Import lvm2 from NetBSD
[dragonfly.git] / contrib / lvm2 / dist / lib / format1 / format1.c
1 /*      $NetBSD: format1.c,v 1.1.1.2 2009/12/02 00:26:48 haad Exp $     */
2
3 /*
4  * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.
5  * Copyright (C) 2004-2007 Red Hat, Inc. All rights reserved.
6  *
7  * This file is part of LVM2.
8  *
9  * This copyrighted material is made available to anyone wishing to use,
10  * modify, copy, or redistribute it subject to the terms and conditions
11  * of the GNU Lesser General Public License v.2.1.
12  *
13  * You should have received a copy of the GNU Lesser General Public License
14  * along with this program; if not, write to the Free Software Foundation,
15  * Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
16  */
17
18 #include "lib.h"
19 #include "disk-rep.h"
20 #include "limits.h"
21 #include "display.h"
22 #include "toolcontext.h"
23 #include "lvm1-label.h"
24 #include "format1.h"
25 #include "segtype.h"
26
27 /* VG consistency checks */
28 static int _check_vgs(struct dm_list *pvs)
29 {
30         struct dm_list *pvh, *t;
31         struct disk_list *dl = NULL;
32         struct disk_list *first = NULL;
33
34         uint32_t pv_count = 0;
35         uint32_t exported = 0;
36         int first_time = 1;
37
38         /*
39          * If there are exported and unexported PVs, ignore exported ones.
40          * This means an active VG won't be affected if disks are inserted
41          * bearing an exported VG with the same name.
42          */
43         dm_list_iterate_items(dl, pvs) {
44                 if (first_time) {
45                         exported = dl->pvd.pv_status & VG_EXPORTED;
46                         first_time = 0;
47                         continue;
48                 }
49
50                 if (exported != (dl->pvd.pv_status & VG_EXPORTED)) {
51                         /* Remove exported PVs */
52                         dm_list_iterate_safe(pvh, t, pvs) {
53                                 dl = dm_list_item(pvh, struct disk_list);
54                                 if (dl->pvd.pv_status & VG_EXPORTED)
55                                         dm_list_del(pvh);
56                         }
57                         break;
58                 }
59         }
60
61         /* Remove any PVs with VG structs that differ from the first */
62         dm_list_iterate_safe(pvh, t, pvs) {
63                 dl = dm_list_item(pvh, struct disk_list);
64
65                 if (!first)
66                         first = dl;
67
68                 else if (memcmp(&first->vgd, &dl->vgd, sizeof(first->vgd))) {
69                         log_error("VG data differs between PVs %s and %s",
70                                   dev_name(first->dev), dev_name(dl->dev));
71                         log_debug("VG data on %s: %s %s %" PRIu32 " %" PRIu32
72                                   "  %" PRIu32 " %" PRIu32 " %" PRIu32 " %"
73                                   PRIu32 " %" PRIu32 " %" PRIu32 " %" PRIu32
74                                   " %" PRIu32 " %" PRIu32 " %" PRIu32 " %"
75                                   PRIu32 " %" PRIu32 " %" PRIu32,
76                                   dev_name(first->dev), first->vgd.vg_uuid,
77                                   first->vgd.vg_name_dummy,
78                                   first->vgd.vg_number, first->vgd.vg_access,
79                                   first->vgd.vg_status, first->vgd.lv_max,
80                                   first->vgd.lv_cur, first->vgd.lv_open,
81                                   first->vgd.pv_max, first->vgd.pv_cur,
82                                   first->vgd.pv_act, first->vgd.dummy,
83                                   first->vgd.vgda, first->vgd.pe_size,
84                                   first->vgd.pe_total, first->vgd.pe_allocated,
85                                   first->vgd.pvg_total);
86                         log_debug("VG data on %s: %s %s %" PRIu32 " %" PRIu32
87                                   "  %" PRIu32 " %" PRIu32 " %" PRIu32 " %"
88                                   PRIu32 " %" PRIu32 " %" PRIu32 " %" PRIu32
89                                   " %" PRIu32 " %" PRIu32 " %" PRIu32 " %"
90                                   PRIu32 " %" PRIu32 " %" PRIu32,
91                                   dev_name(dl->dev), dl->vgd.vg_uuid,
92                                   dl->vgd.vg_name_dummy, dl->vgd.vg_number,
93                                   dl->vgd.vg_access, dl->vgd.vg_status,
94                                   dl->vgd.lv_max, dl->vgd.lv_cur,
95                                   dl->vgd.lv_open, dl->vgd.pv_max,
96                                   dl->vgd.pv_cur, dl->vgd.pv_act, dl->vgd.dummy,
97                                   dl->vgd.vgda, dl->vgd.pe_size,
98                                   dl->vgd.pe_total, dl->vgd.pe_allocated,
99                                   dl->vgd.pvg_total);
100                         dm_list_del(pvh);
101                         return 0;
102                 }
103                 pv_count++;
104         }
105
106         /* On entry to fn, list known to be non-empty */
107         if (pv_count != first->vgd.pv_cur) {
108                 log_error("%d PV(s) found for VG %s: expected %d",
109                           pv_count, first->pvd.vg_name, first->vgd.pv_cur);
110         }
111
112         return 1;
113 }
114
115 static struct volume_group *_build_vg(struct format_instance *fid,
116                                       struct dm_list *pvs,
117                                       struct dm_pool *mem)
118 {
119         struct volume_group *vg = dm_pool_alloc(mem, sizeof(*vg));
120         struct disk_list *dl;
121
122         if (!vg)
123                 goto_bad;
124
125         if (dm_list_empty(pvs))
126                 goto_bad;
127
128         memset(vg, 0, sizeof(*vg));
129
130         vg->cmd = fid->fmt->cmd;
131         vg->vgmem = mem;
132         vg->fid = fid;
133         vg->seqno = 0;
134         dm_list_init(&vg->pvs);
135         dm_list_init(&vg->lvs);
136         dm_list_init(&vg->tags);
137         dm_list_init(&vg->removed_pvs);
138
139         if (!_check_vgs(pvs))
140                 goto_bad;
141
142         dl = dm_list_item(pvs->n, struct disk_list);
143
144         if (!import_vg(mem, vg, dl))
145                 goto_bad;
146
147         if (!import_pvs(fid->fmt, mem, vg, pvs, &vg->pvs, &vg->pv_count))
148                 goto_bad;
149
150         if (!import_lvs(mem, vg, pvs))
151                 goto_bad;
152
153         if (!import_extents(fid->fmt->cmd, vg, pvs))
154                 goto_bad;
155
156         if (!import_snapshots(mem, vg, pvs))
157                 goto_bad;
158
159         return vg;
160
161       bad:
162         dm_pool_free(mem, vg);
163         return NULL;
164 }
165
166 static struct volume_group *_format1_vg_read(struct format_instance *fid,
167                                      const char *vg_name,
168                                      struct metadata_area *mda __attribute((unused)))
169 {
170         struct dm_pool *mem = dm_pool_create("lvm1 vg_read", VG_MEMPOOL_CHUNK);
171         struct dm_list pvs;
172         struct volume_group *vg = NULL;
173         dm_list_init(&pvs);
174
175         if (!mem)
176                 return_NULL;
177
178         /* Strip dev_dir if present */
179         vg_name = strip_dir(vg_name, fid->fmt->cmd->dev_dir);
180
181         if (!read_pvs_in_vg
182             (fid->fmt, vg_name, fid->fmt->cmd->filter, mem, &pvs))
183                 goto_bad;
184
185         if (!(vg = _build_vg(fid, &pvs, mem)))
186                 goto_bad;
187
188         return vg;
189 bad:
190         dm_pool_destroy(mem);
191         return NULL;
192 }
193
194 static struct disk_list *_flatten_pv(struct format_instance *fid,
195                                      struct dm_pool *mem, struct volume_group *vg,
196                                      struct physical_volume *pv,
197                                      const char *dev_dir)
198 {
199         struct disk_list *dl = dm_pool_alloc(mem, sizeof(*dl));
200
201         if (!dl)
202                 return_NULL;
203
204         dl->mem = mem;
205         dl->dev = pv->dev;
206
207         dm_list_init(&dl->uuids);
208         dm_list_init(&dl->lvds);
209
210         if (!export_pv(fid->fmt->cmd, mem, vg, &dl->pvd, pv) ||
211             !export_vg(&dl->vgd, vg) ||
212             !export_uuids(dl, vg) ||
213             !export_lvs(dl, vg, pv, dev_dir) || !calculate_layout(dl)) {
214                 dm_pool_free(mem, dl);
215                 return_NULL;
216         }
217
218         return dl;
219 }
220
221 static int _flatten_vg(struct format_instance *fid, struct dm_pool *mem,
222                        struct volume_group *vg,
223                        struct dm_list *pvds, const char *dev_dir,
224                        struct dev_filter *filter)
225 {
226         struct pv_list *pvl;
227         struct disk_list *data;
228
229         dm_list_iterate_items(pvl, &vg->pvs) {
230                 if (!(data = _flatten_pv(fid, mem, vg, pvl->pv, dev_dir)))
231                         return_0;
232
233                 dm_list_add(pvds, &data->list);
234         }
235
236         export_numbers(pvds, vg);
237         export_pv_act(pvds);
238
239         if (!export_vg_number(fid, pvds, vg->name, filter))
240                 return_0;
241
242         return 1;
243 }
244
245 static int _format1_vg_write(struct format_instance *fid, struct volume_group *vg,
246                      struct metadata_area *mda __attribute((unused)))
247 {
248         struct dm_pool *mem = dm_pool_create("lvm1 vg_write", VG_MEMPOOL_CHUNK);
249         struct dm_list pvds;
250         int r = 0;
251
252         if (!mem)
253                 return_0;
254
255         dm_list_init(&pvds);
256
257         r = (_flatten_vg(fid, mem, vg, &pvds, fid->fmt->cmd->dev_dir,
258                          fid->fmt->cmd->filter) &&
259              write_disks(fid->fmt, &pvds));
260
261         lvmcache_update_vg(vg, 0);
262         dm_pool_destroy(mem);
263         return r;
264 }
265
266 static int _format1_pv_read(const struct format_type *fmt, const char *pv_name,
267                     struct physical_volume *pv, struct dm_list *mdas __attribute((unused)),
268                     int scan_label_only __attribute((unused)))
269 {
270         struct dm_pool *mem = dm_pool_create("lvm1 pv_read", 1024);
271         struct disk_list *dl;
272         struct device *dev;
273         int r = 0;
274
275         log_very_verbose("Reading physical volume data %s from disk", pv_name);
276
277         if (!mem)
278                 return_0;
279
280         if (!(dev = dev_cache_get(pv_name, fmt->cmd->filter)))
281                 goto_out;
282
283         if (!(dl = read_disk(fmt, dev, mem, NULL)))
284                 goto_out;
285
286         if (!import_pv(fmt, fmt->cmd->mem, dl->dev, NULL, pv, &dl->pvd, &dl->vgd))
287                 goto_out;
288
289         pv->fmt = fmt;
290
291         r = 1;
292
293       out:
294         dm_pool_destroy(mem);
295         return r;
296 }
297
298 static int _format1_pv_setup(const struct format_type *fmt,
299                      uint64_t pe_start, uint32_t extent_count,
300                      uint32_t extent_size,
301                      unsigned long data_alignment __attribute((unused)),
302                      unsigned long data_alignment_offset __attribute((unused)),
303                      int pvmetadatacopies __attribute((unused)),
304                      uint64_t pvmetadatasize __attribute((unused)), struct dm_list *mdas __attribute((unused)),
305                      struct physical_volume *pv, struct volume_group *vg __attribute((unused)))
306 {
307         if (pv->size > MAX_PV_SIZE)
308                 pv->size--;
309         if (pv->size > MAX_PV_SIZE) {
310                 log_error("Physical volumes cannot be bigger than %s",
311                           display_size(fmt->cmd, (uint64_t) MAX_PV_SIZE));
312                 return 0;
313         }
314
315         /* Nothing more to do if extent size isn't provided */
316         if (!extent_size)
317                 return 1;
318
319         /*
320          * This works out pe_start and pe_count.
321          */
322         if (!calculate_extent_count(pv, extent_size, extent_count, pe_start))
323                 return_0;
324
325         /* Retain existing extent locations exactly */
326         if (((pe_start || extent_count) && (pe_start != pv->pe_start)) ||
327             (extent_count && (extent_count != pv->pe_count))) {
328                 log_error("Metadata would overwrite physical extents");
329                 return 0;
330         }
331
332         return 1;
333 }
334
335 static int _format1_lv_setup(struct format_instance *fid, struct logical_volume *lv)
336 {
337         uint64_t max_size = UINT_MAX;
338
339         if (!*lv->lvid.s)
340                 lvid_from_lvnum(&lv->lvid, &lv->vg->id, find_free_lvnum(lv));
341
342         if (lv->le_count > MAX_LE_TOTAL) {
343                 log_error("logical volumes cannot contain more than "
344                           "%d extents.", MAX_LE_TOTAL);
345                 return 0;
346         }
347         if (lv->size > max_size) {
348                 log_error("logical volumes cannot be larger than %s",
349                           display_size(fid->fmt->cmd, max_size));
350                 return 0;
351         }
352
353         return 1;
354 }
355
356 static int _format1_pv_write(const struct format_type *fmt, struct physical_volume *pv,
357                      struct dm_list *mdas __attribute((unused)), int64_t sector __attribute((unused)))
358 {
359         struct dm_pool *mem;
360         struct disk_list *dl;
361         struct dm_list pvs;
362         struct label *label;
363         struct lvmcache_info *info;
364
365         if (!(info = lvmcache_add(fmt->labeller, (char *) &pv->id, pv->dev,
366                                   pv->vg_name, NULL, 0)))
367                 return_0;
368         label = info->label;
369         info->device_size = pv->size << SECTOR_SHIFT;
370         info->fmt = fmt;
371
372         dm_list_init(&info->mdas);
373
374         dm_list_init(&pvs);
375
376         /* Ensure any residual PE structure is gone */
377         pv->pe_size = pv->pe_count = 0;
378         pv->pe_start = LVM1_PE_ALIGN;
379
380         if (!(mem = dm_pool_create("lvm1 pv_write", 1024)))
381                 return_0;
382
383         if (!(dl = dm_pool_alloc(mem, sizeof(*dl))))
384                 goto_bad;
385
386         dl->mem = mem;
387         dl->dev = pv->dev;
388
389         if (!export_pv(fmt->cmd, mem, NULL, &dl->pvd, pv))
390                 goto_bad;
391
392         /* must be set to be able to zero gap after PV structure in
393            dev_write in order to make other disk tools happy */
394         dl->pvd.pv_on_disk.base = METADATA_BASE;
395         dl->pvd.pv_on_disk.size = PV_SIZE;
396         dl->pvd.pe_on_disk.base = LVM1_PE_ALIGN << SECTOR_SHIFT;
397
398         dm_list_add(&pvs, &dl->list);
399         if (!write_disks(fmt, &pvs))
400                 goto_bad;
401
402         dm_pool_destroy(mem);
403         return 1;
404
405       bad:
406         dm_pool_destroy(mem);
407         return 0;
408 }
409
410 static int _format1_vg_setup(struct format_instance *fid, struct volume_group *vg)
411 {
412         /* just check max_pv and max_lv */
413         if (!vg->max_lv || vg->max_lv >= MAX_LV)
414                 vg->max_lv = MAX_LV - 1;
415
416         if (!vg->max_pv || vg->max_pv >= MAX_PV)
417                 vg->max_pv = MAX_PV - 1;
418
419         if (vg->extent_size > MAX_PE_SIZE || vg->extent_size < MIN_PE_SIZE) {
420                 log_error("Extent size must be between %s and %s",
421                           display_size(fid->fmt->cmd, (uint64_t) MIN_PE_SIZE),
422                           display_size(fid->fmt->cmd, (uint64_t) MAX_PE_SIZE));
423
424                 return 0;
425         }
426
427         if (vg->extent_size % MIN_PE_SIZE) {
428                 log_error("Extent size must be multiple of %s",
429                           display_size(fid->fmt->cmd, (uint64_t) MIN_PE_SIZE));
430                 return 0;
431         }
432
433         /* Redundant? */
434         if (vg->extent_size & (vg->extent_size - 1)) {
435                 log_error("Extent size must be power of 2");
436                 return 0;
437         }
438
439         return 1;
440 }
441
442 static int _format1_segtype_supported(struct format_instance *fid __attribute((unused)),
443                                       const struct segment_type *segtype)
444 {
445         if (!(segtype->flags & SEG_FORMAT1_SUPPORT))
446                 return_0;
447
448         return 1;
449 }
450
451 static struct metadata_area_ops _metadata_format1_ops = {
452         .vg_read = _format1_vg_read,
453         .vg_write = _format1_vg_write,
454 };
455
456 static struct format_instance *_format1_create_instance(const struct format_type *fmt,
457                                                 const char *vgname __attribute((unused)),
458                                                 const char *vgid __attribute((unused)),
459                                                 void *private __attribute((unused)))
460 {
461         struct format_instance *fid;
462         struct metadata_area *mda;
463
464         if (!(fid = dm_pool_alloc(fmt->cmd->mem, sizeof(*fid))))
465                 return_NULL;
466
467         fid->fmt = fmt;
468         dm_list_init(&fid->metadata_areas);
469
470         /* Define a NULL metadata area */
471         if (!(mda = dm_pool_alloc(fmt->cmd->mem, sizeof(*mda)))) {
472                 dm_pool_free(fmt->cmd->mem, fid);
473                 return_NULL;
474         }
475
476         mda->ops = &_metadata_format1_ops;
477         mda->metadata_locn = NULL;
478         dm_list_add(&fid->metadata_areas, &mda->list);
479
480         return fid;
481 }
482
483 static void _format1_destroy_instance(struct format_instance *fid __attribute((unused)))
484 {
485         return;
486 }
487
488 static void _format1_destroy(const struct format_type *fmt)
489 {
490         dm_free((void *) fmt);
491 }
492
493 static struct format_handler _format1_ops = {
494         .pv_read = _format1_pv_read,
495         .pv_setup = _format1_pv_setup,
496         .pv_write = _format1_pv_write,
497         .lv_setup = _format1_lv_setup,
498         .vg_setup = _format1_vg_setup,
499         .segtype_supported = _format1_segtype_supported,
500         .create_instance = _format1_create_instance,
501         .destroy_instance = _format1_destroy_instance,
502         .destroy = _format1_destroy,
503 };
504
505 #ifdef LVM1_INTERNAL
506 struct format_type *init_lvm1_format(struct cmd_context *cmd)
507 #else                           /* Shared */
508 struct format_type *init_format(struct cmd_context *cmd);
509 struct format_type *init_format(struct cmd_context *cmd)
510 #endif
511 {
512         struct format_type *fmt = dm_malloc(sizeof(*fmt));
513
514         if (!fmt)
515                 return_NULL;
516
517         fmt->cmd = cmd;
518         fmt->ops = &_format1_ops;
519         fmt->name = FMT_LVM1_NAME;
520         fmt->alias = NULL;
521         fmt->orphan_vg_name = FMT_LVM1_ORPHAN_VG_NAME;
522         fmt->features = FMT_RESTRICTED_LVIDS | FMT_ORPHAN_ALLOCATABLE |
523                         FMT_RESTRICTED_READAHEAD;
524         fmt->private = NULL;
525
526         if (!(fmt->labeller = lvm1_labeller_create(fmt))) {
527                 log_error("Couldn't create lvm1 label handler.");
528                 return NULL;
529         }
530
531         if (!(label_register_handler(FMT_LVM1_NAME, fmt->labeller))) {
532                 log_error("Couldn't register lvm1 label handler.");
533                 return NULL;
534         }
535
536         log_very_verbose("Initialised format: %s", fmt->name);
537
538         return fmt;
539 }