Re: [PATCH v16 06/12] bpf: Add bpf_lookup_*_key() and bpf_key_put() kfuncs

[Alexei Starovoitov]

On Tue, Sep 6, 2022 at 1:01 AM Roberto Sassu
<roberto.sassu@xxxxxxxxxxxxxxx> wrote:
>
> > > +struct bpf_key *bpf_lookup_user_key(u32 serial, u64 flags)
> > > +{
> > > +       key_ref_t key_ref;
> > > +       struct bpf_key *bkey;
> > > +
> > > +       if (flags & ~KEY_LOOKUP_ALL)
> > > +               return NULL;
> > > +
> > > +       /*
> > > +        * Permission check is deferred until the key is used, as
> > > the
> > > +        * intent of the caller is unknown here.
> > > +        */
> > > +       key_ref = lookup_user_key(serial, flags,
> > > KEY_DEFER_PERM_CHECK);
> > > +       if (IS_ERR(key_ref))
> > > +               return NULL;
> > > +
> > > +       bkey = kmalloc(sizeof(*bkey), GFP_ATOMIC);
> >
> > Since this function (due to lookup_user_key) is sleepable, do we
> > really need GFP_ATOMIC here?
>
> Daniel suggested it for bpf_lookup_system_key(), so that the kfunc does
> not have to be sleepable.

Hold on. It has to be sleepable. Just take a look
at what lookup_user_key is doing inside.

> For symmetry, I did the same to
> bpf_lookup_user_key(). Will switch back to GFP_KERNEL.
>
> Thanks
>
> Roberto
>