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 | ||
c0a9290e | 232 | static void make_ide_entries(const char *dev_name) |
1da177e4 LT |
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; | |
c0a9290e | 247 | snprintf(name, sizeof(name), "ide0/%s", dev_name); |
1da177e4 LT |
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" | |
20ede453 JD |
440 | " an 's' will cause data to be written to disk on the host immediately.\n" |
441 | " 'c' will cause the device to be treated as being shared between multiple\n" | |
442 | " UMLs and file locking will be turned off - this is appropriate for a\n" | |
443 | " cluster filesystem and inappropriate at almost all other times.\n\n" | |
1da177e4 LT |
444 | ); |
445 | ||
c0a9290e | 446 | static int udb_setup(const char *str) |
1da177e4 LT |
447 | { |
448 | printk("udb%s specified on command line is almost certainly a ubd -> " | |
449 | "udb TYPO\n", str); | |
dc764e50 | 450 | return 1; |
1da177e4 LT |
451 | } |
452 | ||
453 | __setup("udb", udb_setup); | |
454 | __uml_help(udb_setup, | |
455 | "udb\n" | |
0894e27e JD |
456 | " This option is here solely to catch ubd -> udb typos, which can be\n" |
457 | " to impossible to catch visually unless you specifically look for\n" | |
458 | " them. The only result of any option starting with 'udb' is an error\n" | |
1da177e4 LT |
459 | " in the boot output.\n\n" |
460 | ); | |
461 | ||
462 | static int fakehd_set = 0; | |
463 | static int fakehd(char *str) | |
464 | { | |
465 | printk(KERN_INFO "fakehd : Changing ubd name to \"hd\".\n"); | |
466 | fakehd_set = 1; | |
467 | return 1; | |
468 | } | |
469 | ||
470 | __setup("fakehd", fakehd); | |
471 | __uml_help(fakehd, | |
472 | "fakehd\n" | |
473 | " Change the ubd device name to \"hd\".\n\n" | |
474 | ); | |
475 | ||
165125e1 | 476 | static void do_ubd_request(struct request_queue * q); |
91acb21f JD |
477 | |
478 | /* Only changed by ubd_init, which is an initcall. */ | |
479 | int thread_fd = -1; | |
1da177e4 | 480 | |
4898b53a | 481 | static void ubd_end_request(struct request *req, int bytes, int error) |
1da177e4 | 482 | { |
4898b53a | 483 | blk_end_request(req, error, bytes); |
1da177e4 LT |
484 | } |
485 | ||
33f775ee PBG |
486 | /* Callable only from interrupt context - otherwise you need to do |
487 | * spin_lock_irq()/spin_lock_irqsave() */ | |
a0044bdf | 488 | static inline void ubd_finish(struct request *req, int bytes) |
1da177e4 | 489 | { |
a0044bdf | 490 | if(bytes < 0){ |
4898b53a | 491 | ubd_end_request(req, 0, -EIO); |
a0044bdf JD |
492 | return; |
493 | } | |
4898b53a | 494 | ubd_end_request(req, bytes, 0); |
1da177e4 LT |
495 | } |
496 | ||
a0044bdf JD |
497 | static LIST_HEAD(restart); |
498 | ||
2fe30a34 | 499 | /* XXX - move this inside ubd_intr. */ |
62f96cb0 | 500 | /* Called without dev->lock held, and only in interrupt context. */ |
91acb21f | 501 | static void ubd_handler(void) |
1da177e4 | 502 | { |
2adcec21 | 503 | struct io_thread_req *req; |
62f96cb0 | 504 | struct request *rq; |
a0044bdf JD |
505 | struct ubd *ubd; |
506 | struct list_head *list, *next_ele; | |
507 | unsigned long flags; | |
91acb21f JD |
508 | int n; |
509 | ||
a0044bdf | 510 | while(1){ |
a6ea4cce JD |
511 | n = os_read_file(thread_fd, &req, |
512 | sizeof(struct io_thread_req *)); | |
a0044bdf JD |
513 | if(n != sizeof(req)){ |
514 | if(n == -EAGAIN) | |
515 | break; | |
516 | printk(KERN_ERR "spurious interrupt in ubd_handler, " | |
517 | "err = %d\n", -n); | |
518 | return; | |
519 | } | |
62f96cb0 | 520 | |
2adcec21 JD |
521 | rq = req->req; |
522 | rq->nr_sectors -= req->length >> 9; | |
a0044bdf JD |
523 | if(rq->nr_sectors == 0) |
524 | ubd_finish(rq, rq->hard_nr_sectors << 9); | |
2adcec21 | 525 | kfree(req); |
a0044bdf | 526 | } |
62f96cb0 | 527 | reactivate_fd(thread_fd, UBD_IRQ); |
a0044bdf JD |
528 | |
529 | list_for_each_safe(list, next_ele, &restart){ | |
530 | ubd = container_of(list, struct ubd, restart); | |
531 | list_del_init(&ubd->restart); | |
532 | spin_lock_irqsave(&ubd->lock, flags); | |
533 | do_ubd_request(ubd->queue); | |
534 | spin_unlock_irqrestore(&ubd->lock, flags); | |
535 | } | |
1da177e4 LT |
536 | } |
537 | ||
7bea96fd | 538 | static irqreturn_t ubd_intr(int irq, void *dev) |
1da177e4 | 539 | { |
91acb21f | 540 | ubd_handler(); |
dc764e50 | 541 | return IRQ_HANDLED; |
91acb21f | 542 | } |
09ace81c | 543 | |
91acb21f JD |
544 | /* Only changed by ubd_init, which is an initcall. */ |
545 | static int io_pid = -1; | |
09ace81c | 546 | |
91acb21f JD |
547 | void kill_io_thread(void) |
548 | { | |
6c29256c | 549 | if(io_pid != -1) |
91acb21f | 550 | os_kill_process(io_pid, 1); |
09ace81c | 551 | } |
1da177e4 | 552 | |
91acb21f JD |
553 | __uml_exitcall(kill_io_thread); |
554 | ||
d8d7c28e | 555 | static inline int ubd_file_size(struct ubd *ubd_dev, __u64 *size_out) |
1da177e4 LT |
556 | { |
557 | char *file; | |
558 | ||
7d314e34 | 559 | file = ubd_dev->cow.file ? ubd_dev->cow.file : ubd_dev->file; |
dc764e50 | 560 | return os_file_size(file, size_out); |
1da177e4 LT |
561 | } |
562 | ||
5f75a4f8 | 563 | static void ubd_close_dev(struct ubd *ubd_dev) |
1da177e4 | 564 | { |
7d314e34 PBG |
565 | os_close_file(ubd_dev->fd); |
566 | if(ubd_dev->cow.file == NULL) | |
1da177e4 LT |
567 | return; |
568 | ||
7d314e34 PBG |
569 | os_close_file(ubd_dev->cow.fd); |
570 | vfree(ubd_dev->cow.bitmap); | |
571 | ubd_dev->cow.bitmap = NULL; | |
1da177e4 LT |
572 | } |
573 | ||
7d314e34 | 574 | static int ubd_open_dev(struct ubd *ubd_dev) |
1da177e4 LT |
575 | { |
576 | struct openflags flags; | |
577 | char **back_ptr; | |
578 | int err, create_cow, *create_ptr; | |
0bf16bff | 579 | int fd; |
1da177e4 | 580 | |
7d314e34 | 581 | ubd_dev->openflags = ubd_dev->boot_openflags; |
1da177e4 | 582 | create_cow = 0; |
7d314e34 PBG |
583 | create_ptr = (ubd_dev->cow.file != NULL) ? &create_cow : NULL; |
584 | back_ptr = ubd_dev->no_cow ? NULL : &ubd_dev->cow.file; | |
0bf16bff PBG |
585 | |
586 | fd = open_ubd_file(ubd_dev->file, &ubd_dev->openflags, ubd_dev->shared, | |
7d314e34 PBG |
587 | back_ptr, &ubd_dev->cow.bitmap_offset, |
588 | &ubd_dev->cow.bitmap_len, &ubd_dev->cow.data_offset, | |
6c29256c | 589 | create_ptr); |
1da177e4 | 590 | |
0bf16bff PBG |
591 | if((fd == -ENOENT) && create_cow){ |
592 | fd = create_cow_file(ubd_dev->file, ubd_dev->cow.file, | |
7d314e34 PBG |
593 | ubd_dev->openflags, 1 << 9, PAGE_SIZE, |
594 | &ubd_dev->cow.bitmap_offset, | |
595 | &ubd_dev->cow.bitmap_len, | |
596 | &ubd_dev->cow.data_offset); | |
0bf16bff | 597 | if(fd >= 0){ |
1da177e4 | 598 | printk(KERN_INFO "Creating \"%s\" as COW file for " |
7d314e34 | 599 | "\"%s\"\n", ubd_dev->file, ubd_dev->cow.file); |
1da177e4 LT |
600 | } |
601 | } | |
602 | ||
0bf16bff | 603 | if(fd < 0){ |
7d314e34 | 604 | printk("Failed to open '%s', errno = %d\n", ubd_dev->file, |
0bf16bff PBG |
605 | -fd); |
606 | return fd; | |
1da177e4 | 607 | } |
0bf16bff | 608 | ubd_dev->fd = fd; |
1da177e4 | 609 | |
7d314e34 | 610 | if(ubd_dev->cow.file != NULL){ |
f4768ffd JD |
611 | blk_queue_max_sectors(ubd_dev->queue, 8 * sizeof(long)); |
612 | ||
1da177e4 | 613 | err = -ENOMEM; |
da2486ba | 614 | ubd_dev->cow.bitmap = vmalloc(ubd_dev->cow.bitmap_len); |
7d314e34 | 615 | if(ubd_dev->cow.bitmap == NULL){ |
1da177e4 LT |
616 | printk(KERN_ERR "Failed to vmalloc COW bitmap\n"); |
617 | goto error; | |
618 | } | |
619 | flush_tlb_kernel_vm(); | |
620 | ||
7d314e34 PBG |
621 | err = read_cow_bitmap(ubd_dev->fd, ubd_dev->cow.bitmap, |
622 | ubd_dev->cow.bitmap_offset, | |
623 | ubd_dev->cow.bitmap_len); | |
1da177e4 LT |
624 | if(err < 0) |
625 | goto error; | |
626 | ||
7d314e34 | 627 | flags = ubd_dev->openflags; |
1da177e4 | 628 | flags.w = 0; |
7d314e34 | 629 | err = open_ubd_file(ubd_dev->cow.file, &flags, ubd_dev->shared, NULL, |
6c29256c | 630 | NULL, NULL, NULL, NULL); |
1da177e4 | 631 | if(err < 0) goto error; |
7d314e34 | 632 | ubd_dev->cow.fd = err; |
1da177e4 | 633 | } |
dc764e50 | 634 | return 0; |
1da177e4 | 635 | error: |
7d314e34 | 636 | os_close_file(ubd_dev->fd); |
dc764e50 | 637 | return err; |
1da177e4 LT |
638 | } |
639 | ||
2e3f5251 JD |
640 | static void ubd_device_release(struct device *dev) |
641 | { | |
642 | struct ubd *ubd_dev = dev->driver_data; | |
643 | ||
644 | blk_cleanup_queue(ubd_dev->queue); | |
645 | *ubd_dev = ((struct ubd) DEFAULT_UBD); | |
646 | } | |
647 | ||
5f75a4f8 | 648 | static int ubd_disk_register(int major, u64 size, int unit, |
b8831a1d | 649 | struct gendisk **disk_out) |
1da177e4 LT |
650 | { |
651 | struct gendisk *disk; | |
1da177e4 LT |
652 | |
653 | disk = alloc_disk(1 << UBD_SHIFT); | |
654 | if(disk == NULL) | |
dc764e50 | 655 | return -ENOMEM; |
1da177e4 LT |
656 | |
657 | disk->major = major; | |
658 | disk->first_minor = unit << UBD_SHIFT; | |
659 | disk->fops = &ubd_blops; | |
660 | set_capacity(disk, size / 512); | |
ce7b0f46 | 661 | if(major == MAJOR_NR) |
1da177e4 | 662 | sprintf(disk->disk_name, "ubd%c", 'a' + unit); |
ce7b0f46 | 663 | else |
1da177e4 | 664 | sprintf(disk->disk_name, "ubd_fake%d", unit); |
1da177e4 LT |
665 | |
666 | /* sysfs register (not for ide fake devices) */ | |
667 | if (major == MAJOR_NR) { | |
7d314e34 PBG |
668 | ubd_devs[unit].pdev.id = unit; |
669 | ubd_devs[unit].pdev.name = DRIVER_NAME; | |
2e3f5251 JD |
670 | ubd_devs[unit].pdev.dev.release = ubd_device_release; |
671 | ubd_devs[unit].pdev.dev.driver_data = &ubd_devs[unit]; | |
7d314e34 PBG |
672 | platform_device_register(&ubd_devs[unit].pdev); |
673 | disk->driverfs_dev = &ubd_devs[unit].pdev.dev; | |
1da177e4 LT |
674 | } |
675 | ||
7d314e34 | 676 | disk->private_data = &ubd_devs[unit]; |
62f96cb0 | 677 | disk->queue = ubd_devs[unit].queue; |
1da177e4 LT |
678 | add_disk(disk); |
679 | ||
680 | *disk_out = disk; | |
681 | return 0; | |
682 | } | |
683 | ||
684 | #define ROUND_BLOCK(n) ((n + ((1 << 9) - 1)) & (-1 << 9)) | |
685 | ||
f28169d2 | 686 | static int ubd_add(int n, char **error_out) |
1da177e4 | 687 | { |
7d314e34 | 688 | struct ubd *ubd_dev = &ubd_devs[n]; |
f28169d2 | 689 | int err = 0; |
1da177e4 | 690 | |
7d314e34 | 691 | if(ubd_dev->file == NULL) |
ec7cf783 | 692 | goto out; |
1da177e4 | 693 | |
7d314e34 | 694 | err = ubd_file_size(ubd_dev, &ubd_dev->size); |
f28169d2 JD |
695 | if(err < 0){ |
696 | *error_out = "Couldn't determine size of device's file"; | |
80c13749 | 697 | goto out; |
f28169d2 | 698 | } |
1da177e4 | 699 | |
7d314e34 | 700 | ubd_dev->size = ROUND_BLOCK(ubd_dev->size); |
1da177e4 | 701 | |
a0044bdf | 702 | INIT_LIST_HEAD(&ubd_dev->restart); |
4f40c055 | 703 | sg_init_table(ubd_dev->sg, MAX_SG); |
a0044bdf | 704 | |
62f96cb0 JD |
705 | err = -ENOMEM; |
706 | ubd_dev->queue = blk_init_queue(do_ubd_request, &ubd_dev->lock); | |
707 | if (ubd_dev->queue == NULL) { | |
708 | *error_out = "Failed to initialize device queue"; | |
80c13749 | 709 | goto out; |
62f96cb0 JD |
710 | } |
711 | ubd_dev->queue->queuedata = ubd_dev; | |
712 | ||
a0044bdf | 713 | blk_queue_max_hw_segments(ubd_dev->queue, MAX_SG); |
62f96cb0 JD |
714 | err = ubd_disk_register(MAJOR_NR, ubd_dev->size, n, &ubd_gendisk[n]); |
715 | if(err){ | |
716 | *error_out = "Failed to register device"; | |
717 | goto out_cleanup; | |
718 | } | |
6c29256c | 719 | |
1da177e4 | 720 | if(fake_major != MAJOR_NR) |
5f75a4f8 | 721 | ubd_disk_register(fake_major, ubd_dev->size, n, |
62f96cb0 | 722 | &fake_gendisk[n]); |
1da177e4 LT |
723 | |
724 | /* perhaps this should also be under the "if (fake_major)" above */ | |
725 | /* using the fake_disk->disk_name and also the fakehd_set name */ | |
726 | if (fake_ide) | |
727 | make_ide_entries(ubd_gendisk[n]->disk_name); | |
728 | ||
ec7cf783 | 729 | err = 0; |
ec7cf783 JD |
730 | out: |
731 | return err; | |
62f96cb0 JD |
732 | |
733 | out_cleanup: | |
734 | blk_cleanup_queue(ubd_dev->queue); | |
735 | goto out; | |
1da177e4 LT |
736 | } |
737 | ||
f28169d2 | 738 | static int ubd_config(char *str, char **error_out) |
1da177e4 | 739 | { |
e7f6552f | 740 | int n, ret; |
1da177e4 | 741 | |
f28169d2 JD |
742 | /* This string is possibly broken up and stored, so it's only |
743 | * freed if ubd_setup_common fails, or if only general options | |
744 | * were set. | |
745 | */ | |
970d6e3a | 746 | str = kstrdup(str, GFP_KERNEL); |
e7f6552f | 747 | if (str == NULL) { |
f28169d2 JD |
748 | *error_out = "Failed to allocate memory"; |
749 | return -ENOMEM; | |
1da177e4 | 750 | } |
f28169d2 JD |
751 | |
752 | ret = ubd_setup_common(str, &n, error_out); | |
753 | if (ret) | |
e7f6552f | 754 | goto err_free; |
f28169d2 | 755 | |
e7f6552f PBG |
756 | if (n == -1) { |
757 | ret = 0; | |
d8d7c28e | 758 | goto err_free; |
1da177e4 | 759 | } |
1da177e4 | 760 | |
dc764e50 | 761 | mutex_lock(&ubd_lock); |
f28169d2 | 762 | ret = ubd_add(n, error_out); |
e7f6552f | 763 | if (ret) |
7d314e34 | 764 | ubd_devs[n].file = NULL; |
dc764e50 | 765 | mutex_unlock(&ubd_lock); |
1da177e4 | 766 | |
e7f6552f | 767 | out: |
dc764e50 | 768 | return ret; |
e7f6552f PBG |
769 | |
770 | err_free: | |
771 | kfree(str); | |
772 | goto out; | |
1da177e4 LT |
773 | } |
774 | ||
775 | static int ubd_get_config(char *name, char *str, int size, char **error_out) | |
776 | { | |
7d314e34 | 777 | struct ubd *ubd_dev; |
1da177e4 LT |
778 | int n, len = 0; |
779 | ||
780 | n = parse_unit(&name); | |
781 | if((n >= MAX_DEV) || (n < 0)){ | |
782 | *error_out = "ubd_get_config : device number out of range"; | |
dc764e50 | 783 | return -1; |
1da177e4 LT |
784 | } |
785 | ||
7d314e34 | 786 | ubd_dev = &ubd_devs[n]; |
d7fb2c38 | 787 | mutex_lock(&ubd_lock); |
1da177e4 | 788 | |
7d314e34 | 789 | if(ubd_dev->file == NULL){ |
1da177e4 LT |
790 | CONFIG_CHUNK(str, size, len, "", 1); |
791 | goto out; | |
792 | } | |
793 | ||
7d314e34 | 794 | CONFIG_CHUNK(str, size, len, ubd_dev->file, 0); |
1da177e4 | 795 | |
7d314e34 | 796 | if(ubd_dev->cow.file != NULL){ |
1da177e4 | 797 | CONFIG_CHUNK(str, size, len, ",", 0); |
7d314e34 | 798 | CONFIG_CHUNK(str, size, len, ubd_dev->cow.file, 1); |
1da177e4 LT |
799 | } |
800 | else CONFIG_CHUNK(str, size, len, "", 1); | |
801 | ||
802 | out: | |
d7fb2c38 | 803 | mutex_unlock(&ubd_lock); |
dc764e50 | 804 | return len; |
1da177e4 LT |
805 | } |
806 | ||
29d56cfe JD |
807 | static int ubd_id(char **str, int *start_out, int *end_out) |
808 | { | |
dc764e50 | 809 | int n; |
29d56cfe JD |
810 | |
811 | n = parse_unit(str); | |
dc764e50 JD |
812 | *start_out = 0; |
813 | *end_out = MAX_DEV - 1; | |
814 | return n; | |
29d56cfe JD |
815 | } |
816 | ||
f28169d2 | 817 | static int ubd_remove(int n, char **error_out) |
1da177e4 | 818 | { |
2e3f5251 | 819 | struct gendisk *disk = ubd_gendisk[n]; |
7d314e34 | 820 | struct ubd *ubd_dev; |
29d56cfe | 821 | int err = -ENODEV; |
1da177e4 | 822 | |
d7fb2c38 | 823 | mutex_lock(&ubd_lock); |
1da177e4 | 824 | |
7d314e34 | 825 | ubd_dev = &ubd_devs[n]; |
1da177e4 | 826 | |
7d314e34 | 827 | if(ubd_dev->file == NULL) |
29d56cfe | 828 | goto out; |
1da177e4 | 829 | |
29d56cfe JD |
830 | /* you cannot remove a open disk */ |
831 | err = -EBUSY; | |
7d314e34 | 832 | if(ubd_dev->count > 0) |
1da177e4 LT |
833 | goto out; |
834 | ||
dc764e50 | 835 | ubd_gendisk[n] = NULL; |
b47d2deb JD |
836 | if(disk != NULL){ |
837 | del_gendisk(disk); | |
838 | put_disk(disk); | |
839 | } | |
1da177e4 LT |
840 | |
841 | if(fake_gendisk[n] != NULL){ | |
842 | del_gendisk(fake_gendisk[n]); | |
843 | put_disk(fake_gendisk[n]); | |
844 | fake_gendisk[n] = NULL; | |
845 | } | |
846 | ||
1da177e4 | 847 | err = 0; |
2e3f5251 | 848 | platform_device_unregister(&ubd_dev->pdev); |
29d56cfe | 849 | out: |
d7fb2c38 | 850 | mutex_unlock(&ubd_lock); |
29d56cfe | 851 | return err; |
1da177e4 LT |
852 | } |
853 | ||
f28169d2 | 854 | /* All these are called by mconsole in process context and without |
b8831a1d | 855 | * ubd-specific locks. The structure itself is const except for .list. |
f28169d2 | 856 | */ |
1da177e4 | 857 | static struct mc_device ubd_mc = { |
84f48d4f | 858 | .list = LIST_HEAD_INIT(ubd_mc.list), |
1da177e4 LT |
859 | .name = "ubd", |
860 | .config = ubd_config, | |
dc764e50 | 861 | .get_config = ubd_get_config, |
29d56cfe | 862 | .id = ubd_id, |
1da177e4 LT |
863 | .remove = ubd_remove, |
864 | }; | |
865 | ||
d8d7c28e | 866 | static int __init ubd_mc_init(void) |
1da177e4 LT |
867 | { |
868 | mconsole_register_dev(&ubd_mc); | |
869 | return 0; | |
870 | } | |
871 | ||
872 | __initcall(ubd_mc_init); | |
873 | ||
d8d7c28e PBG |
874 | static int __init ubd0_init(void) |
875 | { | |
876 | struct ubd *ubd_dev = &ubd_devs[0]; | |
877 | ||
b8831a1d | 878 | mutex_lock(&ubd_lock); |
d8d7c28e PBG |
879 | if(ubd_dev->file == NULL) |
880 | ubd_dev->file = "root_fs"; | |
b8831a1d JD |
881 | mutex_unlock(&ubd_lock); |
882 | ||
dc764e50 | 883 | return 0; |
d8d7c28e PBG |
884 | } |
885 | ||
886 | __initcall(ubd0_init); | |
887 | ||
b8831a1d | 888 | /* Used in ubd_init, which is an initcall */ |
3ae5eaec RK |
889 | static struct platform_driver ubd_driver = { |
890 | .driver = { | |
891 | .name = DRIVER_NAME, | |
892 | }, | |
1da177e4 LT |
893 | }; |
894 | ||
d8d7c28e | 895 | static int __init ubd_init(void) |
1da177e4 | 896 | { |
f28169d2 JD |
897 | char *error; |
898 | int i, err; | |
1da177e4 | 899 | |
1da177e4 LT |
900 | if (register_blkdev(MAJOR_NR, "ubd")) |
901 | return -1; | |
902 | ||
1da177e4 LT |
903 | if (fake_major != MAJOR_NR) { |
904 | char name[sizeof("ubd_nnn\0")]; | |
905 | ||
906 | snprintf(name, sizeof(name), "ubd_%d", fake_major); | |
1da177e4 LT |
907 | if (register_blkdev(fake_major, "ubd")) |
908 | return -1; | |
909 | } | |
3ae5eaec | 910 | platform_driver_register(&ubd_driver); |
dc764e50 | 911 | mutex_lock(&ubd_lock); |
f28169d2 JD |
912 | for (i = 0; i < MAX_DEV; i++){ |
913 | err = ubd_add(i, &error); | |
914 | if(err) | |
915 | printk(KERN_ERR "Failed to initialize ubd device %d :" | |
916 | "%s\n", i, error); | |
917 | } | |
dc764e50 | 918 | mutex_unlock(&ubd_lock); |
1da177e4 LT |
919 | return 0; |
920 | } | |
921 | ||
922 | late_initcall(ubd_init); | |
923 | ||
d8d7c28e | 924 | static int __init ubd_driver_init(void){ |
91acb21f JD |
925 | unsigned long stack; |
926 | int err; | |
927 | ||
928 | /* Set by CONFIG_BLK_DEV_UBD_SYNC or ubd=sync.*/ | |
929 | if(global_openflags.s){ | |
930 | printk(KERN_INFO "ubd: Synchronous mode\n"); | |
931 | /* Letting ubd=sync be like using ubd#s= instead of ubd#= is | |
932 | * enough. So use anyway the io thread. */ | |
933 | } | |
934 | stack = alloc_stack(0, 0); | |
6c29256c | 935 | io_pid = start_io_thread(stack + PAGE_SIZE - sizeof(void *), |
91acb21f JD |
936 | &thread_fd); |
937 | if(io_pid < 0){ | |
6c29256c | 938 | printk(KERN_ERR |
91acb21f JD |
939 | "ubd : Failed to start I/O thread (errno = %d) - " |
940 | "falling back to synchronous I/O\n", -io_pid); | |
941 | io_pid = -1; | |
dc764e50 | 942 | return 0; |
91acb21f | 943 | } |
6c29256c | 944 | err = um_request_irq(UBD_IRQ, thread_fd, IRQ_READ, ubd_intr, |
7d314e34 | 945 | IRQF_DISABLED, "ubd", ubd_devs); |
91acb21f JD |
946 | if(err != 0) |
947 | printk(KERN_ERR "um_request_irq failed - errno = %d\n", -err); | |
f4c57a78 | 948 | return 0; |
91acb21f JD |
949 | } |
950 | ||
951 | device_initcall(ubd_driver_init); | |
952 | ||
1da177e4 LT |
953 | static int ubd_open(struct inode *inode, struct file *filp) |
954 | { | |
955 | struct gendisk *disk = inode->i_bdev->bd_disk; | |
7d314e34 | 956 | struct ubd *ubd_dev = disk->private_data; |
1da177e4 LT |
957 | int err = 0; |
958 | ||
7d314e34 PBG |
959 | if(ubd_dev->count == 0){ |
960 | err = ubd_open_dev(ubd_dev); | |
1da177e4 LT |
961 | if(err){ |
962 | printk(KERN_ERR "%s: Can't open \"%s\": errno = %d\n", | |
7d314e34 | 963 | disk->disk_name, ubd_dev->file, -err); |
1da177e4 LT |
964 | goto out; |
965 | } | |
966 | } | |
7d314e34 PBG |
967 | ubd_dev->count++; |
968 | set_disk_ro(disk, !ubd_dev->openflags.w); | |
2c49be99 PBG |
969 | |
970 | /* This should no more be needed. And it didn't work anyway to exclude | |
971 | * read-write remounting of filesystems.*/ | |
7d314e34 | 972 | /*if((filp->f_mode & FMODE_WRITE) && !ubd_dev->openflags.w){ |
5f75a4f8 | 973 | if(--ubd_dev->count == 0) ubd_close_dev(ubd_dev); |
1da177e4 | 974 | err = -EROFS; |
2c49be99 | 975 | }*/ |
1da177e4 | 976 | out: |
dc764e50 | 977 | return err; |
1da177e4 LT |
978 | } |
979 | ||
980 | static int ubd_release(struct inode * inode, struct file * file) | |
981 | { | |
982 | struct gendisk *disk = inode->i_bdev->bd_disk; | |
7d314e34 | 983 | struct ubd *ubd_dev = disk->private_data; |
1da177e4 | 984 | |
7d314e34 | 985 | if(--ubd_dev->count == 0) |
5f75a4f8 | 986 | ubd_close_dev(ubd_dev); |
dc764e50 | 987 | return 0; |
1da177e4 LT |
988 | } |
989 | ||
91acb21f JD |
990 | static void cowify_bitmap(__u64 io_offset, int length, unsigned long *cow_mask, |
991 | __u64 *cow_offset, unsigned long *bitmap, | |
992 | __u64 bitmap_offset, unsigned long *bitmap_words, | |
993 | __u64 bitmap_len) | |
1da177e4 | 994 | { |
91acb21f JD |
995 | __u64 sector = io_offset >> 9; |
996 | int i, update_bitmap = 0; | |
997 | ||
998 | for(i = 0; i < length >> 9; i++){ | |
999 | if(cow_mask != NULL) | |
1000 | ubd_set_bit(i, (unsigned char *) cow_mask); | |
1001 | if(ubd_test_bit(sector + i, (unsigned char *) bitmap)) | |
1002 | continue; | |
1da177e4 | 1003 | |
91acb21f JD |
1004 | update_bitmap = 1; |
1005 | ubd_set_bit(sector + i, (unsigned char *) bitmap); | |
1006 | } | |
1007 | ||
1008 | if(!update_bitmap) | |
1009 | return; | |
1da177e4 | 1010 | |
91acb21f | 1011 | *cow_offset = sector / (sizeof(unsigned long) * 8); |
1da177e4 | 1012 | |
91acb21f JD |
1013 | /* This takes care of the case where we're exactly at the end of the |
1014 | * device, and *cow_offset + 1 is off the end. So, just back it up | |
1015 | * by one word. Thanks to Lynn Kerby for the fix and James McMechan | |
1016 | * for the original diagnosis. | |
1017 | */ | |
1018 | if(*cow_offset == ((bitmap_len + sizeof(unsigned long) - 1) / | |
1019 | sizeof(unsigned long) - 1)) | |
1020 | (*cow_offset)--; | |
1021 | ||
1022 | bitmap_words[0] = bitmap[*cow_offset]; | |
1023 | bitmap_words[1] = bitmap[*cow_offset + 1]; | |
1024 | ||
1025 | *cow_offset *= sizeof(unsigned long); | |
1026 | *cow_offset += bitmap_offset; | |
1027 | } | |
1028 | ||
1029 | static void cowify_req(struct io_thread_req *req, unsigned long *bitmap, | |
1030 | __u64 bitmap_offset, __u64 bitmap_len) | |
1031 | { | |
1032 | __u64 sector = req->offset >> 9; | |
1033 | int i; | |
1034 | ||
1035 | if(req->length > (sizeof(req->sector_mask) * 8) << 9) | |
1036 | panic("Operation too long"); | |
1037 | ||
1038 | if(req->op == UBD_READ) { | |
1039 | for(i = 0; i < req->length >> 9; i++){ | |
1040 | if(ubd_test_bit(sector + i, (unsigned char *) bitmap)) | |
6c29256c | 1041 | ubd_set_bit(i, (unsigned char *) |
91acb21f | 1042 | &req->sector_mask); |
dc764e50 | 1043 | } |
91acb21f JD |
1044 | } |
1045 | else cowify_bitmap(req->offset, req->length, &req->sector_mask, | |
1046 | &req->cow_offset, bitmap, bitmap_offset, | |
1047 | req->bitmap_words, bitmap_len); | |
1da177e4 LT |
1048 | } |
1049 | ||
62f96cb0 | 1050 | /* Called with dev->lock held */ |
a0044bdf JD |
1051 | static void prepare_request(struct request *req, struct io_thread_req *io_req, |
1052 | unsigned long long offset, int page_offset, | |
1053 | int len, struct page *page) | |
1da177e4 LT |
1054 | { |
1055 | struct gendisk *disk = req->rq_disk; | |
7d314e34 | 1056 | struct ubd *ubd_dev = disk->private_data; |
91acb21f | 1057 | |
62f96cb0 | 1058 | io_req->req = req; |
a0044bdf JD |
1059 | io_req->fds[0] = (ubd_dev->cow.file != NULL) ? ubd_dev->cow.fd : |
1060 | ubd_dev->fd; | |
7d314e34 | 1061 | io_req->fds[1] = ubd_dev->fd; |
91acb21f | 1062 | io_req->cow_offset = -1; |
1da177e4 LT |
1063 | io_req->offset = offset; |
1064 | io_req->length = len; | |
1065 | io_req->error = 0; | |
91acb21f JD |
1066 | io_req->sector_mask = 0; |
1067 | ||
1068 | io_req->op = (rq_data_dir(req) == READ) ? UBD_READ : UBD_WRITE; | |
1da177e4 | 1069 | io_req->offsets[0] = 0; |
7d314e34 | 1070 | io_req->offsets[1] = ubd_dev->cow.data_offset; |
a0044bdf | 1071 | io_req->buffer = page_address(page) + page_offset; |
1da177e4 LT |
1072 | io_req->sectorsize = 1 << 9; |
1073 | ||
7d314e34 | 1074 | if(ubd_dev->cow.file != NULL) |
a0044bdf JD |
1075 | cowify_req(io_req, ubd_dev->cow.bitmap, |
1076 | ubd_dev->cow.bitmap_offset, ubd_dev->cow.bitmap_len); | |
91acb21f | 1077 | |
1da177e4 LT |
1078 | } |
1079 | ||
62f96cb0 | 1080 | /* Called with dev->lock held */ |
165125e1 | 1081 | static void do_ubd_request(struct request_queue *q) |
1da177e4 | 1082 | { |
2adcec21 | 1083 | struct io_thread_req *io_req; |
1da177e4 | 1084 | struct request *req; |
0a6d3a2a | 1085 | int n, last_sectors; |
a0044bdf JD |
1086 | |
1087 | while(1){ | |
2a9529a0 | 1088 | struct ubd *dev = q->queuedata; |
a0044bdf JD |
1089 | if(dev->end_sg == 0){ |
1090 | struct request *req = elv_next_request(q); | |
1091 | if(req == NULL) | |
1092 | return; | |
1093 | ||
1094 | dev->request = req; | |
1095 | blkdev_dequeue_request(req); | |
1096 | dev->start_sg = 0; | |
1097 | dev->end_sg = blk_rq_map_sg(q, req, dev->sg); | |
1098 | } | |
1099 | ||
1100 | req = dev->request; | |
0a6d3a2a | 1101 | last_sectors = 0; |
a0044bdf JD |
1102 | while(dev->start_sg < dev->end_sg){ |
1103 | struct scatterlist *sg = &dev->sg[dev->start_sg]; | |
1104 | ||
0a6d3a2a | 1105 | req->sector += last_sectors; |
2adcec21 | 1106 | io_req = kmalloc(sizeof(struct io_thread_req), |
990c5587 | 1107 | GFP_ATOMIC); |
2adcec21 JD |
1108 | if(io_req == NULL){ |
1109 | if(list_empty(&dev->restart)) | |
1110 | list_add(&dev->restart, &restart); | |
1111 | return; | |
1112 | } | |
1113 | prepare_request(req, io_req, | |
a0044bdf | 1114 | (unsigned long long) req->sector << 9, |
45711f1a | 1115 | sg->offset, sg->length, sg_page(sg)); |
a0044bdf | 1116 | |
0a6d3a2a | 1117 | last_sectors = sg->length >> 9; |
a6ea4cce JD |
1118 | n = os_write_file(thread_fd, &io_req, |
1119 | sizeof(struct io_thread_req *)); | |
2adcec21 | 1120 | if(n != sizeof(struct io_thread_req *)){ |
a0044bdf JD |
1121 | if(n != -EAGAIN) |
1122 | printk("write to io thread failed, " | |
1123 | "errno = %d\n", -n); | |
1124 | else if(list_empty(&dev->restart)) | |
1125 | list_add(&dev->restart, &restart); | |
12429bf9 | 1126 | kfree(io_req); |
a0044bdf JD |
1127 | return; |
1128 | } | |
1129 | ||
a0044bdf | 1130 | dev->start_sg++; |
1da177e4 | 1131 | } |
a0044bdf JD |
1132 | dev->end_sg = 0; |
1133 | dev->request = NULL; | |
1da177e4 LT |
1134 | } |
1135 | } | |
1136 | ||
a885c8c4 CH |
1137 | static int ubd_getgeo(struct block_device *bdev, struct hd_geometry *geo) |
1138 | { | |
7d314e34 | 1139 | struct ubd *ubd_dev = bdev->bd_disk->private_data; |
a885c8c4 CH |
1140 | |
1141 | geo->heads = 128; | |
1142 | geo->sectors = 32; | |
7d314e34 | 1143 | geo->cylinders = ubd_dev->size / (128 * 32 * 512); |
a885c8c4 CH |
1144 | return 0; |
1145 | } | |
1146 | ||
1da177e4 LT |
1147 | static int ubd_ioctl(struct inode * inode, struct file * file, |
1148 | unsigned int cmd, unsigned long arg) | |
1149 | { | |
7d314e34 | 1150 | struct ubd *ubd_dev = inode->i_bdev->bd_disk->private_data; |
1da177e4 LT |
1151 | struct hd_driveid ubd_id = { |
1152 | .cyls = 0, | |
1153 | .heads = 128, | |
1154 | .sectors = 32, | |
1155 | }; | |
1156 | ||
1157 | switch (cmd) { | |
1da177e4 | 1158 | struct cdrom_volctrl volume; |
1da177e4 | 1159 | case HDIO_GET_IDENTITY: |
7d314e34 | 1160 | ubd_id.cyls = ubd_dev->size / (128 * 32 * 512); |
1da177e4 LT |
1161 | if(copy_to_user((char __user *) arg, (char *) &ubd_id, |
1162 | sizeof(ubd_id))) | |
dc764e50 JD |
1163 | return -EFAULT; |
1164 | return 0; | |
b8831a1d | 1165 | |
1da177e4 LT |
1166 | case CDROMVOLREAD: |
1167 | if(copy_from_user(&volume, (char __user *) arg, sizeof(volume))) | |
dc764e50 | 1168 | return -EFAULT; |
1da177e4 LT |
1169 | volume.channel0 = 255; |
1170 | volume.channel1 = 255; | |
1171 | volume.channel2 = 255; | |
1172 | volume.channel3 = 255; | |
1173 | if(copy_to_user((char __user *) arg, &volume, sizeof(volume))) | |
dc764e50 JD |
1174 | return -EFAULT; |
1175 | return 0; | |
1da177e4 | 1176 | } |
dc764e50 | 1177 | return -EINVAL; |
1da177e4 LT |
1178 | } |
1179 | ||
4833aff7 | 1180 | static int path_requires_switch(char *from_cmdline, char *from_cow, char *cow) |
1da177e4 LT |
1181 | { |
1182 | struct uml_stat buf1, buf2; | |
1183 | int err; | |
1184 | ||
4833aff7 PBG |
1185 | if(from_cmdline == NULL) |
1186 | return 0; | |
1187 | if(!strcmp(from_cmdline, from_cow)) | |
1188 | return 0; | |
1da177e4 LT |
1189 | |
1190 | err = os_stat_file(from_cmdline, &buf1); | |
1191 | if(err < 0){ | |
1192 | printk("Couldn't stat '%s', err = %d\n", from_cmdline, -err); | |
4833aff7 | 1193 | return 0; |
1da177e4 LT |
1194 | } |
1195 | err = os_stat_file(from_cow, &buf2); | |
1196 | if(err < 0){ | |
1197 | printk("Couldn't stat '%s', err = %d\n", from_cow, -err); | |
4833aff7 | 1198 | return 1; |
1da177e4 LT |
1199 | } |
1200 | if((buf1.ust_dev == buf2.ust_dev) && (buf1.ust_ino == buf2.ust_ino)) | |
4833aff7 | 1201 | return 0; |
1da177e4 LT |
1202 | |
1203 | printk("Backing file mismatch - \"%s\" requested,\n" | |
1204 | "\"%s\" specified in COW header of \"%s\"\n", | |
1205 | from_cmdline, from_cow, cow); | |
4833aff7 | 1206 | return 1; |
1da177e4 LT |
1207 | } |
1208 | ||
1209 | static int backing_file_mismatch(char *file, __u64 size, time_t mtime) | |
1210 | { | |
1211 | unsigned long modtime; | |
fe1db50c | 1212 | unsigned long long actual; |
1da177e4 LT |
1213 | int err; |
1214 | ||
1215 | err = os_file_modtime(file, &modtime); | |
1216 | if(err < 0){ | |
1217 | printk("Failed to get modification time of backing file " | |
1218 | "\"%s\", err = %d\n", file, -err); | |
dc764e50 | 1219 | return err; |
1da177e4 LT |
1220 | } |
1221 | ||
1222 | err = os_file_size(file, &actual); | |
1223 | if(err < 0){ | |
1224 | printk("Failed to get size of backing file \"%s\", " | |
1225 | "err = %d\n", file, -err); | |
dc764e50 | 1226 | return err; |
1da177e4 LT |
1227 | } |
1228 | ||
dc764e50 | 1229 | if(actual != size){ |
1da177e4 LT |
1230 | /*__u64 can be a long on AMD64 and with %lu GCC complains; so |
1231 | * the typecast.*/ | |
1232 | printk("Size mismatch (%llu vs %llu) of COW header vs backing " | |
1233 | "file\n", (unsigned long long) size, actual); | |
dc764e50 | 1234 | return -EINVAL; |
1da177e4 LT |
1235 | } |
1236 | if(modtime != mtime){ | |
1237 | printk("mtime mismatch (%ld vs %ld) of COW header vs backing " | |
1238 | "file\n", mtime, modtime); | |
dc764e50 | 1239 | return -EINVAL; |
1da177e4 | 1240 | } |
dc764e50 | 1241 | return 0; |
1da177e4 LT |
1242 | } |
1243 | ||
1244 | int read_cow_bitmap(int fd, void *buf, int offset, int len) | |
1245 | { | |
1246 | int err; | |
1247 | ||
1248 | err = os_seek_file(fd, offset); | |
1249 | if(err < 0) | |
dc764e50 | 1250 | return err; |
1da177e4 | 1251 | |
dc764e50 | 1252 | err = os_read_file(fd, buf, len); |
1da177e4 | 1253 | if(err < 0) |
dc764e50 | 1254 | return err; |
1da177e4 | 1255 | |
dc764e50 | 1256 | return 0; |
1da177e4 LT |
1257 | } |
1258 | ||
6c29256c | 1259 | int open_ubd_file(char *file, struct openflags *openflags, int shared, |
1da177e4 LT |
1260 | char **backing_file_out, int *bitmap_offset_out, |
1261 | unsigned long *bitmap_len_out, int *data_offset_out, | |
1262 | int *create_cow_out) | |
1263 | { | |
1264 | time_t mtime; | |
1265 | unsigned long long size; | |
1266 | __u32 version, align; | |
1267 | char *backing_file; | |
4833aff7 | 1268 | int fd, err, sectorsize, asked_switch, mode = 0644; |
1da177e4 LT |
1269 | |
1270 | fd = os_open_file(file, *openflags, mode); | |
a374a48f PBG |
1271 | if (fd < 0) { |
1272 | if ((fd == -ENOENT) && (create_cow_out != NULL)) | |
1da177e4 | 1273 | *create_cow_out = 1; |
dc764e50 JD |
1274 | if (!openflags->w || |
1275 | ((fd != -EROFS) && (fd != -EACCES))) | |
a374a48f | 1276 | return fd; |
1da177e4 LT |
1277 | openflags->w = 0; |
1278 | fd = os_open_file(file, *openflags, mode); | |
a374a48f PBG |
1279 | if (fd < 0) |
1280 | return fd; | |
dc764e50 | 1281 | } |
1da177e4 | 1282 | |
6c29256c JD |
1283 | if(shared) |
1284 | printk("Not locking \"%s\" on the host\n", file); | |
1285 | else { | |
1286 | err = os_lock_file(fd, openflags->w); | |
1287 | if(err < 0){ | |
1288 | printk("Failed to lock '%s', err = %d\n", file, -err); | |
1289 | goto out_close; | |
1290 | } | |
1da177e4 LT |
1291 | } |
1292 | ||
d6e05edc | 1293 | /* Successful return case! */ |
a374a48f | 1294 | if(backing_file_out == NULL) |
dc764e50 | 1295 | return fd; |
1da177e4 LT |
1296 | |
1297 | err = read_cow_header(file_reader, &fd, &version, &backing_file, &mtime, | |
1298 | &size, §orsize, &align, bitmap_offset_out); | |
1299 | if(err && (*backing_file_out != NULL)){ | |
1300 | printk("Failed to read COW header from COW file \"%s\", " | |
1301 | "errno = %d\n", file, -err); | |
1302 | goto out_close; | |
1303 | } | |
a374a48f | 1304 | if(err) |
dc764e50 | 1305 | return fd; |
1da177e4 | 1306 | |
4833aff7 | 1307 | asked_switch = path_requires_switch(*backing_file_out, backing_file, file); |
1da177e4 | 1308 | |
4833aff7 PBG |
1309 | /* Allow switching only if no mismatch. */ |
1310 | if (asked_switch && !backing_file_mismatch(*backing_file_out, size, mtime)) { | |
1da177e4 LT |
1311 | printk("Switching backing file to '%s'\n", *backing_file_out); |
1312 | err = write_cow_header(file, fd, *backing_file_out, | |
1313 | sectorsize, align, &size); | |
a374a48f | 1314 | if (err) { |
1da177e4 | 1315 | printk("Switch failed, errno = %d\n", -err); |
4833aff7 | 1316 | goto out_close; |
1da177e4 | 1317 | } |
a374a48f | 1318 | } else { |
1da177e4 LT |
1319 | *backing_file_out = backing_file; |
1320 | err = backing_file_mismatch(*backing_file_out, size, mtime); | |
a374a48f PBG |
1321 | if (err) |
1322 | goto out_close; | |
1da177e4 LT |
1323 | } |
1324 | ||
1325 | cow_sizes(version, size, sectorsize, align, *bitmap_offset_out, | |
1326 | bitmap_len_out, data_offset_out); | |
1327 | ||
dc764e50 | 1328 | return fd; |
1da177e4 LT |
1329 | out_close: |
1330 | os_close_file(fd); | |
a374a48f | 1331 | return err; |
1da177e4 LT |
1332 | } |
1333 | ||
1334 | int create_cow_file(char *cow_file, char *backing_file, struct openflags flags, | |
1335 | int sectorsize, int alignment, int *bitmap_offset_out, | |
1336 | unsigned long *bitmap_len_out, int *data_offset_out) | |
1337 | { | |
1338 | int err, fd; | |
1339 | ||
1340 | flags.c = 1; | |
6c29256c | 1341 | fd = open_ubd_file(cow_file, &flags, 0, NULL, NULL, NULL, NULL, NULL); |
1da177e4 LT |
1342 | if(fd < 0){ |
1343 | err = fd; | |
1344 | printk("Open of COW file '%s' failed, errno = %d\n", cow_file, | |
1345 | -err); | |
1346 | goto out; | |
1347 | } | |
1348 | ||
1349 | err = init_cow_file(fd, cow_file, backing_file, sectorsize, alignment, | |
1350 | bitmap_offset_out, bitmap_len_out, | |
1351 | data_offset_out); | |
1352 | if(!err) | |
dc764e50 | 1353 | return fd; |
1da177e4 LT |
1354 | os_close_file(fd); |
1355 | out: | |
dc764e50 | 1356 | return err; |
1da177e4 LT |
1357 | } |
1358 | ||
91acb21f | 1359 | static int update_bitmap(struct io_thread_req *req) |
1da177e4 | 1360 | { |
91acb21f | 1361 | int n; |
1da177e4 | 1362 | |
91acb21f | 1363 | if(req->cow_offset == -1) |
dc764e50 | 1364 | return 0; |
1da177e4 | 1365 | |
91acb21f JD |
1366 | n = os_seek_file(req->fds[1], req->cow_offset); |
1367 | if(n < 0){ | |
1368 | printk("do_io - bitmap lseek failed : err = %d\n", -n); | |
dc764e50 | 1369 | return 1; |
91acb21f | 1370 | } |
1da177e4 | 1371 | |
a6ea4cce JD |
1372 | n = os_write_file(req->fds[1], &req->bitmap_words, |
1373 | sizeof(req->bitmap_words)); | |
91acb21f JD |
1374 | if(n != sizeof(req->bitmap_words)){ |
1375 | printk("do_io - bitmap update failed, err = %d fd = %d\n", -n, | |
1376 | req->fds[1]); | |
dc764e50 | 1377 | return 1; |
91acb21f | 1378 | } |
1da177e4 | 1379 | |
dc764e50 | 1380 | return 0; |
91acb21f | 1381 | } |
1da177e4 | 1382 | |
91acb21f JD |
1383 | void do_io(struct io_thread_req *req) |
1384 | { | |
1385 | char *buf; | |
1386 | unsigned long len; | |
1387 | int n, nsectors, start, end, bit; | |
1388 | int err; | |
1389 | __u64 off; | |
1390 | ||
1391 | nsectors = req->length / req->sectorsize; | |
1392 | start = 0; | |
1393 | do { | |
1394 | bit = ubd_test_bit(start, (unsigned char *) &req->sector_mask); | |
1395 | end = start; | |
1396 | while((end < nsectors) && | |
1397 | (ubd_test_bit(end, (unsigned char *) | |
1398 | &req->sector_mask) == bit)) | |
1399 | end++; | |
1400 | ||
1401 | off = req->offset + req->offsets[bit] + | |
1402 | start * req->sectorsize; | |
1403 | len = (end - start) * req->sectorsize; | |
1404 | buf = &req->buffer[start * req->sectorsize]; | |
1405 | ||
1406 | err = os_seek_file(req->fds[bit], off); | |
1407 | if(err < 0){ | |
1408 | printk("do_io - lseek failed : err = %d\n", -err); | |
1409 | req->error = 1; | |
1410 | return; | |
1411 | } | |
1412 | if(req->op == UBD_READ){ | |
1413 | n = 0; | |
1414 | do { | |
1415 | buf = &buf[n]; | |
1416 | len -= n; | |
a6ea4cce | 1417 | n = os_read_file(req->fds[bit], buf, len); |
91acb21f JD |
1418 | if (n < 0) { |
1419 | printk("do_io - read failed, err = %d " | |
1420 | "fd = %d\n", -n, req->fds[bit]); | |
1421 | req->error = 1; | |
1422 | return; | |
1423 | } | |
1424 | } while((n < len) && (n != 0)); | |
1425 | if (n < len) memset(&buf[n], 0, len - n); | |
1426 | } else { | |
a6ea4cce | 1427 | n = os_write_file(req->fds[bit], buf, len); |
91acb21f JD |
1428 | if(n != len){ |
1429 | printk("do_io - write failed err = %d " | |
1430 | "fd = %d\n", -n, req->fds[bit]); | |
1431 | req->error = 1; | |
1432 | return; | |
1433 | } | |
1434 | } | |
1435 | ||
1436 | start = end; | |
1437 | } while(start < nsectors); | |
1da177e4 | 1438 | |
91acb21f | 1439 | req->error = update_bitmap(req); |
1da177e4 | 1440 | } |
91acb21f JD |
1441 | |
1442 | /* Changed in start_io_thread, which is serialized by being called only | |
1443 | * from ubd_init, which is an initcall. | |
1444 | */ | |
1445 | int kernel_fd = -1; | |
1446 | ||
d8d7c28e PBG |
1447 | /* Only changed by the io thread. XXX: currently unused. */ |
1448 | static int io_count = 0; | |
91acb21f JD |
1449 | |
1450 | int io_thread(void *arg) | |
1451 | { | |
2adcec21 | 1452 | struct io_thread_req *req; |
91acb21f JD |
1453 | int n; |
1454 | ||
1455 | ignore_sigwinch_sig(); | |
1456 | while(1){ | |
a6ea4cce | 1457 | n = os_read_file(kernel_fd, &req, |
2adcec21 JD |
1458 | sizeof(struct io_thread_req *)); |
1459 | if(n != sizeof(struct io_thread_req *)){ | |
91acb21f JD |
1460 | if(n < 0) |
1461 | printk("io_thread - read failed, fd = %d, " | |
1462 | "err = %d\n", kernel_fd, -n); | |
1463 | else { | |
1464 | printk("io_thread - short read, fd = %d, " | |
1465 | "length = %d\n", kernel_fd, n); | |
1466 | } | |
1467 | continue; | |
1468 | } | |
1469 | io_count++; | |
2adcec21 | 1470 | do_io(req); |
a6ea4cce | 1471 | n = os_write_file(kernel_fd, &req, |
2adcec21 JD |
1472 | sizeof(struct io_thread_req *)); |
1473 | if(n != sizeof(struct io_thread_req *)) | |
91acb21f JD |
1474 | printk("io_thread - write failed, fd = %d, err = %d\n", |
1475 | kernel_fd, -n); | |
1476 | } | |
91acb21f | 1477 | |
1b57e9c2 JD |
1478 | return 0; |
1479 | } |