1.. SPDX-License-Identifier: GPL-2.0
2
3#########
4UML HowTo
5#########
6
7.. contents:: :local:
8
9************
10Introduction
11************
12
13Welcome to User Mode Linux
14
15User Mode Linux is the first Open Source virtualization platform (first
16release date 1991) and second virtualization platform for an x86 PC.
17
18How is UML Different from a VM using Virtualization package X?
19==============================================================
20
21We have come to assume that virtualization also means some level of
22hardware emulation. In fact, it does not. As long as a virtualization
23package provides the OS with devices which the OS can recognize and
24has a driver for, the devices do not need to emulate real hardware.
25Most OSes today have built-in support for a number of "fake"
26devices used only under virtualization.
27User Mode Linux takes this concept to the ultimate extreme - there
28is not a single real device in sight. It is 100% artificial or if
29we use the correct term 100% paravirtual. All UML devices are abstract
30concepts which map onto something provided by the host - files, sockets,
31pipes, etc.
32
33The other major difference between UML and various virtualization
34packages is that there is a distinct difference between the way the UML
35kernel and the UML programs operate.
36The UML kernel is just a process running on Linux - same as any other
37program. It can be run by an unprivileged user and it does not require
38anything in terms of special CPU features.
39The UML userspace, however, is a bit different. The Linux kernel on the
40host machine assists UML in intercepting everything the program running
41on a UML instance is trying to do and making the UML kernel handle all
42of its requests.
43This is different from other virtualization packages which do not make any
44difference between the guest kernel and guest programs. This difference
45results in a number of advantages and disadvantages of UML over let's say
46QEMU which we will cover later in this document.
47
48
49Why Would I Want User Mode Linux?
50=================================
51
52
53* If User Mode Linux kernel crashes, your host kernel is still fine. It
54  is not accelerated in any way (vhost, kvm, etc) and it is not trying to
55  access any devices directly.  It is, in fact, a process like any other.
56
57* You can run a usermode kernel as a non-root user (you may need to
58  arrange appropriate permissions for some devices).
59
60* You can run a very small VM with a minimal footprint for a specific
61  task (for example 32M or less).
62
63* You can get extremely high performance for anything which is a "kernel
64  specific task" such as forwarding, firewalling, etc while still being
65  isolated from the host kernel.
66
67* You can play with kernel concepts without breaking things.
68
69* You are not bound by "emulating" hardware, so you can try weird and
70  wonderful concepts which are very difficult to support when emulating
71  real hardware such as time travel and making your system clock
72  dependent on what UML does (very useful for things like tests).
73
74* It's fun.
75
76Why not to run UML
77==================
78
79* The syscall interception technique used by UML makes it inherently
80  slower for any userspace applications. While it can do kernel tasks
81  on par with most other virtualization packages, its userspace is
82  **slow**. The root cause is that UML has a very high cost of creating
83  new processes and threads (something most Unix/Linux applications
84  take for granted).
85
86* UML is strictly uniprocessor at present. If you want to run an
87  application which needs many CPUs to function, it is clearly the
88  wrong choice.
89
90***********************
91Building a UML instance
92***********************
93
94There is no UML installer in any distribution. While you can use off
95the shelf install media to install into a blank VM using a virtualization
96package, there is no UML equivalent. You have to use appropriate tools on
97your host to build a viable filesystem image.
98
99This is extremely easy on Debian - you can do it using debootstrap. It is
100also easy on OpenWRT - the build process can build UML images. All other
101distros - YMMV.
102
103Creating an image
104=================
105
106Create a sparse raw disk image::
107
108   # dd if=/dev/zero of=disk_image_name bs=1 count=1 seek=16G
109
110This will create a 16G disk image. The OS will initially allocate only one
111block and will allocate more as they are written by UML. As of kernel
112version 4.19 UML fully supports TRIM (as usually used by flash drives).
113Using TRIM inside the UML image by specifying discard as a mount option
114or by running ``tune2fs -o discard /dev/ubdXX`` will request UML to
115return any unused blocks to the OS.
116
117Create a filesystem on the disk image and mount it::
118
119   # mkfs.ext4 ./disk_image_name && mount ./disk_image_name /mnt
120
121This example uses ext4, any other filesystem such as ext3, btrfs, xfs,
122jfs, etc will work too.
123
124Create a minimal OS installation on the mounted filesystem::
125
126   # debootstrap buster /mnt http://deb.debian.org/debian
127
128debootstrap does not set up the root password, fstab, hostname or
129anything related to networking. It is up to the user to do that.
130
131Set the root password - the easiest way to do that is to chroot into the
132mounted image::
133
134   # chroot /mnt
135   # passwd
136   # exit
137
138Edit key system files
139=====================
140
141UML block devices are called ubds. The fstab created by debootstrap
142will be empty and it needs an entry for the root file system::
143
144   /dev/ubd0   ext4    discard,errors=remount-ro  0       1
145
146The image hostname will be set to the same as the host on which you
147are creating its image. It is a good idea to change that to avoid
148"Oh, bummer, I rebooted the wrong machine".
149
150UML supports two classes of network devices - the older uml_net ones
151which are scheduled for obsoletion. These are called ethX. It also
152supports the newer vector IO devices which are significantly faster
153and have support for some standard virtual network encapsulations like
154Ethernet over GRE and Ethernet over L2TPv3. These are called vec0.
155
156Depending on which one is in use, ``/etc/network/interfaces`` will
157need entries like::
158
159   # legacy UML network devices
160   auto eth0
161   iface eth0 inet dhcp
162
163   # vector UML network devices
164   auto vec0
165   iface vec0 inet dhcp
166
167We now have a UML image which is nearly ready to run, all we need is a
168UML kernel and modules for it.
169
170Most distributions have a UML package. Even if you intend to use your own
171kernel, testing the image with a stock one is always a good start. These
172packages come with a set of modules which should be copied to the target
173filesystem. The location is distribution dependent. For Debian these
174reside under /usr/lib/uml/modules. Copy recursively the content of this
175directory to the mounted UML filesystem::
176
177   # cp -rax /usr/lib/uml/modules /mnt/lib/modules
178
179If you have compiled your own kernel, you need to use the usual "install
180modules to a location" procedure by running::
181
182  # make INSTALL_MOD_PATH=/mnt/lib/modules modules_install
183
184This will install modules into /mnt/lib/modules/$(KERNELRELEASE).
185To specify the full module installation path, use::
186
187  # make MODLIB=/mnt/lib/modules modules_install
188
189At this point the image is ready to be brought up.
190
191*************************
192Setting Up UML Networking
193*************************
194
195UML networking is designed to emulate an Ethernet connection. This
196connection may be either point-to-point (similar to a connection
197between machines using a back-to-back cable) or a connection to a
198switch. UML supports a wide variety of means to build these
199connections to all of: local machine, remote machine(s), local and
200remote UML and other VM instances.
201
202
203+-----------+--------+------------------------------------+------------+
204| Transport |  Type  |        Capabilities                | Throughput |
205+===========+========+====================================+============+
206| tap       | vector | checksum, tso                      | > 8Gbit    |
207+-----------+--------+------------------------------------+------------+
208| hybrid    | vector | checksum, tso, multipacket rx      | > 6GBit    |
209+-----------+--------+------------------------------------+------------+
210| raw       | vector | checksum, tso, multipacket rx, tx" | > 6GBit    |
211+-----------+--------+------------------------------------+------------+
212| EoGRE     | vector | multipacket rx, tx                 | > 3Gbit    |
213+-----------+--------+------------------------------------+------------+
214| Eol2tpv3  | vector | multipacket rx, tx                 | > 3Gbit    |
215+-----------+--------+------------------------------------+------------+
216| bess      | vector | multipacket rx, tx                 | > 3Gbit    |
217+-----------+--------+------------------------------------+------------+
218| fd        | vector | dependent on fd type               | varies     |
219+-----------+--------+------------------------------------+------------+
220| tuntap    | legacy | none                               | ~ 500Mbit  |
221+-----------+--------+------------------------------------+------------+
222| daemon    | legacy | none                               | ~ 450Mbit  |
223+-----------+--------+------------------------------------+------------+
224| socket    | legacy | none                               | ~ 450Mbit  |
225+-----------+--------+------------------------------------+------------+
226| pcap      | legacy | rx only                            | ~ 450Mbit  |
227+-----------+--------+------------------------------------+------------+
228| ethertap  | legacy | obsolete                           | ~ 500Mbit  |
229+-----------+--------+------------------------------------+------------+
230| vde       | legacy | obsolete                           | ~ 500Mbit  |
231+-----------+--------+------------------------------------+------------+
232
233* All transports which have tso and checksum offloads can deliver speeds
234  approaching 10G on TCP streams.
235
236* All transports which have multi-packet rx and/or tx can deliver pps
237  rates of up to 1Mps or more.
238
239* All legacy transports are generally limited to ~600-700MBit and 0.05Mps.
240
241* GRE and L2TPv3 allow connections to all of: local machine, remote
242  machines, remote network devices and remote UML instances.
243
244* Socket allows connections only between UML instances.
245
246* Daemon and bess require running a local switch. This switch may be
247  connected to the host as well.
248
249
250Network configuration privileges
251================================
252
253The majority of the supported networking modes need ``root`` privileges.
254For example, in the legacy tuntap networking mode, users were required
255to be part of the group associated with the tunnel device.
256
257For newer network drivers like the vector transports, ``root`` privilege
258is required to fire an ioctl to setup the tun interface and/or use
259raw sockets where needed.
260
261This can be achieved by granting the user a particular capability instead
262of running UML as root.  In case of vector transport, a user can add the
263capability ``CAP_NET_ADMIN`` or ``CAP_NET_RAW`` to the uml binary.
264Thenceforth, UML can be run with normal user privilges, along with
265full networking.
266
267For example::
268
269   # sudo setcap cap_net_raw,cap_net_admin+ep linux
270
271Configuring vector transports
272===============================
273
274All vector transports support a similar syntax:
275
276If X is the interface number as in vec0, vec1, vec2, etc, the general
277syntax for options is::
278
279   vecX:transport="Transport Name",option=value,option=value,...,option=value
280
281Common options
282--------------
283
284These options are common for all transports:
285
286* ``depth=int`` - sets the queue depth for vector IO. This is the
287  amount of packets UML will attempt to read or write in a single
288  system call. The default number is 64 and is generally sufficient
289  for most applications that need throughput in the 2-4 Gbit range.
290  Higher speeds may require larger values.
291
292* ``mac=XX:XX:XX:XX:XX`` - sets the interface MAC address value.
293
294* ``gro=[0,1]`` - sets GRO off or on. Enables receive/transmit offloads.
295  The effect of this option depends on the host side support in the transport
296  which is being configured. In most cases it will enable TCP segmentation and
297  RX/TX checksumming offloads. The setting must be identical on the host side
298  and the UML side. The UML kernel will produce warnings if it is not.
299  For example, GRO is enabled by default on local machine interfaces
300  (e.g. veth pairs, bridge, etc), so it should be enabled in UML in the
301  corresponding UML transports (raw, tap, hybrid) in order for networking to
302  operate correctly.
303
304* ``mtu=int`` - sets the interface MTU
305
306* ``headroom=int`` - adjusts the default headroom (32 bytes) reserved
307  if a packet will need to be re-encapsulated into for instance VXLAN.
308
309* ``vec=0`` - disable multipacket IO and fall back to packet at a
310  time mode
311
312Shared Options
313--------------
314
315* ``ifname=str`` Transports which bind to a local network interface
316  have a shared option - the name of the interface to bind to.
317
318* ``src, dst, src_port, dst_port`` - all transports which use sockets
319  which have the notion of source and destination and/or source port
320  and destination port use these to specify them.
321
322* ``v6=[0,1]`` to specify if a v6 connection is desired for all
323  transports which operate over IP. Additionally, for transports that
324  have some differences in the way they operate over v4 and v6 (for example
325  EoL2TPv3), sets the correct mode of operation. In the absence of this
326  option, the socket type is determined based on what do the src and dst
327  arguments resolve/parse to.
328
329tap transport
330-------------
331
332Example::
333
334   vecX:transport=tap,ifname=tap0,depth=128,gro=1
335
336This will connect vec0 to tap0 on the host. Tap0 must already exist (for example
337created using tunctl) and UP.
338
339tap0 can be configured as a point-to-point interface and given an IP
340address so that UML can talk to the host. Alternatively, it is possible
341to connect UML to a tap interface which is connected to a bridge.
342
343While tap relies on the vector infrastructure, it is not a true vector
344transport at this point, because Linux does not support multi-packet
345IO on tap file descriptors for normal userspace apps like UML. This
346is a privilege which is offered only to something which can hook up
347to it at kernel level via specialized interfaces like vhost-net. A
348vhost-net like helper for UML is planned at some point in the future.
349
350Privileges required: tap transport requires either:
351
352* tap interface to exist and be created persistent and owned by the
353  UML user using tunctl. Example ``tunctl -u uml-user -t tap0``
354
355* binary to have ``CAP_NET_ADMIN`` privilege
356
357hybrid transport
358----------------
359
360Example::
361
362   vecX:transport=hybrid,ifname=tap0,depth=128,gro=1
363
364This is an experimental/demo transport which couples tap for transmit
365and a raw socket for receive. The raw socket allows multi-packet
366receive resulting in significantly higher packet rates than normal tap.
367
368Privileges required: hybrid requires ``CAP_NET_RAW`` capability by
369the UML user as well as the requirements for the tap transport.
370
371raw socket transport
372--------------------
373
374Example::
375
376   vecX:transport=raw,ifname=p-veth0,depth=128,gro=1
377
378
379This transport uses vector IO on raw sockets. While you can bind to any
380interface including a physical one, the most common use it to bind to
381the "peer" side of a veth pair with the other side configured on the
382host.
383
384Example host configuration for Debian:
385
386**/etc/network/interfaces**::
387
388   auto veth0
389   iface veth0 inet static
390	address 192.168.4.1
391	netmask 255.255.255.252
392	broadcast 192.168.4.3
393	pre-up ip link add veth0 type veth peer name p-veth0 && \
394          ifconfig p-veth0 up
395
396UML can now bind to p-veth0 like this::
397
398   vec0:transport=raw,ifname=p-veth0,depth=128,gro=1
399
400
401If the UML guest is configured with 192.168.4.2 and netmask 255.255.255.0
402it can talk to the host on 192.168.4.1
403
404The raw transport also provides some support for offloading some of the
405filtering to the host. The two options to control it are:
406
407* ``bpffile=str`` filename of raw bpf code to be loaded as a socket filter
408
409* ``bpfflash=int`` 0/1 allow loading of bpf from inside User Mode Linux.
410  This option allows the use of the ethtool load firmware command to
411  load bpf code.
412
413In either case the bpf code is loaded into the host kernel. While this is
414presently limited to legacy bpf syntax (not ebpf), it is still a security
415risk. It is not recommended to allow this unless the User Mode Linux
416instance is considered trusted.
417
418Privileges required: raw socket transport requires `CAP_NET_RAW`
419capability.
420
421GRE socket transport
422--------------------
423
424Example::
425
426   vecX:transport=gre,src=$src_host,dst=$dst_host
427
428
429This will configure an Ethernet over ``GRE`` (aka ``GRETAP`` or
430``GREIRB``) tunnel which will connect the UML instance to a ``GRE``
431endpoint at host dst_host. ``GRE`` supports the following additional
432options:
433
434* ``rx_key=int`` - GRE 32-bit integer key for rx packets, if set,
435  ``txkey`` must be set too
436
437* ``tx_key=int`` - GRE 32-bit integer key for tx packets, if set
438  ``rx_key`` must be set too
439
440* ``sequence=[0,1]`` - enable GRE sequence
441
442* ``pin_sequence=[0,1]`` - pretend that the sequence is always reset
443  on each packet (needed to interoperate with some really broken
444  implementations)
445
446* ``v6=[0,1]`` - force IPv4 or IPv6 sockets respectively
447
448* GRE checksum is not presently supported
449
450GRE has a number of caveats:
451
452* You can use only one GRE connection per IP address. There is no way to
453  multiplex connections as each GRE tunnel is terminated directly on
454  the UML instance.
455
456* The key is not really a security feature. While it was intended as such
457  its "security" is laughable. It is, however, a useful feature to
458  ensure that the tunnel is not misconfigured.
459
460An example configuration for a Linux host with a local address of
461192.168.128.1 to connect to a UML instance at 192.168.129.1
462
463**/etc/network/interfaces**::
464
465   auto gt0
466   iface gt0 inet static
467    address 10.0.0.1
468    netmask 255.255.255.0
469    broadcast 10.0.0.255
470    mtu 1500
471    pre-up ip link add gt0 type gretap local 192.168.128.1 \
472           remote 192.168.129.1 || true
473    down ip link del gt0 || true
474
475Additionally, GRE has been tested versus a variety of network equipment.
476
477Privileges required: GRE requires ``CAP_NET_RAW``
478
479l2tpv3 socket transport
480-----------------------
481
482_Warning_. L2TPv3 has a "bug". It is the "bug" known as "has more
483options than GNU ls". While it has some advantages, there are usually
484easier (and less verbose) ways to connect a UML instance to something.
485For example, most devices which support L2TPv3 also support GRE.
486
487Example::
488
489    vec0:transport=l2tpv3,udp=1,src=$src_host,dst=$dst_host,srcport=$src_port,dstport=$dst_port,depth=128,rx_session=0xffffffff,tx_session=0xffff
490
491This will configure an Ethernet over L2TPv3 fixed tunnel which will
492connect the UML instance to a L2TPv3 endpoint at host $dst_host using
493the L2TPv3 UDP flavour and UDP destination port $dst_port.
494
495L2TPv3 always requires the following additional options:
496
497* ``rx_session=int`` - l2tpv3 32-bit integer session for rx packets
498
499* ``tx_session=int`` - l2tpv3 32-bit integer session for tx packets
500
501As the tunnel is fixed these are not negotiated and they are
502preconfigured on both ends.
503
504Additionally, L2TPv3 supports the following optional parameters.
505
506* ``rx_cookie=int`` - l2tpv3 32-bit integer cookie for rx packets - same
507  functionality as GRE key, more to prevent misconfiguration than provide
508  actual security
509
510* ``tx_cookie=int`` - l2tpv3 32-bit integer cookie for tx packets
511
512* ``cookie64=[0,1]`` - use 64-bit cookies instead of 32-bit.
513
514* ``counter=[0,1]`` - enable l2tpv3 counter
515
516* ``pin_counter=[0,1]`` - pretend that the counter is always reset on
517  each packet (needed to interoperate with some really broken
518  implementations)
519
520* ``v6=[0,1]`` - force v6 sockets
521
522* ``udp=[0,1]`` - use raw sockets (0) or UDP (1) version of the protocol
523
524L2TPv3 has a number of caveats:
525
526* you can use only one connection per IP address in raw mode. There is
527  no way to multiplex connections as each L2TPv3 tunnel is terminated
528  directly on the UML instance. UDP mode can use different ports for
529  this purpose.
530
531Here is an example of how to configure a Linux host to connect to UML
532via L2TPv3:
533
534**/etc/network/interfaces**::
535
536   auto l2tp1
537   iface l2tp1 inet static
538    address 192.168.126.1
539    netmask 255.255.255.0
540    broadcast 192.168.126.255
541    mtu 1500
542    pre-up ip l2tp add tunnel remote 127.0.0.1 \
543           local 127.0.0.1 encap udp tunnel_id 2 \
544           peer_tunnel_id 2 udp_sport 1706 udp_dport 1707 && \
545           ip l2tp add session name l2tp1 tunnel_id 2 \
546           session_id 0xffffffff peer_session_id 0xffffffff
547    down ip l2tp del session tunnel_id 2 session_id 0xffffffff && \
548           ip l2tp del tunnel tunnel_id 2
549
550
551Privileges required: L2TPv3 requires ``CAP_NET_RAW`` for raw IP mode and
552no special privileges for the UDP mode.
553
554BESS socket transport
555---------------------
556
557BESS is a high performance modular network switch.
558
559https://github.com/NetSys/bess
560
561It has support for a simple sequential packet socket mode which in the
562more recent versions is using vector IO for high performance.
563
564Example::
565
566   vecX:transport=bess,src=$unix_src,dst=$unix_dst
567
568This will configure a BESS transport using the unix_src Unix domain
569socket address as source and unix_dst socket address as destination.
570
571For BESS configuration and how to allocate a BESS Unix domain socket port
572please see the BESS documentation.
573
574https://github.com/NetSys/bess/wiki/Built-In-Modules-and-Ports
575
576BESS transport does not require any special privileges.
577
578Configuring Legacy transports
579=============================
580
581Legacy transports are now considered obsolete. Please use the vector
582versions.
583
584***********
585Running UML
586***********
587
588This section assumes that either the user-mode-linux package from the
589distribution or a custom built kernel has been installed on the host.
590
591These add an executable called linux to the system. This is the UML
592kernel. It can be run just like any other executable.
593It will take most normal linux kernel arguments as command line
594arguments.  Additionally, it will need some UML-specific arguments
595in order to do something useful.
596
597Arguments
598=========
599
600Mandatory Arguments:
601--------------------
602
603* ``mem=int[K,M,G]`` - amount of memory. By default in bytes. It will
604  also accept K, M or G qualifiers.
605
606* ``ubdX[s,d,c,t]=`` virtual disk specification. This is not really
607  mandatory, but it is likely to be needed in nearly all cases so we can
608  specify a root file system.
609  The simplest possible image specification is the name of the image
610  file for the filesystem (created using one of the methods described
611  in `Creating an image`_).
612
613  * UBD devices support copy on write (COW). The changes are kept in
614    a separate file which can be discarded allowing a rollback to the
615    original pristine image.  If COW is desired, the UBD image is
616    specified as: ``cow_file,master_image``.
617    Example:``ubd0=Filesystem.cow,Filesystem.img``
618
619  * UBD devices can be set to use synchronous IO. Any writes are
620    immediately flushed to disk. This is done by adding ``s`` after
621    the ``ubdX`` specification.
622
623  * UBD performs some heuristics on devices specified as a single
624    filename to make sure that a COW file has not been specified as
625    the image. To turn them off, use the ``d`` flag after ``ubdX``.
626
627  * UBD supports TRIM - asking the Host OS to reclaim any unused
628    blocks in the image. To turn it off, specify the ``t`` flag after
629    ``ubdX``.
630
631* ``root=`` root device - most likely ``/dev/ubd0`` (this is a Linux
632  filesystem image)
633
634Important Optional Arguments
635----------------------------
636
637If UML is run as "linux" with no extra arguments, it will try to start an
638xterm for every console configured inside the image (up to 6 in most
639Linux distributions). Each console is started inside an
640xterm. This makes it nice and easy to use UML on a host with a GUI. It is,
641however, the wrong approach if UML is to be used as a testing harness or run
642in a text-only environment.
643
644In order to change this behaviour we need to specify an alternative console
645and wire it to one of the supported "line" channels. For this we need to map a
646console to use something different from the default xterm.
647
648Example which will divert console number 1 to stdin/stdout::
649
650   con1=fd:0,fd:1
651
652UML supports a wide variety of serial line channels which are specified using
653the following syntax
654
655   conX=channel_type:options[,channel_type:options]
656
657
658If the channel specification contains two parts separated by comma, the first
659one is input, the second one output.
660
661* The null channel - Discard all input or output. Example ``con=null`` will set
662  all consoles to null by default.
663
664* The fd channel - use file descriptor numbers for input/output. Example:
665  ``con1=fd:0,fd:1.``
666
667* The port channel - start a telnet server on TCP port number. Example:
668  ``con1=port:4321``.  The host must have /usr/sbin/in.telnetd (usually part of
669  a telnetd package) and the port-helper from the UML utilities (see the
670  information for the xterm channel below).  UML will not boot until a client
671  connects.
672
673* The pty and pts channels - use system pty/pts.
674
675* The tty channel - bind to an existing system tty. Example: ``con1=/dev/tty8``
676  will make UML use the host 8th console (usually unused).
677
678* The xterm channel - this is the default - bring up an xterm on this channel
679  and direct IO to it. Note that in order for xterm to work, the host must
680  have the UML distribution package installed. This usually contains the
681  port-helper and other utilities needed for UML to communicate with the xterm.
682  Alternatively, these need to be complied and installed from source. All
683  options applicable to consoles also apply to UML serial lines which are
684  presented as ttyS inside UML.
685
686Starting UML
687============
688
689We can now run UML.
690::
691
692   # linux mem=2048M umid=TEST \
693    ubd0=Filesystem.img \
694    vec0:transport=tap,ifname=tap0,depth=128,gro=1 \
695    root=/dev/ubda con=null con0=null,fd:2 con1=fd:0,fd:1
696
697This will run an instance with ``2048M RAM`` and try to use the image file
698called ``Filesystem.img`` as root. It will connect to the host using tap0.
699All consoles except ``con1`` will be disabled and console 1 will
700use standard input/output making it appear in the same terminal it was started.
701
702Logging in
703============
704
705If you have not set up a password when generating the image, you will have to
706shut down the UML instance, mount the image, chroot into it and set it - as
707described in the Generating an Image section.  If the password is already set,
708you can just log in.
709
710The UML Management Console
711============================
712
713In addition to managing the image from "the inside" using normal sysadmin tools,
714it is possible to perform a number of low-level operations using the UML
715management console. The UML management console is a low-level interface to the
716kernel on a running UML instance, somewhat like the i386 SysRq interface. Since
717there is a full-blown operating system under UML, there is much greater
718flexibility possible than with the SysRq mechanism.
719
720There are a number of things you can do with the mconsole interface:
721
722* get the kernel version
723* add and remove devices
724* halt or reboot the machine
725* Send SysRq commands
726* Pause and resume the UML
727* Inspect processes running inside UML
728* Inspect UML internal /proc state
729
730You need the mconsole client (uml\_mconsole) which is a part of the UML
731tools package available in most Linux distritions.
732
733You also need ``CONFIG_MCONSOLE`` (under 'General Setup') enabled in the UML
734kernel.  When you boot UML, you'll see a line like::
735
736   mconsole initialized on /home/jdike/.uml/umlNJ32yL/mconsole
737
738If you specify a unique machine id on the UML command line, i.e.
739``umid=debian``, you'll see this::
740
741   mconsole initialized on /home/jdike/.uml/debian/mconsole
742
743
744That file is the socket that uml_mconsole will use to communicate with
745UML.  Run it with either the umid or the full path as its argument::
746
747   # uml_mconsole debian
748
749or
750
751   # uml_mconsole /home/jdike/.uml/debian/mconsole
752
753
754You'll get a prompt, at which you can run one of these commands:
755
756* version
757* help
758* halt
759* reboot
760* config
761* remove
762* sysrq
763* help
764* cad
765* stop
766* go
767* proc
768* stack
769
770version
771-------
772
773This command takes no arguments.  It prints the UML version::
774
775   (mconsole)  version
776   OK Linux OpenWrt 4.14.106 #0 Tue Mar 19 08:19:41 2019 x86_64
777
778
779There are a couple actual uses for this.  It's a simple no-op which
780can be used to check that a UML is running.  It's also a way of
781sending a device interrupt to the UML. UML mconsole is treated internally as
782a UML device.
783
784help
785----
786
787This command takes no arguments. It prints a short help screen with the
788supported mconsole commands.
789
790
791halt and reboot
792---------------
793
794These commands take no arguments.  They shut the machine down immediately, with
795no syncing of disks and no clean shutdown of userspace.  So, they are
796pretty close to crashing the machine::
797
798   (mconsole)  halt
799   OK
800
801config
802------
803
804"config" adds a new device to the virtual machine. This is supported
805by most UML device drivers. It takes one argument, which is the
806device to add, with the same syntax as the kernel command line::
807
808   (mconsole) config ubd3=/home/jdike/incoming/roots/root_fs_debian22
809
810remove
811------
812
813"remove" deletes a device from the system.  Its argument is just the
814name of the device to be removed. The device must be idle in whatever
815sense the driver considers necessary.  In the case of the ubd driver,
816the removed block device must not be mounted, swapped on, or otherwise
817open, and in the case of the network driver, the device must be down::
818
819   (mconsole)  remove ubd3
820
821sysrq
822-----
823
824This command takes one argument, which is a single letter.  It calls the
825generic kernel's SysRq driver, which does whatever is called for by
826that argument.  See the SysRq documentation in
827Documentation/admin-guide/sysrq.rst in your favorite kernel tree to
828see what letters are valid and what they do.
829
830cad
831---
832
833This invokes the ``Ctl-Alt-Del`` action in the running image.  What exactly
834this ends up doing is up to init, systemd, etc.  Normally, it reboots the
835machine.
836
837stop
838----
839
840This puts the UML in a loop reading mconsole requests until a 'go'
841mconsole command is received. This is very useful as a
842debugging/snapshotting tool.
843
844go
845--
846
847This resumes a UML after being paused by a 'stop' command. Note that
848when the UML has resumed, TCP connections may have timed out and if
849the UML is paused for a long period of time, crond might go a little
850crazy, running all the jobs it didn't do earlier.
851
852proc
853----
854
855This takes one argument - the name of a file in /proc which is printed
856to the mconsole standard output
857
858stack
859-----
860
861This takes one argument - the pid number of a process. Its stack is
862printed to a standard output.
863
864*******************
865Advanced UML Topics
866*******************
867
868Sharing Filesystems between Virtual Machines
869============================================
870
871Don't attempt to share filesystems simply by booting two UMLs from the
872same file.  That's the same thing as booting two physical machines
873from a shared disk.  It will result in filesystem corruption.
874
875Using layered block devices
876---------------------------
877
878The way to share a filesystem between two virtual machines is to use
879the copy-on-write (COW) layering capability of the ubd block driver.
880Any changed blocks are stored in the private COW file, while reads come
881from either device - the private one if the requested block is valid in
882it, the shared one if not.  Using this scheme, the majority of data
883which is unchanged is shared between an arbitrary number of virtual
884machines, each of which has a much smaller file containing the changes
885that it has made.  With a large number of UMLs booting from a large root
886filesystem, this leads to a huge disk space saving.
887
888Sharing file system data will also help performance, since the host will
889be able to cache the shared data using a much smaller amount of memory,
890so UML disk requests will be served from the host's memory rather than
891its disks.  There is a major caveat in doing this on multisocket NUMA
892machines.  On such hardware, running many UML instances with a shared
893master image and COW changes may cause issues like NMIs from excess of
894inter-socket traffic.
895
896If you are running UML on high-end hardware like this, make sure to
897bind UML to a set of logical CPUs residing on the same socket using the
898``taskset`` command or have a look at the "tuning" section.
899
900To add a copy-on-write layer to an existing block device file, simply
901add the name of the COW file to the appropriate ubd switch::
902
903   ubd0=root_fs_cow,root_fs_debian_22
904
905where ``root_fs_cow`` is the private COW file and ``root_fs_debian_22`` is
906the existing shared filesystem.  The COW file need not exist.  If it
907doesn't, the driver will create and initialize it.
908
909Disk Usage
910----------
911
912UML has TRIM support which will release any unused space in its disk
913image files to the underlying OS. It is important to use either ls -ls
914or du to verify the actual file size.
915
916COW validity.
917-------------
918
919Any changes to the master image will invalidate all COW files. If this
920happens, UML will *NOT* automatically delete any of the COW files and
921will refuse to boot. In this case the only solution is to either
922restore the old image (including its last modified timestamp) or remove
923all COW files which will result in their recreation. Any changes in
924the COW files will be lost.
925
926Cows can moo - uml_moo : Merging a COW file with its backing file
927-----------------------------------------------------------------
928
929Depending on how you use UML and COW devices, it may be advisable to
930merge the changes in the COW file into the backing file every once in
931a while.
932
933The utility that does this is uml_moo.  Its usage is::
934
935   uml_moo COW_file new_backing_file
936
937
938There's no need to specify the backing file since that information is
939already in the COW file header.  If you're paranoid, boot the new
940merged file, and if you're happy with it, move it over the old backing
941file.
942
943``uml_moo`` creates a new backing file by default as a safety measure.
944It also has a destructive merge option which will merge the COW file
945directly into its current backing file.  This is really only usable
946when the backing file only has one COW file associated with it.  If
947there are multiple COWs associated with a backing file, a -d merge of
948one of them will invalidate all of the others.  However, it is
949convenient if you're short of disk space, and it should also be
950noticeably faster than a non-destructive merge.
951
952``uml_moo`` is installed with the UML distribution packages and is
953available as a part of UML utilities.
954
955Host file access
956==================
957
958If you want to access files on the host machine from inside UML, you
959can treat it as a separate machine and either nfs mount directories
960from the host or copy files into the virtual machine with scp.
961However, since UML is running on the host, it can access those
962files just like any other process and make them available inside the
963virtual machine without the need to use the network.
964This is possible with the hostfs virtual filesystem.  With it, you
965can mount a host directory into the UML filesystem and access the
966files contained in it just as you would on the host.
967
968*SECURITY WARNING*
969
970Hostfs without any parameters to the UML Image will allow the image
971to mount any part of the host filesystem and write to it. Always
972confine hostfs to a specific "harmless" directory (for example ``/var/tmp``)
973if running UML. This is especially important if UML is being run as root.
974
975Using hostfs
976------------
977
978To begin with, make sure that hostfs is available inside the virtual
979machine with::
980
981   # cat /proc/filesystems
982
983``hostfs`` should be listed.  If it's not, either rebuild the kernel
984with hostfs configured into it or make sure that hostfs is built as a
985module and available inside the virtual machine, and insmod it.
986
987
988Now all you need to do is run mount::
989
990   # mount none /mnt/host -t hostfs
991
992will mount the host's ``/`` on the virtual machine's ``/mnt/host``.
993If you don't want to mount the host root directory, then you can
994specify a subdirectory to mount with the -o switch to mount::
995
996   # mount none /mnt/home -t hostfs -o /home
997
998will mount the host's /home on the virtual machine's /mnt/home.
999
1000hostfs as the root filesystem
1001-----------------------------
1002
1003It's possible to boot from a directory hierarchy on the host using
1004hostfs rather than using the standard filesystem in a file.
1005To start, you need that hierarchy.  The easiest way is to loop mount
1006an existing root_fs file::
1007
1008   #  mount root_fs uml_root_dir -o loop
1009
1010
1011You need to change the filesystem type of ``/`` in ``etc/fstab`` to be
1012'hostfs', so that line looks like this::
1013
1014   /dev/ubd/0       /        hostfs      defaults          1   1
1015
1016Then you need to chown to yourself all the files in that directory
1017that are owned by root.  This worked for me::
1018
1019   #  find . -uid 0 -exec chown jdike {} \;
1020
1021Next, make sure that your UML kernel has hostfs compiled in, not as a
1022module.  Then run UML with the boot device pointing at that directory::
1023
1024   ubd0=/path/to/uml/root/directory
1025
1026UML should then boot as it does normally.
1027
1028Hostfs Caveats
1029--------------
1030
1031Hostfs does not support keeping track of host filesystem changes on the
1032host (outside UML). As a result, if a file is changed without UML's
1033knowledge, UML will not know about it and its own in-memory cache of
1034the file may be corrupt. While it is possible to fix this, it is not
1035something which is being worked on at present.
1036
1037Tuning UML
1038============
1039
1040UML at present is strictly uniprocessor. It will, however spin up a
1041number of threads to handle various functions.
1042
1043The UBD driver, SIGIO and the MMU emulation do that. If the system is
1044idle, these threads will be migrated to other processors on a SMP host.
1045This, unfortunately, will usually result in LOWER performance because of
1046all of the cache/memory synchronization traffic between cores. As a
1047result, UML will usually benefit from being pinned on a single CPU,
1048especially on a large system. This can result in performance differences
1049of 5 times or higher on some benchmarks.
1050
1051Similarly, on large multi-node NUMA systems UML will benefit if all of
1052its memory is allocated from the same NUMA node it will run on. The
1053OS will *NOT* do that by default. In order to do that, the sysadmin
1054needs to create a suitable tmpfs ramdisk bound to a particular node
1055and use that as the source for UML RAM allocation by specifying it
1056in the TMP or TEMP environment variables. UML will look at the values
1057of ``TMPDIR``, ``TMP`` or ``TEMP`` for that. If that fails, it will
1058look for shmfs mounted under ``/dev/shm``. If everything else fails use
1059``/tmp/`` regardless of the filesystem type used for it::
1060
1061   mount -t tmpfs -ompol=bind:X none /mnt/tmpfs-nodeX
1062   TEMP=/mnt/tmpfs-nodeX taskset -cX linux options options options..
1063
1064*******************************************
1065Contributing to UML and Developing with UML
1066*******************************************
1067
1068UML is an excellent platform to develop new Linux kernel concepts -
1069filesystems, devices, virtualization, etc. It provides unrivalled
1070opportunities to create and test them without being constrained to
1071emulating specific hardware.
1072
1073Example - want to try how Linux will work with 4096 "proper" network
1074devices?
1075
1076Not an issue with UML. At the same time, this is something which
1077is difficult with other virtualization packages - they are
1078constrained by the number of devices allowed on the hardware bus
1079they are trying to emulate (for example 16 on a PCI bus in qemu).
1080
1081If you have something to contribute such as a patch, a bugfix, a
1082new feature, please send it to ``linux-um@lists.infradead.org``.
1083
1084Please follow all standard Linux patch guidelines such as cc-ing
1085relevant maintainers and run ``./scripts/checkpatch.pl`` on your patch.
1086For more details see ``Documentation/process/submitting-patches.rst``
1087
1088Note - the list does not accept HTML or attachments, all emails must
1089be formatted as plain text.
1090
1091Developing always goes hand in hand with debugging. First of all,
1092you can always run UML under gdb and there will be a whole section
1093later on on how to do that. That, however, is not the only way to
1094debug a Linux kernel. Quite often adding tracing statements and/or
1095using UML specific approaches such as ptracing the UML kernel process
1096are significantly more informative.
1097
1098Tracing UML
1099=============
1100
1101When running, UML consists of a main kernel thread and a number of
1102helper threads. The ones of interest for tracing are NOT the ones
1103that are already ptraced by UML as a part of its MMU emulation.
1104
1105These are usually the first three threads visible in a ps display.
1106The one with the lowest PID number and using most CPU is usually the
1107kernel thread. The other threads are the disk
1108(ubd) device helper thread and the SIGIO helper thread.
1109Running ptrace on this thread usually results in the following picture::
1110
1111   host$ strace -p 16566
1112   --- SIGIO {si_signo=SIGIO, si_code=POLL_IN, si_band=65} ---
1113   epoll_wait(4, [{EPOLLIN, {u32=3721159424, u64=3721159424}}], 64, 0) = 1
1114   epoll_wait(4, [], 64, 0)                = 0
1115   rt_sigreturn({mask=[PIPE]})             = 16967
1116   ptrace(PTRACE_GETREGS, 16967, NULL, 0xd5f34f38) = 0
1117   ptrace(PTRACE_GETREGSET, 16967, NT_X86_XSTATE, [{iov_base=0xd5f35010, iov_len=832}]) = 0
1118   ptrace(PTRACE_GETSIGINFO, 16967, NULL, {si_signo=SIGTRAP, si_code=0x85, si_pid=16967, si_uid=0}) = 0
1119   ptrace(PTRACE_SETREGS, 16967, NULL, 0xd5f34f38) = 0
1120   ptrace(PTRACE_SETREGSET, 16967, NT_X86_XSTATE, [{iov_base=0xd5f35010, iov_len=2696}]) = 0
1121   ptrace(PTRACE_SYSEMU, 16967, NULL, 0)   = 0
1122   --- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_TRAPPED, si_pid=16967, si_uid=0, si_status=SIGTRAP, si_utime=65, si_stime=89} ---
1123   wait4(16967, [{WIFSTOPPED(s) && WSTOPSIG(s) == SIGTRAP | 0x80}], WSTOPPED|__WALL, NULL) = 16967
1124   ptrace(PTRACE_GETREGS, 16967, NULL, 0xd5f34f38) = 0
1125   ptrace(PTRACE_GETREGSET, 16967, NT_X86_XSTATE, [{iov_base=0xd5f35010, iov_len=832}]) = 0
1126   ptrace(PTRACE_GETSIGINFO, 16967, NULL, {si_signo=SIGTRAP, si_code=0x85, si_pid=16967, si_uid=0}) = 0
1127   timer_settime(0, 0, {it_interval={tv_sec=0, tv_nsec=0}, it_value={tv_sec=0, tv_nsec=2830912}}, NULL) = 0
1128   getpid()                                = 16566
1129   clock_nanosleep(CLOCK_MONOTONIC, 0, {tv_sec=1, tv_nsec=0}, NULL) = ? ERESTART_RESTARTBLOCK (Interrupted by signal)
1130   --- SIGALRM {si_signo=SIGALRM, si_code=SI_TIMER, si_timerid=0, si_overrun=0, si_value={int=1631716592, ptr=0x614204f0}} ---
1131   rt_sigreturn({mask=[PIPE]})             = -1 EINTR (Interrupted system call)
1132
1133This is a typical picture from a mostly idle UML instance.
1134
1135* UML interrupt controller uses epoll - this is UML waiting for IO
1136  interrupts:
1137
1138   epoll_wait(4, [{EPOLLIN, {u32=3721159424, u64=3721159424}}], 64, 0) = 1
1139
1140* The sequence of ptrace calls is part of MMU emulation and running the
1141  UML userspace.
1142* ``timer_settime`` is part of the UML high res timer subsystem mapping
1143  timer requests from inside UML onto the host high resolution timers.
1144* ``clock_nanosleep`` is UML going into idle (similar to the way a PC
1145  will execute an ACPI idle).
1146
1147As you can see UML will generate quite a bit of output even in idle. The output
1148can be very informative when observing IO. It shows the actual IO calls, their
1149arguments and returns values.
1150
1151Kernel debugging
1152================
1153
1154You can run UML under gdb now, though it will not necessarily agree to
1155be started under it. If you are trying to track a runtime bug, it is
1156much better to attach gdb to a running UML instance and let UML run.
1157
1158Assuming the same PID number as in the previous example, this would be::
1159
1160   # gdb -p 16566
1161
1162This will STOP the UML instance, so you must enter `cont` at the GDB
1163command line to request it to continue. It may be a good idea to make
1164this into a gdb script and pass it to gdb as an argument.
1165
1166Developing Device Drivers
1167=========================
1168
1169Nearly all UML drivers are monolithic. While it is possible to build a
1170UML driver as a kernel module, that limits the possible functionality
1171to in-kernel only and non-UML specific.  The reason for this is that
1172in order to really leverage UML, one needs to write a piece of
1173userspace code which maps driver concepts onto actual userspace host
1174calls.
1175
1176This forms the so-called "user" portion of the driver. While it can
1177reuse a lot of kernel concepts, it is generally just another piece of
1178userspace code. This portion needs some matching "kernel" code which
1179resides inside the UML image and which implements the Linux kernel part.
1180
1181*Note: There are very few limitations in the way "kernel" and "user" interact*.
1182
1183UML does not have a strictly defined kernel-to-host API. It does not
1184try to emulate a specific architecture or bus. UML's "kernel" and
1185"user" can share memory, code and interact as needed to implement
1186whatever design the software developer has in mind. The only
1187limitations are purely technical. Due to a lot of functions and
1188variables having the same names, the developer should be careful
1189which includes and libraries they are trying to refer to.
1190
1191As a result a lot of userspace code consists of simple wrappers.
1192E.g. ``os_close_file()`` is just a wrapper around ``close()``
1193which ensures that the userspace function close does not clash
1194with similarly named function(s) in the kernel part.
1195
1196Using UML as a Test Platform
1197============================
1198
1199UML is an excellent test platform for device driver development. As
1200with most things UML, "some user assembly may be required". It is
1201up to the user to build their emulation environment. UML at present
1202provides only the kernel infrastructure.
1203
1204Part of this infrastructure is the ability to load and parse fdt
1205device tree blobs as used in Arm or Open Firmware platforms. These
1206are supplied as an optional extra argument to the kernel command
1207line::
1208
1209    dtb=filename
1210
1211The device tree is loaded and parsed at boottime and is accessible by
1212drivers which query it. At this moment in time this facility is
1213intended solely for development purposes. UML's own devices do not
1214query the device tree.
1215
1216Security Considerations
1217-----------------------
1218
1219Drivers or any new functionality should default to not
1220accepting arbitrary filename, bpf code or other parameters
1221which can affect the host from inside the UML instance.
1222For example, specifying the socket used for IPC communication
1223between a driver and the host at the UML command line is OK
1224security-wise. Allowing it as a loadable module parameter
1225isn't.
1226
1227If such functionality is desireable for a particular application
1228(e.g. loading BPF "firmware" for raw socket network transports),
1229it should be off by default and should be explicitly turned on
1230as a command line parameter at startup.
1231
1232Even with this in mind, the level of isolation between UML
1233and the host is relatively weak. If the UML userspace is
1234allowed to load arbitrary kernel drivers, an attacker can
1235use this to break out of UML. Thus, if UML is used in
1236a production application, it is recommended that all modules
1237are loaded at boot and kernel module loading is disabled
1238afterwards.
1239