Re: [PATCH v2] PCI: dw-rockchip: Enable async probe by default

From: Niklas Cassel <cassel@kernel.org>
To: Anand Moon <linux.amoon@gmail.com>
Cc: Grimmauld <grimmauld@grimmauld.de>,
	mani@kernel.org, "Lorenzo Pieralisi" <lpieralisi@kernel.org>,
	"Krzysztof Wilczyński" <kw@linux.com>,
	"Rob Herring" <robh@kernel.org>,
	"Bjorn Helgaas" <bhelgaas@google.com>,
	"Heiko Stuebner" <heiko@sntech.de>,
	linux-pci@vger.kernel.org, linux-arm-kernel@lists.infradead.org,
	linux-rockchip@lists.infradead.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2] PCI: dw-rockchip: Enable async probe by default
Date: Tue, 3 Feb 2026 12:01:41 +0100	[thread overview]
Message-ID: <aYHVlS1nbCMMyF04@ryzen> (raw)
In-Reply-To: <CANAwSgSR7AYCvMXSo5UskTF3LJToOnPV9JE4XBAUzrCbWGKD_w@mail.gmail.com>

On Mon, Feb 02, 2026 at 11:35:48PM +0530, Anand Moon wrote:
> On Mon, 2 Feb 2026 at 15:25, Niklas Cassel <cassel@kernel.org> wrote:
> >
> > On Sat, Jan 31, 2026 at 03:08:42PM +0530, Anand Moon wrote:
> > > >
> > > I’ve attempted to reproduce the warning but was unable to trigger it locally.
> > >
> > > I have tested with the built-in module
> > > CONFIG_R8169=y
> > > CONFIG_R8169_LEDS=y
> > >
> > > As well as the module
> > > CONFIG_R8169=m
> > > CONFIG_R8169_LEDS=y
> > >
> >
> > I'm running with:
> > CONFIG_R8169=y
> > CONFIG_PHYLIB=y
> > CONFIG_REALTEK_PHY=y
> > CONFIG_REALTEK_PHY_HWMON=y
> >
> I feel CONFIG_R8169 should not be built into the kernel image.
> Since the driver is registered via module_pci_driver(rtl8169_pci_driver),
> it is intended to be loaded as a module. In addition, this driver
> requires external firmware
> during initialization, which could make a built‑in configuration problematic.
> Keeping it modular ensures proper firmware loading and avoids
> early‑boot failures.

For what it is worth, I strongly disagree that you should not be allowed
to build something as built-in just because the driver is registered using
module_pci_driver().

In order to use NFS root, and to avoid the hassle of using an initramfs,
having the driver as built-in is a logical option.
Anyway, this is currently working fine with my setup today, and is
unrelated to the phylib splat that I have reported in this thread.

I know that the R8169 driver tries to load a firmware blob, I used to have
an initramfs just to be able to load this firmware blob, but I found out
that the driver probes and works perfectly fine even without the loading
any firmware, therefore I no longer use an initramfs.
Again, this is working perfectly fine today, and the R8169 loading or not
loading any firmware is unrealated to the phylib splat.

(Just for clarity, I don't think that the phylib splat should stop this
patch from being accepted.)

Kind regards,
Niklas