freebsd-skq/sbin/dumpon/dumpon.8
cem 935cac69d7 EKCD: Add Chacha20 encryption mode
Add Chacha20 mode to Encrypted Kernel Crash Dumps.

Chacha20 does not require messages to be multiples of block size, so it is
valid to use the cipher on non-block-sized messages without the explicit
padding AES-CBC would require.  Therefore, allow use with simultaneous dump
compression.  (Continue to disallow use of AES-CBC EKCD with compression.)

dumpon(8) gains a -C cipher flag to select between chacha and aes-cbc.
It defaults to chacha if no -C option is provided.  The man page documents this
behavior.

Relnotes:	sure
Sponsored by:	Dell EMC Isilon
2019-05-23 20:12:24 +00:00

455 lines
11 KiB
Groff

.\" Copyright (c) 1980, 1991, 1993
.\" The Regents of the University of California. 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.
.\" 3. Neither the name of the University nor the names of its contributors
.\" may be used to endorse or promote products derived from this software
.\" without specific prior written permission.
.\"
.\" THIS SOFTWARE IS PROVIDED BY THE REGENTS 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 REGENTS 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.
.\"
.\" From: @(#)swapon.8 8.1 (Berkeley) 6/5/93
.\" $FreeBSD$
.\"
.Dd May 23, 2019
.Dt DUMPON 8
.Os
.Sh NAME
.Nm dumpon
.Nd "specify a device for crash dumps"
.Sh SYNOPSIS
.Nm
.Op Fl i Ar index
.Op Fl r
.Op Fl v
.Op Fl C Ar cipher
.Op Fl k Ar pubkey
.Op Fl Z
.Op Fl z
.Ar device
.Nm
.Op Fl i Ar index
.Op Fl r
.Op Fl v
.Op Fl C Ar cipher
.Op Fl k Ar pubkey
.Op Fl Z
.Op Fl z
.Op Fl g Ar gateway
.Fl s Ar server
.Fl c Ar client
.Ar iface
.Nm
.Op Fl v
.Cm off
.Nm
.Op Fl v
.Fl l
.Sh DESCRIPTION
The
.Nm
utility is used to configure where the kernel can save a crash dump in the case
of a panic.
.Pp
System administrators should typically configure
.Nm
in a persistent fashion using the
.Xr rc.conf 5
variables
.Va dumpdev
and
.Va dumpon_flags .
For more information on this usage, see
.Xr rc.conf 5 .
.Pp
Starting in
.Fx 13.0 ,
.Nm
can configure a series of fallback dump devices.
For example, an administrator may prefer
.Xr netdump 4
by default, but if the
.Xr netdump 4
service cannot be reached or some other failure occurs, they might choose a
local disk dump as a second choice option.
.Ss General options
.Bl -tag -width _k_pubkey
.It Fl i Ar index
Insert the specified dump configuration into the prioritized fallback dump
device list at the specified index, starting at zero.
.Pp
If
.Fl i
is not specified, the configured dump device is appended to the prioritized
list.
.It Fl r
Remove the specified dump device configuration or configurations from the
fallback dump device list rather than inserting or appending it.
In contrast,
.Do
.Nm
off
.Dc
removes all configured devices.
Conflicts with
.Fl i .
.It Fl k Ar pubkey
Configure encrypted kernel dumps.
.Pp
A random, one-time symmetric key is automatically generated for bulk kernel
dump encryption every time
.Nm
is used.
The provided
.Ar pubkey
is used to encrypt a copy of the symmetric key.
The encrypted dump contents consist of a standard dump header, the
pubkey-encrypted symmetric key contents, and the symmetric key encrypted core
dump contents.
.Pp
As a result, only someone with the corresponding private key can decrypt the symmetric key.
The symmetric key is necessary to decrypt the kernel core.
The goal of the mechanism is to provide confidentiality.
.Pp
The
.Va pubkey
file should be a PEM-formatted RSA key of at least 1024 bits.
.It Fl C Ar cipher
Select the symmetric algorithm used for encrypted kernel crash dump.
The default is
.Dq chacha20
but
.Dq aes256-cbc
is also available.
(AES256-CBC mode does not work in conjunction with compression.)
.It Fl l
List the currently configured dump device(s), or /dev/null if no devices are
configured.
.It Fl v
Enable verbose mode.
.It Fl Z
Enable compression (Zstandard).
.It Fl z
Enable compression (gzip).
Only one compression method may be enabled at a time, so
.Fl z
is incompatible with
.Fl Z .
.Pp
Zstandard provides superior compression ratio and performance.
.El
.Ss Netdump
.Nm
may also configure the kernel to dump to a remote
.Xr netdumpd 8
server.
(The
.Xr netdumpd 8
server is available in ports.)
.Xr netdump 4
eliminates the need to reserve space for crash dumps.
It is especially useful in diskless environments.
When
.Nm
is used to configure netdump, the
.Ar device
(or
.Ar iface )
parameter should specify a network interface (e.g.,
.Va igb1 ) .
The specified NIC must be up (online) to configure netdump.
.Pp
.Xr netdump 4
specific options include:
.Bl -tag -width _g_gateway
.It Fl c Ar client
The local IP address of the
.Xr netdump 4
client.
.It Fl g Ar gateway
The first-hop router between
.Ar client
and
.Ar server .
If the
.Fl g
option is not specified and the system has a default route, the default
router is used as the
.Xr netdump 4
gateway.
If the
.Fl g
option is not specified and the system does not have a default route,
.Ar server
is assumed to be on the same link as
.Ar client .
.It Fl s Ar server
The IP address of the
.Xr netdumpd 8
server.
.El
.Pp
All of these options can be specified in the
.Xr rc.conf 5
variable
.Va dumpon_flags .
.Ss Minidumps
The default type of kernel crash dump is the mini crash dump.
Mini crash dumps hold only memory pages in use by the kernel.
Alternatively, full memory dumps can be enabled by setting the
.Va debug.minidump
.Xr sysctl 8
variable to 0.
.Ss Full dumps
For systems using full memory dumps, the size of the specified dump
device must be at least the size of physical memory.
Even though an additional 64 kB header is added to the dump, the BIOS for a
platform typically holds back some memory, so it is not usually
necessary to size the dump device larger than the actual amount of RAM
available in the machine.
Also, when using full memory dumps, the
.Nm
utility will refuse to enable a dump device which is smaller than the
total amount of physical memory as reported by the
.Va hw.physmem
.Xr sysctl 8
variable.
.Sh IMPLEMENTATION NOTES
Because the file system layer is already dead by the time a crash dump
is taken, it is not possible to send crash dumps directly to a file.
.Pp
The
.Xr loader 8
variable
.Va dumpdev
may be used to enable early kernel core dumps for system panics which occur
before userspace starts.
.Sh EXAMPLES
In order to generate an RSA private key, a user can use the
.Xr genrsa 1
tool:
.Pp
.Dl # openssl genrsa -out private.pem 4096
.Pp
A public key can be extracted from the private key using the
.Xr rsa 1
tool:
.Pp
.Dl # openssl rsa -in private.pem -out public.pem -pubout
.Pp
Once the RSA keys are created in a safe place, the public key may be moved to
the untrusted netdump client machine.
Now
.Pa public.pem
can be used by
.Nm
to configure encrypted kernel crash dumps:
.Pp
.Dl # dumpon -k public.pem /dev/ada0s1b
.Pp
It is recommended to test if the kernel saves encrypted crash dumps using the
current configuration.
The easiest way to do that is to cause a kernel panic using the
.Xr ddb 4
debugger:
.Pp
.Dl # sysctl debug.kdb.panic=1
.Pp
In the debugger the following commands should be typed to write a core dump and
reboot:
.Pp
.Dl db> call doadump(0)
.Dl db> reset
.Pp
After reboot
.Xr savecore 8
should be able to save the core dump in the
.Va Dq dumpdir
directory, which is
.Pa /var/crash
by default:
.Pp
.Dl # savecore /dev/ada0s1b
.Pp
Three files should be created in the core directory:
.Pa info.# ,
.Pa key.#
and
.Pa vmcore_encrypted.#
(where
.Dq #
is the number of the last core dump saved by
.Xr savecore 8 ) .
The
.Pa vmcore_encrypted.#
can be decrypted using the
.Xr decryptcore 8
utility:
.Pp
.Dl # decryptcore -p private.pem -k key.# -e vmcore_encrypted.# -c vmcore.#
.Pp
or shorter:
.Pp
.Dl # decryptcore -p private.pem -n #
.Pp
The
.Pa vmcore.#
can be now examined using
.Xr kgdb 1 :
.Pp
.Dl # kgdb /boot/kernel/kernel vmcore.#
.Pp
or shorter:
.Pp
.Dl # kgdb -n #
.Pp
The core was decrypted properly if
.Xr kgdb 1
does not print any errors.
Note that the live kernel might be at a different path
which can be examined by looking at the
.Va kern.bootfile
.Xr sysctl 8 .
.Pp
The
.Nm
.Xr rc 8
script runs early during boot, typically before networking is configured.
This makes it unsuitable for configuring
.Xr netdump
when the client address is dynamic.
To configure
.Xr netdump
when
.Xr dhclient
binds to a server,
.Xr dhclient-script
can be used to run
.Xr dumpon .
For example, to automatically configure
.Xr netdump
on the vtnet0 interface, add the following to
.Pa /etc/dhclient-exit-hooks .
.Bd -literal
case $reason in
BOUND|REBIND|REBOOT|RENEW)
if [ "$interface" != vtnet0 ] || [ -n "$old_ip_address" -a \\
"$old_ip_address" = "$new_ip_address" ]; then
break
fi
if [ -n "$new_routers" ]; then
# Take the first router in the list.
gateway_flag="-g ${new_routers%% *}"
fi
# Configure as the highest-priority dump device.
dumpon -i 0 -c $new_ip_address -s $server $gateway_flag vtnet0
;;
esac
.Ed
.Pp
Be sure to fill in the server IP address and change the interface name if
needed.
.Sh SEE ALSO
.Xr gzip 1 ,
.Xr kgdb 1 ,
.Xr zstd 1 ,
.Xr ddb 4 ,
.Xr netdump 4 ,
.Xr fstab 5 ,
.Xr rc.conf 5 ,
.Xr config 8 ,
.Xr decryptcore 8 ,
.Xr init 8 ,
.Xr loader 8 ,
.Xr rc 8 ,
.Xr savecore 8 ,
.Xr swapon 8 ,
.Xr panic 9
.Sh HISTORY
The
.Nm
utility appeared in
.Fx 2.0.5 .
.Pp
Support for encrypted kernel core dumps and netdump was added in
.Fx 12.0 .
.Sh AUTHORS
The
.Nm
manual page was written by
.An Mark Johnston Aq Mt markj@FreeBSD.org ,
.An Conrad Meyer Aq Mt cem@FreeBSD.org ,
.An Konrad Witaszczyk Aq Mt def@FreeBSD.org ,
and countless others.
.Sh CAVEATS
To configure encrypted kernel core dumps, the running kernel must have been
compiled with the
.Dv EKCD
option.
.Pp
Netdump does not automatically update the configured
.Ar gateway
if routing topology changes.
.Pp
The size of a compressed dump or a minidump is not a fixed function of RAM
size.
Therefore, when at least one of these options is enabled, the
.Nm
utility cannot verify that the
.Ar device
has sufficient space for a dump.
.Nm
is also unable to verify that a configured
.Xr netdumpd 8
server has sufficient space for a dump.
.Pp
.Fl Z
requires a kernel compiled with the
.Dv ZSTDIO
kernel option.
Similarly,
.Fl z
requires the
.Dv GZIO
option.
.Sh BUGS
Netdump only supports IPv4 at this time.
.Sh SECURITY CONSIDERATIONS
The current encrypted kernel core dump scheme does not provide integrity nor
authentication.
That is, the recipient of an encrypted kernel core dump cannot know if they
received an intact core dump, nor can they verify the provenance of the dump.
.Pp
RSA keys smaller than 1024 bits are practical to factor and therefore weak.
Even 1024 bit keys may not be large enough to ensure privacy for many
years, so NIST recommends a minimum of 2048 bit RSA keys.
As a seatbelt,
.Nm
prevents users from configuring encrypted kernel dumps with extremely weak RSA
keys.
If you do not care for cryptographic privacy guarantees, just use
.Nm
without specifying a
.Fl k Ar pubkey
option.
.Pp
This process is sandboxed using
.Xr capsicum 4 .