Updating Linux host software causes host OS crash => Fixed in SVN

[CChrisA]

I am currently automatically picking up updates from you Debian repositories.

When the latest update arrived, it attempted to recompile its kernel module and then replace the module whilst my m/chine was up and there were active VMs. This caused a crash of the host system. (Total freeze, reset button required.) Fortunately no important data was lost, though minor use of backups was required.

This is essentially a show stopper as an automated process can, without warning, knock over the host machine.

[Frank Mehnert]

The question is how this can happen. If you have running VMs then it is normally not possible to remove the kernel modules. Did you enable something like CONFIG_MODULE_FORCE_LOAD in your kernel config?

[CChrisA]

Good point. Am running a stock debian kernel. config snip follows:

CONFIG_MODULES=y CONFIG_MODULE_FORCE_LOAD=y CONFIG_MODULE_UNLOAD=y CONFIG_MODULE_FORCE_UNLOAD=y

Perhaps you could check to see if there are any processes running? Will also be raising bug over at debian.

[CChrisA]

I have had a look at the actual .deb package. I recall it barreling through without actually asking if it could replace the kernel, though I can't find where it would insist on that in the source. Its possible that it is using an old question answer regarding whether or not it should build a module rather than asking again if its OK to chance its luck.

Many people run aptitude update/safe-upgrade regularly for patches and fixes. It needs to clearly signal its intentions so there are no mishaps.

[Frank Mehnert]

Fixed in SVN. Actually we didn't properly lock the kernel module as long as an internal network was active. See r17344 + r17375.