Re: [PATCH bpf-next v6 3/8] bpf: lsm: provide attachment points for BPF LSM programs

[KP Singh]

On 25-Mär 12:28, Kees Cook wrote:
> On Wed, Mar 25, 2020 at 04:26:24PM +0100, KP Singh wrote:
> > From: KP Singh <kpsingh@xxxxxxxxxx>
> > 
> > When CONFIG_BPF_LSM is enabled, nop functions, bpf_lsm_<hook_name>, are
> > generated for each LSM hook. These functions are initialized as LSM
> > hooks in a subsequent patch.
> > 
> > Signed-off-by: KP Singh <kpsingh@xxxxxxxxxx>
> > Reviewed-by: Brendan Jackman <jackmanb@xxxxxxxxxx>
> > Reviewed-by: Florent Revest <revest@xxxxxxxxxx>
> > Reviewed-by: Kees Cook <keescook@xxxxxxxxxxxx>
> > Acked-by: Yonghong Song <yhs@xxxxxx>
> > ---
> >  include/linux/bpf_lsm.h | 22 ++++++++++++++++++++++
> >  kernel/bpf/bpf_lsm.c    | 14 ++++++++++++++
> >  2 files changed, 36 insertions(+)
> >  create mode 100644 include/linux/bpf_lsm.h
> > 
> > diff --git a/include/linux/bpf_lsm.h b/include/linux/bpf_lsm.h
> > new file mode 100644
> > index 000000000000..83b96895829f
> > --- /dev/null
> > +++ b/include/linux/bpf_lsm.h
> > @@ -0,0 +1,22 @@
> > +/* SPDX-License-Identifier: GPL-2.0 */
> > +
> > +/*
> > + * Copyright (C) 2020 Google LLC.
> > + */
> > +
> > +#ifndef _LINUX_BPF_LSM_H
> > +#define _LINUX_BPF_LSM_H
> > +
> > +#include <linux/bpf.h>
> > +#include <linux/lsm_hooks.h>
> > +
> > +#ifdef CONFIG_BPF_LSM
> > +
> > +#define LSM_HOOK(RET, DEFAULT, NAME, ...) \
> > +	RET bpf_lsm_##NAME(__VA_ARGS__);
> > +#include <linux/lsm_hook_defs.h>
> > +#undef LSM_HOOK
> > +
> > +#endif /* CONFIG_BPF_LSM */
> > +
> > +#endif /* _LINUX_BPF_LSM_H */
> > diff --git a/kernel/bpf/bpf_lsm.c b/kernel/bpf/bpf_lsm.c
> > index 82875039ca90..1210a819ca52 100644
> > --- a/kernel/bpf/bpf_lsm.c
> > +++ b/kernel/bpf/bpf_lsm.c
> > @@ -7,6 +7,20 @@
> >  #include <linux/filter.h>
> >  #include <linux/bpf.h>
> >  #include <linux/btf.h>
> > +#include <linux/lsm_hooks.h>
> > +#include <linux/bpf_lsm.h>
> > +
> > +/* For every LSM hook that allows attachment of BPF programs, declare a nop
> > + * function where a BPF program can be attached.
> > + */
> > +#define LSM_HOOK(RET, DEFAULT, NAME, ...) 	\
> > +noinline __weak RET bpf_lsm_##NAME(__VA_ARGS__)	\
> 
> I don't think the __weak is needed any more here?

This was suggested in:

 https://lore.kernel.org/bpf/20200221022537.wbmhdfkdbfvw2pww@ast-mbp/

"I think I saw cases when gcc ignored 'noinline' when function is
defined in the same file and still performed inlining while keeping
the function body.  To be safe I think __weak is necessary. That will
guarantee noinline."

It happened to work nicely with the previous approach for the special
hooks but the actual reason for adding the __weak was to guarrantee
that these functions don't get inlined.

> 
> > +{						\
> > +	return DEFAULT;				\
> 
> I'm impressed that LSM_RET_VOID actually works. :)

All the credit goes to Andrii :)

- KP

> 
> -Kees
> 
> > +}
> > +
> > +#include <linux/lsm_hook_defs.h>
> > +#undef LSM_HOOK
> >  
> >  const struct bpf_prog_ops lsm_prog_ops = {
> >  };
> > -- 
> > 2.20.1
> > 
> 
> -- 
> Kees Cook