e_padlock-x86[_64].pl: SHA fixes, comply with specification and fix bug.

diff --git a/engines/asm/e_padlock-x86.pl b/engines/asm/e_padlock-x86.pl
index 61e91d889f82b73b82d5dc0e19d7a7a9238164b9..e211706ae1b8511e3db929c8ff59eaf6f1513757 100644 (file)
--- a/engines/asm/e_padlock-x86.pl
+++ b/engines/asm/e_padlock-x86.pl
@@ -352,19 +352,34 @@ my ($mode,$opcode) = @_;
        &push   ("edi");
        &push   ("esi");
        &xor    ("eax","eax");
+       &mov    ("edi",&wparam(0));
+       &mov    ("esi",&wparam(1));
+       &mov    ("ecx",&wparam(2));
     if ($::win32 or $::coff) {
        &push   (&::islabel("_win32_segv_handler"));
        &data_byte(0x64,0xff,0x30);             # push  %fs:(%eax)
        &data_byte(0x64,0x89,0x20);             # mov   %esp,%fs:(%eax)
     }
-       &mov    ("edi",&wparam(0));
-       &mov    ("esi",&wparam(1));
-       &mov    ("ecx",&wparam(2));
+       &mov    ("edx","esp");                  # put aside %esp
+       &add    ("esp",-128);                   # 32 is enough but spec says 128
+       &movups ("xmm0",&QWP(0,"edi"));         # copy-in context
+       &and    ("esp",-16);
+       &mov    ("eax",&DWP(16,"edi"));
+       &movaps (&QWP(0,"esp"),"xmm0");
+       &mov    ("edi","esp");
+       &mov    (&DWP(16,"esp"),"eax");
+       &xor    ("eax","eax");
        &data_byte(0xf3,0x0f,0xa6,0xc8);        # rep xsha1
+       &movaps ("xmm0",&QWP(0,"esp"));
+       &mov    ("eax",&DWP(16,"esp"));
+       &mov    ("esp","edx");                  # restore %esp
     if ($::win32 or $::coff) {
        &data_byte(0x64,0x8f,0x05,0,0,0,0);     # pop   %fs:0
        &lea    ("esp",&DWP(4,"esp"));
     }
+       &mov    ("edi",&wparam(0));
+       &movups (&QWP(0,"edi"),"xmm0");         # copy-out context
+       &mov    (&DWP(16,"edi"),"eax");
        &pop    ("esi");
        &pop    ("edi");
        &ret    ();
@@ -373,12 +388,26 @@ my ($mode,$opcode) = @_;
 &function_begin_B("padlock_sha1_blocks");
        &push   ("edi");
        &push   ("esi");
-       &mov    ("eax",-1);
        &mov    ("edi",&wparam(0));
        &mov    ("esi",&wparam(1));
+       &mov    ("edx","esp");                  # put aside %esp
        &mov    ("ecx",&wparam(2));
+       &add    ("esp",-128);
+       &movups ("xmm0",&QWP(0,"edi"));         # copy-in context
+       &and    ("esp",-16);
+       &mov    ("eax",&DWP(16,"edi"));
+       &movaps (&QWP(0,"esp"),"xmm0");
+       &mov    ("edi","esp");
+       &mov    (&DWP(16,"esp"),"eax");
+       &mov    ("eax",-1);
        &data_byte(0xf3,0x0f,0xa6,0xc8);        # rep xsha1
-       &pop    ("esi");
+       &movaps ("xmm0",&QWP(0,"esp"));
+       &mov    ("eax",&DWP(16,"esp"));
+       &mov    ("esp","edx");                  # restore %esp
+       &mov    ("edi",&wparam(0));
+       &movups (&QWP(0,"edi"),"xmm0");         # copy-out context
+       &mov    (&DWP(16,"edi"),"eax");
+       &pop    ("esi");
        &pop    ("edi");
        &ret    ();
 &function_end_B("padlock_sha1_blocks");
@@ -387,19 +416,34 @@ my ($mode,$opcode) = @_;
        &push   ("edi");
        &push   ("esi");
        &xor    ("eax","eax");
+       &mov    ("edi",&wparam(0));
+       &mov    ("esi",&wparam(1));
+       &mov    ("ecx",&wparam(2));
     if ($::win32 or $::coff) {
        &push   (&::islabel("_win32_segv_handler"));
        &data_byte(0x64,0xff,0x30);             # push  %fs:(%eax)
        &data_byte(0x64,0x89,0x20);             # mov   %esp,%fs:(%eax)
     }
-       &mov    ("edi",&wparam(0));
-       &mov    ("esi",&wparam(1));
-       &mov    ("ecx",&wparam(2));
+       &mov    ("edx","esp");                  # put aside %esp
+       &add    ("esp",-128);
+       &movups ("xmm0",&QWP(0,"edi"));         # copy-in context
+       &and    ("esp",-16);
+       &movups ("xmm1",&QWP(16,"edi"));
+       &movaps (&QWP(0,"esp"),"xmm0");
+       &mov    ("edi","esp");
+       &movaps (&QWP(16,"esp"),"xmm1");
+       &xor    ("eax","eax");
        &data_byte(0xf3,0x0f,0xa6,0xd0);        # rep xsha256
+       &movaps ("xmm0",&QWP(0,"esp"));
+       &movaps ("xmm1",&QWP(16,"esp"));
+       &mov    ("esp","edx");                  # restore %esp
     if ($::win32 or $::coff) {
        &data_byte(0x64,0x8f,0x05,0,0,0,0);     # pop   %fs:0
        &lea    ("esp",&DWP(4,"esp"));
     }
+       &mov    ("edi",&wparam(0));
+       &movups (&QWP(0,"edi"),"xmm0");         # copy-out context
+       &movups (&QWP(16,"edi"),"xmm1");
        &pop    ("esi");
        &pop    ("edi");
        &ret    ();
@@ -408,11 +452,25 @@ my ($mode,$opcode) = @_;
 &function_begin_B("padlock_sha256_blocks");
        &push   ("edi");
        &push   ("esi");
-       &mov    ("eax",-1);
        &mov    ("edi",&wparam(0));
        &mov    ("esi",&wparam(1));
        &mov    ("ecx",&wparam(2));
+       &mov    ("edx","esp");                  # put aside %esp
+       &add    ("esp",-128);
+       &movups ("xmm0",&QWP(0,"edi"));         # copy-in context
+       &and    ("esp",-16);
+       &movups ("xmm1",&QWP(16,"edi"));
+       &movaps (&QWP(0,"esp"),"xmm0");
+       &mov    ("edi","esp");
+       &movaps (&QWP(16,"esp"),"xmm1");
+       &mov    ("eax",-1);
        &data_byte(0xf3,0x0f,0xa6,0xd0);        # rep xsha256
+       &movaps ("xmm0",&QWP(0,"esp"));
+       &movaps ("xmm1",&QWP(16,"esp"));
+       &mov    ("esp","edx");                  # restore %esp
+       &mov    ("edi",&wparam(0));
+       &movups (&QWP(0,"edi"),"xmm0");         # copy-out context
+       &movups (&QWP(16,"edi"),"xmm1");
        &pop    ("esi");
        &pop    ("edi");
        &ret    ();
@@ -424,7 +482,29 @@ my ($mode,$opcode) = @_;
        &mov    ("edi",&wparam(0));
        &mov    ("esi",&wparam(1));
        &mov    ("ecx",&wparam(2));
+       &mov    ("edx","esp");                  # put aside %esp
+       &add    ("esp",-128);
+       &movups ("xmm0",&QWP(0,"edi"));         # copy-in context
+       &and    ("esp",-16);
+       &movups ("xmm1",&QWP(16,"edi"));
+       &movups ("xmm2",&QWP(32,"edi"));
+       &movups ("xmm3",&QWP(48,"edi"));
+       &movaps (&QWP(0,"esp"),"xmm0");
+       &mov    ("edi","esp");
+       &movaps (&QWP(16,"esp"),"xmm1");
+       &movaps (&QWP(32,"esp"),"xmm2");
+       &movaps (&QWP(48,"esp"),"xmm3");
        &data_byte(0xf3,0x0f,0xa6,0xe0);        # rep xsha512
+       &movaps ("xmm0",&QWP(0,"esp"));
+       &movaps ("xmm1",&QWP(16,"esp"));
+       &movaps ("xmm2",&QWP(32,"esp"));
+       &movaps ("xmm3",&QWP(48,"esp"));
+       &mov    ("esp","edx");                  # restore %esp
+       &mov    ("edi",&wparam(0));
+       &movups (&QWP(0,"edi"),"xmm0");         # copy-out context
+       &movups (&QWP(16,"edi"),"xmm1");
+       &movups (&QWP(32,"edi"),"xmm2");
+       &movups (&QWP(48,"edi"),"xmm3");
        &pop    ("esi");
        &pop    ("edi");
        &ret    ();

diff --git a/engines/asm/e_padlock-x86_64.pl b/engines/asm/e_padlock-x86_64.pl
index 13c371be67ed5aa7cac3054efcc51ed4141e31ce..ad61974988155beed15c27a1c27851dc98311d4e 100644 (file)
--- a/engines/asm/e_padlock-x86_64.pl
+++ b/engines/asm/e_padlock-x86_64.pl
@@ -146,9 +146,20 @@ padlock_xstore:
 .type  padlock_sha1_oneshot,\@function,3
 .align 16
 padlock_sha1_oneshot:
-       xor     %rax,%rax
        mov     %rdx,%rcx
+       mov     %rdi,%rdx               # put aside %rdi
+       movups  (%rdi),%xmm0            # copy-in context
+       sub     \$128+8,%rsp
+       mov     16(%rdi),%eax
+       movaps  %xmm0,(%rsp)
+       mov     %rsp,%rdi
+       mov     %eax,16(%rsp)
+       xor     %rax,%rax
        .byte   0xf3,0x0f,0xa6,0xc8     # rep xsha1
+       movaps  (%rsp),%xmm0
+       mov     16(%rsp),%eax
+       movups  %xmm0,(%rdx)            # copy-out context
+       mov     %eax,16(%rdx)
        ret
 .size  padlock_sha1_oneshot,.-padlock_sha1_oneshot
 
@@ -156,9 +167,20 @@ padlock_sha1_oneshot:
 .type  padlock_sha1_blocks,\@function,3
 .align 16
 padlock_sha1_blocks:
-       mov     \$-1,%rax
        mov     %rdx,%rcx
+       mov     %rdi,%rdx               # put aside %rdi
+       movups  (%rdi),%xmm0            # copy-in context
+       sub     \$128+8,%rsp
+       mov     16(%rdi),%eax
+       movaps  %xmm0,(%rsp)
+       mov     %rsp,%rdi
+       mov     %eax,16(%rsp)
+       mov     \$-1,%rax
        .byte   0xf3,0x0f,0xa6,0xc8     # rep xsha1
+       movaps  (%rsp),%xmm0
+       mov     16(%rsp),%eax
+       movups  %xmm0,(%rdx)            # copy-out context
+       mov     %eax,16(%rdx)
        ret
 .size  padlock_sha1_blocks,.-padlock_sha1_blocks
 
@@ -166,9 +188,20 @@ padlock_sha1_blocks:
 .type  padlock_sha256_oneshot,\@function,3
 .align 16
 padlock_sha256_oneshot:
-       xor     %rax,%rax
        mov     %rdx,%rcx
+       mov     %rdi,%rdx               # put aside %rdi
+       movups  (%rdi),%xmm0            # copy-in context
+       sub     \$128+8,%rsp
+       movups  16(%rdi),%xmm1
+       movaps  %xmm0,(%rsp)
+       mov     %rsp,%rdi
+       movaps  %xmm1,16(%rsp)
+       xor     %rax,%rax
        .byte   0xf3,0x0f,0xa6,0xd0     # rep xsha256
+       movaps  (%rsp),%xmm0
+       movaps  16(%rsp),%xmm1
+       movups  %xmm0,(%rdx)            # copy-out context
+       movups  %xmm1,16(%rdx)
        ret
 .size  padlock_sha256_oneshot,.-padlock_sha256_oneshot
 
@@ -176,9 +209,20 @@ padlock_sha256_oneshot:
 .type  padlock_sha256_blocks,\@function,3
 .align 16
 padlock_sha256_blocks:
-       mov     \$-1,%rax
        mov     %rdx,%rcx
+       mov     %rdi,%rdx               # put aside %rdi
+       movups  (%rdi),%xmm0            # copy-in context
+       sub     \$128+8,%rsp
+       movups  16(%rdi),%xmm1
+       movaps  %xmm0,(%rsp)
+       mov     %rsp,%rdi
+       movaps  %xmm1,16(%rsp)
+       mov     \$-1,%rax
        .byte   0xf3,0x0f,0xa6,0xd0     # rep xsha256
+       movaps  (%rsp),%xmm0
+       movaps  16(%rsp),%xmm1
+       movups  %xmm0,(%rdx)            # copy-out context
+       movups  %xmm1,16(%rdx)
        ret
 .size  padlock_sha256_blocks,.-padlock_sha256_blocks
 
@@ -187,7 +231,26 @@ padlock_sha256_blocks:
 .align 16
 padlock_sha512_blocks:
        mov     %rdx,%rcx
+       mov     %rdi,%rdx               # put aside %rdi
+       movups  (%rdi),%xmm0            # copy-in context
+       sub     \$128+8,%rsp
+       movups  16(%rdi),%xmm1
+       movups  32(%rdi),%xmm2
+       movups  48(%rdi),%xmm3
+       movaps  %xmm0,(%rsp)
+       mov     %rsp,%rdi
+       movaps  %xmm1,16(%rsp)
+       movaps  %xmm2,32(%rsp)
+       movaps  %xmm3,48(%rsp)
        .byte   0xf3,0x0f,0xa6,0xe0     # rep xha512
+       movaps  (%rsp),%xmm0
+       movaps  16(%rsp),%xmm1
+       movaps  32(%rsp),%xmm2
+       movaps  48(%rsp),%xmm3
+       movups  %xmm0,(%rdx)            # copy-out context
+       movups  %xmm1,16(%rdx)
+       movups  %xmm2,32(%rdx)
+       movups  %xmm3,48(%rdx)
        ret
 .size  padlock_sha512_blocks,.-padlock_sha512_blocks
 ___