Re: utrace comments

[Christoph Hellwig <hch@infradead.org>]

On Mon, Apr 30, 2007 at 10:45:10AM +0100, Russell King wrote:
> For the sake of avoiding too much rehash, here's Roland's reply to my
> initial forrey into utrace:
> 
>   http://marc.info/?l=linux-kernel&m=117309251916053&w=2

In that mail Roland suggests keeping the singlestep code entirely
in the arm ptrace code.  After a brief look at the arm code this
looks easily possible.  From a brief look the arm software singlestep
consist of the following pieces:

 - PTRACE_SINGLESTEP implementation.  Sets the PT_SINGLESTEP flag,
   clears TIF_SYSCALL_TRACE, sets ->exit_code in the traced code
   to the singlestepping signal and wakes the traced process up.

   This can easily be implemented by putting alsmost equivalent code
   into arch_ptrace.
 - clearing PT_SINGLESTEP and cancelling the breakpoint in ptrace_disable.

   Equivalent code can go into tracehook_disable_single_step.

 - Various places in signal.c that check PT_SINGLESTEP to set/clear
   the special singlestep breakpoint.  This can stay, it just needs
   a different place to store the singlestep flag.

Do I miss something?