crypto: x86/salsa20 - remove x86 salsa20 implementations

The x86 assembly implementations of Salsa20 use the frame base pointer
register (%ebp or %rbp), which breaks frame pointer convention and
breaks stack traces when unwinding from an interrupt in the crypto code.
Recent (v4.10+) kernels will warn about this, e.g.

WARNING: kernel stack regs at 00000000a8291e69 in syzkaller047086:4677 has bad 'bp' value 000000001077994c
[...]

But after looking into it, I believe there's very little reason to still
retain the x86 Salsa20 code.  First, these are *not* vectorized
(SSE2/SSSE3/AVX2) implementations, which would be needed to get anywhere
close to the best Salsa20 performance on any remotely modern x86
processor; they're just regular x86 assembly.  Second, it's still
unclear that anyone is actually using the kernel's Salsa20 at all,
especially given that now ChaCha20 is supported too, and with much more
efficient SSSE3 and AVX2 implementations.  Finally, in benchmarks I did
on both Intel and AMD processors with both gcc 8.1.0 and gcc 4.9.4, the
x86_64 salsa20-asm is actually slightly *slower* than salsa20-generic
(~3% slower on Skylake, ~10% slower on Zen), while the i686 salsa20-asm
is only slightly faster than salsa20-generic (~15% faster on Skylake,
~20% faster on Zen).  The gcc version made little difference.

So, the x86_64 salsa20-asm is pretty clearly useless.  That leaves just
the i686 salsa20-asm, which based on my tests provides a 15-20% speed
boost.  But that's without updating the code to not use %ebp.  And given
the maintenance cost, the small speed difference vs. salsa20-generic,
the fact that few people still use i686 kernels, the doubt that anyone
is even using the kernel's Salsa20 at all, and the fact that a SSE2
implementation would almost certainly be much faster on any remotely
modern x86 processor yet no one has cared enough to add one yet, I don't
think it's worthwhile to keep.

Thus, just remove both the x86_64 and i686 salsa20-asm implementations.

Reported-by: syzbot+ffa3a158337bbc01ff09@syzkaller.appspotmail.com
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

arch/x86/crypto/Makefile

@@ -15,7 +15,6 @@ obj-$(CONFIG_CRYPTO_GLUE_HELPER_X86) += glue_helper.o
 
 obj-$(CONFIG_CRYPTO_AES_586) += aes-i586.o
 obj-$(CONFIG_CRYPTO_TWOFISH_586) += twofish-i586.o
-obj-$(CONFIG_CRYPTO_SALSA20_586) += salsa20-i586.o
 obj-$(CONFIG_CRYPTO_SERPENT_SSE2_586) += serpent-sse2-i586.o
 
 obj-$(CONFIG_CRYPTO_AES_X86_64) += aes-x86_64.o
@@ -24,7 +23,6 @@ obj-$(CONFIG_CRYPTO_CAMELLIA_X86_64) += camellia-x86_64.o
 obj-$(CONFIG_CRYPTO_BLOWFISH_X86_64) += blowfish-x86_64.o
 obj-$(CONFIG_CRYPTO_TWOFISH_X86_64) += twofish-x86_64.o
 obj-$(CONFIG_CRYPTO_TWOFISH_X86_64_3WAY) += twofish-x86_64-3way.o
-obj-$(CONFIG_CRYPTO_SALSA20_X86_64) += salsa20-x86_64.o
 obj-$(CONFIG_CRYPTO_CHACHA20_X86_64) += chacha20-x86_64.o
 obj-$(CONFIG_CRYPTO_SERPENT_SSE2_X86_64) += serpent-sse2-x86_64.o
 obj-$(CONFIG_CRYPTO_AES_NI_INTEL) += aesni-intel.o
@@ -71,7 +69,6 @@ endif
 
 aes-i586-y := aes-i586-asm_32.o aes_glue.o
 twofish-i586-y := twofish-i586-asm_32.o twofish_glue.o
-salsa20-i586-y := salsa20-i586-asm_32.o salsa20_glue.o
 serpent-sse2-i586-y := serpent-sse2-i586-asm_32.o serpent_sse2_glue.o
 
 aes-x86_64-y := aes-x86_64-asm_64.o aes_glue.o
@@ -80,7 +77,6 @@ camellia-x86_64-y := camellia-x86_64-asm_64.o camellia_glue.o
 blowfish-x86_64-y := blowfish-x86_64-asm_64.o blowfish_glue.o
 twofish-x86_64-y := twofish-x86_64-asm_64.o twofish_glue.o
 twofish-x86_64-3way-y := twofish-x86_64-asm_64-3way.o twofish_glue_3way.o
-salsa20-x86_64-y := salsa20-x86_64-asm_64.o salsa20_glue.o
 chacha20-x86_64-y := chacha20-ssse3-x86_64.o chacha20_glue.o
 serpent-sse2-x86_64-y := serpent-sse2-x86_64-asm_64.o serpent_sse2_glue.o
 

arch/x86/crypto/salsa20-i586-asm_32.S

@@ -1,938 +0,0 @@
-# Derived from:
-#	salsa20_pm.s version 20051229
-#	D. J. Bernstein
-#	Public domain.
-
-#include <linux/linkage.h>
-
-.text
-
-# enter salsa20_encrypt_bytes
-ENTRY(salsa20_encrypt_bytes)
-	mov	%esp,%eax
-	and	$31,%eax
-	add	$256,%eax
-	sub	%eax,%esp
-	# eax_stack = eax
-	movl	%eax,80(%esp)
-	# ebx_stack = ebx
-	movl	%ebx,84(%esp)
-	# esi_stack = esi
-	movl	%esi,88(%esp)
-	# edi_stack = edi
-	movl	%edi,92(%esp)
-	# ebp_stack = ebp
-	movl	%ebp,96(%esp)
-	# x = arg1
-	movl	4(%esp,%eax),%edx
-	# m = arg2
-	movl	8(%esp,%eax),%esi
-	# out = arg3
-	movl	12(%esp,%eax),%edi
-	# bytes = arg4
-	movl	16(%esp,%eax),%ebx
-	# bytes -= 0
-	sub	$0,%ebx
-	# goto done if unsigned<=
-	jbe	._done
-._start:
-	# in0 = *(uint32 *) (x + 0)
-	movl	0(%edx),%eax
-	# in1 = *(uint32 *) (x + 4)
-	movl	4(%edx),%ecx
-	# in2 = *(uint32 *) (x + 8)
-	movl	8(%edx),%ebp
-	# j0 = in0
-	movl	%eax,164(%esp)
-	# in3 = *(uint32 *) (x + 12)
-	movl	12(%edx),%eax
-	# j1 = in1
-	movl	%ecx,168(%esp)
-	# in4 = *(uint32 *) (x + 16)
-	movl	16(%edx),%ecx
-	# j2 = in2
-	movl	%ebp,172(%esp)
-	# in5 = *(uint32 *) (x + 20)
-	movl	20(%edx),%ebp
-	# j3 = in3
-	movl	%eax,176(%esp)
-	# in6 = *(uint32 *) (x + 24)
-	movl	24(%edx),%eax
-	# j4 = in4
-	movl	%ecx,180(%esp)
-	# in7 = *(uint32 *) (x + 28)
-	movl	28(%edx),%ecx
-	# j5 = in5
-	movl	%ebp,184(%esp)
-	# in8 = *(uint32 *) (x + 32)
-	movl	32(%edx),%ebp
-	# j6 = in6
-	movl	%eax,188(%esp)
-	# in9 = *(uint32 *) (x + 36)
-	movl	36(%edx),%eax
-	# j7 = in7
-	movl	%ecx,192(%esp)
-	# in10 = *(uint32 *) (x + 40)
-	movl	40(%edx),%ecx
-	# j8 = in8
-	movl	%ebp,196(%esp)
-	# in11 = *(uint32 *) (x + 44)
-	movl	44(%edx),%ebp
-	# j9 = in9
-	movl	%eax,200(%esp)
-	# in12 = *(uint32 *) (x + 48)
-	movl	48(%edx),%eax
-	# j10 = in10
-	movl	%ecx,204(%esp)
-	# in13 = *(uint32 *) (x + 52)
-	movl	52(%edx),%ecx
-	# j11 = in11
-	movl	%ebp,208(%esp)
-	# in14 = *(uint32 *) (x + 56)
-	movl	56(%edx),%ebp
-	# j12 = in12
-	movl	%eax,212(%esp)
-	# in15 = *(uint32 *) (x + 60)
-	movl	60(%edx),%eax
-	# j13 = in13
-	movl	%ecx,216(%esp)
-	# j14 = in14
-	movl	%ebp,220(%esp)
-	# j15 = in15
-	movl	%eax,224(%esp)
-	# x_backup = x
-	movl	%edx,64(%esp)
-._bytesatleast1:
-	#   bytes - 64
-	cmp	$64,%ebx
-	#   goto nocopy if unsigned>=
-	jae	._nocopy
-	#     ctarget = out
-	movl	%edi,228(%esp)
-	#     out = &tmp
-	leal	0(%esp),%edi
-	#     i = bytes
-	mov	%ebx,%ecx
-	#     while (i) { *out++ = *m++; --i }
-	rep	movsb
-	#     out = &tmp
-	leal	0(%esp),%edi
-	#     m = &tmp
-	leal	0(%esp),%esi
-._nocopy:
-	#   out_backup = out
-	movl	%edi,72(%esp)
-	#   m_backup = m
-	movl	%esi,68(%esp)
-	#   bytes_backup = bytes
-	movl	%ebx,76(%esp)
-	#   in0 = j0
-	movl	164(%esp),%eax
-	#   in1 = j1
-	movl	168(%esp),%ecx
-	#   in2 = j2
-	movl	172(%esp),%edx
-	#   in3 = j3
-	movl	176(%esp),%ebx
-	#   x0 = in0
-	movl	%eax,100(%esp)
-	#   x1 = in1
-	movl	%ecx,104(%esp)
-	#   x2 = in2
-	movl	%edx,108(%esp)
-	#   x3 = in3
-	movl	%ebx,112(%esp)
-	#   in4 = j4
-	movl	180(%esp),%eax
-	#   in5 = j5
-	movl	184(%esp),%ecx
-	#   in6 = j6
-	movl	188(%esp),%edx
-	#   in7 = j7
-	movl	192(%esp),%ebx
-	#   x4 = in4
-	movl	%eax,116(%esp)
-	#   x5 = in5
-	movl	%ecx,120(%esp)
-	#   x6 = in6
-	movl	%edx,124(%esp)
-	#   x7 = in7
-	movl	%ebx,128(%esp)
-	#   in8 = j8
-	movl	196(%esp),%eax
-	#   in9 = j9
-	movl	200(%esp),%ecx
-	#   in10 = j10
-	movl	204(%esp),%edx
-	#   in11 = j11
-	movl	208(%esp),%ebx
-	#   x8 = in8
-	movl	%eax,132(%esp)
-	#   x9 = in9
-	movl	%ecx,136(%esp)
-	#   x10 = in10
-	movl	%edx,140(%esp)
-	#   x11 = in11
-	movl	%ebx,144(%esp)
-	#   in12 = j12
-	movl	212(%esp),%eax
-	#   in13 = j13
-	movl	216(%esp),%ecx
-	#   in14 = j14
-	movl	220(%esp),%edx
-	#   in15 = j15
-	movl	224(%esp),%ebx
-	#   x12 = in12
-	movl	%eax,148(%esp)
-	#   x13 = in13
-	movl	%ecx,152(%esp)
-	#   x14 = in14
-	movl	%edx,156(%esp)
-	#   x15 = in15
-	movl	%ebx,160(%esp)
-	#   i = 20
-	mov	$20,%ebp
-	# p = x0
-	movl	100(%esp),%eax
-	# s = x5
-	movl	120(%esp),%ecx
-	# t = x10
-	movl	140(%esp),%edx
-	# w = x15
-	movl	160(%esp),%ebx
-._mainloop:
-	# x0 = p
-	movl	%eax,100(%esp)
-	# 				x10 = t
-	movl	%edx,140(%esp)
-	# p += x12
-	addl	148(%esp),%eax
-	# 		x5 = s
-	movl	%ecx,120(%esp)
-	# 				t += x6
-	addl	124(%esp),%edx
-	# 						x15 = w
-	movl	%ebx,160(%esp)
-	# 		r = x1
-	movl	104(%esp),%esi
-	# 		r += s
-	add	%ecx,%esi
-	# 						v = x11
-	movl	144(%esp),%edi
-	# 						v += w
-	add	%ebx,%edi
-	# p <<<= 7
-	rol	$7,%eax
-	# p ^= x4
-	xorl	116(%esp),%eax
-	# 				t <<<= 7
-	rol	$7,%edx
-	# 				t ^= x14
-	xorl	156(%esp),%edx
-	# 		r <<<= 7
-	rol	$7,%esi
-	# 		r ^= x9
-	xorl	136(%esp),%esi
-	# 						v <<<= 7
-	rol	$7,%edi
-	# 						v ^= x3
-	xorl	112(%esp),%edi
-	# x4 = p
-	movl	%eax,116(%esp)
-	# 				x14 = t
-	movl	%edx,156(%esp)
-	# p += x0
-	addl	100(%esp),%eax
-	# 		x9 = r
-	movl	%esi,136(%esp)
-	# 				t += x10
-	addl	140(%esp),%edx
-	# 						x3 = v
-	movl	%edi,112(%esp)
-	# p <<<= 9
-	rol	$9,%eax
-	# p ^= x8
-	xorl	132(%esp),%eax
-	# 				t <<<= 9
-	rol	$9,%edx
-	# 				t ^= x2
-	xorl	108(%esp),%edx
-	# 		s += r
-	add	%esi,%ecx
-	# 		s <<<= 9
-	rol	$9,%ecx
-	# 		s ^= x13
-	xorl	152(%esp),%ecx
-	# 						w += v
-	add	%edi,%ebx
-	# 						w <<<= 9
-	rol	$9,%ebx
-	# 						w ^= x7
-	xorl	128(%esp),%ebx
-	# x8 = p
-	movl	%eax,132(%esp)
-	# 				x2 = t
-	movl	%edx,108(%esp)
-	# p += x4
-	addl	116(%esp),%eax
-	# 		x13 = s
-	movl	%ecx,152(%esp)
-	# 				t += x14
-	addl	156(%esp),%edx
-	# 						x7 = w
-	movl	%ebx,128(%esp)
-	# p <<<= 13
-	rol	$13,%eax
-	# p ^= x12
-	xorl	148(%esp),%eax
-	# 				t <<<= 13
-	rol	$13,%edx
-	# 				t ^= x6
-	xorl	124(%esp),%edx
-	# 		r += s
-	add	%ecx,%esi
-	# 		r <<<= 13
-	rol	$13,%esi
-	# 		r ^= x1
-	xorl	104(%esp),%esi
-	# 						v += w
-	add	%ebx,%edi
-	# 						v <<<= 13
-	rol	$13,%edi
-	# 						v ^= x11
-	xorl	144(%esp),%edi
-	# x12 = p
-	movl	%eax,148(%esp)
-	# 				x6 = t
-	movl	%edx,124(%esp)
-	# p += x8
-	addl	132(%esp),%eax
-	# 		x1 = r
-	movl	%esi,104(%esp)
-	# 				t += x2
-	addl	108(%esp),%edx
-	# 						x11 = v
-	movl	%edi,144(%esp)
-	# p <<<= 18
-	rol	$18,%eax
-	# p ^= x0
-	xorl	100(%esp),%eax
-	# 				t <<<= 18
-	rol	$18,%edx
-	# 				t ^= x10
-	xorl	140(%esp),%edx
-	# 		s += r
-	add	%esi,%ecx
-	# 		s <<<= 18
-	rol	$18,%ecx
-	# 		s ^= x5
-	xorl	120(%esp),%ecx
-	# 						w += v
-	add	%edi,%ebx
-	# 						w <<<= 18
-	rol	$18,%ebx
-	# 						w ^= x15
-	xorl	160(%esp),%ebx
-	# x0 = p
-	movl	%eax,100(%esp)
-	# 				x10 = t
-	movl	%edx,140(%esp)
-	# p += x3
-	addl	112(%esp),%eax
-	# p <<<= 7
-	rol	$7,%eax
-	# 		x5 = s
-	movl	%ecx,120(%esp)
-	# 				t += x9
-	addl	136(%esp),%edx
-	# 						x15 = w
-	movl	%ebx,160(%esp)
-	# 		r = x4
-	movl	116(%esp),%esi
-	# 		r += s
-	add	%ecx,%esi
-	# 						v = x14
-	movl	156(%esp),%edi
-	# 						v += w
-	add	%ebx,%edi
-	# p ^= x1
-	xorl	104(%esp),%eax
-	# 				t <<<= 7
-	rol	$7,%edx
-	# 				t ^= x11
-	xorl	144(%esp),%edx
-	# 		r <<<= 7
-	rol	$7,%esi
-	# 		r ^= x6
-	xorl	124(%esp),%esi
-	# 						v <<<= 7
-	rol	$7,%edi
-	# 						v ^= x12
-	xorl	148(%esp),%edi
-	# x1 = p
-	movl	%eax,104(%esp)
-	# 				x11 = t
-	movl	%edx,144(%esp)
-	# p += x0
-	addl	100(%esp),%eax
-	# 		x6 = r
-	movl	%esi,124(%esp)
-	# 				t += x10
-	addl	140(%esp),%edx
-	# 						x12 = v
-	movl	%edi,148(%esp)
-	# p <<<= 9
-	rol	$9,%eax
-	# p ^= x2
-	xorl	108(%esp),%eax
-	# 				t <<<= 9
-	rol	$9,%edx
-	# 				t ^= x8
-	xorl	132(%esp),%edx
-	# 		s += r
-	add	%esi,%ecx
-	# 		s <<<= 9
-	rol	$9,%ecx
-	# 		s ^= x7
-	xorl	128(%esp),%ecx
-	# 						w += v
-	add	%edi,%ebx
-	# 						w <<<= 9
-	rol	$9,%ebx
-	# 						w ^= x13
-	xorl	152(%esp),%ebx
-	# x2 = p
-	movl	%eax,108(%esp)
-	# 				x8 = t
-	movl	%edx,132(%esp)
-	# p += x1
-	addl	104(%esp),%eax
-	# 		x7 = s
-	movl	%ecx,128(%esp)
-	# 				t += x11
-	addl	144(%esp),%edx
-	# 						x13 = w
-	movl	%ebx,152(%esp)
-	# p <<<= 13
-	rol	$13,%eax
-	# p ^= x3
-	xorl	112(%esp),%eax
-	# 				t <<<= 13
-	rol	$13,%edx
-	# 				t ^= x9
-	xorl	136(%esp),%edx
-	# 		r += s
-	add	%ecx,%esi
-	# 		r <<<= 13
-	rol	$13,%esi
-	# 		r ^= x4
-	xorl	116(%esp),%esi
-	# 						v += w
-	add	%ebx,%edi
-	# 						v <<<= 13
-	rol	$13,%edi
-	# 						v ^= x14
-	xorl	156(%esp),%edi
-	# x3 = p
-	movl	%eax,112(%esp)
-	# 				x9 = t
-	movl	%edx,136(%esp)
-	# p += x2
-	addl	108(%esp),%eax
-	# 		x4 = r
-	movl	%esi,116(%esp)
-	# 				t += x8
-	addl	132(%esp),%edx
-	# 						x14 = v
-	movl	%edi,156(%esp)
-	# p <<<= 18
-	rol	$18,%eax
-	# p ^= x0
-	xorl	100(%esp),%eax
-	# 				t <<<= 18
-	rol	$18,%edx
-	# 				t ^= x10
-	xorl	140(%esp),%edx
-	# 		s += r
-	add	%esi,%ecx
-	# 		s <<<= 18
-	rol	$18,%ecx
-	# 		s ^= x5
-	xorl	120(%esp),%ecx
-	# 						w += v
-	add	%edi,%ebx
-	# 						w <<<= 18
-	rol	$18,%ebx
-	# 						w ^= x15
-	xorl	160(%esp),%ebx
-	# x0 = p
-	movl	%eax,100(%esp)
-	# 				x10 = t
-	movl	%edx,140(%esp)
-	# p += x12
-	addl	148(%esp),%eax
-	# 		x5 = s
-	movl	%ecx,120(%esp)
-	# 				t += x6
-	addl	124(%esp),%edx
-	# 						x15 = w
-	movl	%ebx,160(%esp)
-	# 		r = x1
-	movl	104(%esp),%esi
-	# 		r += s
-	add	%ecx,%esi
-	# 						v = x11
-	movl	144(%esp),%edi
-	# 						v += w
-	add	%ebx,%edi
-	# p <<<= 7
-	rol	$7,%eax
-	# p ^= x4
-	xorl	116(%esp),%eax
-	# 				t <<<= 7
-	rol	$7,%edx
-	# 				t ^= x14
-	xorl	156(%esp),%edx
-	# 		r <<<= 7
-	rol	$7,%esi
-	# 		r ^= x9
-	xorl	136(%esp),%esi
-	# 						v <<<= 7
-	rol	$7,%edi
-	# 						v ^= x3
-	xorl	112(%esp),%edi
-	# x4 = p
-	movl	%eax,116(%esp)
-	# 				x14 = t
-	movl	%edx,156(%esp)
-	# p += x0
-	addl	100(%esp),%eax
-	# 		x9 = r
-	movl	%esi,136(%esp)
-	# 				t += x10
-	addl	140(%esp),%edx
-	# 						x3 = v
-	movl	%edi,112(%esp)
-	# p <<<= 9
-	rol	$9,%eax
-	# p ^= x8
-	xorl	132(%esp),%eax
-	# 				t <<<= 9
-	rol	$9,%edx
-	# 				t ^= x2
-	xorl	108(%esp),%edx
-	# 		s += r
-	add	%esi,%ecx
-	# 		s <<<= 9
-	rol	$9,%ecx
-	# 		s ^= x13
-	xorl	152(%esp),%ecx
-	# 						w += v
-	add	%edi,%ebx
-	# 						w <<<= 9
-	rol	$9,%ebx
-	# 						w ^= x7
-	xorl	128(%esp),%ebx
-	# x8 = p
-	movl	%eax,132(%esp)
-	# 				x2 = t
-	movl	%edx,108(%esp)
-	# p += x4
-	addl	116(%esp),%eax
-	# 		x13 = s
-	movl	%ecx,152(%esp)
-	# 				t += x14
-	addl	156(%esp),%edx
-	# 						x7 = w
-	movl	%ebx,128(%esp)
-	# p <<<= 13
-	rol	$13,%eax
-	# p ^= x12
-	xorl	148(%esp),%eax
-	# 				t <<<= 13
-	rol	$13,%edx
-	# 				t ^= x6
-	xorl	124(%esp),%edx
-	# 		r += s
-	add	%ecx,%esi
-	# 		r <<<= 13
-	rol	$13,%esi
-	# 		r ^= x1
-	xorl	104(%esp),%esi
-	# 						v += w
-	add	%ebx,%edi
-	# 						v <<<= 13
-	rol	$13,%edi
-	# 						v ^= x11
-	xorl	144(%esp),%edi
-	# x12 = p
-	movl	%eax,148(%esp)
-	# 				x6 = t
-	movl	%edx,124(%esp)
-	# p += x8
-	addl	132(%esp),%eax
-	# 		x1 = r
-	movl	%esi,104(%esp)
-	# 				t += x2
-	addl	108(%esp),%edx
-	# 						x11 = v
-	movl	%edi,144(%esp)
-	# p <<<= 18
-	rol	$18,%eax
-	# p ^= x0
-	xorl	100(%esp),%eax
-	# 				t <<<= 18
-	rol	$18,%edx
-	# 				t ^= x10
-	xorl	140(%esp),%edx
-	# 		s += r
-	add	%esi,%ecx
-	# 		s <<<= 18
-	rol	$18,%ecx
-	# 		s ^= x5
-	xorl	120(%esp),%ecx
-	# 						w += v
-	add	%edi,%ebx
-	# 						w <<<= 18
-	rol	$18,%ebx
-	# 						w ^= x15
-	xorl	160(%esp),%ebx
-	# x0 = p
-	movl	%eax,100(%esp)
-	# 				x10 = t
-	movl	%edx,140(%esp)
-	# p += x3
-	addl	112(%esp),%eax
-	# p <<<= 7
-	rol	$7,%eax
-	# 		x5 = s
-	movl	%ecx,120(%esp)
-	# 				t += x9
-	addl	136(%esp),%edx
-	# 						x15 = w
-	movl	%ebx,160(%esp)
-	# 		r = x4
-	movl	116(%esp),%esi
-	# 		r += s
-	add	%ecx,%esi
-	# 						v = x14
-	movl	156(%esp),%edi
-	# 						v += w
-	add	%ebx,%edi
-	# p ^= x1
-	xorl	104(%esp),%eax
-	# 				t <<<= 7
-	rol	$7,%edx
-	# 				t ^= x11
-	xorl	144(%esp),%edx
-	# 		r <<<= 7
-	rol	$7,%esi
-	# 		r ^= x6
-	xorl	124(%esp),%esi
-	# 						v <<<= 7
-	rol	$7,%edi
-	# 						v ^= x12
-	xorl	148(%esp),%edi
-	# x1 = p
-	movl	%eax,104(%esp)
-	# 				x11 = t
-	movl	%edx,144(%esp)
-	# p += x0
-	addl	100(%esp),%eax
-	# 		x6 = r
-	movl	%esi,124(%esp)
-	# 				t += x10
-	addl	140(%esp),%edx
-	# 						x12 = v
-	movl	%edi,148(%esp)
-	# p <<<= 9
-	rol	$9,%eax
-	# p ^= x2
-	xorl	108(%esp),%eax
-	# 				t <<<= 9
-	rol	$9,%edx
-	# 				t ^= x8
-	xorl	132(%esp),%edx
-	# 		s += r
-	add	%esi,%ecx
-	# 		s <<<= 9
-	rol	$9,%ecx
-	# 		s ^= x7
-	xorl	128(%esp),%ecx
-	# 						w += v
-	add	%edi,%ebx
-	# 						w <<<= 9
-	rol	$9,%ebx
-	# 						w ^= x13
-	xorl	152(%esp),%ebx
-	# x2 = p
-	movl	%eax,108(%esp)
-	# 				x8 = t
-	movl	%edx,132(%esp)
-	# p += x1
-	addl	104(%esp),%eax
-	# 		x7 = s
-	movl	%ecx,128(%esp)
-	# 				t += x11
-	addl	144(%esp),%edx
-	# 						x13 = w
-	movl	%ebx,152(%esp)
-	# p <<<= 13
-	rol	$13,%eax
-	# p ^= x3
-	xorl	112(%esp),%eax
-	# 				t <<<= 13
-	rol	$13,%edx
-	# 				t ^= x9
-	xorl	136(%esp),%edx
-	# 		r += s
-	add	%ecx,%esi
-	# 		r <<<= 13
-	rol	$13,%esi
-	# 		r ^= x4
-	xorl	116(%esp),%esi
-	# 						v += w
-	add	%ebx,%edi
-	# 						v <<<= 13
-	rol	$13,%edi
-	# 						v ^= x14
-	xorl	156(%esp),%edi
-	# x3 = p
-	movl	%eax,112(%esp)
-	# 				x9 = t
-	movl	%edx,136(%esp)
-	# p += x2
-	addl	108(%esp),%eax
-	# 		x4 = r
-	movl	%esi,116(%esp)
-	# 				t += x8
-	addl	132(%esp),%edx
-	# 						x14 = v
-	movl	%edi,156(%esp)
-	# p <<<= 18
-	rol	$18,%eax
-	# p ^= x0
-	xorl	100(%esp),%eax
-	# 				t <<<= 18
-	rol	$18,%edx
-	# 				t ^= x10
-	xorl	140(%esp),%edx
-	# 		s += r
-	add	%esi,%ecx
-	# 		s <<<= 18
-	rol	$18,%ecx
-	# 		s ^= x5
-	xorl	120(%esp),%ecx
-	# 						w += v
-	add	%edi,%ebx
-	# 						w <<<= 18
-	rol	$18,%ebx
-	# 						w ^= x15
-	xorl	160(%esp),%ebx
-	# i -= 4
-	sub	$4,%ebp
-	# goto mainloop if unsigned >
-	ja	._mainloop
-	# x0 = p
-	movl	%eax,100(%esp)
-	# x5 = s
-	movl	%ecx,120(%esp)
-	# x10 = t
-	movl	%edx,140(%esp)
-	# x15 = w
-	movl	%ebx,160(%esp)
-	#   out = out_backup
-	movl	72(%esp),%edi
-	#   m = m_backup
-	movl	68(%esp),%esi
-	#   in0 = x0
-	movl	100(%esp),%eax
-	#   in1 = x1
-	movl	104(%esp),%ecx
-	#   in0 += j0
-	addl	164(%esp),%eax
-	#   in1 += j1
-	addl	168(%esp),%ecx
-	#   in0 ^= *(uint32 *) (m + 0)
-	xorl	0(%esi),%eax
-	#   in1 ^= *(uint32 *) (m + 4)
-	xorl	4(%esi),%ecx
-	#   *(uint32 *) (out + 0) = in0
-	movl	%eax,0(%edi)
-	#   *(uint32 *) (out + 4) = in1
-	movl	%ecx,4(%edi)
-	#   in2 = x2
-	movl	108(%esp),%eax
-	#   in3 = x3
-	movl	112(%esp),%ecx
-	#   in2 += j2
-	addl	172(%esp),%eax
-	#   in3 += j3
-	addl	176(%esp),%ecx
-	#   in2 ^= *(uint32 *) (m + 8)
-	xorl	8(%esi),%eax
-	#   in3 ^= *(uint32 *) (m + 12)
-	xorl	12(%esi),%ecx
-	#   *(uint32 *) (out + 8) = in2
-	movl	%eax,8(%edi)
-	#   *(uint32 *) (out + 12) = in3
-	movl	%ecx,12(%edi)
-	#   in4 = x4
-	movl	116(%esp),%eax
-	#   in5 = x5
-	movl	120(%esp),%ecx
-	#   in4 += j4
-	addl	180(%esp),%eax
-	#   in5 += j5
-	addl	184(%esp),%ecx
-	#   in4 ^= *(uint32 *) (m + 16)
-	xorl	16(%esi),%eax
-	#   in5 ^= *(uint32 *) (m + 20)
-	xorl	20(%esi),%ecx
-	#   *(uint32 *) (out + 16) = in4
-	movl	%eax,16(%edi)
-	#   *(uint32 *) (out + 20) = in5
-	movl	%ecx,20(%edi)
-	#   in6 = x6
-	movl	124(%esp),%eax
-	#   in7 = x7
-	movl	128(%esp),%ecx
-	#   in6 += j6
-	addl	188(%esp),%eax
-	#   in7 += j7
-	addl	192(%esp),%ecx
-	#   in6 ^= *(uint32 *) (m + 24)
-	xorl	24(%esi),%eax
-	#   in7 ^= *(uint32 *) (m + 28)
-	xorl	28(%esi),%ecx
-	#   *(uint32 *) (out + 24) = in6
-	movl	%eax,24(%edi)
-	#   *(uint32 *) (out + 28) = in7
-	movl	%ecx,28(%edi)
-	#   in8 = x8
-	movl	132(%esp),%eax
-	#   in9 = x9
-	movl	136(%esp),%ecx
-	#   in8 += j8
-	addl	196(%esp),%eax
-	#   in9 += j9
-	addl	200(%esp),%ecx
-	#   in8 ^= *(uint32 *) (m + 32)
-	xorl	32(%esi),%eax
-	#   in9 ^= *(uint32 *) (m + 36)
-	xorl	36(%esi),%ecx
-	#   *(uint32 *) (out + 32) = in8
-	movl	%eax,32(%edi)
-	#   *(uint32 *) (out + 36) = in9
-	movl	%ecx,36(%edi)
-	#   in10 = x10
-	movl	140(%esp),%eax
-	#   in11 = x11
-	movl	144(%esp),%ecx
-	#   in10 += j10
-	addl	204(%esp),%eax
-	#   in11 += j11
-	addl	208(%esp),%ecx
-	#   in10 ^= *(uint32 *) (m + 40)
-	xorl	40(%esi),%eax
-	#   in11 ^= *(uint32 *) (m + 44)
-	xorl	44(%esi),%ecx
-	#   *(uint32 *) (out + 40) = in10
-	movl	%eax,40(%edi)
-	#   *(uint32 *) (out + 44) = in11
-	movl	%ecx,44(%edi)
-	#   in12 = x12
-	movl	148(%esp),%eax
-	#   in13 = x13
-	movl	152(%esp),%ecx
-	#   in12 += j12
-	addl	212(%esp),%eax
-	#   in13 += j13
-	addl	216(%esp),%ecx
-	#   in12 ^= *(uint32 *) (m + 48)
-	xorl	48(%esi),%eax
-	#   in13 ^= *(uint32 *) (m + 52)
-	xorl	52(%esi),%ecx
-	#   *(uint32 *) (out + 48) = in12
-	movl	%eax,48(%edi)
-	#   *(uint32 *) (out + 52) = in13
-	movl	%ecx,52(%edi)
-	#   in14 = x14
-	movl	156(%esp),%eax
-	#   in15 = x15
-	movl	160(%esp),%ecx
-	#   in14 += j14
-	addl	220(%esp),%eax
-	#   in15 += j15
-	addl	224(%esp),%ecx
-	#   in14 ^= *(uint32 *) (m + 56)
-	xorl	56(%esi),%eax
-	#   in15 ^= *(uint32 *) (m + 60)
-	xorl	60(%esi),%ecx
-	#   *(uint32 *) (out + 56) = in14
-	movl	%eax,56(%edi)
-	#   *(uint32 *) (out + 60) = in15
-	movl	%ecx,60(%edi)
-	#   bytes = bytes_backup
-	movl	76(%esp),%ebx
-	#   in8 = j8
-	movl	196(%esp),%eax
-	#   in9 = j9
-	movl	200(%esp),%ecx
-	#   in8 += 1
-	add	$1,%eax
-	#   in9 += 0 + carry
-	adc	$0,%ecx
-	#   j8 = in8
-	movl	%eax,196(%esp)
-	#   j9 = in9
-	movl	%ecx,200(%esp)
-	#   bytes - 64
-	cmp	$64,%ebx
-	#   goto bytesatleast65 if unsigned>
-	ja	._bytesatleast65
-	#     goto bytesatleast64 if unsigned>=
-	jae	._bytesatleast64
-	#       m = out
-	mov	%edi,%esi
-	#       out = ctarget
-	movl	228(%esp),%edi
-	#       i = bytes
-	mov	%ebx,%ecx
-	#       while (i) { *out++ = *m++; --i }
-	rep	movsb
-._bytesatleast64:
-	#     x = x_backup
-	movl	64(%esp),%eax
-	#     in8 = j8
-	movl	196(%esp),%ecx
-	#     in9 = j9
-	movl	200(%esp),%edx
-	#     *(uint32 *) (x + 32) = in8
-	movl	%ecx,32(%eax)
-	#     *(uint32 *) (x + 36) = in9
-	movl	%edx,36(%eax)
-._done:
-	#     eax = eax_stack
-	movl	80(%esp),%eax
-	#     ebx = ebx_stack
-	movl	84(%esp),%ebx
-	#     esi = esi_stack
-	movl	88(%esp),%esi
-	#     edi = edi_stack
-	movl	92(%esp),%edi
-	#     ebp = ebp_stack
-	movl	96(%esp),%ebp
-	#     leave
-	add	%eax,%esp
-	ret
-._bytesatleast65:
-	#   bytes -= 64
-	sub	$64,%ebx
-	#   out += 64
-	add	$64,%edi
-	#   m += 64
-	add	$64,%esi
-	# goto bytesatleast1
-	jmp	._bytesatleast1
-ENDPROC(salsa20_encrypt_bytes)

arch/x86/crypto/salsa20-x86_64-asm_64.S

@@ -1,805 +0,0 @@
-/* SPDX-License-Identifier: GPL-2.0 */
-#include <linux/linkage.h>
-
-# enter salsa20_encrypt_bytes
-ENTRY(salsa20_encrypt_bytes)
-	mov	%rsp,%r11
-	and	$31,%r11
-	add	$256,%r11
-	sub	%r11,%rsp
-	# x = arg1
-	mov	%rdi,%r8
-	# m = arg2
-	mov	%rsi,%rsi
-	# out = arg3
-	mov	%rdx,%rdi
-	# bytes = arg4
-	mov	%rcx,%rdx
-	#               unsigned>? bytes - 0
-	cmp	$0,%rdx
-	# comment:fp stack unchanged by jump
-	# goto done if !unsigned>
-	jbe	._done
-	# comment:fp stack unchanged by fallthrough
-# start:
-._start:
-	# r11_stack = r11
-	movq	%r11,0(%rsp)
-	# r12_stack = r12
-	movq	%r12,8(%rsp)
-	# r13_stack = r13
-	movq	%r13,16(%rsp)
-	# r14_stack = r14
-	movq	%r14,24(%rsp)
-	# r15_stack = r15
-	movq	%r15,32(%rsp)
-	# rbx_stack = rbx
-	movq	%rbx,40(%rsp)
-	# rbp_stack = rbp
-	movq	%rbp,48(%rsp)
-	# in0 = *(uint64 *) (x + 0)
-	movq	0(%r8),%rcx
-	# in2 = *(uint64 *) (x + 8)
-	movq	8(%r8),%r9
-	# in4 = *(uint64 *) (x + 16)
-	movq	16(%r8),%rax
-	# in6 = *(uint64 *) (x + 24)
-	movq	24(%r8),%r10
-	# in8 = *(uint64 *) (x + 32)
-	movq	32(%r8),%r11
-	# in10 = *(uint64 *) (x + 40)
-	movq	40(%r8),%r12
-	# in12 = *(uint64 *) (x + 48)
-	movq	48(%r8),%r13
-	# in14 = *(uint64 *) (x + 56)
-	movq	56(%r8),%r14
-	# j0 = in0
-	movq	%rcx,56(%rsp)
-	# j2 = in2
-	movq	%r9,64(%rsp)
-	# j4 = in4
-	movq	%rax,72(%rsp)
-	# j6 = in6
-	movq	%r10,80(%rsp)
-	# j8 = in8
-	movq	%r11,88(%rsp)
-	# j10 = in10
-	movq	%r12,96(%rsp)
-	# j12 = in12
-	movq	%r13,104(%rsp)
-	# j14 = in14
-	movq	%r14,112(%rsp)
-	# x_backup = x
-	movq	%r8,120(%rsp)
-# bytesatleast1:
-._bytesatleast1:
-	#                   unsigned<? bytes - 64
-	cmp	$64,%rdx
-	# comment:fp stack unchanged by jump
-	#   goto nocopy if !unsigned<
-	jae	._nocopy
-	#     ctarget = out
-	movq	%rdi,128(%rsp)
-	#     out = &tmp
-	leaq	192(%rsp),%rdi
-	#     i = bytes
-	mov	%rdx,%rcx
-	#     while (i) { *out++ = *m++; --i }
-	rep	movsb
-	#     out = &tmp
-	leaq	192(%rsp),%rdi
-	#     m = &tmp
-	leaq	192(%rsp),%rsi
-	# comment:fp stack unchanged by fallthrough
-#   nocopy:
-._nocopy:
-	#   out_backup = out
-	movq	%rdi,136(%rsp)
-	#   m_backup = m
-	movq	%rsi,144(%rsp)
-	#   bytes_backup = bytes
-	movq	%rdx,152(%rsp)
-	#   x1 = j0
-	movq	56(%rsp),%rdi
-	#   x0 = x1
-	mov	%rdi,%rdx
-	#   (uint64) x1 >>= 32
-	shr	$32,%rdi
-	#   		x3 = j2
-	movq	64(%rsp),%rsi
-	#   		x2 = x3
-	mov	%rsi,%rcx
-	#   		(uint64) x3 >>= 32
-	shr	$32,%rsi
-	#   x5 = j4
-	movq	72(%rsp),%r8
-	#   x4 = x5
-	mov	%r8,%r9
-	#   (uint64) x5 >>= 32
-	shr	$32,%r8
-	#   x5_stack = x5
-	movq	%r8,160(%rsp)
-	#   		x7 = j6
-	movq	80(%rsp),%r8
-	#   		x6 = x7
-	mov	%r8,%rax
-	#   		(uint64) x7 >>= 32
-	shr	$32,%r8
-	#   x9 = j8
-	movq	88(%rsp),%r10
-	#   x8 = x9
-	mov	%r10,%r11
-	#   (uint64) x9 >>= 32
-	shr	$32,%r10
-	#   		x11 = j10
-	movq	96(%rsp),%r12
-	#   		x10 = x11
-	mov	%r12,%r13
-	#   		x10_stack = x10
-	movq	%r13,168(%rsp)
-	#   		(uint64) x11 >>= 32
-	shr	$32,%r12
-	#   x13 = j12
-	movq	104(%rsp),%r13
-	#   x12 = x13
-	mov	%r13,%r14
-	#   (uint64) x13 >>= 32
-	shr	$32,%r13
-	#   		x15 = j14
-	movq	112(%rsp),%r15
-	#   		x14 = x15
-	mov	%r15,%rbx
-	#   		(uint64) x15 >>= 32
-	shr	$32,%r15
-	#   		x15_stack = x15
-	movq	%r15,176(%rsp)
-	#   i = 20
-	mov	$20,%r15
-#   mainloop:
-._mainloop:
-	#   i_backup = i
-	movq	%r15,184(%rsp)
-	# 		x5 = x5_stack
-	movq	160(%rsp),%r15
-	# a = x12 + x0
-	lea	(%r14,%rdx),%rbp
-	# (uint32) a <<<= 7
-	rol	$7,%ebp
-	# x4 ^= a
-	xor	%rbp,%r9
-	# 		b = x1 + x5
-	lea	(%rdi,%r15),%rbp
-	# 		(uint32) b <<<= 7
-	rol	$7,%ebp
-	# 		x9 ^= b
-	xor	%rbp,%r10
-	# a = x0 + x4
-	lea	(%rdx,%r9),%rbp
-	# (uint32) a <<<= 9
-	rol	$9,%ebp
-	# x8 ^= a
-	xor	%rbp,%r11
-	# 		b = x5 + x9
-	lea	(%r15,%r10),%rbp
-	# 		(uint32) b <<<= 9
-	rol	$9,%ebp
-	# 		x13 ^= b
-	xor	%rbp,%r13
-	# a = x4 + x8
-	lea	(%r9,%r11),%rbp
-	# (uint32) a <<<= 13
-	rol	$13,%ebp
-	# x12 ^= a
-	xor	%rbp,%r14
-	# 		b = x9 + x13
-	lea	(%r10,%r13),%rbp
-	# 		(uint32) b <<<= 13
-	rol	$13,%ebp
-	# 		x1 ^= b
-	xor	%rbp,%rdi
-	# a = x8 + x12
-	lea	(%r11,%r14),%rbp
-	# (uint32) a <<<= 18
-	rol	$18,%ebp
-	# x0 ^= a
-	xor	%rbp,%rdx
-	# 		b = x13 + x1
-	lea	(%r13,%rdi),%rbp
-	# 		(uint32) b <<<= 18
-	rol	$18,%ebp
-	# 		x5 ^= b
-	xor	%rbp,%r15
-	# 				x10 = x10_stack
-	movq	168(%rsp),%rbp
-	# 		x5_stack = x5
-	movq	%r15,160(%rsp)
-	# 				c = x6 + x10
-	lea	(%rax,%rbp),%r15
-	# 				(uint32) c <<<= 7
-	rol	$7,%r15d
-	# 				x14 ^= c
-	xor	%r15,%rbx
-	# 				c = x10 + x14
-	lea	(%rbp,%rbx),%r15
-	# 				(uint32) c <<<= 9
-	rol	$9,%r15d
-	# 				x2 ^= c
-	xor	%r15,%rcx
-	# 				c = x14 + x2
-	lea	(%rbx,%rcx),%r15
-	# 				(uint32) c <<<= 13
-	rol	$13,%r15d
-	# 				x6 ^= c
-	xor	%r15,%rax
-	# 				c = x2 + x6
-	lea	(%rcx,%rax),%r15
-	# 				(uint32) c <<<= 18
-	rol	$18,%r15d
-	# 				x10 ^= c
-	xor	%r15,%rbp
-	# 						x15 = x15_stack
-	movq	176(%rsp),%r15
-	# 				x10_stack = x10
-	movq	%rbp,168(%rsp)
-	# 						d = x11 + x15
-	lea	(%r12,%r15),%rbp
-	# 						(uint32) d <<<= 7
-	rol	$7,%ebp
-	# 						x3 ^= d
-	xor	%rbp,%rsi
-	# 						d = x15 + x3
-	lea	(%r15,%rsi),%rbp
-	# 						(uint32) d <<<= 9
-	rol	$9,%ebp
-	# 						x7 ^= d
-	xor	%rbp,%r8
-	# 						d = x3 + x7
-	lea	(%rsi,%r8),%rbp
-	# 						(uint32) d <<<= 13
-	rol	$13,%ebp
-	# 						x11 ^= d
-	xor	%rbp,%r12
-	# 						d = x7 + x11
-	lea	(%r8,%r12),%rbp
-	# 						(uint32) d <<<= 18
-	rol	$18,%ebp
-	# 						x15 ^= d
-	xor	%rbp,%r15
-	# 						x15_stack = x15
-	movq	%r15,176(%rsp)
-	# 		x5 = x5_stack
-	movq	160(%rsp),%r15
-	# a = x3 + x0
-	lea	(%rsi,%rdx),%rbp
-	# (uint32) a <<<= 7
-	rol	$7,%ebp
-	# x1 ^= a
-	xor	%rbp,%rdi
-	# 		b = x4 + x5
-	lea	(%r9,%r15),%rbp
-	# 		(uint32) b <<<= 7
-	rol	$7,%ebp
-	# 		x6 ^= b
-	xor	%rbp,%rax
-	# a = x0 + x1
-	lea	(%rdx,%rdi),%rbp
-	# (uint32) a <<<= 9
-	rol	$9,%ebp
-	# x2 ^= a
-	xor	%rbp,%rcx
-	# 		b = x5 + x6
-	lea	(%r15,%rax),%rbp
-	# 		(uint32) b <<<= 9
-	rol	$9,%ebp
-	# 		x7 ^= b
-	xor	%rbp,%r8
-	# a = x1 + x2
-	lea	(%rdi,%rcx),%rbp
-	# (uint32) a <<<= 13
-	rol	$13,%ebp
-	# x3 ^= a
-	xor	%rbp,%rsi
-	# 		b = x6 + x7
-	lea	(%rax,%r8),%rbp
-	# 		(uint32) b <<<= 13
-	rol	$13,%ebp
-	# 		x4 ^= b
-	xor	%rbp,%r9
-	# a = x2 + x3
-	lea	(%rcx,%rsi),%rbp
-	# (uint32) a <<<= 18
-	rol	$18,%ebp
-	# x0 ^= a
-	xor	%rbp,%rdx
-	# 		b = x7 + x4
-	lea	(%r8,%r9),%rbp
-	# 		(uint32) b <<<= 18
-	rol	$18,%ebp
-	# 		x5 ^= b
-	xor	%rbp,%r15
-	# 				x10 = x10_stack
-	movq	168(%rsp),%rbp
-	# 		x5_stack = x5
-	movq	%r15,160(%rsp)
-	# 				c = x9 + x10
-	lea	(%r10,%rbp),%r15
-	# 				(uint32) c <<<= 7
-	rol	$7,%r15d
-	# 				x11 ^= c
-	xor	%r15,%r12
-	# 				c = x10 + x11
-	lea	(%rbp,%r12),%r15
-	# 				(uint32) c <<<= 9
-	rol	$9,%r15d
-	# 				x8 ^= c
-	xor	%r15,%r11
-	# 				c = x11 + x8
-	lea	(%r12,%r11),%r15
-	# 				(uint32) c <<<= 13
-	rol	$13,%r15d
-	# 				x9 ^= c
-	xor	%r15,%r10
-	# 				c = x8 + x9
-	lea	(%r11,%r10),%r15
-	# 				(uint32) c <<<= 18
-	rol	$18,%r15d
-	# 				x10 ^= c
-	xor	%r15,%rbp
-	# 						x15 = x15_stack
-	movq	176(%rsp),%r15
-	# 				x10_stack = x10
-	movq	%rbp,168(%rsp)
-	# 						d = x14 + x15
-	lea	(%rbx,%r15),%rbp
-	# 						(uint32) d <<<= 7
-	rol	$7,%ebp
-	# 						x12 ^= d
-	xor	%rbp,%r14
-	# 						d = x15 + x12
-	lea	(%r15,%r14),%rbp
-	# 						(uint32) d <<<= 9
-	rol	$9,%ebp
-	# 						x13 ^= d
-	xor	%rbp,%r13
-	# 						d = x12 + x13
-	lea	(%r14,%r13),%rbp
-	# 						(uint32) d <<<= 13
-	rol	$13,%ebp
-	# 						x14 ^= d
-	xor	%rbp,%rbx
-	# 						d = x13 + x14
-	lea	(%r13,%rbx),%rbp
-	# 						(uint32) d <<<= 18
-	rol	$18,%ebp
-	# 						x15 ^= d
-	xor	%rbp,%r15
-	# 						x15_stack = x15
-	movq	%r15,176(%rsp)
-	# 		x5 = x5_stack
-	movq	160(%rsp),%r15
-	# a = x12 + x0
-	lea	(%r14,%rdx),%rbp
-	# (uint32) a <<<= 7
-	rol	$7,%ebp
-	# x4 ^= a
-	xor	%rbp,%r9
-	# 		b = x1 + x5
-	lea	(%rdi,%r15),%rbp
-	# 		(uint32) b <<<= 7
-	rol	$7,%ebp
-	# 		x9 ^= b
-	xor	%rbp,%r10
-	# a = x0 + x4
-	lea	(%rdx,%r9),%rbp
-	# (uint32) a <<<= 9
-	rol	$9,%ebp
-	# x8 ^= a
-	xor	%rbp,%r11
-	# 		b = x5 + x9
-	lea	(%r15,%r10),%rbp
-	# 		(uint32) b <<<= 9
-	rol	$9,%ebp
-	# 		x13 ^= b
-	xor	%rbp,%r13
-	# a = x4 + x8
-	lea	(%r9,%r11),%rbp
-	# (uint32) a <<<= 13
-	rol	$13,%ebp
-	# x12 ^= a
-	xor	%rbp,%r14
-	# 		b = x9 + x13
-	lea	(%r10,%r13),%rbp
-	# 		(uint32) b <<<= 13
-	rol	$13,%ebp
-	# 		x1 ^= b
-	xor	%rbp,%rdi
-	# a = x8 + x12
-	lea	(%r11,%r14),%rbp
-	# (uint32) a <<<= 18
-	rol	$18,%ebp
-	# x0 ^= a
-	xor	%rbp,%rdx
-	# 		b = x13 + x1
-	lea	(%r13,%rdi),%rbp
-	# 		(uint32) b <<<= 18
-	rol	$18,%ebp
-	# 		x5 ^= b
-	xor	%rbp,%r15
-	# 				x10 = x10_stack
-	movq	168(%rsp),%rbp
-	# 		x5_stack = x5
-	movq	%r15,160(%rsp)
-	# 				c = x6 + x10
-	lea	(%rax,%rbp),%r15
-	# 				(uint32) c <<<= 7
-	rol	$7,%r15d
-	# 				x14 ^= c
-	xor	%r15,%rbx
-	# 				c = x10 + x14
-	lea	(%rbp,%rbx),%r15
-	# 				(uint32) c <<<= 9
-	rol	$9,%r15d
-	# 				x2 ^= c
-	xor	%r15,%rcx
-	# 				c = x14 + x2
-	lea	(%rbx,%rcx),%r15
-	# 				(uint32) c <<<= 13
-	rol	$13,%r15d
-	# 				x6 ^= c
-	xor	%r15,%rax
-	# 				c = x2 + x6
-	lea	(%rcx,%rax),%r15
-	# 				(uint32) c <<<= 18
-	rol	$18,%r15d
-	# 				x10 ^= c
-	xor	%r15,%rbp
-	# 						x15 = x15_stack
-	movq	176(%rsp),%r15
-	# 				x10_stack = x10
-	movq	%rbp,168(%rsp)
-	# 						d = x11 + x15
-	lea	(%r12,%r15),%rbp
-	# 						(uint32) d <<<= 7
-	rol	$7,%ebp
-	# 						x3 ^= d
-	xor	%rbp,%rsi
-	# 						d = x15 + x3
-	lea	(%r15,%rsi),%rbp
-	# 						(uint32) d <<<= 9
-	rol	$9,%ebp
-	# 						x7 ^= d
-	xor	%rbp,%r8
-	# 						d = x3 + x7
-	lea	(%rsi,%r8),%rbp
-	# 						(uint32) d <<<= 13
-	rol	$13,%ebp
-	# 						x11 ^= d
-	xor	%rbp,%r12
-	# 						d = x7 + x11
-	lea	(%r8,%r12),%rbp
-	# 						(uint32) d <<<= 18
-	rol	$18,%ebp
-	# 						x15 ^= d
-	xor	%rbp,%r15
-	# 						x15_stack = x15
-	movq	%r15,176(%rsp)
-	# 		x5 = x5_stack
-	movq	160(%rsp),%r15
-	# a = x3 + x0
-	lea	(%rsi,%rdx),%rbp
-	# (uint32) a <<<= 7
-	rol	$7,%ebp
-	# x1 ^= a
-	xor	%rbp,%rdi
-	# 		b = x4 + x5
-	lea	(%r9,%r15),%rbp
-	# 		(uint32) b <<<= 7
-	rol	$7,%ebp
-	# 		x6 ^= b
-	xor	%rbp,%rax
-	# a = x0 + x1
-	lea	(%rdx,%rdi),%rbp
-	# (uint32) a <<<= 9
-	rol	$9,%ebp
-	# x2 ^= a
-	xor	%rbp,%rcx
-	# 		b = x5 + x6
-	lea	(%r15,%rax),%rbp
-	# 		(uint32) b <<<= 9
-	rol	$9,%ebp
-	# 		x7 ^= b
-	xor	%rbp,%r8
-	# a = x1 + x2
-	lea	(%rdi,%rcx),%rbp
-	# (uint32) a <<<= 13
-	rol	$13,%ebp
-	# x3 ^= a
-	xor	%rbp,%rsi
-	# 		b = x6 + x7
-	lea	(%rax,%r8),%rbp
-	# 		(uint32) b <<<= 13
-	rol	$13,%ebp
-	# 		x4 ^= b
-	xor	%rbp,%r9
-	# a = x2 + x3
-	lea	(%rcx,%rsi),%rbp
-	# (uint32) a <<<= 18
-	rol	$18,%ebp
-	# x0 ^= a
-	xor	%rbp,%rdx
-	# 		b = x7 + x4
-	lea	(%r8,%r9),%rbp
-	# 		(uint32) b <<<= 18
-	rol	$18,%ebp
-	# 		x5 ^= b
-	xor	%rbp,%r15
-	# 				x10 = x10_stack
-	movq	168(%rsp),%rbp
-	# 		x5_stack = x5
-	movq	%r15,160(%rsp)
-	# 				c = x9 + x10
-	lea	(%r10,%rbp),%r15
-	# 				(uint32) c <<<= 7
-	rol	$7,%r15d
-	# 				x11 ^= c
-	xor	%r15,%r12
-	# 				c = x10 + x11
-	lea	(%rbp,%r12),%r15
-	# 				(uint32) c <<<= 9
-	rol	$9,%r15d
-	# 				x8 ^= c
-	xor	%r15,%r11
-	# 				c = x11 + x8
-	lea	(%r12,%r11),%r15
-	# 				(uint32) c <<<= 13
-	rol	$13,%r15d
-	# 				x9 ^= c
-	xor	%r15,%r10
-	# 				c = x8 + x9
-	lea	(%r11,%r10),%r15
-	# 				(uint32) c <<<= 18
-	rol	$18,%r15d
-	# 				x10 ^= c
-	xor	%r15,%rbp
-	# 						x15 = x15_stack
-	movq	176(%rsp),%r15
-	# 				x10_stack = x10
-	movq	%rbp,168(%rsp)
-	# 						d = x14 + x15
-	lea	(%rbx,%r15),%rbp
-	# 						(uint32) d <<<= 7
-	rol	$7,%ebp
-	# 						x12 ^= d
-	xor	%rbp,%r14
-	# 						d = x15 + x12
-	lea	(%r15,%r14),%rbp
-	# 						(uint32) d <<<= 9
-	rol	$9,%ebp
-	# 						x13 ^= d
-	xor	%rbp,%r13
-	# 						d = x12 + x13
-	lea	(%r14,%r13),%rbp
-	# 						(uint32) d <<<= 13
-	rol	$13,%ebp
-	# 						x14 ^= d
-	xor	%rbp,%rbx
-	# 						d = x13 + x14
-	lea	(%r13,%rbx),%rbp
-	# 						(uint32) d <<<= 18
-	rol	$18,%ebp
-	# 						x15 ^= d
-	xor	%rbp,%r15
-	# 						x15_stack = x15
-	movq	%r15,176(%rsp)
-	#   i = i_backup
-	movq	184(%rsp),%r15
-	#                  unsigned>? i -= 4
-	sub	$4,%r15
-	# comment:fp stack unchanged by jump
-	# goto mainloop if unsigned>
-	ja	._mainloop
-	#   (uint32) x2 += j2
-	addl	64(%rsp),%ecx
-	#   x3 <<= 32
-	shl	$32,%rsi
-	#   x3 += j2
-	addq	64(%rsp),%rsi
-	#   (uint64) x3 >>= 32
-	shr	$32,%rsi
-	#   x3 <<= 32
-	shl	$32,%rsi
-	#   x2 += x3
-	add	%rsi,%rcx
-	#   (uint32) x6 += j6
-	addl	80(%rsp),%eax
-	#   x7 <<= 32
-	shl	$32,%r8
-	#   x7 += j6
-	addq	80(%rsp),%r8
-	#   (uint64) x7 >>= 32
-	shr	$32,%r8
-	#   x7 <<= 32
-	shl	$32,%r8
-	#   x6 += x7
-	add	%r8,%rax
-	#   (uint32) x8 += j8
-	addl	88(%rsp),%r11d
-	#   x9 <<= 32
-	shl	$32,%r10
-	#   x9 += j8
-	addq	88(%rsp),%r10
-	#   (uint64) x9 >>= 32
-	shr	$32,%r10
-	#   x9 <<= 32
-	shl	$32,%r10
-	#   x8 += x9
-	add	%r10,%r11
-	#   (uint32) x12 += j12
-	addl	104(%rsp),%r14d
-	#   x13 <<= 32
-	shl	$32,%r13
-	#   x13 += j12
-	addq	104(%rsp),%r13
-	#   (uint64) x13 >>= 32
-	shr	$32,%r13
-	#   x13 <<= 32
-	shl	$32,%r13
-	#   x12 += x13
-	add	%r13,%r14
-	#   (uint32) x0 += j0
-	addl	56(%rsp),%edx
-	#   x1 <<= 32
-	shl	$32,%rdi
-	#   x1 += j0
-	addq	56(%rsp),%rdi
-	#   (uint64) x1 >>= 32
-	shr	$32,%rdi
-	#   x1 <<= 32
-	shl	$32,%rdi
-	#   x0 += x1
-	add	%rdi,%rdx
-	#   x5 = x5_stack
-	movq	160(%rsp),%rdi
-	#   (uint32) x4 += j4
-	addl	72(%rsp),%r9d
-	#   x5 <<= 32
-	shl	$32,%rdi
-	#   x5 += j4
-	addq	72(%rsp),%rdi
-	#   (uint64) x5 >>= 32
-	shr	$32,%rdi
-	#   x5 <<= 32
-	shl	$32,%rdi
-	#   x4 += x5
-	add	%rdi,%r9
-	#   x10 = x10_stack
-	movq	168(%rsp),%r8
-	#   (uint32) x10 += j10
-	addl	96(%rsp),%r8d
-	#   x11 <<= 32
-	shl	$32,%r12
-	#   x11 += j10
-	addq	96(%rsp),%r12
-	#   (uint64) x11 >>= 32
-	shr	$32,%r12
-	#   x11 <<= 32
-	shl	$32,%r12
-	#   x10 += x11
-	add	%r12,%r8
-	#   x15 = x15_stack
-	movq	176(%rsp),%rdi
-	#   (uint32) x14 += j14
-	addl	112(%rsp),%ebx
-	#   x15 <<= 32
-	shl	$32,%rdi
-	#   x15 += j14
-	addq	112(%rsp),%rdi
-	#   (uint64) x15 >>= 32
-	shr	$32,%rdi
-	#   x15 <<= 32
-	shl	$32,%rdi
-	#   x14 += x15
-	add	%rdi,%rbx
-	#   out = out_backup
-	movq	136(%rsp),%rdi
-	#   m = m_backup
-	movq	144(%rsp),%rsi
-	#   x0 ^= *(uint64 *) (m + 0)
-	xorq	0(%rsi),%rdx
-	#   *(uint64 *) (out + 0) = x0
-	movq	%rdx,0(%rdi)
-	#   x2 ^= *(uint64 *) (m + 8)
-	xorq	8(%rsi),%rcx
-	#   *(uint64 *) (out + 8) = x2
-	movq	%rcx,8(%rdi)
-	#   x4 ^= *(uint64 *) (m + 16)
-	xorq	16(%rsi),%r9
-	#   *(uint64 *) (out + 16) = x4
-	movq	%r9,16(%rdi)
-	#   x6 ^= *(uint64 *) (m + 24)
-	xorq	24(%rsi),%rax
-	#   *(uint64 *) (out + 24) = x6
-	movq	%rax,24(%rdi)
-	#   x8 ^= *(uint64 *) (m + 32)
-	xorq	32(%rsi),%r11
-	#   *(uint64 *) (out + 32) = x8
-	movq	%r11,32(%rdi)
-	#   x10 ^= *(uint64 *) (m + 40)
-	xorq	40(%rsi),%r8
-	#   *(uint64 *) (out + 40) = x10
-	movq	%r8,40(%rdi)
-	#   x12 ^= *(uint64 *) (m + 48)
-	xorq	48(%rsi),%r14
-	#   *(uint64 *) (out + 48) = x12
-	movq	%r14,48(%rdi)
-	#   x14 ^= *(uint64 *) (m + 56)
-	xorq	56(%rsi),%rbx
-	#   *(uint64 *) (out + 56) = x14
-	movq	%rbx,56(%rdi)
-	#   bytes = bytes_backup
-	movq	152(%rsp),%rdx
-	#   in8 = j8
-	movq	88(%rsp),%rcx
-	#   in8 += 1
-	add	$1,%rcx
-	#   j8 = in8
-	movq	%rcx,88(%rsp)
-	#                          unsigned>? unsigned<? bytes - 64
-	cmp	$64,%rdx
-	# comment:fp stack unchanged by jump
-	#   goto bytesatleast65 if unsigned>
-	ja	._bytesatleast65
-	# comment:fp stack unchanged by jump
-	#     goto bytesatleast64 if !unsigned<
-	jae	._bytesatleast64
-	#       m = out
-	mov	%rdi,%rsi
-	#       out = ctarget
-	movq	128(%rsp),%rdi
-	#       i = bytes
-	mov	%rdx,%rcx
-	#       while (i) { *out++ = *m++; --i }
-	rep	movsb
-	# comment:fp stack unchanged by fallthrough
-#     bytesatleast64:
-._bytesatleast64:
-	#     x = x_backup
-	movq	120(%rsp),%rdi
-	#     in8 = j8
-	movq	88(%rsp),%rsi
-	#     *(uint64 *) (x + 32) = in8
-	movq	%rsi,32(%rdi)
-	#     r11 = r11_stack
-	movq	0(%rsp),%r11
-	#     r12 = r12_stack
-	movq	8(%rsp),%r12
-	#     r13 = r13_stack
-	movq	16(%rsp),%r13
-	#     r14 = r14_stack
-	movq	24(%rsp),%r14
-	#     r15 = r15_stack
-	movq	32(%rsp),%r15
-	#     rbx = rbx_stack
-	movq	40(%rsp),%rbx
-	#     rbp = rbp_stack
-	movq	48(%rsp),%rbp
-	# comment:fp stack unchanged by fallthrough
-#     done:
-._done:
-	#     leave
-	add	%r11,%rsp
-	mov	%rdi,%rax
-	mov	%rsi,%rdx
-	ret
-#   bytesatleast65:
-._bytesatleast65:
-	#   bytes -= 64
-	sub	$64,%rdx
-	#   out += 64
-	add	$64,%rdi
-	#   m += 64
-	add	$64,%rsi
-	# comment:fp stack unchanged by jump
-	# goto bytesatleast1
-	jmp	._bytesatleast1
-ENDPROC(salsa20_encrypt_bytes)

arch/x86/crypto/salsa20_glue.c

@@ -1,91 +0,0 @@
-/*
- * Glue code for optimized assembly version of  Salsa20.
- *
- * Copyright (c) 2007 Tan Swee Heng <thesweeheng@gmail.com>
- *
- * The assembly codes are public domain assembly codes written by Daniel. J.
- * Bernstein <djb@cr.yp.to>. The codes are modified to include indentation
- * and to remove extraneous comments and functions that are not needed.
- * - i586 version, renamed as salsa20-i586-asm_32.S
- *   available from <http://cr.yp.to/snuffle/salsa20/x86-pm/salsa20.s>
- * - x86-64 version, renamed as salsa20-x86_64-asm_64.S
- *   available from <http://cr.yp.to/snuffle/salsa20/amd64-3/salsa20.s>
- *
- * Also modified to set up the initial state using the generic C code rather
- * than in assembly.
- *
- * This program is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License as published by the Free
- * Software Foundation; either version 2 of the License, or (at your option)
- * any later version.
- *
- */
-
-#include <asm/unaligned.h>
-#include <crypto/internal/skcipher.h>
-#include <crypto/salsa20.h>
-#include <linux/module.h>
-
-asmlinkage void salsa20_encrypt_bytes(u32 state[16], const u8 *src, u8 *dst,
-				      u32 bytes);
-
-static int salsa20_asm_crypt(struct skcipher_request *req)
-{
-	struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
-	const struct salsa20_ctx *ctx = crypto_skcipher_ctx(tfm);
-	struct skcipher_walk walk;
-	u32 state[16];
-	int err;
-
-	err = skcipher_walk_virt(&walk, req, true);
-
-	crypto_salsa20_init(state, ctx, walk.iv);
-
-	while (walk.nbytes > 0) {
-		unsigned int nbytes = walk.nbytes;
-
-		if (nbytes < walk.total)
-			nbytes = round_down(nbytes, walk.stride);
-
-		salsa20_encrypt_bytes(state, walk.src.virt.addr,
-				      walk.dst.virt.addr, nbytes);
-		err = skcipher_walk_done(&walk, walk.nbytes - nbytes);
-	}
-
-	return err;
-}
-
-static struct skcipher_alg alg = {
-	.base.cra_name		= "salsa20",
-	.base.cra_driver_name	= "salsa20-asm",
-	.base.cra_priority	= 200,
-	.base.cra_blocksize	= 1,
-	.base.cra_ctxsize	= sizeof(struct salsa20_ctx),
-	.base.cra_module	= THIS_MODULE,
-
-	.min_keysize		= SALSA20_MIN_KEY_SIZE,
-	.max_keysize		= SALSA20_MAX_KEY_SIZE,
-	.ivsize			= SALSA20_IV_SIZE,
-	.chunksize		= SALSA20_BLOCK_SIZE,
-	.setkey			= crypto_salsa20_setkey,
-	.encrypt		= salsa20_asm_crypt,
-	.decrypt		= salsa20_asm_crypt,
-};
-
-static int __init init(void)
-{
-	return crypto_register_skcipher(&alg);
-}
-
-static void __exit fini(void)
-{
-	crypto_unregister_skcipher(&alg);
-}
-
-module_init(init);
-module_exit(fini);
-
-MODULE_LICENSE("GPL");
-MODULE_DESCRIPTION ("Salsa20 stream cipher algorithm (optimized assembly version)");
-MODULE_ALIAS_CRYPTO("salsa20");
-MODULE_ALIAS_CRYPTO("salsa20-asm");

crypto/Kconfig

@@ -1436,34 +1436,6 @@ config CRYPTO_SALSA20
 	  The Salsa20 stream cipher algorithm is designed by Daniel J.
 	  Bernstein <djb@cr.yp.to>. See <http://cr.yp.to/snuffle.html>
 
-config CRYPTO_SALSA20_586
-	tristate "Salsa20 stream cipher algorithm (i586)"
-	depends on (X86 || UML_X86) && !64BIT
-	select CRYPTO_BLKCIPHER
-	select CRYPTO_SALSA20
-	help
-	  Salsa20 stream cipher algorithm.
-
-	  Salsa20 is a stream cipher submitted to eSTREAM, the ECRYPT
-	  Stream Cipher Project. See <http://www.ecrypt.eu.org/stream/>
-
-	  The Salsa20 stream cipher algorithm is designed by Daniel J.
-	  Bernstein <djb@cr.yp.to>. See <http://cr.yp.to/snuffle.html>
-
-config CRYPTO_SALSA20_X86_64
-	tristate "Salsa20 stream cipher algorithm (x86_64)"
-	depends on (X86 || UML_X86) && 64BIT
-	select CRYPTO_BLKCIPHER
-	select CRYPTO_SALSA20
-	help
-	  Salsa20 stream cipher algorithm.
-
-	  Salsa20 is a stream cipher submitted to eSTREAM, the ECRYPT
-	  Stream Cipher Project. See <http://www.ecrypt.eu.org/stream/>
-
-	  The Salsa20 stream cipher algorithm is designed by Daniel J.
-	  Bernstein <djb@cr.yp.to>. See <http://cr.yp.to/snuffle.html>
-
 config CRYPTO_CHACHA20
 	tristate "ChaCha20 cipher algorithm"
 	select CRYPTO_BLKCIPHER