Messages in this thread | | | From | Mel Gorman <> | Subject | [PATCH 0/2] Relax restrictions on setting CONFIG_NUMA on x86 | Date | Fri, 18 Jan 2008 15:35:29 +0000 (GMT) |
| |
A fix[1] was merged to the x86.git tree that allowed NUMA kernels to boot on normal x86 machines (and not just NUMA-Q, Summit etc.). I took a look at the restrictions on setting NUMA on x86 to see if they could be lifted.
The following two patches remove the restrictions on pagetable layout and architecture type when setting NUMA on x86. This is aimed at improving the testing coverage of NUMA code paths by allowing it to be set in more situations. The dependency on CONFIG_ACPI is left due to possible SRAT parsing (although this could also be lifted) and on EXPERIMENTAL as the testing coverage for NUMA on x86 is so weak. The one potential gotcha is that a definition of NR_NODE_MEMBLKS is moved to an arch-specific file. From what I can see, this value was expected to be defined on a per-arch basis and the definition in include/linux/acpi.h was an anomaly.
The patches in combination with the boot-numa-x86 fix have been boot-tested on a bog-standard laptop with 512MB RAM, QEMU-i386 with 1324MB in a variety of different configuarations and a NUMA-Q with its standard .config.
[1] For others watching, this fix was considered controversial as a potentially better solution existed as discussed in http://lkml.org/lkml/2007/8/24/220. However, this better alternative was never investigated properly and booting NUMA remained broken. The merged fix is a variation and while it does waste memory, it is considered better than crashing. Wider testing coverage may help motivate fixing this paths. Mel Gorman Part-time Phd Student Linux Technology Center University of Limerick IBM Dublin Software Lab
| |