| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <CAEf4BzY=ZrS1+7RDPFyzLheZv-C4HqBEh_iLWgnUFPh95TvRFg@mail.gmail.com>
Date: Wed, 14 Aug 2019 13:22:34 -0700
From: Andrii Nakryiko <andrii.nakryiko@...il.com>
To: Stanislav Fomichev <sdf@...ichev.me>
Cc: Stanislav Fomichev <sdf@...gle.com>,
Networking <netdev@...r.kernel.org>, bpf <bpf@...r.kernel.org>,
"David S. Miller" <davem@...emloft.net>,
Alexei Starovoitov <ast@...nel.org>,
Daniel Borkmann <daniel@...earbox.net>,
Andrii Nakryiko <andriin@...com>
Subject: Re: [PATCH bpf-next 3/4] selftests/bpf: test_progs: remove global
fail/success counts
On Wed, Aug 14, 2019 at 1:07 PM Stanislav Fomichev <sdf@...ichev.me> wrote:
>
> On 08/14, Andrii Nakryiko wrote:
> > On Wed, Aug 14, 2019 at 9:48 AM Stanislav Fomichev <sdf@...gle.com> wrote:
> > >
> > > Now that we have a global per-test/per-environment state, there
> > > is no longer the need to have global fail/success counters
> > > (and there is no need to save/get the diff before/after the
> > > test).
> > >
> > > Cc: Andrii Nakryiko <andriin@...com>
> > > Signed-off-by: Stanislav Fomichev <sdf@...gle.com>
> > > ---
[...]
> > > +void __test__fail(const char *file, int line)
> > > +{
> > > + if (env.test->subtest_name)
> > > + fprintf(stderr, "%s:%s failed at %s:%d, errno=%d\n",
> >
> > Nit: let's keep <test>/<subtest> convention here as well?
> >
> > Failure doesn't always set errno, this will be quite confusing
> > sometimes. Especially for higher-level libbpf APIs, which don't set
> > errno at all.
> > If test wants to log additional information, let it do it explicitly.
> SG. Maybe we can adapt log_err from cgroup_helpers.h for error reporting
> once I move sockopt tests into test_progs.
>
> > Also, _CHECK already logs error message, so this is going to be
> > double-verbose for typical case. I'd say let's drop these error
> > messages and instead slightly extend _CHECK one with line number (it
> > already logs func name).
> Not everything goes through the _CHECK macro unfortunately, see
Well, it should (at least eventually). If existing macro doesn't cover
typical use case, we can add one that does cover.
> all the cases where I did s/error_cnt++/test__fail/. How about I
> remove the error message from _CHECK and leave it in __test_fail?
I'd keep test__fail() and test__success() as a low-level building
block. And move all the logging into corresponding high-level macro.
This still gives flexibility to do one-off crazy tests, if necessary,
while having consistent approach for everything else.
>
> > > + env.test->test_name, env.test->subtest_name,
> > > + file, line, errno);
> > > + else
> > > + fprintf(stderr, "%s failed at %s:%d, errno=%d\n",
> > > + env.test->test_name, file, line, errno);
> > > +
> > > + env.test->fail_cnt++;
> > > +}
> > > +
> > > struct ipv4_packet pkt_v4 = {
> > > .eth.h_proto = __bpf_constant_htons(ETH_P_IP),
> > > .iph.ihl = 5,
> > > @@ -150,7 +145,7 @@ int bpf_find_map(const char *test, struct bpf_object *obj, const char *name)
> > > map = bpf_object__find_map_by_name(obj, name);
> > > if (!map) {
> > > printf("%s:FAIL:map '%s' not found\n", test, name);
> > > - error_cnt++;
> > > + test__fail();
> > > return -1;
> > > }
> > > return bpf_map__fd(map);
> > > @@ -509,8 +504,6 @@ int main(int argc, char **argv)
> > > stdio_hijack();
> > > for (i = 0; i < prog_test_cnt; i++) {
> > > struct prog_test_def *test = &prog_test_defs[i];
> > > - int old_pass_cnt = pass_cnt;
> > > - int old_error_cnt = error_cnt;
> > >
> > > env.test = test;
> > > test->test_num = i + 1;
> > > @@ -525,14 +518,12 @@ int main(int argc, char **argv)
> > > test__end_subtest();
> > >
> > > test->tested = true;
> > > - test->pass_cnt = pass_cnt - old_pass_cnt;
> > > - test->error_cnt = error_cnt - old_error_cnt;
> > > - if (test->error_cnt)
> > > + if (test->fail_cnt)
> > > env.fail_cnt++;
> > > else
> > > env.succ_cnt++;
> > >
> > > - dump_test_log(test, test->error_cnt);
> > > + dump_test_log(test, test->fail_cnt);
> > >
> > > fprintf(env.stdout, "#%3d %4s %s\n",
> > > test->test_num,
> > > @@ -546,5 +537,5 @@ int main(int argc, char **argv)
> > > free(env.test_selector.num_set);
> > > free(env.subtest_selector.num_set);
> > >
> > > - return error_cnt ? EXIT_FAILURE : EXIT_SUCCESS;
> > > + return env.fail_cnt ? EXIT_FAILURE : EXIT_SUCCESS;
> > > }
> > > diff --git a/tools/testing/selftests/bpf/test_progs.h b/tools/testing/selftests/bpf/test_progs.h
> > > index 9defd35cb6c0..7b05921784a4 100644
> > > --- a/tools/testing/selftests/bpf/test_progs.h
> > > +++ b/tools/testing/selftests/bpf/test_progs.h
> > > @@ -38,7 +38,23 @@ typedef __u16 __sum16;
> > > #include "trace_helpers.h"
> > > #include "flow_dissector_load.h"
> > >
> > > -struct prog_test_def;
> > > +struct prog_test_def {
> > > + const char *test_name;
> > > + int test_num;
> > > + void (*run_test)(void);
> > > + bool force_log;
> > > + bool tested;
> > > +
> > > + const char *subtest_name;
> > > + int subtest_num;
> > > +
> > > + int succ_cnt;
> > > + int fail_cnt;
> >
> > So I'm neutral on this rename, I even considered it myself initially.
> > But keep in mind, that succ/fail in env means number of tests, while
> > test->succ/fail means number of assertions. We don't report total
> > number of failed checks anymore, so it doesn't matter, but if we ever
> > want to keep track of that at env level, it will be very confusing and
> > inconvenient.
> Point taken, I didn't think about it, let me undo the rename. I'll
> try to add a comment instead to highlight the difference.
>
> > > +
> > > + /* store counts before subtest started */
> > > + int old_succ_cnt;
> > > + int old_fail_cnt;
> > > +};
> >
> > Did you move it here just to access env.test->succ_cnt in _CHECK()?
> > Maybe add test__success() counterpart to test__fail() instead?
> Yeah, good point, will do.
>
> > >
> > > struct test_selector {
> > > const char *name;
> > > @@ -67,13 +83,13 @@ struct test_env {
> > > int skip_cnt; /* skipped tests */
> > > };
> > >
> > > -extern int error_cnt;
> > > -extern int pass_cnt;
> > > extern struct test_env env;
> > >
> > > extern void test__force_log();
> > > extern bool test__start_subtest(const char *name);
> > > extern void test__skip(void);
> > > +#define test__fail() __test__fail(__FILE__, __LINE__)
> > > +extern void __test__fail(const char *file, int line);
> >
> > Given my comment above about too verbose logging, I'd say let's keep
> > this simple and have just
> >
> > extern void test__fail()
> >
> > And let _CHECK log file:line.
> See above about test__fail without _CHECK. Maybe we should do QCHECK
> as you suggested in the other email.
>
> So those lonely:
>
> if (err) {
> error_cnt++;
> return;
> }
>
> checks can instead be converted to:
>
> if (QCHECK(err))
> return;
>
> Let me play with it a bit and see how it goes.
Yeah, give it a go. Try keeping file:line logging in macro, where it's
more natural, IMO.
>
> > >
> > > #define MAGIC_BYTES 123
> > >
> > > @@ -96,11 +112,11 @@ extern struct ipv6_packet pkt_v6;
> > > #define _CHECK(condition, tag, duration, format...) ({ \
> > > int __ret = !!(condition); \
> > > if (__ret) { \
> > > - error_cnt++; \
> > > + test__fail(); \
> > > printf("%s:FAIL:%s ", __func__, tag); \
> > > printf(format); \
> > > } else { \
> > > - pass_cnt++; \
> > > + env.test->succ_cnt++; \
> > > printf("%s:PASS:%s %d nsec\n", \
> > > __func__, tag, duration); \
> > > } \
> > > --
> > > 2.23.0.rc1.153.gdeed80330f-goog
> > >
Powered by blists - more mailing lists