Illumos 4891 - want zdb option to dump all metadata

4891 want zdb option to dump all metadata
Reviewed by: Sonu Pillai <sonu.pillai@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Christopher Siden <christopher.siden@delphix.com>
Reviewed by: Dan McDonald <danmcd@omniti.com>
Reviewed by: Richard Lowe <richlowe@richlowe.net>
Approved by: Garrett D'Amore <garrett@damore.org>

We'd like a way for zdb to dump metadata in a machine-readable
format, so that we can bring that back from a customer site for
in-house diagnosis.  Think of it as a crash dump for zpools,
which can be used for post-mortem analysis of a malfunctioning
pool

References:
  https://www.illumos.org/issues/4891
  illumos/illumos-gate@df15e41

Porting notes:
- [cmd/zdb/zdb.c]
  - a5778ea zdb: Introduce -V for verbatim import
  - In main() getopt 'opt' variable removed and the code was
    brought back in line with illumos.
- [lib/libzpool/kernel.c]
  - 1e33ac1 Fix Solaris thread dependency by using pthreads
  - f0e324f Update utsname support
  - 4d58b69 Fix vn_open/vn_rdwr error handling
  - In vn_open() allocate 'dumppath' on heap instead of stack
  - Properly handle 'dump_fd == -1' error path
  - Free 'realpath' after added vn_dumpdir_code block

Ported-by: kernelOfTruth kerneloftruth@gmail.com
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>

Author: ahrens

cmd/zdb/zdb.c

@@ -118,7 +118,7 @@ usage(void)
 {
 	(void) fprintf(stderr,
 	    "Usage: %s [-CumMdibcsDvhLXFPA] [-t txg] [-e [-p path...]] "
-	    "[-U config] [-I inflight I/Os] poolname [object...]\n"
+	    "[-U config] [-I inflight I/Os] [-x dumpdir] poolname [object...]\n"
 	    "       %s [-divPA] [-e -p path...] [-U config] dataset "
 	    "[object...]\n"
 	    "       %s -mM [-LXFPA] [-t txg] [-e [-p path...]] [-U config] "
@@ -157,7 +157,7 @@ usage(void)
 	(void) fprintf(stderr, "        -R read and display block from a "
 	    "device\n\n");
 	(void) fprintf(stderr, "    Below options are intended for use "
-	    "with other options (except -l):\n");
+	    "with other options:\n");
 	(void) fprintf(stderr, "        -A ignore assertions (-A), enable "
 	    "panic recovery (-AA) or both (-AAA)\n");
 	(void) fprintf(stderr, "        -F attempt automatic rewind within "
@@ -170,12 +170,14 @@ usage(void)
 	    "has altroot/not in a cachefile\n");
 	(void) fprintf(stderr, "        -p <path> -- use one or more with "
 	    "-e to specify path to vdev dir\n");
+	(void) fprintf(stderr, "        -x <dumpdir> -- "
+	    "dump all read blocks into specified directory\n");
 	(void) fprintf(stderr, "        -P print numbers in parseable form\n");
 	(void) fprintf(stderr, "        -t <txg> -- highest txg to use when "
 	    "searching for uberblocks\n");
 	(void) fprintf(stderr, "        -I <number of inflight I/Os> -- "
-	    "specify the maximum number of checksumming I/Os "
-	    "[default is 200]\n");
+	    "specify the maximum number of "
+	    "checksumming I/Os [default is 200]\n");
 	(void) fprintf(stderr, "Specify an option more than once (e.g. -bb) "
 	    "to make only that option verbose\n");
 	(void) fprintf(stderr, "Default is to dump everything non-verbosely\n");
@@ -3626,7 +3628,6 @@ main(int argc, char **argv)
 	int flags = ZFS_IMPORT_MISSING_LOG;
 	int rewind = ZPOOL_NEVER_REWIND;
 	char *spa_config_path_env;
-	const char *opts = "bcdhilmMI:suCDRSAFLXevp:t:U:PV";
 	boolean_t target_is_spa = B_TRUE;
 
 	(void) setrlimit(RLIMIT_NOFILE, &rl);
@@ -3643,7 +3644,8 @@ main(int argc, char **argv)
 	if (spa_config_path_env != NULL)
 		spa_config_path = spa_config_path_env;
 
-	while ((c = getopt(argc, argv, opts)) != -1) {
+	while ((c = getopt(argc, argv,
+	    "bcdhilmMI:suCDRSAFLXx:evp:t:U:PV")) != -1) {
 		switch (c) {
 		case 'b':
 		case 'c':
@@ -3697,6 +3699,9 @@ main(int argc, char **argv)
 			}
 			searchdirs[nsearch++] = optarg;
 			break;
+		case 'x':
+			vn_dumpdir = optarg;
+			break;
 		case 't':
 			max_txg = strtoull(optarg, NULL, 0);
 			if (max_txg < TXG_INITIAL) {

include/sys/zfs_context.h

@@ -500,8 +500,10 @@ typedef struct vnode {
 	uint64_t	v_size;
 	int		v_fd;
 	char		*v_path;
+	int		v_dump_fd;
 } vnode_t;
 
+extern char *vn_dumpdir;
 #define	AV_SCANSTAMP_SZ	32		/* length of anti-virus scanstamp */
 
 typedef struct xoptattr {

lib/libzpool/kernel.c

@@ -29,6 +29,7 @@
 #include <stdlib.h>
 #include <string.h>
 #include <zlib.h>
+#include <libgen.h>
 #include <sys/signal.h>
 #include <sys/spa.h>
 #include <sys/stat.h>
@@ -50,6 +51,9 @@ char hw_serial[HW_HOSTID_LEN];
 struct utsname hw_utsname;
 vmem_t *zio_arena = NULL;
 
+/* If set, all blocks read will be copied to the specified directory. */
+char *vn_dumpdir = NULL;
+
 /* this only exists to have its address taken */
 struct proc p0;
 
@@ -588,6 +592,7 @@ int
 vn_open(char *path, int x1, int flags, int mode, vnode_t **vpp, int x2, int x3)
 {
 	int fd;
+	int dump_fd;
 	vnode_t *vp;
 	int old_umask = 0;
 	char *realpath;
@@ -655,13 +660,31 @@ vn_open(char *path, int x1, int flags, int mode, vnode_t **vpp, int x2, int x3)
 	 * FREAD and FWRITE to the corresponding O_RDONLY, O_WRONLY, and O_RDWR.
 	 */
 	fd = open64(realpath, flags - FREAD, mode);
-	free(realpath);
+	err = errno;
 
 	if (flags & FCREAT)
 		(void) umask(old_umask);
 
+	if (vn_dumpdir != NULL) {
+		char *dumppath = umem_zalloc(MAXPATHLEN, UMEM_NOFAIL);
+		(void) snprintf(dumppath, MAXPATHLEN,
+		    "%s/%s", vn_dumpdir, basename(realpath));
+		dump_fd = open64(dumppath, O_CREAT | O_WRONLY, 0666);
+		umem_free(dumppath, MAXPATHLEN);
+		if (dump_fd == -1) {
+			err = errno;
+			free(realpath);
+			close(fd);
+			return (err);
+		}
+	} else {
+		dump_fd = -1;
+	}
+
+	free(realpath);
+
 	if (fd == -1)
-		return (errno);
+		return (err);
 
 	if (fstat64_blk(fd, &st) == -1) {
 		err = errno;
@@ -676,6 +699,7 @@ vn_open(char *path, int x1, int flags, int mode, vnode_t **vpp, int x2, int x3)
 	vp->v_fd = fd;
 	vp->v_size = st.st_size;
 	vp->v_path = spa_strdup(path);
+	vp->v_dump_fd = dump_fd;
 
 	return (0);
 }
@@ -708,6 +732,11 @@ vn_rdwr(int uio, vnode_t *vp, void *addr, ssize_t len, offset_t offset,
 
 	if (uio == UIO_READ) {
 		rc = pread64(vp->v_fd, addr, len, offset);
+		if (vp->v_dump_fd != -1) {
+			int status =
+			    pwrite64(vp->v_dump_fd, addr, rc, offset);
+			ASSERT(status != -1);
+		}
 	} else {
 		/*
 		 * To simulate partial disk writes, we split writes into two
@@ -750,6 +779,8 @@ void
 vn_close(vnode_t *vp)
 {
 	close(vp->v_fd);
+	if (vp->v_dump_fd != -1)
+		close(vp->v_dump_fd);
 	spa_strfree(vp->v_path);
 	umem_free(vp, sizeof (vnode_t));
 }

man/man8/zdb.8

@@ -11,7 +11,7 @@
 .\"
 .\"
 .\" Copyright 2012, Richard Lowe.
-.\" Copyright (c) 2012 by Delphix. All rights reserved.
+.\" Copyright (c) 2012, 2014 by Delphix. All rights reserved.
 .\"
 .TH "ZDB" "8" "February 15, 2012" "" ""
 
@@ -20,7 +20,7 @@
 
 .SH "SYNOPSIS"
 \fBzdb\fR [-CumdibcsDvhLMXFPA] [-e [-p \fIpath\fR...]] [-t \fItxg\fR]
-    [-U \fIcache\fR] [-I \fIinflight I/Os\fR]
+    [-U \fIcache\fR] [-I \fIinflight I/Os\fR] [-x \fIdumpdir\fR]
     [\fIpoolname\fR [\fIobject\fR ...]]
 
 .P
@@ -372,6 +372,20 @@ Operate on an exported pool, not present in \fB/etc/zfs/zpool.cache\fR. The
 \fB-p\fR flag specifies the path under which devices are to be searched.
 .RE
 
+.sp
+.ne 2
+.na
+\fB-x\fR \fIdumpdir\fR
+.ad
+.sp .6
+.RS 4n
+All blocks accessed will be copied to files in the specified directory.
+The blocks will be placed in sparse files whose name is the same as
+that of the file or device read.  zdb can be then run on the generated files.
+Note that the \fB-bbc\fR flags are sufficient to access (and thus copy)
+all metadata on the pool.
+.RE
+
 .sp
 .ne 2
 .na