Add a kern_setsockopt and kern_getsockopt which can read the option

values from either user land or from the kernel. Use them for
[gs]etsockopt and to clean up some calls to [gs]etsockopt in the
Linux emulation code that uses the stackgap.

3 files changed, 110 insertions, 99 deletions

diff --git a/sys/compat/linux/linux_socket.c b/sys/compat/linux/linux_socket.c
index c73a57a..85c71a6 100644
--- a/sys/compat/linux/linux_socket.c
+++ b/sys/compat/linux/linux_socket.c
@@ -402,35 +402,14 @@ bad:
 
 /* Return 0 if IP_HDRINCL is set for the given socket. */
 static int
-linux_check_hdrincl(struct thread *td, caddr_t *sg, int s)
+linux_check_hdrincl(struct thread *td, int s)
 {
-	struct getsockopt_args /* {
-		int	s;
-		int	level;
-		int	name;
-		void * __restrict val;
-		socklen_t * __restrict avalsize;
-	} */ bsd_args;
-	void * __restrict val;
-	socklen_t * __restrict valsize;
 	int error, optval, size_val;
 
-	val = stackgap_alloc(sg, sizeof(size_val));
-	valsize = stackgap_alloc(sg, sizeof(socklen_t));
-
-	size_val = sizeof(val);
-	if ((error = copyout(&size_val, valsize, sizeof(size_val))))
-		return (error);
-
-	bsd_args.s = s;
-	bsd_args.level = IPPROTO_IP;
-	bsd_args.name = IP_HDRINCL;
-	bsd_args.val = val;
-	bsd_args.avalsize = valsize;
-	if ((error = getsockopt(td, &bsd_args)))
-		return (error);
-
-	if ((error = copyin(val, &optval, sizeof(optval))))
+	size_val = sizeof(optval);
+	error = kern_getsockopt(td, s, IPPROTO_IP, IP_HDRINCL,
+	    &optval, UIO_SYSSPACE, &size_val);
+	if (error)
 		return (error);
 
 	return (optval == 0);
@@ -461,6 +440,7 @@ linux_sendto_hdrincl(struct thread *td, caddr_t *sg, struct linux_sendto_args *l
 #define linux_ip_copysize	8
 
 	struct ip *packet;
+	caddr_t sg = stackgap_init();
 	struct msghdr msg;
 	struct iovec aiov[2];
 	int error;
@@ -515,13 +495,6 @@ linux_socket(struct thread *td, struct linux_socket_args *args)
 		int type;
 		int protocol;
 	} */ bsd_args;
-	struct setsockopt_args /* {
-		int s;
-		int level;
-		int name;
-		caddr_t val;
-		int valsize;
-	} */ bsd_setsockopt_args;
 	int error;
 	int retval_socket;
 
@@ -540,21 +513,12 @@ linux_socket(struct thread *td, struct linux_socket_args *args)
 	    && bsd_args.domain == AF_INET
 	    && retval_socket >= 0) {
 		/* It's a raw IP socket: set the IP_HDRINCL option. */
-		caddr_t sg;
-		int *hdrincl;
-
-		sg = stackgap_init();
-		hdrincl = (int *)stackgap_alloc(&sg, sizeof(*hdrincl));
-		*hdrincl = 1;
-		bsd_setsockopt_args.s = td->td_retval[0];
-		bsd_setsockopt_args.level = IPPROTO_IP;
-		bsd_setsockopt_args.name = IP_HDRINCL;
-		bsd_setsockopt_args.val = (caddr_t)hdrincl;
-		bsd_setsockopt_args.valsize = sizeof(*hdrincl);
-		/* We ignore any error returned by setsockopt() */
-		setsockopt(td, &bsd_setsockopt_args);
-		/* Copy back the return value from socket() */
-		td->td_retval[0] = bsd_setsockopt_args.s;
+		int hdrincl;
+
+		hdrincl = 1;
+		/* We ignore any error returned by kern_setsockopt() */
+		kern_setsockopt(td, td->td_retval[0], IPPROTO_IP, IP_HDRINCL,
+		    &hdrincl, UIO_SYSSPACE, sizeof(hdrincl));
 	}
 #ifdef INET6
 	/*
@@ -571,21 +535,12 @@ linux_socket(struct thread *td, struct linux_socket_args *args)
 	    && ip6_v6only
 #endif
 	    ) {
-		caddr_t sg;
-		int *v6only;
-
-		sg = stackgap_init();
-		v6only = (int *)stackgap_alloc(&sg, sizeof(*v6only));
-		*v6only = 0;
-		bsd_setsockopt_args.s = td->td_retval[0];
-		bsd_setsockopt_args.level = IPPROTO_IPV6;
-		bsd_setsockopt_args.name = IPV6_V6ONLY;
-		bsd_setsockopt_args.val = (caddr_t)v6only;
-		bsd_setsockopt_args.valsize = sizeof(*v6only);
+		int v6only;
+
+		v6only = 0;
 		/* We ignore any error returned by setsockopt() */
-		setsockopt(td, &bsd_setsockopt_args);
-		/* Copy back the return value from socket() */
-		td->td_retval[0] = bsd_setsockopt_args.s;
+		kern_setsockopt(td, td->td_retval[0], IPPROTO_IPV6, IPV6_V6ONLY,
+		    &v6only, UIO_SYSSPACE, sizeof(v6only));
 	}
 #endif
 
@@ -916,15 +871,14 @@ linux_sendto(struct thread *td, struct linux_sendto_args *args)
 	struct linux_sendto_args linux_args;
 	struct msghdr msg;
 	struct iovec aiov;
-	caddr_t sg = stackgap_init();
 	int error;
 
 	if ((error = copyin(args, &linux_args, sizeof(linux_args))))
 		return (error);
 
-	if (linux_check_hdrincl(td, &sg, linux_args.s) == 0)
+	if (linux_check_hdrincl(td, linux_args.s) == 0)
 		/* IP_HDRINCL set, tweak the packet before sending */
-		return (linux_sendto_hdrincl(td, &sg, &linux_args));
+		return (linux_sendto_hdrincl(td, &linux_args));
 
 	msg.msg_name = linux_args.to;
 	msg.msg_namelen = linux_args.tolen;
diff --git a/sys/kern/uipc_syscalls.c b/sys/kern/uipc_syscalls.c
index eeff60d..238dee4 100644
--- a/sys/kern/uipc_syscalls.c
+++ b/sys/kern/uipc_syscalls.c
@@ -1242,23 +1242,48 @@ setsockopt(td, uap)
 		int	valsize;
 	} */ *uap;
 {
+
+	return (kern_setsockopt(td, uap->s, uap->level, uap->name,
+	    uap->val, UIO_USERSPACE, uap->valsize));
+}
+
+int
+kern_setsockopt(td, s, level, name, val, valseg, valsize)
+	struct thread *td;
+	int s;
+	int level;
+	int name;
+	void *val;
+	enum uio_seg valseg;
+	socklen_t valsize;
+{
+	int error;
 	struct socket *so;
 	struct sockopt sopt;
-	int error;
 
-	if (uap->val == 0 && uap->valsize != 0)
+	if (val == NULL && valsize != 0)
 		return (EFAULT);
-	if (uap->valsize < 0)
+	if (valsize < 0)
 		return (EINVAL);
 
-	NET_LOCK_GIANT();
-	if ((error = fgetsock(td, uap->s, &so, NULL)) == 0) {
-		sopt.sopt_dir = SOPT_SET;
-		sopt.sopt_level = uap->level;
-		sopt.sopt_name = uap->name;
-		sopt.sopt_val = uap->val;
-		sopt.sopt_valsize = uap->valsize;
+	sopt.sopt_dir = SOPT_SET;
+	sopt.sopt_level = level;
+	sopt.sopt_name = name;
+	sopt.sopt_val = val;
+	sopt.sopt_valsize = valsize;
+	switch (valseg) {
+	case UIO_USERSPACE:
 		sopt.sopt_td = td;
+		break;
+	case UIO_SYSSPACE:
+		sopt.sopt_td = NULL;
+		break;
+	default:
+		panic("kern_setsockopt called with bad valseg");
+	}
+
+	NET_LOCK_GIANT();
+	if ((error = fgetsock(td, s, &so, NULL)) == 0) {
 		error = sosetopt(so, &sopt);
 		fputsock(so);
 	}
@@ -1283,39 +1308,66 @@ getsockopt(td, uap)
 {
 	socklen_t valsize;
 	int	error;
-	struct  socket *so;
-	struct	sockopt sopt;
 
-	NET_LOCK_GIANT();
-	if ((error = fgetsock(td, uap->s, &so, NULL)) != 0)
-		goto done2;
 	if (uap->val) {
 		error = copyin(uap->avalsize, &valsize, sizeof (valsize));
 		if (error)
-			goto done1;
-		if (valsize < 0) {
-			error = EINVAL;
-			goto done1;
-		}
-	} else {
-		valsize = 0;
+			return (error);
 	}
 
-	sopt.sopt_dir = SOPT_GET;
-	sopt.sopt_level = uap->level;
-	sopt.sopt_name = uap->name;
-	sopt.sopt_val = uap->val;
-	sopt.sopt_valsize = (size_t)valsize; /* checked non-negative above */
-	sopt.sopt_td = td;
+	error = kern_getsockopt(td, uap->s, uap->level, uap->name,
+	    uap->val, UIO_USERSPACE, &valsize);
 
-	error = sogetopt(so, &sopt);
-	if (error == 0) {
-		valsize = sopt.sopt_valsize;
+	if (error == 0)
 		error = copyout(&valsize, uap->avalsize, sizeof (valsize));
+	return (error);
+}
+
+/*
+ * Kernel version of getsockopt.
+ * optval can be a userland or userspace. optlen is always a kernel pointer.
+ */
+int
+kern_getsockopt(td, s, level, name, val, valseg, valsize)
+	struct thread *td;
+	int s;
+	int level;
+	int name;
+	void *val;
+	enum uio_seg valseg;
+	socklen_t *valsize;
+{
+	int error;
+	struct  socket *so;
+	struct	sockopt sopt;
+
+	if (val == NULL)
+		*valsize = 0;
+	if (*valsize < 0)
+		return (EINVAL);
+
+	sopt.sopt_dir = SOPT_GET;
+	sopt.sopt_level = level;
+	sopt.sopt_name = name;
+	sopt.sopt_val = val;
+	sopt.sopt_valsize = (size_t)*valsize; /* checked non-negative above */
+	switch (valseg) {
+	case UIO_USERSPACE:
+		sopt.sopt_td = td;
+		break;
+	case UIO_SYSSPACE:
+		sopt.sopt_td = NULL;
+		break;
+	default:
+		panic("kern_getsockopt called with bad valseg");
+	}
+
+	NET_LOCK_GIANT();
+	if ((error = fgetsock(td, s, &so, NULL)) == 0) {
+		error = sogetopt(so, &sopt);
+		*valsize = sopt.sopt_valsize;
+		fputsock(so);
 	}
-done1:
-	fputsock(so);
-done2:
 	NET_UNLOCK_GIANT();
 	return (error);
 }
diff --git a/sys/sys/syscallsubr.h b/sys/sys/syscallsubr.h
index 8c52027..82ca801 100644
--- a/sys/sys/syscallsubr.h
+++ b/sys/sys/syscallsubr.h
@@ -30,6 +30,7 @@
 
 #include <sys/signal.h>
 #include <sys/uio.h>
+#include <sys/socket.h>
 
 struct sockaddr;
 struct msghdr;
@@ -49,6 +50,8 @@ int	kern_connect(struct thread *td, int fd, struct sockaddr *sa);
 int	kern_fcntl(struct thread *td, int fd, int cmd, intptr_t arg);
 int	kern_futimes(struct thread *td, int fd, struct timeval *tptr,
 	    enum uio_seg tptrseg);
+int	kern_getsockopt(struct thread *td, int s, int level, int name,
+	    void *optval, enum uio_seg valseg, socklen_t *valsize);
 int	kern_lchown(struct thread *td, char *path, enum uio_seg pathseg,
 	    int uid, int gid);
 int	kern_link(struct thread *td, char *path, char *link,
@@ -74,6 +77,8 @@ int	kern_select(struct thread *td, int nd, fd_set *fd_in, fd_set *fd_ou,
 	    fd_set *fd_ex, struct timeval *tvp);
 int	kern_sendit(struct thread *td, int s, struct msghdr *mp, int flags,
 	    struct mbuf *control);
+int	kern_setsockopt(struct thread *td, int s, int level, int name,
+	    void *optval, enum uio_seg valseg, socklen_t valsize);
 int	kern_shmat(struct thread *td, int shmid, const void *shmaddr,
 	    int shmflg);
 int	kern_shmctl(struct thread *td, int shmid, int cmd, void *buf,