[ Note that XFS has a similar problem, with cpu_possible() calls ]
On Wed, Oct 20, 2004 at 07:49:05PM +0200, Christoph Hellwig wrote:
> On Wed, Oct 20, 2004 at 10:16:26AM -0700, Tom Rini wrote:
> > In net/ipv6/icmp.c::icmpv6_init() there is a call to cpu_possible()
> > which preprocesses down to "test_bit(((i)), (phys_cpu_present_map).bits)"
> > If ipv6 is a module, phys_cpu_present_map (or cpu_possible_map which is
> > defined t phys_cpu_present_map) needs to be exported.
>
> The loop in there should be rewritten as for_each_cpu which doesn't need
> this export.
Here's what I did:
--- linux-2.6.9.orig/net/ipv6/icmp.c
+++ linux-2.6.9/net/ipv6/icmp.c
@@ -691,10 +691,7 @@ int __init icmpv6_init(struct net_proto_
struct sock *sk;
int err, i, j;
- for (i = 0; i < NR_CPUS; i++) {
- if (!cpu_possible(i))
- continue;
-
+ for_each_cpu(i) {
But I still get the problem. Further, on i386 cpu_possible() becomes
cpu_callout_map which is exported as well.
--
Tom Rini
http://gate.crashing.org/~trini/
