Messages in this thread | | | Date | Tue, 6 May 2003 13:28:44 +0100 (BST) | From | Hugh Dickins <> | Subject | initcall kmem_cache cpu 1 oops |
| |
I don't know if it's right or wrong for initcalls to happen on cpus other than 0. But with CONFIG_NR_SIBLINGS_2 on 2*HT*P4 in 2.5.69-mm1 (I didn't try 2.5.68-mm, but 2.5.69 okay) I have sock_init run on cpu 0 setting sock_inode_cachep and its array[0], then later rtnetlink_init on cpu 1, causing kmem_cache_alloc oops on NULL sock_inode_cachep->array[1]: g_cpucache_up is still PARTIAL, cpucache_init is yet to be called.
Before 2.5.68-mm3 rtnetlink_init was invoked from within sock_init, but it's now from a separate netlink_proto_init: so that's another source of doubt. For now I have CONFIG_NR_SIBLINGS_0 instead (from a vague suspicion that CONFIG_SHARE_RUNQUEUE work might have caused unexpectedly early switch from cpu 0 to 1), and that works around it for me: but I'm not pointing any finger of blame, I don't know the axioms here at all.
Hugh
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |