Skip to contentSkip to conversation

libbpf: don't require full struct enum64 in UAPI headers

PATCH In review Andrii Nakryiko 3 participants 4 messages Sep 27, 2022

Subsystems (4)
lore
Share & tools
lore

Andrii Nakryiko

author maintainer
#0 lore raw

PATCH 1/2 libbpf: don't require full struct enum64 in UAPI headers

Drop the requirement for system-wide kernel UAPI headers to provide full struct btf_enum64 definition. This is an unexpected requirement that slipped in libbpf 1.0 and put unnecessary pressure ([0]) on users to have a bleeding-edge kernel UAPI header from unreleased Linux 6.0.

To achieve this, we forward declare struct btf_enum64. But that's not enough as there is btf_enum64_value() helper that expects to know the layout of struct btf_enum64. So we get a bit creative with reinterpreting memory layout as array of __u32 and accesing lo32/hi32 fields as array elements. Alternative way would be to have a local pointer variable for anonymous struct with exactly the same layout as struct btf_enum64, but that gets us into C++ compiler errors complaining about invalid type casts. So play it safe, if ugly.

[0] Closes: https://github.com/libbpf/libbpf/issues/562

Fixes d90ec262b35b ("libbpf: Add enum64 support for btf_dump")
---
 tools/lib/bpf/btf.h | 25 ++++++++++++++++++++++++-
 1 file changed, 24 insertions(+), 1 deletion(-)
tools/lib/bpf/btf.h +24 -1
index ae543144ee30..8e6880d91c84 100644
@@ -486,6 +486,8 @@ static inline struct btf_enum *btf_enum(const struct btf_type *t)
 	return (struct btf_enum *)(t + 1);
 }
 
+struct btf_enum64;
+
 static inline struct btf_enum64 *btf_enum64(const struct btf_type *t)
 {
 	return (struct btf_enum64 *)(t + 1);
@@ -493,7 +495,28 @@ static inline struct btf_enum64 *btf_enum64(const struct btf_type *t)
 
 static inline __u64 btf_enum64_value(const struct btf_enum64 *e)
 {
-	return ((__u64)e->val_hi32 << 32) | e->val_lo32;
+	/* struct btf_enum64 is introduced in Linux 6.0, which is very
+	 * bleeding-edge. Here we are avoiding relying on struct btf_enum64
+	 * definition coming from kernel UAPI headers to support wider range
+	 * of system-wide kernel headers.
+	 *
+	 * Given this header can be also included from C++ applications, that
+	 * further restricts C tricks we can use (like using compatible
+	 * anonymous struct). So just treat struct btf_enum64 as
+	 * a three-element array of u32 and access second (lo32) and third
+	 * (hi32) elements directly.
+	 *
+	 * For reference, here is a struct btf_enum64 definition:
+	 *
+	 * const struct btf_enum64 {
+	 *	__u32	name_off;
+	 *	__u32	val_lo32;
+	 *	__u32	val_hi32;
+	 * };
+	 */
+	const __u32 *e64 = (const __u32 *)e;
+
+	return ((__u64)e64[2] << 32) | e64[1];
 }
 
 static inline struct btf_member *btf_members(const struct btf_type *t)
Signature · 1 line
2.30.2
author maintainer
#1 lore raw

PATCH 2/2 selftests/bpf: enforce C++11 mode for test_cpp test

Setting -std=c++11 seems to catch more potential C++-only problems. Also BPF skeleton isn't rely compilable with any older standard due to the use of nullptr.

---
 tools/testing/selftests/bpf/Makefile | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/testing/selftests/bpf/Makefile +1 -1
index e6cf21fad69f..d52069c70e49 100644
@@ -570,7 +570,7 @@ $(OUTPUT)/test_verifier: test_verifier.c verifier/tests.h $(BPFOBJ) | $(OUTPUT)
 # Make sure we are able to include and link libbpf against c++.
 $(OUTPUT)/test_cpp: test_cpp.cpp $(OUTPUT)/test_core_extern.skel.h $(BPFOBJ)
 	$(call msg,CXX,,$@)
-	$(Q)$(CXX) $(CFLAGS) $(filter %.a %.o %.cpp,$^) $(LDLIBS) -o $@
+	$(Q)$(CXX) $(CFLAGS) -std=c++11 $(filter %.a %.o %.cpp,$^) $(LDLIBS) -o $@
 
 # Benchmark runner
 $(OUTPUT)/bench_%.o: benchs/bench_%.c bench.h $(BPFOBJ)
Signature · 1 line
2.30.2
maintainer
#2 lore raw

Re: [PATCH bpf-next 2/2] selftests/bpf: enforce C++11 mode for test_cpp test

Quoted 15 lines · Andrii Nakryiko

Setting -std=c++11 seems to catch more potential C++-only problems. Also BPF skeleton isn't rely compilable with any older standard due to the use of nullptr.

---
  tools/testing/selftests/bpf/Makefile | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
tools/testing/selftests/bpf/Makefile +0 -0
index e6cf21fad69f..d52069c70e49 100644

tools/testing/selftests/bpf/Makefile $(OUTPUT)/test_verifier: test_verifier.c verifier/tests.h $(BPFOBJ) | $(OUTPUT) · line 576

Quoted 8 lines
tools/testing/selftests/bpf/Makefile +1 -1
@@ -570,7 +570,7 @@ $(OUTPUT)/test_verifier: test_verifier.c verifier/tests.h $(BPFOBJ) | $(OUTPUT)
  # Make sure we are able to include and link libbpf against c++.
  $(OUTPUT)/test_cpp: test_cpp.cpp $(OUTPUT)/test_core_extern.skel.h $(BPFOBJ)
  	$(call msg,CXX,,$@)
-	$(Q)$(CXX) $(CFLAGS) $(filter %.a %.o %.cpp,$^) $(LDLIBS) -o $@
+	$(Q)$(CXX) $(CFLAGS) -std=c++11 $(filter %.a %.o %.cpp,$^) $(LDLIBS) -o $@
 
  # Benchmark runner
  $(OUTPUT)/bench_%.o: benchs/bench_%.c bench.h $(BPFOBJ)

Seems CI triggers build error (https://github.com/kernel-patches/bpf/actions/runs/3133037577/jobs/5085992832) :

    [...]
    GEN-SKEL [test_progs] pyperf180.skel.h
    GEN-SKEL [test_progs] pyperf600.skel.h
    GEN-SKEL [test_progs-no_alu32] pyperf180.skel.h
    GEN-SKEL [test_progs-no_alu32] pyperf600.skel.h
    BINARY   bench
  In file included from test_cpp.cpp:9:
  /tmp/work/bpf/bpf/tools/testing/selftests/bpf/test_core_extern.skel.h:45:3: error: unknown type name '_Bool'
                  _Bool CONFIG_BOOL;
                  ^
  /tmp/work/bpf/bpf/tools/testing/selftests/bpf/test_core_extern.skel.h:46:3: error: unknown type name '_Bool'
                  _Bool CONFIG_BPF_SYSCALL;
                  ^

2 errors generated.
make: *** [Makefile:573: /tmp/work/bpf/bpf/tools/testing/selftests/bpf/test_cpp] Error 1
make: *** Waiting for unfinished jobs....
make: Leaving directory '/tmp/work/bpf/bpf/tools/testing/selftests/bpf'
Error: Process completed with exit code 2.

Thanks,
Daniel

Acked-by on v1 1/2
#3 lore raw
In reply to Andrii Nakryiko
Quoted 14 lines · Andrii Nakryiko

Drop the requirement for system-wide kernel UAPI headers to provide full struct btf_enum64 definition. This is an unexpected requirement that slipped in libbpf 1.0 and put unnecessary pressure ([0]) on users to have a bleeding-edge kernel UAPI header from unreleased Linux 6.0.

To achieve this, we forward declare struct btf_enum64. But that's not enough as there is btf_enum64_value() helper that expects to know the layout of struct btf_enum64. So we get a bit creative with reinterpreting memory layout as array of __u32 and accesing lo32/hi32 fields as array elements. Alternative way would be to have a local pointer variable for anonymous struct with exactly the same layout as struct btf_enum64, but that gets us into C++ compiler errors complaining about invalid type casts. So play it safe, if ugly.

Quoted 5 lines

[0] Closes: https://github.com/libbpf/libbpf/issues/562

Fixes d90ec262b35b ("libbpf: Add enum64 support for btf_dump")

This seems like a reasonable (and only a bit ugly) workaround; thanks!