Add bulk OCB for Camellia AES-NI/AVX and AES-NI/AVX2 implementations

* cipher/camellia-aesni-avx-amd64.S: Add OCB assembly functions.
* cipher/camellia-aesni-avx2-amd64.S: Add OCB assembly functions.
* cipher/camellia-glue.c (_gcry_camellia_aesni_avx_ocb_enc)
(_gcry_camellia_aesni_avx_ocb_dec, _gcry_camellia_aesni_avx_ocb_auth)
(_gcry_camellia_aesni_avx2_ocb_enc, _gcry_camellia_aesni_avx2_ocb_dec)
(_gcry_camellia_aesni_avx2_ocb_auth): New prototypes.
(get_l, _gcry_camellia_ocb_crypt, _gcry_camellia_ocb_auth): New.
* cipher/cipher.c (_gcry_cipher_open_internal): Setup OCB bulk
functions for Camellia.
* src/cipher.h (_gcry_camellia_ocb_crypt)
(_gcry_camellia_ocb_auth): New.
* tests/basic.c (check_ocb_cipher): Add test-vector for Camellia.
--

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@iki.fi>

4 files changed, 1252 insertions, 6 deletions

diff --git a/cipher/camellia-aesni-avx-amd64.S b/cipher/camellia-aesni-avx-amd64.S
index c047a21e..5a3a3cbc 100644
--- a/cipher/camellia-aesni-avx-amd64.S
+++ b/cipher/camellia-aesni-avx-amd64.S
@@ -1,6 +1,6 @@
 /* camellia-avx-aesni-amd64.S  -  AES-NI/AVX implementation of Camellia cipher
  *
- * Copyright (C) 2013 Jussi Kivilinna <jussi.kivilinna@iki.fi>
+ * Copyright (C) 2013-2015 Jussi Kivilinna <jussi.kivilinna@iki.fi>
  *
  * This file is part of Libgcrypt.
  *
@@ -1211,6 +1211,428 @@ _gcry_camellia_aesni_avx_cfb_dec:
 	ret;
 ELF(.size _gcry_camellia_aesni_avx_cfb_dec,.-_gcry_camellia_aesni_avx_cfb_dec;)
 
+.align 8
+.globl _gcry_camellia_aesni_avx_ocb_enc
+ELF(.type   _gcry_camellia_aesni_avx_ocb_enc,@function;)
+
+_gcry_camellia_aesni_avx_ocb_enc:
+	/* input:
+	 *	%rdi: ctx, CTX
+	 *	%rsi: dst (16 blocks)
+	 *	%rdx: src (16 blocks)
+	 *	%rcx: offset
+	 *	%r8 : checksum
+	 *	%r9 : L pointers (void *L[16])
+	 */
+
+	pushq %rbp;
+	movq %rsp, %rbp;
+
+	vzeroupper;
+
+	subq $(16 * 16 + 4 * 8), %rsp;
+	andq $~31, %rsp;
+	movq %rsp, %rax;
+
+	movq %r10, (16 * 16 + 0 * 8)(%rax);
+	movq %r11, (16 * 16 + 1 * 8)(%rax);
+	movq %r12, (16 * 16 + 2 * 8)(%rax);
+	movq %r13, (16 * 16 + 3 * 8)(%rax);
+
+	vmovdqu (%rcx), %xmm14;
+	vmovdqu (%r8), %xmm15;
+
+	/* Offset_i = Offset_{i-1} xor L_{ntz(i)} */
+	/* Checksum_i = Checksum_{i-1} xor P_i  */
+	/* C_i = Offset_i xor ENCIPHER(K, P_i xor Offset_i)  */
+
+#define OCB_INPUT(n, lreg, xreg) \
+	  vmovdqu (n * 16)(%rdx), xreg; \
+	  vpxor (lreg), %xmm14, %xmm14; \
+	  vpxor xreg, %xmm15, %xmm15; \
+	  vpxor xreg, %xmm14, xreg; \
+	  vmovdqu %xmm14, (n * 16)(%rsi);
+	movq (0 * 8)(%r9), %r10;
+	movq (1 * 8)(%r9), %r11;
+	movq (2 * 8)(%r9), %r12;
+	movq (3 * 8)(%r9), %r13;
+	OCB_INPUT(0, %r10, %xmm0);
+	vmovdqu %xmm0, (15 * 16)(%rax);
+	OCB_INPUT(1, %r11, %xmm0);
+	vmovdqu %xmm0, (14 * 16)(%rax);
+	OCB_INPUT(2, %r12, %xmm13);
+	OCB_INPUT(3, %r13, %xmm12);
+	movq (4 * 8)(%r9), %r10;
+	movq (5 * 8)(%r9), %r11;
+	movq (6 * 8)(%r9), %r12;
+	movq (7 * 8)(%r9), %r13;
+	OCB_INPUT(4, %r10, %xmm11);
+	OCB_INPUT(5, %r11, %xmm10);
+	OCB_INPUT(6, %r12, %xmm9);
+	OCB_INPUT(7, %r13, %xmm8);
+	movq (8 * 8)(%r9), %r10;
+	movq (9 * 8)(%r9), %r11;
+	movq (10 * 8)(%r9), %r12;
+	movq (11 * 8)(%r9), %r13;
+	OCB_INPUT(8, %r10, %xmm7);
+	OCB_INPUT(9, %r11, %xmm6);
+	OCB_INPUT(10, %r12, %xmm5);
+	OCB_INPUT(11, %r13, %xmm4);
+	movq (12 * 8)(%r9), %r10;
+	movq (13 * 8)(%r9), %r11;
+	movq (14 * 8)(%r9), %r12;
+	movq (15 * 8)(%r9), %r13;
+	OCB_INPUT(12, %r10, %xmm3);
+	OCB_INPUT(13, %r11, %xmm2);
+	OCB_INPUT(14, %r12, %xmm1);
+	OCB_INPUT(15, %r13, %xmm0);
+#undef OCB_INPUT
+
+	vmovdqu %xmm14, (%rcx);
+	vmovdqu %xmm15, (%r8);
+
+	/* inpack16_pre: */
+	vmovq (key_table)(CTX), %xmm15;
+	vpshufb .Lpack_bswap RIP, %xmm15, %xmm15;
+	vpxor %xmm0, %xmm15, %xmm0;
+	vpxor %xmm1, %xmm15, %xmm1;
+	vpxor %xmm2, %xmm15, %xmm2;
+	vpxor %xmm3, %xmm15, %xmm3;
+	vpxor %xmm4, %xmm15, %xmm4;
+	vpxor %xmm5, %xmm15, %xmm5;
+	vpxor %xmm6, %xmm15, %xmm6;
+	vpxor %xmm7, %xmm15, %xmm7;
+	vpxor %xmm8, %xmm15, %xmm8;
+	vpxor %xmm9, %xmm15, %xmm9;
+	vpxor %xmm10, %xmm15, %xmm10;
+	vpxor %xmm11, %xmm15, %xmm11;
+	vpxor %xmm12, %xmm15, %xmm12;
+	vpxor %xmm13, %xmm15, %xmm13;
+	vpxor 14 * 16(%rax), %xmm15, %xmm14;
+	vpxor 15 * 16(%rax), %xmm15, %xmm15;
+
+	call __camellia_enc_blk16;
+
+	vpxor 0 * 16(%rsi), %xmm7, %xmm7;
+	vpxor 1 * 16(%rsi), %xmm6, %xmm6;
+	vpxor 2 * 16(%rsi), %xmm5, %xmm5;
+	vpxor 3 * 16(%rsi), %xmm4, %xmm4;
+	vpxor 4 * 16(%rsi), %xmm3, %xmm3;
+	vpxor 5 * 16(%rsi), %xmm2, %xmm2;
+	vpxor 6 * 16(%rsi), %xmm1, %xmm1;
+	vpxor 7 * 16(%rsi), %xmm0, %xmm0;
+	vpxor 8 * 16(%rsi), %xmm15, %xmm15;
+	vpxor 9 * 16(%rsi), %xmm14, %xmm14;
+	vpxor 10 * 16(%rsi), %xmm13, %xmm13;
+	vpxor 11 * 16(%rsi), %xmm12, %xmm12;
+	vpxor 12 * 16(%rsi), %xmm11, %xmm11;
+	vpxor 13 * 16(%rsi), %xmm10, %xmm10;
+	vpxor 14 * 16(%rsi), %xmm9, %xmm9;
+	vpxor 15 * 16(%rsi), %xmm8, %xmm8;
+
+	write_output(%xmm7, %xmm6, %xmm5, %xmm4, %xmm3, %xmm2, %xmm1, %xmm0,
+		     %xmm15, %xmm14, %xmm13, %xmm12, %xmm11, %xmm10, %xmm9,
+		     %xmm8, %rsi);
+
+	vzeroall;
+
+	movq (16 * 16 + 0 * 8)(%rax), %r10;
+	movq (16 * 16 + 1 * 8)(%rax), %r11;
+	movq (16 * 16 + 2 * 8)(%rax), %r12;
+	movq (16 * 16 + 3 * 8)(%rax), %r13;
+
+	leave;
+	ret;
+ELF(.size _gcry_camellia_aesni_avx_ocb_enc,.-_gcry_camellia_aesni_avx_ocb_enc;)
+
+.align 8
+.globl _gcry_camellia_aesni_avx_ocb_dec
+ELF(.type   _gcry_camellia_aesni_avx_ocb_dec,@function;)
+
+_gcry_camellia_aesni_avx_ocb_dec:
+	/* input:
+	 *	%rdi: ctx, CTX
+	 *	%rsi: dst (16 blocks)
+	 *	%rdx: src (16 blocks)
+	 *	%rcx: offset
+	 *	%r8 : checksum
+	 *	%r9 : L pointers (void *L[16])
+	 */
+
+	pushq %rbp;
+	movq %rsp, %rbp;
+
+	vzeroupper;
+
+	subq $(16 * 16 + 4 * 8), %rsp;
+	andq $~31, %rsp;
+	movq %rsp, %rax;
+
+	movq %r10, (16 * 16 + 0 * 8)(%rax);
+	movq %r11, (16 * 16 + 1 * 8)(%rax);
+	movq %r12, (16 * 16 + 2 * 8)(%rax);
+	movq %r13, (16 * 16 + 3 * 8)(%rax);
+
+	vmovdqu (%rcx), %xmm15;
+
+	/* Offset_i = Offset_{i-1} xor L_{ntz(i)} */
+	/* P_i = Offset_i xor DECIPHER(K, C_i xor Offset_i)  */
+
+#define OCB_INPUT(n, lreg, xreg) \
+	  vmovdqu (n * 16)(%rdx), xreg; \
+	  vpxor (lreg), %xmm15, %xmm15; \
+	  vpxor xreg, %xmm15, xreg; \
+	  vmovdqu %xmm15, (n * 16)(%rsi);
+	movq (0 * 8)(%r9), %r10;
+	movq (1 * 8)(%r9), %r11;
+	movq (2 * 8)(%r9), %r12;
+	movq (3 * 8)(%r9), %r13;
+	OCB_INPUT(0, %r10, %xmm0);
+	vmovdqu %xmm0, (15 * 16)(%rax);
+	OCB_INPUT(1, %r11, %xmm14);
+	OCB_INPUT(2, %r12, %xmm13);
+	OCB_INPUT(3, %r13, %xmm12);
+	movq (4 * 8)(%r9), %r10;
+	movq (5 * 8)(%r9), %r11;
+	movq (6 * 8)(%r9), %r12;
+	movq (7 * 8)(%r9), %r13;
+	OCB_INPUT(4, %r10, %xmm11);
+	OCB_INPUT(5, %r11, %xmm10);
+	OCB_INPUT(6, %r12, %xmm9);
+	OCB_INPUT(7, %r13, %xmm8);
+	movq (8 * 8)(%r9), %r10;
+	movq (9 * 8)(%r9), %r11;
+	movq (10 * 8)(%r9), %r12;
+	movq (11 * 8)(%r9), %r13;
+	OCB_INPUT(8, %r10, %xmm7);
+	OCB_INPUT(9, %r11, %xmm6);
+	OCB_INPUT(10, %r12, %xmm5);
+	OCB_INPUT(11, %r13, %xmm4);
+	movq (12 * 8)(%r9), %r10;
+	movq (13 * 8)(%r9), %r11;
+	movq (14 * 8)(%r9), %r12;
+	movq (15 * 8)(%r9), %r13;
+	OCB_INPUT(12, %r10, %xmm3);
+	OCB_INPUT(13, %r11, %xmm2);
+	OCB_INPUT(14, %r12, %xmm1);
+	OCB_INPUT(15, %r13, %xmm0);
+#undef OCB_INPUT
+
+	vmovdqu %xmm15, (%rcx);
+
+	movq %r8, %r10;
+
+	cmpl $128, key_bitlength(CTX);
+	movl $32, %r8d;
+	movl $24, %r9d;
+	cmovel %r9d, %r8d; /* max */
+
+	/* inpack16_pre: */
+	vmovq (key_table)(CTX, %r8, 8), %xmm15;
+	vpshufb .Lpack_bswap RIP, %xmm15, %xmm15;
+	vpxor %xmm0, %xmm15, %xmm0;
+	vpxor %xmm1, %xmm15, %xmm1;
+	vpxor %xmm2, %xmm15, %xmm2;
+	vpxor %xmm3, %xmm15, %xmm3;
+	vpxor %xmm4, %xmm15, %xmm4;
+	vpxor %xmm5, %xmm15, %xmm5;
+	vpxor %xmm6, %xmm15, %xmm6;
+	vpxor %xmm7, %xmm15, %xmm7;
+	vpxor %xmm8, %xmm15, %xmm8;
+	vpxor %xmm9, %xmm15, %xmm9;
+	vpxor %xmm10, %xmm15, %xmm10;
+	vpxor %xmm11, %xmm15, %xmm11;
+	vpxor %xmm12, %xmm15, %xmm12;
+	vpxor %xmm13, %xmm15, %xmm13;
+	vpxor %xmm14, %xmm15, %xmm14;
+	vpxor 15 * 16(%rax), %xmm15, %xmm15;
+
+	call __camellia_dec_blk16;
+
+	vpxor 0 * 16(%rsi), %xmm7, %xmm7;
+	vpxor 1 * 16(%rsi), %xmm6, %xmm6;
+	vpxor 2 * 16(%rsi), %xmm5, %xmm5;
+	vpxor 3 * 16(%rsi), %xmm4, %xmm4;
+	vpxor 4 * 16(%rsi), %xmm3, %xmm3;
+	vpxor 5 * 16(%rsi), %xmm2, %xmm2;
+	vpxor 6 * 16(%rsi), %xmm1, %xmm1;
+	vpxor 7 * 16(%rsi), %xmm0, %xmm0;
+	vmovdqu %xmm7, (7 * 16)(%rax);
+	vpxor 8 * 16(%rsi), %xmm15, %xmm15;
+	vpxor 9 * 16(%rsi), %xmm14, %xmm14;
+	vpxor 10 * 16(%rsi), %xmm13, %xmm13;
+	vpxor 11 * 16(%rsi), %xmm12, %xmm12;
+	vpxor 12 * 16(%rsi), %xmm11, %xmm11;
+	vpxor 13 * 16(%rsi), %xmm10, %xmm10;
+	vpxor 14 * 16(%rsi), %xmm9, %xmm9;
+	vpxor 15 * 16(%rsi), %xmm8, %xmm8;
+
+	/* Checksum_i = Checksum_{i-1} xor P_i  */
+
+	vpxor (%r10), %xmm7, %xmm7;
+	vpxor %xmm6, %xmm7, %xmm7;
+	vpxor %xmm5, %xmm7, %xmm7;
+	vpxor %xmm4, %xmm7, %xmm7;
+	vpxor %xmm3, %xmm7, %xmm7;
+	vpxor %xmm2, %xmm7, %xmm7;
+	vpxor %xmm1, %xmm7, %xmm7;
+	vpxor %xmm0, %xmm7, %xmm7;
+	vpxor %xmm15, %xmm7, %xmm7;
+	vpxor %xmm14, %xmm7, %xmm7;
+	vpxor %xmm13, %xmm7, %xmm7;
+	vpxor %xmm12, %xmm7, %xmm7;
+	vpxor %xmm11, %xmm7, %xmm7;
+	vpxor %xmm10, %xmm7, %xmm7;
+	vpxor %xmm9, %xmm7, %xmm7;
+	vpxor %xmm8, %xmm7, %xmm7;
+	vmovdqu %xmm7, (%r10);
+	vmovdqu (7 * 16)(%rax), %xmm7;
+
+	write_output(%xmm7, %xmm6, %xmm5, %xmm4, %xmm3, %xmm2, %xmm1, %xmm0,
+		     %xmm15, %xmm14, %xmm13, %xmm12, %xmm11, %xmm10, %xmm9,
+		     %xmm8, %rsi);
+
+	vzeroall;
+
+	movq (16 * 16 + 0 * 8)(%rax), %r10;
+	movq (16 * 16 + 1 * 8)(%rax), %r11;
+	movq (16 * 16 + 2 * 8)(%rax), %r12;
+	movq (16 * 16 + 3 * 8)(%rax), %r13;
+
+	leave;
+	ret;
+ELF(.size _gcry_camellia_aesni_avx_ocb_dec,.-_gcry_camellia_aesni_avx_ocb_dec;)
+
+.align 8
+.globl _gcry_camellia_aesni_avx_ocb_auth
+ELF(.type   _gcry_camellia_aesni_avx_ocb_auth,@function;)
+
+_gcry_camellia_aesni_avx_ocb_auth:
+	/* input:
+	 *	%rdi: ctx, CTX
+	 *	%rsi: abuf (16 blocks)
+	 *	%rdx: offset
+	 *	%rcx: checksum
+	 *	%r8 : L pointers (void *L[16])
+	 */
+
+	pushq %rbp;
+	movq %rsp, %rbp;
+
+	vzeroupper;
+
+	subq $(16 * 16 + 4 * 8), %rsp;
+	andq $~31, %rsp;
+	movq %rsp, %rax;
+
+	movq %r10, (16 * 16 + 0 * 8)(%rax);
+	movq %r11, (16 * 16 + 1 * 8)(%rax);
+	movq %r12, (16 * 16 + 2 * 8)(%rax);
+	movq %r13, (16 * 16 + 3 * 8)(%rax);
+
+	vmovdqu (%rdx), %xmm15;
+
+	/* Offset_i = Offset_{i-1} xor L_{ntz(i)} */
+	/* Sum_i = Sum_{i-1} xor ENCIPHER(K, A_i xor Offset_i)  */
+
+#define OCB_INPUT(n, lreg, xreg) \
+	  vmovdqu (n * 16)(%rsi), xreg; \
+	  vpxor (lreg), %xmm15, %xmm15; \
+	  vpxor xreg, %xmm15, xreg;
+
+	movq (0 * 8)(%r8), %r10;
+	movq (1 * 8)(%r8), %r11;
+	movq (2 * 8)(%r8), %r12;
+	movq (3 * 8)(%r8), %r13;
+	OCB_INPUT(0, %r10, %xmm0);
+	vmovdqu %xmm0, (15 * 16)(%rax);
+	OCB_INPUT(1, %r11, %xmm14);
+	OCB_INPUT(2, %r12, %xmm13);
+	OCB_INPUT(3, %r13, %xmm12);
+	movq (4 * 8)(%r8), %r10;
+	movq (5 * 8)(%r8), %r11;
+	movq (6 * 8)(%r8), %r12;
+	movq (7 * 8)(%r8), %r13;
+	OCB_INPUT(4, %r10, %xmm11);
+	OCB_INPUT(5, %r11, %xmm10);
+	OCB_INPUT(6, %r12, %xmm9);
+	OCB_INPUT(7, %r13, %xmm8);
+	movq (8 * 8)(%r8), %r10;
+	movq (9 * 8)(%r8), %r11;
+	movq (10 * 8)(%r8), %r12;
+	movq (11 * 8)(%r8), %r13;
+	OCB_INPUT(8, %r10, %xmm7);
+	OCB_INPUT(9, %r11, %xmm6);
+	OCB_INPUT(10, %r12, %xmm5);
+	OCB_INPUT(11, %r13, %xmm4);
+	movq (12 * 8)(%r8), %r10;
+	movq (13 * 8)(%r8), %r11;
+	movq (14 * 8)(%r8), %r12;
+	movq (15 * 8)(%r8), %r13;
+	OCB_INPUT(12, %r10, %xmm3);
+	OCB_INPUT(13, %r11, %xmm2);
+	OCB_INPUT(14, %r12, %xmm1);
+	OCB_INPUT(15, %r13, %xmm0);
+#undef OCB_INPUT
+
+	vmovdqu %xmm15, (%rdx);
+
+	movq %rcx, %r10;
+
+	/* inpack16_pre: */
+	vmovq (key_table)(CTX), %xmm15;
+	vpshufb .Lpack_bswap RIP, %xmm15, %xmm15;
+	vpxor %xmm0, %xmm15, %xmm0;
+	vpxor %xmm1, %xmm15, %xmm1;
+	vpxor %xmm2, %xmm15, %xmm2;
+	vpxor %xmm3, %xmm15, %xmm3;
+	vpxor %xmm4, %xmm15, %xmm4;
+	vpxor %xmm5, %xmm15, %xmm5;
+	vpxor %xmm6, %xmm15, %xmm6;
+	vpxor %xmm7, %xmm15, %xmm7;
+	vpxor %xmm8, %xmm15, %xmm8;
+	vpxor %xmm9, %xmm15, %xmm9;
+	vpxor %xmm10, %xmm15, %xmm10;
+	vpxor %xmm11, %xmm15, %xmm11;
+	vpxor %xmm12, %xmm15, %xmm12;
+	vpxor %xmm13, %xmm15, %xmm13;
+	vpxor %xmm14, %xmm15, %xmm14;
+	vpxor 15 * 16(%rax), %xmm15, %xmm15;
+
+	call __camellia_enc_blk16;
+
+	vpxor %xmm7, %xmm6, %xmm6;
+	vpxor %xmm5, %xmm4, %xmm4;
+	vpxor %xmm3, %xmm2, %xmm2;
+	vpxor %xmm1, %xmm0, %xmm0;
+	vpxor %xmm15, %xmm14, %xmm14;
+	vpxor %xmm13, %xmm12, %xmm12;
+	vpxor %xmm11, %xmm10, %xmm10;
+	vpxor %xmm9, %xmm8, %xmm8;
+
+	vpxor %xmm6, %xmm4, %xmm4;
+	vpxor %xmm2, %xmm0, %xmm0;
+	vpxor %xmm14, %xmm12, %xmm12;
+	vpxor %xmm10, %xmm8, %xmm8;
+
+	vpxor %xmm4, %xmm0, %xmm0;
+	vpxor %xmm12, %xmm8, %xmm8;
+
+	vpxor %xmm0, %xmm8, %xmm0;
+	vpxor (%r10), %xmm0, %xmm0;
+	vmovdqu %xmm0, (%r10);
+
+	vzeroall;
+
+	movq (16 * 16 + 0 * 8)(%rax), %r10;
+	movq (16 * 16 + 1 * 8)(%rax), %r11;
+	movq (16 * 16 + 2 * 8)(%rax), %r12;
+	movq (16 * 16 + 3 * 8)(%rax), %r13;
+
+	leave;
+	ret;
+ELF(.size _gcry_camellia_aesni_avx_ocb_auth,.-_gcry_camellia_aesni_avx_ocb_auth;)
+
 /*
  * IN:
  *  ab: 64-bit AB state
diff --git a/cipher/camellia-aesni-avx2-amd64.S b/cipher/camellia-aesni-avx2-amd64.S
index a3fa229d..26381df0 100644
--- a/cipher/camellia-aesni-avx2-amd64.S
+++ b/cipher/camellia-aesni-avx2-amd64.S
@@ -1,6 +1,6 @@
 /* camellia-avx2-aesni-amd64.S  -  AES-NI/AVX2 implementation of Camellia cipher
  *
- * Copyright (C) 2013 Jussi Kivilinna <jussi.kivilinna@iki.fi>
+ * Copyright (C) 2013-2015 Jussi Kivilinna <jussi.kivilinna@iki.fi>
  *
  * This file is part of Libgcrypt.
  *
@@ -1127,8 +1127,8 @@ ELF(.type   _gcry_camellia_aesni_avx2_cbc_dec,@function;)
 _gcry_camellia_aesni_avx2_cbc_dec:
 	/* input:
 	 *	%rdi: ctx, CTX
-	 *	%rsi: dst (16 blocks)
-	 *	%rdx: src (16 blocks)
+	 *	%rsi: dst (32 blocks)
+	 *	%rdx: src (32 blocks)
 	 *	%rcx: iv
 	 */
 
@@ -1199,8 +1199,8 @@ ELF(.type   _gcry_camellia_aesni_avx2_cfb_dec,@function;)
 _gcry_camellia_aesni_avx2_cfb_dec:
 	/* input:
 	 *	%rdi: ctx, CTX
-	 *	%rsi: dst (16 blocks)
-	 *	%rdx: src (16 blocks)
+	 *	%rsi: dst (32 blocks)
+	 *	%rdx: src (32 blocks)
 	 *	%rcx: iv
 	 */
 
@@ -1266,5 +1266,498 @@ _gcry_camellia_aesni_avx2_cfb_dec:
 	ret;
 ELF(.size _gcry_camellia_aesni_avx2_cfb_dec,.-_gcry_camellia_aesni_avx2_cfb_dec;)
 
+.align 8
+.globl _gcry_camellia_aesni_avx2_ocb_enc
+ELF(.type   _gcry_camellia_aesni_avx2_ocb_enc,@function;)
+
+_gcry_camellia_aesni_avx2_ocb_enc:
+	/* input:
+	 *	%rdi: ctx, CTX
+	 *	%rsi: dst (32 blocks)
+	 *	%rdx: src (32 blocks)
+	 *	%rcx: offset
+	 *	%r8 : checksum
+	 *	%r9 : L pointers (void *L[32])
+	 */
+
+	pushq %rbp;
+	movq %rsp, %rbp;
+
+	vzeroupper;
+
+	subq $(16 * 32 + 4 * 8), %rsp;
+	andq $~63, %rsp;
+	movq %rsp, %rax;
+
+	movq %r10, (16 * 32 + 0 * 8)(%rax);
+	movq %r11, (16 * 32 + 1 * 8)(%rax);
+	movq %r12, (16 * 32 + 2 * 8)(%rax);
+	movq %r13, (16 * 32 + 3 * 8)(%rax);
+
+	vmovdqu (%rcx), %xmm14;
+	vmovdqu (%r8), %xmm13;
+
+	/* Offset_i = Offset_{i-1} xor L_{ntz(i)} */
+	/* Checksum_i = Checksum_{i-1} xor P_i  */
+	/* C_i = Offset_i xor ENCIPHER(K, P_i xor Offset_i)  */
+
+#define OCB_INPUT(n, l0reg, l1reg, yreg) \
+	  vmovdqu (n * 32)(%rdx), yreg; \
+	  vpxor (l0reg), %xmm14, %xmm15; \
+	  vpxor (l1reg), %xmm15, %xmm14; \
+	  vinserti128 $1, %xmm14, %ymm15, %ymm15; \
+	  vpxor yreg, %ymm13, %ymm13; \
+	  vpxor yreg, %ymm15, yreg; \
+	  vmovdqu %ymm15, (n * 32)(%rsi);
+
+	movq (0 * 8)(%r9), %r10;
+	movq (1 * 8)(%r9), %r11;
+	movq (2 * 8)(%r9), %r12;
+	movq (3 * 8)(%r9), %r13;
+	OCB_INPUT(0, %r10, %r11, %ymm0);
+	vmovdqu %ymm0, (15 * 32)(%rax);
+	OCB_INPUT(1, %r12, %r13, %ymm0);
+	vmovdqu %ymm0, (14 * 32)(%rax);
+	movq (4 * 8)(%r9), %r10;
+	movq (5 * 8)(%r9), %r11;
+	movq (6 * 8)(%r9), %r12;
+	movq (7 * 8)(%r9), %r13;
+	OCB_INPUT(2, %r10, %r11, %ymm0);
+	vmovdqu %ymm0, (13 * 32)(%rax);
+	OCB_INPUT(3, %r12, %r13, %ymm12);
+	movq (8 * 8)(%r9), %r10;
+	movq (9 * 8)(%r9), %r11;
+	movq (10 * 8)(%r9), %r12;
+	movq (11 * 8)(%r9), %r13;
+	OCB_INPUT(4, %r10, %r11, %ymm11);
+	OCB_INPUT(5, %r12, %r13, %ymm10);
+	movq (12 * 8)(%r9), %r10;
+	movq (13 * 8)(%r9), %r11;
+	movq (14 * 8)(%r9), %r12;
+	movq (15 * 8)(%r9), %r13;
+	OCB_INPUT(6, %r10, %r11, %ymm9);
+	OCB_INPUT(7, %r12, %r13, %ymm8);
+	movq (16 * 8)(%r9), %r10;
+	movq (17 * 8)(%r9), %r11;
+	movq (18 * 8)(%r9), %r12;
+	movq (19 * 8)(%r9), %r13;
+	OCB_INPUT(8, %r10, %r11, %ymm7);
+	OCB_INPUT(9, %r12, %r13, %ymm6);
+	movq (20 * 8)(%r9), %r10;
+	movq (21 * 8)(%r9), %r11;
+	movq (22 * 8)(%r9), %r12;
+	movq (23 * 8)(%r9), %r13;
+	OCB_INPUT(10, %r10, %r11, %ymm5);
+	OCB_INPUT(11, %r12, %r13, %ymm4);
+	movq (24 * 8)(%r9), %r10;
+	movq (25 * 8)(%r9), %r11;
+	movq (26 * 8)(%r9), %r12;
+	movq (27 * 8)(%r9), %r13;
+	OCB_INPUT(12, %r10, %r11, %ymm3);
+	OCB_INPUT(13, %r12, %r13, %ymm2);
+	movq (28 * 8)(%r9), %r10;
+	movq (29 * 8)(%r9), %r11;
+	movq (30 * 8)(%r9), %r12;
+	movq (31 * 8)(%r9), %r13;
+	OCB_INPUT(14, %r10, %r11, %ymm1);
+	OCB_INPUT(15, %r12, %r13, %ymm0);
+#undef OCB_INPUT
+
+	vextracti128 $1, %ymm13, %xmm15;
+	vmovdqu %xmm14, (%rcx);
+	vpxor %xmm13, %xmm15, %xmm15;
+	vmovdqu %xmm15, (%r8);
+
+	/* inpack16_pre: */
+	vpbroadcastq (key_table)(CTX), %ymm15;
+	vpshufb .Lpack_bswap RIP, %ymm15, %ymm15;
+	vpxor %ymm0, %ymm15, %ymm0;
+	vpxor %ymm1, %ymm15, %ymm1;
+	vpxor %ymm2, %ymm15, %ymm2;
+	vpxor %ymm3, %ymm15, %ymm3;
+	vpxor %ymm4, %ymm15, %ymm4;
+	vpxor %ymm5, %ymm15, %ymm5;
+	vpxor %ymm6, %ymm15, %ymm6;
+	vpxor %ymm7, %ymm15, %ymm7;
+	vpxor %ymm8, %ymm15, %ymm8;
+	vpxor %ymm9, %ymm15, %ymm9;
+	vpxor %ymm10, %ymm15, %ymm10;
+	vpxor %ymm11, %ymm15, %ymm11;
+	vpxor %ymm12, %ymm15, %ymm12;
+	vpxor 13 * 32(%rax), %ymm15, %ymm13;
+	vpxor 14 * 32(%rax), %ymm15, %ymm14;
+	vpxor 15 * 32(%rax), %ymm15, %ymm15;
+
+	call __camellia_enc_blk32;
+
+	vpxor 0 * 32(%rsi), %ymm7, %ymm7;
+	vpxor 1 * 32(%rsi), %ymm6, %ymm6;
+	vpxor 2 * 32(%rsi), %ymm5, %ymm5;
+	vpxor 3 * 32(%rsi), %ymm4, %ymm4;
+	vpxor 4 * 32(%rsi), %ymm3, %ymm3;
+	vpxor 5 * 32(%rsi), %ymm2, %ymm2;
+	vpxor 6 * 32(%rsi), %ymm1, %ymm1;
+	vpxor 7 * 32(%rsi), %ymm0, %ymm0;
+	vpxor 8 * 32(%rsi), %ymm15, %ymm15;
+	vpxor 9 * 32(%rsi), %ymm14, %ymm14;
+	vpxor 10 * 32(%rsi), %ymm13, %ymm13;
+	vpxor 11 * 32(%rsi), %ymm12, %ymm12;
+	vpxor 12 * 32(%rsi), %ymm11, %ymm11;
+	vpxor 13 * 32(%rsi), %ymm10, %ymm10;
+	vpxor 14 * 32(%rsi), %ymm9, %ymm9;
+	vpxor 15 * 32(%rsi), %ymm8, %ymm8;
+
+	write_output(%ymm7, %ymm6, %ymm5, %ymm4, %ymm3, %ymm2, %ymm1, %ymm0,
+		     %ymm15, %ymm14, %ymm13, %ymm12, %ymm11, %ymm10, %ymm9,
+		     %ymm8, %rsi);
+
+	vzeroall;
+
+	movq (16 * 32 + 0 * 8)(%rax), %r10;
+	movq (16 * 32 + 1 * 8)(%rax), %r11;
+	movq (16 * 32 + 2 * 8)(%rax), %r12;
+	movq (16 * 32 + 3 * 8)(%rax), %r13;
+
+	leave;
+	ret;
+ELF(.size _gcry_camellia_aesni_avx2_ocb_enc,.-_gcry_camellia_aesni_avx2_ocb_enc;)
+
+.align 8
+.globl _gcry_camellia_aesni_avx2_ocb_dec
+ELF(.type   _gcry_camellia_aesni_avx2_ocb_dec,@function;)
+
+_gcry_camellia_aesni_avx2_ocb_dec:
+	/* input:
+	 *	%rdi: ctx, CTX
+	 *	%rsi: dst (32 blocks)
+	 *	%rdx: src (32 blocks)
+	 *	%rcx: offset
+	 *	%r8 : checksum
+	 *	%r9 : L pointers (void *L[32])
+	 */
+
+	pushq %rbp;
+	movq %rsp, %rbp;
+
+	vzeroupper;
+
+	subq $(16 * 32 + 4 * 8), %rsp;
+	andq $~63, %rsp;
+	movq %rsp, %rax;
+
+	movq %r10, (16 * 32 + 0 * 8)(%rax);
+	movq %r11, (16 * 32 + 1 * 8)(%rax);
+	movq %r12, (16 * 32 + 2 * 8)(%rax);
+	movq %r13, (16 * 32 + 3 * 8)(%rax);
+
+	vmovdqu (%rcx), %xmm14;
+
+	/* Offset_i = Offset_{i-1} xor L_{ntz(i)} */
+	/* P_i = Offset_i xor DECIPHER(K, C_i xor Offset_i)  */
+
+#define OCB_INPUT(n, l0reg, l1reg, yreg) \
+	  vmovdqu (n * 32)(%rdx), yreg; \
+	  vpxor (l0reg), %xmm14, %xmm15; \
+	  vpxor (l1reg), %xmm15, %xmm14; \
+	  vinserti128 $1, %xmm14, %ymm15, %ymm15; \
+	  vpxor yreg, %ymm15, yreg; \
+	  vmovdqu %ymm15, (n * 32)(%rsi);
+
+	movq (0 * 8)(%r9), %r10;
+	movq (1 * 8)(%r9), %r11;
+	movq (2 * 8)(%r9), %r12;
+	movq (3 * 8)(%r9), %r13;
+	OCB_INPUT(0, %r10, %r11, %ymm0);
+	vmovdqu %ymm0, (15 * 32)(%rax);
+	OCB_INPUT(1, %r12, %r13, %ymm0);
+	vmovdqu %ymm0, (14 * 32)(%rax);
+	movq (4 * 8)(%r9), %r10;
+	movq (5 * 8)(%r9), %r11;
+	movq (6 * 8)(%r9), %r12;
+	movq (7 * 8)(%r9), %r13;
+	OCB_INPUT(2, %r10, %r11, %ymm13);
+	OCB_INPUT(3, %r12, %r13, %ymm12);
+	movq (8 * 8)(%r9), %r10;
+	movq (9 * 8)(%r9), %r11;
+	movq (10 * 8)(%r9), %r12;
+	movq (11 * 8)(%r9), %r13;
+	OCB_INPUT(4, %r10, %r11, %ymm11);
+	OCB_INPUT(5, %r12, %r13, %ymm10);
+	movq (12 * 8)(%r9), %r10;
+	movq (13 * 8)(%r9), %r11;
+	movq (14 * 8)(%r9), %r12;
+	movq (15 * 8)(%r9), %r13;
+	OCB_INPUT(6, %r10, %r11, %ymm9);
+	OCB_INPUT(7, %r12, %r13, %ymm8);
+	movq (16 * 8)(%r9), %r10;
+	movq (17 * 8)(%r9), %r11;
+	movq (18 * 8)(%r9), %r12;
+	movq (19 * 8)(%r9), %r13;
+	OCB_INPUT(8, %r10, %r11, %ymm7);
+	OCB_INPUT(9, %r12, %r13, %ymm6);
+	movq (20 * 8)(%r9), %r10;
+	movq (21 * 8)(%r9), %r11;
+	movq (22 * 8)(%r9), %r12;
+	movq (23 * 8)(%r9), %r13;
+	OCB_INPUT(10, %r10, %r11, %ymm5);
+	OCB_INPUT(11, %r12, %r13, %ymm4);
+	movq (24 * 8)(%r9), %r10;
+	movq (25 * 8)(%r9), %r11;
+	movq (26 * 8)(%r9), %r12;
+	movq (27 * 8)(%r9), %r13;
+	OCB_INPUT(12, %r10, %r11, %ymm3);
+	OCB_INPUT(13, %r12, %r13, %ymm2);
+	movq (28 * 8)(%r9), %r10;
+	movq (29 * 8)(%r9), %r11;
+	movq (30 * 8)(%r9), %r12;
+	movq (31 * 8)(%r9), %r13;
+	OCB_INPUT(14, %r10, %r11, %ymm1);
+	OCB_INPUT(15, %r12, %r13, %ymm0);
+#undef OCB_INPUT
+
+	vmovdqu %xmm14, (%rcx);
+
+	movq %r8, %r10;
+
+	cmpl $128, key_bitlength(CTX);
+	movl $32, %r8d;
+	movl $24, %r9d;
+	cmovel %r9d, %r8d; /* max */
+
+	/* inpack16_pre: */
+	vpbroadcastq (key_table)(CTX, %r8, 8), %ymm15;
+	vpshufb .Lpack_bswap RIP, %ymm15, %ymm15;
+	vpxor %ymm0, %ymm15, %ymm0;
+	vpxor %ymm1, %ymm15, %ymm1;
+	vpxor %ymm2, %ymm15, %ymm2;
+	vpxor %ymm3, %ymm15, %ymm3;
+	vpxor %ymm4, %ymm15, %ymm4;
+	vpxor %ymm5, %ymm15, %ymm5;
+	vpxor %ymm6, %ymm15, %ymm6;
+	vpxor %ymm7, %ymm15, %ymm7;
+	vpxor %ymm8, %ymm15, %ymm8;
+	vpxor %ymm9, %ymm15, %ymm9;
+	vpxor %ymm10, %ymm15, %ymm10;
+	vpxor %ymm11, %ymm15, %ymm11;
+	vpxor %ymm12, %ymm15, %ymm12;
+	vpxor %ymm13, %ymm15, %ymm13;
+	vpxor 14 * 32(%rax), %ymm15, %ymm14;
+	vpxor 15 * 32(%rax), %ymm15, %ymm15;
+
+	call __camellia_dec_blk32;
+
+	vpxor 0 * 32(%rsi), %ymm7, %ymm7;
+	vpxor 1 * 32(%rsi), %ymm6, %ymm6;
+	vpxor 2 * 32(%rsi), %ymm5, %ymm5;
+	vpxor 3 * 32(%rsi), %ymm4, %ymm4;
+	vpxor 4 * 32(%rsi), %ymm3, %ymm3;
+	vpxor 5 * 32(%rsi), %ymm2, %ymm2;
+	vpxor 6 * 32(%rsi), %ymm1, %ymm1;
+	vpxor 7 * 32(%rsi), %ymm0, %ymm0;
+	vmovdqu %ymm7, (7 * 32)(%rax);
+	vmovdqu %ymm6, (6 * 32)(%rax);
+	vpxor 8 * 32(%rsi), %ymm15, %ymm15;
+	vpxor 9 * 32(%rsi), %ymm14, %ymm14;
+	vpxor 10 * 32(%rsi), %ymm13, %ymm13;
+	vpxor 11 * 32(%rsi), %ymm12, %ymm12;
+	vpxor 12 * 32(%rsi), %ymm11, %ymm11;
+	vpxor 13 * 32(%rsi), %ymm10, %ymm10;
+	vpxor 14 * 32(%rsi), %ymm9, %ymm9;
+	vpxor 15 * 32(%rsi), %ymm8, %ymm8;
+
+	/* Checksum_i = Checksum_{i-1} xor P_i  */
+
+	vpxor %ymm5, %ymm7, %ymm7;
+	vpxor %ymm4, %ymm6, %ymm6;
+	vpxor %ymm3, %ymm7, %ymm7;
+	vpxor %ymm2, %ymm6, %ymm6;
+	vpxor %ymm1, %ymm7, %ymm7;
+	vpxor %ymm0, %ymm6, %ymm6;
+	vpxor %ymm15, %ymm7, %ymm7;
+	vpxor %ymm14, %ymm6, %ymm6;
+	vpxor %ymm13, %ymm7, %ymm7;
+	vpxor %ymm12, %ymm6, %ymm6;
+	vpxor %ymm11, %ymm7, %ymm7;
+	vpxor %ymm10, %ymm6, %ymm6;
+	vpxor %ymm9, %ymm7, %ymm7;
+	vpxor %ymm8, %ymm6, %ymm6;
+	vpxor %ymm7, %ymm6, %ymm7;
+
+	vextracti128 $1, %ymm7, %xmm6;
+	vpxor %xmm6, %xmm7, %xmm7;
+	vpxor (%r10), %xmm7, %xmm7;
+	vmovdqu %xmm7, (%r10);
+
+	vmovdqu 7 * 32(%rax), %ymm7;
+	vmovdqu 6 * 32(%rax), %ymm6;
+
+	write_output(%ymm7, %ymm6, %ymm5, %ymm4, %ymm3, %ymm2, %ymm1, %ymm0,
+		     %ymm15, %ymm14, %ymm13, %ymm12, %ymm11, %ymm10, %ymm9,
+		     %ymm8, %rsi);
+
+	vzeroall;
+
+	movq (16 * 32 + 0 * 8)(%rax), %r10;
+	movq (16 * 32 + 1 * 8)(%rax), %r11;
+	movq (16 * 32 + 2 * 8)(%rax), %r12;
+	movq (16 * 32 + 3 * 8)(%rax), %r13;
+
+	leave;
+	ret;
+ELF(.size _gcry_camellia_aesni_avx2_ocb_dec,.-_gcry_camellia_aesni_avx2_ocb_dec;)
+
+.align 8
+.globl _gcry_camellia_aesni_avx2_ocb_auth
+ELF(.type   _gcry_camellia_aesni_avx2_ocb_auth,@function;)
+
+_gcry_camellia_aesni_avx2_ocb_auth:
+	/* input:
+	 *	%rdi: ctx, CTX
+	 *	%rsi: abuf (16 blocks)
+	 *	%rdx: offset
+	 *	%rcx: checksum
+	 *	%r8 : L pointers (void *L[16])
+	 */
+
+	pushq %rbp;
+	movq %rsp, %rbp;
+
+	vzeroupper;
+
+	subq $(16 * 32 + 4 * 8), %rsp;
+	andq $~63, %rsp;
+	movq %rsp, %rax;
+
+	movq %r10, (16 * 32 + 0 * 8)(%rax);
+	movq %r11, (16 * 32 + 1 * 8)(%rax);
+	movq %r12, (16 * 32 + 2 * 8)(%rax);
+	movq %r13, (16 * 32 + 3 * 8)(%rax);
+
+	vmovdqu (%rdx), %xmm14;
+
+	/* Offset_i = Offset_{i-1} xor L_{ntz(i)} */
+	/* Checksum_i = Checksum_{i-1} xor P_i  */
+	/* C_i = Offset_i xor ENCIPHER(K, P_i xor Offset_i)  */
+
+#define OCB_INPUT(n, l0reg, l1reg, yreg) \
+	  vmovdqu (n * 32)(%rsi), yreg; \
+	  vpxor (l0reg), %xmm14, %xmm15; \
+	  vpxor (l1reg), %xmm15, %xmm14; \
+	  vinserti128 $1, %xmm14, %ymm15, %ymm15; \
+	  vpxor yreg, %ymm15, yreg;
+
+	movq (0 * 8)(%r8), %r10;
+	movq (1 * 8)(%r8), %r11;
+	movq (2 * 8)(%r8), %r12;
+	movq (3 * 8)(%r8), %r13;
+	OCB_INPUT(0, %r10, %r11, %ymm0);
+	vmovdqu %ymm0, (15 * 32)(%rax);
+	OCB_INPUT(1, %r12, %r13, %ymm0);
+	vmovdqu %ymm0, (14 * 32)(%rax);
+	movq (4 * 8)(%r8), %r10;
+	movq (5 * 8)(%r8), %r11;
+	movq (6 * 8)(%r8), %r12;
+	movq (7 * 8)(%r8), %r13;
+	OCB_INPUT(2, %r10, %r11, %ymm13);
+	OCB_INPUT(3, %r12, %r13, %ymm12);
+	movq (8 * 8)(%r8), %r10;
+	movq (9 * 8)(%r8), %r11;
+	movq (10 * 8)(%r8), %r12;
+	movq (11 * 8)(%r8), %r13;
+	OCB_INPUT(4, %r10, %r11, %ymm11);
+	OCB_INPUT(5, %r12, %r13, %ymm10);
+	movq (12 * 8)(%r8), %r10;
+	movq (13 * 8)(%r8), %r11;
+	movq (14 * 8)(%r8), %r12;
+	movq (15 * 8)(%r8), %r13;
+	OCB_INPUT(6, %r10, %r11, %ymm9);
+	OCB_INPUT(7, %r12, %r13, %ymm8);
+	movq (16 * 8)(%r8), %r10;
+	movq (17 * 8)(%r8), %r11;
+	movq (18 * 8)(%r8), %r12;
+	movq (19 * 8)(%r8), %r13;
+	OCB_INPUT(8, %r10, %r11, %ymm7);
+	OCB_INPUT(9, %r12, %r13, %ymm6);
+	movq (20 * 8)(%r8), %r10;
+	movq (21 * 8)(%r8), %r11;
+	movq (22 * 8)(%r8), %r12;
+	movq (23 * 8)(%r8), %r13;
+	OCB_INPUT(10, %r10, %r11, %ymm5);
+	OCB_INPUT(11, %r12, %r13, %ymm4);
+	movq (24 * 8)(%r8), %r10;
+	movq (25 * 8)(%r8), %r11;
+	movq (26 * 8)(%r8), %r12;
+	movq (27 * 8)(%r8), %r13;
+	OCB_INPUT(12, %r10, %r11, %ymm3);
+	OCB_INPUT(13, %r12, %r13, %ymm2);
+	movq (28 * 8)(%r8), %r10;
+	movq (29 * 8)(%r8), %r11;
+	movq (30 * 8)(%r8), %r12;
+	movq (31 * 8)(%r8), %r13;
+	OCB_INPUT(14, %r10, %r11, %ymm1);
+	OCB_INPUT(15, %r12, %r13, %ymm0);
+#undef OCB_INPUT
+
+	vmovdqu %xmm14, (%rdx);
+
+	movq %rcx, %r10;
+
+	/* inpack16_pre: */
+	vpbroadcastq (key_table)(CTX), %ymm15;
+	vpshufb .Lpack_bswap RIP, %ymm15, %ymm15;
+	vpxor %ymm0, %ymm15, %ymm0;
+	vpxor %ymm1, %ymm15, %ymm1;
+	vpxor %ymm2, %ymm15, %ymm2;
+	vpxor %ymm3, %ymm15, %ymm3;
+	vpxor %ymm4, %ymm15, %ymm4;
+	vpxor %ymm5, %ymm15, %ymm5;
+	vpxor %ymm6, %ymm15, %ymm6;
+	vpxor %ymm7, %ymm15, %ymm7;
+	vpxor %ymm8, %ymm15, %ymm8;
+	vpxor %ymm9, %ymm15, %ymm9;
+	vpxor %ymm10, %ymm15, %ymm10;
+	vpxor %ymm11, %ymm15, %ymm11;
+	vpxor %ymm12, %ymm15, %ymm12;
+	vpxor %ymm13, %ymm15, %ymm13;
+	vpxor 14 * 32(%rax), %ymm15, %ymm14;
+	vpxor 15 * 32(%rax), %ymm15, %ymm15;
+
+	call __camellia_enc_blk32;
+
+	vpxor %ymm7, %ymm6, %ymm6;
+	vpxor %ymm5, %ymm4, %ymm4;
+	vpxor %ymm3, %ymm2, %ymm2;
+	vpxor %ymm1, %ymm0, %ymm0;
+	vpxor %ymm15, %ymm14, %ymm14;
+	vpxor %ymm13, %ymm12, %ymm12;
+	vpxor %ymm11, %ymm10, %ymm10;
+	vpxor %ymm9, %ymm8, %ymm8;
+
+	vpxor %ymm6, %ymm4, %ymm4;
+	vpxor %ymm2, %ymm0, %ymm0;
+	vpxor %ymm14, %ymm12, %ymm12;
+	vpxor %ymm10, %ymm8, %ymm8;
+
+	vpxor %ymm4, %ymm0, %ymm0;
+	vpxor %ymm12, %ymm8, %ymm8;
+
+	vpxor %ymm0, %ymm8, %ymm0;
+
+	vextracti128 $1, %ymm0, %xmm1;
+	vpxor (%r10), %xmm0, %xmm0;
+	vpxor %xmm0, %xmm1, %xmm0;
+	vmovdqu %xmm0, (%r10);
+
+	vzeroall;
+
+	movq (16 * 32 + 0 * 8)(%rax), %r10;
+	movq (16 * 32 + 1 * 8)(%rax), %r11;
+	movq (16 * 32 + 2 * 8)(%rax), %r12;
+	movq (16 * 32 + 3 * 8)(%rax), %r13;
+
+	leave;
+	ret;
+ELF(.size _gcry_camellia_aesni_avx2_ocb_auth,.-_gcry_camellia_aesni_avx2_ocb_auth;)
+
 #endif /*defined(ENABLE_AESNI_SUPPORT) && defined(ENABLE_AVX2_SUPPORT)*/
 #endif /*__x86_64*/
diff --git a/cipher/camellia-glue.c b/cipher/camellia-glue.c
index 50323218..197e1b39 100644
--- a/cipher/camellia-glue.c
+++ b/cipher/camellia-glue.c
@@ -63,6 +63,7 @@
 #include "cipher.h"
 #include "camellia.h"
 #include "bufhelp.h"
+#include "cipher-internal.h"
 #include "cipher-selftest.h"
 
 /* Helper macro to force alignment to 16 bytes.  */
@@ -135,6 +136,26 @@ extern void _gcry_camellia_aesni_avx_cfb_dec(CAMELLIA_context *ctx,
 					     const unsigned char *in,
 					     unsigned char *iv) ASM_FUNC_ABI;
 
+extern void _gcry_camellia_aesni_avx_ocb_enc(CAMELLIA_context *ctx,
+					     unsigned char *out,
+					     const unsigned char *in,
+					     unsigned char *offset,
+					     unsigned char *checksum,
+					     const void *Ls[16]) ASM_FUNC_ABI;
+
+extern void _gcry_camellia_aesni_avx_ocb_dec(CAMELLIA_context *ctx,
+					     unsigned char *out,
+					     const unsigned char *in,
+					     unsigned char *offset,
+					     unsigned char *checksum,
+					     const void *Ls[16]) ASM_FUNC_ABI;
+
+extern void _gcry_camellia_aesni_avx_ocb_auth(CAMELLIA_context *ctx,
+					     const unsigned char *abuf,
+					     unsigned char *offset,
+					     unsigned char *checksum,
+					     const void *Ls[16]) ASM_FUNC_ABI;
+
 extern void _gcry_camellia_aesni_avx_keygen(CAMELLIA_context *ctx,
 					    const unsigned char *key,
 					    unsigned int keylen) ASM_FUNC_ABI;
@@ -158,6 +179,26 @@ extern void _gcry_camellia_aesni_avx2_cfb_dec(CAMELLIA_context *ctx,
 					      unsigned char *out,
 					      const unsigned char *in,
 					      unsigned char *iv) ASM_FUNC_ABI;
+
+extern void _gcry_camellia_aesni_avx2_ocb_enc(CAMELLIA_context *ctx,
+					      unsigned char *out,
+					      const unsigned char *in,
+					      unsigned char *offset,
+					      unsigned char *checksum,
+					      const void *Ls[32]) ASM_FUNC_ABI;
+
+extern void _gcry_camellia_aesni_avx2_ocb_dec(CAMELLIA_context *ctx,
+					      unsigned char *out,
+					      const unsigned char *in,
+					      unsigned char *offset,
+					      unsigned char *checksum,
+					      const void *Ls[32]) ASM_FUNC_ABI;
+
+extern void _gcry_camellia_aesni_avx2_ocb_auth(CAMELLIA_context *ctx,
+					       const unsigned char *abuf,
+					       unsigned char *offset,
+					       unsigned char *checksum,
+					       const void *Ls[32]) ASM_FUNC_ABI;
 #endif
 
 static const char *selftest(void);
@@ -563,6 +604,294 @@ _gcry_camellia_cfb_dec(void *context, unsigned char *iv,
   _gcry_burn_stack(burn_stack_depth);
 }
 
+static inline const unsigned char *
+get_l (gcry_cipher_hd_t c, unsigned char *l_tmp, u64 i)
+{
+  unsigned int ntz = _gcry_ctz64 (i);
+
+  if (ntz < OCB_L_TABLE_SIZE)
+      return c->u_mode.ocb.L[ntz];
+  else
+      return _gcry_cipher_ocb_get_l (c, l_tmp, i);
+}
+
+/* Bulk encryption/decryption of complete blocks in OCB mode. */
+void
+_gcry_camellia_ocb_crypt (gcry_cipher_hd_t c, void *outbuf_arg,
+			  const void *inbuf_arg, size_t nblocks, int encrypt)
+{
+  CAMELLIA_context *ctx = (void *)&c->context.c;
+  unsigned char *outbuf = outbuf_arg;
+  const unsigned char *inbuf = inbuf_arg;
+  unsigned char l_tmp[CAMELLIA_BLOCK_SIZE];
+  const unsigned char *l;
+  int burn_stack_depth;
+  u64 blkn = c->u_mode.ocb.data_nblocks;
+
+  burn_stack_depth = encrypt ? CAMELLIA_encrypt_stack_burn_size :
+			      CAMELLIA_decrypt_stack_burn_size;
+
+#ifdef USE_AESNI_AVX2
+  if (ctx->use_aesni_avx2)
+    {
+      int did_use_aesni_avx2 = 0;
+      const void *Ls[32];
+      int i;
+
+      /* Process data in 32 block chunks. */
+      while (nblocks >= 32)
+	{
+	  /* l_tmp will be used only every 65536-th block. */
+	  for (i = 0; i < 32; i += 4)
+	    {
+	      Ls[i + 0] = get_l(c, l_tmp, blkn + 1);
+	      Ls[i + 1] = get_l(c, l_tmp, blkn + 2);
+	      Ls[i + 2] = get_l(c, l_tmp, blkn + 3);
+	      Ls[i + 3] = get_l(c, l_tmp, blkn + 4);
+	      blkn += 4;
+	    }
+
+	  if (encrypt)
+	    _gcry_camellia_aesni_avx2_ocb_enc(ctx, outbuf, inbuf, c->u_iv.iv,
+					      c->u_ctr.ctr, Ls);
+	  else
+	    _gcry_camellia_aesni_avx2_ocb_dec(ctx, outbuf, inbuf, c->u_iv.iv,
+					      c->u_ctr.ctr, Ls);
+
+	  nblocks -= 32;
+	  outbuf += 32 * CAMELLIA_BLOCK_SIZE;
+	  inbuf  += 32 * CAMELLIA_BLOCK_SIZE;
+	  did_use_aesni_avx2 = 1;
+	}
+
+      if (did_use_aesni_avx2)
+	{
+	  int avx2_burn_stack_depth = 32 * CAMELLIA_BLOCK_SIZE +
+				      2 * sizeof(void *) + ASM_EXTRA_STACK;
+
+	  if (burn_stack_depth < avx2_burn_stack_depth)
+	    burn_stack_depth = avx2_burn_stack_depth;
+	}
+
+      /* Use generic code to handle smaller chunks... */
+    }
+#endif
+
+#ifdef USE_AESNI_AVX
+  if (ctx->use_aesni_avx)
+    {
+      int did_use_aesni_avx = 0;
+      const void *Ls[16];
+      int i;
+
+      /* Process data in 16 block chunks. */
+      while (nblocks >= 16)
+	{
+	  /* l_tmp will be used only every 65536-th block. */
+	  for (i = 0; i < 16; i += 4)
+	    {
+	      Ls[i + 0] = get_l(c, l_tmp, blkn + 1);
+	      Ls[i + 1] = get_l(c, l_tmp, blkn + 2);
+	      Ls[i + 2] = get_l(c, l_tmp, blkn + 3);
+	      Ls[i + 3] = get_l(c, l_tmp, blkn + 4);
+	      blkn += 4;
+	    }
+
+	  if (encrypt)
+	    _gcry_camellia_aesni_avx_ocb_enc(ctx, outbuf, inbuf, c->u_iv.iv,
+					    c->u_ctr.ctr, Ls);
+	  else
+	    _gcry_camellia_aesni_avx_ocb_dec(ctx, outbuf, inbuf, c->u_iv.iv,
+					    c->u_ctr.ctr, Ls);
+
+	  nblocks -= 16;
+	  outbuf += 16 * CAMELLIA_BLOCK_SIZE;
+	  inbuf  += 16 * CAMELLIA_BLOCK_SIZE;
+	  did_use_aesni_avx = 1;
+	}
+
+      if (did_use_aesni_avx)
+	{
+	  int avx_burn_stack_depth = 16 * CAMELLIA_BLOCK_SIZE +
+				      2 * sizeof(void *) + ASM_EXTRA_STACK;
+
+	  if (burn_stack_depth < avx_burn_stack_depth)
+	    burn_stack_depth = avx_burn_stack_depth;
+	}
+
+      /* Use generic code to handle smaller chunks... */
+    }
+#endif
+
+  if (encrypt)
+    {
+      for (; nblocks; nblocks--)
+	{
+	  l = get_l(c, l_tmp, ++blkn);
+
+	  /* Offset_i = Offset_{i-1} xor L_{ntz(i)} */
+	  buf_xor_1 (c->u_iv.iv, l, CAMELLIA_BLOCK_SIZE);
+	  buf_cpy (l_tmp, inbuf, CAMELLIA_BLOCK_SIZE);
+	  /* Checksum_i = Checksum_{i-1} xor P_i  */
+	  buf_xor_1 (c->u_ctr.ctr, l_tmp, CAMELLIA_BLOCK_SIZE);
+	  /* C_i = Offset_i xor ENCIPHER(K, P_i xor Offset_i)  */
+	  buf_xor_1 (l_tmp, c->u_iv.iv, CAMELLIA_BLOCK_SIZE);
+	  Camellia_EncryptBlock(ctx->keybitlength, l_tmp, ctx->keytable, l_tmp);
+	  buf_xor_1 (l_tmp, c->u_iv.iv, CAMELLIA_BLOCK_SIZE);
+	  buf_cpy (outbuf, l_tmp, CAMELLIA_BLOCK_SIZE);
+
+	  inbuf += CAMELLIA_BLOCK_SIZE;
+	  outbuf += CAMELLIA_BLOCK_SIZE;
+	}
+    }
+  else
+    {
+      for (; nblocks; nblocks--)
+	{
+	  l = get_l(c, l_tmp, ++blkn);
+
+	  /* Offset_i = Offset_{i-1} xor L_{ntz(i)} */
+	  buf_xor_1 (c->u_iv.iv, l, CAMELLIA_BLOCK_SIZE);
+	  buf_cpy (l_tmp, inbuf, CAMELLIA_BLOCK_SIZE);
+	  /* C_i = Offset_i xor ENCIPHER(K, P_i xor Offset_i)  */
+	  buf_xor_1 (l_tmp, c->u_iv.iv, CAMELLIA_BLOCK_SIZE);
+	  Camellia_DecryptBlock(ctx->keybitlength, l_tmp, ctx->keytable, l_tmp);
+	  buf_xor_1 (l_tmp, c->u_iv.iv, CAMELLIA_BLOCK_SIZE);
+	  /* Checksum_i = Checksum_{i-1} xor P_i  */
+	  buf_xor_1 (c->u_ctr.ctr, l_tmp, CAMELLIA_BLOCK_SIZE);
+	  buf_cpy (outbuf, l_tmp, CAMELLIA_BLOCK_SIZE);
+
+	  inbuf += CAMELLIA_BLOCK_SIZE;
+	  outbuf += CAMELLIA_BLOCK_SIZE;
+	}
+    }
+
+  c->u_mode.ocb.data_nblocks = blkn;
+
+  wipememory(&l_tmp, sizeof(l_tmp));
+
+  if (burn_stack_depth)
+    _gcry_burn_stack (burn_stack_depth + 4 * sizeof(void *));
+}
+
+/* Bulk authentication of complete blocks in OCB mode. */
+void
+_gcry_camellia_ocb_auth (gcry_cipher_hd_t c, const void *abuf_arg,
+			size_t nblocks)
+{
+  CAMELLIA_context *ctx = (void *)&c->context.c;
+  const unsigned char *abuf = abuf_arg;
+  unsigned char l_tmp[CAMELLIA_BLOCK_SIZE];
+  const unsigned char *l;
+  int burn_stack_depth;
+  u64 blkn = c->u_mode.ocb.aad_nblocks;
+
+  burn_stack_depth = CAMELLIA_encrypt_stack_burn_size;
+
+#ifdef USE_AESNI_AVX2
+  if (ctx->use_aesni_avx2)
+    {
+      int did_use_aesni_avx2 = 0;
+      const void *Ls[32];
+      int i;
+
+      /* Process data in 32 block chunks. */
+      while (nblocks >= 32)
+	{
+	  /* l_tmp will be used only every 65536-th block. */
+	  for (i = 0; i < 32; i += 4)
+	    {
+	      Ls[i + 0] = get_l(c, l_tmp, blkn + 1);
+	      Ls[i + 1] = get_l(c, l_tmp, blkn + 2);
+	      Ls[i + 2] = get_l(c, l_tmp, blkn + 3);
+	      Ls[i + 3] = get_l(c, l_tmp, blkn + 4);
+	      blkn += 4;
+	    }
+
+	  _gcry_camellia_aesni_avx2_ocb_auth(ctx, abuf, c->u_mode.ocb.aad_offset,
+					    c->u_mode.ocb.aad_sum, Ls);
+
+	  nblocks -= 32;
+	  abuf += 32 * CAMELLIA_BLOCK_SIZE;
+	  did_use_aesni_avx2 = 1;
+	}
+
+      if (did_use_aesni_avx2)
+	{
+	  int avx2_burn_stack_depth = 32 * CAMELLIA_BLOCK_SIZE +
+				      2 * sizeof(void *) + ASM_EXTRA_STACK;
+
+	  if (burn_stack_depth < avx2_burn_stack_depth)
+	    burn_stack_depth = avx2_burn_stack_depth;
+	}
+
+      /* Use generic code to handle smaller chunks... */
+    }
+#endif
+
+#ifdef USE_AESNI_AVX
+  if (ctx->use_aesni_avx)
+    {
+      int did_use_aesni_avx = 0;
+      const void *Ls[16];
+      int i;
+
+      /* Process data in 16 block chunks. */
+      while (nblocks >= 16)
+	{
+	  /* l_tmp will be used only every 65536-th block. */
+	  for (i = 0; i < 16; i += 4)
+	    {
+	      Ls[i + 0] = get_l(c, l_tmp, blkn + 1);
+	      Ls[i + 1] = get_l(c, l_tmp, blkn + 2);
+	      Ls[i + 2] = get_l(c, l_tmp, blkn + 3);
+	      Ls[i + 3] = get_l(c, l_tmp, blkn + 4);
+	      blkn += 4;
+	    }
+
+	  _gcry_camellia_aesni_avx_ocb_auth(ctx, abuf, c->u_mode.ocb.aad_offset,
+					    c->u_mode.ocb.aad_sum, Ls);
+
+	  nblocks -= 16;
+	  abuf += 16 * CAMELLIA_BLOCK_SIZE;
+	  did_use_aesni_avx = 1;
+	}
+
+      if (did_use_aesni_avx)
+	{
+	  int avx_burn_stack_depth = 16 * CAMELLIA_BLOCK_SIZE +
+				      2 * sizeof(void *) + ASM_EXTRA_STACK;
+
+	  if (burn_stack_depth < avx_burn_stack_depth)
+	    burn_stack_depth = avx_burn_stack_depth;
+	}
+
+      /* Use generic code to handle smaller chunks... */
+    }
+#endif
+
+  for (; nblocks; nblocks--)
+    {
+      l = get_l(c, l_tmp, ++blkn);
+
+      /* Offset_i = Offset_{i-1} xor L_{ntz(i)} */
+      buf_xor_1 (c->u_mode.ocb.aad_offset, l, CAMELLIA_BLOCK_SIZE);
+      /* Sum_i = Sum_{i-1} xor ENCIPHER(K, A_i xor Offset_i)  */
+      buf_xor (l_tmp, c->u_mode.ocb.aad_offset, abuf, CAMELLIA_BLOCK_SIZE);
+      Camellia_EncryptBlock(ctx->keybitlength, l_tmp, ctx->keytable, l_tmp);
+      buf_xor_1 (c->u_mode.ocb.aad_sum, l_tmp, CAMELLIA_BLOCK_SIZE);
+
+      abuf += CAMELLIA_BLOCK_SIZE;
+    }
+
+  c->u_mode.ocb.aad_nblocks = blkn;
+
+  wipememory(&l_tmp, sizeof(l_tmp));
+
+  if (burn_stack_depth)
+    _gcry_burn_stack (burn_stack_depth + 4 * sizeof(void *));
+}
+
 /* Run the self-tests for CAMELLIA-CTR-128, tests IV increment of bulk CTR
    encryption.  Returns NULL on success. */
 static const char*
diff --git a/cipher/cipher.c b/cipher/cipher.c
index 7a29824c..2d2b0ade 100644
--- a/cipher/cipher.c
+++ b/cipher/cipher.c
@@ -535,6 +535,8 @@ _gcry_cipher_open_internal (gcry_cipher_hd_t *handle,
               h->bulk.cbc_dec = _gcry_camellia_cbc_dec;
               h->bulk.cfb_dec = _gcry_camellia_cfb_dec;
               h->bulk.ctr_enc = _gcry_camellia_ctr_enc;
+              h->bulk.ocb_crypt = _gcry_camellia_ocb_crypt;
+              h->bulk.ocb_auth  = _gcry_camellia_ocb_auth;
               break;
 #endif /*USE_CAMELLIA*/
 #ifdef USE_DES