Correct a problem reported by Holger Mikolon that turns out to be

a bug in the initial EP80579 commit from dms that was exposed by gcc4.

Lots of help tracking down the block of code at fault from
Mike Belopuhov, but I spotted the problem in the end :)

ok kettenis@ deraadt@

diff --git a/sys/dev/pci/if_em.c b/sys/dev/pci/if_em.c
index 660a271..5e58cef 100644 (file)
--- a/sys/dev/pci/if_em.c
+++ b/sys/dev/pci/if_em.c
@@ -31,7 +31,7 @@ POSSIBILITY OF SUCH DAMAGE.
 
 ***************************************************************************/
 
-/* $OpenBSD: if_em.c,v 1.242 2010/08/03 16:21:52 jsg Exp $ */
+/* $OpenBSD: if_em.c,v 1.243 2010/08/04 17:10:34 jsg Exp $ */
 /* $FreeBSD: if_em.c,v 1.46 2004/09/29 18:28:28 mlaier Exp $ */
 
 #include <dev/pci/if_em.h>
@@ -1648,12 +1648,11 @@ em_allocate_pci_resources(struct em_softc *sc)
         * can confuse the system
         */
        if(sc->hw.mac_type == em_icp_xxxx) {
-               uint8_t offset;
+               int offset;
                pcireg_t val;
                
                if (!pci_get_capability(sc->osdep.em_pa.pa_pc, 
-                   sc->osdep.em_pa.pa_tag, PCI_CAP_ID_ST, (int*) &offset, 
-                   &val)) {
+                   sc->osdep.em_pa.pa_tag, PCI_CAP_ID_ST, &offset, &val)) {
                        return (0);
                }
                offset += PCI_ST_SMIA_OFFSET;