Re: [PATCH 1/2] PCI: dwc: ep: Fix MSI-X configuration to write to correct physical function

[Aksh Garg]

On 02/02/26 13:27, Niklas Cassel wrote:
> On Mon, Feb 02, 2026 at 08:55:52AM +0100, Niklas Cassel wrote:
> > On Mon, Feb 02, 2026 at 12:57:57PM +0530, Aksh Garg wrote:
> > > The MSI-X configuration code reads from the correct physical function's
> > > register space using dw_pcie_ep_readw_dbi(), but writes back only to
> > > PF0 using the old dw_pcie_writew_dbi() helper. This causes incorrect
> > > MSI-X configuration for other PFs.
> > > 
> > > Fix this by using dw_pcie_ep_writew_dbi() to write to the correct PF's
> > > register space, matching the read operation.
> > > 
> > > Fixes: 70fa02ca1446 ("PCI: dwc: Add dw_pcie_ep_{read,write}_dbi[2] helpers")
> > > Signed-off-by: Aksh Garg <a-garg7@xxxxxx>
> > > ---
> > >  drivers/pci/controller/dwc/pcie-designware-ep.c | 2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > 
> > > diff --git a/drivers/pci/controller/dwc/pcie-designware-ep.c b/drivers/pci/controller/dwc/pcie-designware-ep.c
> > > index 7e7844ff0f7e..771241e1a2c9 100644
> > > --- a/drivers/pci/controller/dwc/pcie-designware-ep.c
> > > +++ b/drivers/pci/controller/dwc/pcie-designware-ep.c
> > > @@ -745,7 +745,7 @@ static int dw_pcie_ep_set_msix(struct pci_epc *epc, u8 func_no, u8 vfunc_no,
> > >  	val = dw_pcie_ep_readw_dbi(ep, func_no, reg);
> > >  	val &= ~PCI_MSIX_FLAGS_QSIZE;
> > >  	val |= nr_irqs - 1; /* encoded as N-1 */
> > > -	dw_pcie_writew_dbi(pci, reg, val);
> > > +	dw_pcie_ep_writew_dbi(ep, func_no, reg, val);
> > >  
> > >  	reg = ep_func->msix_cap + PCI_MSIX_TABLE;
> > >  	val = offset | bir;
> > > -- 
> > > 2.34.1
> > > 
> >
> > Reviewed-by: Niklas Cassel <cassel@xxxxxxxxxx>
> >
> >
> > However, please also fix the only remaining place in this file which still
> > uses the incorrect dw_pcie_writew_dbi() instead of dw_pcie_ep_writew_dbi()
> >
> > i.e.:
> > dw_pcie_ep_raise_msix_irq_doorbell().
>
> Perhaps doorbell is special, and there is only one register for this...

Yes, we have only one doorbell register.

> But then, perhaps add a comment why this is the only place that needs
> to not supply func_no.

However, I think the offset (0x948) of the register is self-explanatory,
as after 0x700, we have port logic registers, which are per-device, not
per-PF.

>