lvm2/lib/label/label.c

/*
 * Copyright (C) 2002-2004 Sistina Software, Inc. All rights reserved.
 * Copyright (C) 2004-2007 Red Hat, Inc. All rights reserved.
 *
 * This file is part of LVM2.
 *
 * This copyrighted material is made available to anyone wishing to use,
 * modify, copy, or redistribute it subject to the terms and conditions
 * of the GNU Lesser General Public License v.2.1.
 *
 * You should have received a copy of the GNU Lesser General Public License
 * along with this program; if not, write to the Free Software Foundation,
 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

#include "lib/misc/lib.h"
#include "base/memory/zalloc.h"
#include "lib/label/label.h"
#include "lib/misc/crc.h"
#include "lib/mm/xlate.h"
#include "lib/cache/lvmcache.h"
#include "lib/device/bcache.h"
#include "lib/commands/toolcontext.h"
#include "lib/activate/activate.h"
#include "lib/label/hints.h"
#include "lib/metadata/metadata.h"
#include "lib/format_text/layout.h"
#include "lib/device/device_id.h"
#include "lib/device/online.h"

#include <sys/stat.h>
#include <fcntl.h>
#include <unistd.h>
#include <sys/types.h>
#include <sys/resource.h>

/* FIXME Allow for larger labels?  Restricted to single sector currently */

static uint64_t _current_bcache_size_bytes;

/*
 * Internal labeller struct.
 */
struct labeller_i {
	struct dm_list list;

	struct labeller *l;
	char name[0];
};

static struct dm_list _labellers;

static struct labeller_i *_alloc_li(const char *name, struct labeller *l)
{
	struct labeller_i *li;
	size_t len;

	len = strlen(name);

	if (!(li = malloc(sizeof(*li) + len + 1))) {
		log_error("Couldn't allocate memory for labeller list object.");
		return NULL;
	}

	li->l = l;
	memcpy(li->name, name, len + 1);

	return li;
}

int label_init(void)
{
	dm_list_init(&_labellers);
	return 1;
}

void label_exit(void)
{
	struct labeller_i *li, *tli;

	dm_list_iterate_items_safe(li, tli, &_labellers) {
		dm_list_del(&li->list);
		li->l->ops->destroy(li->l);
		free(li);
	}

	dm_list_init(&_labellers);
}

int label_register_handler(struct labeller *handler)
{
	struct labeller_i *li;

	if (!(li = _alloc_li(handler->fmt->name, handler)))
		return_0;

	dm_list_add(&_labellers, &li->list);
	return 1;
}

struct labeller *label_get_handler(const char *name)
{
	struct labeller_i *li;

	dm_list_iterate_items(li, &_labellers)
		if (!strcmp(li->name, name))
			return li->l;

	return NULL;
}

/* FIXME Also wipe associated metadata area headers? */
int label_remove(struct device *dev)
{
	char readbuf[LABEL_SIZE] __attribute__((aligned(8)));
	int r = 1;
	uint64_t sector;
	int wipe;
	struct labeller_i *li;
	struct label_header *lh;
	struct lvmcache_info *info;

	log_very_verbose("Scanning for labels to wipe from %s", dev_name(dev));

	if (!label_scan_open_excl(dev)) {
		log_error("Failed to open device %s", dev_name(dev));
		return 0;
	}

	/* Scan first few sectors for anything looking like a label */
	for (sector = 0; sector < LABEL_SCAN_SECTORS;
	     sector += LABEL_SIZE >> SECTOR_SHIFT) {

		memset(readbuf, 0, sizeof(readbuf));

		if (!dev_read_bytes(dev, sector << SECTOR_SHIFT, LABEL_SIZE, readbuf)) {
			log_error("Failed to read label from %s sector %llu",
				  dev_name(dev), (unsigned long long)sector);
			continue;
		}

		lh = (struct label_header *)readbuf;

		wipe = 0;

		if (!memcmp(lh->id, LABEL_ID, sizeof(lh->id))) {
			if (xlate64(lh->sector_xl) == sector)
				wipe = 1;
		} else {
			dm_list_iterate_items(li, &_labellers) {
				if (li->l->ops->can_handle(li->l, (char *)lh, sector)) {
					wipe = 1;
					break;
				}
			}
		}

		if (wipe) {
			log_very_verbose("%s: Wiping label at sector %llu",
					 dev_name(dev), (unsigned long long)sector);

			if (!dev_write_zeros(dev, sector << SECTOR_SHIFT, LABEL_SIZE)) {
				log_error("Failed to remove label from %s at sector %llu",
					  dev_name(dev), (unsigned long long)sector);
				r = 0;
			} else {
				/* Also remove the PV record from cache. */
				info = lvmcache_info_from_pvid(dev->pvid, dev, 0);
				if (info)
					lvmcache_del(info);
			}
		}
	}

	return r;
}

/* Caller may need to use label_get_handler to create label struct! */
int label_write(struct device *dev, struct label *label)
{
	char buf[LABEL_SIZE] __attribute__((aligned(8)));
	struct label_header *lh = (struct label_header *) buf;
	uint64_t offset;
	int r = 1;

	if (!label->labeller->ops->write) {
		log_error("Label handler does not support label writes");
		return 0;
	}

	if ((LABEL_SIZE + (label->sector << SECTOR_SHIFT)) > LABEL_SCAN_SIZE) {
		log_error("Label sector %" PRIu64 " beyond range (%ld)",
			  label->sector, LABEL_SCAN_SECTORS);
		return 0;
	}

	memset(buf, 0, LABEL_SIZE);

	memcpy(lh->id, LABEL_ID, sizeof(lh->id));
	lh->sector_xl = xlate64(label->sector);
	lh->offset_xl = xlate32(sizeof(*lh));

	if (!(label->labeller->ops->write)(label, buf))
		return_0;

	lh->crc_xl = xlate32(calc_crc(INITIAL_CRC, (uint8_t *)&lh->offset_xl, LABEL_SIZE -
				      ((uint8_t *) &lh->offset_xl - (uint8_t *) lh)));

	log_very_verbose("%s: Writing label to sector %" PRIu64 " with stored offset %"
			 PRIu32 ".", dev_name(dev), label->sector,
			 xlate32(lh->offset_xl));

	if (!label_scan_open(dev)) {
		log_error("Failed to open device %s", dev_name(dev));
		return 0;
	}

	offset = label->sector << SECTOR_SHIFT;

	dev_set_last_byte(dev, offset + LABEL_SIZE);

	if (!dev_write_bytes(dev, offset, LABEL_SIZE, buf)) {
		log_debug_devs("Failed to write label to %s", dev_name(dev));
		return 0;
	}

	dev_unset_last_byte(dev);

	return r;
}

void label_destroy(struct label *label)
{
	label->labeller->ops->destroy_label(label->labeller, label);
	free(label);
}

struct label *label_create(struct labeller *labeller)
{
	struct label *label;

	if (!(label = zalloc(sizeof(*label)))) {
		log_error("label allocaction failed");
		return NULL;
	}

	label->labeller = labeller;

	labeller->ops->initialise_label(labeller, label);

	return label;
}


/* global variable for accessing the bcache populated by label scan */
/* FIXME/TODO convert to cmd_context */
static struct bcache *scan_bcache;

#define BCACHE_BLOCK_SIZE_IN_SECTORS 256 /* 256*512 = 128K */

static bool _in_bcache(struct device *dev)
{
	if (!dev)
		return NULL;
	return (dev->flags & DEV_IN_BCACHE) ? true : false;
}

static struct labeller *_find_lvm_header(struct device *dev,
				   char *headers_buf,
				   size_t headers_buf_size,
				   uint64_t *label_sector,
				   uint64_t block_sector,
				   uint64_t start_sector)
{
	struct labeller_i *li;
	struct labeller *labeller_ret = NULL;
	struct label_header *lh;
	uint64_t sector;
	int found = 0;

	for (sector = start_sector; sector < start_sector + LABEL_SCAN_SECTORS;
	     sector += LABEL_SIZE >> SECTOR_SHIFT) {

		if ((sector * 512) >= headers_buf_size)
			break;

		lh = (struct label_header *) (headers_buf + (sector << SECTOR_SHIFT));

		if (!memcmp(lh->id, LABEL_ID, sizeof(lh->id))) {
			if (found) {
				log_error("Ignoring additional label on %s at sector %llu",
					  dev_name(dev), (unsigned long long)(block_sector + sector));
			}
			if (xlate64(lh->sector_xl) != sector) {
				log_warn("%s: Label for sector %llu found at sector %llu - ignoring.",
					 dev_name(dev),
					 (unsigned long long)xlate64(lh->sector_xl),
					 (unsigned long long)(block_sector + sector));
				continue;
			}
			if (calc_crc(INITIAL_CRC, (uint8_t *)&lh->offset_xl,
				     LABEL_SIZE - ((uint8_t *) &lh->offset_xl - (uint8_t *) lh)) != xlate32(lh->crc_xl)) {
				log_very_verbose("Label checksum incorrect on %s - ignoring", dev_name(dev));
				continue;
			}
			if (found)
				continue;
		}

		dm_list_iterate_items(li, &_labellers) {
			if (li->l->ops->can_handle(li->l, (char *) lh, block_sector + sector)) {
				log_debug("Found label at sector %llu on %s",
					  (unsigned long long)(block_sector + sector), dev_name(dev));
				if (found) {
					log_error("Ignoring additional label on %s at sector %llu",
						  dev_name(dev),
						  (unsigned long long)(block_sector + sector));
					continue;
				}

				labeller_ret = li->l;
				found = 1;

				if (label_sector)
					*label_sector = block_sector + sector;
				break;
			}
		}
	}

	return labeller_ret;
}

/*
 * Process/parse the headers from the data read from a device.
 * Populates lvmcache with device / mda locations / vgname
 * so that vg_read(vgname) will know which devices/locations
 * to read metadata from.
 *
 * If during processing, headers/metadata are found to be needed
 * beyond the range of the scanned block, then additional reads
 * are performed in the processing functions to get that data.
 */
static int _process_block(struct cmd_context *cmd, struct dev_filter *f,
			  struct device *dev, char *headers_buf, size_t headers_buf_size,
			  uint64_t block_sector, uint64_t start_sector,
			  int *is_lvm_device)
{
	char *label_buf;
	struct labeller *labeller;
	uint64_t label_sector = 0;
	int is_duplicate = 0;
	int ret = 0;

	dev->flags &= ~DEV_SCAN_FOUND_LABEL;

	/*
	 * The device may have signatures that exclude it from being processed,
	 * even if it might look like a PV.  Now that the device has been read
	 * and data is available in bcache for it, recheck filters, including
	 * those that use data.  The device needs to be excluded before it
	 * begins to be processed as a PV.
	 */
	if (f) {
		if (!f->passes_filter(cmd, f, dev, NULL)) {
			/*
			 * If this device was previously scanned (not common)
			 * and if it passed filters at that point, lvmcache
			 * info may have been saved for it.  Now the same
			 * device is being scanned again, and it may fail
			 * filters this time.  If the caller did not clear
			 * lvmcache info for this dev before rescanning, do
			 * that now.  It's unlikely this is actually needed.
			 */
			if (dev->pvid[0]) {
				log_print_unless_silent("Clear pvid and info for filtered dev %s.", dev_name(dev));
				lvmcache_del_dev(dev);
				memset(dev->pvid, 0, sizeof(dev->pvid));
			}

			*is_lvm_device = 0;
			goto_out;
		}
	}

	/*
	 * Finds the data sector containing the label.
	 */
	if (!(labeller = _find_lvm_header(dev, headers_buf, headers_buf_size, &label_sector, block_sector, start_sector))) {

		/*
		 * Non-PVs exit here
		 *
		 * FIXME: check for PVs with errors that also exit here!
		 * i.e. this code cannot distinguish between a non-lvm
		 * device an an lvm device with errors.
		 */

		log_very_verbose("%s: No lvm label detected", dev_name(dev));

		/* See comment above */
		if (dev->pvid[0]) {
			log_print_unless_silent("Clear pvid and info for no lvm header %s", dev_name(dev));
			lvmcache_del_dev(dev);
			memset(dev->pvid, 0, sizeof(dev->pvid));
		}

		dev->flags |= DEV_SCAN_FOUND_NOLABEL;
		*is_lvm_device = 0;
		goto out;
	}

	dev->flags |= DEV_SCAN_FOUND_LABEL;
	*is_lvm_device = 1;
	label_buf = headers_buf + (label_sector * 512);

	/*
	 * This is the point where the scanning code dives into the rest of
	 * lvm.  ops->read() is _text_read() which reads the pv_header, mda
	 * locations, and metadata text.  All of the info it finds about the PV
	 * and VG is stashed in lvmcache which saves it in the form of
	 * info/vginfo structs.  That lvmcache info is used later when the
	 * command wants to read the VG to do something to it.
	 */
	ret = labeller->ops->read(cmd, labeller, dev, label_buf, label_sector, &is_duplicate);

	if (!ret) {
		if (is_duplicate) {
			/*
			 * _text_read() called lvmcache_add() which found an
			 * existing info struct for this PVID but for a
			 * different dev.  lvmcache_add() did not add an info
			 * struct for this dev, but added this dev to the list
			 * of duplicate devs.
			 */
			log_debug("label scan found duplicate PVID %s on %s", dev->pvid, dev_name(dev));
		} else {
			/*
			 * Leave the info in lvmcache because the device is
			 * present and can still be used even if it has
			 * metadata that we can't process (we can get metadata
			 * from another PV/mda.) _text_read only saves mdas
			 * with good metadata in lvmcache (this includes old
			 * metadata), and if a PV has no mdas with good
			 * metadata, then the info for the PV will be in
			 * lvmcache with empty info->mdas, and it will behave
			 * like a PV with no mdas (a common configuration.)
			 */
			log_warn("WARNING: scan failed to get metadata summary from %s PVID %s", dev_name(dev), dev->pvid);
		}
	}
 out:
	return ret;
}

static int _scan_dev_open(struct device *dev)
{
	struct dm_list *name_list;
	struct dm_str_list *name_sl;
	const char *name;
	const char *modestr;
	struct stat sbuf;
	int flags = 0;
	int fd, di;

	if (!dev)
		return 0;

	if (dev->flags & DEV_IN_BCACHE) {
		/* Shouldn't happen */
		log_error("Device open %s has DEV_IN_BCACHE already set", dev_name(dev));
		dev->flags &= ~DEV_IN_BCACHE;
	}

	if (dev->bcache_di != -1) {
		/* Shouldn't happen */
		log_error("Device open %s already open with di %d fd %d",
			  dev_name(dev), dev->bcache_di, dev->bcache_fd);
		return 0;
	}

 next_name:
	/*
	 * All the names for this device (major:minor) are kept on
	 * dev->aliases, the first one is the primary/preferred name.
	 *
	 * The default name preferences in dev-cache mean that the first
	 * name in dev->aliases is not a symlink for scsi devices, but is
	 * the /dev/mapper/ symlink for mpath devices.
	 *
	 * If preferred names are set to symlinks, should this
	 * first attempt to open using a non-symlink?
	 *
	 * dm_list_first() returns NULL if the list is empty.
	 */
	if (!(name_list = dm_list_first(&dev->aliases))) {
		log_error("Device open %u:%u has no path names.",
			  MAJOR(dev->dev), MINOR(dev->dev));
		return 0;
	}
	name_sl = dm_list_item(name_list, struct dm_str_list);
	name = name_sl->str;

	flags |= O_DIRECT;
	flags |= O_NOATIME;

	/*
	 * FIXME: udev is a train wreck when we open RDWR and close, so we
	 * need to only use RDWR when we actually need to write, and use
	 * RDONLY otherwise.  Fix, disable or scrap udev nonsense so we can
	 * just open with RDWR by default.
	 */

	if (dev->flags & DEV_BCACHE_EXCL) {
		flags |= O_EXCL;
		flags |= O_RDWR;
		modestr = "rwex";
	} else if (dev->flags & DEV_BCACHE_WRITE) {
		flags |= O_RDWR;
		modestr = "rw";
	} else {
		flags |= O_RDONLY;
		modestr = "ro";
	}

	fd = open(name, flags, 0777);
	if (fd < 0) {
		if ((errno == EBUSY) && (flags & O_EXCL)) {
			log_error("Can't open %s exclusively.  Mounted filesystem?",
				  dev_name(dev));
			return 0;
		} else {
			/*
			 * drop name from dev->aliases and use verify_aliases to
			 * drop any other invalid aliases before retrying open with
			 * any remaining valid paths.
			 */
			log_debug("Drop alias for %u:%u failed open %s (%d).",
				  MAJOR(dev->dev), MINOR(dev->dev), name, errno);
			dev_cache_failed_path(dev, name);
			dev_cache_verify_aliases(dev);
			goto next_name;
		}
	}

	/* Verify that major:minor from the path still match dev. */
	if ((fstat(fd, &sbuf) < 0) || (sbuf.st_rdev != dev->dev)) {
		log_warn("Invalid path %s for device %u:%u, trying different path.",
			 name, MAJOR(dev->dev), MINOR(dev->dev));
		(void)close(fd);
		dev_cache_failed_path(dev, name);
		dev_cache_verify_aliases(dev);
		goto next_name;
	}

	dev->flags |= DEV_IN_BCACHE;
	dev->bcache_fd = fd;

	di = bcache_set_fd(fd);

	if (di == -1) {
		log_error("Failed to set bcache fd.");
		if (close(fd))
			log_sys_debug("close", name);
		dev->bcache_fd = -1;
		return 0;
	}

	log_debug("open %s %s di %d fd %d", dev_name(dev), modestr, di, fd);

	dev->bcache_di = di;

	return 1;
}

static int _scan_dev_close(struct device *dev)
{
	if (!(dev->flags & DEV_IN_BCACHE))
		log_error("scan_dev_close %s no DEV_IN_BCACHE set", dev_name(dev));

	dev->flags &= ~DEV_IN_BCACHE;
	dev->flags &= ~DEV_BCACHE_EXCL;
	dev->flags &= ~DEV_BCACHE_WRITE;

	if (dev->bcache_di == -1) {
		log_error("scan_dev_close %s already closed", dev_name(dev));
		return 0;
	}