd_struct *part = dev_to_part(dev);
74 if (!part->info)
75 goto no_match;
77 if (memcmp(uuid, part->info->uuid, sizeof(part->info->uuid)))
78 goto no_match;
80 return 1;
81 no_match:
82 return 0;
86 /**
87 * devt_from_partuuid - looks up the dev_t of a partition by its UUID
88 * @uuid: min 36 byte char array containing a hex ascii UUID
90 * The function will return the first partition which contains a matching
91 * UUID value in its partition_meta_info struct. This does not search
92 * by filesystem UUIDs.
94 * If @uuid is followed by a "/PARTNROFF=%d", then the number will be
95 * extracted and used as an offset from the partition identified by the UUID.
97 * Returns the matching dev_t on success or 0 on failure.
99 static dev_t devt_from_partuuid(char *uuid_str)
101 dev_t res = 0;
102 struct device *dev = NULL;
103 u8 uuid[16];
104 struct gendisk *disk;
105 struct hd_struct *part;
106 int offset = 0;
108 if (strlen(uuid_str) < 36)
109 goto done;
111 /* Check for optional partition number offset attributes. */
112 if (uuid_str[36]) {
113 char c = 0;
114 /* Explicitly fail on poor PARTUUID syntax. */
115 if (sscanf(&uuid_str[36],
116 "/PARTNROFF=%d%c", &offset, &c) != 1) {
117 printk(KERN_ERR "VFS: PARTUUID= is invalid.\n"
118 "Expected PARTUUID=<valid-uuid-id>[/PARTNROFF=%%d]\n");
119 if (root_wait)
120 printk(KERN_ERR
121 "Disabling rootwait; root= is invalid.\n");
122 root_wait = 0;
123 goto done;
127 /* Pack the requested UUID in the expected format. */
128 part_pack_uuid(uuid_str, uuid);
130 dev = class_find_device(&block_class, NULL, uuid, &match_dev_by_uuid);
131 if (!dev)
132 goto done;
134 res = dev->devt;
136 /* Attempt to find the partition by offset. */
137 if (!offset)
138 goto no_offset;
140 res = 0;
141 disk = part_to_disk(dev_to_part(dev));
142 part = disk_get_part(disk, dev_to_part(dev)->partno + offset);
143 if (part) {
144 res = part_devt(part);
145 put_device(part_to_dev(part));
148 no_offset:
149 put_device(dev);
150 done:
151 return res;
153 #endif
156 * Convert a name into device number. We accept the following variants:
158 * 1) device number in hexadecimal represents itself
159 * 2) /dev/nfs represents Root_NFS (0xff)
160 * 3) /dev/<disk_name> represents the device number of disk
161 * 4) /dev/<disk_name><decimal> represents the device number
162 * of partition - device number of disk plus the partition number
163 * 5) /dev/<disk_name>p<decimal> - same as the above, that form is
164 * used when disk name of partitioned disk ends on a digit.
165 * 6) PARTUUID=00112233-4455-6677-8899-AABBCCDDEEFF representing the
166 * unique id of a partition if the partition table provides it.
167 * 7) PARTUUID=<UUID>/PARTNROFF=<int> to select a partition in relation to
168 * a partition with a known unique id.
170 * If name doesn't have fall into the categories above, we return (0,0).
171 * block_class is used to check if something is a disk name. If the disk
172 * name contains slashes, the device name has them replaced with
173 * bangs.
176 dev_t name_to_dev_t(char *name)
178 char s[32];
179 char *p;
180 dev_t res = 0;
181 int part;
183 #ifdef CONFIG_BLOCK
184 if (strncmp(name, "PARTUUID=", 9) == 0) {
185 name += 9;
186 res = devt_from_partuuid(name);
187 if (!res)
188 goto fail;
189 goto done;
191 #endif
193 if (strncmp(name, "/dev/", 5) != 0) {
194 unsigned maj, min;
196 if (sscanf(name, "%u:%u", &maj, &min) == 2) {
197 res = MKDEV(maj, min);
198 if (maj != MAJOR(res) || min != MINOR(res))
199 goto fail;
200 } else {
201 res = new_decode_dev(simple_strtoul(name, &p, 16));
202 if (*p)
203 goto fail;
205 goto done;
208 name += 5;
209 res = Root_NFS;
210 if (strcmp(name, "nfs") == 0)
211 goto done;
212 res = Root_RAM0;
213 if (strcmp(name, "ram") == 0)
214 goto done;
216 if (strlen(name) > 31)
217 goto fail;
218 strcpy(s, name);
219 for (p = s; *p; p++)
220 if (*p == '/')
221 *p = '!';
222 res = blk_lookup_devt(s, 0);
223 if (res)
224 goto done;
227 * try non-existent, but valid partition, which may only exist
228 * after revalidating the disk, like partitioned md devices
230 while (p > s && isdigit(p[-1]))
231 p--;
232 if (p == s || !*p || *p == '0')
233 goto fail;
235 /* try disk name without <part number> */
236 part = simple_strtoul(p, NULL, 10);
237 *p = '\0';
238 res = blk_lookup_devt(s, part);
239 if (res)
240 goto done;
242 /* try disk name without p<part number> */
243 if (p < s + 2 || !isdigit(p[-2]) || p[-1] != 'p')
244 goto fail;
245 p[-1] = '\0';
246 res = blk_lookup_devt(s, part);
247 if (res)
248 goto done;
250 fail:
251 return 0;
252 done:
253 return res;
256 static int __init root_dev_setup(char *line)
258 strlcpy(saved_root_name, line, sizeof(saved_root_name));
259 return 1;
262 __setup("root=", root_dev_setup);
264 static int __init rootwait_setup(char *str)
266 if (*str)
267 return 0;
268 root_wait = 1;
269 return 1;
272 __setup("rootwait", rootwait_setup);
274 static char * __initdata root_mount_data;
275 static int __init root_data_setup(char *str)
277 root_mount_data = str;
278 return 1;
281 static char * __initdata root_fs_names;
282 static int __init fs_names_setup(char *str)
284 root_fs_names = str;
285 return 1;
288 static unsigned int __initdata root_delay;
289 static int __init root_delay_setup(char *str)
291 root_delay = simple_strtoul(str, NULL, 0);
292 return 1;
295 __setup("rootflags=", root_data_setup);
296 __setup("rootfstype=", fs_names_setup);
297 __setup("rootdelay=", root_delay_setup);
299 static void __init get_fs_names(char *page)
301 char *s = page;
303 if (root_fs_names) {
304 strcpy(page, root_fs_names);
305 while (*s++) {
306 if (s[-1] == ',')
307 s[-1] = '\0';
309 } else {
310 int len = get_filesystem_list(page);
311 char *p, *next;
313 page[len] = '\0';
314 for (p = page-1; p; p = next) {
315 next = strchr(++p, '\n');
316 if (*p++ != '\t')
317 continue;
318 while ((*s++ = *p++) != '\n')
320 s[-1] = '\0';
323 *s = '\0';
326 static int __init do_mount_root(char *name, char *fs, int flags, void *data)
328 struct super_block *s;
329 int err = sys_mount(name, "/root", fs, flags, data);
330 if (err)
331 return err;
333 sys_chdir((const char __user __force *)"/root");
334 s = current->fs->pwd.dentry->d_sb;
335 ROOT_DEV = s->s_dev;
336 printk(KERN_INFO
337 "VFS: Mounted root (%s filesystem)%s on device %u:%u.\n",
338 s->s_type->name,
339 s->s_flags & MS_RDONLY ? " readonly" : "",
340 MAJOR(ROOT_DEV), MINOR(ROOT_DEV));
341 return 0;
344 void __init mount_block_root(char *name, int flags)
346 char *fs_names = __getname_gfp(GFP_KERNEL
347 | __GFP_NOTRACK_FALSE_POSITIVE);
348 char *p;
349 #ifdef CONFIG_BLOCK
350 char b[BDEVNAME_SIZE];
351 #else
352 const char *b = name;
353 #endif
355 get_fs_names(fs_names);
356 retry:
357 for (p = fs_names; *p; p += strlen(p)+1) {
358 int err = do_mount_root(name, p, flags, root_mount_data);
359 switch (err) {
360 case 0:
361 goto out;
362 case -EACCES:
363 flags |= MS_RDONLY;
364 goto retry;
365 case -EINVAL:
366 continue;
369 * Allow the user to distinguish between failed sys_open
370 * and bad superblock on root device.
371 * and give them a list of the available devices
373 #ifdef CONFIG_BLOCK
374 __bdevname(ROOT_DEV, b);
375 #endif
376 printk("VFS: Cannot open root device \"%s\" or %s\n",
377 root_device_name, b);
378 printk("Please append a correct \"root=\" boot option; here are the available partitions:\n");
380 printk_all_partitions();
381 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
382 printk("DEBUG_BLOCK_EXT_DEVT is enabled, you need to specify "
383 "explicit textual name for \"root=\" boot option.\n");
384 #endif
385 panic("VFS: Unable to mount root fs on %s", b);
388 printk("List of all partitions:\n");
389 printk_all_partitions();
390 printk("No filesystem could mount root, tried: ");
391 for (p = fs_names; *p; p += strlen(p)+1)
392 printk(" %s", p);
393 printk("\n");
394 #ifdef CONFIG_BLOCK
395 __bdevname(ROOT_DEV, b);
396 #endif
397 panic("VFS: Unable to mount root fs on %s", b);
398 out:
399 putname(fs_names);
402 #ifdef CONFIG_ROOT_NFS
404 #define NFSROOT_TIMEOUT_MIN 5
405 #define NFSROOT_TIMEOUT_MAX 30
406 #define NFSROOT_RETRY_MAX 5
408 static int __init mount_nfs_root(void)
410 char *root_dev, *root_data;
411 unsigned int timeout;
412 int try, err;
414 err = nfs_root_data(&root_dev, &root_data);
415 if (err != 0)
416 return 0;
419 * The server or network may not be ready, so try several
420 * times. Stop after a few tries in case the client wants
421 * to fall back to other boot methods.
423 timeout = NFSROOT_TIMEOUT_MIN;
424 for (try = 1; ; try++) {
425 err = do_mount_root(root_dev, "nfs",
426 root_mountflags, root_data);
427 if (err == 0)
428 return 1;
429 if (try > NFSROOT_RETRY_MAX)
430 break;
432 /* Wait, in case the server refused us immediately */
433 ssleep(timeout);
434 timeout <<= 1;
435 if (timeout > NFSROOT_TIMEOUT_MAX)
436 timeout = NFSROOT_TIMEOUT_MAX;
438 return 0;
440 #endif
442 #if defined(CONFIG_BLK_DEV_RAM) || defined(CONFIG_BLK_DEV_FD)
443 void __init change_floppy(char *fmt, ...)
445 struct termios termios;
446 char buf[80];
447 char c;
448 int fd;
449 va_list args;
450 va_start(args, fmt);
451 vsprintf(buf, fmt, args);
452 va_end(args);
453 fd = sys_open("/dev/root", O_RDWR | O_NDELAY, 0);
454 if (fd >= 0) {
455 sys_ioctl(fd, FDEJECT, 0);
456 sys_close(fd);
458 printk(KERN_NOTICE "VFS: Insert %s and press ENTER\n", buf);
459 fd = sys_open("/dev/console", O_RDWR, 0);
460 if (fd >= 0) {
461 sys_ioctl(fd, TCGETS, (long)&termios);
462 termios.c_lflag &= ~ICANON;
463 sys_ioctl(fd, TCSETSF, (long)&termios);
464 sys_read(fd, &c, 1);
465 termios.c_lflag |= ICANON;
466 sys_ioctl(fd, TCSETSF, (long)&termios);
467 sys_close(fd);
470 #endif
472 void __init mount_root(void)
474 #ifdef CONFIG_ROOT_NFS
475 if (MAJOR(ROOT_DEV) == UNNAMED_MAJOR) {
476 if (mount_nfs_root())
477 return;
479 printk(KERN_ERR "VFS: Unable to mount root fs via NFS, trying floppy.\n");
480 ROOT_DEV = Root_FD0;
482 #endif
483 #ifdef CONFIG_BLK_DEV_FD
484 if (MAJOR(ROOT_DEV) == FLOPPY_MAJOR) {
485 /* rd_doload is 2 for a dual initrd/ramload setup */
486 if (rd_doload==2) {
487 if (rd_load_disk(1)) {
488 ROOT_DEV = Root_RAM1;
489 root_device_name = NULL;
491 } else
492 change_floppy("root floppy");
494 #endif
495 #ifdef CONFIG_BLOCK
496 create_dev("/dev/root", ROOT_DEV);
497 mount_block_root("/dev/root", root_mountflags);
498 #endif
502 * Prepare the namespace - decide what/where to mount, load ramdisks, etc.
504 void __init prepare_namespace(void)
506 int is_floppy;
508 if (root_delay) {
509 printk(KERN_INFO "Waiting %dsec before mounting root device...\n",
510 root_delay);
511 ssleep(root_delay);
515 * wait for the known devices to complete their probing
517 * Note: this is a potential source of long boot delays.
518 * For example, it is not atypical to wait 5 seconds here
519 * for the touchpad of a laptop to initialize.
521 wait_for_device_probe();
523 md_run_setup();
525 if (saved_root_name[0]) {
526 root_device_name = saved_root_name;
527 if (!strncmp(root_device_name, "mtd", 3) ||
528 !strncmp(root_device_name, "ubi", 3)) {
529 mount_block_root(root_device_name, root_mountflags);
530 goto out;
532 ROOT_DEV = name_to_dev_t(root_device_name);
533 if (strncmp(root_device_name, "/dev/", 5) == 0)
534 root_device_name += 5;
537 if (initrd_load())
538 goto out;
540 /* wait for any asynchronous scanning to complete */
541 if ((ROOT_DEV == 0) && root_wait) {
542 printk(KERN_INFO "Waiting for root device %s...\n",
543 saved_root_name);
544 while (driver_probe_done() != 0 ||
545 (ROOT_DEV = name_to_dev_t(saved_root_name)) == 0)
546 msleep(100);
547 async_synchronize_full();
550 is_floppy = MAJOR(ROOT_DEV) == FLOPPY_MAJOR;
552 if (is_floppy && rd_doload && rd_load_disk(0))
553 ROOT_DEV = Root_RAM0;
555 mount_root();
556 out:
557 devtmpfs_mount("dev");
558 sys_mount(".", "/", NULL, MS_MOVE, NULL);
559 sys_chroot((const char __user __force *)".");