Existing scripts and associated config such as rc.initdiskless, rc.d/var,
and others, use mdmfs to create memory filesystems. That program accepts a
size argument which allows SI suffixes and treats an unsuffixed number as a
count of 512 byte sectors. That makes it difficult to convert existing
scripts to use tmpfs instead of mdmfs, because tmpfs treats unsuffixed
numbers as a count of bytes. The script logic to deal with existing user
config that might include suffixed and unsuffixed numbers is... unpleasant.
Also, there is no g'tee that tmpfs will be available. It is sometimes
configured out of small-resource embedded systems to save memory and flash
storage space.
These changes enhance mdmfs(8) so that it accepts two new values for the
'md-device' arg: 'tmpfs' and 'auto'. With tmpfs, the program always uses
tmpfs(5) (and fails if it's not available). With 'auto' the program prefers
tmpfs, but falls back to using md(4) if tmpfs isn't available. It also
handles the -s <size> argument so that the mdconfig interpetation of
unsuffixed numbers applies when tmpfs is used as well, so that existing user
config keeps working after a switch to tmpfs.
A new rc setting, mfs_type, is added to etc/defaults/rc.conf to let users
force the use of tmpfs or md; the default value is "auto".
Differential Revision: https://reviews.freebsd.org/D12301
Need to multiply the size of the disk passed to mount_md by 512 as mdmfs
expects number of 512-byte blocks while tmpfs size option wants number of
bytes.
Reviewed by: brooks
Approved by: sjg (mentor)
Obtained from: Juniper Networks, Inc.
Differential Revision: https://reviews.freebsd.org/D11106
This allows the location of the configuration data to be relocated
within the filesystem containing it. A nullfs mount is used in order
to achieve this.
Obtained from: XORP, Inc.
The rc.initdiskless functionality is used by NanoBSD to allow configuration
files to live on a separate configuration slice, which acts as NVRAM, whilst
the system image is mounted read-only.
Normally, if the remount command fails during boot, this is regarded as
a fatal error. If /conf/T/M/remount_optional is present, this error is
non-fatal. If the file is not present, the default behaviour is unchanged.
This is very useful for people building live CD images using FreeBSD,
where the NVRAM lives somewhere completely differently from the system image,
and may be present on removable media which is not present during the
initial boot.
- add better checks on non-existing directories to prevent error
messages at run time;
- introduce a function log() to help debugging diskless booting
when things don't work;
- modify the parsing of diskless_remount so you can add mount_nfs
options after the pathname. You could use 'remount' to do something
similar, but this way is more convenient because you don't have to
hardwire the server name in the command.
- document the above.
I have been running the above in a diskless lab since february on RELENG_6.
MFC after: 1 week
- Note that diskless_remount files may use ".." to support mounts above
the root path.
- Copy dot files when populating directories from /conf. [1]
PR: misc/102724 [1]
Submitted by: Attila Nagy <bra at fsn.hu> [1]
extraction.
This will allow cpio archive support to work, at least in situations
where /tmp is writable. Because pax requires a writable /tmp it is
unsuitable for this task, but replacing it will come in a later commit.
Submitted by: Joerg Pulz <Joerg dot Pulz at frm2 dot tum dot de>
PR: conf/88293
to run initdiskless before we run rcorder on /etc/rc.d. To allow this,
move /etc/rc.d/initdiskless to /etc/rc.initdiskless and run it directly
from /etc/rc.
Remove /etc/rc.d/preseedrandom as it is no longer necessicary (we start
with entropy unblocked) and was only used by initdiskless when it
was needed.
Discussed on: freebsd-rc
Repocopy by: peter
use the conf/* stuff for their firewall configuration.
Running ipfw before could seem to make sense in that it would allow
one to setup access to the NFS server on a "default-deny" kernel,
but it is pretty obvious to the casual observer that such a configuration
never makes it far enough to mount the NFS-root in the first place.
initdiskless. The output of several commands and if available the
contents of /entropy are feed into /dev/random to kickstart the PRNG.
/etc/rc.d/initrandom is left alone to maintain the previous behavior as
much as possiable.
Further work in this area is probably needed.
Discussed with: markm
comments and empty lines have been touched.
All of this should go in the diskless(8) manpage, now if we had
some kind of 'literate programming' tool to extract the comments
from the script and put them in a reasonable nroff format, it
would be a lot easier to keep code and docs in sync
+ SUBDIR.cpio.gz prevents files from SUBDIR/ to be copied when
priming the memory filesystems. This restores the old behaviour
and makes the copy process a lot more efficient
+ look for templates also in the list of directories supplied by
bootp/dhcp via the T134 option aka kern.bootp_cookie
+ keep track of directories temporarily mounted with "remount"
or "diskless_remount" commands and unmount them once we are done
with them (at the end of this script).
Trigger not only on diskless booting sysctls being set, but also
on the existence of the file "/etc/diskless". But do not try to
extract IP# related keywords in that case.
Add a general "remount" facility to allow non-NFS remounting.
Instead, load them as part of the rc.d system. This allows us to prioritize
the initidiskless script so it runs before the configuration files are loaded
and allows us to get rid of the exit 2 hack in /etc/rc. The exit 2 never
worked anyway since it did not unset the prior configuration, causing the
diskless code to not operate properly.
Do a major cleanup and revamping of the diskless code for RCNG. This will
be backported to the non-RCNG scripts as well as -stable.
With suggestions from: Mike Makonnen <mtm@identd.net>
MFC after: 7 days
* Space -> tabs conversion.
* Removed blanks before semicolon in "if ... ; then".
* Proper indentation of misindented lines.
* Put a full stop after some comments.
* Removed whitespace at end of line.
Approved by: silence from gordon
a per-machine or per-cluster (with different ways of expressing what's
part of a cluster) basis. In order for this to be effective, rc.conf
has to be reread after initdiskless is finished. Implement this by
adding a hook to etc/rc which rereads rc.conf by request. This can
also be implemented by renaming initdiskless to initdiskless.sh and
sourcing rc.conf there manually, but it was decided that, that would
be uglier than a hook in etc/rc.
Developed in concert with: gordon
/etc filesystem isn't enough; consequently, add "-i 4096" to the newfs
command for /etc. This results in 1022 inodes, which should be enough
for the forseeable future (although I don't know why we would ever
have more than 1000 files in a default /etc).
Silence by: -current
while. This is only the script pieces, the glue for the build comes next.
Submitted by: Mike Makonnen <makonnen@pacbell.net>
Reviewed by: silence on -current and -hackers
Prodded by: rwatson
and looks like no other Unix diskless configuration I've ever seen.
Thus allow a more traditional /etc.
Note, the use of an MFS /var should also be settable.
Otherwise installing ports(packages) is just a total PITA.
