From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-ig0-f170.google.com ([209.85.213.170]:64712 "EHLO mail-ig0-f170.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752580AbaJNRPY (ORCPT ); Tue, 14 Oct 2014 13:15:24 -0400 Received: by mail-ig0-f170.google.com with SMTP id hn15so18145660igb.1 for ; Tue, 14 Oct 2014 10:15:23 -0700 (PDT) Date: Tue, 14 Oct 2014 11:15:19 -0600 From: Bjorn Helgaas To: matt@masarand.com Cc: linux-pci@vger.kernel.org Subject: Re: [PATCH 01/18] Added way to register deferred PCI IRQ assignment handlers Message-ID: <20141014171519.GA10125@google.com> References: <1412222866-21068-1-git-send-email-matt@masarand.com> <1412222866-21068-2-git-send-email-matt@masarand.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii In-Reply-To: <1412222866-21068-2-git-send-email-matt@masarand.com> Sender: linux-pci-owner@vger.kernel.org List-ID: I like the overall idea here and I think this will be a major improvement. But we need to tweak a few procedural things to integrate it cleanly. On Thu, Oct 02, 2014 at 05:07:29AM +0100, matt@masarand.com wrote: > From: Matthew Minter Please run "git log --oneline --no-merges drivers/pci" and follow the style there for subject lines, i.e., PCI: Add ... For the patches that touch things outside drivers/pci, e.g., the architecture patches, run git log on those as well and try to do something consistent with previous patch descriptions. > These are the main infrastructure changes to allow the registration of > PCI IRQ assignment functions which can then be called during the device > enabling stage. This replaces the pre-initialisation of PCI IRQs at boot > time which caused limitations such as failing for devices which are > hot-plugged after the machine has booted and resulted in highly > fragmented PCI initialisation paths. > > The pdev_assign_irq function becomes a centralised method of assigning > IRQs to PCI devices in a non platform specific way. The pci_host_bridge > structure has had some function pointers added to it where the boot code > can register the platform specific ways of assigning PCI IRQs to be used > to allow them to be used by pdev_assign_irq. > > Some small adjustements have also been made to makefiles in order to > accomodate these changes. > > Note this code completely obsolites the pci_fixup_irqs code path which > has thus been removed. > > Signed-off-by: Matthew Minter > > --- > drivers/of/of_pci_irq.c | 2 +- > drivers/pci/Makefile | 15 ++------------- > drivers/pci/host-bridge.c | 2 +- > drivers/pci/pci.c | 5 ++++- > drivers/pci/pci.h | 1 + > drivers/pci/probe.c | 12 ------------ > drivers/pci/setup-irq.c | 34 ++++++++++++++++------------------ > include/linux/pci.h | 6 ++++-- > 8 files changed, 29 insertions(+), 48 deletions(-) There's a lot going on in this first patch. I'd like to split it up so it's easier to see where the critical changes are. For example, I think it is safe to build setup-irq.o for all arches, even if they don't actually use it, so the makefile changes could be split out. Similarly for the change to make find_pci_host_bridge() non-static. > diff --git a/drivers/of/of_pci_irq.c b/drivers/of/of_pci_irq.c > index 1710d9d..205eb7a 100644 > --- a/drivers/of/of_pci_irq.c > +++ b/drivers/of/of_pci_irq.c > @@ -97,7 +97,7 @@ EXPORT_SYMBOL_GPL(of_irq_parse_pci); > * @pin: PCI irq pin number; passed when used as map_irq callback. Unused > * > * @slot and @pin are unused, but included in the function so that this > - * function can be used directly as the map_irq callback to pci_fixup_irqs(). > + * function can be used directly as the map_irq callback to pdev_assign_irq(). > */ > int of_irq_parse_and_map_pci(const struct pci_dev *dev, u8 slot, u8 pin) > { > diff --git a/drivers/pci/Makefile b/drivers/pci/Makefile > index e04fe2d..38c4cb0 100644 > --- a/drivers/pci/Makefile > +++ b/drivers/pci/Makefile > @@ -4,7 +4,8 @@ > > obj-y += access.o bus.o probe.o host-bridge.o remove.o pci.o \ > pci-driver.o search.o pci-sysfs.o rom.o setup-res.o \ > - irq.o vpd.o setup-bus.o vc.o > + irq.o vpd.o setup-bus.o vc.o setup-irq.o > + > obj-$(CONFIG_PROC_FS) += proc.o > obj-$(CONFIG_SYSFS) += slot.o > > @@ -31,18 +32,6 @@ obj-$(CONFIG_PCI_ATS) += ats.o > obj-$(CONFIG_PCI_IOV) += iov.o > > # > -# Some architectures use the generic PCI setup functions > -# > -obj-$(CONFIG_ALPHA) += setup-irq.o > -obj-$(CONFIG_ARM) += setup-irq.o > -obj-$(CONFIG_UNICORE32) += setup-irq.o > -obj-$(CONFIG_SUPERH) += setup-irq.o > -obj-$(CONFIG_MIPS) += setup-irq.o > -obj-$(CONFIG_TILE) += setup-irq.o > -obj-$(CONFIG_SPARC_LEON) += setup-irq.o > -obj-$(CONFIG_M68K) += setup-irq.o > - > -# > # ACPI Related PCI FW Functions > # ACPI _DSM provided firmware instance and string name > # > diff --git a/drivers/pci/host-bridge.c b/drivers/pci/host-bridge.c > index 0e5f3c9..8ed186f 100644 > --- a/drivers/pci/host-bridge.c > +++ b/drivers/pci/host-bridge.c > @@ -16,7 +16,7 @@ static struct pci_bus *find_pci_root_bus(struct pci_bus *bus) > return bus; > } > > -static struct pci_host_bridge *find_pci_host_bridge(struct pci_bus *bus) > +struct pci_host_bridge *find_pci_host_bridge(struct pci_bus *bus) > { > struct pci_bus *root_bus = find_pci_root_bus(bus); > > diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c > index 2c9ac70..2dd28d9 100644 > --- a/drivers/pci/pci.c > +++ b/drivers/pci/pci.c > @@ -1192,11 +1192,15 @@ static int do_pci_enable_device(struct pci_dev *dev, int bars) > struct pci_dev *bridge; > u16 cmd; > u8 pin; > + struct pci_host_bridge *hbrg = find_pci_host_bridge(dev->bus); > > err = pci_set_power_state(dev, PCI_D0); > if (err < 0 && err != -EIO) > return err; > > + pdev_assign_irq(dev, hbrg->swizzle_irq, hbrg->map_irq); We don't need the host bridge for anything else here, so why not just look it up inside pdev_assign_irq()? > + pci_read_config_byte(dev, PCI_INTERRUPT_PIN, &pin); > + > bridge = pci_upstream_bridge(dev); > if (bridge) > pcie_aspm_powersave_config_link(bridge); > @@ -1209,7 +1213,6 @@ static int do_pci_enable_device(struct pci_dev *dev, int bars) > if (dev->msi_enabled || dev->msix_enabled) > return 0; > > - pci_read_config_byte(dev, PCI_INTERRUPT_PIN, &pin); > if (pin) { > pci_read_config_word(dev, PCI_COMMAND, &cmd); > if (cmd & PCI_COMMAND_INTX_DISABLE) > diff --git a/drivers/pci/pci.h b/drivers/pci/pci.h > index 0601890..a6cf445 100644 > --- a/drivers/pci/pci.h > +++ b/drivers/pci/pci.h > @@ -70,6 +70,7 @@ void pci_config_pm_runtime_put(struct pci_dev *dev); > void pci_pm_init(struct pci_dev *dev); > void pci_allocate_cap_save_buffers(struct pci_dev *dev); > void pci_free_cap_save_buffers(struct pci_dev *dev); > +struct pci_host_bridge *find_pci_host_bridge(struct pci_bus *bus); > > static inline void pci_wakeup_event(struct pci_dev *dev) > { > diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c > index 4170113..314e9e3 100644 > --- a/drivers/pci/probe.c > +++ b/drivers/pci/probe.c > @@ -1719,18 +1719,6 @@ unsigned int pci_scan_child_bus(struct pci_bus *bus) > } > EXPORT_SYMBOL_GPL(pci_scan_child_bus); > > -/** > - * pcibios_root_bridge_prepare - Platform-specific host bridge setup. > - * @bridge: Host bridge to set up. > - * > - * Default empty implementation. Replace with an architecture-specific setup > - * routine, if necessary. > - */ > -int __weak pcibios_root_bridge_prepare(struct pci_host_bridge *bridge) > -{ > - return 0; > -} We can't just remove this, because it will cause link failures for any arch that doesn't supply its own implementation of pcibios_root_bridge_prepare(). Future patches add such implementations, but the tree should build and work correctly after every patch. That is required for bisectability. If your intent is that every arch must supply its own pcibios_root_bridge_prepare(), you should - keep the __weak version here, - add a non-weak version for every arch, - then finally remove this __weak default version at the end. > - > void __weak pcibios_add_bus(struct pci_bus *bus) > { > } > diff --git a/drivers/pci/setup-irq.c b/drivers/pci/setup-irq.c > index 4e2d595..203bf7e 100644 > --- a/drivers/pci/setup-irq.c > +++ b/drivers/pci/setup-irq.c > @@ -22,13 +22,19 @@ void __weak pcibios_update_irq(struct pci_dev *dev, int irq) > pci_write_config_byte(dev, PCI_INTERRUPT_LINE, irq); > } > > -static void pdev_fixup_irq(struct pci_dev *dev, > +void pdev_assign_irq(struct pci_dev *dev, > u8 (*swizzle)(struct pci_dev *, u8 *), > - int (*map_irq)(const struct pci_dev *, u8, u8)) > + int (*map_irq)(struct pci_dev *, u8, u8)) > { > - u8 pin, slot; > + u8 pin; > + u8 slot = -1; > int irq = 0; > > + if (!map_irq) { > + dev_dbg(&dev->dev, "runtime irq mapping not provided by arch\n"); > + return; > + } > + > /* If this device is not on the primary bus, we need to figure out > which interrupt pin it will come in on. We know which slot it > will come in on 'cos that slot is where the bridge is. Each > @@ -40,28 +46,20 @@ static void pdev_fixup_irq(struct pci_dev *dev, > if (pin > 4) > pin = 1; > > - if (pin != 0) { > - /* Follow the chain of bridges, swizzling as we go. */ > - slot = (*swizzle)(dev, &pin); > + if (pin) { > + /* Follow the chain of bridges, swizzling as we go. */ > + if(swizzle) > + slot = (*swizzle)(dev, &pin); > > + /* If a swizzling function is not used map_irq must ignore slot */ > irq = (*map_irq)(dev, slot, pin); > if (irq == -1) > irq = 0; > } > - dev->irq = irq; > - > - dev_dbg(&dev->dev, "fixup irq: got %d\n", dev->irq); > > + dev_dbg(&dev->dev, "assign irq: got %d\n", dev->irq); > + dev->irq = irq; > /* Always tell the device, so the driver knows what is > the real IRQ to use; the device does not use it. */ > pcibios_update_irq(dev, irq); > } > - > -void pci_fixup_irqs(u8 (*swizzle)(struct pci_dev *, u8 *), > - int (*map_irq)(const struct pci_dev *, u8, u8)) > -{ > - struct pci_dev *dev = NULL; > - > - for_each_pci_dev(dev) > - pdev_fixup_irq(dev, swizzle, map_irq); > -} This removes pci_fixup_irqs(), which is still used (at this point) by many architectures. I see that you remove those uses in later patches, but that means the tree is broken in the interim, so it's not bisectable. I think you're going to have to build up the new assign-IRQ-at-enable-time path while leaving the old path intact, in such a way that the new path does nothing until an arch is converted to use it. Then when you convert each arch to the new strategy, you can remove its use of the old path. After all arches are converted and there are no users of the old path, you can remove it completely. > diff --git a/include/linux/pci.h b/include/linux/pci.h > index 96453f9..5426d11 100644 > --- a/include/linux/pci.h > +++ b/include/linux/pci.h > @@ -403,6 +403,8 @@ struct pci_host_bridge { > struct device dev; > struct pci_bus *bus; /* root bus */ > struct list_head windows; /* pci_host_bridge_windows */ > + u8 (*swizzle_irq)(struct pci_dev *, u8 *); /* platform irq swizzler */ > + int (*map_irq)(struct pci_dev *, u8, u8); > void (*release_fn)(struct pci_host_bridge *); > void *release_data; > }; > @@ -1064,8 +1066,8 @@ void pci_assign_unassigned_bus_resources(struct pci_bus *bus); > void pci_assign_unassigned_root_bus_resources(struct pci_bus *bus); > void pdev_enable_device(struct pci_dev *); > int pci_enable_resources(struct pci_dev *, int mask); > -void pci_fixup_irqs(u8 (*)(struct pci_dev *, u8 *), > - int (*)(const struct pci_dev *, u8, u8)); > +void pdev_assign_irq(struct pci_dev *dev, u8 (*swizzle)(struct pci_dev *, u8 *), For things exported via include/linux/pci.h, I think we should use names prefixed by "pci_", not "pdev_". > + int (*map_irq)(struct pci_dev *, u8, u8)); > #define HAVE_PCI_REQ_REGIONS 2 > int __must_check pci_request_regions(struct pci_dev *, const char *); > int __must_check pci_request_regions_exclusive(struct pci_dev *, const char *); Bjorn