[Bug libc/31349] New: backtrace fails to look beyond null pointer when resolving stack (x86_64)

[Bug libc/31349] New: backtrace fails to look beyond null pointer when resolving stack (x86_64)

[daniel at mariadb dot org]

https://sourceware.org/bugzilla/show_bug.cgi?id=31349

            Bug ID: 31349
           Summary: backtrace fails to look beyond null pointer when
                    resolving stack (x86_64)
           Product: glibc
           Version: 2.38
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: libc
          Assignee: unassigned at sourceware dot org
          Reporter: daniel at mariadb dot org
                CC: drepper.fsp at gmail dot com
  Target Milestone: ---

Created attachment 15352
  --> https://sourceware.org/bugzilla/attachment.cgi?id=15352&action=edit
bt.c

Using backtrace in a signal hander, a null pointer dereference is a typical
cause.

Backtrace stops at the null pointer and fails to show information of the higher
stack that caused it.

$ gcc -g -o /tmp/bt /tmp/bt.c

$ /tmp/bt
Obtained 2 stack frames.
/tmp/bt() [0x4011a2]
/lib64/libc.so.6(+0x3e9a0) [0x7f98f2b759a0]

Expected results closer to what gdb does:

$ gdb /tmp/bt
(gdb) bt
#0  0x0000000000000000 in ?? ()
#1  0x0000000000401235 in dummy_function () at /tmp/bt.c:35
#2  0x0000000000401250 in main () at /tmp/bt.c:42

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug libc/31349] backtrace fails to look beyond null pointer when resolving stack (x86_64)

[fweimer at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=31349

Florian Weimer <fweimer at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2024-02-07
                 CC|                            |fweimer at redhat dot com
             Status|UNCONFIRMED                 |WAITING
     Ever confirmed|0                           |1

--- Comment #1 from Florian Weimer <fweimer at redhat dot com> ---
The summary does not accurately describe what is happening here, I think.

Is your concern that the frame containing dummy_function is skipped?

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug libc/31349] backtrace fails to look beyond null pointer when resolving stack (x86_64)

[daniel at mariadb dot org]

https://sourceware.org/bugzilla/show_bug.cgi?id=31349

--- Comment #2 from daniel at mariadb dot org <daniel at mariadb dot org> ---
Right, so the gdb has it intercepted before the signal hander.

The backtrace has the signal handler address and I assume jump point in libc
reflecting the signal hander invokation by the kernel(?).

But yes, I would expect the 0x00000 and dummy_function in the frames returned
by backtrace as they form part of the call stack.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug libc/31349] backtrace fails to look beyond null pointer when resolving stack (x86_64)

[schwab@linux-m68k.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=31349

--- Comment #3 from Andreas Schwab <schwab@linux-m68k.org> ---
Note that the backtrace is actually produced by _Unwind_Backtrace in libgcc, so
this needs to be reported to gcc.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug libc/31349] backtrace fails to look beyond null pointer when resolving stack (x86_64)

[schwab@linux-m68k.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=31349

--- Comment #4 from Andreas Schwab <schwab@linux-m68k.org> ---
gdb has an outside view of the situation, access to extra debug information,
and can construct artificial frames that the pure unwind info does not
describe.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug libc/31349] backtrace fails to look beyond null pointer when resolving stack (x86_64)

[fweimer at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=31349

Florian Weimer <fweimer at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |MOVED
           See Also|                            |https://gcc.gnu.org/bugzill
                   |                            |a/show_bug.cgi?id=113803
             Status|WAITING                     |RESOLVED

--- Comment #5 from Florian Weimer <fweimer at redhat dot com> ---
I was confused, the summary is accurate. I've filed a GCC bug:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113803

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug libc/31349] backtrace fails to look beyond null pointer when resolving stack (x86_64)

[schwab@linux-m68k.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=31349

--- Comment #6 from Andreas Schwab <schwab@linux-m68k.org> ---
0x7f98f2b759a0 is the signal trampoline (__restore_rt), but _Unwind_Backtrace
cannot unwind through it.

-- 
You are receiving this mail because:
You are on the CC list for the bug.