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

Re: [PATCH 5/5] booty changes for iswmd



Hans de Goede wrote:

On 06/11/2009 06:54 PM, Dan Williams wrote:

<snip>

We also need to consider the case where the user has disabled raid
support in the bios, but the metadata and the 'iswmd' command line
parameter still exist. I would expect that the OS ignores raid arrays in
that case. Currently the bios-raid-enabled state can be determined by
"mdadm --detail-platform -e imsm" returning 'true'.


Hmm,

Currently with dmraid we're not handling this, and I'm not sure we should,
what if the user does:
* reset bios to defaults
* boot
* remember oh, darn I needed to durn on raid in the bios
* reboot

Now we have a fscked up raidset, because if for example it was a mirror,
we will have used the partitions from the first disk (most likely) without
updating the second, and without marking things dirty in the metadata.

To me, if someone wants disks to stop being seen as raid, they should first
mark them as not part of a raidset in the orom, and then after that if
they have no other raidsets they can disable the orom.


Could there be a prompt in the boot process to ask the user to resolve the ambiguity? Being safe by default is nice, but it's also annoying when you know what you are doing.

It can definitely be a prompt at installation time if the user would like to install to raw disks, but did not clear the metadata yet... right?

There is another platform configuration issue that will need to be addressed down the road. What to do when a user moves a raid member to an "unsupported" port, as it will be missing as far as the option-rom is concerned. I sketched out a possible mdadm configuration file parameter for managing this policy here:

	http://article.gmane.org/gmane.linux.raid/22237.

Thanks,
Dan


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