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

Re: nslu2-utils, update-initramfs and triggers



On Sun, 27 Jul 2008, Joey Hess wrote:

> maximilian attems wrote:
> > the linux image in postinst does *not* call yet update-initramfs,
> > but mkinitramfs-kpkg, which is not triggered.
> > so that loophole does not yet exist.
> 
> Thanks for explaining that (and for your patience).
> 
> > right i'd have to rethink to scrap mkinitramfs-kpkg postlenny.
> 
> Yes, it will need some real care to avoid the power outage gap when
> dropping it.

ok rechecked the trigger code in update-initramfs.
the only triggered action is -u aka update, thus if i call
create -c in linux image postinst we'd have the same state
as now. also create does not call run_bootloader().
happy. :)
 
> > for now there is just the flash_kernel missing after the generated
> > initramfs in update-initramfs.
> 
> Well, my understanding now is:
> 
> - kernel package installed
> - mkinitramfs-kpkg generates initramfs #1
> - postinst_hook=flash-kernel is run, flashing new kernel+initramfs
> - <stuff>
> - update-initramfs triggered, generates initramfs #2

ack, the run #2 happens due to several hook scripts dropping in
and having the update-initramfs -u call in postinst.
 
> If so, there's no pressing need to reflash the kernel at the end. It's
> only confusing/ugly that as Paul Collins reported it can seem to flash
> before updating. But not dangerous.
> 
> We can modify update-initramfs to call flash-kernel, but it will just
> flash the second build of the initramfs. Is there any reason to perhaps
> want to use that one? Or will we be just be doubling the write cycles to
> the flash?

well it depends, as initramfs-tools does *not* depend on for example
cryptsetup, but if the user had configured a cryptoroot (i'm not sure
about when cryptsetup gets installed by d-i). would that be before
initramfs-tools? then the second call would be gratious.

 
> FWIW, I've committed a patch to flash-kernel svn so that you could do
> somehing like this:
> 
> if flash-kernel --supported >/dev/null 2>&1; then
> 	flash-kernel
> fi

ok cool. it depends on the order of installed of packages if that
call is necessary (lvm2, mdadm, cryptsetup, dmraid, loop-aes, uswsusp).

will wait on your feedback.

-- 
maks


Reply to: