New: [rhinstaller/anaconda/pulls/21 master] find valid mounted boot stage1 parts on any disk (#1168118)

AdamWill installerbot-noreply at redhat.com
Mon Mar 2 23:25:44 UTC 2015


This is another refinement to the patch I and Anne worked on
around F21 final which aims to fix RHBZ #1168118. When the
target platform requires the bootloader stage1 device to be
a partition with a specific mount point, this patch causes
the boot disk selection logic to prefer a disk with a correct
partition on it as the 'boot disk' if there is one. If the
platform also requires the partition to have a particular
filesystem it also checks that. If no matching partition can
be found the first disk in the list of available disks will be
chosen, as before.

This version of the patch does not apply the logic to
platforms which require the stage1 device to be a partition
but do not require it to be mounted, because that behaviour
is more questionable. For instance, in the case where the
user does an install to two disks, the second containing
a partition of the appropriate type, and chooses guided
partitioning, we would currently create a new stage1 partition
on the first disk. If we applied this logic to such a platform,
we would instead re-use the stage1 partition from the second
disk. That's a somewhat more arguable behaviour change, so
for the sake of conservatism this patch does not do it.

The case where a mount point is involved is more clear-cut,
because specifying a mount point *must* be done explicitly by
the user, and if they explicitly specify a partition's mount
point as the stage1 bootloader target mount point (and give
it the correct filesystem), it seems inarguably correct to
infer that they wish to use that partition as the stage1 target
device. Requiring them to *also* explicitly select the disk on
which it resides as the 'boot disk' and failing otherwise just
seems obtuse.

Revised by: mulhern <amulhern at redhat.com>
-- 
To view this pull request on github, visit https://github.com/rhinstaller/anaconda/pull/21


More information about the anaconda-patches mailing list