| Message ID | 1374823418-1550-1-git-send-email-Chunhe.Lan@freescale.com (mailing list archive) |
|---|---|
| State | New, archived |
| Delegated to: | Bjorn Helgaas |
| Headers | show |
[+cc Ben, Paul, Minghuan, Kumar, Anton, linuxppc-dev On Fri, Jul 26, 2013 at 1:23 AM, Chunhe Lan <Chunhe.Lan@freescale.com> wrote: > Freescale platform has class code = 0x0b2000, when it boots. This makes > kernel PCI bus code to setup these devices resulting into the following > notice information when trying to enable them: > > pci 0000:00:00.0: ignoring class 0x0b2000 (doesn't match header type 01) > > This patch adds a ID specific(PCI_VENDOR_ID_FREESCALE & PCI_ANY_ID based) > 'early' fixup quirk to replace class code with PCI_CLASS_BRIDGE_PCI as > class. Then the above information disappears. > > Signed-off-by: Chunhe Lan <Chunhe.Lan@freescale.com> > Cc: Bjorn Helgaas <bhelgaas@google.com> > --- > drivers/pci/quirks.c | 10 ++++++++++ > 1 files changed, 10 insertions(+), 0 deletions(-) > > diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c > index e85d230..906a04a3 100644 > --- a/drivers/pci/quirks.c > +++ b/drivers/pci/quirks.c > @@ -2772,6 +2772,16 @@ DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_INTEL, 0x342e, vtd_mask_spec_errors); > DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_INTEL, 0x3c28, vtd_mask_spec_errors); > #endif > > +#ifdef CONFIG_FSL_PCI > +static void quirk_freescale_class(struct pci_dev *dev) > +{ > + dev_info(&dev->dev, "Setting PCI class for FSL PCI host bridge\n"); > + dev->class = (PCI_CLASS_BRIDGE_PCI << 8) | (dev->class & 0xff); > +} > +DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_FREESCALE, PCI_ANY_ID, > + quirk_freescale_class); 1. This looks an awful lot like quirk_fsl_pcie_head(). quirk_fsl_pcie_head() is currently a HEADER quirk, but that runs *after* pci_setup_device(), where dev->class is used. I suspect this should be an EARLY quirk instead. But this quirk has a long history as a HEADER quirk, and I'm sure many people have tested it, so I don't know what's going on. 2. I assume the Freescale vendor ID might be used for devices other than bridges, so I doubt it would be correct to overwrite the class for *every* Freescale device as you're doing here. > +#endif > + > static void fixup_ti816x_class(struct pci_dev *dev) > { > /* TI 816x devices do not have class code set when in PCIe boot mode */ > -- > 1.7.6.5 > > -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
[-cc Anton (bounced)] On Fri, Jul 26, 2013 at 7:07 AM, Bjorn Helgaas <bhelgaas@google.com> wrote: > [+cc Ben, Paul, Minghuan, Kumar, Anton, linuxppc-dev > > On Fri, Jul 26, 2013 at 1:23 AM, Chunhe Lan <Chunhe.Lan@freescale.com> wrote: >> Freescale platform has class code = 0x0b2000, when it boots. This makes >> kernel PCI bus code to setup these devices resulting into the following >> notice information when trying to enable them: >> >> pci 0000:00:00.0: ignoring class 0x0b2000 (doesn't match header type 01) >> >> This patch adds a ID specific(PCI_VENDOR_ID_FREESCALE & PCI_ANY_ID based) >> 'early' fixup quirk to replace class code with PCI_CLASS_BRIDGE_PCI as >> class. Then the above information disappears. >> >> Signed-off-by: Chunhe Lan <Chunhe.Lan@freescale.com> >> Cc: Bjorn Helgaas <bhelgaas@google.com> >> --- >> drivers/pci/quirks.c | 10 ++++++++++ >> 1 files changed, 10 insertions(+), 0 deletions(-) >> >> diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c >> index e85d230..906a04a3 100644 >> --- a/drivers/pci/quirks.c >> +++ b/drivers/pci/quirks.c >> @@ -2772,6 +2772,16 @@ DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_INTEL, 0x342e, vtd_mask_spec_errors); >> DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_INTEL, 0x3c28, vtd_mask_spec_errors); >> #endif >> >> +#ifdef CONFIG_FSL_PCI >> +static void quirk_freescale_class(struct pci_dev *dev) >> +{ >> + dev_info(&dev->dev, "Setting PCI class for FSL PCI host bridge\n"); >> + dev->class = (PCI_CLASS_BRIDGE_PCI << 8) | (dev->class & 0xff); >> +} >> +DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_FREESCALE, PCI_ANY_ID, >> + quirk_freescale_class); > > 1. This looks an awful lot like quirk_fsl_pcie_head(). > quirk_fsl_pcie_head() is currently a HEADER quirk, but that runs > *after* pci_setup_device(), where dev->class is used. I suspect this > should be an EARLY quirk instead. But this quirk has a long history > as a HEADER quirk, and I'm sure many people have tested it, so I don't > know what's going on. > > 2. I assume the Freescale vendor ID might be used for devices other > than bridges, so I doubt it would be correct to overwrite the class > for *every* Freescale device as you're doing here. > >> +#endif >> + >> static void fixup_ti816x_class(struct pci_dev *dev) >> { >> /* TI 816x devices do not have class code set when in PCIe boot mode */ >> -- >> 1.7.6.5 >> >> -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Fri, 2013-07-26 at 07:07 -0600, Bjorn Helgaas wrote: > > 1. This looks an awful lot like quirk_fsl_pcie_head(). > quirk_fsl_pcie_head() is currently a HEADER quirk, but that runs > *after* pci_setup_device(), where dev->class is used. I suspect this > should be an EARLY quirk instead. But this quirk has a long history > as a HEADER quirk, and I'm sure many people have tested it, so I don't > know what's going on. > > 2. I assume the Freescale vendor ID might be used for devices other > than bridges, so I doubt it would be correct to overwrite the class > for *every* Freescale device as you're doing here. That wouldn't be the first time we need a quirk like that, it looks like HW folks love playing with the class codes in stupid ways. I remember once trying to change the FSP header quirk to an early quirk and breaking something else doing so ... At some point I ran out of time debugging that stuff and put it on my TODO list ... you know the file you open in O_APPEND mode :-) That was a year ago or so... So yes, something is fishy here and in need of a better fix, and at this stage I'd say FSL folks should sort it out. These things need to be early quirks and if that breaks something we need to know what and why. Also we tend to have them in arch/powerpc next to the relevant driver code... that quirks.c file is just a mess and certainly not the right place for something like that. Cheers, Ben. -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c index e85d230..906a04a3 100644 --- a/drivers/pci/quirks.c +++ b/drivers/pci/quirks.c @@ -2772,6 +2772,16 @@ DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_INTEL, 0x342e, vtd_mask_spec_errors); DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_INTEL, 0x3c28, vtd_mask_spec_errors); #endif +#ifdef CONFIG_FSL_PCI +static void quirk_freescale_class(struct pci_dev *dev) +{ + dev_info(&dev->dev, "Setting PCI class for FSL PCI host bridge\n"); + dev->class = (PCI_CLASS_BRIDGE_PCI << 8) | (dev->class & 0xff); +} +DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_FREESCALE, PCI_ANY_ID, + quirk_freescale_class); +#endif + static void fixup_ti816x_class(struct pci_dev *dev) { /* TI 816x devices do not have class code set when in PCIe boot mode */
Freescale platform has class code = 0x0b2000, when it boots. This makes kernel PCI bus code to setup these devices resulting into the following notice information when trying to enable them: pci 0000:00:00.0: ignoring class 0x0b2000 (doesn't match header type 01) This patch adds a ID specific(PCI_VENDOR_ID_FREESCALE & PCI_ANY_ID based) 'early' fixup quirk to replace class code with PCI_CLASS_BRIDGE_PCI as class. Then the above information disappears. Signed-off-by: Chunhe Lan <Chunhe.Lan@freescale.com> Cc: Bjorn Helgaas <bhelgaas@google.com> --- drivers/pci/quirks.c | 10 ++++++++++ 1 files changed, 10 insertions(+), 0 deletions(-)