Re: libsa: udp read changes endianness

2018-03-31 Thread Theo de Raadt
> I think we should not convert headers into a host-readable format, the > upper layers need to know the network layers if they do this kind of > layer overreach. I agree. It probably comes from a time when ntohs and family were true functions (during the inline asm conversion days); it was seen

libsa: udp read changes endianness

2018-03-31 Thread Patrick Wildt
Hi, I have been working on TFTP boot support for arm64 and armv7 on top of u-boot. One thing that set me back was an endianness issue. TFTP works the way that you send to port 69, but when the server answers he chooses a new source port. So when you reply again, you have to reply to the new sou