On Mon, Jan 30, 2023 at 09:15:26AM -0600, David Vernet wrote:
> On Mon, Jan 30, 2023 at 09:55:34AM +0100, Jiri Olsa wrote:
> > Move all kfunc exports into separate header file and include it
> > in tests that need it.
> >
> > We will move all test kfuncs into bpf_testmod in following change,
> > so it's convenient to have declarations in single place.
>
> Nice, good call.
>
> >
> > Signed-off-by: Jiri Olsa <jolsa@xxxxxxxxxx>
> > ---
> > .../bpf/bpf_testmod/bpf_testmod_kfunc.h | 92 +++++++++++++++++++
> > tools/testing/selftests/bpf/progs/cb_refs.c | 4 +-
> > .../selftests/bpf/progs/jit_probe_mem.c | 4 +-
> > .../bpf/progs/kfunc_call_destructive.c | 3 +-
> > .../selftests/bpf/progs/kfunc_call_fail.c | 9 +-
> > .../selftests/bpf/progs/kfunc_call_race.c | 3 +-
> > .../selftests/bpf/progs/kfunc_call_test.c | 16 +---
> > .../bpf/progs/kfunc_call_test_subprog.c | 17 +++-
> > tools/testing/selftests/bpf/progs/map_kptr.c | 6 +-
> > .../selftests/bpf/progs/map_kptr_fail.c | 5 +-
> > 10 files changed, 114 insertions(+), 45 deletions(-)
> > create mode 100644 tools/testing/selftests/bpf/bpf_testmod/bpf_testmod_kfunc.h
> >
> > diff --git a/tools/testing/selftests/bpf/bpf_testmod/bpf_testmod_kfunc.h b/tools/testing/selftests/bpf/bpf_testmod/bpf_testmod_kfunc.h
> > new file mode 100644
> > index 000000000000..164cbae2b0d7
> > --- /dev/null
> > +++ b/tools/testing/selftests/bpf/bpf_testmod/bpf_testmod_kfunc.h
>
> Should we update the selftests Makefile to rebuild progs when the testmod
> changes? Something like:
>
> diff --git a/tools/testing/selftests/bpf/Makefile b/tools/testing/selftests/bpf/Makefile
> index e79039726173..ed0fb32aa855 100644
> --- a/tools/testing/selftests/bpf/Makefile
> +++ b/tools/testing/selftests/bpf/Makefile
> @@ -438,6 +438,7 @@ $(TRUNNER_BPF_OBJS): $(TRUNNER_OUTPUT)/%.bpf.o: \
> $(TRUNNER_BPF_PROGS_DIR)/%.c \
> $(TRUNNER_BPF_PROGS_DIR)/*.h \
> $$(INCLUDE_DIR)/vmlinux.h \
> + $(OUTPUT)/bpf_testmod.ko \
> $(wildcard $(BPFDIR)/bpf_*.h) \
> $(wildcard $(BPFDIR)/*.bpf.h) \
> | $(TRUNNER_OUTPUT) $$(BPFOBJ)
ok, looks good will add it
>
> > @@ -0,0 +1,92 @@
> > +/* SPDX-License-Identifier: GPL-2.0 */
> > +
> > +#ifndef _BPF_TESTMOD_KFUNC_H
> > +#define _BPF_TESTMOD_KFUNC_H
> > +
> > +#ifndef __ksym
> > +#define __ksym __attribute__((section(".ksyms")))
> > +#endif
>
> What about doing something like this:
>
> #ifndef __KERNEL__
> #include <vmlinux.h>
> #include <bpf/bpf_helpers.h>
> #else
> #define __ksym
> #endif
>
> Thoughts?
that goes nicely along with the solution for extra typedef you suggested below
SNIP
> > +extern struct prog_test_ref_kfunc *
> > +bpf_kfunc_call_test_acquire(unsigned long *scalar_ptr) __ksym;
> > +extern struct prog_test_ref_kfunc *
> > +bpf_kfunc_call_test_kptr_get(struct prog_test_ref_kfunc **p, int a, int b) __ksym;
> > +extern void bpf_kfunc_call_test_release(struct prog_test_ref_kfunc *p) __ksym;
> > +
> > +extern void bpf_kfunc_call_test_mem_len_pass1(void *mem, int len) __ksym;
> > +extern int *bpf_kfunc_call_test_get_rdwr_mem(struct prog_test_ref_kfunc *p, const int rdwr_buf_size) __ksym;
> > +extern int *bpf_kfunc_call_test_get_rdonly_mem(struct prog_test_ref_kfunc *p, const int rdonly_buf_size) __ksym;
> > +extern int *bpf_kfunc_call_test_acq_rdonly_mem(struct prog_test_ref_kfunc *p, const int rdonly_buf_size) __ksym;
> > +extern void bpf_kfunc_call_int_mem_release(int *p) __ksym;
> > +
> > +extern void bpf_testmod_test_mod_kfunc(int i) __ksym;
> > +
> > +extern __u64 bpf_kfunc_call_test1(struct sock *sk, __u32 a, __u64 b,
> > + __u32 c, __u64 d) __ksym;
> > +extern int bpf_kfunc_call_test2(struct sock *sk, __u32 a, __u32 b) __ksym;
> > +extern struct sock *bpf_kfunc_call_test3(struct sock *sk) __ksym;
> > +extern long bpf_kfunc_call_test4(signed char a, short b, int c, long d) __ksym;
> > +
> > +extern void bpf_kfunc_call_test_pass_ctx(struct __sk_buff *skb) __ksym;
> > +extern void bpf_kfunc_call_test_pass1(struct prog_test_pass1 *p) __ksym;
> > +extern void bpf_kfunc_call_test_pass2(struct prog_test_pass2 *p) __ksym;
> > +extern void bpf_kfunc_call_test_mem_len_fail2(__u64 *mem, int len) __ksym;
> > +
> > +extern void bpf_kfunc_call_test_destructive(void) __ksym;
>
> nit: Can we remove extern from all of these function signatures? Doesn't
> really matter much to leave it there, but given that the keyword does
> nothing for functions it feels unnecessary / noisy.
np, I can remove it
SNIP
> > diff --git a/tools/testing/selftests/bpf/progs/kfunc_call_test_subprog.c b/tools/testing/selftests/bpf/progs/kfunc_call_test_subprog.c
> > index c1fdecabeabf..f74c78bb5efd 100644
> > --- a/tools/testing/selftests/bpf/progs/kfunc_call_test_subprog.c
> > +++ b/tools/testing/selftests/bpf/progs/kfunc_call_test_subprog.c
> > @@ -4,10 +4,21 @@
> > #include <bpf/bpf_helpers.h>
> > #include "bpf_tcp_helpers.h"
> >
> > +/*
> > + * We can't include vmlinux.h, because it conflicts with bpf_tcp_helpers.h,
> > + * but we need refcount_t typedef for bpf_testmod_kfunc.h.
> > + * Adding it directly.
> > + */
> > +typedef struct {
> > + int counter;
> > +} atomic_t;
> > +typedef struct refcount_struct {
> > + atomic_t refs;
> > +} refcount_t;
>
> Meh, this is kind of unfortunate, but OK, not the end of the world.
> Don't really see an easy way to resolve these types of typedef / include
> spaghetti issues in a general way.
>
> As an alternative, it looks like this also works:
>
> diff --git a/tools/testing/selftests/bpf/progs/kfunc_call_test_subprog.c b/tools/testing/selftests/bpf/progs/kfunc_call_test_subprog.c
> index f74c78bb5efd..7b3472ebc445 100644
> --- a/tools/testing/selftests/bpf/progs/kfunc_call_test_subprog.c
> +++ b/tools/testing/selftests/bpf/progs/kfunc_call_test_subprog.c
> @@ -1,21 +1,8 @@
> // SPDX-License-Identifier: GPL-2.0
> /* Copyright (c) 2021 Facebook */
> -#include <linux/bpf.h>
> +#include <linux/types.h>
> #include <bpf/bpf_helpers.h>
> -#include "bpf_tcp_helpers.h"
> -
> -/*
> - * We can't include vmlinux.h, because it conflicts with bpf_tcp_helpers.h,
> - * but we need refcount_t typedef for bpf_testmod_kfunc.h.
> - * Adding it directly.
> - */
> -typedef struct {
> - int counter;
> -} atomic_t;
> -typedef struct refcount_struct {
> - atomic_t refs;
> -} refcount_t;
> -
> +#include <vmlinux.h>
> #include "bpf_testmod/bpf_testmod_kfunc.h"
>
> extern const int bpf_prog_active __ksym;
> @@ -39,7 +26,7 @@ int __noinline f1(struct __sk_buff *skb)
> if (active)
> active_res = *active;
>
> - sk_state_res = bpf_kfunc_call_test3((struct sock *)sk)->sk_state;
> + sk_state_res = bpf_kfunc_call_test3((struct sock *)sk)->__sk_common.skc_state;
great, I was wondering why the sock type was different when I was trying similar fix,
that looks much better, will try that
thanks,
jirka
