Update libunwind to LLVM 18.1.2 (#21607)

Most of the changes seem irrelevant to us given that we only use
`Uniwnd-wasm.cpp`. Changes related to Wasm are what I submitted upstream
for adding the file to `CMakeLists.txt` and fixing some guards:
llvm/llvm-project#67770
llvm/llvm-project#73196
llvm/llvm-project#78230

system/lib/libunwind/include/__libunwind_config.h

@@ -36,6 +36,9 @@
 # if defined(__linux__)
 #  define _LIBUNWIND_TARGET_LINUX 1
 # endif
+# if defined(__HAIKU__)
+#  define _LIBUNWIND_TARGET_HAIKU 1
+# endif
 # if defined(__i386__)
 #  define _LIBUNWIND_TARGET_I386
 #  define _LIBUNWIND_CONTEXT_SIZE 8
@@ -196,7 +199,7 @@
 # define _LIBUNWIND_TARGET_RISCV 1
 # define _LIBUNWIND_TARGET_VE 1
 # define _LIBUNWIND_TARGET_S390X 1
- #define _LIBUNWIND_TARGET_LOONGARCH 1
+# define _LIBUNWIND_TARGET_LOONGARCH 1
 # define _LIBUNWIND_CONTEXT_SIZE 167
 # define _LIBUNWIND_CURSOR_SIZE 204
 # define _LIBUNWIND_HIGHEST_DWARF_REGISTER 287

system/lib/libunwind/include/libunwind.h

@@ -876,6 +876,9 @@ enum {
   UNW_MIPS_F29 = 61,
   UNW_MIPS_F30 = 62,
   UNW_MIPS_F31 = 63,
+  // HI,LO have been dropped since r6, we keep them here.
+  // So, when we add DSP/MSA etc, we can use the same register indexes
+  // for r6 and pre-r6.
   UNW_MIPS_HI = 64,
   UNW_MIPS_LO = 65,
 };

system/lib/libunwind/include/mach-o/compact_unwind_encoding.h

@@ -108,7 +108,7 @@ enum {
 //    are encoded in the UNWIND_X86_EBP_FRAME_REGISTERS bits as five 3-bit entries.
 //    Each entry contains which register to restore.
 // UNWIND_X86_MODE_STACK_IMMD:
-//    A "frameless" (EBP not used as frame pointer) function with a small 
+//    A "frameless" (EBP not used as frame pointer) function with a small
 //    constant stack size.  To return, a constant (encoded in the compact
 //    unwind encoding) is added to the ESP. Then the return is done by
 //    popping the stack into the pc.
@@ -119,16 +119,16 @@ enum {
 //    UNWIND_X86_FRAMELESS_STACK_REG_PERMUTATION contains which registers were
 //    saved and their order.
 // UNWIND_X86_MODE_STACK_IND:
-//    A "frameless" (EBP not used as frame pointer) function large constant 
+//    A "frameless" (EBP not used as frame pointer) function large constant
 //    stack size.  This case is like the previous, except the stack size is too
-//    large to encode in the compact unwind encoding.  Instead it requires that 
-//    the function contains "subl $nnnnnnnn,ESP" in its prolog.  The compact 
+//    large to encode in the compact unwind encoding.  Instead it requires that
+//    the function contains "subl $nnnnnnnn,ESP" in its prolog.  The compact
 //    encoding contains the offset to the nnnnnnnn value in the function in
-//    UNWIND_X86_FRAMELESS_STACK_SIZE.  
+//    UNWIND_X86_FRAMELESS_STACK_SIZE.
 // UNWIND_X86_MODE_DWARF:
 //    No compact unwind encoding is available.  Instead the low 24-bits of the
 //    compact encoding is the offset of the DWARF FDE in the __eh_frame section.
-//    This mode is never used in object files.  It is only generated by the 
+//    This mode is never used in object files.  It is only generated by the
 //    linker in final linked images which have only DWARF unwind info for a
 //    function.
 //
@@ -233,36 +233,36 @@ enum {
 // For x86_64 there are four modes for the compact unwind encoding:
 // UNWIND_X86_64_MODE_RBP_FRAME:
 //    RBP based frame where RBP is push on stack immediately after return address,
-//    then RSP is moved to RBP. Thus, to unwind RSP is restored with the current 
-//    EPB value, then RBP is restored by popping off the stack, and the return 
+//    then RSP is moved to RBP. Thus, to unwind RSP is restored with the current
+//    EPB value, then RBP is restored by popping off the stack, and the return
 //    is done by popping the stack once more into the pc.
 //    All non-volatile registers that need to be restored must have been saved
-//    in a small range in the stack that starts RBP-8 to RBP-2040.  The offset/8 
+//    in a small range in the stack that starts RBP-8 to RBP-2040.  The offset/8
 //    is encoded in the UNWIND_X86_64_RBP_FRAME_OFFSET bits.  The registers saved
 //    are encoded in the UNWIND_X86_64_RBP_FRAME_REGISTERS bits as five 3-bit entries.
-//    Each entry contains which register to restore.  
+//    Each entry contains which register to restore.
 // UNWIND_X86_64_MODE_STACK_IMMD:
-//    A "frameless" (RBP not used as frame pointer) function with a small 
-//    constant stack size.  To return, a constant (encoded in the compact 
-//    unwind encoding) is added to the RSP. Then the return is done by 
+//    A "frameless" (RBP not used as frame pointer) function with a small
+//    constant stack size.  To return, a constant (encoded in the compact
+//    unwind encoding) is added to the RSP. Then the return is done by
 //    popping the stack into the pc.
 //    All non-volatile registers that need to be restored must have been saved
 //    on the stack immediately after the return address.  The stack_size/8 is
 //    encoded in the UNWIND_X86_64_FRAMELESS_STACK_SIZE (max stack size is 2048).
 //    The number of registers saved is encoded in UNWIND_X86_64_FRAMELESS_STACK_REG_COUNT.
 //    UNWIND_X86_64_FRAMELESS_STACK_REG_PERMUTATION contains which registers were
-//    saved and their order.  
+//    saved and their order.
 // UNWIND_X86_64_MODE_STACK_IND:
-//    A "frameless" (RBP not used as frame pointer) function large constant 
+//    A "frameless" (RBP not used as frame pointer) function large constant
 //    stack size.  This case is like the previous, except the stack size is too
-//    large to encode in the compact unwind encoding.  Instead it requires that 
-//    the function contains "subq $nnnnnnnn,RSP" in its prolog.  The compact 
+//    large to encode in the compact unwind encoding.  Instead it requires that
+//    the function contains "subq $nnnnnnnn,RSP" in its prolog.  The compact
 //    encoding contains the offset to the nnnnnnnn value in the function in
-//    UNWIND_X86_64_FRAMELESS_STACK_SIZE.  
+//    UNWIND_X86_64_FRAMELESS_STACK_SIZE.
 // UNWIND_X86_64_MODE_DWARF:
 //    No compact unwind encoding is available.  Instead the low 24-bits of the
 //    compact encoding is the offset of the DWARF FDE in the __eh_frame section.
-//    This mode is never used in object files.  It is only generated by the 
+//    This mode is never used in object files.  It is only generated by the
 //    linker in final linked images which have only DWARF unwind info for a
 //    function.
 //
@@ -307,20 +307,20 @@ enum {
 //    This is a standard arm64 prolog where FP/LR are immediately pushed on the
 //    stack, then SP is copied to FP. If there are any non-volatile registers
 //    saved, then are copied into the stack frame in pairs in a contiguous
-//    range right below the saved FP/LR pair.  Any subset of the five X pairs 
+//    range right below the saved FP/LR pair.  Any subset of the five X pairs
 //    and four D pairs can be saved, but the memory layout must be in register
-//    number order.  
+//    number order.
 // UNWIND_ARM64_MODE_FRAMELESS:
-//    A "frameless" leaf function, where FP/LR are not saved. The return address 
+//    A "frameless" leaf function, where FP/LR are not saved. The return address
 //    remains in LR throughout the function. If any non-volatile registers
 //    are saved, they must be pushed onto the stack before any stack space is
 //    allocated for local variables.  The stack sized (including any saved
-//    non-volatile registers) divided by 16 is encoded in the bits 
+//    non-volatile registers) divided by 16 is encoded in the bits
 //    UNWIND_ARM64_FRAMELESS_STACK_SIZE_MASK.
 // UNWIND_ARM64_MODE_DWARF:
 //    No compact unwind encoding is available.  Instead the low 24-bits of the
 //    compact encoding is the offset of the DWARF FDE in the __eh_frame section.
-//    This mode is never used in object files.  It is only generated by the 
+//    This mode is never used in object files.  It is only generated by the
 //    linker in final linked images which have only DWARF unwind info for a
 //    function.
 //
@@ -337,19 +337,19 @@ enum {
 
 //
 // A compiler can generated compact unwind information for a function by adding
-// a "row" to the __LD,__compact_unwind section.  This section has the 
-// S_ATTR_DEBUG bit set, so the section will be ignored by older linkers. 
-// It is removed by the new linker, so never ends up in final executables. 
-// This section is a table, initially with one row per function (that needs 
+// a "row" to the __LD,__compact_unwind section.  This section has the
+// S_ATTR_DEBUG bit set, so the section will be ignored by older linkers.
+// It is removed by the new linker, so never ends up in final executables.
+// This section is a table, initially with one row per function (that needs
 // unwind info).  The table columns and some conceptual entries are:
 //
 //     range-start               pointer to start of function/range
-//     range-length              
-//     compact-unwind-encoding   32-bit encoding  
+//     range-length
+//     compact-unwind-encoding   32-bit encoding
 //     personality-function      or zero if no personality function
 //     lsda                      or zero if no LSDA data
 //
-// The length and encoding fields are 32-bits.  The other are all pointer sized. 
+// The length and encoding fields are 32-bits.  The other are all pointer sized.
 //
 // In x86_64 assembly, these entry would look like:
 //
@@ -372,23 +372,23 @@ enum {
 //     .quad    except_tab1
 //
 //
-// Notes: There is no need for any labels in the the __compact_unwind section.  
-//        The use of the .set directive is to force the evaluation of the 
+// Notes: There is no need for any labels in the __compact_unwind section.
+//        The use of the .set directive is to force the evaluation of the
 //        range-length at assembly time, instead of generating relocations.
 //
-// To support future compiler optimizations where which non-volatile registers 
+// To support future compiler optimizations where which non-volatile registers
 // are saved changes within a function (e.g. delay saving non-volatiles until
 // necessary), there can by multiple lines in the __compact_unwind table for one
-// function, each with a different (non-overlapping) range and each with 
-// different compact unwind encodings that correspond to the non-volatiles 
+// function, each with a different (non-overlapping) range and each with
+// different compact unwind encodings that correspond to the non-volatiles
 // saved at that range of the function.
 //
 // If a particular function is so wacky that there is no compact unwind way
-// to encode it, then the compiler can emit traditional DWARF unwind info.  
+// to encode it, then the compiler can emit traditional DWARF unwind info.
 // The runtime will use which ever is available.
 //
-// Runtime support for compact unwind encodings are only available on 10.6 
-// and later.  So, the compiler should not generate it when targeting pre-10.6. 
+// Runtime support for compact unwind encodings are only available on 10.6
+// and later.  So, the compiler should not generate it when targeting pre-10.6.
 
 
 
@@ -402,7 +402,7 @@ enum {
 //
 // The __TEXT,__unwind_info section is laid out for an efficient two level lookup.
 // The header of the section contains a coarse index that maps function address
-// to the page (4096 byte block) containing the unwind info for that function.  
+// to the page (4096 byte block) containing the unwind info for that function.
 //
 
 #define UNWIND_SECTION_VERSION 1

system/lib/libunwind/src/AddressSpace.hpp

@@ -414,8 +414,8 @@ static bool checkForUnwindInfoSegment(const Elf_Phdr *phdr, size_t image_base,
     cbdata->sects->dwarf_index_section = eh_frame_hdr_start;
     cbdata->sects->dwarf_index_section_length = phdr->p_memsz;
     if (EHHeaderParser<LocalAddressSpace>::decodeEHHdr(
-            *cbdata->addressSpace, eh_frame_hdr_start, phdr->p_memsz,
-            hdrInfo)) {
+            *cbdata->addressSpace, eh_frame_hdr_start,
+            eh_frame_hdr_start + phdr->p_memsz, hdrInfo)) {
       // .eh_frame_hdr records the start of .eh_frame, but not its size.
       // Rely on a zero terminator to find the end of the section.
       cbdata->sects->dwarf_section = hdrInfo.eh_frame_ptr;
@@ -638,7 +638,8 @@ inline bool LocalAddressSpace::findUnwindSections(pint_t targetAddr,
     info.dwarf_index_section_length = SIZE_MAX;
     EHHeaderParser<LocalAddressSpace>::EHHeaderInfo hdrInfo;
     if (!EHHeaderParser<LocalAddressSpace>::decodeEHHdr(
-            *this, info.dwarf_index_section, info.dwarf_index_section_length,
+            *this, info.dwarf_index_section,
+            info.dwarf_index_section + info.dwarf_index_section_length,
             hdrInfo)) {
       return false;
     }

system/lib/libunwind/src/DwarfInstructions.hpp

@@ -68,7 +68,7 @@ class DwarfInstructions {
       return (pint_t)((sint_t)registers.getRegister((int)prolog.cfaRegister) +
              prolog.cfaRegisterOffset);
     if (prolog.cfaExpression != 0)
-      return evaluateExpression((pint_t)prolog.cfaExpression, addressSpace, 
+      return evaluateExpression((pint_t)prolog.cfaExpression, addressSpace,
                                 registers, 0);
     assert(0 && "getCFA(): unknown location");
     __builtin_unreachable();

system/lib/libunwind/src/EHHeaderParser.hpp

@@ -55,6 +55,19 @@ template <typename A>
 bool EHHeaderParser<A>::decodeEHHdr(A &addressSpace, pint_t ehHdrStart,
                                     pint_t ehHdrEnd, EHHeaderInfo &ehHdrInfo) {
   pint_t p = ehHdrStart;
+
+  // Ensure that we don't read data beyond the end of .eh_frame_hdr
+  if (ehHdrEnd - ehHdrStart < 4) {
+    // Don't print a message for an empty .eh_frame_hdr (this can happen if
+    // the linker script defines symbols for it even in the empty case).
+    if (ehHdrEnd == ehHdrStart)
+      return false;
+    _LIBUNWIND_LOG("unsupported .eh_frame_hdr at %" PRIx64
+                   ": need at least 4 bytes of data but only got %zd",
+                   static_cast<uint64_t>(ehHdrStart),
+                   static_cast<size_t>(ehHdrEnd - ehHdrStart));
+    return false;
+  }
   uint8_t version = addressSpace.get8(p++);
   if (version != 1) {
     _LIBUNWIND_LOG("unsupported .eh_frame_hdr version: %" PRIu8 " at %" PRIx64,

system/lib/libunwind/src/FrameHeaderCache.hpp

@@ -31,8 +31,8 @@
 
 class _LIBUNWIND_HIDDEN FrameHeaderCache {
   struct CacheEntry {
-    uintptr_t LowPC() { return Info.dso_base; };
-    uintptr_t HighPC() { return Info.dso_base + Info.text_segment_length; };
+    uintptr_t LowPC() { return Info.dso_base; }
+    uintptr_t HighPC() { return Info.dso_base + Info.text_segment_length; }
     UnwindInfoSections Info;
     CacheEntry *Next;
   };
@@ -41,7 +41,7 @@ class _LIBUNWIND_HIDDEN FrameHeaderCache {
 
   // Can't depend on the C++ standard library in libunwind, so use an array to
   // allocate the entries, and two linked lists for ordering unused and recently
-  // used entries.  FIXME: Would the the extra memory for a doubly-linked list
+  // used entries.  FIXME: Would the extra memory for a doubly-linked list
   // be better than the runtime cost of traversing a very short singly-linked
   // list on a cache miss? The entries themselves are all small and consecutive,
   // so unlikely to cause page faults when following the pointers. The memory

system/lib/libunwind/src/Registers.hpp

@@ -619,6 +619,8 @@ class _LIBUNWIND_HIDDEN Registers_ppc {
   void      setIP(uint32_t value) { _registers.__srr0 = value; }
   uint64_t  getCR() const         { return _registers.__cr; }
   void      setCR(uint32_t value) { _registers.__cr = value; }
+  uint64_t  getLR() const         { return _registers.__lr; }
+  void      setLR(uint32_t value) { _registers.__lr = value; }
 
 private:
   struct ppc_thread_state_t {
@@ -1189,6 +1191,8 @@ class _LIBUNWIND_HIDDEN Registers_ppc64 {
   void      setIP(uint64_t value) { _registers.__srr0 = value; }
   uint64_t  getCR() const         { return _registers.__cr; }
   void      setCR(uint64_t value) { _registers.__cr = value; }
+  uint64_t  getLR() const         { return _registers.__lr; }
+  void      setLR(uint64_t value) { _registers.__lr = value; }
 
 private:
   struct ppc64_thread_state_t {
@@ -2869,7 +2873,7 @@ inline bool Registers_mips_o32::validRegister(int regNum) const {
     return false;
   if (regNum <= UNW_MIPS_R31)
     return true;
-#if __mips_isa_rev != 6
+#if __mips_isa_rev < 6
   if (regNum == UNW_MIPS_HI)
     return true;
   if (regNum == UNW_MIPS_LO)
@@ -2903,10 +2907,12 @@ inline uint32_t Registers_mips_o32::getRegister(int regNum) const {
     return _registers.__pc;
   case UNW_REG_SP:
     return _registers.__r[29];
+#if __mips_isa_rev < 6
   case UNW_MIPS_HI:
     return _registers.__hi;
   case UNW_MIPS_LO:
     return _registers.__lo;
+#endif
   }
   _LIBUNWIND_ABORT("unsupported mips_o32 register");
 }
@@ -2936,11 +2942,13 @@ inline void Registers_mips_o32::setRegister(int regNum, uint32_t value) {
   case UNW_REG_SP:
     _registers.__r[29] = value;
     return;
+#if __mips_isa_rev < 6
   case UNW_MIPS_HI:
     _registers.__hi = value;
     return;
   case UNW_MIPS_LO:
     _registers.__lo = value;
+#endif
     return;
   }
   _LIBUNWIND_ABORT("unsupported mips_o32 register");
@@ -3120,10 +3128,12 @@ inline const char *Registers_mips_o32::getRegisterName(int regNum) {
     return "$f30";
   case UNW_MIPS_F31:
     return "$f31";
+#if __mips_isa_rev < 6
   case UNW_MIPS_HI:
     return "$hi";
   case UNW_MIPS_LO:
     return "$lo";
+#endif
   default:
     return "unknown register";
   }
@@ -3193,7 +3203,7 @@ inline bool Registers_mips_newabi::validRegister(int regNum) const {
     return false;
   if (regNum <= UNW_MIPS_R31)
     return true;
-#if __mips_isa_rev != 6
+#if __mips_isa_rev < 6
   if (regNum == UNW_MIPS_HI)
     return true;
   if (regNum == UNW_MIPS_LO)
@@ -3212,10 +3222,12 @@ inline uint64_t Registers_mips_newabi::getRegister(int regNum) const {
     return _registers.__pc;
   case UNW_REG_SP:
     return _registers.__r[29];
+#if __mips_isa_rev < 6
   case UNW_MIPS_HI:
     return _registers.__hi;
   case UNW_MIPS_LO:
     return _registers.__lo;
+#endif
   }
   _LIBUNWIND_ABORT("unsupported mips_newabi register");
 }
@@ -3233,12 +3245,14 @@ inline void Registers_mips_newabi::setRegister(int regNum, uint64_t value) {
   case UNW_REG_SP:
     _registers.__r[29] = value;
     return;
+#if __mips_isa_rev < 6
   case UNW_MIPS_HI:
     _registers.__hi = value;
     return;
   case UNW_MIPS_LO:
     _registers.__lo = value;
     return;
+#endif
   }
   _LIBUNWIND_ABORT("unsupported mips_newabi register");
 }
@@ -3417,10 +3431,12 @@ inline const char *Registers_mips_newabi::getRegisterName(int regNum) {
     return "$f30";
   case UNW_MIPS_F31:
     return "$f31";
+#if __mips_isa_rev < 6
   case UNW_MIPS_HI:
     return "$hi";
   case UNW_MIPS_LO:
     return "$lo";
+#endif
   default:
     return "unknown register";
   }