What should I do to enable DWARF on a new backend in LLVM?

Question

I am learning LLVM and trying to write a backend for a specific target architecture.

When I compile a hello-world with clang -g, the result of llvm-dwarfdump --verify is as follows.

llvm-dwarfdump --verify

I traced the process of llvm-dwarfdump with gdb and it looks like the .debug_info section isn’t generated correctly.

I’ve had a look at the documentation relating to LLVM and DWARF. It looks like they are mostly about how to adapt DWARF for a new language on the front end.

I also googled a combination of keywords like LLVM and DWARF and didn’t find anyone with the same problem.

I grep’d the keywords under Target/X86 and Target/AArch64. It doesn’t look like there is any code related to the .debug_info section either.

I’m wondering if I need to write some additional code for a new backend to support DWARF and if so, is there a document that tells me what code I need to write? If not, where should I start to debug this problem?

Here is part of the result of llvm-objdump -s. I'm not familiar with ELF. Let me know if additional information is needed.

Contents of section .debug_info:
 0000 00000000 05000102 00000000 01000c00  ................
 0010 01000000 00000000 00020123 00000000  ...........#....
 0020 00000002 2d000000 000302a1 00033900  ....-.........9.
 0030 0000043d 00000005 00050306 01060408  ...=............
 0040 07070123 00000002 90230500 02510000  ...#.....#...Q..
 0050 00050605 0400

As a comparison, here are the results for X86_64.

Contents of section .debug_info:
 0000 51000000 05000108 00000000 01000c00  Q...............
 0010 01080000 00000000 00020125 00000008  ...........%....
 0020 00000002 2d000000 000302a1 00033900  ....-.........9.
 0030 0000043d 00000005 00050306 01060408  ...=............
 0040 07070125 00000001 56050002 50000000  ...%....V...P...
 0050 05060504 00

Here is the assembly for my target:

.Lcu_begin0:
    .long   .Ldebug_info_end0-.Ldebug_info_start0
.Ldebug_info_start0:
    .short  5
    .byte   1
    .byte   2
    .long   .debug_abbrev
    0x01    .byte   0
    .short  12
    .byte   1
    .long   .Lstr_offsets_base0
    .long   .Lline_table_start0
    .byte   2
    0x01    .long   .Lfunc_end0-.Lfunc_begin0
    .long   .Laddr_table_base0
    0x02    .long   45
    .byte   0
    .byte   3
    0x02    .byte   161
    0x00    0x03    .long   57
    0x04    .long   61
    .byte   5
    .byte   0
    0x05    .byte   3
    .byte   6
    .byte   1
    0x06    .byte   4
    .byte   8
    .byte   7
    0x07    0x01    .long   .Lfunc_end0-.Lfunc_begin0
    0x02    .byte   144
    0x23    .byte   5
    .byte   0
    .byte   2
    .long   81

    0x05    .byte   6
    .byte   5
    .byte   4
    .byte   0
.Ldebug_info_end0:
    .section    .debug_str_offsets,"",@progbits
    .long   32
    .short  5
    .short  0

And here is X86_64:

.Lcu_begin0:
    .long   .Ldebug_info_end0-.Ldebug_info_start0 # Length of Unit
.Ldebug_info_start0:
    .short  5                               # DWARF version number
    .byte   1                               # DWARF Unit Type
    .byte   8                               # Address Size (in bytes)
    .long   .debug_abbrev                   # Offset Into Abbrev. Section
    .byte   1                               # Abbrev [1] 0xc:0x49 DW_TAG_compile_unit
    .byte   0                               # DW_AT_producer
    .short  12                              # DW_AT_language
    .byte   1                               # DW_AT_name
    .long   .Lstr_offsets_base0             # DW_AT_str_offsets_base
    .long   .Lline_table_start0             # DW_AT_stmt_list
    .byte   2                               # DW_AT_comp_dir
    .byte   1                               # DW_AT_low_pc
    .long   .Lfunc_end0-.Lfunc_begin0       # DW_AT_high_pc
    .long   .Laddr_table_base0              # DW_AT_addr_base
    .byte   2                               # Abbrev [2] 0x23:0xa DW_TAG_variable
    .long   45                              # DW_AT_type
    .byte   0                               # DW_AT_decl_file
    .byte   3                               # DW_AT_decl_line
    .byte   2                               # DW_AT_location
    .byte   161
    .byte   0
    .byte   3                               # Abbrev [3] 0x2d:0xc DW_TAG_array_type
    .long   57                              # DW_AT_type
    .byte   4                               # Abbrev [4] 0x32:0x6 DW_TAG_subrange_type
    .long   61                              # DW_AT_type
    .byte   5                               # DW_AT_count
    .byte   0                               # End Of Children Mark
    .byte   5                               # Abbrev [5] 0x39:0x4 DW_TAG_base_type
    .byte   3                               # DW_AT_name
    .byte   6                               # DW_AT_encoding
    .byte   1                               # DW_AT_byte_size
    .byte   6                               # Abbrev [6] 0x3d:0x4 DW_TAG_base_type
    .byte   4                               # DW_AT_name
    .byte   8                               # DW_AT_byte_size
    .byte   7                               # DW_AT_encoding
    .byte   7                               # Abbrev [7] 0x41:0xf DW_TAG_subprogram
    .byte   1                               # DW_AT_low_pc
    .long   .Lfunc_end0-.Lfunc_begin0       # DW_AT_high_pc
    .byte   1                               # DW_AT_frame_base
    .byte   86
    .byte   5                               # DW_AT_name
    .byte   0                               # DW_AT_decl_file
    .byte   2                               # DW_AT_decl_line
    .long   80                              # DW_AT_type
                                        # DW_AT_external
    .byte   5                               # Abbrev [5] 0x50:0x4 DW_TAG_base_type
    .byte   6                               # DW_AT_name
    .byte   5                               # DW_AT_encoding
    .byte   4                               # DW_AT_byte_size
    .byte   0                               # End Of Children Mark
.Ldebug_info_end0:
    .section    .debug_str_offsets,"",@progbits
    .long   32                              # Length of String Offsets Set
    .short  5
    .short  0

Although the assembly code for my target seems odd, it looks like the start and end tags for .debug_info are emitted correctly, as is the unit length .Ldebug_info_end0-.Ldebug_info_start0. It looks like something may have gone wrong converting from assembly to binary.

Are you sure -g is the last -g option on the command line? I would also try with -O0 since "Clang debug information works best at -O0" : https://clang.llvm.org/docs/CommandGuide/clang.html#cmdoption-g — Tanner Firl, Apr 18 '23 at 03:03
@TannerFirl Yes, I'm sure. I can understand "Clang debug information works best at -O0". It is conceivable that debug information could easily be lost in the optimization pass. But I don't think this would raise a functional issue that would cause an error in the generation of .debug_info. I have tried compiling to my target architecture with `-O0` and the problem still exists, while compiling to x86_64 is fine. — , Apr 18 '23 at 03:21
oh, are you just asking for the DWARF standard? This is where DWARF is defined: https://dwarfstd.org/doc/DWARF5.pdf — Tanner Firl, Apr 18 '23 at 04:01
@TannerFirl Thanks for the answer but this will be my last solution (learning DWARF Spec and LLVM code emission) :D. What I'm wondering is if I need to do something to make DWARF work on the new backend, as it doesn't look like the other backends are doing anything special. — , Apr 18 '23 at 04:11

score 0 · Answer 1 · answered Apr 18 '23 at 03:58

I can't reproduce this error without more information. This worked for me using github's codespaces blank template:

hello.c contents:

#include <stdio.h>

int main(int argc, char *argv[]){
    printf("hi\n");
}

Commands:

clang -g -o hello hello.c
@TannerFirl ➜ /workspaces/codespaces-blank $  /usr/bin/llvm-dwarfdump --verify hello
Verifying hello:        file format ELF64-x86-64
Verifying .debug_abbrev...
Verifying .debug_info Unit Header Chain...
Verifying .debug_info references...
Verifying .debug_types Unit Header Chain...
No errors.
@TannerFirl ➜ /workspaces/codespaces-blank $ clang --version
clang version 10.0.0-4ubuntu1 
Target: x86_64-pc-linux-gnu
Thread model: posix
InstalledDir: /usr/bin

What should I do to enable DWARF on a new backend in LLVM?

1 Answers1