This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [PATCH 0/6] S390: Watchpoint enhancements and hardware breakpoints
- From: "Ulrich Weigand" <uweigand at de dot ibm dot com>
- To: palves at redhat dot com (Pedro Alves)
- Cc: arnez at linux dot vnet dot ibm dot com (Andreas Arnez), gdb-patches at sourceware dot org
- Date: Fri, 16 Sep 2016 14:43:23 +0200 (CEST)
- Subject: Re: [PATCH 0/6] S390: Watchpoint enhancements and hardware breakpoints
- Authentication-results: sourceware.org; auth=none
Pedro Alves wrote:
> On 09/15/2016 12:52 PM, Andreas Arnez wrote:
> > This series adds some enhancements to the current s390-linux native
> > support for watchpoints. It also adds hardware breakpoint support.
> >
> > While this is mostly an s390-specific series, it includes one
> > common-code patch: patch #5, "linux-nat: Add function lwp_is_stepping".
> > I'm looking for feedback on that. Comments to the s390-specific patches
> > are certainly welcome as well.
>
> I'm curious on the need to disable hw breakpoints when single-stepping.
This is more of a hardware restriction than a GDB issue. The problem is
that s390 hardware doesn't really have a separate hardware single-step
feature; to implement ptrace single-stepping, the kernel uses the s390
hardware breakpoint feature (in effect, enabling hardware breakpoints
on the full address space, which will hit on the next instruction that
is executed).
This use of the feature by kernel single-stepping can conflict with
explicit use of the same feature by GDB; that's why it is better to
not attempt such explicit use while at the same time also requesting
kernel single-stepping.
Bye,
Ulrich
--
Dr. Ulrich Weigand
GNU/Linux compilers and toolchain
Ulrich.Weigand@de.ibm.com