Re: [PATCH] PCI: brcmstb: Assign pcie->gen from pcie_get_link_speed()

[Florian Fainelli]

On 5/4/26 10:26, Hans Zhang wrote:
> On 5/5/26 00:58, Florian Fainelli wrote:
> > On 5/2/26 04:40, Bjorn Helgaas wrote:
> > > On Fri, May 01, 2026 at 01:24:38PM -0700, Florian Fainelli wrote:
> > > > After commit 03f920936977 ("PCI: controller: Validate max-link-speed"),
> > > > pcie->gen stopped being assigned and as a result the established PCIe
> > > > link would stop supporting Gen3 speeds on 2712 since pcie->gen is used
> > > > to populate LnkCntl2 and LnkCap in brcm_pcie_set_gen().
> > > >
> > > > Link: https://github.com/raspberrypi/linux/issues/7343
> > > > Reported-by: Dom Cobley <popcornmix@xxxxxxxxx>
> > > > Reported-by: Phil Elwell <phil@xxxxxxxxxxxxxxx>
> > > > Fixes: 03f920936977 ("PCI: controller: Validate max-link-speed")
> > > > Signed-off-by: Florian Fainelli <florian.fainelli@xxxxxxxxxxxx>
> > > > ---
> > > >   drivers/pci/controller/pcie-brcmstb.c | 3 +--
> > > >   1 file changed, 1 insertion(+), 2 deletions(-)
> > > >
> > > > diff --git a/drivers/pci/controller/pcie-brcmstb.c b/drivers/pci/ 
> > > > controller/pcie-brcmstb.c
> > > > index 714bcab97b60..6138fc4bc064 100644
> > > > --- a/drivers/pci/controller/pcie-brcmstb.c
> > > > +++ b/drivers/pci/controller/pcie-brcmstb.c
> > > > @@ -2072,8 +2072,7 @@ static int brcm_pcie_probe(struct 
> > > > platform_device *pdev)
> > > >           return PTR_ERR(pcie->clk);
> > > >       ret = of_pci_get_max_link_speed(np);
> > > > -    if (pcie_get_link_speed(ret) == PCI_SPEED_UNKNOWN)
> > > > -        pcie->gen = 0;
> > > > +    pcie->gen = pcie_get_link_speed(ret);
> > >
> > > Take a look at https://sashiko.dev/#/ 
> > > patchset/20260501202438.376033-1- florian.fainelli%40broadcom.com
> > >
> > > The notes at https://github.com/raspberrypi/linux/issues/7343 assumed
> > > PCI_SPEED_UNKNOWN was 0, but in fact it is 0xff, which means you might
> > > want the more defensive patch instead.
> > >
> > > I'll be happy to replace what's on pci/for-linus if so.
> >
> > I am starting to think a revert is the simplest path forward, it's not 
> > clear what pcie_get_link_speed() brings to the table honestly.
>
> Hi Florian,
>
> The pcie_get_link_speed function is designed to prevent other Root Port 
> drivers from accessing the array pcie_link_speed out of bounds.

Yes, so that's useful in the first hunk of your commit where we were 
printing the link speed without checking that the 'max-link-speed' would 
be bounds check, however it is not really useful for assigning to 
pcie->gen in our case, so I would prefer the second option:


          ret = of_pci_get_max_link_speed(np);
 -       if (pcie_get_link_speed(ret) == PCI_SPEED_UNKNOWN)
 -               pcie->gen = 0;
 +       pcie->gen = (ret < 0) ? 0 : ret;

Thank you for stepping in!
--
Florian