[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [PATCH] Enable support for bootloader on md RAID1 partitions.

On 06/29/2009 02:33 PM, Radek Vykydal wrote:
Hans de Goede wrote:

Ah, good catch, looking further I cannot help but notice, that for the
other use of gtPart, one can just as well use bootDev, as
searches for the first physical device which is on the same disk
as bootDev, and then we use getDiskPart() to get only the diskname, so
if we just pass in gtPart directly, nothing changes.

I was looking at this, because if we do this, we can then get rid of
the gtPart variable and of the getMatchingPart() method (which
seems to be a bullshit method anyways).

Yes, I was thinking the same but I wanted to keep this patch
expecting discussion about whether we really want to start to install
grub on
md raid1 boot partitions.

Ah, well AFAIK, that has always been the intention, and once upon a time
we actually did that, so it is not that controversial.

+ elif path.startswith('md') and not flags.cmdline.has_key("iswmd"):
+ target = "partition"

I'm afraid this check is not 100% bullet proof, when we have the
key iswmd, we can still be installing grub on a partition, or despite
the presence of the flag, the mdraid device node path, may actually point
to a regular mdraid set, not one using isw metadata.

You are right, maybe this kind check (i.e. for iswmd) should go somewhere
to partitioning and/or bootloader selection so that this part of code
can expect
grub target /dev/md* to be regular md raid 1.

That wont work, as target will be /dev/md126 for example when asking
booty to use the mbr, in combination with isw BIOS raid using

I think this part needs a bit more thinking through, we probably have
more places in booty where we expect md* to mean this is 100% software
raid on partitions, where as with imsm mdraid it could also mean
its BIOS raid using whole disks,



[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]