2023-06-14 08:00:31

by Yonggang Wu

[permalink] [raw]
Subject: [PATCH] selftests/bpf: Fix the address is NULL

Fix the following coccicheck error:

tools/testing/selftests/bpf/progs/test_ksyms_weak.c:53:6-20: ERROR: test
of a variable/field address

Signed-off-by: Yonggang Wu <[email protected]>
---
tools/testing/selftests/bpf/progs/test_ksyms_weak.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tools/testing/selftests/bpf/progs/test_ksyms_weak.c
b/tools/testing/selftests/bpf/progs/test_ksyms_weak.c
index d00268c91e19..768a4d6ee6f5 100644
--- a/tools/testing/selftests/bpf/progs/test_ksyms_weak.c
+++ b/tools/testing/selftests/bpf/progs/test_ksyms_weak.c
@@ -50,7 +50,7 @@ int pass_handler(const void *ctx)
/* tests non-existent symbols. */
out__non_existent_typed = (__u64)&bpf_link_fops2;

- if (&bpf_link_fops2) /* can't happen */
+ if (&bpf_link_fops2 != NULL) /* can't happen */
out__non_existent_typed =
(__u64)bpf_per_cpu_ptr(&bpf_link_fops2, 0);

if (!bpf_ksym_exists(bpf_task_acquire))


2023-06-14 14:41:46

by Yonghong Song

[permalink] [raw]
Subject: Re: [PATCH] selftests/bpf: Fix the address is NULL



On 6/14/23 12:42 AM, [email protected] wrote:
> Fix the following coccicheck error:
>
> tools/testing/selftests/bpf/progs/test_ksyms_weak.c:53:6-20: ERROR: test
> of a variable/field address

I didn't see clang/gcc compiler warns about this. Maybe need some
additional flags beyond what current selftest/bpf already has
in order to trigger this warning?
If you feel this warning has some merit, could you propose
it to gcc/llvm community?

>
> Signed-off-by: Yonggang Wu <[email protected]>
> ---
>  tools/testing/selftests/bpf/progs/test_ksyms_weak.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/tools/testing/selftests/bpf/progs/test_ksyms_weak.c
> b/tools/testing/selftests/bpf/progs/test_ksyms_weak.c
> index d00268c91e19..768a4d6ee6f5 100644
> --- a/tools/testing/selftests/bpf/progs/test_ksyms_weak.c
> +++ b/tools/testing/selftests/bpf/progs/test_ksyms_weak.c
> @@ -50,7 +50,7 @@ int pass_handler(const void *ctx)
>      /* tests non-existent symbols. */
>      out__non_existent_typed = (__u64)&bpf_link_fops2;
>
> -    if (&bpf_link_fops2) /* can't happen */
> +    if (&bpf_link_fops2 != NULL) /* can't happen */
>          out__non_existent_typed =
> (__u64)bpf_per_cpu_ptr(&bpf_link_fops2, 0);
>
>      if (!bpf_ksym_exists(bpf_task_acquire))
>

2023-06-16 16:49:34

by Andrii Nakryiko

[permalink] [raw]
Subject: Re: [PATCH] selftests/bpf: Fix the address is NULL

On Wed, Jun 14, 2023 at 7:09 AM Yonghong Song <[email protected]> wrote:
>
>
>
> On 6/14/23 12:42 AM, [email protected] wrote:
> > Fix the following coccicheck error:
> >
> > tools/testing/selftests/bpf/progs/test_ksyms_weak.c:53:6-20: ERROR: test
> > of a variable/field address
>
> I didn't see clang/gcc compiler warns about this. Maybe need some
> additional flags beyond what current selftest/bpf already has
> in order to trigger this warning?
> If you feel this warning has some merit, could you propose
> it to gcc/llvm community?

bpf_link_fops2 is a weak symbol, this check is totally valid and reasonable.

There are two problems here, though:

a) coccicheck shouldn't warn about "test of a variable/field address"
for weak symbols, because they can be NULL.

b) this patch is not even fixing that warning, it does a no-op change
from implicit non-NULL check to explicit non-NULL check. And the
former is actually the preferred style.

So this patch is doubly wrong.

>
> >
> > Signed-off-by: Yonggang Wu <[email protected]>
> > ---
> > tools/testing/selftests/bpf/progs/test_ksyms_weak.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/tools/testing/selftests/bpf/progs/test_ksyms_weak.c
> > b/tools/testing/selftests/bpf/progs/test_ksyms_weak.c
> > index d00268c91e19..768a4d6ee6f5 100644
> > --- a/tools/testing/selftests/bpf/progs/test_ksyms_weak.c
> > +++ b/tools/testing/selftests/bpf/progs/test_ksyms_weak.c
> > @@ -50,7 +50,7 @@ int pass_handler(const void *ctx)
> > /* tests non-existent symbols. */
> > out__non_existent_typed = (__u64)&bpf_link_fops2;
> >
> > - if (&bpf_link_fops2) /* can't happen */
> > + if (&bpf_link_fops2 != NULL) /* can't happen */
> > out__non_existent_typed =
> > (__u64)bpf_per_cpu_ptr(&bpf_link_fops2, 0);
> >
> > if (!bpf_ksym_exists(bpf_task_acquire))
> >