On Wed, Sep 21, 2016 at 06:51:55AM +0000, Po Liu wrote:

Hi Bjorn,

quoted

 -----Original Message-----
 From: Bjorn Helgaas [mailto:helgaas@kernel.org]
 Sent: Wednesday, September 21, 2016 4:47 AM
 To: Po Liu
 Cc: Roy Zang; Arnd Bergmann; devicetree@vger.kernel.org; Marc Zyngier;
 linux-pci@vger.kernel.org; linux-kernel@vger.kernel.org; Stuart Yoder;
 M.H. Lian; Murali Karicheri; Mingkai Hu; Bjorn Helgaas; Leo Li; Shawn
 Guo; linux-arm-kernel@lists.infradead.org
 Subject: Re: [PATCH v3 2/2] pci/aer: interrupt fixup in the quirk
 
 On Mon, Aug 22, 2016 at 10:09:18AM +0000, Po Liu wrote:
 > Hi Bjorn,
 >
 > Sorry for late reply.
 >
 > I checked the updated kernel with Dongdong mentioned ACPI patch which
 was truly affected my quirk patch uploaded. So I suppose the quirk patch
 is not qualify to fix the bug.
 
 I don't understand what you're saying here.
 
 The quirk worked on your machine.  It apparently didn't work on
 Dongdong's machine because of_irq_parse_and_map_pci() is run after the
 quirk in this path:
 
   pci_device_probe
     pcibios_alloc_irq                 # arm64
       dev->irq = of_irq_parse_and_map_pci
 
 and of_irq_parse_and_map_pci() returned zero, probably because
 of_irq_parse_pci() failed.  My guess is that the reason it works on your
 machine but not Dongdong's is that your DTs are different such that
 of_irq_parse_pci() works for you but not for Dongdong.
 
 I think the idea of of_irq_parse_and_map_pci() is to set up a device's
 INTx line.  But that doesn't quite apply here because your device
 doesn't actually *use* INTx.  So I don't know why of_irq_parse_pci()
 works for you.  Maybe that's a symptom of a problem in your DT.
 
 Or maybe you're saying that the quirk *didn't* work on your machine when
 you tested it in a kernel that included d8ed75d59332 ("ARM64:
 PCI: ACPI support for legacy IRQs parsing and consolidation with DT
 code"). 

Yes, this point is what I mean. After this patch my quirk patch would not work. 
Since I discussed with Dongdong, the patches d8ed75d59332 ACPI related were not be merged yet.

quoted

 But that doesn't make sense either, because prior to
 d8ed75d59332, we *always* set
 
   dev->irq = of_irq_parse_and_map_pci(dev, 0, 0);
 
 and after the patch we only do it if "acpi_disabled".  I guess I just
 don't understand what you're saying.

Before the patch merged. pcibios_add_device()(which run the ->irq =
of_irq_parse_and_map_pci(dev, 0, 0);) was loaded before the
pci_fixup_device(pci_fixup_final).  But after the patch
d8ed75d59332("ARM64: PCI: ACPI support for legacy IRQs parsing and
consolidation with DT code") merged, the
pci_fixup_device(pci_fixup_final) run BEFORE the
pcibios_alloc_irq()(which run the ->irq =
of_irq_parse_and_map_pci(dev, 0, 0);). So the dev->irq were
overwhelm by the pcibios_alloc_irq().

OK.  Prior to d8ed75d59332, arm64 overrode the default empty
pcibios_add_device() implementation, and called
of_irq_parse_and_map_pci() there.  d8ed75d59332 changed that function
to pcibios_alloc_irq(), which is called later, in the driver probe
path.

When I test, the acpi_disabled is '1' although my kernel config
default is CONFIG_ACPI=y. And no setting in the uboot with apci=xxx.
But this is another issue, I didn't deep to check it. 

Likely your platform just doesn't have ACPI or something's wrong in
the initial ACPI setup.

quoted

 > I were keep thinking what your "explicitly checking for a root port
 device" meaning. Do you mean I should upload again the first version
 patch which fix it in the portdrv_core.c ? I would upload again if yes.
 
 No, I did not mean you should go back to the first version of the patch.
 If we *can* do this in a quirk, I think that would be much better than
 doing it in the PCIe port driver.  I meant that Dongdong's suggestion of
 adding this:
 
   if (pci_pcie_type(dev) != PCI_EXP_TYPE_ROOT_PORT)
     return;
 
 to your quirk made sense to me.

If the quirk patch could make workaround. It should be the better way.

It doesn't sound like a quirk is going to work because all the quirks
run too early.  I'll respond to the patch itself with more ideas.

quoted

 > >  -----Original Message-----
 > >  From: Bjorn Helgaas [mailto:helgaas@kernel.org]
 > >  Sent: Saturday, July 30, 2016 6:42 AM
 > >  To: Po Liu
 > >  Cc: linux-pci@vger.kernel.org;
 > > linux-arm-kernel@lists.infradead.org;
 > >  linux-kernel@vger.kernel.org; devicetree@vger.kernel.org; Roy Zang;
 > > Arnd  Bergmann; Marc Zyngier; Stuart Yoder; Yang-Leo Li; Minghuan
 > > Lian; Murali  Karicheri; Bjorn Helgaas; Shawn Guo; Mingkai Hu
 > >  Subject: Re: [PATCH v3 2/2] pci/aer: interrupt fixup in the quirk
 > >
 > >  On Tue, Jun 14, 2016 at 04:24:05PM +0800, Po Liu wrote:
 > >  > On some platforms, root port doesn't support MSI/MSI-X/INTx in RC
 mode.
 > >  > When chip support the aer interrupt with none MSI/MSI-X/INTx
 > > mode,  > maybe there is interrupt line for aer pme etc. Search the
 > > interrupt  > number in the fdt file. Then fixup the dev->irq with it.
 > >  >
 > >  > Signed-off-by: Po Liu [off-list ref]
 > >
 > >  I'm not sure where we're at with this.  Dongdong had some issue
 > > (possibly with a version of the quirk on a different platform?), and
 > > I  think the suggestion of explicitly checking for a root port
 > > device was a  good one.
 > >
 > >  So please update and repost this for next cycle.
 > >
 > >  > ---
 > >  > changes for V3:
 > >  > 	- Move to quirk;
 > >  > 	- Only correct the irq in RC mode;
 > >  >
 > >  >  drivers/pci/quirks.c | 29 +++++++++++++++++++++++++++++  >  1
 > > file changed, 29 insertions(+)  >  > diff --git
 > > a/drivers/pci/quirks.c b/drivers/pci/quirks.c index  >
 > > ee72ebe..8b39cce 100644  > --- a/drivers/pci/quirks.c  > +++
 > > b/drivers/pci/quirks.c  > @@ -25,6 +25,7 @@  >  #include
 > > <linux/sched.h>  >  #include <linux/ktime.h>  >  #include
 > > <linux/mm.h>  > +#include <linux/of_irq.h>
 > >  >  #include <asm/dma.h>	/* isa_dma_bridge_buggy */
 > >  >  #include "pci.h"
 > >  >
 > >  > @@ -4419,3 +4420,31 @@ static void quirk_intel_qat_vf_cap(struct
 > > pci_dev *pdev)
 > >  >  	}
 > >  >  }
 > >  >  DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_INTEL, 0x443,  >
 > > quirk_intel_qat_vf_cap);  > +  > +/* If root port doesn't support
 > > MSI/MSI-X/INTx in RC mode,  > + * but use standalone irq. Read the
 > > device tree for the aer  > + * interrupt number.
 > >  > + */
 > >  > +static void quirk_aer_interrupt(struct pci_dev *dev) {
 > >  > +	int ret;
 > >  > +	u8 header_type;
 > >  > +	struct device_node *np = NULL;
 > >  > +
 > >  > +	/* Only for the RC mode device */
 > >  > +	pci_read_config_byte(dev, PCI_HEADER_TYPE, &header_type);
 > >  > +	if ((header_type & 0x7F) != PCI_HEADER_TYPE_BRIDGE)
 > >  > +		return;
 > >  > +
 > >  > +	if (dev->bus->dev.of_node)
 > >  > +		np = dev->bus->dev.of_node;
 > >  > +
 > >  > +	if (IS_ENABLED(CONFIG_OF_IRQ) && np) {
 > >  > +		ret = of_irq_get_byname(np, "aer");
 > >  > +		if (ret > 0) {
 > >  > +			dev->no_msi = 1;
 > >  > +			dev->irq = ret;
 > >  > +		}
 > >  > +	}
 > >  > +}
 > >  > +DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_FREESCALE, PCI_ANY_ID,  >
 > > +quirk_aer_interrupt);  > --  > 2.1.0.27.g96db324  >  >  >
 > > _______________________________________________
 > >  > linux-arm-kernel mailing list
 > >  > linux-arm-kernel@lists.infradead.org
 > >  > http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
 > _______________________________________________
 > linux-arm-kernel mailing list
 > linux-arm-kernel@lists.infradead.org
 > http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel