wait4() and waitid() only allow a parent or subreaper to collect exit status, so an external manager holding a userland file descriptor cannot wait on a userland init process forked by another process or observe exit status multiple times across observers. Add P_USERLAND to waitid() and wire kernel_waitid() to userland_waitid() to wait freezably and interruptibly on a userland file descriptor until the userland reaches USERLAND_DEAD and populate siginfo_t with init's caller-visible TGID, munged UID, and exit status without consuming it. Co-developed-by: David Howells Signed-off-by: David Howells Signed-off-by: Jarkko Sakkinen --- include/linux/userland.h | 10 +++++ include/uapi/linux/wait.h | 1 + kernel/exit.c | 4 ++ kernel/userland.c | 83 +++++++++++++++++++++++++++++++++++++++ 4 files changed, 98 insertions(+) diff --git a/include/linux/userland.h b/include/linux/userland.h index a92709957529..16988e10f623 100644 --- a/include/linux/userland.h +++ b/include/linux/userland.h @@ -16,8 +16,10 @@ struct file; struct nsproxy; struct pid; struct pid_namespace; +struct rusage; struct task_struct; struct user_namespace; +struct waitid_info; /** * enum userland_state - Userland lifecycle states @@ -73,6 +75,8 @@ bool userland_is_file(struct file *file); void userland_copy(struct task_struct *tsk, struct pid *pid, struct userland *userland); void userland_exit(struct pid_namespace *pid_ns, struct task_struct *tsk); +long userland_waitid(int upid, struct waitid_info *infop, int options, + struct rusage *ru); #else static inline void userland_put(struct userland *c) { @@ -92,6 +96,12 @@ static inline void userland_exit(struct pid_namespace *pid_ns, struct task_struct *tsk) { } + +static inline long userland_waitid(int upid, struct waitid_info *infop, + int options, struct rusage *ru) +{ + return -EINVAL; +} #endif /* CONFIG_USERLAND */ #endif /* _LINUX_USERLAND_H */ diff --git a/include/uapi/linux/wait.h b/include/uapi/linux/wait.h index 85b809fc9f11..edd39c2cab96 100644 --- a/include/uapi/linux/wait.h +++ b/include/uapi/linux/wait.h @@ -18,6 +18,7 @@ #define P_PID 1 #define P_PGID 2 #define P_PIDFD 3 +#define P_USERLAND 4 #endif /* _UAPI_LINUX_WAIT_H */ diff --git a/kernel/exit.c b/kernel/exit.c index d5e7bc2af804..5b0bd3a9187c 100644 --- a/kernel/exit.c +++ b/kernel/exit.c @@ -71,6 +71,7 @@ #include #include #include +#include #include @@ -1803,6 +1804,9 @@ static long kernel_waitid(int which, pid_t upid, struct waitid_info *infop, struct wait_opts wo; long ret; + if (which == P_USERLAND) + return userland_waitid(upid, infop, options, ru); + ret = kernel_waitid_prepare(&wo, which, upid, infop, options, ru); if (ret) return ret; diff --git a/kernel/userland.c b/kernel/userland.c index 279b92e7b7f8..3ded663b71af 100644 --- a/kernel/userland.c +++ b/kernel/userland.c @@ -15,6 +15,7 @@ #include #include #include +#include #include #include #include @@ -40,6 +41,8 @@ #include #include +#include "exit.h" + void userland_put(struct userland *c) { if (!c || !refcount_dec_and_test(&c->usage)) @@ -213,6 +216,86 @@ static long userland_share_nsfd(struct userland *c, unsigned long arg) return 0; } +/** + * DOC: P_USERLAND + * + * ``waitid(P_USERLAND, userland_fd, infop, options, NULL)`` waits freezably until + * ``userland_fd`` reaches ``USERLAND_DEAD`` (unless ``WNOHANG`` or ``O_NONBLOCK`` is + * set) and writes init's exit information into ``infop`` without consuming it. + * ``options`` must include ``WEXITED`` and may include ``WNOHANG`` and + * ``WNOWAIT``; ``rusage`` must be ``NULL``, and the caller must run in an + * ancestor active PID namespace. + */ +long userland_waitid(int upid, struct waitid_info *infop, int options, + struct rusage *ru) +{ + struct pid_namespace *pid_ns; + unsigned int f_flags; + struct userland *userland; + bool nonblock; + int status; + kuid_t uid; + long ret; + + if (upid < 0 || ru) + return -EINVAL; + if (options & ~(WEXITED | WNOHANG | WNOWAIT)) + return -EINVAL; + if (!(options & WEXITED)) + return -EINVAL; + + CLASS(fd, f)(upid); + if (fd_empty(f) || !userland_is_file(fd_file(f))) + return -EBADF; + + f_flags = READ_ONCE(fd_file(f)->f_flags); + nonblock = (options & WNOHANG) || (f_flags & O_NONBLOCK); + userland = fd_file(f)->private_data; + pid_ns = task_active_pid_ns(current); + + if (!pidns_is_ancestor(userland->ns->pid_ns_for_children, pid_ns)) + return -ECHILD; + for (;;) { + enum userland_state state; + + scoped_guard(spinlock, &userland->lock) { + state = userland->state; + if (state == USERLAND_INIT) + return -ECHILD; + if (state == USERLAND_DEAD) { + ret = pid_nr_ns(userland->init_pid, pid_ns); + if (ret <= 0) + return -ECHILD; + status = userland->exit_code; + uid = userland->init_uid; + } + } + if (state == USERLAND_DEAD) + break; + if (nonblock) + return (options & WNOHANG) ? 0 : -EAGAIN; + + ret = wait_event_freezable(userland->waitq, + READ_ONCE(userland->state) == + USERLAND_DEAD || + READ_ONCE(userland->state) == + USERLAND_INIT); + if (ret) + return ret; + } + + if ((status & 0x7f) == 0) { + infop->cause = CLD_EXITED; + infop->status = status >> 8; + } else { + infop->cause = (status & 0x80) ? CLD_DUMPED : CLD_KILLED; + infop->status = status & 0x7f; + } + infop->pid = ret; + infop->uid = from_kuid_munged(current_user_ns(), uid); + return ret; +} + static long userland_ioctl(struct file *file, unsigned int cmd, unsigned long arg) { struct userland *userland = file->private_data; -- 2.47.3