Re: [PATCH] build some drivers only when compile-testing

On Wed, 2013-05-22 at 19:23 -0700, Greg Kroah-Hartman wrote:

On Wed, May 22, 2013 at 11:18:46AM +0200, Jiri Slaby wrote:

quoted

Some drivers can be built on more platforms than they run on. This
causes users and distributors packaging burden when they have to
manually deselect some drivers from their allmodconfigs. Or sometimes
it is even impossible to disable the drivers without patching the
kernel.

Introduce a new config option COMPILE_TEST and make all those drivers
to depend on the platform they run on, or on the COMPILE_TEST option.
Now, when users/distributors choose COMPILE_TEST=n they will not have
the drivers in their allmodconfig setups, but developers still can
compile-test them with COMPILE_TEST=y.

I understand the urge, and it's getting hard for distros to handle these
drivers that just don't work on other architectures, but it's really
valuable to ensure that they build properly, for those of us that don't
have many/any cross compilers set up.

quoted

Now the drivers where we use this new option:
* PTP_1588_CLOCK_PCH: The PCH EG20T is only compatible with Intel Atom
  processors so it should depend on x86.
* FB_GEODE: Geode is 32-bit only so only enable it for X86_32.
* USB_CHIPIDEA_IMX: The OF_DEVICE dependency will be met on powerpc
  systems -- which do not actually support the hardware via that
  method.

This seems ripe to start to get really messy, really quickly.  Shouldn't
"default configs" handle if this "should" be enabled for a platform or
not, and let the rest of us just build them with no problems?

Debian aims to provide a consistent feature set across all supported
architectures so far as possible, and I would expect other distributions
to do the same.  I don't believe anyone is coordinating defconfigs
across architectures to ensure that.

Distributions may also have particular requirements from userland that a
distribution-agnostic defconfig obviously doesn't cover.  (Isn't that
why we had the discussion about 'configure for my distribution' some
months back?)

For the driver configuration, what we provide in Debian is closer to
allmodconfig, only with some attempt to exclude those useless drivers.
Dependencies like this would make it easier to do that.

What problems is this causing you?  Are you running out of space in
kernel packages with drivers that will never be actually used?

On most x86 systems this isn't going to be an issue.  But if packages
are built natively (as they are for most distributions) then building
useless drivers for some architecture which doesn't have fast processors
available is a waste of precious resources.

This is particularly bad where the architecture doesn't support generic
kernels that boot on a wide range of machines and therefore we must
build many times over.  These unfortunately tend to be among those with
slower processors (ARM[1], MIPS, SH, ...).  Currently, Debian's linux
package takes ~48 hours to build for armel (ARM processors without FPU)
- and that's without building many of the PCI drivers we could for those
platforms which have PCI support.  So that's a minimum of 2 days to
provide a security update across all architectures[2].

[1] I'm aware that ARM is getting better in this regard and most ARMv7
machines are likely to be supportable by a single kernel image soon.
That doesn't mean the whole problem is solved.

[2] We don't always wait for all builds before releasing/announcing an
update; for example <http://www.debian.org/security/2013/dsa-2669>.  But
that's no comfort for those using the slower architecture.

quoted

+config COMPILE_TEST
+     bool "Compile also drivers which will not load" if EXPERT

EXPERT is getting to be the "let's hide it here" option, isn't it...

This little detail seems likely to reduce the usefulness of randconfig
as many drivers will become dependent on both EXPERT and COMPILE_TEST
being selected.

I don't know, if no one else strongly objects, I can be convinced that
this is needed, but so far, I don't see why it really is, or what this
is going to help with.

Ben.

-- 
Ben Hutchings
friends: People who know you well, but like you anyway.