Re: RFA: infrun.c, breakpoint.c: Kludge for Solaris x86 hardware watchpoint support

[Michael Snyder <msnyder at cygnus dot com>]

Peter.Schauer wrote:
> 
> Unfortunately I'd need the following kludge to work around a Solaris x86
> kernel problem with hardware watchpoint support.
> See the comment in the patches for a description of the problem.

OK, committed (although doesn't this sort of defeat the advantage
of a hardware watchpoint?)

BTW, next time could you also include the ChangeLog entry in the diffs?

				Thanks,
				Michael

> 2000-03-12  Peter Schauer  <pes@regent.e-technik.tu-muenchen.de>
> 
>         breakpoint.c, breakpoint.h (remove_hw_watchpoints):  New function.
>         infrun.c (resume):  Remove hardware watchpoints before stepping
>         when CANNOT_STEP_HW_WATCHPOINTS is nonzero.
> 
> *** gdb/breakpoint.c.orig       Thu Feb 24 13:41:33 2000
> --- gdb/breakpoint.c    Sat Mar  4 11:42:48 2000
> ***************
> *** 1080,1085 ****
> --- 1081,1107 ----
>   }
> 
>   int
> + remove_hw_watchpoints ()
> + {
> +   register struct breakpoint *b;
> +   int val;
> +
> +   ALL_BREAKPOINTS (b)
> +   {
> +     if (b->inserted
> +       && (b->type == bp_hardware_watchpoint
> +           || b->type == bp_read_watchpoint
> +           || b->type == bp_access_watchpoint))
> +       {
> +       val = remove_breakpoint (b, mark_uninserted);
> +       if (val != 0)
> +         return val;
> +       }
> +   }
> +   return 0;
> + }
> +
> + int
>   reattach_breakpoints (pid)
>        int pid;
>   {
> *** ./gdb/breakpoint.h.orig     Thu Feb  3 05:14:27 2000
> --- ./gdb/breakpoint.h  Sat Mar  4 11:42:48 2000
> ***************
> *** 562,567 ****
> --- 562,568 ----
>   extern int insert_breakpoints PARAMS ((void));
> 
>   extern int remove_breakpoints PARAMS ((void));
> + extern int remove_hw_watchpoints PARAMS ((void));
> 
>   /* This function can be used to physically insert eventpoints from the
>      specified traced inferior process, without modifying the breakpoint
> *** gdb/infrun.c.orig   Thu Feb 24 13:41:46 2000
> --- gdb/infrun.c        Mon Mar 13 09:35:54 2000
> ***************
> *** 296,301 ****
> --- 296,308 ----
>   #define HAVE_CONTINUABLE_WATCHPOINT 1
>   #endif
> 
> + #ifndef CANNOT_STEP_HW_WATCHPOINTS
> + #define CANNOT_STEP_HW_WATCHPOINTS 0
> + #else
> + #undef  CANNOT_STEP_HW_WATCHPOINTS
> + #define CANNOT_STEP_HW_WATCHPOINTS 1
> + #endif
> +
>   /* Tables of how to react to signals; the user sets them.  */
> 
>   static unsigned char *signal_stop;
> ***************
> *** 796,801 ****
> --- 803,820 ----
>     if (step && breakpoints_inserted && breakpoint_here_p (read_pc ()))
>       step = 0;
>   #endif
> +
> +   /* Some targets (e.g. Solaris x86) have a kernel bug when stepping
> +      over an instruction that causes a page fault without triggering
> +      a hardware watchpoint. The kernel properly notices that it shouldn't
> +      stop, because the hardware watchpoint is not triggered, but it forgets
> +      the step request and continues the program normally.
> +      Work around the problem by removing hardware watchpoints if a step is
> +      requested, GDB will check for a hardware watchpoint trigger after the
> +      step anyway.  */
> +   if (CANNOT_STEP_HW_WATCHPOINTS && step && breakpoints_inserted)
> +     remove_hw_watchpoints ();
> +
> 
>     /* Normally, by the time we reach `resume', the breakpoints are either
>        removed or inserted, as appropriate.  The exception is if we're sitting
> 
> --
> Peter Schauer                   pes@regent.e-technik.tu-muenchen.de