mirror of
git://sourceware.org/git/lvm2.git
synced 2024-12-22 17:35:59 +03:00
f53c6aa66e
o When volume group does not have write flag set, prevent changes to it. o Preparation for partial activation (not completed yet).
657 lines
13 KiB
C
657 lines
13 KiB
C
/*
|
|
* Copyright (C) 2001 Sistina Software (UK) Limited.
|
|
*
|
|
* This file is released under the LGPL.
|
|
*/
|
|
|
|
#include "disk-rep.h"
|
|
#include "dbg_malloc.h"
|
|
#include "pool.h"
|
|
#include "xlate.h"
|
|
#include "log.h"
|
|
#include "vgcache.h"
|
|
#include "filter.h"
|
|
|
|
#include <sys/types.h>
|
|
#include <sys/stat.h>
|
|
#include <fcntl.h>
|
|
#include <linux/kdev_t.h>
|
|
|
|
#define fail do {stack; return 0;} while(0)
|
|
#define xx16(v) disk->v = xlate16(disk->v)
|
|
#define xx32(v) disk->v = xlate32(disk->v)
|
|
#define xx64(v) disk->v = xlate64(disk->v)
|
|
|
|
/*
|
|
* Functions to perform the endian conversion
|
|
* between disk and core. The same code works
|
|
* both ways of course.
|
|
*/
|
|
static void _xlate_pvd(struct pv_disk *disk)
|
|
{
|
|
xx16(version);
|
|
|
|
xx32(pv_on_disk.base);
|
|
xx32(pv_on_disk.size);
|
|
xx32(vg_on_disk.base);
|
|
xx32(vg_on_disk.size);
|
|
xx32(pv_uuidlist_on_disk.base);
|
|
xx32(pv_uuidlist_on_disk.size);
|
|
xx32(lv_on_disk.base);
|
|
xx32(lv_on_disk.size);
|
|
xx32(pe_on_disk.base);
|
|
xx32(pe_on_disk.size);
|
|
|
|
xx32(pv_major);
|
|
xx32(pv_number);
|
|
xx32(pv_status);
|
|
xx32(pv_allocatable);
|
|
xx32(pv_size);
|
|
xx32(lv_cur);
|
|
xx32(pe_size);
|
|
xx32(pe_total);
|
|
xx32(pe_allocated);
|
|
xx32(pe_start);
|
|
}
|
|
|
|
static void _xlate_lvd(struct lv_disk *disk)
|
|
{
|
|
xx32(lv_access);
|
|
xx32(lv_status);
|
|
xx32(lv_open);
|
|
xx32(lv_dev);
|
|
xx32(lv_number);
|
|
xx32(lv_mirror_copies);
|
|
xx32(lv_recovery);
|
|
xx32(lv_schedule);
|
|
xx32(lv_size);
|
|
xx32(lv_snapshot_minor);
|
|
xx16(lv_chunk_size);
|
|
xx16(dummy);
|
|
xx32(lv_allocated_le);
|
|
xx32(lv_stripes);
|
|
xx32(lv_stripesize);
|
|
xx32(lv_badblock);
|
|
xx32(lv_allocation);
|
|
xx32(lv_io_timeout);
|
|
xx32(lv_read_ahead);
|
|
}
|
|
|
|
static void _xlate_vgd(struct vg_disk *disk)
|
|
{
|
|
xx32(vg_number);
|
|
xx32(vg_access);
|
|
xx32(vg_status);
|
|
xx32(lv_max);
|
|
xx32(lv_cur);
|
|
xx32(lv_open);
|
|
xx32(pv_max);
|
|
xx32(pv_cur);
|
|
xx32(pv_act);
|
|
xx32(dummy);
|
|
xx32(vgda);
|
|
xx32(pe_size);
|
|
xx32(pe_total);
|
|
xx32(pe_allocated);
|
|
xx32(pvg_total);
|
|
}
|
|
|
|
static void _xlate_extents(struct pe_disk *extents, int count)
|
|
{
|
|
int i;
|
|
|
|
for (i = 0; i < count; i++) {
|
|
extents[i].lv_num = xlate16(extents[i].lv_num);
|
|
extents[i].le_num = xlate16(extents[i].le_num);
|
|
}
|
|
}
|
|
|
|
/*
|
|
* Handle both minor metadata formats.
|
|
*/
|
|
static int _munge_formats(struct pv_disk *pvd)
|
|
{
|
|
uint32_t pe_start;
|
|
|
|
switch (pvd->version) {
|
|
case 1:
|
|
pvd->pe_start = ((pvd->pe_on_disk.base +
|
|
pvd->pe_on_disk.size) / SECTOR_SIZE);
|
|
break;
|
|
|
|
case 2:
|
|
pvd->version = 1;
|
|
pe_start = pvd->pe_start * SECTOR_SIZE;
|
|
pvd->pe_on_disk.size = pe_start - pvd->pe_on_disk.base;
|
|
break;
|
|
|
|
default:
|
|
return 0;
|
|
}
|
|
|
|
return 1;
|
|
}
|
|
|
|
int read_pvd(struct device *dev, struct pv_disk *pvd)
|
|
{
|
|
if (dev_read(dev, 0, sizeof(*pvd), pvd) != sizeof(*pvd)) {
|
|
log_very_verbose("Failed to read PV data from %s",
|
|
dev_name(dev));
|
|
return 0;
|
|
}
|
|
|
|
_xlate_pvd(pvd);
|
|
|
|
if (pvd->id[0] != 'H' || pvd->id[1] != 'M') {
|
|
log_very_verbose("%s does not have a valid PV identifier",
|
|
dev_name(dev));
|
|
return 0;
|
|
}
|
|
|
|
if (!_munge_formats(pvd)) {
|
|
log_very_verbose("Unknown metadata version %d found on %s",
|
|
pvd->version, dev_name(dev));
|
|
return 0;
|
|
}
|
|
|
|
return 1;
|
|
}
|
|
|
|
static int _read_lvd(struct device *dev, ulong pos, struct lv_disk *disk)
|
|
{
|
|
if (dev_read(dev, pos, sizeof(*disk), disk) != sizeof(*disk))
|
|
fail;
|
|
|
|
_xlate_lvd(disk);
|
|
|
|
return 1;
|
|
}
|
|
|
|
static int _read_vgd(struct disk_list *data)
|
|
{
|
|
struct vg_disk *vgd = &data->vgd;
|
|
ulong pos = data->pvd.vg_on_disk.base;
|
|
if (dev_read(data->dev, pos, sizeof(*vgd), vgd) != sizeof(*vgd))
|
|
fail;
|
|
|
|
_xlate_vgd(vgd);
|
|
|
|
return 1;
|
|
}
|
|
|
|
static int _read_uuids(struct disk_list *data)
|
|
{
|
|
int num_read = 0;
|
|
struct uuid_list *ul;
|
|
char buffer[NAME_LEN];
|
|
ulong pos = data->pvd.pv_uuidlist_on_disk.base;
|
|
ulong end = pos + data->pvd.pv_uuidlist_on_disk.size;
|
|
|
|
while (pos < end && num_read < data->vgd.pv_cur) {
|
|
if (dev_read(data->dev, pos, sizeof(buffer), buffer) !=
|
|
sizeof(buffer))
|
|
fail;
|
|
|
|
if (!(ul = pool_alloc(data->mem, sizeof(*ul))))
|
|
fail;
|
|
|
|
memcpy(ul->uuid, buffer, NAME_LEN);
|
|
ul->uuid[NAME_LEN - 1] = '\0';
|
|
|
|
list_add(&data->uuids, &ul->list);
|
|
|
|
pos += NAME_LEN;
|
|
num_read++;
|
|
}
|
|
|
|
return 1;
|
|
}
|
|
|
|
static inline int _check_lvd(struct lv_disk *lvd)
|
|
{
|
|
return !(lvd->lv_name[0] == '\0');
|
|
}
|
|
|
|
static int _read_lvs(struct disk_list *data)
|
|
{
|
|
int i, read = 0;
|
|
ulong pos;
|
|
struct lvd_list *ll;
|
|
struct vg_disk *vgd = &data->vgd;
|
|
|
|
for (i = 0; (i < vgd->lv_max) && (read < vgd->lv_cur); i++) {
|
|
pos = data->pvd.lv_on_disk.base + (i * sizeof(struct lv_disk));
|
|
ll = pool_alloc(data->mem, sizeof(*ll));
|
|
|
|
if (!ll)
|
|
fail;
|
|
|
|
if (!_read_lvd(data->dev, pos, &ll->lvd))
|
|
fail;
|
|
|
|
if (!_check_lvd(&ll->lvd))
|
|
continue;
|
|
|
|
read++;
|
|
list_add(&data->lvds, &ll->list);
|
|
}
|
|
|
|
return 1;
|
|
}
|
|
|
|
static int _read_extents(struct disk_list *data)
|
|
{
|
|
size_t len = sizeof(struct pe_disk) * data->pvd.pe_total;
|
|
struct pe_disk *extents = pool_alloc(data->mem, len);
|
|
ulong pos = data->pvd.pe_on_disk.base;
|
|
|
|
if (!extents)
|
|
fail;
|
|
|
|
if (dev_read(data->dev, pos, len, extents) != len)
|
|
fail;
|
|
|
|
_xlate_extents(extents, data->pvd.pe_total);
|
|
data->extents = extents;
|
|
|
|
return 1;
|
|
}
|
|
|
|
/*
|
|
* If exported, remove "PV_EXP" from end of VG name
|
|
*/
|
|
static void _munge_exported_vg(struct disk_list *data)
|
|
{
|
|
int l, s;
|
|
|
|
/* Return if PV not in a VG or VG not exported */
|
|
if ((!*data->pvd.vg_name) ||
|
|
!(data->vgd.vg_status & VG_EXPORTED))
|
|
return;
|
|
|
|
l = strlen(data->pvd.vg_name);
|
|
s = sizeof(EXPORTED_TAG);
|
|
if (!strncmp(data->pvd.vg_name + l - s + 1, EXPORTED_TAG, s))
|
|
data->pvd.vg_name[l - s + 1] = '\0';
|
|
|
|
data->pvd.pv_status |= VG_EXPORTED;
|
|
}
|
|
|
|
static struct disk_list *__read_disk(struct device *dev, struct pool *mem,
|
|
const char *vg_name)
|
|
{
|
|
struct disk_list *data = pool_alloc(mem, sizeof(*data));
|
|
const char *name = dev_name(dev);
|
|
|
|
if (!data) {
|
|
stack;
|
|
return NULL;
|
|
}
|
|
|
|
data->dev = dev;
|
|
data->mem = mem;
|
|
list_init(&data->uuids);
|
|
list_init(&data->lvds);
|
|
|
|
if (!read_pvd(dev, &data->pvd)) {
|
|
stack;
|
|
goto bad;
|
|
}
|
|
|
|
/*
|
|
* is it an orphan ?
|
|
*/
|
|
if (!*data->pvd.vg_name) {
|
|
log_very_verbose("%s is not a member of any VG", name);
|
|
|
|
/* Update VG cache */
|
|
vgcache_add(data->pvd.vg_name, dev);
|
|
|
|
return (vg_name) ? NULL : data;
|
|
}
|
|
|
|
if (!_read_vgd(data)) {
|
|
log_error("Failed to read VG data from PV (%s)", name);
|
|
goto bad;
|
|
}
|
|
|
|
/* If VG is exported, set VG name back to the real name */
|
|
_munge_exported_vg(data);
|
|
|
|
/* Update VG cache with what we found */
|
|
vgcache_add(data->pvd.vg_name, dev);
|
|
|
|
if (vg_name && strcmp(vg_name, data->pvd.vg_name)) {
|
|
log_very_verbose("%s is not a member of the VG %s",
|
|
name, vg_name);
|
|
goto bad;
|
|
}
|
|
|
|
if (!_read_uuids(data)) {
|
|
log_error("Failed to read PV uuid list from %s", name);
|
|
goto bad;
|
|
}
|
|
|
|
if (!_read_lvs(data)) {
|
|
log_error("Failed to read LV's from %s", name);
|
|
goto bad;
|
|
}
|
|
|
|
if (!_read_extents(data)) {
|
|
log_error("Failed to read extents from %s", name);
|
|
goto bad;
|
|
}
|
|
|
|
log_very_verbose("Found %s in %sVG %s", name,
|
|
(data->vgd.vg_status & VG_EXPORTED) ? "exported " : "",
|
|
data->pvd.vg_name);
|
|
|
|
return data;
|
|
|
|
bad:
|
|
pool_free(data->mem, data);
|
|
return NULL;
|
|
}
|
|
|
|
struct disk_list *read_disk(struct device *dev, struct pool *mem,
|
|
const char *vg_name)
|
|
{
|
|
struct disk_list *r;
|
|
|
|
if (!dev_open(dev, O_RDONLY)) {
|
|
stack;
|
|
return NULL;
|
|
}
|
|
|
|
r = __read_disk(dev, mem, vg_name);
|
|
|
|
if (!dev_close(dev))
|
|
stack;
|
|
|
|
return r;
|
|
}
|
|
|
|
static void _add_pv_to_list(struct list *head, struct disk_list *data)
|
|
{
|
|
struct list *pvdh;
|
|
struct pv_disk *pvd;
|
|
|
|
list_iterate(pvdh, head) {
|
|
pvd = &list_item(pvdh, struct disk_list)->pvd;
|
|
if (!strncmp(data->pvd.pv_uuid, pvd->pv_uuid,
|
|
sizeof(pvd->pv_uuid))) {
|
|
if (MAJOR(data->dev->dev) != md_major()) {
|
|
log_very_verbose("Ignoring duplicate PV %s on "
|
|
"%s", pvd->pv_uuid,
|
|
dev_name(data->dev));
|
|
return;
|
|
}
|
|
log_very_verbose("Duplicate PV %s - using md %s",
|
|
pvd->pv_uuid, dev_name(data->dev));
|
|
list_del(pvdh);
|
|
break;
|
|
}
|
|
}
|
|
list_add(head, &data->list);
|
|
}
|
|
|
|
/*
|
|
* Build a list of pv_d's structures, allocated from mem.
|
|
* We keep track of the first object allocated form the pool
|
|
* so we can free off all the memory if something goes wrong.
|
|
*/
|
|
int read_pvs_in_vg(const char *vg_name, struct dev_filter *filter,
|
|
struct pool *mem, struct list *head)
|
|
{
|
|
struct dev_iter *iter;
|
|
struct device *dev;
|
|
struct disk_list *data = NULL;
|
|
|
|
struct list *pvdh, *pvdh2;
|
|
|
|
/* Fast path if we already saw this VG and cached the list of PVs */
|
|
if ((pvdh = vgcache_find(vg_name))) {
|
|
list_iterate(pvdh2, pvdh) {
|
|
dev = list_item(pvdh2, struct pvdev_list)->dev;
|
|
if (!(data = read_disk(dev, mem, vg_name)))
|
|
break;
|
|
_add_pv_to_list(head, data);
|
|
}
|
|
|
|
/* Did we find the whole VG? */
|
|
if (!vg_name || !*vg_name ||
|
|
(data && *data->pvd.vg_name &&
|
|
list_size(head) == data->vgd.pv_cur))
|
|
return 1;
|
|
|
|
/* Something changed. Remove the hints. */
|
|
list_init(head);
|
|
vgcache_del(vg_name);
|
|
}
|
|
|
|
if (!(iter = dev_iter_create(filter))) {
|
|
log_error("read_pvs_in_vg: dev_iter_create failed");
|
|
return 0;
|
|
}
|
|
|
|
/* Otherwise do a complete scan */
|
|
for (dev = dev_iter_get(iter); dev; dev = dev_iter_get(iter)) {
|
|
if ((data = read_disk(dev, mem, vg_name))) {
|
|
_add_pv_to_list(head, data);
|
|
}
|
|
}
|
|
dev_iter_destroy(iter);
|
|
|
|
if (list_empty(head))
|
|
return 0;
|
|
|
|
return 1;
|
|
}
|
|
|
|
static int _write_vgd(struct disk_list *data)
|
|
{
|
|
struct vg_disk *vgd = &data->vgd;
|
|
ulong pos = data->pvd.vg_on_disk.base;
|
|
|
|
_xlate_vgd(vgd);
|
|
if (dev_write(data->dev, pos, sizeof(*vgd), vgd) != sizeof(*vgd))
|
|
fail;
|
|
|
|
_xlate_vgd(vgd);
|
|
|
|
return 1;
|
|
}
|
|
|
|
static int _write_uuids(struct disk_list *data)
|
|
{
|
|
struct uuid_list *ul;
|
|
struct list *uh;
|
|
ulong pos = data->pvd.pv_uuidlist_on_disk.base;
|
|
ulong end = pos + data->pvd.pv_uuidlist_on_disk.size;
|
|
|
|
list_iterate(uh, &data->uuids) {
|
|
if (pos >= end) {
|
|
log_error("Too many uuids to fit on %s",
|
|
dev_name(data->dev));
|
|
return 0;
|
|
}
|
|
|
|
ul = list_item(uh, struct uuid_list);
|
|
if (dev_write(data->dev, pos, NAME_LEN, ul->uuid) != NAME_LEN)
|
|
fail;
|
|
|
|
pos += NAME_LEN;
|
|
}
|
|
|
|
return 1;
|
|
}
|
|
|
|
static int _write_lvd(struct device *dev, ulong pos, struct lv_disk *disk)
|
|
{
|
|
_xlate_lvd(disk);
|
|
if (dev_write(dev, pos, sizeof(*disk), disk) != sizeof(*disk))
|
|
fail;
|
|
|
|
_xlate_lvd(disk);
|
|
|
|
return 1;
|
|
}
|
|
|
|
static int _write_lvs(struct disk_list *data)
|
|
{
|
|
struct list *lvh;
|
|
ulong pos;
|
|
|
|
pos = data->pvd.lv_on_disk.base;
|
|
|
|
if (!dev_zero(data->dev, pos, data->pvd.lv_on_disk.size)) {
|
|
log_error("Couldn't zero lv area on device '%s'",
|
|
dev_name(data->dev));
|
|
return 0;
|
|
}
|
|
|
|
list_iterate(lvh, &data->lvds) {
|
|
struct lvd_list *ll = list_item(lvh, struct lvd_list);
|
|
|
|
if (!_write_lvd(data->dev, pos, &ll->lvd))
|
|
fail;
|
|
|
|
pos += sizeof(struct lv_disk);
|
|
}
|
|
|
|
return 1;
|
|
}
|
|
|
|
static int _write_extents(struct disk_list *data)
|
|
{
|
|
size_t len = sizeof(struct pe_disk) * data->pvd.pe_total;
|
|
struct pe_disk *extents = data->extents;
|
|
ulong pos = data->pvd.pe_on_disk.base;
|
|
|
|
_xlate_extents(extents, data->pvd.pe_total);
|
|
if (dev_write(data->dev, pos, len, extents) != len)
|
|
fail;
|
|
|
|
_xlate_extents(extents, data->pvd.pe_total);
|
|
|
|
return 1;
|
|
}
|
|
|
|
static int _write_pvd(struct disk_list *data)
|
|
{
|
|
char *buf;
|
|
ulong pos = data->pvd.pv_on_disk.base;
|
|
ulong size = data->pvd.pv_on_disk.size;
|
|
|
|
if(size < sizeof(struct pv_disk)) {
|
|
log_error("Invalid PV structure size.");
|
|
return 0;
|
|
}
|
|
|
|
/* Make sure that the gap between the PV structure and
|
|
the next one is zeroed in order to make non LVM tools
|
|
happy (idea from AED) */
|
|
buf = dbg_malloc(size);
|
|
if(!buf) {
|
|
log_err("Couldn't allocate temporary PV buffer.");
|
|
return 0;
|
|
}
|
|
|
|
memset(buf, 0, size);
|
|
memcpy(buf, &data->pvd, sizeof(struct pv_disk));
|
|
|
|
_xlate_pvd((struct pv_disk *)buf);
|
|
if (dev_write(data->dev, pos, size, buf) != size) {
|
|
dbg_free(buf);
|
|
fail;
|
|
}
|
|
|
|
dbg_free(buf);
|
|
return 1;
|
|
}
|
|
|
|
/*
|
|
* assumes the device has been opened.
|
|
*/
|
|
static int __write_all_pvd(struct disk_list *data)
|
|
{
|
|
const char *pv_name = dev_name(data->dev);
|
|
|
|
if (!_write_pvd(data)) {
|
|
log_error("Failed to write PV structure onto %s", pv_name);
|
|
return 0;
|
|
}
|
|
|
|
if (!test_mode())
|
|
vgcache_add(data->pvd.vg_name, data->dev);
|
|
/*
|
|
* Stop here for orphan pv's.
|
|
*/
|
|
if (data->pvd.vg_name[0] == '\0')
|
|
return 1;
|
|
|
|
if (!_write_vgd(data)) {
|
|
log_error("Failed to write VG data to %s", pv_name);
|
|
return 0;
|
|
}
|
|
|
|
if (!_write_uuids(data)) {
|
|
log_error("Failed to write PV uuid list to %s", pv_name);
|
|
return 0;
|
|
}
|
|
|
|
if (!_write_lvs(data)) {
|
|
log_error("Failed to write LV's to %s", pv_name);
|
|
return 0;
|
|
}
|
|
|
|
if (!_write_extents(data)) {
|
|
log_error("Failed to write extents to %s", pv_name);
|
|
return 0;
|
|
}
|
|
|
|
return 1;
|
|
}
|
|
|
|
/*
|
|
* opens the device and hands to the above fn.
|
|
*/
|
|
static int _write_all_pvd(struct disk_list *data)
|
|
{
|
|
int r;
|
|
|
|
if (!dev_open(data->dev, O_WRONLY)) {
|
|
stack;
|
|
return 0;
|
|
}
|
|
|
|
r = __write_all_pvd(data);
|
|
|
|
if (!dev_close(data->dev))
|
|
stack;
|
|
|
|
return r;
|
|
}
|
|
|
|
/*
|
|
* Writes all the given pv's to disk. Does very
|
|
* little sanity checking, so make sure correct
|
|
* data is passed to here.
|
|
*/
|
|
int write_disks(struct list *pvs)
|
|
{
|
|
struct list *pvh;
|
|
struct disk_list *dl;
|
|
|
|
list_iterate(pvh, pvs) {
|
|
dl = list_item(pvh, struct disk_list);
|
|
if (!(_write_all_pvd(dl)))
|
|
fail;
|
|
|
|
log_very_verbose("Successfully wrote data to %s",
|
|
dev_name(dl->dev));
|
|
}
|
|
|
|
return 1;
|
|
}
|