Hi,
On 8/31/2023 7:46 PM, Daniel Borkmann wrote:
> On 8/31/23 1:00 PM, Jiri Olsa wrote:
>> Recent commit [1] broken d_path test, because now filp_close is not
>> called
>> directly from sys_close, but eventually later when the file is finally
>> released.
>>
>> As suggested by Hou Tao we don't need to re-hook the bpf program, but
>> just
>> instead we can use sys_close_range to trigger filp_close synchronously.
>>
>> [1] 021a160abf62 ("fs: use __fput_sync in close(2)")
>> Suggested-by: Hou Tao <houtao@xxxxxxxxxxxxxxx>
>> Signed-off-by: Jiri Olsa <jolsa@xxxxxxxxxx>
>> ---
>> tools/testing/selftests/bpf/prog_tests/d_path.c | 8 +++++++-
>> 1 file changed, 7 insertions(+), 1 deletion(-)
>>
>> diff --git a/tools/testing/selftests/bpf/prog_tests/d_path.c
>> b/tools/testing/selftests/bpf/prog_tests/d_path.c
>> index 911345c526e6..81e34a4a05d1 100644
>> --- a/tools/testing/selftests/bpf/prog_tests/d_path.c
>> +++ b/tools/testing/selftests/bpf/prog_tests/d_path.c
>> @@ -90,7 +90,11 @@ static int trigger_fstat_events(pid_t pid)
>> fstat(indicatorfd, &fileStat);
>> out_close:
>> - /* triggers filp_close */
>> + /* sys_close no longer triggers filp_close, but we can
>> + * call sys_close_range instead which still does
>> + */
>> +#define close(fd) close_range(fd, fd, 0)
>> +
>
> The BPF CI selftest build says:
>
> [...]
> TEST-OBJ [test_progs] lookup_key.test.o
> TEST-OBJ [test_progs] migrate_reuseport.test.o
> TEST-OBJ [test_progs] user_ringbuf.test.o
> /tmp/work/bpf/bpf/tools/testing/selftests/bpf/prog_tests/d_path.c:
> In function ‘trigger_fstat_events’:
>
> /tmp/work/bpf/bpf/tools/testing/selftests/bpf/prog_tests/d_path.c:96:19:
> error: implicit declaration of function ‘close_range’
> [-Werror=implicit-function-declaration]
> 96 | #define close(fd) close_range(fd, fd, 0)
> | ^~~~~~~~~~~
>
> /tmp/work/bpf/bpf/tools/testing/selftests/bpf/prog_tests/d_path.c:98:2:
> note: in expansion of macro ‘close’
> 98 | close(pipefd[0]);
> | ^~~~~
> TEST-OBJ [test_progs] task_pt_regs.test.o
> [...]
>
> Perhaps #include <linux/close_range.h> missing ?
Including <linux/close_range.h> doesn't help because it only defines two
macros.
I got the same error when testing locally. It seems close_range() was
introduced by glibc 2.34 [1] and it was defined in <unistd.h>, but the
version of glibc in my local VM is 2.29. I modify d_path locally to call
close_range through syscall():
diff --git a/tools/testing/selftests/bpf/prog_tests/d_path.c
b/tools/testing/selftests/bpf/prog_tests/d_path.c
index 81e34a4a05d1..c5811843ce7e 100644
--- a/tools/testing/selftests/bpf/prog_tests/d_path.c
+++ b/tools/testing/selftests/bpf/prog_tests/d_path.c
@@ -12,6 +12,14 @@
#include "test_d_path_check_rdonly_mem.skel.h"
#include "test_d_path_check_types.skel.h"
+#ifndef __NR_close_range
+#ifdef __alpha__
+#define __NR_close_range 546
+#else
+#define __NR_close_range 436
+#endif
+#endif
+
static int duration;
static struct {
@@ -93,7 +101,7 @@ static int trigger_fstat_events(pid_t pid)
/* sys_close no longer triggers filp_close, but we can
* call sys_close_range instead which still does
*/
-#define close(fd) close_range(fd, fd, 0)
+#define close(fd) syscall(__NR_close_range, fd, fd, 0)
[1]: 9b4feb630e8e arch: wire-up close_range()
>
>> close(pipefd[0]);
>> close(pipefd[1]);
>> close(sockfd);
>> @@ -98,6 +102,8 @@ static int trigger_fstat_events(pid_t pid)
>> close(devfd);
>> close(localfd);
>> close(indicatorfd);
>> +
>> +#undef close
>> return ret;
>> }
>>
