device-mapper.c revision 1.4.4.4 1 /* $NetBSD: device-mapper.c,v 1.4.4.4 2009/04/28 07:35:21 skrll Exp $ */
2
3 /*
4 * Copyright (c) 2008 The NetBSD Foundation, Inc.
5 * All rights reserved.
6 *
7 * This code is derived from software contributed to The NetBSD Foundation
8 * by Adam Hamsik.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29 * POSSIBILITY OF SUCH DAMAGE.
30 */
31
32 /*
33 * I want to say thank you to all people who helped me with this project.
34 */
35
36 #include <sys/types.h>
37 #include <sys/param.h>
38
39 #include <sys/buf.h>
40 #include <sys/conf.h>
41 #include <sys/dkio.h>
42 #include <sys/disk.h>
43 #include <sys/disklabel.h>
44 #include <sys/ioctl.h>
45 #include <sys/ioccom.h>
46 #include <sys/kmem.h>
47 #include <sys/module.h>
48
49 #include "netbsd-dm.h"
50 #include "dm.h"
51
52 static dev_type_open(dmopen);
53 static dev_type_close(dmclose);
54 static dev_type_read(dmread);
55 static dev_type_write(dmwrite);
56 static dev_type_ioctl(dmioctl);
57 static dev_type_strategy(dmstrategy);
58 static dev_type_size(dmsize);
59
60 /* attach and detach routines */
61 int dmattach(void);
62 int dmdestroy(void);
63
64 static int dm_cmd_to_fun(prop_dictionary_t);
65 static int disk_ioctl_switch(dev_t, u_long, void *);
66 static int dm_ioctl_switch(u_long);
67 static void dmminphys(struct buf *);
68
69 /* ***Variable-definitions*** */
70 const struct bdevsw dm_bdevsw = {
71 .d_open = dmopen,
72 .d_close = dmclose,
73 .d_strategy = dmstrategy,
74 .d_ioctl = dmioctl,
75 .d_dump = nodump,
76 .d_psize = dmsize,
77 .d_flag = D_DISK | D_MPSAFE
78 };
79
80 const struct cdevsw dm_cdevsw = {
81 .d_open = dmopen,
82 .d_close = dmclose,
83 .d_read = dmread,
84 .d_write = dmwrite,
85 .d_ioctl = dmioctl,
86 .d_stop = nostop,
87 .d_tty = notty,
88 .d_poll = nopoll,
89 .d_mmap = nommap,
90 .d_kqfilter = nokqfilter,
91 .d_flag = D_DISK | D_MPSAFE
92 };
93
94 const struct dkdriver dmdkdriver = {
95 .d_strategy = dmstrategy
96 };
97
98 extern uint64_t dev_counter;
99
100 /*
101 * This array is used to translate cmd to function pointer.
102 *
103 * Interface between libdevmapper and lvm2tools uses different
104 * names for one IOCTL call because libdevmapper do another thing
105 * then. When I run "info" or "mknodes" libdevmapper will send same
106 * ioctl to kernel but will do another things in userspace.
107 *
108 */
109 struct cmd_function cmd_fn[] = {
110 { .cmd = "version", .fn = dm_get_version_ioctl},
111 { .cmd = "targets", .fn = dm_list_versions_ioctl},
112 { .cmd = "create", .fn = dm_dev_create_ioctl},
113 { .cmd = "info", .fn = dm_dev_status_ioctl},
114 { .cmd = "mknodes", .fn = dm_dev_status_ioctl},
115 { .cmd = "names", .fn = dm_dev_list_ioctl},
116 { .cmd = "suspend", .fn = dm_dev_suspend_ioctl},
117 { .cmd = "remove", .fn = dm_dev_remove_ioctl},
118 { .cmd = "rename", .fn = dm_dev_rename_ioctl},
119 { .cmd = "resume", .fn = dm_dev_resume_ioctl},
120 { .cmd = "clear", .fn = dm_table_clear_ioctl},
121 { .cmd = "deps", .fn = dm_table_deps_ioctl},
122 { .cmd = "reload", .fn = dm_table_load_ioctl},
123 { .cmd = "status", .fn = dm_table_status_ioctl},
124 { .cmd = "table", .fn = dm_table_status_ioctl},
125 {NULL, NULL}
126 };
127
128
129 MODULE(MODULE_CLASS_DRIVER, dm, NULL);
130
131 /* New module handle routine */
132 static int
133 dm_modcmd(modcmd_t cmd, void *arg)
134 {
135 #ifdef _MODULE
136 int bmajor = -1, cmajor = -1;
137
138 switch (cmd) {
139 case MODULE_CMD_INIT:
140 dmattach();
141 return devsw_attach("dm", &dm_bdevsw, &bmajor,
142 &dm_cdevsw, &cmajor);
143 break;
144
145 case MODULE_CMD_FINI:
146 /*
147 * Disable unloading of dm module if there are any devices
148 * defined in driver. This is probably too strong we need
149 * to disable auto-unload only if there is mounted dm device
150 * present.
151 */
152 if (dev_counter > 0)
153 return EBUSY;
154 dmdestroy();
155 return devsw_detach(&dm_bdevsw, &dm_cdevsw);
156 break;
157 case MODULE_CMD_STAT:
158 return ENOTTY;
159
160 default:
161 return ENOTTY;
162 }
163
164 return 0;
165 #else
166
167 if (cmd == MODULE_CMD_INIT)
168 return 0;
169 return ENOTTY;
170
171 #endif /* _MODULE */
172 }
173
174
175 /* attach routine */
176 int
177 dmattach(void)
178 {
179 dm_target_init();
180 dm_dev_init();
181 dm_pdev_init();
182
183 return 0;
184 }
185
186 /* Destroy routine */
187 int
188 dmdestroy(void)
189 {
190 dm_dev_destroy();
191 dm_pdev_destroy();
192 dm_target_destroy();
193
194 return 0;
195 }
196
197 static int
198 dmopen(dev_t dev, int flags, int mode, struct lwp *l)
199 {
200 aprint_debug("open routine called %" PRIu32 "\n", minor(dev));
201 return 0;
202 }
203
204 static int
205 dmclose(dev_t dev, int flags, int mode, struct lwp *l)
206 {
207 aprint_debug("CLOSE routine called\n");
208
209 return 0;
210 }
211
212
213 static int
214 dmioctl(dev_t dev, const u_long cmd, void *data, int flag, struct lwp *l)
215 {
216 int r;
217 prop_dictionary_t dm_dict_in;
218
219 r = 0;
220
221 aprint_debug("dmioctl called\n");
222
223 KASSERT(data != NULL);
224
225 if (disk_ioctl_switch(dev, cmd, data) != 0) {
226 struct plistref *pref = (struct plistref *) data;
227
228 if((r = prop_dictionary_copyin_ioctl(pref, cmd, &dm_dict_in)) != 0)
229 return r;
230
231 dm_check_version(dm_dict_in);
232
233 /* call cmd selected function */
234 if ((r = dm_ioctl_switch(cmd)) != 0) {
235 prop_object_release(dm_dict_in);
236 return r;
237 }
238
239 /* run ioctl routine */
240 if ((r = dm_cmd_to_fun(dm_dict_in)) != 0) {
241 prop_object_release(dm_dict_in);
242 return r;
243 }
244
245 r = prop_dictionary_copyout_ioctl(pref, cmd, dm_dict_in);
246
247 prop_object_release(dm_dict_in);
248 }
249
250 return r;
251 }
252
253 /*
254 * Translate command sent from libdevmapper to func.
255 */
256 static int
257 dm_cmd_to_fun(prop_dictionary_t dm_dict){
258 int i, r;
259 prop_string_t command;
260
261 r = 0;
262
263 if ((command = prop_dictionary_get(dm_dict, DM_IOCTL_COMMAND)) == NULL)
264 return EINVAL;
265
266 for(i = 0; cmd_fn[i].cmd != NULL; i++)
267 if (prop_string_equals_cstring(command, cmd_fn[i].cmd))
268 break;
269
270 if (cmd_fn[i].cmd == NULL)
271 return EINVAL;
272
273 aprint_debug("ioctl %s called\n", cmd_fn[i].cmd);
274 r = cmd_fn[i].fn(dm_dict);
275
276 return r;
277 }
278
279 /* Call apropriate ioctl handler function. */
280 static int
281 dm_ioctl_switch(u_long cmd)
282 {
283 int r;
284
285 r = 0;
286
287 switch(cmd) {
288
289 case NETBSD_DM_IOCTL:
290 aprint_debug("NetBSD_DM_IOCTL called\n");
291 break;
292
293 default:
294 aprint_debug("unknown ioctl called\n");
295 return ENOTTY;
296 break; /* NOT REACHED */
297 }
298
299 return r;
300 }
301
302 /*
303 * Check for disk specific ioctls.
304 */
305
306 static int
307 disk_ioctl_switch(dev_t dev, u_long cmd, void *data)
308 {
309 dm_dev_t *dmv;
310
311 switch(cmd) {
312 case DIOCGWEDGEINFO:
313 {
314 struct dkwedge_info *dkw = (void *) data;
315
316 if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL)
317 return ENOENT;
318
319 aprint_normal("DIOCGWEDGEINFO ioctl called\n");
320
321 strlcpy(dkw->dkw_devname, dmv->name, 16);
322 strlcpy(dkw->dkw_wname, dmv->name, DM_NAME_LEN);
323 strlcpy(dkw->dkw_parent, dmv->name, 16);
324
325 dkw->dkw_offset = 0;
326 dkw->dkw_size = dm_table_size(&dmv->table_head);
327 strcpy(dkw->dkw_ptype, DKW_PTYPE_FFS);
328
329 dm_dev_unbusy(dmv);
330 break;
331 }
332
333 case DIOCGDINFO:
334 {
335 if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL)
336 return ENOENT;
337
338 aprint_debug("DIOCGDINFO %d\n", dmv->diskp->dk_label->d_secsize);
339
340 *(struct disklabel *)data = *(dmv->diskp->dk_label);
341
342 dm_dev_unbusy(dmv);
343 break;
344 }
345
346 case DIOCGPART:
347 {
348 if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL)
349 return ENOENT;
350
351 ((struct partinfo *)data)->disklab = dmv->diskp->dk_label;
352 ((struct partinfo *)data)->part = &dmv->diskp->dk_label->d_partitions[0];
353
354 dm_dev_unbusy(dmv);
355 break;
356 }
357 case DIOCWDINFO:
358 case DIOCSDINFO:
359 case DIOCKLABEL:
360 case DIOCWLABEL:
361 case DIOCGDEFLABEL:
362
363 default:
364 aprint_debug("unknown disk_ioctl called\n");
365 return 1;
366 break; /* NOT REACHED */
367 }
368
369 return 0;
370 }
371
372 /*
373 * Do all IO operations on dm logical devices.
374 */
375 static void
376 dmstrategy(struct buf *bp)
377 {
378 dm_dev_t *dmv;
379 dm_table_t *tbl;
380 dm_table_entry_t *table_en;
381 struct buf *nestbuf;
382
383 uint32_t dev_type;
384
385 uint64_t buf_start, buf_len, issued_len;
386 uint64_t table_start, table_end;
387 uint64_t start, end;
388
389 buf_start = bp->b_blkno * DEV_BSIZE;
390 buf_len = bp->b_bcount;
391
392 tbl = NULL;
393
394 table_end = 0;
395 dev_type = 0;
396 issued_len = 0;
397
398 if ((dmv = dm_dev_lookup(NULL, NULL, minor(bp->b_dev))) == NULL) {
399 bp->b_error = EIO;
400 bp->b_resid = bp->b_bcount;
401 biodone(bp);
402 return;
403 }
404
405 /* FIXME: have to be called with IPL_BIO*/
406 disk_busy(dmv->diskp);
407
408 /* Select active table */
409 tbl = dm_table_get_entry(&dmv->table_head, DM_TABLE_ACTIVE);
410
411 /* Nested buffers count down to zero therefore I have
412 to set bp->b_resid to maximal value. */
413 bp->b_resid = bp->b_bcount;
414
415 /*
416 * Find out what tables I want to select.
417 */
418 SLIST_FOREACH(table_en, tbl, next)
419 {
420 /* I need need number of bytes not blocks. */
421 table_start = table_en->start * DEV_BSIZE;
422 /*
423 * I have to sub 1 from table_en->length to prevent
424 * off by one error
425 */
426 table_end = table_start + (table_en->length)* DEV_BSIZE;
427
428 start = MAX(table_start, buf_start);
429
430 end = MIN(table_end, buf_start + buf_len);
431
432 aprint_debug("----------------------------------------\n");
433 aprint_debug("table_start %010" PRIu64", table_end %010"
434 PRIu64 "\n", table_start, table_end);
435 aprint_debug("buf_start %010" PRIu64", buf_len %010"
436 PRIu64"\n", buf_start, buf_len);
437 aprint_debug("start-buf_start %010"PRIu64", end %010"
438 PRIu64"\n", start - buf_start, end);
439 aprint_debug("start %010" PRIu64" , end %010"
440 PRIu64"\n", start, end);
441 aprint_debug("\n----------------------------------------\n");
442
443 if (start < end) {
444 /* create nested buffer */
445 nestbuf = getiobuf(NULL, true);
446
447 nestiobuf_setup(bp, nestbuf, start - buf_start,
448 (end - start));
449
450 issued_len += end - start;
451
452 /* I need number of blocks. */
453 nestbuf->b_blkno = (start - table_start) / DEV_BSIZE;
454
455 table_en->target->strategy(table_en, nestbuf);
456 }
457 }
458
459 if (issued_len < buf_len)
460 nestiobuf_done(bp, buf_len - issued_len, EINVAL);
461
462 /* FIXME have to be called with SPL_BIO*/
463 disk_unbusy(dmv->diskp, buf_len, bp != NULL ? bp->b_flags & B_READ : 0);
464
465 dm_table_release(&dmv->table_head, DM_TABLE_ACTIVE);
466 dm_dev_unbusy(dmv);
467
468 return;
469 }
470
471
472 static int
473 dmread(dev_t dev, struct uio *uio, int flag)
474 {
475 return (physio(dmstrategy, NULL, dev, B_READ, dmminphys, uio));
476 }
477
478 static int
479 dmwrite(dev_t dev, struct uio *uio, int flag)
480 {
481 return (physio(dmstrategy, NULL, dev, B_WRITE, dmminphys, uio));
482 }
483
484 static int
485 dmsize(dev_t dev)
486 {
487 dm_dev_t *dmv;
488 uint64_t size;
489
490 size = 0;
491
492 if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL)
493 return -ENOENT;
494
495 size = dm_table_size(&dmv->table_head);
496 dm_dev_unbusy(dmv);
497
498 return size;
499 }
500
501 static void
502 dmminphys(struct buf *bp)
503 {
504 bp->b_bcount = MIN(bp->b_bcount, MAXPHYS);
505 }
506
507 /*
508 * Load the label information on the named device
509 * Actually fabricate a disklabel.
510 *
511 * EVENTUALLY take information about different
512 * data tracks from the TOC and put it in the disklabel
513 *
514 * Copied from vnd code.
515 */
516 void
517 dmgetdisklabel(struct disklabel *lp, dm_table_head_t *head)
518 {
519 struct partition *pp;
520 int dmp_size;
521
522 dmp_size = dm_table_size(head);
523
524 /*
525 * Size must be at least 2048 DEV_BSIZE blocks
526 * (1M) in order to use this geometry.
527 */
528
529 lp->d_secperunit = dmp_size;
530 lp->d_secsize = DEV_BSIZE;
531 lp->d_nsectors = 32;
532 lp->d_ntracks = 64;
533 lp->d_ncylinders = dmp_size / (lp->d_nsectors * lp->d_ntracks);
534 lp->d_secpercyl = lp->d_ntracks * lp->d_nsectors;
535
536 strncpy(lp->d_typename, "lvm", sizeof(lp->d_typename));
537 lp->d_type = DTYPE_DM;
538 strncpy(lp->d_packname, "fictitious", sizeof(lp->d_packname));
539 lp->d_rpm = 3600;
540 lp->d_interleave = 1;
541 lp->d_flags = 0;
542
543 pp = &lp->d_partitions[0];
544 /*
545 * This is logical offset and therefore it can be 0
546 * I will consider table offsets later in dmstrategy.
547 */
548 pp->p_offset = 0;
549 pp->p_size = dmp_size * DEV_BSIZE;
550 pp->p_fstype = FS_BSDFFS; /* default value */
551 lp->d_npartitions = 1;
552
553 lp->d_magic = DISKMAGIC;
554 lp->d_magic2 = DISKMAGIC;
555 lp->d_checksum = dkcksum(lp);
556
557 return;
558 }
559