Re: [PATCH v4 net-next 06/15] selftests: xsk: move the AF_XDP test suite to selftests/net

[Jakub Kicinski <kuba@kernel.org>]

On Fri, 16 Oct 2026 19:43:29 +0200 Maciej Fijalkowski wrote:
> Move the xskxceiver engine, the AF_XDP userspace library, the test
> definitions and the XDP program to net/lib/xsk, and test_xsk.sh with
> its prerequisites script to selftests/net. A later patch drives the
> same engine against real NICs from drivers/net/hw, and net/lib is where
> both suites take shared helpers from.
> 
> net/lib builds xskxceiver with its XDP program and skeleton, using the
> bpftool in PATH, and skips it with a warning when bpftool is missing or
> does not run; test_xsk.sh then skips too. xskxceiver links -lpthread,
> which glibc older than 2.34 needs for the threads of test_xsk.c.
> 
> Like the rest of net/lib, xskxceiver builds against the installed uapi
> headers, so include linux/kernel.h for min_t() directly rather than via
> the tools copy of linux/netlink.h. tools/include goes ahead of CFLAGS,
> so that O= builds take linux/kernel.h and linux/types.h from it rather
> than from a source tree usr/include.

To be clear the expectation is that user runs

 make headers

first, so installed != system headers. There's no requirement 
that selftests build without running make headers first.

> Move the ethtool ring helper from network_helpers into xsk.c and
> point the kselftest.h include of test_xsk.h at the net/lib include
> path. The XDP program moves unchanged; its Ethernet header bound check
> compares pointers of distinct types, so net/lib builds it with
> -Wno-compare-distinct-pointer-types, as the bpf selftests do.
> xdp_hw_metadata and the xdp_metadata test keep using xsk.c and xsk.h
> from their new location. Update the AF_XDP entry in MAINTAINERS.
> 

> +# AF_XDP test engine used by net/test_xsk.sh.
> +XSK_DIR := xsk
> +XSK_BPF_OBJ := $(OUTPUT)/xsk_xdp_progs.bpf.o
> +XSK_SKEL := $(OUTPUT)/xsk_xdp_progs.skel.h
> +XSK_SOURCES := $(addprefix $(XSK_DIR)/,xskxceiver.c xsk.c test_xsk.c) \
> +	$(top_srcdir)/tools/lib/find_bit.c
> +XSK_HEADERS := $(wildcard $(XSK_DIR)/*.h)
> +
> +PKG_CONFIG ?= $(CROSS_COMPILE)pkg-config
> +XSK_LDLIBS := $(shell $(PKG_CONFIG) libelf --libs 2>/dev/null || echo -lelf) -lz
> +XSK_LDLIBS += -lpthread
> +
> +$(XSK_BPF_OBJ): $(XSK_DIR)/xsk_xdp_progs.bpf.c $(XSK_DIR)/xsk_xdp_common.h $(BPFOBJ)
> +	$(call msg,BPF_PROG,,$@)
> +	$(Q)$(CLANG) -O2 -g --target=bpf $(CCINCLUDE) $(CLANG_SYS_INCLUDES) \
> +		-Wno-compare-distinct-pointer-types -c $< -o $@
> +
> +$(OUTPUT)/xskxceiver: $(XSK_SOURCES) $(XSK_HEADERS) $(XSK_SKEL) $(BPFOBJ)
> +	$(call msg,CC,,$@)
> +	$(Q)$(CC) -I$(top_srcdir)/tools/include $(CFLAGS) -I$(OUTPUT) \
> +		-I$(SCRATCH_DIR)/include \
> +		$(XSK_SOURCES) $(BPFOBJ) $(LDLIBS) $(XSK_LDLIBS) -o $@

Why do we need all these extra rules to build the xsk stuff?
bpf.mk should build all %.bpf.c into objects
You seem to do similar thing for skel.
We shouldn't need 20 lines of Makefile for each test.
We are not using any bleeding edge toolchain features here

> +EXTRA_CLEAN += $(XSK_BPF_OBJ) $(XSK_SKEL)

This should also somehow be covered by wildcards not obj by obj.