On Sat, 1 Feb 2020 11:21:55 -0800, Cong Wang wrote:

On Fri, Jan 31, 2020 at 3:27 PM 'Eric Dumazet' via syzkaller
[off-list ref] wrote:

quoted

NLA_BINARY can be confusing, since .len value represents
the max size of the blob.

cls_rsvp really wants user space to provide long enough data
for TCA_RSVP_DST and TCA_RSVP_SRC attributes.  

To clarify: for NLA_UNSPEC (0), len is the minimum length.

Nice clarification, that makes the patch obvious.

Looks like we have NLA_MIN_LEN now, from Wireguard, but I understand
for backports it's easier to stick to the original way of handling 
this.

Applied, thank you!