The BPF program for the snprintf selftest runs on all syscall entries.
On busy multicore systems this can cause concurrency issues.
For example it was observed that sometimes the userspace part of the
test reads " 4 0000" instead of " 4 000" (extra '0' at the end)
which seems to happen just before snprintf on another core sets
end[-1] = '\0'.
This patch adds a pid filter to the test to ensure that no
bpf_snprintf() will write over the test's output buffers while the
userspace reads the values.
Fixes: c2e39c6bdc7e ("selftests/bpf: Add a series of tests for bpf_snprintf")
Reported-by: Andrii Nakryiko <[email protected]>
Signed-off-by: Florent Revest <[email protected]>
---
tools/testing/selftests/bpf/prog_tests/snprintf.c | 2 ++
tools/testing/selftests/bpf/progs/test_snprintf.c | 5 +++++
2 files changed, 7 insertions(+)
diff --git a/tools/testing/selftests/bpf/prog_tests/snprintf.c b/tools/testing/selftests/bpf/prog_tests/snprintf.c
index a958c22aec75..dffbcaa1ec98 100644
--- a/tools/testing/selftests/bpf/prog_tests/snprintf.c
+++ b/tools/testing/selftests/bpf/prog_tests/snprintf.c
@@ -43,6 +43,8 @@ void test_snprintf_positive(void)
if (!ASSERT_OK_PTR(skel, "skel_open"))
return;
+ skel->bss->pid = getpid();
+
if (!ASSERT_OK(test_snprintf__attach(skel), "skel_attach"))
goto cleanup;
diff --git a/tools/testing/selftests/bpf/progs/test_snprintf.c b/tools/testing/selftests/bpf/progs/test_snprintf.c
index 951a0301c553..e35129bea0a0 100644
--- a/tools/testing/selftests/bpf/progs/test_snprintf.c
+++ b/tools/testing/selftests/bpf/progs/test_snprintf.c
@@ -5,6 +5,8 @@
#include <bpf/bpf_helpers.h>
#include <bpf/bpf_tracing.h>
+__u32 pid = 0;
+
char num_out[64] = {};
long num_ret = 0;
@@ -42,6 +44,9 @@ int handler(const void *ctx)
static const char str1[] = "str1";
static const char longstr[] = "longstr";
+ if ((int)bpf_get_current_pid_tgid() != pid)
+ return 0;
+
/* Integer types */
num_ret = BPF_SNPRINTF(num_out, sizeof(num_out),
"%d %u %x %li %llu %lX",
--
2.31.1.498.g6c1eba8ee3d-goog
On Wed, Apr 28, 2021 at 8:25 AM Florent Revest <[email protected]> wrote:
>
> The BPF program for the snprintf selftest runs on all syscall entries.
> On busy multicore systems this can cause concurrency issues.
>
> For example it was observed that sometimes the userspace part of the
> test reads " 4 0000" instead of " 4 000" (extra '0' at the end)
> which seems to happen just before snprintf on another core sets
> end[-1] = '\0'.
>
> This patch adds a pid filter to the test to ensure that no
> bpf_snprintf() will write over the test's output buffers while the
> userspace reads the values.
>
> Fixes: c2e39c6bdc7e ("selftests/bpf: Add a series of tests for bpf_snprintf")
> Reported-by: Andrii Nakryiko <[email protected]>
> Signed-off-by: Florent Revest <[email protected]>
> ---
Applied to bpf tree. Thanks.
> tools/testing/selftests/bpf/prog_tests/snprintf.c | 2 ++
> tools/testing/selftests/bpf/progs/test_snprintf.c | 5 +++++
> 2 files changed, 7 insertions(+)
>
> diff --git a/tools/testing/selftests/bpf/prog_tests/snprintf.c b/tools/testing/selftests/bpf/prog_tests/snprintf.c
> index a958c22aec75..dffbcaa1ec98 100644
> --- a/tools/testing/selftests/bpf/prog_tests/snprintf.c
> +++ b/tools/testing/selftests/bpf/prog_tests/snprintf.c
> @@ -43,6 +43,8 @@ void test_snprintf_positive(void)
> if (!ASSERT_OK_PTR(skel, "skel_open"))
> return;
>
> + skel->bss->pid = getpid();
> +
> if (!ASSERT_OK(test_snprintf__attach(skel), "skel_attach"))
> goto cleanup;
>
> diff --git a/tools/testing/selftests/bpf/progs/test_snprintf.c b/tools/testing/selftests/bpf/progs/test_snprintf.c
> index 951a0301c553..e35129bea0a0 100644
> --- a/tools/testing/selftests/bpf/progs/test_snprintf.c
> +++ b/tools/testing/selftests/bpf/progs/test_snprintf.c
> @@ -5,6 +5,8 @@
> #include <bpf/bpf_helpers.h>
> #include <bpf/bpf_tracing.h>
>
> +__u32 pid = 0;
> +
> char num_out[64] = {};
> long num_ret = 0;
>
> @@ -42,6 +44,9 @@ int handler(const void *ctx)
> static const char str1[] = "str1";
> static const char longstr[] = "longstr";
>
> + if ((int)bpf_get_current_pid_tgid() != pid)
> + return 0;
> +
> /* Integer types */
> num_ret = BPF_SNPRINTF(num_out, sizeof(num_out),
> "%d %u %x %li %llu %lX",
> --
> 2.31.1.498.g6c1eba8ee3d-goog
>
Hello:
This patch was applied to bpf/bpf.git (refs/heads/master):
On Wed, 28 Apr 2021 17:25:01 +0200 you wrote:
> The BPF program for the snprintf selftest runs on all syscall entries.
> On busy multicore systems this can cause concurrency issues.
>
> For example it was observed that sometimes the userspace part of the
> test reads " 4 0000" instead of " 4 000" (extra '0' at the end)
> which seems to happen just before snprintf on another core sets
> end[-1] = '\0'.
>
> [...]
Here is the summary with links:
- [bpf-next] selftests/bpf: Fix the snprintf test
https://git.kernel.org/bpf/bpf/c/f80f88f0e2f2
You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html