[CRIU] [PATCH v3 01/33] zdtm: Add userns00 test
Andrei Vagin
avagin at virtuozzo.com
Mon Feb 20 21:28:53 PST 2017
On Thu, Feb 16, 2017 at 03:06:43PM +0300, Kirill Tkhai wrote:
> Create two children, and unshare() user_ns in one of them (C1).
> The second child creates one more process, which switches to C1's
> namespace and unshares.
>
> v3: Unshare net ns in grand child
>
> Signed-off-by: Kirill Tkhai <ktkhai at virtuozzo.com>
> ---
> test/zdtm/static/Makefile | 1
> test/zdtm/static/userns00.c | 295 ++++++++++++++++++++++++++++++++++++++++
> test/zdtm/static/userns00.desc | 1
> 3 files changed, 297 insertions(+)
> create mode 100644 test/zdtm/static/userns00.c
> create mode 100644 test/zdtm/static/userns00.desc
>
> diff --git a/test/zdtm/static/Makefile b/test/zdtm/static/Makefile
> index de35e8fe7..f7c46da8c 100644
> --- a/test/zdtm/static/Makefile
> +++ b/test/zdtm/static/Makefile
> @@ -175,6 +175,7 @@ TST_NOFILE := \
> sock_peercred \
> uffd-events \
> netns_sub \
> + userns00 \
> # jobctl00 \
>
> ifneq ($(SRCARCH),arm)
> diff --git a/test/zdtm/static/userns00.c b/test/zdtm/static/userns00.c
> new file mode 100644
> index 000000000..53925d644
> --- /dev/null
> +++ b/test/zdtm/static/userns00.c
> @@ -0,0 +1,295 @@
> +#define _GNU_SOURCE
> +#include <stdbool.h>
> +#include <string.h>
> +#include <fcntl.h>
> +#include <unistd.h>
> +#include <signal.h>
> +#include <stdio.h>
> +#include <sys/mount.h>
> +#include <sys/stat.h>
> +#include <sys/mman.h>
> +#include <sched.h>
> +#include <sys/wait.h>
> +#include <stdlib.h>
> +#include <limits.h>
> +#include <dirent.h>
> +
> +#include "zdtmtst.h"
> +#include "lock.h"
> +
> +/*
> + * The test creates the following items.
> + *
> + * Processes hierarhy with their namespaces (in brackets):
> + * Parent (ns_p)
> + * / \
> + * (ns_c1) Child1 Child2 (ns_p)
> + * |
> + * GrandChild (ns_gc)
> + *
> + * Namespaces hierarhy:
> + * ns_p
> + * |
> + * ns_c1
> + * |
> + * ns_gc
> + */
> +const char *test_doc = "Check user namespaces remain the same over process tree";
> +const char *test_author = "Kirill Tkhai <ktkhai at virtuozzo.com>";
> +
> +enum {
> + FUTEX_INITIALIZED = 0,
> + CHILD1_CREATED,
> + GRAND_CHILD_PID_WRITTEN,
> + GRAND_CHILD_CREATED,
> + GRAND_CHILD_SETUP,
> + POST_RESTORE_CHECK,
> + EMERGENCY_ABORT,
> +};
> +
> +volatile pid_t *grand_child_pid;
> +futex_t *futex;
> +
> +int get_user_ns(pid_t pid, unsigned int *ns_id)
> +{
> + char path[PATH_MAX], buf[PATH_MAX];
> + int len;
> +
> + sprintf(path, "/proc/%d/ns/user", pid);
> + len = readlink(path, buf, PATH_MAX);
> + if (len < 0) {
> + pr_perror("Can't read link %s\n", path);
> + return -1;
> + }
> +
readlink() may return PATH_MAX, and buf[len] will access out of buffer
memory.
> + buf[len] = '\0';
> + if (sscanf(buf, "user:[%u", ns_id) < 1) {
> + pr_err("Can't get id: %s\n", buf);
> + return -1;
> + }
> +
> + return 0;
> +}
> +
> +int write_map(pid_t pid, char *map)
> +{
> + char path[PATH_MAX];
> + int fd, ret;
> +
> + sprintf(path, "/proc/%d/%s", pid, map);
> + fd = open(path, O_WRONLY);
> + if (fd < 0) {
> + fail("Can't open");
> + return -1;
> + }
> + ret = write(fd, "0 0 1\n", 6);
I think the mapping has to be more complect and should not be "one to one".
For example:
0 10 1000
1000 1000 1000
> + if (ret != 6) {
> + fail("Can't write");
> + return -1;
> + }
> + close(fd);
> +
> + return 0;
> +}
> +
> +/* Child1 creates its own namespace */
> +int child1(void)
> +{
> + int ret;
> +
> + ret = unshare(CLONE_NEWUSER);
> + if (ret < 0) {
> + pr_perror("unshare");
> + futex_set_and_wake(futex, EMERGENCY_ABORT);
> + return 1;
> + }
> +
> + futex_set_and_wake(futex, CHILD1_CREATED);
> + futex_wait_while_lt(futex, GRAND_CHILD_CREATED);
> +
> + if (write_map(*grand_child_pid, "uid_map") < 0 ||
> + write_map(*grand_child_pid, "gid_map") < 0) {
> + fail("write map");
> + futex_set_and_wake(futex, EMERGENCY_ABORT);
> + return 2;
> + }
> +
> + futex_set_and_wake(futex, GRAND_CHILD_SETUP);
> + futex_wait_while_lt(futex, POST_RESTORE_CHECK);
> +
> + return 0;
> +}
> +
> +/* GrandChild switches to Child1 namespace and unshares */
> +int grand_child(pid_t pid1)
> +{
> + char path[PATH_MAX];
> + int fd = -1, ret;
> +
> + futex_wait_while_lt(futex, GRAND_CHILD_PID_WRITTEN);
> +
> + sprintf(path, "/proc/%d/ns/user", pid1);
> + fd = open(path, O_RDONLY);
> + if (fd < 0) {
> + pr_perror("open");
> + ret = 1;
> + goto err;
> + }
> +
> + if (setns(fd, CLONE_NEWUSER) < 0) {
> + pr_perror("setns()");
> + ret = 2;
> + goto err;
> + }
> + close(fd);
> + fd = -1;
> +
> + if (setuid(0) || setgid(0)) {
> + pr_perror("setxids");
> + ret = 3;
> + goto err;
> + }
> +
> + if (unshare(CLONE_NEWUSER) < 0) {
> + pr_perror("unshare");
> + ret = 3;
> + goto err;
> + }
> +
> + if (unshare(CLONE_NEWNET) < 0) {
> + pr_perror("unshare");
> + ret = 4;
> + goto err;
> + }
> +
can we set a non-zero uid ana gid here
> + futex_set_and_wake(futex, GRAND_CHILD_CREATED);
> + futex_wait_while_lt(futex, POST_RESTORE_CHECK);
and check here that uid and gid are restored correctly
> +
> + return 0;
> +err:
> + futex_set_and_wake(futex, EMERGENCY_ABORT);
> + if (fd >= 0)
> + close(fd);
> + return ret;
> +}
> +
> +/*
> + * Child2 remains in the namespace of the parent task,
> + * while its descendant GrandChild enters to Child1's
> + * namespace and unshares.
> + */
> +int child2(pid_t pid1)
> +{
> + int status, ret;
> + pid_t gc_pid;
> +
> + gc_pid = fork();
> + if (gc_pid < 0) {
> + pr_perror("Can't fork");
> + ret = 1;
> + goto err;
> + } else if (gc_pid == 0)
> + exit(grand_child(pid1));
> +
> + *grand_child_pid = gc_pid;
> + futex_set_and_wake(futex, GRAND_CHILD_PID_WRITTEN);
> + futex_wait_while_lt(futex, POST_RESTORE_CHECK);
> +
> + if (wait(&status) != *grand_child_pid) {
> + pr_perror("Failed to wait grand child");
> + ret = 3;
> + goto err;
> + }
> +
> + if (WEXITSTATUS(status)) {
> + pr_err("Grand child exited with %d\n", WEXITSTATUS(status));
> + ret = 4;
> + goto err;
> + }
> +
> + return 0;
> +err:
> + futex_set_and_wake(futex, EMERGENCY_ABORT);
> + if (*grand_child_pid > 0)
> + wait(&status);
> + return ret;
> +}
> +
> +
> +int main(int argc, char **argv)
> +{
> + pid_t my_pid, pid1 = -1, pid2 = -1;
> + unsigned int ns_p, ns_c1, ns_c2;
> + int status;
> +
> + test_init(argc, argv);
> + futex = mmap(NULL, sizeof(*futex) + sizeof(*grand_child_pid), PROT_WRITE | PROT_READ,
> + MAP_SHARED | MAP_ANONYMOUS, -1, 0);
> + grand_child_pid = (void *)futex + sizeof(*futex);
> + if (futex == MAP_FAILED) {
> + fail("mmap futex\n");
> + return 1;
> + }
> + futex_init(futex);
> + my_pid = getpid();
> +
> + pid1 = fork();
> + if (pid1 == -1) {
> + fail("fork");
> + return 1;
> + } else if (pid1 == 0)
> + exit(child1());
> +
> + futex_wait_while_lt(futex, CHILD1_CREATED);
> +
> + if (write_map(pid1, "uid_map") < 0 ||
> + write_map(pid1, "gid_map") < 0) {
> + fail("write map");
> + goto err;
> + }
> +
> + pid2 = fork();
> + if (pid2 == -1) {
> + fail("fork");
> + goto err;
> + } else if (pid2 == 0)
> + exit(child2(pid1));
> +
> + futex_wait_while_lt(futex, GRAND_CHILD_SETUP);
> + test_daemon();
> + test_waitsig();
> +
> + if (get_user_ns(my_pid, &ns_p) < 0 ||
> + get_user_ns(pid1, &ns_c1) < 0 ||
> + get_user_ns(pid2, &ns_c2) < 0) {
> + fail("Can't get user ns\n");
> + goto err;
> + }
> +
> + if (ns_p == ns_c1 || ns_p != ns_c2) {
> + fail("ns_p=%u, ns_c1=%u, ns_c2=%u\n", ns_p, ns_c1, ns_c2);
> + goto err;
> + }
> +
> + futex_set_and_wake(futex, POST_RESTORE_CHECK);
> +
> + errno = 0;
> + if (waitpid(pid1, &status, 0) < 0 || WEXITSTATUS(status)) {
> + fail("pid1: status=%d\n", WEXITSTATUS(status));
it is better to print status, because a process can be killed
> + goto err;
> + }
> +
> + if (waitpid(pid2, &status, 0) < 0 || WEXITSTATUS(status)) {
> + fail("pid2: status=%d\n", WEXITSTATUS(status));
> + goto err;
> + }
> + pass();
> + return 0;
> +err:
> + futex_set_and_wake(futex, EMERGENCY_ABORT);
> + if (pid1 > 0)
> + wait(&status);
> + if (pid2 > 0)
> + wait(&status);
> + return 1;
> +}
> diff --git a/test/zdtm/static/userns00.desc b/test/zdtm/static/userns00.desc
> new file mode 100644
> index 000000000..1f8bec515
> --- /dev/null
> +++ b/test/zdtm/static/userns00.desc
> @@ -0,0 +1 @@
> +{'flavor': 'uns', 'flags': 'suid noauto'}
>
> _______________________________________________
> CRIU mailing list
> CRIU at openvz.org
> https://lists.openvz.org/mailman/listinfo/criu
More information about the CRIU
mailing list