Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.37 as permitted sender) client-ip=23.128.96.37;
MIME-Version: 1.0
References: <20231005230851.3666908-1-irogers@google.com> <20231005230851.3666908-6-irogers@google.com>
 <CAM9d7chmVRrFgEZMYk3EWG+1wjXqLC3suu-xrX64hmfBSAFi0A@mail.gmail.com>
In-Reply-To: <CAM9d7chmVRrFgEZMYk3EWG+1wjXqLC3suu-xrX64hmfBSAFi0A@mail.gmail.com>
From:   Ian Rogers <irogers@google.com>
Date:   Mon, 9 Oct 2023 09:13:05 -0700
Message-ID: <CAP-5=fXbU_hckJp=evE5ja7BA4sS30E6hHryXXAEvwXs=1qkBA@mail.gmail.com>
Subject: Re: [PATCH v2 05/18] perf bench uprobe: Fix potential use of memory
 after free
To:     Namhyung Kim <namhyung@kernel.org>
Cc:     Nathan Chancellor <nathan@kernel.org>,
        Nick Desaulniers <ndesaulniers@google.com>,
        Tom Rix <trix@redhat.com>,
        Peter Zijlstra <peterz@infradead.org>,
        Ingo Molnar <mingo@redhat.com>,
        Arnaldo Carvalho de Melo <acme@kernel.org>,
        Mark Rutland <mark.rutland@arm.com>,
        Alexander Shishkin <alexander.shishkin@linux.intel.com>,
        Jiri Olsa <jolsa@kernel.org>,
        Adrian Hunter <adrian.hunter@intel.com>,
        Yicong Yang <yangyicong@hisilicon.com>,
        Jonathan Cameron <jonathan.cameron@huawei.com>,
        Yang Jihong <yangjihong1@huawei.com>,
        Kan Liang <kan.liang@linux.intel.com>,
        Ming Wang <wangming01@loongson.cn>,
        Huacai Chen <chenhuacai@kernel.org>,
        Sean Christopherson <seanjc@google.com>,
        K Prateek Nayak <kprateek.nayak@amd.com>,
        Yanteng Si <siyanteng@loongson.cn>,
        Yuan Can <yuancan@huawei.com>,
        Ravi Bangoria <ravi.bangoria@amd.com>,
        James Clark <james.clark@arm.com>, llvm@lists.linux.dev,
        linux-kernel@vger.kernel.org, linux-perf-users@vger.kernel.org,
        bpf@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Sun, Oct 8, 2023 at 10:51=E2=80=AFPM Namhyung Kim <namhyung@kernel.org> =
wrote:
>
> On Thu, Oct 5, 2023 at 4:09=E2=80=AFPM Ian Rogers <irogers@google.com> wr=
ote:
> >
> > Found by clang-tidy:
> > ```
> > bench/uprobe.c:98:3: warning: Use of memory after it is freed [clang-an=
alyzer-unix.Malloc]
> >                 bench_uprobe_bpf__destroy(skel);
> > ```
> >
> > Signed-off-by: Ian Rogers <irogers@google.com>
>
> I'm ok with the change but I think it won't call
> bench_uprobe__teardown_bpf_skel() if the setup function
> returns a negative value.  Maybe we also need to set the
> err in the default case of `switch (bench)` statement.

Yes, the analysis (I'll put it below) assumes that the err can be
positive yielding destroy being called twice, the second creating a
use-after-free. I think it is worth cleaning the code up and making
the analyzer's job easier.

Thanks,
Ian

```
bench/uprobe.c:98:3: warning: Use of memory after it is freed
[clang-analyzer-unix.Malloc]
               bench_uprobe_bpf__destroy(skel);
               ^
tools/perf/bench/uprobe.c:197:9: note: Calling 'bench_uprobe'
       return bench_uprobe(argc, argv, BENCH_UPROBE__TRACE_PRINTK);
              ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
tools/perf/bench/uprobe.c:150:6: note: 'bench' is not equal to
BENCH_UPROBE__BASELINE
       if (bench !=3D BENCH_UPROBE__BASELINE &&
bench_uprobe__setup_bpf_skel(bench) < 0)
           ^~~~~
tools/perf/bench/uprobe.c:150:6: note: Left side of '&&' is true
tools/perf/bench/uprobe.c:150:41: note: Calling 'bench_uprobe__setup_bpf_sk=
el'
       if (bench !=3D BENCH_UPROBE__BASELINE &&
bench_uprobe__setup_bpf_skel(bench) < 0)

^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
tools/perf/bench/uprobe.c:68:7: note: 'skel' is non-null
       if (!skel) {
            ^~~~
tools/perf/bench/uprobe.c:68:2: note: Taking false branch
       if (!skel) {
       ^
tools/perf/bench/uprobe.c:74:6: note: Assuming 'err' is not equal to 0
       if (err) {
           ^~~
tools/perf/bench/uprobe.c:74:2: note: Taking true branch
       if (err) {
       ^
tools/perf/bench/uprobe.c:76:3: note: Control jumps to line 91
               goto cleanup;
               ^
tools/perf/bench/uprobe.c:91:2: note: Calling 'bench_uprobe_bpf__destroy'
       bench_uprobe_bpf__destroy(skel);
       ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
tools/perf/util/bpf_skel/bench_uprobe.skel.h:44:6: note: Assuming
'obj' is non-null
       if (!obj)
           ^~~~
tools/perf/util/bpf_skel/bench_uprobe.skel.h:44:2: note: Taking false branc=
h
       if (!obj)
       ^
tools/perf/util/bpf_skel/bench_uprobe.skel.h:46:6: note: Assuming
field 'skeleton' is null
       if (obj->skeleton)
           ^~~~~~~~~~~~~
tools/perf/util/bpf_skel/bench_uprobe.skel.h:46:2: note: Taking false branc=
h
       if (obj->skeleton)
       ^
tools/perf/util/bpf_skel/bench_uprobe.skel.h:48:2: note: Memory is released
       free(obj);
       ^~~~~~~~~
tools/perf/bench/uprobe.c:91:2: note: Returning; memory was released
via 1st parameter
       bench_uprobe_bpf__destroy(skel);
       ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
tools/perf/bench/uprobe.c:150:41: note: Returning; memory was released
       if (bench !=3D BENCH_UPROBE__BASELINE &&
bench_uprobe__setup_bpf_skel(bench) < 0)

^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
tools/perf/bench/uprobe.c:150:41: note: Assuming the condition is false
       if (bench !=3D BENCH_UPROBE__BASELINE &&
bench_uprobe__setup_bpf_skel(bench) < 0)

^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
tools/perf/bench/uprobe.c:150:2: note: Taking false branch
       if (bench !=3D BENCH_UPROBE__BASELINE &&
bench_uprobe__setup_bpf_skel(bench) < 0)
       ^
tools/perf/bench/uprobe.c:155:14: note: Assuming 'i' is >=3D 'loops'
       for (i =3D 0; i < loops; i++) {
                   ^~~~~~~~~
tools/perf/bench/uprobe.c:155:2: note: Loop condition is false.
Execution continues on line 159
       for (i =3D 0; i < loops; i++) {
       ^
tools/perf/bench/uprobe.c:164:2: note: Control jumps to 'case 1:'  at line =
169
       switch (bench_format) {
       ^
tools/perf/bench/uprobe.c:171:3: note:  Execution continues on line 179
               break;
               ^
tools/perf/bench/uprobe.c:179:6: note: 'bench' is not equal to
BENCH_UPROBE__BASELINE
       if (bench !=3D BENCH_UPROBE__BASELINE)
           ^~~~~
tools/perf/bench/uprobe.c:179:2: note: Taking true branch
       if (bench !=3D BENCH_UPROBE__BASELINE)
       ^
tools/perf/bench/uprobe.c:180:3: note: Calling 'bench_uprobe__teardown_bpf_=
skel'
               bench_uprobe__teardown_bpf_skel();
               ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
tools/perf/bench/uprobe.c:97:6: note: 'skel' is non-null
       if (skel) {
           ^~~~
tools/perf/bench/uprobe.c:97:2: note: Taking true branch
       if (skel) {
       ^
tools/perf/bench/uprobe.c:98:3: note: Use of memory after it is freed
               bench_uprobe_bpf__destroy(skel);
               ^                         ~~~~
1 warning generated.
```

> Thanks,
> Namhyung
>
>
> > ---
> >  tools/perf/bench/uprobe.c | 1 +
> >  1 file changed, 1 insertion(+)
> >
> > diff --git a/tools/perf/bench/uprobe.c b/tools/perf/bench/uprobe.c
> > index 914c0817fe8a..5c71fdc419dd 100644
> > --- a/tools/perf/bench/uprobe.c
> > +++ b/tools/perf/bench/uprobe.c
> > @@ -89,6 +89,7 @@ static int bench_uprobe__setup_bpf_skel(enum bench_up=
robe bench)
> >         return err;
> >  cleanup:
> >         bench_uprobe_bpf__destroy(skel);
> > +       skel =3D NULL;
> >         return err;
> >  }
> >
> > --
> > 2.42.0.609.gbb76f46606-goog
> >