Re: [PATCH v3] coccinelle: semantic code search for missing of_node_put

From: Julia Lawall
Date: Tue Jul 16 2019 - 07:09:12 EST




On Tue, 16 Jul 2019, Markus Elfring wrote:

> > We find these functions by using the following script:
>
> Why would you like to keep this SmPL code in the commit description?

I don't know indetail what you are proposing, but I would prefer not to
put semantic patches that involve iteration into the kernel, for
simplicity.

julia


>
> I would prefer software evolution in an other direction.
> https://lore.kernel.org/lkml/44be5924-26ca-5106-aa25-3cbc3343aa2c@xxxxxx/
> https://lkml.org/lkml/2019/7/4/21
>
>
> > @initialize:ocaml@
> > @@
> >
> > let relevant_str = "use of_node_put() on it when done"
>
> I see further possibilities to improve this data processing approach.
> https://lore.kernel.org/lkml/904b9362-cd01-ffc9-600b-0c48848617a0@xxxxxx/
> https://lore.kernel.org/patchwork/patch/1095169/#1291378
> https://lkml.org/lkml/2019/6/28/326
>
>
> I am missing more constructive answers for mentioned development concerns.
>
>
> > And this patch also looks for places â
>
> Does a SmPL script perform an action?
>
>
> > Finally, this patch finds use-after-free issues for a node.
> > (implemented by the r_use_after_put rule)
>
> This software extension is another interesting contribution.
> But I imagine that a separate SmPL script can be more helpful for
> this source code search pattern.
>
>
> > v3: delete the global set, â
>
> To which previous implementation detail do you refer here?
>
>
> > +virtual report
> > +virtual org
> > +
> > +@initialize:python@
> > +@@
> > +
> > +report_miss_prefix = "ERROR: missing of_node_put; acquired a node pointer with refcount incremented on line "
> > +report_miss_suffix = ", but without a corresponding object release within this function."
> > +org_miss_main = "acquired a node pointer with refcount incremented"
> > +org_miss_sec = "needed of_node_put"
> > +report_use_after_put = "ERROR: use-after-free; reference preceded by of_node_put on line "
> > +org_use_after_put_main = "of_node_put"
> > +org_use_after_put_sec = "reference"
>
> If you would insist on the usage of these variables, they should be applied
> only for the selected analysis operation mode.
> I would expect corresponding SmPL dependency specifications.
> https://github.com/coccinelle/coccinelle/blob/b4509f6e7fb06d5616bb19dd5a110b203fd0e566/docs/manual/cocci_syntax.tex#L559
>
>
> > +@r_miss_put exists@
> > +local idexpression struct device_node *x;
> > +expression e, e1;
> > +position p1, p2;
> > +statement S;
> > +type T, T1;
> > +@@
> > +
> > +* x = @p1\(of_find_all_nodes\|
>
> The usage of the SmPL asterisk functionality can fit to the operation mode âcontextâ.
> https://bottest.wiki.kernel.org/coccicheck#modes
> Would you like to add any corresponding SmPL details?
>
> Under which circumstances will remaining programming concerns be clarified
> for such SmPL disjunctions?
>
>
> > +... when != e = (T)x
> > + when != true x == NULL
>
> Will assignment exclusions get any more software development attention?
> https://lore.kernel.org/lkml/03cc4df5-ce7f-ba91-36b5-687fec8c7297@xxxxxx/
> https://lore.kernel.org/patchwork/patch/1095169/#1291892
> https://lkml.org/lkml/2019/6/29/193
>
>
> > + when != of_node_put(x)
> â
> > +)
> > +&
> > +x = f(...)
> > +...
> > +if (<+...x...+>) S
> > +...
> > +of_node_put(x);
> > +)
>
> You propose once more to use a SmPL conjunction in the rule âr_miss_put_extâ.
> I am also still waiting for a definitive explanation on the applicability
> of this combination.
>
>
> > +@r_put@
> > +expression E;
> > +position p1;
> > +@@
> > +
> > +* of_node_put@p1(E)
>
> I guess that this SmPL code will need further adjustments.
>
>
> > +@r_use_after_put exists@
> > +expression r_put.E, subE<=r_put.E;
>
> I have got an understanding difficulty around the interpretation
> of the shown SmPL constraint.
> How will the clarification be continued?
>
> Regards,
> Markus
>