bug#48266: support dynamic loading of modules from initrd

bug#48266: support dynamic loading of modules from initrd

[Vagrant Cascadian]

[-- Attachment #1: Type: text/plain, Size: 4528 bytes --]

There should be an option to support dynamic loading of modules from the
initrd.

I recently pushed some changes to use the "linux-libre" kernel with
pinebook pro:

  https://git.savannah.gnu.org/cgit/guix.git/commit/?id=d330d63a29f35ebcbebce19b13d49c69d38a5815

But in order to even boot, I need to add a lot of modules to
initrd-modules:

  (initrd-modules 
    (append
      (list
       "rtc-rk808"
       "dw_mmc"
       "dw_mmc-pltfm"
       "dw_mmc-rockchip"
       "joydev"
       "bluetooth"
       "jitterentropy_rng"
       "hid_generic"
       "videodev"
       "ghash_ce"
       "gf128mul"
       "ansi_cprng"
       "mc"
       "sha2_ce"
       "usbhid"
       "panfrost"
       "ecdh_generic"
       ;; "fusb302"
       "ofpart"
       ;; "tcpm"
       "hid"
       "ecc"
       "evdev"
       "leds_gpio"
       "io_domain"
       "dw_wdt"
       ;; "rockchip-thermal"
       "cw2015_battery"
       "pwm-rockchip"
       ;; "gpio_charger"
       "cpufreq_dt"
       "configfs"
       "xhci-plat-hcd"
       "xhci-hcd"
       "rk808-regulator"
       "clk-rk808"
       "udc-core"
       "ulpi"
       "fan53555"
       "rk808"
       "pwm-regulator"
       "fixed"
       "gpio_keys"
       "cec"
       ;; "phy-rockchip-typec"
       "phy-rockchip-emmc"
       "phy-rockchip-inno-usb2"
       ;; "analogix-dp"
       "sdhci-of-arasan"
       "sdhci-pltfm"
       "cqhci"
       "drm_kms_helper"
       "ohci-platform"
       "ohci-hcd"
       "ehci-platform"
       "panel-simple"
       "ehci-hcd"
       "sdhci"
       "drm"
       "i2c-rk3x"
       "usbcore"
       "pl330"
       "pwm_bl"
       "dwc3"
       )
  %base-initrd-modules))

Initially, I tried adding just the obviously mmc related modules, but
this gave me guile prompt from the initramfs as it failed to find the
rootfs. Notably, even with the above list, I still need to explore
additional modules to load in order to get the display and keyboard to
work from the initramfs, in case I wanted to use this with encrypted
rootfs...

The above list of modules could almost certainly be trimmed, but even
getting a bootable system for pinebook pro took about 20 tries, and the
process of defining the modules is further complicated by several
factors...

* The filesystem names of the modules (e.g. dw_mmc-pltfm) do not
  necessarily match the runtime name from lsmod (e.g. dw_mmc_pltfm).
  This becomes a good deal of trial and error to figure out which
  modules to add.

* One needs to manually resolve the soft and hard dependencies of the
  modules, and ensure they are loaded, and include them in the list.

* If upstream changes the module name (which does happen from time to
  time), you have to update the system config.scm to the new module
  names.

* If some functionality changes from a module to a built-in, or
  vice-versa, the system config.scm needs manual updating.

* Switching system between two different arm boards potentially requires
  entirely different lists of modules.


Rather than handling modules one at a time, I would propose to at least
add an option that can add whole directory trees of modules to the
initrd (e.g.  kernel/drivers/usb/)... and then use modprobe (or udev?)
to handle the dependencies. Maybe opt-in at first, but longer-term,
explore making it default?


In Debian, adding modules to the intird is done in initramfs-tools:

  https://salsa.debian.org/kernel-team/initramfs-tools/-/blob/f350f122a244c60f91a3e3e5f8b58f9cb75308b6/hook-functions#L599

As for which modules to load at runtime, initramfs-tools uses udev;
maybe that could be integrated into the guix initramfs as an option?

Obviously, adding more modules than a strict bare minimum to the initrd
will increase boot times a bit, and adding further dependencies
(modprobe, udev) to the initrd will add to that even more, but the
current hard-coded list of modules to load, that you can extend with
another hard-coded list, is a bit painful when trying to get a new
system working...


Maybe the Guix way to do this is to write some guile code that can
generate the list of modules to include in the initrd at build time? But
that still doesn't resolve the dynamic loading of modules at runtime,
and it would be impractical to load *all* the modules at runtime... and
maybe impractical to re-implement modprobe and/or udev in guile?


Well, thanks for considering!


live well,
  vagrant

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 227 bytes --]

bug#48266: support dynamic loading of modules from initrd

[Ludovic Courtès]

Hi!

Vagrant Cascadian <vagrant@debian.org> skribis:

> Initially, I tried adding just the obviously mmc related modules, but
> this gave me guile prompt from the initramfs as it failed to find the
> rootfs. Notably, even with the above list, I still need to explore
> additional modules to load in order to get the display and keyboard to
> work from the initramfs, in case I wanted to use this with encrypted
> rootfs...
>
> The above list of modules could almost certainly be trimmed, but even
> getting a bootable system for pinebook pro took about 20 tries, and the
> process of defining the modules is further complicated by several
> factors...
>
> * The filesystem names of the modules (e.g. dw_mmc-pltfm) do not
>   necessarily match the runtime name from lsmod (e.g. dw_mmc_pltfm).
>   This becomes a good deal of trial and error to figure out which
>   modules to add.
>
> * One needs to manually resolve the soft and hard dependencies of the
>   modules, and ensure they are loaded, and include them in the list.
>
> * If upstream changes the module name (which does happen from time to
>   time), you have to update the system config.scm to the new module
>   names.
>
> * If some functionality changes from a module to a built-in, or
>   vice-versa, the system config.scm needs manual updating.
>
> * Switching system between two different arm boards potentially requires
>   entirely different lists of modules.

Note that ‘guix system init’, ‘reconfigure’, and ‘deploy’ error out if
drivers for a storage device are missing (see
‘check-device-initrd-modules’).

Now, that doesn’t help if you’re using ‘guix system image’, which
perhaps is what you were doing?  I wonder how we could take advantage of
that code in such a scenario.

> Rather than handling modules one at a time, I would propose to at least
> add an option that can add whole directory trees of modules to the
> initrd (e.g.  kernel/drivers/usb/)... and then use modprobe (or udev?)
> to handle the dependencies. Maybe opt-in at first, but longer-term,
> explore making it default?

I remember Danny and I worked on something along these lines in the past
but it didn’t completely materialize (some of the machinery is already
here, though).  That said, we still wouldn’t want to include too much in
the initrd, would we?

Thanks,
Ludo’.

bug#48266: support dynamic loading of modules from initrd

[Vagrant Cascadian]

[-- Attachment #1: Type: text/plain, Size: 3199 bytes --]

On 2021-05-11, Ludovic Courtès wrote:
> Vagrant Cascadian <vagrant@debian.org> skribis:
>
>> Initially, I tried adding just the obviously mmc related modules, but
>> this gave me guile prompt from the initramfs as it failed to find the
>> rootfs. Notably, even with the above list, I still need to explore
>> additional modules to load in order to get the display and keyboard to
>> work from the initramfs, in case I wanted to use this with encrypted
>> rootfs...
>>
>> The above list of modules could almost certainly be trimmed, but even
>> getting a bootable system for pinebook pro took about 20 tries, and the
>> process of defining the modules is further complicated by several
>> factors...
>>
>> * The filesystem names of the modules (e.g. dw_mmc-pltfm) do not
>>   necessarily match the runtime name from lsmod (e.g. dw_mmc_pltfm).
>>   This becomes a good deal of trial and error to figure out which
>>   modules to add.
>>
>> * One needs to manually resolve the soft and hard dependencies of the
>>   modules, and ensure they are loaded, and include them in the list.
>>
>> * If upstream changes the module name (which does happen from time to
>>   time), you have to update the system config.scm to the new module
>>   names.
>>
>> * If some functionality changes from a module to a built-in, or
>>   vice-versa, the system config.scm needs manual updating.
>>
>> * Switching system between two different arm boards potentially requires
>>   entirely different lists of modules.
>
> Note that ‘guix system init’, ‘reconfigure’, and ‘deploy’ error out if
> drivers for a storage device are missing (see
> ‘check-device-initrd-modules’).

Yes, I often have to tell it to skip those checks when using 'guix
system init', as I'm installing to a microSD card by way of a
usb-to-microSD adapter, and it guesses all the wrong modules.


> Now, that doesn’t help if you’re using ‘guix system image’, which
> perhaps is what you were doing?  I wonder how we could take advantage of
> that code in such a scenario.

I sometimes do 'guix system image' for the initial pass, and then
follow-up with init or reconfigure...


>> Rather than handling modules one at a time, I would propose to at least
>> add an option that can add whole directory trees of modules to the
>> initrd (e.g.  kernel/drivers/usb/)... and then use modprobe (or udev?)
>> to handle the dependencies. Maybe opt-in at first, but longer-term,
>> explore making it default?
>
> I remember Danny and I worked on something along these lines in the past
> but it didn’t completely materialize (some of the machinery is already
> here, though).  That said, we still wouldn’t want to include too much in
> the initrd, would we?

Notably, at the moment it loads various virtio modules on all my
baremetal systems, which is a bit uneccesary. :)

Honestly, when debugging support for a new arm system, I'm of the
opinion that more is better than less, as it takes way too many
iterations to get to a working modular configuration. So at least an
option to include even the kitchen sink drivers would be nice. :)


live well,
  vagrant

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 227 bytes --]