s390x: Add optimized chacha20

It adds vectorized ChaCha20 implementation based on libgcrypt
cipher/chacha20-s390x.S.  The final state register clearing is
omitted.

On a z15 it shows the following improvements (using formatted
bench-arc4random data):

GENERIC                                    MB/s
-----------------------------------------------
arc4random [single-thread]               198.92
arc4random_buf(16) [single-thread]       244.49
arc4random_buf(32) [single-thread]       282.73
arc4random_buf(48) [single-thread]       286.64
arc4random_buf(64) [single-thread]       320.06
arc4random_buf(80) [single-thread]       297.43
arc4random_buf(96) [single-thread]       310.96
arc4random_buf(112) [single-thread]      308.10
arc4random_buf(128) [single-thread]      309.90
-----------------------------------------------

VX.                                        MB/s
-----------------------------------------------
arc4random [single-thread]               430.26
arc4random_buf(16) [single-thread]       735.14
arc4random_buf(32) [single-thread]      1029.99
arc4random_buf(48) [single-thread]      1206.76
arc4random_buf(64) [single-thread]      1311.92
arc4random_buf(80) [single-thread]      1378.74
arc4random_buf(96) [single-thread]      1445.06
arc4random_buf(112) [single-thread]     1484.32
arc4random_buf(128) [single-thread]     1517.30
-----------------------------------------------

Checked on s390x-linux-gnu.

4 files changed, 626 insertions, 1 deletions

diff --git a/LICENSES b/LICENSES
index e177af6..cd04fb6 100644
--- a/LICENSES
+++ b/LICENSES
@@ -392,7 +392,8 @@ Copyright 2001 by Stephen L. Moshier <moshier@na-net.ornl.gov>
 
 sysdeps/aarch64/chacha20-aarch64.S, sysdeps/x86_64/chacha20-amd64-sse2.S,
 sysdeps/x86_64/chacha20-amd64-avx2.S, and
-sysdeps/powerpc/powerpc64/power8/chacha20-ppc.c imports code from libgcrypt,
+sysdeps/powerpc/powerpc64/power8/chacha20-ppc.c, and
+sysdeps/s390/s390-64/chacha20-s390x.S imports code from libgcrypt,
 with the following notices:
 
 Copyright (C) 2017-2019 Jussi Kivilinna <jussi.kivilinna@iki.fi>
diff --git a/sysdeps/s390/s390-64/Makefile b/sysdeps/s390/s390-64/Makefile
index 66ed844..96c110f 100644
--- a/sysdeps/s390/s390-64/Makefile
+++ b/sysdeps/s390/s390-64/Makefile
@@ -67,3 +67,9 @@ tests-container += tst-glibc-hwcaps-cache
 endif
 
 endif # $(subdir) == elf
+
+ifeq ($(subdir),stdlib)
+sysdep_routines += \
+  chacha20-s390x \
+  # sysdep_routines
+endif
diff --git a/sysdeps/s390/s390-64/chacha20-s390x.S b/sysdeps/s390/s390-64/chacha20-s390x.S
new file mode 100644
index 0000000..e38504d
--- /dev/null
+++ b/sysdeps/s390/s390-64/chacha20-s390x.S
@@ -0,0 +1,573 @@
+/* Optimized s390x implementation of ChaCha20 cipher.
+   Copyright (C) 2022 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <https://www.gnu.org/licenses/>.  */
+
+/* chacha20-s390x.S  -  zSeries implementation of ChaCha20 cipher
+
+   Copyright (C) 2020 Jussi Kivilinna <jussi.kivilinna@iki.fi>
+
+   This file is part of Libgcrypt.
+
+   Libgcrypt is free software; you can redistribute it and/or modify
+   it under the terms of the GNU Lesser General Public License as
+   published by the Free Software Foundation; either version 2.1 of
+   the License, or (at your option) any later version.
+
+   Libgcrypt is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with this program; if not, see <https://www.gnu.org/licenses/>.
+ */
+
+#include <sysdep.h>
+
+#ifdef HAVE_S390_VX_ASM_SUPPORT
+
+/* CFA expressions are used for pointing CFA and registers to
+ * SP relative offsets. */
+# define DW_REGNO_SP 15
+
+/* Fixed length encoding used for integers for now. */
+# define DW_SLEB128_7BIT(value) \
+        0x00|((value) & 0x7f)
+# define DW_SLEB128_28BIT(value) \
+        0x80|((value)&0x7f), \
+        0x80|(((value)>>7)&0x7f), \
+        0x80|(((value)>>14)&0x7f), \
+        0x00|(((value)>>21)&0x7f)
+
+# define cfi_cfa_on_stack(rsp_offs,cfa_depth) \
+        .cfi_escape \
+          0x0f, /* DW_CFA_def_cfa_expression */ \
+            DW_SLEB128_7BIT(11), /* length */ \
+          0x7f, /* DW_OP_breg15, rsp + constant */ \
+            DW_SLEB128_28BIT(rsp_offs), \
+          0x06, /* DW_OP_deref */ \
+          0x23, /* DW_OP_plus_constu */ \
+            DW_SLEB128_28BIT((cfa_depth)+160)
+
+.machine "z13+vx"
+.text
+
+.balign 16
+.Lconsts:
+.Lwordswap:
+	.byte 12, 13, 14, 15, 8, 9, 10, 11, 4, 5, 6, 7, 0, 1, 2, 3
+.Lbswap128:
+	.byte 15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0
+.Lbswap32:
+	.byte 3, 2, 1, 0, 7, 6, 5, 4, 11, 10, 9, 8, 15, 14, 13, 12
+.Lone:
+	.long 0, 0, 0, 1
+.Ladd_counter_0123:
+	.long 0, 1, 2, 3
+.Ladd_counter_4567:
+	.long 4, 5, 6, 7
+
+/* register macros */
+#define INPUT %r2
+#define DST   %r3
+#define SRC   %r4
+#define NBLKS %r0
+#define ROUND %r1
+
+/* stack structure */
+
+#define STACK_FRAME_STD    (8 * 16 + 8 * 4)
+#define STACK_FRAME_F8_F15 (8 * 8)
+#define STACK_FRAME_Y0_Y15 (16 * 16)
+#define STACK_FRAME_CTR    (4 * 16)
+#define STACK_FRAME_PARAMS (6 * 8)
+
+#define STACK_MAX   (STACK_FRAME_STD + STACK_FRAME_F8_F15 + \
+		     STACK_FRAME_Y0_Y15 + STACK_FRAME_CTR + \
+		     STACK_FRAME_PARAMS)
+
+#define STACK_F8     (STACK_MAX - STACK_FRAME_F8_F15)
+#define STACK_F9     (STACK_F8 + 8)
+#define STACK_F10    (STACK_F9 + 8)
+#define STACK_F11    (STACK_F10 + 8)
+#define STACK_F12    (STACK_F11 + 8)
+#define STACK_F13    (STACK_F12 + 8)
+#define STACK_F14    (STACK_F13 + 8)
+#define STACK_F15    (STACK_F14 + 8)
+#define STACK_Y0_Y15 (STACK_F8 - STACK_FRAME_Y0_Y15)
+#define STACK_CTR    (STACK_Y0_Y15 - STACK_FRAME_CTR)
+#define STACK_INPUT  (STACK_CTR - STACK_FRAME_PARAMS)
+#define STACK_DST    (STACK_INPUT + 8)
+#define STACK_SRC    (STACK_DST + 8)
+#define STACK_NBLKS  (STACK_SRC + 8)
+#define STACK_POCTX  (STACK_NBLKS + 8)
+#define STACK_POSRC  (STACK_POCTX + 8)
+
+#define STACK_G0_H3  STACK_Y0_Y15
+
+/* vector registers */
+#define A0 %v0
+#define A1 %v1
+#define A2 %v2
+#define A3 %v3
+
+#define B0 %v4
+#define B1 %v5
+#define B2 %v6
+#define B3 %v7
+
+#define C0 %v8
+#define C1 %v9
+#define C2 %v10
+#define C3 %v11
+
+#define D0 %v12
+#define D1 %v13
+#define D2 %v14
+#define D3 %v15
+
+#define E0 %v16
+#define E1 %v17
+#define E2 %v18
+#define E3 %v19
+
+#define F0 %v20
+#define F1 %v21
+#define F2 %v22
+#define F3 %v23
+
+#define G0 %v24
+#define G1 %v25
+#define G2 %v26
+#define G3 %v27
+
+#define H0 %v28
+#define H1 %v29
+#define H2 %v30
+#define H3 %v31
+
+#define IO0 E0
+#define IO1 E1
+#define IO2 E2
+#define IO3 E3
+#define IO4 F0
+#define IO5 F1
+#define IO6 F2
+#define IO7 F3
+
+#define S0 G0
+#define S1 G1
+#define S2 G2
+#define S3 G3
+
+#define TMP0 H0
+#define TMP1 H1
+#define TMP2 H2
+#define TMP3 H3
+
+#define X0 A0
+#define X1 A1
+#define X2 A2
+#define X3 A3
+#define X4 B0
+#define X5 B1
+#define X6 B2
+#define X7 B3
+#define X8 C0
+#define X9 C1
+#define X10 C2
+#define X11 C3
+#define X12 D0
+#define X13 D1
+#define X14 D2
+#define X15 D3
+
+#define Y0 E0
+#define Y1 E1
+#define Y2 E2
+#define Y3 E3
+#define Y4 F0
+#define Y5 F1
+#define Y6 F2
+#define Y7 F3
+#define Y8 G0
+#define Y9 G1
+#define Y10 G2
+#define Y11 G3
+#define Y12 H0
+#define Y13 H1
+#define Y14 H2
+#define Y15 H3
+
+/**********************************************************************
+  helper macros
+ **********************************************************************/
+
+#define _ /*_*/
+
+#define START_STACK(last_r) \
+	lgr %r0, %r15; \
+	lghi %r1, ~15; \
+	stmg %r6, last_r, 6 * 8(%r15); \
+	aghi %r0, -STACK_MAX; \
+	ngr %r0, %r1; \
+	lgr %r1, %r15; \
+	cfi_def_cfa_register(1); \
+	lgr %r15, %r0; \
+	stg %r1, 0(%r15); \
+	cfi_cfa_on_stack(0, 0); \
+	std %f8, STACK_F8(%r15); \
+	std %f9, STACK_F9(%r15); \
+	std %f10, STACK_F10(%r15); \
+	std %f11, STACK_F11(%r15); \
+	std %f12, STACK_F12(%r15); \
+	std %f13, STACK_F13(%r15); \
+	std %f14, STACK_F14(%r15); \
+	std %f15, STACK_F15(%r15);
+
+#define END_STACK(last_r) \
+	lg %r1, 0(%r15); \
+	ld %f8, STACK_F8(%r15); \
+	ld %f9, STACK_F9(%r15); \
+	ld %f10, STACK_F10(%r15); \
+	ld %f11, STACK_F11(%r15); \
+	ld %f12, STACK_F12(%r15); \
+	ld %f13, STACK_F13(%r15); \
+	ld %f14, STACK_F14(%r15); \
+	ld %f15, STACK_F15(%r15); \
+	lmg %r6, last_r, 6 * 8(%r1); \
+	lgr %r15, %r1; \
+	cfi_def_cfa_register(DW_REGNO_SP);
+
+#define PLUS(dst,src) \
+	vaf dst, dst, src;
+
+#define XOR(dst,src) \
+	vx dst, dst, src;
+
+#define ROTATE(v1,c) \
+	verllf v1, v1, (c)(0);
+
+#define WORD_ROTATE(v1,s) \
+	vsldb v1, v1, v1, ((s) * 4);
+
+#define DST_8(OPER, I, J) \
+	OPER(A##I, J); OPER(B##I, J); OPER(C##I, J); OPER(D##I, J); \
+	OPER(E##I, J); OPER(F##I, J); OPER(G##I, J); OPER(H##I, J);
+
+/**********************************************************************
+  round macros
+ **********************************************************************/
+
+/**********************************************************************
+  8-way chacha20 ("vertical")
+ **********************************************************************/
+
+#define QUARTERROUND4_V8_POLY(x0,x1,x2,x3,x4,x5,x6,x7,\
+			      x8,x9,x10,x11,x12,x13,x14,x15,\
+			      y0,y1,y2,y3,y4,y5,y6,y7,\
+			      y8,y9,y10,y11,y12,y13,y14,y15,\
+			      op1,op2,op3,op4,op5,op6,op7,op8,\
+			      op9,op10,op11,op12) \
+	op1;							\
+	PLUS(x0, x1); PLUS(x4, x5);				\
+	PLUS(x8, x9); PLUS(x12, x13);				\
+	PLUS(y0, y1); PLUS(y4, y5);				\
+	PLUS(y8, y9); PLUS(y12, y13);				\
+	    op2;						\
+	    XOR(x3, x0);  XOR(x7, x4);				\
+	    XOR(x11, x8); XOR(x15, x12);			\
+	    XOR(y3, y0);  XOR(y7, y4);				\
+	    XOR(y11, y8); XOR(y15, y12);			\
+		op3;						\
+		ROTATE(x3, 16); ROTATE(x7, 16);			\
+		ROTATE(x11, 16); ROTATE(x15, 16);		\
+		ROTATE(y3, 16); ROTATE(y7, 16);			\
+		ROTATE(y11, 16); ROTATE(y15, 16);		\
+	op4;							\
+	PLUS(x2, x3); PLUS(x6, x7);				\
+	PLUS(x10, x11); PLUS(x14, x15);				\
+	PLUS(y2, y3); PLUS(y6, y7);				\
+	PLUS(y10, y11); PLUS(y14, y15);				\
+	    op5;						\
+	    XOR(x1, x2); XOR(x5, x6);				\
+	    XOR(x9, x10); XOR(x13, x14);			\
+	    XOR(y1, y2); XOR(y5, y6);				\
+	    XOR(y9, y10); XOR(y13, y14);			\
+		op6;						\
+		ROTATE(x1,12); ROTATE(x5,12);			\
+		ROTATE(x9,12); ROTATE(x13,12);			\
+		ROTATE(y1,12); ROTATE(y5,12);			\
+		ROTATE(y9,12); ROTATE(y13,12);			\
+	op7;							\
+	PLUS(x0, x1); PLUS(x4, x5);				\
+	PLUS(x8, x9); PLUS(x12, x13);				\
+	PLUS(y0, y1); PLUS(y4, y5);				\
+	PLUS(y8, y9); PLUS(y12, y13);				\
+	    op8;						\
+	    XOR(x3, x0); XOR(x7, x4);				\
+	    XOR(x11, x8); XOR(x15, x12);			\
+	    XOR(y3, y0); XOR(y7, y4);				\
+	    XOR(y11, y8); XOR(y15, y12);			\
+		op9;						\
+		ROTATE(x3,8); ROTATE(x7,8);			\
+		ROTATE(x11,8); ROTATE(x15,8);			\
+		ROTATE(y3,8); ROTATE(y7,8);			\
+		ROTATE(y11,8); ROTATE(y15,8);			\
+	op10;							\
+	PLUS(x2, x3); PLUS(x6, x7);				\
+	PLUS(x10, x11); PLUS(x14, x15);				\
+	PLUS(y2, y3); PLUS(y6, y7);				\
+	PLUS(y10, y11); PLUS(y14, y15);				\
+	    op11;						\
+	    XOR(x1, x2); XOR(x5, x6);				\
+	    XOR(x9, x10); XOR(x13, x14);			\
+	    XOR(y1, y2); XOR(y5, y6);				\
+	    XOR(y9, y10); XOR(y13, y14);			\
+		op12;						\
+		ROTATE(x1,7); ROTATE(x5,7);			\
+		ROTATE(x9,7); ROTATE(x13,7);			\
+		ROTATE(y1,7); ROTATE(y5,7);			\
+		ROTATE(y9,7); ROTATE(y13,7);
+
+#define QUARTERROUND4_V8(x0,x1,x2,x3,x4,x5,x6,x7,x8,x9,x10,x11,x12,x13,x14,x15,\
+			 y0,y1,y2,y3,y4,y5,y6,y7,y8,y9,y10,y11,y12,y13,y14,y15) \
+	QUARTERROUND4_V8_POLY(x0,x1,x2,x3,x4,x5,x6,x7,\
+			      x8,x9,x10,x11,x12,x13,x14,x15,\
+			      y0,y1,y2,y3,y4,y5,y6,y7,\
+			      y8,y9,y10,y11,y12,y13,y14,y15,\
+			      ,,,,,,,,,,,)
+
+#define TRANSPOSE_4X4_2(v0,v1,v2,v3,va,vb,vc,vd,tmp0,tmp1,tmp2,tmpa,tmpb,tmpc) \
+	  vmrhf tmp0, v0, v1;					\
+	  vmrhf tmp1, v2, v3;					\
+	  vmrlf tmp2, v0, v1;					\
+	  vmrlf   v3, v2, v3;					\
+	  vmrhf tmpa, va, vb;					\
+	  vmrhf tmpb, vc, vd;					\
+	  vmrlf tmpc, va, vb;					\
+	  vmrlf   vd, vc, vd;					\
+	  vpdi v0, tmp0, tmp1, 0;				\
+	  vpdi v1, tmp0, tmp1, 5;				\
+	  vpdi v2, tmp2,   v3, 0;				\
+	  vpdi v3, tmp2,   v3, 5;				\
+	  vpdi va, tmpa, tmpb, 0;				\
+	  vpdi vb, tmpa, tmpb, 5;				\
+	  vpdi vc, tmpc,   vd, 0;				\
+	  vpdi vd, tmpc,   vd, 5;
+
+.balign 8
+.globl __chacha20_s390x_vx_blocks8
+ENTRY (__chacha20_s390x_vx_blocks8)
+	/* input:
+	 *	%r2: input
+	 *	%r3: dst
+	 *	%r4: src
+	 *	%r5: nblks (multiple of 8)
+	 */
+
+	START_STACK(%r8);
+	lgr NBLKS, %r5;
+
+	larl %r7, .Lconsts;
+
+	/* Load counter. */
+	lg %r8, (12 * 4)(INPUT);
+	rllg %r8, %r8, 32;
+
+.balign 4
+	/* Process eight chacha20 blocks per loop. */
+.Lloop8:
+	vlm Y0, Y3, 0(INPUT);
+
+	slgfi NBLKS, 8;
+	lghi ROUND, (20 / 2);
+
+	/* Construct counter vectors X12/X13 & Y12/Y13. */
+	vl X4, (.Ladd_counter_0123 - .Lconsts)(%r7);
+	vl Y4, (.Ladd_counter_4567 - .Lconsts)(%r7);
+	vrepf Y12, Y3, 0;
+	vrepf Y13, Y3, 1;
+	vaccf X5, Y12, X4;
+	vaccf Y5, Y12, Y4;
+	vaf X12, Y12, X4;
+	vaf Y12, Y12, Y4;
+	vaf X13, Y13, X5;
+	vaf Y13, Y13, Y5;
+
+	vrepf X0, Y0, 0;
+	vrepf X1, Y0, 1;
+	vrepf X2, Y0, 2;
+	vrepf X3, Y0, 3;
+	vrepf X4, Y1, 0;
+	vrepf X5, Y1, 1;
+	vrepf X6, Y1, 2;
+	vrepf X7, Y1, 3;
+	vrepf X8, Y2, 0;
+	vrepf X9, Y2, 1;
+	vrepf X10, Y2, 2;
+	vrepf X11, Y2, 3;
+	vrepf X14, Y3, 2;
+	vrepf X15, Y3, 3;
+
+	/* Store counters for blocks 0-7. */
+	vstm X12, X13, (STACK_CTR + 0 * 16)(%r15);
+	vstm Y12, Y13, (STACK_CTR + 2 * 16)(%r15);
+
+	vlr Y0, X0;
+	vlr Y1, X1;
+	vlr Y2, X2;
+	vlr Y3, X3;
+	vlr Y4, X4;
+	vlr Y5, X5;
+	vlr Y6, X6;
+	vlr Y7, X7;
+	vlr Y8, X8;
+	vlr Y9, X9;
+	vlr Y10, X10;
+	vlr Y11, X11;
+	vlr Y14, X14;
+	vlr Y15, X15;
+
+	/* Update and store counter. */
+	agfi %r8, 8;
+	rllg %r5, %r8, 32;
+	stg %r5, (12 * 4)(INPUT);
+
+.balign 4
+.Lround2_8:
+	QUARTERROUND4_V8(X0, X4,  X8, X12,   X1, X5,  X9, X13,
+			 X2, X6, X10, X14,   X3, X7, X11, X15,
+			 Y0, Y4,  Y8, Y12,   Y1, Y5,  Y9, Y13,
+			 Y2, Y6, Y10, Y14,   Y3, Y7, Y11, Y15);
+	QUARTERROUND4_V8(X0, X5, X10, X15,   X1, X6, X11, X12,
+			 X2, X7,  X8, X13,   X3, X4,  X9, X14,
+			 Y0, Y5, Y10, Y15,   Y1, Y6, Y11, Y12,
+			 Y2, Y7,  Y8, Y13,   Y3, Y4,  Y9, Y14);
+	brctg ROUND, .Lround2_8;
+
+	/* Store blocks 4-7. */
+	vstm Y0, Y15, STACK_Y0_Y15(%r15);
+
+	/* Load counters for blocks 0-3. */
+	vlm Y0, Y1, (STACK_CTR + 0 * 16)(%r15);
+
+	lghi ROUND, 1;
+	j .Lfirst_output_4blks_8;
+
+.balign 4
+.Lsecond_output_4blks_8:
+	/* Load blocks 4-7. */
+	vlm X0, X15, STACK_Y0_Y15(%r15);
+
+	/* Load counters for blocks 4-7. */
+	vlm Y0, Y1, (STACK_CTR + 2 * 16)(%r15);
+
+	lghi ROUND, 0;
+
+.balign 4
+	/* Output four chacha20 blocks per loop. */
+.Lfirst_output_4blks_8:
+	vlm Y12, Y15, 0(INPUT);
+	PLUS(X12, Y0);
+	PLUS(X13, Y1);
+	vrepf Y0, Y12, 0;
+	vrepf Y1, Y12, 1;
+	vrepf Y2, Y12, 2;
+	vrepf Y3, Y12, 3;
+	vrepf Y4, Y13, 0;
+	vrepf Y5, Y13, 1;
+	vrepf Y6, Y13, 2;
+	vrepf Y7, Y13, 3;
+	vrepf Y8, Y14, 0;
+	vrepf Y9, Y14, 1;
+	vrepf Y10, Y14, 2;
+	vrepf Y11, Y14, 3;
+	vrepf Y14, Y15, 2;
+	vrepf Y15, Y15, 3;
+	PLUS(X0, Y0);
+	PLUS(X1, Y1);
+	PLUS(X2, Y2);
+	PLUS(X3, Y3);
+	PLUS(X4, Y4);
+	PLUS(X5, Y5);
+	PLUS(X6, Y6);
+	PLUS(X7, Y7);
+	PLUS(X8, Y8);
+	PLUS(X9, Y9);
+	PLUS(X10, Y10);
+	PLUS(X11, Y11);
+	PLUS(X14, Y14);
+	PLUS(X15, Y15);
+
+	vl Y15, (.Lbswap32 - .Lconsts)(%r7);
+	TRANSPOSE_4X4_2(X0, X1, X2, X3, X4, X5, X6, X7,
+			Y9, Y10, Y11, Y12, Y13, Y14);
+	TRANSPOSE_4X4_2(X8, X9, X10, X11, X12, X13, X14, X15,
+			Y9, Y10, Y11, Y12, Y13, Y14);
+
+	vlm Y0, Y14, 0(SRC);
+	vperm X0, X0, X0, Y15;
+	vperm X1, X1, X1, Y15;
+	vperm X2, X2, X2, Y15;
+	vperm X3, X3, X3, Y15;
+	vperm X4, X4, X4, Y15;
+	vperm X5, X5, X5, Y15;
+	vperm X6, X6, X6, Y15;
+	vperm X7, X7, X7, Y15;
+	vperm X8, X8, X8, Y15;
+	vperm X9, X9, X9, Y15;
+	vperm X10, X10, X10, Y15;
+	vperm X11, X11, X11, Y15;
+	vperm X12, X12, X12, Y15;
+	vperm X13, X13, X13, Y15;
+	vperm X14, X14, X14, Y15;
+	vperm X15, X15, X15, Y15;
+	vl Y15, (15 * 16)(SRC);
+
+	XOR(Y0, X0);
+	XOR(Y1, X4);
+	XOR(Y2, X8);
+	XOR(Y3, X12);
+	XOR(Y4, X1);
+	XOR(Y5, X5);
+	XOR(Y6, X9);
+	XOR(Y7, X13);
+	XOR(Y8, X2);
+	XOR(Y9, X6);
+	XOR(Y10, X10);
+	XOR(Y11, X14);
+	XOR(Y12, X3);
+	XOR(Y13, X7);
+	XOR(Y14, X11);
+	XOR(Y15, X15);
+	vstm Y0, Y15, 0(DST);
+
+	aghi SRC, 256;
+	aghi DST, 256;
+
+	clgije ROUND, 1, .Lsecond_output_4blks_8;
+
+	clgijhe NBLKS, 8, .Lloop8;
+
+
+	END_STACK(%r8);
+	xgr %r2, %r2;
+	br %r14;
+END (__chacha20_s390x_vx_blocks8)
+
+#endif /* HAVE_S390_VX_ASM_SUPPORT */
diff --git a/sysdeps/s390/s390-64/chacha20_arch.h b/sysdeps/s390/s390-64/chacha20_arch.h
new file mode 100644
index 0000000..0c6abf7
--- /dev/null
+++ b/sysdeps/s390/s390-64/chacha20_arch.h
@@ -0,0 +1,45 @@
+/* s390x optimization for ChaCha20.VE_S390_VX_ASM_SUPPORT
+   Copyright (C) 2022 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <https://www.gnu.org/licenses/>.  */
+
+#include <stdbool.h>
+#include <ldsodefs.h>
+#include <sys/auxv.h>
+
+unsigned int __chacha20_s390x_vx_blocks8 (uint32_t *state, uint8_t *dst,
+					  const uint8_t *src, size_t nblks)
+     attribute_hidden;
+
+static inline void
+chacha20_crypt (uint32_t *state, uint8_t *dst, const uint8_t *src,
+		size_t bytes)
+{
+#ifdef HAVE_S390_VX_ASM_SUPPORT
+  _Static_assert (CHACHA20_BUFSIZE % 8 == 0,
+		  "CHACHA20_BUFSIZE not multiple of 8");
+  _Static_assert (CHACHA20_BUFSIZE >= CHACHA20_BLOCK_SIZE * 8,
+		  "CHACHA20_BUFSIZE < CHACHA20_BLOCK_SIZE * 8");
+
+  if (GLRO(dl_hwcap) & HWCAP_S390_VX)
+    {
+      __chacha20_s390x_vx_blocks8 (state, dst, src,
+				   CHACHA20_BUFSIZE / CHACHA20_BLOCK_SIZE);
+      return;
+    }
+#endif
+  chacha20_crypt_generic (state, dst, src, bytes);
+}