On Sun, Nov 17, 2019 at 02:03:21PM -0700, Todd C. Miller wrote: > On Sun, 17 Nov 2019 20:38:59 +0100, Alexander Bluhm wrote: > > > I think the best way to handle it, is to make the kernel strict and > > fix userland. If the kernel would allow the sloppiest userland > > program to succeed, creating security would be hard. > > Sorry, I don't agree. We cannot expect userland to fill in a > non-standard length field. The kernel ioctl handler should set > sa_len appropriately instead. >
I agree. It should do the same check and assignment as for bind(), etc. -- :wq Claudio