]>
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 | ||
6c29256c | 74 | extern int open_ubd_file(char *file, struct openflags *openflags, int shared, |
1da177e4 LT |
75 | char **backing_file_out, int *bitmap_offset_out, |
76 | unsigned long *bitmap_len_out, int *data_offset_out, | |
77 | int *create_cow_out); | |
78 | extern int create_cow_file(char *cow_file, char *backing_file, | |
79 | struct openflags flags, int sectorsize, | |
80 | int alignment, int *bitmap_offset_out, | |
81 | unsigned long *bitmap_len_out, | |
82 | int *data_offset_out); | |
83 | extern int read_cow_bitmap(int fd, void *buf, int offset, int len); | |
91acb21f | 84 | extern void do_io(struct io_thread_req *req); |
1da177e4 | 85 | |
91acb21f | 86 | static inline int ubd_test_bit(__u64 bit, unsigned char *data) |
1da177e4 LT |
87 | { |
88 | __u64 n; | |
89 | int bits, off; | |
90 | ||
91acb21f | 91 | bits = sizeof(data[0]) * 8; |
1da177e4 LT |
92 | n = bit / bits; |
93 | off = bit % bits; | |
dc764e50 | 94 | return (data[n] & (1 << off)) != 0; |
1da177e4 LT |
95 | } |
96 | ||
91acb21f | 97 | static inline void ubd_set_bit(__u64 bit, unsigned char *data) |
1da177e4 LT |
98 | { |
99 | __u64 n; | |
100 | int bits, off; | |
101 | ||
91acb21f | 102 | bits = sizeof(data[0]) * 8; |
1da177e4 LT |
103 | n = bit / bits; |
104 | off = bit % bits; | |
91acb21f | 105 | data[n] |= (1 << off); |
1da177e4 LT |
106 | } |
107 | /*End stuff from ubd_user.h*/ | |
108 | ||
109 | #define DRIVER_NAME "uml-blkdev" | |
110 | ||
d7fb2c38 | 111 | static DEFINE_MUTEX(ubd_lock); |
1da177e4 | 112 | |
1da177e4 LT |
113 | static int ubd_open(struct inode * inode, struct file * filp); |
114 | static int ubd_release(struct inode * inode, struct file * file); | |
115 | static int ubd_ioctl(struct inode * inode, struct file * file, | |
116 | unsigned int cmd, unsigned long arg); | |
a885c8c4 | 117 | static int ubd_getgeo(struct block_device *bdev, struct hd_geometry *geo); |
1da177e4 | 118 | |
97d88ac8 | 119 | #define MAX_DEV (16) |
1da177e4 | 120 | |
1da177e4 LT |
121 | static struct block_device_operations ubd_blops = { |
122 | .owner = THIS_MODULE, | |
123 | .open = ubd_open, | |
124 | .release = ubd_release, | |
125 | .ioctl = ubd_ioctl, | |
a885c8c4 | 126 | .getgeo = ubd_getgeo, |
1da177e4 LT |
127 | }; |
128 | ||
1da177e4 LT |
129 | /* Protected by ubd_lock */ |
130 | static int fake_major = MAJOR_NR; | |
1da177e4 LT |
131 | static struct gendisk *ubd_gendisk[MAX_DEV]; |
132 | static struct gendisk *fake_gendisk[MAX_DEV]; | |
6c29256c | 133 | |
1da177e4 LT |
134 | #ifdef CONFIG_BLK_DEV_UBD_SYNC |
135 | #define OPEN_FLAGS ((struct openflags) { .r = 1, .w = 1, .s = 1, .c = 0, \ | |
136 | .cl = 1 }) | |
137 | #else | |
138 | #define OPEN_FLAGS ((struct openflags) { .r = 1, .w = 1, .s = 0, .c = 0, \ | |
139 | .cl = 1 }) | |
140 | #endif | |
1da177e4 LT |
141 | static struct openflags global_openflags = OPEN_FLAGS; |
142 | ||
143 | struct cow { | |
2a9d32f6 | 144 | /* backing file name */ |
1da177e4 | 145 | char *file; |
2a9d32f6 | 146 | /* backing file fd */ |
1da177e4 LT |
147 | int fd; |
148 | unsigned long *bitmap; | |
149 | unsigned long bitmap_len; | |
150 | int bitmap_offset; | |
dc764e50 | 151 | int data_offset; |
1da177e4 LT |
152 | }; |
153 | ||
a0044bdf JD |
154 | #define MAX_SG 64 |
155 | ||
1da177e4 | 156 | struct ubd { |
a0044bdf | 157 | struct list_head restart; |
2a9d32f6 PBG |
158 | /* name (and fd, below) of the file opened for writing, either the |
159 | * backing or the cow file. */ | |
1da177e4 LT |
160 | char *file; |
161 | int count; | |
162 | int fd; | |
163 | __u64 size; | |
164 | struct openflags boot_openflags; | |
165 | struct openflags openflags; | |
84e945e3 PBG |
166 | unsigned shared:1; |
167 | unsigned no_cow:1; | |
1da177e4 LT |
168 | struct cow cow; |
169 | struct platform_device pdev; | |
62f96cb0 JD |
170 | struct request_queue *queue; |
171 | spinlock_t lock; | |
a0044bdf JD |
172 | struct scatterlist sg[MAX_SG]; |
173 | struct request *request; | |
174 | int start_sg, end_sg; | |
1da177e4 LT |
175 | }; |
176 | ||
177 | #define DEFAULT_COW { \ | |
178 | .file = NULL, \ | |
dc764e50 JD |
179 | .fd = -1, \ |
180 | .bitmap = NULL, \ | |
1da177e4 | 181 | .bitmap_offset = 0, \ |
dc764e50 | 182 | .data_offset = 0, \ |
1da177e4 LT |
183 | } |
184 | ||
185 | #define DEFAULT_UBD { \ | |
186 | .file = NULL, \ | |
187 | .count = 0, \ | |
188 | .fd = -1, \ | |
189 | .size = -1, \ | |
190 | .boot_openflags = OPEN_FLAGS, \ | |
191 | .openflags = OPEN_FLAGS, \ | |
dc764e50 | 192 | .no_cow = 0, \ |
6c29256c | 193 | .shared = 0, \ |
dc764e50 | 194 | .cow = DEFAULT_COW, \ |
62f96cb0 | 195 | .lock = SPIN_LOCK_UNLOCKED, \ |
a0044bdf JD |
196 | .request = NULL, \ |
197 | .start_sg = 0, \ | |
198 | .end_sg = 0, \ | |
1da177e4 LT |
199 | } |
200 | ||
b8831a1d | 201 | /* Protected by ubd_lock */ |
7d314e34 | 202 | struct ubd ubd_devs[MAX_DEV] = { [ 0 ... MAX_DEV - 1 ] = DEFAULT_UBD }; |
1da177e4 | 203 | |
1da177e4 LT |
204 | /* Only changed by fake_ide_setup which is a setup */ |
205 | static int fake_ide = 0; | |
206 | static struct proc_dir_entry *proc_ide_root = NULL; | |
207 | static struct proc_dir_entry *proc_ide = NULL; | |
208 | ||
209 | static void make_proc_ide(void) | |
210 | { | |
211 | proc_ide_root = proc_mkdir("ide", NULL); | |
212 | proc_ide = proc_mkdir("ide0", proc_ide_root); | |
213 | } | |
214 | ||
215 | static int proc_ide_read_media(char *page, char **start, off_t off, int count, | |
216 | int *eof, void *data) | |
217 | { | |
218 | int len; | |
219 | ||
220 | strcpy(page, "disk\n"); | |
221 | len = strlen("disk\n"); | |
222 | len -= off; | |
223 | if (len < count){ | |
224 | *eof = 1; | |
225 | if (len <= 0) return 0; | |
226 | } | |
227 | else len = count; | |
228 | *start = page + off; | |
229 | return len; | |
230 | } | |
231 | ||
232 | static void make_ide_entries(char *dev_name) | |
233 | { | |
234 | struct proc_dir_entry *dir, *ent; | |
235 | char name[64]; | |
236 | ||
237 | if(proc_ide_root == NULL) make_proc_ide(); | |
238 | ||
239 | dir = proc_mkdir(dev_name, proc_ide); | |
240 | if(!dir) return; | |
241 | ||
242 | ent = create_proc_entry("media", S_IFREG|S_IRUGO, dir); | |
243 | if(!ent) return; | |
1da177e4 LT |
244 | ent->data = NULL; |
245 | ent->read_proc = proc_ide_read_media; | |
246 | ent->write_proc = NULL; | |
247 | sprintf(name,"ide0/%s", dev_name); | |
248 | proc_symlink(dev_name, proc_ide_root, name); | |
249 | } | |
250 | ||
251 | static int fake_ide_setup(char *str) | |
252 | { | |
253 | fake_ide = 1; | |
dc764e50 | 254 | return 1; |
1da177e4 LT |
255 | } |
256 | ||
257 | __setup("fake_ide", fake_ide_setup); | |
258 | ||
259 | __uml_help(fake_ide_setup, | |
260 | "fake_ide\n" | |
261 | " Create ide0 entries that map onto ubd devices.\n\n" | |
262 | ); | |
263 | ||
264 | static int parse_unit(char **ptr) | |
265 | { | |
266 | char *str = *ptr, *end; | |
267 | int n = -1; | |
268 | ||
269 | if(isdigit(*str)) { | |
270 | n = simple_strtoul(str, &end, 0); | |
271 | if(end == str) | |
dc764e50 | 272 | return -1; |
1da177e4 LT |
273 | *ptr = end; |
274 | } | |
97d88ac8 | 275 | else if (('a' <= *str) && (*str <= 'z')) { |
1da177e4 LT |
276 | n = *str - 'a'; |
277 | str++; | |
278 | *ptr = str; | |
279 | } | |
dc764e50 | 280 | return n; |
1da177e4 LT |
281 | } |
282 | ||
d8d7c28e PBG |
283 | /* If *index_out == -1 at exit, the passed option was a general one; |
284 | * otherwise, the str pointer is used (and owned) inside ubd_devs array, so it | |
285 | * should not be freed on exit. | |
286 | */ | |
f28169d2 | 287 | static int ubd_setup_common(char *str, int *index_out, char **error_out) |
1da177e4 | 288 | { |
7d314e34 | 289 | struct ubd *ubd_dev; |
1da177e4 LT |
290 | struct openflags flags = global_openflags; |
291 | char *backing_file; | |
b8831a1d | 292 | int n, err = 0, i; |
1da177e4 LT |
293 | |
294 | if(index_out) *index_out = -1; | |
295 | n = *str; | |
296 | if(n == '='){ | |
297 | char *end; | |
298 | int major; | |
299 | ||
300 | str++; | |
1da177e4 LT |
301 | if(!strcmp(str, "sync")){ |
302 | global_openflags = of_sync(global_openflags); | |
b8831a1d | 303 | goto out1; |
1da177e4 | 304 | } |
b8831a1d JD |
305 | |
306 | err = -EINVAL; | |
1da177e4 LT |
307 | major = simple_strtoul(str, &end, 0); |
308 | if((*end != '\0') || (end == str)){ | |
f28169d2 | 309 | *error_out = "Didn't parse major number"; |
b8831a1d | 310 | goto out1; |
1da177e4 LT |
311 | } |
312 | ||
f28169d2 JD |
313 | mutex_lock(&ubd_lock); |
314 | if(fake_major != MAJOR_NR){ | |
315 | *error_out = "Can't assign a fake major twice"; | |
316 | goto out1; | |
317 | } | |
6c29256c | 318 | |
f28169d2 | 319 | fake_major = major; |
1da177e4 LT |
320 | |
321 | printk(KERN_INFO "Setting extra ubd major number to %d\n", | |
322 | major); | |
f28169d2 JD |
323 | err = 0; |
324 | out1: | |
325 | mutex_unlock(&ubd_lock); | |
326 | return err; | |
1da177e4 LT |
327 | } |
328 | ||
329 | n = parse_unit(&str); | |
330 | if(n < 0){ | |
f28169d2 JD |
331 | *error_out = "Couldn't parse device number"; |
332 | return -EINVAL; | |
1da177e4 LT |
333 | } |
334 | if(n >= MAX_DEV){ | |
f28169d2 JD |
335 | *error_out = "Device number out of range"; |
336 | return 1; | |
1da177e4 LT |
337 | } |
338 | ||
f28169d2 | 339 | err = -EBUSY; |
d7fb2c38 | 340 | mutex_lock(&ubd_lock); |
1da177e4 | 341 | |
7d314e34 PBG |
342 | ubd_dev = &ubd_devs[n]; |
343 | if(ubd_dev->file != NULL){ | |
f28169d2 | 344 | *error_out = "Device is already configured"; |
1da177e4 LT |
345 | goto out; |
346 | } | |
347 | ||
348 | if (index_out) | |
349 | *index_out = n; | |
350 | ||
f28169d2 | 351 | err = -EINVAL; |
6c29256c | 352 | for (i = 0; i < sizeof("rscd="); i++) { |
1da177e4 LT |
353 | switch (*str) { |
354 | case 'r': | |
355 | flags.w = 0; | |
356 | break; | |
357 | case 's': | |
358 | flags.s = 1; | |
359 | break; | |
360 | case 'd': | |
7d314e34 | 361 | ubd_dev->no_cow = 1; |
1da177e4 | 362 | break; |
6c29256c | 363 | case 'c': |
7d314e34 | 364 | ubd_dev->shared = 1; |
6c29256c | 365 | break; |
1da177e4 LT |
366 | case '=': |
367 | str++; | |
368 | goto break_loop; | |
369 | default: | |
f28169d2 JD |
370 | *error_out = "Expected '=' or flag letter " |
371 | "(r, s, c, or d)"; | |
1da177e4 LT |
372 | goto out; |
373 | } | |
374 | str++; | |
375 | } | |
376 | ||
f28169d2 JD |
377 | if (*str == '=') |
378 | *error_out = "Too many flags specified"; | |
379 | else | |
380 | *error_out = "Missing '='"; | |
1da177e4 LT |
381 | goto out; |
382 | ||
383 | break_loop: | |
1da177e4 LT |
384 | backing_file = strchr(str, ','); |
385 | ||
f28169d2 | 386 | if (backing_file == NULL) |
1da177e4 | 387 | backing_file = strchr(str, ':'); |
1da177e4 | 388 | |
f28169d2 JD |
389 | if(backing_file != NULL){ |
390 | if(ubd_dev->no_cow){ | |
391 | *error_out = "Can't specify both 'd' and a cow file"; | |
392 | goto out; | |
393 | } | |
1da177e4 LT |
394 | else { |
395 | *backing_file = '\0'; | |
396 | backing_file++; | |
397 | } | |
398 | } | |
f28169d2 | 399 | err = 0; |
7d314e34 PBG |
400 | ubd_dev->file = str; |
401 | ubd_dev->cow.file = backing_file; | |
402 | ubd_dev->boot_openflags = flags; | |
1da177e4 | 403 | out: |
d7fb2c38 | 404 | mutex_unlock(&ubd_lock); |
f28169d2 | 405 | return err; |
1da177e4 LT |
406 | } |
407 | ||
408 | static int ubd_setup(char *str) | |
409 | { | |
f28169d2 JD |
410 | char *error; |
411 | int err; | |
412 | ||
413 | err = ubd_setup_common(str, NULL, &error); | |
414 | if(err) | |
415 | printk(KERN_ERR "Failed to initialize device with \"%s\" : " | |
416 | "%s\n", str, error); | |
417 | return 1; | |
1da177e4 LT |
418 | } |
419 | ||
420 | __setup("ubd", ubd_setup); | |
421 | __uml_help(ubd_setup, | |
422 | "ubd<n><flags>=<filename>[(:|,)<filename2>]\n" | |
423 | " This is used to associate a device with a file in the underlying\n" | |
424 | " filesystem. When specifying two filenames, the first one is the\n" | |
425 | " COW name and the second is the backing file name. As separator you can\n" | |
426 | " use either a ':' or a ',': the first one allows writing things like;\n" | |
427 | " ubd0=~/Uml/root_cow:~/Uml/root_backing_file\n" | |
428 | " while with a ',' the shell would not expand the 2nd '~'.\n" | |
f28169d2 | 429 | " When using only one filename, UML will detect whether to treat it like\n" |
1da177e4 LT |
430 | " a COW file or a backing file. To override this detection, add the 'd'\n" |
431 | " flag:\n" | |
432 | " ubd0d=BackingFile\n" | |
433 | " Usually, there is a filesystem in the file, but \n" | |
434 | " that's not required. Swap devices containing swap files can be\n" | |
435 | " specified like this. Also, a file which doesn't contain a\n" | |
436 | " filesystem can have its contents read in the virtual \n" | |
437 | " machine by running 'dd' on the device. <n> must be in the range\n" | |
438 | " 0 to 7. Appending an 'r' to the number will cause that device\n" | |
439 | " to be mounted read-only. For example ubd1r=./ext_fs. Appending\n" | |
440 | " an 's' will cause data to be written to disk on the host immediately.\n\n" | |
441 | ); | |
442 | ||
443 | static int udb_setup(char *str) | |
444 | { | |
445 | printk("udb%s specified on command line is almost certainly a ubd -> " | |
446 | "udb TYPO\n", str); | |
dc764e50 | 447 | return 1; |
1da177e4 LT |
448 | } |
449 | ||
450 | __setup("udb", udb_setup); | |
451 | __uml_help(udb_setup, | |
452 | "udb\n" | |
0894e27e JD |
453 | " This option is here solely to catch ubd -> udb typos, which can be\n" |
454 | " to impossible to catch visually unless you specifically look for\n" | |
455 | " them. The only result of any option starting with 'udb' is an error\n" | |
1da177e4 LT |
456 | " in the boot output.\n\n" |
457 | ); | |
458 | ||
459 | static int fakehd_set = 0; | |
460 | static int fakehd(char *str) | |
461 | { | |
462 | printk(KERN_INFO "fakehd : Changing ubd name to \"hd\".\n"); | |
463 | fakehd_set = 1; | |
464 | return 1; | |
465 | } | |
466 | ||
467 | __setup("fakehd", fakehd); | |
468 | __uml_help(fakehd, | |
469 | "fakehd\n" | |
470 | " Change the ubd device name to \"hd\".\n\n" | |
471 | ); | |
472 | ||
165125e1 | 473 | static void do_ubd_request(struct request_queue * q); |
91acb21f JD |
474 | |
475 | /* Only changed by ubd_init, which is an initcall. */ | |
476 | int thread_fd = -1; | |
1da177e4 | 477 | |
a0044bdf | 478 | static void ubd_end_request(struct request *req, int bytes, int uptodate) |
1da177e4 | 479 | { |
a0044bdf JD |
480 | if (!end_that_request_first(req, uptodate, bytes >> 9)) { |
481 | struct ubd *dev = req->rq_disk->private_data; | |
482 | unsigned long flags; | |
483 | ||
484 | add_disk_randomness(req->rq_disk); | |
485 | spin_lock_irqsave(&dev->lock, flags); | |
486 | end_that_request_last(req, uptodate); | |
487 | spin_unlock_irqrestore(&dev->lock, flags); | |
1da177e4 | 488 | } |
1da177e4 LT |
489 | } |
490 | ||
33f775ee PBG |
491 | /* Callable only from interrupt context - otherwise you need to do |
492 | * spin_lock_irq()/spin_lock_irqsave() */ | |
a0044bdf | 493 | static inline void ubd_finish(struct request *req, int bytes) |
1da177e4 | 494 | { |
a0044bdf JD |
495 | if(bytes < 0){ |
496 | ubd_end_request(req, 0, 0); | |
497 | return; | |
498 | } | |
499 | ubd_end_request(req, bytes, 1); | |
1da177e4 LT |
500 | } |
501 | ||
a0044bdf JD |
502 | static LIST_HEAD(restart); |
503 | ||
2fe30a34 | 504 | /* XXX - move this inside ubd_intr. */ |
62f96cb0 | 505 | /* Called without dev->lock held, and only in interrupt context. */ |
91acb21f | 506 | static void ubd_handler(void) |
1da177e4 | 507 | { |
2adcec21 | 508 | struct io_thread_req *req; |
62f96cb0 | 509 | struct request *rq; |
a0044bdf JD |
510 | struct ubd *ubd; |
511 | struct list_head *list, *next_ele; | |
512 | unsigned long flags; | |
91acb21f JD |
513 | int n; |
514 | ||
a0044bdf | 515 | while(1){ |
a6ea4cce JD |
516 | n = os_read_file(thread_fd, &req, |
517 | sizeof(struct io_thread_req *)); | |
a0044bdf JD |
518 | if(n != sizeof(req)){ |
519 | if(n == -EAGAIN) | |
520 | break; | |
521 | printk(KERN_ERR "spurious interrupt in ubd_handler, " | |
522 | "err = %d\n", -n); | |
523 | return; | |
524 | } | |
62f96cb0 | 525 | |
2adcec21 JD |
526 | rq = req->req; |
527 | rq->nr_sectors -= req->length >> 9; | |
a0044bdf JD |
528 | if(rq->nr_sectors == 0) |
529 | ubd_finish(rq, rq->hard_nr_sectors << 9); | |
2adcec21 | 530 | kfree(req); |
a0044bdf | 531 | } |
62f96cb0 | 532 | reactivate_fd(thread_fd, UBD_IRQ); |
a0044bdf JD |
533 | |
534 | list_for_each_safe(list, next_ele, &restart){ | |
535 | ubd = container_of(list, struct ubd, restart); | |
536 | list_del_init(&ubd->restart); | |
537 | spin_lock_irqsave(&ubd->lock, flags); | |
538 | do_ubd_request(ubd->queue); | |
539 | spin_unlock_irqrestore(&ubd->lock, flags); | |
540 | } | |
1da177e4 LT |
541 | } |
542 | ||
7bea96fd | 543 | static irqreturn_t ubd_intr(int irq, void *dev) |
1da177e4 | 544 | { |
91acb21f | 545 | ubd_handler(); |
dc764e50 | 546 | return IRQ_HANDLED; |
91acb21f | 547 | } |
09ace81c | 548 | |
91acb21f JD |
549 | /* Only changed by ubd_init, which is an initcall. */ |
550 | static int io_pid = -1; | |
09ace81c | 551 | |
91acb21f JD |
552 | void kill_io_thread(void) |
553 | { | |
6c29256c | 554 | if(io_pid != -1) |
91acb21f | 555 | os_kill_process(io_pid, 1); |
09ace81c | 556 | } |
1da177e4 | 557 | |
91acb21f JD |
558 | __uml_exitcall(kill_io_thread); |
559 | ||
d8d7c28e | 560 | static inline int ubd_file_size(struct ubd *ubd_dev, __u64 *size_out) |
1da177e4 LT |
561 | { |
562 | char *file; | |
563 | ||
7d314e34 | 564 | file = ubd_dev->cow.file ? ubd_dev->cow.file : ubd_dev->file; |
dc764e50 | 565 | return os_file_size(file, size_out); |
1da177e4 LT |
566 | } |
567 | ||
5f75a4f8 | 568 | static void ubd_close_dev(struct ubd *ubd_dev) |
1da177e4 | 569 | { |
7d314e34 PBG |
570 | os_close_file(ubd_dev->fd); |
571 | if(ubd_dev->cow.file == NULL) | |
1da177e4 LT |
572 | return; |
573 | ||
7d314e34 PBG |
574 | os_close_file(ubd_dev->cow.fd); |
575 | vfree(ubd_dev->cow.bitmap); | |
576 | ubd_dev->cow.bitmap = NULL; | |
1da177e4 LT |
577 | } |
578 | ||
7d314e34 | 579 | static int ubd_open_dev(struct ubd *ubd_dev) |
1da177e4 LT |
580 | { |
581 | struct openflags flags; | |
582 | char **back_ptr; | |
583 | int err, create_cow, *create_ptr; | |
0bf16bff | 584 | int fd; |
1da177e4 | 585 | |
7d314e34 | 586 | ubd_dev->openflags = ubd_dev->boot_openflags; |
1da177e4 | 587 | create_cow = 0; |
7d314e34 PBG |
588 | create_ptr = (ubd_dev->cow.file != NULL) ? &create_cow : NULL; |
589 | back_ptr = ubd_dev->no_cow ? NULL : &ubd_dev->cow.file; | |
0bf16bff PBG |
590 | |
591 | fd = open_ubd_file(ubd_dev->file, &ubd_dev->openflags, ubd_dev->shared, | |
7d314e34 PBG |
592 | back_ptr, &ubd_dev->cow.bitmap_offset, |
593 | &ubd_dev->cow.bitmap_len, &ubd_dev->cow.data_offset, | |
6c29256c | 594 | create_ptr); |
1da177e4 | 595 | |
0bf16bff PBG |
596 | if((fd == -ENOENT) && create_cow){ |
597 | fd = create_cow_file(ubd_dev->file, ubd_dev->cow.file, | |
7d314e34 PBG |
598 | ubd_dev->openflags, 1 << 9, PAGE_SIZE, |
599 | &ubd_dev->cow.bitmap_offset, | |
600 | &ubd_dev->cow.bitmap_len, | |
601 | &ubd_dev->cow.data_offset); | |
0bf16bff | 602 | if(fd >= 0){ |
1da177e4 | 603 | printk(KERN_INFO "Creating \"%s\" as COW file for " |
7d314e34 | 604 | "\"%s\"\n", ubd_dev->file, ubd_dev->cow.file); |
1da177e4 LT |
605 | } |
606 | } | |
607 | ||
0bf16bff | 608 | if(fd < 0){ |
7d314e34 | 609 | printk("Failed to open '%s', errno = %d\n", ubd_dev->file, |
0bf16bff PBG |
610 | -fd); |
611 | return fd; | |
1da177e4 | 612 | } |
0bf16bff | 613 | ubd_dev->fd = fd; |
1da177e4 | 614 | |
7d314e34 | 615 | if(ubd_dev->cow.file != NULL){ |
f4768ffd JD |
616 | blk_queue_max_sectors(ubd_dev->queue, 8 * sizeof(long)); |
617 | ||
1da177e4 | 618 | err = -ENOMEM; |
da2486ba | 619 | ubd_dev->cow.bitmap = vmalloc(ubd_dev->cow.bitmap_len); |
7d314e34 | 620 | if(ubd_dev->cow.bitmap == NULL){ |
1da177e4 LT |
621 | printk(KERN_ERR "Failed to vmalloc COW bitmap\n"); |
622 | goto error; | |
623 | } | |
624 | flush_tlb_kernel_vm(); | |
625 | ||
7d314e34 PBG |
626 | err = read_cow_bitmap(ubd_dev->fd, ubd_dev->cow.bitmap, |
627 | ubd_dev->cow.bitmap_offset, | |
628 | ubd_dev->cow.bitmap_len); | |
1da177e4 LT |
629 | if(err < 0) |
630 | goto error; | |
631 | ||
7d314e34 | 632 | flags = ubd_dev->openflags; |
1da177e4 | 633 | flags.w = 0; |
7d314e34 | 634 | err = open_ubd_file(ubd_dev->cow.file, &flags, ubd_dev->shared, NULL, |
6c29256c | 635 | NULL, NULL, NULL, NULL); |
1da177e4 | 636 | if(err < 0) goto error; |
7d314e34 | 637 | ubd_dev->cow.fd = err; |
1da177e4 | 638 | } |
dc764e50 | 639 | return 0; |
1da177e4 | 640 | error: |
7d314e34 | 641 | os_close_file(ubd_dev->fd); |
dc764e50 | 642 | return err; |
1da177e4 LT |
643 | } |
644 | ||
2e3f5251 JD |
645 | static void ubd_device_release(struct device *dev) |
646 | { | |
647 | struct ubd *ubd_dev = dev->driver_data; | |
648 | ||
649 | blk_cleanup_queue(ubd_dev->queue); | |
650 | *ubd_dev = ((struct ubd) DEFAULT_UBD); | |
651 | } | |
652 | ||
5f75a4f8 | 653 | static int ubd_disk_register(int major, u64 size, int unit, |
b8831a1d | 654 | struct gendisk **disk_out) |
1da177e4 LT |
655 | { |
656 | struct gendisk *disk; | |
1da177e4 LT |
657 | |
658 | disk = alloc_disk(1 << UBD_SHIFT); | |
659 | if(disk == NULL) | |
dc764e50 | 660 | return -ENOMEM; |
1da177e4 LT |
661 | |
662 | disk->major = major; | |
663 | disk->first_minor = unit << UBD_SHIFT; | |
664 | disk->fops = &ubd_blops; | |
665 | set_capacity(disk, size / 512); | |
ce7b0f46 | 666 | if(major == MAJOR_NR) |
1da177e4 | 667 | sprintf(disk->disk_name, "ubd%c", 'a' + unit); |
ce7b0f46 | 668 | else |
1da177e4 | 669 | sprintf(disk->disk_name, "ubd_fake%d", unit); |
1da177e4 LT |
670 | |
671 | /* sysfs register (not for ide fake devices) */ | |
672 | if (major == MAJOR_NR) { | |
7d314e34 PBG |
673 | ubd_devs[unit].pdev.id = unit; |
674 | ubd_devs[unit].pdev.name = DRIVER_NAME; | |
2e3f5251 JD |
675 | ubd_devs[unit].pdev.dev.release = ubd_device_release; |
676 | ubd_devs[unit].pdev.dev.driver_data = &ubd_devs[unit]; | |
7d314e34 PBG |
677 | platform_device_register(&ubd_devs[unit].pdev); |
678 | disk->driverfs_dev = &ubd_devs[unit].pdev.dev; | |
1da177e4 LT |
679 | } |
680 | ||
7d314e34 | 681 | disk->private_data = &ubd_devs[unit]; |
62f96cb0 | 682 | disk->queue = ubd_devs[unit].queue; |
1da177e4 LT |
683 | add_disk(disk); |
684 | ||
685 | *disk_out = disk; | |
686 | return 0; | |
687 | } | |
688 | ||
689 | #define ROUND_BLOCK(n) ((n + ((1 << 9) - 1)) & (-1 << 9)) | |
690 | ||
f28169d2 | 691 | static int ubd_add(int n, char **error_out) |
1da177e4 | 692 | { |
7d314e34 | 693 | struct ubd *ubd_dev = &ubd_devs[n]; |
f28169d2 | 694 | int err = 0; |
1da177e4 | 695 | |
7d314e34 | 696 | if(ubd_dev->file == NULL) |
ec7cf783 | 697 | goto out; |
1da177e4 | 698 | |
7d314e34 | 699 | err = ubd_file_size(ubd_dev, &ubd_dev->size); |
f28169d2 JD |
700 | if(err < 0){ |
701 | *error_out = "Couldn't determine size of device's file"; | |
80c13749 | 702 | goto out; |
f28169d2 | 703 | } |
1da177e4 | 704 | |
7d314e34 | 705 | ubd_dev->size = ROUND_BLOCK(ubd_dev->size); |
1da177e4 | 706 | |
a0044bdf | 707 | INIT_LIST_HEAD(&ubd_dev->restart); |
4f40c055 | 708 | sg_init_table(ubd_dev->sg, MAX_SG); |
a0044bdf | 709 | |
62f96cb0 JD |
710 | err = -ENOMEM; |
711 | ubd_dev->queue = blk_init_queue(do_ubd_request, &ubd_dev->lock); | |
712 | if (ubd_dev->queue == NULL) { | |
713 | *error_out = "Failed to initialize device queue"; | |
80c13749 | 714 | goto out; |
62f96cb0 JD |
715 | } |
716 | ubd_dev->queue->queuedata = ubd_dev; | |
717 | ||
a0044bdf | 718 | blk_queue_max_hw_segments(ubd_dev->queue, MAX_SG); |
62f96cb0 JD |
719 | err = ubd_disk_register(MAJOR_NR, ubd_dev->size, n, &ubd_gendisk[n]); |
720 | if(err){ | |
721 | *error_out = "Failed to register device"; | |
722 | goto out_cleanup; | |
723 | } | |
6c29256c | 724 | |
1da177e4 | 725 | if(fake_major != MAJOR_NR) |
5f75a4f8 | 726 | ubd_disk_register(fake_major, ubd_dev->size, n, |
62f96cb0 | 727 | &fake_gendisk[n]); |
1da177e4 LT |
728 | |
729 | /* perhaps this should also be under the "if (fake_major)" above */ | |
730 | /* using the fake_disk->disk_name and also the fakehd_set name */ | |
731 | if (fake_ide) | |
732 | make_ide_entries(ubd_gendisk[n]->disk_name); | |
733 | ||
ec7cf783 | 734 | err = 0; |
ec7cf783 JD |
735 | out: |
736 | return err; | |
62f96cb0 JD |
737 | |
738 | out_cleanup: | |
739 | blk_cleanup_queue(ubd_dev->queue); | |
740 | goto out; | |
1da177e4 LT |
741 | } |
742 | ||
f28169d2 | 743 | static int ubd_config(char *str, char **error_out) |
1da177e4 | 744 | { |
e7f6552f | 745 | int n, ret; |
1da177e4 | 746 | |
f28169d2 JD |
747 | /* This string is possibly broken up and stored, so it's only |
748 | * freed if ubd_setup_common fails, or if only general options | |
749 | * were set. | |
750 | */ | |
970d6e3a | 751 | str = kstrdup(str, GFP_KERNEL); |
e7f6552f | 752 | if (str == NULL) { |
f28169d2 JD |
753 | *error_out = "Failed to allocate memory"; |
754 | return -ENOMEM; | |
1da177e4 | 755 | } |
f28169d2 JD |
756 | |
757 | ret = ubd_setup_common(str, &n, error_out); | |
758 | if (ret) | |
e7f6552f | 759 | goto err_free; |
f28169d2 | 760 | |
e7f6552f PBG |
761 | if (n == -1) { |
762 | ret = 0; | |
d8d7c28e | 763 | goto err_free; |
1da177e4 | 764 | } |
1da177e4 | 765 | |
dc764e50 | 766 | mutex_lock(&ubd_lock); |
f28169d2 | 767 | ret = ubd_add(n, error_out); |
e7f6552f | 768 | if (ret) |
7d314e34 | 769 | ubd_devs[n].file = NULL; |
dc764e50 | 770 | mutex_unlock(&ubd_lock); |
1da177e4 | 771 | |
e7f6552f | 772 | out: |
dc764e50 | 773 | return ret; |
e7f6552f PBG |
774 | |
775 | err_free: | |
776 | kfree(str); | |
777 | goto out; | |
1da177e4 LT |
778 | } |
779 | ||
780 | static int ubd_get_config(char *name, char *str, int size, char **error_out) | |
781 | { | |
7d314e34 | 782 | struct ubd *ubd_dev; |
1da177e4 LT |
783 | int n, len = 0; |
784 | ||
785 | n = parse_unit(&name); | |
786 | if((n >= MAX_DEV) || (n < 0)){ | |
787 | *error_out = "ubd_get_config : device number out of range"; | |
dc764e50 | 788 | return -1; |
1da177e4 LT |
789 | } |
790 | ||
7d314e34 | 791 | ubd_dev = &ubd_devs[n]; |
d7fb2c38 | 792 | mutex_lock(&ubd_lock); |
1da177e4 | 793 | |
7d314e34 | 794 | if(ubd_dev->file == NULL){ |
1da177e4 LT |
795 | CONFIG_CHUNK(str, size, len, "", 1); |
796 | goto out; | |
797 | } | |
798 | ||
7d314e34 | 799 | CONFIG_CHUNK(str, size, len, ubd_dev->file, 0); |
1da177e4 | 800 | |
7d314e34 | 801 | if(ubd_dev->cow.file != NULL){ |
1da177e4 | 802 | CONFIG_CHUNK(str, size, len, ",", 0); |
7d314e34 | 803 | CONFIG_CHUNK(str, size, len, ubd_dev->cow.file, 1); |
1da177e4 LT |
804 | } |
805 | else CONFIG_CHUNK(str, size, len, "", 1); | |
806 | ||
807 | out: | |
d7fb2c38 | 808 | mutex_unlock(&ubd_lock); |
dc764e50 | 809 | return len; |
1da177e4 LT |
810 | } |
811 | ||
29d56cfe JD |
812 | static int ubd_id(char **str, int *start_out, int *end_out) |
813 | { | |
dc764e50 | 814 | int n; |
29d56cfe JD |
815 | |
816 | n = parse_unit(str); | |
dc764e50 JD |
817 | *start_out = 0; |
818 | *end_out = MAX_DEV - 1; | |
819 | return n; | |
29d56cfe JD |
820 | } |
821 | ||
f28169d2 | 822 | static int ubd_remove(int n, char **error_out) |
1da177e4 | 823 | { |
2e3f5251 | 824 | struct gendisk *disk = ubd_gendisk[n]; |
7d314e34 | 825 | struct ubd *ubd_dev; |
29d56cfe | 826 | int err = -ENODEV; |
1da177e4 | 827 | |
d7fb2c38 | 828 | mutex_lock(&ubd_lock); |
1da177e4 | 829 | |
7d314e34 | 830 | ubd_dev = &ubd_devs[n]; |
1da177e4 | 831 | |
7d314e34 | 832 | if(ubd_dev->file == NULL) |
29d56cfe | 833 | goto out; |
1da177e4 | 834 | |
29d56cfe JD |
835 | /* you cannot remove a open disk */ |
836 | err = -EBUSY; | |
7d314e34 | 837 | if(ubd_dev->count > 0) |
1da177e4 LT |
838 | goto out; |
839 | ||
dc764e50 | 840 | ubd_gendisk[n] = NULL; |
b47d2deb JD |
841 | if(disk != NULL){ |
842 | del_gendisk(disk); | |
843 | put_disk(disk); | |
844 | } | |
1da177e4 LT |
845 | |
846 | if(fake_gendisk[n] != NULL){ | |
847 | del_gendisk(fake_gendisk[n]); | |
848 | put_disk(fake_gendisk[n]); | |
849 | fake_gendisk[n] = NULL; | |
850 | } | |
851 | ||
1da177e4 | 852 | err = 0; |
2e3f5251 | 853 | platform_device_unregister(&ubd_dev->pdev); |
29d56cfe | 854 | out: |
d7fb2c38 | 855 | mutex_unlock(&ubd_lock); |
29d56cfe | 856 | return err; |
1da177e4 LT |
857 | } |
858 | ||
f28169d2 | 859 | /* All these are called by mconsole in process context and without |
b8831a1d | 860 | * ubd-specific locks. The structure itself is const except for .list. |
f28169d2 | 861 | */ |
1da177e4 | 862 | static struct mc_device ubd_mc = { |
84f48d4f | 863 | .list = LIST_HEAD_INIT(ubd_mc.list), |
1da177e4 LT |
864 | .name = "ubd", |
865 | .config = ubd_config, | |
dc764e50 | 866 | .get_config = ubd_get_config, |
29d56cfe | 867 | .id = ubd_id, |
1da177e4 LT |
868 | .remove = ubd_remove, |
869 | }; | |
870 | ||
d8d7c28e | 871 | static int __init ubd_mc_init(void) |
1da177e4 LT |
872 | { |
873 | mconsole_register_dev(&ubd_mc); | |
874 | return 0; | |
875 | } | |
876 | ||
877 | __initcall(ubd_mc_init); | |
878 | ||
d8d7c28e PBG |
879 | static int __init ubd0_init(void) |
880 | { | |
881 | struct ubd *ubd_dev = &ubd_devs[0]; | |
882 | ||
b8831a1d | 883 | mutex_lock(&ubd_lock); |
d8d7c28e PBG |
884 | if(ubd_dev->file == NULL) |
885 | ubd_dev->file = "root_fs"; | |
b8831a1d JD |
886 | mutex_unlock(&ubd_lock); |
887 | ||
dc764e50 | 888 | return 0; |
d8d7c28e PBG |
889 | } |
890 | ||
891 | __initcall(ubd0_init); | |
892 | ||
b8831a1d | 893 | /* Used in ubd_init, which is an initcall */ |
3ae5eaec RK |
894 | static struct platform_driver ubd_driver = { |
895 | .driver = { | |
896 | .name = DRIVER_NAME, | |
897 | }, | |
1da177e4 LT |
898 | }; |
899 | ||
d8d7c28e | 900 | static int __init ubd_init(void) |
1da177e4 | 901 | { |
f28169d2 JD |
902 | char *error; |
903 | int i, err; | |
1da177e4 | 904 | |
1da177e4 LT |
905 | if (register_blkdev(MAJOR_NR, "ubd")) |
906 | return -1; | |
907 | ||
1da177e4 LT |
908 | if (fake_major != MAJOR_NR) { |
909 | char name[sizeof("ubd_nnn\0")]; | |
910 | ||
911 | snprintf(name, sizeof(name), "ubd_%d", fake_major); | |
1da177e4 LT |
912 | if (register_blkdev(fake_major, "ubd")) |
913 | return -1; | |
914 | } | |
3ae5eaec | 915 | platform_driver_register(&ubd_driver); |
dc764e50 | 916 | mutex_lock(&ubd_lock); |
f28169d2 JD |
917 | for (i = 0; i < MAX_DEV; i++){ |
918 | err = ubd_add(i, &error); | |
919 | if(err) | |
920 | printk(KERN_ERR "Failed to initialize ubd device %d :" | |
921 | "%s\n", i, error); | |
922 | } | |
dc764e50 | 923 | mutex_unlock(&ubd_lock); |
1da177e4 LT |
924 | return 0; |
925 | } | |
926 | ||
927 | late_initcall(ubd_init); | |
928 | ||
d8d7c28e | 929 | static int __init ubd_driver_init(void){ |
91acb21f JD |
930 | unsigned long stack; |
931 | int err; | |
932 | ||
933 | /* Set by CONFIG_BLK_DEV_UBD_SYNC or ubd=sync.*/ | |
934 | if(global_openflags.s){ | |
935 | printk(KERN_INFO "ubd: Synchronous mode\n"); | |
936 | /* Letting ubd=sync be like using ubd#s= instead of ubd#= is | |
937 | * enough. So use anyway the io thread. */ | |
938 | } | |
939 | stack = alloc_stack(0, 0); | |
6c29256c | 940 | io_pid = start_io_thread(stack + PAGE_SIZE - sizeof(void *), |
91acb21f JD |
941 | &thread_fd); |
942 | if(io_pid < 0){ | |
6c29256c | 943 | printk(KERN_ERR |
91acb21f JD |
944 | "ubd : Failed to start I/O thread (errno = %d) - " |
945 | "falling back to synchronous I/O\n", -io_pid); | |
946 | io_pid = -1; | |
dc764e50 | 947 | return 0; |
91acb21f | 948 | } |
6c29256c | 949 | err = um_request_irq(UBD_IRQ, thread_fd, IRQ_READ, ubd_intr, |
7d314e34 | 950 | IRQF_DISABLED, "ubd", ubd_devs); |
91acb21f JD |
951 | if(err != 0) |
952 | printk(KERN_ERR "um_request_irq failed - errno = %d\n", -err); | |
f4c57a78 | 953 | return 0; |
91acb21f JD |
954 | } |
955 | ||
956 | device_initcall(ubd_driver_init); | |
957 | ||
1da177e4 LT |
958 | static int ubd_open(struct inode *inode, struct file *filp) |
959 | { | |
960 | struct gendisk *disk = inode->i_bdev->bd_disk; | |
7d314e34 | 961 | struct ubd *ubd_dev = disk->private_data; |
1da177e4 LT |
962 | int err = 0; |
963 | ||
7d314e34 PBG |
964 | if(ubd_dev->count == 0){ |
965 | err = ubd_open_dev(ubd_dev); | |
1da177e4 LT |
966 | if(err){ |
967 | printk(KERN_ERR "%s: Can't open \"%s\": errno = %d\n", | |
7d314e34 | 968 | disk->disk_name, ubd_dev->file, -err); |
1da177e4 LT |
969 | goto out; |
970 | } | |
971 | } | |
7d314e34 PBG |
972 | ubd_dev->count++; |
973 | set_disk_ro(disk, !ubd_dev->openflags.w); | |
2c49be99 PBG |
974 | |
975 | /* This should no more be needed. And it didn't work anyway to exclude | |
976 | * read-write remounting of filesystems.*/ | |
7d314e34 | 977 | /*if((filp->f_mode & FMODE_WRITE) && !ubd_dev->openflags.w){ |
5f75a4f8 | 978 | if(--ubd_dev->count == 0) ubd_close_dev(ubd_dev); |
1da177e4 | 979 | err = -EROFS; |
2c49be99 | 980 | }*/ |
1da177e4 | 981 | out: |
dc764e50 | 982 | return err; |
1da177e4 LT |
983 | } |
984 | ||
985 | static int ubd_release(struct inode * inode, struct file * file) | |
986 | { | |
987 | struct gendisk *disk = inode->i_bdev->bd_disk; | |
7d314e34 | 988 | struct ubd *ubd_dev = disk->private_data; |
1da177e4 | 989 | |
7d314e34 | 990 | if(--ubd_dev->count == 0) |
5f75a4f8 | 991 | ubd_close_dev(ubd_dev); |
dc764e50 | 992 | return 0; |
1da177e4 LT |
993 | } |
994 | ||
91acb21f JD |
995 | static void cowify_bitmap(__u64 io_offset, int length, unsigned long *cow_mask, |
996 | __u64 *cow_offset, unsigned long *bitmap, | |
997 | __u64 bitmap_offset, unsigned long *bitmap_words, | |
998 | __u64 bitmap_len) | |
1da177e4 | 999 | { |
91acb21f JD |
1000 | __u64 sector = io_offset >> 9; |
1001 | int i, update_bitmap = 0; | |
1002 | ||
1003 | for(i = 0; i < length >> 9; i++){ | |
1004 | if(cow_mask != NULL) | |
1005 | ubd_set_bit(i, (unsigned char *) cow_mask); | |
1006 | if(ubd_test_bit(sector + i, (unsigned char *) bitmap)) | |
1007 | continue; | |
1da177e4 | 1008 | |
91acb21f JD |
1009 | update_bitmap = 1; |
1010 | ubd_set_bit(sector + i, (unsigned char *) bitmap); | |
1011 | } | |
1012 | ||
1013 | if(!update_bitmap) | |
1014 | return; | |
1da177e4 | 1015 | |
91acb21f | 1016 | *cow_offset = sector / (sizeof(unsigned long) * 8); |
1da177e4 | 1017 | |
91acb21f JD |
1018 | /* This takes care of the case where we're exactly at the end of the |
1019 | * device, and *cow_offset + 1 is off the end. So, just back it up | |
1020 | * by one word. Thanks to Lynn Kerby for the fix and James McMechan | |
1021 | * for the original diagnosis. | |
1022 | */ | |
1023 | if(*cow_offset == ((bitmap_len + sizeof(unsigned long) - 1) / | |
1024 | sizeof(unsigned long) - 1)) | |
1025 | (*cow_offset)--; | |
1026 | ||
1027 | bitmap_words[0] = bitmap[*cow_offset]; | |
1028 | bitmap_words[1] = bitmap[*cow_offset + 1]; | |
1029 | ||
1030 | *cow_offset *= sizeof(unsigned long); | |
1031 | *cow_offset += bitmap_offset; | |
1032 | } | |
1033 | ||
1034 | static void cowify_req(struct io_thread_req *req, unsigned long *bitmap, | |
1035 | __u64 bitmap_offset, __u64 bitmap_len) | |
1036 | { | |
1037 | __u64 sector = req->offset >> 9; | |
1038 | int i; | |
1039 | ||
1040 | if(req->length > (sizeof(req->sector_mask) * 8) << 9) | |
1041 | panic("Operation too long"); | |
1042 | ||
1043 | if(req->op == UBD_READ) { | |
1044 | for(i = 0; i < req->length >> 9; i++){ | |
1045 | if(ubd_test_bit(sector + i, (unsigned char *) bitmap)) | |
6c29256c | 1046 | ubd_set_bit(i, (unsigned char *) |
91acb21f | 1047 | &req->sector_mask); |
dc764e50 | 1048 | } |
91acb21f JD |
1049 | } |
1050 | else cowify_bitmap(req->offset, req->length, &req->sector_mask, | |
1051 | &req->cow_offset, bitmap, bitmap_offset, | |
1052 | req->bitmap_words, bitmap_len); | |
1da177e4 LT |
1053 | } |
1054 | ||
62f96cb0 | 1055 | /* Called with dev->lock held */ |
a0044bdf JD |
1056 | static void prepare_request(struct request *req, struct io_thread_req *io_req, |
1057 | unsigned long long offset, int page_offset, | |
1058 | int len, struct page *page) | |
1da177e4 LT |
1059 | { |
1060 | struct gendisk *disk = req->rq_disk; | |
7d314e34 | 1061 | struct ubd *ubd_dev = disk->private_data; |
91acb21f | 1062 | |
62f96cb0 | 1063 | io_req->req = req; |
a0044bdf JD |
1064 | io_req->fds[0] = (ubd_dev->cow.file != NULL) ? ubd_dev->cow.fd : |
1065 | ubd_dev->fd; | |
7d314e34 | 1066 | io_req->fds[1] = ubd_dev->fd; |
91acb21f | 1067 | io_req->cow_offset = -1; |
1da177e4 LT |
1068 | io_req->offset = offset; |
1069 | io_req->length = len; | |
1070 | io_req->error = 0; | |
91acb21f JD |
1071 | io_req->sector_mask = 0; |
1072 | ||
1073 | io_req->op = (rq_data_dir(req) == READ) ? UBD_READ : UBD_WRITE; | |
1da177e4 | 1074 | io_req->offsets[0] = 0; |
7d314e34 | 1075 | io_req->offsets[1] = ubd_dev->cow.data_offset; |
a0044bdf | 1076 | io_req->buffer = page_address(page) + page_offset; |
1da177e4 LT |
1077 | io_req->sectorsize = 1 << 9; |
1078 | ||
7d314e34 | 1079 | if(ubd_dev->cow.file != NULL) |
a0044bdf JD |
1080 | cowify_req(io_req, ubd_dev->cow.bitmap, |
1081 | ubd_dev->cow.bitmap_offset, ubd_dev->cow.bitmap_len); | |
91acb21f | 1082 | |
1da177e4 LT |
1083 | } |
1084 | ||
62f96cb0 | 1085 | /* Called with dev->lock held */ |
165125e1 | 1086 | static void do_ubd_request(struct request_queue *q) |
1da177e4 | 1087 | { |
2adcec21 | 1088 | struct io_thread_req *io_req; |
1da177e4 | 1089 | struct request *req; |
0a6d3a2a | 1090 | int n, last_sectors; |
a0044bdf JD |
1091 | |
1092 | while(1){ | |
2a9529a0 | 1093 | struct ubd *dev = q->queuedata; |
a0044bdf JD |
1094 | if(dev->end_sg == 0){ |
1095 | struct request *req = elv_next_request(q); | |
1096 | if(req == NULL) | |
1097 | return; | |
1098 | ||
1099 | dev->request = req; | |
1100 | blkdev_dequeue_request(req); | |
1101 | dev->start_sg = 0; | |
1102 | dev->end_sg = blk_rq_map_sg(q, req, dev->sg); | |
1103 | } | |
1104 | ||
1105 | req = dev->request; | |
0a6d3a2a | 1106 | last_sectors = 0; |
a0044bdf JD |
1107 | while(dev->start_sg < dev->end_sg){ |
1108 | struct scatterlist *sg = &dev->sg[dev->start_sg]; | |
1109 | ||
0a6d3a2a | 1110 | req->sector += last_sectors; |
2adcec21 | 1111 | io_req = kmalloc(sizeof(struct io_thread_req), |
990c5587 | 1112 | GFP_ATOMIC); |
2adcec21 JD |
1113 | if(io_req == NULL){ |
1114 | if(list_empty(&dev->restart)) | |
1115 | list_add(&dev->restart, &restart); | |
1116 | return; | |
1117 | } | |
1118 | prepare_request(req, io_req, | |
a0044bdf | 1119 | (unsigned long long) req->sector << 9, |
45711f1a | 1120 | sg->offset, sg->length, sg_page(sg)); |
a0044bdf | 1121 | |
0a6d3a2a | 1122 | last_sectors = sg->length >> 9; |
a6ea4cce JD |
1123 | n = os_write_file(thread_fd, &io_req, |
1124 | sizeof(struct io_thread_req *)); | |
2adcec21 | 1125 | if(n != sizeof(struct io_thread_req *)){ |
a0044bdf JD |
1126 | if(n != -EAGAIN) |
1127 | printk("write to io thread failed, " | |
1128 | "errno = %d\n", -n); | |
1129 | else if(list_empty(&dev->restart)) | |
1130 | list_add(&dev->restart, &restart); | |
1131 | return; | |
1132 | } | |
1133 | ||
a0044bdf | 1134 | dev->start_sg++; |
1da177e4 | 1135 | } |
a0044bdf JD |
1136 | dev->end_sg = 0; |
1137 | dev->request = NULL; | |
1da177e4 LT |
1138 | } |
1139 | } | |
1140 | ||
a885c8c4 CH |
1141 | static int ubd_getgeo(struct block_device *bdev, struct hd_geometry *geo) |
1142 | { | |
7d314e34 | 1143 | struct ubd *ubd_dev = bdev->bd_disk->private_data; |
a885c8c4 CH |
1144 | |
1145 | geo->heads = 128; | |
1146 | geo->sectors = 32; | |
7d314e34 | 1147 | geo->cylinders = ubd_dev->size / (128 * 32 * 512); |
a885c8c4 CH |
1148 | return 0; |
1149 | } | |
1150 | ||
1da177e4 LT |
1151 | static int ubd_ioctl(struct inode * inode, struct file * file, |
1152 | unsigned int cmd, unsigned long arg) | |
1153 | { | |
7d314e34 | 1154 | struct ubd *ubd_dev = inode->i_bdev->bd_disk->private_data; |
1da177e4 LT |
1155 | struct hd_driveid ubd_id = { |
1156 | .cyls = 0, | |
1157 | .heads = 128, | |
1158 | .sectors = 32, | |
1159 | }; | |
1160 | ||
1161 | switch (cmd) { | |
1da177e4 | 1162 | struct cdrom_volctrl volume; |
1da177e4 | 1163 | case HDIO_GET_IDENTITY: |
7d314e34 | 1164 | ubd_id.cyls = ubd_dev->size / (128 * 32 * 512); |
1da177e4 LT |
1165 | if(copy_to_user((char __user *) arg, (char *) &ubd_id, |
1166 | sizeof(ubd_id))) | |
dc764e50 JD |
1167 | return -EFAULT; |
1168 | return 0; | |
b8831a1d | 1169 | |
1da177e4 LT |
1170 | case CDROMVOLREAD: |
1171 | if(copy_from_user(&volume, (char __user *) arg, sizeof(volume))) | |
dc764e50 | 1172 | return -EFAULT; |
1da177e4 LT |
1173 | volume.channel0 = 255; |
1174 | volume.channel1 = 255; | |
1175 | volume.channel2 = 255; | |
1176 | volume.channel3 = 255; | |
1177 | if(copy_to_user((char __user *) arg, &volume, sizeof(volume))) | |
dc764e50 JD |
1178 | return -EFAULT; |
1179 | return 0; | |
1da177e4 | 1180 | } |
dc764e50 | 1181 | return -EINVAL; |
1da177e4 LT |
1182 | } |
1183 | ||
4833aff7 | 1184 | static int path_requires_switch(char *from_cmdline, char *from_cow, char *cow) |
1da177e4 LT |
1185 | { |
1186 | struct uml_stat buf1, buf2; | |
1187 | int err; | |
1188 | ||
4833aff7 PBG |
1189 | if(from_cmdline == NULL) |
1190 | return 0; | |
1191 | if(!strcmp(from_cmdline, from_cow)) | |
1192 | return 0; | |
1da177e4 LT |
1193 | |
1194 | err = os_stat_file(from_cmdline, &buf1); | |
1195 | if(err < 0){ | |
1196 | printk("Couldn't stat '%s', err = %d\n", from_cmdline, -err); | |
4833aff7 | 1197 | return 0; |
1da177e4 LT |
1198 | } |
1199 | err = os_stat_file(from_cow, &buf2); | |
1200 | if(err < 0){ | |
1201 | printk("Couldn't stat '%s', err = %d\n", from_cow, -err); | |
4833aff7 | 1202 | return 1; |
1da177e4 LT |
1203 | } |
1204 | if((buf1.ust_dev == buf2.ust_dev) && (buf1.ust_ino == buf2.ust_ino)) | |
4833aff7 | 1205 | return 0; |
1da177e4 LT |
1206 | |
1207 | printk("Backing file mismatch - \"%s\" requested,\n" | |
1208 | "\"%s\" specified in COW header of \"%s\"\n", | |
1209 | from_cmdline, from_cow, cow); | |
4833aff7 | 1210 | return 1; |
1da177e4 LT |
1211 | } |
1212 | ||
1213 | static int backing_file_mismatch(char *file, __u64 size, time_t mtime) | |
1214 | { | |
1215 | unsigned long modtime; | |
fe1db50c | 1216 | unsigned long long actual; |
1da177e4 LT |
1217 | int err; |
1218 | ||
1219 | err = os_file_modtime(file, &modtime); | |
1220 | if(err < 0){ | |
1221 | printk("Failed to get modification time of backing file " | |
1222 | "\"%s\", err = %d\n", file, -err); | |
dc764e50 | 1223 | return err; |
1da177e4 LT |
1224 | } |
1225 | ||
1226 | err = os_file_size(file, &actual); | |
1227 | if(err < 0){ | |
1228 | printk("Failed to get size of backing file \"%s\", " | |
1229 | "err = %d\n", file, -err); | |
dc764e50 | 1230 | return err; |
1da177e4 LT |
1231 | } |
1232 | ||
dc764e50 | 1233 | if(actual != size){ |
1da177e4 LT |
1234 | /*__u64 can be a long on AMD64 and with %lu GCC complains; so |
1235 | * the typecast.*/ | |
1236 | printk("Size mismatch (%llu vs %llu) of COW header vs backing " | |
1237 | "file\n", (unsigned long long) size, actual); | |
dc764e50 | 1238 | return -EINVAL; |
1da177e4 LT |
1239 | } |
1240 | if(modtime != mtime){ | |
1241 | printk("mtime mismatch (%ld vs %ld) of COW header vs backing " | |
1242 | "file\n", mtime, modtime); | |
dc764e50 | 1243 | return -EINVAL; |
1da177e4 | 1244 | } |
dc764e50 | 1245 | return 0; |
1da177e4 LT |
1246 | } |
1247 | ||
1248 | int read_cow_bitmap(int fd, void *buf, int offset, int len) | |
1249 | { | |
1250 | int err; | |
1251 | ||
1252 | err = os_seek_file(fd, offset); | |
1253 | if(err < 0) | |
dc764e50 | 1254 | return err; |
1da177e4 | 1255 | |
dc764e50 | 1256 | err = os_read_file(fd, buf, len); |
1da177e4 | 1257 | if(err < 0) |
dc764e50 | 1258 | return err; |
1da177e4 | 1259 | |
dc764e50 | 1260 | return 0; |
1da177e4 LT |
1261 | } |
1262 | ||
6c29256c | 1263 | int open_ubd_file(char *file, struct openflags *openflags, int shared, |
1da177e4 LT |
1264 | char **backing_file_out, int *bitmap_offset_out, |
1265 | unsigned long *bitmap_len_out, int *data_offset_out, | |
1266 | int *create_cow_out) | |
1267 | { | |
1268 | time_t mtime; | |
1269 | unsigned long long size; | |
1270 | __u32 version, align; | |
1271 | char *backing_file; | |
4833aff7 | 1272 | int fd, err, sectorsize, asked_switch, mode = 0644; |
1da177e4 LT |
1273 | |
1274 | fd = os_open_file(file, *openflags, mode); | |
a374a48f PBG |
1275 | if (fd < 0) { |
1276 | if ((fd == -ENOENT) && (create_cow_out != NULL)) | |
1da177e4 | 1277 | *create_cow_out = 1; |
dc764e50 JD |
1278 | if (!openflags->w || |
1279 | ((fd != -EROFS) && (fd != -EACCES))) | |
a374a48f | 1280 | return fd; |
1da177e4 LT |
1281 | openflags->w = 0; |
1282 | fd = os_open_file(file, *openflags, mode); | |
a374a48f PBG |
1283 | if (fd < 0) |
1284 | return fd; | |
dc764e50 | 1285 | } |
1da177e4 | 1286 | |
6c29256c JD |
1287 | if(shared) |
1288 | printk("Not locking \"%s\" on the host\n", file); | |
1289 | else { | |
1290 | err = os_lock_file(fd, openflags->w); | |
1291 | if(err < 0){ | |
1292 | printk("Failed to lock '%s', err = %d\n", file, -err); | |
1293 | goto out_close; | |
1294 | } | |
1da177e4 LT |
1295 | } |
1296 | ||
d6e05edc | 1297 | /* Successful return case! */ |
a374a48f | 1298 | if(backing_file_out == NULL) |
dc764e50 | 1299 | return fd; |
1da177e4 LT |
1300 | |
1301 | err = read_cow_header(file_reader, &fd, &version, &backing_file, &mtime, | |
1302 | &size, §orsize, &align, bitmap_offset_out); | |
1303 | if(err && (*backing_file_out != NULL)){ | |
1304 | printk("Failed to read COW header from COW file \"%s\", " | |
1305 | "errno = %d\n", file, -err); | |
1306 | goto out_close; | |
1307 | } | |
a374a48f | 1308 | if(err) |
dc764e50 | 1309 | return fd; |
1da177e4 | 1310 | |
4833aff7 | 1311 | asked_switch = path_requires_switch(*backing_file_out, backing_file, file); |
1da177e4 | 1312 | |
4833aff7 PBG |
1313 | /* Allow switching only if no mismatch. */ |
1314 | if (asked_switch && !backing_file_mismatch(*backing_file_out, size, mtime)) { | |
1da177e4 LT |
1315 | printk("Switching backing file to '%s'\n", *backing_file_out); |
1316 | err = write_cow_header(file, fd, *backing_file_out, | |
1317 | sectorsize, align, &size); | |
a374a48f | 1318 | if (err) { |
1da177e4 | 1319 | printk("Switch failed, errno = %d\n", -err); |
4833aff7 | 1320 | goto out_close; |
1da177e4 | 1321 | } |
a374a48f | 1322 | } else { |
1da177e4 LT |
1323 | *backing_file_out = backing_file; |
1324 | err = backing_file_mismatch(*backing_file_out, size, mtime); | |
a374a48f PBG |
1325 | if (err) |
1326 | goto out_close; | |
1da177e4 LT |
1327 | } |
1328 | ||
1329 | cow_sizes(version, size, sectorsize, align, *bitmap_offset_out, | |
1330 | bitmap_len_out, data_offset_out); | |
1331 | ||
dc764e50 | 1332 | return fd; |
1da177e4 LT |
1333 | out_close: |
1334 | os_close_file(fd); | |
a374a48f | 1335 | return err; |
1da177e4 LT |
1336 | } |
1337 | ||
1338 | int create_cow_file(char *cow_file, char *backing_file, struct openflags flags, | |
1339 | int sectorsize, int alignment, int *bitmap_offset_out, | |
1340 | unsigned long *bitmap_len_out, int *data_offset_out) | |
1341 | { | |
1342 | int err, fd; | |
1343 | ||
1344 | flags.c = 1; | |
6c29256c | 1345 | fd = open_ubd_file(cow_file, &flags, 0, NULL, NULL, NULL, NULL, NULL); |
1da177e4 LT |
1346 | if(fd < 0){ |
1347 | err = fd; | |
1348 | printk("Open of COW file '%s' failed, errno = %d\n", cow_file, | |
1349 | -err); | |
1350 | goto out; | |
1351 | } | |
1352 | ||
1353 | err = init_cow_file(fd, cow_file, backing_file, sectorsize, alignment, | |
1354 | bitmap_offset_out, bitmap_len_out, | |
1355 | data_offset_out); | |
1356 | if(!err) | |
dc764e50 | 1357 | return fd; |
1da177e4 LT |
1358 | os_close_file(fd); |
1359 | out: | |
dc764e50 | 1360 | return err; |
1da177e4 LT |
1361 | } |
1362 | ||
91acb21f | 1363 | static int update_bitmap(struct io_thread_req *req) |
1da177e4 | 1364 | { |
91acb21f | 1365 | int n; |
1da177e4 | 1366 | |
91acb21f | 1367 | if(req->cow_offset == -1) |
dc764e50 | 1368 | return 0; |
1da177e4 | 1369 | |
91acb21f JD |
1370 | n = os_seek_file(req->fds[1], req->cow_offset); |
1371 | if(n < 0){ | |
1372 | printk("do_io - bitmap lseek failed : err = %d\n", -n); | |
dc764e50 | 1373 | return 1; |
91acb21f | 1374 | } |
1da177e4 | 1375 | |
a6ea4cce JD |
1376 | n = os_write_file(req->fds[1], &req->bitmap_words, |
1377 | sizeof(req->bitmap_words)); | |
91acb21f JD |
1378 | if(n != sizeof(req->bitmap_words)){ |
1379 | printk("do_io - bitmap update failed, err = %d fd = %d\n", -n, | |
1380 | req->fds[1]); | |
dc764e50 | 1381 | return 1; |
91acb21f | 1382 | } |
1da177e4 | 1383 | |
dc764e50 | 1384 | return 0; |
91acb21f | 1385 | } |
1da177e4 | 1386 | |
91acb21f JD |
1387 | void do_io(struct io_thread_req *req) |
1388 | { | |
1389 | char *buf; | |
1390 | unsigned long len; | |
1391 | int n, nsectors, start, end, bit; | |
1392 | int err; | |
1393 | __u64 off; | |
1394 | ||
1395 | nsectors = req->length / req->sectorsize; | |
1396 | start = 0; | |
1397 | do { | |
1398 | bit = ubd_test_bit(start, (unsigned char *) &req->sector_mask); | |
1399 | end = start; | |
1400 | while((end < nsectors) && | |
1401 | (ubd_test_bit(end, (unsigned char *) | |
1402 | &req->sector_mask) == bit)) | |
1403 | end++; | |
1404 | ||
1405 | off = req->offset + req->offsets[bit] + | |
1406 | start * req->sectorsize; | |
1407 | len = (end - start) * req->sectorsize; | |
1408 | buf = &req->buffer[start * req->sectorsize]; | |
1409 | ||
1410 | err = os_seek_file(req->fds[bit], off); | |
1411 | if(err < 0){ | |
1412 | printk("do_io - lseek failed : err = %d\n", -err); | |
1413 | req->error = 1; | |
1414 | return; | |
1415 | } | |
1416 | if(req->op == UBD_READ){ | |
1417 | n = 0; | |
1418 | do { | |
1419 | buf = &buf[n]; | |
1420 | len -= n; | |
a6ea4cce | 1421 | n = os_read_file(req->fds[bit], buf, len); |
91acb21f JD |
1422 | if (n < 0) { |
1423 | printk("do_io - read failed, err = %d " | |
1424 | "fd = %d\n", -n, req->fds[bit]); | |
1425 | req->error = 1; | |
1426 | return; | |
1427 | } | |
1428 | } while((n < len) && (n != 0)); | |
1429 | if (n < len) memset(&buf[n], 0, len - n); | |
1430 | } else { | |
a6ea4cce | 1431 | n = os_write_file(req->fds[bit], buf, len); |
91acb21f JD |
1432 | if(n != len){ |
1433 | printk("do_io - write failed err = %d " | |
1434 | "fd = %d\n", -n, req->fds[bit]); | |
1435 | req->error = 1; | |
1436 | return; | |
1437 | } | |
1438 | } | |
1439 | ||
1440 | start = end; | |
1441 | } while(start < nsectors); | |
1da177e4 | 1442 | |
91acb21f | 1443 | req->error = update_bitmap(req); |
1da177e4 | 1444 | } |
91acb21f JD |
1445 | |
1446 | /* Changed in start_io_thread, which is serialized by being called only | |
1447 | * from ubd_init, which is an initcall. | |
1448 | */ | |
1449 | int kernel_fd = -1; | |
1450 | ||
d8d7c28e PBG |
1451 | /* Only changed by the io thread. XXX: currently unused. */ |
1452 | static int io_count = 0; | |
91acb21f JD |
1453 | |
1454 | int io_thread(void *arg) | |
1455 | { | |
2adcec21 | 1456 | struct io_thread_req *req; |
91acb21f JD |
1457 | int n; |
1458 | ||
1459 | ignore_sigwinch_sig(); | |
1460 | while(1){ | |
a6ea4cce | 1461 | n = os_read_file(kernel_fd, &req, |
2adcec21 JD |
1462 | sizeof(struct io_thread_req *)); |
1463 | if(n != sizeof(struct io_thread_req *)){ | |
91acb21f JD |
1464 | if(n < 0) |
1465 | printk("io_thread - read failed, fd = %d, " | |
1466 | "err = %d\n", kernel_fd, -n); | |
1467 | else { | |
1468 | printk("io_thread - short read, fd = %d, " | |
1469 | "length = %d\n", kernel_fd, n); | |
1470 | } | |
1471 | continue; | |
1472 | } | |
1473 | io_count++; | |
2adcec21 | 1474 | do_io(req); |
a6ea4cce | 1475 | n = os_write_file(kernel_fd, &req, |
2adcec21 JD |
1476 | sizeof(struct io_thread_req *)); |
1477 | if(n != sizeof(struct io_thread_req *)) | |
91acb21f JD |
1478 | printk("io_thread - write failed, fd = %d, err = %d\n", |
1479 | kernel_fd, -n); | |
1480 | } | |
91acb21f | 1481 | |
1b57e9c2 JD |
1482 | return 0; |
1483 | } |