| Message ID | 1440715146-16578-2-git-send-email-keith.busch@intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Delegated to: | Bjorn Helgaas |
| Headers | show |
On Thu, 27 Aug 2015, Keith Busch wrote: > This patch adds struct x86_msi_ops to x86's PCI sysdata. This gives a > host bridge driver the option to provide alternate MSI Data Register > and MSI-X Table Entry programming for devices in PCI domains that do > not subscribe to usual "IOAPIC" format. I'm not too fond about more ad hoc indirection and special casing. We should be able to handle this with hierarchical irq domains. Jiang might have an idea how to do that for your case. Thanks, tglx -- 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, 28 Aug 2015, Thomas Gleixner wrote: > On Thu, 27 Aug 2015, Keith Busch wrote: > >> This patch adds struct x86_msi_ops to x86's PCI sysdata. This gives a >> host bridge driver the option to provide alternate MSI Data Register >> and MSI-X Table Entry programming for devices in PCI domains that do >> not subscribe to usual "IOAPIC" format. > > I'm not too fond about more ad hoc indirection and special casing. We > should be able to handle this with hierarchical irq domains. Jiang > might have an idea how to do that for your case. Thank you for the suggestion, I will take a closer look at this again. All the better if we don't require an arch specific dependency. I asked Jiang about domain hierarchies a few weeks ago and understood these are suited for irq controllers, but the VMD device is an aggregator. Here's a little more h/w info in case it helps steer me in the right direction: VMD muxes all the device interrupts in its PCI domain into one of several VMD h/w irqs. The VMD driver then has to de-mux that into CPU side irqs for each device. -- 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 2015/8/29 0:54, Thomas Gleixner wrote: > On Thu, 27 Aug 2015, Keith Busch wrote: > >> This patch adds struct x86_msi_ops to x86's PCI sysdata. This gives a >> host bridge driver the option to provide alternate MSI Data Register >> and MSI-X Table Entry programming for devices in PCI domains that do >> not subscribe to usual "IOAPIC" format. > > I'm not too fond about more ad hoc indirection and special casing. We > should be able to handle this with hierarchical irq domains. Jiang > might have an idea how to do that for your case. Hi Thomas and Keith, I have noticed this patch set yesterday, but still investigating the better way to handle this. Basically I think we should build per-domain/per-bus/per-device PCI MSI irqdomain, just like what ARM have done. That will give us a clear picture. But I need more information about the hardware topology to correctly build up the hierarchical irqdomain, especially the relationship between the embedded host bridge and IOMMU units. Keith, could you please help to provide some doc with hardware details? Thanks! Gerry -- 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/arch/x86/include/asm/pci.h b/arch/x86/include/asm/pci.h index 4625943..98f3802 100644 --- a/arch/x86/include/asm/pci.h +++ b/arch/x86/include/asm/pci.h @@ -20,6 +20,9 @@ struct pci_sysdata { #ifdef CONFIG_X86_64 void *iommu; /* IOMMU private data */ #endif +#ifdef CONFIG_PCI_MSI + struct x86_msi_ops *msi_ops; +#endif }; extern int pci_routeirq; diff --git a/arch/x86/kernel/x86_init.c b/arch/x86/kernel/x86_init.c index 3839628..416de84 100644 --- a/arch/x86/kernel/x86_init.c +++ b/arch/x86/kernel/x86_init.c @@ -118,21 +118,40 @@ struct x86_msi_ops x86_msi = { /* MSI arch specific hooks */ int arch_setup_msi_irqs(struct pci_dev *dev, int nvec, int type) { + struct pci_sysdata *sysdata = dev->bus->sysdata; + + if (sysdata && sysdata->msi_ops && sysdata->msi_ops->setup_msi_irqs) + return sysdata->msi_ops->setup_msi_irqs(dev, nvec, type); return x86_msi.setup_msi_irqs(dev, nvec, type); } void arch_teardown_msi_irqs(struct pci_dev *dev) { + struct pci_sysdata *sysdata = dev->bus->sysdata; + + if (sysdata && sysdata->msi_ops && sysdata->msi_ops->teardown_msi_irqs) + return sysdata->msi_ops->teardown_msi_irqs(dev); x86_msi.teardown_msi_irqs(dev); } void arch_teardown_msi_irq(unsigned int irq) { + struct msi_desc *desc = irq_get_msi_desc(irq); + struct pci_sysdata *sysdata = NULL; + + if (desc) + sysdata = desc->dev->bus->sysdata; + if (sysdata && sysdata->msi_ops && sysdata->msi_ops->teardown_msi_irq) + return sysdata->msi_ops->teardown_msi_irq(irq); x86_msi.teardown_msi_irq(irq); } void arch_restore_msi_irqs(struct pci_dev *dev) { + struct pci_sysdata *sysdata = dev->bus->sysdata; + + if (sysdata && sysdata->msi_ops && sysdata->msi_ops->restore_msi_irqs) + return sysdata->msi_ops->restore_msi_irqs(dev); x86_msi.restore_msi_irqs(dev); } #endif
This patch adds struct x86_msi_ops to x86's PCI sysdata. This gives a host bridge driver the option to provide alternate MSI Data Register and MSI-X Table Entry programming for devices in PCI domains that do not subscribe to usual "IOAPIC" format. Signed-off-by: Keith Busch <keith.busch@intel.com> CC: Bryan Veal <bryan.e.veal@intel.com> CC: Dan Williams <dan.j.williams@intel.com> CC: x86@kernel.org CC: linux-kernel@vger.kernel.org CC: linux-pci@vger.kernel.org --- arch/x86/include/asm/pci.h | 3 +++ arch/x86/kernel/x86_init.c | 19 +++++++++++++++++++ 2 files changed, 22 insertions(+)