Re: [PATCH] PCI: Disallow retraining link for Atheros QCA98xx chips on non-Gen1 PCIe bridges

From: "Pali Rohár" <pali@kernel.org>
To: "Krzysztof Wilczyński" <kw@linux.com>
Cc: "Bjorn Helgaas" <bhelgaas@google.com>,
	"Kalle Valo" <kvalo@codeaurora.org>,
	"Toke Høiland-Jørgensen" <toke@redhat.com>,
	"Marek Behún" <kabel@kernel.org>,
	vtolkm@gmail.com, "Rob Herring" <robh@kernel.org>,
	"Ilias Apalodimas" <ilias.apalodimas@linaro.org>,
	"Thomas Petazzoni" <thomas.petazzoni@bootlin.com>,
	"Jason Cooper" <jason@lakedaemon.net>,
	linux-pci@vger.kernel.org, ath10k@lists.infradead.org,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH] PCI: Disallow retraining link for Atheros QCA98xx chips on non-Gen1 PCIe bridges
Date: Sat, 27 Mar 2021 14:29:59 +0100	[thread overview]
Message-ID: <20210327132959.fwkphna7gg57aove@pali> (raw)
In-Reply-To: <YF540gjh156QIirA@rocinante>

Hello!

On Saturday 27 March 2021 01:14:10 Krzysztof Wilczyński wrote:
> Hi Pali,
> 
> Thank you for sending the patch over!
> 
> [...]
> > +static int pcie_change_tls_to_gen1(struct pci_dev *parent)
> 
> Just a nitpick, so feel free to ignore it.  I would just call the
> variable "dev" as we pass a pointer to a particular device, but it does
> not matter as much, so I am leaving this to you.

I called it 'parent' because it is called 'parent' also in caller
function. Link consists of two devices, so 'dev' could be ambiguous.

> [...]
> > +	if (ret == 0) {
> 
> You prefer this style over "if (!ret)"?  Just asking in the view of the
> style that seem to be preferred in the code base at the moment.

I can change this to 'if (!ret)' if needed, no problem.

I use 'if (!val)' mostly for boolean and pointer variables. If variable
can contain more integer values then I lot of times I use '=='.

> > +		/* Verify that new value was really set */
> > +		pcie_capability_read_word(parent, PCI_EXP_LNKCTL2, &reg16);
> > +		if ((reg16 & PCI_EXP_LNKCTL2_TLS) != PCI_EXP_LNKCTL2_TLS_2_5GT)
> > +			ret = -EINVAL;
> 
> I am wondering about this verification - did you have a case where the
> device would not properly set its capability, or accept the write and do
> nothing?

I do not know any real device which is doing this thing.

But this issue can happen with kernel's PCIe emulated root bridge:
drivers/pci/pci-bridge-emul.c

Drivers which are using this emulated root bridge (and both pci-mvebu.c
and pci-aardvark.c are using it!) do not have to implement callback for
every read and every write operation of every register.

Note that both pci-mvebu.c and pci-aardvark.c currently does not
implement access to HW register PCI_EXP_LNKCTL2. So currently it is not
possible to set PCI_EXP_LNKCTL2_TLS_2_5GT. And above code correctly
fails and disallow ASPM code to retrain link.

I have some WIP patches which implement LNKCAP2, LNKCTL2 and LNKSTA2
read/write callbacks on emulated bridge for pci-mvebu.c and
pci-aardvark.c. And I have tested that with those WIP patches ASPM code
can correctly switch link to 2.5GT/s and enable ASPM.

> > +	if (ret != 0)
> 
> I think "if (ret)" would be fine to use here, unless you prefer being
> more explicit.  See my question about style above.
> 
> >  static bool pcie_retrain_link(struct pcie_link_state *link)
> >  {
> >  	struct pci_dev *parent = link->pdev;
> >  	unsigned long end_jiffies;
> >  	u16 reg16;
> > +	u32 reg32;
> > +
> > +		/* Check if link is capable of higher speed than 2.5 GT/s and needs quirk */
> > +		pcie_capability_read_dword(parent, PCI_EXP_LNKCAP, &reg32);
> > +		if ((reg32 & PCI_EXP_LNKCAP_SLS) > PCI_EXP_LNKCAP_SLS_2_5GB) {
> 
> I wonder if moving this check to pcie_change_tls_to_gen1() would make
> more sense?  It would then make this function a little cleaner.  What do
> you think?

Ok, no problem. But then function needs to be renamed. Any idea how
should be this function called?

> [...]
> > +static void quirk_no_bus_reset_and_no_retrain_link(struct pci_dev *dev)
> > +{
> > +	dev->dev_flags |= PCI_DEV_FLAGS_NO_BUS_RESET | PCI_DEV_FLAGS_NO_RETRAIN_LINK_WHEN_NOT_GEN1;
> > +}
> [...]
> 
> I know that the style has been changed to allow 100 characters width and
> that checkpatch.pl now also does not warn about line length, as per
> commit bdc48fa11e46 ("checkpatch/coding-style: deprecate 80-column
> warning"), but I think Bjorn still prefers 80 characters, thus this line
> above might have to be aligned.

Ok! If needed I can reformat patch to 80 characters per line.