samples: show race-free pidfd metadata access

This is a sample program showing userspace how to get race-free access
to process metadata from a pidfd.  It is rather easy to do and userspace
can actually simply reuse code that currently parses a process's status
file in procfs.
The program can easily be extended into a generic helper suitable for
inclusion in a libc to make it even easier for userspace to gain metadata
access.

Since this came up in a discussion because this API is going to be used
in various service managers: A lot of programs will have a whitelist
seccomp filter that returns <some-errno> for all new syscalls.  This
means that programs might get confused if CLONE_PIDFD works but the
later pidfd_send_signal() syscall doesn't.  Hence, here's a ahead of
time check that pidfd_send_signal() is supported:

bool pidfd_send_signal_supported()
{
        int procfd = open("/proc/self", O_DIRECTORY | O_RDONLY | O_CLOEXEC);
        if (procfd < 0)
                return false;

        /*
         * A process is always allowed to signal itself so
         * pidfd_send_signal() should never fail this test. If it does
         * it must mean it is not available, blocked by an LSM, seccomp,
         * or other.
         */
        return pidfd_send_signal(procfd, 0, NULL, 0) == 0;
}

Signed-off-by: Christian Brauner <christian@brauner.io>
Co-developed-by: Jann Horn <jannh@google.com>
Signed-off-by: Jann Horn <jannh@google.com>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: David Howells <dhowells@redhat.com>
Cc: "Michael Kerrisk (man-pages)" <mtk.manpages@gmail.com>
Cc: Andy Lutomirsky <luto@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Aleksa Sarai <cyphar@cyphar.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>

3 files changed, 119 insertions, 1 deletions

diff --git a/samples/Makefile b/samples/Makefile
index b1142a958811..fadadb1c3b05 100644
--- a/samples/Makefile
+++ b/samples/Makefile
@@ -3,4 +3,4 @@
 obj-$(CONFIG_SAMPLES)   += kobject/ kprobes/ trace_events/ livepatch/ \
                            hw_breakpoint/ kfifo/ kdb/ hidraw/ rpmsg/ seccomp/ \
                            configfs/ connector/ v4l/ trace_printk/ \
-                           vfio-mdev/ statx/ qmi/ binderfs/
+                           vfio-mdev/ statx/ qmi/ binderfs/ pidfd/
diff --git a/samples/pidfd/Makefile b/samples/pidfd/Makefile
new file mode 100644
index 000000000000..0ff97784177a
--- /dev/null
+++ b/samples/pidfd/Makefile
@@ -0,0 +1,6 @@
+# SPDX-License-Identifier: GPL-2.0
+
+hostprogs-y := pidfd-metadata
+always := $(hostprogs-y)
+HOSTCFLAGS_pidfd-metadata.o += -I$(objtree)/usr/include
+all: pidfd-metadata
diff --git a/samples/pidfd/pidfd-metadata.c b/samples/pidfd/pidfd-metadata.c
new file mode 100644
index 000000000000..640f5f757c57
--- /dev/null
+++ b/samples/pidfd/pidfd-metadata.c
@@ -0,0 +1,112 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#define _GNU_SOURCE
+#include <err.h>
+#include <errno.h>
+#include <fcntl.h>
+#include <inttypes.h>
+#include <limits.h>
+#include <sched.h>
+#include <signal.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/stat.h>
+#include <sys/syscall.h>
+#include <sys/types.h>
+#include <sys/wait.h>
+#include <unistd.h>
+
+#ifndef CLONE_PIDFD
+#define CLONE_PIDFD 0x00001000
+#endif
+
+static int do_child(void *args)
+{
+        printf("%d\n", getpid());
+        _exit(EXIT_SUCCESS);
+}
+
+static pid_t pidfd_clone(int flags, int *pidfd)
+{
+        size_t stack_size = 1024;
+        char *stack[1024] = { 0 };
+
+#ifdef __ia64__
+        return __clone2(do_child, stack, stack_size, flags | SIGCHLD, NULL, pidfd);
+#else
+        return clone(do_child, stack + stack_size, flags | SIGCHLD, NULL, pidfd);
+#endif
+}
+
+static inline int sys_pidfd_send_signal(int pidfd, int sig, siginfo_t *info,
+                                        unsigned int flags)
+{
+        return syscall(__NR_pidfd_send_signal, pidfd, sig, info, flags);
+}
+
+static int pidfd_metadata_fd(pid_t pid, int pidfd)
+{
+        int procfd, ret;
+        char path[100];
+
+        snprintf(path, sizeof(path), "/proc/%d", pid);
+        procfd = open(path, O_DIRECTORY | O_RDONLY | O_CLOEXEC);
+        if (procfd < 0) {
+                warn("Failed to open %s\n", path);
+                return -1;
+        }
+
+        /*
+         * Verify that the pid has not been recycled and our /proc/<pid> handle
+         * is still valid.
+         */
+        ret = sys_pidfd_send_signal(pidfd, 0, NULL, 0);
+        if (ret < 0) {
+                switch (errno) {
+                case EPERM:
+                        /* Process exists, just not allowed to signal it. */
+                        break;
+                default:
+                        warn("Failed to signal process\n");
+                        close(procfd);
+                        procfd = -1;
+                }
+        }
+
+        return procfd;
+}
+
+int main(int argc, char *argv[])
+{
+        int pidfd = 0, ret = EXIT_FAILURE;
+        char buf[4096] = { 0 };
+        pid_t pid;
+        int procfd, statusfd;
+        ssize_t bytes;
+
+        pid = pidfd_clone(CLONE_PIDFD, &pidfd);
+        if (pid < 0)
+                exit(ret);
+
+        procfd = pidfd_metadata_fd(pid, pidfd);
+        close(pidfd);
+        if (procfd < 0)
+                goto out;
+
+        statusfd = openat(procfd, "status", O_RDONLY | O_CLOEXEC);
+        close(procfd);
+        if (statusfd < 0)
+                goto out;
+
+        bytes = read(statusfd, buf, sizeof(buf));
+        if (bytes > 0)
+                bytes = write(STDOUT_FILENO, buf, bytes);
+        close(statusfd);
+        ret = EXIT_SUCCESS;
+
+out:
+        (void)wait(NULL);
+
+        exit(ret);
+}