Commit | Line | Data |
---|---|---|
6c29256c | 1 | /* |
1da177e4 LT |
2 | * Copyright (C) 2000 Jeff Dike (jdike@karaya.com) |
3 | * Licensed under the GPL | |
4 | */ | |
5 | ||
6 | /* 2001-09-28...2002-04-17 | |
7 | * Partition stuff by James_McMechan@hotmail.com | |
8 | * old style ubd by setting UBD_SHIFT to 0 | |
9 | * 2002-09-27...2002-10-18 massive tinkering for 2.5 | |
10 | * partitions have changed in 2.5 | |
11 | * 2003-01-29 more tinkering for 2.5.59-1 | |
12 | * This should now address the sysfs problems and has | |
13 | * the symlink for devfs to allow for booting with | |
14 | * the common /dev/ubd/discX/... names rather than | |
15 | * only /dev/ubdN/discN this version also has lots of | |
16 | * clean ups preparing for ubd-many. | |
17 | * James McMechan | |
18 | */ | |
19 | ||
20 | #define MAJOR_NR UBD_MAJOR | |
21 | #define UBD_SHIFT 4 | |
22 | ||
e16f5350 | 23 | #include "linux/kernel.h" |
1da177e4 LT |
24 | #include "linux/module.h" |
25 | #include "linux/blkdev.h" | |
26 | #include "linux/hdreg.h" | |
27 | #include "linux/init.h" | |
1da177e4 LT |
28 | #include "linux/cdrom.h" |
29 | #include "linux/proc_fs.h" | |
30 | #include "linux/ctype.h" | |
31 | #include "linux/capability.h" | |
32 | #include "linux/mm.h" | |
33 | #include "linux/vmalloc.h" | |
34 | #include "linux/blkpg.h" | |
35 | #include "linux/genhd.h" | |
36 | #include "linux/spinlock.h" | |
d052d1be | 37 | #include "linux/platform_device.h" |
23464ffa | 38 | #include "linux/scatterlist.h" |
1da177e4 LT |
39 | #include "asm/segment.h" |
40 | #include "asm/uaccess.h" | |
41 | #include "asm/irq.h" | |
42 | #include "asm/types.h" | |
43 | #include "asm/tlbflush.h" | |
1da177e4 LT |
44 | #include "mem_user.h" |
45 | #include "kern_util.h" | |
46 | #include "kern.h" | |
47 | #include "mconsole_kern.h" | |
48 | #include "init.h" | |
49 | #include "irq_user.h" | |
50 | #include "irq_kern.h" | |
51 | #include "ubd_user.h" | |
1da177e4 LT |
52 | #include "os.h" |
53 | #include "mem.h" | |
54 | #include "mem_kern.h" | |
55 | #include "cow.h" | |
56 | ||
7b9014c1 | 57 | enum ubd_req { UBD_READ, UBD_WRITE }; |
1da177e4 LT |
58 | |
59 | struct io_thread_req { | |
62f96cb0 | 60 | struct request *req; |
91acb21f | 61 | enum ubd_req op; |
1da177e4 LT |
62 | int fds[2]; |
63 | unsigned long offsets[2]; | |
64 | unsigned long long offset; | |
65 | unsigned long length; | |
66 | char *buffer; | |
67 | int sectorsize; | |
91acb21f JD |
68 | unsigned long sector_mask; |
69 | unsigned long long cow_offset; | |
70 | unsigned long bitmap_words[2]; | |
1da177e4 LT |
71 | int error; |
72 | }; | |
73 | ||
91acb21f | 74 | static inline int ubd_test_bit(__u64 bit, unsigned char *data) |
1da177e4 LT |
75 | { |
76 | __u64 n; | |
77 | int bits, off; | |
78 | ||
91acb21f | 79 | bits = sizeof(data[0]) * 8; |
1da177e4 LT |
80 | n = bit / bits; |
81 | off = bit % bits; | |
dc764e50 | 82 | return (data[n] & (1 << off)) != 0; |
1da177e4 LT |
83 | } |
84 | ||
91acb21f | 85 | static inline void ubd_set_bit(__u64 bit, unsigned char *data) |
1da177e4 LT |
86 | { |
87 | __u64 n; | |
88 | int bits, off; | |
89 | ||
91acb21f | 90 | bits = sizeof(data[0]) * 8; |
1da177e4 LT |
91 | n = bit / bits; |
92 | off = bit % bits; | |
91acb21f | 93 | data[n] |= (1 << off); |
1da177e4 LT |
94 | } |
95 | /*End stuff from ubd_user.h*/ | |
96 | ||
97 | #define DRIVER_NAME "uml-blkdev" | |
98 | ||
d7fb2c38 | 99 | static DEFINE_MUTEX(ubd_lock); |
1da177e4 | 100 | |
a625c998 AV |
101 | static int ubd_open(struct block_device *bdev, fmode_t mode); |
102 | static int ubd_release(struct gendisk *disk, fmode_t mode); | |
103 | static int ubd_ioctl(struct block_device *bdev, fmode_t mode, | |
1da177e4 | 104 | unsigned int cmd, unsigned long arg); |
a885c8c4 | 105 | static int ubd_getgeo(struct block_device *bdev, struct hd_geometry *geo); |
1da177e4 | 106 | |
97d88ac8 | 107 | #define MAX_DEV (16) |
1da177e4 | 108 | |
1da177e4 LT |
109 | static struct block_device_operations ubd_blops = { |
110 | .owner = THIS_MODULE, | |
a625c998 AV |
111 | .open = ubd_open, |
112 | .release = ubd_release, | |
113 | .ioctl = ubd_ioctl, | |
a885c8c4 | 114 | .getgeo = ubd_getgeo, |
1da177e4 LT |
115 | }; |
116 | ||
1da177e4 LT |
117 | /* Protected by ubd_lock */ |
118 | static int fake_major = MAJOR_NR; | |
1da177e4 LT |
119 | static struct gendisk *ubd_gendisk[MAX_DEV]; |
120 | static struct gendisk *fake_gendisk[MAX_DEV]; | |
6c29256c | 121 | |
1da177e4 LT |
122 | #ifdef CONFIG_BLK_DEV_UBD_SYNC |
123 | #define OPEN_FLAGS ((struct openflags) { .r = 1, .w = 1, .s = 1, .c = 0, \ | |
124 | .cl = 1 }) | |
125 | #else | |
126 | #define OPEN_FLAGS ((struct openflags) { .r = 1, .w = 1, .s = 0, .c = 0, \ | |
127 | .cl = 1 }) | |
128 | #endif | |
1da177e4 LT |
129 | static struct openflags global_openflags = OPEN_FLAGS; |
130 | ||
131 | struct cow { | |
2a9d32f6 | 132 | /* backing file name */ |
1da177e4 | 133 | char *file; |
2a9d32f6 | 134 | /* backing file fd */ |
1da177e4 LT |
135 | int fd; |
136 | unsigned long *bitmap; | |
137 | unsigned long bitmap_len; | |
138 | int bitmap_offset; | |
dc764e50 | 139 | int data_offset; |
1da177e4 LT |
140 | }; |
141 | ||
a0044bdf JD |
142 | #define MAX_SG 64 |
143 | ||
1da177e4 | 144 | struct ubd { |
a0044bdf | 145 | struct list_head restart; |
2a9d32f6 PBG |
146 | /* name (and fd, below) of the file opened for writing, either the |
147 | * backing or the cow file. */ | |
1da177e4 LT |
148 | char *file; |
149 | int count; | |
150 | int fd; | |
151 | __u64 size; | |
152 | struct openflags boot_openflags; | |
153 | struct openflags openflags; | |
84e945e3 PBG |
154 | unsigned shared:1; |
155 | unsigned no_cow:1; | |
1da177e4 LT |
156 | struct cow cow; |
157 | struct platform_device pdev; | |
62f96cb0 JD |
158 | struct request_queue *queue; |
159 | spinlock_t lock; | |
a0044bdf JD |
160 | struct scatterlist sg[MAX_SG]; |
161 | struct request *request; | |
162 | int start_sg, end_sg; | |
1da177e4 LT |
163 | }; |
164 | ||
165 | #define DEFAULT_COW { \ | |
166 | .file = NULL, \ | |
dc764e50 JD |
167 | .fd = -1, \ |
168 | .bitmap = NULL, \ | |
1da177e4 | 169 | .bitmap_offset = 0, \ |
dc764e50 | 170 | .data_offset = 0, \ |
1da177e4 LT |
171 | } |
172 | ||
173 | #define DEFAULT_UBD { \ | |
174 | .file = NULL, \ | |
175 | .count = 0, \ | |
176 | .fd = -1, \ | |
177 | .size = -1, \ | |
178 | .boot_openflags = OPEN_FLAGS, \ | |
179 | .openflags = OPEN_FLAGS, \ | |
dc764e50 | 180 | .no_cow = 0, \ |
6c29256c | 181 | .shared = 0, \ |
dc764e50 | 182 | .cow = DEFAULT_COW, \ |
62f96cb0 | 183 | .lock = SPIN_LOCK_UNLOCKED, \ |
a0044bdf JD |
184 | .request = NULL, \ |
185 | .start_sg = 0, \ | |
186 | .end_sg = 0, \ | |
1da177e4 LT |
187 | } |
188 | ||
b8831a1d | 189 | /* Protected by ubd_lock */ |
5dc62b1b | 190 | static struct ubd ubd_devs[MAX_DEV] = { [0 ... MAX_DEV - 1] = DEFAULT_UBD }; |
1da177e4 | 191 | |
1da177e4 LT |
192 | /* Only changed by fake_ide_setup which is a setup */ |
193 | static int fake_ide = 0; | |
194 | static struct proc_dir_entry *proc_ide_root = NULL; | |
195 | static struct proc_dir_entry *proc_ide = NULL; | |
196 | ||
197 | static void make_proc_ide(void) | |
198 | { | |
199 | proc_ide_root = proc_mkdir("ide", NULL); | |
200 | proc_ide = proc_mkdir("ide0", proc_ide_root); | |
201 | } | |
202 | ||
203 | static int proc_ide_read_media(char *page, char **start, off_t off, int count, | |
204 | int *eof, void *data) | |
205 | { | |
206 | int len; | |
207 | ||
208 | strcpy(page, "disk\n"); | |
209 | len = strlen("disk\n"); | |
210 | len -= off; | |
211 | if (len < count){ | |
212 | *eof = 1; | |
213 | if (len <= 0) return 0; | |
214 | } | |
215 | else len = count; | |
216 | *start = page + off; | |
217 | return len; | |
218 | } | |
219 | ||
c0a9290e | 220 | static void make_ide_entries(const char *dev_name) |
1da177e4 LT |
221 | { |
222 | struct proc_dir_entry *dir, *ent; | |
223 | char name[64]; | |
224 | ||
225 | if(proc_ide_root == NULL) make_proc_ide(); | |
226 | ||
227 | dir = proc_mkdir(dev_name, proc_ide); | |
228 | if(!dir) return; | |
229 | ||
230 | ent = create_proc_entry("media", S_IFREG|S_IRUGO, dir); | |
231 | if(!ent) return; | |
1da177e4 LT |
232 | ent->data = NULL; |
233 | ent->read_proc = proc_ide_read_media; | |
234 | ent->write_proc = NULL; | |
c0a9290e | 235 | snprintf(name, sizeof(name), "ide0/%s", dev_name); |
1da177e4 LT |
236 | proc_symlink(dev_name, proc_ide_root, name); |
237 | } | |
238 | ||
239 | static int fake_ide_setup(char *str) | |
240 | { | |
241 | fake_ide = 1; | |
dc764e50 | 242 | return 1; |
1da177e4 LT |
243 | } |
244 | ||
245 | __setup("fake_ide", fake_ide_setup); | |
246 | ||
247 | __uml_help(fake_ide_setup, | |
248 | "fake_ide\n" | |
249 | " Create ide0 entries that map onto ubd devices.\n\n" | |
250 | ); | |
251 | ||
252 | static int parse_unit(char **ptr) | |
253 | { | |
254 | char *str = *ptr, *end; | |
255 | int n = -1; | |
256 | ||
257 | if(isdigit(*str)) { | |
258 | n = simple_strtoul(str, &end, 0); | |
259 | if(end == str) | |
dc764e50 | 260 | return -1; |
1da177e4 LT |
261 | *ptr = end; |
262 | } | |
97d88ac8 | 263 | else if (('a' <= *str) && (*str <= 'z')) { |
1da177e4 LT |
264 | n = *str - 'a'; |
265 | str++; | |
266 | *ptr = str; | |
267 | } | |
dc764e50 | 268 | return n; |
1da177e4 LT |
269 | } |
270 | ||
d8d7c28e PBG |
271 | /* If *index_out == -1 at exit, the passed option was a general one; |
272 | * otherwise, the str pointer is used (and owned) inside ubd_devs array, so it | |
273 | * should not be freed on exit. | |
274 | */ | |
f28169d2 | 275 | static int ubd_setup_common(char *str, int *index_out, char **error_out) |
1da177e4 | 276 | { |
7d314e34 | 277 | struct ubd *ubd_dev; |
1da177e4 LT |
278 | struct openflags flags = global_openflags; |
279 | char *backing_file; | |
b8831a1d | 280 | int n, err = 0, i; |
1da177e4 LT |
281 | |
282 | if(index_out) *index_out = -1; | |
283 | n = *str; | |
284 | if(n == '='){ | |
285 | char *end; | |
286 | int major; | |
287 | ||
288 | str++; | |
1da177e4 LT |
289 | if(!strcmp(str, "sync")){ |
290 | global_openflags = of_sync(global_openflags); | |
b8831a1d | 291 | goto out1; |
1da177e4 | 292 | } |
b8831a1d JD |
293 | |
294 | err = -EINVAL; | |
1da177e4 LT |
295 | major = simple_strtoul(str, &end, 0); |
296 | if((*end != '\0') || (end == str)){ | |
f28169d2 | 297 | *error_out = "Didn't parse major number"; |
b8831a1d | 298 | goto out1; |
1da177e4 LT |
299 | } |
300 | ||
f28169d2 JD |
301 | mutex_lock(&ubd_lock); |
302 | if(fake_major != MAJOR_NR){ | |
303 | *error_out = "Can't assign a fake major twice"; | |
304 | goto out1; | |
305 | } | |
6c29256c | 306 | |
f28169d2 | 307 | fake_major = major; |
1da177e4 LT |
308 | |
309 | printk(KERN_INFO "Setting extra ubd major number to %d\n", | |
310 | major); | |
f28169d2 JD |
311 | err = 0; |
312 | out1: | |
313 | mutex_unlock(&ubd_lock); | |
314 | return err; | |
1da177e4 LT |
315 | } |
316 | ||
317 | n = parse_unit(&str); | |
318 | if(n < 0){ | |
f28169d2 JD |
319 | *error_out = "Couldn't parse device number"; |
320 | return -EINVAL; | |
1da177e4 LT |
321 | } |
322 | if(n >= MAX_DEV){ | |
f28169d2 JD |
323 | *error_out = "Device number out of range"; |
324 | return 1; | |
1da177e4 LT |
325 | } |
326 | ||
f28169d2 | 327 | err = -EBUSY; |
d7fb2c38 | 328 | mutex_lock(&ubd_lock); |
1da177e4 | 329 | |
7d314e34 PBG |
330 | ubd_dev = &ubd_devs[n]; |
331 | if(ubd_dev->file != NULL){ | |
f28169d2 | 332 | *error_out = "Device is already configured"; |
1da177e4 LT |
333 | goto out; |
334 | } | |
335 | ||
336 | if (index_out) | |
337 | *index_out = n; | |
338 | ||
f28169d2 | 339 | err = -EINVAL; |
6c29256c | 340 | for (i = 0; i < sizeof("rscd="); i++) { |
1da177e4 LT |
341 | switch (*str) { |
342 | case 'r': | |
343 | flags.w = 0; | |
344 | break; | |
345 | case 's': | |
346 | flags.s = 1; | |
347 | break; | |
348 | case 'd': | |
7d314e34 | 349 | ubd_dev->no_cow = 1; |
1da177e4 | 350 | break; |
6c29256c | 351 | case 'c': |
7d314e34 | 352 | ubd_dev->shared = 1; |
6c29256c | 353 | break; |
1da177e4 LT |
354 | case '=': |
355 | str++; | |
356 | goto break_loop; | |
357 | default: | |
f28169d2 JD |
358 | *error_out = "Expected '=' or flag letter " |
359 | "(r, s, c, or d)"; | |
1da177e4 LT |
360 | goto out; |
361 | } | |
362 | str++; | |
363 | } | |
364 | ||
f28169d2 JD |
365 | if (*str == '=') |
366 | *error_out = "Too many flags specified"; | |
367 | else | |
368 | *error_out = "Missing '='"; | |
1da177e4 LT |
369 | goto out; |
370 | ||
371 | break_loop: | |
1da177e4 LT |
372 | backing_file = strchr(str, ','); |
373 | ||
f28169d2 | 374 | if (backing_file == NULL) |
1da177e4 | 375 | backing_file = strchr(str, ':'); |
1da177e4 | 376 | |
f28169d2 JD |
377 | if(backing_file != NULL){ |
378 | if(ubd_dev->no_cow){ | |
379 | *error_out = "Can't specify both 'd' and a cow file"; | |
380 | goto out; | |
381 | } | |
1da177e4 LT |
382 | else { |
383 | *backing_file = '\0'; | |
384 | backing_file++; | |
385 | } | |
386 | } | |
f28169d2 | 387 | err = 0; |
7d314e34 PBG |
388 | ubd_dev->file = str; |
389 | ubd_dev->cow.file = backing_file; | |
390 | ubd_dev->boot_openflags = flags; | |
1da177e4 | 391 | out: |
d7fb2c38 | 392 | mutex_unlock(&ubd_lock); |
f28169d2 | 393 | return err; |
1da177e4 LT |
394 | } |
395 | ||
396 | static int ubd_setup(char *str) | |
397 | { | |
f28169d2 JD |
398 | char *error; |
399 | int err; | |
400 | ||
401 | err = ubd_setup_common(str, NULL, &error); | |
402 | if(err) | |
403 | printk(KERN_ERR "Failed to initialize device with \"%s\" : " | |
404 | "%s\n", str, error); | |
405 | return 1; | |
1da177e4 LT |
406 | } |
407 | ||
408 | __setup("ubd", ubd_setup); | |
409 | __uml_help(ubd_setup, | |
410 | "ubd<n><flags>=<filename>[(:|,)<filename2>]\n" | |
411 | " This is used to associate a device with a file in the underlying\n" | |
412 | " filesystem. When specifying two filenames, the first one is the\n" | |
413 | " COW name and the second is the backing file name. As separator you can\n" | |
414 | " use either a ':' or a ',': the first one allows writing things like;\n" | |
415 | " ubd0=~/Uml/root_cow:~/Uml/root_backing_file\n" | |
416 | " while with a ',' the shell would not expand the 2nd '~'.\n" | |
f28169d2 | 417 | " When using only one filename, UML will detect whether to treat it like\n" |
1da177e4 LT |
418 | " a COW file or a backing file. To override this detection, add the 'd'\n" |
419 | " flag:\n" | |
420 | " ubd0d=BackingFile\n" | |
421 | " Usually, there is a filesystem in the file, but \n" | |
422 | " that's not required. Swap devices containing swap files can be\n" | |
423 | " specified like this. Also, a file which doesn't contain a\n" | |
424 | " filesystem can have its contents read in the virtual \n" | |
425 | " machine by running 'dd' on the device. <n> must be in the range\n" | |
426 | " 0 to 7. Appending an 'r' to the number will cause that device\n" | |
427 | " to be mounted read-only. For example ubd1r=./ext_fs. Appending\n" | |
20ede453 JD |
428 | " an 's' will cause data to be written to disk on the host immediately.\n" |
429 | " 'c' will cause the device to be treated as being shared between multiple\n" | |
430 | " UMLs and file locking will be turned off - this is appropriate for a\n" | |
431 | " cluster filesystem and inappropriate at almost all other times.\n\n" | |
1da177e4 LT |
432 | ); |
433 | ||
8299ca5c | 434 | static int udb_setup(char *str) |
1da177e4 LT |
435 | { |
436 | printk("udb%s specified on command line is almost certainly a ubd -> " | |
437 | "udb TYPO\n", str); | |
dc764e50 | 438 | return 1; |
1da177e4 LT |
439 | } |
440 | ||
441 | __setup("udb", udb_setup); | |
442 | __uml_help(udb_setup, | |
443 | "udb\n" | |
0894e27e JD |
444 | " This option is here solely to catch ubd -> udb typos, which can be\n" |
445 | " to impossible to catch visually unless you specifically look for\n" | |
446 | " them. The only result of any option starting with 'udb' is an error\n" | |
1da177e4 LT |
447 | " in the boot output.\n\n" |
448 | ); | |
449 | ||
165125e1 | 450 | static void do_ubd_request(struct request_queue * q); |
91acb21f JD |
451 | |
452 | /* Only changed by ubd_init, which is an initcall. */ | |
5dc62b1b | 453 | static int thread_fd = -1; |
1da177e4 | 454 | |
4898b53a | 455 | static void ubd_end_request(struct request *req, int bytes, int error) |
1da177e4 | 456 | { |
4898b53a | 457 | blk_end_request(req, error, bytes); |
1da177e4 LT |
458 | } |
459 | ||
33f775ee PBG |
460 | /* Callable only from interrupt context - otherwise you need to do |
461 | * spin_lock_irq()/spin_lock_irqsave() */ | |
a0044bdf | 462 | static inline void ubd_finish(struct request *req, int bytes) |
1da177e4 | 463 | { |
a0044bdf | 464 | if(bytes < 0){ |
4898b53a | 465 | ubd_end_request(req, 0, -EIO); |
a0044bdf JD |
466 | return; |
467 | } | |
4898b53a | 468 | ubd_end_request(req, bytes, 0); |
1da177e4 LT |
469 | } |
470 | ||
a0044bdf JD |
471 | static LIST_HEAD(restart); |
472 | ||
2fe30a34 | 473 | /* XXX - move this inside ubd_intr. */ |
62f96cb0 | 474 | /* Called without dev->lock held, and only in interrupt context. */ |
91acb21f | 475 | static void ubd_handler(void) |
1da177e4 | 476 | { |
2adcec21 | 477 | struct io_thread_req *req; |
62f96cb0 | 478 | struct request *rq; |
a0044bdf JD |
479 | struct ubd *ubd; |
480 | struct list_head *list, *next_ele; | |
481 | unsigned long flags; | |
91acb21f JD |
482 | int n; |
483 | ||
a0044bdf | 484 | while(1){ |
a6ea4cce JD |
485 | n = os_read_file(thread_fd, &req, |
486 | sizeof(struct io_thread_req *)); | |
a0044bdf JD |
487 | if(n != sizeof(req)){ |
488 | if(n == -EAGAIN) | |
489 | break; | |
490 | printk(KERN_ERR "spurious interrupt in ubd_handler, " | |
491 | "err = %d\n", -n); | |
492 | return; | |
493 | } | |
62f96cb0 | 494 | |
2adcec21 JD |
495 | rq = req->req; |
496 | rq->nr_sectors -= req->length >> 9; | |
a0044bdf JD |
497 | if(rq->nr_sectors == 0) |
498 | ubd_finish(rq, rq->hard_nr_sectors << 9); | |
2adcec21 | 499 | kfree(req); |
a0044bdf | 500 | } |
62f96cb0 | 501 | reactivate_fd(thread_fd, UBD_IRQ); |
a0044bdf JD |
502 | |
503 | list_for_each_safe(list, next_ele, &restart){ | |
504 | ubd = container_of(list, struct ubd, restart); | |
505 | list_del_init(&ubd->restart); | |
506 | spin_lock_irqsave(&ubd->lock, flags); | |
507 | do_ubd_request(ubd->queue); | |
508 | spin_unlock_irqrestore(&ubd->lock, flags); | |
509 | } | |
1da177e4 LT |
510 | } |
511 | ||
7bea96fd | 512 | static irqreturn_t ubd_intr(int irq, void *dev) |
1da177e4 | 513 | { |
91acb21f | 514 | ubd_handler(); |
dc764e50 | 515 | return IRQ_HANDLED; |
91acb21f | 516 | } |
09ace81c | 517 | |
91acb21f JD |
518 | /* Only changed by ubd_init, which is an initcall. */ |
519 | static int io_pid = -1; | |
09ace81c | 520 | |
5dc62b1b | 521 | static void kill_io_thread(void) |
91acb21f | 522 | { |
6c29256c | 523 | if(io_pid != -1) |
91acb21f | 524 | os_kill_process(io_pid, 1); |
09ace81c | 525 | } |
1da177e4 | 526 | |
91acb21f JD |
527 | __uml_exitcall(kill_io_thread); |
528 | ||
d8d7c28e | 529 | static inline int ubd_file_size(struct ubd *ubd_dev, __u64 *size_out) |
1da177e4 LT |
530 | { |
531 | char *file; | |
532 | ||
7d314e34 | 533 | file = ubd_dev->cow.file ? ubd_dev->cow.file : ubd_dev->file; |
dc764e50 | 534 | return os_file_size(file, size_out); |
1da177e4 LT |
535 | } |
536 | ||
5dc62b1b WC |
537 | static int read_cow_bitmap(int fd, void *buf, int offset, int len) |
538 | { | |
539 | int err; | |
540 | ||
541 | err = os_seek_file(fd, offset); | |
542 | if (err < 0) | |
543 | return err; | |
544 | ||
545 | err = os_read_file(fd, buf, len); | |
546 | if (err < 0) | |
547 | return err; | |
548 | ||
549 | return 0; | |
550 | } | |
551 | ||
552 | static int backing_file_mismatch(char *file, __u64 size, time_t mtime) | |
553 | { | |
554 | unsigned long modtime; | |
555 | unsigned long long actual; | |
556 | int err; | |
557 | ||
558 | err = os_file_modtime(file, &modtime); | |
559 | if (err < 0) { | |
560 | printk(KERN_ERR "Failed to get modification time of backing " | |
561 | "file \"%s\", err = %d\n", file, -err); | |
562 | return err; | |
563 | } | |
564 | ||
565 | err = os_file_size(file, &actual); | |
566 | if (err < 0) { | |
567 | printk(KERN_ERR "Failed to get size of backing file \"%s\", " | |
568 | "err = %d\n", file, -err); | |
569 | return err; | |
570 | } | |
571 | ||
572 | if (actual != size) { | |
573 | /*__u64 can be a long on AMD64 and with %lu GCC complains; so | |
574 | * the typecast.*/ | |
575 | printk(KERN_ERR "Size mismatch (%llu vs %llu) of COW header " | |
576 | "vs backing file\n", (unsigned long long) size, actual); | |
577 | return -EINVAL; | |
578 | } | |
579 | if (modtime != mtime) { | |
580 | printk(KERN_ERR "mtime mismatch (%ld vs %ld) of COW header vs " | |
581 | "backing file\n", mtime, modtime); | |
582 | return -EINVAL; | |
583 | } | |
584 | return 0; | |
585 | } | |
586 | ||
587 | static int path_requires_switch(char *from_cmdline, char *from_cow, char *cow) | |
588 | { | |
589 | struct uml_stat buf1, buf2; | |
590 | int err; | |
591 | ||
592 | if (from_cmdline == NULL) | |
593 | return 0; | |
594 | if (!strcmp(from_cmdline, from_cow)) | |
595 | return 0; | |
596 | ||
597 | err = os_stat_file(from_cmdline, &buf1); | |
598 | if (err < 0) { | |
599 | printk(KERN_ERR "Couldn't stat '%s', err = %d\n", from_cmdline, | |
600 | -err); | |
601 | return 0; | |
602 | } | |
603 | err = os_stat_file(from_cow, &buf2); | |
604 | if (err < 0) { | |
605 | printk(KERN_ERR "Couldn't stat '%s', err = %d\n", from_cow, | |
606 | -err); | |
607 | return 1; | |
608 | } | |
609 | if ((buf1.ust_dev == buf2.ust_dev) && (buf1.ust_ino == buf2.ust_ino)) | |
610 | return 0; | |
611 | ||
612 | printk(KERN_ERR "Backing file mismatch - \"%s\" requested, " | |
613 | "\"%s\" specified in COW header of \"%s\"\n", | |
614 | from_cmdline, from_cow, cow); | |
615 | return 1; | |
616 | } | |
617 | ||
618 | static int open_ubd_file(char *file, struct openflags *openflags, int shared, | |
619 | char **backing_file_out, int *bitmap_offset_out, | |
620 | unsigned long *bitmap_len_out, int *data_offset_out, | |
621 | int *create_cow_out) | |
622 | { | |
623 | time_t mtime; | |
624 | unsigned long long size; | |
625 | __u32 version, align; | |
626 | char *backing_file; | |
627 | int fd, err, sectorsize, asked_switch, mode = 0644; | |
628 | ||
629 | fd = os_open_file(file, *openflags, mode); | |
630 | if (fd < 0) { | |
631 | if ((fd == -ENOENT) && (create_cow_out != NULL)) | |
632 | *create_cow_out = 1; | |
633 | if (!openflags->w || | |
634 | ((fd != -EROFS) && (fd != -EACCES))) | |
635 | return fd; | |
636 | openflags->w = 0; | |
637 | fd = os_open_file(file, *openflags, mode); | |
638 | if (fd < 0) | |
639 | return fd; | |
640 | } | |
641 | ||
642 | if (shared) | |
643 | printk(KERN_INFO "Not locking \"%s\" on the host\n", file); | |
644 | else { | |
645 | err = os_lock_file(fd, openflags->w); | |
646 | if (err < 0) { | |
647 | printk(KERN_ERR "Failed to lock '%s', err = %d\n", | |
648 | file, -err); | |
649 | goto out_close; | |
650 | } | |
651 | } | |
652 | ||
653 | /* Successful return case! */ | |
654 | if (backing_file_out == NULL) | |
655 | return fd; | |
656 | ||
657 | err = read_cow_header(file_reader, &fd, &version, &backing_file, &mtime, | |
658 | &size, §orsize, &align, bitmap_offset_out); | |
659 | if (err && (*backing_file_out != NULL)) { | |
660 | printk(KERN_ERR "Failed to read COW header from COW file " | |
661 | "\"%s\", errno = %d\n", file, -err); | |
662 | goto out_close; | |
663 | } | |
664 | if (err) | |
665 | return fd; | |
666 | ||
667 | asked_switch = path_requires_switch(*backing_file_out, backing_file, | |
668 | file); | |
669 | ||
670 | /* Allow switching only if no mismatch. */ | |
671 | if (asked_switch && !backing_file_mismatch(*backing_file_out, size, | |
672 | mtime)) { | |
673 | printk(KERN_ERR "Switching backing file to '%s'\n", | |
674 | *backing_file_out); | |
675 | err = write_cow_header(file, fd, *backing_file_out, | |
676 | sectorsize, align, &size); | |
677 | if (err) { | |
678 | printk(KERN_ERR "Switch failed, errno = %d\n", -err); | |
679 | goto out_close; | |
680 | } | |
681 | } else { | |
682 | *backing_file_out = backing_file; | |
683 | err = backing_file_mismatch(*backing_file_out, size, mtime); | |
684 | if (err) | |
685 | goto out_close; | |
686 | } | |
687 | ||
688 | cow_sizes(version, size, sectorsize, align, *bitmap_offset_out, | |
689 | bitmap_len_out, data_offset_out); | |
690 | ||
691 | return fd; | |
692 | out_close: | |
693 | os_close_file(fd); | |
694 | return err; | |
695 | } | |
696 | ||
697 | static int create_cow_file(char *cow_file, char *backing_file, | |
698 | struct openflags flags, | |
699 | int sectorsize, int alignment, int *bitmap_offset_out, | |
700 | unsigned long *bitmap_len_out, int *data_offset_out) | |
701 | { | |
702 | int err, fd; | |
703 | ||
704 | flags.c = 1; | |
705 | fd = open_ubd_file(cow_file, &flags, 0, NULL, NULL, NULL, NULL, NULL); | |
706 | if (fd < 0) { | |
707 | err = fd; | |
708 | printk(KERN_ERR "Open of COW file '%s' failed, errno = %d\n", | |
709 | cow_file, -err); | |
710 | goto out; | |
711 | } | |
712 | ||
713 | err = init_cow_file(fd, cow_file, backing_file, sectorsize, alignment, | |
714 | bitmap_offset_out, bitmap_len_out, | |
715 | data_offset_out); | |
716 | if (!err) | |
717 | return fd; | |
718 | os_close_file(fd); | |
719 | out: | |
720 | return err; | |
721 | } | |
722 | ||
5f75a4f8 | 723 | static void ubd_close_dev(struct ubd *ubd_dev) |
1da177e4 | 724 | { |
7d314e34 PBG |
725 | os_close_file(ubd_dev->fd); |
726 | if(ubd_dev->cow.file == NULL) | |
1da177e4 LT |
727 | return; |
728 | ||
7d314e34 PBG |
729 | os_close_file(ubd_dev->cow.fd); |
730 | vfree(ubd_dev->cow.bitmap); | |
731 | ubd_dev->cow.bitmap = NULL; | |
1da177e4 LT |
732 | } |
733 | ||
7d314e34 | 734 | static int ubd_open_dev(struct ubd *ubd_dev) |
1da177e4 LT |
735 | { |
736 | struct openflags flags; | |
737 | char **back_ptr; | |
738 | int err, create_cow, *create_ptr; | |
0bf16bff | 739 | int fd; |
1da177e4 | 740 | |
7d314e34 | 741 | ubd_dev->openflags = ubd_dev->boot_openflags; |
1da177e4 | 742 | create_cow = 0; |
7d314e34 PBG |
743 | create_ptr = (ubd_dev->cow.file != NULL) ? &create_cow : NULL; |
744 | back_ptr = ubd_dev->no_cow ? NULL : &ubd_dev->cow.file; | |
0bf16bff PBG |
745 | |
746 | fd = open_ubd_file(ubd_dev->file, &ubd_dev->openflags, ubd_dev->shared, | |
7d314e34 PBG |
747 | back_ptr, &ubd_dev->cow.bitmap_offset, |
748 | &ubd_dev->cow.bitmap_len, &ubd_dev->cow.data_offset, | |
6c29256c | 749 | create_ptr); |
1da177e4 | 750 | |
0bf16bff PBG |
751 | if((fd == -ENOENT) && create_cow){ |
752 | fd = create_cow_file(ubd_dev->file, ubd_dev->cow.file, | |
7d314e34 PBG |
753 | ubd_dev->openflags, 1 << 9, PAGE_SIZE, |
754 | &ubd_dev->cow.bitmap_offset, | |
755 | &ubd_dev->cow.bitmap_len, | |
756 | &ubd_dev->cow.data_offset); | |
0bf16bff | 757 | if(fd >= 0){ |
1da177e4 | 758 | printk(KERN_INFO "Creating \"%s\" as COW file for " |
7d314e34 | 759 | "\"%s\"\n", ubd_dev->file, ubd_dev->cow.file); |
1da177e4 LT |
760 | } |
761 | } | |
762 | ||
0bf16bff | 763 | if(fd < 0){ |
7d314e34 | 764 | printk("Failed to open '%s', errno = %d\n", ubd_dev->file, |
0bf16bff PBG |
765 | -fd); |
766 | return fd; | |
1da177e4 | 767 | } |
0bf16bff | 768 | ubd_dev->fd = fd; |
1da177e4 | 769 | |
7d314e34 | 770 | if(ubd_dev->cow.file != NULL){ |
f4768ffd JD |
771 | blk_queue_max_sectors(ubd_dev->queue, 8 * sizeof(long)); |
772 | ||
1da177e4 | 773 | err = -ENOMEM; |
da2486ba | 774 | ubd_dev->cow.bitmap = vmalloc(ubd_dev->cow.bitmap_len); |
7d314e34 | 775 | if(ubd_dev->cow.bitmap == NULL){ |
1da177e4 LT |
776 | printk(KERN_ERR "Failed to vmalloc COW bitmap\n"); |
777 | goto error; | |
778 | } | |
779 | flush_tlb_kernel_vm(); | |
780 | ||
7d314e34 PBG |
781 | err = read_cow_bitmap(ubd_dev->fd, ubd_dev->cow.bitmap, |
782 | ubd_dev->cow.bitmap_offset, | |
783 | ubd_dev->cow.bitmap_len); | |
1da177e4 LT |
784 | if(err < 0) |
785 | goto error; | |
786 | ||
7d314e34 | 787 | flags = ubd_dev->openflags; |
1da177e4 | 788 | flags.w = 0; |
7d314e34 | 789 | err = open_ubd_file(ubd_dev->cow.file, &flags, ubd_dev->shared, NULL, |
6c29256c | 790 | NULL, NULL, NULL, NULL); |
1da177e4 | 791 | if(err < 0) goto error; |
7d314e34 | 792 | ubd_dev->cow.fd = err; |
1da177e4 | 793 | } |
dc764e50 | 794 | return 0; |
1da177e4 | 795 | error: |
7d314e34 | 796 | os_close_file(ubd_dev->fd); |
dc764e50 | 797 | return err; |
1da177e4 LT |
798 | } |
799 | ||
2e3f5251 JD |
800 | static void ubd_device_release(struct device *dev) |
801 | { | |
802 | struct ubd *ubd_dev = dev->driver_data; | |
803 | ||
804 | blk_cleanup_queue(ubd_dev->queue); | |
805 | *ubd_dev = ((struct ubd) DEFAULT_UBD); | |
806 | } | |
807 | ||
5f75a4f8 | 808 | static int ubd_disk_register(int major, u64 size, int unit, |
b8831a1d | 809 | struct gendisk **disk_out) |
1da177e4 LT |
810 | { |
811 | struct gendisk *disk; | |
1da177e4 LT |
812 | |
813 | disk = alloc_disk(1 << UBD_SHIFT); | |
814 | if(disk == NULL) | |
dc764e50 | 815 | return -ENOMEM; |
1da177e4 LT |
816 | |
817 | disk->major = major; | |
818 | disk->first_minor = unit << UBD_SHIFT; | |
819 | disk->fops = &ubd_blops; | |
820 | set_capacity(disk, size / 512); | |
ce7b0f46 | 821 | if(major == MAJOR_NR) |
1da177e4 | 822 | sprintf(disk->disk_name, "ubd%c", 'a' + unit); |
ce7b0f46 | 823 | else |
1da177e4 | 824 | sprintf(disk->disk_name, "ubd_fake%d", unit); |
1da177e4 LT |
825 | |
826 | /* sysfs register (not for ide fake devices) */ | |
827 | if (major == MAJOR_NR) { | |
7d314e34 PBG |
828 | ubd_devs[unit].pdev.id = unit; |
829 | ubd_devs[unit].pdev.name = DRIVER_NAME; | |
2e3f5251 JD |
830 | ubd_devs[unit].pdev.dev.release = ubd_device_release; |
831 | ubd_devs[unit].pdev.dev.driver_data = &ubd_devs[unit]; | |
7d314e34 PBG |
832 | platform_device_register(&ubd_devs[unit].pdev); |
833 | disk->driverfs_dev = &ubd_devs[unit].pdev.dev; | |
1da177e4 LT |
834 | } |
835 | ||
7d314e34 | 836 | disk->private_data = &ubd_devs[unit]; |
62f96cb0 | 837 | disk->queue = ubd_devs[unit].queue; |
1da177e4 LT |
838 | add_disk(disk); |
839 | ||
840 | *disk_out = disk; | |
841 | return 0; | |
842 | } | |
843 | ||
844 | #define ROUND_BLOCK(n) ((n + ((1 << 9) - 1)) & (-1 << 9)) | |
845 | ||
f28169d2 | 846 | static int ubd_add(int n, char **error_out) |
1da177e4 | 847 | { |
7d314e34 | 848 | struct ubd *ubd_dev = &ubd_devs[n]; |
f28169d2 | 849 | int err = 0; |
1da177e4 | 850 | |
7d314e34 | 851 | if(ubd_dev->file == NULL) |
ec7cf783 | 852 | goto out; |
1da177e4 | 853 | |
7d314e34 | 854 | err = ubd_file_size(ubd_dev, &ubd_dev->size); |
f28169d2 JD |
855 | if(err < 0){ |
856 | *error_out = "Couldn't determine size of device's file"; | |
80c13749 | 857 | goto out; |
f28169d2 | 858 | } |
1da177e4 | 859 | |
7d314e34 | 860 | ubd_dev->size = ROUND_BLOCK(ubd_dev->size); |
1da177e4 | 861 | |
a0044bdf | 862 | INIT_LIST_HEAD(&ubd_dev->restart); |
4f40c055 | 863 | sg_init_table(ubd_dev->sg, MAX_SG); |
a0044bdf | 864 | |
62f96cb0 JD |
865 | err = -ENOMEM; |
866 | ubd_dev->queue = blk_init_queue(do_ubd_request, &ubd_dev->lock); | |
867 | if (ubd_dev->queue == NULL) { | |
868 | *error_out = "Failed to initialize device queue"; | |
80c13749 | 869 | goto out; |
62f96cb0 JD |
870 | } |
871 | ubd_dev->queue->queuedata = ubd_dev; | |
872 | ||
a0044bdf | 873 | blk_queue_max_hw_segments(ubd_dev->queue, MAX_SG); |
62f96cb0 JD |
874 | err = ubd_disk_register(MAJOR_NR, ubd_dev->size, n, &ubd_gendisk[n]); |
875 | if(err){ | |
876 | *error_out = "Failed to register device"; | |
877 | goto out_cleanup; | |
878 | } | |
6c29256c | 879 | |
1da177e4 | 880 | if(fake_major != MAJOR_NR) |
5f75a4f8 | 881 | ubd_disk_register(fake_major, ubd_dev->size, n, |
62f96cb0 | 882 | &fake_gendisk[n]); |
1da177e4 | 883 | |
83380cc1 JD |
884 | /* |
885 | * Perhaps this should also be under the "if (fake_major)" above | |
886 | * using the fake_disk->disk_name | |
887 | */ | |
1da177e4 LT |
888 | if (fake_ide) |
889 | make_ide_entries(ubd_gendisk[n]->disk_name); | |
890 | ||
ec7cf783 | 891 | err = 0; |
ec7cf783 JD |
892 | out: |
893 | return err; | |
62f96cb0 JD |
894 | |
895 | out_cleanup: | |
896 | blk_cleanup_queue(ubd_dev->queue); | |
897 | goto out; | |
1da177e4 LT |
898 | } |
899 | ||
f28169d2 | 900 | static int ubd_config(char *str, char **error_out) |
1da177e4 | 901 | { |
e7f6552f | 902 | int n, ret; |
1da177e4 | 903 | |
f28169d2 JD |
904 | /* This string is possibly broken up and stored, so it's only |
905 | * freed if ubd_setup_common fails, or if only general options | |
906 | * were set. | |
907 | */ | |
970d6e3a | 908 | str = kstrdup(str, GFP_KERNEL); |
e7f6552f | 909 | if (str == NULL) { |
f28169d2 JD |
910 | *error_out = "Failed to allocate memory"; |
911 | return -ENOMEM; | |
1da177e4 | 912 | } |
f28169d2 JD |
913 | |
914 | ret = ubd_setup_common(str, &n, error_out); | |
915 | if (ret) | |
e7f6552f | 916 | goto err_free; |
f28169d2 | 917 | |
e7f6552f PBG |
918 | if (n == -1) { |
919 | ret = 0; | |
d8d7c28e | 920 | goto err_free; |
1da177e4 | 921 | } |
1da177e4 | 922 | |
dc764e50 | 923 | mutex_lock(&ubd_lock); |
f28169d2 | 924 | ret = ubd_add(n, error_out); |
e7f6552f | 925 | if (ret) |
7d314e34 | 926 | ubd_devs[n].file = NULL; |
dc764e50 | 927 | mutex_unlock(&ubd_lock); |
1da177e4 | 928 | |
e7f6552f | 929 | out: |
dc764e50 | 930 | return ret; |
e7f6552f PBG |
931 | |
932 | err_free: | |
933 | kfree(str); | |
934 | goto out; | |
1da177e4 LT |
935 | } |
936 | ||
937 | static int ubd_get_config(char *name, char *str, int size, char **error_out) | |
938 | { | |
7d314e34 | 939 | struct ubd *ubd_dev; |
1da177e4 LT |
940 | int n, len = 0; |
941 | ||
942 | n = parse_unit(&name); | |
943 | if((n >= MAX_DEV) || (n < 0)){ | |
944 | *error_out = "ubd_get_config : device number out of range"; | |
dc764e50 | 945 | return -1; |
1da177e4 LT |
946 | } |
947 | ||
7d314e34 | 948 | ubd_dev = &ubd_devs[n]; |
d7fb2c38 | 949 | mutex_lock(&ubd_lock); |
1da177e4 | 950 | |
7d314e34 | 951 | if(ubd_dev->file == NULL){ |
1da177e4 LT |
952 | CONFIG_CHUNK(str, size, len, "", 1); |
953 | goto out; | |
954 | } | |
955 | ||
7d314e34 | 956 | CONFIG_CHUNK(str, size, len, ubd_dev->file, 0); |
1da177e4 | 957 | |
7d314e34 | 958 | if(ubd_dev->cow.file != NULL){ |
1da177e4 | 959 | CONFIG_CHUNK(str, size, len, ",", 0); |
7d314e34 | 960 | CONFIG_CHUNK(str, size, len, ubd_dev->cow.file, 1); |
1da177e4 LT |
961 | } |
962 | else CONFIG_CHUNK(str, size, len, "", 1); | |
963 | ||
964 | out: | |
d7fb2c38 | 965 | mutex_unlock(&ubd_lock); |
dc764e50 | 966 | return len; |
1da177e4 LT |
967 | } |
968 | ||
29d56cfe JD |
969 | static int ubd_id(char **str, int *start_out, int *end_out) |
970 | { | |
dc764e50 | 971 | int n; |
29d56cfe JD |
972 | |
973 | n = parse_unit(str); | |
dc764e50 JD |
974 | *start_out = 0; |
975 | *end_out = MAX_DEV - 1; | |
976 | return n; | |
29d56cfe JD |
977 | } |
978 | ||
f28169d2 | 979 | static int ubd_remove(int n, char **error_out) |
1da177e4 | 980 | { |
2e3f5251 | 981 | struct gendisk *disk = ubd_gendisk[n]; |
7d314e34 | 982 | struct ubd *ubd_dev; |
29d56cfe | 983 | int err = -ENODEV; |
1da177e4 | 984 | |
d7fb2c38 | 985 | mutex_lock(&ubd_lock); |
1da177e4 | 986 | |
7d314e34 | 987 | ubd_dev = &ubd_devs[n]; |
1da177e4 | 988 | |
7d314e34 | 989 | if(ubd_dev->file == NULL) |
29d56cfe | 990 | goto out; |
1da177e4 | 991 | |
29d56cfe JD |
992 | /* you cannot remove a open disk */ |
993 | err = -EBUSY; | |
7d314e34 | 994 | if(ubd_dev->count > 0) |
1da177e4 LT |
995 | goto out; |
996 | ||
dc764e50 | 997 | ubd_gendisk[n] = NULL; |
b47d2deb JD |
998 | if(disk != NULL){ |
999 | del_gendisk(disk); | |
1000 | put_disk(disk); | |
1001 | } | |
1da177e4 LT |
1002 | |
1003 | if(fake_gendisk[n] != NULL){ | |
1004 | del_gendisk(fake_gendisk[n]); | |
1005 | put_disk(fake_gendisk[n]); | |
1006 | fake_gendisk[n] = NULL; | |
1007 | } | |
1008 | ||
1da177e4 | 1009 | err = 0; |
2e3f5251 | 1010 | platform_device_unregister(&ubd_dev->pdev); |
29d56cfe | 1011 | out: |
d7fb2c38 | 1012 | mutex_unlock(&ubd_lock); |
29d56cfe | 1013 | return err; |
1da177e4 LT |
1014 | } |
1015 | ||
f28169d2 | 1016 | /* All these are called by mconsole in process context and without |
b8831a1d | 1017 | * ubd-specific locks. The structure itself is const except for .list. |
f28169d2 | 1018 | */ |
1da177e4 | 1019 | static struct mc_device ubd_mc = { |
84f48d4f | 1020 | .list = LIST_HEAD_INIT(ubd_mc.list), |
1da177e4 LT |
1021 | .name = "ubd", |
1022 | .config = ubd_config, | |
dc764e50 | 1023 | .get_config = ubd_get_config, |
29d56cfe | 1024 | .id = ubd_id, |
1da177e4 LT |
1025 | .remove = ubd_remove, |
1026 | }; | |
1027 | ||
d8d7c28e | 1028 | static int __init ubd_mc_init(void) |
1da177e4 LT |
1029 | { |
1030 | mconsole_register_dev(&ubd_mc); | |
1031 | return 0; | |
1032 | } | |
1033 | ||
1034 | __initcall(ubd_mc_init); | |
1035 | ||
d8d7c28e PBG |
1036 | static int __init ubd0_init(void) |
1037 | { | |
1038 | struct ubd *ubd_dev = &ubd_devs[0]; | |
1039 | ||
b8831a1d | 1040 | mutex_lock(&ubd_lock); |
d8d7c28e PBG |
1041 | if(ubd_dev->file == NULL) |
1042 | ubd_dev->file = "root_fs"; | |
b8831a1d JD |
1043 | mutex_unlock(&ubd_lock); |
1044 | ||
dc764e50 | 1045 | return 0; |
d8d7c28e PBG |
1046 | } |
1047 | ||
1048 | __initcall(ubd0_init); | |
1049 | ||
b8831a1d | 1050 | /* Used in ubd_init, which is an initcall */ |
3ae5eaec RK |
1051 | static struct platform_driver ubd_driver = { |
1052 | .driver = { | |
1053 | .name = DRIVER_NAME, | |
1054 | }, | |
1da177e4 LT |
1055 | }; |
1056 | ||
d8d7c28e | 1057 | static int __init ubd_init(void) |
1da177e4 | 1058 | { |
f28169d2 JD |
1059 | char *error; |
1060 | int i, err; | |
1da177e4 | 1061 | |
1da177e4 LT |
1062 | if (register_blkdev(MAJOR_NR, "ubd")) |
1063 | return -1; | |
1064 | ||
1da177e4 LT |
1065 | if (fake_major != MAJOR_NR) { |
1066 | char name[sizeof("ubd_nnn\0")]; | |
1067 | ||
1068 | snprintf(name, sizeof(name), "ubd_%d", fake_major); | |
1da177e4 LT |
1069 | if (register_blkdev(fake_major, "ubd")) |
1070 | return -1; | |
1071 | } | |
3ae5eaec | 1072 | platform_driver_register(&ubd_driver); |
dc764e50 | 1073 | mutex_lock(&ubd_lock); |
f28169d2 JD |
1074 | for (i = 0; i < MAX_DEV; i++){ |
1075 | err = ubd_add(i, &error); | |
1076 | if(err) | |
1077 | printk(KERN_ERR "Failed to initialize ubd device %d :" | |
1078 | "%s\n", i, error); | |
1079 | } | |
dc764e50 | 1080 | mutex_unlock(&ubd_lock); |
1da177e4 LT |
1081 | return 0; |
1082 | } | |
1083 | ||
1084 | late_initcall(ubd_init); | |
1085 | ||
d8d7c28e | 1086 | static int __init ubd_driver_init(void){ |
91acb21f JD |
1087 | unsigned long stack; |
1088 | int err; | |
1089 | ||
1090 | /* Set by CONFIG_BLK_DEV_UBD_SYNC or ubd=sync.*/ | |
1091 | if(global_openflags.s){ | |
1092 | printk(KERN_INFO "ubd: Synchronous mode\n"); | |
1093 | /* Letting ubd=sync be like using ubd#s= instead of ubd#= is | |
1094 | * enough. So use anyway the io thread. */ | |
1095 | } | |
1096 | stack = alloc_stack(0, 0); | |
6c29256c | 1097 | io_pid = start_io_thread(stack + PAGE_SIZE - sizeof(void *), |
91acb21f JD |
1098 | &thread_fd); |
1099 | if(io_pid < 0){ | |
6c29256c | 1100 | printk(KERN_ERR |
91acb21f JD |
1101 | "ubd : Failed to start I/O thread (errno = %d) - " |
1102 | "falling back to synchronous I/O\n", -io_pid); | |
1103 | io_pid = -1; | |
dc764e50 | 1104 | return 0; |
91acb21f | 1105 | } |
6c29256c | 1106 | err = um_request_irq(UBD_IRQ, thread_fd, IRQ_READ, ubd_intr, |
7d314e34 | 1107 | IRQF_DISABLED, "ubd", ubd_devs); |
91acb21f JD |
1108 | if(err != 0) |
1109 | printk(KERN_ERR "um_request_irq failed - errno = %d\n", -err); | |
f4c57a78 | 1110 | return 0; |
91acb21f JD |
1111 | } |
1112 | ||
1113 | device_initcall(ubd_driver_init); | |
1114 | ||
a625c998 | 1115 | static int ubd_open(struct block_device *bdev, fmode_t mode) |
1da177e4 | 1116 | { |
a625c998 | 1117 | struct gendisk *disk = bdev->bd_disk; |
7d314e34 | 1118 | struct ubd *ubd_dev = disk->private_data; |
1da177e4 LT |
1119 | int err = 0; |
1120 | ||
7d314e34 PBG |
1121 | if(ubd_dev->count == 0){ |
1122 | err = ubd_open_dev(ubd_dev); | |
1da177e4 LT |
1123 | if(err){ |
1124 | printk(KERN_ERR "%s: Can't open \"%s\": errno = %d\n", | |
7d314e34 | 1125 | disk->disk_name, ubd_dev->file, -err); |
1da177e4 LT |
1126 | goto out; |
1127 | } | |
1128 | } | |
7d314e34 PBG |
1129 | ubd_dev->count++; |
1130 | set_disk_ro(disk, !ubd_dev->openflags.w); | |
2c49be99 PBG |
1131 | |
1132 | /* This should no more be needed. And it didn't work anyway to exclude | |
1133 | * read-write remounting of filesystems.*/ | |
a625c998 | 1134 | /*if((mode & FMODE_WRITE) && !ubd_dev->openflags.w){ |
5f75a4f8 | 1135 | if(--ubd_dev->count == 0) ubd_close_dev(ubd_dev); |
1da177e4 | 1136 | err = -EROFS; |
2c49be99 | 1137 | }*/ |
1da177e4 | 1138 | out: |
dc764e50 | 1139 | return err; |
1da177e4 LT |
1140 | } |
1141 | ||
a625c998 | 1142 | static int ubd_release(struct gendisk *disk, fmode_t mode) |
1da177e4 | 1143 | { |
7d314e34 | 1144 | struct ubd *ubd_dev = disk->private_data; |
1da177e4 | 1145 | |
7d314e34 | 1146 | if(--ubd_dev->count == 0) |
5f75a4f8 | 1147 | ubd_close_dev(ubd_dev); |
dc764e50 | 1148 | return 0; |
1da177e4 LT |
1149 | } |
1150 | ||
91acb21f JD |
1151 | static void cowify_bitmap(__u64 io_offset, int length, unsigned long *cow_mask, |
1152 | __u64 *cow_offset, unsigned long *bitmap, | |
1153 | __u64 bitmap_offset, unsigned long *bitmap_words, | |
1154 | __u64 bitmap_len) | |
1da177e4 | 1155 | { |
91acb21f JD |
1156 | __u64 sector = io_offset >> 9; |
1157 | int i, update_bitmap = 0; | |
1158 | ||
1159 | for(i = 0; i < length >> 9; i++){ | |
1160 | if(cow_mask != NULL) | |
1161 | ubd_set_bit(i, (unsigned char *) cow_mask); | |
1162 | if(ubd_test_bit(sector + i, (unsigned char *) bitmap)) | |
1163 | continue; | |
1da177e4 | 1164 | |
91acb21f JD |
1165 | update_bitmap = 1; |
1166 | ubd_set_bit(sector + i, (unsigned char *) bitmap); | |
1167 | } | |
1168 | ||
1169 | if(!update_bitmap) | |
1170 | return; | |
1da177e4 | 1171 | |
91acb21f | 1172 | *cow_offset = sector / (sizeof(unsigned long) * 8); |
1da177e4 | 1173 | |
91acb21f JD |
1174 | /* This takes care of the case where we're exactly at the end of the |
1175 | * device, and *cow_offset + 1 is off the end. So, just back it up | |
1176 | * by one word. Thanks to Lynn Kerby for the fix and James McMechan | |
1177 | * for the original diagnosis. | |
1178 | */ | |
6d074242 JO |
1179 | if (*cow_offset == (DIV_ROUND_UP(bitmap_len, |
1180 | sizeof(unsigned long)) - 1)) | |
91acb21f JD |
1181 | (*cow_offset)--; |
1182 | ||
1183 | bitmap_words[0] = bitmap[*cow_offset]; | |
1184 | bitmap_words[1] = bitmap[*cow_offset + 1]; | |
1185 | ||
1186 | *cow_offset *= sizeof(unsigned long); | |
1187 | *cow_offset += bitmap_offset; | |
1188 | } | |
1189 | ||
1190 | static void cowify_req(struct io_thread_req *req, unsigned long *bitmap, | |
1191 | __u64 bitmap_offset, __u64 bitmap_len) | |
1192 | { | |
1193 | __u64 sector = req->offset >> 9; | |
1194 | int i; | |
1195 | ||
1196 | if(req->length > (sizeof(req->sector_mask) * 8) << 9) | |
1197 | panic("Operation too long"); | |
1198 | ||
1199 | if(req->op == UBD_READ) { | |
1200 | for(i = 0; i < req->length >> 9; i++){ | |
1201 | if(ubd_test_bit(sector + i, (unsigned char *) bitmap)) | |
6c29256c | 1202 | ubd_set_bit(i, (unsigned char *) |
91acb21f | 1203 | &req->sector_mask); |
dc764e50 | 1204 | } |
91acb21f JD |
1205 | } |
1206 | else cowify_bitmap(req->offset, req->length, &req->sector_mask, | |
1207 | &req->cow_offset, bitmap, bitmap_offset, | |
1208 | req->bitmap_words, bitmap_len); | |
1da177e4 LT |
1209 | } |
1210 | ||
62f96cb0 | 1211 | /* Called with dev->lock held */ |
a0044bdf JD |
1212 | static void prepare_request(struct request *req, struct io_thread_req *io_req, |
1213 | unsigned long long offset, int page_offset, | |
1214 | int len, struct page *page) | |
1da177e4 LT |
1215 | { |
1216 | struct gendisk *disk = req->rq_disk; | |
7d314e34 | 1217 | struct ubd *ubd_dev = disk->private_data; |
91acb21f | 1218 | |
62f96cb0 | 1219 | io_req->req = req; |
a0044bdf JD |
1220 | io_req->fds[0] = (ubd_dev->cow.file != NULL) ? ubd_dev->cow.fd : |
1221 | ubd_dev->fd; | |
7d314e34 | 1222 | io_req->fds[1] = ubd_dev->fd; |
91acb21f | 1223 | io_req->cow_offset = -1; |
1da177e4 LT |
1224 | io_req->offset = offset; |
1225 | io_req->length = len; | |
1226 | io_req->error = 0; | |
91acb21f JD |
1227 | io_req->sector_mask = 0; |
1228 | ||
1229 | io_req->op = (rq_data_dir(req) == READ) ? UBD_READ : UBD_WRITE; | |
1da177e4 | 1230 | io_req->offsets[0] = 0; |
7d314e34 | 1231 | io_req->offsets[1] = ubd_dev->cow.data_offset; |
a0044bdf | 1232 | io_req->buffer = page_address(page) + page_offset; |
1da177e4 LT |
1233 | io_req->sectorsize = 1 << 9; |
1234 | ||
7d314e34 | 1235 | if(ubd_dev->cow.file != NULL) |
a0044bdf JD |
1236 | cowify_req(io_req, ubd_dev->cow.bitmap, |
1237 | ubd_dev->cow.bitmap_offset, ubd_dev->cow.bitmap_len); | |
91acb21f | 1238 | |
1da177e4 LT |
1239 | } |
1240 | ||
62f96cb0 | 1241 | /* Called with dev->lock held */ |
165125e1 | 1242 | static void do_ubd_request(struct request_queue *q) |
1da177e4 | 1243 | { |
2adcec21 | 1244 | struct io_thread_req *io_req; |
1da177e4 | 1245 | struct request *req; |
0a6d3a2a | 1246 | int n, last_sectors; |
a0044bdf JD |
1247 | |
1248 | while(1){ | |
2a9529a0 | 1249 | struct ubd *dev = q->queuedata; |
a0044bdf JD |
1250 | if(dev->end_sg == 0){ |
1251 | struct request *req = elv_next_request(q); | |
1252 | if(req == NULL) | |
1253 | return; | |
1254 | ||
1255 | dev->request = req; | |
1256 | blkdev_dequeue_request(req); | |
1257 | dev->start_sg = 0; | |
1258 | dev->end_sg = blk_rq_map_sg(q, req, dev->sg); | |
1259 | } | |
1260 | ||
1261 | req = dev->request; | |
0a6d3a2a | 1262 | last_sectors = 0; |
a0044bdf JD |
1263 | while(dev->start_sg < dev->end_sg){ |
1264 | struct scatterlist *sg = &dev->sg[dev->start_sg]; | |
1265 | ||
0a6d3a2a | 1266 | req->sector += last_sectors; |
2adcec21 | 1267 | io_req = kmalloc(sizeof(struct io_thread_req), |
990c5587 | 1268 | GFP_ATOMIC); |
2adcec21 JD |
1269 | if(io_req == NULL){ |
1270 | if(list_empty(&dev->restart)) | |
1271 | list_add(&dev->restart, &restart); | |
1272 | return; | |
1273 | } | |
1274 | prepare_request(req, io_req, | |
a0044bdf | 1275 | (unsigned long long) req->sector << 9, |
45711f1a | 1276 | sg->offset, sg->length, sg_page(sg)); |
a0044bdf | 1277 | |
0a6d3a2a | 1278 | last_sectors = sg->length >> 9; |
a6ea4cce JD |
1279 | n = os_write_file(thread_fd, &io_req, |
1280 | sizeof(struct io_thread_req *)); | |
2adcec21 | 1281 | if(n != sizeof(struct io_thread_req *)){ |
a0044bdf JD |
1282 | if(n != -EAGAIN) |
1283 | printk("write to io thread failed, " | |
1284 | "errno = %d\n", -n); | |
1285 | else if(list_empty(&dev->restart)) | |
1286 | list_add(&dev->restart, &restart); | |
12429bf9 | 1287 | kfree(io_req); |
a0044bdf JD |
1288 | return; |
1289 | } | |
1290 | ||
a0044bdf | 1291 | dev->start_sg++; |
1da177e4 | 1292 | } |
a0044bdf JD |
1293 | dev->end_sg = 0; |
1294 | dev->request = NULL; | |
1da177e4 LT |
1295 | } |
1296 | } | |
1297 | ||
a885c8c4 CH |
1298 | static int ubd_getgeo(struct block_device *bdev, struct hd_geometry *geo) |
1299 | { | |
7d314e34 | 1300 | struct ubd *ubd_dev = bdev->bd_disk->private_data; |
a885c8c4 CH |
1301 | |
1302 | geo->heads = 128; | |
1303 | geo->sectors = 32; | |
7d314e34 | 1304 | geo->cylinders = ubd_dev->size / (128 * 32 * 512); |
a885c8c4 CH |
1305 | return 0; |
1306 | } | |
1307 | ||
a625c998 | 1308 | static int ubd_ioctl(struct block_device *bdev, fmode_t mode, |
1da177e4 LT |
1309 | unsigned int cmd, unsigned long arg) |
1310 | { | |
a625c998 | 1311 | struct ubd *ubd_dev = bdev->bd_disk->private_data; |
1da177e4 LT |
1312 | struct hd_driveid ubd_id = { |
1313 | .cyls = 0, | |
1314 | .heads = 128, | |
1315 | .sectors = 32, | |
1316 | }; | |
1317 | ||
1318 | switch (cmd) { | |
1da177e4 | 1319 | struct cdrom_volctrl volume; |
1da177e4 | 1320 | case HDIO_GET_IDENTITY: |
7d314e34 | 1321 | ubd_id.cyls = ubd_dev->size / (128 * 32 * 512); |
1da177e4 LT |
1322 | if(copy_to_user((char __user *) arg, (char *) &ubd_id, |
1323 | sizeof(ubd_id))) | |
dc764e50 JD |
1324 | return -EFAULT; |
1325 | return 0; | |
b8831a1d | 1326 | |
1da177e4 LT |
1327 | case CDROMVOLREAD: |
1328 | if(copy_from_user(&volume, (char __user *) arg, sizeof(volume))) | |
dc764e50 | 1329 | return -EFAULT; |
1da177e4 LT |
1330 | volume.channel0 = 255; |
1331 | volume.channel1 = 255; | |
1332 | volume.channel2 = 255; | |
1333 | volume.channel3 = 255; | |
1334 | if(copy_to_user((char __user *) arg, &volume, sizeof(volume))) | |
dc764e50 JD |
1335 | return -EFAULT; |
1336 | return 0; | |
1da177e4 | 1337 | } |
dc764e50 | 1338 | return -EINVAL; |
1da177e4 LT |
1339 | } |
1340 | ||
91acb21f | 1341 | static int update_bitmap(struct io_thread_req *req) |
1da177e4 | 1342 | { |
91acb21f | 1343 | int n; |
1da177e4 | 1344 | |
91acb21f | 1345 | if(req->cow_offset == -1) |
dc764e50 | 1346 | return 0; |
1da177e4 | 1347 | |
91acb21f JD |
1348 | n = os_seek_file(req->fds[1], req->cow_offset); |
1349 | if(n < 0){ | |
1350 | printk("do_io - bitmap lseek failed : err = %d\n", -n); | |
dc764e50 | 1351 | return 1; |
91acb21f | 1352 | } |
1da177e4 | 1353 | |
a6ea4cce JD |
1354 | n = os_write_file(req->fds[1], &req->bitmap_words, |
1355 | sizeof(req->bitmap_words)); | |
91acb21f JD |
1356 | if(n != sizeof(req->bitmap_words)){ |
1357 | printk("do_io - bitmap update failed, err = %d fd = %d\n", -n, | |
1358 | req->fds[1]); | |
dc764e50 | 1359 | return 1; |
91acb21f | 1360 | } |
1da177e4 | 1361 | |
dc764e50 | 1362 | return 0; |
91acb21f | 1363 | } |
1da177e4 | 1364 | |
5dc62b1b | 1365 | static void do_io(struct io_thread_req *req) |
91acb21f JD |
1366 | { |
1367 | char *buf; | |
1368 | unsigned long len; | |
1369 | int n, nsectors, start, end, bit; | |
1370 | int err; | |
1371 | __u64 off; | |
1372 | ||
1373 | nsectors = req->length / req->sectorsize; | |
1374 | start = 0; | |
1375 | do { | |
1376 | bit = ubd_test_bit(start, (unsigned char *) &req->sector_mask); | |
1377 | end = start; | |
1378 | while((end < nsectors) && | |
1379 | (ubd_test_bit(end, (unsigned char *) | |
1380 | &req->sector_mask) == bit)) | |
1381 | end++; | |
1382 | ||
1383 | off = req->offset + req->offsets[bit] + | |
1384 | start * req->sectorsize; | |
1385 | len = (end - start) * req->sectorsize; | |
1386 | buf = &req->buffer[start * req->sectorsize]; | |
1387 | ||
1388 | err = os_seek_file(req->fds[bit], off); | |
1389 | if(err < 0){ | |
1390 | printk("do_io - lseek failed : err = %d\n", -err); | |
1391 | req->error = 1; | |
1392 | return; | |
1393 | } | |
1394 | if(req->op == UBD_READ){ | |
1395 | n = 0; | |
1396 | do { | |
1397 | buf = &buf[n]; | |
1398 | len -= n; | |
a6ea4cce | 1399 | n = os_read_file(req->fds[bit], buf, len); |
91acb21f JD |
1400 | if (n < 0) { |
1401 | printk("do_io - read failed, err = %d " | |
1402 | "fd = %d\n", -n, req->fds[bit]); | |
1403 | req->error = 1; | |
1404 | return; | |
1405 | } | |
1406 | } while((n < len) && (n != 0)); | |
1407 | if (n < len) memset(&buf[n], 0, len - n); | |
1408 | } else { | |
a6ea4cce | 1409 | n = os_write_file(req->fds[bit], buf, len); |
91acb21f JD |
1410 | if(n != len){ |
1411 | printk("do_io - write failed err = %d " | |
1412 | "fd = %d\n", -n, req->fds[bit]); | |
1413 | req->error = 1; | |
1414 | return; | |
1415 | } | |
1416 | } | |
1417 | ||
1418 | start = end; | |
1419 | } while(start < nsectors); | |
1da177e4 | 1420 | |
91acb21f | 1421 | req->error = update_bitmap(req); |
1da177e4 | 1422 | } |
91acb21f JD |
1423 | |
1424 | /* Changed in start_io_thread, which is serialized by being called only | |
1425 | * from ubd_init, which is an initcall. | |
1426 | */ | |
1427 | int kernel_fd = -1; | |
1428 | ||
d8d7c28e PBG |
1429 | /* Only changed by the io thread. XXX: currently unused. */ |
1430 | static int io_count = 0; | |
91acb21f JD |
1431 | |
1432 | int io_thread(void *arg) | |
1433 | { | |
2adcec21 | 1434 | struct io_thread_req *req; |
91acb21f JD |
1435 | int n; |
1436 | ||
1437 | ignore_sigwinch_sig(); | |
1438 | while(1){ | |
a6ea4cce | 1439 | n = os_read_file(kernel_fd, &req, |
2adcec21 JD |
1440 | sizeof(struct io_thread_req *)); |
1441 | if(n != sizeof(struct io_thread_req *)){ | |
91acb21f JD |
1442 | if(n < 0) |
1443 | printk("io_thread - read failed, fd = %d, " | |
1444 | "err = %d\n", kernel_fd, -n); | |
1445 | else { | |
1446 | printk("io_thread - short read, fd = %d, " | |
1447 | "length = %d\n", kernel_fd, n); | |
1448 | } | |
1449 | continue; | |
1450 | } | |
1451 | io_count++; | |
2adcec21 | 1452 | do_io(req); |
a6ea4cce | 1453 | n = os_write_file(kernel_fd, &req, |
2adcec21 JD |
1454 | sizeof(struct io_thread_req *)); |
1455 | if(n != sizeof(struct io_thread_req *)) | |
91acb21f JD |
1456 | printk("io_thread - write failed, fd = %d, err = %d\n", |
1457 | kernel_fd, -n); | |
1458 | } | |
91acb21f | 1459 | |
1b57e9c2 JD |
1460 | return 0; |
1461 | } |