c/mirror_zfs
1
0
Fork 0
mirror of https://git.proxmox.com/git/mirror_zfs.git synced 2026-05-26 12:12:13 +03:00
Code Issues Packages Projects Releases Wiki Activity

Unify Assembler files between Linux and Windows

Browse Source 
Add new macro ASMABI used by Windows to change
calling API to "sysv_abi".

Reviewed-by: Attila Fülöp <attila@fueloep.org>
Reviewed-by: Richard Yao <richard.yao@alumni.stonybrook.edu>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Jorgen Lundman <lundman@lundman.net>
Closes #14228
This commit is contained in:
Jorgen Lundman
2023-01-18 04:09:19 +09:00
committed by GitHub
parent 19d3961589
commit 68c0771cc9
32 changed files with 815 additions and 488 deletions
Download Patch File Download Diff File Expand all files Collapse all files
module/icp/asm-x86_64/modes/aesni-gcm-x86_64.S
+27 -39
View File
@@ -50,14 +50,16 @@
#define _ASM
#include <sys/asm_linkage.h>
/* Windows userland links with OpenSSL */
#if !defined (_WIN32) || defined (_KERNEL)
.extern gcm_avx_can_use_movbe
.text
#ifdef HAVE_MOVBE
.type _aesni_ctr32_ghash_6x,@function
.align 32
_aesni_ctr32_ghash_6x:
.align 32
FUNCTION(_aesni_ctr32_ghash_6x)
.cfi_startproc
ENDBR
vmovdqu 32(%r11),%xmm2
@@ -369,12 +371,11 @@ _aesni_ctr32_ghash_6x:
RET
.cfi_endproc
.size _aesni_ctr32_ghash_6x,.-_aesni_ctr32_ghash_6x
SET_SIZE(_aesni_ctr32_ghash_6x)
#endif /* ifdef HAVE_MOVBE */
.type _aesni_ctr32_ghash_no_movbe_6x,@function
.align 32
_aesni_ctr32_ghash_no_movbe_6x:
.align 32
FUNCTION(_aesni_ctr32_ghash_no_movbe_6x)
.cfi_startproc
ENDBR
vmovdqu 32(%r11),%xmm2
@@ -698,12 +699,9 @@ _aesni_ctr32_ghash_no_movbe_6x:
RET
.cfi_endproc
.size _aesni_ctr32_ghash_no_movbe_6x,.-_aesni_ctr32_ghash_no_movbe_6x
SET_SIZE(_aesni_ctr32_ghash_no_movbe_6x)
.globl aesni_gcm_decrypt
.type aesni_gcm_decrypt,@function
.align 32
aesni_gcm_decrypt:
ENTRY_ALIGN(aesni_gcm_decrypt, 32)
.cfi_startproc
ENDBR
xorq %r10,%r10
@@ -818,10 +816,10 @@ aesni_gcm_decrypt:
movq %r10,%rax
RET
.cfi_endproc
.size aesni_gcm_decrypt,.-aesni_gcm_decrypt
.type _aesni_ctr32_6x,@function
.align 32
_aesni_ctr32_6x:
SET_SIZE(aesni_gcm_decrypt)
.align 32
FUNCTION(_aesni_ctr32_6x)
.cfi_startproc
ENDBR
vmovdqu 0-128(%rcx),%xmm4
@@ -911,12 +909,9 @@ _aesni_ctr32_6x:
vpxor %xmm4,%xmm14,%xmm14
jmp .Loop_ctr32
.cfi_endproc
.size _aesni_ctr32_6x,.-_aesni_ctr32_6x
SET_SIZE(_aesni_ctr32_6x)
.globl aesni_gcm_encrypt
.type aesni_gcm_encrypt,@function
.align 32
aesni_gcm_encrypt:
ENTRY_ALIGN(aesni_gcm_encrypt, 32)
.cfi_startproc
ENDBR
xorq %r10,%r10
@@ -1196,7 +1191,9 @@ aesni_gcm_encrypt:
movq %r10,%rax
RET
.cfi_endproc
.size aesni_gcm_encrypt,.-aesni_gcm_encrypt
SET_SIZE(aesni_gcm_encrypt)
#endif /* !_WIN32 || _KERNEL */
/* Some utility routines */
@@ -1204,13 +1201,10 @@ aesni_gcm_encrypt:
* 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:
ENTRY_ALIGN(clear_fpu_regs_avx, 32)
vzeroall
RET
.size clear_fpu_regs_avx,.-clear_fpu_regs_avx
SET_SIZE(clear_fpu_regs_avx)
/*
* void gcm_xor_avx(const uint8_t *src, uint8_t *dst);
@@ -1219,25 +1213,19 @@ clear_fpu_regs_avx:
* 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:
ENTRY_ALIGN(gcm_xor_avx, 32)
movdqu (%rdi), %xmm0
movdqu (%rsi), %xmm1
pxor %xmm1, %xmm0
movdqu %xmm0, (%rsi)
RET
.size gcm_xor_avx,.-gcm_xor_avx
SET_SIZE(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:
ENTRY_ALIGN(atomic_toggle_boolean_nv, 32)
xorl %eax, %eax
lock
xorl $1, (%rdi)
@@ -1245,9 +1233,10 @@ atomic_toggle_boolean_nv:
movl $1, %eax
1:
RET
.size atomic_toggle_boolean_nv,.-atomic_toggle_boolean_nv
SET_SIZE(atomic_toggle_boolean_nv)
SECTION_STATIC
.pushsection .rodata
.align 64
.Lbswap_mask:
.byte 15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0
@@ -1261,7 +1250,6 @@ atomic_toggle_boolean_nv:
.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
.popsection
/* Mark the stack non-executable. */
#if defined(__linux__) && defined(__ELF__)
module/icp/asm-x86_64/modes/ghash-x86_64.S
+20 -24
View File
@@ -102,12 +102,13 @@
.text
.globl gcm_gmult_clmul
.type gcm_gmult_clmul,@function
.align 16
gcm_gmult_clmul:
/* Windows userland links with OpenSSL */
#if !defined (_WIN32) || defined (_KERNEL)
ENTRY_ALIGN(gcm_gmult_clmul, 16)
.cfi_startproc
ENDBR
.L_gmult_clmul:
movdqu (%rdi),%xmm0
movdqa .Lbswap_mask(%rip),%xmm5
@@ -155,12 +156,10 @@ gcm_gmult_clmul:
movdqu %xmm0,(%rdi)
RET
.cfi_endproc
.size gcm_gmult_clmul,.-gcm_gmult_clmul
SET_SIZE(gcm_gmult_clmul)
#endif /* !_WIN32 || _KERNEL */
.globl gcm_init_htab_avx
.type gcm_init_htab_avx,@function
.align 32
gcm_init_htab_avx:
ENTRY_ALIGN(gcm_init_htab_avx, 32)
.cfi_startproc
ENDBR
vzeroupper
@@ -269,21 +268,17 @@ gcm_init_htab_avx:
vzeroupper
RET
.cfi_endproc
.size gcm_init_htab_avx,.-gcm_init_htab_avx
SET_SIZE(gcm_init_htab_avx)
.globl gcm_gmult_avx
.type gcm_gmult_avx,@function
.align 32
gcm_gmult_avx:
#if !defined (_WIN32) || defined (_KERNEL)
ENTRY_ALIGN(gcm_gmult_avx, 32)
.cfi_startproc
ENDBR
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:
SET_SIZE(gcm_gmult_avx)
ENTRY_ALIGN(gcm_ghash_avx, 32)
.cfi_startproc
ENDBR
vzeroupper
@@ -658,9 +653,11 @@ gcm_ghash_avx:
vzeroupper
RET
.cfi_endproc
.size gcm_ghash_avx,.-gcm_ghash_avx
SET_SIZE(gcm_ghash_avx)
.pushsection .rodata
#endif /* !_WIN32 || _KERNEL */
SECTION_STATIC
.align 64
.Lbswap_mask:
.byte 15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0
@@ -671,13 +668,13 @@ gcm_ghash_avx:
.L7_mask_poly:
.long 7,0,450,0
.align 64
.type .Lrem_4bit,@object
SET_OBJ(.Lrem_4bit)
.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
SET_OBJ(.Lrem_8bit)
.Lrem_8bit:
.value 0x0000,0x01C2,0x0384,0x0246,0x0708,0x06CA,0x048C,0x054E
.value 0x0E10,0x0FD2,0x0D94,0x0C56,0x0918,0x08DA,0x0A9C,0x0B5E
@@ -714,7 +711,6 @@ gcm_ghash_avx:
.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
.popsection
/* Mark the stack non-executable. */
#if defined(__linux__) && defined(__ELF__)