Re: [PATCH] KVM: replace DEFINE_SIMPLE_ATTRIBUTE with DEFINE_DEBUGFS_ATTRIBUTE

[Sean Christopherson <seanjc@google.com>]

On Wed, Nov 02, 2022, Paolo Bonzini wrote:
> On 11/1/22 08:25, Bo Liu wrote:
> > Fix the following coccicheck warning:
> >    virt/kvm/kvm_main.c:3847:0-23: WARNING
> >      vcpu_get_pid_fops should be defined with DEFINE_DEBUGFS_ATTRIBUTE
> > 
> > Signed-off-by: Bo Liu <liubo03@inspur.com>
> > ---
> >   virt/kvm/kvm_main.c | 2 +-
> >   1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
> > index f1df24c2bc84..3f383f27d3d7 100644
> > --- a/virt/kvm/kvm_main.c
> > +++ b/virt/kvm/kvm_main.c
> > @@ -3844,7 +3844,7 @@ static int vcpu_get_pid(void *data, u64 *val)
> >   	return 0;
> >   }
> > -DEFINE_SIMPLE_ATTRIBUTE(vcpu_get_pid_fops, vcpu_get_pid, NULL, "%llu\n");
> > +DEFINE_DEBUGFS_ATTRIBUTE(vcpu_get_pid_fops, vcpu_get_pid, NULL, "%llu\n");
> >   static void kvm_create_vcpu_debugfs(struct kvm_vcpu *vcpu)
> >   {
> 
> If you really wanted to do this, you would also have to replace
> debugfs_create_file with debugfs_create_file_unsafe.
> 
> However, this is not a good idea.  The rationale in the .cocci file is that
> "DEFINE_SIMPLE_ATTRIBUTE + debugfs_create_file() imposes some significant
> overhead", but this should not really be relevant for a debugfs file.
> 
> Such a patch would only make sense if there was a version of
> debugfs_create_file_unsafe() with a less-terrible name (e.g.
> debugfs_create_simple_attr?), which could _only_ be used with fops created
> by DEFINE_DEBUGFS_ATTRIBUTE.  Without such a type-safe trick, the .cocci
> file is only adding confusion to perfectly fine code.

Heh, some serious deja vu here[1].  This is the second case of identical, flawed
patches being sent in response to misguided coccinelle warnings in a rather short
amount of time, the "return min(r, 0)" horror being the other case[2][3].

The min() thing is supposed to be fixed by commit aeb300c1dbfc ("coccinelle: misc:
minmax: suppress patch generation for err returns").  Is that patch broken, or are
folks just running old scripts?

As for the DEFINE_DEBUGFS_ATTRIBUTE check, can that warning be downgraded (is that
even a thing?) or even deleted?  As much as I enjoyed the opportunity to learn more
about debugfs, the unnecessary confusion and wasted time was/is annoying.

[1] https://lore.kernel.org/all/Yxoo1A2fmlAWruyV@google.com
[2] https://lore.kernel.org/all/8881d7b4-0c31-cafd-1158-0d42c1c7f43a@redhat.com
[3] https://lore.kernel.org/all/d8a518c4a4014307b30020b38022d633@AcuMS.aculab.com