[Hexagon] `llvm-objdump -D` crashes in `.debug_info` on a Hexagon binary

[alexrp]

❯ cat test.c
int main(){}
❯ zig cc test.c -target hexagon-linux # or clang
❯ llvm-objdump --disassemble-all a.out
...snip...
00000000 <.debug_info>:
...snip...
   2ba94:       00 0a b6 01     01b60a00 {      r6 = memw(r19+#0x4);    r0 = memw(r0+#0x28) }
   2ba98:       00 00 8d 72     728d0000 {      r21 = and(r16,#1);      memw(r0+#0x0) = r0 }
   2ba9c:       14 00 0c 0a     0a0c0014 {      r20 = memw(r0+#0x28);   r4 = memw(r1+#0x0) }
   2baa0:       50 00 00 d5     d5000050 {      r0 = memub(r0+#0x5);    memh(r5+#0x0) = r0 }
   2baa4:       a2 00 00 0c     0c0000a2 {      r0 = memw(r0+#0x30);    r2 = memw(r18+#0x0) }
   2baa8:       0e 54 00 00     0000540e <unknown>
   2bab0:       0c b1 a6 02     02a6b10c <unknown>
 #0 0x000075d9cbdc49c0 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) /home/alexrp/Source/llvm-project/llvm/lib/Support/Unix/Signals.inc:803:3
 #1 0x000075d9cbdc21c4 llvm::sys::RunSignalHandlers() /home/alexrp/Source/llvm-project/llvm/lib/Support/Signals.cpp:105:20
 #2 0x000075d9cbdc25bd SignalHandler(int) /home/alexrp/Source/llvm-project/llvm/lib/Support/Unix/Signals.inc:415:1
 #3 0x000075d9ca245810 (/lib/x86_64-linux-gnu/libc.so.6+0x45810)
 #4 0x000075d9ca2a50bc __pthread_kill_implementation ./nptl/pthread_kill.c:44:76
 #5 0x000075d9ca2a50bc __pthread_kill_internal ./nptl/pthread_kill.c:89:10
 #6 0x000075d9ca2a50bc pthread_kill ./nptl/pthread_kill.c:100:10
 #7 0x000075d9ca2456de raise ./signal/../sysdeps/posix/raise.c:27:6
 #8 0x000075d9ca2288cd internal_signal_block_all ./stdlib/../sysdeps/unix/sysv/linux/internal-signals.h:81:3
 #9 0x000075d9ca2288cd __abort_lock_wrlock ./stdlib/abort.c:58:3
#10 0x000075d9ca2288cd abort ./stdlib/abort.c:81:3
#11 0x000075d9ca228830 __assert_perror_fail ./assert/assert-perr.c:31:1
#12 0x00006285be4846ee llvm::ArrayRef<unsigned char>::slice(unsigned long, unsigned long) const (.part.0) /home/alexrp/Source/llvm-project/llvm/include/llvm/ADT/ArrayRef.h:198:17
#13 0x00006285be484b37 llvm::SmallVectorTemplateCommon<char const*, void>::operator[](unsigned long) const (/opt/llvm-20/bin/llvm-objdump+0x1ab37)
#14 0x00006285be484b37 llvm::opt::Arg::getValue(unsigned int) const /home/alexrp/Source/llvm-project/llvm/include/llvm/Option/Arg.h:126:20
#15 0x00006285be484b37 llvm::objdump::invalidArgValue(llvm::opt::Arg const*) (.cold) /home/alexrp/Source/llvm-project/llvm/tools/llvm-objdump/llvm-objdump.cpp:3435:51
#16 0x00006285be4b13b5 disassembleObject(llvm::object::ObjectFile&, llvm::object::ObjectFile const&, (anonymous namespace)::DisassemblerTarget&, std::optional<(anonymous namespace)::DisassemblerTarget>&, llvm::objdump::SourcePrinter&, bool) /home/alexrp/Source/llvm-project/llvm/tools/llvm-objdump/llvm-objdump.cpp:2316:44
#17 0x00006285be4b31e7 llvm::objdump::SourcePrinter::~SourcePrinter() /home/alexrp/Source/llvm-project/llvm/tools/llvm-objdump/SourcePrinter.h:161:11
#18 0x00006285be4b31e7 disassembleObject(llvm::object::ObjectFile*, bool) /home/alexrp/Source/llvm-project/llvm/tools/llvm-objdump/llvm-objdump.cpp:2660:1
#19 0x00006285be4b630d dumpObject(llvm::object::ObjectFile*, llvm::object::Archive const*, llvm::object::Archive::Child const*) /home/alexrp/Source/llvm-project/llvm/tools/llvm-objdump/llvm-objdump.cpp:3321:3
#20 0x00006285be4b84c4 dumpInput(llvm::StringRef) /home/alexrp/Source/llvm-project/llvm/tools/llvm-objdump/llvm-objdump.cpp:3403:15
#21 0x00006285be4b9320 void (*std::for_each<__gnu_cxx::__normal_iterator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>*, std::vector<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>, std::allocator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>>>>, void (*)(llvm::StringRef)>(__gnu_cxx::__normal_iterator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>*, std::vector<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>, std::allocator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>>>>, __gnu_cxx::__normal_iterator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>*, std::vector<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>, std::allocator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>>>>, void (*)(llvm::StringRef)))(llvm::StringRef) /usr/include/c++/14/bits/stl_algo.h:3785:22
#22 0x00006285be4b9320 void (*llvm::for_each<std::vector<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>, std::allocator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>>>&, void (*)(llvm::StringRef)>(std::vector<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>, std::allocator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>>>&, void (*)(llvm::StringRef)))(llvm::StringRef) /home/alexrp/Source/llvm-project/llvm/include/llvm/ADT/STLExtras.h:1733:23
#23 0x00006285be4b9320 llvm_objdump_main(int, char**, llvm::ToolContext const&) /home/alexrp/Source/llvm-project/llvm/tools/llvm-objdump/llvm-objdump.cpp:3748:17
#24 0x00006285be486d18 main /home/alexrp/Source/llvm-project/build-20/tools/llvm-objdump/llvm-objdump-driver.cpp:18:1
#25 0x000075d9ca22a338 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:74:3
#26 0x000075d9ca22a3fb call_init ./csu/../csu/libc-start.c:128:20
#27 0x000075d9ca22a3fb __libc_start_main ./csu/../csu/libc-start.c:347:5
#28 0x00006285be486db5 _start (/opt/llvm-20/bin/llvm-objdump+0x1cdb5)
❯ llvm-objdump --version | head -n2
LLVM (http://llvm.org/):
  LLVM version 20.1.6

[alexrp]

cc @androm3da

[androm3da]

@quic-akaryaki can you investigate this failure?

[quic-akaryaki]

@alexrp would it be possible to attach the object file?

[alexrp]

@alexrp would it be possible to attach the object file?

https://files.alexrp.com/llvm-141740.elf

[svs-quic]

The PrintLead function in the HexagonPrettyPrinter expects that there is always >= 4 bytes available. Looks like that isn't the case while disassembling .debug_info leading to the failure we are seeing.

The following diff seems to fix the issue:

diff --git a/llvm/tools/llvm-objdump/llvm-objdump.cpp b/llvm/tools/llvm-objdump/llvm-objdump.cpp
index 85c455b2891a..325f655a20a2 100644
--- a/llvm/tools/llvm-objdump/llvm-objdump.cpp
+++ b/llvm/tools/llvm-objdump/llvm-objdump.cpp
@@ -700,14 +700,18 @@ class HexagonPrettyPrinter : public PrettyPrinter {
 public:
   void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address,
                  formatted_raw_ostream &OS) {
-    uint32_t opcode =
-      (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0];
     if (LeadingAddr)
       OS << format("%8" PRIx64 ":", Address);
     if (ShowRawInsn) {
       OS << "\t";
-      dumpBytes(Bytes.slice(0, 4), OS);
-      OS << format("\t%08" PRIx32, opcode);
+      if (Bytes.size() >= 4) {
+        uint32_t opcode =
+            (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0];
+        dumpBytes(Bytes.slice(0, 4), OS);
+        OS << format("\t%08" PRIx32, opcode);
+      } else {
+        dumpBytes(Bytes, OS);
+      }
     }
   }

If I look at the disassembly of .debug_info with the above patch it looks like this in the end:

   314d8:	b1 a6 02 00	0002a6b1 <unknown>
   314e0:	00 <unknown>

   Disassembly of section .debug_str:

[llvmbot]

@llvm/issue-subscribers-tools-llvm-objdump

Author: Alex Rønne Petersen (alexrp)

``` ❯ cat test.c int main(){} ❯ zig cc test.c -target hexagon-linux # or clang ❯ llvm-objdump --disassemble-all a.out ...snip... 00000000 <.debug_info>: ...snip... 2ba94: 00 0a b6 01 01b60a00 { r6 = memw(r19+#0x4); r0 = memw(r0+#0x28) } 2ba98: 00 00 8d 72 728d0000 { r21 = and(r16,#1); memw(r0+#0x0) = r0 } 2ba9c: 14 00 0c 0a 0a0c0014 { r20 = memw(r0+#0x28); r4 = memw(r1+#0x0) } 2baa0: 50 00 00 d5 d5000050 { r0 = memub(r0+#0x5); memh(r5+#0x0) = r0 } 2baa4: a2 00 00 0c 0c0000a2 { r0 = memw(r0+#0x30); r2 = memw(r18+#0x0) } 2baa8: 0e 54 00 00 0000540e <unknown> 2bab0: 0c b1 a6 02 02a6b10c <unknown> #0 0x000075d9cbdc49c0 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) /home/alexrp/Source/llvm-project/llvm/lib/Support/Unix/Signals.inc:803:3 #1 0x000075d9cbdc21c4 llvm::sys::RunSignalHandlers() /home/alexrp/Source/llvm-project/llvm/lib/Support/Signals.cpp:105:20 #2 0x000075d9cbdc25bd SignalHandler(int) /home/alexrp/Source/llvm-project/llvm/lib/Support/Unix/Signals.inc:415:1 #3 0x000075d9ca245810 (/lib/x86_64-linux-gnu/libc.so.6+0x45810) #4 0x000075d9ca2a50bc __pthread_kill_implementation ./nptl/pthread_kill.c:44:76 #5 0x000075d9ca2a50bc __pthread_kill_internal ./nptl/pthread_kill.c:89:10 #6 0x000075d9ca2a50bc pthread_kill ./nptl/pthread_kill.c:100:10 #7 0x000075d9ca2456de raise ./signal/../sysdeps/posix/raise.c:27:6 #8 0x000075d9ca2288cd internal_signal_block_all ./stdlib/../sysdeps/unix/sysv/linux/internal-signals.h:81:3 #9 0x000075d9ca2288cd __abort_lock_wrlock ./stdlib/abort.c:58:3 #10 0x000075d9ca2288cd abort ./stdlib/abort.c:81:3 #11 0x000075d9ca228830 __assert_perror_fail ./assert/assert-perr.c:31:1 #12 0x00006285be4846ee llvm::ArrayRef<unsigned char>::slice(unsigned long, unsigned long) const (.part.0) /home/alexrp/Source/llvm-project/llvm/include/llvm/ADT/ArrayRef.h:198:17 #13 0x00006285be484b37 llvm::SmallVectorTemplateCommon<char const*, void>::operator[](unsigned long) const (/opt/llvm-20/bin/llvm-objdump+0x1ab37) #14 0x00006285be484b37 llvm::opt::Arg::getValue(unsigned int) const /home/alexrp/Source/llvm-project/llvm/include/llvm/Option/Arg.h:126:20 #15 0x00006285be484b37 llvm::objdump::invalidArgValue(llvm::opt::Arg const*) (.cold) /home/alexrp/Source/llvm-project/llvm/tools/llvm-objdump/llvm-objdump.cpp:3435:51 #16 0x00006285be4b13b5 disassembleObject(llvm::object::ObjectFile&, llvm::object::ObjectFile const&, (anonymous namespace)::DisassemblerTarget&, std::optional<(anonymous namespace)::DisassemblerTarget>&, llvm::objdump::SourcePrinter&, bool) /home/alexrp/Source/llvm-project/llvm/tools/llvm-objdump/llvm-objdump.cpp:2316:44 #17 0x00006285be4b31e7 llvm::objdump::SourcePrinter::~SourcePrinter() /home/alexrp/Source/llvm-project/llvm/tools/llvm-objdump/SourcePrinter.h:161:11 #18 0x00006285be4b31e7 disassembleObject(llvm::object::ObjectFile*, bool) /home/alexrp/Source/llvm-project/llvm/tools/llvm-objdump/llvm-objdump.cpp:2660:1 #19 0x00006285be4b630d dumpObject(llvm::object::ObjectFile*, llvm::object::Archive const*, llvm::object::Archive::Child const*) /home/alexrp/Source/llvm-project/llvm/tools/llvm-objdump/llvm-objdump.cpp:3321:3 #20 0x00006285be4b84c4 dumpInput(llvm::StringRef) /home/alexrp/Source/llvm-project/llvm/tools/llvm-objdump/llvm-objdump.cpp:3403:15 #21 0x00006285be4b9320 void (*std::for_each<__gnu_cxx::__normal_iterator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>*, std::vector<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>, std::allocator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>>>>, void (*)(llvm::StringRef)>(__gnu_cxx::__normal_iterator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>*, std::vector<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>, std::allocator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>>>>, __gnu_cxx::__normal_iterator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>*, std::vector<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>, std::allocator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>>>>, void (*)(llvm::StringRef)))(llvm::StringRef) /usr/include/c++/14/bits/stl_algo.h:3785:22 #22 0x00006285be4b9320 void (*llvm::for_each<std::vector<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>, std::allocator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>>>&, void (*)(llvm::StringRef)>(std::vector<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>, std::allocator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>>>&, void (*)(llvm::StringRef)))(llvm::StringRef) /home/alexrp/Source/llvm-project/llvm/include/llvm/ADT/STLExtras.h:1733:23 #23 0x00006285be4b9320 llvm_objdump_main(int, char**, llvm::ToolContext const&) /home/alexrp/Source/llvm-project/llvm/tools/llvm-objdump/llvm-objdump.cpp:3748:17 #24 0x00006285be486d18 main /home/alexrp/Source/llvm-project/build-20/tools/llvm-objdump/llvm-objdump-driver.cpp:18:1 #25 0x000075d9ca22a338 __libc_start_call_main ./csu/../sysdeps/nptl/libc_start_call_main.h:74:3 #26 0x000075d9ca22a3fb call_init ./csu/../csu/libc-start.c:128:20 #27 0x000075d9ca22a3fb __libc_start_main ./csu/../csu/libc-start.c:347:5 #28 0x00006285be486db5 _start (/opt/llvm-20/bin/llvm-objdump+0x1cdb5) ❯ llvm-objdump --version | head -n2 LLVM (http://llvm.org/): LLVM version 20.1.6 ```

[androm3da]

Looks great, thanks @svs-quic ! @quic-akaryaki can you make a PR with that change? Or Sudharsan if you'd rather, that's fine too.