Re: http://lkml.org/lkml/2007/1/28/146 Just got a similar OOPS on a system under heavy load (transcode + p2p), with kernel 2.6.20.3 / x86_64 (in free_block). [] free_block+0xae/0x13c [] drain_array+0x93/0xd1 [] cache_reap+0xea/0x239 [] cache_reap+0x0/0x239 [] run_workqueue+0x95/0x140 [] worker_thread+0x0/0x150 [] worker_thread+0x11a/0x150 [] default_wake_function+0x0/0xe [] kthread+0xd1/0x100 [] child_rip+0xa/0x12 [] kthread+0x0/0x100 [] child_rip+0x0/0x12 Not sure this is relevant, but transcode uses nice/ionice: ionice -n 7 -c 3 nice -n 19 transcode [...] No forced pre-emption, 100HZ, etc... Note: this is not an SMP system, just a SMP kernel. So if it is a race, it is not an SMP triggered race. (or I could just be unlucky again - damn those cosmic rays) Antoine