device mapper - implement dumping
[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
50 #include "netbsd-dm.h"
51 #include "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
103 /*
104  * This array is used to translate cmd to function pointer.
105  *
106  * Interface between libdevmapper and lvm2tools uses different 
107  * names for one IOCTL call because libdevmapper do another thing
108  * then. When I run "info" or "mknodes" libdevmapper will send same
109  * ioctl to kernel but will do another things in userspace.
110  *
111  */
112 static struct cmd_function cmd_fn[] = {
113                 { .cmd = "version", .fn = dm_get_version_ioctl},
114                 { .cmd = "targets", .fn = dm_list_versions_ioctl},
115                 { .cmd = "create",  .fn = dm_dev_create_ioctl},
116                 { .cmd = "info",    .fn = dm_dev_status_ioctl},
117                 { .cmd = "mknodes", .fn = dm_dev_status_ioctl},         
118                 { .cmd = "names",   .fn = dm_dev_list_ioctl},
119                 { .cmd = "suspend", .fn = dm_dev_suspend_ioctl},
120                 { .cmd = "remove",  .fn = dm_dev_remove_ioctl}, 
121                 { .cmd = "rename",  .fn = dm_dev_rename_ioctl},
122                 { .cmd = "resume",  .fn = dm_dev_resume_ioctl},
123                 { .cmd = "clear",   .fn = dm_table_clear_ioctl},
124                 { .cmd = "deps",    .fn = dm_table_deps_ioctl},
125                 { .cmd = "reload",  .fn = dm_table_load_ioctl},
126                 { .cmd = "status",  .fn = dm_table_status_ioctl},
127                 { .cmd = "table",   .fn = dm_table_status_ioctl},
128                 {NULL, NULL}    
129 };
130
131 /* New module handle routine */
132 static int
133 dm_modcmd(module_t mod, int cmd, void *unused)
134 {
135         int error, bmajor, cmajor;
136
137         error = 0;
138         bmajor = -1;
139         cmajor = -1;
140
141         switch (cmd) {
142         case MOD_LOAD:
143                 dm_doinit();
144                 kprintf("Device Mapper version %d.%d.%d loaded\n",
145                     DM_VERSION_MAJOR, DM_VERSION_MINOR, DM_VERSION_PATCHLEVEL);
146                 break;
147
148         case MOD_UNLOAD:
149                 /*
150                  * Disable unloading of dm module if there are any devices
151                  * defined in driver. This is probably too strong we need
152                  * to disable auto-unload only if there is mounted dm device
153                  * present.
154                  */ 
155                 if (dm_dev_counter > 0)
156                         return EBUSY;
157
158                 error = dmdestroy();
159                 if (error)
160                         break;
161                 kprintf("Device Mapper unloaded\n");
162                 break;
163
164         default:
165                 break;
166         }
167
168         return error;
169 }
170
171 /*
172  * dm_detach:
173  *
174  *      Autoconfiguration detach function for pseudo-device glue.
175  * This routine is called by dm_ioctl::dm_dev_remove_ioctl and by autoconf to
176  * remove devices created in device-mapper. 
177  */
178 int
179 dm_detach(dm_dev_t *dmv)
180 {
181         disable_dev(dmv);
182
183         /* Destroy active table first.  */
184         dm_table_destroy(&dmv->table_head, DM_TABLE_ACTIVE);
185
186         /* Destroy inactive table if exits, too. */
187         dm_table_destroy(&dmv->table_head, DM_TABLE_INACTIVE);
188         
189         dm_table_head_destroy(&dmv->table_head);
190
191         destroy_dev(dmv->devt);
192
193         /* Destroy device */
194         (void)dm_dev_free(dmv);
195
196         /* Decrement device counter After removing device */
197         --dm_dev_counter; /* XXX: was atomic 64 */
198
199         return 0;
200 }
201
202 static void
203 dm_doinit(void)
204 {
205         dm_target_init();
206         dm_dev_init();
207         dm_pdev_init();
208         dmcdev = make_dev(&dm_ops, 0, UID_ROOT, GID_OPERATOR, 0640, "mapper/control");
209 }
210
211 /* Destroy routine */
212 static int
213 dmdestroy(void)
214 {
215         destroy_dev(dmcdev);
216         
217         dm_dev_destroy();
218         dm_pdev_destroy();
219         dm_target_destroy();
220
221         return 0;
222 }
223
224 static int
225 dmopen(struct dev_open_args *ap)
226 {
227
228         aprint_debug("dm open routine called %" PRIu32 "\n",
229             minor(ap->a_head.a_dev));
230         return 0;
231 }
232
233 static int
234 dmclose(struct dev_close_args *ap)
235 {
236
237         aprint_debug("dm close routine called %" PRIu32 "\n",
238             minor(ap->a_head.a_dev));
239         return 0;
240 }
241
242
243 static int
244 dmioctl(struct dev_ioctl_args *ap)
245 {
246         cdev_t dev = ap->a_head.a_dev;
247         u_long cmd = ap->a_cmd;
248         void *data = ap->a_data;
249
250         int r;
251         prop_dictionary_t dm_dict_in;
252
253         r = 0;
254
255         aprint_debug("dmioctl called\n");
256         
257         KKASSERT(data != NULL);
258         
259         if (( r = disk_ioctl_switch(dev, cmd, data)) == ENOTTY) {
260                 struct plistref *pref = (struct plistref *) data;
261
262                 /* Check if we were called with NETBSD_DM_IOCTL ioctl
263                    otherwise quit. */
264                 if ((r = dm_ioctl_switch(cmd)) != 0)
265                         return r;
266
267                 if((r = prop_dictionary_copyin_ioctl(pref, cmd, &dm_dict_in)) != 0)
268                         return r;
269
270                 if ((r = dm_check_version(dm_dict_in)) != 0)
271                         goto cleanup_exit;
272
273                 /* run ioctl routine */
274                 if ((r = dm_cmd_to_fun(dm_dict_in)) != 0)
275                         goto cleanup_exit;
276
277 cleanup_exit:
278                 r = prop_dictionary_copyout_ioctl(pref, cmd, dm_dict_in);
279                 prop_object_release(dm_dict_in);
280         }
281
282         return 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
447         /*
448          * Find out what tables I want to select.
449          */
450         SLIST_FOREACH(table_en, tbl, next) {
451                 /*
452                  * I need need number of bytes not blocks.
453                  */
454                 table_start = table_en->start * DEV_BSIZE;
455                 table_end = table_start + (table_en->length) * DEV_BSIZE;
456
457                 /*
458                  * Calculate the start and end
459                  */
460                 start = MAX(table_start, buf_start);
461                 end = MIN(table_end, buf_start + buf_len);
462
463                 aprint_debug("----------------------------------------\n");
464                 aprint_debug("table_start %010" PRIu64", table_end %010"
465                     PRIu64 "\n", table_start, table_end);
466                 aprint_debug("buf_start %010" PRIu64", buf_len %010"
467                     PRIu64"\n", buf_start, buf_len);
468                 aprint_debug("start-buf_start %010"PRIu64", end %010"
469                     PRIu64"\n", start - buf_start, end);
470                 aprint_debug("start %010" PRIu64" , end %010"
471                     PRIu64"\n", start, end);
472                 aprint_debug("\n----------------------------------------\n");
473
474                 if (bypass) {
475                         nestbuf = getpbuf(NULL);
476                         nestbuf->b_flags |= bio->bio_buf->b_flags & B_HASBOGUS;
477
478                         nestiobuf_add(bio, nestbuf, 0, 0);
479                         nestbuf->b_bio1.bio_offset = 0;
480                         table_en->target->strategy(table_en, nestbuf);
481                 } else if (start < end) {
482                         nestbuf = getpbuf(NULL);
483                         nestbuf->b_flags |= bio->bio_buf->b_flags & B_HASBOGUS;
484
485                         nestiobuf_add(bio, nestbuf,
486                                       start - buf_start, (end - start));
487                         issued_len += end - start;
488
489                         nestbuf->b_bio1.bio_offset = (start - table_start);
490                         table_en->target->strategy(table_en, nestbuf);
491                 }
492         }
493
494         if (issued_len < buf_len)
495                 nestiobuf_error(bio, EINVAL);
496         nestiobuf_start(bio);
497         dm_table_release(&dmv->table_head, DM_TABLE_ACTIVE);
498         dm_dev_unbusy(dmv);
499
500         return 0;
501 }
502
503 static int
504 dmdump(struct dev_dump_args *ap)
505 {
506         cdev_t dev = ap->a_head.a_dev;
507         dm_dev_t *dmv;
508         dm_table_t  *tbl;
509         dm_table_entry_t *table_en;
510         uint32_t dev_type;
511         uint64_t buf_start, buf_len, issued_len;
512         uint64_t table_start, table_end;
513         uint64_t start, end, data_offset;
514         off_t offset;
515         size_t length;
516         int error = 0;
517
518         buf_start = ap->a_offset;
519         buf_len = ap->a_length;
520
521         tbl = NULL; 
522
523         table_end = 0;
524         dev_type = 0;
525         issued_len = 0;
526
527         if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL) {
528                 return EIO;
529         } 
530
531         /* Select active table */
532         tbl = dm_table_get_entry(&dmv->table_head, DM_TABLE_ACTIVE);
533
534
535         /*
536          * Find out what tables I want to select.
537          */
538         SLIST_FOREACH(table_en, tbl, next) {
539                 /*
540                  * I need need number of bytes not blocks.
541                  */
542                 table_start = table_en->start * DEV_BSIZE;
543                 table_end = table_start + (table_en->length) * DEV_BSIZE;
544
545                 /*
546                  * Calculate the start and end
547                  */
548                 start = MAX(table_start, buf_start);
549                 end = MIN(table_end, buf_start + buf_len);
550
551                 if (ap->a_length == 0) {
552                         if (table_en->target->dump == NULL) {
553                                 error = ENXIO;
554                                 goto out;
555                         }
556
557                         table_en->target->dump(table_en, NULL, 0, 0);
558                 } else if (start < end) {
559                         data_offset = start - buf_start;
560                         offset = start - table_start;
561                         length = end - start;
562
563                         if (table_en->target->dump == NULL) {
564                                 error = ENXIO;
565                                 goto out;
566                         }
567
568                         table_en->target->dump(table_en,
569                             (char *)ap->a_virtual + data_offset,
570                             length, offset);
571
572                         issued_len += end - start;
573                 }
574         }
575
576         if (issued_len < buf_len)
577                 error = EINVAL;
578
579 out:
580         dm_table_release(&dmv->table_head, DM_TABLE_ACTIVE);
581         dm_dev_unbusy(dmv);
582
583         return error;
584 }
585
586 static int
587 dmsize(struct dev_psize_args *ap)
588 {
589         cdev_t dev = ap->a_head.a_dev;
590         dm_dev_t *dmv;
591         uint64_t size;
592
593         size = 0;
594
595         if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL)
596                         return ENOENT;
597
598         size = dm_table_size(&dmv->table_head);
599         dm_dev_unbusy(dmv);
600
601         ap->a_result = (int64_t)size;
602
603         return 0;
604 }
605
606 #if 0
607 static void
608 dmminphys(struct buf *bp)
609 {
610
611         bp->b_bcount = MIN(bp->b_bcount, MAXPHYS);
612 }
613 #endif
614
615 void
616 dmsetdiskinfo(struct disk *disk, dm_table_head_t *head)
617 {
618         struct disk_info info;
619         int dmp_size;
620
621         dmp_size = dm_table_size(head);
622
623         bzero(&info, sizeof(struct disk_info));
624         info.d_media_blksize = DEV_BSIZE;
625         info.d_media_blocks = dmp_size;
626         info.d_media_size = dmp_size * DEV_BSIZE;
627         info.d_dsflags = DSO_MBRQUIET; /* XXX */
628         info.d_secpertrack = 32;
629         info.d_nheads = 64;
630         info.d_secpercyl = info.d_secpertrack * info.d_nheads;
631         info.d_ncylinders = dmp_size / 2048;
632         bcopy(&info, &disk->d_info, sizeof(disk->d_info));
633 }
634
635 prop_dictionary_t
636 dmgetdiskinfo(struct disk *disk)
637 {
638         prop_dictionary_t disk_info, geom;
639         struct disk_info *pinfo;
640
641         pinfo = &disk->d_info;
642
643         disk_info = prop_dictionary_create();
644         geom = prop_dictionary_create();
645
646         prop_dictionary_set_cstring_nocopy(disk_info, "type", "ESDI");
647         prop_dictionary_set_uint64(geom, "sectors-per-unit", pinfo->d_media_blocks);
648         prop_dictionary_set_uint32(geom, "sector-size",
649             DEV_BSIZE /* XXX 512? */);
650         prop_dictionary_set_uint32(geom, "sectors-per-track", 32);
651         prop_dictionary_set_uint32(geom, "tracks-per-cylinder", 64);
652         prop_dictionary_set_uint32(geom, "cylinders-per-unit",
653             pinfo->d_media_blocks / 2048);
654         prop_dictionary_set(disk_info, "geometry", geom);
655         prop_object_release(geom);
656
657         return disk_info;
658 }
659
660 void
661 dmgetproperties(struct disk *disk, dm_table_head_t *head)
662 {
663 #if 0
664         prop_dictionary_t disk_info, odisk_info, geom;
665         int dmp_size;
666
667         dmp_size = dm_table_size(head);
668         disk_info = prop_dictionary_create();
669         geom = prop_dictionary_create();
670
671         prop_dictionary_set_cstring_nocopy(disk_info, "type", "ESDI");
672         prop_dictionary_set_uint64(geom, "sectors-per-unit", dmp_size);
673         prop_dictionary_set_uint32(geom, "sector-size",
674             DEV_BSIZE /* XXX 512? */);
675         prop_dictionary_set_uint32(geom, "sectors-per-track", 32);
676         prop_dictionary_set_uint32(geom, "tracks-per-cylinder", 64);
677         prop_dictionary_set_uint32(geom, "cylinders-per-unit", dmp_size / 2048);
678         prop_dictionary_set(disk_info, "geometry", geom);
679         prop_object_release(geom);
680
681         odisk_info = disk->dk_info;
682         disk->dk_info = disk_info;
683
684         if (odisk_info != NULL)
685                 prop_object_release(odisk_info);
686 #endif
687 }
688
689 TUNABLE_INT("debug.dm_debug", &dm_debug_level);
690 SYSCTL_INT(_debug, OID_AUTO, dm_debug, CTLFLAG_RW, &dm_debug_level,
691                0, "Eanble device mapper debugging");
692