A sub-Hurd is like a neighbor Hurd, however, makes use of some resources provided by another Hurd. For instance, backing store and the console.

Sub-hurds are extremely useful for debugging core servers as it is possible to attach to them with gdb from the parent (debugging via subhurds). This avoids deadlock, e.g., when the instance of gdb stops the server but requires its use. (Note: it is possible to use noninvasive debugging, but this is less flexible.) Vice versa, it is also possible to use a subhurd to debug the main Hurd system, for example, the latter's root file system, but that requires a privileged subhurd.

Howto

Quick hack: Use an existing image

You can download a Debian/Hurd image here and directly boot it:

$ wget --no-check-certificate https://cdimage.debian.org/cdimage/ports/latest/hurd-i386/debian-hurd.img.gz
$ gunzip debian-hurd.img.gz
$ boot --kernel-command-line="fastboot root=pseudo-root" -T typed part:1:file:debian-hurd.img

The 'fastboot' is necessary to skip the filesystem check which fails because the image assumes the root filesystem to be /etc/hd0s1. Once booted, you can correct this:

root@debian:~# settrans -c /dev/pseudo-root /hurd/storeio pseudo-root
root@debian:~# echo /dev/pseudo-root / ext2 defaults 0 1 >/etc/fstab
root@debian:~# halt

From now on, boot -T typed part:1:file:debian-hurd.img should suffice.

Preparing your own image

To run a subhurd, you need an additional partition (or a file) with an installed Hurd system. In principle, you can also use your main partition in read-only mode; but this obviously will create severe limitations. Usually, you will want a complete independent system.

The following instructions assume that the system is installed on /dev/hd0s1. You can use any partition instead, or in fact just a simple file created e.g. using dd.

Creating a file used as root filesystem for the Subhurd

To create a file my.image of size 2 gigabytes, do

$ dd if=/dev/zero of=my.image bs=1M count=1 seek=2000

(Note that it is currently problematic to create files larger than 2 gigabytes this way.)

Using debootstrap

The system for the subhurd is a normal Hurd installation, which could just as well run standalone. You can use any of the various possible installation methods, or reuse an existing installation if you already have several. If using Debian GNU/Hurd, the easiest is probably to use debootstrap as root:

mke2fs /dev/hd0s6
settrans -ca mnt /hurd/ext2fs /dev/hd0s6
debootstrap sid mnt/ http://httpredir.debian.org/debian
settrans -fga mnt

Booting

If you are using a recent version of the Hurd (>= 0.9), then you can simply boot the subhurd as an unprivileged user by issuing

   boot /dev/hd0s6

(Replace hd0s6 by the name of your partition or backing file for the subhurd.)

The partition must be unmounted (or mounted read-only) before you boot from it!

Networking

You can provide the subhurd with a network card by passing a -f option to boot. For instance, if you have a second network card /dev/eth1 in your host hurd, pass -f eth0=/dev/eth1 to make it appear as device eth0 in the subhurd.

If you don't have a second network card, you can setup the eth-multiplexer to share one network card. To do so, install the multiplexer

settrans -c /dev/eth0m /hurd/eth-multiplexer --interface=/dev/eth0

Then configure your main Hurd system to use a virtual network interface (e.g. /dev/eth0m/0) instead. On Debian/Hurd, this can be accomplished using

ifdown /dev/eth0
sed -i -e s#/dev/eth0#/dev/eth0m/0# /etc/network/interfaces
ifup /dev/eth0m/0

You can now pass -f eth0=/dev/eth0m/1 to boot.

Booting and shutting down

Now the subhurd should boot just like a normal Hurd started directly from GRUB, finally presenting a login prompt. The boot program serves as proxy for the subhurd, so you can control it from the terminal where you issued the boot command.

To exit the subhurd, issue halt or reboot. This should exit it cleanly, but for some reason it doesn't always work; sometimes it will output various errors and then hang. If that happens, you need to kill the boot process manually from a different terminal. If the boot process dies, the proc server will kill all tasks belonging to the Subhurd.

Using

In the subhurd, you can do basically all the same things as in the main Hurd.

You can even set up networking: Just invoke settrans on the /servers/socket/2 as usual inside the subhurd, using /dev/eth0, only using a different local IP than in the main Hurd. This way, the subhurd will be able to communicate to the outside world with its own IP -- allowing for example to do apt-get inside the subhurd, or to ssh directly into the subhurd.

If you want to access the subhurd processes from the outside, e.g. for debugging purposes (or to get rid of a subhurd that didn't exit cleanly...), you need to find out how main Hurd ?PIDs correspond to subhurd processes: the subhurd processes appear in the main Hurd (e.g. if doing ps -e) as unknown processes, but the ?PIDs are different! To find out which process is which, you can simply compare the order -- while the numbers are different, the order should usually match. Often it also helps to look at the number of threads (e.g. using ps -l), as many servers have very characteristic thread counts.

IRC, freenode, #hurd, 2013-08-09

<teythoon> btw, is there a way to get dde-based networking into a subhurd?
<teythoon> the wiki instructions look like they're for the mach driver
<teythoon> and starting the dde translator inside the subhurd does not work
  for me
<teythoon> that's probably a good thing though
<youpi> the netdde process will need privileged access to mach
<youpi> for hardware access
<braunr> you can't easily use netdde from a subhurd, unless with a
  different nic
<braunr> i usually rebuild mach with in kernel devices so both the main and
  subhurd can share on nic
<braunr> one*
<youpi> could a port to netdde perhaps forwarded to the subhurd?
<braunr> zengh da wrote the eth-multiplexer for that iirc
<youpi> it's a matter of making it appear as an eth0 device on the master
  port aiui
<braunr> zheng*
<teythoon> yes, I looked at that
<teythoon> what is the master port?
<youpi> on a plain hurd system it's the port that privileged processes can
  use to access mach devices
<youpi> in a subhurd, it's the same for the subhurd, to access some devices
  that you choose to give access to
<braunr> its real name is the "device master port"
<teythoon> ah yes

IRC, freenode, #hurd, 2013-08-10

<antrik> teythoon: use eth-multiplexer to use the NIC within a
  subhurd. that's exactly what it was created for.
<antrik> I don't remember whether it's even possible to share a "raw"
  netdde device... I don't think I ever tried that; and I don't remember
  enough of the theory to tell whether it should be possible
<antrik> but I really don't see the reason to, when eth-multiplexer is
  available
<antrik> (IMHO running an eth-multiplexer on top of netdde should be the
  default setup in fact)
<antrik> as for actually passing on the device, that should be perfectly
  possible with zhengda's modified subhurd... but I don't remember whether
  that was ever merged upstream
<antrik> (you will definitely need that for using netdde in a subhurd,
  regardless whether through eth-multiplexer or directly)

IRC, freenode, #hurd, 2013-09-15

<teythoon> I wonder if we can modify the boot program so that it passes
  ports from the mother hurd to the subhurd
<teythoon> so that we could pass in a port to the eth-multiplexer
<teythoon> or use like /hurd/remap as the root translator for the subhurd

remap root translator.

<braunr> eth-multiplexer was created exactly for that iirc,
<braunr> so it's probably already done somewhere

IRC, freenode, #hurd, 2013-09-16

<gnu_srs> braunr: regarding subhurd did you mean to install
  sthibault/hurd-i386/debian-hurd.img.tar.gz
<gnu_srs> on a separate partition and booting using the instructions for
  subhurds on the web.
<braunr> gnu_srs: yes
<braunr> be careful that the subhurd doesn't use the same partition as the
  main hurd, that's all
<gnu_srs> what about changing fstab?
<braunr> 12:17 < braunr> be careful that the subhurd doesn't use the same
  partition as the main hurd, that's all
<teythoon> gnu_srs: yes, you need to change the fstab
<teythoon> currently it is used for fscking stuff, so if it points to your
  main partition it will cause severe corruption
<teythoon> gnu_srs: you also have to specify the right partition in the
  servers.boot file
<gnu_srs> fstab of the subhurd image?
<teythoon> yes
<gnu_srs> how to unpack the .img file (just to be sure)?
<teythoon> gnu_srs: you don't need to, just use the img file as secondary
  hard disk image
<gnu_srs> Then how should I be able to change fstab of the image?
<teythoon> boot your hurd box, mount the partition and change it
<gnu_srs> I missed something here: on my partition /my_chroot I have have
  the file debian-hurd-20130504.img
<teythoon> gnu_srs: ah, you copied it to the partition, braunr meant to use
  it as the secondary disk, e.g. qemu ... -hdb debian-hurd-20130504.img ...
<gnu_srs> That is the same as installing another cd image, where does the
  subhurd come into play?
<teythoon> mount the partition on the secondary hd, fix the fstab there,
  mount it r/o, get the servers.boot file from the wiki, modify it so that
  it points to the right partition, execute boot servers.boot /dev/<your
  partition>, probably /dev/hd1s1
<gnu_srs> BTW: unpacking was problematic: tar: debian-hurd-20130504.img:
  Cannot seek to 2147696640 (2G limitations)
<teythoon> I wonder why you did this on your hurd system in the first
  place...
<gnu_srs> I thought I could use that partition, /my_chroot as a chroot
  place. So it won't work for subhurds?
<teythoon> well, there are several ways to setup a subhurd. one is to
  already have a spare partition for that and use crosshurd or as I did
  debootstrap to install a debian system there
<teythoon> braunr suggested an even easier way, download the .img file and
  use it as secondary hard disk
<teythoon> you ended up doing kind of both
<gnu_srs> I tried once with debootstrap and that created a disaster...
<teythoon> how so?
<gnu_srs> The install errored out, and the whole filesystem (including /)
  was left in a broken state. Maybe I tried 
<gnu_srs> that without using a separate partition. Don't remember any
  longer. So you say it's safe now?
<teythoon> I used it successfully to setup my subhurd
<gnu_srs> and you have your subhurd in a separate partition, installed from
  there too, as root?
<gnu_srs> the web page only mentions crosshurd, and that failed for you?
<teythoon> yes, having a separate partition is (currently) necessary to run
  a subhurd
<teythoon> yes, I used debootstrap as root, afaics that is necessary
<teythoon> and yes, as I said the other day, I tried crosshurd first and it
  failed
<teythoon> then again, I fail to see any reason to use crosshurd these days
<teythoon> it's only a wrapper around debootstrap anyway, using it with
  --foreign and fixing up stuff later
<teythoon> one has more control over the process if one uses debootstrap
  directly
<gnu_srs> I still don't dare to do it yet. I'll create another image using
  netinst with a separate partition and try out first.
<gnu_srs> When installing a new image using netinst.iso (2013-06-30) and
  rebooting /proc does not get mounted?
<teythoon> gnu_srs: is that a statement or a question?
<gnu_srs> A statement.
<teythoon> it's not customary to end statements with question marks ;)
<gnu_srs> s/mounted?/mounted, why?/
<teythoon> well, you seem to be the last person to perform such an
  installation, so you are in the perfect position to answer this question.
<gnu_srs> cat /var/log/dmesg?
<gnu_srs> On other images I have: fsysopts /proc; /hurd/procfs
  --clk-tck=100 --stat-mode=444 --fake-self=1
<youpi> gnu_srs: no, check the installation log
<youpi> gnu_srs: and what does showtrans say?
<gnu_srs> showtrans /proc; <empty>
<gnu_srs> which log file to look for?
<youpi> the installation log, somewhere in /var/log probably
<gnu_srs> I only find /proc in /var/log/installer/syslog, mainly printing
  out errors not finding /proc/mounts
<youpi> iirc the /proc translator should be set during the hurd package
  configuration
<youpi> you should probably look for that part in the log
<youpi> Setting up translators: /hurd/exec /hurd/proxy-defpager
  /hurd/pflocal (+link) /hurd/pfinet (+link) (+link) /hurd/procfs -c
  /hurd/password crash-kill crash-suspend crash-dump-core crash.
<youpi> that part
<gnu_srs>  debootstrap:  /hurd/procfs -c and in-target:  /hurd/procfs -c No
  errors
<youpi> I don't understand what that means
<youpi> please explain in more details
<gnu_srs> see: http://paste.debian.net/41195/
<youpi> makes much more sense :)
<gnu_srs> Where is the 'Setting up translators' done? I cannot find
  anything in /var/lib/dpkg/info/hurd* or /etc/init.d/...
<pinotree> /usr/lib/hurd/setup-translators, called in hurd.postinst
<gnu_srs> tks:)
<gnu_srs> Hi, when installing a new image with debootstrap to /chroot the
  script boot/servers.boot is already there (as well as in /boot/ + grub)
<gnu_srs> Is it OK to use that file to boot the subhurd? 
<gnu_srs> using /boot/servers.boot or /chroot/boot/servers.boot (if the
  /chroot partition is unmounted it cannot be used?)
<gnu_srs> and how to unmount /chroot: umount does not work?
<gnu_srs> braunr: I'm also trying to find out what's wrong with glibc, when
  my subhurd is up and running 2.13-39 (if possible)
<gnu_srs> I know I should issue settrans command, but I'm not yet fluent in
  translators.
<gnu_srs> sorry:-/
<gnu_srs> Now this, after a reboot: unknown code P 30 while trying to open
  /dev/hd0s3 (/chroot)
<gnu_srs> Disk write protected: use the -n option to do a read-only check
  of the device.
<gnu_srs> fsysopts /dev/&hd0s1 --writable: Operation not supported??
<gnu_srs> OK, I'm giving up for now, no subhurd:-( and a broken install.
<gnu_srs> Which terminal to use in rescue mode, TERM is not set,
  dumb,mach,hurd does not work with nano?
<gnu_srs> e2fsck /dev/ho0s3; e2fsck: Unknown code P 2 while trying to open
  /dev/ho0s3; Possibly non-existent device?
<gnu_srs> mke2fs /dev/hd0s3; /dev/hd0s3 is not a block special device.;
  Proceed anyway? (y,n) n: What's going on (hd0s3 not mounted)??
<gnu_srs> anybody, help?
<gnu_srs> after removing and creating the partition again:mke2fs
  /dev/hd0s3, <same>, mke2fs: Unknown code P 13 while trying to determine
  filesystem size: What's going on?
<gnu_srs> Where to find the glibc-2.13 versions which used to be at
  debian-ports?.
<gnu_srs> seems they can be found on snapshot.debian.org

IRC, freenode, #hurd, 2013-09-17

<gnu_srs> teythoon: Installing subhurd via debootstrap on partition
  /chroot fails miserably. Install hangs, and after reboot \rm -r
  /chroot/* fails for dev and proc
<gnu_srs> Are there translators running there already? I have not
  booted the subhurd. 
<gnu_srs> translators for hd0s3 (/chroot) are storeio and
  ex2fs.static. Do I have to stop them to be able to clean out
  /chroot?
<gnu_srs>  mount -v /chroot; settrans -a /chroot /hurd/ext2fs
  /dev/hd0s3;
<gnu_srs> ext2fs: /dev/hd0s3: panic: main: device too small for
  superblock (0 bytes);
<gnu_srs> mount: cannot start translator /hurd/ext2fs: Translator
  died
<gnu_srs> Please, somebody!
<gnu_srs> don't ask to ask, just ask, right?
<braunr> we've already told you everything you need
<braunr> just get it right
<braunr> for example, i told you to be careful about fstab so that
  the subhurd wouldn't use the main hurd partition
<braunr> but you managed to screw that
<braunr> good job
<gnu_srs> I installed the subhurd in a partition /chroot /dev/hd0s3
  using debootstrap
<braunr> i don't know deboostrap, it may be broken, use the disk
  image youpi maintains
<gnu_srs> ant the install screwed up with debootstrap
<gnu_srs> ok; then I cannot use a partition, but another disk in
  kvm, e.g. hdb?
<braunr> gnu_srs: hd1
<gnu_srs> something is fishy with glibc, definitely, that's why I'm
  trying to set up a subhurd to revert to 2.13-39
<gnu_srs> hi, when trying to boot a subhurd: /hurd/ext2fs.static:
  hd0s3: Gratuitous error; bye
<braunr> gnu_srs: why hd0s3 ?
<braunr> it should be hd1s1
<gnu_srs> I'm still using a separate partition /my_chroot
  /hd0s3. Will switch to hd1 next. teythoon?
<gnu_srs> the servers.boot script use absolute
  paths:/hurd/ext2fs.static and /lib/ld.so.1 /hurd/exec, 
<gnu_srs> shouldn't they be relative to /my_chroot?
<braunr> no
<braunr> they're actually from your host
<gnu_srs> teythoon: please, how did you succeed to boot a subhurd
  in a partition?
<gnu_srs> using debootstrap
<teythoon> gnu_srs: from my shell history:
<teythoon> : 1374672426:0;debootstrap sid /mnt
  http://http.debian.net/debian/
<teythoon> : 1374673020:0;cp /etc/hosts /etc/resolv.conf /mnt/etc
<teythoon> : 1374673048:0;cp /etc/passwd /etc/shadow /mnt/etc
<braunr> teythoon: so it does work fine ?
<braunr> great
<teythoon> yes, why wouldn't it?
<teythoon> gnu_srs: I then remounted that partition r/o and used
  the servers.boot file from the wiki to boot it
<teythoon> braunr: why wouldn't it? (you do mean the debootstrap
  part, don't you?)
<braunr> teythoon: i don't know
<braunr> i've heard it wasn't maintained any more
<braunr> not being maintained is a good reason for something to
  become unusable/untrustable with time
<teythoon> o_O it is at the heart of d-i, isn't it?
<teythoon> I actually do most Debian installations using
  debootstrap directly
<braunr> ah
<braunr> ok :)
<braunr> teythoon: even hurd ones ?
<teythoon> braunr: well, just the subhurd installation, but that
  went as expected
<braunr> good
<gnu_srs> Finally: I found the reason for Gratuitous error, I used
  the /boot/servers.boot script, 
<gnu_srs> that being different to the one on the wiki:-/
<gnu_srs> Is it possible to copy files between a host hurd and
  subhurd, what about access to eth0? 
<gnu_srs> Hi, when starting the subhurd I see some warnings/error:
  http://paste.debian.net/41963/
<gnu_srs> 1) A spelling error execunable-> executable
<gnu_srs> 2) libports: invalid destination port
<gnu_srs> 3) mach-defpager: another already running
<pinotree> "execunable" is not a typo, but just "exec" and "unable
  ..." without a space-type character
<gnu_srs> OK, sounds more plausible
<gnu_srs> Ah, the printouts are mixed, no bug
<gnu_srs> When setting up nework in the subhurd: /hurd/pfinet:
  file_name_lookup /dev/eth0: Translator died
<gnu_srs> /hurd/pfinet: device_open(/dev/eth0): (os/device) no such
  device
<gnu_srs> settrans: /hurd/pfinet: Translator died

IRC, freenode, #hurd, 2013-09-18

<youpi> priority does not matter much
<youpi> memory manager is not really surprising, there's indeed already one
<youpi> what is actually the problem?
<gnu_srs> So these are merely warnings?
<youpi> gnu_srs: yes
<gnu_srs> Real problems are I cannot set up networking, e.g. wget ...:
  Connecting to ... failed: Address family not supported by protocol.
<youpi> gnu_srs: did you give the subhurd a network card?
<gnu_srs> How?
<gnu_srs> and do I need to set up fstab, for now it's empty.
<gnu_srs> I just installed the base with dbootstrap
<youpi> gnu_srs: -f option of boot
<youpi> e2fsck will need fstab for sure
<youpi> otherwise it can't divine what should be checked
<gnu_srs> Why is the /boot/servers.boot different from the subhurd one on
  the wiki? Is it used at all, I thought grub was in charge.
<youpi> it's not used at all
<gnu_srs> maybe better to put in the subhurd one there then, with a
  comment?
<youpi> no, since /boot/servers.boot is supposed to be used for machine
  boot
<youpi> not subhurd boot
<gnu_srs> what about putting a copy of the suhurd one there, with a
  different name?
<youpi> probably a good idea, yes
<youpi> matter of making it happen
<gnu_srs> the wiki page on subhurd does not say how to set up networking,
  only that you can do it.
<youpi> matter of adding the information
<youpi> I remember it's the -f option of boot
<youpi> make it work, and add the information for others
<gnu_srs> I could try, but don't know how to add a network card to the
  subhurd, and e.g. how to set up swap
<youpi> see -f option
<gnu_srs> of boot?
<youpi> "gnu_srs: -f option of boot"
<youpi> if you could read what we write, that'd make things happen way
  faster
<gnu_srs> yes I saw your comment above, it was just to be 100% sure:-D
<gnu_srs> device_file=/dev/eth0 or something else?
<gnu_srs> eth0 is used by the host already
<youpi> did you read boot --help ?
<youpi> iirc it's not a problem, both will receive all frames
<gnu_srs> yes I did
<youpi> then I don't see where you took device_file from
<youpi> at least in that form
<youpi> --device=device_name=device_file
<youpi> that means rather something like --device=foo=bar
<gnu_srs> so -f /dev/eth0 is correct usage then?
<youpi> didn't you see that in what I wrote, there was a "=" in there?
<gnu_srs> -f is the short option, --device is the long, I don't see the
  need for = in the short option?
<youpi> in the long option there are *two* =
<gnu_srs> yes, but in the short no?
<youpi> why not?
<youpi> long -> short usually drops one =
<gnu_srs> to summarize: -f=/dev/eth0 or --device=eth_sub=/dev/eth0?
<youpi> why shouldn't there be a eth_sub in the short version?
<gnu_srs> 10:15:49) youpi: long -> short usually drops one =
<youpi> yes, it drops the =
<youpi> but  nothing else
<youpi> if the long option needs some information, the short needs it too?
<youpi> -?
<gnu_srs> correct now? -f eth_sub=/dev/eth0 or --device=eth_sub=/dev/eth0?
<youpi> yes
<gnu_srs> k!

Further Info

Read about using a subhurd for debugging purposes.

Roland's tutorial about running a subhurd.

Use Cases

Debugging the Main Hurd System

Note: This only works with privileged subhurds.

A subhurd can be used for debugging the main Hurd system. This works as long as the subhurd doesn't use any services provided by the main Hurd. For example, if you already have a subhurd running at the time it happens, you can use that one to debug a deadlocked ext2fs root file system in the main Hurd.

For this, you need to get a handle to the main Hurd's ext2fs translator's ?PID, but this is no problem, as currently ?PIDs are visible across subhurd boundaries. (It is a open issue hurd whether this is the right thing to do in virtualization contexts, but that's how it currently is.)

Unit Testing

freenode, #hurd channel, 2011-03-06:

From unit testing.

<youpi> it could be interesting to have scripts that automatically start a
  sub-hurd to do the tests
<youpi> though that'd catch sub-hurd issues :)
<foocraft> so a sub-hurd is a hurd that I can run on something that I know
  works(like linux)?
<foocraft> Virtual machine I would think
<foocraft> and over a network connection it would submit results back to
  the host :p
* foocraft brain damage
<youpi> sub-hurd is a bit like chroot
<youpi> except that it's more complete
<foocraft> oh okay
<youpi> i.e. almost everything gets replaced with what you want, except the
  micro-kernel
<youpi> that way you can even test the exec server for instance, without
  risks of damaging the host OS
<foocraft> and we know the micro-kernel works correctly, right youpi?
<youpi> well, at least it's small enough that most bugs are not there
<foocraft> 1) all tests run in subhurd 2) output results in a place in the
  subhurd 3) tester in the host checks the result and pretty-prints it 4)
  rinse & repeat
<youpi> the output can actually be redirected iirc
<youpi> since you give the sub-hurd a "console"
<foocraft> youpi, yup yeah, so now it's more like chroot if that's the case
<youpi> it really looks like chroot, yes
<foocraft> but again, there's this subset of tests that we need to have
  that ensures that even the tester running on the subhurd is valid, and it
  didn't break because of a bug in the subhurd
<tschwinge> As long as you do in-system testing, you'll always (have to)
  rely on some functionality provided by the host system.
<foocraft> the worst thing that could happen with unit testing is false
  results that lead someone to try to fix something that isn't broken :p
<tschwinge> Yes.
<youpi> usually one tries to repeat the test by hand in a normal
  environment