On Fri, Oct 6, 2023 at 10:55 AM Andrii Nakryiko <andrii.nakryiko@xxxxxxxxx> wrote: > > On Thu, Oct 5, 2023 at 2:04 AM Alan Maguire <alan.maguire@xxxxxxxxxx> wrote: > > > > On 05/10/2023 08:19, Jiri Olsa wrote: > > > On Wed, Oct 04, 2023 at 10:21:12AM -0700, Andrii Nakryiko wrote: > > >> On Wed, Oct 4, 2023 at 1:27 AM Jiri Olsa <olsajiri@xxxxxxxxx> wrote: > > >>> > > >>> On Tue, Oct 03, 2023 at 05:17:49PM -0700, Andrii Nakryiko wrote: > > >>>> Add support for building selftests with -O2 level of optimization, which > > >>>> allows more compiler warnings detection (like lots of potentially > > >>>> uninitialized usage), but also is useful to have a faster-running test > > >>>> for some CPU-intensive tests. > > >>>> > > >>>> One can build optimized versions of libbpf and selftests by running: > > >>>> > > >>>> $ make RELEASE=1 > > >>>> > > >>>> There is a measurable speed up of about 10 seconds for me locally, > > >>>> though it's mostly capped by non-parallelized serial tests. User CPU > > >>>> time goes down by total 40 seconds, from 1m10s to 0m28s. > > >>>> > > >>>> Unoptimized build (-O0) > > >>>> ======================= > > >>>> Summary: 430/3544 PASSED, 25 SKIPPED, 4 FAILED > > >>>> > > >>>> real 1m59.937s > > >>>> user 1m10.877s > > >>>> sys 3m14.880s > > >>>> > > >>>> Optimized build (-O2) > > >>>> ===================== > > >>>> Summary: 425/3543 PASSED, 25 SKIPPED, 9 FAILED > > >>>> > > >>>> real 1m50.540s > > >>>> user 0m28.406s > > >>>> sys 3m13.198s > > >>> > > >>> hi, > > >>> I get following error when running selftest compiled with RELEASE=1 > > >>> > > >>> # ./test_progs -t attach_probe/manual-legacy > > >>> test_attach_probe:PASS:skel_open 0 nsec > > >>> test_attach_probe:PASS:skel_load 0 nsec > > >>> test_attach_probe:PASS:check_bss 0 nsec > > >>> test_attach_probe:PASS:uprobe_ref_ctr_cleanup 0 nsec > > >>> test_attach_probe_manual:PASS:skel_kprobe_manual_open_and_load 0 nsec > > >>> test_attach_probe_manual:PASS:uprobe_offset 0 nsec > > >>> test_attach_probe_manual:PASS:attach_kprobe 0 nsec > > >>> test_attach_probe_manual:PASS:attach_kretprobe 0 nsec > > >>> test_attach_probe_manual:PASS:attach_uprobe 0 nsec > > >>> test_attach_probe_manual:PASS:attach_uretprobe 0 nsec > > >>> libbpf: failed to add legacy uprobe event for /proc/self/exe:0x19020: -17 > > >>> libbpf: prog 'handle_uprobe_byname': failed to create uprobe '/proc/self/exe:0x19020' perf event: File exists > > >>> test_attach_probe_manual:FAIL:attach_uprobe_byname unexpected error: -17 > > >>> #8/2 attach_probe/manual-legacy:FAIL > > >>> #8 attach_probe:FAIL > > >>> > > >>> > > >>> it looks like -O2 can merge some of the trigger functions: > > >>> > > >>> [root@qemu bpf]# nm test_progs | grep trigger_func > > >>> 0000000000558f30 t autoattach_trigger_func.constprop.0 > > >>> 000000000041d240 t trigger_func > > >>> 0000000000419020 t trigger_func > > >>> 0000000000420e70 t trigger_func > > >>> 0000000000507aa0 t trigger_func > > >>> 0000000000419020 t trigger_func2 > > >>> 0000000000419020 t trigger_func3 > > >>> 0000000000419030 t trigger_func4 > > >>> [root@qemu bpf]# nm test_progs | grep 0000000000419020 > > >>> 0000000000419020 t trigger_func > > >>> 0000000000419020 t trigger_func2 > > >>> 0000000000419020 t trigger_func3 > > >>> > > >>> I got more tests fails, but I suspect it's all for similar > > >>> reason like above > > >>> > > >> > > > > This one is an interesting failure that definitely seems worth fixing; > > as above trigger_func2 and trigger_func3 were merged it looks like, and > > the legacy perf_event_kprobe_open_legacy()-based attach failed due to > > add_uprobe_event_legacy() failing. The latter seems to be down to the > > way that gen_uprobe_legacy_event_name() constructs legacy event names > > via pid, binary_path and offset. In this case it will construct the > > same name for trigger_func2 and trigger_func3, hence the -EEXIST. > > > > It _seems_ like a fix here would be to add an optional func_name to > > gen_uprobe_legacy_event_name(). At least it appears that the non-legacy > > attach handles this case, which is great. Regardless, we can follow > > up with some of this stuff later. > > Yeah, let's fix this up as a follow up. I'm not sure about using > function name as part of uprobe name, mostly because these symbol > names can be super long, and I don't know what's kernel size limits. > So we probably want to keep them bounded in size. Having said that, > seems like we use binary path and we also don't sanitize that. So I Scratch that, we do sanitization at the end (I don't know how I missed it the first time), so the concern is only about the potentially very large binary path length. I think either way we should have a global counter for all uprobes to avoid any probability of a naming conflict. > think we should try to fix all that as a follow up: sanitize and maybe > truncate binary_path, and generally make sure that each uprobe name is > unique (probably with atomic counter). This atomic static counter will > take care of all such issues, right? Maybe we should drop the binary > path from the uprobe name altogether with that? > > Either way, thanks for taking a deeper look into failures! > > > > > >> yes, I didn't say that -O2 version passes all tests :) at least there > > >> are complicated USDT cases under -O2 which libbpf can't support (if I > > >> remember correctly, it was offset relative to global symbol case). But > > >> it's the first step. And once we have ability to build with RELEASE=1, > > >> we can add it as a separate test in CI and catch more of these > > >> uninitialized usage errors. Initially we can denylist tests that are > > >> broken due to -O2 and work to fix them. > > > > > > > Sounds good to me; it's a great change as we're more likely to hit > > more problems that users run into. > > yep > > > > > For the series: > > > > Reviewed-by: Alan Maguire <alan.maguire@xxxxxxxxxx> > > > > > I see ;-) sounds good > > > > > > Acked-by: Jiri Olsa <jolsa@xxxxxxxxxx> > > > > > > jirka > > > > > >> > > >>> jirka > > >>> > > >>> > > >>>> > > >>>> Signed-off-by: Andrii Nakryiko <andrii@xxxxxxxxxx> > > >>>> --- > > >>>> tools/testing/selftests/bpf/Makefile | 14 ++++++++------ > > >>>> 1 file changed, 8 insertions(+), 6 deletions(-) > > >>>> > > >>>> diff --git a/tools/testing/selftests/bpf/Makefile b/tools/testing/selftests/bpf/Makefile > > >>>> index a25e262dbc69..55d1b1848e6c 100644 > > >>>> --- a/tools/testing/selftests/bpf/Makefile > > >>>> +++ b/tools/testing/selftests/bpf/Makefile > > >>>> @@ -27,7 +27,9 @@ endif > > >>>> BPF_GCC ?= $(shell command -v bpf-gcc;) > > >>>> SAN_CFLAGS ?= > > >>>> SAN_LDFLAGS ?= $(SAN_CFLAGS) > > >>>> -CFLAGS += -g -O0 -rdynamic \ > > >>>> +RELEASE ?= > > >>>> +OPT_FLAGS ?= $(if $(RELEASE),-O2,-O0) > > >>>> +CFLAGS += -g $(OPT_FLAGS) -rdynamic \ > > >>>> -Wall -Werror \ > > >>>> $(GENFLAGS) $(SAN_CFLAGS) \ > > >>>> -I$(CURDIR) -I$(INCLUDE_DIR) -I$(GENDIR) -I$(LIBDIR) \ > > >>>> @@ -241,7 +243,7 @@ $(OUTPUT)/runqslower: $(BPFOBJ) | $(DEFAULT_BPFTOOL) $(RUNQSLOWER_OUTPUT) > > >>>> BPFTOOL_OUTPUT=$(HOST_BUILD_DIR)/bpftool/ \ > > >>>> BPFOBJ_OUTPUT=$(BUILD_DIR)/libbpf \ > > >>>> BPFOBJ=$(BPFOBJ) BPF_INCLUDE=$(INCLUDE_DIR) \ > > >>>> - EXTRA_CFLAGS='-g -O0 $(SAN_CFLAGS)' \ > > >>>> + EXTRA_CFLAGS='-g $(OPT_FLAGS) $(SAN_CFLAGS)' \ > > >>>> EXTRA_LDFLAGS='$(SAN_LDFLAGS)' && \ > > >>>> cp $(RUNQSLOWER_OUTPUT)runqslower $@ > > >>>> > > >>>> @@ -279,7 +281,7 @@ $(DEFAULT_BPFTOOL): $(wildcard $(BPFTOOLDIR)/*.[ch] $(BPFTOOLDIR)/Makefile) \ > > >>>> $(HOST_BPFOBJ) | $(HOST_BUILD_DIR)/bpftool > > >>>> $(Q)$(MAKE) $(submake_extras) -C $(BPFTOOLDIR) \ > > >>>> ARCH= CROSS_COMPILE= CC="$(HOSTCC)" LD="$(HOSTLD)" \ > > >>>> - EXTRA_CFLAGS='-g -O0' \ > > >>>> + EXTRA_CFLAGS='-g $(OPT_FLAGS)' \ > > >>>> OUTPUT=$(HOST_BUILD_DIR)/bpftool/ \ > > >>>> LIBBPF_OUTPUT=$(HOST_BUILD_DIR)/libbpf/ \ > > >>>> LIBBPF_DESTDIR=$(HOST_SCRATCH_DIR)/ \ > > >>>> @@ -290,7 +292,7 @@ $(CROSS_BPFTOOL): $(wildcard $(BPFTOOLDIR)/*.[ch] $(BPFTOOLDIR)/Makefile) \ > > >>>> $(BPFOBJ) | $(BUILD_DIR)/bpftool > > >>>> $(Q)$(MAKE) $(submake_extras) -C $(BPFTOOLDIR) \ > > >>>> ARCH=$(ARCH) CROSS_COMPILE=$(CROSS_COMPILE) \ > > >>>> - EXTRA_CFLAGS='-g -O0' \ > > >>>> + EXTRA_CFLAGS='-g $(OPT_FLAGS)' \ > > >>>> OUTPUT=$(BUILD_DIR)/bpftool/ \ > > >>>> LIBBPF_OUTPUT=$(BUILD_DIR)/libbpf/ \ > > >>>> LIBBPF_DESTDIR=$(SCRATCH_DIR)/ \ > > >>>> @@ -313,7 +315,7 @@ $(BPFOBJ): $(wildcard $(BPFDIR)/*.[ch] $(BPFDIR)/Makefile) \ > > >>>> $(APIDIR)/linux/bpf.h \ > > >>>> | $(BUILD_DIR)/libbpf > > >>>> $(Q)$(MAKE) $(submake_extras) -C $(BPFDIR) OUTPUT=$(BUILD_DIR)/libbpf/ \ > > >>>> - EXTRA_CFLAGS='-g -O0 $(SAN_CFLAGS)' \ > > >>>> + EXTRA_CFLAGS='-g $(OPT_FLAGS) $(SAN_CFLAGS)' \ > > >>>> EXTRA_LDFLAGS='$(SAN_LDFLAGS)' \ > > >>>> DESTDIR=$(SCRATCH_DIR) prefix= all install_headers > > >>>> > > >>>> @@ -322,7 +324,7 @@ $(HOST_BPFOBJ): $(wildcard $(BPFDIR)/*.[ch] $(BPFDIR)/Makefile) \ > > >>>> $(APIDIR)/linux/bpf.h \ > > >>>> | $(HOST_BUILD_DIR)/libbpf > > >>>> $(Q)$(MAKE) $(submake_extras) -C $(BPFDIR) \ > > >>>> - EXTRA_CFLAGS='-g -O0' ARCH= CROSS_COMPILE= \ > > >>>> + EXTRA_CFLAGS='-g $(OPT_FLAGS)' ARCH= CROSS_COMPILE= \ > > >>>> OUTPUT=$(HOST_BUILD_DIR)/libbpf/ \ > > >>>> CC="$(HOSTCC)" LD="$(HOSTLD)" \ > > >>>> DESTDIR=$(HOST_SCRATCH_DIR)/ prefix= all install_headers > > >>>> -- > > >>>> 2.34.1 > > >>>> > > >>>> > > >
