| Message ID | 20170421050500.13957-10-yinghai@kernel.org |
|---|---|
| State | Not Applicable |
| Delegated to: | David Miller |
| Headers | show |
On Thu, Apr 20, 2017 at 10:04:56PM -0700, Yinghai Lu wrote: > We still get "no compatible bridge window" warning on sparc T5-8 > after we add support for 64bit resource parsing for root bus. > > PCI: scan_bus[/pci@300/pci@1/pci@0/pci@6] bus no 8 > PCI: Claiming 0000:00:01.0: Resource 15: 0000800100000000..00008004afffffff [220c] > PCI: Claiming 0000:01:00.0: Resource 15: 0000800100000000..00008004afffffff [220c] > PCI: Claiming 0000:02:04.0: Resource 15: 0000800100000000..000080012fffffff [220c] > PCI: Claiming 0000:03:00.0: Resource 15: 0000800100000000..000080012fffffff [220c] > PCI: Claiming 0000:04:06.0: Resource 14: 0000800100000000..000080010fffffff [220c] > PCI: Claiming 0000:05:00.0: Resource 0: 0000800100000000..0000800100001fff [204] > pci 0000:05:00.0: can't claim BAR 0 [mem 0x800100000000-0x800100001fff]: no compatible bridge window > > All the bridges 64-bit resource have pref bit, but the device resource does not > have pref set, then we can not find parent for the device resource, > as we can not put non-pref mmio under pref mmio. > > According to pcie spec errta > https://www.pcisig.com/specifications/pciexpress/base2/PCIe_Base_r2.1_Errata_08Jun10.pdf > page 13, in some case it is ok to mark some as pref. This link is broken. The text is included as an implementation note in the PCIe r3.1 spec, sec 7.5.2.1. If you cite that, I don't think we need the link. Please also fix the broken link in the patch below. This changelog needs to recap the conditions in that implementation note, i.e., entire path is PCIe, no conventional PCI or PCI-X devices perform peer-to-peer, no byte merging, etc. And we need something about why we believe these conditions all hold. In particular, I'm a little concerned about the peer-to-peer question and the TH bit. There are peer-to-peer patches being discussed, though I don't know whether they include conventional PCI and PCI-X. I don't think Linux currently does anything with the TH bit, but if we do in the future, or if BIOS enabled TH via the TPH Requester Capability (PCIe r3.1, sec 7.26) it seems like we could break something. Maybe we need to check for TPH being enabled? I'm interested in your opinion here. The changelog should show that you've considered the issue. If we support peer-to-peer, I guess that if we do put a non-prefetchable BAR in a prefetchable window, we would have to prevent any device in the hierarchy from enabling TPH? I have no idea how we know whether the BAR could be a target of a speculative Memory Read. I guess maybe the CPU ioremap properties are such that speculative reads are prohibited for MMIO apertures? Maybe you could investigate that and include the results? > diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c > index 5548044..676b55f 100644 > --- a/drivers/pci/probe.c > +++ b/drivers/pci/probe.c > @@ -1920,6 +1920,36 @@ static void pci_dma_configure(struct pci_dev *dev) > pci_put_host_bridge_device(bridge); > } > > +static bool pci_up_path_over_pcie(struct pci_bus *bus) > +{ > + if (pci_is_root_bus(bus)) > + return true; > + > + if (bus->self && !pci_is_pcie(bus->self)) > + return false; > + > + return pci_up_path_over_pcie(bus->parent); > +} > + > +/* > + * According to > + * https://www.pcisig.com/specifications/pciexpress/base2/PCIe_Base_r2.1_Errata_08Jun10.pdf > + * page 13, system firmware could put some 64bit non-pref under 64bit pref, > + * on some cases. > + * Let's mark if entire path from the host to the adapter is over PCI > + * Express. later will use that compute pref compaitable bit. > + */ > +static void pci_set_on_all_pcie_path(struct pci_dev *dev) > +{ > + if (!pci_is_pcie(dev)) > + return; > + > + if (!pci_up_path_over_pcie(dev->bus)) > + return; > + > + dev->on_all_pcie_path = 1; I think this can be set more simply as: if (pci_is_pcie(dev)) { if (pci_is_root_bus(dev->bus)) dev->pcie_path = 1; else if (dev->bus->self && dev->bus->self->pcie_path) dev->pcie_path = 1; } How about if you just put this code in set_pcie_port_type() so you don't have to add another function and worry about whether it's called after pcie_cap is assigned? Then you wouldn't need the pci_is_pcie() test either. > +} > + > void pci_device_add(struct pci_dev *dev, struct pci_bus *bus) > { > int ret; > @@ -1950,6 +1980,9 @@ void pci_device_add(struct pci_dev *dev, struct pci_bus *bus) > /* Initialize various capabilities */ > pci_init_capabilities(dev); > > + /* After pcie_cap is assigned */ > + pci_set_on_all_pcie_path(dev); > + > /* > * Add the device to our list of discovered devices > * and the bus list for fixup functions, etc. -- To unsubscribe from this list: send the line "unsubscribe sparclinux" 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/sparc/kernel/pci_common.c b/arch/sparc/kernel/pci_common.c index 1ebc7ff..6f206a1 100644 --- a/arch/sparc/kernel/pci_common.c +++ b/arch/sparc/kernel/pci_common.c @@ -343,7 +343,7 @@ static void pci_register_region(struct pci_bus *bus, const char *name, region.start = rstart; region.end = rstart + size - 1UL; pcibios_bus_to_resource(bus, res, ®ion); - bus_res = pci_find_bus_resource(bus, res); + bus_res = pci_find_bus_resource(bus, res, res->flags); if (!bus_res) { kfree(res); return; diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c index deb828f..bdb70b7 100644 --- a/drivers/pci/pci.c +++ b/drivers/pci/pci.c @@ -438,7 +438,7 @@ int pci_find_ht_capability(struct pci_dev *dev, int ht_cap) EXPORT_SYMBOL_GPL(pci_find_ht_capability); struct resource *pci_find_bus_resource(const struct pci_bus *bus, - struct resource *res) + struct resource *res, int flags) { struct resource *r; int i; @@ -453,7 +453,7 @@ struct resource *pci_find_bus_resource(const struct pci_bus *bus, * not, the allocator made a mistake. */ if (r->flags & IORESOURCE_PREFETCH && - !(res->flags & IORESOURCE_PREFETCH)) + !(flags & IORESOURCE_PREFETCH)) return NULL; /* @@ -481,7 +481,9 @@ struct resource *pci_find_bus_resource(const struct pci_bus *bus, struct resource *pci_find_parent_resource(const struct pci_dev *dev, struct resource *res) { - return pci_find_bus_resource(dev->bus, res); + int flags = pci_resource_pref_compatible(dev, res); + + return pci_find_bus_resource(dev->bus, res, flags); } EXPORT_SYMBOL(pci_find_parent_resource); diff --git a/drivers/pci/pci.h b/drivers/pci/pci.h index 586e63f..eb57780 100644 --- a/drivers/pci/pci.h +++ b/drivers/pci/pci.h @@ -368,4 +368,6 @@ int acpi_get_rc_resources(struct device *dev, const char *hid, u16 segment, struct resource *res); #endif +int pci_resource_pref_compatible(const struct pci_dev *dev, + struct resource *res); #endif /* DRIVERS_PCI_H */ diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c index 5548044..676b55f 100644 --- a/drivers/pci/probe.c +++ b/drivers/pci/probe.c @@ -1920,6 +1920,36 @@ static void pci_dma_configure(struct pci_dev *dev) pci_put_host_bridge_device(bridge); } +static bool pci_up_path_over_pcie(struct pci_bus *bus) +{ + if (pci_is_root_bus(bus)) + return true; + + if (bus->self && !pci_is_pcie(bus->self)) + return false; + + return pci_up_path_over_pcie(bus->parent); +} + +/* + * According to + * https://www.pcisig.com/specifications/pciexpress/base2/PCIe_Base_r2.1_Errata_08Jun10.pdf + * page 13, system firmware could put some 64bit non-pref under 64bit pref, + * on some cases. + * Let's mark if entire path from the host to the adapter is over PCI + * Express. later will use that compute pref compaitable bit. + */ +static void pci_set_on_all_pcie_path(struct pci_dev *dev) +{ + if (!pci_is_pcie(dev)) + return; + + if (!pci_up_path_over_pcie(dev->bus)) + return; + + dev->on_all_pcie_path = 1; +} + void pci_device_add(struct pci_dev *dev, struct pci_bus *bus) { int ret; @@ -1950,6 +1980,9 @@ void pci_device_add(struct pci_dev *dev, struct pci_bus *bus) /* Initialize various capabilities */ pci_init_capabilities(dev); + /* After pcie_cap is assigned */ + pci_set_on_all_pcie_path(dev); + /* * Add the device to our list of discovered devices * and the bus list for fixup functions, etc. diff --git a/drivers/pci/setup-bus.c b/drivers/pci/setup-bus.c index 958da7d..3de66e6 100644 --- a/drivers/pci/setup-bus.c +++ b/drivers/pci/setup-bus.c @@ -735,6 +735,20 @@ int pci_claim_bridge_resource(struct pci_dev *bridge, int i) return -EINVAL; } +int pci_resource_pref_compatible(const struct pci_dev *dev, + struct resource *res) +{ + if (res->flags & IORESOURCE_PREFETCH) + return res->flags; + + if ((res->flags & IORESOURCE_MEM) && + (res->flags & IORESOURCE_MEM_64) && + dev->on_all_pcie_path) + return res->flags | IORESOURCE_PREFETCH; + + return res->flags; +} + /* Check whether the bridge supports optional I/O and prefetchable memory ranges. If not, the respective base/limit registers must be read-only and read as 0. */ @@ -1032,11 +1046,12 @@ static int pbus_size_mem(struct pci_bus *bus, unsigned long mask, for (i = 0; i < PCI_NUM_RESOURCES; i++) { struct resource *r = &dev->resource[i]; resource_size_t r_size; + int flags = pci_resource_pref_compatible(dev, r); - if (r->parent || (r->flags & IORESOURCE_PCI_FIXED) || - ((r->flags & mask) != type && - (r->flags & mask) != type2 && - (r->flags & mask) != type3)) + if (r->parent || (flags & IORESOURCE_PCI_FIXED) || + ((flags & mask) != type && + (flags & mask) != type2 && + (flags & mask) != type3)) continue; r_size = resource_size(r); #ifdef CONFIG_PCI_IOV diff --git a/drivers/pci/setup-res.c b/drivers/pci/setup-res.c index 85774b7..2aeb4bc 100644 --- a/drivers/pci/setup-res.c +++ b/drivers/pci/setup-res.c @@ -285,15 +285,19 @@ static int __pci_assign_resource(struct pci_bus *bus, struct pci_dev *dev, static int _pci_assign_resource(struct pci_dev *dev, int resno, resource_size_t size, resource_size_t min_align) { + struct resource *res = dev->resource + resno; + int old_flags = res->flags; struct pci_bus *bus; int ret; + res->flags = pci_resource_pref_compatible(dev, res); bus = dev->bus; while ((ret = __pci_assign_resource(bus, dev, resno, size, min_align))) { if (!bus->parent || !bus->self->transparent) break; bus = bus->parent; } + res->flags = old_flags; return ret; } diff --git a/include/linux/pci.h b/include/linux/pci.h index 817786b..b14dd94 100644 --- a/include/linux/pci.h +++ b/include/linux/pci.h @@ -321,6 +321,7 @@ struct pci_dev { unsigned int hotplug_user_indicators:1; /* SlotCtl indicators controlled exclusively by user sysfs */ + unsigned int on_all_pcie_path:1; /* up to host-bridge all pcie */ unsigned int d3_delay; /* D3->D0 transition time in ms */ unsigned int d3cold_delay; /* D3cold->D0 transition time in ms */ @@ -837,7 +838,7 @@ void pcibios_resource_to_bus(struct pci_bus *bus, struct pci_bus_region *region, void pcibios_bus_to_resource(struct pci_bus *bus, struct resource *res, struct pci_bus_region *region); struct resource *pci_find_bus_resource(const struct pci_bus *bus, - struct resource *res); + struct resource *res, int flags); void pcibios_scan_specific_bus(int busn); struct pci_bus *pci_find_bus(int domain, int busnr); void pci_bus_add_devices(const struct pci_bus *bus);
We still get "no compatible bridge window" warning on sparc T5-8 after we add support for 64bit resource parsing for root bus. PCI: scan_bus[/pci@300/pci@1/pci@0/pci@6] bus no 8 PCI: Claiming 0000:00:01.0: Resource 15: 0000800100000000..00008004afffffff [220c] PCI: Claiming 0000:01:00.0: Resource 15: 0000800100000000..00008004afffffff [220c] PCI: Claiming 0000:02:04.0: Resource 15: 0000800100000000..000080012fffffff [220c] PCI: Claiming 0000:03:00.0: Resource 15: 0000800100000000..000080012fffffff [220c] PCI: Claiming 0000:04:06.0: Resource 14: 0000800100000000..000080010fffffff [220c] PCI: Claiming 0000:05:00.0: Resource 0: 0000800100000000..0000800100001fff [204] pci 0000:05:00.0: can't claim BAR 0 [mem 0x800100000000-0x800100001fff]: no compatible bridge window All the bridges 64-bit resource have pref bit, but the device resource does not have pref set, then we can not find parent for the device resource, as we can not put non-pref mmio under pref mmio. According to pcie spec errta https://www.pcisig.com/specifications/pciexpress/base2/PCIe_Base_r2.1_Errata_08Jun10.pdf page 13, in some case it is ok to mark some as pref. Mark if the entire path from the host to the adapter is over PCI Express. Set pref compatible bit for claim/sizing/assign for 64bit mem resource on that pcie device. -v2: set pref for mmio 64 when whole path is PCI Express, according to David Miller. -v3: don't set pref directly, change to UNDER_PREF, and set PREF before sizing and assign resource, and cleart PREF afterwards. requested by BenH. -v4: use on_all_pcie_path device flag instead. -v6: update after pci_find_bus_resource() change Link: http://lkml.kernel.org/r/CAE9FiQU1gJY1LYrxs+ma5LCTEEe4xmtjRG0aXJ9K_Tsu+m9Wuw@mail.gmail.com Reported-by: David Ahern <david.ahern@oracle.com> Tested-by: David Ahern <david.ahern@oracle.com> Link: https://bugzilla.kernel.org/show_bug.cgi?id=81431 Tested-by: TJ <linux@iam.tj> Signed-off-by: Yinghai Lu <yinghai@kernel.org> Tested-by: Khalid Aziz <khalid.aziz@oracle.com> Cc: sparclinux@vger.kernel.org --- arch/sparc/kernel/pci_common.c | 2 +- drivers/pci/pci.c | 8 +++++--- drivers/pci/pci.h | 2 ++ drivers/pci/probe.c | 33 +++++++++++++++++++++++++++++++++ drivers/pci/setup-bus.c | 23 +++++++++++++++++++---- drivers/pci/setup-res.c | 4 ++++ include/linux/pci.h | 3 ++- 7 files changed, 66 insertions(+), 9 deletions(-)