[Bug server/29652] New: [gdbserver] Keeping symbols between two gdbserver sessions breaks second session

[Bug server/29652] New: [gdbserver] Keeping symbols between two gdbserver sessions breaks second session

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=29652

            Bug ID: 29652
           Summary: [gdbserver] Keeping symbols between two gdbserver
                    sessions breaks second session
           Product: gdb
           Version: HEAD
            Status: NEW
          Severity: normal
          Priority: P2
         Component: server
          Assignee: unassigned at sourceware dot org
          Reporter: vries at gcc dot gnu.org
  Target Milestone: ---

Consider the following 2 consecutive gdbserver sessions, separated by "file":
...
$ gdb -q -batch \
  -ex "target remote | gdbserver - ~/a.out" \
  -ex continue \
  -ex "file" \
  -ex "target remote | gdbserver - ~/a.out" \
  -ex continue
stdin/stdout redirected
Process /home/vries/a.out created; pid = 28666
Remote debugging using stdio
Reading /home/vries/a.out from remote target...
warning: File transfers from remote targets can be slow. Use "set sysroot" to
access files locally instead.
Reading /home/vries/a.out from remote target...
Reading symbols from target:/home/vries/a.out...
Reading /lib64/ld-linux-x86-64.so.2 from remote target...
Reading /lib64/ld-linux-x86-64.so.2 from remote target...
Reading
/usr/lib/debug/.build-id/5c/abd6dddf6c61859687ee6a3c3273fe4c01c68b.debug from
remote target...
Reading /lib64/ld-2.31.so-2.31-150300.41.1.x86_64.debug from remote target...
Reading /lib64/.debug/ld-2.31.so-2.31-150300.41.1.x86_64.debug from remote
target...
Reading /usr/lib/debug//lib64/ld-2.31.so-2.31-150300.41.1.x86_64.debug from
remote target...
Reading /usr/lib/debug/lib64//ld-2.31.so-2.31-150300.41.1.x86_64.debug from
remote target...
Reading target:/usr/lib/debug/lib64//ld-2.31.so-2.31-150300.41.1.x86_64.debug
from remote target...
Reading
/usr/lib/debug/.build-id/26/7a3e00bc5c234ec8c0c4c68ca478a2844e3338.debug from
remote target...
0x00007ffff7dd4550 in _start () from target:/lib64/ld-linux-x86-64.so.2
Reading /lib64/libc.so.6 from remote target...
Reading
/usr/lib/debug/.build-id/c8/417d767baccfadb39b474e484d46947915cd8f.debug from
remote target...
Reading /lib64/libc-2.31.so-2.31-150300.41.1.x86_64.debug from remote target...
Reading /lib64/.debug/libc-2.31.so-2.31-150300.41.1.x86_64.debug from remote
target...
Reading /usr/lib/debug//lib64/libc-2.31.so-2.31-150300.41.1.x86_64.debug from
remote target...
Reading /usr/lib/debug/lib64//libc-2.31.so-2.31-150300.41.1.x86_64.debug from
remote target...
Reading target:/usr/lib/debug/lib64//libc-2.31.so-2.31-150300.41.1.x86_64.debug
from remote target...
hello
[Inferior 1 (process 28666) exited normally]
stdin/stdout redirected
Process /home/vries/a.out created; pid = 28720
Remote debugging using stdio
Reading /home/vries/a.out from remote target...
Reading /home/vries/a.out from remote target...
Reading symbols from target:/home/vries/a.out...
Reading /lib64/ld-linux-x86-64.so.2 from remote target...
Reading /lib64/ld-linux-x86-64.so.2 from remote target...
Reading
/usr/lib/debug/.build-id/5c/abd6dddf6c61859687ee6a3c3273fe4c01c68b.debug from
remote target...
Reading /lib64/ld-2.31.so-2.31-150300.41.1.x86_64.debug from remote target...
Reading /lib64/.debug/ld-2.31.so-2.31-150300.41.1.x86_64.debug from remote
target...
Reading /usr/lib/debug//lib64/ld-2.31.so-2.31-150300.41.1.x86_64.debug from
remote target...
Reading /usr/lib/debug/lib64//ld-2.31.so-2.31-150300.41.1.x86_64.debug from
remote target...
Reading target:/usr/lib/debug/lib64//ld-2.31.so-2.31-150300.41.1.x86_64.debug
from remote target...
Reading
/usr/lib/debug/.build-id/26/7a3e00bc5c234ec8c0c4c68ca478a2844e3338.debug from
remote target...
0x00007ffff7dd4550 in _start () from target:/lib64/ld-linux-x86-64.so.2
Reading /lib64/libc.so.6 from remote target...
Reading
/usr/lib/debug/.build-id/c8/417d767baccfadb39b474e484d46947915cd8f.debug from
remote target...
Reading /lib64/libc-2.31.so-2.31-150300.41.1.x86_64.debug from remote target...
Reading /lib64/.debug/libc-2.31.so-2.31-150300.41.1.x86_64.debug from remote
target...
Reading /usr/lib/debug//lib64/libc-2.31.so-2.31-150300.41.1.x86_64.debug from
remote target...
Reading /usr/lib/debug/lib64//libc-2.31.so-2.31-150300.41.1.x86_64.debug from
remote target...
Reading target:/usr/lib/debug/lib64//libc-2.31.so-2.31-150300.41.1.x86_64.debug
from remote target...
hello
[Inferior 1 (process 28720) exited normally]
$ 
...

The two sessions are fairly similar.

The 'warning: File transfers from remote targets can be slow. Use "set sysroot"
to access files locally instead' only occurs in the first session.

Also not sure why in both cases we're claiming to load a.out twice (more clear
in the second session, in the first session the warning is inbetween):
...
Reading /home/vries/a.out from remote target...
Reading /home/vries/a.out from remote target...
...

Anyway, if we remove the file command, we have instead:
...
$ gdb -q -batch \
  -ex "target remote | gdbserver - ~/a.out" \
  -ex continue \
  -ex "target remote | gdbserver - ~/a.out" \
  -ex continue
stdin/stdout redirected
Process /home/vries/a.out created; pid = 27584
Remote debugging using stdio
Reading /home/vries/a.out from remote target...
warning: File transfers from remote targets can be slow. Use "set sysroot" to
access files locally instead.
Reading /home/vries/a.out from remote target...
Reading symbols from target:/home/vries/a.out...
Reading /lib64/ld-linux-x86-64.so.2 from remote target...
Reading /lib64/ld-linux-x86-64.so.2 from remote target...
Reading
/usr/lib/debug/.build-id/5c/abd6dddf6c61859687ee6a3c3273fe4c01c68b.debug from
remote target...
Reading /lib64/ld-2.31.so-2.31-150300.41.1.x86_64.debug from remote target...
Reading /lib64/.debug/ld-2.31.so-2.31-150300.41.1.x86_64.debug from remote
target...
Reading /usr/lib/debug//lib64/ld-2.31.so-2.31-150300.41.1.x86_64.debug from
remote target...
Reading /usr/lib/debug/lib64//ld-2.31.so-2.31-150300.41.1.x86_64.debug from
remote target...
Reading target:/usr/lib/debug/lib64//ld-2.31.so-2.31-150300.41.1.x86_64.debug
from remote target...
Reading
/usr/lib/debug/.build-id/26/7a3e00bc5c234ec8c0c4c68ca478a2844e3338.debug from
remote target...
0x00007ffff7dd4550 in _start () from target:/lib64/ld-linux-x86-64.so.2
Reading /lib64/libc.so.6 from remote target...
Reading
/usr/lib/debug/.build-id/c8/417d767baccfadb39b474e484d46947915cd8f.debug from
remote target...
Reading /lib64/libc-2.31.so-2.31-150300.41.1.x86_64.debug from remote target...
Reading /lib64/.debug/libc-2.31.so-2.31-150300.41.1.x86_64.debug from remote
target...
Reading /usr/lib/debug//lib64/libc-2.31.so-2.31-150300.41.1.x86_64.debug from
remote target...
Reading /usr/lib/debug/lib64//libc-2.31.so-2.31-150300.41.1.x86_64.debug from
remote target...
Reading target:/usr/lib/debug/lib64//libc-2.31.so-2.31-150300.41.1.x86_64.debug
from remote target...
hello
[Inferior 1 (process 27584) exited normally]
`target:/home/vries/a.out' has disappeared; keeping its symbols.
stdin/stdout redirected
Process /home/vries/a.out created; pid = 27588
Remote debugging using stdio
warning: Unable to find dynamic linker breakpoint function.
GDB will be unable to debug shared library initializers
and track explicitly loaded dynamic code.
Reading
/usr/lib/debug/.build-id/26/7a3e00bc5c234ec8c0c4c68ca478a2844e3338.debug from
remote target...
0x00007ffff7dd4550 in ?? ()
Reading /lib64/libc.so.6 from remote target...
Reading /lib64/ld-linux-x86-64.so.2 from remote target...
Reading
/usr/lib/debug/.build-id/c8/417d767baccfadb39b474e484d46947915cd8f.debug from
remote target...
Reading /lib64/libc-2.31.so-2.31-150300.41.1.x86_64.debug from remote target...
Reading /lib64/.debug/libc-2.31.so-2.31-150300.41.1.x86_64.debug from remote
target...
Reading /usr/lib/debug//lib64/libc-2.31.so-2.31-150300.41.1.x86_64.debug from
remote target...
Reading /usr/lib/debug/lib64//libc-2.31.so-2.31-150300.41.1.x86_64.debug from
remote target...
Reading target:/usr/lib/debug/lib64//libc-2.31.so-2.31-150300.41.1.x86_64.debug
from remote target...
Reading
/usr/lib/debug/.build-id/5c/abd6dddf6c61859687ee6a3c3273fe4c01c68b.debug from
remote target...
Reading /lib64/ld-2.31.so-2.31-150300.41.1.x86_64.debug from remote target...
Reading /lib64/.debug/ld-2.31.so-2.31-150300.41.1.x86_64.debug from remote
target...
Reading /usr/lib/debug//lib64/ld-2.31.so-2.31-150300.41.1.x86_64.debug from
remote target...
Reading /usr/lib/debug/lib64//ld-2.31.so-2.31-150300.41.1.x86_64.debug from
remote target...
Reading target:/usr/lib/debug/lib64//ld-2.31.so-2.31-150300.41.1.x86_64.debug
from remote target...
hello
[Inferior 1 (process 27588) exited normally]
...

Now we have:
...
`target:/home/vries/a.out' has disappeared; keeping its symbols.
...
and then we run into trouble:
...
warning: Unable to find dynamic linker breakpoint function.
GDB will be unable to debug shared library initializers
and track explicitly loaded dynamic code.
Reading
/usr/lib/debug/.build-id/26/7a3e00bc5c234ec8c0c4c68ca478a2844e3338.debug from
remote target...
0x00007ffff7dd4550 in ?? ()
...
in other words, gdb doesn't realize that we're stopped at _start.

I'd expect the first and second session to look roughly the same again.

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

[Bug server/29652] [gdbserver] Keeping symbols between two gdbserver sessions breaks second session

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=29652

--- Comment #1 from Tom de Vries <vries at gcc dot gnu.org> ---
Playing around a bit more with the example, using "symbol-file" instead of
"file" doesn't fix the problem but using "exec-file" does.

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

[Bug server/29652] [gdbserver] Keeping symbols between two gdbserver sessions breaks second session

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=29652

--- Comment #2 from Tom de Vries <vries at gcc dot gnu.org> ---
This fixes it:
...
diff --git a/gdb/solib-svr4.c b/gdb/solib-svr4.c
index a6a9ec5c86b..ce3278f5e89 100644
--- a/gdb/solib-svr4.c
+++ b/gdb/solib-svr4.c
@@ -571,9 +571,11 @@ find_program_interpreter (void)
        int sect_size = bfd_section_size (interp_sect);

        gdb::byte_vector buf (sect_size);
-       bfd_get_section_contents (current_program_space->exec_bfd (),
-                                 interp_sect, buf.data (), 0, sect_size);
-       return buf;
+       bool res
+         = bfd_get_section_contents (current_program_space->exec_bfd (),
+                                     interp_sect, buf.data (), 0, sect_size);
+       if (res)
+         return buf;
       }
    }

...

It makes the function fall back to:
...
  /* If we didn't find it, use the target auxiliary vector.  */
  return read_program_header (PT_INTERP, NULL, NULL);
...
which works for this case.

But perhaps the root cause of the problem is that bfd_get_section_contents
returns false in the first place, and we need "gdb_assert (res)" instead, and
fix the assert?

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

[Bug server/29652] [gdbserver] Keeping symbols between two gdbserver sessions breaks second session

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=29652

--- Comment #3 from Tom de Vries <vries at gcc dot gnu.org> ---
(In reply to Tom de Vries from comment #2)
> But perhaps the root cause of the problem is that bfd_get_section_contents
> returns false in the first place,

Drilling down, this happens here in target_fileio_pread:
...
  else if (fh->target == NULL)
    *target_errno = FILEIO_EIO;
...
because:
...
(gdb) p *fh
$8 = {target = 0x0, target_fd = 4}
...

In contrast, the first time, when the read succeeds, we have instead:
...
(gdb) p *fh
$15 = {target = 0x16d43a0, target_fd = 4}
...

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

[Bug server/29652] [gdbserver] Keeping symbols between two gdbserver sessions breaks second session

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=29652

--- Comment #4 from Tom de Vries <vries at gcc dot gnu.org> ---
Triggered "gdb_assert (res)" using this as well:
...
$ gdb -q -batch \
  -ex "shell gdbserver 127.0.0.1:2345 ~/a.out &" \
  -ex "target remote 127.0.0.1:2345" \
  -ex continue \
  -ex "shell gdbserver 127.0.0.1:2345 ~/a.out &" \
  -ex "target remote 127.0.0.1:2345" \
  -ex continue
...
so it's not specific to using gdbserver stdin/stdout.

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

[Bug server/29652] [gdbserver] Keeping symbols between two gdbserver sessions breaks second session

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=29652

--- Comment #5 from Tom de Vries <vries at gcc dot gnu.org> ---
Could be a duplicate of PR22417.

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

[Bug server/29652] [gdbserver] Keeping symbols between two gdbserver sessions breaks second session

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=29652

--- Comment #6 from Tom de Vries <vries at gcc dot gnu.org> ---
(In reply to Tom de Vries from comment #5)
> Could be a duplicate of PR22417.

Confirmed, running into the same tsan error.

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

[Bug server/29652] [gdbserver] Keeping symbols between two gdbserver sessions breaks second session

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=29652

--- Comment #7 from Tom de Vries <vries at gcc dot gnu.org> ---
(In reply to Tom de Vries from comment #6)
> Confirmed, running into the same tsan error.

I meant asan.

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

[Bug server/29652] [gdbserver] Keeping symbols between two gdbserver sessions breaks second session

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=29652

--- Comment #8 from Tom de Vries <vries at gcc dot gnu.org> ---
(In reply to Tom de Vries from comment #3)
> Drilling down, this happens here in target_fileio_pread:
> ...
>   else if (fh->target == NULL)
>     *target_errno = FILEIO_EIO;
> ...
> because:
> ...
> (gdb) p *fh
> $8 = {target = 0x0, target_fd = 4}
> ...

The target = NULL is done by fileio_handles_invalidate_target, which is called
from target_close.

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

[Bug server/29652] [gdbserver] Keeping symbols between two gdbserver sessions breaks second session

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=29652

--- Comment #9 from Tom de Vries <vries at gcc dot gnu.org> ---
Well, this is a different way of dealing with it:
...
diff --git a/gdb/remote.c b/gdb/remote.c
index 2f6cb2d01ee..3ae6ad688b8 100644
--- a/gdb/remote.c
+++ b/gdb/remote.c
@@ -2543,14 +2543,13 @@ remote_target::remote_add_inferior (bool fake_pid_p,
int pid, 
int attached,
   inf->attach_flag = attached;
   inf->fake_pid_p = fake_pid_p;

+  current_program_space->exec_close ();
+
   /* If no main executable is currently open then attempt to
      open the file that was executed to create this inferior.  */
-  if (try_open_exec && get_exec_file (0) == NULL)
+  if (try_open_exec)
     exec_file_locate_attach (pid, 0, 1);

-  /* Check for exec file mismatch, and let the user solve it.  */
-  validate_exec_file (1);
-
   return inf;
 }

...

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

[Bug server/29652] [gdbserver] Keeping symbols between two gdbserver sessions breaks second session

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=29652

--- Comment #10 from Tom de Vries <vries at gcc dot gnu.org> ---
https://sourceware.org/pipermail/gdb-patches/2022-October/192615.html

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

[Bug server/29652] [gdbserver] Keeping symbols between two gdbserver sessions breaks second session

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=29652

Tom de Vries <vries at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
             Status|NEW                         |RESOLVED
   Target Milestone|---                         |13.1

--- Comment #11 from Tom de Vries <vries at gcc dot gnu.org> ---
https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=8e94bb3e3a478544c0d8abfad8404af015f7130b

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

[Bug server/29652] [gdbserver] Keeping symbols between two gdbserver sessions breaks second session

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=29652

Tom de Vries <vries at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |qiyao at gcc dot gnu.org

--- Comment #12 from Tom de Vries <vries at gcc dot gnu.org> ---
*** Bug 22417 has been marked as a duplicate of this bug. ***

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