[Fedora-livecd-list] related to 6/7, mayflower loop device cleanup proposal

[Douglas McClendon]

Jeremy Katz wrote:
> On Mon, 2007-09-17 at 13:55 -0500, Douglas McClendon wrote:
>> Now that you're into the guts of genMinInstDelta (patch 6/7), maybe this 
>> topic is very timely-
> 
> Heh, indeed :-)  -ENEEDMORECAFFEINE first though before I finish looking
> at 6 and 7.

Yeah, I'm at the end of my caffeine run, so I'll respond to further 
stuff much later tonight or tomorrow.

> 
>> All those hard coded (117/118/119/120/121) loop devices used are very ugly.
>>
>> Probably they should get replaced with dynamic choices via losetup -f, 
>> as my recent livecd-creator patch did (though it was already using 
>> dynamic via losetup -a prior, different issue).
> 
> Agreed.  Given that most people don't use loop devs explicitly, I can't
> see how this would be problematic.  Too bad losetup doesn't have a "find
> the first available loopdev backwards" option, though.  Because that
> would be even slicker as if there _are_ things lying around with
> loopdevs hard-coded, they wouldn't break.

Between that and the *nod* I'll take that as "go ahead and make the 
patch, and for the rare people that have broken hardcoded loop stuff 
that can't handle taken low-numbered loop devices, *and* want to run 
those broken things in a LiveOS environment, too bad"

> 
>> I think I can put together a patch that does this.  Will probably even 
>> look pretty slick, as they info would get conveyed via udev rules 
>> dynamically generated in mayflower-init (just like existing loop120 and 
>> loop121 rules, but based on dynamic rather than hard-coded loop devices).
> 
> *nod*  
> 
>> The other thing, is I seemed to notice that, at least for my test cases, 
>> the mknod-s of the /dev/loop??? devices in the mayflower-initramfs seem 
>> unnecessary, as udev seems to create them just fine.  Anybody (david? 
>> jeremy?) know of why they are perhaps needed?
> 
> pre-udev version I suspect.  There shouldn't be a need for it any more
> that I can see
> 
>> I know you mentioned Jeremy, about merging mayflower into mkinitrd.  In 
>> fact, that is one reason why I did the rewrite, as I need for it to 
>> remain runnable as non-root.  You should probably get mkinitrd that way 
>> too, but...?  But my point is that it's still probably good to do these 
>> cleanups to mayflower, as they will make the merge into mkinitrd that 
>> much easier/better.
> 
> FWIW, I have a branch of mkinitrd up that has the changes so that
> booting a live image with an initrd generated by it "works".  The caveat
> being that some things like the udev rules aren't currently there and I
> sort of want to figure out a different way of conveying that.  But ran
> out of time.  For anyone interested, 'git clone
> http://katzj.fedorapeople.org/git/mkinitrd.git'.
> 
> And yeah, the amount of mkinitrd requiring being run as root is so small
> and piddly that it really is quite annoying.  Just haven't had the round
> tuits to fix them.

Is it anything that can't be handled by the fakeroot tool that Colin 
recently pointed me at?  (and perhaps mtools?)

-dmc