OpenZFS 8067 - zdb should be able to dump literal embedded block pointer

Authored by: Matthew Ahrens <[email protected]>
Reviewed by: George Wilson <[email protected]>
Reviewed by: Alex Reece <[email protected]>
Reviewed by: Yuri Pankov <[email protected]>
Approved by: Robert Mustacchi <[email protected]>
Reviewed-by: Brian Behlendorf <[email protected]>
Ported-by: Giuseppe Di Natale <[email protected]>

OpenZFS-issue: https://www.illumos.org/issues/8067
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/8173085
Closes #6319

4 files changed, 95 insertions, 5 deletions

diff --git a/cmd/zdb/zdb.c b/cmd/zdb/zdb.c
index 6c4071060..55a6c4fa1 100644
--- a/cmd/zdb/zdb.c
+++ b/cmd/zdb/zdb.c
@@ -63,6 +63,7 @@
 #include <sys/ddt.h>
 #include <sys/zfeature.h>
 #include <sys/abd.h>
+#include <sys/blkptr.h>
 #include <zfs_comutil.h>
 #include <libzfs.h>
 
@@ -137,10 +138,11 @@ usage(void)
 	    "\t%s -O <dataset> <path>\n"
 	    "\t%s -R [-A] [-e [-V] [-p <path> ...]] [-U <cache>]\n"
 	    "\t\t<poolname> <vdev>:<offset>:<size>[:<flags>]\n"
+	    "\t%s -E [-A] word0:word1:...:word15\n"
 	    "\t%s -S [-AP] [-e [-V] [-p <path> ...]] [-U <cache>] "
 	    "<poolname>\n\n",
 	    cmdname, cmdname, cmdname, cmdname, cmdname, cmdname, cmdname,
-	    cmdname);
+	    cmdname, cmdname);
 
 	(void) fprintf(stderr, "    Dataset name must include at least one "
 	    "separator character '/' or '@'\n");
@@ -155,6 +157,8 @@ usage(void)
 	(void) fprintf(stderr, "        -C config (or cachefile if alone)\n");
 	(void) fprintf(stderr, "        -d dataset(s)\n");
 	(void) fprintf(stderr, "        -D dedup statistics\n");
+	(void) fprintf(stderr, "        -E decode and display block from an "
+	    "embedded block pointer\n");
 	(void) fprintf(stderr, "        -h pool history\n");
 	(void) fprintf(stderr, "        -i intent logs\n");
 	(void) fprintf(stderr, "        -l read label contents\n");
@@ -4092,6 +4096,35 @@ out:
 	free(dup);
 }
 
+static void
+zdb_embedded_block(char *thing)
+{
+	blkptr_t bp;
+	unsigned long long *words = (void *)&bp;
+	char buf[SPA_MAXBLOCKSIZE];
+	int err;
+
+	memset(&bp, 0, sizeof (blkptr_t));
+
+	err = sscanf(thing, "%llx:%llx:%llx:%llx:%llx:%llx:%llx:%llx:"
+	    "%llx:%llx:%llx:%llx:%llx:%llx:%llx:%llx",
+	    words + 0, words + 1, words + 2, words + 3,
+	    words + 4, words + 5, words + 6, words + 7,
+	    words + 8, words + 9, words + 10, words + 11,
+	    words + 12, words + 13, words + 14, words + 15);
+	if (err != 16) {
+		(void) printf("invalid input format\n");
+		exit(1);
+	}
+	ASSERT3U(BPE_GET_LSIZE(&bp), <=, SPA_MAXBLOCKSIZE);
+	err = decode_embedded_bp(&bp, buf, BPE_GET_LSIZE(&bp));
+	if (err != 0) {
+		(void) printf("decode failed: %u\n", err);
+		exit(1);
+	}
+	zdb_dump_block_raw(buf, BPE_GET_LSIZE(&bp), 0);
+}
+
 static boolean_t
 pool_match(nvlist_t *cfg, char *tgt)
 {
@@ -4210,13 +4243,14 @@ main(int argc, char **argv)
 		spa_config_path = spa_config_path_env;
 
 	while ((c = getopt(argc, argv,
-	    "AbcCdDeFGhiI:lLmMo:Op:PqRsSt:uU:vVx:X")) != -1) {
+	    "AbcCdDeEFGhiI:lLmMo:Op:PqRsSt:uU:vVx:X")) != -1) {
 		switch (c) {
 		case 'b':
 		case 'c':
 		case 'C':
 		case 'd':
 		case 'D':
+		case 'E':
 		case 'G':
 		case 'h':
 		case 'i':
@@ -4280,6 +4314,12 @@ main(int argc, char **argv)
 			break;
 		case 'U':
 			spa_config_path = optarg;
+			if (spa_config_path[0] != '/') {
+				(void) fprintf(stderr,
+				    "cachefile must be an absolute path "
+				    "(i.e. start with a slash)\n");
+				usage();
+			}
 			break;
 		case 'v':
 			verbose++;
@@ -4331,7 +4371,7 @@ main(int argc, char **argv)
 		verbose = MAX(verbose, 1);
 
 	for (c = 0; c < 256; c++) {
-		if (dump_all && strchr("AeFlLOPRSX", c) == NULL)
+		if (dump_all && strchr("AeEFlLOPRSX", c) == NULL)
 			dump_opt[c] = 1;
 		if (dump_opt[c])
 			dump_opt[c] += verbose;
@@ -4345,6 +4385,14 @@ main(int argc, char **argv)
 
 	if (argc < 2 && dump_opt['R'])
 		usage();
+
+	if (dump_opt['E']) {
+		if (argc != 1)
+			usage();
+		zdb_embedded_block(argv[0]);
+		return (0);
+	}
+
 	if (argc < 1) {
 		if (!dump_opt['e'] && dump_opt['C']) {
 			dump_cachefile(spa_config_path);
diff --git a/include/sys/blkptr.h b/include/sys/blkptr.h
index b720482a7..77b1b827a 100644
--- a/include/sys/blkptr.h
+++ b/include/sys/blkptr.h
@@ -30,6 +30,7 @@ extern "C" {
 void encode_embedded_bp_compressed(blkptr_t *, void *,
     enum zio_compress, int, int);
 void decode_embedded_bp_compressed(const blkptr_t *, void *);
+int decode_embedded_bp(const blkptr_t *, void *, int);
 
 #ifdef	__cplusplus
 }
diff --git a/man/man8/zdb.8 b/man/man8/zdb.8
index c899db395..4e47de7be 100644
--- a/man/man8/zdb.8
+++ b/man/man8/zdb.8
@@ -10,12 +10,12 @@
 .\"
 .\"
 .\" Copyright 2012, Richard Lowe.
-.\" Copyright (c) 2012, 2016 by Delphix. All rights reserved.
+.\" Copyright (c) 2012, 2017 by Delphix. All rights reserved.
 .\" Copyright 2017 Nexenta Systems, Inc.
 .\" Copyright (c) 2017 Lawrence Livermore National Security, LLC.
 .\" Copyright (c) 2017 Intel Corporation.
 .\"
-.Dd April 11, 2017
+.Dd April 14, 2017
 .Dt ZDB 8 SMM
 .Os Linux
 .Sh NAME
@@ -41,6 +41,10 @@
 .Op Fl A
 .Op Fl U Ar cache
 .Nm
+.Fl E
+.Op Fl A
+.Ar word0 Ns : Ns Ar word1 Ns :...: Ns Ar word15
+.Nm
 .Fl l
 .Op Fl Aqu
 .Ar device
@@ -155,6 +159,10 @@ Display the statistics independently for each deduplication table.
 Dump the contents of the deduplication tables describing duplicate blocks.
 .It Fl DDDDD
 Also dump the contents of the deduplication tables describing unique blocks.
+.It Fl E Ar word0 Ns : Ns Ar word1 Ns :...: Ns Ar word15
+Decode and display block from an embedded block pointer specified by the
+.Ar word
+arguments.
 .It Fl h
 Display pool history similar to
 .Nm zpool Cm history ,
diff --git a/module/zfs/blkptr.c b/module/zfs/blkptr.c
index 99accfa0f..bb407af03 100644
--- a/module/zfs/blkptr.c
+++ b/module/zfs/blkptr.c
@@ -119,3 +119,36 @@ decode_embedded_bp_compressed(const blkptr_t *bp, void *buf)
 		buf8[i] = BF64_GET(w, (i % sizeof (w)) * NBBY, NBBY);
 	}
 }
+
+/*
+ * Fill in the buffer with the (decompressed) payload of the embedded
+ * blkptr_t.  Takes into account compression and byteorder (the payload is
+ * treated as a stream of bytes).
+ * Return 0 on success, or ENOSPC if it won't fit in the buffer.
+ */
+int
+decode_embedded_bp(const blkptr_t *bp, void *buf, int buflen)
+{
+	int lsize, psize;
+
+	ASSERT(BP_IS_EMBEDDED(bp));
+
+	lsize = BPE_GET_LSIZE(bp);
+	psize = BPE_GET_PSIZE(bp);
+
+	if (lsize > buflen)
+		return (ENOSPC);
+	ASSERT3U(lsize, ==, buflen);
+
+	if (BP_GET_COMPRESS(bp) != ZIO_COMPRESS_OFF) {
+		uint8_t dstbuf[BPE_PAYLOAD_SIZE];
+		decode_embedded_bp_compressed(bp, dstbuf);
+		VERIFY0(zio_decompress_data_buf(BP_GET_COMPRESS(bp),
+		    dstbuf, buf, psize, buflen));
+	} else {
+		ASSERT3U(lsize, ==, psize);
+		decode_embedded_bp_compressed(bp, buf);
+	}
+
+	return (0);
+}