Add textdump(4) facility, which provides an alternative form of kernel

dump using mechanically generated/extracted debugging output rather than
a simple memory dump.  Current sources of debugging output are:

- DDB output capture buffer, if there is captured output to save
- Kernel message buffer
- Kernel configuration, if included in kernel
- Kernel version string
- Panic message

Textdumps are stored in swap/dump partitions as with regular dumps, but
are laid out as ustar files in order to allow multiple parts to be stored
as a stream of sequentially written blocks.  Blocks are written out in
reverse order, as the size of a textdump isn't known a priori.  As with
regular dumps, they will be extracted using savecore(8).

One new DDB(4) command is added, "textdump", which accepts "set",
"unset", and "status" arguments.  By default, normal kernel dumps are
generated unless "textdump set" is run in order to schedule a textdump.
It can be canceled using "textdump unset" to restore generation of a
normal kernel dump.

Several sysctls exist to configure aspects of textdumps;
debug.ddb.textdump.pending can be set to check whether a textdump is
pending, or set/unset in order to control whether the next kernel dump
will be a textdump from userspace.

While textdumps don't have to be generated as a result of a DDB script
run automatically as part of a kernel panic, this is a particular useful
way to use them, as instead of generating a complete memory dump, a
simple transcript of an automated DDB session can be captured using the
DDB output capture and textdump facilities.  This can be used to
generate quite brief kernel bug reports rich in debugging information
but not dependent on kernel symbol tables or precisely synchronized
source code.  Most textdumps I generate are less than 100k including
the full message buffer.  Using textdumps with an interactive debugging
session is also useful, with capture being enabled/disabled in order to
record some but not all of the DDB session.

MFC after:	3 months

4 files changed, 634 insertions, 3 deletions

diff --git a/sys/ddb/db_capture.c b/sys/ddb/db_capture.c
index 4e6dd7c..bbc7925 100644
--- a/sys/ddb/db_capture.c
+++ b/sys/ddb/db_capture.c
@@ -57,11 +57,13 @@ static MALLOC_DEFINE(M_DB_CAPTURE, "db_capture", "DDB capture buffer");
 
 #define	DB_CAPTURE_DEFAULTBUFSIZE	48*1024
 #define	DB_CAPTURE_MAXBUFSIZE	512*1024
+#define	DB_CAPTURE_FILENAME	"ddb.txt"	/* Captured DDB output. */
 
 static char *db_capture_buf;
 static u_int db_capture_bufsize = DB_CAPTURE_DEFAULTBUFSIZE;
 static u_int db_capture_maxbufsize = DB_CAPTURE_MAXBUFSIZE; /* Read-only. */
 static u_int db_capture_bufoff;		/* Next location to write in buffer. */
+static u_int db_capture_bufpadding;	/* Amount of zero padding. */
 static int db_capture_inpager;		/* Suspend capture in pager. */
 static int db_capture_inprogress;	/* DDB capture currently in progress. */
 
@@ -79,6 +81,14 @@ SYSCTL_UINT(_debug_ddb_capture, OID_AUTO, maxbufsize, CTLFLAG_RD,
     "Maximum value for debug.ddb.capture.bufsize");
 
 /*
+ * Various compile-time assertions: defaults must be even multiples of
+ * textdump block size.  We also perform run-time checking of
+ * user-configurable values.
+ */
+CTASSERT(DB_CAPTURE_DEFAULTBUFSIZE % TEXTDUMP_BLOCKSIZE == 0);
+CTASSERT(DB_CAPTURE_MAXBUFSIZE % TEXTDUMP_BLOCKSIZE == 0);
+
+/*
  * Boot-time allocation of the DDB capture buffer, if any.
  */
 static void
@@ -86,9 +96,9 @@ db_capture_sysinit(__unused void *dummy)
 {
 
 	TUNABLE_INT_FETCH("debug.ddb.capture.bufsize", &db_capture_bufsize);
+	db_capture_bufsize = roundup(db_capture_bufsize, TEXTDUMP_BLOCKSIZE);
 	if (db_capture_bufsize > DB_CAPTURE_MAXBUFSIZE)
 		db_capture_bufsize = DB_CAPTURE_MAXBUFSIZE;
-
 	if (db_capture_bufsize != 0)
 		db_capture_buf = malloc(db_capture_bufsize, M_DB_CAPTURE,
 		    M_WAITOK);
@@ -110,9 +120,9 @@ sysctl_debug_ddb_capture_bufsize(SYSCTL_HANDLER_ARGS)
 	error = sysctl_handle_int(oidp, &size, 0, req);
 	if (error || req->newptr == NULL)
 		return (error);
+	size = roundup(size, TEXTDUMP_BLOCKSIZE);
 	if (size > DB_CAPTURE_MAXBUFSIZE)
 		return (EINVAL);
-
 	sx_xlock(&db_capture_sx);
 	if (size != 0) {
 		/*
@@ -216,6 +226,23 @@ db_capture_exitpager(void)
 }
 
 /*
+ * Zero out any bytes left in the last block of the DDB capture buffer.  This
+ * is run shortly before writing the blocks to disk, rather than when output
+ * capture is stopped, in order to avoid injecting nul's into the middle of
+ * output.
+ */
+static void
+db_capture_zeropad(void)
+{
+	u_int len;
+
+	len = min(TEXTDUMP_BLOCKSIZE, (db_capture_bufsize -
+	    db_capture_bufoff) % TEXTDUMP_BLOCKSIZE);
+	bzero(db_capture_buf + db_capture_bufoff, len);
+	db_capture_bufpadding = len;
+}
+
+/*
  * Reset capture state, which flushes buffers.
  */
 static void
@@ -224,6 +251,7 @@ db_capture_reset(void)
 
 	db_capture_inprogress = 0;
 	db_capture_bufoff = 0;
+	db_capture_bufpadding = 0;
 }
 
 /*
@@ -242,7 +270,9 @@ db_capture_start(void)
 }
 
 /*
- * Terminate DDB output capture.
+ * Terminate DDB output capture--real work is deferred to db_capture_dump,
+ * which executes outside of the DDB context.  We don't zero pad here because
+ * capture may be started again before the dump takes place.
  */
 static void
 db_capture_stop(void)
@@ -255,6 +285,28 @@ db_capture_stop(void)
 	db_capture_inprogress = 0;
 }
 
+/*
+ * Dump DDB(4) captured output (and resets capture buffers).
+ */
+void
+db_capture_dump(struct dumperinfo *di)
+{
+	u_int offset;
+
+	if (db_capture_bufoff == 0)
+		return;
+
+	db_capture_zeropad();
+	textdump_mkustar(textdump_block_buffer, DB_CAPTURE_FILENAME,
+	    db_capture_bufoff);
+	(void)textdump_writenextblock(di, textdump_block_buffer);
+	for (offset = 0; offset < db_capture_bufoff + db_capture_bufpadding;
+	    offset += TEXTDUMP_BLOCKSIZE)
+		(void)textdump_writenextblock(di, db_capture_buf + offset);
+	db_capture_bufoff = 0;
+	db_capture_bufpadding = 0;
+}
+
 /*-
  * DDB(4) command to manage capture:
  *
diff --git a/sys/ddb/db_command.c b/sys/ddb/db_command.c
index c1a062b..5ebcc15 100644
--- a/sys/ddb/db_command.c
+++ b/sys/ddb/db_command.c
@@ -146,6 +146,7 @@ static struct command db_commands[] = {
 	{ "scripts",	db_scripts_cmd,		0,	0 },
 	{ "unscript",	db_unscript_cmd,	CS_OWN,	0 },
 	{ "capture",	db_capture_cmd,		CS_OWN,	0 },
+	{ "textdump",	db_textdump_cmd,	CS_OWN, 0 },
 	{ (char *)0, }
 };
 
diff --git a/sys/ddb/db_textdump.c b/sys/ddb/db_textdump.c
new file mode 100644
index 0000000..237040c
--- /dev/null
+++ b/sys/ddb/db_textdump.c
@@ -0,0 +1,555 @@
+/*-
+ * Copyright (c) 2007 Robert N. M. Watson
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+/*-
+ * Kernel text-dump support: allow a series of text files to be written to
+ * the dump partition for later recovery, including captured DDB output, the
+ * kernel configuration, message buffer, panic message, etc.  This allows for
+ * a more compact representation of critical debugging information than
+ * traditional binary dumps, as well as allowing dump information to be used
+ * without access to kernel symbols, source code, etc.
+ *
+ * Storage Layout
+ * --------------
+ *
+ * Crash dumps are aligned to the end of the dump or swap partition in order
+ * to minimize the chances of swap duing fsck eating into the dump.  However,
+ * unlike a memory dump, we don't know the size of the textdump a priori, so
+ * can't just write it out sequentially in order from a known starting point
+ * calculated with respect to the end of the partition.  In order to address
+ * this, we actually write out the textdump in reverse block order, allowing
+ * us to directly align it to the end of the partition and then write out the
+ * dump header and trailer before and after it once done.  savecore(8) must
+ * know to reverse the order of the blocks in order to produce a readable
+ * file.
+ *
+ * Data is written out in the 'tar' file format, as it provides the facility
+ * to write data incrementally as a stream without reference to previous
+ * files.
+ *
+ * TODO
+ * ----
+ *
+ * - Allow subsytems to register to submit files for inclusion in the text
+ *   dump in a generic way.
+ */
+
+#include <sys/cdefs.h>
+__FBSDID("$FreeBSD$");
+
+#include "opt_config.h"
+
+#include <sys/param.h>
+#include <sys/conf.h>
+#include <sys/kernel.h>
+#include <sys/kerneldump.h>
+#include <sys/msgbuf.h>
+#include <sys/sysctl.h>
+#include <sys/systm.h>
+
+#include <ddb/ddb.h>
+#include <ddb/db_lex.h>
+
+static SYSCTL_NODE(_debug_ddb, OID_AUTO, textdump, CTLFLAG_RW, 0,
+    "DDB textdump options");
+
+/*
+ * Don't touch the first SIZEOF_METADATA bytes on the dump device.  This is
+ * to protect us from metadata and metadata from us.
+ */
+#define	SIZEOF_METADATA		(64*1024)
+
+/*
+ * Data is written out as a series of files in the ustar tar format.  ustar
+ * is a simple streamed format consiting of a series of files prefixed with
+ * headers, and all padded to 512-byte block boundaries, which maps
+ * conveniently to our requirements.
+ */
+struct ustar_header {
+	char	uh_filename[100];
+	char	uh_mode[8];
+	char	uh_tar_owner[8];
+	char	uh_tar_group[8];
+	char	uh_size[12];
+	char	uh_mtime[12];
+	char	uh_sum[8];
+	char	uh_type;
+	char	uh_linkfile[100];
+	char	uh_ustar[6];
+	char	uh_version[2];
+	char	uh_owner[32];
+	char	uh_group[32];
+	char	uh_major[8];
+	char	uh_minor[8];
+	char	uh_filenameprefix[155];
+	char	uh_zeropad[12];
+} __packed;
+
+/*
+ * Various size assertions -- pretty much everything must be one block in
+ * size.
+ */
+CTASSERT(sizeof(struct kerneldumpheader) == TEXTDUMP_BLOCKSIZE);
+CTASSERT(sizeof(struct ustar_header) == TEXTDUMP_BLOCKSIZE);
+
+/*
+ * Is a textdump scheduled?  If so, the shutdown code will invoke our dumpsys
+ * routine instead of the machine-dependent kernel dump routine.
+ */
+int	textdump_pending;
+SYSCTL_INT(_debug_ddb_textdump, OID_AUTO, pending, CTLFLAG_RW,
+    &textdump_pending, 0,
+    "Perform textdump instead of regular kernel dump.");
+
+/*
+ * Various constants for tar headers and contents.
+ */
+#define	TAR_USER	"root"
+#define	TAR_GROUP	"wheel"
+#define	TAR_UID		"0"
+#define	TAR_GID		"0"
+#define	TAR_MODE	"0600"
+#define	TAR_USTAR	"ustar"
+
+#define	TAR_CONFIG_FILENAME	"config.txt"	/* Kernel configuration. */
+#define	TAR_MSGBUF_FILENAME	"msgbuf.txt"	/* Kernel messsage buffer. */
+#define	TAR_PANIC_FILENAME	"panic.txt"	/* Panic message. */
+#define	TAR_VERSION_FILENAME	"version.txt"	/* Kernel version. */
+
+/*
+ * Configure which files will be dumped.
+ */
+#ifdef INCLUDE_CONFIG_FILE
+static int textdump_do_config = 1;
+SYSCTL_INT(_debug_ddb_textdump, OID_AUTO, do_config, CTLFLAG_RW,
+    &textdump_do_config, 0, "Dump kernel configuration in textdump");
+#endif
+
+static int textdump_do_ddb = 1;
+SYSCTL_INT(_debug_ddb_textdump, OID_AUTO, do_ddb, CTLFLAG_RW,
+    &textdump_do_ddb, 0, "Dump DDB captured output in textdump");
+
+static int textdump_do_msgbuf = 1;
+SYSCTL_INT(_debug_ddb_textdump, OID_AUTO, do_msgbuf, CTLFLAG_RW,
+    &textdump_do_msgbuf, 0, "Dump kernel message buffer in textdump");
+
+static int textdump_do_panic = 1;
+SYSCTL_INT(_debug_ddb_textdump, OID_AUTO, do_panic, CTLFLAG_RW,
+    &textdump_do_panic, 0, "Dump kernel panic message in textdump");
+
+static int textdump_do_version = 1;
+SYSCTL_INT(_debug_ddb_textdump, OID_AUTO, do_version, CTLFLAG_RW,
+    &textdump_do_version, 0, "Dump kernel version string in textdump");
+
+/*
+ * State related to incremental writing of blocks to disk.
+ */
+static off_t textdump_offset;		/* Offset of next sequential write. */
+static int textdump_error;		/* Carried write error, if any. */
+
+/*
+ * Statically allocate space to prepare block-sized headers and data.
+ */
+char textdump_block_buffer[TEXTDUMP_BLOCKSIZE];
+static struct kerneldumpheader kdh;
+
+/*
+ * Text dumps are prefixed with a normal kernel dump header but with a
+ * different magic number to allow them to be uniquely identified.
+ */
+static void
+mkdumpheader(struct kerneldumpheader *kdh, uint32_t archver,
+    uint64_t dumplen, uint32_t blksz)
+{
+
+	bzero(kdh, sizeof(*kdh));
+	strncpy(kdh->magic, TEXTDUMPMAGIC, sizeof(kdh->magic));
+	strncpy(kdh->architecture, MACHINE_ARCH, sizeof(kdh->architecture));
+	kdh->version = htod32(KERNELDUMPVERSION);
+	kdh->architectureversion = htod32(archver);
+	kdh->dumplength = htod64(dumplen);
+	kdh->dumptime = htod64(time_second);
+	kdh->blocksize = htod32(blksz);
+	strncpy(kdh->hostname, hostname, sizeof(kdh->hostname));
+	strncpy(kdh->versionstring, version, sizeof(kdh->versionstring));
+	if (panicstr != NULL)
+		strncpy(kdh->panicstring, panicstr, sizeof(kdh->panicstring));
+	kdh->parity = kerneldump_parity(kdh);
+}
+
+/*
+ * Calculate and fill in the checksum for a tar header.
+ */
+static void
+ustar_checksum(struct ustar_header *uhp)
+{
+	u_int sum;
+	int i;
+
+	for (i = 0; i < sizeof(uhp->uh_sum); i++)
+		uhp->uh_sum[i] = ' ';
+	sum = 0;
+	for (i = 0; i < sizeof(*uhp); i++)
+		sum += ((u_char *)uhp)[i];
+	snprintf(uhp->uh_sum, sizeof(uhp->uh_sum), "%6o", sum);
+}
+
+/*
+ * Each file in the tarball has a block-sized header with its name and other,
+ * largely hard-coded, properties.
+ */
+void
+textdump_mkustar(char *block_buffer, const char *filename, u_int size)
+{
+	struct ustar_header *uhp;
+
+	uhp = (struct ustar_header *)block_buffer;
+	bzero(uhp, sizeof(*uhp));
+	strlcpy(uhp->uh_filename, filename, sizeof(uhp->uh_filename));
+	strlcpy(uhp->uh_mode, TAR_MODE, sizeof(uhp->uh_mode));
+	snprintf(uhp->uh_size, sizeof(uhp->uh_size), "%o", size);
+	strlcpy(uhp->uh_tar_owner, TAR_UID, sizeof(uhp->uh_tar_owner));
+	strlcpy(uhp->uh_tar_group, TAR_GID, sizeof(uhp->uh_tar_group));
+	strlcpy(uhp->uh_owner, TAR_USER, sizeof(uhp->uh_owner));
+	strlcpy(uhp->uh_group, TAR_GROUP, sizeof(uhp->uh_group));
+	snprintf(uhp->uh_mtime, sizeof(uhp->uh_mtime), "%lo",
+	    (unsigned long)time_second);
+	uhp->uh_type = 0;
+	strlcpy(uhp->uh_ustar, TAR_USTAR, sizeof(uhp->uh_ustar));
+	ustar_checksum(uhp);
+}
+
+/*
+ * textdump_writeblock() writes TEXTDUMP_BLOCKSIZE-sized blocks of data to
+ * the space between di->mediaoffset and di->mediaoffset + di->mediasize.  It
+ * accepts an offset relative to di->mediaoffset.  If we're carrying any
+ * error from previous I/O, return that error and don't continue to try to
+ * write.  Most writers ignore the error and forge ahead on the basis that
+ * there's not much you can do.
+ */
+static int
+textdump_writeblock(struct dumperinfo *di, off_t offset, char *buffer)
+{
+
+	if (textdump_error)
+		return (textdump_error);
+	if (offset + TEXTDUMP_BLOCKSIZE > di->mediasize)
+		return (EIO);
+	if (offset < SIZEOF_METADATA)
+		return (ENOSPC);
+	textdump_error = di->dumper(di->priv, buffer, 0, offset +
+	    di->mediaoffset, TEXTDUMP_BLOCKSIZE);
+	return (textdump_error);
+}
+
+/*
+ * Interfaces to save and restore the dump offset, so that printers can go
+ * back to rewrite a header if required, while avoiding their knowing about
+ * the global layout of the blocks.
+ */
+void
+textdump_saveoff(off_t *offsetp)
+{
+
+	*offsetp = textdump_offset;
+}
+
+void
+textdump_restoreoff(off_t offset)
+{
+
+	textdump_offset = offset;
+}
+
+/*
+ * Interface to write the "next block" relative to the current offset; since
+ * we write backwards from the end of the partition, we subtract, but there's
+ * no reason for the caller to know this.
+ */
+int
+textdump_writenextblock(struct dumperinfo *di, char *buffer)
+{
+	int error;
+
+	error = textdump_writeblock(di, textdump_offset, buffer);
+	textdump_offset -= TEXTDUMP_BLOCKSIZE;
+	return (error);
+}
+
+#ifdef INCLUDE_CONFIG_FILE
+extern char kernconfstring[];
+
+/*
+ * Dump kernel configuration.
+ */
+static void
+textdump_dump_config(struct dumperinfo *di)
+{
+	u_int count, fullblocks, len;
+
+	len = strlen(kernconfstring);
+	textdump_mkustar(textdump_block_buffer, TAR_CONFIG_FILENAME, len);
+	(void)textdump_writenextblock(di, textdump_block_buffer);
+
+	/*
+	 * Write out all full blocks directly from the string, and handle any
+	 * left-over bits by copying it to out to the local buffer and
+	 * zero-padding it.
+	 */
+	fullblocks = len / TEXTDUMP_BLOCKSIZE;
+	for (count = 0; count < fullblocks; count++)
+		(void)textdump_writenextblock(di, kernconfstring + count *
+		    TEXTDUMP_BLOCKSIZE);
+	if (len % TEXTDUMP_BLOCKSIZE != 0) {
+		bzero(textdump_block_buffer, TEXTDUMP_BLOCKSIZE);
+		bcopy(kernconfstring + count * TEXTDUMP_BLOCKSIZE,
+		    textdump_block_buffer, len % TEXTDUMP_BLOCKSIZE);
+		(void)textdump_writenextblock(di, textdump_block_buffer);
+	}
+}
+#endif /* INCLUDE_CONFIG_FILE */
+
+/*
+ * Dump kernel message buffer.
+ */
+static void
+textdump_dump_msgbuf(struct dumperinfo *di)
+{
+	off_t end_offset, tarhdr_offset;
+	u_int i, len, offset, seq, total_len;
+	char buf[16];
+
+	/*
+	 * Write out a dummy tar header to advance the offset; we'll rewrite
+	 * it later once we know the true size.
+	 */
+	textdump_saveoff(&tarhdr_offset);
+	textdump_mkustar(textdump_block_buffer, TAR_MSGBUF_FILENAME, 0);
+	(void)textdump_writenextblock(di, textdump_block_buffer);
+
+	/*
+	 * Copy out the data in small chunks, but don't copy nuls that may be
+	 * present if the message buffer has not yet completely filled at
+	 * least once.
+	 */
+	total_len = 0;
+	offset = 0;
+        msgbuf_peekbytes(msgbufp, NULL, 0, &seq);
+        while ((len = msgbuf_peekbytes(msgbufp, buf, sizeof(buf), &seq)) > 0) {
+		for (i = 0; i < len; i++) {
+			if (buf[i] == '\0')
+				continue;
+			textdump_block_buffer[offset] = buf[i];
+			offset++;
+			if (offset != sizeof(textdump_block_buffer))
+				continue;
+			(void)textdump_writenextblock(di,
+			    textdump_block_buffer);
+			total_len += offset;
+			offset = 0;
+		}
+        }
+	total_len += offset;	/* Without the zero-padding. */
+	if (offset != 0) {
+		bzero(textdump_block_buffer + offset,
+		    sizeof(textdump_block_buffer) - offset);
+		(void)textdump_writenextblock(di, textdump_block_buffer);
+	}
+
+	/*
+	 * Rewrite tar header to reflect how much was actually written.
+	 */
+	textdump_saveoff(&end_offset);
+	textdump_restoreoff(tarhdr_offset);
+	textdump_mkustar(textdump_block_buffer, TAR_MSGBUF_FILENAME,
+	    total_len);
+	(void)textdump_writenextblock(di, textdump_block_buffer);
+	textdump_restoreoff(end_offset);
+}
+
+static void
+textdump_dump_panic(struct dumperinfo *di)
+{
+	u_int len;
+
+	/*
+	 * Write out tar header -- we store up to one block of panic message.
+	 */
+	len = min(strlen(panicstr), TEXTDUMP_BLOCKSIZE);
+	textdump_mkustar(textdump_block_buffer, TAR_PANIC_FILENAME, len);
+	(void)textdump_writenextblock(di, textdump_block_buffer);
+
+	/*
+	 * Zero-pad the panic string and write out block.
+	 */
+	bzero(textdump_block_buffer, sizeof(textdump_block_buffer));
+	bcopy(panicstr, textdump_block_buffer, len);
+	(void)textdump_writenextblock(di, textdump_block_buffer);
+}
+
+static void
+textdump_dump_version(struct dumperinfo *di)
+{
+	u_int len;
+
+	/*
+	 * Write out tar header -- at most one block of version information.
+	 */
+	len = min(strlen(version), TEXTDUMP_BLOCKSIZE);
+	textdump_mkustar(textdump_block_buffer, TAR_VERSION_FILENAME, len);
+	(void)textdump_writenextblock(di, textdump_block_buffer);
+
+	/*
+	 * Zero pad the version string and write out block.
+	 */
+	bzero(textdump_block_buffer, sizeof(textdump_block_buffer));
+	bcopy(version, textdump_block_buffer, len);
+	(void)textdump_writenextblock(di, textdump_block_buffer);
+}
+
+/*
+ * Commit text dump to disk.
+ */
+void
+textdump_dumpsys(struct dumperinfo *di)
+{
+	off_t dumplen, trailer_offset;
+
+	if (di->blocksize != TEXTDUMP_BLOCKSIZE) {
+		printf("Dump partition block size (%ju) not textdump "
+		    "block size (%ju)", (uintmax_t)di->blocksize,
+		    (uintmax_t)TEXTDUMP_BLOCKSIZE);
+		return;
+	}
+
+	/*
+	 * We don't know a priori how large the dump will be, but we do know
+	 * that we need to reserve space for metadata and that we need two
+	 * dump headers.  Also leave room for one ustar header and one block
+	 * of data.
+	 */
+	if (di->mediasize < SIZEOF_METADATA + 2 * sizeof(kdh)) {
+		printf("Insufficient space on dump partition.\n");
+		return;
+	}
+	textdump_error = 0;
+
+	/*
+	 * Position the start of the dump so that we'll write the kernel dump
+	 * trailer immediately before the end of the partition, and then work
+	 * our way back.  We will rewrite this header later to reflect the
+	 * true size if things go well.
+	 */
+	textdump_offset = di->mediasize - sizeof(kdh);
+	textdump_saveoff(&trailer_offset);
+	mkdumpheader(&kdh, KERNELDUMP_TEXT_VERSION, 0, TEXTDUMP_BLOCKSIZE);
+	(void)textdump_writenextblock(di, (char *)&kdh);
+
+	/*
+	 * Write a series of files in ustar format.
+	 */
+	if (textdump_do_ddb)
+		db_capture_dump(di);
+#ifdef INCLUDE_CONFIG_FILE
+	if (textdump_do_config)
+		textdump_dump_config(di);
+#endif
+	if (textdump_do_msgbuf)
+		textdump_dump_msgbuf(di);
+	if (textdump_do_panic && panicstr != NULL)
+		textdump_dump_panic(di);
+	if (textdump_do_version)
+		textdump_dump_version(di);
+
+	/*
+	 * Now that we know the true size, we can write out the header, then
+	 * seek back to the end and rewrite the trailer with the correct
+	 * size.
+	 */
+	dumplen = trailer_offset - (textdump_offset + TEXTDUMP_BLOCKSIZE);
+	mkdumpheader(&kdh, KERNELDUMP_TEXT_VERSION, dumplen,
+	    TEXTDUMP_BLOCKSIZE);
+	(void)textdump_writenextblock(di, (char *)&kdh);
+	textdump_restoreoff(trailer_offset);
+	(void)textdump_writenextblock(di, (char *)&kdh);
+
+	/*
+	 * Terminate the dump, report any errors, and clear the pending flag.
+	 */
+	if (textdump_error == 0)
+		(void)di->dumper(di->priv, NULL, 0, 0, 0);
+	if (textdump_error == ENOSPC)
+		printf("Insufficient space on dump partition\n");
+	else if (textdump_error != 0)
+		printf("Error %d writing dump\n", textdump_error);
+	else
+		printf("Textdump complete.\n");
+	textdump_pending = 0;
+}
+
+/*-
+ * DDB(4) command to manage textdumps:
+ *
+ * textdump set        - request a textdump
+ * textdump status     - print DDB output textdump status
+ * textdump unset      - clear textdump request
+ */
+static void
+db_textdump_usage(void)
+{
+
+	db_printf("textdump [unset|set|status]\n");
+}
+
+void
+db_textdump_cmd(db_expr_t addr, boolean_t have_addr, db_expr_t count,
+    char *modif)
+{
+	int t;
+
+	t = db_read_token();
+	if (t != tIDENT) {
+		db_textdump_usage();
+		return;
+	}
+	if (db_read_token() != tEOL) {
+		db_textdump_usage();
+		return;
+	}
+	if (strcmp(db_tok_string, "set") == 0) {
+		textdump_pending = 1;
+		db_printf("textdump set\n");
+	} else if (strcmp(db_tok_string, "status") == 0) {
+		if (textdump_pending)
+			db_printf("textdump is set\n");
+		else
+			db_printf("textdump is not set\n");
+	} else if (strcmp(db_tok_string, "unset") == 0) {
+		textdump_pending = 0;
+		db_printf("textdump unset\n");
+	} else
+		db_textdump_usage();
+}
diff --git a/sys/ddb/ddb.h b/sys/ddb/ddb.h
index ebec60a..d559978 100644
--- a/sys/ddb/ddb.h
+++ b/sys/ddb/ddb.h
@@ -169,6 +169,7 @@ db_cmdfcn_t	db_set_thread;
 db_cmdfcn_t	db_show_regs;
 db_cmdfcn_t	db_show_threads;
 db_cmdfcn_t	db_single_step_cmd;
+db_cmdfcn_t	db_textdump_cmd;
 db_cmdfcn_t	db_trace_until_call_cmd;
 db_cmdfcn_t	db_trace_until_matching_cmd;
 db_cmdfcn_t	db_unscript_cmd;
@@ -212,4 +213,26 @@ void	db_capture_writech(char ch);
  */
 void	db_script_kdbenter(const char *eventname);	/* KDB enter event. */
 
+/*
+ * Interface between DDB and the textdump facility.
+ *
+ * Text dump blocks are of a fixed size; textdump_block_buffer is a
+ * statically allocated buffer that code interacting with textdumps can use
+ * to prepare and hold a pending block in when calling writenextblock().
+ */
+#define	TEXTDUMP_BLOCKSIZE	512
+extern char	textdump_block_buffer[TEXTDUMP_BLOCKSIZE];
+
+void	textdump_mkustar(char *block_buffer, const char *filename,
+	    u_int size);
+void	textdump_restoreoff(off_t offset);
+void	textdump_saveoff(off_t *offsetp);
+int	textdump_writenextblock(struct dumperinfo *di, char *buffer);
+
+/*
+ * Interface between the kernel and textdumps.
+ */
+extern int	textdump_pending;	/* Call textdump_dumpsys() instead. */
+void	textdump_dumpsys(struct dumperinfo *di);
+
 #endif /* !_DDB_DDB_H_ */