[Bug libbacktrace/104463] New: Split debug info not loaded from .debug/ if .gnu_debuglink points to binary itself

[Bug libbacktrace/104463] New: Split debug info not loaded from .debug/ if .gnu_debuglink points to binary itself

[matti.niemenmaa+gccbugs at iki dot fi]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104463

            Bug ID: 104463
           Summary: Split debug info not loaded from .debug/ if
                    .gnu_debuglink points to binary itself
           Product: gcc
           Version: 11.1.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: libbacktrace
          Assignee: unassigned at gcc dot gnu.org
          Reporter: matti.niemenmaa+gccbugs at iki dot fi
                CC: ian at gcc dot gnu.org
  Target Milestone: ---

(Related to PR77631; it seems this issue has been there since the initial
implementation.)

Some example code triggering UBSan just to get a stack trace:

$ cat bug.c
#include <limits.h>
int main(void) {
    int x = INT_MAX;
    return x + 1;
}

With normal (non-split) debug info, no problem:

$ gcc -o bug bug.c -ggdb3 -fsanitize=undefined
$ UBSAN_OPTIONS=print_stacktrace=1 ./bug
bug.c:4:14: runtime error: signed integer overflow: 2147483647 + 1 cannot be
represented in type 'int'
    #0 0x55d52e60e182 in main /path/to/bug.c:4
    #1 0x7fccca3fb0b2 in __libc_start_main
(/lib/x86_64-linux-gnu/libc.so.6+0x270b2)
    #2 0x55d52e60e08d in _start (/path/to/bug+0x108d)

With debug info split into a different filename in .debug, no problem:

$ gcc -o bug bug.c -ggdb3 -fsanitize=undefined
$ ( rm -rf .debug && mkdir .debug && \
  cd .debug && \
  objcopy --only-keep-debug ../bug bug.dbg && \
  objcopy --strip-debug ../bug && \
  objcopy --add-gnu-debuglink=bug.dbg ../bug )
$ UBSAN_OPTIONS=print_stacktrace=1 ./bug
bug.c:4:14: runtime error: signed integer overflow: 2147483647 + 1 cannot be
represented in type 'int'
    #0 0x55d52e60e182 in main /path/to/bug.c:4
    #1 0x7fccca3fb0b2 in __libc_start_main
(/lib/x86_64-linux-gnu/libc.so.6+0x270b2)
    #2 0x55d52e60e08d in _start (/path/to/bug+0x108d)

But when using the same filename, we lose the file/line number info for main:

$ gcc -o bug bug.c -ggdb3 -fsanitize=undefined
$ ( rm -rf .debug && mkdir .debug && \
  cd .debug && \
  objcopy --only-keep-debug ../bug bug && \
  objcopy --strip-debug ../bug && \
  objcopy --add-gnu-debuglink=bug ../bug )
$ UBSAN_OPTIONS=print_stacktrace=1 ./bug
bug.c:4:14: runtime error: signed integer overflow: 2147483647 + 1 cannot be
represented in type 'int'
    #0 0x55cbd7c4117f in main (/path/to/bug+0x117f)
    #1 0x7f863518c0b2 in __libc_start_main
(/lib/x86_64-linux-gnu/libc.so.6+0x270b2)
    #2 0x55cbd7c4108d in _start (/path/to/bug+0x108d)

And e.g. strace can verify that it never even looks at the file in .debug/.

gdb, on the other hand, does load the debug info correctly even in this case:

$ gdb -q ./bug </dev/null
Reading symbols from ./bug...
Reading symbols from /path/to/.debug/bug...
(gdb) quit

In my "real world" case all shared libraries on the system are built like this,
for example libubsan itself:

$ ls -l /usr/lib/{.debug/,}libubsan* | cut -d/ -f2-
usr/lib/.debug/libubsan.so.1.0.0
usr/lib/libubsan.so.1 -> libubsan.so.1.0.0
usr/lib/libubsan.so.1.0.0
$ readelf -p .gnu_debuglink /usr/lib/libubsan.so.1

String dump of section '.gnu_debuglink':
  [     0]  libubsan.so.1.0.0
  [    16]  n

I'm not sure what the most appropriate fix is, but perhaps one of the following
would make sense:

- if the .gnu_debuglink points to the file itself, ignore it and only look in
.debug/
- if the .gnu_debuglink points to the file itself, prefer a .debug/ file to it
- if the .gnu_debuglink pointee is successfully loaded but contains no debug
info, continue searching in .debug/

I haven't checked what gdb itself does.

$ gcc -v
Using built-in specs.
COLLECT_GCC=gcc-11
COLLECT_LTO_WRAPPER=/usr/lib/gcc/x86_64-linux-gnu/11/lto-wrapper
OFFLOAD_TARGET_NAMES=nvptx-none:amdgcn-amdhsa
OFFLOAD_TARGET_DEFAULT=1
Target: x86_64-linux-gnu
Configured with: ../src/configure -v --with-pkgversion='Ubuntu
11.1.0-1ubuntu1~20.04' --with-bugurl=file:///usr/share/doc/gcc-11/README.Bugs
--enable-languages=c,ada,c++,go,brig,d,fortran,objc,obj-c++,m2 --prefix=/usr
--with-gcc-major-version-only --program-suffix=-11
--program-prefix=x86_64-linux-gnu- --enable-shared --enable-linker-build-id
--libexecdir=/usr/lib --without-included-gettext --enable-threads=posix
--libdir=/usr/lib --enable-nls --enable-bootstrap --enable-clocale=gnu
--enable-libstdcxx-debug --enable-libstdcxx-time=yes
--with-default-libstdcxx-abi=new --enable-gnu-unique-object
--disable-vtable-verify --enable-plugin --enable-default-pie --with-system-zlib
--enable-libphobos-checking=release --with-target-system-zlib=auto
--enable-objc-gc=auto --enable-multiarch --disable-werror --disable-cet
--with-arch-32=i686 --with-abi=m64 --with-multilib-list=m32,m64,mx32
--enable-multilib --with-tune=generic
--enable-offload-targets=nvptx-none=/build/gcc-11-2V7zgg/gcc-11-11.1.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-11-2V7zgg/gcc-11-11.1.0/debian/tmp-gcn/usr
--without-cuda-driver --enable-checking=release --build=x86_64-linux-gnu
--host=x86_64-linux-gnu --target=x86_64-linux-gnu
--with-build-config=bootstrap-lto-lean --enable-link-serialization=2
Thread model: posix
Supported LTO compression algorithms: zlib zstd
gcc version 11.1.0 (Ubuntu 11.1.0-1ubuntu1~20.04)

[Bug libbacktrace/104463] Split debug info not loaded from .debug/ if .gnu_debuglink points to binary itself

[mark at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104463

Mark Wielaard <mark at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2022-02-25

--- Comment #1 from Mark Wielaard <mark at gcc dot gnu.org> ---
Note. The term "split dwarf" is often associated with the -gsplit-dwarf option,
which uses .dwo sections (and files) to split DWARF debuginfo. This bug isn't
about that.

I can replicate the issue, but haven't fully traced why it happens.

It seems libbacktrace gets confused about where /proc/self/exe points to and
tries to open /proc/self/bug, /proc/self/.debug/bug and
/usr/lib/debug//proc/self/bug