On 2/5/2024 8:17 AM, Kalle Valo wrote: > Kang Yang <quic_kangyang@xxxxxxxxxxx> writes: >> + if (WARN_ON(skb->len < ie_offset)) >> + return -EINVAL; >> + >> + ie = (u8 *)cfg80211_find_vendor_ie(oui, oui_type, >> + skb->data + ie_offset, >> + skb->len - ie_offset); > > So cfg80211_find_vendor_ie() returns: > > static inline const u8 * > cfg80211_find_vendor_ie(unsigned int oui, int oui_type, > const u8 *ies, unsigned int len) > > You are casting away the const here. I see that other drivers do the > same as the assumption is that the pointer points to the same buffer, > but I still don't really like casting away const. Thoughts? I'm not a fan of typecasting either. However, the problem here is that from the perspective of cfg80211_find_vendor_ie() the buffer is const (not modified in any way by the function), but from the perspective of every caller the buffer is part of a message that is in RAM. So you have to decide if you want to change the prototype to remove the const from cfg80211_find_vendor_ie() or if you want to keep the current semantic to cast away the const. This is one of the rare cases where I think the typecast is ok. /jeff
