hi!
provides a nice feature. But: on live systems (like grml,
debian-live,...) this feature usually won't be activated by default
(because dhcp might not be available, causing delays, having
ssh-server/cryptsetup/... available later on anyway, etc).
Disabling the feature via initramfs.conf works fine of course, but
it would be nice if the feature could be *enabled* on request.
This would provide a setup where dropbear isn't enabled by default
in the initramfs but could be enabled for example if the initramfs
doesn't find a root filesystem.
Would be nice if the dropbear setup could be triggered by another
initscript of initramfs-tools (for example when locating the root
filesystem fails) and/or by using a bootoption like "dropbear=yes"
to enable it manually.
the keyword in initramfs.conf actually determines (among other
parameters) whether dropbear (binary and stuff) are installed into
the initrd image when it is built.
if it isn't included in the image, it ofc can't simply be 'switched
on' on boot.
there has just been a bugreport that the ip config, which was
unconfigured in that case (which meant it would automatically try
dhcp), prevented the system from booting (no network was attached).
currently, to prevent this from happening, the 'ip='-config has to
be explicitly set (or dropbear disabled). that's not exactly nice.
therefore, i guess, pushing the ip-config to the background, i.e.
making it non-blocking, would qualify as 'nice' (for all cases,
like: missing configuration, no network attached, no dhcpd running,
etc. - and your live-system setup, too, i guess?).
i've set this issue on my 2do. but i can't say if/when i have the
time to work on it. if you want to work on it:
in '/usr/share/initramfs-tools/scripts/functions' the function
'configure_networking' is probably the starting point. just adding a
trailing '&' to all 'ipconfig' lines should basically do the job for
our case. BUT: one should think about the output (e.g. just throw it
away, or collect it and display it at some point in a nice
fashion?), and what's much more important: currently all scripts
using this function can rely on it to block until the network is
configured. so a mechanism is needed to provide the
blocking-behaviour if it's needed (i'm thinking of e.g. nfs-root)...
(maybe this bugreport should be moved to package initramfs-tools?)
regards,
Chris
--
To UNSUBSCRIBE, email to [email protected]
with a subject of "unsubscribe". Trouble? Contact [email protected]
