[PATCH] drm/i915/fbc: Avoid full proxy f_ops for FBC debug attributes

Rodrigo Vivi rodrigo.vivi at intel.com
Mon Jan 9 19:06:13 UTC 2023


On Sun, Jan 08, 2023 at 01:33:41AM +0530, Deepak R Varma wrote:
> On Thu, Jan 05, 2023 at 09:13:35AM +0100, Julia Lawall wrote:
> > > Hi Julia, thanks for helping here.
> > >
> > > So, my question is why this
> > >
> > > make coccicheck M=drivers/gpu/drm/i915/ MODE=context COCCI=./scripts/coccinelle/api/debugfs/debugfs_simple_attr.cocci
> > >
> > > didn't catch this chunck:
> > >
> > > -		debugfs_create_file("i915_fbc_false_color", 0644, parent,
> > > -				    fbc, &intel_fbc_debugfs_false_color_fops);
> > > +		debugfs_create_file_unsafe("i915_fbc_false_color", 0644, parent,
> > > +					   fbc, &intel_fbc_debugfs_false_color_fops);
> > >
> > > When I run it it only catches and replaces this:
> > >
> > > - DEFINE_SIMPLE_ATTRIBUTE(dsa_fops, dsa_get, dsa_set, dsa_fmt);
> > > + DEFINE_DEBUGFS_ATTRIBUTE(dsa_fops, dsa_get, dsa_set, dsa_fmt);
> >
> > There is something strange in your question.  You have MODE=context but
> > you show the output for MODE=patch.  The rule dcf matches a call to
> > debugfs_create_file, and the context rule matching DEFINE_SIMPLE_ATTRIBUTE
> > is only activated if dcf succeeds.  So when the context rule gives a
> > report, there is always a corresponding call to debugfs_create_file in the
> > same file, it is just not highlighted.  So the request is that it should
> > be highlighted as well?
> 
> Hello Rodrigo,
> Not trying to speak for you, but I think Julia's comment appears to be the
> correct interpretation of your observation. Would you mind confirming/clarifying
> and suggest next steps for this proposal?

doh! newby coccinelle user detected! My bad, sorry!

make coccicheck M=drivers/gpu/drm/i915/ MODE=patch COCCI=./scripts/coccinelle/api/debugfs/debugfs_simple_attr.cocci

do shows everything.

So, could you please mention this line in the commit message so we don't forget that?

Also could you please provide patches for the other cases?
1 patch for each file is desirable in this case since it touches different areas.

> 
> Thank you,
> ./drv
> 
> >
> > julia
> >
> 
> 


More information about the dri-devel mailing list