On Tue, Oct 15, 2019 at 4:50 PM Andrii Nakryiko <andrii.nakryiko@xxxxxxxxx> wrote: > > On Tue, Oct 15, 2019 at 4:41 PM Alexei Starovoitov <ast@xxxxxx> wrote: > > > > On 10/15/19 3:03 PM, Andrii Nakryiko wrote: > > > Define test runner generation meta-rule that codifies dependencies > > > between test runner, its tests, and its dependent BPF programs. Use that > > > for defining test_progs and test_maps test-runners. Also additionally define > > > 3 flavors of test_progs: > > > - alu32, which builds BPF programs with 32-bit registers codegen; > > > - bpf_gcc, which build BPF programs using GCC, if it supports BPF target; > > > - native, which uses a mix of native Clang target and BPF target for LLC. > > > > Great improvement, but it's taking it too far. > > (clang -I. -I/data/users/ast/net-next/tools/testing/selftests/bpf -g > > -D__TARGET_ARCH_x86 -I. -I./include/uapi > > -I/data/users/ast/net-next/tools/include/uapi > > -I/data/users/ast/net-next/tools/lib/bpf > > -I/data/users/ast/net-next/tools/testing/selftests/usr/include > > -idirafter /usr/local/include -idirafter > > /data/users/ast/llvm/bld/lib/clang/10.0.0/include -idirafter > > /usr/include -Wno-compare-distinct-pointer-types -O2 -emit-llvm -c > > progs/test_core_reloc_existence.c -o - || echo "BPF obj compilation > > failed") | llc -march=bpf -mcpu=probe -filetype=obj -o > > /data/users/ast/net-next/tools/testing/selftests/bpf/native/test_core_reloc_existence.o > > progs/test_core_reloc_existence.c:47:18: error: use of unknown builtin > > '__builtin_preserve_field_info' [-Wimplicit-function-declaration] > > out->a_exists = bpf_core_field_exists(in->a); > > Do you use latest clang that supports __builtin_preserve_field_info()? > All the flavors are building just fine for me with latest clang. Ok, so I didn't have __builtin_preserve_field_info() tests together with these Makefile changes and thus test_progs-native was compiling just fine for me. __builtin_preserve_access_index() is not BPF-target-only built-in, so it was compiling fine for non-BPF target. I've dropped test_progs-native and added back test_xdp.o override, but now **after** we define generic rule for test_progs, which will cause it to override previous test_xdp.o recipe. This causes make to emit warning about rule re-definition, which I'm not excited about, but avoiding it would require some further checks and filterings in DEFINE_TEST_RUNNER_RULES just for the sake of avoiding this warning, which seems to be an overkill... I'm wondering if we can just drop that mixed native Clang/bpf LLC mode altogether? But for now we have it with a warning in v2. > > > > > native clang + llc is useful for old school tracing only (before CO-RE). > > Don't disagree (I actually have little context why we needed this > special case at all), but I had no errors or warnings whatsoever. I > think in this particular case it's not specific to test_progs-native > build, can you please double-check on your side?
