> Right hard_header_len is being overloaded with two meanings. > On the one hand it's used to indicate the length of the header > managed by the hardware header cache, on the other hand we use > it to indicate the head room for those devices that don't have > a cache. > > Since your device wants both, it fails miserably. Yeah I guess that is actually the root cause. > There are three ways out of this. You could have your own cache > (I haven't looked at your headers so I don't know whether that > makes sense), have no cache at all, or add a new field that gets > added to hard_header_len for the purposes of calculating head > room. In fact, I should probably add another field for the needed tail length too since we need that as well. johannes
Attachment:
signature.asc
Description: This is a digitally signed message part
