You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
After a reboot, ssh to the cirros instance is failing because the hostkeys have changed. The reason seems to be that /etc/dropbear is a symlink to /var/run/dropbear, and /var/run again is a symlink to /run, which is a tmpfs. The expected result would be for the hostkeys to be stored permanently so that they are persisted after a reboot.
Seems the current handling was introduced in #45 , which missed the "remove the symlink on RW fs" part introduced in buildroot/buildroot@e7d04dd. Will create a patch that tries to do both, although I'm not sure where in the cirros context the root fs would actually be RO.
buildroot starts out with /etc/dropbear being a symlink to
/var/run/dropbear in order to allow dropbear to start even when /etc is
located on a RO filesystem. The startup script is then to check whether
the symlink can be replaced by an actual directory, which then allows to
permanently store the hostkeys that dropbear generates, ensuring the
keys will persist across reboots. This patch amends the S50dropbear init
script to copy this functionality from the original buildroot script.
Also add ed25519 to the list of hostkey types that are generated before
dropbear starts.
Fixes: #118
Signed-off-by: Dr. Jens Harbott <[email protected]>
After a reboot, ssh to the cirros instance is failing because the hostkeys have changed. The reason seems to be that
/etc/dropbear
is a symlink to/var/run/dropbear
, and/var/run
again is a symlink to/run
, which is a tmpfs. The expected result would be for the hostkeys to be stored permanently so that they are persisted after a reboot.This may also be related to #110.
The text was updated successfully, but these errors were encountered: