Linux user gdbserver does not respond to remote `Ctrl-C' interrupts

Bug #1031920 reported by Meador Inge on 2012-08-01
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
QEMU
Undecided
Unassigned

Bug Description

The bug was reproduced in a recent mainline build for ARM Linux by starting emulation with a gdbserver:

$ qemu-arm -g 1234 a.out

and then connecting from gdb:

(gdb) target remote :1234
Remote debugging using :1234
[New Remote target]
[Switching to Remote target]
0x00008ba8 in _start ()
(gdb) b main
Breakpoint 1 at 0x8cb0: file hello.c, line 5.
(gdb) cont
Continuing.

Breakpoint 1, main (argc=1, argv=0xf6fff24c) at hello.c:5
5 int n = 0;
(gdb) l
1 #include <stdio.h>
2
3 int main (int argc, char **argv)
4 {
5 int n = 0;
6
7 for (;;) {
8 printf ("Hello, World!\n");
9 sleep (5);
10 }
(gdb) cont
Continuing.
^C^CInterrupted while waiting for the program.
Give up (and stop debugging it)? (y or n) y

Notice that the `Ctrl-C' interrupts are ignored.

Meador Inge (meadori) on 2012-08-01
description: updated
thierry bultel (tbultel) wrote :

I have encountered that issue recently, and started some analysis.

The issue is due to the fact that in qemu, gdbstub no longer reads the communication channel once
the debugged process is resumed with "cont".
Another way to say that, is that communication with gdb is only possible once the process thread execution
is re-routed in the gdb handler.

I am working on a fix.

The fix will consist in having an additional thread, launched that the internal gdbserver startup,
that will be wakeup when the debugged process is resumed.
That thread will be waiting on the communication channel for the eventually incoming CTRL-C request (0x3)
I start to have promising results but it still needs some testing.

Meanwhile, (this should likely be a separate discussion thread, but is somehow related with the named issue above) ,
I also noticed that with multithreaded
processes, a breakpoint does not suspend all the threads when it is hit.
This is a little bit annoying, and does not match the expected behaviour that is typically seen on a pure native gdb debugging
session.

Once it is ready I will post a patch to qemu mailing list

Best regards
Thierry

Thomas Huth (th-huth) wrote :

Has the fix mentioned in comment #1 been included in the QEMU repository?

Changed in qemu:
status: New → Incomplete
Launchpad Janitor (janitor) wrote :

[Expired for QEMU because there has been no activity for 60 days.]

Changed in qemu:
status: Incomplete → Expired
To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Other bug subscribers