Re: [PATCH bpf-next v3 09/11] selftests: Remove tools/lib/bpf from include path

From: Toke HÃiland-JÃrgensen
Date: Fri Jan 17 2020 - 04:49:47 EST


Andrii Nakryiko <andrii.nakryiko@xxxxxxxxx> writes:

> On Thu, Jan 16, 2020 at 5:28 AM Toke HÃiland-JÃrgensen <toke@xxxxxxxxxx> wrote:
>>
>> From: Toke HÃiland-JÃrgensen <toke@xxxxxxxxxx>
>>
>> To make sure no new files are introduced that doesn't include the bpf/
>> prefix in its #include, remove tools/lib/bpf from the include path
>> entirely.
>>
>> Instead, we introduce a new header files directory under the scratch tools/
>> dir, and add a rule to run the 'install_headers' rule from libbpf to have a
>> full set of consistent libbpf headers in $(OUTPUT)/tools/include/bpf, and
>> then use $(OUTPUT)/tools/include as the include path for selftests.
>>
>> For consistency we also make sure we put all the scratch build files from
>> other bpftool and libbpf into tools/build/, so everything stays within
>> selftests/.
>>
>> Signed-off-by: Toke HÃiland-JÃrgensen <toke@xxxxxxxxxx>
>> ---
>> tools/testing/selftests/bpf/.gitignore | 1 +
>> tools/testing/selftests/bpf/Makefile | 50 +++++++++++++++++++-------------
>> 2 files changed, 31 insertions(+), 20 deletions(-)
>>
>> diff --git a/tools/testing/selftests/bpf/.gitignore b/tools/testing/selftests/bpf/.gitignore
>> index 1d14e3ab70be..849be9990ad2 100644
>> --- a/tools/testing/selftests/bpf/.gitignore
>> +++ b/tools/testing/selftests/bpf/.gitignore
>> @@ -40,3 +40,4 @@ test_cpp
>> /bpf_gcc
>> /tools
>> bpf_helper_defs.h
>> +/include/bpf
>
> Isn't the real path (within selftests/bpf) a tools/include/bpf, which
> is already ignored through /tools rule?

Yeah, you're correct. I started out with having it in include/bpf, but
ended up moving it, and guess I forgot to remove the .gitignore. Will fix.

>> diff --git a/tools/testing/selftests/bpf/Makefile b/tools/testing/selftests/bpf/Makefile
>> index 1fd7da49bd56..c3fa695bb028 100644
>> --- a/tools/testing/selftests/bpf/Makefile
>> +++ b/tools/testing/selftests/bpf/Makefile
>> @@ -20,8 +20,8 @@ CLANG ?= clang
>> LLC ?= llc
>> LLVM_OBJCOPY ?= llvm-objcopy
>> BPF_GCC ?= $(shell command -v bpf-gcc;)
>> -CFLAGS += -g -Wall -O2 $(GENFLAGS) -I$(CURDIR) -I$(APIDIR) -I$(LIBDIR) \
>> - -I$(BPFDIR) -I$(GENDIR) -I$(TOOLSINCDIR) \
>> +CFLAGS += -g -Wall -O2 $(GENFLAGS) -I$(CURDIR) -I$(APIDIR) \
>> + -I$(INCLUDE_DIR) -I$(GENDIR) -I$(LIBDIR) -I$(TOOLSINCDIR) \
>> -Dbpf_prog_load=bpf_prog_test_load \
>> -Dbpf_load_program=bpf_test_load_program
>> LDLIBS += -lcap -lelf -lz -lrt -lpthread
>> @@ -97,11 +97,15 @@ OVERRIDE_TARGETS := 1
>> override define CLEAN
>> $(call msg,CLEAN)
>> $(RM) -r $(TEST_GEN_PROGS) $(TEST_GEN_PROGS_EXTENDED) $(TEST_GEN_FILES) $(EXTRA_CLEAN)
>> - $(MAKE) -C $(BPFDIR) OUTPUT=$(OUTPUT)/ clean
>> endef
>>
>> include ../lib.mk
>>
>> +SCRATCH_DIR := $(OUTPUT)/tools
>> +BUILD_DIR := $(SCRATCH_DIR)/build
>> +INCLUDE_DIR := $(SCRATCH_DIR)/include
>> +INCLUDE_BPF := $(INCLUDE_DIR)/bpf/bpf.h
>> +
>> # Define simple and short `make test_progs`, `make test_sysctl`, etc targets
>> # to build individual tests.
>> # NOTE: Semicolon at the end is critical to override lib.mk's default static
>> @@ -120,7 +124,7 @@ $(OUTPUT)/urandom_read: urandom_read.c
>> $(call msg,BINARY,,$@)
>> $(CC) -o $@ $< -Wl,--build-id
>>
>> -$(OUTPUT)/test_stub.o: test_stub.c
>> +$(OUTPUT)/test_stub.o: test_stub.c $(INCLUDE_BPF)
>> $(call msg,CC,,$@)
>> $(CC) -c $(CFLAGS) -o $@ $<
>>
>> @@ -129,7 +133,7 @@ $(OUTPUT)/runqslower: force
>> $(Q)$(MAKE) $(submake_extras) -C $(TOOLSDIR)/bpf/runqslower \
>> OUTPUT=$(CURDIR)/tools/ VMLINUX_BTF=$(abspath ../../../../vmlinux)
>>
>> -BPFOBJ := $(OUTPUT)/libbpf.a
>> +BPFOBJ := $(BUILD_DIR)/libbpf/libbpf.a
>>
>> $(TEST_GEN_PROGS) $(TEST_GEN_PROGS_EXTENDED): $(OUTPUT)/test_stub.o $(BPFOBJ)
>>
>> @@ -155,17 +159,23 @@ force:
>> DEFAULT_BPFTOOL := $(OUTPUT)/tools/sbin/bpftool
>> BPFTOOL ?= $(DEFAULT_BPFTOOL)
>>
>> -$(DEFAULT_BPFTOOL): force
>> - $(Q)$(MAKE) $(submake_extras) -C $(BPFTOOLDIR) \
>> +$(BUILD_DIR)/libbpf $(BUILD_DIR)/bpftool $(INCLUDE_DIR):
>> + $(call msg,MKDIR,,$@)
>> + mkdir -p $@
>> +
>> +$(DEFAULT_BPFTOOL): force $(BUILD_DIR)/bpftool
>
> directories should be included as order-only dependencies (after | )

OK.

>> + $(Q)$(MAKE) $(submake_extras) -C $(BPFTOOLDIR) \
>> + OUTPUT=$(BUILD_DIR)/bpftool/ \
>> prefix= DESTDIR=$(OUTPUT)/tools/ install
>>
>> -$(BPFOBJ): force
>> - $(Q)$(MAKE) $(submake_extras) -C $(BPFDIR) OUTPUT=$(OUTPUT)/
>> +$(BPFOBJ): force $(BUILD_DIR)/libbpf
>
> same
>
>> + $(Q)$(MAKE) $(submake_extras) -C $(BPFDIR) \
>> + OUTPUT=$(BUILD_DIR)/libbpf/
>>
>> -BPF_HELPERS := $(OUTPUT)/bpf_helper_defs.h $(wildcard $(BPFDIR)/bpf_*.h)
>> -$(OUTPUT)/bpf_helper_defs.h: $(BPFOBJ)
>> - $(Q)$(MAKE) $(submake_extras) -C $(BPFDIR) \
>> - OUTPUT=$(OUTPUT)/ $(OUTPUT)/bpf_helper_defs.h
>> +BPF_HELPERS := $(wildcard $(BPFDIR)/bpf_*.h) $(INCLUDE_BPF)
>
> Shouldn't all BPF_HELPERS come from $(INCLUDE_DIR)/bpf now?
>
>> +$(INCLUDE_BPF): force $(BPFOBJ)
>
> And this can be more properly a $(BPF_HELPERS): force $(BPFOBJ)?
>
>> + $(Q)$(MAKE) $(submake_extras) -C $(BPFDIR) install_headers \
>> + OUTPUT=$(BUILD_DIR)/libbpf/ DESTDIR=$(SCRATCH_DIR) prefix=
>>
>> # Get Clang's default includes on this system, as opposed to those seen by
>> # '-target bpf'. This fixes "missing" files on some architectures/distros,
>> @@ -185,8 +195,8 @@ MENDIAN=$(if $(IS_LITTLE_ENDIAN),-mlittle-endian,-mbig-endian)
>>
>> CLANG_SYS_INCLUDES = $(call get_sys_includes,$(CLANG))
>> BPF_CFLAGS = -g -D__TARGET_ARCH_$(SRCARCH) $(MENDIAN) \
>> - -I$(OUTPUT) -I$(CURDIR) -I$(CURDIR)/include/uapi \
>> - -I$(APIDIR) -I$(LIBDIR) -I$(BPFDIR) -I$(abspath $(OUTPUT)/../usr/include)
>> + -I$(INCLUDE_DIR) -I$(CURDIR) -I$(CURDIR)/include/uapi \
>> + -I$(APIDIR) -I$(abspath $(OUTPUT)/../usr/include)
>>
>> CLANG_CFLAGS = $(CLANG_SYS_INCLUDES) \
>> -Wno-compare-distinct-pointer-types
>> @@ -306,7 +316,7 @@ $(TRUNNER_TEST_OBJS): $(TRUNNER_OUTPUT)/%.test.o: \
>> $(TRUNNER_EXTRA_HDRS) \
>> $(TRUNNER_BPF_OBJS) \
>> $(TRUNNER_BPF_SKELS) \
>> - $$(BPFOBJ) | $(TRUNNER_OUTPUT)
>> + $$(BPFOBJ) $$(INCLUDE_BPF) | $(TRUNNER_OUTPUT)
>
> singling out $(INCLUDE_BPF) looks weird? But I think $(BPFOBJ)
> achieves the same effect, so this change can be probably dropped? Same
> below.

I was having some trouble getting the dependency order right here.
$(INCLUDE_BPF) depends on $(BPFOBJ), not the other way around. May be
fixable though, I'll take another look.

-Toke