Re: [PATCH] PCI: qcom: Enable MSI interrupts together with Link up if global IRQ is supported

[Manivannan Sadhasivam]

On Mon, Sep 30, 2024 at 12:11:01PM -0500, Bjorn Helgaas wrote:
> On Mon, Sep 30, 2024 at 07:14:09PM +0530, Manivannan Sadhasivam wrote:
> > Currently, if global IRQ is supported by the platform, only the Link up
> > interrupt is enabled in the PARF_INT_ALL_MASK register. But on some Qcom
> > platforms like SM8250, and X1E80100, MSIs are getting masked due to this.
> > They require enabling the MSI interrupt bits in the register to unmask
> > (enable) the MSIs.
> 
> "global IRQ" is a very generic name.  If that's the official name, it
> should at least be capitalized, e.g., "Global IRQ", to show that it is
> a proper noun that refers to a specific IRQ.
> 

Sure.

> > Even though the MSI interrupt enable bits in PARF_INT_ALL_MASK are
> > described as 'diagnostic' interrupts in the internal documentation,
> > disabling them masks MSI on these platforms. Due to this,
> 
> > MSIs were not
> > reported to be received these platforms while supporting global IRQ.
> 
> I'm trying to parse "while supporting global IRQ."  We basically
> support global IRQ by installing qcom_pcie_global_irq_thread(), but of
> course the device doesn't see that, so I assume it would be more
> informative to say that MSIs are masked by some register setting.
> 

Hmm, this is what I mentioned in the above paragraph referencing
PARF_INT_ALL_MASK register. Is that not clear enough?

> The patch suggests that MSIs are masked internally unless
> PARF_INT_MSI_DEV_0_7 is set in PARF_INT_ALL_MASK.
> 
> Are you saying that prior to 4581403f6792, MSIs did work?  Does that
> mean PARF_INT_MSI_DEV_0_7 was set by a bootloader or something, so
> MSIs worked?  And then 4581403f6792 came along and implicitly cleared
> PARF_INT_MSI_DEV_0_7, so MSIs were then masked?
> 

Yeah. Those bits were enabled by default in hardware, but since they were
mentioned as 'diagnostic interrupts' in documentation, commit 4581403f6792
intentionally disabled them. But that results in MSIs getting masked in
*some* platforms.

- Mani

> > So enable the MSI interrupts along with the Link up interrupt in the
> > PARF_INT_ALL_MASK register if global IRQ is supported. This ensures that
> > the MSIs continue to work and also the driver is able to catch the Link
> > up interrupt for enumerating endpoint devices.
> > 
> > Fixes: 4581403f6792 ("PCI: qcom: Enumerate endpoints based on Link up event in 'global_irq' interrupt")
> > Reported-by: Konrad Dybcio <konradybcio@xxxxxxxxxx>
> > Closes: https://lore.kernel.org/linux-pci/9a692c98-eb0a-4d86-b642-ea655981ff53@xxxxxxxxxx/
> > Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@xxxxxxxxxx>
> > ---
> >  drivers/pci/controller/dwc/pcie-qcom.c | 4 +++-
> >  1 file changed, 3 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/pci/controller/dwc/pcie-qcom.c b/drivers/pci/controller/dwc/pcie-qcom.c
> > index ef44a82be058..2b33d03ed054 100644
> > --- a/drivers/pci/controller/dwc/pcie-qcom.c
> > +++ b/drivers/pci/controller/dwc/pcie-qcom.c
> > @@ -133,6 +133,7 @@
> >  
> >  /* PARF_INT_ALL_{STATUS/CLEAR/MASK} register fields */
> >  #define PARF_INT_ALL_LINK_UP			BIT(13)
> > +#define PARF_INT_MSI_DEV_0_7			GENMASK(30, 23)
> >  
> >  /* PARF_NO_SNOOP_OVERIDE register fields */
> >  #define WR_NO_SNOOP_OVERIDE_EN			BIT(1)
> > @@ -1716,7 +1717,8 @@ static int qcom_pcie_probe(struct platform_device *pdev)
> >  			goto err_host_deinit;
> >  		}
> >  
> > -		writel_relaxed(PARF_INT_ALL_LINK_UP, pcie->parf + PARF_INT_ALL_MASK);
> > +		writel_relaxed(PARF_INT_ALL_LINK_UP | PARF_INT_MSI_DEV_0_7,
> > +			       pcie->parf + PARF_INT_ALL_MASK);
> >  	}
> >  
> >  	qcom_pcie_icc_opp_update(pcie);
> > -- 
> > 2.25.1
> > 

-- 
மணிவண்ணன் சதாசிவம்