On Tue, Oct 26, 2021 at 9:43 PM Gabriel Krisman Bertazi
<krisman@xxxxxxxxxxxxx> wrote:
>
> fanotify20 is a new test validating the FAN_FS_ERROR file system error
> event. This adds some basic structure for the next patches.
>
> The strategy for error reporting testing in fanotify20 goes like this:
>
> - Generate a broken filesystem
> - Start FAN_FS_ERROR monitoring group
> - Make the file system notice the error through ordinary operations
> - Observe the event generated
>
> Signed-off-by: Gabriel Krisman Bertazi <krisman@xxxxxxxxxxxxx>
>
> ---
> Changes since v1:
> - Move defines to header file.
> ---
> testcases/kernel/syscalls/fanotify/.gitignore | 1 +
> testcases/kernel/syscalls/fanotify/fanotify.h | 3 +
> .../kernel/syscalls/fanotify/fanotify20.c | 128 ++++++++++++++++++
> 3 files changed, 132 insertions(+)
> create mode 100644 testcases/kernel/syscalls/fanotify/fanotify20.c
>
> diff --git a/testcases/kernel/syscalls/fanotify/.gitignore b/testcases/kernel/syscalls/fanotify/.gitignore
> index 9554b16b196e..c99e6fff76d6 100644
> --- a/testcases/kernel/syscalls/fanotify/.gitignore
> +++ b/testcases/kernel/syscalls/fanotify/.gitignore
> @@ -17,4 +17,5 @@
> /fanotify17
> /fanotify18
> /fanotify19
> +/fanotify20
> /fanotify_child
> diff --git a/testcases/kernel/syscalls/fanotify/fanotify.h b/testcases/kernel/syscalls/fanotify/fanotify.h
> index b2b56466d028..8828b53532a2 100644
> --- a/testcases/kernel/syscalls/fanotify/fanotify.h
> +++ b/testcases/kernel/syscalls/fanotify/fanotify.h
> @@ -124,6 +124,9 @@ static inline int safe_fanotify_mark(const char *file, const int lineno,
> #ifndef FAN_OPEN_EXEC_PERM
> #define FAN_OPEN_EXEC_PERM 0x00040000
> #endif
> +#ifndef FAN_FS_ERROR
> +#define FAN_FS_ERROR 0x00008000
> +#endif
>
> /* Flags required for unprivileged user group */
> #define FANOTIFY_REQUIRED_USER_INIT_FLAGS (FAN_REPORT_FID)
> diff --git a/testcases/kernel/syscalls/fanotify/fanotify20.c b/testcases/kernel/syscalls/fanotify/fanotify20.c
> new file mode 100644
> index 000000000000..7a522aad4386
> --- /dev/null
> +++ b/testcases/kernel/syscalls/fanotify/fanotify20.c
> @@ -0,0 +1,128 @@
> +// SPDX-License-Identifier: GPL-2.0-or-later
> +/*
> + * Copyright (c) 2021 Collabora Ltd.
> + *
> + * Author: Gabriel Krisman Bertazi <gabriel@xxxxxxxxxx>
> + * Based on previous work by Amir Goldstein <amir73il@xxxxxxxxx>
> + */
> +
> +/*\
> + * [Description]
> + * Check fanotify FAN_ERROR_FS events triggered by intentionally
> + * corrupted filesystems:
> + *
> + * - Generate a broken filesystem
> + * - Start FAN_FS_ERROR monitoring group
> + * - Make the file system notice the error through ordinary operations
> + * - Observe the event generated
> + */
> +
> +#define _GNU_SOURCE
> +#include "config.h"
> +
> +#include <stdio.h>
> +#include <sys/types.h>
> +#include <errno.h>
> +#include <string.h>
> +#include <sys/mount.h>
> +#include <sys/syscall.h>
> +#include "tst_test.h"
> +#include <sys/fanotify.h>
> +#include <sys/types.h>
> +#include <fcntl.h>
> +
> +#ifdef HAVE_SYS_FANOTIFY_H
> +#include "fanotify.h"
> +
> +#define BUF_SIZE 256
> +static char event_buf[BUF_SIZE];
> +int fd_notify;
> +
> +#define MOUNT_PATH "test_mnt"
> +
> +static struct test_case {
> + char *name;
> + void (*trigger_error)(void);
> +} testcases[] = {
> +};
> +
Does LTP accept .tcnt = 0 gracefully?
or maybe LTP project does not care much about failing tests during bisection?
> +int check_error_event_metadata(struct fanotify_event_metadata *event)
> +{
> + int fail = 0;
> +
> + if (event->mask != FAN_FS_ERROR) {
> + fail++;
> + tst_res(TFAIL, "got unexpected event %llx",
> + (unsigned long long)event->mask);
> + }
> +
> + if (event->fd != FAN_NOFD) {
> + fail++;
> + tst_res(TFAIL, "Weird FAN_FD %llx",
> + (unsigned long long)event->mask);
> + }
> + return fail;
> +}
> +
> +void check_event(char *buf, size_t len, const struct test_case *ex)
> +{
> + struct fanotify_event_metadata *event =
> + (struct fanotify_event_metadata *) buf;
> +
> + if (len < FAN_EVENT_METADATA_LEN) {
> + tst_res(TFAIL, "No event metadata found");
> + return;
> + }
> +
> + if (check_error_event_metadata(event))
> + return;
> +
> + tst_res(TPASS, "Successfully received: %s", ex->name);
> +}
> +
> +static void do_test(unsigned int i)
> +{
> + const struct test_case *tcase = &testcases[i];
> + size_t read_len;
> +
> + tcase->trigger_error();
> +
> + read_len = SAFE_READ(0, fd_notify, event_buf, BUF_SIZE);
> +
> + check_event(event_buf, read_len, tcase);
> +}
> +
> +static void setup(void)
> +{
> + REQUIRE_FANOTIFY_EVENTS_SUPPORTED_ON_FS(FAN_CLASS_NOTIF|FAN_REPORT_FID,
> + FAN_MARK_FILESYSTEM,
> + FAN_FS_ERROR, ".");
> +
> + fd_notify = SAFE_FANOTIFY_INIT(FAN_CLASS_NOTIF|FAN_REPORT_FID,
> + O_RDONLY);
> +
> + SAFE_FANOTIFY_MARK(fd_notify, FAN_MARK_ADD|FAN_MARK_FILESYSTEM,
> + FAN_FS_ERROR, AT_FDCWD, MOUNT_PATH);
I think it is better to have the mark add/remove inside do_test
This way when running fanotify -i 10 (which testers do)
we also get test coverage for add/remove of mark with FS_ERROR mask.
Thanks,
Amir.
