On 10/09/2014 12:14 PM, Tadeusz Struk wrote: > On 10/09/2014 04:23 AM, Prarit Bhargava wrote: >>> int numa_node; /* NUMA node this device is close to */ >>>> ... >> That's just bad english. The numa node value (for pci devices) is >> read from the ACPI tables on the system and represents the node that >> the pci_dev is connected to. >> >>>> }; >>>> >>>> In case when there are two nodes and only node 0 has memory, >>>> dev->numa_node will be 0 even though the device will be connected to the >>>> pci root port of node 1. >> Your calculation completely falls apart and returns incorrect values when >> cpu hotplug is used or if there are multi-socket nodes (as was the case >> on the system that panicked), or if one uses the new cluster-on-die mode. > > This calculation is sole for multi-socket configuration. This is why is > was introduced and what it was tested for. > There is no point discussing NUMA for single-socket configuration. > Single socket configurations are not NUMA. In this case dev->numa_node > is usually equal to NUMA_NO_NODE (-1) and adf_get_dev_node_id(pdev) will > always return 0;
The fact that you return an incorrect value here for any configuration is simply put, bad. You shouldn't do that. > Please confirm that, but I think the system it panicked on was a two > sockets system with only node 0 populated with memory and accelerator > plugged it to node 1 (phys_proc_id == 1). > In this case adf_get_dev_node_id(pdev) returned 1 and this was passed to > kzalloc_node(size, GFP_KERNEL, 1) and because there was no memory on > node 1 kzalloc_node() panicked. Yep; but my interpretation was that node 1 didn't exist at all and it panicked. > This patch will make sure that this will not happen and that the > configuration will be optimal. > Yep, it will. But what about cpu hotplug? P. -- To unsubscribe from this list: send the line "unsubscribe linux-crypto" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
