On 2019/10/14 17:25, Greg KH wrote: > On Mon, Oct 14, 2019 at 04:00:46PM +0800, Yunsheng Lin wrote: >> On 2019/10/12 18:47, Greg KH wrote: >>> On Sat, Oct 12, 2019 at 12:40:01PM +0200, Greg KH wrote: >>>> On Sat, Oct 12, 2019 at 05:47:56PM +0800, Yunsheng Lin wrote: >>>>> On 2019/10/12 15:40, Greg KH wrote: >>>>>> On Sat, Oct 12, 2019 at 02:17:26PM +0800, Yunsheng Lin wrote: >>>>>>> add pci and acpi maintainer >>>>>>> cc linux-pci@xxxxxxxxxxxxxxx and linux-acpi@xxxxxxxxxxxxxxx >>>>>>> >>>>>>> On 2019/10/11 19:15, Peter Zijlstra wrote: >>>>>>>> On Fri, Oct 11, 2019 at 11:27:54AM +0800, Yunsheng Lin wrote: >>>>>>>>> But I failed to see why the above is related to making node_to_cpumask_map() >>>>>>>>> NUMA_NO_NODE aware? >>>>>>>> >>>>>>>> Your initial bug is for hns3, which is a PCI device, which really _MUST_ >>>>>>>> have a node assigned. >>>>>>>> >>>>>>>> It not having one, is a straight up bug. We must not silently accept >>>>>>>> NO_NODE there, ever. >>>>>>>> >>>>>>> >>>>>>> I suppose you mean reporting a lack of affinity when the node of a pcie >>>>>>> device is not set by "not silently accept NO_NODE". >>>>>> >>>>>> If the firmware of a pci device does not provide the node information, >>>>>> then yes, warn about that. >>>>>> >>>>>>> As Greg has asked about in [1]: >>>>>>> what is a user to do when the user sees the kernel reporting that? >>>>>>> >>>>>>> We may tell user to contact their vendor for info or updates about >>>>>>> that when they do not know about their system well enough, but their >>>>>>> vendor may get away with this by quoting ACPI spec as the spec >>>>>>> considering this optional. Should the user believe this is indeed a >>>>>>> fw bug or a misreport from the kernel? >>>>>> >>>>>> Say it is a firmware bug, if it is a firmware bug, that's simple. >>>>>> >>>>>>> If this kind of reporting is common pratice and will not cause any >>>>>>> misunderstanding, then maybe we can report that. >>>>>> >>>>>> Yes, please do so, that's the only way those boxes are ever going to get >>>>>> fixed. And go add the test to the "firmware testing" tool that is based >>>>>> on Linux that Intel has somewhere, to give vendors a chance to fix this >>>>>> before they ship hardware. >>>>>> >>>>>> This shouldn't be a big deal, we warn of other hardware bugs all the >>>>>> time. >>>>> >>>>> Ok, thanks for clarifying. >>>>> >>>>> Will send a patch to catch the case when a pcie device without numa node >>>>> being set and warn about it. >>>>> >>>>> Maybe use dev->bus to verify if it is a pci device? >>>> >>>> No, do that in the pci bus core code itself, when creating the devices >>>> as that is when you know, or do not know, the numa node, right? >>>> >>>> This can't be in the driver core only, as each bus type will have a >>>> different way of determining what the node the device is on. For some >>>> reason, I thought the PCI core code already does this, right? >>> >>> Yes, pci_irq_get_node(), which NO ONE CALLS! I should go delete that >>> thing... >>> >>> Anyway, it looks like the pci core code does call set_dev_node() based >>> on the PCI bridge, so if that is set up properly, all should be fine. >>> >>> If not, well, you have buggy firmware and you need to warn about that at >>> the time you are creating the bridge. Look at the call to >>> pcibus_to_node() in pci_register_host_bridge(). >> >> Thanks for pointing out the specific function. >> Maybe we do not need to warn about the case when the device has a parent, >> because we must have warned about the parent if the device has a parent >> and the parent also has a node of NO_NODE, so do not need to warn the child >> device anymore? like blew: >> >> @@ -932,6 +932,10 @@ static int pci_register_host_bridge(struct pci_host_bridge *bridge) >> list_add_tail(&bus->node, &pci_root_buses); >> up_write(&pci_bus_sem); >> >> + if (nr_node_ids > 1 && !parent && > > Why do you need to check this? If you have a parent, it's your node > should be set, if not, that's an error, right? If the device has parent and the parent device also has a node of NUMA_NO_NODE, then maybe we have warned about the parent device, so we do not have to warn about the child device? In pci_register_host_bridge(): if (!parent) set_dev_node(bus->bridge, pcibus_to_node(bus)); The above only set the node of the bridge device to the node of bus if the bridge device does not have a parent. bus->dev.parent = bus->bridge; dev_set_name(&bus->dev, "%04x:%02x", pci_domain_nr(bus), bus->number); name = dev_name(&bus->dev); err = device_register(&bus->dev); The above then set the bus device's parent to bridge device, and then call device_register(), which will set the bus device's node according to bridge device' node. > >> + dev_to_node(bus->bridge) == NUMA_NO_NODE) >> + dev_err(bus->bridge, FW_BUG "No node assigned on NUMA capable HW. Please contact your vendor for updates.\n"); >> + >> return 0; > > Who set that bus->bridge node to NUMA_NO_NODE? It seems x86 and arm64 may have different implemention of pcibus_to_node(): For arm64: int pcibus_to_node(struct pci_bus *bus) { return dev_to_node(&bus->dev); } And the node of bus is set in: int pcibios_root_bridge_prepare(struct pci_host_bridge *bridge) { if (!acpi_disabled) { struct pci_config_window *cfg = bridge->bus->sysdata; struct acpi_device *adev = to_acpi_device(cfg->parent); struct device *bus_dev = &bridge->bus->dev; ACPI_COMPANION_SET(&bridge->dev, adev); set_dev_node(bus_dev, acpi_get_node(acpi_device_handle(adev))); } return 0; } acpi_get_node() may return NUMA_NO_NODE in pcibios_root_bridge_prepare(), which will set the node of bus_dev to NUMA_NO_NODE x86: static inline int __pcibus_to_node(const struct pci_bus *bus) { const struct pci_sysdata *sd = bus->sysdata; return sd->node; } And the node of bus is set in pci_acpi_scan_root(), which uses pci_acpi_root_get_node() get the node of a bus. And it also may return NUMA_NO_NODE. > If that is set, the firmware is broken, as you say, but you need to tell > the user what firmware is broken. Maybe mentioning the BIOS in log? dev_err(bus->bridge, FW_BUG "No node assigned on NUMA capable HW by BIOS. Please contact your vendor for updates.\n"); > > Try something like this out and see what happens on your machine that > had things "broken". What does it say? Does not have a older bios right now. But always returning NUMA_NO_NODE by below patch: --- a/drivers/acpi/numa.c +++ b/drivers/acpi/numa.c @@ -484,6 +484,7 @@ int acpi_get_node(acpi_handle handle) pxm = acpi_get_pxm(handle); - return acpi_map_pxm_to_node(pxm); + return -1; + //return acpi_map_pxm_to_node(pxm); it gives the blow warning in my machine: [ 16.126136] pci0000:00: [Firmware Bug]: No node assigned on NUMA capable HW by BIOS. Please contact your vendor for updates. [ 17.733831] pci0000:7b: [Firmware Bug]: No node assigned on NUMA capable HW by BIOS. Please contact your vendor for updates. [ 18.020924] pci0000:7a: [Firmware Bug]: No node assigned on NUMA capable HW by BIOS. Please contact your vendor for updates. [ 18.552832] pci0000:78: [Firmware Bug]: No node assigned on NUMA capable HW by BIOS. Please contact your vendor for updates. [ 19.514948] pci0000:7c: [Firmware Bug]: No node assigned on NUMA capable HW by BIOS. Please contact your vendor for updates. [ 20.652990] pci0000:74: [Firmware Bug]: No node assigned on NUMA capable HW by BIOS. Please contact your vendor for updates. [ 22.573200] pci0000:80: [Firmware Bug]: No node assigned on NUMA capable HW by BIOS. Please contact your vendor for updates. [ 23.225355] pci0000:bb: [Firmware Bug]: No node assigned on NUMA capable HW by BIOS. Please contact your vendor for updates. [ 23.514040] pci0000:ba: [Firmware Bug]: No node assigned on NUMA capable HW by BIOS. Please contact your vendor for updates. [ 24.050107] pci0000:b8: [Firmware Bug]: No node assigned on NUMA capable HW by BIOS. Please contact your vendor for updates. [ 25.017491] pci0000:bc: [Firmware Bug]: No node assigned on NUMA capable HW by BIOS. Please contact your vendor for updates. [ 25.557974] pci0000:b4: [Firmware Bug]: No node assigned on NUMA capable HW by BIOS. Please contact your vendor for updates. > >> Also, we do not need to warn about that in pci_device_add(), Right? >> Because we must have warned about the pci host bridge of the pci device. > > That should be true, yes. > >> I may be wrong about above because I am not so familiar with the pci. >> >>> >>> And yes, you need to do this all on a per-bus-type basis, as has been >>> pointed out. It's up to the bus to create the device and set this up >>> properly. >> >> Thanks. >> Will do that on per-bus-type basis. > > Good luck, I don't really think that most, if any, of this is needed, > but hey, it's nice to clean it up where it can be :) > > greg k-h > > . >