ZFS: make new pool from disk detached from mirror

zhack.c

/*
 * CDDL HEADER START
 *
 * The contents of this file are subject to the terms of the
 * Common Development and Distribution License, Version 1.0 only
 * (the "License").  You may not use this file except in compliance
 * with the License.
 *
 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
 * or http://www.opensolaris.org/os/licensing.
 * See the License for the specific language governing permissions
 * and limitations under the License.
 *
 * When distributing Covered Code, include this CDDL HEADER in each
 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
 * If applicable, add the following below this CDDL HEADER, with the
 * fields enclosed by brackets "[]" replaced with your own identifying
 * information: Portions Copyright [yyyy] [name of copyright owner]
 *
 * CDDL HEADER END
 */
/*
 * Copyright 2005 Sun Microsystems, Inc.  All rights reserved.
 * Use is subject to license terms.
 */
/*
 * Copyright 2013 Saso Kiselkov.  All rights reserved.
 * Copyright 2015 Toomas Soome <tsoome@me.com>
 */

/*
 * SHA-256 and SHA-512/256 hashes, as specified in FIPS 180-4, available at:
 * http://csrc.nist.gov/cryptval
 *
 * This is a very compact implementation of SHA-256 and SHA-512/256.
 * It is designed to be simple and portable, not to be fast.
 */

#include <stdio.h>
#include <stdlib.h>
#include <sys/fs/zfs.h>
#include <libzutil.h>
#include <sys/stat.h>
#include <unistd.h>
#include <fcntl.h>
#include <stddef.h>
#include <sys/zio.h>
#include <sys/vdev_impl.h>
#include <libzfs/libnvpair.h>


/*
 * The literal definitions according to FIPS180-4 would be:
 *
 * 	Ch(x, y, z)     (((x) & (y)) ^ ((~(x)) & (z)))
 * 	Maj(x, y, z)    (((x) & (y)) | ((x) & (z)) | ((y) & (z)))
 *
 * We use logical equivalents which require one less op.
 */
#include <stdint.h>
#include <sys/spa_checksum.h>

#define __unused __attribute__((unused))

#define        Ch(x, y, z)        ((z) ^ ((x) & ((y) ^ (z))))
#define        Maj(x, y, z)        (((x) & (y)) ^ ((z) & ((x) ^ (y))))
#define        ROTR(x, n)        (((x) >> (n)) | ((x) << ((sizeof (x) * NBBY)-(n))))

/* SHA-224/256 operations */
#define        BIGSIGMA0_256(x)        (ROTR(x, 2) ^ ROTR(x, 13) ^ ROTR(x, 22))
#define        BIGSIGMA1_256(x)        (ROTR(x, 6) ^ ROTR(x, 11) ^ ROTR(x, 25))
#define        SIGMA0_256(x)                (ROTR(x, 7) ^ ROTR(x, 18) ^ ((x) >> 3))
#define        SIGMA1_256(x)                (ROTR(x, 17) ^ ROTR(x, 19) ^ ((x) >> 10))

/* SHA-256 round constants */
static const uint32_t SHA256_K[64] = {
	0x428a2f98, 0x71374491, 0xb5c0fbcf, 0xe9b5dba5,
	0x3956c25b, 0x59f111f1, 0x923f82a4, 0xab1c5ed5,
	0xd807aa98, 0x12835b01, 0x243185be, 0x550c7dc3,
	0x72be5d74, 0x80deb1fe, 0x9bdc06a7, 0xc19bf174,
	0xe49b69c1, 0xefbe4786, 0x0fc19dc6, 0x240ca1cc,
	0x2de92c6f, 0x4a7484aa, 0x5cb0a9dc, 0x76f988da,
	0x983e5152, 0xa831c66d, 0xb00327c8, 0xbf597fc7,
	0xc6e00bf3, 0xd5a79147, 0x06ca6351, 0x14292967,
	0x27b70a85, 0x2e1b2138, 0x4d2c6dfc, 0x53380d13,
	0x650a7354, 0x766a0abb, 0x81c2c92e, 0x92722c85,
	0xa2bfe8a1, 0xa81a664b, 0xc24b8b70, 0xc76c51a3,
	0xd192e819, 0xd6990624, 0xf40e3585, 0x106aa070,
	0x19a4c116, 0x1e376c08, 0x2748774c, 0x34b0bcb5,
	0x391c0cb3, 0x4ed8aa4a, 0x5b9cca4f, 0x682e6ff3,
	0x748f82ee, 0x78a5636f, 0x84c87814, 0x8cc70208,
	0x90befffa, 0xa4506ceb, 0xbef9a3f7, 0xc67178f2
};

static void
SHA256Transform(uint32_t *H, const uint8_t *cp) {
	uint32_t a, b, c, d, e, f, g, h, t, T1, T2, W[64];

	/* copy chunk into the first 16 words of the message schedule */
	for (t = 0; t < 16; t++, cp += sizeof(uint32_t))
		W[t] = (cp[0] << 24) | (cp[1] << 16) | (cp[2] << 8) | cp[3];

	/* extend the first 16 words into the remaining 48 words */
	for (t = 16; t < 64; t++)
		W[t] = SIGMA1_256(W[t - 2]) + W[t - 7] +
		       SIGMA0_256(W[t - 15]) + W[t - 16];

	/* init working variables to the current hash value */
	a = H[0];
	b = H[1];
	c = H[2];
	d = H[3];
	e = H[4];
	f = H[5];
	g = H[6];
	h = H[7];

	/* iterate the compression function for all rounds of the hash */
	for (t = 0; t < 64; t++) {
		T1 = h + BIGSIGMA1_256(e) + Ch(e, f, g) + SHA256_K[t] + W[t];
		T2 = BIGSIGMA0_256(a) + Maj(a, b, c);
		h = g;
		g = f;
		f = e;
		e = d + T1;
		d = c;
		c = b;
		b = a;
		a = T1 + T2;
	}

	/* add the compressed chunk to the current hash value */
	H[0] += a;
	H[1] += b;
	H[2] += c;
	H[3] += d;
	H[4] += e;
	H[5] += f;
	H[6] += g;
	H[7] += h;
}

/*
 * Implements the SHA-224 and SHA-256 hash algos - to select between them
 * pass the appropriate initial values of 'H' and truncate the last 32 bits
 * in case of SHA-224.
 */
static void
SHA256(uint32_t *H, const void *buf, uint64_t size, zio_cksum_t *zcp) {
	uint8_t pad[128];
	unsigned padsize = size & 63;
	unsigned i, k;

	/* process all blocks up to the last one */
	for (i = 0; i < size - padsize; i += 64)
		SHA256Transform(H, (uint8_t *) buf + i);

	/* process the last block and padding */
	for (k = 0; k < padsize; k++)
		pad[k] = ((uint8_t *) buf)[k + i];

	for (pad[padsize++] = 0x80; (padsize & 63) != 56; padsize++)
		pad[padsize] = 0;

	for (i = 0; i < 8; i++)
		pad[padsize++] = (size << 3) >> (56 - 8 * i);

	for (i = 0; i < padsize; i += 64)
		SHA256Transform(H, pad + i);

	ZIO_SET_CHECKSUM(zcp,
			 (uint64_t) H[0] << 32 | H[1],
			 (uint64_t) H[2] << 32 | H[3],
			 (uint64_t) H[4] << 32 | H[5],
			 (uint64_t) H[6] << 32 | H[7]);
}

static void
zio_checksum_SHA256(const void *buf, uint64_t size, zio_cksum_t *zcp) {
	/* SHA-256 as per FIPS 180-4. */
	uint32_t H[] = {
		0x6a09e667, 0xbb67ae85, 0x3c6ef372, 0xa54ff53a,
		0x510e527f, 0x9b05688c, 0x1f83d9ab, 0x5be0cd19
	};
	SHA256(H, buf, size, zcp);
}

static void label_write(int fd, uint64_t offset, uint64_t size, void *buf) {
	zio_eck_t *zbt, zbt_orig;
	zbt = (zio_eck_t *) ((char *) buf + size) - 1;
	zbt_orig = *zbt;
	zio_cksum_t zc;

	ZIO_SET_CHECKSUM(&zbt->zec_cksum, offset, 0, 0, 0);
	zio_checksum_SHA256(buf, size, &zc);
	zbt->zec_cksum = zc;

	printf("writing new label... ");
	VERIFY(pwrite(fd, buf, size, offset) == size);
	printf("Ok\n");

	*zbt = zbt_orig;
}

void usage(void) {
	fprintf(stderr, "error: no device specified\n");
	printf("Read/write label from detached zfs vdev\n");
	printf("Usage: zhack [-s] [-w] device\n");
	printf("        -w               write new label\n");
	printf("        -s               show pool config in text format\n");
	printf("        device           path to device or file with pool data\n\n");
	printf("Examples:\n");
	printf("	check and show label config from file with label copy (made with dd)\n");
	printf("		zhack -s ada2p4.eli.256K.copy\n\n");
	printf("	write new label in file\n");
	printf("		zhack -w ada2p4.eli.256K.copy\n\n");
	printf("	write new label directly to device\n");
	printf("		zhack -w /dev/ada2p4.eli\n\n");

	exit(1);
}

// based on https://gist.github.com/jjwhitney/baaa63144da89726e482
int main(int argc, char **argv) {
	int fd;
	vdev_label_t vl;
	nvlist_t *config;
	uberblock_t *ub = (uberblock_t *) vl.vl_uberblock;
	uint64_t txg;
	char *buf;
	size_t buflen;
	int c;
	int write_flag = 0;
	int show_config = 0;

	while ((c = getopt(argc, argv, "ws")) != -1) {
		switch (c) {
			case 'w':
				write_flag = 1;
				break;
			case 's':
				show_config = 1;
				break;
			default:
				usage();
				break;
		}
	}

	argc -= optind;
	argv += optind;
	optind = 1;

	if (argc == 0) {
		usage();
	}

	// open device
	VERIFY((fd = open(argv[0], O_RDWR)) != -1);

	// read 256K as vdev label `vl`
	VERIFY(pread(fd, &vl, sizeof(vdev_label_t), 0) == sizeof(vdev_label_t));

	// unpack config
	config = fnvlist_unpack(vl.vl_vdev_phys.vp_nvlist, sizeof(vl.vl_vdev_phys.vp_nvlist));

	if (show_config) {
		// print current pool config
		printf("Current pool config:\n");
		nvlist_print(stdout, config);
	}

	// get current txg
	txg = fnvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG);
	printf("current txg: %lu (==0)\n", txg);

	/* check this pool is detached */
	VERIFY(txg == 0);
	VERIFY(ub->ub_txg == 0);
	VERIFY(ub->ub_rootbp.blk_birth != 0); // we will make this value new txg
	printf("blk_birth: %lu\n", ub->ub_rootbp.blk_birth);

	// set new txg in uberblock
	txg = ub->ub_rootbp.blk_birth;
	ub->ub_txg = txg;
	printf("new txg set in uberblock: %lu\n", ub->ub_txg);

	// set current timestamp
	ub->ub_timestamp = gethrestime_sec();
	time_t timestamp = ub->ub_timestamp;
	printf("new timestamp: %llu UTC: %s", (u_longlong_t) ub->ub_timestamp, asctime(localtime(&timestamp)));

	// remove old txg value...
	fnvlist_remove(config, ZPOOL_CONFIG_POOL_TXG);
	// ...and set new one
	fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_TXG, txg);

	// get current ashift
	nvlist_t *vdev_tree = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE);
	uint64_t ashift = fnvlist_lookup_uint64(vdev_tree, ZPOOL_CONFIG_ASHIFT);
	printf("ashift: %lu\n", ashift);

	VERIFY(ashift != 0);

	// pack config
	buf = vl.vl_vdev_phys.vp_nvlist;
	buflen = sizeof(vl.vl_vdev_phys.vp_nvlist);
	VERIFY(nvlist_pack(config, &buf, &buflen, NV_ENCODE_XDR, 0) == 0);

	if (show_config) {
		// print changed pool config
		printf("Changed pool config:\n");
		nvlist_print(stdout, config);
	}

	if (write_flag) {
		// write new label
		label_write(fd, offsetof(vdev_label_t, vl_uberblock), 1ULL << (ashift), ub);
		label_write(fd, offsetof(vdev_label_t, vl_vdev_phys), VDEV_PHYS_SIZE, &vl.vl_vdev_phys);
		fsync(fd);
	}

	return (0);
}

works with zfs 0.8.3 source tree and FreeBSD 12.1 zfs labels.

replace zhack utility in zfs source tree with this file and run make.

based on https://gist.github.com/jjwhitney/baaa63144da89726e482
more info:

• https://unix.stackexchange.com/a/566238/172911
• https://thr3ads.net/zfs-discuss/2008/04/351635-recovering-data-from-a-dettach-mirrored-vdev
• http://www.giis.co.in/Zfs_ondiskformat.pdf

This utility works with label0 (first 256K of the vdev disk), so it's better to create a copy of the label with:

dd if=/dev/ada2p4.eli of=./ada2p4.256K.orig bs=256K count=1

and restore it with

dd if=./ada2p4.256K.orig of=/dev/ada2p4.eli bs=256K count=1 conv=notrunc

tip: import this pool with -N option to avoid mount filesystems.