Re: [PATCH 2/5] ftrace: use code patching for ftrace graph tracer

[Andrew Morton <akpm-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org>]

On Tue, 25 Nov 2008 22:52:29 -0800 Harvey Harrison <harvey.harrison-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> wrote:

> On Tue, 2008-11-25 at 21:35 -0800, Andrew Morton wrote:
> > On Wed, 26 Nov 2008 00:16:24 -0500 Steven Rostedt <rostedt-nx8X9YLhiw1AfugRpC6u6w@public.gmane.org> wrote:
> > 
> > > From: Steven Rostedt <rostedt-nx8X9YLhiw1AfugRpC6u6w@public.gmane.org>
> > > 
> > > Impact: more efficient code for ftrace graph tracer
> > > 
> > > This patch uses the dynamic patching, when available, to patch
> > > the function graph code into the kernel.
> > > 
> > > This patch will ease the way for letting both function tracing
> > > and function graph tracing run together.
> > > 
> > > ...
> > >
> > > +static int ftrace_mod_jmp(unsigned long ip,
> > > +			  int old_offset, int new_offset)
> > > +{
> > > +	unsigned char code[MCOUNT_INSN_SIZE];
> > > +
> > > +	if (probe_kernel_read(code, (void *)ip, MCOUNT_INSN_SIZE))
> > > +		return -EFAULT;
> > > +
> > > +	if (code[0] != 0xe9 || old_offset != *(int *)(&code[1]))
> > 
> > erk.  I suspect that there's a nicer way of doing this amongst our
> > forest of get_unaligned_foo() interfaces.  Harvey will know.
> > 
> 
> if (code[0] != 0xe9 || old_offset != get_unaligned((int *)(&code[1])))

urgh, OK, that didn't really improve anything except to document
something which was already rather obvious.

> > > +		return -EINVAL;
> > > +
> > > +	*(int *)(&code[1]) = new_offset;
> > 
> > Might be able to use put_unaligned_foo() here.
> > 
> 
> 	put_unaligned(new_offset, (int *)(&code[1]));
> 
> > The problem is that these functions use sizeof(*ptr) to work out what
> > to do, so a cast is still needed.  A get_unaligned32(ptr) would be
> > nice.  One which takes a void* and assumes CPU ordering.
> 
> I've been thinking similarly, I could investigate something that
> goes in with the _noalign stuff?

If it's a commonly used pattern (you'd know better than I) then sure,
it would be clean to have

u32 just_gimme_the_u32_at(void *this_address);

> I'll finish the documentation patch for the _noalign stuff and then see
> about doing the host-order bits to fit in as well.
> 
> Harvey