static int blake2b_compress( blake2b_state *S, const uint8_t block[BLAKE2B_BLOCKBYTES] )
{
uint64_t m[16];
uint64_t v[16];
int i;
for( i = 0; i < 16; ++i )
m[i] = load64( block + i * sizeof( m[i] ) );
for( i = 0; i < 8; ++i )
v[i] = S->h[i];
v[ 8] = blake2b_IV[0];
v[ 9] = blake2b_IV[1];
v[10] = blake2b_IV[2];
v[11] = blake2b_IV[3];
v[12] = S->t[0] ^ blake2b_IV[4];
v[13] = S->t[1] ^ blake2b_IV[5];
v[14] = S->f[0] ^ blake2b_IV[6];
v[15] = S->f[1] ^ blake2b_IV[7];
#define G(r,i,a,b,c,d) \
do { \
a = a + b + m[blake2b_sigma[r][2*i+0]]; \
d = rotr64(d ^ a, 32); \
c = c + d; \
b = rotr64(b ^ c, 24); \
a = a + b + m[blake2b_sigma[r][2*i+1]]; \
d = rotr64(d ^ a, 16); \
c = c + d; \
b = rotr64(b ^ c, 63); \
} while(0)
#define ROUND(r) \
do { \
G(r,0,v[ 0],v[ 4],v[ 8],v[12]); \
G(r,1,v[ 1],v[ 5],v[ 9],v[13]); \
G(r,2,v[ 2],v[ 6],v[10],v[14]); \
G(r,3,v[ 3],v[ 7],v[11],v[15]); \
G(r,4,v[ 0],v[ 5],v[10],v[15]); \
G(r,5,v[ 1],v[ 6],v[11],v[12]); \
G(r,6,v[ 2],v[ 7],v[ 8],v[13]); \
G(r,7,v[ 3],v[ 4],v[ 9],v[14]); \
} while(0)
ROUND( 0 );
ROUND( 1 );
ROUND( 2 );
ROUND( 3 );
ROUND( 4 );
ROUND( 5 );
ROUND( 6 );
ROUND( 7 );
ROUND( 8 );
ROUND( 9 );
ROUND( 10 );
ROUND( 11 );
for( i = 0; i < 8; ++i )
S->h[i] = S->h[i] ^ v[i] ^ v[i + 8];
#undef G
#undef ROUND
return 0;
}
void JIT::compileLoadVarargs(Instruction* instruction)
{
int thisValue = instruction[2].u.operand;
int arguments = instruction[3].u.operand;
int firstFreeRegister = instruction[4].u.operand;
killLastResultRegister();
JumpList slowCase;
JumpList end;
bool canOptimize = m_codeBlock->usesArguments()
&& arguments == m_codeBlock->argumentsRegister()
&& !m_codeBlock->symbolTable()->slowArguments();
if (canOptimize) {
emitGetVirtualRegister(arguments, regT0);
slowCase.append(branch64(NotEqual, regT0, TrustedImm64(JSValue::encode(JSValue()))));
emitGetFromCallFrameHeader32(JSStack::ArgumentCount, regT0);
slowCase.append(branch32(Above, regT0, TrustedImm32(Arguments::MaxArguments + 1)));
// regT0: argumentCountIncludingThis
move(regT0, regT1);
add32(TrustedImm32(firstFreeRegister + JSStack::CallFrameHeaderSize), regT1);
lshift32(TrustedImm32(3), regT1);
addPtr(callFrameRegister, regT1);
// regT1: newCallFrame
slowCase.append(branchPtr(Below, AbsoluteAddress(m_globalData->interpreter->stack().addressOfEnd()), regT1));
// Initialize ArgumentCount.
store32(regT0, Address(regT1, JSStack::ArgumentCount * static_cast<int>(sizeof(Register)) + OBJECT_OFFSETOF(EncodedValueDescriptor, asBits.payload)));
// Initialize 'this'.
emitGetVirtualRegister(thisValue, regT2);
store64(regT2, Address(regT1, CallFrame::thisArgumentOffset() * static_cast<int>(sizeof(Register))));
// Copy arguments.
neg32(regT0);
signExtend32ToPtr(regT0, regT0);
end.append(branchAdd64(Zero, TrustedImm32(1), regT0));
// regT0: -argumentCount
Label copyLoop = label();
load64(BaseIndex(callFrameRegister, regT0, TimesEight, CallFrame::thisArgumentOffset() * static_cast<int>(sizeof(Register))), regT2);
store64(regT2, BaseIndex(regT1, regT0, TimesEight, CallFrame::thisArgumentOffset() * static_cast<int>(sizeof(Register))));
branchAdd64(NonZero, TrustedImm32(1), regT0).linkTo(copyLoop, this);
end.append(jump());
}
if (canOptimize)
slowCase.link(this);
JITStubCall stubCall(this, cti_op_load_varargs);
stubCall.addArgument(thisValue, regT0);
stubCall.addArgument(arguments, regT0);
stubCall.addArgument(Imm32(firstFreeRegister));
stubCall.call(regT1);
if (canOptimize)
end.link(this);
}
static int blake2b_compress( blake2b_state *S,
const byte block[BLAKE2B_BLOCKBYTES] )
{
int i;
#ifdef WOLFSSL_SMALL_STACK
word64* m;
word64* v;
m = (word64*)XMALLOC(sizeof(word64) * 16, NULL, DYNAMIC_TYPE_TMP_BUFFER);
if ( m == NULL ) return -1;
v = (word64*)XMALLOC(sizeof(word64) * 16, NULL, DYNAMIC_TYPE_TMP_BUFFER);
if ( v == NULL )
{
XFREE(m, NULL, DYNAMIC_TYPE_TMP_BUFFER);
return -1;
}
#else
word64 m[16];
word64 v[16];
#endif
for( i = 0; i < 16; ++i )
m[i] = load64( block + i * sizeof( m[i] ) );
for( i = 0; i < 8; ++i )
v[i] = S->h[i];
v[ 8] = blake2b_IV[0];
v[ 9] = blake2b_IV[1];
v[10] = blake2b_IV[2];
v[11] = blake2b_IV[3];
v[12] = S->t[0] ^ blake2b_IV[4];
v[13] = S->t[1] ^ blake2b_IV[5];
v[14] = S->f[0] ^ blake2b_IV[6];
v[15] = S->f[1] ^ blake2b_IV[7];
#define G(r,i,a,b,c,d) \
do { \
a = a + b + m[blake2b_sigma[r][2*i+0]]; \
d = rotr64(d ^ a, 32); \
c = c + d; \
b = rotr64(b ^ c, 24); \
a = a + b + m[blake2b_sigma[r][2*i+1]]; \
d = rotr64(d ^ a, 16); \
c = c + d; \
b = rotr64(b ^ c, 63); \
} while(0)
#define ROUND(r) \
do { \
G(r,0,v[ 0],v[ 4],v[ 8],v[12]); \
G(r,1,v[ 1],v[ 5],v[ 9],v[13]); \
G(r,2,v[ 2],v[ 6],v[10],v[14]); \
G(r,3,v[ 3],v[ 7],v[11],v[15]); \
G(r,4,v[ 0],v[ 5],v[10],v[15]); \
G(r,5,v[ 1],v[ 6],v[11],v[12]); \
G(r,6,v[ 2],v[ 7],v[ 8],v[13]); \
G(r,7,v[ 3],v[ 4],v[ 9],v[14]); \
} while(0)
ROUND( 0 );
ROUND( 1 );
ROUND( 2 );
ROUND( 3 );
ROUND( 4 );
ROUND( 5 );
ROUND( 6 );
ROUND( 7 );
ROUND( 8 );
ROUND( 9 );
ROUND( 10 );
ROUND( 11 );
for( i = 0; i < 8; ++i )
S->h[i] = S->h[i] ^ v[i] ^ v[i + 8];
#undef G
#undef ROUND
#ifdef WOLFSSL_SMALL_STACK
XFREE(m, NULL, DYNAMIC_TYPE_TMP_BUFFER);
XFREE(v, NULL, DYNAMIC_TYPE_TMP_BUFFER);
#endif
return 0;
}
void JIT::compileLoadVarargs(Instruction* instruction)
{
int thisValue = instruction[3].u.operand;
int arguments = instruction[4].u.operand;
int firstFreeRegister = instruction[5].u.operand;
JumpList slowCase;
JumpList end;
bool canOptimize = m_codeBlock->usesArguments()
&& arguments == m_codeBlock->argumentsRegister().offset()
&& !m_codeBlock->symbolTable()->slowArguments();
if (canOptimize) {
emitGetVirtualRegister(arguments, regT0);
slowCase.append(branch64(NotEqual, regT0, TrustedImm64(JSValue::encode(JSValue()))));
emitGetFromCallFrameHeader32(JSStack::ArgumentCount, regT0);
slowCase.append(branch32(Above, regT0, TrustedImm32(Arguments::MaxArguments + 1)));
// regT0: argumentCountIncludingThis
move(regT0, regT1);
neg64(regT1);
add64(TrustedImm32(firstFreeRegister - JSStack::CallFrameHeaderSize), regT1);
lshift64(TrustedImm32(3), regT1);
addPtr(callFrameRegister, regT1);
// regT1: newCallFrame
slowCase.append(branchPtr(Above, AbsoluteAddress(m_vm->addressOfJSStackLimit()), regT1));
// Initialize ArgumentCount.
store32(regT0, Address(regT1, JSStack::ArgumentCount * static_cast<int>(sizeof(Register)) + OBJECT_OFFSETOF(EncodedValueDescriptor, asBits.payload)));
// Initialize 'this'.
emitGetVirtualRegister(thisValue, regT2);
store64(regT2, Address(regT1, CallFrame::thisArgumentOffset() * static_cast<int>(sizeof(Register))));
// Copy arguments.
signExtend32ToPtr(regT0, regT0);
end.append(branchSub64(Zero, TrustedImm32(1), regT0));
// regT0: argumentCount
Label copyLoop = label();
load64(BaseIndex(callFrameRegister, regT0, TimesEight, CallFrame::thisArgumentOffset() * static_cast<int>(sizeof(Register))), regT2);
store64(regT2, BaseIndex(regT1, regT0, TimesEight, CallFrame::thisArgumentOffset() * static_cast<int>(sizeof(Register))));
branchSub64(NonZero, TrustedImm32(1), regT0).linkTo(copyLoop, this);
end.append(jump());
}
if (canOptimize)
slowCase.link(this);
emitGetVirtualRegister(arguments, regT1);
callOperation(operationSizeAndAllocFrameForVarargs, regT1, firstFreeRegister);
emitGetVirtualRegister(thisValue, regT1);
emitGetVirtualRegister(arguments, regT2);
callOperation(operationLoadVarargs, returnValueGPR, regT1, regT2);
move(returnValueGPR, regT1);
if (canOptimize)
end.link(this);
}
/* Relocate a non-PLT object with addend. */
static int
reloc_non_plt_obj(Obj_Entry *obj_rtld, Obj_Entry *obj, const Elf_Rela *rela,
SymCache *cache, int flags, RtldLockState *lockstate)
{
struct fptr **fptrs;
Elf_Addr *where = (Elf_Addr *) (obj->relocbase + rela->r_offset);
switch (ELF_R_TYPE(rela->r_info)) {
case R_IA_64_REL64LSB:
/*
* We handle rtld's relocations in rtld_start.S
*/
if (obj != obj_rtld)
store64(where,
load64(where) + (Elf_Addr) obj->relocbase);
break;
case R_IA_64_DIR64LSB: {
const Elf_Sym *def;
const Obj_Entry *defobj;
Elf_Addr target;
def = find_symdef(ELF_R_SYM(rela->r_info), obj, &defobj,
flags, cache, lockstate);
if (def == NULL)
return -1;
target = (def->st_shndx != SHN_UNDEF)
? (Elf_Addr)(defobj->relocbase + def->st_value) : 0;
store64(where, target + rela->r_addend);
break;
}
case R_IA_64_FPTR64LSB: {
/*
* We have to make sure that all @fptr references to
* the same function are identical so that code can
* compare function pointers.
*/
const Elf_Sym *def;
const Obj_Entry *defobj;
struct fptr *fptr = 0;
Elf_Addr target, gp;
int sym_index;
def = find_symdef(ELF_R_SYM(rela->r_info), obj, &defobj,
SYMLOOK_IN_PLT | flags, cache, lockstate);
if (def == NULL) {
/*
* XXX r_debug_state is problematic and find_symdef()
* returns NULL for it. This probably has something to
* do with symbol versioning (r_debug_state is in the
* symbol map). If we return -1 in that case we abort
* relocating rtld, which typically is fatal. So, for
* now just skip the symbol when we're relocating
* rtld. We don't care about r_debug_state unless we
* are being debugged.
*/
if (obj != obj_rtld)
return -1;
break;
}
if (def->st_shndx != SHN_UNDEF) {
target = (Elf_Addr)(defobj->relocbase + def->st_value);
gp = (Elf_Addr)defobj->pltgot;
/* rtld is allowed to reference itself only */
assert(!obj->rtld || obj == defobj);
fptrs = defobj->priv;
if (fptrs == NULL)
fptrs = alloc_fptrs((Obj_Entry *) defobj,
obj->rtld);
sym_index = def - defobj->symtab;
/*
* Find the @fptr, using fptrs as a helper.
*/
if (fptrs)
fptr = fptrs[sym_index];
if (!fptr) {
fptr = alloc_fptr(target, gp);
if (fptrs)
fptrs[sym_index] = fptr;
}
} else
fptr = NULL;
store64(where, (Elf_Addr)fptr);
break;
}
case R_IA_64_IPLTLSB: {
/*
* Relocation typically used to populate C++ virtual function
* tables. It creates a 128-bit function descriptor at the
* specified memory address.
*/
const Elf_Sym *def;
const Obj_Entry *defobj;
struct fptr *fptr;
Elf_Addr target, gp;
def = find_symdef(ELF_R_SYM(rela->r_info), obj, &defobj,
flags, cache, lockstate);
if (def == NULL)
return -1;
if (def->st_shndx != SHN_UNDEF) {
target = (Elf_Addr)(defobj->relocbase + def->st_value);
gp = (Elf_Addr)defobj->pltgot;
} else {
target = 0;
gp = 0;
}
fptr = (void*)where;
store64(&fptr->target, target);
store64(&fptr->gp, gp);
break;
}
case R_IA_64_DTPMOD64LSB: {
const Elf_Sym *def;
const Obj_Entry *defobj;
def = find_symdef(ELF_R_SYM(rela->r_info), obj, &defobj,
flags, cache, lockstate);
if (def == NULL)
return -1;
store64(where, defobj->tlsindex);
break;
}
case R_IA_64_DTPREL64LSB: {
const Elf_Sym *def;
const Obj_Entry *defobj;
def = find_symdef(ELF_R_SYM(rela->r_info), obj, &defobj,
flags, cache, lockstate);
if (def == NULL)
return -1;
store64(where, def->st_value + rela->r_addend);
break;
}
case R_IA_64_TPREL64LSB: {
const Elf_Sym *def;
const Obj_Entry *defobj;
def = find_symdef(ELF_R_SYM(rela->r_info), obj, &defobj,
flags, cache, lockstate);
if (def == NULL)
return -1;
/*
* We lazily allocate offsets for static TLS as we
* see the first relocation that references the
* TLS block. This allows us to support (small
* amounts of) static TLS in dynamically loaded
* modules. If we run out of space, we generate an
* error.
*/
if (!defobj->tls_done) {
if (!allocate_tls_offset((Obj_Entry*) defobj)) {
_rtld_error("%s: No space available for static "
"Thread Local Storage", obj->path);
return -1;
}
}
store64(where, defobj->tlsoffset + def->st_value + rela->r_addend);
break;
}
case R_IA_64_NONE:
break;
default:
_rtld_error("%s: Unsupported relocation type %u"
" in non-PLT relocations\n", obj->path,
(unsigned int)ELF_R_TYPE(rela->r_info));
return -1;
}
return(0);
}
/* Permute the state while xoring in the block of data. */
static void blake2b_compress(BLAKE2B_CTX *S,
const uint8_t *blocks,
size_t len)
{
uint64_t m[16];
uint64_t v[16];
int i;
size_t increment;
/*
* There are two distinct usage vectors for this function:
*
* a) BLAKE2b_Update uses it to process complete blocks,
* possibly more than one at a time;
*
* b) BLAK2b_Final uses it to process last block, always
* single but possibly incomplete, in which case caller
* pads input with zeros.
*/
assert(len < BLAKE2B_BLOCKBYTES || len % BLAKE2B_BLOCKBYTES == 0);
/*
* Since last block is always processed with separate call,
* |len| not being multiple of complete blocks can be observed
* only with |len| being less than BLAKE2B_BLOCKBYTES ("less"
* including even zero), which is why following assignment doesn't
* have to reside inside the main loop below.
*/
increment = len < BLAKE2B_BLOCKBYTES ? len : BLAKE2B_BLOCKBYTES;
for (i = 0; i < 8; ++i) {
v[i] = S->h[i];
}
do {
for (i = 0; i < 16; ++i) {
m[i] = load64(blocks + i * sizeof(m[i]));
}
/* blake2b_increment_counter */
S->t[0] += increment;
S->t[1] += (S->t[0] < increment);
v[8] = blake2b_IV[0];
v[9] = blake2b_IV[1];
v[10] = blake2b_IV[2];
v[11] = blake2b_IV[3];
v[12] = S->t[0] ^ blake2b_IV[4];
v[13] = S->t[1] ^ blake2b_IV[5];
v[14] = S->f[0] ^ blake2b_IV[6];
v[15] = S->f[1] ^ blake2b_IV[7];
#define G(r,i,a,b,c,d) \
do { \
a = a + b + m[blake2b_sigma[r][2*i+0]]; \
d = rotr64(d ^ a, 32); \
c = c + d; \
b = rotr64(b ^ c, 24); \
a = a + b + m[blake2b_sigma[r][2*i+1]]; \
d = rotr64(d ^ a, 16); \
c = c + d; \
b = rotr64(b ^ c, 63); \
} while (0)
#define ROUND(r) \
do { \
G(r,0,v[ 0],v[ 4],v[ 8],v[12]); \
G(r,1,v[ 1],v[ 5],v[ 9],v[13]); \
G(r,2,v[ 2],v[ 6],v[10],v[14]); \
G(r,3,v[ 3],v[ 7],v[11],v[15]); \
G(r,4,v[ 0],v[ 5],v[10],v[15]); \
G(r,5,v[ 1],v[ 6],v[11],v[12]); \
G(r,6,v[ 2],v[ 7],v[ 8],v[13]); \
G(r,7,v[ 3],v[ 4],v[ 9],v[14]); \
} while (0)
#if defined(OPENSSL_SMALL_FOOTPRINT)
/* 3x size reduction on x86_64, almost 7x on ARMv8, 9x on ARMv4 */
for (i = 0; i < 12; i++) {
ROUND(i);
}
#else
ROUND(0);
ROUND(1);
ROUND(2);
ROUND(3);
ROUND(4);
ROUND(5);
ROUND(6);
ROUND(7);
ROUND(8);
ROUND(9);
ROUND(10);
ROUND(11);
#endif
for (i = 0; i < 8; ++i) {
S->h[i] = v[i] ^= v[i + 8] ^ S->h[i];
}
#undef G
#undef ROUND
blocks += increment;
len -= increment;
} while (len);
}
int crypto_aead_decrypt(
unsigned char *m,unsigned long long *mlen,
unsigned char *nsec,
const unsigned char *c,unsigned long long clen,
const unsigned char *ad,unsigned long long adlen,
const unsigned char *npub,
const unsigned char *k
)
{
ICESTATE S;
uint64_t Tcomp[2]; /* computed authentication tag */
uint64_t Trecv[2]; /* received authentication tag */
unsigned int frameBit;
/* ciphertext cannot be shorter than the tag length */
if (clen < ICEPOLETAGLEN) {
return -1;
}
initState128a(S, k, npub);
/* secret message number is zero-length */
frameBit = 0;
processIceBlockRev(S, NULL, NULL, 0, frameBit);
/* process associated data blocks */
do {
unsigned long long blocklen = 128;
frameBit = (adlen <= blocklen ? 1 : 0);
if (adlen < blocklen) {
blocklen = adlen;
}
processIceBlock(S, ad, NULL, blocklen, frameBit);
ad += blocklen;
adlen -= blocklen;
} while (adlen > 0);
/* process ciphertext blocks to get auth tag */
*mlen = 0;
clen -= ICEPOLETAGLEN; /* need to stop before auth tag*/
do {
unsigned long long blocklen = 128;
frameBit = (clen <= blocklen ? 0 : 1);
if (clen < blocklen) {
blocklen = clen;
}
processIceBlockRev(S, c, &m, blocklen, frameBit);
c += blocklen;
*mlen += blocklen;
clen -= blocklen;
} while (clen > 0);
/* compare computed and received auth tags */
Tcomp[0] = S[0][0];
Tcomp[1] = S[1][0];
Trecv[0] = load64(c, 8);
Trecv[1] = load64(c+8, 8);
if (Tcomp[0] != Trecv[0] || Tcomp[1] != Trecv[1]) {
*mlen = 0;
return -1;
}
return 0;
}
