[Intel-gfx] [PATCH 1/2] trace: Export anonymous tracing

[Steven Rostedt]

On Sun, 01 Mar 2020 22:22:25 +0000
Chris Wilson <chris at chris-wilson.co.uk> wrote:

> Quoting Steven Rostedt (2020-03-01 18:18:16)
> > On Sun,  1 Mar 2020 15:52:47 +0000
> > Chris Wilson <chris at chris-wilson.co.uk> wrote:
> >   
> > > To facilitate construction of per-client event ringbuffers, in
> > > particular for a per-client debug and error report log, it would be
> > > extremely useful to create an anonymous file that can be handed to
> > > userspace so that it can see its and only its events. trace already
> > > provides a means of encapsulating the trace ringbuffer into a struct
> > > file that can be opened via the tracefs, and so with a couple of minor
> > > tweaks can provide the same access via an anonymous inode.  
> > 
> > I'm curious to why we need it to be anonymous. Why not allow them to be
> > visible from the tracing directory. This could allow for easier
> > debugging. Note, the trace instances have ref counters thus they can't
> > be removed if something has a reference to it.  
> 
> Do you really want a few thousand (or even tens) i915-client-%d? That
> does not particularly seem like it adds ease-of-use, and would need to be
> restricted to the client [or root]. The intent is for the client to have
> a private channel for detailed debug/error reporting of its own calls
> into the kernel.

Fair enough,

I would still want "trace_array_create()" to take a name. If it is NULL, it
becomes anonymous, but if you want it to appear in the tracing directory,
you can add a name to it.

Again, adding kernel doc comments to the global functions is still
necessary.

-- Steve