Most recent kernel where this bug did *NOT* occur: 2.6.21-rc6.mm1 with SLAB Distribution: Fedora Devel Hardware Environment: AMD X2 on CK804 Software Environment: N/A Problem Description: page allocation failure during heavy system activity (mesa + kernel build + video capture) Steps to reproduce: one-time
Created attachment 11446 [details] kernel logs since initial boot
Created attachment 11447 [details] kernel config
Created attachment 11448 [details] lspci
A know problem with too optimistic page order use by SLUB if antifragmentation is enabled. I experienced it myself a couple of times. 2.6.21-mm2 has the max order reduced to 3. See also the recent conversation with Mel on linux-mm.