NB: the variable 'directory' specifies that the image should be
mounted, when initrd-bootchain feature using.

This one fixes the "cooler goes off during installation"
10.2rc problem on multi-CPU e8c2 motherboards that follows
misinitialization of sensor drivers (lacking i2c mux init
_before_ them) and subsequent misreading of zero degree
temperature for CPUs (would need no cooling if it was true).

See-also: mcst#8627
Suggested-by: mcst.ru folks
Tested-by: elbrus.ru folks

It only makes sense on engineering samples of .404 board,
and 201-PC is .480 with only a single 4X PCIe gen3 slot
that needs riser to get a reasonable videocard connected
(manual multiseat setup is a minor hurdle it seems).

We don't do ship vivante drivers since p9_e2k
as those behaved erratically with gtk+3 at least
(and were a major PITA to handle in the first place,
and nobody asked in the end: folks just needed 2D+Xv);
so no sense to bring extra noise into installation.

See the .318-01 related story in a nearby pkg.in commit.

e1cp and e2c3 have tricky GPU driver modules that can misinit
given the lack of other modules they don't directly depend on;
let's provide softdep lists so that the prereqs do get loaded.

See-also: mcst#8098

The problem is with indirect module initialization dependencies
(the infamous "Driver probe function unexpectedly returned 256"
error when loading the module from initrd and not full rootfs);
let's provide modprobe with a hint.
Suggested-by: Vitaly Chikunov <vt@altlinux.org>
See-also: mcst#8089
See-also: mcst#8392

...accordingly to the package getting renamed.

201-PC has been developed using .404 motherboards
(IIRC using the two middle PCIe slots for GPUs);
1601-PC one has been done with .466 motherboard
by sciberia.io guys who have outrun me at that!

It's actually set within the corresponding apt-conf
but referring to that here is better than tossing
REPO=http/pvt all over the make runs.

It should do some work: nothing else might have pulled use/x11 in
(which brings another topic to the table: we'd probably better
have some platform-independent targets that pull archdep pieces in
under the hood).

Fixes: 2002e625

Looks like it can get stuck for quite some time on serial ports
that are present and might be occupied by other processes already.

NB: the missing modules break image build,
    avoid mga2/galcore/vivante by default
    (absent in some CPU-specific kernels).

To provide priority: installer, live, rescue.

Those comments were either outdated or unneeded here.

I've lost enough hair when this regression surfaced
and almost bothered legion@ but finally saw the

  UUID=...

instead of

  root=UUID=...

in kernel cmdline; this looked as a result of a thinko
that was pretty hard to spot at once but finally
vm/alt-rescue.img works again on Elbrus.

Fixes: 36d25f43

...to all the installers.

Fixed a bug with moving the core in a cycle. This only needs to be
done once.

...not just e2kv3; a home Elbrus-8CB user just asked about
Simply 10.1 beta for v5 as well.

We still didn't do multiseat on 4C (although it's technically
possible using either PCI videocard or an onboard SM7xx one
as the secondary head).

With this change, we simplify the created initrd image. Kernel
modules and firmware are added by make-Initrd itself. The layer
in Initrd is now one. It also allows you to unifyed configs for
bootloading with propagator and bootchain.
Config files for refind and elilo is creates in mkimage, so have
to copy initrd.img in efi/boot/ with the name full.cz.

This MCST utility is on the repos already, and it's going to be
useful both standalone and hopefully via other sysinfo utilities.

Looks like this helps installers with 5.4.0-2.20 based kernel
to install fine on Elbrus 801-PC with Radeon RX 580.

pwmd is finally available on all of v3/v4/v5 subarches,
and there's an installer feature for v4/v5 specifically.

This one is somewhat different for readonly optical media
(where [auto]mounted disk is a nuisanse) and writable USB
Flash or even fixed disk (where it's too easy to destroy
data by missing the right letter).

...so that it's more applicable to a wider set of images
(those lacking e2k-alt-linux' specific labels just won't
get broken default boot target label name for no reason).

...otherwise a whole-disk block device would be accepted
by the script but not by the firmware's bootloader.

Optical media isn't to be checked for those of course.

use/e2k/multiseat/full invoked from many targets, regardless of the
fact that it is e2k.

No need to switch udisks2 into "shared mode" anymore;
see the previous commit.

There might be a more generic approach
but let's at least make this possible
in the first place.

...to the multiseat target.

Not that it's exported anyways!

This one is going to need adaptation for each private repo
but eventually should be gone in favour of sisyphus-mirror.

This one should ensure we can actually see propagator
(given dri/kms kernel drivers built as modules).