Second-to-last commit implementing Capsicum capabilities in the FreeBSD

kernel for FreeBSD 9.0:

Add a new capability mask argument to fget(9) and friends, allowing system
call code to declare what capabilities are required when an integer file
descriptor is converted into an in-kernel struct file *.  With options
CAPABILITIES compiled into the kernel, this enforces capability
protection; without, this change is effectively a no-op.

Some cases require special handling, such as mmap(2), which must preserve
information about the maximum rights at the time of mapping in the memory
map so that they can later be enforced in mprotect(2) -- this is done by
narrowing the rights in the existing max_protection field used for similar
purposes with file permissions.

In namei(9), we assert that the code is not reached from within capability
mode, as we're not yet ready to enforce namespace capabilities there.
This will follow in a later commit.

Update two capability names: CAP_EVENT and CAP_KEVENT become
CAP_POST_KEVENT and CAP_POLL_KEVENT to more accurately indicate what they
represent.

Approved by:	re (bz)
Submitted by:	jonathan
Sponsored by:	Google Inc

10 files changed, 62 insertions, 37 deletions

diff --git a/sys/compat/freebsd32/freebsd32_ioctl.c b/sys/compat/freebsd32/freebsd32_ioctl.c
index e662a5d..1d773ca 100644
--- a/sys/compat/freebsd32/freebsd32_ioctl.c
+++ b/sys/compat/freebsd32/freebsd32_ioctl.c
@@ -33,6 +33,7 @@ __FBSDID("$FreeBSD$");
 #include "opt_compat.h"
 
 #include <sys/param.h>
+#include <sys/capability.h>
 #include <sys/cdio.h>
 #include <sys/fcntl.h>
 #include <sys/filio.h>
@@ -354,7 +355,7 @@ freebsd32_ioctl(struct thread *td, struct freebsd32_ioctl_args *uap)
 	struct file *fp;
 	int error;
 
-	if ((error = fget(td, uap->fd, &fp)) != 0)
+	if ((error = fget(td, uap->fd, CAP_IOCTL, &fp)) != 0)
 		return (error);
 	if ((fp->f_flag & (FREAD | FWRITE)) == 0) {
 		fdrop(fp, td);
diff --git a/sys/compat/linux/linux_file.c b/sys/compat/linux/linux_file.c
index 44ad193..e923032 100644
--- a/sys/compat/linux/linux_file.c
+++ b/sys/compat/linux/linux_file.c
@@ -33,6 +33,7 @@ __FBSDID("$FreeBSD$");
 
 #include <sys/param.h>
 #include <sys/systm.h>
+#include <sys/capability.h>
 #include <sys/conf.h>
 #include <sys/dirent.h>
 #include <sys/fcntl.h>
@@ -141,7 +142,7 @@ linux_common_open(struct thread *td, int dirfd, char *path, int l_flags, int mod
 	     * having the same filedesc could use that fd without
 	     * checking below.
 	     */
-	    error = fget(td, fd, &fp);
+	    error = fget(td, fd, CAP_IOCTL, &fp);
 	    if (!error) {
 		    sx_slock(&proctree_lock);
 		    PROC_LOCK(p);
@@ -345,7 +346,7 @@ getdents_common(struct thread *td, struct linux_getdents64_args *args,
 	} else
 		justone = 0;
 
-	if ((error = getvnode(td->td_proc->p_fd, args->fd, &fp)) != 0)
+	if ((error = getvnode(td->td_proc->p_fd, args->fd, CAP_READ, &fp)) != 0)
 		return (error);
 
 	if ((fp->f_flag & FREAD) == 0) {
@@ -1041,7 +1042,7 @@ linux_pread(td, uap)
 
 	if (error == 0) {
    	   	/* This seems to violate POSIX but linux does it */
-   	   	if ((error = fgetvp(td, uap->fd, &vp)) != 0)
+		if ((error = fgetvp(td, uap->fd, CAP_READ, &vp)) != 0)
    		   	return (error);
 		if (vp->v_type == VDIR) {
    		   	vrele(vp);
@@ -1390,7 +1391,7 @@ fcntl_common(struct thread *td, struct linux_fcntl64_args *args)
 		 * significant effect for pipes (SIGIO is not delivered for
 		 * pipes under Linux-2.2.35 at least).
 		 */
-		error = fget(td, args->fd, &fp);
+		error = fget(td, args->fd, CAP_FCNTL, &fp);
 		if (error)
 			return (error);
 		if (fp->f_type == DTYPE_PIPE) {
diff --git a/sys/compat/linux/linux_ioctl.c b/sys/compat/linux/linux_ioctl.c
index 5532c93..d021fba 100644
--- a/sys/compat/linux/linux_ioctl.c
+++ b/sys/compat/linux/linux_ioctl.c
@@ -34,6 +34,7 @@ __FBSDID("$FreeBSD$");
 #include <sys/param.h>
 #include <sys/systm.h>
 #include <sys/sysproto.h>
+#include <sys/capability.h>
 #include <sys/cdio.h>
 #include <sys/dvdio.h>
 #include <sys/conf.h>
@@ -193,7 +194,7 @@ linux_ioctl_hdio(struct thread *td, struct linux_ioctl_args *args)
 	u_int sectorsize, fwcylinders, fwheads, fwsectors;
 	off_t mediasize, bytespercyl;
 
-	if ((error = fget(td, args->fd, &fp)) != 0)
+	if ((error = fget(td, args->fd, CAP_IOCTL, &fp)) != 0)
 		return (error);
 	switch (args->cmd & 0xffff) {
 	case LINUX_HDIO_GET_GEO:
@@ -274,7 +275,7 @@ linux_ioctl_disk(struct thread *td, struct linux_ioctl_args *args)
 	u_int sectorsize;
 	off_t mediasize;
 
-	if ((error = fget(td, args->fd, &fp)) != 0)
+	if ((error = fget(td, args->fd, CAP_IOCTL, &fp)) != 0)
 		return (error);
 	switch (args->cmd & 0xffff) {
 	case LINUX_BLKGETSIZE:
@@ -700,7 +701,7 @@ linux_ioctl_termio(struct thread *td, struct linux_ioctl_args *args)
 	struct file *fp;
 	int error;
 
-	if ((error = fget(td, args->fd, &fp)) != 0)
+	if ((error = fget(td, args->fd, CAP_IOCTL, &fp)) != 0)
 		return (error);
 
 	switch (args->cmd & 0xffff) {
@@ -1440,7 +1441,7 @@ linux_ioctl_cdrom(struct thread *td, struct linux_ioctl_args *args)
 	struct file *fp;
 	int error;
 
-	if ((error = fget(td, args->fd, &fp)) != 0)
+	if ((error = fget(td, args->fd, CAP_IOCTL, &fp)) != 0)
 		return (error);
 	switch (args->cmd & 0xffff) {
 
@@ -1965,7 +1966,7 @@ linux_ioctl_console(struct thread *td, struct linux_ioctl_args *args)
 	struct file *fp;
 	int error;
 
-	if ((error = fget(td, args->fd, &fp)) != 0)
+	if ((error = fget(td, args->fd, CAP_IOCTL, &fp)) != 0)
 		return (error);
 	switch (args->cmd & 0xffff) {
 
@@ -2356,7 +2357,7 @@ linux_ioctl_socket(struct thread *td, struct linux_ioctl_args *args)
 	ifp = NULL;
 	error = 0;
 
-	if ((error = fget(td, args->fd, &fp)) != 0)
+	if ((error = fget(td, args->fd, CAP_IOCTL, &fp)) != 0)
 		return (error);
 	type = fp->f_type;
 	fdrop(fp, td);
@@ -2582,7 +2583,7 @@ linux_ioctl_private(struct thread *td, struct linux_ioctl_args *args)
 	struct file *fp;
 	int error, type;
 
-	if ((error = fget(td, args->fd, &fp)) != 0)
+	if ((error = fget(td, args->fd, CAP_IOCTL, &fp)) != 0)
 		return (error);
 	type = fp->f_type;
 	fdrop(fp, td);
@@ -2608,7 +2609,7 @@ linux_ioctl_sg(struct thread *td, struct linux_ioctl_args *args)
 	u_long cmd;
 	int error;
 
-	if ((error = fget(td, args->fd, &fp)) != 0) {
+	if ((error = fget(td, args->fd, CAP_IOCTL, &fp)) != 0) {
 		printf("sg_linux_ioctl: fget returned %d\n", error);
 		return (error);
 	}
@@ -2843,7 +2844,7 @@ linux_ioctl_v4l(struct thread *td, struct linux_ioctl_args *args)
 	case LINUX_VIDIOCSCHAN:		args->cmd = VIDIOCSCHAN; break;
 
 	case LINUX_VIDIOCGTUNER:
-		if ((error = fget(td, args->fd, &fp)) != 0)
+		if ((error = fget(td, args->fd, CAP_IOCTL, &fp)) != 0)
 			return (error);
 		error = copyin((void *) args->arg, &l_vtun, sizeof(l_vtun));
 		if (error) {
@@ -2861,7 +2862,7 @@ linux_ioctl_v4l(struct thread *td, struct linux_ioctl_args *args)
 		return (error);
 
 	case LINUX_VIDIOCSTUNER:
-		if ((error = fget(td, args->fd, &fp)) != 0)
+		if ((error = fget(td, args->fd, CAP_IOCTL, &fp)) != 0)
 			return (error);
 		error = copyin((void *) args->arg, &l_vtun, sizeof(l_vtun));
 		if (error) {
@@ -2878,7 +2879,7 @@ linux_ioctl_v4l(struct thread *td, struct linux_ioctl_args *args)
 	case LINUX_VIDIOCCAPTURE:	args->cmd = VIDIOCCAPTURE; break;
 
 	case LINUX_VIDIOCGWIN:
-		if ((error = fget(td, args->fd, &fp)) != 0)
+		if ((error = fget(td, args->fd, CAP_IOCTL, &fp)) != 0)
 			return (error);
 		error = fo_ioctl(fp, VIDIOCGWIN, &vwin, td->td_ucred, td);
 		if (!error) {
@@ -2890,7 +2891,7 @@ linux_ioctl_v4l(struct thread *td, struct linux_ioctl_args *args)
 		return (error);
 
 	case LINUX_VIDIOCSWIN:
-		if ((error = fget(td, args->fd, &fp)) != 0)
+		if ((error = fget(td, args->fd, CAP_IOCTL, &fp)) != 0)
 			return (error);
 		error = copyin((void *) args->arg, &l_vwin, sizeof(l_vwin));
 		if (error) {
@@ -2913,7 +2914,7 @@ linux_ioctl_v4l(struct thread *td, struct linux_ioctl_args *args)
 		return (error);
 
 	case LINUX_VIDIOCGFBUF:
-		if ((error = fget(td, args->fd, &fp)) != 0)
+		if ((error = fget(td, args->fd, CAP_IOCTL, &fp)) != 0)
 			return (error);
 		error = fo_ioctl(fp, VIDIOCGFBUF, &vbuf, td->td_ucred, td);
 		if (!error) {
@@ -2925,7 +2926,7 @@ linux_ioctl_v4l(struct thread *td, struct linux_ioctl_args *args)
 		return (error);
 
 	case LINUX_VIDIOCSFBUF:
-		if ((error = fget(td, args->fd, &fp)) != 0)
+		if ((error = fget(td, args->fd, CAP_IOCTL, &fp)) != 0)
 			return (error);
 		error = copyin((void *) args->arg, &l_vbuf, sizeof(l_vbuf));
 		if (error) {
@@ -2953,7 +2954,7 @@ linux_ioctl_v4l(struct thread *td, struct linux_ioctl_args *args)
 	case LINUX_VIDIOCGPLAYINFO:	args->cmd = VIDIOCGPLAYINFO; break;
 
 	case LINUX_VIDIOCSMICROCODE:
-		if ((error = fget(td, args->fd, &fp)) != 0)
+		if ((error = fget(td, args->fd, CAP_IOCTL, &fp)) != 0)
 			return (error);
 		error = copyin((void *) args->arg, &l_vcode, sizeof(l_vcode));
 		if (error) {
@@ -3197,7 +3198,7 @@ linux_ioctl_v4l2(struct thread *td, struct linux_ioctl_args *args)
 		error = copyin((void *)args->arg, &l_vformat, sizeof(l_vformat));
 		if (error)
 			return (error);
-		if ((error = fget(td, args->fd, &fp)) != 0)
+		if ((error = fget(td, args->fd, CAP_IOCTL, &fp)) != 0)
 			return (error);
 		if (linux_to_bsd_v4l2_format(&l_vformat, &vformat) != 0)
 			error = EINVAL;
@@ -3220,7 +3221,7 @@ linux_ioctl_v4l2(struct thread *td, struct linux_ioctl_args *args)
 		if (error)
 			return (error);
 		linux_to_bsd_v4l2_standard(&l_vstd, &vstd);
-		if ((error = fget(td, args->fd, &fp)) != 0)
+		if ((error = fget(td, args->fd, CAP_IOCTL, &fp)) != 0)
 			return (error);
 		error = fo_ioctl(fp, VIDIOC_ENUMSTD, (caddr_t)&vstd,
 		    td->td_ucred, td);
@@ -3242,7 +3243,7 @@ linux_ioctl_v4l2(struct thread *td, struct linux_ioctl_args *args)
 				sizeof(struct l_v4l2_input));
 		if (error != 0)
 			return (error);
-		if ((error = fget(td, args->fd, &fp)) != 0)
+		if ((error = fget(td, args->fd, CAP_IOCTL, &fp)) != 0)
 			return (error);
 		error = fo_ioctl(fp, VIDIOC_ENUMINPUT, (caddr_t)&vinp,
 		    td->td_ucred, td);
@@ -3261,7 +3262,7 @@ linux_ioctl_v4l2(struct thread *td, struct linux_ioctl_args *args)
 		error = copyin((void *)args->arg, &l_vbuf, sizeof(l_vbuf));
 		if (error)
 			return (error);
-		if ((error = fget(td, args->fd, &fp)) != 0)
+		if ((error = fget(td, args->fd, CAP_IOCTL, &fp)) != 0)
 			return (error);
 		linux_to_bsd_v4l2_buffer(&l_vbuf, &vbuf);
 		if ((args->cmd & 0xffff) == LINUX_VIDIOC_QUERYBUF)
@@ -3431,7 +3432,7 @@ linux_ioctl(struct thread *td, struct linux_ioctl_args *args)
 		    (unsigned long)args->cmd);
 #endif
 
-	if ((error = fget(td, args->fd, &fp)) != 0)
+	if ((error = fget(td, args->fd, CAP_IOCTL, &fp)) != 0)
 		return (error);
 	if ((fp->f_flag & (FREAD|FWRITE)) == 0) {
 		fdrop(fp, td);
diff --git a/sys/compat/linux/linux_socket.c b/sys/compat/linux/linux_socket.c
index 6940e45..08728a1 100644
--- a/sys/compat/linux/linux_socket.c
+++ b/sys/compat/linux/linux_socket.c
@@ -37,6 +37,7 @@ __FBSDID("$FreeBSD$");
 #include <sys/proc.h>
 #include <sys/systm.h>
 #include <sys/sysproto.h>
+#include <sys/capability.h>
 #include <sys/fcntl.h>
 #include <sys/file.h>
 #include <sys/limits.h>
@@ -743,7 +744,7 @@ linux_connect(struct thread *td, struct linux_connect_args *args)
 	 * socket and use the file descriptor reference instead of
 	 * creating a new one.
 	 */
-	error = fgetsock(td, args->s, &so, &fflag);
+	error = fgetsock(td, args->s, CAP_CONNECT, &so, &fflag);
 	if (error == 0) {
 		error = EISCONN;
 		if (fflag & FNONBLOCK) {
diff --git a/sys/compat/linux/linux_stats.c b/sys/compat/linux/linux_stats.c
index 8fa08b6..90f860d 100644
--- a/sys/compat/linux/linux_stats.c
+++ b/sys/compat/linux/linux_stats.c
@@ -141,8 +141,11 @@ translate_fd_major_minor(struct thread *td, int fd, struct stat *buf)
 	struct vnode *vp;
 	int major, minor;
 
+	/*
+	 * No capability rights required here.
+	 */
 	if ((!S_ISCHR(buf->st_mode) && !S_ISBLK(buf->st_mode)) ||
-	    fget(td, fd, &fp) != 0)
+	    fget(td, fd, 0, &fp) != 0)
 		return;
 	vp = fp->f_vnode;
 	if (vp != NULL && vp->v_rdev != NULL &&
diff --git a/sys/compat/svr4/svr4_fcntl.c b/sys/compat/svr4/svr4_fcntl.c
index 88e4fc2..ce1452a 100644
--- a/sys/compat/svr4/svr4_fcntl.c
+++ b/sys/compat/svr4/svr4_fcntl.c
@@ -33,6 +33,7 @@
 __FBSDID("$FreeBSD$");
 
 #include <sys/param.h>
+#include <sys/capability.h>
 #include <sys/systm.h>
 #include <sys/file.h>
 #include <sys/filedesc.h>
@@ -261,7 +262,17 @@ fd_revoke(td, fd)
 	int error, *retval;
 
 	retval = td->td_retval;
-	if ((error = fgetvp(td, fd, &vp)) != 0)
+	/*
+	 * If we ever want to support Capsicum on SVR4 processes (unlikely)
+	 * or FreeBSD grows a native frevoke() (more likely), we will need a
+	 * CAP_REVOKE here.
+	 *
+	 * In the meantime, use CAP_MASK_VALID: if a SVR4 process wants to
+	 * do an frevoke(), it needs to do it on either a regular file
+	 * descriptor or a fully-privileged capability (which is effectively
+	 * the same as a non-capability-restricted file descriptor).
+	 */
+	if ((error = fgetvp(td, fd, CAP_MASK_VALID, &vp)) != 0)
 		return (error);
 
 	if (vp->v_type != VCHR && vp->v_type != VBLK) {
@@ -313,7 +324,7 @@ fd_truncate(td, fd, flp)
 	/*
 	 * We only support truncating the file.
 	 */
-	if ((error = fget(td, fd, &fp)) != 0)
+	if ((error = fget(td, fd, CAP_FTRUNCATE, &fp)) != 0)
 		return (error);
 
 	vp = fp->f_vnode;
@@ -392,7 +403,7 @@ svr4_sys_open(td, uap)
 #if defined(NOTYET)
 		struct file	*fp;
 
-		error = fget(td, retval, &fp);
+		error = fget(td, retval, CAP_IOCTL, &fp);
 		PROC_UNLOCK(p);
 		/*
 		 * we may have lost a race the above open() and
diff --git a/sys/compat/svr4/svr4_filio.c b/sys/compat/svr4/svr4_filio.c
index ca85653..cb7cada 100644
--- a/sys/compat/svr4/svr4_filio.c
+++ b/sys/compat/svr4/svr4_filio.c
@@ -32,6 +32,7 @@ __FBSDID("$FreeBSD$");
 #include <sys/param.h>
 #include <sys/proc.h>
 #include <sys/systm.h>
+#include <sys/capability.h>
 #include <sys/file.h>
 #include <sys/filio.h>
 #include <sys/lock.h>
@@ -113,7 +114,7 @@ svr4_sys_read(td, uap)
      ra.buf = uap->buf;
      ra.nbyte = uap->nbyte;
 
-     if (fget(td, uap->fd, &fp) != 0) {
+     if (fget(td, uap->fd, CAP_READ, &fp) != 0) {
        DPRINTF(("Something fishy with the user-supplied file descriptor...\n"));
        return EBADF;
      }
diff --git a/sys/compat/svr4/svr4_ioctl.c b/sys/compat/svr4/svr4_ioctl.c
index 1cea41a..36b0580 100644
--- a/sys/compat/svr4/svr4_ioctl.c
+++ b/sys/compat/svr4/svr4_ioctl.c
@@ -31,6 +31,7 @@ __FBSDID("$FreeBSD$");
 
 #include <sys/param.h>
 #include <sys/proc.h>
+#include <sys/capability.h>
 #include <sys/file.h>
 #include <sys/filedesc.h>
 #include <sys/fcntl.h>
@@ -102,7 +103,7 @@ svr4_sys_ioctl(td, uap)
 	retval = td->td_retval;
 	cmd = uap->com;
 
-	if ((error = fget(td, uap->fd, &fp)) != 0)
+	if ((error = fget(td, uap->fd, CAP_IOCTL, &fp)) != 0)
 		return (error);
 
 	if ((fp->f_flag & (FREAD | FWRITE)) == 0) {
diff --git a/sys/compat/svr4/svr4_misc.c b/sys/compat/svr4/svr4_misc.c
index 6f80fe6..c0a7478 100644
--- a/sys/compat/svr4/svr4_misc.c
+++ b/sys/compat/svr4/svr4_misc.c
@@ -37,6 +37,7 @@ __FBSDID("$FreeBSD$");
 
 #include <sys/param.h>
 #include <sys/systm.h>
+#include <sys/capability.h>
 #include <sys/dirent.h>
 #include <sys/fcntl.h>
 #include <sys/filedesc.h>
@@ -246,7 +247,8 @@ svr4_sys_getdents64(td, uap)
 
 	DPRINTF(("svr4_sys_getdents64(%d, *, %d)\n",
 		uap->fd, uap->nbytes));
-	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0) {
+	if ((error = getvnode(td->td_proc->p_fd, uap->fd,
+	    CAP_READ | CAP_SEEK, &fp)) != 0) {
 		return (error);
 	}
 
@@ -427,7 +429,8 @@ svr4_sys_getdents(td, uap)
 	if (uap->nbytes < 0)
 		return (EINVAL);
 
-	if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
+	if ((error = getvnode(td->td_proc->p_fd, uap->fd,
+	    CAP_READ | CAP_SEEK, &fp)) != 0)
 		return (error);
 
 	if ((fp->f_flag & FREAD) == 0) {
@@ -615,7 +618,8 @@ svr4_sys_fchroot(td, uap)
 
 	if ((error = priv_check(td, PRIV_VFS_FCHROOT)) != 0)
 		return error;
-	if ((error = getvnode(fdp, uap->fd, &fp)) != 0)
+	/* XXX: we have the chroot priv... what cap might we need? all? */
+	if ((error = getvnode(fdp, uap->fd, 0, &fp)) != 0)
 		return error;
 	vp = fp->f_vnode;
 	VREF(vp);
diff --git a/sys/compat/svr4/svr4_stream.c b/sys/compat/svr4/svr4_stream.c
index ec95eec..a1a42c0 100644
--- a/sys/compat/svr4/svr4_stream.c
+++ b/sys/compat/svr4/svr4_stream.c
@@ -43,6 +43,7 @@ __FBSDID("$FreeBSD$");
 
 #include <sys/param.h>
 #include <sys/systm.h>
+#include <sys/capability.h>
 #include <sys/fcntl.h>
 #include <sys/filedesc.h>
 #include <sys/filio.h>
@@ -1448,7 +1449,7 @@ svr4_sys_putmsg(td, uap)
 	struct file     *fp;
 	int error;
 
-	if ((error = fget(td, uap->fd, &fp)) != 0) {
+	if ((error = fget(td, uap->fd, CAP_WRITE, &fp)) != 0) {
 #ifdef DEBUG_SVR4
 	        uprintf("putmsg: bad fp\n");
 #endif
@@ -1620,7 +1621,7 @@ svr4_sys_getmsg(td, uap)
 	struct file     *fp;
 	int error;
 
-	if ((error = fget(td, uap->fd, &fp)) != 0) {
+	if ((error = fget(td, uap->fd, CAP_READ, &fp)) != 0) {
 #ifdef DEBUG_SVR4
 	        uprintf("getmsg: bad fp\n");
 #endif