Add experimental support for process descriptors

A "process descriptor" file descriptor is used to manage processes
without using the PID namespace. This is required for Capsicum's
Capability Mode, where the PID namespace is unavailable.

New system calls pdfork(2) and pdkill(2) offer the functional equivalents
of fork(2) and kill(2). pdgetpid(2) allows querying the PID of the remote
process for debugging purposes. The currently-unimplemented pdwait(2) will,
in the future, allow querying rusage/exit status. In the interim, poll(2)
may be used to check (and wait for) process termination.

When a process is referenced by a process descriptor, it does not issue
SIGCHLD to the parent, making it suitable for use in libraries---a common
scenario when using library compartmentalisation from within large
applications (such as web browsers). Some observers may note a similarity
to Mach task ports; process descriptors provide a subset of this behaviour,
but in a UNIX style.

This feature is enabled by "options PROCDESC", but as with several other
Capsicum kernel features, is not enabled by default in GENERIC 9.0.

Reviewed by: jhb, kib
Approved by: re (kib), mentor (rwatson)
Sponsored by: Google Inc

1 files changed, 524 insertions, 0 deletions

diff --git a/sys/kern/sys_procdesc.c b/sys/kern/sys_procdesc.c
new file mode 100644
index 0000000..9993732
--- /dev/null
+++ b/sys/kern/sys_procdesc.c
@@ -0,0 +1,524 @@
+/*-
+ * Copyright (c) 2009 Robert N. M. Watson
+ * All rights reserved.
+ *
+ * This software was developed at the University of Cambridge Computer
+ * Laboratory with support from a grant from Google, Inc.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+/*-
+ * FreeBSD process descriptor facility.
+ *
+ * Some processes are represented by a file descriptor, which will be used in
+ * preference to signaling and pids for the purposes of process management,
+ * and is, in effect, a form of capability.  When a process descriptor is
+ * used with a process, it ceases to be visible to certain traditional UNIX
+ * process facilities, such as waitpid(2).
+ *
+ * Some semantics:
+ *
+ * - At most one process descriptor will exist for any process, although
+ *   references to that descriptor may be held from many processes (or even
+ *   be in flight between processes over a local domain socket).
+ * - Last close on the process descriptor will terminate the process using
+ *   SIGKILL and reparent it to init so that there's a process to reap it
+ *   when it's done exiting.
+ * - If the process exits before the descriptor is closed, it will not
+ *   generate SIGCHLD on termination, or be picked up by waitpid().
+ * - The pdkill(2) system call may be used to deliver a signal to the process
+ *   using its process descriptor.
+ * - The pdwait4(2) system call may be used to block (or not) on a process
+ *   descriptor to collect termination information.
+ *
+ * Open questions:
+ *
+ * - How to handle ptrace(2)?
+ * - Will we want to add a pidtoprocdesc(2) system call to allow process
+ *   descriptors to be created for processes without pfork(2)?
+ */
+
+#include <sys/cdefs.h>
+__FBSDID("$FreeBSD$");
+
+#include "opt_procdesc.h"
+
+#include <sys/param.h>
+#include <sys/capability.h>
+#include <sys/fcntl.h>
+#include <sys/file.h>
+#include <sys/filedesc.h>
+#include <sys/kernel.h>
+#include <sys/lock.h>
+#include <sys/mutex.h>
+#include <sys/poll.h>
+#include <sys/proc.h>
+#include <sys/procdesc.h>
+#include <sys/resourcevar.h>
+#include <sys/stat.h>
+#include <sys/sysproto.h>
+#include <sys/sysctl.h>
+#include <sys/systm.h>
+#include <sys/ucred.h>
+
+#include <security/audit/audit.h>
+
+#include <vm/uma.h>
+
+#ifdef PROCDESC
+
+FEATURE(process_descriptors, "Process Descriptors");
+
+static uma_zone_t procdesc_zone;
+
+static fo_rdwr_t	procdesc_read;
+static fo_rdwr_t	procdesc_write;
+static fo_truncate_t	procdesc_truncate;
+static fo_ioctl_t	procdesc_ioctl;
+static fo_poll_t	procdesc_poll;
+static fo_kqfilter_t	procdesc_kqfilter;
+static fo_stat_t	procdesc_stat;
+static fo_close_t	procdesc_close;
+static fo_chmod_t	procdesc_chmod;
+static fo_chown_t	procdesc_chown;
+
+static struct fileops procdesc_ops = {
+	.fo_read = procdesc_read,
+	.fo_write = procdesc_write,
+	.fo_truncate = procdesc_truncate,
+	.fo_ioctl = procdesc_ioctl,
+	.fo_poll = procdesc_poll,
+	.fo_kqfilter = procdesc_kqfilter,
+	.fo_stat = procdesc_stat,
+	.fo_close = procdesc_close,
+	.fo_chmod = procdesc_chmod,
+	.fo_chown = procdesc_chown,
+	.fo_flags = DFLAG_PASSABLE,
+};
+
+/*
+ * Initialize with VFS so that process descriptors are available along with
+ * other file descriptor types.  As long as it runs before init(8) starts,
+ * there shouldn't be a problem.
+ */
+static void
+procdesc_init(void *dummy __unused)
+{
+
+	procdesc_zone = uma_zcreate("procdesc", sizeof(struct procdesc),
+	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
+	if (procdesc_zone == NULL)
+		panic("procdesc_init: procdesc_zone not initialized");
+}
+SYSINIT(vfs, SI_SUB_VFS, SI_ORDER_ANY, procdesc_init, NULL);
+
+/*
+ * Return a locked process given a process descriptor, or ESRCH if it has
+ * died.
+ */
+int
+procdesc_find(struct thread *td, int fd, cap_rights_t rights,
+    struct proc **p)
+{
+	struct procdesc *pd;
+	struct file *fp;
+	int error;
+
+	error = fget(td, fd, rights, &fp);
+	if (error)
+		return (error);
+	if (fp->f_type != DTYPE_PROCDESC) {
+		error = EBADF;
+		goto out;
+	}
+	pd = fp->f_data;
+	sx_slock(&proctree_lock);
+	if (pd->pd_proc != NULL) {
+		*p = pd->pd_proc;
+		PROC_LOCK(*p);
+	} else
+		error = ESRCH;
+	sx_sunlock(&proctree_lock);
+out:
+	fdrop(fp, td);
+	return (error);
+}
+
+/*
+ * Function to be used by procstat(1) sysctls when returning procdesc
+ * information.
+ */
+pid_t
+procdesc_pid(struct file *fp_procdesc)
+{
+	struct procdesc *pd;
+
+	KASSERT(fp_procdesc->f_type == DTYPE_PROCDESC,
+	   ("procdesc_pid: !procdesc"));
+
+	pd = fp_procdesc->f_data;
+	return (pd->pd_pid);
+}
+
+/*
+ * Retrieve the PID associated with a process descriptor.
+ */
+int
+kern_pdgetpid(struct thread *td, int fd, cap_rights_t rights, pid_t *pidp)
+{
+	struct file *fp;
+	int error;
+
+	error = fget(td, fd, rights, &fp);
+	if (error)
+		return (error);
+	if (fp->f_type != DTYPE_PROCDESC) {
+		error = EBADF;
+		goto out;
+	}
+	*pidp = procdesc_pid(fp);
+out:
+	fdrop(fp, td);
+	return (error);
+}
+
+/*
+ * System call to return the pid of a process given its process descriptor.
+ */
+int
+pdgetpid(struct thread *td, struct pdgetpid_args *uap)
+{
+	pid_t pid;
+	int error;
+
+	AUDIT_ARG_FD(uap->fd);
+	error = kern_pdgetpid(td, uap->fd, CAP_PDGETPID, &pid);
+	if (error == 0)
+		error = copyout(&pid, uap->pidp, sizeof(pid));
+	return (error);
+}
+
+/*
+ * When a new process is forked by pdfork(), a file descriptor is allocated
+ * by the fork code first, then the process is forked, and then we get a
+ * chance to set up the process descriptor.  Failure is not permitted at this
+ * point, so procdesc_new() must succeed.
+ */
+void
+procdesc_new(struct proc *p, int flags)
+{
+	struct procdesc *pd;
+
+	pd = uma_zalloc(procdesc_zone, M_WAITOK | M_ZERO);
+	pd->pd_proc = p;
+	pd->pd_pid = p->p_pid;
+	p->p_procdesc = pd;
+	pd->pd_flags = 0;
+	if (flags & PD_DAEMON)
+		pd->pd_flags |= PDF_DAEMON;
+	PROCDESC_LOCK_INIT(pd);
+
+	/*
+	 * Process descriptors start out with two references: one from their
+	 * struct file, and the other from their struct proc.
+	 */
+	refcount_init(&pd->pd_refcount, 2);
+}
+
+/*
+ * Initialize a file with a process descriptor.
+ */
+void
+procdesc_finit(struct procdesc *pdp, struct file *fp)
+{
+
+	finit(fp, FREAD | FWRITE, DTYPE_PROCDESC, pdp, &procdesc_ops);
+}
+
+static void
+procdesc_free(struct procdesc *pd)
+{
+
+	/*
+	 * When the last reference is released, we assert that the descriptor
+	 * has been closed, but not that the process has exited, as we will
+	 * detach the descriptor before the process dies if the descript is
+	 * closed, as we can't wait synchronously.
+	 */
+	if (refcount_release(&pd->pd_refcount)) {
+		KASSERT(pd->pd_proc == NULL,
+		    ("procdesc_free: pd_proc != NULL"));
+		KASSERT((pd->pd_flags & PDF_CLOSED),
+		    ("procdesc_free: !PDF_CLOSED"));
+
+		PROCDESC_LOCK_DESTROY(pd);
+		uma_zfree(procdesc_zone, pd);
+	}
+}
+
+/*
+ * procdesc_exit() - notify a process descriptor that its process is exiting.
+ * We use the proctree_lock to ensure that process exit either happens
+ * strictly before or strictly after a concurrent call to procdesc_close().
+ */
+int
+procdesc_exit(struct proc *p)
+{
+	struct procdesc *pd;
+
+	sx_assert(&proctree_lock, SA_XLOCKED);
+	PROC_LOCK_ASSERT(p, MA_OWNED);
+	KASSERT(p->p_procdesc != NULL, ("procdesc_exit: p_procdesc NULL"));
+
+	pd = p->p_procdesc;
+
+	PROCDESC_LOCK(pd);
+	KASSERT((pd->pd_flags & PDF_CLOSED) == 0 || p->p_pptr == initproc,
+	    ("procdesc_exit: closed && parent not init"));
+
+	pd->pd_flags |= PDF_EXITED;
+
+	/*
+	 * If the process descriptor has been closed, then we have nothing
+	 * to do; return 1 so that init will get SIGCHLD and do the reaping.
+	 * Clean up the procdesc now rather than letting it happen during
+	 * that reap.
+	 */
+	if (pd->pd_flags & PDF_CLOSED) {
+		PROCDESC_UNLOCK(pd);
+		pd->pd_proc = NULL;
+		p->p_procdesc = NULL;
+		procdesc_free(pd);
+		return (1);
+	}
+	if (pd->pd_flags & PDF_SELECTED) {
+		pd->pd_flags &= ~PDF_SELECTED;
+		selwakeup(&pd->pd_selinfo);
+	}
+	PROCDESC_UNLOCK(pd);
+	return (0);
+}
+
+/*
+ * When a process descriptor is reaped, perhaps as a result of close() or
+ * pdwait4(), release the process's reference on the process descriptor.
+ */
+void
+procdesc_reap(struct proc *p)
+{
+	struct procdesc *pd;
+
+	sx_assert(&proctree_lock, SA_XLOCKED);
+	KASSERT(p->p_procdesc != NULL, ("procdesc_reap: p_procdesc == NULL"));
+
+	pd = p->p_procdesc;
+	pd->pd_proc = NULL;
+	procdesc_free(pd);
+}
+
+/*
+ * procdesc_close() - last close on a process descriptor.  If the process is
+ * still running, terminate with SIGKILL (unless PD_DAEMON is set) and let
+ * init(8) clean up the mess; if not, we have to clean up the zombie ourselves.
+ */
+static int
+procdesc_close(struct file *fp, struct thread *td)
+{
+	struct procdesc *pd;
+	struct proc *p;
+
+	KASSERT(fp->f_type == DTYPE_PROCDESC, ("procdesc_close: !procdesc"));
+
+	pd = fp->f_data;
+	fp->f_ops = &badfileops;
+	fp->f_data = NULL;
+
+	sx_xlock(&proctree_lock);
+	PROCDESC_LOCK(pd);
+	pd->pd_flags |= PDF_CLOSED;
+	PROCDESC_UNLOCK(pd);
+	p = pd->pd_proc;
+	PROC_LOCK(p);
+	if (p->p_state == PRS_ZOMBIE) {
+		/*
+		 * If the process is already dead and just awaiting reaping,
+		 * do that now.  This will release the process's reference to
+		 * the process descriptor when it calls back into
+		 * procdesc_reap().
+		 */
+		PROC_SLOCK(p);
+		proc_reap(curthread, p, NULL, 0, NULL);
+	} else {
+		/*
+		 * If the process is not yet dead, we need to kill it, but we
+		 * can't wait around synchronously for it to go away, as that
+		 * path leads to madness (and deadlocks).  First, detach the
+		 * process from its descriptor so that its exit status will
+		 * be reported normally.
+		 */
+		pd->pd_proc = NULL;
+		p->p_procdesc = NULL;
+		procdesc_free(pd);
+
+		/*
+		 * Next, reparent it to init(8) so that there's someone to
+		 * pick up the pieces; finally, terminate with prejudice.
+		 */
+		p->p_sigparent = SIGCHLD;
+		proc_reparent(p, initproc);
+		if ((pd->pd_flags & PD_DAEMON) == 0)
+			psignal(p, SIGKILL);
+		PROC_UNLOCK(p);
+		sx_xunlock(&proctree_lock);
+	}
+
+	/*
+	 * Release the file descriptor's reference on the process descriptor.
+	 */
+	procdesc_free(pd);
+	return (0);
+}
+
+static int
+procdesc_read(struct file *fp, struct uio *uio, struct ucred *active_cred,
+    int flags, struct thread *td)
+{
+
+	return (EOPNOTSUPP);
+}
+
+static int
+procdesc_write(struct file *fp, struct uio *uio, struct ucred *active_cred,
+    int flags, struct thread *td)
+{
+
+	return (EOPNOTSUPP);
+}
+
+static int
+procdesc_truncate(struct file *fp, off_t length, struct ucred *active_cred,
+    struct thread *td)
+{
+
+	return (EOPNOTSUPP);
+}
+
+static int
+procdesc_ioctl(struct file *fp, u_long com, void *data,
+    struct ucred *active_cred, struct thread *td)
+{
+
+	return (EOPNOTSUPP);
+}
+
+static int
+procdesc_poll(struct file *fp, int events, struct ucred *active_cred,
+    struct thread *td)
+{
+	struct procdesc *pd;
+	int revents;
+
+	revents = 0;
+	pd = fp->f_data;
+	PROCDESC_LOCK(pd);
+	if (pd->pd_flags & PDF_EXITED)
+		revents |= POLLHUP;
+	if (revents == 0) {
+		selrecord(td, &pd->pd_selinfo);
+		pd->pd_flags |= PDF_SELECTED;
+	}
+	PROCDESC_UNLOCK(pd);
+	return (revents);
+}
+
+static int
+procdesc_kqfilter(struct file *fp, struct knote *kn)
+{
+
+	return (EOPNOTSUPP);
+}
+
+static int
+procdesc_stat(struct file *fp, struct stat *sb, struct ucred *active_cred,
+    struct thread *td)
+{
+	struct procdesc *pd;
+	struct timeval pstart;
+
+	/*
+	 * XXXRW: Perhaps we should cache some more information from the
+	 * process so that we can return it reliably here even after it has
+	 * died.  For example, caching its credential data.
+	 */
+	bzero(sb, sizeof(*sb));
+	pd = fp->f_data;
+	sx_slock(&proctree_lock);
+	if (pd->pd_proc != NULL) {
+		PROC_LOCK(pd->pd_proc);
+
+		/* Set birth and [acm] times to process start time. */
+		pstart = pd->pd_proc->p_stats->p_start;
+		timevaladd(&pstart, &boottime);
+		TIMEVAL_TO_TIMESPEC(&pstart, &sb->st_birthtim);
+		sb->st_atim = sb->st_birthtim;
+		sb->st_ctim = sb->st_birthtim;
+		sb->st_mtim = sb->st_birthtim;
+		if (pd->pd_proc->p_state != PRS_ZOMBIE)
+			sb->st_mode = S_IFREG | S_IRWXU;
+		else
+			sb->st_mode = S_IFREG;
+		sb->st_uid = pd->pd_proc->p_ucred->cr_ruid;
+		sb->st_gid = pd->pd_proc->p_ucred->cr_rgid;
+		PROC_UNLOCK(pd->pd_proc);
+	} else
+		sb->st_mode = S_IFREG;
+	sx_sunlock(&proctree_lock);
+	return (0);
+}
+
+static int
+procdesc_chmod(struct file *fp, mode_t mode, struct ucred *active_cred,
+    struct thread *td)
+{
+
+	return (EOPNOTSUPP);
+}
+
+static int
+procdesc_chown(struct file *fp, uid_t uid, gid_t gid, struct ucred *active_cred,
+    struct thread *td)
+{
+
+	return (EOPNOTSUPP);
+}
+
+#else /* !PROCDESC */
+
+int
+pdgetpid(struct thread *td, struct pdgetpid_args *uap)
+{
+
+	return (ENOSYS);
+}
+
+#endif /* PROCDESC */