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

[Rodrigo Vivi]

On Wed, Jan 04, 2023 at 06:51:37PM +0100, Julia Lawall wrote:
> 
> 
> On Tue, 3 Jan 2023, Deepak R Varma wrote:
> 
> > On Wed, Dec 28, 2022 at 06:18:12AM -0500, Rodrigo Vivi wrote:
> > > On Tue, Dec 27, 2022 at 11:36:13PM +0530, Deepak R Varma wrote:
> > > > On Tue, Dec 27, 2022 at 12:13:56PM -0500, Rodrigo Vivi wrote:
> > > > > On Tue, Dec 27, 2022 at 01:30:53PM +0530, Deepak R Varma wrote:
> > > > > > Using DEFINE_SIMPLE_ATTRIBUTE macro with the debugfs_create_file()
> > > > > > function adds the overhead of introducing a proxy file operation
> > > > > > functions to wrap the original read/write inside file removal protection
> > > > > > functions. This adds significant overhead in terms of introducing and
> > > > > > managing the proxy factory file operations structure and function
> > > > > > wrapping at runtime.
> > > > > > As a replacement, a combination of DEFINE_DEBUGFS_ATTRIBUTE macro paired
> > > > > > with debugfs_create_file_unsafe() is suggested to be used instead.  The
> > > > > > DEFINE_DEBUGFS_ATTRIBUTE utilises debugfs_file_get() and
> > > > > > debugfs_file_put() wrappers to protect the original read and write
> > > > > > function calls for the debug attributes. There is no need for any
> > > > > > runtime proxy file operations to be managed by the debugfs core.
> > > > > >
> > > > > > This Change is reported by the debugfs_simple_attr.cocci Coccinelle
> > > > > > semantic patch.
> > > > >
> > > > > I just checked here with
> > > > > $ make coccicheck M=drivers/gpu/drm/i915/ MODE=context COCCI=./scripts/coccinelle/api/debugfs/debugfs_simple_attr.cocci
> > > >
> > > > Hello Rodrigo,
> > > > Thank you so much for your review and feedback on the patch proposal.
> > > >
> > > > >
> > > > > The part reported by the this script is the s/SIMPLE/DEBUGFS
> > > > > but the change to the unsafe option is not.
> > > >
> > > > If you look at the original commit of this coccinelle file, it calls out the
> > > > need for pairing debugfs_create_file_unsafe() as well. Please review this
> > > >
> > > > commitID: 5103068eaca2: ("debugfs, coccinelle: check for obsolete DEFINE_SIMPLE_ATTRIBUTE() usage")
> > >
> > > +Nicolai and Julia.
> > >
> > > It looks like coccinelle got right the
> > > - DEFINE_SIMPLE_ATTRIBUTE(dsa_fops, dsa_get, dsa_set, dsa_fmt);
> > > + DEFINE_DEBUGFS_ATTRIBUTE(dsa_fops, dsa_get, dsa_set, dsa_fmt);
> > >
> > > but it failed badly on
> > > - debugfs_create_file(name, mode, parent, data, &dsa_fops)
> > > + debugfs_create_file_unsafe(name, mode, parent, data, &dsa_fops)
> > >
> > > >
> > > > Based on my review of the code, the functions debugfs_create_file() and
> > > > debugfs_create_file_unsafe(), both internally call __debugfs_create_file().
> > > > However, they pass debugfs_full_proxy_file_operations and
> > > > debugfs_open_proxy_file_operations respectively to it. The former represents the
> > > > full proxy factory, where as the later one is lightweight open proxy
> > > > implementation of the file operations structure.
> > > >
> > > > >
> > > > > This commit message is not explaining why the unsafe is the suggested
> > > > > or who suggested it.
> > > >
> > > > If you find the response above accurate, I will include these details about
> > > > the _unsafe() function in my commit message in v2.
> > > >
> > > > >
> > > > > If you remove the unsafe part feel free to resend adding:
> > > >
> > > > Please confirm you still believe switching to _unsafe() is not necessary.
> > >
> > > Based on the coccinelle commit it looks like you are right, but cocinelle
> > > just failed to detect the case. Let's see what Nicolai and Julia respond
> > > before we move with any patch here.
> >
> > Hello Nicolai and Julia,
> > Can you please review this proposed patch and the feedback comments from Rodrigo
> > please?
> 
> I'm not an expert on this issue.  If the semantic patch needs to change in
> some way, I would be happy to take any improvements.

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);

But looking to the .cocci script or at least to its description,
I believe it should catch both cases.

But if it is not a bug in the cocci script, then I'd like to hear
from Nicolai why. And have this documented in the script.

Thanks,
Rodrigo.

> 
> julia
> 
> 
> >
> > Thank you,
> > ./drv
> >
> > >
> > > >
> > > > >
> > > > > Reviewed-by: Rodrigo Vivi <rodrigo.vivi@xxxxxxxxx>
> > > > > (to both patches, this and the drrs one.
> > > > >
> > > > > Also, it looks like you could contribute with other 2 patches:
> > > > > drivers/gpu/drm/i915/pxp/intel_pxp_debugfs.c:64:0-23: WARNING: pxp_terminate_fops should be defined with DEFINE_DEBUGFS_ATTRIBUTE
> > > > > drivers/gpu/drm/i915/gvt/debugfs.c:150:0-23: WARNING: vgpu_scan_nonprivbb_fops should be defined with DEFINE_DEBUGFS_ATTRIBUTE
> > > >
> > > > Yes, these are on my list. Was waiting for a feedback on the first submission
> > > > before I send more similar patches.
> > > >
> > > > Appreciate your time and the feedback.
> > > >
> > > >
> > > > Regards,
> > > > ./drv
> > > >
> > > > >
> > > > > >
> > > > > > Signed-off-by: Deepak R Varma <drv@xxxxxxxxx>
> > > > > > ---
> > > > > >  drivers/gpu/drm/i915/display/intel_fbc.c | 12 ++++++------
> > > > > >  1 file changed, 6 insertions(+), 6 deletions(-)
> > > > > >
> > > > > > diff --git a/drivers/gpu/drm/i915/display/intel_fbc.c b/drivers/gpu/drm/i915/display/intel_fbc.c
> > > > > > index b5ee5ea0d010..4b481e2f908b 100644
> > > > > > --- a/drivers/gpu/drm/i915/display/intel_fbc.c
> > > > > > +++ b/drivers/gpu/drm/i915/display/intel_fbc.c
> > > > > > @@ -1809,10 +1809,10 @@ static int intel_fbc_debugfs_false_color_set(void *data, u64 val)
> > > > > >  	return 0;
> > > > > >  }
> > > > > >
> > > > > > -DEFINE_SIMPLE_ATTRIBUTE(intel_fbc_debugfs_false_color_fops,
> > > > > > -			intel_fbc_debugfs_false_color_get,
> > > > > > -			intel_fbc_debugfs_false_color_set,
> > > > > > -			"%llu\n");
> > > > > > +DEFINE_DEBUGFS_ATTRIBUTE(intel_fbc_debugfs_false_color_fops,
> > > > > > +			 intel_fbc_debugfs_false_color_get,
> > > > > > +			 intel_fbc_debugfs_false_color_set,
> > > > > > +			 "%llu\n");
> > > > > >
> > > > > >  static void intel_fbc_debugfs_add(struct intel_fbc *fbc,
> > > > > >  				  struct dentry *parent)
> > > > > > @@ -1821,8 +1821,8 @@ static void intel_fbc_debugfs_add(struct intel_fbc *fbc,
> > > > > >  			    fbc, &intel_fbc_debugfs_status_fops);
> > > > > >
> > > > > >  	if (fbc->funcs->set_false_color)
> > > > > > -		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);
> > > > > >  }
> > > > > >
> > > > > >  void intel_fbc_crtc_debugfs_add(struct intel_crtc *crtc)
> > > > > > --
> > > > > > 2.34.1
> > > > > >
> > > > > >
> > > > > >
> > > >
> > > >
> > >
> >
> >
> >