Re: [PATCH v2 11/11] test_firmware: test three firmware kernel configs using a proc knob

[Josh Triplett]

On Thu, Mar 01, 2018 at 12:38:16AM +0000, Luis R. Rodriguez wrote:
> On Wed, Feb 28, 2018 at 04:00:58PM -0800, Josh Triplett wrote:
> > On Wed, Feb 28, 2018 at 06:26:03PM +0000, Luis R. Rodriguez wrote:
> > > So for folks who enable CONFIG_FW_LOADER=y, they'd now be forced to gain an
> > > extra 13436 bytes broken down as follows:
> > 
> > Ah, I see.
> > 
> > If you have CONFIG_FW_LOADER and not CONFIG_FW_LOADER_USER_HELPER, then
> > you only have the in-kernel firmware loading mechanism?
> 
> Right, we don't have the old fallback mechanism (which BTW used to be
> the default way back in the hayday).
> 
> > Given the
> > *substantial* size difference between the two, it seems useful to have
> > that option.
> 
> That's what I wanted to get to, is 13436 bytes is*substantial* enough to
> merit a kernel configuration option? It seems like that is the case.

By at least an order of magnitude, yes.

> > What would it gain to combine the two?
> 
> Well Android enables CONFIG_FW_LOADER_USER_HELPER, and if they do, I was trying
> to think if there really was any point in having CONFIG_FW_LOADER_USER_HELPER
> as an option. Who would enable CONFIG_FW_LOADER but not
> CONFIG_FW_LOADER_USER_HELPER?

An embedded system with a fixed set of hardware that needs exclusively a
fixed set of firmware files known at system build time.

> The less hairball of mess of kconfig options the better to test. Even
> though this series has reduced being able to consolidating being
> able to make a kernel now which lets us test all configurations in
> one build.
> 
> Who would save some 13436 bytes in the real world?

*raises hand*