dm - add remove_all, refactor
[dragonfly.git] / sys / dev / disk / dm / device-mapper.c
1 /*        $NetBSD: device-mapper.c,v 1.22 2010/03/26 15:46:04 jakllsch Exp $ */
2
3 /*
4  * Copyright (c) 2010 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/device.h>
42 #include <sys/disk.h>
43 #include <sys/disklabel.h>
44 #include <sys/dtype.h>
45 #include <sys/ioccom.h>
46 #include <sys/malloc.h>
47 #include <sys/module.h>
48 #include <sys/sysctl.h>
49 #include <dev/disk/dm/dm.h>
50
51 #include "netbsd-dm.h"
52
53 static  d_ioctl_t       dmioctl;
54 static  d_open_t        dmopen;
55 static  d_close_t       dmclose;
56 static  d_psize_t       dmsize;
57 static  d_strategy_t    dmstrategy;
58 static  d_dump_t        dmdump;
59
60 /* attach and detach routines */
61 void dmattach(int);
62 static int dm_modcmd(module_t mod, int cmd, void *unused);
63 static int dmdestroy(void);
64
65 static void dm_doinit(void);
66
67 static int dm_cmd_to_fun(prop_dictionary_t);
68 static int disk_ioctl_switch(cdev_t, u_long, void *);
69 static int dm_ioctl_switch(u_long);
70 #if 0
71 static void dmminphys(struct buf *);
72 #endif
73
74 /* ***Variable-definitions*** */
75 struct dev_ops dm_ops = {
76         { "dm", 0, D_DISK | D_MPSAFE },
77         .d_open         = dmopen,
78         .d_close        = dmclose,
79         .d_read         = physread,
80         .d_write        = physwrite,
81         .d_ioctl        = dmioctl,
82         .d_strategy     = dmstrategy,
83         .d_psize        = dmsize,
84         .d_dump         = dmdump,
85 /* D_DISK */
86 };
87
88 MALLOC_DEFINE(M_DM, "dm", "Device Mapper allocations");
89
90 int dm_debug_level = 0;
91
92 extern uint64_t dm_dev_counter;
93
94 static cdev_t dmcdev;
95
96 static moduledata_t dm_mod = {
97     "dm",
98     dm_modcmd,
99     NULL
100 };
101 DECLARE_MODULE(dm, dm_mod, SI_SUB_RAID, SI_ORDER_ANY);
102 MODULE_VERSION(dm, 1);
103
104 /*
105  * This array is used to translate cmd to function pointer.
106  *
107  * Interface between libdevmapper and lvm2tools uses different
108  * names for one IOCTL call because libdevmapper do another thing
109  * then. When I run "info" or "mknodes" libdevmapper will send same
110  * ioctl to kernel but will do another things in userspace.
111  *
112  */
113 static struct cmd_function cmd_fn[] = {
114                 { .cmd = "version", .fn = dm_get_version_ioctl},
115                 { .cmd = "targets", .fn = dm_list_versions_ioctl},
116                 { .cmd = "create",  .fn = dm_dev_create_ioctl},
117                 { .cmd = "info",    .fn = dm_dev_status_ioctl},
118                 { .cmd = "mknodes", .fn = dm_dev_status_ioctl},
119                 { .cmd = "names",   .fn = dm_dev_list_ioctl},
120                 { .cmd = "suspend", .fn = dm_dev_suspend_ioctl},
121                 { .cmd = "remove",  .fn = dm_dev_remove_ioctl},
122                 { .cmd = "remove_all", .fn = dm_dev_remove_all_ioctl},
123                 { .cmd = "rename",  .fn = dm_dev_rename_ioctl},
124                 { .cmd = "resume",  .fn = dm_dev_resume_ioctl},
125                 { .cmd = "clear",   .fn = dm_table_clear_ioctl},
126                 { .cmd = "deps",    .fn = dm_table_deps_ioctl},
127                 { .cmd = "reload",  .fn = dm_table_load_ioctl},
128                 { .cmd = "status",  .fn = dm_table_status_ioctl},
129                 { .cmd = "table",   .fn = dm_table_status_ioctl},
130                 {NULL, NULL}
131 };
132
133 /* New module handle routine */
134 static int
135 dm_modcmd(module_t mod, int cmd, void *unused)
136 {
137         int error, bmajor, cmajor;
138
139         error = 0;
140         bmajor = -1;
141         cmajor = -1;
142
143         switch (cmd) {
144         case MOD_LOAD:
145                 dm_doinit();
146                 kprintf("Device Mapper version %d.%d.%d loaded\n",
147                     DM_VERSION_MAJOR, DM_VERSION_MINOR, DM_VERSION_PATCHLEVEL);
148                 break;
149
150         case MOD_UNLOAD:
151                 /*
152                  * Disable unloading of dm module if there are any devices
153                  * defined in driver. This is probably too strong we need
154                  * to disable auto-unload only if there is mounted dm device
155                  * present.
156                  */
157                 if (dm_dev_counter > 0)
158                         return EBUSY;
159
160                 error = dmdestroy();
161                 if (error)
162                         break;
163                 kprintf("Device Mapper unloaded\n");
164                 break;
165
166         default:
167                 break;
168         }
169
170         return error;
171 }
172
173 static void
174 dm_doinit(void)
175 {
176         dm_target_init();
177         dm_dev_init();
178         dm_pdev_init();
179         dmcdev = make_dev(&dm_ops, 0, UID_ROOT, GID_OPERATOR, 0640, "mapper/control");
180 }
181
182 /* Destroy routine */
183 static int
184 dmdestroy(void)
185 {
186         destroy_dev(dmcdev);
187
188         dm_dev_uninit();
189         dm_pdev_uninit();
190         dm_target_uninit();
191
192         return 0;
193 }
194
195 static int
196 dmopen(struct dev_open_args *ap)
197 {
198         cdev_t dev = ap->a_head.a_dev;
199         dm_dev_t *dmv;
200
201         /* Shortcut for the control device */
202         if (minor(dev) == 0)
203                 return 0;
204
205         if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL)
206                 return ENXIO;
207
208         dmv->is_open = 1;
209         dm_dev_unbusy(dmv);
210
211         aprint_debug("dm open routine called %" PRIu32 "\n",
212             minor(ap->a_head.a_dev));
213         return 0;
214 }
215
216 static int
217 dmclose(struct dev_close_args *ap)
218 {
219         cdev_t dev = ap->a_head.a_dev;
220         dm_dev_t *dmv;
221
222         /* Shortcut for the control device */
223         if (minor(dev) == 0)
224                 return 0;
225
226         if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL)
227                 return ENXIO;
228
229         dmv->is_open = 0;
230         dm_dev_unbusy(dmv);
231
232         aprint_debug("dm close routine called %" PRIu32 "\n",
233             minor(ap->a_head.a_dev));
234         return 0;
235 }
236
237
238 static int
239 dmioctl(struct dev_ioctl_args *ap)
240 {
241         cdev_t dev = ap->a_head.a_dev;
242         u_long cmd = ap->a_cmd;
243         void *data = ap->a_data;
244
245         int r, err;
246         prop_dictionary_t dm_dict_in;
247
248         err = r = 0;
249
250         aprint_debug("dmioctl called\n");
251
252         KKASSERT(data != NULL);
253
254         if (( r = disk_ioctl_switch(dev, cmd, data)) == ENOTTY) {
255                 struct plistref *pref = (struct plistref *) data;
256
257                 /* Check if we were called with NETBSD_DM_IOCTL ioctl
258                    otherwise quit. */
259                 if ((r = dm_ioctl_switch(cmd)) != 0)
260                         return r;
261
262                 if((r = prop_dictionary_copyin_ioctl(pref, cmd, &dm_dict_in)) != 0)
263                         return r;
264
265                 if ((r = dm_check_version(dm_dict_in)) != 0)
266                         goto cleanup_exit;
267
268                 /* run ioctl routine */
269                 if ((err = dm_cmd_to_fun(dm_dict_in)) != 0)
270                         goto cleanup_exit;
271
272 cleanup_exit:
273                 r = prop_dictionary_copyout_ioctl(pref, cmd, dm_dict_in);
274                 prop_object_release(dm_dict_in);
275         }
276
277         /*
278          * Return the error of the actual command if one one has
279          * happened. Otherwise return 'r' which indicates errors
280          * that occurred during helper operations.
281          */
282         return (err != 0)?err:r;
283 }
284
285 /*
286  * Translate command sent from libdevmapper to func.
287  */
288 static int
289 dm_cmd_to_fun(prop_dictionary_t dm_dict){
290         int i, r;
291         prop_string_t command;
292
293         r = 0;
294
295         if ((command = prop_dictionary_get(dm_dict, DM_IOCTL_COMMAND)) == NULL)
296                 return EINVAL;
297
298         for(i = 0; cmd_fn[i].cmd != NULL; i++)
299                 if (prop_string_equals_cstring(command, cmd_fn[i].cmd))
300                         break;
301
302         if (cmd_fn[i].cmd == NULL)
303                 return EINVAL;
304
305         aprint_debug("ioctl %s called\n", cmd_fn[i].cmd);
306         r = cmd_fn[i].fn(dm_dict);
307
308         return r;
309 }
310
311 /* Call apropriate ioctl handler function. */
312 static int
313 dm_ioctl_switch(u_long cmd)
314 {
315
316         switch(cmd) {
317
318         case NETBSD_DM_IOCTL:
319                 aprint_debug("dm NetBSD_DM_IOCTL called\n");
320                 break;
321         default:
322                  aprint_debug("dm unknown ioctl called\n");
323                  return ENOTTY;
324                  break; /* NOT REACHED */
325         }
326
327          return 0;
328 }
329
330  /*
331   * Check for disk specific ioctls.
332   */
333
334 static int
335 disk_ioctl_switch(cdev_t dev, u_long cmd, void *data)
336 {
337         dm_dev_t *dmv;
338
339         /* disk ioctls make sense only on block devices */
340         if (minor(dev) == 0)
341                 return ENOTTY;
342
343         switch(cmd) {
344         case DIOCGPART:
345         {
346                 struct partinfo *dpart;
347                 u_int64_t size;
348                 dpart = (void *)data;
349                 bzero(dpart, sizeof(*dpart));
350
351                 if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL)
352                         return ENODEV;
353                 if (dmv->diskp->d_info.d_media_blksize == 0) {
354                         dm_dev_unbusy(dmv);
355                         return ENOTSUP;
356                 } else {
357                         size = dm_table_size(&dmv->table_head);
358                         dpart->media_offset  = 0;
359                         dpart->media_size    = size * DEV_BSIZE;
360                         dpart->media_blocks  = size;
361                         dpart->media_blksize = DEV_BSIZE;
362                         dpart->fstype = FS_BSDFFS;
363                 }
364                 dm_dev_unbusy(dmv);
365                 break;
366         }
367
368         default:
369                 aprint_debug("unknown disk_ioctl called\n");
370                 return ENOTTY;
371                 break; /* NOT REACHED */
372         }
373
374         return 0;
375 }
376
377 /*
378  * Do all IO operations on dm logical devices.
379  */
380 static int
381 dmstrategy(struct dev_strategy_args *ap)
382 {
383         cdev_t dev = ap->a_head.a_dev;
384         struct bio *bio = ap->a_bio;
385         struct buf *bp = bio->bio_buf;
386         int bypass;
387
388         dm_dev_t *dmv;
389         dm_table_t  *tbl;
390         dm_table_entry_t *table_en;
391         struct buf *nestbuf;
392
393         uint32_t dev_type;
394
395         uint64_t buf_start, buf_len, issued_len;
396         uint64_t table_start, table_end;
397         uint64_t start, end;
398
399         buf_start = bio->bio_offset;
400         buf_len = bp->b_bcount;
401
402         tbl = NULL;
403
404         table_end = 0;
405         dev_type = 0;
406         issued_len = 0;
407
408         if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL) {
409                 bp->b_error = EIO;
410                 bp->b_resid = bp->b_bcount;
411                 biodone(bio);
412                 return 0;
413         }
414
415         switch(bp->b_cmd) {
416         case BUF_CMD_READ:
417         case BUF_CMD_WRITE:
418         case BUF_CMD_FREEBLKS:
419                 bypass = 0;
420                 break;
421         case BUF_CMD_FLUSH:
422                 bypass = 1;
423                 KKASSERT(buf_len == 0);
424                 break;
425         default:
426                 dm_dev_unbusy(dmv);
427                 bp->b_error = EIO;
428                 bp->b_resid = bp->b_bcount;
429                 biodone(bio);
430                 return 0;
431         }
432
433         if (bypass == 0 &&
434             bounds_check_with_mediasize(bio, DEV_BSIZE,
435                                         dm_table_size(&dmv->table_head)) <= 0) {
436                 dm_dev_unbusy(dmv);
437                 bp->b_resid = bp->b_bcount;
438                 biodone(bio);
439                 return 0;
440         }
441
442         /* Select active table */
443         tbl = dm_table_get_entry(&dmv->table_head, DM_TABLE_ACTIVE);
444
445         nestiobuf_init(bio);
446         devstat_start_transaction(&dmv->stats);
447
448         /*
449          * Find out what tables I want to select.
450          */
451         SLIST_FOREACH(table_en, tbl, next) {
452                 /*
453                  * I need need number of bytes not blocks.
454                  */
455                 table_start = table_en->start * DEV_BSIZE;
456                 table_end = table_start + (table_en->length) * DEV_BSIZE;
457
458                 /*
459                  * Calculate the start and end
460                  */
461                 start = MAX(table_start, buf_start);
462                 end = MIN(table_end, buf_start + buf_len);
463
464                 aprint_debug("----------------------------------------\n");
465                 aprint_debug("table_start %010" PRIu64", table_end %010"
466                     PRIu64 "\n", table_start, table_end);
467                 aprint_debug("buf_start %010" PRIu64", buf_len %010"
468                     PRIu64"\n", buf_start, buf_len);
469                 aprint_debug("start-buf_start %010"PRIu64", end %010"
470                     PRIu64"\n", start - buf_start, end);
471                 aprint_debug("start %010" PRIu64" , end %010"
472                     PRIu64"\n", start, end);
473                 aprint_debug("\n----------------------------------------\n");
474
475                 if (bypass) {
476                         nestbuf = getpbuf(NULL);
477                         nestbuf->b_flags |= bio->bio_buf->b_flags & B_HASBOGUS;
478
479                         nestiobuf_add(bio, nestbuf, 0, 0, &dmv->stats);
480                         nestbuf->b_bio1.bio_offset = 0;
481                         table_en->target->strategy(table_en, nestbuf);
482                 } else if (start < end) {
483                         nestbuf = getpbuf(NULL);
484                         nestbuf->b_flags |= bio->bio_buf->b_flags & B_HASBOGUS;
485
486                         nestiobuf_add(bio, nestbuf,
487                                       start - buf_start, (end - start),
488                                       &dmv->stats);
489                         issued_len += end - start;
490
491                         nestbuf->b_bio1.bio_offset = (start - table_start);
492                         table_en->target->strategy(table_en, nestbuf);
493                 }
494         }
495
496         if (issued_len < buf_len)
497                 nestiobuf_error(bio, EINVAL);
498         nestiobuf_start(bio);
499         dm_table_release(&dmv->table_head, DM_TABLE_ACTIVE);
500         dm_dev_unbusy(dmv);
501
502         return 0;
503 }
504
505 static int
506 dmdump(struct dev_dump_args *ap)
507 {
508         cdev_t dev = ap->a_head.a_dev;
509         dm_dev_t *dmv;
510         dm_table_t  *tbl;
511         dm_table_entry_t *table_en;
512         uint32_t dev_type;
513         uint64_t buf_start, buf_len, issued_len;
514         uint64_t table_start, table_end;
515         uint64_t start, end, data_offset;
516         off_t offset;
517         size_t length;
518         int error = 0;
519
520         buf_start = ap->a_offset;
521         buf_len = ap->a_length;
522
523         tbl = NULL;
524
525         table_end = 0;
526         dev_type = 0;
527         issued_len = 0;
528
529         if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL) {
530                 return EIO;
531         }
532
533         /* Select active table */
534         tbl = dm_table_get_entry(&dmv->table_head, DM_TABLE_ACTIVE);
535
536
537         /*
538          * Find out what tables I want to select.
539          */
540         SLIST_FOREACH(table_en, tbl, next) {
541                 /*
542                  * I need need number of bytes not blocks.
543                  */
544                 table_start = table_en->start * DEV_BSIZE;
545                 table_end = table_start + (table_en->length) * DEV_BSIZE;
546
547                 /*
548                  * Calculate the start and end
549                  */
550                 start = MAX(table_start, buf_start);
551                 end = MIN(table_end, buf_start + buf_len);
552
553                 if (ap->a_length == 0) {
554                         if (table_en->target->dump == NULL) {
555                                 error = ENXIO;
556                                 goto out;
557                         }
558
559                         table_en->target->dump(table_en, NULL, 0, 0);
560                 } else if (start < end) {
561                         data_offset = start - buf_start;
562                         offset = start - table_start;
563                         length = end - start;
564
565                         if (table_en->target->dump == NULL) {
566                                 error = ENXIO;
567                                 goto out;
568                         }
569
570                         table_en->target->dump(table_en,
571                             (char *)ap->a_virtual + data_offset,
572                             length, offset);
573
574                         issued_len += end - start;
575                 }
576         }
577
578         if (issued_len < buf_len)
579                 error = EINVAL;
580
581 out:
582         dm_table_release(&dmv->table_head, DM_TABLE_ACTIVE);
583         dm_dev_unbusy(dmv);
584
585         return error;
586 }
587
588 static int
589 dmsize(struct dev_psize_args *ap)
590 {
591         cdev_t dev = ap->a_head.a_dev;
592         dm_dev_t *dmv;
593         uint64_t size;
594
595         size = 0;
596
597         if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL)
598                         return ENOENT;
599
600         size = dm_table_size(&dmv->table_head);
601         dm_dev_unbusy(dmv);
602
603         ap->a_result = (int64_t)size;
604
605         return 0;
606 }
607
608 #if 0
609 static void
610 dmminphys(struct buf *bp)
611 {
612
613         bp->b_bcount = MIN(bp->b_bcount, MAXPHYS);
614 }
615 #endif
616
617 void
618 dmsetdiskinfo(struct disk *disk, dm_table_head_t *head)
619 {
620         struct disk_info info;
621         uint64_t dmp_size;
622
623         dmp_size = dm_table_size(head);
624
625         bzero(&info, sizeof(struct disk_info));
626         info.d_media_blksize = DEV_BSIZE;
627         info.d_media_blocks = dmp_size;
628 #if 0
629         /* this is set by disk_setdiskinfo */
630         info.d_media_size = dmp_size * DEV_BSIZE;
631 #endif
632         info.d_dsflags = DSO_MBRQUIET | DSO_DEVICEMAPPER;
633
634         info.d_secpertrack = 32;
635         info.d_nheads = 64;
636         info.d_secpercyl = info.d_secpertrack * info.d_nheads;
637         info.d_ncylinders = dmp_size / info.d_secpercyl;
638
639         disk_setdiskinfo(disk, &info);
640 }
641
642 /*
643  * Transform char s to uint64_t offset number.
644  */
645 uint64_t
646 atoi64(const char *s)
647 {
648         uint64_t n;
649         n = 0;
650
651         while (*s != '\0') {
652                 if (!isdigit(*s))
653                         break;
654
655                 n = (10 * n) + (*s - '0');
656                 s++;
657         }
658
659         return n;
660 }
661
662 void
663 dm_builtin_init(void *arg)
664 {
665         modeventhand_t evh = (modeventhand_t)arg;
666
667         KKASSERT(evh != NULL);
668         evh(NULL, MOD_LOAD, NULL);
669 }
670
671 void
672 dm_builtin_uninit(void *arg)
673 {
674         modeventhand_t evh = (modeventhand_t)arg;
675
676         KKASSERT(evh != NULL);
677         evh(NULL, MOD_UNLOAD, NULL);
678 }
679
680 TUNABLE_INT("debug.dm_debug", &dm_debug_level);
681 SYSCTL_INT(_debug, OID_AUTO, dm_debug, CTLFLAG_RW, &dm_debug_level,
682                0, "Eanble device mapper debugging");
683