On Wed, Jul 11, 2018 at 09:27:56AM -0700, Eric Biggers wrote: > > The callers do check for errors, but at the point of the proposed BUG_ON() a > buffer overflow may have already occurred, so I think a BUG_ON() would be more > appropriate than a WARN_ON(). Of course, it would be better to prevent any > buffer overflow from occurring in the first place, but that's already the > purpose of the 'len != crypto_dh_key_len(params)' check; the issue was that > 'crypto_dh_key_len()' calculated the wrong length. How about adding an end argument to dh_pack_data so that it can check that the buffer isn't overflown each time it's called? Then if you cared you could also have one final check at the end of the function to ensure the buffer is fully used. As we can easily avoid having a BUG_ON here I think we should since a BUG_ON would leave the machine unresponsive which should be a last resort. Thanks, -- Email: Herbert Xu <herbert@xxxxxxxxxxxxxxxxxxx> Home Page: http://gondor.apana.org.au/~herbert/ PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt