On Tue, 2008-07-08 at 16:13 -0700, Harvey Harrison wrote:
> On Wed, 2008-07-09 at 01:04 +0200, Johannes Berg wrote:
> > On Tue, 2008-07-08 at 15:25 -0700, Harvey Harrison wrote:
> > > Those functions that still use ieee80211_get_hdrlen are moved over
> > > to use the little endian frame control.
> >
> > > - switch (fc & (IEEE80211_FCTL_TODS | IEEE80211_FCTL_FROMDS)) {
> > > - case IEEE80211_FCTL_TODS:
> > > - /* BSSID SA DA */
> > > - memcpy(dst, hdr->addr3, ETH_ALEN);
> > > - memcpy(src, hdr->addr2, ETH_ALEN);
> > > -
> > > + switch (hdr->frame_control &
> > > + cpu_to_le16(IEEE80211_FCTL_TODS | IEEE80211_FCTL_FROMDS)) {
> > > + case __constant_cpu_to_le16(IEEE80211_FCTL_TODS):
> > > if (unlikely(sdata->vif.type != IEEE80211_IF_TYPE_AP &&
> > > sdata->vif.type != IEEE80211_IF_TYPE_VLAN))
> > > return -1;
> > > break;
> >
> > Looks good, but didn't you just lose all the memcpys?
> >
>
> See the three lines immediately before the part you quoted.
Oh I see, sorry. Wouldn't it be more efficient to keep it inlined in the
switch though? Or is the compiler actually smart enough to move the
inline into there?
johannes
Attachment:
signature.asc
Description: This is a digitally signed message part
