[PATCH] x86_64 early numa init fix
authorRavikiran G Thirumalai <kiran@scalex86.org>
Fri, 30 Sep 2005 18:59:22 +0000 (11:59 -0700)
committerLinus Torvalds <torvalds@g5.osdl.org>
Fri, 30 Sep 2005 19:41:20 +0000 (12:41 -0700)
commit85cc5135ace4c8b75d7b4e1ea9fe15a7fcbd1516
tree9b671671331c8da39dc2a9b75e3d34d79aeac86e
parente6a045a5b89037ae87c8c1bc84403f1d498e52a1
[PATCH] x86_64 early numa init fix

The tests Alok carried out on Petr's box confirmed that cpu_to_node[BP] is
not setup early enough by numa_init_array due to the x86_64 changes in
2.6.14-rc*, and unfortunately set wrongly by the work around code in
numa_init_array().  cpu_to_node[0] gets set with 1 early and later gets set
properly to 0 during identify_cpu() when all cpus are brought up, but
confusing the numa slab in the process.

Here is a quick fix for this.  The right fix obviously is to have
cpu_to_node[bsp] setup early for numa_init_array().  The following patch
will fix the problem now, and the code can stay on even when
cpu_to_node{BP] gets fixed early correctly.

Thanks to Petr for access to his box.

Signed off by: Ravikiran Thirumalai <kiran@scalex86.org>
Signed-off-by: Alok N Kataria <alokk@calsoftinc.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
arch/x86_64/mm/numa.c