Restructure the unwinder a bit; use visitStackElems and simplify some things

Hopefully a little bit simpler, and duplicating less logic
with Stack::toString.  Contains some fixes to visitStackElems to make
it work for this (it is printing some stacks slightly incorrectly
right now, too ... :)
Esse commit está contido em:
Jordan DeLong
2013-06-04 22:51:47 -07:00
commit de sgolemon
commit 960f36f836
4 arquivos alterados com 260 adições e 248 exclusões
+4 -4
Ver Arquivo
@@ -602,6 +602,10 @@ public:
const Stack& getStack() const { checkRegState(); return m_stack; }
Stack& getStack() { checkRegState(); return m_stack; }
Offset pcOff() const {
return getFP()->m_func->unit()->offsetOf(m_pc);
}
ActRec* m_firstAR;
std::vector<Fault> m_faults;
@@ -772,10 +776,6 @@ public:
static int64_t s_threadIdxCounter;
Variant m_setprofileCallback;
bool m_executingSetprofileCallback;
inline Offset pcOff() const {
return m_fp->m_func->unit()->offsetOf(m_pc);
}
};
class ExecutionContext : public VMExecutionContext {};
+20 -11
Ver Arquivo
@@ -1613,26 +1613,35 @@ resume:
always_assert(action == UnwindAction::Propagate);
}
// Here we have to propagate an exception out of this VM's nesting
// level.
/*
* Here we have to propagate an exception out of this VM's nesting
* level.
*/
if (g_vmContext->m_nestedVMs.empty()) {
m_fp = nullptr;
m_pc = nullptr;
}
assert(m_faults.size() > 0);
Fault fault = m_faults.back();
m_faults.pop_back();
switch (fault.m_faultType) {
case Fault::UserException: {
Object obj = fault.m_userException;
fault.m_userException->decRefCount();
throw obj;
}
case Fault::UserException:
{
Object obj = fault.m_userException;
fault.m_userException->decRefCount();
throw obj;
}
case Fault::CppException:
// throwException() will take care of deleting heap-allocated
// exception object for us
fault.m_cppException->throwException();
NOT_REACHED();
default:
not_implemented();
not_reached();
}
NOT_REACHED();
not_reached();
}
void VMExecutionContext::reenterVM(TypedValue* retval,
+235 -228
Ver Arquivo
@@ -15,6 +15,10 @@
*/
#include "hphp/runtime/vm/unwind.h"
#include <boost/implicit_cast.hpp>
#include "folly/ScopeGuard.h"
#include "hphp/util/trace.h"
#include "hphp/runtime/vm/core_types.h"
#include "hphp/runtime/vm/bytecode.h"
@@ -26,95 +30,131 @@
namespace HPHP {
TRACE_SET_MOD(unwind);
using boost::implicit_cast;
namespace {
//////////////////////////////////////////////////////////////////////
UnwindAction unwindFrag(Stack& stack, ActRec* fp, int offset,
PC& pc, Fault& fault) {
const Func* func = fp->m_func;
FTRACE(1, "unwindFrag: func {} ({})\n",
func->fullName()->data(), func->unit()->filepath()->data());
const bool unwindingGeneratorFrame = func->isGenerator();
auto const curOp = *reinterpret_cast<const Opcode*>(pc);
using namespace HPHP;
const bool unwindingReturningFrame = curOp == OpRetC || curOp == OpRetV;
TypedValue* evalTop;
if (UNLIKELY(unwindingGeneratorFrame)) {
assert(!stack.isValidAddress((uintptr_t)fp));
evalTop = Stack::generatorStackBase(fp);
} else {
assert(stack.isValidAddress((uintptr_t)fp));
evalTop = Stack::frameStackBase(fp);
std::string describeFault(const Fault& f) {
switch (f.m_faultType) {
case Fault::UserException:
return folly::format("[user exception] {}",
implicit_cast<void*>(f.m_userException)).str();
case Fault::CppException:
return folly::format("[cpp exception] {}",
implicit_cast<void*>(f.m_cppException)).str();
}
assert(stack.isValidAddress((uintptr_t)evalTop));
assert(evalTop >= stack.top());
not_reached();
}
while (stack.top() < evalTop) {
stack.popTV();
}
void discardStackTemps(const ActRec* const fp,
Stack& stack,
Offset const bcOffset) {
FTRACE(2, "discardStackTemps with fp {} sp {} pc {}\n",
implicit_cast<const void*>(fp),
implicit_cast<void*>(stack.top()),
bcOffset);
visitStackElems(
fp, stack.top(), bcOffset,
[&] (ActRec* ar) {
assert(ar == reinterpret_cast<ActRec*>(stack.top()));
if (ar->isFromFPushCtor()) {
assert(ar->hasThis());
ar->getThis()->setNoDestruct();
}
stack.popAR();
},
[&] (TypedValue* tv) {
assert(tv == stack.top());
stack.popTV();
}
);
FTRACE(2, "discardStackTemps ends with sp = {}\n",
implicit_cast<void*>(stack.top()));
}
UnwindAction checkHandlers(const EHEnt* eh,
const ActRec* const fp,
PC& pc,
Fault& fault) {
auto const func = fp->m_func;
FTRACE(1, "checkHandlers: func {} ({})\n",
func->fullName()->data(),
func->unit()->filepath()->data());
/*
* This code is repeatedly called with the same offset when an
* exception is raised and rethrown by fault handlers. This
* exception is raised and rethrown by fault handlers. The
* `faultNest' iterator is here to skip the EHEnt handlers that have
* already been run for this in-flight exception.
*/
if (const EHEnt* eh = func->findEH(offset)) {
int faultNest = 0;
for (;;) {
assert(faultNest <= fault.m_handledCount);
if (faultNest == fault.m_handledCount) {
++fault.m_handledCount;
int faultNest = 0;
for (;;) {
assert(faultNest <= fault.m_handledCount);
if (faultNest == fault.m_handledCount) {
++fault.m_handledCount;
switch (eh->m_ehtype) {
case EHEnt::EHType_Fault:
FTRACE(1, "unwindFrag: entering fault at {}: save {}\n",
eh->m_fault,
func->unit()->offsetOf(pc));
fault.m_savedRaiseOffset = func->unit()->offsetOf(pc);
pc = (uchar*)(func->unit()->entry() + eh->m_fault);
DEBUGGER_ATTACHED_ONLY(phpDebuggerExceptionHandlerHook());
return UnwindAction::ResumeVM;
case EHEnt::EHType_Catch:
// Note: we skip catch clauses if we have a pending C++ exception
// as part of our efforts to avoid running more PHP code in the
// face of such exceptions.
if ((fault.m_faultType == Fault::UserException) &&
(ThreadInfo::s_threadInfo->m_pendingException == nullptr)) {
ObjectData* obj = fault.m_userException;
for (auto& idOff : eh->m_catches) {
auto handler = func->unit()->at(idOff.second);
FTRACE(1, "unwindFrag: catch candidate {}\n", handler);
Class* cls = Unit::lookupClass(
func->unit()->lookupNamedEntityId(idOff.first)
);
if (cls && obj->instanceof(cls)) {
pc = handler;
FTRACE(1, "unwindFrag: entering catch at {}\n", pc);
DEBUGGER_ATTACHED_ONLY(phpDebuggerExceptionHandlerHook());
return UnwindAction::ResumeVM;
}
}
switch (eh->m_ehtype) {
case EHEnt::EHType_Fault:
FTRACE(1, "checkHandlers: entering fault at {}: save {}\n",
eh->m_fault,
func->unit()->offsetOf(pc));
fault.m_savedRaiseOffset = func->unit()->offsetOf(pc);
pc = func->unit()->entry() + eh->m_fault;
DEBUGGER_ATTACHED_ONLY(phpDebuggerExceptionHandlerHook());
return UnwindAction::ResumeVM;
case EHEnt::EHType_Catch:
// Note: we skip catch clauses if we have a pending C++ exception
// as part of our efforts to avoid running more PHP code in the
// face of such exceptions.
if (fault.m_faultType == Fault::UserException &&
ThreadInfo::s_threadInfo->m_pendingException == nullptr) {
auto const obj = fault.m_userException;
for (auto& idOff : eh->m_catches) {
auto handler = func->unit()->at(idOff.second);
FTRACE(1, "checkHandlers: catch candidate {}\n", handler);
auto const cls = Unit::lookupClass(
func->unit()->lookupNamedEntityId(idOff.first)
);
if (!cls || !obj->instanceof(cls)) continue;
FTRACE(1, "checkHandlers: entering catch at {}\n", pc);
pc = handler;
DEBUGGER_ATTACHED_ONLY(phpDebuggerExceptionHandlerHook());
return UnwindAction::ResumeVM;
}
break;
}
}
if (eh->m_parentIndex != -1) {
eh = &func->ehtab()[eh->m_parentIndex];
} else {
break;
}
++faultNest;
}
if (eh->m_parentIndex != -1) {
eh = &func->ehtab()[eh->m_parentIndex];
} else {
break;
}
++faultNest;
}
// We found no more handlers in this frame, so the nested fault
// count starts over for the caller frame.
fault.m_handledCount = 0;
return UnwindAction::Propagate;
}
void tearDownFrame(ActRec*& fp, Stack& stack, PC& pc, Offset& faultOffset) {
auto const func = fp->m_func;
auto const curOp = *reinterpret_cast<const Opcode*>(pc);
auto const unwindingGeneratorFrame = func->isGenerator();
auto const unwindingReturningFrame = curOp == OpRetC || curOp == OpRetV;
auto const prevFp = fp->arGetSfp();
FTRACE(1, "tearDownFrame: {} ({})\n fp {} prevFp {}\n",
func->fullName()->data(),
func->unit()->filepath()->data(),
implicit_cast<void*>(fp),
implicit_cast<void*>(prevFp));
if (fp->isFromFPushCtor() && fp->hasThis()) {
fp->getThis()->setNoDestruct();
@@ -154,174 +194,110 @@ UnwindAction unwindFrag(Stack& stack, ActRec* fp, int offset,
} catch (...) {}
}
stack.ndiscard(func->numSlotsInFrame());
stack.discardAR();
}
FTRACE(1, "unwindFrag: propagate\n");
return UnwindAction::Propagate;
assert(stack.isValidAddress(reinterpret_cast<uintptr_t>(prevFp)) ||
prevFp->m_func->isGenerator());
auto const prevOff = fp->m_soff + prevFp->m_func->base();
pc = prevFp->m_func->unit()->at(prevOff);
fp = prevFp;
faultOffset = prevOff;
}
// Pops everything between the current stack pointer and the passed ActRec*.
// It assumes everything there is values, not ActRecs.
void unwindARFrag(Stack& stack, ActRec* ar) {
while (stack.top() < (TypedValue*)ar) {
stack.popTV();
}
}
/*
* Unwinding proceeds as follows:
*
* - Discard all evaluation stack temporaries (including pre-live
* activation records).
*
* - Check if the faultOffset that raised the exception is inside a
* protected region, if so, if it can handle the Fault resume the
* VM at the handler.
*
* - Failing any of the above, pop the frame for the current
* function. If the current function was the last frame in the
* current VM nesting level, return UnwindAction::Propagate,
* otherwise go to the first step and repeat this process in the
* caller's frame.
*
* Note: it's important that the unwinder makes a copy of the Fault
* it's currently operating on, as the underlying faults vector may
* reallocate due to nested exception handling.
*/
UnwindAction unwind(ActRec*& fp,
Stack& stack,
PC& pc,
Offset faultOffset, // distinct from pc after iopUnwind
Fault fault) {
FTRACE(1, "entering unwinder for fault: {}\n", describeFault(fault));
SCOPE_EXIT {
FTRACE(1, "leaving unwinder for fault: {}\n", describeFault(fault));
};
// Pops everything up to and including the outermost unactivated ActRec. Since
// it's impossible to have more than one chain of nested unactivated ActRecs
// on the stack, this means that after this function returns, everything
// between the stack pointer and frame pointer is a value, Iter or local.
void unwindAR(Stack& stack, ActRec* fp, const FPIEnt* fe) {
while (true) {
TRACE(1, "unwindAR: function %s, pIdx %d\n",
fp->m_func->name()->data(), fe->m_parentIndex);
ActRec* ar;
if (LIKELY(!fp->m_func->isGenerator())) {
ar = arAtOffset(fp, -fe->m_fpOff);
} else {
// FIXME: duplicated logic from visitStackElems
TypedValue* genStackBase = Stack::generatorStackBase(fp);
ActRec* fakePrevFP =
(ActRec*)(genStackBase + fp->m_func->numSlotsInFrame());
ar = arAtOffset(fakePrevFP, -fe->m_fpOff);
}
assert((TypedValue*)ar >= stack.top());
unwindARFrag(stack, ar);
if (ar->isFromFPushCtor()) {
assert(ar->hasThis());
ar->getThis()->setNoDestruct();
}
stack.popAR();
if (fe->m_parentIndex != -1) {
fe = &fp->m_func->fpitab()[fe->m_parentIndex];
} else {
return;
}
}
}
UnwindAction unwindFrame(Stack& stack,
ActRec*& fp,
Offset offset,
PC& pc,
Fault fault) {
VMExecutionContext* context = g_vmContext;
while (true) {
FTRACE(1, "unwindFrame: func {}, offset {} fp {}\n",
for (;;) {
FTRACE(1, "unwind: func {}, faultOffset {} fp {}\n",
fp->m_func->name()->data(),
offset,
static_cast<void*>(fp));
faultOffset,
implicit_cast<void*>(fp));
// If the exception is already propagating, if it was in any FPI
// region we already handled unwinding it the first time around.
/*
* If the handledCount is non-zero, we've already this fault once
* while unwinding this frema, and popped all eval stack
* temporaries the first time it was thrown (before entering a
* fault funclet). When the Unwind instruction was executed in
* the funclet, the eval stack must have been left empty again.
*
* (We have to skip discardStackTemps in this case because it will
* look for FPI regions and assume the stack offsets correspond to
* what the FPI table expects.)
*/
if (fault.m_handledCount == 0) {
if (const FPIEnt *fe = fp->m_func->findFPI(offset)) {
unwindAR(stack, fp, fe);
discardStackTemps(fp, stack, faultOffset);
}
if (const EHEnt* eh = fp->m_func->findEH(faultOffset)) {
switch (checkHandlers(eh, fp, pc, fault)) {
case UnwindAction::ResumeVM:
// We've kept our own copy of the Fault, because m_faults may
// change if we have a reentry during unwinding. When we're
// ready to resume, we need to replace the fault to reflect
// any state changes we've made (handledCount, etc).
g_vmContext->m_faults.back() = fault;
return UnwindAction::ResumeVM;
case UnwindAction::Propagate:
break;
}
}
if (unwindFrag(stack, fp, offset, pc, fault) == UnwindAction::ResumeVM) {
// We've kept our own copy of the Fault, because m_faults may
// change if we have a reentry during unwinding. When we're
// ready to resume, we need to replace the current fault to
// reflect any state changes we've made (handledCount, etc).
assert(!context->m_faults.empty());
context->m_faults.back() = fault;
return UnwindAction::ResumeVM;
}
// We found no more handlers in this frame, so the nested fault
// count starts over for the caller frame.
fault.m_handledCount = 0;
ActRec *prevFp = fp->arGetSfp();
FTRACE(1, "unwindFrame: fp {} prevFp {}\n",
static_cast<void*>(fp),
static_cast<void*>(prevFp));
if (LIKELY(!fp->m_func->isGenerator())) {
// We don't need to refcount the AR's refcounted members; that was
// taken care of in frame_free_locals, called from unwindFrag().
// If it's a generator, the AR doesn't live on this stack.
stack.discardAR();
}
if (prevFp == fp) {
TRACE(1, "unwindFrame: reached the end of this nesting's ActRec "
"chain\n");
auto const lastFrameForNesting = fp == fp->arGetSfp();
tearDownFrame(fp, stack, pc, faultOffset);
if (lastFrameForNesting) {
FTRACE(1, "unwind: reached the end of this nesting's ActRec chain\n");
break;
}
// Keep the pc up to date while unwinding.
Offset prevOff = fp->m_soff + prevFp->m_func->base();
const Func *prevF = prevFp->m_func;
assert(stack.isValidAddress((uintptr_t)prevFp) || prevF->isGenerator());
pc = prevF->unit()->at(prevOff);
fp = prevFp;
offset = prevOff;
}
return UnwindAction::Propagate;
}
void pushFault(Fault::Type t, Exception* e, const Object* o = nullptr) {
FTRACE(1, "pushing new fault: {} {} {}\n",
t == Fault::UserException ? "[user exception]" : "[cpp exception]",
e, o);
VMExecutionContext* ec = g_vmContext;
Fault fault;
fault.m_faultType = t;
if (t == Fault::UserException) {
// User object.
assert(o);
fault.m_userException = o->get();
fault.m_userException->incRefCount();
} else {
fault.m_cppException = e;
}
ec->m_faults.push_back(fault);
}
UnwindAction handleUnwind(UnwindAction unwindType) {
UnwindAction longJumpType;
if (unwindType == UnwindAction::Propagate) {
longJumpType = UnwindAction::Propagate;
if (g_vmContext->m_nestedVMs.empty()) {
g_vmContext->m_fp = nullptr;
g_vmContext->m_pc = nullptr;
}
} else {
assert(unwindType == UnwindAction::ResumeVM);
longJumpType = UnwindAction::ResumeVM;
}
return longJumpType;
}
UnwindAction hhvmPrepareThrow() {
Fault& fault = g_vmContext->m_faults.back();
TRACE(2, "hhvmPrepareThrow: %p(\"%s\") {\n",
g_vmContext->getFP(),
g_vmContext->getFP()->m_func->name()->data());
UnwindAction unwindType;
unwindType = unwindFrame(g_vmContext->getStack(),
g_vmContext->m_fp, // by ref
g_vmContext->pcOff(),
g_vmContext->m_pc, // by ref
fault);
return handleUnwind(unwindType);
}
const StaticString s_hphpd_break("hphpd_break");
const StaticString s_fb_enable_code_coverage("fb_enable_code_coverage");
// Unwind the frame for a builtin. Currently only used when switching
// modes for hphpd_break and fb_enable_code_coverage.
void unwindBuiltinFrame() {
auto& stack = g_vmContext->getStack();
auto& fp = g_vmContext->m_fp;
assert(fp->m_func->info());
assert(fp->m_func->name()->isame(s_hphpd_break.get()) ||
fp->m_func->name()->isame(s_fb_enable_code_coverage.get()));
auto& stack = g_vmContext->getStack();
// Free any values that may be on the eval stack. We know there
// can't be FPI regions and it can't be a generator body because
// it's a builtin frame.
@@ -332,6 +308,7 @@ void unwindBuiltinFrame() {
// Free the locals and VarEnv if there is one
frame_free_locals_inl(fp, fp->m_func->numLocals());
// Tear down the frame
Offset pc = -1;
ActRec* sfp = g_vmContext->getPrevVMState(fp, &pc);
@@ -341,38 +318,70 @@ void unwindBuiltinFrame() {
stack.discardAR();
}
void pushFault(Exception* e) {
Fault f;
f.m_faultType = Fault::CppException;
f.m_cppException = e;
g_vmContext->m_faults.push_back(f);
FTRACE(1, "pushing new fault: {}\n", describeFault(f));
}
void pushFault(const Object& o) {
Fault f;
f.m_faultType = Fault::UserException;
f.m_userException = o.get();
f.m_userException->incRefCount();
g_vmContext->m_faults.push_back(f);
FTRACE(1, "pushing new fault: {}\n", describeFault(f));
}
UnwindAction enterUnwinder() {
auto fault = g_vmContext->m_faults.back();
return unwind(
g_vmContext->m_fp, // by ref
g_vmContext->getStack(),// by ref
g_vmContext->m_pc, // by ref
g_vmContext->pcOff(),
fault
);
}
//////////////////////////////////////////////////////////////////////
}
UnwindAction exception_handler() noexcept {
FTRACE(1, "unwind exception_handler\n");
g_vmContext->checkRegState();
try { throw; }
/*
* Unwind (repropagating from a fault funclet) is slightly different
* from the throw case, because we need to re-raise the exception as
* if it came from the same offset to handle nested fault handlers
* correctly.
* from the throw cases, because we need to re-raise the exception
* as if it came from the same offset to handle nested fault
* handlers correctly, and we continue propagating the current Fault
* instead of pushing a new one.
*/
catch (const VMPrepareUnwind&) {
Fault fault = g_vmContext->m_faults.back();
Offset faultPC = fault.m_savedRaiseOffset;
FTRACE(1, "unwind: restoring offset {}\n", faultPC);
assert(faultPC != kInvalidOffset);
Offset faultOffset = fault.m_savedRaiseOffset;
FTRACE(1, "unwind: restoring offset {}\n", faultOffset);
assert(faultOffset != kInvalidOffset);
fault.m_savedRaiseOffset = kInvalidOffset;
UnwindAction unwindType = unwindFrame(
g_vmContext->getStack(),
return unwind(
g_vmContext->m_fp,
faultPC,
g_vmContext->getStack(),
g_vmContext->m_pc,
faultOffset,
fault
);
return handleUnwind(unwindType);
}
catch (const Object& e) {
pushFault(Fault::UserException, nullptr, &e);
return hhvmPrepareThrow();
catch (const Object& o) {
pushFault(o);
return enterUnwinder();
}
catch (VMSwitchMode&) {
@@ -386,20 +395,18 @@ UnwindAction exception_handler() noexcept {
}
catch (Exception& e) {
pushFault(Fault::CppException, e.clone());
return hhvmPrepareThrow();
pushFault(e.clone());;
return enterUnwinder();
}
catch (std::exception& e) {
pushFault(Fault::CppException,
new Exception("unexpected %s: %s", typeid(e).name(), e.what()));
return hhvmPrepareThrow();
pushFault(new Exception("unexpected %s: %s", typeid(e).name(), e.what()));
return enterUnwinder();
}
catch (...) {
pushFault(Fault::CppException,
new Exception("unknown exception"));
return hhvmPrepareThrow();
pushFault(new Exception("unknown exception"));
return enterUnwinder();
}
not_reached();
+1 -5
Ver Arquivo
@@ -16,16 +16,12 @@
#ifndef incl_HPHP_UNWIND_H_
#define incl_HPHP_UNWIND_H_
#include "hphp/runtime/vm/core_types.h"
#include <stdexcept>
namespace HPHP {
//////////////////////////////////////////////////////////////////////
struct Stack;
struct ActRec;
struct Fault;
/*
* Enumerates actions that should be taken by the enterVM loop after
* unwinding an exception.