Dina, Comments below. I updated the webrev. - Dan On 17/11/2009 10:52 PM, Dina wrote: > DKN-1 > bignumimpl.c, line 188 (new numbering) > arcfourcrypt.c, line 240: > nit: ... checks for an AMD-64 ...
FIXED. > DKN-2 > aes_intel.s, line 479/480 (new numbers): > is there a reason why there is no > aeskeygenassist $0x40, %xmm0, %xmm1 > call _key_expansion 256b > after the preceding block? > > line 444 says 14 rounds, but the following code looks like 13 > line 494 says 12 rounds, but the following code looks like 8 > line 531 says 10 rounds, and it actually looks like 10 rounds > > I don't get it. I noticed that too and so did someone reviewing the Linux port of similar code. After looking it over for a few months (among other things), I found the comments misleading. The term "rounds" is used in two different ways: * "ROUNDS" of encryption/decryption required The 14, 12, 10 rounds for aes-256, -192, -128 refer to the encryption and decryption rounds performed later. * "ROUNDS"--number of calls to AESKEYGENASSIST The 13, 8, and 10 are the number of calls to AESKEYGENASSIST required to setup the key expansion (from the user-provided key). I removed the comments that called this "rounds" and replaced it with "expanding the key" (the comment used in the orignal Intel white paper). > DKN-3 > intel/sha1/Makefile > intel/sha2/Makefile > LINTFLAGS += -I$(COMDIR) -I../../common/sys > but > sparc/sha1/Makefile > sparc/sha2/Makefile > sun4u/sha1/Makefile > LINTFLAGS += -I$(COMDIR) > You're sure about this? Yeah. The Intel version uses a header file that the SPARC-ites (not to be confused with the Hittites) do not need. -- This message posted from opensolaris.org
