Socket-Class
view release on metacpan or search on metacpan
xs/sc_ssl/openssl/source/crypto/engine/eng_padlock.c view on Meta::CPAN
asm volatile (
"pushl %%ebx\n"
"cpuid\n"
"movl %%ebx,(%%edi)\n"
"movl %%edx,4(%%edi)\n"
"movl %%ecx,8(%%edi)\n"
"popl %%ebx"
: "+a"(eax) : "D"(vendor_string) : "ecx", "edx");
if (strcmp(vendor_string, "CentaurHauls") != 0)
return 0;
/* Check for Centaur Extended Feature Flags presence */
eax = 0xC0000000;
asm volatile ("pushl %%ebx; cpuid; popl %%ebx"
: "+a"(eax) : : "ecx", "edx");
if (eax < 0xC0000001)
return 0;
/* Read the Centaur Extended Feature Flags */
eax = 0xC0000001;
asm volatile ("pushl %%ebx; cpuid; popl %%ebx"
: "+a"(eax), "=d"(edx) : : "ecx");
/* Fill up some flags */
padlock_use_ace = ((edx & (0x3<<6)) == (0x3<<6));
padlock_use_rng = ((edx & (0x3<<2)) == (0x3<<2));
return padlock_use_ace + padlock_use_rng;
}
#ifndef OPENSSL_NO_AES
/* Our own htonl()/ntohl() */
static inline void
padlock_bswapl(AES_KEY *ks)
{
size_t i = sizeof(ks->rd_key)/sizeof(ks->rd_key[0]);
unsigned int *key = ks->rd_key;
while (i--) {
asm volatile ("bswapl %0" : "+r"(*key));
key++;
}
}
#endif
/* Force key reload from memory to the CPU microcode.
Loading EFLAGS from the stack clears EFLAGS[30]
which does the trick. */
static inline void
padlock_reload_key(void)
{
asm volatile ("pushfl; popfl");
}
#ifndef OPENSSL_NO_AES
/*
* This is heuristic key context tracing. At first one
* believes that one should use atomic swap instructions,
* but it's not actually necessary. Point is that if
* padlock_saved_context was changed by another thread
* after we've read it and before we compare it with cdata,
* our key *shall* be reloaded upon thread context switch
* and we are therefore set in either case...
*/
static inline void
padlock_verify_context(struct padlock_cipher_data *cdata)
{
asm volatile (
"pushfl\n"
" btl $30,(%%esp)\n"
" jnc 1f\n"
" cmpl %2,%1\n"
" je 1f\n"
" popfl\n"
" subl $4,%%esp\n"
"1: addl $4,%%esp\n"
" movl %2,%0"
:"+m"(padlock_saved_context)
: "r"(padlock_saved_context), "r"(cdata) : "cc");
}
/* Template for padlock_xcrypt_* modes */
/* BIG FAT WARNING:
* The offsets used with 'leal' instructions
* describe items of the 'padlock_cipher_data'
* structure.
*/
#define PADLOCK_XCRYPT_ASM(name,rep_xcrypt) \
static inline void *name(size_t cnt, \
struct padlock_cipher_data *cdata, \
void *out, const void *inp) \
{ void *iv; \
asm volatile ( "pushl %%ebx\n" \
" leal 16(%0),%%edx\n" \
" leal 32(%0),%%ebx\n" \
rep_xcrypt "\n" \
" popl %%ebx" \
: "=a"(iv), "=c"(cnt), "=D"(out), "=S"(inp) \
: "0"(cdata), "1"(cnt), "2"(out), "3"(inp) \
: "edx", "cc", "memory"); \
return iv; \
}
/* Generate all functions with appropriate opcodes */
PADLOCK_XCRYPT_ASM(padlock_xcrypt_ecb, ".byte 0xf3,0x0f,0xa7,0xc8") /* rep xcryptecb */
PADLOCK_XCRYPT_ASM(padlock_xcrypt_cbc, ".byte 0xf3,0x0f,0xa7,0xd0") /* rep xcryptcbc */
PADLOCK_XCRYPT_ASM(padlock_xcrypt_cfb, ".byte 0xf3,0x0f,0xa7,0xe0") /* rep xcryptcfb */
PADLOCK_XCRYPT_ASM(padlock_xcrypt_ofb, ".byte 0xf3,0x0f,0xa7,0xe8") /* rep xcryptofb */
#endif
/* The RNG call itself */
static inline unsigned int
padlock_xstore(void *addr, unsigned int edx_in)
{
unsigned int eax_out;
asm volatile (".byte 0x0f,0xa7,0xc0" /* xstore */
: "=a"(eax_out),"=m"(*(unsigned *)addr)
: "D"(addr), "d" (edx_in)
);
return eax_out;
}
/* Why not inline 'rep movsd'? I failed to find information on what
* value in Direction Flag one can expect and consequently have to
* apply "better-safe-than-sorry" approach and assume "undefined."
* I could explicitly clear it and restore the original value upon
* return from padlock_aes_cipher, but it's presumably too much
* trouble for too little gain...
*
* In case you wonder 'rep xcrypt*' instructions above are *not*
* affected by the Direction Flag and pointers advance toward
* larger addresses unconditionally.
*/
static inline unsigned char *
padlock_memcpy(void *dst,const void *src,size_t n)
{
long *d=dst;
const long *s=src;
n /= sizeof(*d);
do { *d++ = *s++; } while (--n);
return dst;
}
#elif defined(_MSC_VER)
/*
* Unlike GCC these are real functions. In order to minimize impact
* on performance we adhere to __fastcall calling convention in
* order to get two first arguments passed through %ecx and %edx.
* Which kind of suits very well, as instructions in question use
* both %ecx and %edx as input:-)
*/
#define REP_XCRYPT(code) \
_asm _emit 0xf3 \
_asm _emit 0x0f _asm _emit 0xa7 \
_asm _emit code
/* BIG FAT WARNING:
* The offsets used with 'lea' instructions
* describe items of the 'padlock_cipher_data'
* structure.
*/
#define PADLOCK_XCRYPT_ASM(name,code) \
static void * __fastcall \
name (size_t cnt, void *cdata, \
void *outp, const void *inp) \
{ _asm mov eax,edx \
_asm lea edx,[eax+16] \
_asm lea ebx,[eax+32] \
_asm mov edi,outp \
_asm mov esi,inp \
REP_XCRYPT(code) \
}
PADLOCK_XCRYPT_ASM(padlock_xcrypt_ecb,0xc8)
PADLOCK_XCRYPT_ASM(padlock_xcrypt_cbc,0xd0)
PADLOCK_XCRYPT_ASM(padlock_xcrypt_cfb,0xe0)
PADLOCK_XCRYPT_ASM(padlock_xcrypt_ofb,0xe8)
static int __fastcall
padlock_xstore(void *outp,unsigned int code)
{ _asm mov edi,ecx
_asm _emit 0x0f _asm _emit 0xa7 _asm _emit 0xc0
}
static void __fastcall
padlock_reload_key(void)
{ _asm pushfd _asm popfd }
static void __fastcall
padlock_verify_context(void *cdata)
{ _asm {
pushfd
bt DWORD PTR[esp],30
jnc skip
cmp ecx,padlock_saved_context
je skip
popfd
sub esp,4
skip: add esp,4
mov padlock_saved_context,ecx
}
}
static int
padlock_available(void)
{ _asm {
pushfd
pop eax
mov ecx,eax
xor eax,1<<21
push eax
popfd
pushfd
pop eax
xor eax,ecx
bt eax,21
jnc noluck
mov eax,0
cpuid
xor eax,eax
cmp ebx,'tneC'
jne noluck
cmp edx,'Hrua'
jne noluck
cmp ecx,'slua'
jne noluck
mov eax,0xC0000000
cpuid
mov edx,eax
xor eax,eax
cmp edx,0xC0000001
jb noluck
mov eax,0xC0000001
cpuid
xor eax,eax
bt edx,6
jnc skip_a
bt edx,7
jnc skip_a
mov padlock_use_ace,1
inc eax
skip_a: bt edx,2
jnc skip_r
bt edx,3
jnc skip_r
mov padlock_use_rng,1
inc eax
skip_r:
noluck:
}
xs/sc_ssl/openssl/source/crypto/engine/eng_padlock.c view on Meta::CPAN
if (!cipher) {
*nids = padlock_cipher_nids;
return padlock_cipher_nids_num;
}
/* ... or the requested "cipher" otherwise */
switch (nid) {
case NID_aes_128_ecb:
*cipher = &padlock_aes_128_ecb;
break;
case NID_aes_128_cbc:
*cipher = &padlock_aes_128_cbc;
break;
case NID_aes_128_cfb:
*cipher = &padlock_aes_128_cfb;
break;
case NID_aes_128_ofb:
*cipher = &padlock_aes_128_ofb;
break;
case NID_aes_192_ecb:
*cipher = &padlock_aes_192_ecb;
break;
case NID_aes_192_cbc:
*cipher = &padlock_aes_192_cbc;
break;
case NID_aes_192_cfb:
*cipher = &padlock_aes_192_cfb;
break;
case NID_aes_192_ofb:
*cipher = &padlock_aes_192_ofb;
break;
case NID_aes_256_ecb:
*cipher = &padlock_aes_256_ecb;
break;
case NID_aes_256_cbc:
*cipher = &padlock_aes_256_cbc;
break;
case NID_aes_256_cfb:
*cipher = &padlock_aes_256_cfb;
break;
case NID_aes_256_ofb:
*cipher = &padlock_aes_256_ofb;
break;
default:
/* Sorry, we don't support this NID */
*cipher = NULL;
return 0;
}
return 1;
}
/* Prepare the encryption key for PadLock usage */
static int
padlock_aes_init_key (EVP_CIPHER_CTX *ctx, const unsigned char *key,
const unsigned char *iv, int enc)
{
struct padlock_cipher_data *cdata;
int key_len = EVP_CIPHER_CTX_key_length(ctx) * 8;
if (key==NULL) return 0; /* ERROR */
cdata = ALIGNED_CIPHER_DATA(ctx);
memset(cdata, 0, sizeof(struct padlock_cipher_data));
/* Prepare Control word. */
if (EVP_CIPHER_CTX_mode(ctx) == EVP_CIPH_OFB_MODE)
cdata->cword.b.encdec = 0;
else
cdata->cword.b.encdec = (ctx->encrypt == 0);
cdata->cword.b.rounds = 10 + (key_len - 128) / 32;
cdata->cword.b.ksize = (key_len - 128) / 64;
switch(key_len) {
case 128:
/* PadLock can generate an extended key for
AES128 in hardware */
memcpy(cdata->ks.rd_key, key, AES_KEY_SIZE_128);
cdata->cword.b.keygen = 0;
break;
case 192:
case 256:
/* Generate an extended AES key in software.
Needed for AES192/AES256 */
/* Well, the above applies to Stepping 8 CPUs
and is listed as hardware errata. They most
likely will fix it at some point and then
a check for stepping would be due here. */
if (EVP_CIPHER_CTX_mode(ctx) == EVP_CIPH_CFB_MODE ||
EVP_CIPHER_CTX_mode(ctx) == EVP_CIPH_OFB_MODE ||
enc)
AES_set_encrypt_key(key, key_len, &cdata->ks);
else
AES_set_decrypt_key(key, key_len, &cdata->ks);
#ifndef AES_ASM
/* OpenSSL C functions use byte-swapped extended key. */
padlock_bswapl(&cdata->ks);
#endif
cdata->cword.b.keygen = 1;
break;
default:
/* ERROR */
return 0;
}
/*
* This is done to cover for cases when user reuses the
* context for new key. The catch is that if we don't do
* this, padlock_eas_cipher might proceed with old key...
*/
padlock_reload_key ();
return 1;
}
/*
* Simplified version of padlock_aes_cipher() used when
* 1) both input and output buffers are at aligned addresses.
* or when
* 2) running on a newer CPU that doesn't require aligned buffers.
*/
static int
padlock_aes_cipher_omnivorous(EVP_CIPHER_CTX *ctx, unsigned char *out_arg,
const unsigned char *in_arg, size_t nbytes)
{
struct padlock_cipher_data *cdata;
void *iv;
cdata = ALIGNED_CIPHER_DATA(ctx);
padlock_verify_context(cdata);
switch (EVP_CIPHER_CTX_mode(ctx)) {
case EVP_CIPH_ECB_MODE:
padlock_xcrypt_ecb(nbytes/AES_BLOCK_SIZE, cdata, out_arg, in_arg);
break;
case EVP_CIPH_CBC_MODE:
memcpy(cdata->iv, ctx->iv, AES_BLOCK_SIZE);
iv = padlock_xcrypt_cbc(nbytes/AES_BLOCK_SIZE, cdata, out_arg, in_arg);
memcpy(ctx->iv, iv, AES_BLOCK_SIZE);
break;
case EVP_CIPH_CFB_MODE:
memcpy(cdata->iv, ctx->iv, AES_BLOCK_SIZE);
iv = padlock_xcrypt_cfb(nbytes/AES_BLOCK_SIZE, cdata, out_arg, in_arg);
memcpy(ctx->iv, iv, AES_BLOCK_SIZE);
break;
case EVP_CIPH_OFB_MODE:
memcpy(cdata->iv, ctx->iv, AES_BLOCK_SIZE);
padlock_xcrypt_ofb(nbytes/AES_BLOCK_SIZE, cdata, out_arg, in_arg);
memcpy(ctx->iv, cdata->iv, AES_BLOCK_SIZE);
break;
default:
return 0;
}
memset(cdata->iv, 0, AES_BLOCK_SIZE);
return 1;
}
#ifndef PADLOCK_CHUNK
# define PADLOCK_CHUNK 512 /* Must be a power of 2 larger than 16 */
#endif
#if PADLOCK_CHUNK<16 || PADLOCK_CHUNK&(PADLOCK_CHUNK-1)
# error "insane PADLOCK_CHUNK..."
#endif
/* Re-align the arguments to 16-Bytes boundaries and run the
encryption function itself. This function is not AES-specific. */
static int
padlock_aes_cipher(EVP_CIPHER_CTX *ctx, unsigned char *out_arg,
const unsigned char *in_arg, size_t nbytes)
{
struct padlock_cipher_data *cdata;
const void *inp;
unsigned char *out;
void *iv;
int inp_misaligned, out_misaligned, realign_in_loop;
size_t chunk, allocated=0;
/* ctx->num is maintained in byte-oriented modes,
such as CFB and OFB... */
if ((chunk = ctx->num)) { /* borrow chunk variable */
unsigned char *ivp=ctx->iv;
switch (EVP_CIPHER_CTX_mode(ctx)) {
case EVP_CIPH_CFB_MODE:
if (chunk >= AES_BLOCK_SIZE)
return 0; /* bogus value */
if (ctx->encrypt)
while (chunk<AES_BLOCK_SIZE && nbytes!=0) {
ivp[chunk] = *(out_arg++) = *(in_arg++) ^ ivp[chunk];
chunk++, nbytes--;
}
else while (chunk<AES_BLOCK_SIZE && nbytes!=0) {
unsigned char c = *(in_arg++);
*(out_arg++) = c ^ ivp[chunk];
ivp[chunk++] = c, nbytes--;
}
ctx->num = chunk%AES_BLOCK_SIZE;
break;
case EVP_CIPH_OFB_MODE:
if (chunk >= AES_BLOCK_SIZE)
return 0; /* bogus value */
while (chunk<AES_BLOCK_SIZE && nbytes!=0) {
*(out_arg++) = *(in_arg++) ^ ivp[chunk];
chunk++, nbytes--;
}
ctx->num = chunk%AES_BLOCK_SIZE;
break;
}
}
if (nbytes == 0)
return 1;
#if 0
if (nbytes % AES_BLOCK_SIZE)
return 0; /* are we expected to do tail processing? */
#else
/* nbytes is always multiple of AES_BLOCK_SIZE in ECB and CBC
modes and arbitrary value in byte-oriented modes, such as
CFB and OFB... */
#endif
/* VIA promises CPUs that won't require alignment in the future.
For now padlock_aes_align_required is initialized to 1 and
the condition is never met... */
/* C7 core is capable to manage unaligned input in non-ECB[!]
mode, but performance penalties appear to be approximately
same as for software alignment below or ~3x. They promise to
improve it in the future, but for now we can just as well
pretend that it can only handle aligned input... */
if (!padlock_aes_align_required && (nbytes%AES_BLOCK_SIZE)==0)
return padlock_aes_cipher_omnivorous(ctx, out_arg, in_arg, nbytes);
inp_misaligned = (((size_t)in_arg) & 0x0F);
out_misaligned = (((size_t)out_arg) & 0x0F);
/* Note that even if output is aligned and input not,
* I still prefer to loop instead of copy the whole
* input and then encrypt in one stroke. This is done
* in order to improve L1 cache utilization... */
realign_in_loop = out_misaligned|inp_misaligned;
if (!realign_in_loop && (nbytes%AES_BLOCK_SIZE)==0)
return padlock_aes_cipher_omnivorous(ctx, out_arg, in_arg, nbytes);
/* this takes one "if" out of the loops */
chunk = nbytes;
chunk %= PADLOCK_CHUNK;
if (chunk==0) chunk = PADLOCK_CHUNK;
if (out_misaligned) {
/* optmize for small input */
allocated = (chunk<nbytes?PADLOCK_CHUNK:nbytes);
out = alloca(0x10 + allocated);
out = NEAREST_ALIGNED(out);
}
else
out = out_arg;
cdata = ALIGNED_CIPHER_DATA(ctx);
padlock_verify_context(cdata);
switch (EVP_CIPHER_CTX_mode(ctx)) {
case EVP_CIPH_ECB_MODE:
do {
if (inp_misaligned)
inp = padlock_memcpy(out, in_arg, chunk);
else
inp = in_arg;
in_arg += chunk;
padlock_xcrypt_ecb(chunk/AES_BLOCK_SIZE, cdata, out, inp);
if (out_misaligned)
out_arg = padlock_memcpy(out_arg, out, chunk) + chunk;
else
out = out_arg+=chunk;
nbytes -= chunk;
chunk = PADLOCK_CHUNK;
} while (nbytes);
break;
case EVP_CIPH_CBC_MODE:
memcpy(cdata->iv, ctx->iv, AES_BLOCK_SIZE);
goto cbc_shortcut;
do {
if (iv != cdata->iv)
memcpy(cdata->iv, iv, AES_BLOCK_SIZE);
chunk = PADLOCK_CHUNK;
cbc_shortcut: /* optimize for small input */
if (inp_misaligned)
inp = padlock_memcpy(out, in_arg, chunk);
else
inp = in_arg;
in_arg += chunk;
iv = padlock_xcrypt_cbc(chunk/AES_BLOCK_SIZE, cdata, out, inp);
if (out_misaligned)
out_arg = padlock_memcpy(out_arg, out, chunk) + chunk;
else
out = out_arg+=chunk;
} while (nbytes -= chunk);
memcpy(ctx->iv, iv, AES_BLOCK_SIZE);
break;
case EVP_CIPH_CFB_MODE:
memcpy (iv = cdata->iv, ctx->iv, AES_BLOCK_SIZE);
chunk &= ~(AES_BLOCK_SIZE-1);
if (chunk) goto cfb_shortcut;
else goto cfb_skiploop;
do {
if (iv != cdata->iv)
memcpy(cdata->iv, iv, AES_BLOCK_SIZE);
chunk = PADLOCK_CHUNK;
cfb_shortcut: /* optimize for small input */
if (inp_misaligned)
inp = padlock_memcpy(out, in_arg, chunk);
else
inp = in_arg;
in_arg += chunk;
iv = padlock_xcrypt_cfb(chunk/AES_BLOCK_SIZE, cdata, out, inp);
if (out_misaligned)
out_arg = padlock_memcpy(out_arg, out, chunk) + chunk;
else
out = out_arg+=chunk;
nbytes -= chunk;
} while (nbytes >= AES_BLOCK_SIZE);
cfb_skiploop:
if (nbytes) {
unsigned char *ivp = cdata->iv;
if (iv != ivp) {
memcpy(ivp, iv, AES_BLOCK_SIZE);
iv = ivp;
}
ctx->num = nbytes;
if (cdata->cword.b.encdec) {
cdata->cword.b.encdec=0;
padlock_reload_key();
padlock_xcrypt_ecb(1,cdata,ivp,ivp);
cdata->cword.b.encdec=1;
padlock_reload_key();
while(nbytes) {
unsigned char c = *(in_arg++);
*(out_arg++) = c ^ *ivp;
*(ivp++) = c, nbytes--;
}
}
else { padlock_reload_key();
padlock_xcrypt_ecb(1,cdata,ivp,ivp);
padlock_reload_key();
while (nbytes) {
*ivp = *(out_arg++) = *(in_arg++) ^ *ivp;
ivp++, nbytes--;
}
}
}
memcpy(ctx->iv, iv, AES_BLOCK_SIZE);
break;
case EVP_CIPH_OFB_MODE:
memcpy(cdata->iv, ctx->iv, AES_BLOCK_SIZE);
chunk &= ~(AES_BLOCK_SIZE-1);
if (chunk) do {
if (inp_misaligned)
inp = padlock_memcpy(out, in_arg, chunk);
else
inp = in_arg;
in_arg += chunk;
padlock_xcrypt_ofb(chunk/AES_BLOCK_SIZE, cdata, out, inp);
if (out_misaligned)
out_arg = padlock_memcpy(out_arg, out, chunk) + chunk;
else
out = out_arg+=chunk;
nbytes -= chunk;
chunk = PADLOCK_CHUNK;
} while (nbytes >= AES_BLOCK_SIZE);
if (nbytes) {
unsigned char *ivp = cdata->iv;
ctx->num = nbytes;
padlock_reload_key(); /* empirically found */
padlock_xcrypt_ecb(1,cdata,ivp,ivp);
padlock_reload_key(); /* empirically found */
while (nbytes) {
*(out_arg++) = *(in_arg++) ^ *ivp;
ivp++, nbytes--;
}
}
memcpy(ctx->iv, cdata->iv, AES_BLOCK_SIZE);
break;
default:
return 0;
}
/* Clean the realign buffer if it was used */
if (out_misaligned) {
volatile unsigned long *p=(void *)out;
size_t n = allocated/sizeof(*p);
while (n--) *p++=0;
}
memset(cdata->iv, 0, AES_BLOCK_SIZE);
return 1;
}
#endif /* OPENSSL_NO_AES */
/* ===== Random Number Generator ===== */
/*
* This code is not engaged. The reason is that it does not comply
* with recommendations for VIA RNG usage for secure applications
* (posted at http://www.via.com.tw/en/viac3/c3.jsp) nor does it
* provide meaningful error control...
*/
/* Wrapper that provides an interface between the API and
the raw PadLock RNG */
static int
padlock_rand_bytes(unsigned char *output, int count)
{
unsigned int eax, buf;
while (count >= 8) {
eax = padlock_xstore(output, 0);
if (!(eax&(1<<6))) return 0; /* RNG disabled */
/* this ---vv--- covers DC bias, Raw Bits and String Filter */
if (eax&(0x1F<<10)) return 0;
if ((eax&0x1F)==0) continue; /* no data, retry... */
if ((eax&0x1F)!=8) return 0; /* fatal failure... */
output += 8;
count -= 8;
}
while (count > 0) {
eax = padlock_xstore(&buf, 3);
if (!(eax&(1<<6))) return 0; /* RNG disabled */
/* this ---vv--- covers DC bias, Raw Bits and String Filter */
if (eax&(0x1F<<10)) return 0;
if ((eax&0x1F)==0) continue; /* no data, retry... */
if ((eax&0x1F)!=1) return 0; /* fatal failure... */
*output++ = (unsigned char)buf;
count--;
}
*(volatile unsigned int *)&buf=0;
return 1;
}
/* Dummy but necessary function */
static int
padlock_rand_status(void)
{
return 1;
}
/* Prepare structure for registration */
static RAND_METHOD padlock_rand = {
NULL, /* seed */
padlock_rand_bytes, /* bytes */
NULL, /* cleanup */
NULL, /* add */
padlock_rand_bytes, /* pseudorand */
padlock_rand_status, /* rand status */
( run in 0.874 second using v1.01-cache-2.11-cpan-97f6503c9c8 )