linux-kernel - Re: [PATCH] PCI/pwrctl: reduce the amount of Kconfig noise

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <CAMRc=MdcUhBe6-P0KH8KfrHfp8wqM0fCWGoUZaVZ7+FAkECtpA@mail.gmail.com>
Date: Wed, 17 Jul 2024 14:31:57 +0200
From: Bartosz Golaszewski <brgl@...ev.pl>
To: Linus Torvalds <torvalds@...ux-foundation.org>
Cc: Bartosz Golaszewski <bartosz.golaszewski@...aro.org>, Bjorn Helgaas <bhelgaas@...gle.com>, 
	linux-pci@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] PCI/pwrctl: reduce the amount of Kconfig noise

On Tue, Jul 16, 2024 at 10:29 PM Linus Torvalds
<torvalds@...ux-foundation.org> wrote:
>
> On Tue, 16 Jul 2024 at 13:10, Bartosz Golaszewski
> <bartosz.golaszewski@...aro.org> wrote:
> >
> >     select PCI_PWRCTL_PWRSEQ if ARCH_QCOM
> >
> > in the ATH11K_PCI and ATH12K Kconfig entries? Am I getting this right?
>
> So on the whole, I'd prefer these things to be done where they are
> actually required.
>
> But I'm not actually entirely sure what the hard _requirements_ from a
> hardware - or a kernel build - standpoint actually are.
>
> If there aren't any hard requirements at all, then maybe the whole "do
> you want pweseq" should be an actual question (but limited only to
> situations where it makes sense).
>
> If the hard requirement is at the level of the driver itself, then the
> "select" should be in the driver.
>
> That doesn't seem to be the case here, since ATH11K_PCI certainly
> works without it, but if that driver requires power sequencing on
> ARCH_QCOM platforms, then maybe that is indeed the right thing.
>
> And if the hard requirement comes from some SoC setup, then optimally
> I think the "select" should be at that level, but we don't actually
> seem to have that level of questions (but maybe something in
>
>    drivers/soc/qcom/Kconfig
>
> might make sense?)
>

The hard requirement really comes from the board level - not SoC. It's
the board that has the BT/WLAN module hardwired and - depending on how
the module is powered - may or may not require power sequencing. But I
don't think we have any per-board infrastructure in Kconfig.

> Anyway, this is not necessarily something where there is "one correct
> answer". This may be a somewhat gray area, and it looks like ARCH_QCOM
> is a very big "any Qualcomm SoC" thing and not very specific.
>
> So I'm not sure what the right answer is, but I *am* pretty sure of
> what the wrong answer is. And this:
>
>         default m if ((ATH11K_PCI || ATH12K) && ARCH_QCOM)
>
> looks like it cannot possibly be right if ATH11K_PCI is built-in,
> since then the probing of the device will happen long before that
> PCI_PWRCTL_PWRSEQ module has been loaded.
>
> And that doesn't sound sensible to me. Does it?
>
> TL;DR:  I do think that the
>
>       select PCI_PWRCTL_PWRSEQ if ARCH_QCOM
>
> in the ATH11K_PCI and ATH12K Kconfig entries *may* be the right thing.
> But again, I'm not actually 100% sure of the hard requirements here.
>
>            Linus

After sleeping on it I really think that it may be better to introduce
a more generic ARCH_HAVE_PCI_PWRCTL symbol so that we don't have to
update the ATH Kconfig entires for every new platform that needs it.
For want of a more fine-grained solution, we would select it from
ARCH_QCOM.

Bart