2013-05-22 04:24:07 +08:00
|
|
|
//===-- RTDyldMemoryManager.cpp - Memory manager for MC-JIT -----*- C++ -*-===//
|
|
|
|
//
|
2019-01-19 16:50:56 +08:00
|
|
|
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
|
|
|
|
// See https://llvm.org/LICENSE.txt for license information.
|
|
|
|
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
|
2013-05-22 04:24:07 +08:00
|
|
|
//
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
//
|
|
|
|
// Implementation of the runtime dynamic memory manager base class.
|
|
|
|
//
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
|
|
|
|
#include "llvm/Config/config.h"
|
|
|
|
#include "llvm/ExecutionEngine/RTDyldMemoryManager.h"
|
2014-12-31 06:52:33 +08:00
|
|
|
#include "llvm/Support/Compiler.h"
|
2013-05-22 04:24:07 +08:00
|
|
|
#include "llvm/Support/DynamicLibrary.h"
|
|
|
|
#include "llvm/Support/ErrorHandling.h"
|
2013-05-25 04:54:11 +08:00
|
|
|
#include <cstdlib>
|
|
|
|
|
2013-05-22 04:24:07 +08:00
|
|
|
#ifdef __linux__
|
|
|
|
// These includes used by RTDyldMemoryManager::getPointerToNamedFunction()
|
|
|
|
// for Glibc trickery. See comments in this function for more information.
|
|
|
|
#ifdef HAVE_SYS_STAT_H
|
|
|
|
#include <sys/stat.h>
|
|
|
|
#endif
|
|
|
|
#include <fcntl.h>
|
|
|
|
#include <unistd.h>
|
|
|
|
#endif
|
|
|
|
|
|
|
|
namespace llvm {
|
|
|
|
|
|
|
|
RTDyldMemoryManager::~RTDyldMemoryManager() {}
|
|
|
|
|
|
|
|
// Determine whether we can register EH tables.
|
2019-05-23 01:41:27 +08:00
|
|
|
#if (defined(__GNUC__) && !defined(__ARM_EABI__) && !defined(__ia64__) && \
|
|
|
|
!(defined(_AIX) && defined(__ibmxl__)) && !defined(__SEH__) && \
|
|
|
|
!defined(__USING_SJLJ_EXCEPTIONS__))
|
2013-05-22 04:24:07 +08:00
|
|
|
#define HAVE_EHTABLE_SUPPORT 1
|
|
|
|
#else
|
|
|
|
#define HAVE_EHTABLE_SUPPORT 0
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#if HAVE_EHTABLE_SUPPORT
|
2015-06-20 01:21:02 +08:00
|
|
|
extern "C" void __register_frame(void *);
|
|
|
|
extern "C" void __deregister_frame(void *);
|
2013-10-16 08:14:21 +08:00
|
|
|
#else
|
|
|
|
// The building compiler does not have __(de)register_frame but
|
|
|
|
// it may be found at runtime in a dynamically-loaded library.
|
|
|
|
// For example, this happens when building LLVM with Visual C++
|
|
|
|
// but using the MingW runtime.
|
Initial implementation of JITLink - A replacement for RuntimeDyld.
Summary:
JITLink is a jit-linker that performs the same high-level task as RuntimeDyld:
it parses relocatable object files and makes their contents runnable in a target
process.
JITLink aims to improve on RuntimeDyld in several ways:
(1) A clear design intended to maximize code-sharing while minimizing coupling.
RuntimeDyld has been developed in an ad-hoc fashion for a number of years and
this had led to intermingling of code for multiple architectures (e.g. in
RuntimeDyldELF::processRelocationRef) in a way that makes the code more
difficult to read, reason about, extend. JITLink is designed to isolate
format and architecture specific code, while still sharing generic code.
(2) Support for native code models.
RuntimeDyld required the use of large code models (where calls to external
functions are made indirectly via registers) for many of platforms due to its
restrictive model for stub generation (one "stub" per symbol). JITLink allows
arbitrary mutation of the atom graph, allowing both GOT and PLT atoms to be
added naturally.
(3) Native support for asynchronous linking.
JITLink uses asynchronous calls for symbol resolution and finalization: these
callbacks are passed a continuation function that they must call to complete the
linker's work. This allows for cleaner interoperation with the new concurrent
ORC JIT APIs, while still being easily implementable in synchronous style if
asynchrony is not needed.
To maximise sharing, the design has a hierarchy of common code:
(1) Generic atom-graph data structure and algorithms (e.g. dead stripping and
| memory allocation) that are intended to be shared by all architectures.
|
+ -- (2) Shared per-format code that utilizes (1), e.g. Generic MachO to
| atom-graph parsing.
|
+ -- (3) Architecture specific code that uses (1) and (2). E.g.
JITLinkerMachO_x86_64, which adds x86-64 specific relocation
support to (2) to build and patch up the atom graph.
To support asynchronous symbol resolution and finalization, the callbacks for
these operations take continuations as arguments:
using JITLinkAsyncLookupContinuation =
std::function<void(Expected<AsyncLookupResult> LR)>;
using JITLinkAsyncLookupFunction =
std::function<void(const DenseSet<StringRef> &Symbols,
JITLinkAsyncLookupContinuation LookupContinuation)>;
using FinalizeContinuation = std::function<void(Error)>;
virtual void finalizeAsync(FinalizeContinuation OnFinalize);
In addition to its headline features, JITLink also makes other improvements:
- Dead stripping support: symbols that are not used (e.g. redundant ODR
definitions) are discarded, and take up no memory in the target process
(In contrast, RuntimeDyld supported pointer equality for weak definitions,
but the redundant definitions stayed resident in memory).
- Improved exception handling support. JITLink provides a much more extensive
eh-frame parser than RuntimeDyld, and is able to correctly fix up many
eh-frame sections that RuntimeDyld currently (silently) fails on.
- More extensive validation and error handling throughout.
This initial patch supports linking MachO/x86-64 only. Work on support for
other architectures and formats will happen in-tree.
Differential Revision: https://reviews.llvm.org/D58704
llvm-svn: 358818
2019-04-21 01:10:34 +08:00
|
|
|
static void __register_frame(void *p) {
|
2013-10-16 08:14:21 +08:00
|
|
|
static bool Searched = false;
|
2015-06-20 01:21:02 +08:00
|
|
|
static void((*rf)(void *)) = 0;
|
2013-10-16 08:14:21 +08:00
|
|
|
|
|
|
|
if (!Searched) {
|
|
|
|
Searched = true;
|
2015-06-20 01:21:02 +08:00
|
|
|
*(void **)&rf =
|
|
|
|
llvm::sys::DynamicLibrary::SearchForAddressOfSymbol("__register_frame");
|
2013-10-16 08:14:21 +08:00
|
|
|
}
|
|
|
|
if (rf)
|
2015-06-20 01:21:02 +08:00
|
|
|
rf(p);
|
2013-10-16 08:14:21 +08:00
|
|
|
}
|
2013-05-22 04:24:07 +08:00
|
|
|
|
Initial implementation of JITLink - A replacement for RuntimeDyld.
Summary:
JITLink is a jit-linker that performs the same high-level task as RuntimeDyld:
it parses relocatable object files and makes their contents runnable in a target
process.
JITLink aims to improve on RuntimeDyld in several ways:
(1) A clear design intended to maximize code-sharing while minimizing coupling.
RuntimeDyld has been developed in an ad-hoc fashion for a number of years and
this had led to intermingling of code for multiple architectures (e.g. in
RuntimeDyldELF::processRelocationRef) in a way that makes the code more
difficult to read, reason about, extend. JITLink is designed to isolate
format and architecture specific code, while still sharing generic code.
(2) Support for native code models.
RuntimeDyld required the use of large code models (where calls to external
functions are made indirectly via registers) for many of platforms due to its
restrictive model for stub generation (one "stub" per symbol). JITLink allows
arbitrary mutation of the atom graph, allowing both GOT and PLT atoms to be
added naturally.
(3) Native support for asynchronous linking.
JITLink uses asynchronous calls for symbol resolution and finalization: these
callbacks are passed a continuation function that they must call to complete the
linker's work. This allows for cleaner interoperation with the new concurrent
ORC JIT APIs, while still being easily implementable in synchronous style if
asynchrony is not needed.
To maximise sharing, the design has a hierarchy of common code:
(1) Generic atom-graph data structure and algorithms (e.g. dead stripping and
| memory allocation) that are intended to be shared by all architectures.
|
+ -- (2) Shared per-format code that utilizes (1), e.g. Generic MachO to
| atom-graph parsing.
|
+ -- (3) Architecture specific code that uses (1) and (2). E.g.
JITLinkerMachO_x86_64, which adds x86-64 specific relocation
support to (2) to build and patch up the atom graph.
To support asynchronous symbol resolution and finalization, the callbacks for
these operations take continuations as arguments:
using JITLinkAsyncLookupContinuation =
std::function<void(Expected<AsyncLookupResult> LR)>;
using JITLinkAsyncLookupFunction =
std::function<void(const DenseSet<StringRef> &Symbols,
JITLinkAsyncLookupContinuation LookupContinuation)>;
using FinalizeContinuation = std::function<void(Error)>;
virtual void finalizeAsync(FinalizeContinuation OnFinalize);
In addition to its headline features, JITLink also makes other improvements:
- Dead stripping support: symbols that are not used (e.g. redundant ODR
definitions) are discarded, and take up no memory in the target process
(In contrast, RuntimeDyld supported pointer equality for weak definitions,
but the redundant definitions stayed resident in memory).
- Improved exception handling support. JITLink provides a much more extensive
eh-frame parser than RuntimeDyld, and is able to correctly fix up many
eh-frame sections that RuntimeDyld currently (silently) fails on.
- More extensive validation and error handling throughout.
This initial patch supports linking MachO/x86-64 only. Work on support for
other architectures and formats will happen in-tree.
Differential Revision: https://reviews.llvm.org/D58704
llvm-svn: 358818
2019-04-21 01:10:34 +08:00
|
|
|
static void __deregister_frame(void *p) {
|
2013-10-16 08:14:21 +08:00
|
|
|
static bool Searched = false;
|
2015-06-20 01:21:02 +08:00
|
|
|
static void((*df)(void *)) = 0;
|
2013-10-16 08:14:21 +08:00
|
|
|
|
|
|
|
if (!Searched) {
|
|
|
|
Searched = true;
|
2015-06-20 01:21:02 +08:00
|
|
|
*(void **)&df = llvm::sys::DynamicLibrary::SearchForAddressOfSymbol(
|
|
|
|
"__deregister_frame");
|
2013-10-16 08:14:21 +08:00
|
|
|
}
|
|
|
|
if (df)
|
2015-06-20 01:21:02 +08:00
|
|
|
df(p);
|
2013-10-16 08:14:21 +08:00
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifdef __APPLE__
|
|
|
|
|
|
|
|
static const char *processFDE(const char *Entry, bool isDeregister) {
|
2013-05-22 04:24:07 +08:00
|
|
|
const char *P = Entry;
|
2013-06-04 17:09:15 +08:00
|
|
|
uint32_t Length = *((const uint32_t *)P);
|
2013-05-22 04:24:07 +08:00
|
|
|
P += 4;
|
2013-06-04 17:09:15 +08:00
|
|
|
uint32_t Offset = *((const uint32_t *)P);
|
2013-10-16 14:50:36 +08:00
|
|
|
if (Offset != 0) {
|
|
|
|
if (isDeregister)
|
2013-10-16 08:14:21 +08:00
|
|
|
__deregister_frame(const_cast<char *>(Entry));
|
2013-10-16 14:50:36 +08:00
|
|
|
else
|
2013-10-16 08:14:21 +08:00
|
|
|
__register_frame(const_cast<char *>(Entry));
|
2013-10-16 14:50:36 +08:00
|
|
|
}
|
2013-05-22 04:24:07 +08:00
|
|
|
return P + Length;
|
|
|
|
}
|
|
|
|
|
2013-10-12 05:25:48 +08:00
|
|
|
// This implementation handles frame registration for local targets.
|
|
|
|
// Memory managers for remote targets should re-implement this function
|
|
|
|
// and use the LoadAddr parameter.
|
2016-01-15 06:02:03 +08:00
|
|
|
void RTDyldMemoryManager::registerEHFramesInProcess(uint8_t *Addr,
|
|
|
|
size_t Size) {
|
2013-10-16 08:14:21 +08:00
|
|
|
// On OS X OS X __register_frame takes a single FDE as an argument.
|
2016-12-22 04:51:42 +08:00
|
|
|
// See http://lists.llvm.org/pipermail/llvm-dev/2013-April/061737.html
|
|
|
|
// and projects/libunwind/src/UnwindLevel1-gcc-ext.c.
|
2013-10-12 05:25:48 +08:00
|
|
|
const char *P = (const char *)Addr;
|
|
|
|
const char *End = P + Size;
|
2013-05-22 04:24:07 +08:00
|
|
|
do {
|
2013-10-16 08:14:21 +08:00
|
|
|
P = processFDE(P, false);
|
2013-05-22 04:24:07 +08:00
|
|
|
} while(P != End);
|
|
|
|
}
|
|
|
|
|
2016-01-15 06:02:03 +08:00
|
|
|
void RTDyldMemoryManager::deregisterEHFramesInProcess(uint8_t *Addr,
|
|
|
|
size_t Size) {
|
2013-10-16 08:14:21 +08:00
|
|
|
const char *P = (const char *)Addr;
|
|
|
|
const char *End = P + Size;
|
|
|
|
do {
|
|
|
|
P = processFDE(P, true);
|
|
|
|
} while(P != End);
|
|
|
|
}
|
|
|
|
|
|
|
|
#else
|
|
|
|
|
2016-01-15 06:02:03 +08:00
|
|
|
void RTDyldMemoryManager::registerEHFramesInProcess(uint8_t *Addr,
|
|
|
|
size_t Size) {
|
2018-07-31 03:41:25 +08:00
|
|
|
// On Linux __register_frame takes a single argument:
|
2013-10-16 08:14:21 +08:00
|
|
|
// a pointer to the start of the .eh_frame section.
|
|
|
|
|
2018-07-31 03:41:25 +08:00
|
|
|
// How can it find the end? Because crtendS.o is linked
|
2013-10-16 08:14:21 +08:00
|
|
|
// in and it has an .eh_frame section with four zero chars.
|
|
|
|
__register_frame(Addr);
|
|
|
|
}
|
|
|
|
|
2016-01-15 06:02:03 +08:00
|
|
|
void RTDyldMemoryManager::deregisterEHFramesInProcess(uint8_t *Addr,
|
|
|
|
size_t Size) {
|
2013-10-16 08:14:21 +08:00
|
|
|
__deregister_frame(Addr);
|
|
|
|
}
|
|
|
|
|
|
|
|
#endif
|
|
|
|
|
2017-05-10 05:32:18 +08:00
|
|
|
void RTDyldMemoryManager::registerEHFrames(uint8_t *Addr, uint64_t LoadAddr,
|
|
|
|
size_t Size) {
|
|
|
|
registerEHFramesInProcess(Addr, Size);
|
|
|
|
EHFrames.push_back({Addr, Size});
|
|
|
|
}
|
|
|
|
|
|
|
|
void RTDyldMemoryManager::deregisterEHFrames() {
|
|
|
|
for (auto &Frame : EHFrames)
|
|
|
|
deregisterEHFramesInProcess(Frame.Addr, Frame.Size);
|
|
|
|
EHFrames.clear();
|
|
|
|
}
|
|
|
|
|
2013-05-22 04:24:07 +08:00
|
|
|
static int jit_noop() {
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2013-11-16 01:59:43 +08:00
|
|
|
// ARM math functions are statically linked on Android from libgcc.a, but not
|
|
|
|
// available at runtime for dynamic linking. On Linux these are usually placed
|
|
|
|
// in libgcc_s.so so can be found by normal dynamic lookup.
|
|
|
|
#if defined(__BIONIC__) && defined(__arm__)
|
|
|
|
// List of functions which are statically linked on Android and can be generated
|
|
|
|
// by LLVM. This is done as a nested macro which is used once to declare the
|
|
|
|
// imported functions with ARM_MATH_DECL and once to compare them to the
|
|
|
|
// user-requested symbol in getSymbolAddress with ARM_MATH_CHECK. The test
|
|
|
|
// assumes that all functions start with __aeabi_ and getSymbolAddress must be
|
|
|
|
// modified if that changes.
|
|
|
|
#define ARM_MATH_IMPORTS(PP) \
|
|
|
|
PP(__aeabi_d2f) \
|
|
|
|
PP(__aeabi_d2iz) \
|
|
|
|
PP(__aeabi_d2lz) \
|
|
|
|
PP(__aeabi_d2uiz) \
|
|
|
|
PP(__aeabi_d2ulz) \
|
|
|
|
PP(__aeabi_dadd) \
|
|
|
|
PP(__aeabi_dcmpeq) \
|
|
|
|
PP(__aeabi_dcmpge) \
|
|
|
|
PP(__aeabi_dcmpgt) \
|
|
|
|
PP(__aeabi_dcmple) \
|
|
|
|
PP(__aeabi_dcmplt) \
|
|
|
|
PP(__aeabi_dcmpun) \
|
|
|
|
PP(__aeabi_ddiv) \
|
|
|
|
PP(__aeabi_dmul) \
|
|
|
|
PP(__aeabi_dsub) \
|
|
|
|
PP(__aeabi_f2d) \
|
|
|
|
PP(__aeabi_f2iz) \
|
|
|
|
PP(__aeabi_f2lz) \
|
|
|
|
PP(__aeabi_f2uiz) \
|
|
|
|
PP(__aeabi_f2ulz) \
|
|
|
|
PP(__aeabi_fadd) \
|
|
|
|
PP(__aeabi_fcmpeq) \
|
|
|
|
PP(__aeabi_fcmpge) \
|
|
|
|
PP(__aeabi_fcmpgt) \
|
|
|
|
PP(__aeabi_fcmple) \
|
|
|
|
PP(__aeabi_fcmplt) \
|
|
|
|
PP(__aeabi_fcmpun) \
|
|
|
|
PP(__aeabi_fdiv) \
|
|
|
|
PP(__aeabi_fmul) \
|
|
|
|
PP(__aeabi_fsub) \
|
|
|
|
PP(__aeabi_i2d) \
|
|
|
|
PP(__aeabi_i2f) \
|
|
|
|
PP(__aeabi_idiv) \
|
|
|
|
PP(__aeabi_idivmod) \
|
|
|
|
PP(__aeabi_l2d) \
|
|
|
|
PP(__aeabi_l2f) \
|
|
|
|
PP(__aeabi_lasr) \
|
|
|
|
PP(__aeabi_ldivmod) \
|
|
|
|
PP(__aeabi_llsl) \
|
|
|
|
PP(__aeabi_llsr) \
|
|
|
|
PP(__aeabi_lmul) \
|
|
|
|
PP(__aeabi_ui2d) \
|
|
|
|
PP(__aeabi_ui2f) \
|
|
|
|
PP(__aeabi_uidiv) \
|
|
|
|
PP(__aeabi_uidivmod) \
|
|
|
|
PP(__aeabi_ul2d) \
|
|
|
|
PP(__aeabi_ul2f) \
|
|
|
|
PP(__aeabi_uldivmod)
|
|
|
|
|
|
|
|
// Declare statically linked math functions on ARM. The function declarations
|
|
|
|
// here do not have the correct prototypes for each function in
|
|
|
|
// ARM_MATH_IMPORTS, but it doesn't matter because only the symbol addresses are
|
|
|
|
// needed. In particular the __aeabi_*divmod functions do not have calling
|
|
|
|
// conventions which match any C prototype.
|
|
|
|
#define ARM_MATH_DECL(name) extern "C" void name();
|
|
|
|
ARM_MATH_IMPORTS(ARM_MATH_DECL)
|
|
|
|
#undef ARM_MATH_DECL
|
|
|
|
#endif
|
|
|
|
|
2014-12-31 02:22:06 +08:00
|
|
|
#if defined(__linux__) && defined(__GLIBC__) && \
|
|
|
|
(defined(__i386__) || defined(__x86_64__))
|
2014-12-31 06:52:33 +08:00
|
|
|
extern "C" LLVM_ATTRIBUTE_WEAK void __morestack();
|
2014-12-31 02:06:52 +08:00
|
|
|
#endif
|
|
|
|
|
2014-10-01 12:11:13 +08:00
|
|
|
uint64_t
|
|
|
|
RTDyldMemoryManager::getSymbolAddressInProcess(const std::string &Name) {
|
2013-10-01 09:47:35 +08:00
|
|
|
// This implementation assumes that the host program is the target.
|
|
|
|
// Clients generating code for a remote target should implement their own
|
|
|
|
// memory manager.
|
2013-11-16 01:59:43 +08:00
|
|
|
#if defined(__linux__) && defined(__GLIBC__)
|
2013-05-22 04:24:07 +08:00
|
|
|
//===--------------------------------------------------------------------===//
|
|
|
|
// Function stubs that are invoked instead of certain library calls
|
|
|
|
//
|
|
|
|
// Force the following functions to be linked in to anything that uses the
|
|
|
|
// JIT. This is a hack designed to work around the all-too-clever Glibc
|
|
|
|
// strategy of making these functions work differently when inlined vs. when
|
|
|
|
// not inlined, and hiding their real definitions in a separate archive file
|
|
|
|
// that the dynamic linker can't see. For more info, search for
|
|
|
|
// 'libc_nonshared.a' on Google, or read http://llvm.org/PR274.
|
2013-10-01 09:47:35 +08:00
|
|
|
if (Name == "stat") return (uint64_t)&stat;
|
|
|
|
if (Name == "fstat") return (uint64_t)&fstat;
|
|
|
|
if (Name == "lstat") return (uint64_t)&lstat;
|
|
|
|
if (Name == "stat64") return (uint64_t)&stat64;
|
|
|
|
if (Name == "fstat64") return (uint64_t)&fstat64;
|
|
|
|
if (Name == "lstat64") return (uint64_t)&lstat64;
|
|
|
|
if (Name == "atexit") return (uint64_t)&atexit;
|
|
|
|
if (Name == "mknod") return (uint64_t)&mknod;
|
2014-12-31 02:06:52 +08:00
|
|
|
|
2014-12-31 02:22:06 +08:00
|
|
|
#if defined(__i386__) || defined(__x86_64__)
|
2014-12-31 02:06:52 +08:00
|
|
|
// __morestack lives in libgcc, a static library.
|
2014-12-31 06:52:33 +08:00
|
|
|
if (&__morestack && Name == "__morestack")
|
|
|
|
return (uint64_t)&__morestack;
|
2014-12-31 02:22:06 +08:00
|
|
|
#endif
|
2013-11-16 01:59:43 +08:00
|
|
|
#endif // __linux__ && __GLIBC__
|
2018-07-31 03:41:25 +08:00
|
|
|
|
2013-11-16 01:59:43 +08:00
|
|
|
// See ARM_MATH_IMPORTS definition for explanation
|
|
|
|
#if defined(__BIONIC__) && defined(__arm__)
|
|
|
|
if (Name.compare(0, 8, "__aeabi_") == 0) {
|
|
|
|
// Check if the user has requested any of the functions listed in
|
|
|
|
// ARM_MATH_IMPORTS, and if so redirect to the statically linked symbol.
|
|
|
|
#define ARM_MATH_CHECK(fn) if (Name == #fn) return (uint64_t)&fn;
|
|
|
|
ARM_MATH_IMPORTS(ARM_MATH_CHECK)
|
|
|
|
#undef ARM_MATH_CHECK
|
|
|
|
}
|
|
|
|
#endif
|
2013-05-22 04:24:07 +08:00
|
|
|
|
|
|
|
// We should not invoke parent's ctors/dtors from generated main()!
|
|
|
|
// On Mingw and Cygwin, the symbol __main is resolved to
|
|
|
|
// callee's(eg. tools/lli) one, to invoke wrong duplicated ctors
|
|
|
|
// (and register wrong callee's dtors with atexit(3)).
|
|
|
|
// We expect ExecutionEngine::runStaticConstructorsDestructors()
|
|
|
|
// is called before ExecutionEngine::runFunctionAsMain() is called.
|
2013-10-01 09:47:35 +08:00
|
|
|
if (Name == "__main") return (uint64_t)&jit_noop;
|
2013-05-22 04:24:07 +08:00
|
|
|
|
|
|
|
const char *NameStr = Name.c_str();
|
|
|
|
|
2016-03-04 05:23:15 +08:00
|
|
|
// DynamicLibrary::SearchForAddresOfSymbol expects an unmangled 'C' symbol
|
2016-08-18 09:33:28 +08:00
|
|
|
// name so ff we're on Darwin, strip the leading '_' off.
|
|
|
|
#ifdef __APPLE__
|
[MCJIT] Make RTDyldMemoryManager::getSymbolAddress's behaviour more consistent.
This patch modifies RTDyldMemoryManager::getSymbolAddress(Name)'s behavior to
make it consistent with how clients are using it: Name should be mangled, and
getSymbolAddress should demangle it on the caller's behalf before looking the
name up in the process. This patch also fixes the one client
(MCJIT::getPointerToFunction) that had been passing unmangled names (by having
it pass mangled names instead).
Background:
RTDyldMemoryManager::getSymbolAddress(Name) has always used a re-try mechanism
when looking up symbol names in the current process. Prior to this patch
getSymbolAddress first tried to look up 'Name' exactly as the user passed it in
and then, if that failed, tried to demangle 'Name' and re-try the look up. The
implication of this behavior is that getSymbolAddress expected to be called with
unmangled names, and that handling mangled names was a fallback for convenience.
This is inconsistent with how clients (particularly the RuntimeDyldImpl
subclasses, but also MCJIT) usually use this API. Most clients pass in mangled
names, and succeed only because of the fallback case. For clients passing in
mangled names, getSymbolAddress's old behavior was actually dangerous, as it
could cause unmangled names in the process to shadow mangled names being looked
up.
For example, consider:
foo.c:
int _x = 7;
int x() { return _x; }
foo.o:
000000000000000c D __x
0000000000000000 T _x
If foo.c becomes part of the process (E.g. via dlopen("libfoo.dylib")) it will
add symbols 'x' (the function) and '_x' (the variable) to the process. However
jit clients looking for the function 'x' will be using the mangled function name
'_x' (note how function 'x' appears in foo.o). When getSymbolAddress goes
looking for '_x' it will find the variable instead, and return its address and
in place of the function, leading to JIT'd code calling the variable and
crashing (if we're lucky).
By requiring that getSymbolAddress be called with mangled names, and demangling
only when we're about to do a lookup in the process, the new behavior
implemented in this patch should eliminate any chance of names being shadowed
during lookup.
There's no good way to test this at the moment: This issue only arrises when
looking up process symbols (not JIT'd symbols). Any test case would have to
generate a platform-appropriate dylib to pass to llvm-rtdyld, and I'm not
aware of any in-tree tool for doing this in a portable way.
llvm-svn: 218187
2014-09-21 01:44:56 +08:00
|
|
|
if (NameStr[0] == '_')
|
2016-03-04 05:23:15 +08:00
|
|
|
++NameStr;
|
|
|
|
#endif
|
[MCJIT] Make RTDyldMemoryManager::getSymbolAddress's behaviour more consistent.
This patch modifies RTDyldMemoryManager::getSymbolAddress(Name)'s behavior to
make it consistent with how clients are using it: Name should be mangled, and
getSymbolAddress should demangle it on the caller's behalf before looking the
name up in the process. This patch also fixes the one client
(MCJIT::getPointerToFunction) that had been passing unmangled names (by having
it pass mangled names instead).
Background:
RTDyldMemoryManager::getSymbolAddress(Name) has always used a re-try mechanism
when looking up symbol names in the current process. Prior to this patch
getSymbolAddress first tried to look up 'Name' exactly as the user passed it in
and then, if that failed, tried to demangle 'Name' and re-try the look up. The
implication of this behavior is that getSymbolAddress expected to be called with
unmangled names, and that handling mangled names was a fallback for convenience.
This is inconsistent with how clients (particularly the RuntimeDyldImpl
subclasses, but also MCJIT) usually use this API. Most clients pass in mangled
names, and succeed only because of the fallback case. For clients passing in
mangled names, getSymbolAddress's old behavior was actually dangerous, as it
could cause unmangled names in the process to shadow mangled names being looked
up.
For example, consider:
foo.c:
int _x = 7;
int x() { return _x; }
foo.o:
000000000000000c D __x
0000000000000000 T _x
If foo.c becomes part of the process (E.g. via dlopen("libfoo.dylib")) it will
add symbols 'x' (the function) and '_x' (the variable) to the process. However
jit clients looking for the function 'x' will be using the mangled function name
'_x' (note how function 'x' appears in foo.o). When getSymbolAddress goes
looking for '_x' it will find the variable instead, and return its address and
in place of the function, leading to JIT'd code calling the variable and
crashing (if we're lucky).
By requiring that getSymbolAddress be called with mangled names, and demangling
only when we're about to do a lookup in the process, the new behavior
implemented in this patch should eliminate any chance of names being shadowed
during lookup.
There's no good way to test this at the moment: This issue only arrises when
looking up process symbols (not JIT'd symbols). Any test case would have to
generate a platform-appropriate dylib to pass to llvm-rtdyld, and I'm not
aware of any in-tree tool for doing this in a portable way.
llvm-svn: 218187
2014-09-21 01:44:56 +08:00
|
|
|
|
2014-09-22 01:21:56 +08:00
|
|
|
return (uint64_t)sys::DynamicLibrary::SearchForAddressOfSymbol(NameStr);
|
2013-10-01 09:47:35 +08:00
|
|
|
}
|
2013-05-22 04:24:07 +08:00
|
|
|
|
2013-10-01 09:47:35 +08:00
|
|
|
void *RTDyldMemoryManager::getPointerToNamedFunction(const std::string &Name,
|
|
|
|
bool AbortOnFailure) {
|
|
|
|
uint64_t Addr = getSymbolAddress(Name);
|
|
|
|
|
|
|
|
if (!Addr && AbortOnFailure)
|
2013-05-22 04:24:07 +08:00
|
|
|
report_fatal_error("Program used external function '" + Name +
|
|
|
|
"' which could not be resolved!");
|
2016-03-04 05:23:15 +08:00
|
|
|
|
2013-10-01 09:47:35 +08:00
|
|
|
return (void*)Addr;
|
2013-05-22 04:24:07 +08:00
|
|
|
}
|
|
|
|
|
Add missing vtable anchors
Summary: This patch adds anchor() for MemoryBuffer, raw_fd_ostream, RTDyldMemoryManager, SectionMemoryManager, etc.
Reviewers: jlebar, eli.friedman, dblaikie
Reviewed By: dblaikie
Subscribers: mehdi_amini, mgorny, dblaikie, weimingz, llvm-commits
Differential Revision: https://reviews.llvm.org/D45244
llvm-svn: 329861
2018-04-12 07:09:20 +08:00
|
|
|
void RTDyldMemoryManager::anchor() {}
|
|
|
|
void MCJITMemoryManager::anchor() {}
|
2013-05-22 04:24:07 +08:00
|
|
|
} // namespace llvm
|