Re: [RFC PATCH v2 1/2] PM: domains: Skip disabling unused domains if provider has sync_state

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

 



Hi,

On Mon, Feb 20, 2023 at 9:12 AM Bjorn Andersson <andersson@xxxxxxxxxx> wrote:
>
> > I suspect that here we'll have to agree to disagree. IMO it's a
> > non-goal to expect hardware to work for which there is no driver. So
> > making the backlight work without a backlight driver isn't really
> > something we should strive for.
> >
>
> Without trying to make you agree ;)
>
> How can you differentiate between "the driver wasn't built" and "the
> driver isn't yet available"?

BTW, when I was responding to Saravana's series [1], I realized that
you _can_ differentiate between these two cases, at least from a
practical point of view. Specifically, when the
"deferred_probe_timeout" expires then you should assume that "the
driver wasn't built". Said another way, once the
"deferred_probe_timeout" expires then you should assume that the
driver won't be available in the future. While you still could try
loading it, in general once that timeout has expired the kernel has
made decisions (like making -EPROBE_DEFER non-retriable) that make it
very awkward to load new drivers.

Of course, one could say "hey, let's get rid of the
deferred_probe_timeout". That might be a tough sell unless you can
come up with an equivalent solution for those currently using this
feature.

[1] https://lore.kernel.org/r/CAD=FV=XQnLpD1P8sRBcizTMjCQyHTjaiNvjcPdgyZc5JCzvOtw@xxxxxxxxxxxxxx


> Consider the case where I boot my laptop, I have some set of builtin
> drivers, some set of drivers in the ramdisk and some set of drivers in
> the root filesystem.
>
> In the event that something goes wrong mounting the rootfs, I will now
> be in the ramdisk console. Given the current timer-based disabling of
> regulators, I have ~25 seconds to solve my problem before the backlight
> goes blank.

I personally don't love the timeout. It feels like the kind of thing
that userspace knows and should be able to tell the kernel. I know we
don't like to put impositions on userspace, but userspace is pretty
definitely involved in things like loading modules. It just makes
sense (in my mind) for userspace to say when it's done and all modules
for cold-plugged devices have been loaded...

-Doug




[Index of Archives]     [Linux ARM Kernel]     [Linux ARM]     [Linux Omap]     [Fedora ARM]     [Linux for Sparc]     [IETF Annouce]     [Security]     [Bugtraq]     [Linux MIPS]     [ECOS]     [Asterisk Internet PBX]     [Linux API]

  Powered by Linux