RE: [Xen-devel] [PATCH] break-in to Xen 1) when gdb is invoked and 2) ^C on gdb

[Caz Yokoyama <cazyokoyama@xxxxxxxxx>]

Hello Simon,
I afraid I don't understand your question. I assume your question is why I
move the declaration of lsr from inside of while loop to outside. The reason
is that I expect lsr is allocated and de-allocated every time the body of
the while loop is executed. The speed of serial port is slow. Therefore,
there may be no difference. Also gcc may optimize allocation and
de-allocation moving toward outside of while loop.
Let me know if I haven't answered your question.
-caz

-----Original Message-----
From: Simon Horman [mailto:horms@xxxxxxxxxxxx] 
Sent: Thursday, August 20, 2009 5:24 PM
To: Caz Yokoyama
Cc: 'Keir Fraser'; xen-devel@xxxxxxxxxxxxxxxxxxx
Subject: Re: [Xen-devel] [PATCH] break-in to Xen 1) when gdb is invoked and
2) ^C on gdb

On Mon, Aug 17, 2009 at 01:56:38PM -0700, Caz Yokoyama wrote:
> Hello Keir,
> This patch makes gdb break-in to Xen when 1) it is invoked and 2) ^C is
> typed on gdb or clicking interrupt on ddd.
> diff -r 8b0f1f37e145 xen/common/keyhandler.c
> --- a/xen/common/keyhandler.c Sun Aug 16 08:46:08 2009 +0100
> +++ b/xen/common/keyhandler.c Mon Aug 17 10:07:04 2009 -0700
> @@ -397,7 +397,11 @@
>      .desc = "print all diagnostics"
>  };
>  
> +#if 1
> +void do_debug_key(unsigned char key, struct cpu_user_regs *regs)
> +#else
>  static void do_debug_key(unsigned char key, struct cpu_user_regs *regs)
> +#endif
>  {
>      printk("'%c' pressed -> trapping into debugger\n", key);
>      (void)debugger_trap_fatal(0xf001, regs);
> diff -r 8b0f1f37e145 xen/drivers/char/ns16550.c
> --- a/xen/drivers/char/ns16550.c      Sun Aug 16 08:46:08 2009 +0100
> +++ b/xen/drivers/char/ns16550.c      Mon Aug 17 10:07:04 2009 -0700
> @@ -16,6 +16,7 @@
>  #include <xen/serial.h>
>  #include <xen/iocap.h>
>  #include <asm/io.h>
> +#include <xen/keyhandler.h>
>  
>  /*
>   * Configure serial port with a string:
> @@ -125,10 +126,16 @@
>  {
>      struct serial_port *port = dev_id;
>      struct ns16550 *uart = port->uart;
> +    char lsr;
>  
>      while ( !(ns_read_reg(uart, IIR) & IIR_NOINT) )
>      {
> -        char lsr = ns_read_reg(uart, LSR);
> +        lsr = ns_read_reg(uart, LSR);
> +#if 1

Is there a reason for #if 1 here?

> +        if ( lsr & LSR_BI ) { /* on BREAK */
> +            do_debug_key('g', regs); /* g is meaningless */
> +        }
> +#endif
>          if ( lsr & LSR_THRE )
>              serial_tx_interrupt(port, regs);
>          if ( lsr & LSR_DR )
> diff -r 8b0f1f37e145 xen/include/xen/keyhandler.h
> --- a/xen/include/xen/keyhandler.h    Sun Aug 16 08:46:08 2009 +0100
> +++ b/xen/include/xen/keyhandler.h    Mon Aug 17 10:07:04 2009 -0700
> @@ -40,6 +40,10 @@
>      char *desc;
>  };
>  
> +#if 1
> +extern void do_debug_key(unsigned char key, struct cpu_user_regs *regs);
> +#endif
> +
>  /* Initialize keytable with default handlers */
>  extern void initialize_keytable(void);
> 
> Xen expects BREAK and 'g' which is the same sequence of the one on Linux
> kernel. I submitted a patch which sends BREAK and 'g' to gdb-patches on
> 08/07, but it is not approved yet.
> You can't use ssplitd/nsplitd because it does not pass BREAK. Therefore,
you
> lose console output and corresponding ^A^A^A functions. My associated
lines
> for the patch are
> menu.lst
>         kernel /boot/xen-3.5-unstable.gz com1=115200,8n1 gdb=com1 
> .gdbinit
>       set remotebreak BREAK-g
>       target remote /dev/ttyS0
>  ddd --gdb --debugger "/home/caz/lightfleet/kgdb/gdb-KgdbLight/gdb/gdb"
> xen/xen-syms
> -caz
> 


> _______________________________________________
> Xen-devel mailing list
> Xen-devel@xxxxxxxxxxxxxxxxxxx
> http://lists.xensource.com/xen-devel


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel