xref: /netbsd-src/external/gpl2/lvm2/dist/lib/format1/format1.c (revision 404fbe5fb94ca1e054339640cabb2801ce52dd30)
1 /*	$NetBSD: format1.c,v 1.1.1.1 2008/12/22 00:17:59 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 {
118 	struct dm_pool *mem = fid->fmt->cmd->mem;
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->fid = fid;
132 	vg->seqno = 0;
133 	dm_list_init(&vg->pvs);
134 	dm_list_init(&vg->lvs);
135 	dm_list_init(&vg->tags);
136 
137 	if (!_check_vgs(pvs))
138 		goto_bad;
139 
140 	dl = dm_list_item(pvs->n, struct disk_list);
141 
142 	if (!import_vg(mem, vg, dl))
143 		goto_bad;
144 
145 	if (!import_pvs(fid->fmt, mem, vg, pvs, &vg->pvs, &vg->pv_count))
146 		goto_bad;
147 
148 	if (!import_lvs(mem, vg, pvs))
149 		goto_bad;
150 
151 	if (!import_extents(fid->fmt->cmd, vg, pvs))
152 		goto_bad;
153 
154 	if (!import_snapshots(mem, vg, pvs))
155 		goto_bad;
156 
157 	return vg;
158 
159       bad:
160 	dm_pool_free(mem, vg);
161 	return NULL;
162 }
163 
164 static struct volume_group *_format1_vg_read(struct format_instance *fid,
165 				     const char *vg_name,
166 				     struct metadata_area *mda __attribute((unused)))
167 {
168 	struct dm_pool *mem = dm_pool_create("lvm1 vg_read", 1024 * 10);
169 	struct dm_list pvs;
170 	struct volume_group *vg = NULL;
171 	dm_list_init(&pvs);
172 
173 	if (!mem)
174 		return_NULL;
175 
176 	/* Strip dev_dir if present */
177 	vg_name = strip_dir(vg_name, fid->fmt->cmd->dev_dir);
178 
179 	if (!read_pvs_in_vg
180 	    (fid->fmt, vg_name, fid->fmt->cmd->filter, mem, &pvs))
181 		goto_bad;
182 
183 	if (!(vg = _build_vg(fid, &pvs)))
184 		goto_bad;
185 
186       bad:
187 	dm_pool_destroy(mem);
188 	return vg;
189 }
190 
191 static struct disk_list *_flatten_pv(struct format_instance *fid,
192 				     struct dm_pool *mem, struct volume_group *vg,
193 				     struct physical_volume *pv,
194 				     const char *dev_dir)
195 {
196 	struct disk_list *dl = dm_pool_alloc(mem, sizeof(*dl));
197 
198 	if (!dl)
199 		return_NULL;
200 
201 	dl->mem = mem;
202 	dl->dev = pv->dev;
203 
204 	dm_list_init(&dl->uuids);
205 	dm_list_init(&dl->lvds);
206 
207 	if (!export_pv(fid->fmt->cmd, mem, vg, &dl->pvd, pv) ||
208 	    !export_vg(&dl->vgd, vg) ||
209 	    !export_uuids(dl, vg) ||
210 	    !export_lvs(dl, vg, pv, dev_dir) || !calculate_layout(dl)) {
211 		dm_pool_free(mem, dl);
212 		return_NULL;
213 	}
214 
215 	return dl;
216 }
217 
218 static int _flatten_vg(struct format_instance *fid, struct dm_pool *mem,
219 		       struct volume_group *vg,
220 		       struct dm_list *pvds, const char *dev_dir,
221 		       struct dev_filter *filter)
222 {
223 	struct pv_list *pvl;
224 	struct disk_list *data;
225 
226 	dm_list_iterate_items(pvl, &vg->pvs) {
227 		if (!(data = _flatten_pv(fid, mem, vg, pvl->pv, dev_dir)))
228 			return_0;
229 
230 		dm_list_add(pvds, &data->list);
231 	}
232 
233 	export_numbers(pvds, vg);
234 	export_pv_act(pvds);
235 
236 	if (!export_vg_number(fid, pvds, vg->name, filter))
237 		return_0;
238 
239 	return 1;
240 }
241 
242 static int _format1_vg_write(struct format_instance *fid, struct volume_group *vg,
243 		     struct metadata_area *mda __attribute((unused)))
244 {
245 	struct dm_pool *mem = dm_pool_create("lvm1 vg_write", 1024 * 10);
246 	struct dm_list pvds;
247 	int r = 0;
248 
249 	if (!mem)
250 		return_0;
251 
252 	dm_list_init(&pvds);
253 
254 	r = (_flatten_vg(fid, mem, vg, &pvds, fid->fmt->cmd->dev_dir,
255 			 fid->fmt->cmd->filter) &&
256 	     write_disks(fid->fmt, &pvds));
257 
258 	lvmcache_update_vg(vg, 0);
259 	dm_pool_destroy(mem);
260 	return r;
261 }
262 
263 static int _format1_pv_read(const struct format_type *fmt, const char *pv_name,
264 		    struct physical_volume *pv, struct dm_list *mdas __attribute((unused)))
265 {
266 	struct dm_pool *mem = dm_pool_create("lvm1 pv_read", 1024);
267 	struct disk_list *dl;
268 	struct device *dev;
269 	int r = 0;
270 
271 	log_very_verbose("Reading physical volume data %s from disk", pv_name);
272 
273 	if (!mem)
274 		return_0;
275 
276 	if (!(dev = dev_cache_get(pv_name, fmt->cmd->filter)))
277 		goto_out;
278 
279 	if (!(dl = read_disk(fmt, dev, mem, NULL)))
280 		goto_out;
281 
282 	if (!import_pv(fmt, fmt->cmd->mem, dl->dev, NULL, pv, &dl->pvd, &dl->vgd))
283 		goto_out;
284 
285 	pv->fmt = fmt;
286 
287 	r = 1;
288 
289       out:
290 	dm_pool_destroy(mem);
291 	return r;
292 }
293 
294 static int _format1_pv_setup(const struct format_type *fmt,
295 		     uint64_t pe_start, uint32_t extent_count,
296 		     uint32_t extent_size,
297 		     int pvmetadatacopies __attribute((unused)),
298 		     uint64_t pvmetadatasize __attribute((unused)), struct dm_list *mdas __attribute((unused)),
299 		     struct physical_volume *pv, struct volume_group *vg __attribute((unused)))
300 {
301 	if (pv->size > MAX_PV_SIZE)
302 		pv->size--;
303 	if (pv->size > MAX_PV_SIZE) {
304 		log_error("Physical volumes cannot be bigger than %s",
305 			  display_size(fmt->cmd, (uint64_t) MAX_PV_SIZE));
306 		return 0;
307 	}
308 
309 	/* Nothing more to do if extent size isn't provided */
310 	if (!extent_size)
311 		return 1;
312 
313 	/*
314 	 * This works out pe_start and pe_count.
315 	 */
316 	if (!calculate_extent_count(pv, extent_size, extent_count, pe_start))
317 		return_0;
318 
319 	/* Retain existing extent locations exactly */
320 	if (((pe_start || extent_count) && (pe_start != pv->pe_start)) ||
321 	    (extent_count && (extent_count != pv->pe_count))) {
322 		log_error("Metadata would overwrite physical extents");
323 		return 0;
324 	}
325 
326 	return 1;
327 }
328 
329 static int _format1_lv_setup(struct format_instance *fid, struct logical_volume *lv)
330 {
331 	uint64_t max_size = UINT_MAX;
332 
333 	if (!*lv->lvid.s)
334 		lvid_from_lvnum(&lv->lvid, &lv->vg->id, find_free_lvnum(lv));
335 
336 	if (lv->le_count > MAX_LE_TOTAL) {
337 		log_error("logical volumes cannot contain more than "
338 			  "%d extents.", MAX_LE_TOTAL);
339 		return 0;
340 	}
341 	if (lv->size > max_size) {
342 		log_error("logical volumes cannot be larger than %s",
343 			  display_size(fid->fmt->cmd, max_size));
344 		return 0;
345 	}
346 
347 	return 1;
348 }
349 
350 static int _format1_pv_write(const struct format_type *fmt, struct physical_volume *pv,
351 		     struct dm_list *mdas __attribute((unused)), int64_t sector __attribute((unused)))
352 {
353 	struct dm_pool *mem;
354 	struct disk_list *dl;
355 	struct dm_list pvs;
356 	struct label *label;
357 	struct lvmcache_info *info;
358 
359 	if (!(info = lvmcache_add(fmt->labeller, (char *) &pv->id, pv->dev,
360 				  pv->vg_name, NULL, 0)))
361 		return_0;
362 	label = info->label;
363 	info->device_size = pv->size << SECTOR_SHIFT;
364 	info->fmt = fmt;
365 
366 	dm_list_init(&info->mdas);
367 
368 	dm_list_init(&pvs);
369 
370 	/* Ensure any residual PE structure is gone */
371 	pv->pe_size = pv->pe_count = 0;
372 	pv->pe_start = LVM1_PE_ALIGN;
373 
374 	if (!(mem = dm_pool_create("lvm1 pv_write", 1024)))
375 		return_0;
376 
377 	if (!(dl = dm_pool_alloc(mem, sizeof(*dl))))
378 		goto_bad;
379 
380 	dl->mem = mem;
381 	dl->dev = pv->dev;
382 
383 	if (!export_pv(fmt->cmd, mem, NULL, &dl->pvd, pv))
384 		goto_bad;
385 
386 	/* must be set to be able to zero gap after PV structure in
387 	   dev_write in order to make other disk tools happy */
388 	dl->pvd.pv_on_disk.base = METADATA_BASE;
389 	dl->pvd.pv_on_disk.size = PV_SIZE;
390 	dl->pvd.pe_on_disk.base = LVM1_PE_ALIGN << SECTOR_SHIFT;
391 
392 	dm_list_add(&pvs, &dl->list);
393 	if (!write_disks(fmt, &pvs))
394 		goto_bad;
395 
396 	dm_pool_destroy(mem);
397 	return 1;
398 
399       bad:
400 	dm_pool_destroy(mem);
401 	return 0;
402 }
403 
404 static int _format1_vg_setup(struct format_instance *fid, struct volume_group *vg)
405 {
406 	/* just check max_pv and max_lv */
407 	if (!vg->max_lv || vg->max_lv >= MAX_LV)
408 		vg->max_lv = MAX_LV - 1;
409 
410 	if (!vg->max_pv || vg->max_pv >= MAX_PV)
411 		vg->max_pv = MAX_PV - 1;
412 
413 	if (vg->extent_size > MAX_PE_SIZE || vg->extent_size < MIN_PE_SIZE) {
414 		log_error("Extent size must be between %s and %s",
415 			  display_size(fid->fmt->cmd, (uint64_t) MIN_PE_SIZE),
416 			  display_size(fid->fmt->cmd, (uint64_t) MAX_PE_SIZE));
417 
418 		return 0;
419 	}
420 
421 	if (vg->extent_size % MIN_PE_SIZE) {
422 		log_error("Extent size must be multiple of %s",
423 			  display_size(fid->fmt->cmd, (uint64_t) MIN_PE_SIZE));
424 		return 0;
425 	}
426 
427 	/* Redundant? */
428 	if (vg->extent_size & (vg->extent_size - 1)) {
429 		log_error("Extent size must be power of 2");
430 		return 0;
431 	}
432 
433 	return 1;
434 }
435 
436 static int _format1_segtype_supported(struct format_instance *fid __attribute((unused)),
437 				      const struct segment_type *segtype)
438 {
439 	if (!(segtype->flags & SEG_FORMAT1_SUPPORT))
440 		return_0;
441 
442 	return 1;
443 }
444 
445 static struct metadata_area_ops _metadata_format1_ops = {
446 	.vg_read = _format1_vg_read,
447 	.vg_write = _format1_vg_write,
448 };
449 
450 static struct format_instance *_format1_create_instance(const struct format_type *fmt,
451 						const char *vgname __attribute((unused)),
452 						const char *vgid __attribute((unused)),
453 						void *private __attribute((unused)))
454 {
455 	struct format_instance *fid;
456 	struct metadata_area *mda;
457 
458 	if (!(fid = dm_pool_alloc(fmt->cmd->mem, sizeof(*fid))))
459 		return_NULL;
460 
461 	fid->fmt = fmt;
462 	dm_list_init(&fid->metadata_areas);
463 
464 	/* Define a NULL metadata area */
465 	if (!(mda = dm_pool_alloc(fmt->cmd->mem, sizeof(*mda)))) {
466 		dm_pool_free(fmt->cmd->mem, fid);
467 		return_NULL;
468 	}
469 
470 	mda->ops = &_metadata_format1_ops;
471 	mda->metadata_locn = NULL;
472 	dm_list_add(&fid->metadata_areas, &mda->list);
473 
474 	return fid;
475 }
476 
477 static void _format1_destroy_instance(struct format_instance *fid __attribute((unused)))
478 {
479 	return;
480 }
481 
482 static void _format1_destroy(const struct format_type *fmt)
483 {
484 	dm_free((void *) fmt);
485 }
486 
487 static struct format_handler _format1_ops = {
488 	.pv_read = _format1_pv_read,
489 	.pv_setup = _format1_pv_setup,
490 	.pv_write = _format1_pv_write,
491 	.lv_setup = _format1_lv_setup,
492 	.vg_setup = _format1_vg_setup,
493 	.segtype_supported = _format1_segtype_supported,
494 	.create_instance = _format1_create_instance,
495 	.destroy_instance = _format1_destroy_instance,
496 	.destroy = _format1_destroy,
497 };
498 
499 #ifdef LVM1_INTERNAL
500 struct format_type *init_lvm1_format(struct cmd_context *cmd)
501 #else				/* Shared */
502 struct format_type *init_format(struct cmd_context *cmd);
503 struct format_type *init_format(struct cmd_context *cmd)
504 #endif
505 {
506 	struct format_type *fmt = dm_malloc(sizeof(*fmt));
507 
508 	if (!fmt)
509 		return_NULL;
510 
511 	fmt->cmd = cmd;
512 	fmt->ops = &_format1_ops;
513 	fmt->name = FMT_LVM1_NAME;
514 	fmt->alias = NULL;
515 	fmt->orphan_vg_name = FMT_LVM1_ORPHAN_VG_NAME;
516 	fmt->features = FMT_RESTRICTED_LVIDS | FMT_ORPHAN_ALLOCATABLE |
517 			FMT_RESTRICTED_READAHEAD;
518 	fmt->private = NULL;
519 
520 	if (!(fmt->labeller = lvm1_labeller_create(fmt))) {
521 		log_error("Couldn't create lvm1 label handler.");
522 		return NULL;
523 	}
524 
525 	if (!(label_register_handler(FMT_LVM1_NAME, fmt->labeller))) {
526 		log_error("Couldn't register lvm1 label handler.");
527 		return NULL;
528 	}
529 
530 	log_very_verbose("Initialised format: %s", fmt->name);
531 
532 	return fmt;
533 }
534