This is the mail archive of the
gdb-prs@sourceware.org
mailing list for the GDB project.
tui/2242: Race condition on run
- From: jneuheisel at gmail dot com
- To: gdb-gnats at sources dot redhat dot com
- Date: 27 Mar 2007 12:30:58 -0000
- Subject: tui/2242: Race condition on run
- Reply-to: jneuheisel at gmail dot com
>Number: 2242
>Category: tui
>Synopsis: Race condition on run
>Confidential: no
>Severity: serious
>Priority: medium
>Responsible: unassigned
>State: open
>Class: sw-bug
>Submitter-Id: net
>Arrival-Date: Tue Mar 27 13:38:01 BST 2007
>Closed-Date:
>Last-Modified:
>Originator: Joshua Neuheisel
>Release: GNU gdb 6.6.50.20070326-cvs
>Organization:
>Environment:
Linux 2.6.12-1-686-smp #1 SMP Tue Sep 27 13:10:31 JST 2005 i686 GNU/Linux
>Description:
Running gdbtui incurs a fatal race condition between ptrace calls on the parent and child.
>How-To-Repeat:
1. Create a C source file sample.c with the following line:
int main (void) { return 42; }
2. Compile the source as follows:
gcc -g -o sample sample.c
3. Run "gdbtui ./sample". The output will be:
Couldn't get registers: No such process.
>Fix:
After the fork in fork-inferior.c, the child will call ptrace(TRACEME) in "(*traceme_fun) ()". This will cause the child to sleep. The parent (gdbtui) should wait for the child to go to sleep, the do a ptrace to begin debugging. Instead, the parent calls starup_inferior which eventually calls tui_registers_changed_hook which eventually calls fetch_regs. This last function calls ptrace(GETREGS) which fails whenever the child has not yet called its ptrace.
A possible fix would be to wait for the child to call its ptrace before calling "fetch_regs", or the "tui_registers_changed_hook" could handle the exception thrown by "fetch_regs" and ignore it.
>Release-Note:
>Audit-Trail:
>Unformatted: