Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Wed, 31 May 2023 15:24:46 -0500
From:   Bjorn Helgaas <helgaas@kernel.org>
To:     Vladimir Oltean <vladimir.oltean@nxp.com>
Cc:     linux-pci@vger.kernel.org, netdev@vger.kernel.org,
        Bjorn Helgaas <bhelgaas@google.com>,
        Rob Herring <robh@kernel.org>,
        Claudiu Manoil <claudiu.manoil@nxp.com>,
        Michael Walle <michael@walle.cc>, linux-kernel@vger.kernel.org,
        Liu Peibao <liupeibao@loongson.cn>,
        Binbin Zhou <zhoubinbin@loongson.cn>,
        Huacai Chen <chenhuacai@loongson.cn>
Subject: Re: [PATCH pci] PCI: don't skip probing entire device if first fn OF
 node has status = "disabled"
Message-ID: <ZHetDo5PozWdtrxP@bhelgaas>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20230531165819.phx7uwlgtvnt3tvb@skbuf>
Precedence: bulk

[+cc Loongson folks, thread at
https://lore.kernel.org/r/20230521115141.2384444-1-vladimir.oltean@nxp.com]

On Wed, May 31, 2023 at 07:58:19PM +0300, Vladimir Oltean wrote:
> On Wed, May 31, 2023 at 11:56:02AM -0500, Bjorn Helgaas wrote:
> > What bad things happen without this patch?
> 
> It's in the commit title: probing the entire device (PCI device!!!) is
> skipped if function 0 has status = "disabled". Aka PCIe functions 1, 2, 3, 4, ...

I guess I should have asked "what bad things happen without this patch
and without the DT 'disabled' status"?

I think 6fffbc7ae137 ("PCI: Honor firmware's device disabled status")
was basically a workaround for Loongson making a device visible in PCI
config space when it shouldn't have been [1].

6fffbc7ae137 [2] means we pretend the PCI device doesn't exist if DT
status is "disabled".  If the device happens to be Function 0, that
means we don't look for any more functions.  I guess that doesn't
matter for Loongson.  But it does matter for this NXP platform, where
we don't want to use Function 0, but we *do* want to use other
Functions.

There are several PCIe things that are required to be in Function 0
(MPS, ASPM, IDE, CMA/SPDM, etc), at least in certain cases.

What would happen if instead of making pci_setup_device() fail (as
both 6fffbc7ae137 and this patch do, which means the device doesn't
even show up in "lspci"), we just prevent drivers from binding to it,
e.g., by making pci_device_probe() fail?  The device would then appear
in "lspci" and the PCI core would configure things as usual, but no
drivers would be able to claim it.

Bjorn

[1] https://lore.kernel.org/all/20221114074346.23008-1-liupeibao@loongson.cn/
[2] https://git.kernel.org/linus/6fffbc7ae137