Currently use_after_iter.cocci generates false positives for code of the
following form:
~~~
list_for_each_entry(d, &ddata->irq_list, node) {
if (irq == d->irq)
break;
}
if (list_entry_is_head(d, &ddata->irq_list, node))
return IRQ_NONE;
~~~
[This specific example comes from drivers/power/supply/cpcap-battery.c]
Most list macros use list_entry_is_head() as loop exit condition meaning it
is not unsafe to reuse pos (a.k.a. d) in the code above.
Let's avoid reporting these cases.
Signed-off-by: Daniel Thompson <[email protected]>
---
Notes:
I'm pretty much a complete beginner w.r.t. SmPL. This is written
entirely by finding previous fixes and emulating them!
However I did test it by running the checker across the current kernel
tree. The changes reduced the error count by four... which was small
enough for me to eyeball each one and check they match the pattern I
was targetting.
scripts/coccinelle/iterators/use_after_iter.cocci | 2 ++
1 file changed, 2 insertions(+)
diff --git a/scripts/coccinelle/iterators/use_after_iter.cocci b/scripts/coccinelle/iterators/use_after_iter.cocci
index 9be48b520879..676edd562eef 100644
--- a/scripts/coccinelle/iterators/use_after_iter.cocci
+++ b/scripts/coccinelle/iterators/use_after_iter.cocci
@@ -123,6 +123,8 @@ hlist_for_each_entry_safe(c,...) S
|
list_remove_head(x,c,...)
|
+list_entry_is_head(c,...)
+|
sizeof(<+...c...+>)
|
&c->member
base-commit: 2734d6c1b1a089fb593ef6a23d4b70903526fe0c
--
2.30.2
On Fri, 30 Jul 2021, Daniel Thompson wrote:
> Currently use_after_iter.cocci generates false positives for code of the
> following form:
> ~~~
> list_for_each_entry(d, &ddata->irq_list, node) {
> if (irq == d->irq)
> break;
> }
>
> if (list_entry_is_head(d, &ddata->irq_list, node))
> return IRQ_NONE;
> ~~~
> [This specific example comes from drivers/power/supply/cpcap-battery.c]
>
> Most list macros use list_entry_is_head() as loop exit condition meaning it
> is not unsafe to reuse pos (a.k.a. d) in the code above.
>
> Let's avoid reporting these cases.
>
> Signed-off-by: Daniel Thompson <[email protected]>
> ---
>
> Notes:
> I'm pretty much a complete beginner w.r.t. SmPL. This is written
> entirely by finding previous fixes and emulating them!
>
> However I did test it by running the checker across the current kernel
> tree. The changes reduced the error count by four... which was small
> enough for me to eyeball each one and check they match the pattern I
> was targetting.
This looks fine. Thanks for the proposal.
julia
>
> scripts/coccinelle/iterators/use_after_iter.cocci | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/scripts/coccinelle/iterators/use_after_iter.cocci b/scripts/coccinelle/iterators/use_after_iter.cocci
> index 9be48b520879..676edd562eef 100644
> --- a/scripts/coccinelle/iterators/use_after_iter.cocci
> +++ b/scripts/coccinelle/iterators/use_after_iter.cocci
> @@ -123,6 +123,8 @@ hlist_for_each_entry_safe(c,...) S
> |
> list_remove_head(x,c,...)
> |
> +list_entry_is_head(c,...)
> +|
> sizeof(<+...c...+>)
> |
> &c->member
>
> base-commit: 2734d6c1b1a089fb593ef6a23d4b70903526fe0c
> --
> 2.30.2
>
>
On Fri, 30 Jul 2021, Daniel Thompson wrote:
> Currently use_after_iter.cocci generates false positives for code of the
> following form:
> ~~~
> list_for_each_entry(d, &ddata->irq_list, node) {
> if (irq == d->irq)
> break;
> }
>
> if (list_entry_is_head(d, &ddata->irq_list, node))
> return IRQ_NONE;
> ~~~
> [This specific example comes from drivers/power/supply/cpcap-battery.c]
>
> Most list macros use list_entry_is_head() as loop exit condition meaning it
> is not unsafe to reuse pos (a.k.a. d) in the code above.
>
> Let's avoid reporting these cases.
Applied.
The function is really nice. The use_after_iter.cocci rule gives a lot of
false positives where people have used random variables as flags to
indicate how the loop was executed...
thanks,
julia
>
> Signed-off-by: Daniel Thompson <[email protected]>
> ---
>
> Notes:
> I'm pretty much a complete beginner w.r.t. SmPL. This is written
> entirely by finding previous fixes and emulating them!
>
> However I did test it by running the checker across the current kernel
> tree. The changes reduced the error count by four... which was small
> enough for me to eyeball each one and check they match the pattern I
> was targetting.
>
> scripts/coccinelle/iterators/use_after_iter.cocci | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/scripts/coccinelle/iterators/use_after_iter.cocci b/scripts/coccinelle/iterators/use_after_iter.cocci
> index 9be48b520879..676edd562eef 100644
> --- a/scripts/coccinelle/iterators/use_after_iter.cocci
> +++ b/scripts/coccinelle/iterators/use_after_iter.cocci
> @@ -123,6 +123,8 @@ hlist_for_each_entry_safe(c,...) S
> |
> list_remove_head(x,c,...)
> |
> +list_entry_is_head(c,...)
> +|
> sizeof(<+...c...+>)
> |
> &c->member
>
> base-commit: 2734d6c1b1a089fb593ef6a23d4b70903526fe0c
> --
> 2.30.2
>
>