Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id A5A25C636D6 for ; Mon, 20 Feb 2023 20:42:51 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S232910AbjBTUmu (ORCPT ); Mon, 20 Feb 2023 15:42:50 -0500 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:33404 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S232864AbjBTUmq (ORCPT ); Mon, 20 Feb 2023 15:42:46 -0500 Received: from mx0a-001b2d01.pphosted.com (mx0a-001b2d01.pphosted.com [148.163.156.1]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id D58C6D536; Mon, 20 Feb 2023 12:42:42 -0800 (PST) Received: from pps.filterd (m0098404.ppops.net [127.0.0.1]) by mx0a-001b2d01.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 31KIVC0X013103; Mon, 20 Feb 2023 20:42:38 GMT DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=ibm.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-transfer-encoding; s=pp1; bh=ecxk4qUHhcfyDq4/UvgC3XUBzVVl89HOpzHCEVSbIK0=; b=FTncpLYYUhBWblby9Axk8gM/G4b4ZzS/+l94WEaiWQ/YBtUavk8I9dGRNFLYVHJ8mW3/ E6FjjwfmI4hH9/vwHdAeEYb8l9G9UoOQHqZBDmN97eu9xHk3jtxJ4zP5/ggdouLvZ/MJ F5Ub72RgBrylocSWTkzjCKkcWECUONmisp43+/EwLwh82gCRldefPA4cu1CgmgOJMlTK YytMbWjnU2tJm+pz0++1rqh/RjJ9Z1U2RNLgj8nBWiTvl4HzbOAOdOFP8K6WB70UfshG vfPqNtYwi1xU1bQvVmEF9OPdyUvgORKuIMOYQb7gZxJTr/FB3VANimOW6qJuKjgCBY7w xg== Received: from ppma01wdc.us.ibm.com (fd.55.37a9.ip4.static.sl-reverse.com [169.55.85.253]) by mx0a-001b2d01.pphosted.com (PPS) with ESMTPS id 3nvbudwrxc-1 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-GCM-SHA384 bits=256 verify=NOT); Mon, 20 Feb 2023 20:42:37 +0000 Received: from pps.filterd (ppma01wdc.us.ibm.com [127.0.0.1]) by ppma01wdc.us.ibm.com (8.17.1.19/8.17.1.19) with ESMTP id 31KJ2kKw015811; Mon, 20 Feb 2023 20:42:36 GMT Received: from smtprelay07.wdc07v.mail.ibm.com ([9.208.129.116]) by ppma01wdc.us.ibm.com (PPS) with ESMTPS id 3ntpa6q36u-1 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-GCM-SHA384 bits=256 verify=NOT); Mon, 20 Feb 2023 20:42:36 +0000 Received: from smtpav03.dal12v.mail.ibm.com (smtpav03.dal12v.mail.ibm.com [10.241.53.102]) by smtprelay07.wdc07v.mail.ibm.com (8.14.9/8.14.9/NCO v10.0) with ESMTP id 31KKgYpU6226644 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-GCM-SHA384 bits=256 verify=OK); Mon, 20 Feb 2023 20:42:34 GMT Received: from smtpav03.dal12v.mail.ibm.com (unknown [127.0.0.1]) by IMSVA (Postfix) with ESMTP id 3FEC758063; Mon, 20 Feb 2023 20:42:34 +0000 (GMT) Received: from smtpav03.dal12v.mail.ibm.com (unknown [127.0.0.1]) by IMSVA (Postfix) with ESMTP id EC45E5805A; Mon, 20 Feb 2023 20:42:33 +0000 (GMT) Received: from ltcden12-lp3.aus.stglabs.ibm.com (unknown [9.40.195.53]) by smtpav03.dal12v.mail.ibm.com (Postfix) with ESMTP; Mon, 20 Feb 2023 20:42:33 +0000 (GMT) From: Danny Tsen To: linux-crypto@vger.kernel.org Cc: herbert@gondor.apana.org.au, leitao@debian.org, nayna@linux.ibm.com, appro@cryptogams.org, linux-kernel@vger.kernel.org, ltcgcw@linux.vnet.ibm.com, dtsen@us.ibm.com, Danny Tsen Subject: [PATCH v4 3/6] An accelerated AES/GCM stitched implementation. Date: Mon, 20 Feb 2023 15:42:21 -0500 Message-Id: <20230220204224.4907-4-dtsen@linux.ibm.com> X-Mailer: git-send-email 2.31.1 In-Reply-To: <20230220204224.4907-1-dtsen@linux.ibm.com> References: <20230220204224.4907-1-dtsen@linux.ibm.com> MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-TM-AS-GCONF: 00 X-Proofpoint-GUID: BAmbSQ0WfWK9jDcJhyhJGaNJodwifudb X-Proofpoint-ORIG-GUID: BAmbSQ0WfWK9jDcJhyhJGaNJodwifudb X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.219,Aquarius:18.0.930,Hydra:6.0.562,FMLib:17.11.170.22 definitions=2023-02-20_17,2023-02-20_02,2023-02-09_01 X-Proofpoint-Spam-Details: rule=outbound_notspam policy=outbound score=0 phishscore=0 bulkscore=0 mlxlogscore=881 clxscore=1015 lowpriorityscore=0 priorityscore=1501 impostorscore=0 malwarescore=0 suspectscore=0 mlxscore=0 spamscore=0 adultscore=0 classifier=spam adjust=0 reason=mlx scancount=1 engine=8.12.0-2212070000 definitions=main-2302200190 Precedence: bulk List-ID: X-Mailing-List: linux-crypto@vger.kernel.org Improve overall performance of AES/GCM encrypt and decrypt operations for Power10 or later CPU. Signed-off-by: Danny Tsen --- arch/powerpc/crypto/aes-gcm-p10.S | 1521 +++++++++++++++++++++++++++++ 1 file changed, 1521 insertions(+) create mode 100644 arch/powerpc/crypto/aes-gcm-p10.S diff --git a/arch/powerpc/crypto/aes-gcm-p10.S b/arch/powerpc/crypto/aes-gcm-p10.S new file mode 100644 index 000000000000..a51f4b265308 --- /dev/null +++ b/arch/powerpc/crypto/aes-gcm-p10.S @@ -0,0 +1,1521 @@ +/* SPDX-License-Identifier: GPL-2.0-or-later */ + # + # Accelerated AES-GCM stitched implementation for ppc64le. + # + # Copyright 2022- IBM Inc. All rights reserved + # + #=================================================================================== + # Written by Danny Tsen + # + # GHASH is based on the Karatsuba multiplication method. + # + # Xi xor X1 + # + # X1 * H^4 + X2 * H^3 + x3 * H^2 + X4 * H = + # (X1.h * H4.h + xX.l * H4.l + X1 * H4) + + # (X2.h * H3.h + X2.l * H3.l + X2 * H3) + + # (X3.h * H2.h + X3.l * H2.l + X3 * H2) + + # (X4.h * H.h + X4.l * H.l + X4 * H) + # + # Xi = v0 + # H Poly = v2 + # Hash keys = v3 - v14 + # ( H.l, H, H.h) + # ( H^2.l, H^2, H^2.h) + # ( H^3.l, H^3, H^3.h) + # ( H^4.l, H^4, H^4.h) + # + # v30 is IV + # v31 - counter 1 + # + # AES used, + # vs0 - vs14 for round keys + # v15, v16, v17, v18, v19, v20, v21, v22 for 8 blocks (encrypted) + # + # This implementation uses stitched AES-GCM approach to improve overall performance. + # AES is implemented with 8x blocks and GHASH is using 2 4x blocks. + # + # =================================================================================== + # + +#include +#include + +.machine "any" +.text + + # 4x loops + # v15 - v18 - input states + # vs1 - vs9 - round keys + # +.macro Loop_aes_middle4x + xxlor 19+32, 1, 1 + xxlor 20+32, 2, 2 + xxlor 21+32, 3, 3 + xxlor 22+32, 4, 4 + + vcipher 15, 15, 19 + vcipher 16, 16, 19 + vcipher 17, 17, 19 + vcipher 18, 18, 19 + + vcipher 15, 15, 20 + vcipher 16, 16, 20 + vcipher 17, 17, 20 + vcipher 18, 18, 20 + + vcipher 15, 15, 21 + vcipher 16, 16, 21 + vcipher 17, 17, 21 + vcipher 18, 18, 21 + + vcipher 15, 15, 22 + vcipher 16, 16, 22 + vcipher 17, 17, 22 + vcipher 18, 18, 22 + + xxlor 19+32, 5, 5 + xxlor 20+32, 6, 6 + xxlor 21+32, 7, 7 + xxlor 22+32, 8, 8 + + vcipher 15, 15, 19 + vcipher 16, 16, 19 + vcipher 17, 17, 19 + vcipher 18, 18, 19 + + vcipher 15, 15, 20 + vcipher 16, 16, 20 + vcipher 17, 17, 20 + vcipher 18, 18, 20 + + vcipher 15, 15, 21 + vcipher 16, 16, 21 + vcipher 17, 17, 21 + vcipher 18, 18, 21 + + vcipher 15, 15, 22 + vcipher 16, 16, 22 + vcipher 17, 17, 22 + vcipher 18, 18, 22 + + xxlor 23+32, 9, 9 + vcipher 15, 15, 23 + vcipher 16, 16, 23 + vcipher 17, 17, 23 + vcipher 18, 18, 23 +.endm + + # 8x loops + # v15 - v22 - input states + # vs1 - vs9 - round keys + # +.macro Loop_aes_middle8x + xxlor 23+32, 1, 1 + xxlor 24+32, 2, 2 + xxlor 25+32, 3, 3 + xxlor 26+32, 4, 4 + + vcipher 15, 15, 23 + vcipher 16, 16, 23 + vcipher 17, 17, 23 + vcipher 18, 18, 23 + vcipher 19, 19, 23 + vcipher 20, 20, 23 + vcipher 21, 21, 23 + vcipher 22, 22, 23 + + vcipher 15, 15, 24 + vcipher 16, 16, 24 + vcipher 17, 17, 24 + vcipher 18, 18, 24 + vcipher 19, 19, 24 + vcipher 20, 20, 24 + vcipher 21, 21, 24 + vcipher 22, 22, 24 + + vcipher 15, 15, 25 + vcipher 16, 16, 25 + vcipher 17, 17, 25 + vcipher 18, 18, 25 + vcipher 19, 19, 25 + vcipher 20, 20, 25 + vcipher 21, 21, 25 + vcipher 22, 22, 25 + + vcipher 15, 15, 26 + vcipher 16, 16, 26 + vcipher 17, 17, 26 + vcipher 18, 18, 26 + vcipher 19, 19, 26 + vcipher 20, 20, 26 + vcipher 21, 21, 26 + vcipher 22, 22, 26 + + xxlor 23+32, 5, 5 + xxlor 24+32, 6, 6 + xxlor 25+32, 7, 7 + xxlor 26+32, 8, 8 + + vcipher 15, 15, 23 + vcipher 16, 16, 23 + vcipher 17, 17, 23 + vcipher 18, 18, 23 + vcipher 19, 19, 23 + vcipher 20, 20, 23 + vcipher 21, 21, 23 + vcipher 22, 22, 23 + + vcipher 15, 15, 24 + vcipher 16, 16, 24 + vcipher 17, 17, 24 + vcipher 18, 18, 24 + vcipher 19, 19, 24 + vcipher 20, 20, 24 + vcipher 21, 21, 24 + vcipher 22, 22, 24 + + vcipher 15, 15, 25 + vcipher 16, 16, 25 + vcipher 17, 17, 25 + vcipher 18, 18, 25 + vcipher 19, 19, 25 + vcipher 20, 20, 25 + vcipher 21, 21, 25 + vcipher 22, 22, 25 + + vcipher 15, 15, 26 + vcipher 16, 16, 26 + vcipher 17, 17, 26 + vcipher 18, 18, 26 + vcipher 19, 19, 26 + vcipher 20, 20, 26 + vcipher 21, 21, 26 + vcipher 22, 22, 26 + + xxlor 23+32, 9, 9 + vcipher 15, 15, 23 + vcipher 16, 16, 23 + vcipher 17, 17, 23 + vcipher 18, 18, 23 + vcipher 19, 19, 23 + vcipher 20, 20, 23 + vcipher 21, 21, 23 + vcipher 22, 22, 23 +.endm + +.macro Loop_aes_middle_1x + xxlor 19+32, 1, 1 + xxlor 20+32, 2, 2 + xxlor 21+32, 3, 3 + xxlor 22+32, 4, 4 + + vcipher 15, 15, 19 + vcipher 15, 15, 20 + vcipher 15, 15, 21 + vcipher 15, 15, 22 + + xxlor 19+32, 5, 5 + xxlor 20+32, 6, 6 + xxlor 21+32, 7, 7 + xxlor 22+32, 8, 8 + + vcipher 15, 15, 19 + vcipher 15, 15, 20 + vcipher 15, 15, 21 + vcipher 15, 15, 22 + + xxlor 19+32, 9, 9 + vcipher 15, 15, 19 +.endm + + # + # Compute 4x hash values based on Karatsuba method. + # +.macro ppc_aes_gcm_ghash + vxor 15, 15, 0 + + vpmsumd 23, 12, 15 # H4.L * X.L + vpmsumd 24, 9, 16 + vpmsumd 25, 6, 17 + vpmsumd 26, 3, 18 + + vxor 23, 23, 24 + vxor 23, 23, 25 + vxor 23, 23, 26 # L + + vpmsumd 24, 13, 15 # H4.L * X.H + H4.H * X.L + vpmsumd 25, 10, 16 # H3.L * X1.H + H3.H * X1.L + vpmsumd 26, 7, 17 + vpmsumd 27, 4, 18 + + vxor 24, 24, 25 + vxor 24, 24, 26 + vxor 24, 24, 27 # M + + # sum hash and reduction with H Poly + vpmsumd 28, 23, 2 # reduction + + vxor 29, 29, 29 + vsldoi 26, 24, 29, 8 # mL + vsldoi 29, 29, 24, 8 # mH + vxor 23, 23, 26 # mL + L + + vsldoi 23, 23, 23, 8 # swap + vxor 23, 23, 28 + + vpmsumd 24, 14, 15 # H4.H * X.H + vpmsumd 25, 11, 16 + vpmsumd 26, 8, 17 + vpmsumd 27, 5, 18 + + vxor 24, 24, 25 + vxor 24, 24, 26 + vxor 24, 24, 27 + + vxor 24, 24, 29 + + # sum hash and reduction with H Poly + vsldoi 27, 23, 23, 8 # swap + vpmsumd 23, 23, 2 + vxor 27, 27, 24 + vxor 23, 23, 27 + + xxlor 32, 23+32, 23+32 # update hash + +.endm + + # + # Combine two 4x ghash + # v15 - v22 - input blocks + # +.macro ppc_aes_gcm_ghash2_4x + # first 4x hash + vxor 15, 15, 0 # Xi + X + + vpmsumd 23, 12, 15 # H4.L * X.L + vpmsumd 24, 9, 16 + vpmsumd 25, 6, 17 + vpmsumd 26, 3, 18 + + vxor 23, 23, 24 + vxor 23, 23, 25 + vxor 23, 23, 26 # L + + vpmsumd 24, 13, 15 # H4.L * X.H + H4.H * X.L + vpmsumd 25, 10, 16 # H3.L * X1.H + H3.H * X1.L + vpmsumd 26, 7, 17 + vpmsumd 27, 4, 18 + + vxor 24, 24, 25 + vxor 24, 24, 26 + + # sum hash and reduction with H Poly + vpmsumd 28, 23, 2 # reduction + + vxor 29, 29, 29 + + vxor 24, 24, 27 # M + vsldoi 26, 24, 29, 8 # mL + vsldoi 29, 29, 24, 8 # mH + vxor 23, 23, 26 # mL + L + + vsldoi 23, 23, 23, 8 # swap + vxor 23, 23, 28 + + vpmsumd 24, 14, 15 # H4.H * X.H + vpmsumd 25, 11, 16 + vpmsumd 26, 8, 17 + vpmsumd 27, 5, 18 + + vxor 24, 24, 25 + vxor 24, 24, 26 + vxor 24, 24, 27 # H + + vxor 24, 24, 29 # H + mH + + # sum hash and reduction with H Poly + vsldoi 27, 23, 23, 8 # swap + vpmsumd 23, 23, 2 + vxor 27, 27, 24 + vxor 27, 23, 27 # 1st Xi + + # 2nd 4x hash + vpmsumd 24, 9, 20 + vpmsumd 25, 6, 21 + vpmsumd 26, 3, 22 + vxor 19, 19, 27 # Xi + X + vpmsumd 23, 12, 19 # H4.L * X.L + + vxor 23, 23, 24 + vxor 23, 23, 25 + vxor 23, 23, 26 # L + + vpmsumd 24, 13, 19 # H4.L * X.H + H4.H * X.L + vpmsumd 25, 10, 20 # H3.L * X1.H + H3.H * X1.L + vpmsumd 26, 7, 21 + vpmsumd 27, 4, 22 + + vxor 24, 24, 25 + vxor 24, 24, 26 + + # sum hash and reduction with H Poly + vpmsumd 28, 23, 2 # reduction + + vxor 29, 29, 29 + + vxor 24, 24, 27 # M + vsldoi 26, 24, 29, 8 # mL + vsldoi 29, 29, 24, 8 # mH + vxor 23, 23, 26 # mL + L + + vsldoi 23, 23, 23, 8 # swap + vxor 23, 23, 28 + + vpmsumd 24, 14, 19 # H4.H * X.H + vpmsumd 25, 11, 20 + vpmsumd 26, 8, 21 + vpmsumd 27, 5, 22 + + vxor 24, 24, 25 + vxor 24, 24, 26 + vxor 24, 24, 27 # H + + vxor 24, 24, 29 # H + mH + + # sum hash and reduction with H Poly + vsldoi 27, 23, 23, 8 # swap + vpmsumd 23, 23, 2 + vxor 27, 27, 24 + vxor 23, 23, 27 + + xxlor 32, 23+32, 23+32 # update hash + +.endm + + # + # Compute update single hash + # +.macro ppc_update_hash_1x + vxor 28, 28, 0 + + vxor 19, 19, 19 + + vpmsumd 22, 3, 28 # L + vpmsumd 23, 4, 28 # M + vpmsumd 24, 5, 28 # H + + vpmsumd 27, 22, 2 # reduction + + vsldoi 25, 23, 19, 8 # mL + vsldoi 26, 19, 23, 8 # mH + vxor 22, 22, 25 # LL + LL + vxor 24, 24, 26 # HH + HH + + vsldoi 22, 22, 22, 8 # swap + vxor 22, 22, 27 + + vsldoi 20, 22, 22, 8 # swap + vpmsumd 22, 22, 2 # reduction + vxor 20, 20, 24 + vxor 22, 22, 20 + + vmr 0, 22 # update hash + +.endm + +.macro SAVE_REGS + stdu 1,-640(1) + mflr 0 + + std 14,112(1) + std 15,120(1) + std 16,128(1) + std 17,136(1) + std 18,144(1) + std 19,152(1) + std 20,160(1) + std 21,168(1) + li 9, 256 + stvx 20, 9, 1 + addi 9, 9, 16 + stvx 21, 9, 1 + addi 9, 9, 16 + stvx 22, 9, 1 + addi 9, 9, 16 + stvx 23, 9, 1 + addi 9, 9, 16 + stvx 24, 9, 1 + addi 9, 9, 16 + stvx 25, 9, 1 + addi 9, 9, 16 + stvx 26, 9, 1 + addi 9, 9, 16 + stvx 27, 9, 1 + addi 9, 9, 16 + stvx 28, 9, 1 + addi 9, 9, 16 + stvx 29, 9, 1 + addi 9, 9, 16 + stvx 30, 9, 1 + addi 9, 9, 16 + stvx 31, 9, 1 + stxv 14, 464(1) + stxv 15, 480(1) + stxv 16, 496(1) + stxv 17, 512(1) + stxv 18, 528(1) + stxv 19, 544(1) + stxv 20, 560(1) + stxv 21, 576(1) + stxv 22, 592(1) + std 0, 656(1) +.endm + +.macro RESTORE_REGS + lxv 14, 464(1) + lxv 15, 480(1) + lxv 16, 496(1) + lxv 17, 512(1) + lxv 18, 528(1) + lxv 19, 544(1) + lxv 20, 560(1) + lxv 21, 576(1) + lxv 22, 592(1) + li 9, 256 + lvx 20, 9, 1 + addi 9, 9, 16 + lvx 21, 9, 1 + addi 9, 9, 16 + lvx 22, 9, 1 + addi 9, 9, 16 + lvx 23, 9, 1 + addi 9, 9, 16 + lvx 24, 9, 1 + addi 9, 9, 16 + lvx 25, 9, 1 + addi 9, 9, 16 + lvx 26, 9, 1 + addi 9, 9, 16 + lvx 27, 9, 1 + addi 9, 9, 16 + lvx 28, 9, 1 + addi 9, 9, 16 + lvx 29, 9, 1 + addi 9, 9, 16 + lvx 30, 9, 1 + addi 9, 9, 16 + lvx 31, 9, 1 + + ld 0, 656(1) + ld 14,112(1) + ld 15,120(1) + ld 16,128(1) + ld 17,136(1) + ld 18,144(1) + ld 19,152(1) + ld 20,160(1) + ld 21,168(1) + + mtlr 0 + addi 1, 1, 640 +.endm + +.macro LOAD_HASH_TABLE + # Load Xi + lxvb16x 32, 0, 8 # load Xi + + # load Hash - h^4, h^3, h^2, h + li 10, 32 + lxvd2x 2+32, 10, 8 # H Poli + li 10, 48 + lxvd2x 3+32, 10, 8 # Hl + li 10, 64 + lxvd2x 4+32, 10, 8 # H + li 10, 80 + lxvd2x 5+32, 10, 8 # Hh + + li 10, 96 + lxvd2x 6+32, 10, 8 # H^2l + li 10, 112 + lxvd2x 7+32, 10, 8 # H^2 + li 10, 128 + lxvd2x 8+32, 10, 8 # H^2h + + li 10, 144 + lxvd2x 9+32, 10, 8 # H^3l + li 10, 160 + lxvd2x 10+32, 10, 8 # H^3 + li 10, 176 + lxvd2x 11+32, 10, 8 # H^3h + + li 10, 192 + lxvd2x 12+32, 10, 8 # H^4l + li 10, 208 + lxvd2x 13+32, 10, 8 # H^4 + li 10, 224 + lxvd2x 14+32, 10, 8 # H^4h +.endm + + # + # aes_p10_gcm_encrypt (const void *inp, void *out, size_t len, + # const char *rk, unsigned char iv[16], void *Xip); + # + # r3 - inp + # r4 - out + # r5 - len + # r6 - AES round keys + # r7 - iv and other data + # r8 - Xi, HPoli, hash keys + # + # rounds is at offset 240 in rk + # Xi is at 0 in gcm_table (Xip). + # +_GLOBAL(aes_p10_gcm_encrypt) +.align 5 + + SAVE_REGS + + LOAD_HASH_TABLE + + # initialize ICB: GHASH( IV ), IV - r7 + lxvb16x 30+32, 0, 7 # load IV - v30 + + mr 12, 5 # length + li 11, 0 # block index + + # counter 1 + vxor 31, 31, 31 + vspltisb 22, 1 + vsldoi 31, 31, 22,1 # counter 1 + + # load round key to VSR + lxv 0, 0(6) + lxv 1, 0x10(6) + lxv 2, 0x20(6) + lxv 3, 0x30(6) + lxv 4, 0x40(6) + lxv 5, 0x50(6) + lxv 6, 0x60(6) + lxv 7, 0x70(6) + lxv 8, 0x80(6) + lxv 9, 0x90(6) + lxv 10, 0xa0(6) + + # load rounds - 10 (128), 12 (192), 14 (256) + lwz 9,240(6) + + # + # vxor state, state, w # addroundkey + xxlor 32+29, 0, 0 + vxor 15, 30, 29 # IV + round key - add round key 0 + + cmpdi 9, 10 + beq Loop_aes_gcm_8x + + # load 2 more round keys (v11, v12) + lxv 11, 0xb0(6) + lxv 12, 0xc0(6) + + cmpdi 9, 12 + beq Loop_aes_gcm_8x + + # load 2 more round keys (v11, v12, v13, v14) + lxv 13, 0xd0(6) + lxv 14, 0xe0(6) + cmpdi 9, 14 + beq Loop_aes_gcm_8x + + b aes_gcm_out + +.align 5 +Loop_aes_gcm_8x: + mr 14, 3 + mr 9, 4 + + # + # check partial block + # +Continue_partial_check: + ld 15, 56(7) + cmpdi 15, 0 + beq Continue + bgt Final_block + cmpdi 15, 16 + blt Final_block + +Continue: + # n blcoks + li 10, 128 + divdu 10, 12, 10 # n 128 bytes-blocks + cmpdi 10, 0 + beq Loop_last_block + + vaddudm 30, 30, 31 # IV + counter + vxor 16, 30, 29 + vaddudm 30, 30, 31 + vxor 17, 30, 29 + vaddudm 30, 30, 31 + vxor 18, 30, 29 + vaddudm 30, 30, 31 + vxor 19, 30, 29 + vaddudm 30, 30, 31 + vxor 20, 30, 29 + vaddudm 30, 30, 31 + vxor 21, 30, 29 + vaddudm 30, 30, 31 + vxor 22, 30, 29 + + mtctr 10 + + li 15, 16 + li 16, 32 + li 17, 48 + li 18, 64 + li 19, 80 + li 20, 96 + li 21, 112 + + lwz 10, 240(6) + +Loop_8x_block: + + lxvb16x 15, 0, 14 # load block + lxvb16x 16, 15, 14 # load block + lxvb16x 17, 16, 14 # load block + lxvb16x 18, 17, 14 # load block + lxvb16x 19, 18, 14 # load block + lxvb16x 20, 19, 14 # load block + lxvb16x 21, 20, 14 # load block + lxvb16x 22, 21, 14 # load block + addi 14, 14, 128 + + Loop_aes_middle8x + + xxlor 23+32, 10, 10 + + cmpdi 10, 10 + beq Do_next_ghash + + # 192 bits + xxlor 24+32, 11, 11 + + vcipher 15, 15, 23 + vcipher 16, 16, 23 + vcipher 17, 17, 23 + vcipher 18, 18, 23 + vcipher 19, 19, 23 + vcipher 20, 20, 23 + vcipher 21, 21, 23 + vcipher 22, 22, 23 + + vcipher 15, 15, 24 + vcipher 16, 16, 24 + vcipher 17, 17, 24 + vcipher 18, 18, 24 + vcipher 19, 19, 24 + vcipher 20, 20, 24 + vcipher 21, 21, 24 + vcipher 22, 22, 24 + + xxlor 23+32, 12, 12 + + cmpdi 10, 12 + beq Do_next_ghash + + # 256 bits + xxlor 24+32, 13, 13 + + vcipher 15, 15, 23 + vcipher 16, 16, 23 + vcipher 17, 17, 23 + vcipher 18, 18, 23 + vcipher 19, 19, 23 + vcipher 20, 20, 23 + vcipher 21, 21, 23 + vcipher 22, 22, 23 + + vcipher 15, 15, 24 + vcipher 16, 16, 24 + vcipher 17, 17, 24 + vcipher 18, 18, 24 + vcipher 19, 19, 24 + vcipher 20, 20, 24 + vcipher 21, 21, 24 + vcipher 22, 22, 24 + + xxlor 23+32, 14, 14 + + cmpdi 10, 14 + beq Do_next_ghash + b aes_gcm_out + +Do_next_ghash: + + # + # last round + vcipherlast 15, 15, 23 + vcipherlast 16, 16, 23 + + xxlxor 47, 47, 15 + stxvb16x 47, 0, 9 # store output + xxlxor 48, 48, 16 + stxvb16x 48, 15, 9 # store output + + vcipherlast 17, 17, 23 + vcipherlast 18, 18, 23 + + xxlxor 49, 49, 17 + stxvb16x 49, 16, 9 # store output + xxlxor 50, 50, 18 + stxvb16x 50, 17, 9 # store output + + vcipherlast 19, 19, 23 + vcipherlast 20, 20, 23 + + xxlxor 51, 51, 19 + stxvb16x 51, 18, 9 # store output + xxlxor 52, 52, 20 + stxvb16x 52, 19, 9 # store output + + vcipherlast 21, 21, 23 + vcipherlast 22, 22, 23 + + xxlxor 53, 53, 21 + stxvb16x 53, 20, 9 # store output + xxlxor 54, 54, 22 + stxvb16x 54, 21, 9 # store output + + addi 9, 9, 128 + + # ghash here + ppc_aes_gcm_ghash2_4x + + xxlor 27+32, 0, 0 + vaddudm 30, 30, 31 # IV + counter + vmr 29, 30 + vxor 15, 30, 27 # add round key + vaddudm 30, 30, 31 + vxor 16, 30, 27 + vaddudm 30, 30, 31 + vxor 17, 30, 27 + vaddudm 30, 30, 31 + vxor 18, 30, 27 + vaddudm 30, 30, 31 + vxor 19, 30, 27 + vaddudm 30, 30, 31 + vxor 20, 30, 27 + vaddudm 30, 30, 31 + vxor 21, 30, 27 + vaddudm 30, 30, 31 + vxor 22, 30, 27 + + addi 12, 12, -128 + addi 11, 11, 128 + + bdnz Loop_8x_block + + vmr 30, 29 + stxvb16x 30+32, 0, 7 # update IV + +Loop_last_block: + cmpdi 12, 0 + beq aes_gcm_out + + # loop last few blocks + li 10, 16 + divdu 10, 12, 10 + + mtctr 10 + + lwz 10, 240(6) + + cmpdi 12, 16 + blt Final_block + +Next_rem_block: + lxvb16x 15, 0, 14 # load block + + Loop_aes_middle_1x + + xxlor 23+32, 10, 10 + + cmpdi 10, 10 + beq Do_next_1x + + # 192 bits + xxlor 24+32, 11, 11 + + vcipher 15, 15, 23 + vcipher 15, 15, 24 + + xxlor 23+32, 12, 12 + + cmpdi 10, 12 + beq Do_next_1x + + # 256 bits + xxlor 24+32, 13, 13 + + vcipher 15, 15, 23 + vcipher 15, 15, 24 + + xxlor 23+32, 14, 14 + + cmpdi 10, 14 + beq Do_next_1x + +Do_next_1x: + vcipherlast 15, 15, 23 + + xxlxor 47, 47, 15 + stxvb16x 47, 0, 9 # store output + addi 14, 14, 16 + addi 9, 9, 16 + + vmr 28, 15 + ppc_update_hash_1x + + addi 12, 12, -16 + addi 11, 11, 16 + xxlor 19+32, 0, 0 + vaddudm 30, 30, 31 # IV + counter + vxor 15, 30, 19 # add round key + + bdnz Next_rem_block + + li 15, 0 + std 15, 56(7) # clear partial? + stxvb16x 30+32, 0, 7 # update IV + cmpdi 12, 0 + beq aes_gcm_out + +Final_block: + lwz 10, 240(6) + Loop_aes_middle_1x + + xxlor 23+32, 10, 10 + + cmpdi 10, 10 + beq Do_final_1x + + # 192 bits + xxlor 24+32, 11, 11 + + vcipher 15, 15, 23 + vcipher 15, 15, 24 + + xxlor 23+32, 12, 12 + + cmpdi 10, 12 + beq Do_final_1x + + # 256 bits + xxlor 24+32, 13, 13 + + vcipher 15, 15, 23 + vcipher 15, 15, 24 + + xxlor 23+32, 14, 14 + + cmpdi 10, 14 + beq Do_final_1x + +Do_final_1x: + vcipherlast 15, 15, 23 + + # check partial block + li 21, 0 # encrypt + ld 15, 56(7) # partial? + cmpdi 15, 0 + beq Normal_block + bl Do_partial_block + + cmpdi 12, 0 + ble aes_gcm_out + + b Continue_partial_check + +Normal_block: + lxvb16x 15, 0, 14 # load last block + xxlxor 47, 47, 15 + + # create partial block mask + li 15, 16 + sub 15, 15, 12 # index to the mask + + vspltisb 16, -1 # first 16 bytes - 0xffff...ff + vspltisb 17, 0 # second 16 bytes - 0x0000...00 + li 10, 192 + stvx 16, 10, 1 + addi 10, 10, 16 + stvx 17, 10, 1 + + addi 10, 1, 192 + lxvb16x 16, 15, 10 # load partial block mask + xxland 47, 47, 16 + + vmr 28, 15 + ppc_update_hash_1x + + # * should store only the remaining bytes. + bl Write_partial_block + + stxvb16x 30+32, 0, 7 # update IV + std 12, 56(7) # update partial? + li 16, 16 + + stxvb16x 32, 0, 8 # write out Xi + stxvb16x 32, 16, 8 # write out Xi + b aes_gcm_out + + # + # Compute data mask + # +.macro GEN_MASK _mask _start _end + vspltisb 16, -1 # first 16 bytes - 0xffff...ff + vspltisb 17, 0 # second 16 bytes - 0x0000...00 + li 10, 192 + stxvb16x 17+32, 10, 1 + add 10, 10, \_start + stxvb16x 16+32, 10, 1 + add 10, 10, \_end + stxvb16x 17+32, 10, 1 + + addi 10, 1, 192 + lxvb16x \_mask, 0, 10 # load partial block mask +.endm + + # + # Handle multiple partial blocks for encrypt and decrypt + # operations. + # +SYM_FUNC_START_LOCAL(Do_partial_block) + add 17, 15, 5 + cmpdi 17, 16 + bgt Big_block + GEN_MASK 18, 15, 5 + b _Partial +SYM_FUNC_END(Do_partial_block) +Big_block: + li 16, 16 + GEN_MASK 18, 15, 16 + +_Partial: + lxvb16x 17+32, 0, 14 # load last block + sldi 16, 15, 3 + mtvsrdd 32+16, 0, 16 + vsro 17, 17, 16 + xxlxor 47, 47, 17+32 + xxland 47, 47, 18 + + vxor 0, 0, 0 # clear Xi + vmr 28, 15 + + cmpdi 21, 0 # encrypt/decrypt ops? + beq Skip_decrypt + xxland 32+28, 32+17, 18 + +Skip_decrypt: + + ppc_update_hash_1x + + li 16, 16 + lxvb16x 32+29, 16, 8 + vxor 0, 0, 29 + stxvb16x 32, 0, 8 # save Xi + stxvb16x 32, 16, 8 # save Xi + + # store partial block + # loop the rest of the stream if any + sldi 16, 15, 3 + mtvsrdd 32+16, 0, 16 + vslo 15, 15, 16 + #stxvb16x 15+32, 0, 9 # last block + + li 16, 16 + sub 17, 16, 15 # 16 - partial + + add 16, 15, 5 + cmpdi 16, 16 + bgt Larger_16 + mr 17, 5 +Larger_16: + + # write partial + li 10, 192 + stxvb16x 15+32, 10, 1 # save current block + + addi 10, 9, -1 + addi 16, 1, 191 + mtctr 17 # move partial byte count + +Write_last_partial: + lbzu 18, 1(16) + stbu 18, 1(10) + bdnz Write_last_partial + # Complete loop partial + + add 14, 14, 17 + add 9, 9, 17 + sub 12, 12, 17 + add 11, 11, 17 + + add 15, 15, 5 + cmpdi 15, 16 + blt Save_partial + + vaddudm 30, 30, 31 + stxvb16x 30+32, 0, 7 # update IV + xxlor 32+29, 0, 0 + vxor 15, 30, 29 # IV + round key - add round key 0 + li 15, 0 + std 15, 56(7) # partial done - clear + b Partial_done +Save_partial: + std 15, 56(7) # partial + +Partial_done: + blr + + # + # Write partial block + # r9 - output + # r12 - remaining bytes + # v15 - partial input data + # +SYM_FUNC_START_LOCAL(Write_partial_block) + li 10, 192 + stxvb16x 15+32, 10, 1 # last block + + addi 10, 9, -1 + addi 16, 1, 191 + + mtctr 12 # remaining bytes + li 15, 0 + +Write_last_byte: + lbzu 14, 1(16) + stbu 14, 1(10) + bdnz Write_last_byte + blr +SYM_FUNC_END(Write_partial_block) + +aes_gcm_out: + # out = state + stxvb16x 32, 0, 8 # write out Xi + add 3, 11, 12 # return count + + RESTORE_REGS + blr + + # + # 8x Decrypt + # +_GLOBAL(aes_p10_gcm_decrypt) +.align 5 + + SAVE_REGS + + LOAD_HASH_TABLE + + # initialize ICB: GHASH( IV ), IV - r7 + lxvb16x 30+32, 0, 7 # load IV - v30 + + mr 12, 5 # length + li 11, 0 # block index + + # counter 1 + vxor 31, 31, 31 + vspltisb 22, 1 + vsldoi 31, 31, 22,1 # counter 1 + + # load round key to VSR + lxv 0, 0(6) + lxv 1, 0x10(6) + lxv 2, 0x20(6) + lxv 3, 0x30(6) + lxv 4, 0x40(6) + lxv 5, 0x50(6) + lxv 6, 0x60(6) + lxv 7, 0x70(6) + lxv 8, 0x80(6) + lxv 9, 0x90(6) + lxv 10, 0xa0(6) + + # load rounds - 10 (128), 12 (192), 14 (256) + lwz 9,240(6) + + # + # vxor state, state, w # addroundkey + xxlor 32+29, 0, 0 + vxor 15, 30, 29 # IV + round key - add round key 0 + + cmpdi 9, 10 + beq Loop_aes_gcm_8x_dec + + # load 2 more round keys (v11, v12) + lxv 11, 0xb0(6) + lxv 12, 0xc0(6) + + cmpdi 9, 12 + beq Loop_aes_gcm_8x_dec + + # load 2 more round keys (v11, v12, v13, v14) + lxv 13, 0xd0(6) + lxv 14, 0xe0(6) + cmpdi 9, 14 + beq Loop_aes_gcm_8x_dec + + b aes_gcm_out + +.align 5 +Loop_aes_gcm_8x_dec: + mr 14, 3 + mr 9, 4 + + # + # check partial block + # +Continue_partial_check_dec: + ld 15, 56(7) + cmpdi 15, 0 + beq Continue_dec + bgt Final_block_dec + cmpdi 15, 16 + blt Final_block_dec + +Continue_dec: + # n blcoks + li 10, 128 + divdu 10, 12, 10 # n 128 bytes-blocks + cmpdi 10, 0 + beq Loop_last_block_dec + + vaddudm 30, 30, 31 # IV + counter + vxor 16, 30, 29 + vaddudm 30, 30, 31 + vxor 17, 30, 29 + vaddudm 30, 30, 31 + vxor 18, 30, 29 + vaddudm 30, 30, 31 + vxor 19, 30, 29 + vaddudm 30, 30, 31 + vxor 20, 30, 29 + vaddudm 30, 30, 31 + vxor 21, 30, 29 + vaddudm 30, 30, 31 + vxor 22, 30, 29 + + mtctr 10 + + li 15, 16 + li 16, 32 + li 17, 48 + li 18, 64 + li 19, 80 + li 20, 96 + li 21, 112 + + lwz 10, 240(6) + +Loop_8x_block_dec: + + lxvb16x 15, 0, 14 # load block + lxvb16x 16, 15, 14 # load block + lxvb16x 17, 16, 14 # load block + lxvb16x 18, 17, 14 # load block + lxvb16x 19, 18, 14 # load block + lxvb16x 20, 19, 14 # load block + lxvb16x 21, 20, 14 # load block + lxvb16x 22, 21, 14 # load block + addi 14, 14, 128 + + Loop_aes_middle8x + + xxlor 23+32, 10, 10 + + cmpdi 10, 10 + beq Do_next_ghash_dec + + # 192 bits + xxlor 24+32, 11, 11 + + vcipher 15, 15, 23 + vcipher 16, 16, 23 + vcipher 17, 17, 23 + vcipher 18, 18, 23 + vcipher 19, 19, 23 + vcipher 20, 20, 23 + vcipher 21, 21, 23 + vcipher 22, 22, 23 + + vcipher 15, 15, 24 + vcipher 16, 16, 24 + vcipher 17, 17, 24 + vcipher 18, 18, 24 + vcipher 19, 19, 24 + vcipher 20, 20, 24 + vcipher 21, 21, 24 + vcipher 22, 22, 24 + + xxlor 23+32, 12, 12 + + cmpdi 10, 12 + beq Do_next_ghash_dec + + # 256 bits + xxlor 24+32, 13, 13 + + vcipher 15, 15, 23 + vcipher 16, 16, 23 + vcipher 17, 17, 23 + vcipher 18, 18, 23 + vcipher 19, 19, 23 + vcipher 20, 20, 23 + vcipher 21, 21, 23 + vcipher 22, 22, 23 + + vcipher 15, 15, 24 + vcipher 16, 16, 24 + vcipher 17, 17, 24 + vcipher 18, 18, 24 + vcipher 19, 19, 24 + vcipher 20, 20, 24 + vcipher 21, 21, 24 + vcipher 22, 22, 24 + + xxlor 23+32, 14, 14 + + cmpdi 10, 14 + beq Do_next_ghash_dec + b aes_gcm_out + +Do_next_ghash_dec: + + # + # last round + vcipherlast 15, 15, 23 + vcipherlast 16, 16, 23 + + xxlxor 47, 47, 15 + stxvb16x 47, 0, 9 # store output + xxlxor 48, 48, 16 + stxvb16x 48, 15, 9 # store output + + vcipherlast 17, 17, 23 + vcipherlast 18, 18, 23 + + xxlxor 49, 49, 17 + stxvb16x 49, 16, 9 # store output + xxlxor 50, 50, 18 + stxvb16x 50, 17, 9 # store output + + vcipherlast 19, 19, 23 + vcipherlast 20, 20, 23 + + xxlxor 51, 51, 19 + stxvb16x 51, 18, 9 # store output + xxlxor 52, 52, 20 + stxvb16x 52, 19, 9 # store output + + vcipherlast 21, 21, 23 + vcipherlast 22, 22, 23 + + xxlxor 53, 53, 21 + stxvb16x 53, 20, 9 # store output + xxlxor 54, 54, 22 + stxvb16x 54, 21, 9 # store output + + addi 9, 9, 128 + + xxlor 15+32, 15, 15 + xxlor 16+32, 16, 16 + xxlor 17+32, 17, 17 + xxlor 18+32, 18, 18 + xxlor 19+32, 19, 19 + xxlor 20+32, 20, 20 + xxlor 21+32, 21, 21 + xxlor 22+32, 22, 22 + + # ghash here + ppc_aes_gcm_ghash2_4x + + xxlor 27+32, 0, 0 + vaddudm 30, 30, 31 # IV + counter + vmr 29, 30 + vxor 15, 30, 27 # add round key + vaddudm 30, 30, 31 + vxor 16, 30, 27 + vaddudm 30, 30, 31 + vxor 17, 30, 27 + vaddudm 30, 30, 31 + vxor 18, 30, 27 + vaddudm 30, 30, 31 + vxor 19, 30, 27 + vaddudm 30, 30, 31 + vxor 20, 30, 27 + vaddudm 30, 30, 31 + vxor 21, 30, 27 + vaddudm 30, 30, 31 + vxor 22, 30, 27 + + addi 12, 12, -128 + addi 11, 11, 128 + + bdnz Loop_8x_block_dec + + vmr 30, 29 + stxvb16x 30+32, 0, 7 # update IV + +Loop_last_block_dec: + cmpdi 12, 0 + beq aes_gcm_out + + # loop last few blocks + li 10, 16 + divdu 10, 12, 10 + + mtctr 10 + + lwz 10, 240(6) + + cmpdi 12, 16 + blt Final_block_dec + +Next_rem_block_dec: + lxvb16x 15, 0, 14 # load block + + Loop_aes_middle_1x + + xxlor 23+32, 10, 10 + + cmpdi 10, 10 + beq Do_next_1x_dec + + # 192 bits + xxlor 24+32, 11, 11 + + vcipher 15, 15, 23 + vcipher 15, 15, 24 + + xxlor 23+32, 12, 12 + + cmpdi 10, 12 + beq Do_next_1x_dec + + # 256 bits + xxlor 24+32, 13, 13 + + vcipher 15, 15, 23 + vcipher 15, 15, 24 + + xxlor 23+32, 14, 14 + + cmpdi 10, 14 + beq Do_next_1x_dec + +Do_next_1x_dec: + vcipherlast 15, 15, 23 + + xxlxor 47, 47, 15 + stxvb16x 47, 0, 9 # store output + addi 14, 14, 16 + addi 9, 9, 16 + + xxlor 28+32, 15, 15 + #vmr 28, 15 + ppc_update_hash_1x + + addi 12, 12, -16 + addi 11, 11, 16 + xxlor 19+32, 0, 0 + vaddudm 30, 30, 31 # IV + counter + vxor 15, 30, 19 # add round key + + bdnz Next_rem_block_dec + + li 15, 0 + std 15, 56(7) # clear partial? + stxvb16x 30+32, 0, 7 # update IV + cmpdi 12, 0 + beq aes_gcm_out + +Final_block_dec: + lwz 10, 240(6) + Loop_aes_middle_1x + + xxlor 23+32, 10, 10 + + cmpdi 10, 10 + beq Do_final_1x_dec + + # 192 bits + xxlor 24+32, 11, 11 + + vcipher 15, 15, 23 + vcipher 15, 15, 24 + + xxlor 23+32, 12, 12 + + cmpdi 10, 12 + beq Do_final_1x_dec + + # 256 bits + xxlor 24+32, 13, 13 + + vcipher 15, 15, 23 + vcipher 15, 15, 24 + + xxlor 23+32, 14, 14 + + cmpdi 10, 14 + beq Do_final_1x_dec + +Do_final_1x_dec: + vcipherlast 15, 15, 23 + + # check partial block + li 21, 1 # decrypt + ld 15, 56(7) # partial? + cmpdi 15, 0 + beq Normal_block_dec + bl Do_partial_block + cmpdi 12, 0 + ble aes_gcm_out + + b Continue_partial_check_dec + +Normal_block_dec: + lxvb16x 15, 0, 14 # load last block + xxlxor 47, 47, 15 + + # create partial block mask + li 15, 16 + sub 15, 15, 12 # index to the mask + + vspltisb 16, -1 # first 16 bytes - 0xffff...ff + vspltisb 17, 0 # second 16 bytes - 0x0000...00 + li 10, 192 + stvx 16, 10, 1 + addi 10, 10, 16 + stvx 17, 10, 1 + + addi 10, 1, 192 + lxvb16x 16, 15, 10 # load partial block mask + xxland 47, 47, 16 + + xxland 32+28, 15, 16 + #vmr 28, 15 + ppc_update_hash_1x + + # * should store only the remaining bytes. + bl Write_partial_block + + stxvb16x 30+32, 0, 7 # update IV + std 12, 56(7) # update partial? + li 16, 16 + + stxvb16x 32, 0, 8 # write out Xi + stxvb16x 32, 16, 8 # write out Xi + b aes_gcm_out -- 2.31.1