Re: [RFC PATCH 2/2] trace: splice support for tracing_pipe

[Frederic Weisbecker <fweisbec@gmail.com>]

On Wed, Feb 11, 2009 at 01:35:26PM +0100, Ingo Molnar wrote:
> 
> * Steven Rostedt <rostedt@goodmis.org> wrote:
> 
> > On Mon, 9 Feb 2009, Eduard - Gabriel Munteanu wrote:
> > > +
> > > +static ssize_t tracing_splice_read_pipe(struct file *filp,
> > > +					loff_t *ppos,
> > > +					struct pipe_inode_info *pipe,
> > > +					size_t len,
> > > +					unsigned int flags)
> > > +{
> > > +	struct page *pages[PIPE_BUFFERS];
> > > +	struct partial_page partial[PIPE_BUFFERS];
> > > +	struct trace_iterator *iter = filp->private_data;
> > > +	struct splice_pipe_desc spd = {
> > > +		.pages = pages,
> > > +		.partial = partial,
> > > +		.nr_pages = 0, /* This gets updated below. */
> > > +		.flags = flags,
> > > +		.ops = &tracing_pipe_buf_ops,
> > > +		.spd_release = tracing_spd_release_pipe,
> > > +	};
> > 
> > Note, this is getting a little stack heavy. It is still in bounds,
> > but I get very nervous when I see structure arrays on the stack.
> > If either the structure or the array grows large, we can be in trouble.
> 
> Good point - but note that this is how splice support is implemented in
> a number of other files - so if PIPE_BUFFERS or partial_page grows in
> size, those places will fail too.
> 
> 	Ingo

I first thought they can be declared as static, since the read_pipe functions
are not supposed to be reentrant, only one reader is allowed inside the pipe
(serialized with a global mutex).
But the mutex is released while waiting, and still, I guess it should be
reentrant one day if we plan to let the user having one trace_pipe per
cpu trace.

Would a dynamic allocation be too much overhead for that?