[cocci] question about kernel cocci and cleanup.h

From: "Brandeburg, Jesse" <jesse.brandeburg@intel.com>
To: Julia Lawall <julia.lawall@inria.fr>
Cc: "cocci@inria.fr" <cocci@inria.fr>,
	"peterz@infradead.org" <peterz@infradead.org>,
	"Kitszel, Przemyslaw" <przemyslaw.kitszel@intel.com>,
	"Brady, Alan" <alan.brady@intel.com>,
	"Keller, Jacob E" <jacob.e.keller@intel.com>
Subject: [cocci] question about kernel cocci and cleanup.h
Date: Thu, 7 Mar 2024 18:25:27 +0000	[thread overview]
Message-ID: <CO1PR11MB49149F1167679926A2917E0997202@CO1PR11MB4914.namprd11.prod.outlook.com> (raw)

[-- Attachment #1: Type: text/plain, Size: 956 bytes --]

Hi Julia and team, I had a question about whether it's possible to use coccinelle/spatch to start to identify functions which are able to use inline destructors (__free or DEFINE_FREE()) from include/linux/cleanup.h

To some extent a missing kfree logic is already part of the scripts/cocci, but what I was curious about is whether spatch can find

  1.  Function that allocates a stack variable which is a pointer (or has one passed as a pointer argument)
  2.  Function calls *alloc
  3.  Function calls kfree() before return in all paths after alloc

There is also a variant of the above cleanup.h macros for functions that acquire a lock but need to release it before returning.

I know this would be nearly impossible to auto-patch, but having a "grep" for this pattern would be super helpful.

Would someone be interested in looking at this for the kernel? It seems like a valuable help to find refactoring locations.

Thoughts?

[-- Attachment #2: Type: text/html, Size: 4976 bytes --]