ICP: Improve AES-GCM performance

Currently SIMD accelerated AES-GCM performance is limited by two
factors:

a. The need to disable preemption and interrupts and save the FPU
state before using it and to do the reverse when done. Due to the
way the code is organized (see (b) below) we have to pay this price
twice for each 16 byte GCM block processed.

b. Most processing is done in C, operating on single GCM blocks.
The use of SIMD instructions is limited to the AES encryption of the
counter block (AES-NI) and the Galois multiplication (PCLMULQDQ).
This leads to the FPU not being fully utilized for crypto
operations.

To solve (a) we do crypto processing in larger chunks while owning
the FPU. An `icp_gcm_avx_chunk_size` module parameter was introduced
to make this chunk size tweakable. It defaults to 32 KiB. This step
alone roughly doubles performance. (b) is tackled by porting and
using the highly optimized openssl AES-GCM assembler routines, which
do all the processing (CTR, AES, GMULT) in a single routine. Both
steps together result in up to 32x reduction of the time spend in
the en/decryption routines, leading up to approximately 12x
throughput increase for large (128 KiB) blocks.

Lastly, this commit changes the default encryption algorithm from
AES-CCM to AES-GCM when setting the `encryption=on` property.

Reviewed-By: Brian Behlendorf <[email protected]>
Reviewed-By: Jason King <[email protected]>
Reviewed-By: Tom Caputi <[email protected]>
Reviewed-By: Richard Laager <[email protected]>
Signed-off-by: Attila Fülöp <[email protected]>
Closes #9749 

6 files changed, 1821 insertions, 0 deletions

diff --git a/module/icp/asm-x86_64/modes/THIRDPARTYLICENSE.cryptogams b/module/icp/asm-x86_64/modes/THIRDPARTYLICENSE.cryptogams
new file mode 100644
index 000000000..0de1883dc
--- /dev/null
+++ b/module/icp/asm-x86_64/modes/THIRDPARTYLICENSE.cryptogams
@@ -0,0 +1,36 @@
+Copyright (c) 2006-2017, CRYPTOGAMS by <[email protected]>
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions
+are met:
+
+      *	Redistributions of source code must retain copyright notices,
+	this list of conditions and the following disclaimer.
+
+      *	Redistributions in binary form must reproduce the above
+	copyright notice, this list of conditions and the following
+	disclaimer in the documentation and/or other materials
+	provided with the distribution.
+
+      *	Neither the name of the CRYPTOGAMS nor the names of its
+	copyright holder and contributors may be used to endorse or
+	promote products derived from this software without specific
+	prior written permission.
+
+ALTERNATIVELY, provided that this notice is retained in full, this
+product may be distributed under the terms of the GNU General Public
+License (GPL), in which case the provisions of the GPL apply INSTEAD OF
+those given above.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDER AND CONTRIBUTORS
+"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
diff --git a/module/icp/asm-x86_64/modes/THIRDPARTYLICENSE.cryptogams.descrip b/module/icp/asm-x86_64/modes/THIRDPARTYLICENSE.cryptogams.descrip
new file mode 100644
index 000000000..6184759c8
--- /dev/null
+++ b/module/icp/asm-x86_64/modes/THIRDPARTYLICENSE.cryptogams.descrip
@@ -0,0 +1 @@
+PORTIONS OF GCM and GHASH FUNCTIONALITY
diff --git a/module/icp/asm-x86_64/modes/THIRDPARTYLICENSE.openssl b/module/icp/asm-x86_64/modes/THIRDPARTYLICENSE.openssl
new file mode 100644
index 000000000..49cc83d2e
--- /dev/null
+++ b/module/icp/asm-x86_64/modes/THIRDPARTYLICENSE.openssl
@@ -0,0 +1,177 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        https://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
diff --git a/module/icp/asm-x86_64/modes/THIRDPARTYLICENSE.openssl.descrip b/module/icp/asm-x86_64/modes/THIRDPARTYLICENSE.openssl.descrip
new file mode 100644
index 000000000..6184759c8
--- /dev/null
+++ b/module/icp/asm-x86_64/modes/THIRDPARTYLICENSE.openssl.descrip
@@ -0,0 +1 @@
+PORTIONS OF GCM and GHASH FUNCTIONALITY
diff --git a/module/icp/asm-x86_64/modes/aesni-gcm-x86_64.S b/module/icp/asm-x86_64/modes/aesni-gcm-x86_64.S
new file mode 100644
index 000000000..bad0b7d23
--- /dev/null
+++ b/module/icp/asm-x86_64/modes/aesni-gcm-x86_64.S
@@ -0,0 +1,892 @@
+# Copyright 2013-2016 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the Apache License 2.0 (the "License").  You may not use
+# this file except in compliance with the License.  You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+#
+# ====================================================================
+# Written by Andy Polyakov <[email protected]> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+#
+#
+# AES-NI-CTR+GHASH stitch.
+#
+# February 2013
+#
+# OpenSSL GCM implementation is organized in such way that its
+# performance is rather close to the sum of its streamed components,
+# in the context parallelized AES-NI CTR and modulo-scheduled
+# PCLMULQDQ-enabled GHASH. Unfortunately, as no stitch implementation
+# was observed to perform significantly better than the sum of the
+# components on contemporary CPUs, the effort was deemed impossible to
+# justify. This module is based on combination of Intel submissions,
+# [1] and [2], with MOVBE twist suggested by Ilya Albrekht and Max
+# Locktyukhin of Intel Corp. who verified that it reduces shuffles
+# pressure with notable relative improvement, achieving 1.0 cycle per
+# byte processed with 128-bit key on Haswell processor, 0.74 - on
+# Broadwell, 0.63 - on Skylake... [Mentioned results are raw profiled
+# measurements for favourable packet size, one divisible by 96.
+# Applications using the EVP interface will observe a few percent
+# worse performance.]
+#
+# Knights Landing processes 1 byte in 1.25 cycles (measured with EVP).
+#
+# [1] http://rt.openssl.org/Ticket/Display.html?id=2900&user=guest&pass=guest
+# [2] http://www.intel.com/content/dam/www/public/us/en/documents/software-support/enabling-high-performance-gcm.pdf
+
+# Generated once from
+# https://github.com/openssl/openssl/blob/5ffc3324/crypto/modes/asm/aesni-gcm-x86_64.pl
+# and modified for ICP. Modification are kept at a bare minimum to ease later
+# upstream merges.
+
+#if defined(__x86_64__) && defined(HAVE_AVX) && \
+    defined(HAVE_AES) && defined(HAVE_PCLMULQDQ) && defined(HAVE_MOVBE)
+
+.text
+
+.type	_aesni_ctr32_ghash_6x,@function
+.align	32
+_aesni_ctr32_ghash_6x:
+	vmovdqu	32(%r11),%xmm2
+	subq	$6,%rdx
+	vpxor	%xmm4,%xmm4,%xmm4
+	vmovdqu	0-128(%rcx),%xmm15
+	vpaddb	%xmm2,%xmm1,%xmm10
+	vpaddb	%xmm2,%xmm10,%xmm11
+	vpaddb	%xmm2,%xmm11,%xmm12
+	vpaddb	%xmm2,%xmm12,%xmm13
+	vpaddb	%xmm2,%xmm13,%xmm14
+	vpxor	%xmm15,%xmm1,%xmm9
+	vmovdqu	%xmm4,16+8(%rsp)
+	jmp	.Loop6x
+
+.align	32
+.Loop6x:
+	addl	$100663296,%ebx
+	jc	.Lhandle_ctr32
+	vmovdqu	0-32(%r9),%xmm3
+	vpaddb	%xmm2,%xmm14,%xmm1
+	vpxor	%xmm15,%xmm10,%xmm10
+	vpxor	%xmm15,%xmm11,%xmm11
+
+.Lresume_ctr32:
+	vmovdqu	%xmm1,(%r8)
+	vpclmulqdq	$0x10,%xmm3,%xmm7,%xmm5
+	vpxor	%xmm15,%xmm12,%xmm12
+	vmovups	16-128(%rcx),%xmm2
+	vpclmulqdq	$0x01,%xmm3,%xmm7,%xmm6
+	xorq	%r12,%r12
+	cmpq	%r14,%r15
+
+	vaesenc	%xmm2,%xmm9,%xmm9
+	vmovdqu	48+8(%rsp),%xmm0
+	vpxor	%xmm15,%xmm13,%xmm13
+	vpclmulqdq	$0x00,%xmm3,%xmm7,%xmm1
+	vaesenc	%xmm2,%xmm10,%xmm10
+	vpxor	%xmm15,%xmm14,%xmm14
+	setnc	%r12b
+	vpclmulqdq	$0x11,%xmm3,%xmm7,%xmm7
+	vaesenc	%xmm2,%xmm11,%xmm11
+	vmovdqu	16-32(%r9),%xmm3
+	negq	%r12
+	vaesenc	%xmm2,%xmm12,%xmm12
+	vpxor	%xmm5,%xmm6,%xmm6
+	vpclmulqdq	$0x00,%xmm3,%xmm0,%xmm5
+	vpxor	%xmm4,%xmm8,%xmm8
+	vaesenc	%xmm2,%xmm13,%xmm13
+	vpxor	%xmm5,%xmm1,%xmm4
+	andq	$0x60,%r12
+	vmovups	32-128(%rcx),%xmm15
+	vpclmulqdq	$0x10,%xmm3,%xmm0,%xmm1
+	vaesenc	%xmm2,%xmm14,%xmm14
+
+	vpclmulqdq	$0x01,%xmm3,%xmm0,%xmm2
+	leaq	(%r14,%r12,1),%r14
+	vaesenc	%xmm15,%xmm9,%xmm9
+	vpxor	16+8(%rsp),%xmm8,%xmm8
+	vpclmulqdq	$0x11,%xmm3,%xmm0,%xmm3
+	vmovdqu	64+8(%rsp),%xmm0
+	vaesenc	%xmm15,%xmm10,%xmm10
+	movbeq	88(%r14),%r13
+	vaesenc	%xmm15,%xmm11,%xmm11
+	movbeq	80(%r14),%r12
+	vaesenc	%xmm15,%xmm12,%xmm12
+	movq	%r13,32+8(%rsp)
+	vaesenc	%xmm15,%xmm13,%xmm13
+	movq	%r12,40+8(%rsp)
+	vmovdqu	48-32(%r9),%xmm5
+	vaesenc	%xmm15,%xmm14,%xmm14
+
+	vmovups	48-128(%rcx),%xmm15
+	vpxor	%xmm1,%xmm6,%xmm6
+	vpclmulqdq	$0x00,%xmm5,%xmm0,%xmm1
+	vaesenc	%xmm15,%xmm9,%xmm9
+	vpxor	%xmm2,%xmm6,%xmm6
+	vpclmulqdq	$0x10,%xmm5,%xmm0,%xmm2
+	vaesenc	%xmm15,%xmm10,%xmm10
+	vpxor	%xmm3,%xmm7,%xmm7
+	vpclmulqdq	$0x01,%xmm5,%xmm0,%xmm3
+	vaesenc	%xmm15,%xmm11,%xmm11
+	vpclmulqdq	$0x11,%xmm5,%xmm0,%xmm5
+	vmovdqu	80+8(%rsp),%xmm0
+	vaesenc	%xmm15,%xmm12,%xmm12
+	vaesenc	%xmm15,%xmm13,%xmm13
+	vpxor	%xmm1,%xmm4,%xmm4
+	vmovdqu	64-32(%r9),%xmm1
+	vaesenc	%xmm15,%xmm14,%xmm14
+
+	vmovups	64-128(%rcx),%xmm15
+	vpxor	%xmm2,%xmm6,%xmm6
+	vpclmulqdq	$0x00,%xmm1,%xmm0,%xmm2
+	vaesenc	%xmm15,%xmm9,%xmm9
+	vpxor	%xmm3,%xmm6,%xmm6
+	vpclmulqdq	$0x10,%xmm1,%xmm0,%xmm3
+	vaesenc	%xmm15,%xmm10,%xmm10
+	movbeq	72(%r14),%r13
+	vpxor	%xmm5,%xmm7,%xmm7
+	vpclmulqdq	$0x01,%xmm1,%xmm0,%xmm5
+	vaesenc	%xmm15,%xmm11,%xmm11
+	movbeq	64(%r14),%r12
+	vpclmulqdq	$0x11,%xmm1,%xmm0,%xmm1
+	vmovdqu	96+8(%rsp),%xmm0
+	vaesenc	%xmm15,%xmm12,%xmm12
+	movq	%r13,48+8(%rsp)
+	vaesenc	%xmm15,%xmm13,%xmm13
+	movq	%r12,56+8(%rsp)
+	vpxor	%xmm2,%xmm4,%xmm4
+	vmovdqu	96-32(%r9),%xmm2
+	vaesenc	%xmm15,%xmm14,%xmm14
+
+	vmovups	80-128(%rcx),%xmm15
+	vpxor	%xmm3,%xmm6,%xmm6
+	vpclmulqdq	$0x00,%xmm2,%xmm0,%xmm3
+	vaesenc	%xmm15,%xmm9,%xmm9
+	vpxor	%xmm5,%xmm6,%xmm6
+	vpclmulqdq	$0x10,%xmm2,%xmm0,%xmm5
+	vaesenc	%xmm15,%xmm10,%xmm10
+	movbeq	56(%r14),%r13
+	vpxor	%xmm1,%xmm7,%xmm7
+	vpclmulqdq	$0x01,%xmm2,%xmm0,%xmm1
+	vpxor	112+8(%rsp),%xmm8,%xmm8
+	vaesenc	%xmm15,%xmm11,%xmm11
+	movbeq	48(%r14),%r12
+	vpclmulqdq	$0x11,%xmm2,%xmm0,%xmm2
+	vaesenc	%xmm15,%xmm12,%xmm12
+	movq	%r13,64+8(%rsp)
+	vaesenc	%xmm15,%xmm13,%xmm13
+	movq	%r12,72+8(%rsp)
+	vpxor	%xmm3,%xmm4,%xmm4
+	vmovdqu	112-32(%r9),%xmm3
+	vaesenc	%xmm15,%xmm14,%xmm14
+
+	vmovups	96-128(%rcx),%xmm15
+	vpxor	%xmm5,%xmm6,%xmm6
+	vpclmulqdq	$0x10,%xmm3,%xmm8,%xmm5
+	vaesenc	%xmm15,%xmm9,%xmm9
+	vpxor	%xmm1,%xmm6,%xmm6
+	vpclmulqdq	$0x01,%xmm3,%xmm8,%xmm1
+	vaesenc	%xmm15,%xmm10,%xmm10
+	movbeq	40(%r14),%r13
+	vpxor	%xmm2,%xmm7,%xmm7
+	vpclmulqdq	$0x00,%xmm3,%xmm8,%xmm2
+	vaesenc	%xmm15,%xmm11,%xmm11
+	movbeq	32(%r14),%r12
+	vpclmulqdq	$0x11,%xmm3,%xmm8,%xmm8
+	vaesenc	%xmm15,%xmm12,%xmm12
+	movq	%r13,80+8(%rsp)
+	vaesenc	%xmm15,%xmm13,%xmm13
+	movq	%r12,88+8(%rsp)
+	vpxor	%xmm5,%xmm6,%xmm6
+	vaesenc	%xmm15,%xmm14,%xmm14
+	vpxor	%xmm1,%xmm6,%xmm6
+
+	vmovups	112-128(%rcx),%xmm15
+	vpslldq	$8,%xmm6,%xmm5
+	vpxor	%xmm2,%xmm4,%xmm4
+	vmovdqu	16(%r11),%xmm3
+
+	vaesenc	%xmm15,%xmm9,%xmm9
+	vpxor	%xmm8,%xmm7,%xmm7
+	vaesenc	%xmm15,%xmm10,%xmm10
+	vpxor	%xmm5,%xmm4,%xmm4
+	movbeq	24(%r14),%r13
+	vaesenc	%xmm15,%xmm11,%xmm11
+	movbeq	16(%r14),%r12
+	vpalignr	$8,%xmm4,%xmm4,%xmm0
+	vpclmulqdq	$0x10,%xmm3,%xmm4,%xmm4
+	movq	%r13,96+8(%rsp)
+	vaesenc	%xmm15,%xmm12,%xmm12
+	movq	%r12,104+8(%rsp)
+	vaesenc	%xmm15,%xmm13,%xmm13
+	vmovups	128-128(%rcx),%xmm1
+	vaesenc	%xmm15,%xmm14,%xmm14
+
+	vaesenc	%xmm1,%xmm9,%xmm9
+	vmovups	144-128(%rcx),%xmm15
+	vaesenc	%xmm1,%xmm10,%xmm10
+	vpsrldq	$8,%xmm6,%xmm6
+	vaesenc	%xmm1,%xmm11,%xmm11
+	vpxor	%xmm6,%xmm7,%xmm7
+	vaesenc	%xmm1,%xmm12,%xmm12
+	vpxor	%xmm0,%xmm4,%xmm4
+	movbeq	8(%r14),%r13
+	vaesenc	%xmm1,%xmm13,%xmm13
+	movbeq	0(%r14),%r12
+	vaesenc	%xmm1,%xmm14,%xmm14
+	vmovups	160-128(%rcx),%xmm1
+	cmpl	$12,%ebp	// ICP uses 10,12,14 not 9,11,13 for rounds.
+	jb	.Lenc_tail
+
+	vaesenc	%xmm15,%xmm9,%xmm9
+	vaesenc	%xmm15,%xmm10,%xmm10
+	vaesenc	%xmm15,%xmm11,%xmm11
+	vaesenc	%xmm15,%xmm12,%xmm12
+	vaesenc	%xmm15,%xmm13,%xmm13
+	vaesenc	%xmm15,%xmm14,%xmm14
+
+	vaesenc	%xmm1,%xmm9,%xmm9
+	vaesenc	%xmm1,%xmm10,%xmm10
+	vaesenc	%xmm1,%xmm11,%xmm11
+	vaesenc	%xmm1,%xmm12,%xmm12
+	vaesenc	%xmm1,%xmm13,%xmm13
+	vmovups	176-128(%rcx),%xmm15
+	vaesenc	%xmm1,%xmm14,%xmm14
+	vmovups	192-128(%rcx),%xmm1
+	cmpl	$14,%ebp	// ICP does not zero key schedule.
+	jb	.Lenc_tail
+
+	vaesenc	%xmm15,%xmm9,%xmm9
+	vaesenc	%xmm15,%xmm10,%xmm10
+	vaesenc	%xmm15,%xmm11,%xmm11
+	vaesenc	%xmm15,%xmm12,%xmm12
+	vaesenc	%xmm15,%xmm13,%xmm13
+	vaesenc	%xmm15,%xmm14,%xmm14
+
+	vaesenc	%xmm1,%xmm9,%xmm9
+	vaesenc	%xmm1,%xmm10,%xmm10
+	vaesenc	%xmm1,%xmm11,%xmm11
+	vaesenc	%xmm1,%xmm12,%xmm12
+	vaesenc	%xmm1,%xmm13,%xmm13
+	vmovups	208-128(%rcx),%xmm15
+	vaesenc	%xmm1,%xmm14,%xmm14
+	vmovups	224-128(%rcx),%xmm1
+	jmp	.Lenc_tail
+
+.align	32
+.Lhandle_ctr32:
+	vmovdqu	(%r11),%xmm0
+	vpshufb	%xmm0,%xmm1,%xmm6
+	vmovdqu	48(%r11),%xmm5
+	vpaddd	64(%r11),%xmm6,%xmm10
+	vpaddd	%xmm5,%xmm6,%xmm11
+	vmovdqu	0-32(%r9),%xmm3
+	vpaddd	%xmm5,%xmm10,%xmm12
+	vpshufb	%xmm0,%xmm10,%xmm10
+	vpaddd	%xmm5,%xmm11,%xmm13
+	vpshufb	%xmm0,%xmm11,%xmm11
+	vpxor	%xmm15,%xmm10,%xmm10
+	vpaddd	%xmm5,%xmm12,%xmm14
+	vpshufb	%xmm0,%xmm12,%xmm12
+	vpxor	%xmm15,%xmm11,%xmm11
+	vpaddd	%xmm5,%xmm13,%xmm1
+	vpshufb	%xmm0,%xmm13,%xmm13
+	vpshufb	%xmm0,%xmm14,%xmm14
+	vpshufb	%xmm0,%xmm1,%xmm1
+	jmp	.Lresume_ctr32
+
+.align	32
+.Lenc_tail:
+	vaesenc	%xmm15,%xmm9,%xmm9
+	vmovdqu	%xmm7,16+8(%rsp)
+	vpalignr	$8,%xmm4,%xmm4,%xmm8
+	vaesenc	%xmm15,%xmm10,%xmm10
+	vpclmulqdq	$0x10,%xmm3,%xmm4,%xmm4
+	vpxor	0(%rdi),%xmm1,%xmm2
+	vaesenc	%xmm15,%xmm11,%xmm11
+	vpxor	16(%rdi),%xmm1,%xmm0
+	vaesenc	%xmm15,%xmm12,%xmm12
+	vpxor	32(%rdi),%xmm1,%xmm5
+	vaesenc	%xmm15,%xmm13,%xmm13
+	vpxor	48(%rdi),%xmm1,%xmm6
+	vaesenc	%xmm15,%xmm14,%xmm14
+	vpxor	64(%rdi),%xmm1,%xmm7
+	vpxor	80(%rdi),%xmm1,%xmm3
+	vmovdqu	(%r8),%xmm1
+
+	vaesenclast	%xmm2,%xmm9,%xmm9
+	vmovdqu	32(%r11),%xmm2
+	vaesenclast	%xmm0,%xmm10,%xmm10
+	vpaddb	%xmm2,%xmm1,%xmm0
+	movq	%r13,112+8(%rsp)
+	leaq	96(%rdi),%rdi
+	vaesenclast	%xmm5,%xmm11,%xmm11
+	vpaddb	%xmm2,%xmm0,%xmm5
+	movq	%r12,120+8(%rsp)
+	leaq	96(%rsi),%rsi
+	vmovdqu	0-128(%rcx),%xmm15
+	vaesenclast	%xmm6,%xmm12,%xmm12
+	vpaddb	%xmm2,%xmm5,%xmm6
+	vaesenclast	%xmm7,%xmm13,%xmm13
+	vpaddb	%xmm2,%xmm6,%xmm7
+	vaesenclast	%xmm3,%xmm14,%xmm14
+	vpaddb	%xmm2,%xmm7,%xmm3
+
+	addq	$0x60,%r10
+	subq	$0x6,%rdx
+	jc	.L6x_done
+
+	vmovups	%xmm9,-96(%rsi)
+	vpxor	%xmm15,%xmm1,%xmm9
+	vmovups	%xmm10,-80(%rsi)
+	vmovdqa	%xmm0,%xmm10
+	vmovups	%xmm11,-64(%rsi)
+	vmovdqa	%xmm5,%xmm11
+	vmovups	%xmm12,-48(%rsi)
+	vmovdqa	%xmm6,%xmm12
+	vmovups	%xmm13,-32(%rsi)
+	vmovdqa	%xmm7,%xmm13
+	vmovups	%xmm14,-16(%rsi)
+	vmovdqa	%xmm3,%xmm14
+	vmovdqu	32+8(%rsp),%xmm7
+	jmp	.Loop6x
+
+.L6x_done:
+	vpxor	16+8(%rsp),%xmm8,%xmm8
+	vpxor	%xmm4,%xmm8,%xmm8
+
+	.byte	0xf3,0xc3
+.size	_aesni_ctr32_ghash_6x,.-_aesni_ctr32_ghash_6x
+.globl	aesni_gcm_decrypt
+.type	aesni_gcm_decrypt,@function
+.align	32
+aesni_gcm_decrypt:
+.cfi_startproc
+	xorq	%r10,%r10
+	cmpq	$0x60,%rdx
+	jb	.Lgcm_dec_abort
+
+	leaq	(%rsp),%rax
+.cfi_def_cfa_register	%rax
+	pushq	%rbx
+.cfi_offset	%rbx,-16
+	pushq	%rbp
+.cfi_offset	%rbp,-24
+	pushq	%r12
+.cfi_offset	%r12,-32
+	pushq	%r13
+.cfi_offset	%r13,-40
+	pushq	%r14
+.cfi_offset	%r14,-48
+	pushq	%r15
+.cfi_offset	%r15,-56
+	vzeroupper
+
+	vmovdqu	(%r8),%xmm1
+	addq	$-128,%rsp
+	movl	12(%r8),%ebx
+	leaq	.Lbswap_mask(%rip),%r11
+	leaq	-128(%rcx),%r14
+	movq	$0xf80,%r15
+	vmovdqu	(%r9),%xmm8
+	andq	$-128,%rsp
+	vmovdqu	(%r11),%xmm0
+	leaq	128(%rcx),%rcx
+	leaq	32+32(%r9),%r9
+	movl	504-128(%rcx),%ebp	// ICP has a larger offset for rounds.
+	vpshufb	%xmm0,%xmm8,%xmm8
+
+	andq	%r15,%r14
+	andq	%rsp,%r15
+	subq	%r14,%r15
+	jc	.Ldec_no_key_aliasing
+	cmpq	$768,%r15
+	jnc	.Ldec_no_key_aliasing
+	subq	%r15,%rsp
+.Ldec_no_key_aliasing:
+
+	vmovdqu	80(%rdi),%xmm7
+	leaq	(%rdi),%r14
+	vmovdqu	64(%rdi),%xmm4
+	leaq	-192(%rdi,%rdx,1),%r15
+	vmovdqu	48(%rdi),%xmm5
+	shrq	$4,%rdx
+	xorq	%r10,%r10
+	vmovdqu	32(%rdi),%xmm6
+	vpshufb	%xmm0,%xmm7,%xmm7
+	vmovdqu	16(%rdi),%xmm2
+	vpshufb	%xmm0,%xmm4,%xmm4
+	vmovdqu	(%rdi),%xmm3
+	vpshufb	%xmm0,%xmm5,%xmm5
+	vmovdqu	%xmm4,48(%rsp)
+	vpshufb	%xmm0,%xmm6,%xmm6
+	vmovdqu	%xmm5,64(%rsp)
+	vpshufb	%xmm0,%xmm2,%xmm2
+	vmovdqu	%xmm6,80(%rsp)
+	vpshufb	%xmm0,%xmm3,%xmm3
+	vmovdqu	%xmm2,96(%rsp)
+	vmovdqu	%xmm3,112(%rsp)
+
+	call	_aesni_ctr32_ghash_6x
+
+	vmovups	%xmm9,-96(%rsi)
+	vmovups	%xmm10,-80(%rsi)
+	vmovups	%xmm11,-64(%rsi)
+	vmovups	%xmm12,-48(%rsi)
+	vmovups	%xmm13,-32(%rsi)
+	vmovups	%xmm14,-16(%rsi)
+
+	vpshufb	(%r11),%xmm8,%xmm8
+	vmovdqu	%xmm8,-64(%r9)
+
+	vzeroupper
+	movq	-48(%rax),%r15
+.cfi_restore	%r15
+	movq	-40(%rax),%r14
+.cfi_restore	%r14
+	movq	-32(%rax),%r13
+.cfi_restore	%r13
+	movq	-24(%rax),%r12
+.cfi_restore	%r12
+	movq	-16(%rax),%rbp
+.cfi_restore	%rbp
+	movq	-8(%rax),%rbx
+.cfi_restore	%rbx
+	leaq	(%rax),%rsp
+.cfi_def_cfa_register	%rsp
+.Lgcm_dec_abort:
+	movq	%r10,%rax
+	.byte	0xf3,0xc3
+.cfi_endproc
+.size	aesni_gcm_decrypt,.-aesni_gcm_decrypt
+.type	_aesni_ctr32_6x,@function
+.align	32
+_aesni_ctr32_6x:
+	vmovdqu	0-128(%rcx),%xmm4
+	vmovdqu	32(%r11),%xmm2
+	leaq	-2(%rbp),%r13	// ICP uses 10,12,14 not 9,11,13 for rounds.
+	vmovups	16-128(%rcx),%xmm15
+	leaq	32-128(%rcx),%r12
+	vpxor	%xmm4,%xmm1,%xmm9
+	addl	$100663296,%ebx
+	jc	.Lhandle_ctr32_2
+	vpaddb	%xmm2,%xmm1,%xmm10
+	vpaddb	%xmm2,%xmm10,%xmm11
+	vpxor	%xmm4,%xmm10,%xmm10
+	vpaddb	%xmm2,%xmm11,%xmm12
+	vpxor	%xmm4,%xmm11,%xmm11
+	vpaddb	%xmm2,%xmm12,%xmm13
+	vpxor	%xmm4,%xmm12,%xmm12
+	vpaddb	%xmm2,%xmm13,%xmm14
+	vpxor	%xmm4,%xmm13,%xmm13
+	vpaddb	%xmm2,%xmm14,%xmm1
+	vpxor	%xmm4,%xmm14,%xmm14
+	jmp	.Loop_ctr32
+
+.align	16
+.Loop_ctr32:
+	vaesenc	%xmm15,%xmm9,%xmm9
+	vaesenc	%xmm15,%xmm10,%xmm10
+	vaesenc	%xmm15,%xmm11,%xmm11
+	vaesenc	%xmm15,%xmm12,%xmm12
+	vaesenc	%xmm15,%xmm13,%xmm13
+	vaesenc	%xmm15,%xmm14,%xmm14
+	vmovups	(%r12),%xmm15
+	leaq	16(%r12),%r12
+	decl	%r13d
+	jnz	.Loop_ctr32
+
+	vmovdqu	(%r12),%xmm3
+	vaesenc	%xmm15,%xmm9,%xmm9
+	vpxor	0(%rdi),%xmm3,%xmm4
+	vaesenc	%xmm15,%xmm10,%xmm10
+	vpxor	16(%rdi),%xmm3,%xmm5
+	vaesenc	%xmm15,%xmm11,%xmm11
+	vpxor	32(%rdi),%xmm3,%xmm6
+	vaesenc	%xmm15,%xmm12,%xmm12
+	vpxor	48(%rdi),%xmm3,%xmm8
+	vaesenc	%xmm15,%xmm13,%xmm13
+	vpxor	64(%rdi),%xmm3,%xmm2
+	vaesenc	%xmm15,%xmm14,%xmm14
+	vpxor	80(%rdi),%xmm3,%xmm3
+	leaq	96(%rdi),%rdi
+
+	vaesenclast	%xmm4,%xmm9,%xmm9
+	vaesenclast	%xmm5,%xmm10,%xmm10
+	vaesenclast	%xmm6,%xmm11,%xmm11
+	vaesenclast	%xmm8,%xmm12,%xmm12
+	vaesenclast	%xmm2,%xmm13,%xmm13
+	vaesenclast	%xmm3,%xmm14,%xmm14
+	vmovups	%xmm9,0(%rsi)
+	vmovups	%xmm10,16(%rsi)
+	vmovups	%xmm11,32(%rsi)
+	vmovups	%xmm12,48(%rsi)
+	vmovups	%xmm13,64(%rsi)
+	vmovups	%xmm14,80(%rsi)
+	leaq	96(%rsi),%rsi
+
+	.byte	0xf3,0xc3
+.align	32
+.Lhandle_ctr32_2:
+	vpshufb	%xmm0,%xmm1,%xmm6
+	vmovdqu	48(%r11),%xmm5
+	vpaddd	64(%r11),%xmm6,%xmm10
+	vpaddd	%xmm5,%xmm6,%xmm11
+	vpaddd	%xmm5,%xmm10,%xmm12
+	vpshufb	%xmm0,%xmm10,%xmm10
+	vpaddd	%xmm5,%xmm11,%xmm13
+	vpshufb	%xmm0,%xmm11,%xmm11
+	vpxor	%xmm4,%xmm10,%xmm10
+	vpaddd	%xmm5,%xmm12,%xmm14
+	vpshufb	%xmm0,%xmm12,%xmm12
+	vpxor	%xmm4,%xmm11,%xmm11
+	vpaddd	%xmm5,%xmm13,%xmm1
+	vpshufb	%xmm0,%xmm13,%xmm13
+	vpxor	%xmm4,%xmm12,%xmm12
+	vpshufb	%xmm0,%xmm14,%xmm14
+	vpxor	%xmm4,%xmm13,%xmm13
+	vpshufb	%xmm0,%xmm1,%xmm1
+	vpxor	%xmm4,%xmm14,%xmm14
+	jmp	.Loop_ctr32
+.size	_aesni_ctr32_6x,.-_aesni_ctr32_6x
+
+.globl	aesni_gcm_encrypt
+.type	aesni_gcm_encrypt,@function
+.align	32
+aesni_gcm_encrypt:
+.cfi_startproc
+	xorq	%r10,%r10
+	cmpq	$288,%rdx
+	jb	.Lgcm_enc_abort
+
+	leaq	(%rsp),%rax
+.cfi_def_cfa_register	%rax
+	pushq	%rbx
+.cfi_offset	%rbx,-16
+	pushq	%rbp
+.cfi_offset	%rbp,-24
+	pushq	%r12
+.cfi_offset	%r12,-32
+	pushq	%r13
+.cfi_offset	%r13,-40
+	pushq	%r14
+.cfi_offset	%r14,-48
+	pushq	%r15
+.cfi_offset	%r15,-56
+	vzeroupper
+
+	vmovdqu	(%r8),%xmm1
+	addq	$-128,%rsp
+	movl	12(%r8),%ebx
+	leaq	.Lbswap_mask(%rip),%r11
+	leaq	-128(%rcx),%r14
+	movq	$0xf80,%r15
+	leaq	128(%rcx),%rcx
+	vmovdqu	(%r11),%xmm0
+	andq	$-128,%rsp
+	movl	504-128(%rcx),%ebp	// ICP has an larger offset for rounds.
+
+	andq	%r15,%r14
+	andq	%rsp,%r15
+	subq	%r14,%r15
+	jc	.Lenc_no_key_aliasing
+	cmpq	$768,%r15
+	jnc	.Lenc_no_key_aliasing
+	subq	%r15,%rsp
+.Lenc_no_key_aliasing:
+
+	leaq	(%rsi),%r14
+	leaq	-192(%rsi,%rdx,1),%r15
+	shrq	$4,%rdx
+
+	call	_aesni_ctr32_6x
+	vpshufb	%xmm0,%xmm9,%xmm8
+	vpshufb	%xmm0,%xmm10,%xmm2
+	vmovdqu	%xmm8,112(%rsp)
+	vpshufb	%xmm0,%xmm11,%xmm4
+	vmovdqu	%xmm2,96(%rsp)
+	vpshufb	%xmm0,%xmm12,%xmm5
+	vmovdqu	%xmm4,80(%rsp)
+	vpshufb	%xmm0,%xmm13,%xmm6
+	vmovdqu	%xmm5,64(%rsp)
+	vpshufb	%xmm0,%xmm14,%xmm7
+	vmovdqu	%xmm6,48(%rsp)
+
+	call	_aesni_ctr32_6x
+
+	vmovdqu	(%r9),%xmm8
+	leaq	32+32(%r9),%r9
+	subq	$12,%rdx
+	movq	$192,%r10
+	vpshufb	%xmm0,%xmm8,%xmm8
+
+	call	_aesni_ctr32_ghash_6x
+	vmovdqu	32(%rsp),%xmm7
+	vmovdqu	(%r11),%xmm0
+	vmovdqu	0-32(%r9),%xmm3
+	vpunpckhqdq	%xmm7,%xmm7,%xmm1
+	vmovdqu	32-32(%r9),%xmm15
+	vmovups	%xmm9,-96(%rsi)
+	vpshufb	%xmm0,%xmm9,%xmm9
+	vpxor	%xmm7,%xmm1,%xmm1
+	vmovups	%xmm10,-80(%rsi)
+	vpshufb	%xmm0,%xmm10,%xmm10
+	vmovups	%xmm11,-64(%rsi)
+	vpshufb	%xmm0,%xmm11,%xmm11
+	vmovups	%xmm12,-48(%rsi)
+	vpshufb	%xmm0,%xmm12,%xmm12
+	vmovups	%xmm13,-32(%rsi)
+	vpshufb	%xmm0,%xmm13,%xmm13
+	vmovups	%xmm14,-16(%rsi)
+	vpshufb	%xmm0,%xmm14,%xmm14
+	vmovdqu	%xmm9,16(%rsp)
+	vmovdqu	48(%rsp),%xmm6
+	vmovdqu	16-32(%r9),%xmm0
+	vpunpckhqdq	%xmm6,%xmm6,%xmm2
+	vpclmulqdq	$0x00,%xmm3,%xmm7,%xmm5
+	vpxor	%xmm6,%xmm2,%xmm2
+	vpclmulqdq	$0x11,%xmm3,%xmm7,%xmm7
+	vpclmulqdq	$0x00,%xmm15,%xmm1,%xmm1
+
+	vmovdqu	64(%rsp),%xmm9
+	vpclmulqdq	$0x00,%xmm0,%xmm6,%xmm4
+	vmovdqu	48-32(%r9),%xmm3
+	vpxor	%xmm5,%xmm4,%xmm4
+	vpunpckhqdq	%xmm9,%xmm9,%xmm5
+	vpclmulqdq	$0x11,%xmm0,%xmm6,%xmm6
+	vpxor	%xmm9,%xmm5,%xmm5
+	vpxor	%xmm7,%xmm6,%xmm6
+	vpclmulqdq	$0x10,%xmm15,%xmm2,%xmm2
+	vmovdqu	80-32(%r9),%xmm15
+	vpxor	%xmm1,%xmm2,%xmm2
+
+	vmovdqu	80(%rsp),%xmm1
+	vpclmulqdq	$0x00,%xmm3,%xmm9,%xmm7
+	vmovdqu	64-32(%r9),%xmm0
+	vpxor	%xmm4,%xmm7,%xmm7
+	vpunpckhqdq	%xmm1,%xmm1,%xmm4
+	vpclmulqdq	$0x11,%xmm3,%xmm9,%xmm9
+	vpxor	%xmm1,%xmm4,%xmm4
+	vpxor	%xmm6,%xmm9,%xmm9
+	vpclmulqdq	$0x00,%xmm15,%xmm5,%xmm5
+	vpxor	%xmm2,%xmm5,%xmm5
+
+	vmovdqu	96(%rsp),%xmm2
+	vpclmulqdq	$0x00,%xmm0,%xmm1,%xmm6
+	vmovdqu	96-32(%r9),%xmm3
+	vpxor	%xmm7,%xmm6,%xmm6
+	vpunpckhqdq	%xmm2,%xmm2,%xmm7
+	vpclmulqdq	$0x11,%xmm0,%xmm1,%xmm1
+	vpxor	%xmm2,%xmm7,%xmm7
+	vpxor	%xmm9,%xmm1,%xmm1
+	vpclmulqdq	$0x10,%xmm15,%xmm4,%xmm4
+	vmovdqu	128-32(%r9),%xmm15
+	vpxor	%xmm5,%xmm4,%xmm4
+
+	vpxor	112(%rsp),%xmm8,%xmm8
+	vpclmulqdq	$0x00,%xmm3,%xmm2,%xmm5
+	vmovdqu	112-32(%r9),%xmm0
+	vpunpckhqdq	%xmm8,%xmm8,%xmm9
+	vpxor	%xmm6,%xmm5,%xmm5
+	vpclmulqdq	$0x11,%xmm3,%xmm2,%xmm2
+	vpxor	%xmm8,%xmm9,%xmm9
+	vpxor	%xmm1,%xmm2,%xmm2
+	vpclmulqdq	$0x00,%xmm15,%xmm7,%xmm7
+	vpxor	%xmm4,%xmm7,%xmm4
+
+	vpclmulqdq	$0x00,%xmm0,%xmm8,%xmm6
+	vmovdqu	0-32(%r9),%xmm3
+	vpunpckhqdq	%xmm14,%xmm14,%xmm1
+	vpclmulqdq	$0x11,%xmm0,%xmm8,%xmm8
+	vpxor	%xmm14,%xmm1,%xmm1
+	vpxor	%xmm5,%xmm6,%xmm5
+	vpclmulqdq	$0x10,%xmm15,%xmm9,%xmm9
+	vmovdqu	32-32(%r9),%xmm15
+	vpxor	%xmm2,%xmm8,%xmm7
+	vpxor	%xmm4,%xmm9,%xmm6
+
+	vmovdqu	16-32(%r9),%xmm0
+	vpxor	%xmm5,%xmm7,%xmm9
+	vpclmulqdq	$0x00,%xmm3,%xmm14,%xmm4
+	vpxor	%xmm9,%xmm6,%xmm6
+	vpunpckhqdq	%xmm13,%xmm13,%xmm2
+	vpclmulqdq	$0x11,%xmm3,%xmm14,%xmm14
+	vpxor	%xmm13,%xmm2,%xmm2
+	vpslldq	$8,%xmm6,%xmm9
+	vpclmulqdq	$0x00,%xmm15,%xmm1,%xmm1
+	vpxor	%xmm9,%xmm5,%xmm8
+	vpsrldq	$8,%xmm6,%xmm6
+	vpxor	%xmm6,%xmm7,%xmm7
+
+	vpclmulqdq	$0x00,%xmm0,%xmm13,%xmm5
+	vmovdqu	48-32(%r9),%xmm3
+	vpxor	%xmm4,%xmm5,%xmm5
+	vpunpckhqdq	%xmm12,%xmm12,%xmm9
+	vpclmulqdq	$0x11,%xmm0,%xmm13,%xmm13
+	vpxor	%xmm12,%xmm9,%xmm9
+	vpxor	%xmm14,%xmm13,%xmm13
+	vpalignr	$8,%xmm8,%xmm8,%xmm14
+	vpclmulqdq	$0x10,%xmm15,%xmm2,%xmm2
+	vmovdqu	80-32(%r9),%xmm15
+	vpxor	%xmm1,%xmm2,%xmm2
+
+	vpclmulqdq	$0x00,%xmm3,%xmm12,%xmm4
+	vmovdqu	64-32(%r9),%xmm0
+	vpxor	%xmm5,%xmm4,%xmm4
+	vpunpckhqdq	%xmm11,%xmm11,%xmm1
+	vpclmulqdq	$0x11,%xmm3,%xmm12,%xmm12
+	vpxor	%xmm11,%xmm1,%xmm1
+	vpxor	%xmm13,%xmm12,%xmm12
+	vxorps	16(%rsp),%xmm7,%xmm7
+	vpclmulqdq	$0x00,%xmm15,%xmm9,%xmm9
+	vpxor	%xmm2,%xmm9,%xmm9
+
+	vpclmulqdq	$0x10,16(%r11),%xmm8,%xmm8
+	vxorps	%xmm14,%xmm8,%xmm8
+
+	vpclmulqdq	$0x00,%xmm0,%xmm11,%xmm5
+	vmovdqu	96-32(%r9),%xmm3
+	vpxor	%xmm4,%xmm5,%xmm5
+	vpunpckhqdq	%xmm10,%xmm10,%xmm2
+	vpclmulqdq	$0x11,%xmm0,%xmm11,%xmm11
+	vpxor	%xmm10,%xmm2,%xmm2
+	vpalignr	$8,%xmm8,%xmm8,%xmm14
+	vpxor	%xmm12,%xmm11,%xmm11
+	vpclmulqdq	$0x10,%xmm15,%xmm1,%xmm1
+	vmovdqu	128-32(%r9),%xmm15
+	vpxor	%xmm9,%xmm1,%xmm1
+
+	vxorps	%xmm7,%xmm14,%xmm14
+	vpclmulqdq	$0x10,16(%r11),%xmm8,%xmm8
+	vxorps	%xmm14,%xmm8,%xmm8
+
+	vpclmulqdq	$0x00,%xmm3,%xmm10,%xmm4
+	vmovdqu	112-32(%r9),%xmm0
+	vpxor	%xmm5,%xmm4,%xmm4
+	vpunpckhqdq	%xmm8,%xmm8,%xmm9
+	vpclmulqdq	$0x11,%xmm3,%xmm10,%xmm10
+	vpxor	%xmm8,%xmm9,%xmm9
+	vpxor	%xmm11,%xmm10,%xmm10
+	vpclmulqdq	$0x00,%xmm15,%xmm2,%xmm2
+	vpxor	%xmm1,%xmm2,%xmm2
+
+	vpclmulqdq	$0x00,%xmm0,%xmm8,%xmm5
+	vpclmulqdq	$0x11,%xmm0,%xmm8,%xmm7
+	vpxor	%xmm4,%xmm5,%xmm5
+	vpclmulqdq	$0x10,%xmm15,%xmm9,%xmm6
+	vpxor	%xmm10,%xmm7,%xmm7
+	vpxor	%xmm2,%xmm6,%xmm6
+
+	vpxor	%xmm5,%xmm7,%xmm4
+	vpxor	%xmm4,%xmm6,%xmm6
+	vpslldq	$8,%xmm6,%xmm1
+	vmovdqu	16(%r11),%xmm3
+	vpsrldq	$8,%xmm6,%xmm6
+	vpxor	%xmm1,%xmm5,%xmm8
+	vpxor	%xmm6,%xmm7,%xmm7
+
+	vpalignr	$8,%xmm8,%xmm8,%xmm2
+	vpclmulqdq	$0x10,%xmm3,%xmm8,%xmm8
+	vpxor	%xmm2,%xmm8,%xmm8
+
+	vpalignr	$8,%xmm8,%xmm8,%xmm2
+	vpclmulqdq	$0x10,%xmm3,%xmm8,%xmm8
+	vpxor	%xmm7,%xmm2,%xmm2
+	vpxor	%xmm2,%xmm8,%xmm8
+	vpshufb	(%r11),%xmm8,%xmm8
+	vmovdqu	%xmm8,-64(%r9)
+
+	vzeroupper
+	movq	-48(%rax),%r15
+.cfi_restore	%r15
+	movq	-40(%rax),%r14
+.cfi_restore	%r14
+	movq	-32(%rax),%r13
+.cfi_restore	%r13
+	movq	-24(%rax),%r12
+.cfi_restore	%r12
+	movq	-16(%rax),%rbp
+.cfi_restore	%rbp
+	movq	-8(%rax),%rbx
+.cfi_restore	%rbx
+	leaq	(%rax),%rsp
+.cfi_def_cfa_register	%rsp
+.Lgcm_enc_abort:
+	movq	%r10,%rax
+	.byte	0xf3,0xc3
+.cfi_endproc
+.size	aesni_gcm_encrypt,.-aesni_gcm_encrypt
+
+/* Some utility routines */
+
+/*
+ * clear all fpu registers
+ * void clear_fpu_regs_avx(void);
+ */
+.globl	clear_fpu_regs_avx
+.type	clear_fpu_regs_avx,@function
+.align	32
+clear_fpu_regs_avx:
+	vzeroall
+	ret
+.size	clear_fpu_regs_avx,.-clear_fpu_regs_avx
+
+/*
+ * void gcm_xor_avx(const uint8_t *src, uint8_t *dst);
+ *
+ * XORs one pair of unaligned 128-bit blocks from `src' and `dst' and
+ * stores the result at `dst'. The XOR is performed using FPU registers,
+ * so make sure FPU state is saved when running this in the kernel.
+ */
+.globl  gcm_xor_avx
+.type	gcm_xor_avx,@function
+.align	32
+gcm_xor_avx:
+	movdqu  (%rdi), %xmm0
+	movdqu  (%rsi), %xmm1
+	pxor    %xmm1, %xmm0
+	movdqu  %xmm0, (%rsi)
+	ret
+.size	gcm_xor_avx,.-gcm_xor_avx
+
+/*
+ * Toggle a boolean_t value atomically and return the new value.
+ * boolean_t atomic_toggle_boolean_nv(volatile boolean_t *);
+ */
+.globl	atomic_toggle_boolean_nv
+.type	atomic_toggle_boolean_nv,@function
+.align	32
+atomic_toggle_boolean_nv:
+	xorl	%eax, %eax
+	lock
+	xorl	$1, (%rdi)
+	jz	1f
+	movl	$1, %eax
+1:
+	ret
+.size	atomic_toggle_boolean_nv,.-atomic_toggle_boolean_nv
+
+.align	64
+.Lbswap_mask:
+.byte	15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0
+.Lpoly:
+.byte	0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0xc2
+.Lone_msb:
+.byte	0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1
+.Ltwo_lsb:
+.byte	2,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
+.Lone_lsb:
+.byte	1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
+.byte	65,69,83,45,78,73,32,71,67,77,32,109,111,100,117,108,101,32,102,111,114,32,120,56,54,95,54,52,44,32,67,82,89,80,84,79,71,65,77,83,32,98,121,32,60,97,112,112,114,111,64,111,112,101,110,115,115,108,46,111,114,103,62,0
+.align	64
+
+/* Mark the stack non-executable. */
+#if defined(__linux__) && defined(__ELF__)
+.section .note.GNU-stack,"",%progbits
+#endif
+
+#endif /* defined(__x86_64__) && defined(HAVE_AVX) && defined(HAVE_AES) ... */
diff --git a/module/icp/asm-x86_64/modes/ghash-x86_64.S b/module/icp/asm-x86_64/modes/ghash-x86_64.S
new file mode 100644
index 000000000..90cc36b43
--- /dev/null
+++ b/module/icp/asm-x86_64/modes/ghash-x86_64.S
@@ -0,0 +1,714 @@
+# Copyright 2010-2016 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the Apache License 2.0 (the "License").  You may not use
+# this file except in compliance with the License.  You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+#
+# ====================================================================
+# Written by Andy Polyakov <[email protected]> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+#
+# March, June 2010
+#
+# The module implements "4-bit" GCM GHASH function and underlying
+# single multiplication operation in GF(2^128). "4-bit" means that
+# it uses 256 bytes per-key table [+128 bytes shared table]. GHASH
+# function features so called "528B" variant utilizing additional
+# 256+16 bytes of per-key storage [+512 bytes shared table].
+# Performance results are for this streamed GHASH subroutine and are
+# expressed in cycles per processed byte, less is better:
+#
+#		gcc 3.4.x(*)	assembler
+#
+# P4		28.6		14.0		+100%
+# Opteron	19.3		7.7		+150%
+# Core2		17.8		8.1(**)		+120%
+# Atom		31.6		16.8		+88%
+# VIA Nano	21.8		10.1		+115%
+#
+# (*)	comparison is not completely fair, because C results are
+#	for vanilla "256B" implementation, while assembler results
+#	are for "528B";-)
+# (**)	it's mystery [to me] why Core2 result is not same as for
+#	Opteron;
+
+# May 2010
+#
+# Add PCLMULQDQ version performing at 2.02 cycles per processed byte.
+# See ghash-x86.pl for background information and details about coding
+# techniques.
+#
+# Special thanks to David Woodhouse for providing access to a
+# Westmere-based system on behalf of Intel Open Source Technology Centre.
+
+# December 2012
+#
+# Overhaul: aggregate Karatsuba post-processing, improve ILP in
+# reduction_alg9, increase reduction aggregate factor to 4x. As for
+# the latter. ghash-x86.pl discusses that it makes lesser sense to
+# increase aggregate factor. Then why increase here? Critical path
+# consists of 3 independent pclmulqdq instructions, Karatsuba post-
+# processing and reduction. "On top" of this we lay down aggregated
+# multiplication operations, triplets of independent pclmulqdq's. As
+# issue rate for pclmulqdq is limited, it makes lesser sense to
+# aggregate more multiplications than it takes to perform remaining
+# non-multiplication operations. 2x is near-optimal coefficient for
+# contemporary Intel CPUs (therefore modest improvement coefficient),
+# but not for Bulldozer. Latter is because logical SIMD operations
+# are twice as slow in comparison to Intel, so that critical path is
+# longer. A CPU with higher pclmulqdq issue rate would also benefit
+# from higher aggregate factor...
+#
+# Westmere	1.78(+13%)
+# Sandy Bridge	1.80(+8%)
+# Ivy Bridge	1.80(+7%)
+# Haswell	0.55(+93%) (if system doesn't support AVX)
+# Broadwell	0.45(+110%)(if system doesn't support AVX)
+# Skylake	0.44(+110%)(if system doesn't support AVX)
+# Bulldozer	1.49(+27%)
+# Silvermont	2.88(+13%)
+# Knights L	2.12(-)    (if system doesn't support AVX)
+# Goldmont	1.08(+24%)
+
+# March 2013
+#
+# ... 8x aggregate factor AVX code path is using reduction algorithm
+# suggested by Shay Gueron[1]. Even though contemporary AVX-capable
+# CPUs such as Sandy and Ivy Bridge can execute it, the code performs
+# sub-optimally in comparison to above mentioned version. But thanks
+# to Ilya Albrekht and Max Locktyukhin of Intel Corp. we knew that
+# it performs in 0.41 cycles per byte on Haswell processor, in
+# 0.29 on Broadwell, and in 0.36 on Skylake.
+#
+# Knights Landing achieves 1.09 cpb.
+#
+# [1] http://rt.openssl.org/Ticket/Display.html?id=2900&user=guest&pass=guest
+
+# Generated once from
+# https://github.com/openssl/openssl/blob/5ffc3324/crypto/modes/asm/ghash-x86_64.pl
+# and modified for ICP. Modification are kept at a bare minimum to ease later
+# upstream merges.
+
+#if defined(__x86_64__) && defined(HAVE_AVX) && \
+    defined(HAVE_AES) && defined(HAVE_PCLMULQDQ)
+
+.text
+
+.globl	gcm_gmult_clmul
+.type	gcm_gmult_clmul,@function
+.align	16
+gcm_gmult_clmul:
+.cfi_startproc
+.L_gmult_clmul:
+	movdqu	(%rdi),%xmm0
+	movdqa	.Lbswap_mask(%rip),%xmm5
+	movdqu	(%rsi),%xmm2
+	movdqu	32(%rsi),%xmm4
+.byte	102,15,56,0,197
+	movdqa	%xmm0,%xmm1
+	pshufd	$78,%xmm0,%xmm3
+	pxor	%xmm0,%xmm3
+.byte	102,15,58,68,194,0
+.byte	102,15,58,68,202,17
+.byte	102,15,58,68,220,0
+	pxor	%xmm0,%xmm3
+	pxor	%xmm1,%xmm3
+
+	movdqa	%xmm3,%xmm4
+	psrldq	$8,%xmm3
+	pslldq	$8,%xmm4
+	pxor	%xmm3,%xmm1
+	pxor	%xmm4,%xmm0
+
+	movdqa	%xmm0,%xmm4
+	movdqa	%xmm0,%xmm3
+	psllq	$5,%xmm0
+	pxor	%xmm0,%xmm3
+	psllq	$1,%xmm0
+	pxor	%xmm3,%xmm0
+	psllq	$57,%xmm0
+	movdqa	%xmm0,%xmm3
+	pslldq	$8,%xmm0
+	psrldq	$8,%xmm3
+	pxor	%xmm4,%xmm0
+	pxor	%xmm3,%xmm1
+
+
+	movdqa	%xmm0,%xmm4
+	psrlq	$1,%xmm0
+	pxor	%xmm4,%xmm1
+	pxor	%xmm0,%xmm4
+	psrlq	$5,%xmm0
+	pxor	%xmm4,%xmm0
+	psrlq	$1,%xmm0
+	pxor	%xmm1,%xmm0
+.byte	102,15,56,0,197
+	movdqu	%xmm0,(%rdi)
+	.byte	0xf3,0xc3
+.cfi_endproc
+.size	gcm_gmult_clmul,.-gcm_gmult_clmul
+
+.globl	gcm_init_htab_avx
+.type	gcm_init_htab_avx,@function
+.align	32
+gcm_init_htab_avx:
+.cfi_startproc
+	vzeroupper
+
+	vmovdqu	(%rsi),%xmm2
+	// KCF/ICP stores H in network byte order with the hi qword first
+	// so we need to swap all bytes, not the 2 qwords.
+	vmovdqu	.Lbswap_mask(%rip),%xmm4
+	vpshufb	%xmm4,%xmm2,%xmm2
+
+
+	vpshufd	$255,%xmm2,%xmm4
+	vpsrlq	$63,%xmm2,%xmm3
+	vpsllq	$1,%xmm2,%xmm2
+	vpxor	%xmm5,%xmm5,%xmm5
+	vpcmpgtd	%xmm4,%xmm5,%xmm5
+	vpslldq	$8,%xmm3,%xmm3
+	vpor	%xmm3,%xmm2,%xmm2
+
+
+	vpand	.L0x1c2_polynomial(%rip),%xmm5,%xmm5
+	vpxor	%xmm5,%xmm2,%xmm2
+
+	vpunpckhqdq	%xmm2,%xmm2,%xmm6
+	vmovdqa	%xmm2,%xmm0
+	vpxor	%xmm2,%xmm6,%xmm6
+	movq	$4,%r10
+	jmp	.Linit_start_avx
+.align	32
+.Linit_loop_avx:
+	vpalignr	$8,%xmm3,%xmm4,%xmm5
+	vmovdqu	%xmm5,-16(%rdi)
+	vpunpckhqdq	%xmm0,%xmm0,%xmm3
+	vpxor	%xmm0,%xmm3,%xmm3
+	vpclmulqdq	$0x11,%xmm2,%xmm0,%xmm1
+	vpclmulqdq	$0x00,%xmm2,%xmm0,%xmm0
+	vpclmulqdq	$0x00,%xmm6,%xmm3,%xmm3
+	vpxor	%xmm0,%xmm1,%xmm4
+	vpxor	%xmm4,%xmm3,%xmm3
+
+	vpslldq	$8,%xmm3,%xmm4
+	vpsrldq	$8,%xmm3,%xmm3
+	vpxor	%xmm4,%xmm0,%xmm0
+	vpxor	%xmm3,%xmm1,%xmm1
+	vpsllq	$57,%xmm0,%xmm3
+	vpsllq	$62,%xmm0,%xmm4
+	vpxor	%xmm3,%xmm4,%xmm4
+	vpsllq	$63,%xmm0,%xmm3
+	vpxor	%xmm3,%xmm4,%xmm4
+	vpslldq	$8,%xmm4,%xmm3
+	vpsrldq	$8,%xmm4,%xmm4
+	vpxor	%xmm3,%xmm0,%xmm0
+	vpxor	%xmm4,%xmm1,%xmm1
+
+	vpsrlq	$1,%xmm0,%xmm4
+	vpxor	%xmm0,%xmm1,%xmm1
+	vpxor	%xmm4,%xmm0,%xmm0
+	vpsrlq	$5,%xmm4,%xmm4
+	vpxor	%xmm4,%xmm0,%xmm0
+	vpsrlq	$1,%xmm0,%xmm0
+	vpxor	%xmm1,%xmm0,%xmm0
+.Linit_start_avx:
+	vmovdqa	%xmm0,%xmm5
+	vpunpckhqdq	%xmm0,%xmm0,%xmm3
+	vpxor	%xmm0,%xmm3,%xmm3
+	vpclmulqdq	$0x11,%xmm2,%xmm0,%xmm1
+	vpclmulqdq	$0x00,%xmm2,%xmm0,%xmm0
+	vpclmulqdq	$0x00,%xmm6,%xmm3,%xmm3
+	vpxor	%xmm0,%xmm1,%xmm4
+	vpxor	%xmm4,%xmm3,%xmm3
+
+	vpslldq	$8,%xmm3,%xmm4
+	vpsrldq	$8,%xmm3,%xmm3
+	vpxor	%xmm4,%xmm0,%xmm0
+	vpxor	%xmm3,%xmm1,%xmm1
+	vpsllq	$57,%xmm0,%xmm3
+	vpsllq	$62,%xmm0,%xmm4
+	vpxor	%xmm3,%xmm4,%xmm4
+	vpsllq	$63,%xmm0,%xmm3
+	vpxor	%xmm3,%xmm4,%xmm4
+	vpslldq	$8,%xmm4,%xmm3
+	vpsrldq	$8,%xmm4,%xmm4
+	vpxor	%xmm3,%xmm0,%xmm0
+	vpxor	%xmm4,%xmm1,%xmm1
+
+	vpsrlq	$1,%xmm0,%xmm4
+	vpxor	%xmm0,%xmm1,%xmm1
+	vpxor	%xmm4,%xmm0,%xmm0
+	vpsrlq	$5,%xmm4,%xmm4
+	vpxor	%xmm4,%xmm0,%xmm0
+	vpsrlq	$1,%xmm0,%xmm0
+	vpxor	%xmm1,%xmm0,%xmm0
+	vpshufd	$78,%xmm5,%xmm3
+	vpshufd	$78,%xmm0,%xmm4
+	vpxor	%xmm5,%xmm3,%xmm3
+	vmovdqu	%xmm5,0(%rdi)
+	vpxor	%xmm0,%xmm4,%xmm4
+	vmovdqu	%xmm0,16(%rdi)
+	leaq	48(%rdi),%rdi
+	subq	$1,%r10
+	jnz	.Linit_loop_avx
+
+	vpalignr	$8,%xmm4,%xmm3,%xmm5
+	vmovdqu	%xmm5,-16(%rdi)
+
+	vzeroupper
+	.byte	0xf3,0xc3
+.cfi_endproc
+.size	gcm_init_htab_avx,.-gcm_init_htab_avx
+
+.globl	gcm_gmult_avx
+.type	gcm_gmult_avx,@function
+.align	32
+gcm_gmult_avx:
+.cfi_startproc
+	jmp	.L_gmult_clmul
+.cfi_endproc
+.size	gcm_gmult_avx,.-gcm_gmult_avx
+.globl	gcm_ghash_avx
+.type	gcm_ghash_avx,@function
+.align	32
+gcm_ghash_avx:
+.cfi_startproc
+	vzeroupper
+
+	vmovdqu	(%rdi),%xmm10
+	leaq	.L0x1c2_polynomial(%rip),%r10
+	leaq	64(%rsi),%rsi
+	vmovdqu	.Lbswap_mask(%rip),%xmm13
+	vpshufb	%xmm13,%xmm10,%xmm10
+	cmpq	$0x80,%rcx
+	jb	.Lshort_avx
+	subq	$0x80,%rcx
+
+	vmovdqu	112(%rdx),%xmm14
+	vmovdqu	0-64(%rsi),%xmm6
+	vpshufb	%xmm13,%xmm14,%xmm14
+	vmovdqu	32-64(%rsi),%xmm7
+
+	vpunpckhqdq	%xmm14,%xmm14,%xmm9
+	vmovdqu	96(%rdx),%xmm15
+	vpclmulqdq	$0x00,%xmm6,%xmm14,%xmm0
+	vpxor	%xmm14,%xmm9,%xmm9
+	vpshufb	%xmm13,%xmm15,%xmm15
+	vpclmulqdq	$0x11,%xmm6,%xmm14,%xmm1
+	vmovdqu	16-64(%rsi),%xmm6
+	vpunpckhqdq	%xmm15,%xmm15,%xmm8
+	vmovdqu	80(%rdx),%xmm14
+	vpclmulqdq	$0x00,%xmm7,%xmm9,%xmm2
+	vpxor	%xmm15,%xmm8,%xmm8
+
+	vpshufb	%xmm13,%xmm14,%xmm14
+	vpclmulqdq	$0x00,%xmm6,%xmm15,%xmm3
+	vpunpckhqdq	%xmm14,%xmm14,%xmm9
+	vpclmulqdq	$0x11,%xmm6,%xmm15,%xmm4
+	vmovdqu	48-64(%rsi),%xmm6
+	vpxor	%xmm14,%xmm9,%xmm9
+	vmovdqu	64(%rdx),%xmm15
+	vpclmulqdq	$0x10,%xmm7,%xmm8,%xmm5
+	vmovdqu	80-64(%rsi),%xmm7
+
+	vpshufb	%xmm13,%xmm15,%xmm15
+	vpxor	%xmm0,%xmm3,%xmm3
+	vpclmulqdq	$0x00,%xmm6,%xmm14,%xmm0
+	vpxor	%xmm1,%xmm4,%xmm4
+	vpunpckhqdq	%xmm15,%xmm15,%xmm8
+	vpclmulqdq	$0x11,%xmm6,%xmm14,%xmm1
+	vmovdqu	64-64(%rsi),%xmm6
+	vpxor	%xmm2,%xmm5,%xmm5
+	vpclmulqdq	$0x00,%xmm7,%xmm9,%xmm2
+	vpxor	%xmm15,%xmm8,%xmm8
+
+	vmovdqu	48(%rdx),%xmm14
+	vpxor	%xmm3,%xmm0,%xmm0
+	vpclmulqdq	$0x00,%xmm6,%xmm15,%xmm3
+	vpxor	%xmm4,%xmm1,%xmm1
+	vpshufb	%xmm13,%xmm14,%xmm14
+	vpclmulqdq	$0x11,%xmm6,%xmm15,%xmm4
+	vmovdqu	96-64(%rsi),%xmm6
+	vpxor	%xmm5,%xmm2,%xmm2
+	vpunpckhqdq	%xmm14,%xmm14,%xmm9
+	vpclmulqdq	$0x10,%xmm7,%xmm8,%xmm5
+	vmovdqu	128-64(%rsi),%xmm7
+	vpxor	%xmm14,%xmm9,%xmm9
+
+	vmovdqu	32(%rdx),%xmm15
+	vpxor	%xmm0,%xmm3,%xmm3
+	vpclmulqdq	$0x00,%xmm6,%xmm14,%xmm0
+	vpxor	%xmm1,%xmm4,%xmm4
+	vpshufb	%xmm13,%xmm15,%xmm15
+	vpclmulqdq	$0x11,%xmm6,%xmm14,%xmm1
+	vmovdqu	112-64(%rsi),%xmm6
+	vpxor	%xmm2,%xmm5,%xmm5
+	vpunpckhqdq	%xmm15,%xmm15,%xmm8
+	vpclmulqdq	$0x00,%xmm7,%xmm9,%xmm2
+	vpxor	%xmm15,%xmm8,%xmm8
+
+	vmovdqu	16(%rdx),%xmm14
+	vpxor	%xmm3,%xmm0,%xmm0
+	vpclmulqdq	$0x00,%xmm6,%xmm15,%xmm3
+	vpxor	%xmm4,%xmm1,%xmm1
+	vpshufb	%xmm13,%xmm14,%xmm14
+	vpclmulqdq	$0x11,%xmm6,%xmm15,%xmm4
+	vmovdqu	144-64(%rsi),%xmm6
+	vpxor	%xmm5,%xmm2,%xmm2
+	vpunpckhqdq	%xmm14,%xmm14,%xmm9
+	vpclmulqdq	$0x10,%xmm7,%xmm8,%xmm5
+	vmovdqu	176-64(%rsi),%xmm7
+	vpxor	%xmm14,%xmm9,%xmm9
+
+	vmovdqu	(%rdx),%xmm15
+	vpxor	%xmm0,%xmm3,%xmm3
+	vpclmulqdq	$0x00,%xmm6,%xmm14,%xmm0
+	vpxor	%xmm1,%xmm4,%xmm4
+	vpshufb	%xmm13,%xmm15,%xmm15
+	vpclmulqdq	$0x11,%xmm6,%xmm14,%xmm1
+	vmovdqu	160-64(%rsi),%xmm6
+	vpxor	%xmm2,%xmm5,%xmm5
+	vpclmulqdq	$0x10,%xmm7,%xmm9,%xmm2
+
+	leaq	128(%rdx),%rdx
+	cmpq	$0x80,%rcx
+	jb	.Ltail_avx
+
+	vpxor	%xmm10,%xmm15,%xmm15
+	subq	$0x80,%rcx
+	jmp	.Loop8x_avx
+
+.align	32
+.Loop8x_avx:
+	vpunpckhqdq	%xmm15,%xmm15,%xmm8
+	vmovdqu	112(%rdx),%xmm14
+	vpxor	%xmm0,%xmm3,%xmm3
+	vpxor	%xmm15,%xmm8,%xmm8
+	vpclmulqdq	$0x00,%xmm6,%xmm15,%xmm10
+	vpshufb	%xmm13,%xmm14,%xmm14
+	vpxor	%xmm1,%xmm4,%xmm4
+	vpclmulqdq	$0x11,%xmm6,%xmm15,%xmm11
+	vmovdqu	0-64(%rsi),%xmm6
+	vpunpckhqdq	%xmm14,%xmm14,%xmm9
+	vpxor	%xmm2,%xmm5,%xmm5
+	vpclmulqdq	$0x00,%xmm7,%xmm8,%xmm12
+	vmovdqu	32-64(%rsi),%xmm7
+	vpxor	%xmm14,%xmm9,%xmm9
+
+	vmovdqu	96(%rdx),%xmm15
+	vpclmulqdq	$0x00,%xmm6,%xmm14,%xmm0
+	vpxor	%xmm3,%xmm10,%xmm10
+	vpshufb	%xmm13,%xmm15,%xmm15
+	vpclmulqdq	$0x11,%xmm6,%xmm14,%xmm1
+	vxorps	%xmm4,%xmm11,%xmm11
+	vmovdqu	16-64(%rsi),%xmm6
+	vpunpckhqdq	%xmm15,%xmm15,%xmm8
+	vpclmulqdq	$0x00,%xmm7,%xmm9,%xmm2
+	vpxor	%xmm5,%xmm12,%xmm12
+	vxorps	%xmm15,%xmm8,%xmm8
+
+	vmovdqu	80(%rdx),%xmm14
+	vpxor	%xmm10,%xmm12,%xmm12
+	vpclmulqdq	$0x00,%xmm6,%xmm15,%xmm3
+	vpxor	%xmm11,%xmm12,%xmm12
+	vpslldq	$8,%xmm12,%xmm9
+	vpxor	%xmm0,%xmm3,%xmm3
+	vpclmulqdq	$0x11,%xmm6,%xmm15,%xmm4
+	vpsrldq	$8,%xmm12,%xmm12
+	vpxor	%xmm9,%xmm10,%xmm10
+	vmovdqu	48-64(%rsi),%xmm6
+	vpshufb	%xmm13,%xmm14,%xmm14
+	vxorps	%xmm12,%xmm11,%xmm11
+	vpxor	%xmm1,%xmm4,%xmm4
+	vpunpckhqdq	%xmm14,%xmm14,%xmm9
+	vpclmulqdq	$0x10,%xmm7,%xmm8,%xmm5
+	vmovdqu	80-64(%rsi),%xmm7
+	vpxor	%xmm14,%xmm9,%xmm9
+	vpxor	%xmm2,%xmm5,%xmm5
+
+	vmovdqu	64(%rdx),%xmm15
+	vpalignr	$8,%xmm10,%xmm10,%xmm12
+	vpclmulqdq	$0x00,%xmm6,%xmm14,%xmm0
+	vpshufb	%xmm13,%xmm15,%xmm15
+	vpxor	%xmm3,%xmm0,%xmm0
+	vpclmulqdq	$0x11,%xmm6,%xmm14,%xmm1
+	vmovdqu	64-64(%rsi),%xmm6
+	vpunpckhqdq	%xmm15,%xmm15,%xmm8
+	vpxor	%xmm4,%xmm1,%xmm1
+	vpclmulqdq	$0x00,%xmm7,%xmm9,%xmm2
+	vxorps	%xmm15,%xmm8,%xmm8
+	vpxor	%xmm5,%xmm2,%xmm2
+
+	vmovdqu	48(%rdx),%xmm14
+	vpclmulqdq	$0x10,(%r10),%xmm10,%xmm10
+	vpclmulqdq	$0x00,%xmm6,%xmm15,%xmm3
+	vpshufb	%xmm13,%xmm14,%xmm14
+	vpxor	%xmm0,%xmm3,%xmm3
+	vpclmulqdq	$0x11,%xmm6,%xmm15,%xmm4
+	vmovdqu	96-64(%rsi),%xmm6
+	vpunpckhqdq	%xmm14,%xmm14,%xmm9
+	vpxor	%xmm1,%xmm4,%xmm4
+	vpclmulqdq	$0x10,%xmm7,%xmm8,%xmm5
+	vmovdqu	128-64(%rsi),%xmm7
+	vpxor	%xmm14,%xmm9,%xmm9
+	vpxor	%xmm2,%xmm5,%xmm5
+
+	vmovdqu	32(%rdx),%xmm15
+	vpclmulqdq	$0x00,%xmm6,%xmm14,%xmm0
+	vpshufb	%xmm13,%xmm15,%xmm15
+	vpxor	%xmm3,%xmm0,%xmm0
+	vpclmulqdq	$0x11,%xmm6,%xmm14,%xmm1
+	vmovdqu	112-64(%rsi),%xmm6
+	vpunpckhqdq	%xmm15,%xmm15,%xmm8
+	vpxor	%xmm4,%xmm1,%xmm1
+	vpclmulqdq	$0x00,%xmm7,%xmm9,%xmm2
+	vpxor	%xmm15,%xmm8,%xmm8
+	vpxor	%xmm5,%xmm2,%xmm2
+	vxorps	%xmm12,%xmm10,%xmm10
+
+	vmovdqu	16(%rdx),%xmm14
+	vpalignr	$8,%xmm10,%xmm10,%xmm12
+	vpclmulqdq	$0x00,%xmm6,%xmm15,%xmm3
+	vpshufb	%xmm13,%xmm14,%xmm14
+	vpxor	%xmm0,%xmm3,%xmm3
+	vpclmulqdq	$0x11,%xmm6,%xmm15,%xmm4
+	vmovdqu	144-64(%rsi),%xmm6
+	vpclmulqdq	$0x10,(%r10),%xmm10,%xmm10
+	vxorps	%xmm11,%xmm12,%xmm12
+	vpunpckhqdq	%xmm14,%xmm14,%xmm9
+	vpxor	%xmm1,%xmm4,%xmm4
+	vpclmulqdq	$0x10,%xmm7,%xmm8,%xmm5
+	vmovdqu	176-64(%rsi),%xmm7
+	vpxor	%xmm14,%xmm9,%xmm9
+	vpxor	%xmm2,%xmm5,%xmm5
+
+	vmovdqu	(%rdx),%xmm15
+	vpclmulqdq	$0x00,%xmm6,%xmm14,%xmm0
+	vpshufb	%xmm13,%xmm15,%xmm15
+	vpclmulqdq	$0x11,%xmm6,%xmm14,%xmm1
+	vmovdqu	160-64(%rsi),%xmm6
+	vpxor	%xmm12,%xmm15,%xmm15
+	vpclmulqdq	$0x10,%xmm7,%xmm9,%xmm2
+	vpxor	%xmm10,%xmm15,%xmm15
+
+	leaq	128(%rdx),%rdx
+	subq	$0x80,%rcx
+	jnc	.Loop8x_avx
+
+	addq	$0x80,%rcx
+	jmp	.Ltail_no_xor_avx
+
+.align	32
+.Lshort_avx:
+	vmovdqu	-16(%rdx,%rcx,1),%xmm14
+	leaq	(%rdx,%rcx,1),%rdx
+	vmovdqu	0-64(%rsi),%xmm6
+	vmovdqu	32-64(%rsi),%xmm7
+	vpshufb	%xmm13,%xmm14,%xmm15
+
+	vmovdqa	%xmm0,%xmm3
+	vmovdqa	%xmm1,%xmm4
+	vmovdqa	%xmm2,%xmm5
+	subq	$0x10,%rcx
+	jz	.Ltail_avx
+
+	vpunpckhqdq	%xmm15,%xmm15,%xmm8
+	vpxor	%xmm0,%xmm3,%xmm3
+	vpclmulqdq	$0x00,%xmm6,%xmm15,%xmm0
+	vpxor	%xmm15,%xmm8,%xmm8
+	vmovdqu	-32(%rdx),%xmm14
+	vpxor	%xmm1,%xmm4,%xmm4
+	vpclmulqdq	$0x11,%xmm6,%xmm15,%xmm1
+	vmovdqu	16-64(%rsi),%xmm6
+	vpshufb	%xmm13,%xmm14,%xmm15
+	vpxor	%xmm2,%xmm5,%xmm5
+	vpclmulqdq	$0x00,%xmm7,%xmm8,%xmm2
+	vpsrldq	$8,%xmm7,%xmm7
+	subq	$0x10,%rcx
+	jz	.Ltail_avx
+
+	vpunpckhqdq	%xmm15,%xmm15,%xmm8
+	vpxor	%xmm0,%xmm3,%xmm3
+	vpclmulqdq	$0x00,%xmm6,%xmm15,%xmm0
+	vpxor	%xmm15,%xmm8,%xmm8
+	vmovdqu	-48(%rdx),%xmm14
+	vpxor	%xmm1,%xmm4,%xmm4
+	vpclmulqdq	$0x11,%xmm6,%xmm15,%xmm1
+	vmovdqu	48-64(%rsi),%xmm6
+	vpshufb	%xmm13,%xmm14,%xmm15
+	vpxor	%xmm2,%xmm5,%xmm5
+	vpclmulqdq	$0x00,%xmm7,%xmm8,%xmm2
+	vmovdqu	80-64(%rsi),%xmm7
+	subq	$0x10,%rcx
+	jz	.Ltail_avx
+
+	vpunpckhqdq	%xmm15,%xmm15,%xmm8
+	vpxor	%xmm0,%xmm3,%xmm3
+	vpclmulqdq	$0x00,%xmm6,%xmm15,%xmm0
+	vpxor	%xmm15,%xmm8,%xmm8
+	vmovdqu	-64(%rdx),%xmm14
+	vpxor	%xmm1,%xmm4,%xmm4
+	vpclmulqdq	$0x11,%xmm6,%xmm15,%xmm1
+	vmovdqu	64-64(%rsi),%xmm6
+	vpshufb	%xmm13,%xmm14,%xmm15
+	vpxor	%xmm2,%xmm5,%xmm5
+	vpclmulqdq	$0x00,%xmm7,%xmm8,%xmm2
+	vpsrldq	$8,%xmm7,%xmm7
+	subq	$0x10,%rcx
+	jz	.Ltail_avx
+
+	vpunpckhqdq	%xmm15,%xmm15,%xmm8
+	vpxor	%xmm0,%xmm3,%xmm3
+	vpclmulqdq	$0x00,%xmm6,%xmm15,%xmm0
+	vpxor	%xmm15,%xmm8,%xmm8
+	vmovdqu	-80(%rdx),%xmm14
+	vpxor	%xmm1,%xmm4,%xmm4
+	vpclmulqdq	$0x11,%xmm6,%xmm15,%xmm1
+	vmovdqu	96-64(%rsi),%xmm6
+	vpshufb	%xmm13,%xmm14,%xmm15
+	vpxor	%xmm2,%xmm5,%xmm5
+	vpclmulqdq	$0x00,%xmm7,%xmm8,%xmm2
+	vmovdqu	128-64(%rsi),%xmm7
+	subq	$0x10,%rcx
+	jz	.Ltail_avx
+
+	vpunpckhqdq	%xmm15,%xmm15,%xmm8
+	vpxor	%xmm0,%xmm3,%xmm3
+	vpclmulqdq	$0x00,%xmm6,%xmm15,%xmm0
+	vpxor	%xmm15,%xmm8,%xmm8
+	vmovdqu	-96(%rdx),%xmm14
+	vpxor	%xmm1,%xmm4,%xmm4
+	vpclmulqdq	$0x11,%xmm6,%xmm15,%xmm1
+	vmovdqu	112-64(%rsi),%xmm6
+	vpshufb	%xmm13,%xmm14,%xmm15
+	vpxor	%xmm2,%xmm5,%xmm5
+	vpclmulqdq	$0x00,%xmm7,%xmm8,%xmm2
+	vpsrldq	$8,%xmm7,%xmm7
+	subq	$0x10,%rcx
+	jz	.Ltail_avx
+
+	vpunpckhqdq	%xmm15,%xmm15,%xmm8
+	vpxor	%xmm0,%xmm3,%xmm3
+	vpclmulqdq	$0x00,%xmm6,%xmm15,%xmm0
+	vpxor	%xmm15,%xmm8,%xmm8
+	vmovdqu	-112(%rdx),%xmm14
+	vpxor	%xmm1,%xmm4,%xmm4
+	vpclmulqdq	$0x11,%xmm6,%xmm15,%xmm1
+	vmovdqu	144-64(%rsi),%xmm6
+	vpshufb	%xmm13,%xmm14,%xmm15
+	vpxor	%xmm2,%xmm5,%xmm5
+	vpclmulqdq	$0x00,%xmm7,%xmm8,%xmm2
+	vmovq	184-64(%rsi),%xmm7
+	subq	$0x10,%rcx
+	jmp	.Ltail_avx
+
+.align	32
+.Ltail_avx:
+	vpxor	%xmm10,%xmm15,%xmm15
+.Ltail_no_xor_avx:
+	vpunpckhqdq	%xmm15,%xmm15,%xmm8
+	vpxor	%xmm0,%xmm3,%xmm3
+	vpclmulqdq	$0x00,%xmm6,%xmm15,%xmm0
+	vpxor	%xmm15,%xmm8,%xmm8
+	vpxor	%xmm1,%xmm4,%xmm4
+	vpclmulqdq	$0x11,%xmm6,%xmm15,%xmm1
+	vpxor	%xmm2,%xmm5,%xmm5
+	vpclmulqdq	$0x00,%xmm7,%xmm8,%xmm2
+
+	vmovdqu	(%r10),%xmm12
+
+	vpxor	%xmm0,%xmm3,%xmm10
+	vpxor	%xmm1,%xmm4,%xmm11
+	vpxor	%xmm2,%xmm5,%xmm5
+
+	vpxor	%xmm10,%xmm5,%xmm5
+	vpxor	%xmm11,%xmm5,%xmm5
+	vpslldq	$8,%xmm5,%xmm9
+	vpsrldq	$8,%xmm5,%xmm5
+	vpxor	%xmm9,%xmm10,%xmm10
+	vpxor	%xmm5,%xmm11,%xmm11
+
+	vpclmulqdq	$0x10,%xmm12,%xmm10,%xmm9
+	vpalignr	$8,%xmm10,%xmm10,%xmm10
+	vpxor	%xmm9,%xmm10,%xmm10
+
+	vpclmulqdq	$0x10,%xmm12,%xmm10,%xmm9
+	vpalignr	$8,%xmm10,%xmm10,%xmm10
+	vpxor	%xmm11,%xmm10,%xmm10
+	vpxor	%xmm9,%xmm10,%xmm10
+
+	cmpq	$0,%rcx
+	jne	.Lshort_avx
+
+	vpshufb	%xmm13,%xmm10,%xmm10
+	vmovdqu	%xmm10,(%rdi)
+	vzeroupper
+	.byte	0xf3,0xc3
+.cfi_endproc
+.size	gcm_ghash_avx,.-gcm_ghash_avx
+.align	64
+.Lbswap_mask:
+.byte	15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0
+.L0x1c2_polynomial:
+.byte	1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0xc2
+.L7_mask:
+.long	7,0,7,0
+.L7_mask_poly:
+.long	7,0,450,0
+.align	64
+.type	.Lrem_4bit,@object
+.Lrem_4bit:
+.long	0,0,0,471859200,0,943718400,0,610271232
+.long	0,1887436800,0,1822425088,0,1220542464,0,1423966208
+.long	0,3774873600,0,4246732800,0,3644850176,0,3311403008
+.long	0,2441084928,0,2376073216,0,2847932416,0,3051356160
+.type	.Lrem_8bit,@object
+.Lrem_8bit:
+.value	0x0000,0x01C2,0x0384,0x0246,0x0708,0x06CA,0x048C,0x054E
+.value	0x0E10,0x0FD2,0x0D94,0x0C56,0x0918,0x08DA,0x0A9C,0x0B5E
+.value	0x1C20,0x1DE2,0x1FA4,0x1E66,0x1B28,0x1AEA,0x18AC,0x196E
+.value	0x1230,0x13F2,0x11B4,0x1076,0x1538,0x14FA,0x16BC,0x177E
+.value	0x3840,0x3982,0x3BC4,0x3A06,0x3F48,0x3E8A,0x3CCC,0x3D0E
+.value	0x3650,0x3792,0x35D4,0x3416,0x3158,0x309A,0x32DC,0x331E
+.value	0x2460,0x25A2,0x27E4,0x2626,0x2368,0x22AA,0x20EC,0x212E
+.value	0x2A70,0x2BB2,0x29F4,0x2836,0x2D78,0x2CBA,0x2EFC,0x2F3E
+.value	0x7080,0x7142,0x7304,0x72C6,0x7788,0x764A,0x740C,0x75CE
+.value	0x7E90,0x7F52,0x7D14,0x7CD6,0x7998,0x785A,0x7A1C,0x7BDE
+.value	0x6CA0,0x6D62,0x6F24,0x6EE6,0x6BA8,0x6A6A,0x682C,0x69EE
+.value	0x62B0,0x6372,0x6134,0x60F6,0x65B8,0x647A,0x663C,0x67FE
+.value	0x48C0,0x4902,0x4B44,0x4A86,0x4FC8,0x4E0A,0x4C4C,0x4D8E
+.value	0x46D0,0x4712,0x4554,0x4496,0x41D8,0x401A,0x425C,0x439E
+.value	0x54E0,0x5522,0x5764,0x56A6,0x53E8,0x522A,0x506C,0x51AE
+.value	0x5AF0,0x5B32,0x5974,0x58B6,0x5DF8,0x5C3A,0x5E7C,0x5FBE
+.value	0xE100,0xE0C2,0xE284,0xE346,0xE608,0xE7CA,0xE58C,0xE44E
+.value	0xEF10,0xEED2,0xEC94,0xED56,0xE818,0xE9DA,0xEB9C,0xEA5E
+.value	0xFD20,0xFCE2,0xFEA4,0xFF66,0xFA28,0xFBEA,0xF9AC,0xF86E
+.value	0xF330,0xF2F2,0xF0B4,0xF176,0xF438,0xF5FA,0xF7BC,0xF67E
+.value	0xD940,0xD882,0xDAC4,0xDB06,0xDE48,0xDF8A,0xDDCC,0xDC0E
+.value	0xD750,0xD692,0xD4D4,0xD516,0xD058,0xD19A,0xD3DC,0xD21E
+.value	0xC560,0xC4A2,0xC6E4,0xC726,0xC268,0xC3AA,0xC1EC,0xC02E
+.value	0xCB70,0xCAB2,0xC8F4,0xC936,0xCC78,0xCDBA,0xCFFC,0xCE3E
+.value	0x9180,0x9042,0x9204,0x93C6,0x9688,0x974A,0x950C,0x94CE
+.value	0x9F90,0x9E52,0x9C14,0x9DD6,0x9898,0x995A,0x9B1C,0x9ADE
+.value	0x8DA0,0x8C62,0x8E24,0x8FE6,0x8AA8,0x8B6A,0x892C,0x88EE
+.value	0x83B0,0x8272,0x8034,0x81F6,0x84B8,0x857A,0x873C,0x86FE
+.value	0xA9C0,0xA802,0xAA44,0xAB86,0xAEC8,0xAF0A,0xAD4C,0xAC8E
+.value	0xA7D0,0xA612,0xA454,0xA596,0xA0D8,0xA11A,0xA35C,0xA29E
+.value	0xB5E0,0xB422,0xB664,0xB7A6,0xB2E8,0xB32A,0xB16C,0xB0AE
+.value	0xBBF0,0xBA32,0xB874,0xB9B6,0xBCF8,0xBD3A,0xBF7C,0xBEBE
+
+.byte	71,72,65,83,72,32,102,111,114,32,120,56,54,95,54,52,44,32,67,82,89,80,84,79,71,65,77,83,32,98,121,32,60,97,112,112,114,111,64,111,112,101,110,115,115,108,46,111,114,103,62,0
+.align	64
+
+/* Mark the stack non-executable. */
+#if defined(__linux__) && defined(__ELF__)
+.section .note.GNU-stack,"",%progbits
+#endif
+
+#endif /* defined(__x86_64__) && defined(HAVE_AVX) && defined(HAVE_AES) ... */