kernel -- MPIPE: Add a constructor argument and priv ptr.
[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 #include <sys/ctype.h>
39
40 #include <sys/buf.h>
41 #include <sys/conf.h>
42 #include <sys/device.h>
43 #include <sys/disk.h>
44 #include <sys/disklabel.h>
45 #include <sys/dtype.h>
46 #include <sys/ioccom.h>
47 #include <sys/malloc.h>
48 #include <sys/module.h>
49 #include <sys/sysctl.h>
50 #include <dev/disk/dm/dm.h>
51
52 #include "netbsd-dm.h"
53
54 static  d_ioctl_t       dmioctl;
55 static  d_open_t        dmopen;
56 static  d_close_t       dmclose;
57 static  d_psize_t       dmsize;
58 static  d_strategy_t    dmstrategy;
59 static  d_dump_t        dmdump;
60
61 /* attach and detach routines */
62 void dmattach(int);
63 static int dm_modcmd(module_t mod, int cmd, void *unused);
64 static int dmdestroy(void);
65
66 static void dm_doinit(void);
67
68 static int dm_cmd_to_fun(prop_dictionary_t);
69 static int disk_ioctl_switch(cdev_t, u_long, void *);
70 static int dm_ioctl_switch(u_long);
71 #if 0
72 static void dmminphys(struct buf *);
73 #endif
74
75 /* ***Variable-definitions*** */
76 struct dev_ops dm_ops = {
77         { "dm", 0, D_DISK | D_MPSAFE },
78         .d_open         = dmopen,
79         .d_close        = dmclose,
80         .d_read         = physread,
81         .d_write        = physwrite,
82         .d_ioctl        = dmioctl,
83         .d_strategy     = dmstrategy,
84         .d_psize        = dmsize,
85         .d_dump         = dmdump,
86 /* D_DISK */
87 };
88
89 MALLOC_DEFINE(M_DM, "dm", "Device Mapper allocations");
90
91 int dm_debug_level = 0;
92
93 extern uint64_t dm_dev_counter;
94
95 static cdev_t dmcdev;
96
97 static moduledata_t dm_mod = {
98     "dm",
99     dm_modcmd,
100     NULL
101 };
102 DECLARE_MODULE(dm, dm_mod, SI_SUB_RAID, SI_ORDER_ANY);
103 MODULE_VERSION(dm, 1);
104
105 /*
106  * This array is used to translate cmd to function pointer.
107  *
108  * Interface between libdevmapper and lvm2tools uses different
109  * names for one IOCTL call because libdevmapper do another thing
110  * then. When I run "info" or "mknodes" libdevmapper will send same
111  * ioctl to kernel but will do another things in userspace.
112  *
113  */
114 static struct cmd_function cmd_fn[] = {
115                 { .cmd = "version", .fn = dm_get_version_ioctl},
116                 { .cmd = "targets", .fn = dm_list_versions_ioctl},
117                 { .cmd = "create",  .fn = dm_dev_create_ioctl},
118                 { .cmd = "info",    .fn = dm_dev_status_ioctl},
119                 { .cmd = "mknodes", .fn = dm_dev_status_ioctl},
120                 { .cmd = "names",   .fn = dm_dev_list_ioctl},
121                 { .cmd = "suspend", .fn = dm_dev_suspend_ioctl},
122                 { .cmd = "remove",  .fn = dm_dev_remove_ioctl},
123                 { .cmd = "remove_all", .fn = dm_dev_remove_all_ioctl},
124                 { .cmd = "rename",  .fn = dm_dev_rename_ioctl},
125                 { .cmd = "resume",  .fn = dm_dev_resume_ioctl},
126                 { .cmd = "clear",   .fn = dm_table_clear_ioctl},
127                 { .cmd = "deps",    .fn = dm_table_deps_ioctl},
128                 { .cmd = "reload",  .fn = dm_table_load_ioctl},
129                 { .cmd = "status",  .fn = dm_table_status_ioctl},
130                 { .cmd = "table",   .fn = dm_table_status_ioctl},
131                 {NULL, NULL}
132 };
133
134 /* New module handle routine */
135 static int
136 dm_modcmd(module_t mod, int cmd, void *unused)
137 {
138         int error, bmajor, cmajor;
139
140         error = 0;
141         bmajor = -1;
142         cmajor = -1;
143
144         switch (cmd) {
145         case MOD_LOAD:
146                 dm_doinit();
147                 kprintf("Device Mapper version %d.%d.%d loaded\n",
148                     DM_VERSION_MAJOR, DM_VERSION_MINOR, DM_VERSION_PATCHLEVEL);
149                 break;
150
151         case MOD_UNLOAD:
152                 /*
153                  * Disable unloading of dm module if there are any devices
154                  * defined in driver. This is probably too strong we need
155                  * to disable auto-unload only if there is mounted dm device
156                  * present.
157                  */
158                 if (dm_dev_counter > 0)
159                         return EBUSY;
160
161                 error = dmdestroy();
162                 if (error)
163                         break;
164                 kprintf("Device Mapper unloaded\n");
165                 break;
166
167         default:
168                 break;
169         }
170
171         return error;
172 }
173
174 static void
175 dm_doinit(void)
176 {
177         dm_target_init();
178         dm_dev_init();
179         dm_pdev_init();
180         dmcdev = make_dev(&dm_ops, 0, UID_ROOT, GID_OPERATOR, 0640, "mapper/control");
181 }
182
183 /* Destroy routine */
184 static int
185 dmdestroy(void)
186 {
187         destroy_dev(dmcdev);
188
189         dm_dev_uninit();
190         dm_pdev_uninit();
191         dm_target_uninit();
192
193         return 0;
194 }
195
196 static int
197 dmopen(struct dev_open_args *ap)
198 {
199         cdev_t dev = ap->a_head.a_dev;
200         dm_dev_t *dmv;
201
202         /* Shortcut for the control device */
203         if (minor(dev) == 0)
204                 return 0;
205
206         if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL)
207                 return ENXIO;
208
209         dmv->is_open = 1;
210         dm_dev_unbusy(dmv);
211
212         aprint_debug("dm open routine called %" PRIu32 "\n",
213             minor(ap->a_head.a_dev));
214         return 0;
215 }
216
217 static int
218 dmclose(struct dev_close_args *ap)
219 {
220         cdev_t dev = ap->a_head.a_dev;
221         dm_dev_t *dmv;
222
223         /* Shortcut for the control device */
224         if (minor(dev) == 0)
225                 return 0;
226
227         if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL)
228                 return ENXIO;
229
230         dmv->is_open = 0;
231         dm_dev_unbusy(dmv);
232
233         aprint_debug("dm close routine called %" PRIu32 "\n",
234             minor(ap->a_head.a_dev));
235         return 0;
236 }
237
238
239 static int
240 dmioctl(struct dev_ioctl_args *ap)
241 {
242         cdev_t dev = ap->a_head.a_dev;
243         u_long cmd = ap->a_cmd;
244         void *data = ap->a_data;
245
246         int r, err;
247         prop_dictionary_t dm_dict_in;
248
249         err = r = 0;
250
251         aprint_debug("dmioctl called\n");
252
253         KKASSERT(data != NULL);
254
255         if (( r = disk_ioctl_switch(dev, cmd, data)) == ENOTTY) {
256                 struct plistref *pref = (struct plistref *) data;
257
258                 /* Check if we were called with NETBSD_DM_IOCTL ioctl
259                    otherwise quit. */
260                 if ((r = dm_ioctl_switch(cmd)) != 0)
261                         return r;
262
263                 if((r = prop_dictionary_copyin_ioctl(pref, cmd, &dm_dict_in)) != 0)
264                         return r;
265
266                 if ((r = dm_check_version(dm_dict_in)) != 0)
267                         goto cleanup_exit;
268
269                 /* run ioctl routine */
270                 if ((err = dm_cmd_to_fun(dm_dict_in)) != 0)
271                         goto cleanup_exit;
272
273 cleanup_exit:
274                 r = prop_dictionary_copyout_ioctl(pref, cmd, dm_dict_in);
275                 prop_object_release(dm_dict_in);
276         }
277
278         /*
279          * Return the error of the actual command if one one has
280          * happened. Otherwise return 'r' which indicates errors
281          * that occurred during helper operations.
282          */
283         return (err != 0)?err:r;
284 }
285
286 /*
287  * Translate command sent from libdevmapper to func.
288  */
289 static int
290 dm_cmd_to_fun(prop_dictionary_t dm_dict){
291         int i, r;
292         prop_string_t command;
293
294         r = 0;
295
296         if ((command = prop_dictionary_get(dm_dict, DM_IOCTL_COMMAND)) == NULL)
297                 return EINVAL;
298
299         for(i = 0; cmd_fn[i].cmd != NULL; i++)
300                 if (prop_string_equals_cstring(command, cmd_fn[i].cmd))
301                         break;
302
303         if (cmd_fn[i].cmd == NULL)
304                 return EINVAL;
305
306         aprint_debug("ioctl %s called\n", cmd_fn[i].cmd);
307         r = cmd_fn[i].fn(dm_dict);
308
309         return r;
310 }
311
312 /* Call apropriate ioctl handler function. */
313 static int
314 dm_ioctl_switch(u_long cmd)
315 {
316
317         switch(cmd) {
318
319         case NETBSD_DM_IOCTL:
320                 aprint_debug("dm NetBSD_DM_IOCTL called\n");
321                 break;
322         default:
323                  aprint_debug("dm unknown ioctl called\n");
324                  return ENOTTY;
325                  break; /* NOT REACHED */
326         }
327
328          return 0;
329 }
330
331  /*
332   * Check for disk specific ioctls.
333   */
334
335 static int
336 disk_ioctl_switch(cdev_t dev, u_long cmd, void *data)
337 {
338         dm_dev_t *dmv;
339
340         /* disk ioctls make sense only on block devices */
341         if (minor(dev) == 0)
342                 return ENOTTY;
343
344         switch(cmd) {
345         case DIOCGPART:
346         {
347                 struct partinfo *dpart;
348                 u_int64_t size;
349                 dpart = (void *)data;
350                 bzero(dpart, sizeof(*dpart));
351
352                 if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL)
353                         return ENODEV;
354                 if (dmv->diskp->d_info.d_media_blksize == 0) {
355                         dm_dev_unbusy(dmv);
356                         return ENOTSUP;
357                 } else {
358                         size = dm_table_size(&dmv->table_head);
359                         dpart->media_offset  = 0;
360                         dpart->media_size    = size * DEV_BSIZE;
361                         dpart->media_blocks  = size;
362                         dpart->media_blksize = DEV_BSIZE;
363                         dpart->fstype = FS_BSDFFS;
364                 }
365                 dm_dev_unbusy(dmv);
366                 break;
367         }
368
369         default:
370                 aprint_debug("unknown disk_ioctl called\n");
371                 return ENOTTY;
372                 break; /* NOT REACHED */
373         }
374
375         return 0;
376 }
377
378 /*
379  * Do all IO operations on dm logical devices.
380  */
381 static int
382 dmstrategy(struct dev_strategy_args *ap)
383 {
384         cdev_t dev = ap->a_head.a_dev;
385         struct bio *bio = ap->a_bio;
386         struct buf *bp = bio->bio_buf;
387         int bypass;
388
389         dm_dev_t *dmv;
390         dm_table_t  *tbl;
391         dm_table_entry_t *table_en;
392         struct buf *nestbuf;
393
394         uint32_t dev_type;
395
396         uint64_t buf_start, buf_len, issued_len;
397         uint64_t table_start, table_end;
398         uint64_t start, end;
399
400         buf_start = bio->bio_offset;
401         buf_len = bp->b_bcount;
402
403         tbl = NULL;
404
405         table_end = 0;
406         dev_type = 0;
407         issued_len = 0;
408
409         if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL) {
410                 bp->b_error = EIO;
411                 bp->b_resid = bp->b_bcount;
412                 biodone(bio);
413                 return 0;
414         }
415
416         switch(bp->b_cmd) {
417         case BUF_CMD_READ:
418         case BUF_CMD_WRITE:
419         case BUF_CMD_FREEBLKS:
420                 bypass = 0;
421                 break;
422         case BUF_CMD_FLUSH:
423                 bypass = 1;
424                 KKASSERT(buf_len == 0);
425                 break;
426         default:
427                 dm_dev_unbusy(dmv);
428                 bp->b_error = EIO;
429                 bp->b_resid = bp->b_bcount;
430                 biodone(bio);
431                 return 0;
432         }
433
434         if (bypass == 0 &&
435             bounds_check_with_mediasize(bio, DEV_BSIZE,
436                                         dm_table_size(&dmv->table_head)) <= 0) {
437                 dm_dev_unbusy(dmv);
438                 bp->b_resid = bp->b_bcount;
439                 biodone(bio);
440                 return 0;
441         }
442
443         /* Select active table */
444         tbl = dm_table_get_entry(&dmv->table_head, DM_TABLE_ACTIVE);
445
446         nestiobuf_init(bio);
447         devstat_start_transaction(&dmv->stats);
448
449         /*
450          * Find out what tables I want to select.
451          */
452         SLIST_FOREACH(table_en, tbl, next) {
453                 /*
454                  * I need need number of bytes not blocks.
455                  */
456                 table_start = table_en->start * DEV_BSIZE;
457                 table_end = table_start + (table_en->length) * DEV_BSIZE;
458
459                 /*
460                  * Calculate the start and end
461                  */
462                 start = MAX(table_start, buf_start);
463                 end = MIN(table_end, buf_start + buf_len);
464
465                 aprint_debug("----------------------------------------\n");
466                 aprint_debug("table_start %010" PRIu64", table_end %010"
467                     PRIu64 "\n", table_start, table_end);
468                 aprint_debug("buf_start %010" PRIu64", buf_len %010"
469                     PRIu64"\n", buf_start, buf_len);
470                 aprint_debug("start-buf_start %010"PRIu64", end %010"
471                     PRIu64"\n", start - buf_start, end);
472                 aprint_debug("start %010" PRIu64" , end %010"
473                     PRIu64"\n", start, end);
474                 aprint_debug("\n----------------------------------------\n");
475
476                 if (bypass) {
477                         nestbuf = getpbuf(NULL);
478                         nestbuf->b_flags |= bio->bio_buf->b_flags & B_HASBOGUS;
479
480                         nestiobuf_add(bio, nestbuf, 0, 0, &dmv->stats);
481                         nestbuf->b_bio1.bio_offset = 0;
482                         table_en->target->strategy(table_en, nestbuf);
483                 } else if (start < end) {
484                         nestbuf = getpbuf(NULL);
485                         nestbuf->b_flags |= bio->bio_buf->b_flags & B_HASBOGUS;
486
487                         nestiobuf_add(bio, nestbuf,
488                                       start - buf_start, (end - start),
489                                       &dmv->stats);
490                         issued_len += end - start;
491
492                         nestbuf->b_bio1.bio_offset = (start - table_start);
493                         table_en->target->strategy(table_en, nestbuf);
494                 }
495         }
496
497         if (issued_len < buf_len)
498                 nestiobuf_error(bio, EINVAL);
499         nestiobuf_start(bio);
500         dm_table_release(&dmv->table_head, DM_TABLE_ACTIVE);
501         dm_dev_unbusy(dmv);
502
503         return 0;
504 }
505
506 static int
507 dmdump(struct dev_dump_args *ap)
508 {
509         cdev_t dev = ap->a_head.a_dev;
510         dm_dev_t *dmv;
511         dm_table_t  *tbl;
512         dm_table_entry_t *table_en;
513         uint32_t dev_type;
514         uint64_t buf_start, buf_len, issued_len;
515         uint64_t table_start, table_end;
516         uint64_t start, end, data_offset;
517         off_t offset;
518         size_t length;
519         int error = 0;
520
521         buf_start = ap->a_offset;
522         buf_len = ap->a_length;
523
524         tbl = NULL;
525
526         table_end = 0;
527         dev_type = 0;
528         issued_len = 0;
529
530         if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL) {
531                 return EIO;
532         }
533
534         /* Select active table */
535         tbl = dm_table_get_entry(&dmv->table_head, DM_TABLE_ACTIVE);
536
537
538         /*
539          * Find out what tables I want to select.
540          */
541         SLIST_FOREACH(table_en, tbl, next) {
542                 /*
543                  * I need need number of bytes not blocks.
544                  */
545                 table_start = table_en->start * DEV_BSIZE;
546                 table_end = table_start + (table_en->length) * DEV_BSIZE;
547
548                 /*
549                  * Calculate the start and end
550                  */
551                 start = MAX(table_start, buf_start);
552                 end = MIN(table_end, buf_start + buf_len);
553
554                 if (ap->a_length == 0) {
555                         if (table_en->target->dump == NULL) {
556                                 error = ENXIO;
557                                 goto out;
558                         }
559
560                         table_en->target->dump(table_en, NULL, 0, 0);
561                 } else if (start < end) {
562                         data_offset = start - buf_start;
563                         offset = start - table_start;
564                         length = end - start;
565
566                         if (table_en->target->dump == NULL) {
567                                 error = ENXIO;
568                                 goto out;
569                         }
570
571                         table_en->target->dump(table_en,
572                             (char *)ap->a_virtual + data_offset,
573                             length, offset);
574
575                         issued_len += end - start;
576                 }
577         }
578
579         if (issued_len < buf_len)
580                 error = EINVAL;
581
582 out:
583         dm_table_release(&dmv->table_head, DM_TABLE_ACTIVE);
584         dm_dev_unbusy(dmv);
585
586         return error;
587 }
588
589 static int
590 dmsize(struct dev_psize_args *ap)
591 {
592         cdev_t dev = ap->a_head.a_dev;
593         dm_dev_t *dmv;
594         uint64_t size;
595
596         size = 0;
597
598         if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL)
599                         return ENOENT;
600
601         size = dm_table_size(&dmv->table_head);
602         dm_dev_unbusy(dmv);
603
604         ap->a_result = (int64_t)size;
605
606         return 0;
607 }
608
609 #if 0
610 static void
611 dmminphys(struct buf *bp)
612 {
613
614         bp->b_bcount = MIN(bp->b_bcount, MAXPHYS);
615 }
616 #endif
617
618 void
619 dmsetdiskinfo(struct disk *disk, dm_table_head_t *head)
620 {
621         struct disk_info info;
622         uint64_t dmp_size;
623
624         dmp_size = dm_table_size(head);
625
626         bzero(&info, sizeof(struct disk_info));
627         info.d_media_blksize = DEV_BSIZE;
628         info.d_media_blocks = dmp_size;
629 #if 0
630         /* this is set by disk_setdiskinfo */
631         info.d_media_size = dmp_size * DEV_BSIZE;
632 #endif
633         info.d_dsflags = DSO_MBRQUIET | DSO_DEVICEMAPPER;
634
635         info.d_secpertrack = 32;
636         info.d_nheads = 64;
637         info.d_secpercyl = info.d_secpertrack * info.d_nheads;
638         info.d_ncylinders = dmp_size / info.d_secpercyl;
639
640         disk_setdiskinfo(disk, &info);
641 }
642
643 /*
644  * Transform char s to uint64_t offset number.
645  */
646 uint64_t
647 atoi64(const char *s)
648 {
649         uint64_t n;
650         n = 0;
651
652         while (*s != '\0') {
653                 if (!isdigit(*s))
654                         break;
655
656                 n = (10 * n) + (*s - '0');
657                 s++;
658         }
659
660         return n;
661 }
662
663 void
664 dm_builtin_init(void *arg)
665 {
666         modeventhand_t evh = (modeventhand_t)arg;
667
668         KKASSERT(evh != NULL);
669         evh(NULL, MOD_LOAD, NULL);
670 }
671
672 void
673 dm_builtin_uninit(void *arg)
674 {
675         modeventhand_t evh = (modeventhand_t)arg;
676
677         KKASSERT(evh != NULL);
678         evh(NULL, MOD_UNLOAD, NULL);
679 }
680
681 TUNABLE_INT("debug.dm_debug", &dm_debug_level);
682 SYSCTL_INT(_debug, OID_AUTO, dm_debug, CTLFLAG_RW, &dm_debug_level,
683                0, "Eanble device mapper debugging");
684