Messages in this thread Patch in this message | | | From | Jann Horn <> | Subject | [PATCH] time: Prevent union confusion from unexpected restart_syscall() | Date | Thu, 5 Jan 2023 14:44:03 +0100 |
| |
The nanosleep syscalls use the restart_block mechanism, with a quirk: The `type` and `rmtp`/`compat_rmtp` fields are set up unconditionally on syscall entry, while the rest of the restart_block is only set up in the unlikely case that the syscall is actually interrupted by a signal (or pseudo-signal) that doesn't have a signal handler.
If the restart_block was set up by a previous syscall (futex(..., FUTEX_WAIT, ...) or poll()) and hasn't been invalidated somehow since then, this will clobber some of the union fields used by futex_wait_restart()/do_restart_poll(). If userspace afterwards wrongly calls the restart_syscall syscall, futex_wait_restart()/do_restart_poll() will read struct fields that have been clobbered.
This doesn't actually lead to anything particularly interesting because none of the union fields contain trusted kernel data, and futex(..., FUTEX_WAIT, ...) and poll() aren't syscalls where it makes much sense to apply seccomp filters to their arguments. So the current consequences are just of the "if userspace does bad stuff, it can damage itself, and that's not a problem" flavor.
But still, it seems like a hazard for future developers, so invalidate the restart_block when partly setting it up in the nanosleep syscalls.
Signed-off-by: Jann Horn <jannh@google.com> --- reproducer, demonstrates nanosleep() clobbering the upper half of current->restart_block.poll.ufds (with TT_NATIVE==1) and current->restart_block.poll.nfds (with 42):
user@vm:~/restart_syscall$ cat restart_syscall_union_confusion.c #define _GNU_SOURCE #include <err.h> #include <errno.h> #include <stdlib.h> #include <time.h> #include <unistd.h> #include <stdio.h> #include <poll.h> #include <fcntl.h> #include <signal.h> #include <sys/mman.h> #include <sys/select.h> #include <sys/wait.h> #include <sys/syscall.h>
#define SYSCHK(x) ({ \ typeof(x) __res = (x); \ if (__res == (typeof(x))-1) \ err(1, "SYSCHK(" #x ")"); \ __res; \ })
int main(void) { int child = SYSCHK(fork()); if (child == 0) { struct pollfd *pollfds = SYSCHK(mmap((void*)0x100000000, 0x1000, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS|MAP_FIXED_NOREPLACE, -1, 0)); int dev_null_fd = SYSCHK(open("/dev/null", O_WRONLY)); for (int i=0; i<100; i++) pollfds[i] = (struct pollfd) { .fd = dev_null_fd, .events = POLLOUT };
errno = 0; int res = poll(NULL, 0, 2000); printf("poll = %d (%m)\n", res);
// this writes current->restart_block.nanosleep.{type,rmtp} struct timespec ts_sleep = { .tv_nsec = 1000 }; SYSCHK(nanosleep(&ts_sleep, (void*)42UL));
errno = 0; int ret = syscall(__NR_restart_syscall); if (ret == -1 && errno == EINTR) { printf("restart_syscall() returned EINTR, probably do_no_restart_syscall\n"); } else { printf("restart_syscall() = %d (%m)\n", ret); } for (int i=0; i<50; i++) printf("pollfds[%d].revents = 0x%x\n", i, pollfds[i].revents);
exit(0); } else { // parent sleep(1); printf("sending SIGSTOP\n"); kill(child, SIGSTOP); sleep(1); printf("sending SIGCONT\n"); kill(child, SIGCONT);
printf("waiting for child...\n"); int status; SYSCHK(waitpid(child, &status, 0)); } } user@vm:~/restart_syscall$ gcc -o restart_syscall_union_confusion restart_syscall_union_confusion.c user@vm:~/restart_syscall$ ./restart_syscall_union_confusion sending SIGSTOP sending SIGCONT waiting for child... poll = 0 (Success) restart_syscall() = 42 (Success) pollfds[0].revents = 0x4 pollfds[1].revents = 0x4 pollfds[2].revents = 0x4 pollfds[3].revents = 0x4 pollfds[4].revents = 0x4 pollfds[5].revents = 0x4 pollfds[6].revents = 0x4 pollfds[7].revents = 0x4 pollfds[8].revents = 0x4 pollfds[9].revents = 0x4 pollfds[10].revents = 0x4 pollfds[11].revents = 0x4 pollfds[12].revents = 0x4 pollfds[13].revents = 0x4 pollfds[14].revents = 0x4 pollfds[15].revents = 0x4 pollfds[16].revents = 0x4 pollfds[17].revents = 0x4 pollfds[18].revents = 0x4 pollfds[19].revents = 0x4 pollfds[20].revents = 0x4 pollfds[21].revents = 0x4 pollfds[22].revents = 0x4 pollfds[23].revents = 0x4 pollfds[24].revents = 0x4 pollfds[25].revents = 0x4 pollfds[26].revents = 0x4 pollfds[27].revents = 0x4 pollfds[28].revents = 0x4 pollfds[29].revents = 0x4 pollfds[30].revents = 0x4 pollfds[31].revents = 0x4 pollfds[32].revents = 0x4 pollfds[33].revents = 0x4 pollfds[34].revents = 0x4 pollfds[35].revents = 0x4 pollfds[36].revents = 0x4 pollfds[37].revents = 0x4 pollfds[38].revents = 0x4 pollfds[39].revents = 0x4 pollfds[40].revents = 0x4 pollfds[41].revents = 0x4 pollfds[42].revents = 0x0 pollfds[43].revents = 0x0 pollfds[44].revents = 0x0 pollfds[45].revents = 0x0 pollfds[46].revents = 0x0 pollfds[47].revents = 0x0 pollfds[48].revents = 0x0 pollfds[49].revents = 0x0
kernel/time/hrtimer.c | 2 ++ kernel/time/posix-stubs.c | 2 ++ kernel/time/posix-timers.c | 2 ++ 3 files changed, 6 insertions(+)
diff --git a/kernel/time/hrtimer.c b/kernel/time/hrtimer.c index 3ae661ab6260..e4f0e3b0c4f4 100644 --- a/kernel/time/hrtimer.c +++ b/kernel/time/hrtimer.c @@ -2126,6 +2126,7 @@ SYSCALL_DEFINE2(nanosleep, struct __kernel_timespec __user *, rqtp, if (!timespec64_valid(&tu)) return -EINVAL; + current->restart_block.fn = do_no_restart_syscall; current->restart_block.nanosleep.type = rmtp ? TT_NATIVE : TT_NONE; current->restart_block.nanosleep.rmtp = rmtp; return hrtimer_nanosleep(timespec64_to_ktime(tu), HRTIMER_MODE_REL, @@ -2147,6 +2148,7 @@ SYSCALL_DEFINE2(nanosleep_time32, struct old_timespec32 __user *, rqtp, if (!timespec64_valid(&tu)) return -EINVAL; + current->restart_block.fn = do_no_restart_syscall; current->restart_block.nanosleep.type = rmtp ? TT_COMPAT : TT_NONE; current->restart_block.nanosleep.compat_rmtp = rmtp; return hrtimer_nanosleep(timespec64_to_ktime(tu), HRTIMER_MODE_REL, diff --git a/kernel/time/posix-stubs.c b/kernel/time/posix-stubs.c index 90ea5f373e50..828aeecbd1e8 100644 --- a/kernel/time/posix-stubs.c +++ b/kernel/time/posix-stubs.c @@ -147,6 +147,7 @@ SYSCALL_DEFINE4(clock_nanosleep, const clockid_t, which_clock, int, flags, return -EINVAL; if (flags & TIMER_ABSTIME) rmtp = NULL; + current->restart_block.fn = do_no_restart_syscall; current->restart_block.nanosleep.type = rmtp ? TT_NATIVE : TT_NONE; current->restart_block.nanosleep.rmtp = rmtp; texp = timespec64_to_ktime(t); @@ -240,6 +241,7 @@ SYSCALL_DEFINE4(clock_nanosleep_time32, clockid_t, which_clock, int, flags, return -EINVAL; if (flags & TIMER_ABSTIME) rmtp = NULL; + current->restart_block.fn = do_no_restart_syscall; current->restart_block.nanosleep.type = rmtp ? TT_COMPAT : TT_NONE; current->restart_block.nanosleep.compat_rmtp = rmtp; texp = timespec64_to_ktime(t); diff --git a/kernel/time/posix-timers.c b/kernel/time/posix-timers.c index 5dead89308b7..0c8a87a11b39 100644 --- a/kernel/time/posix-timers.c +++ b/kernel/time/posix-timers.c @@ -1270,6 +1270,7 @@ SYSCALL_DEFINE4(clock_nanosleep, const clockid_t, which_clock, int, flags, return -EINVAL; if (flags & TIMER_ABSTIME) rmtp = NULL; + current->restart_block.fn = do_no_restart_syscall; current->restart_block.nanosleep.type = rmtp ? TT_NATIVE : TT_NONE; current->restart_block.nanosleep.rmtp = rmtp; @@ -1297,6 +1298,7 @@ SYSCALL_DEFINE4(clock_nanosleep_time32, clockid_t, which_clock, int, flags, return -EINVAL; if (flags & TIMER_ABSTIME) rmtp = NULL; + current->restart_block.fn = do_no_restart_syscall; current->restart_block.nanosleep.type = rmtp ? TT_COMPAT : TT_NONE; current->restart_block.nanosleep.compat_rmtp = rmtp; base-commit: 41c03ba9beea760bd2d2ac9250b09a2e192da2dc -- 2.39.0.314.g84b9a713c41-goog
| |