Re: [PATCH] Support for HWbreak/watchpoint accross fork/vfork on arm-native

[Pedro Alves <palves@redhat.com>]

> +static int
> +update_registers_callback (struct lwp_info *lwp, void *arg)
> +{
> +  struct update_registers_data *data = (struct update_registers_data *) arg;
> +
> +  /* Force iterate_over_lwps to return matched lwp_info*.  */
> +  if (arg == NULL)
> +    return 1;
> +

You don't seem to have addressed my comment here?  It was:

"I don't understand this.  It seems nothing passes a NULL arg."


>    if (watchpoint)
>      {
>        count = arm_linux_get_hw_watchpoint_count ();
> -      bpts = t->wpts;
> -      dir = -1;
> +      bpts = arm_linux_get_debug_reg_state (pid)->wpts;
> +      if (count > MAX_WPTS)
> +        warning (_("arm-linux-gdb supports %d hardware watchpoints but target \
> +                 supports %d"), MAX_WPTS, count);

Like in the Aarch64 port, add a 'count = MAX_WPTS' as well, otherwise
GDB would access the arrays out of bounds when that happens.

>      }
>    else
>      {
>        count = arm_linux_get_hw_breakpoint_count ();
> -      bpts = t->bpts;
> -      dir = 1;
> +      bpts = arm_linux_get_debug_reg_state (pid)->bpts;
> +      if (count > MAX_BPTS)
> +        warning (_("arm-linux-gdb supports %d hardware breakpoints but target \
> +                 supports %d"), MAX_BPTS, count);

Likewise.

>      }
>  
>    for (i = 0; i < count; ++i)
>      if (!arm_hwbp_control_is_enabled (bpts[i].control))

-- 
Pedro Alves