On Wed, May 20, 2009 at 03:18:34PM -0400, James Carlson wrote: > Nicolas Williams writes: > > There could be times where updating the archive won't be feasible > > because, say, a driver needed for mounting / needs to be updated to > > match a firmware update, and a panic occurs before the archive update -- > > if the driver and firmware have to be updated together then you have a > > brick. To fix this should require using a recovery CD or net boot. > > > > More complex cases might involve heavy hacking on /etc/system vis-a-vis > > the root filesystem. You'd be on your own then though. > > If so, then that argues for having either a flag that keeps the system > from going into a potentially 'bad' rebuild / reboot / panic loop.
Bad rebuild should cause a sulogin or a hang, not reboot. > (Whether the system sticks in the completely unusable maintenance mode > or skips the update and runs crippled for this one corner seems like a > coin toss to me.) If it can mount / rw (and /usr, if it's separate) then the boot archive can be rebuilt, otherwise the best you can do is give the user a sulogin/shell, or maybe not even (in which case rebooting won't help, so better just spew on console and wait for user input). > I've rebuilt more archives than I care to count, and I can't say I've > ever encountered such a problem, but I guess anything's possible. Me too -- I was just imagining what cases could fail to rebuild boot archives. All such cases would be the result of something bad having happened that is so rare that forcing fallback on a recovery CD or net boot seems reasonable. Nico --
