/*
- * Copyright (c) 2000-2010 Apple Inc. All rights reserved.
+ * Copyright (c) 2000-2015 Apple Inc. All rights reserved.
*
* @APPLE_OSREFERENCE_LICENSE_HEADER_START@
- *
+ *
* This file contains Original Code and/or Modifications of Original Code
* as defined in and that are subject to the Apple Public Source License
* Version 2.0 (the 'License'). You may not use this file except in
* unlawful or unlicensed copies of an Apple operating system, or to
* circumvent, violate, or enable the circumvention or violation of, any
* terms of an Apple operating system software license agreement.
- *
+ *
* Please obtain a copy of the License at
* http://www.opensource.apple.com/apsl/ and read it before using this file.
- *
+ *
* The Original Code and all software distributed under the License are
* distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
* EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
* FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
* Please see the License for the specific language governing rights and
* limitations under the License.
- *
+ *
* @APPLE_OSREFERENCE_LICENSE_HEADER_END@
*/
/*
#include <sys/file_internal.h>
#include <sys/vnode_internal.h>
#include <sys/malloc.h>
+#include <sys/mcache.h>
#include <sys/mbuf.h>
-#include <kern/lock.h>
+#include <kern/locks.h>
#include <sys/domain.h>
#include <sys/protosw.h>
#include <sys/signalvar.h>
#include <sys/uio_internal.h>
#include <sys/kauth.h>
#include <kern/task.h>
+#include <sys/priv.h>
+#include <sys/sysctl.h>
#include <security/audit/audit.h>
#endif /* MAC_SOCKET_SUBSET */
#define f_flag f_fglob->fg_flag
-#define f_type f_fglob->fg_type
+#define f_type f_fglob->fg_ops->fo_type
#define f_msgcount f_fglob->fg_msgcount
#define f_cred f_fglob->fg_cred
#define f_ops f_fglob->fg_ops
#define f_offset f_fglob->fg_offset
#define f_data f_fglob->fg_data
-
#define DBG_LAYER_IN_BEG NETDBG_CODE(DBG_NETSOCK, 0)
#define DBG_LAYER_IN_END NETDBG_CODE(DBG_NETSOCK, 2)
#define DBG_LAYER_OUT_BEG NETDBG_CODE(DBG_NETSOCK, 1)
#define DBG_FNC_SENDFILE_WAIT NETDBG_CODE(DBG_NETSOCK, ((10 << 8) | 1))
#define DBG_FNC_SENDFILE_READ NETDBG_CODE(DBG_NETSOCK, ((10 << 8) | 2))
#define DBG_FNC_SENDFILE_SEND NETDBG_CODE(DBG_NETSOCK, ((10 << 8) | 3))
+#define DBG_FNC_SENDMSG_X NETDBG_CODE(DBG_NETSOCK, (11 << 8))
+#define DBG_FNC_RECVMSG_X NETDBG_CODE(DBG_NETSOCK, (12 << 8))
-
-#define HACK_FOR_4056224 1
-#if HACK_FOR_4056224
-static pid_t last_pid_4056224 = 0;
-#endif /* HACK_FOR_4056224 */
+#if DEBUG || DEVELOPMENT
+#define DEBUG_KERNEL_ADDRPERM(_v) (_v)
+#define DBG_PRINTF(...) printf(__VA_ARGS__)
+#else
+#define DEBUG_KERNEL_ADDRPERM(_v) VM_KERNEL_ADDRPERM(_v)
+#define DBG_PRINTF(...) do { } while (0)
+#endif
/* TODO: should be in header file */
int falloc_locked(proc_t, struct fileproc **, int *, vfs_context_t, int);
-static int sendit(struct proc *, int, struct user_msghdr *, uio_t, int,
- int32_t *);
+static int sendit(struct proc *, struct socket *, struct user_msghdr *, uio_t,
+ int, int32_t *);
static int recvit(struct proc *, int, struct user_msghdr *, uio_t, user_addr_t,
int32_t *);
+static int connectit(struct socket *, struct sockaddr *);
static int getsockaddr(struct socket *, struct sockaddr **, user_addr_t,
size_t, boolean_t);
static int getsockaddr_s(struct socket *, struct sockaddr_storage *,
user_addr_t, size_t, boolean_t);
+static int getsockaddrlist(struct socket *, struct sockaddr_list **,
+ user_addr_t, socklen_t, boolean_t);
#if SENDFILE
static void alloc_sendpkt(int, size_t, unsigned int *, struct mbuf **,
boolean_t);
#endif /* SENDFILE */
+static int connectx_nocancel(struct proc *, struct connectx_args *, int *);
+static int connectitx(struct socket *, struct sockaddr_list **,
+ struct sockaddr_list **, struct proc *, uint32_t, sae_associd_t,
+ sae_connid_t *, uio_t, unsigned int, user_ssize_t *);
+static int peeloff_nocancel(struct proc *, struct peeloff_args *, int *);
+static int disconnectx_nocancel(struct proc *, struct disconnectx_args *,
+ int *);
+static int socket_common(struct proc *, int, int, int, pid_t, int32_t *, int);
+
+static int internalize_user_msghdr_array(const void *, int, int, u_int,
+ struct user_msghdr_x *, struct uio **);
+static u_int externalize_user_msghdr_array(void *, int, int, u_int,
+ const struct user_msghdr_x *, struct uio **);
+
+static void free_uio_array(struct uio **, u_int);
+static int uio_array_is_valid(struct uio **, u_int);
+static int recv_msg_array_is_valid(struct recv_msg_elem *, u_int);
+static int internalize_recv_msghdr_array(const void *, int, int,
+ u_int, struct user_msghdr_x *, struct recv_msg_elem *);
+static u_int externalize_recv_msghdr_array(void *, int, int, u_int,
+ const struct user_msghdr_x *, struct recv_msg_elem *);
+static struct recv_msg_elem *alloc_recv_msg_array(u_int count);
+static void free_recv_msg_array(struct recv_msg_elem *, u_int);
+
+SYSCTL_DECL(_kern_ipc);
+
+static u_int somaxsendmsgx = 100;
+SYSCTL_UINT(_kern_ipc, OID_AUTO, maxsendmsgx,
+ CTLFLAG_RW | CTLFLAG_LOCKED, &somaxsendmsgx, 0, "");
+static u_int somaxrecvmsgx = 100;
+SYSCTL_UINT(_kern_ipc, OID_AUTO, maxrecvmsgx,
+ CTLFLAG_RW | CTLFLAG_LOCKED, &somaxrecvmsgx, 0, "");
/*
* System call interface to the socket abstraction.
*/
-extern struct fileops socketops;
+extern const struct fileops socketops;
/*
* Returns: 0 Success
* socreate:EPROTONOSUPPORT
* socreate:ENOBUFS
* socreate:ENOMEM
- * socreate:EISCONN
* socreate:??? [other protocol families, IPSEC]
*/
int
-socket(struct proc *p, struct socket_args *uap, int32_t *retval)
+socket(struct proc *p,
+ struct socket_args *uap,
+ int32_t *retval)
+{
+ return (socket_common(p, uap->domain, uap->type, uap->protocol,
+ proc_selfpid(), retval, 0));
+}
+
+int
+socket_delegate(struct proc *p,
+ struct socket_delegate_args *uap,
+ int32_t *retval)
+{
+ return socket_common(p, uap->domain, uap->type, uap->protocol,
+ uap->epid, retval, 1);
+}
+
+static int
+socket_common(struct proc *p,
+ int domain,
+ int type,
+ int protocol,
+ pid_t epid,
+ int32_t *retval,
+ int delegate)
{
struct socket *so;
struct fileproc *fp;
int fd, error;
- AUDIT_ARG(socket, uap->domain, uap->type, uap->protocol);
+ AUDIT_ARG(socket, domain, type, protocol);
#if CONFIG_MACF_SOCKET_SUBSET
- if ((error = mac_socket_check_create(kauth_cred_get(), uap->domain,
- uap->type, uap->protocol)) != 0)
+ if ((error = mac_socket_check_create(kauth_cred_get(), domain,
+ type, protocol)) != 0)
return (error);
#endif /* MAC_SOCKET_SUBSET */
+ if (delegate) {
+ error = priv_check_cred(kauth_cred_get(),
+ PRIV_NET_PRIVILEGED_SOCKET_DELEGATE, 0);
+ if (error)
+ return (EACCES);
+ }
+
error = falloc(p, &fp, &fd, vfs_context_current());
if (error) {
return (error);
}
fp->f_flag = FREAD|FWRITE;
- fp->f_type = DTYPE_SOCKET;
fp->f_ops = &socketops;
- error = socreate(uap->domain, &so, uap->type, uap->protocol);
+ if (delegate)
+ error = socreate_delegate(domain, &so, type, protocol, epid);
+ else
+ error = socreate(domain, &so, type, protocol);
+
if (error) {
fp_free(p, fd, fp);
} else {
- thread_t thread;
- struct uthread *ut;
-
- thread = current_thread();
- ut = get_bsdthread_info(thread);
-
- /* if this is a backgrounded thread then throttle all new sockets */
-#if !CONFIG_EMBEDDED
- if (proc_get_selfthread_isbackground() != 0)
-#else /* !CONFIG_EMBEDDED */
- if ( (ut->uu_flag & UT_BACKGROUND) != 0 )
-#endif /* !CONFIG_EMBEDDED */
- {
- so->so_traffic_mgt_flags |= TRAFFIC_MGT_SO_BACKGROUND;
- so->so_background_thread = thread;
- }
fp->f_data = (caddr_t)so;
proc_fdlock(p);
proc_fdunlock(p);
*retval = fd;
+ if (ENTR_SHOULDTRACE) {
+ KERNEL_ENERGYTRACE(kEnTrActKernSocket, DBG_FUNC_START,
+ fd, 0, (int64_t)VM_KERNEL_ADDRPERM(so));
+ }
}
return (error);
}
* getsockaddr:EINVAL Invalid argument
* getsockaddr:ENOMEM Not enough space
* getsockaddr:EFAULT Bad address
- * sobind:???
+ * sobindlock:???
*/
/* ARGSUSED */
int
AUDIT_ARG(sockaddr, vfs_context_cwd(vfs_context_current()), sa);
#if CONFIG_MACF_SOCKET_SUBSET
if ((error = mac_socket_check_bind(kauth_cred_get(), so, sa)) == 0)
- error = sobind(so, sa);
+ error = sobindlock(so, sa, 1); /* will lock socket */
#else
- error = sobind(so, sa);
+ error = sobindlock(so, sa, 1); /* will lock socket */
#endif /* MAC_SOCKET_SUBSET */
if (want_free)
FREE(sa, M_SONAME);
* the file descriptor should the MAC check fails.
*/
if ((error = mac_socket_check_accepted(kauth_cred_get(), so)) != 0) {
+ socket_lock(so, 1);
so->so_state &= ~(SS_NOFDREF | SS_COMP);
so->so_head = NULL;
+ socket_unlock(so, 1);
soclose(so);
/* Drop reference on listening socket */
sodereference(head);
error = falloc(p, &fp, &newfd, vfs_context_current());
if (error) {
/*
- * Probably ran out of file descriptors. Put the
- * unaccepted connection back onto the queue and
- * do another wakeup so some other process might
- * have a chance at it.
+ * Probably ran out of file descriptors.
+ *
+ * <rdar://problem/8554930>
+ * Don't put this back on the socket like we used to, that
+ * just causes the client to spin. Drop the socket.
*/
- socket_lock(head, 0);
- TAILQ_INSERT_HEAD(&head->so_comp, so, so_list);
- head->so_qlen++;
- wakeup_one((caddr_t)&head->so_timeo);
- socket_unlock(head, 1);
+ socket_lock(so, 1);
+ so->so_state &= ~(SS_NOFDREF | SS_COMP);
+ so->so_head = NULL;
+ socket_unlock(so, 1);
+ soclose(so);
+ sodereference(head);
goto out;
}
*retval = newfd;
- fp->f_type = DTYPE_SOCKET;
fp->f_flag = fflag;
fp->f_ops = &socketops;
fp->f_data = (caddr_t)so;
+
socket_lock(head, 0);
if (dosocklock)
socket_lock(so, 1);
+
so->so_state &= ~SS_COMP;
so->so_head = NULL;
+
+ /* Sync socket non-blocking/async state with file flags */
+ if (fp->f_flag & FNONBLOCK) {
+ so->so_state |= SS_NBIO;
+ } else {
+ so->so_state &= ~SS_NBIO;
+ }
+
+ if (fp->f_flag & FASYNC) {
+ so->so_state |= SS_ASYNC;
+ so->so_rcv.sb_flags |= SB_ASYNC;
+ so->so_snd.sb_flags |= SB_ASYNC;
+ } else {
+ so->so_state &= ~SS_ASYNC;
+ so->so_rcv.sb_flags &= ~SB_ASYNC;
+ so->so_snd.sb_flags &= ~SB_ASYNC;
+ }
+
(void) soacceptlock(so, &sa, 0);
socket_unlock(head, 1);
if (sa == NULL) {
out:
file_drop(fd);
+
+ if (error == 0 && ENTR_SHOULDTRACE) {
+ KERNEL_ENERGYTRACE(kEnTrActKernSocket, DBG_FUNC_START,
+ newfd, 0, (int64_t)VM_KERNEL_ADDRPERM(so));
+ }
return (error);
}
accept(struct proc *p, struct accept_args *uap, int32_t *retval)
{
__pthread_testcancel(1);
- return(accept_nocancel(p, (struct accept_nocancel_args *)uap, retval));
+ return (accept_nocancel(p, (struct accept_nocancel_args *)uap,
+ retval));
}
/*
connect(struct proc *p, struct connect_args *uap, int32_t *retval)
{
__pthread_testcancel(1);
- return(connect_nocancel(p, (struct connect_nocancel_args *)uap, retval));
+ return (connect_nocancel(p, (struct connect_nocancel_args *)uap,
+ retval));
}
int
-connect_nocancel(__unused proc_t p, struct connect_nocancel_args *uap, __unused int32_t *retval)
+connect_nocancel(proc_t p, struct connect_nocancel_args *uap, int32_t *retval)
{
+#pragma unused(p, retval)
struct socket *so;
struct sockaddr_storage ss;
struct sockaddr *sa = NULL;
- lck_mtx_t *mutex_held;
- boolean_t want_free = TRUE;
int error;
int fd = uap->s;
boolean_t dgram;
error = getsockaddr(so, &sa, uap->name, uap->namelen, !dgram);
} else {
error = getsockaddr_s(so, &ss, uap->name, uap->namelen, !dgram);
- if (error == 0) {
+ if (error == 0)
sa = (struct sockaddr *)&ss;
- want_free = FALSE;
- }
}
if (error != 0)
goto out;
- AUDIT_ARG(sockaddr, vfs_context_cwd(vfs_context_current()), sa);
-#if CONFIG_MACF_SOCKET_SUBSET
- if ((error = mac_socket_check_connect(kauth_cred_get(), so, sa)) != 0) {
- if (want_free)
- FREE(sa, M_SONAME);
+ error = connectit(so, sa);
+
+ if (sa != NULL && sa != SA(&ss))
+ FREE(sa, M_SONAME);
+ if (error == ERESTART)
+ error = EINTR;
+out:
+ file_drop(fd);
+ return (error);
+}
+
+static int
+connectx_nocancel(struct proc *p, struct connectx_args *uap, int *retval)
+{
+#pragma unused(p, retval)
+ struct sockaddr_list *src_sl = NULL, *dst_sl = NULL;
+ struct socket *so;
+ int error, error1, fd = uap->socket;
+ boolean_t dgram;
+ sae_connid_t cid = SAE_CONNID_ANY;
+ struct user32_sa_endpoints ep32;
+ struct user64_sa_endpoints ep64;
+ struct user_sa_endpoints ep;
+ user_ssize_t bytes_written = 0;
+ struct user_iovec *iovp;
+ uio_t auio = NULL;
+
+ AUDIT_ARG(fd, uap->socket);
+ error = file_socket(fd, &so);
+ if (error != 0)
+ return (error);
+ if (so == NULL) {
+ error = EBADF;
+ goto out;
+ }
+
+ if (uap->endpoints == USER_ADDR_NULL) {
+ error = EINVAL;
+ goto out;
+ }
+
+ if (IS_64BIT_PROCESS(p)) {
+ error = copyin(uap->endpoints, (caddr_t)&ep64, sizeof(ep64));
+ if (error != 0)
+ goto out;
+
+ ep.sae_srcif = ep64.sae_srcif;
+ ep.sae_srcaddr = ep64.sae_srcaddr;
+ ep.sae_srcaddrlen = ep64.sae_srcaddrlen;
+ ep.sae_dstaddr = ep64.sae_dstaddr;
+ ep.sae_dstaddrlen = ep64.sae_dstaddrlen;
+ } else {
+ error = copyin(uap->endpoints, (caddr_t)&ep32, sizeof(ep32));
+ if (error != 0)
+ goto out;
+
+ ep.sae_srcif = ep32.sae_srcif;
+ ep.sae_srcaddr = ep32.sae_srcaddr;
+ ep.sae_srcaddrlen = ep32.sae_srcaddrlen;
+ ep.sae_dstaddr = ep32.sae_dstaddr;
+ ep.sae_dstaddrlen = ep32.sae_dstaddrlen;
+ }
+
+ /*
+ * Ask getsockaddr{_s} to not translate AF_UNSPEC to AF_INET
+ * if this is a datagram socket; translate for other types.
+ */
+ dgram = (so->so_type == SOCK_DGRAM);
+
+ /*
+ * Get socket address(es) now before we obtain socket lock; use
+ * sockaddr_list for src address for convenience, if present,
+ * even though it won't hold more than one.
+ */
+ if (ep.sae_srcaddr != USER_ADDR_NULL && (error = getsockaddrlist(so,
+ &src_sl, (user_addr_t)(caddr_t)ep.sae_srcaddr, ep.sae_srcaddrlen,
+ dgram)) != 0)
goto out;
+
+ if (ep.sae_dstaddr == USER_ADDR_NULL) {
+ error = EINVAL;
+ goto out;
+ }
+
+ error = getsockaddrlist(so, &dst_sl, (user_addr_t)(caddr_t)ep.sae_dstaddr,
+ ep.sae_dstaddrlen, dgram);
+ if (error != 0)
+ goto out;
+
+ VERIFY(dst_sl != NULL &&
+ !TAILQ_EMPTY(&dst_sl->sl_head) && dst_sl->sl_cnt > 0);
+
+ if (uap->iov != USER_ADDR_NULL) {
+ /* Verify range before calling uio_create() */
+ if (uap->iovcnt <= 0 || uap->iovcnt > UIO_MAXIOV)
+ return (EINVAL);
+
+ if (uap->len == USER_ADDR_NULL)
+ return (EINVAL);
+
+ /* allocate a uio to hold the number of iovecs passed */
+ auio = uio_create(uap->iovcnt, 0,
+ (IS_64BIT_PROCESS(p) ? UIO_USERSPACE64 : UIO_USERSPACE32),
+ UIO_WRITE);
+
+ if (auio == NULL) {
+ error = ENOMEM;
+ goto out;
+ }
+
+ /*
+ * get location of iovecs within the uio.
+ * then copyin the iovecs from user space.
+ */
+ iovp = uio_iovsaddr(auio);
+ if (iovp == NULL) {
+ error = ENOMEM;
+ goto out;
+ }
+ error = copyin_user_iovec_array(uap->iov,
+ IS_64BIT_PROCESS(p) ? UIO_USERSPACE64 : UIO_USERSPACE32,
+ uap->iovcnt, iovp);
+ if (error != 0)
+ goto out;
+
+ /* finish setup of uio_t */
+ error = uio_calculateresid(auio);
+ if (error != 0) {
+ goto out;
+ }
+ }
+
+ error = connectitx(so, &src_sl, &dst_sl, p, ep.sae_srcif, uap->associd,
+ &cid, auio, uap->flags, &bytes_written);
+ if (error == ERESTART)
+ error = EINTR;
+
+ if (uap->len != USER_ADDR_NULL) {
+ error1 = copyout(&bytes_written, uap->len, sizeof (uap->len));
+ /* give precedence to connectitx errors */
+ if ((error1 != 0) && (error == 0))
+ error = error1;
+ }
+
+ if (uap->connid != USER_ADDR_NULL) {
+ error1 = copyout(&cid, uap->connid, sizeof (cid));
+ /* give precedence to connectitx errors */
+ if ((error1 != 0) && (error == 0))
+ error = error1;
+ }
+out:
+ file_drop(fd);
+ if (auio != NULL) {
+ uio_free(auio);
}
+ if (src_sl != NULL)
+ sockaddrlist_free(src_sl);
+ if (dst_sl != NULL)
+ sockaddrlist_free(dst_sl);
+ return (error);
+}
+
+int
+connectx(struct proc *p, struct connectx_args *uap, int *retval)
+{
+ /*
+ * Due to similiarity with a POSIX interface, define as
+ * an unofficial cancellation point.
+ */
+ __pthread_testcancel(1);
+ return (connectx_nocancel(p, uap, retval));
+}
+
+static int
+connectit(struct socket *so, struct sockaddr *sa)
+{
+ int error;
+
+ AUDIT_ARG(sockaddr, vfs_context_cwd(vfs_context_current()), sa);
+#if CONFIG_MACF_SOCKET_SUBSET
+ if ((error = mac_socket_check_connect(kauth_cred_get(), so, sa)) != 0)
+ return (error);
#endif /* MAC_SOCKET_SUBSET */
- socket_lock(so, 1);
+ socket_lock(so, 1);
if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
- if (want_free)
- FREE(sa, M_SONAME);
- socket_unlock(so, 1);
error = EALREADY;
goto out;
}
error = soconnectlock(so, sa, 0);
- if (error)
- goto bad;
+ if (error != 0) {
+ so->so_state &= ~SS_ISCONNECTING;
+ goto out;
+ }
if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
- if (want_free)
- FREE(sa, M_SONAME);
- socket_unlock(so, 1);
error = EINPROGRESS;
goto out;
}
while ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
+ lck_mtx_t *mutex_held;
+
if (so->so_proto->pr_getlock != NULL)
mutex_held = (*so->so_proto->pr_getlock)(so, 0);
else
mutex_held = so->so_proto->pr_domain->dom_mtx;
error = msleep((caddr_t)&so->so_timeo, mutex_held,
- PSOCK | PCATCH, "connect", 0);
- if ((so->so_state & SS_DRAINING)) {
+ PSOCK | PCATCH, __func__, 0);
+ if (so->so_state & SS_DRAINING) {
error = ECONNABORTED;
}
- if (error)
+ if (error != 0)
break;
}
if (error == 0) {
error = so->so_error;
so->so_error = 0;
}
-bad:
- so->so_state &= ~SS_ISCONNECTING;
- socket_unlock(so, 1);
- if (want_free)
- FREE(sa, M_SONAME);
- if (error == ERESTART)
- error = EINTR;
out:
- file_drop(fd);
+ socket_unlock(so, 1);
return (error);
}
-/*
- * Returns: 0 Success
- * socreate:EAFNOSUPPORT
- * socreate:EPROTOTYPE
- * socreate:EPROTONOSUPPORT
- * socreate:ENOBUFS
- * socreate:ENOMEM
- * socreate:EISCONN
- * socreate:??? [other protocol families, IPSEC]
- * falloc:ENFILE
- * falloc:EMFILE
- * falloc:ENOMEM
- * copyout:EFAULT
- * soconnect2:EINVAL
- * soconnect2:EPROTOTYPE
- * soconnect2:??? [other protocol families[
- */
-int
-socketpair(struct proc *p, struct socketpair_args *uap,
- __unused int32_t *retval)
+static int
+connectitx(struct socket *so, struct sockaddr_list **src_sl,
+ struct sockaddr_list **dst_sl, struct proc *p, uint32_t ifscope,
+ sae_associd_t aid, sae_connid_t *pcid, uio_t auio, unsigned int flags,
+ user_ssize_t *bytes_written)
{
- struct fileproc *fp1, *fp2;
- struct socket *so1, *so2;
- int fd, error, sv[2];
+ struct sockaddr_entry *se;
+ int error;
+#pragma unused (flags)
- AUDIT_ARG(socket, uap->domain, uap->type, uap->protocol);
- error = socreate(uap->domain, &so1, uap->type, uap->protocol);
- if (error)
- return (error);
- error = socreate(uap->domain, &so2, uap->type, uap->protocol);
- if (error)
- goto free1;
+ VERIFY(dst_sl != NULL && *dst_sl != NULL);
- error = falloc(p, &fp1, &fd, vfs_context_current());
- if (error) {
- goto free2;
+ TAILQ_FOREACH(se, &(*dst_sl)->sl_head, se_link) {
+ VERIFY(se->se_addr != NULL);
+ AUDIT_ARG(sockaddr, vfs_context_cwd(vfs_context_current()),
+ se->se_addr);
+#if CONFIG_MACF_SOCKET_SUBSET
+ if ((error = mac_socket_check_connect(kauth_cred_get(),
+ so, se->se_addr)) != 0)
+ return (error);
+#endif /* MAC_SOCKET_SUBSET */
}
- fp1->f_flag = FREAD|FWRITE;
- fp1->f_type = DTYPE_SOCKET;
- fp1->f_ops = &socketops;
- fp1->f_data = (caddr_t)so1;
- sv[0] = fd;
- error = falloc(p, &fp2, &fd, vfs_context_current());
- if (error) {
- goto free3;
+ socket_lock(so, 1);
+ if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
+ error = EALREADY;
+ goto out;
}
- fp2->f_flag = FREAD|FWRITE;
- fp2->f_type = DTYPE_SOCKET;
- fp2->f_ops = &socketops;
- fp2->f_data = (caddr_t)so2;
- sv[1] = fd;
- error = soconnect2(so1, so2);
- if (error) {
- goto free4;
- }
- if (uap->type == SOCK_DGRAM) {
- /*
- * Datagram socket connection is asymmetric.
- */
- error = soconnect2(so2, so1);
- if (error) {
- goto free4;
- }
- }
+ if ((so->so_proto->pr_flags & PR_DATA_IDEMPOTENT) &&
+ (flags & CONNECT_DATA_IDEMPOTENT))
+ so->so_flags1 |= SOF1_DATA_IDEMPOTENT;
- if ((error = copyout(sv, uap->rsv, 2 * sizeof (int))) != 0)
- goto free4;
+ /*
+ * Case 1: CONNECT_RESUME_ON_READ_WRITE set, no data.
+ * Case 2: CONNECT_RESUME_ON_READ_WRITE set, with data (user error)
+ * Case 3: CONNECT_RESUME_ON_READ_WRITE not set, with data
+ * Case 3 allows user to combine write with connect even if they have
+ * no use for TFO (such as regular TCP, and UDP).
+ * Case 4: CONNECT_RESUME_ON_READ_WRITE not set, no data (regular case)
+ */
+ if ((so->so_proto->pr_flags & PR_PRECONN_WRITE) &&
+ ((flags & CONNECT_RESUME_ON_READ_WRITE) || auio))
+ so->so_flags1 |= SOF1_PRECONNECT_DATA;
+
+ /*
+ * If a user sets data idempotent and does not pass an uio, or
+ * sets CONNECT_RESUME_ON_READ_WRITE, this is an error, reset
+ * SOF1_DATA_IDEMPOTENT.
+ */
+ if (!(so->so_flags1 & SOF1_PRECONNECT_DATA) &&
+ (so->so_flags1 & SOF1_DATA_IDEMPOTENT)) {
+ /* We should return EINVAL instead perhaps. */
+ so->so_flags1 &= ~SOF1_DATA_IDEMPOTENT;
+ }
+
+ error = soconnectxlocked(so, src_sl, dst_sl, p, ifscope,
+ aid, pcid, 0, NULL, 0, auio, bytes_written);
+ if (error != 0) {
+ so->so_state &= ~SS_ISCONNECTING;
+ goto out;
+ }
+ /*
+ * If, after the call to soconnectxlocked the flag is still set (in case
+ * data has been queued and the connect() has actually been triggered,
+ * it will have been unset by the transport), we exit immediately. There
+ * is no reason to wait on any event.
+ */
+ if (so->so_flags1 & SOF1_PRECONNECT_DATA) {
+ error = 0;
+ goto out;
+ }
+ if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
+ error = EINPROGRESS;
+ goto out;
+ }
+ while ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
+ lck_mtx_t *mutex_held;
+
+ if (so->so_proto->pr_getlock != NULL)
+ mutex_held = (*so->so_proto->pr_getlock)(so, 0);
+ else
+ mutex_held = so->so_proto->pr_domain->dom_mtx;
+ error = msleep((caddr_t)&so->so_timeo, mutex_held,
+ PSOCK | PCATCH, __func__, 0);
+ if (so->so_state & SS_DRAINING) {
+ error = ECONNABORTED;
+ }
+ if (error != 0)
+ break;
+ }
+ if (error == 0) {
+ error = so->so_error;
+ so->so_error = 0;
+ }
+out:
+ socket_unlock(so, 1);
+ return (error);
+}
+
+int
+peeloff(struct proc *p, struct peeloff_args *uap, int *retval)
+{
+ /*
+ * Due to similiarity with a POSIX interface, define as
+ * an unofficial cancellation point.
+ */
+ __pthread_testcancel(1);
+ return (peeloff_nocancel(p, uap, retval));
+}
+
+static int
+peeloff_nocancel(struct proc *p, struct peeloff_args *uap, int *retval)
+{
+ struct fileproc *fp;
+ struct socket *mp_so, *so = NULL;
+ int newfd, fd = uap->s;
+ short fflag; /* type must match fp->f_flag */
+ int error;
+
+ *retval = -1;
+
+ error = fp_getfsock(p, fd, &fp, &mp_so);
+ if (error != 0) {
+ if (error == EOPNOTSUPP)
+ error = ENOTSOCK;
+ goto out_nofile;
+ }
+ if (mp_so == NULL) {
+ error = EBADF;
+ goto out;
+ }
+
+ socket_lock(mp_so, 1);
+ error = sopeelofflocked(mp_so, uap->aid, &so);
+ if (error != 0) {
+ socket_unlock(mp_so, 1);
+ goto out;
+ }
+ VERIFY(so != NULL);
+ socket_unlock(mp_so, 0); /* keep ref on mp_so for us */
+
+ fflag = fp->f_flag;
+ error = falloc(p, &fp, &newfd, vfs_context_current());
+ if (error != 0) {
+ /* drop this socket (probably ran out of file descriptors) */
+ soclose(so);
+ sodereference(mp_so); /* our mp_so ref */
+ goto out;
+ }
+
+ fp->f_flag = fflag;
+ fp->f_ops = &socketops;
+ fp->f_data = (caddr_t)so;
+
+ /*
+ * If the socket has been marked as inactive by sosetdefunct(),
+ * disallow further operations on it.
+ */
+ if (so->so_flags & SOF_DEFUNCT) {
+ sodefunct(current_proc(), so,
+ SHUTDOWN_SOCKET_LEVEL_DISCONNECT_INTERNAL);
+ }
+
+ proc_fdlock(p);
+ procfdtbl_releasefd(p, newfd, NULL);
+ fp_drop(p, newfd, fp, 1);
+ proc_fdunlock(p);
+
+ sodereference(mp_so); /* our mp_so ref */
+ *retval = newfd;
+
+out:
+ file_drop(fd);
+
+out_nofile:
+ return (error);
+}
+
+int
+disconnectx(struct proc *p, struct disconnectx_args *uap, int *retval)
+{
+ /*
+ * Due to similiarity with a POSIX interface, define as
+ * an unofficial cancellation point.
+ */
+ __pthread_testcancel(1);
+ return (disconnectx_nocancel(p, uap, retval));
+}
+
+static int
+disconnectx_nocancel(struct proc *p, struct disconnectx_args *uap, int *retval)
+{
+#pragma unused(p, retval)
+ struct socket *so;
+ int fd = uap->s;
+ int error;
+
+ error = file_socket(fd, &so);
+ if (error != 0)
+ return (error);
+ if (so == NULL) {
+ error = EBADF;
+ goto out;
+ }
+
+ error = sodisconnectx(so, uap->aid, uap->cid);
+out:
+ file_drop(fd);
+ return (error);
+}
+
+/*
+ * Returns: 0 Success
+ * socreate:EAFNOSUPPORT
+ * socreate:EPROTOTYPE
+ * socreate:EPROTONOSUPPORT
+ * socreate:ENOBUFS
+ * socreate:ENOMEM
+ * socreate:EISCONN
+ * socreate:??? [other protocol families, IPSEC]
+ * falloc:ENFILE
+ * falloc:EMFILE
+ * falloc:ENOMEM
+ * copyout:EFAULT
+ * soconnect2:EINVAL
+ * soconnect2:EPROTOTYPE
+ * soconnect2:??? [other protocol families[
+ */
+int
+socketpair(struct proc *p, struct socketpair_args *uap,
+ __unused int32_t *retval)
+{
+ struct fileproc *fp1, *fp2;
+ struct socket *so1, *so2;
+ int fd, error, sv[2];
+
+ AUDIT_ARG(socket, uap->domain, uap->type, uap->protocol);
+ error = socreate(uap->domain, &so1, uap->type, uap->protocol);
+ if (error)
+ return (error);
+ error = socreate(uap->domain, &so2, uap->type, uap->protocol);
+ if (error)
+ goto free1;
+
+ error = falloc(p, &fp1, &fd, vfs_context_current());
+ if (error) {
+ goto free2;
+ }
+ fp1->f_flag = FREAD|FWRITE;
+ fp1->f_ops = &socketops;
+ fp1->f_data = (caddr_t)so1;
+ sv[0] = fd;
+
+ error = falloc(p, &fp2, &fd, vfs_context_current());
+ if (error) {
+ goto free3;
+ }
+ fp2->f_flag = FREAD|FWRITE;
+ fp2->f_ops = &socketops;
+ fp2->f_data = (caddr_t)so2;
+ sv[1] = fd;
+
+ error = soconnect2(so1, so2);
+ if (error) {
+ goto free4;
+ }
+ if (uap->type == SOCK_DGRAM) {
+ /*
+ * Datagram socket connection is asymmetric.
+ */
+ error = soconnect2(so2, so1);
+ if (error) {
+ goto free4;
+ }
+ }
+
+ if ((error = copyout(sv, uap->rsv, 2 * sizeof (int))) != 0)
+ goto free4;
proc_fdlock(p);
procfdtbl_releasefd(p, sv[0], NULL);
* sockargs:???
*/
static int
-sendit(struct proc *p, int s, struct user_msghdr *mp, uio_t uiop,
+sendit(struct proc *p, struct socket *so, struct user_msghdr *mp, uio_t uiop,
int flags, int32_t *retval)
{
struct mbuf *control = NULL;
struct sockaddr *to = NULL;
boolean_t want_free = TRUE;
int error;
- struct socket *so;
user_ssize_t len;
KERNEL_DEBUG(DBG_FNC_SENDIT | DBG_FUNC_START, 0, 0, 0, 0, 0);
- error = file_socket(s, &so);
- if (error) {
- KERNEL_DEBUG(DBG_FNC_SENDIT | DBG_FUNC_END, error, 0, 0, 0, 0);
- return (error);
- }
- if (so == NULL) {
- error = EBADF;
- goto out;
- }
if (mp->msg_name != USER_ADDR_NULL) {
if (mp->msg_namelen > sizeof (ss)) {
error = getsockaddr(so, &to, mp->msg_name,
* if a race condition occurs, it would simply result
* in an extra call to the MAC check function.
*/
- if (!(so->so_state & SS_ISCONNECTED) &&
+ if (to != NULL &&
+ !(so->so_state & SS_DEFUNCT) &&
(error = mac_socket_check_send(kauth_cred_get(), so, to)) != 0)
goto bad;
#endif /* MAC_SOCKET_SUBSET */
len = uio_resid(uiop);
- error = so->so_proto->pr_usrreqs->pru_sosend(so, to, uiop, 0, control,
- flags);
+ error = so->so_proto->pr_usrreqs->pru_sosend(so, to, uiop, 0,
+ control, flags);
if (error != 0) {
if (uio_resid(uiop) != len && (error == ERESTART ||
error == EINTR || error == EWOULDBLOCK))
FREE(to, M_SONAME);
out:
KERNEL_DEBUG(DBG_FNC_SENDIT | DBG_FUNC_END, error, 0, 0, 0, 0);
- file_drop(s);
+
return (error);
}
sendto(struct proc *p, struct sendto_args *uap, int32_t *retval)
{
__pthread_testcancel(1);
- return(sendto_nocancel(p, (struct sendto_nocancel_args *)uap, retval));
+ return (sendto_nocancel(p, (struct sendto_nocancel_args *)uap, retval));
}
int
-sendto_nocancel(struct proc *p, struct sendto_nocancel_args *uap, int32_t *retval)
+sendto_nocancel(struct proc *p,
+ struct sendto_nocancel_args *uap,
+ int32_t *retval)
{
struct user_msghdr msg;
int error;
uio_t auio = NULL;
+ struct socket *so;
KERNEL_DEBUG(DBG_FNC_SENDTO | DBG_FUNC_START, 0, 0, 0, 0, 0);
AUDIT_ARG(fd, uap->s);
(IS_64BIT_PROCESS(p) ? UIO_USERSPACE64 : UIO_USERSPACE32),
UIO_WRITE);
if (auio == NULL) {
- return (ENOMEM);
+ error = ENOMEM;
+ goto done;
}
uio_addiov(auio, uap->buf, uap->len);
msg.msg_control = 0;
msg.msg_flags = 0;
- error = sendit(p, uap->s, &msg, auio, uap->flags, retval);
+ error = file_socket(uap->s, &so);
+ if (error)
+ goto done;
- if (auio != NULL) {
- uio_free(auio);
+ if (so == NULL) {
+ error = EBADF;
+ } else {
+ error = sendit(p, so, &msg, auio, uap->flags, retval);
}
-#if HACK_FOR_4056224
- /*
- * Radar 4056224
- * Temporary workaround to let send() and recv() work over
- * a pipe for binary compatibility
- * This will be removed in the release following Tiger
- */
- if (error == ENOTSOCK) {
- struct fileproc *fp;
-
- if (fp_lookup(p, uap->s, &fp, 0) == 0) {
- (void) fp_drop(p, uap->s, fp, 0);
-
- if (fp->f_type == DTYPE_PIPE) {
- struct write_args write_uap;
- user_ssize_t write_retval;
-
- if (p->p_pid > last_pid_4056224) {
- last_pid_4056224 = p->p_pid;
-
- printf("%s[%d] uses send/recv "
- "on a pipe\n", p->p_comm, p->p_pid);
- }
-
- bzero(&write_uap, sizeof (struct write_args));
- write_uap.fd = uap->s;
- write_uap.cbuf = uap->buf;
- write_uap.nbyte = uap->len;
-
- error = write(p, &write_uap, &write_retval);
- *retval = (int)write_retval;
- }
- }
- }
-#endif /* HACK_FOR_4056224 */
+ file_drop(uap->s);
+done:
+ if (auio != NULL)
+ uio_free(auio);
KERNEL_DEBUG(DBG_FNC_SENDTO | DBG_FUNC_END, error, *retval, 0, 0, 0);
sendmsg(struct proc *p, struct sendmsg_args *uap, int32_t *retval)
{
__pthread_testcancel(1);
- return(sendmsg_nocancel(p, (struct sendmsg_nocancel_args *)uap, retval));
+ return (sendmsg_nocancel(p, (struct sendmsg_nocancel_args *)uap,
+ retval));
}
int
-sendmsg_nocancel(struct proc *p, struct sendmsg_nocancel_args *uap, int32_t *retval)
+sendmsg_nocancel(struct proc *p, struct sendmsg_nocancel_args *uap,
+ int32_t *retval)
{
struct user32_msghdr msg32;
struct user64_msghdr msg64;
int error;
uio_t auio = NULL;
struct user_iovec *iovp;
+ struct socket *so;
KERNEL_DEBUG(DBG_FNC_SENDMSG | DBG_FUNC_START, 0, 0, 0, 0, 0);
AUDIT_ARG(fd, uap->s);
user_msg.msg_iov = CAST_USER_ADDR_T(iovp);
/* finish setup of uio_t */
- uio_calculateresid(auio);
+ error = uio_calculateresid(auio);
+ if (error) {
+ goto done;
+ }
} else {
user_msg.msg_iov = 0;
}
/* msg_flags is ignored for send */
user_msg.msg_flags = 0;
- error = sendit(p, uap->s, &user_msg, auio, uap->flags, retval);
+ error = file_socket(uap->s, &so);
+ if (error) {
+ goto done;
+ }
+ if (so == NULL) {
+ error = EBADF;
+ } else {
+ error = sendit(p, so, &user_msg, auio, uap->flags, retval);
+ }
+ file_drop(uap->s);
done:
if (auio != NULL) {
uio_free(auio);
return (error);
}
-/*
- * Returns: 0 Success
- * ENOTSOCK
- * EINVAL
- * EBADF
- * EACCES Mandatory Access Control failure
- * copyout:EFAULT
- * fp_lookup:EBADF
- * <pru_soreceive>:ENOBUFS
- * <pru_soreceive>:ENOTCONN
- * <pru_soreceive>:EWOULDBLOCK
- * <pru_soreceive>:EFAULT
- * <pru_soreceive>:EINTR
+int
+sendmsg_x(struct proc *p, struct sendmsg_x_args *uap, user_ssize_t *retval)
+{
+ int error = 0;
+ struct user_msghdr_x *user_msg_x = NULL;
+ struct uio **uiop = NULL;
+ struct socket *so;
+ u_int i;
+ struct sockaddr *to = NULL;
+ user_ssize_t len_before = 0, len_after;
+ int need_drop = 0;
+ size_t size_of_msghdr;
+ void *umsgp = NULL;
+ u_int uiocnt;
+ int has_addr_or_ctl = 0;
+
+ KERNEL_DEBUG(DBG_FNC_SENDMSG_X | DBG_FUNC_START, 0, 0, 0, 0, 0);
+
+ error = file_socket(uap->s, &so);
+ if (error) {
+ goto out;
+ }
+ need_drop = 1;
+ if (so == NULL) {
+ error = EBADF;
+ goto out;
+ }
+
+ /*
+ * Input parameter range check
+ */
+ if (uap->cnt == 0 || uap->cnt > UIO_MAXIOV) {
+ error = EINVAL;
+ goto out;
+ }
+ /*
+ * Clip to max currently allowed
+ */
+ if (uap->cnt > somaxsendmsgx)
+ uap->cnt = somaxsendmsgx;
+
+ user_msg_x = _MALLOC(uap->cnt * sizeof(struct user_msghdr_x),
+ M_TEMP, M_WAITOK | M_ZERO);
+ if (user_msg_x == NULL) {
+ DBG_PRINTF("%s _MALLOC() user_msg_x failed\n", __func__);
+ error = ENOMEM;
+ goto out;
+ }
+ uiop = _MALLOC(uap->cnt * sizeof(struct uio *),
+ M_TEMP, M_WAITOK | M_ZERO);
+ if (uiop == NULL) {
+ DBG_PRINTF("%s _MALLOC() uiop failed\n", __func__);
+ error = ENOMEM;
+ goto out;
+ }
+
+ size_of_msghdr = IS_64BIT_PROCESS(p) ?
+ sizeof(struct user64_msghdr_x) : sizeof(struct user32_msghdr_x);
+
+ umsgp = _MALLOC(uap->cnt * size_of_msghdr,
+ M_TEMP, M_WAITOK | M_ZERO);
+ if (umsgp == NULL) {
+ printf("%s _MALLOC() user_msg_x failed\n", __func__);
+ error = ENOMEM;
+ goto out;
+ }
+ error = copyin(uap->msgp, umsgp, uap->cnt * size_of_msghdr);
+ if (error) {
+ DBG_PRINTF("%s copyin() failed\n", __func__);
+ goto out;
+ }
+ error = internalize_user_msghdr_array(umsgp,
+ IS_64BIT_PROCESS(p) ? UIO_USERSPACE64 : UIO_USERSPACE32,
+ UIO_WRITE, uap->cnt, user_msg_x, uiop);
+ if (error) {
+ DBG_PRINTF("%s copyin_user_msghdr_array() failed\n", __func__);
+ goto out;
+ }
+ /*
+ * Make sure the size of each message iovec and
+ * the aggregate size of all the iovec is valid
+ */
+ if (uio_array_is_valid(uiop, uap->cnt) == 0) {
+ error = EINVAL;
+ goto out;
+ }
+
+ /*
+ * Sanity check on passed arguments
+ */
+ for (i = 0; i < uap->cnt; i++) {
+ struct user_msghdr_x *mp = user_msg_x + i;
+
+ /*
+ * No flags on send message
+ */
+ if (mp->msg_flags != 0) {
+ error = EINVAL;
+ goto out;
+ }
+ /*
+ * No support for address or ancillary data (yet)
+ */
+ if (mp->msg_name != USER_ADDR_NULL || mp->msg_namelen != 0)
+ has_addr_or_ctl = 1;
+
+ if (mp->msg_control != USER_ADDR_NULL ||
+ mp->msg_controllen != 0)
+ has_addr_or_ctl = 1;
+
+#if CONFIG_MACF_SOCKET_SUBSET
+ /*
+ * We check the state without holding the socket lock;
+ * if a race condition occurs, it would simply result
+ * in an extra call to the MAC check function.
+ *
+ * Note: The following check is never true taken with the
+ * current limitation that we do not accept to pass an address,
+ * this is effectively placeholder code. If we add support for
+ * addresses, we will have to check every address.
+ */
+ if (to != NULL &&
+ !(so->so_state & SS_DEFUNCT) &&
+ (error = mac_socket_check_send(kauth_cred_get(), so, to))
+ != 0)
+ goto out;
+#endif /* MAC_SOCKET_SUBSET */
+ }
+
+ len_before = uio_array_resid(uiop, uap->cnt);
+
+ /*
+ * Feed list of packets at once only for connected socket without
+ * control message
+ */
+ if (so->so_proto->pr_usrreqs->pru_sosend_list !=
+ pru_sosend_list_notsupp &&
+ has_addr_or_ctl == 0 && somaxsendmsgx == 0) {
+ error = so->so_proto->pr_usrreqs->pru_sosend_list(so, uiop,
+ uap->cnt, uap->flags);
+ } else {
+ for (i = 0; i < uap->cnt; i++) {
+ struct user_msghdr_x *mp = user_msg_x + i;
+ struct user_msghdr user_msg;
+ uio_t auio = uiop[i];
+ int32_t tmpval;
+
+ user_msg.msg_flags = mp->msg_flags;
+ user_msg.msg_controllen = mp->msg_controllen;
+ user_msg.msg_control = mp->msg_control;
+ user_msg.msg_iovlen = mp->msg_iovlen;
+ user_msg.msg_iov = mp->msg_iov;
+ user_msg.msg_namelen = mp->msg_namelen;
+ user_msg.msg_name = mp->msg_name;
+
+ error = sendit(p, so, &user_msg, auio, uap->flags,
+ &tmpval);
+ if (error != 0)
+ break;
+ }
+ }
+ len_after = uio_array_resid(uiop, uap->cnt);
+
+ VERIFY(len_after <= len_before);
+
+ if (error != 0) {
+ if (len_after != len_before && (error == ERESTART ||
+ error == EINTR || error == EWOULDBLOCK ||
+ error == ENOBUFS))
+ error = 0;
+ /* Generation of SIGPIPE can be controlled per socket */
+ if (error == EPIPE && !(so->so_flags & SOF_NOSIGPIPE))
+ psignal(p, SIGPIPE);
+ }
+ if (error == 0) {
+ uiocnt = externalize_user_msghdr_array(umsgp,
+ IS_64BIT_PROCESS(p) ? UIO_USERSPACE64 : UIO_USERSPACE32,
+ UIO_WRITE, uap->cnt, user_msg_x, uiop);
+
+ *retval = (int)(uiocnt);
+ }
+out:
+ if (need_drop)
+ file_drop(uap->s);
+ if (umsgp != NULL)
+ _FREE(umsgp, M_TEMP);
+ if (uiop != NULL) {
+ free_uio_array(uiop, uap->cnt);
+ _FREE(uiop, M_TEMP);
+ }
+ if (user_msg_x != NULL)
+ _FREE(user_msg_x, M_TEMP);
+
+ KERNEL_DEBUG(DBG_FNC_SENDMSG_X | DBG_FUNC_END, error, 0, 0, 0, 0);
+
+ return (error);
+}
+
+
+static int
+copyout_sa(struct sockaddr *fromsa, user_addr_t name, socklen_t *namelen)
+{
+ int error = 0;
+ socklen_t sa_len = 0;
+ ssize_t len;
+
+ len = *namelen;
+ if (len <= 0 || fromsa == 0) {
+ len = 0;
+ } else {
+#ifndef MIN
+#define MIN(a, b) ((a) > (b) ? (b) : (a))
+#endif
+ sa_len = fromsa->sa_len;
+ len = MIN((unsigned int)len, sa_len);
+ error = copyout(fromsa, name, (unsigned)len);
+ if (error)
+ goto out;
+ }
+ *namelen = sa_len;
+out:
+ return (0);
+}
+
+static int
+copyout_control(struct proc *p, struct mbuf *m, user_addr_t control,
+ socklen_t *controllen, int *flags)
+{
+ int error = 0;
+ ssize_t len;
+ user_addr_t ctlbuf;
+
+ len = *controllen;
+ *controllen = 0;
+ ctlbuf = control;
+
+ while (m && len > 0) {
+ unsigned int tocopy;
+ struct cmsghdr *cp = mtod(m, struct cmsghdr *);
+ int cp_size = CMSG_ALIGN(cp->cmsg_len);
+ int buflen = m->m_len;
+
+ while (buflen > 0 && len > 0) {
+ /*
+ * SCM_TIMESTAMP hack because struct timeval has a
+ * different size for 32 bits and 64 bits processes
+ */
+ if (cp->cmsg_level == SOL_SOCKET && cp->cmsg_type == SCM_TIMESTAMP) {
+ unsigned char tmp_buffer[CMSG_SPACE(sizeof(struct user64_timeval))];
+ struct cmsghdr *tmp_cp = (struct cmsghdr *)(void *)tmp_buffer;
+ int tmp_space;
+ struct timeval *tv = (struct timeval *)(void *)CMSG_DATA(cp);
+
+ tmp_cp->cmsg_level = SOL_SOCKET;
+ tmp_cp->cmsg_type = SCM_TIMESTAMP;
+
+ if (proc_is64bit(p)) {
+ struct user64_timeval *tv64 = (struct user64_timeval *)(void *)CMSG_DATA(tmp_cp);
+
+ tv64->tv_sec = tv->tv_sec;
+ tv64->tv_usec = tv->tv_usec;
+
+ tmp_cp->cmsg_len = CMSG_LEN(sizeof(struct user64_timeval));
+ tmp_space = CMSG_SPACE(sizeof(struct user64_timeval));
+ } else {
+ struct user32_timeval *tv32 = (struct user32_timeval *)(void *)CMSG_DATA(tmp_cp);
+
+ tv32->tv_sec = tv->tv_sec;
+ tv32->tv_usec = tv->tv_usec;
+
+ tmp_cp->cmsg_len = CMSG_LEN(sizeof(struct user32_timeval));
+ tmp_space = CMSG_SPACE(sizeof(struct user32_timeval));
+ }
+ if (len >= tmp_space) {
+ tocopy = tmp_space;
+ } else {
+ *flags |= MSG_CTRUNC;
+ tocopy = len;
+ }
+ error = copyout(tmp_buffer, ctlbuf, tocopy);
+ if (error)
+ goto out;
+ } else {
+ if (cp_size > buflen) {
+ panic("cp_size > buflen, something"
+ "wrong with alignment!");
+ }
+ if (len >= cp_size) {
+ tocopy = cp_size;
+ } else {
+ *flags |= MSG_CTRUNC;
+ tocopy = len;
+ }
+ error = copyout((caddr_t) cp, ctlbuf, tocopy);
+ if (error)
+ goto out;
+ }
+
+ ctlbuf += tocopy;
+ len -= tocopy;
+
+ buflen -= cp_size;
+ cp = (struct cmsghdr *)(void *)
+ ((unsigned char *) cp + cp_size);
+ cp_size = CMSG_ALIGN(cp->cmsg_len);
+ }
+
+ m = m->m_next;
+ }
+ *controllen = ctlbuf - control;
+out:
+ return (error);
+}
+
+/*
+ * Returns: 0 Success
+ * ENOTSOCK
+ * EINVAL
+ * EBADF
+ * EACCES Mandatory Access Control failure
+ * copyout:EFAULT
+ * fp_lookup:EBADF
+ * <pru_soreceive>:ENOBUFS
+ * <pru_soreceive>:ENOTCONN
+ * <pru_soreceive>:EWOULDBLOCK
+ * <pru_soreceive>:EFAULT
+ * <pru_soreceive>:EINTR
* <pru_soreceive>:EBADF
* <pru_soreceive>:EINVAL
* <pru_soreceive>:EMSGSIZE
recvit(struct proc *p, int s, struct user_msghdr *mp, uio_t uiop,
user_addr_t namelenp, int32_t *retval)
{
- int len, error;
- struct mbuf *m, *control = 0;
- user_addr_t ctlbuf;
+ ssize_t len;
+ int error;
+ struct mbuf *control = 0;
struct socket *so;
struct sockaddr *fromsa = 0;
struct fileproc *fp;
* if a race condition occurs, it would simply result
* in an extra call to the MAC check function.
*/
- if (!(so->so_state & SS_ISCONNECTED) &&
+ if (!(so->so_state & SS_DEFUNCT) &&
+ !(so->so_state & SS_ISCONNECTED) &&
+ !(so->so_proto->pr_flags & PR_CONNREQUIRED) &&
(error = mac_socket_check_receive(kauth_cred_get(), so)) != 0)
goto out1;
#endif /* MAC_SOCKET_SUBSET */
error == EINTR || error == EWOULDBLOCK))
error = 0;
}
-
if (error)
goto out;
*retval = len - uio_resid(uiop);
- if (mp->msg_name) {
- socklen_t sa_len = 0;
- len = mp->msg_namelen;
- if (len <= 0 || fromsa == 0) {
- len = 0;
- } else {
-#ifndef MIN
-#define MIN(a, b) ((a) > (b) ? (b) : (a))
-#endif
- sa_len = fromsa->sa_len;
- len = MIN((unsigned int)len, sa_len);
- error = copyout(fromsa, mp->msg_name, (unsigned)len);
- if (error)
- goto out;
- }
- mp->msg_namelen = sa_len;
+ if (mp->msg_name) {
+ error = copyout_sa(fromsa, mp->msg_name, &mp->msg_namelen);
+ if (error)
+ goto out;
/* return the actual, untruncated address length */
if (namelenp &&
- (error = copyout((caddr_t)&sa_len, namelenp,
+ (error = copyout((caddr_t)&mp->msg_namelen, namelenp,
sizeof (int)))) {
goto out;
}
}
+
if (mp->msg_control) {
- len = mp->msg_controllen;
- m = control;
- mp->msg_controllen = 0;
- ctlbuf = mp->msg_control;
-
- while (m && len > 0) {
- unsigned int tocopy;
- struct cmsghdr *cp = mtod(m, struct cmsghdr *);
- int cp_size = CMSG_ALIGN(cp->cmsg_len);
- int buflen = m->m_len;
-
- while (buflen > 0 && len > 0) {
-
- /*
- SCM_TIMESTAMP hack because struct timeval has a
- * different size for 32 bits and 64 bits processes
- */
- if (cp->cmsg_level == SOL_SOCKET && cp->cmsg_type == SCM_TIMESTAMP) {
- unsigned char tmp_buffer[CMSG_SPACE(sizeof(struct user64_timeval))];
- struct cmsghdr *tmp_cp = (struct cmsghdr *)tmp_buffer;
- int tmp_space;
- struct timeval *tv = (struct timeval *)CMSG_DATA(cp);
-
- tmp_cp->cmsg_level = SOL_SOCKET;
- tmp_cp->cmsg_type = SCM_TIMESTAMP;
-
- if (proc_is64bit(p)) {
- struct user64_timeval *tv64 = (struct user64_timeval *)CMSG_DATA(tmp_cp);
-
- tv64->tv_sec = tv->tv_sec;
- tv64->tv_usec = tv->tv_usec;
-
- tmp_cp->cmsg_len = CMSG_LEN(sizeof(struct user64_timeval));
- tmp_space = CMSG_SPACE(sizeof(struct user64_timeval));
- } else {
- struct user32_timeval *tv32 = (struct user32_timeval *)CMSG_DATA(tmp_cp);
-
- tv32->tv_sec = tv->tv_sec;
- tv32->tv_usec = tv->tv_usec;
-
- tmp_cp->cmsg_len = CMSG_LEN(sizeof(struct user32_timeval));
- tmp_space = CMSG_SPACE(sizeof(struct user32_timeval));
- }
- if (len >= tmp_space) {
- tocopy = tmp_space;
- } else {
- mp->msg_flags |= MSG_CTRUNC;
- tocopy = len;
- }
- error = copyout(tmp_buffer, ctlbuf, tocopy);
- if (error)
- goto out;
-
- } else {
-
- if (cp_size > buflen) {
- panic("cp_size > buflen, something wrong with alignment!");
- }
-
- if (len >= cp_size) {
- tocopy = cp_size;
- } else {
- mp->msg_flags |= MSG_CTRUNC;
- tocopy = len;
- }
-
- error = copyout((caddr_t) cp, ctlbuf,
- tocopy);
- if (error)
- goto out;
- }
-
-
- ctlbuf += tocopy;
- len -= tocopy;
-
- buflen -= cp_size;
- cp = (struct cmsghdr *) ((unsigned char *) cp + cp_size);
- cp_size = CMSG_ALIGN(cp->cmsg_len);
- }
-
- m = m->m_next;
- }
- mp->msg_controllen = ctlbuf - mp->msg_control;
+ error = copyout_control(p, control, mp->msg_control,
+ &mp->msg_controllen, &mp->msg_flags);
}
out:
if (fromsa)
recvfrom(struct proc *p, struct recvfrom_args *uap, int32_t *retval)
{
__pthread_testcancel(1);
- return(recvfrom_nocancel(p, (struct recvfrom_nocancel_args *)uap, retval));
+ return (recvfrom_nocancel(p, (struct recvfrom_nocancel_args *)uap,
+ retval));
}
int
-recvfrom_nocancel(struct proc *p, struct recvfrom_nocancel_args *uap, int32_t *retval)
+recvfrom_nocancel(struct proc *p, struct recvfrom_nocancel_args *uap,
+ int32_t *retval)
{
struct user_msghdr msg;
int error;
uio_free(auio);
}
-#if HACK_FOR_4056224
- /*
- * Radar 4056224
- * Temporary workaround to let send() and recv() work over
- * a pipe for binary compatibility
- * This will be removed in the release following Tiger
- */
- if (error == ENOTSOCK && proc_is64bit(p) == 0) {
- struct fileproc *fp;
-
- if (fp_lookup(p, uap->s, &fp, 0) == 0) {
- (void) fp_drop(p, uap->s, fp, 0);
-
- if (fp->f_type == DTYPE_PIPE) {
- struct read_args read_uap;
- user_ssize_t read_retval;
-
- if (p->p_pid > last_pid_4056224) {
- last_pid_4056224 = p->p_pid;
-
- printf("%s[%d] uses send/recv on "
- "a pipe\n", p->p_comm, p->p_pid);
- }
-
- bzero(&read_uap, sizeof (struct read_args));
- read_uap.fd = uap->s;
- read_uap.cbuf = uap->buf;
- read_uap.nbyte = uap->len;
-
- error = read(p, &read_uap, &read_retval);
- *retval = (int)read_retval;
- }
- }
- }
-#endif /* HACK_FOR_4056224 */
-
KERNEL_DEBUG(DBG_FNC_RECVFROM | DBG_FUNC_END, error, 0, 0, 0, 0);
return (error);
recvmsg(struct proc *p, struct recvmsg_args *uap, int32_t *retval)
{
__pthread_testcancel(1);
- return(recvmsg_nocancel(p, (struct recvmsg_nocancel_args *)uap, retval));
+ return (recvmsg_nocancel(p, (struct recvmsg_nocancel_args *)uap,
+ retval));
}
int
-recvmsg_nocancel(struct proc *p, struct recvmsg_nocancel_args *uap, int32_t *retval)
+recvmsg_nocancel(struct proc *p, struct recvmsg_nocancel_args *uap,
+ int32_t *retval)
{
struct user32_msghdr msg32;
struct user64_msghdr msg64;
goto done;
}
- /*
- * get location of iovecs within the uio. then copyin the iovecs from
- * user space.
- */
- iovp = uio_iovsaddr(auio);
- if (iovp == NULL) {
- error = ENOMEM;
- goto done;
+ /*
+ * get location of iovecs within the uio. then copyin the iovecs from
+ * user space.
+ */
+ iovp = uio_iovsaddr(auio);
+ if (iovp == NULL) {
+ error = ENOMEM;
+ goto done;
+ }
+ uiov = user_msg.msg_iov;
+ user_msg.msg_iov = CAST_USER_ADDR_T(iovp);
+ error = copyin_user_iovec_array(uiov,
+ IS_64BIT_PROCESS(p) ? UIO_USERSPACE64 : UIO_USERSPACE32,
+ user_msg.msg_iovlen, iovp);
+ if (error)
+ goto done;
+
+ /* finish setup of uio_t */
+ error = uio_calculateresid(auio);
+ if (error) {
+ goto done;
+ }
+
+ error = recvit(p, uap->s, &user_msg, auio, 0, retval);
+ if (!error) {
+ user_msg.msg_iov = uiov;
+ if (IS_64BIT_PROCESS(p)) {
+ msg64.msg_flags = user_msg.msg_flags;
+ msg64.msg_controllen = user_msg.msg_controllen;
+ msg64.msg_control = user_msg.msg_control;
+ msg64.msg_iovlen = user_msg.msg_iovlen;
+ msg64.msg_iov = user_msg.msg_iov;
+ msg64.msg_namelen = user_msg.msg_namelen;
+ msg64.msg_name = user_msg.msg_name;
+ } else {
+ msg32.msg_flags = user_msg.msg_flags;
+ msg32.msg_controllen = user_msg.msg_controllen;
+ msg32.msg_control = user_msg.msg_control;
+ msg32.msg_iovlen = user_msg.msg_iovlen;
+ msg32.msg_iov = user_msg.msg_iov;
+ msg32.msg_namelen = user_msg.msg_namelen;
+ msg32.msg_name = user_msg.msg_name;
+ }
+ error = copyout(msghdrp, uap->msg, size_of_msghdr);
+ }
+done:
+ if (auio != NULL) {
+ uio_free(auio);
+ }
+ KERNEL_DEBUG(DBG_FNC_RECVMSG | DBG_FUNC_END, error, 0, 0, 0, 0);
+ return (error);
+}
+
+int
+recvmsg_x(struct proc *p, struct recvmsg_x_args *uap, user_ssize_t *retval)
+{
+ int error = EOPNOTSUPP;
+ struct user_msghdr_x *user_msg_x = NULL;
+ struct recv_msg_elem *recv_msg_array = NULL;
+ struct socket *so;
+ user_ssize_t len_before = 0, len_after;
+ int need_drop = 0;
+ size_t size_of_msghdr;
+ void *umsgp = NULL;
+ u_int i;
+ u_int uiocnt;
+
+ KERNEL_DEBUG(DBG_FNC_RECVMSG_X | DBG_FUNC_START, 0, 0, 0, 0, 0);
+
+ error = file_socket(uap->s, &so);
+ if (error) {
+ goto out;
+ }
+ need_drop = 1;
+ if (so == NULL) {
+ error = EBADF;
+ goto out;
+ }
+ /*
+ * Input parameter range check
+ */
+ if (uap->cnt == 0 || uap->cnt > UIO_MAXIOV) {
+ error = EINVAL;
+ goto out;
+ }
+ if (uap->cnt > somaxrecvmsgx)
+ uap->cnt = somaxrecvmsgx;
+
+ user_msg_x = _MALLOC(uap->cnt * sizeof(struct user_msghdr_x),
+ M_TEMP, M_WAITOK | M_ZERO);
+ if (user_msg_x == NULL) {
+ DBG_PRINTF("%s _MALLOC() user_msg_x failed\n", __func__);
+ error = ENOMEM;
+ goto out;
+ }
+ recv_msg_array = alloc_recv_msg_array(uap->cnt);
+ if (recv_msg_array == NULL) {
+ DBG_PRINTF("%s alloc_recv_msg_array() failed\n", __func__);
+ error = ENOMEM;
+ goto out;
+ }
+ size_of_msghdr = IS_64BIT_PROCESS(p) ?
+ sizeof(struct user64_msghdr_x) : sizeof(struct user32_msghdr_x);
+
+ umsgp = _MALLOC(uap->cnt * size_of_msghdr, M_TEMP, M_WAITOK | M_ZERO);
+ if (umsgp == NULL) {
+ DBG_PRINTF("%s _MALLOC() umsgp failed\n", __func__);
+ error = ENOMEM;
+ goto out;
+ }
+ error = copyin(uap->msgp, umsgp, uap->cnt * size_of_msghdr);
+ if (error) {
+ DBG_PRINTF("%s copyin() failed\n", __func__);
+ goto out;
+ }
+ error = internalize_recv_msghdr_array(umsgp,
+ IS_64BIT_PROCESS(p) ? UIO_USERSPACE64 : UIO_USERSPACE32,
+ UIO_READ, uap->cnt, user_msg_x, recv_msg_array);
+ if (error) {
+ DBG_PRINTF("%s copyin_user_msghdr_array() failed\n", __func__);
+ goto out;
+ }
+ /*
+ * Make sure the size of each message iovec and
+ * the aggregate size of all the iovec is valid
+ */
+ if (recv_msg_array_is_valid(recv_msg_array, uap->cnt) == 0) {
+ error = EINVAL;
+ goto out;
+ }
+ /*
+ * Sanity check on passed arguments
+ */
+ for (i = 0; i < uap->cnt; i++) {
+ struct user_msghdr_x *mp = user_msg_x + i;
+
+ if (mp->msg_flags != 0) {
+ error = EINVAL;
+ goto out;
+ }
+ }
+#if CONFIG_MACF_SOCKET_SUBSET
+ /*
+ * We check the state without holding the socket lock;
+ * if a race condition occurs, it would simply result
+ * in an extra call to the MAC check function.
+ */
+ if (!(so->so_state & SS_DEFUNCT) &&
+ !(so->so_state & SS_ISCONNECTED) &&
+ !(so->so_proto->pr_flags & PR_CONNREQUIRED) &&
+ (error = mac_socket_check_receive(kauth_cred_get(), so)) != 0)
+ goto out;
+#endif /* MAC_SOCKET_SUBSET */
+
+ len_before = recv_msg_array_resid(recv_msg_array, uap->cnt);
+
+ if (so->so_proto->pr_usrreqs->pru_soreceive_list !=
+ pru_soreceive_list_notsupp &&
+ somaxrecvmsgx == 0) {
+ error = so->so_proto->pr_usrreqs->pru_soreceive_list(so,
+ recv_msg_array, uap->cnt, &uap->flags);
+ } else {
+ int flags = uap->flags;
+
+ for (i = 0; i < uap->cnt; i++) {
+ struct recv_msg_elem *recv_msg_elem;
+ uio_t auio;
+ struct sockaddr **psa;
+ struct mbuf **controlp;
+
+ recv_msg_elem = recv_msg_array + i;
+ auio = recv_msg_elem->uio;
+
+ /*
+ * Do not block if we got at least one packet
+ */
+ if (i > 0)
+ flags |= MSG_DONTWAIT;
+
+ psa = (recv_msg_elem->which & SOCK_MSG_SA) ?
+ &recv_msg_elem->psa : NULL;
+ controlp = (recv_msg_elem->which & SOCK_MSG_CONTROL) ?
+ &recv_msg_elem->controlp : NULL;
+
+ error = so->so_proto->pr_usrreqs->pru_soreceive(so, psa,
+ auio, (struct mbuf **)0, controlp, &flags);
+ if (error)
+ break;
+ /*
+ * We have some data
+ */
+ recv_msg_elem->which |= SOCK_MSG_DATA;
+ /*
+ * Stop on partial copy
+ */
+ if (flags & (MSG_RCVMORE | MSG_TRUNC))
+ break;
+ }
+ if ((uap->flags & MSG_DONTWAIT) == 0)
+ flags &= ~MSG_DONTWAIT;
+ uap->flags = flags;
+ }
+
+ len_after = recv_msg_array_resid(recv_msg_array, uap->cnt);
+
+ if (error) {
+ if (len_after != len_before && (error == ERESTART ||
+ error == EINTR || error == EWOULDBLOCK))
+ error = 0;
+ else
+ goto out;
+ }
+
+ uiocnt = externalize_recv_msghdr_array(umsgp,
+ IS_64BIT_PROCESS(p) ? UIO_USERSPACE64 : UIO_USERSPACE32,
+ UIO_READ, uap->cnt, user_msg_x, recv_msg_array);
+
+ error = copyout(umsgp, uap->msgp, uap->cnt * size_of_msghdr);
+ if (error) {
+ DBG_PRINTF("%s copyout() failed\n", __func__);
+ goto out;
}
- uiov = user_msg.msg_iov;
- user_msg.msg_iov = CAST_USER_ADDR_T(iovp);
- error = copyin_user_iovec_array(uiov,
- IS_64BIT_PROCESS(p) ? UIO_USERSPACE64 : UIO_USERSPACE32,
- user_msg.msg_iovlen, iovp);
- if (error)
- goto done;
+ *retval = (int)(uiocnt);
- /* finish setup of uio_t */
- uio_calculateresid(auio);
+ for (i = 0; i < uap->cnt; i++) {
+ struct user_msghdr_x *mp = user_msg_x + i;
+ struct recv_msg_elem *recv_msg_elem = recv_msg_array + i;
+ struct sockaddr *fromsa = recv_msg_elem->psa;
- error = recvit(p, uap->s, &user_msg, auio, 0, retval);
- if (!error) {
- user_msg.msg_iov = uiov;
- if (IS_64BIT_PROCESS(p)) {
- msg64.msg_flags = user_msg.msg_flags;
- msg64.msg_controllen = user_msg.msg_controllen;
- msg64.msg_control = user_msg.msg_control;
- msg64.msg_iovlen = user_msg.msg_iovlen;
- msg64.msg_iov = user_msg.msg_iov;
- msg64.msg_namelen = user_msg.msg_namelen;
- msg64.msg_name = user_msg.msg_name;
- } else {
- msg32.msg_flags = user_msg.msg_flags;
- msg32.msg_controllen = user_msg.msg_controllen;
- msg32.msg_control = user_msg.msg_control;
- msg32.msg_iovlen = user_msg.msg_iovlen;
- msg32.msg_iov = user_msg.msg_iov;
- msg32.msg_namelen = user_msg.msg_namelen;
- msg32.msg_name = user_msg.msg_name;
+ if (mp->msg_name) {
+ error = copyout_sa(fromsa, mp->msg_name,
+ &mp->msg_namelen);
+ if (error)
+ goto out;
+ }
+ if (mp->msg_control) {
+ error = copyout_control(p, recv_msg_elem->controlp,
+ mp->msg_control, &mp->msg_controllen,
+ &mp->msg_flags);
+ if (error)
+ goto out;
}
- error = copyout(msghdrp, uap->msg, size_of_msghdr);
- }
-done:
- if (auio != NULL) {
- uio_free(auio);
}
- KERNEL_DEBUG(DBG_FNC_RECVMSG | DBG_FUNC_END, error, 0, 0, 0, 0);
+out:
+ if (need_drop)
+ file_drop(uap->s);
+ if (umsgp != NULL)
+ _FREE(umsgp, M_TEMP);
+ if (recv_msg_array != NULL)
+ free_recv_msg_array(recv_msg_array, uap->cnt);
+ if (user_msg_x != NULL)
+ _FREE(user_msg_x, M_TEMP);
+
+ KERNEL_DEBUG(DBG_FNC_RECVMSG_X | DBG_FUNC_END, error, 0, 0, 0, 0);
+
return (error);
}
&sopt)) != 0)
goto out;
#endif /* MAC_SOCKET_SUBSET */
- error = sosetopt(so, &sopt);
+ error = sosetoptlock(so, &sopt, 1); /* will lock socket */
out:
file_drop(uap->s);
return (error);
&sopt)) != 0)
goto out;
#endif /* MAC_SOCKET_SUBSET */
- error = sogetopt((struct socket *)so, &sopt);
+ error = sogetoptlock((struct socket *)so, &sopt, 1); /* will lock */
if (error == 0) {
valsize = sopt.sopt_valsize;
error = copyout((caddr_t)&valsize, uap->avalsize,
struct mbuf *m;
int error;
- int alloc_buflen = buflen;
+ size_t alloc_buflen = (size_t)buflen;
+
+ if (alloc_buflen > INT_MAX/2)
+ return (EINVAL);
#ifdef __LP64__
- /* The fd's in the buffer must expand to be pointers, thus we need twice as much space */
- if(type == MT_CONTROL)
- alloc_buflen = ((buflen - sizeof(struct cmsghdr))*2) + sizeof(struct cmsghdr);
+ /*
+ * The fd's in the buffer must expand to be pointers, thus we need twice
+ * as much space
+ */
+ if (type == MT_CONTROL)
+ alloc_buflen = ((buflen - sizeof(struct cmsghdr))*2) +
+ sizeof(struct cmsghdr);
#endif
- if ((u_int)alloc_buflen > MLEN) {
- if (type == MT_SONAME && (u_int)alloc_buflen <= 112)
- alloc_buflen = MLEN; /* unix domain compat. hack */
- else if ((u_int)alloc_buflen > MCLBYTES)
+ if (alloc_buflen > MLEN) {
+ if (type == MT_SONAME && alloc_buflen <= 112)
+ alloc_buflen = MLEN; /* unix domain compat. hack */
+ else if (alloc_buflen > MCLBYTES)
return (EINVAL);
}
m = m_get(M_WAIT, type);
if (m == NULL)
return (ENOBUFS);
- if ((u_int)alloc_buflen > MLEN) {
+ if (alloc_buflen > MLEN) {
MCLGET(m, M_WAIT);
if ((m->m_flags & M_EXT) == 0) {
m_free(m);
return (ENOBUFS);
}
}
- /* K64: We still copyin the original buflen because it gets expanded later
- * and we lie about the size of the mbuf because it only affects unp_* functions
+ /*
+ * K64: We still copyin the original buflen because it gets expanded
+ * later and we lie about the size of the mbuf because it only affects
+ * unp_* functions
*/
m->m_len = buflen;
error = copyin(data, mtod(m, caddr_t), (u_int)buflen);
* handle it.
*/
if (translate_unspec && sa->sa_family == AF_UNSPEC &&
- INP_CHECK_SOCKAF(so, AF_INET) &&
+ SOCK_CHECK_DOM(so, PF_INET) &&
len == sizeof (struct sockaddr_in))
sa->sa_family = AF_INET;
* handle it.
*/
if (translate_unspec && ss->ss_family == AF_UNSPEC &&
- INP_CHECK_SOCKAF(so, AF_INET) &&
+ SOCK_CHECK_DOM(so, PF_INET) &&
len == sizeof (struct sockaddr_in))
ss->ss_family = AF_INET;
return (error);
}
-#if SENDFILE
+/*
+ * Hard limit on the number of source and/or destination addresses
+ * that can be specified by an application.
+ */
+#define SOCKADDRLIST_MAX_ENTRIES 64
-SYSCTL_DECL(_kern_ipc);
+static int
+getsockaddrlist(struct socket *so, struct sockaddr_list **slp,
+ user_addr_t uaddr, socklen_t uaddrlen, boolean_t xlate_unspec)
+{
+ struct sockaddr_list *sl;
+ int error = 0;
-#define SFUIOBUFS 64
-static int sendfileuiobufs = SFUIOBUFS;
-SYSCTL_INT(_kern_ipc, OID_AUTO, sendfileuiobufs, CTLFLAG_RW | CTLFLAG_LOCKED, &sendfileuiobufs,
- 0, "");
+ *slp = NULL;
-/* Macros to compute the number of mbufs needed depending on cluster size */
-#define HOWMANY_16K(n) ((((unsigned int)(n) - 1) >> (PGSHIFT + 2)) + 1)
-#define HOWMANY_4K(n) ((((unsigned int)(n) - 1) >> PGSHIFT) + 1)
+ if (uaddr == USER_ADDR_NULL || uaddrlen == 0 ||
+ uaddrlen > (sizeof(struct sockaddr_in6) * SOCKADDRLIST_MAX_ENTRIES))
+ return (EINVAL);
+
+ sl = sockaddrlist_alloc(M_WAITOK);
+ if (sl == NULL)
+ return (ENOMEM);
-/* Upper send limit in bytes (sendfileuiobufs * PAGESIZE) */
-#define SENDFILE_MAX_BYTES (sendfileuiobufs << PGSHIFT)
+ VERIFY(sl->sl_cnt == 0);
+ while (uaddrlen > 0 && sl->sl_cnt < SOCKADDRLIST_MAX_ENTRIES) {
+ struct sockaddr_storage ss;
+ struct sockaddr_entry *se;
+ struct sockaddr *sa;
-/* Upper send limit in the number of mbuf clusters */
-#define SENDFILE_MAX_16K HOWMANY_16K(SENDFILE_MAX_BYTES)
-#define SENDFILE_MAX_4K HOWMANY_4K(SENDFILE_MAX_BYTES)
+ if (uaddrlen < sizeof (struct sockaddr)) {
+ error = EINVAL;
+ break;
+ }
+
+ bzero(&ss, sizeof (ss));
+ error = copyin(uaddr, (caddr_t)&ss, sizeof (struct sockaddr));
+ if (error != 0)
+ break;
+
+ /* getsockaddr does the same but we need them now */
+ if (uaddrlen < ss.ss_len ||
+ ss.ss_len < offsetof(struct sockaddr, sa_data[0])) {
+ error = EINVAL;
+ break;
+ } else if (ss.ss_len > sizeof (ss)) {
+ /*
+ * sockaddr_storage size is less than SOCK_MAXADDRLEN,
+ * so the check here is inclusive. We could use the
+ * latter instead, but seems like an overkill for now.
+ */
+ error = ENAMETOOLONG;
+ break;
+ }
+
+ se = sockaddrentry_alloc(M_WAITOK);
+ if (se == NULL) {
+ error = ENOBUFS;
+ break;
+ }
+
+ sockaddrlist_insert(sl, se);
+
+ error = getsockaddr(so, &sa, uaddr, ss.ss_len, xlate_unspec);
+ if (error != 0)
+ break;
+
+ VERIFY(sa != NULL && sa->sa_len == ss.ss_len);
+ se->se_addr = sa;
+
+ uaddr += ss.ss_len;
+ VERIFY(((signed)uaddrlen - ss.ss_len) >= 0);
+ uaddrlen -= ss.ss_len;
+ }
+
+ if (error != 0)
+ sockaddrlist_free(sl);
+ else
+ *slp = sl;
+
+ return (error);
+}
+
+int
+internalize_user_msghdr_array(const void *src, int spacetype, int direction,
+ u_int count, struct user_msghdr_x *dst, struct uio **uiop)
+{
+ int error = 0;
+ u_int i;
+ u_int namecnt = 0;
+ u_int ctlcnt = 0;
+
+ for (i = 0; i < count; i++) {
+ uio_t auio;
+ struct user_iovec *iovp;
+ struct user_msghdr_x *user_msg = dst + i;
+
+ if (spacetype == UIO_USERSPACE64) {
+ const struct user64_msghdr_x *msghdr64;
+
+ msghdr64 = ((const struct user64_msghdr_x *)src) + i;
+
+ user_msg->msg_name = msghdr64->msg_name;
+ user_msg->msg_namelen = msghdr64->msg_namelen;
+ user_msg->msg_iov = msghdr64->msg_iov;
+ user_msg->msg_iovlen = msghdr64->msg_iovlen;
+ user_msg->msg_control = msghdr64->msg_control;
+ user_msg->msg_controllen = msghdr64->msg_controllen;
+ user_msg->msg_flags = msghdr64->msg_flags;
+ user_msg->msg_datalen = msghdr64->msg_datalen;
+ } else {
+ const struct user32_msghdr_x *msghdr32;
+
+ msghdr32 = ((const struct user32_msghdr_x *)src) + i;
+
+ user_msg->msg_name = msghdr32->msg_name;
+ user_msg->msg_namelen = msghdr32->msg_namelen;
+ user_msg->msg_iov = msghdr32->msg_iov;
+ user_msg->msg_iovlen = msghdr32->msg_iovlen;
+ user_msg->msg_control = msghdr32->msg_control;
+ user_msg->msg_controllen = msghdr32->msg_controllen;
+ user_msg->msg_flags = msghdr32->msg_flags;
+ user_msg->msg_datalen = msghdr32->msg_datalen;
+ }
+
+ if (user_msg->msg_iovlen <= 0 ||
+ user_msg->msg_iovlen > UIO_MAXIOV) {
+ error = EMSGSIZE;
+ goto done;
+ }
+ auio = uio_create(user_msg->msg_iovlen, 0, spacetype,
+ direction);
+ if (auio == NULL) {
+ error = ENOMEM;
+ goto done;
+ }
+ uiop[i] = auio;
+
+ iovp = uio_iovsaddr(auio);
+ if (iovp == NULL) {
+ error = ENOMEM;
+ goto done;
+ }
+ error = copyin_user_iovec_array(user_msg->msg_iov,
+ spacetype, user_msg->msg_iovlen, iovp);
+ if (error)
+ goto done;
+ user_msg->msg_iov = CAST_USER_ADDR_T(iovp);
+
+ error = uio_calculateresid(auio);
+ if (error)
+ goto done;
+ user_msg->msg_datalen = uio_resid(auio);
+
+ if (user_msg->msg_name && user_msg->msg_namelen)
+ namecnt++;
+ if (user_msg->msg_control && user_msg->msg_controllen)
+ ctlcnt++;
+ }
+done:
+
+ return (error);
+}
+
+int
+internalize_recv_msghdr_array(const void *src, int spacetype, int direction,
+ u_int count, struct user_msghdr_x *dst,
+ struct recv_msg_elem *recv_msg_array)
+{
+ int error = 0;
+ u_int i;
+
+ for (i = 0; i < count; i++) {
+ struct user_iovec *iovp;
+ struct user_msghdr_x *user_msg = dst + i;
+ struct recv_msg_elem *recv_msg_elem = recv_msg_array + i;
+
+ if (spacetype == UIO_USERSPACE64) {
+ const struct user64_msghdr_x *msghdr64;
+
+ msghdr64 = ((const struct user64_msghdr_x *)src) + i;
+
+ user_msg->msg_name = msghdr64->msg_name;
+ user_msg->msg_namelen = msghdr64->msg_namelen;
+ user_msg->msg_iov = msghdr64->msg_iov;
+ user_msg->msg_iovlen = msghdr64->msg_iovlen;
+ user_msg->msg_control = msghdr64->msg_control;
+ user_msg->msg_controllen = msghdr64->msg_controllen;
+ user_msg->msg_flags = msghdr64->msg_flags;
+ user_msg->msg_datalen = msghdr64->msg_datalen;
+ } else {
+ const struct user32_msghdr_x *msghdr32;
+
+ msghdr32 = ((const struct user32_msghdr_x *)src) + i;
+
+ user_msg->msg_name = msghdr32->msg_name;
+ user_msg->msg_namelen = msghdr32->msg_namelen;
+ user_msg->msg_iov = msghdr32->msg_iov;
+ user_msg->msg_iovlen = msghdr32->msg_iovlen;
+ user_msg->msg_control = msghdr32->msg_control;
+ user_msg->msg_controllen = msghdr32->msg_controllen;
+ user_msg->msg_flags = msghdr32->msg_flags;
+ user_msg->msg_datalen = msghdr32->msg_datalen;
+ }
+
+ if (user_msg->msg_iovlen <= 0 ||
+ user_msg->msg_iovlen > UIO_MAXIOV) {
+ error = EMSGSIZE;
+ goto done;
+ }
+ recv_msg_elem->uio = uio_create(user_msg->msg_iovlen, 0,
+ spacetype, direction);
+ if (recv_msg_elem->uio == NULL) {
+ error = ENOMEM;
+ goto done;
+ }
+
+ iovp = uio_iovsaddr(recv_msg_elem->uio);
+ if (iovp == NULL) {
+ error = ENOMEM;
+ goto done;
+ }
+ error = copyin_user_iovec_array(user_msg->msg_iov,
+ spacetype, user_msg->msg_iovlen, iovp);
+ if (error)
+ goto done;
+ user_msg->msg_iov = CAST_USER_ADDR_T(iovp);
+
+ error = uio_calculateresid(recv_msg_elem->uio);
+ if (error)
+ goto done;
+ user_msg->msg_datalen = uio_resid(recv_msg_elem->uio);
+
+ if (user_msg->msg_name && user_msg->msg_namelen)
+ recv_msg_elem->which |= SOCK_MSG_SA;
+ if (user_msg->msg_control && user_msg->msg_controllen)
+ recv_msg_elem->which |= SOCK_MSG_CONTROL;
+ }
+done:
+
+ return (error);
+}
+
+u_int
+externalize_user_msghdr_array(void *dst, int spacetype, int direction,
+ u_int count, const struct user_msghdr_x *src, struct uio **uiop)
+{
+#pragma unused(direction)
+ u_int i;
+ int seenlast = 0;
+ u_int retcnt = 0;
+
+ for (i = 0; i < count; i++) {
+ const struct user_msghdr_x *user_msg = src + i;
+ uio_t auio = uiop[i];
+ user_ssize_t len = user_msg->msg_datalen - uio_resid(auio);
+
+ if (user_msg->msg_datalen != 0 && len == 0)
+ seenlast = 1;
+
+ if (seenlast == 0)
+ retcnt ++;
+
+ if (spacetype == UIO_USERSPACE64) {
+ struct user64_msghdr_x *msghdr64;
+
+ msghdr64 = ((struct user64_msghdr_x *)dst) + i;
+
+ msghdr64->msg_flags = user_msg->msg_flags;
+ msghdr64->msg_datalen = len;
+
+ } else {
+ struct user32_msghdr_x *msghdr32;
+
+ msghdr32 = ((struct user32_msghdr_x *)dst) + i;
+
+ msghdr32->msg_flags = user_msg->msg_flags;
+ msghdr32->msg_datalen = len;
+ }
+ }
+ return (retcnt);
+}
+
+u_int
+externalize_recv_msghdr_array(void *dst, int spacetype, int direction,
+ u_int count, const struct user_msghdr_x *src,
+ struct recv_msg_elem *recv_msg_array)
+{
+ u_int i;
+ int seenlast = 0;
+ u_int retcnt = 0;
+
+ for (i = 0; i < count; i++) {
+ const struct user_msghdr_x *user_msg = src + i;
+ struct recv_msg_elem *recv_msg_elem = recv_msg_array + i;
+ user_ssize_t len;
+
+ len = user_msg->msg_datalen - uio_resid(recv_msg_elem->uio);
+
+ if (direction == UIO_READ) {
+ if ((recv_msg_elem->which & SOCK_MSG_DATA) == 0)
+ seenlast = 1;
+ } else {
+ if (user_msg->msg_datalen != 0 && len == 0)
+ seenlast = 1;
+ }
+
+ if (seenlast == 0)
+ retcnt ++;
+
+ if (spacetype == UIO_USERSPACE64) {
+ struct user64_msghdr_x *msghdr64;
+
+ msghdr64 = ((struct user64_msghdr_x *)dst) + i;
+
+ msghdr64->msg_flags = user_msg->msg_flags;
+ msghdr64->msg_datalen = len;
+
+ } else {
+ struct user32_msghdr_x *msghdr32;
+
+ msghdr32 = ((struct user32_msghdr_x *)dst) + i;
+
+ msghdr32->msg_flags = user_msg->msg_flags;
+ msghdr32->msg_datalen = len;
+ }
+ }
+ return (retcnt);
+}
+
+void
+free_uio_array(struct uio **uiop, u_int count)
+{
+ u_int i;
+
+ for (i = 0; i < count; i++) {
+ if (uiop[i] != NULL)
+ uio_free(uiop[i]);
+ }
+}
+
+__private_extern__ user_ssize_t
+uio_array_resid(struct uio **uiop, u_int count)
+{
+ user_ssize_t len = 0;
+ u_int i;
+
+ for (i = 0; i < count; i++) {
+ struct uio *auio = uiop[i];
+
+ if (auio != NULL)
+ len += uio_resid(auio);
+ }
+ return (len);
+}
+
+int
+uio_array_is_valid(struct uio **uiop, u_int count)
+{
+ user_ssize_t len = 0;
+ u_int i;
+
+ for (i = 0; i < count; i++) {
+ struct uio *auio = uiop[i];
+
+ if (auio != NULL) {
+ user_ssize_t resid = uio_resid(auio);
+
+ /*
+ * Sanity check on the validity of the iovec:
+ * no point of going over sb_max
+ */
+ if (resid < 0 || (u_int32_t)resid > sb_max)
+ return (0);
+
+ len += resid;
+ if (len < 0 || (u_int32_t)len > sb_max)
+ return (0);
+ }
+ }
+ return (1);
+}
+
+
+struct recv_msg_elem *
+alloc_recv_msg_array(u_int count)
+{
+ struct recv_msg_elem *recv_msg_array;
+
+ recv_msg_array = _MALLOC(count * sizeof(struct recv_msg_elem),
+ M_TEMP, M_WAITOK | M_ZERO);
+
+ return (recv_msg_array);
+}
+
+void
+free_recv_msg_array(struct recv_msg_elem *recv_msg_array, u_int count)
+{
+ u_int i;
+
+ for (i = 0; i < count; i++) {
+ struct recv_msg_elem *recv_msg_elem = recv_msg_array + i;
+
+ if (recv_msg_elem->uio != NULL)
+ uio_free(recv_msg_elem->uio);
+ if (recv_msg_elem->psa != NULL)
+ _FREE(recv_msg_elem->psa, M_TEMP);
+ if (recv_msg_elem->controlp != NULL)
+ m_freem(recv_msg_elem->controlp);
+ }
+ _FREE(recv_msg_array, M_TEMP);
+}
-size_t mbuf_pkt_maxlen(mbuf_t m);
-__private_extern__ size_t
-mbuf_pkt_maxlen(mbuf_t m)
+__private_extern__ user_ssize_t
+recv_msg_array_resid(struct recv_msg_elem *recv_msg_array, u_int count)
{
- size_t maxlen = 0;
+ user_ssize_t len = 0;
+ u_int i;
- while (m) {
- maxlen += mbuf_maxlen(m);
- m = mbuf_next(m);
+ for (i = 0; i < count; i++) {
+ struct recv_msg_elem *recv_msg_elem = recv_msg_array + i;
+
+ if (recv_msg_elem->uio != NULL)
+ len += uio_resid(recv_msg_elem->uio);
+ }
+ return (len);
+}
+
+int
+recv_msg_array_is_valid(struct recv_msg_elem *recv_msg_array, u_int count)
+{
+ user_ssize_t len = 0;
+ u_int i;
+
+ for (i = 0; i < count; i++) {
+ struct recv_msg_elem *recv_msg_elem = recv_msg_array + i;
+
+ if (recv_msg_elem->uio != NULL) {
+ user_ssize_t resid = uio_resid(recv_msg_elem->uio);
+
+ /*
+ * Sanity check on the validity of the iovec:
+ * no point of going over sb_max
+ */
+ if (resid < 0 || (u_int32_t)resid > sb_max)
+ return (0);
+
+ len += resid;
+ if (len < 0 || (u_int32_t)len > sb_max)
+ return (0);
+ }
}
- return (maxlen);
+ return (1);
}
+#if SENDFILE
+
+#define SFUIOBUFS 64
+
+/* Macros to compute the number of mbufs needed depending on cluster size */
+#define HOWMANY_16K(n) ((((unsigned int)(n) - 1) >> M16KCLSHIFT) + 1)
+#define HOWMANY_4K(n) ((((unsigned int)(n) - 1) >> MBIGCLSHIFT) + 1)
+
+/* Upper send limit in bytes (SFUIOBUFS * PAGESIZE) */
+#define SENDFILE_MAX_BYTES (SFUIOBUFS << PGSHIFT)
+
+/* Upper send limit in the number of mbuf clusters */
+#define SENDFILE_MAX_16K HOWMANY_16K(SENDFILE_MAX_BYTES)
+#define SENDFILE_MAX_4K HOWMANY_4K(SENDFILE_MAX_BYTES)
+
static void
alloc_sendpkt(int how, size_t pktlen, unsigned int *maxchunks,
struct mbuf **m, boolean_t jumbocl)
nuap.iovp = user_hdtr.headers;
nuap.iovcnt = user_hdtr.hdr_cnt;
error = writev_nocancel(p, &nuap, &writev_retval);
- if (error)
+ if (error) {
goto done2;
+ }
sbytes += writev_retval;
}
}
* 1. We don't want to allocate more mbufs than necessary
* 2. We don't want to read past the end of file
*/
- if ((error = vnode_size(vp, &file_size, vfs_context_current())) != 0)
+ if ((error = vnode_size(vp, &file_size, vfs_context_current())) != 0) {
goto done2;
+ }
/*
* Simply read file data into a chain of mbufs that used with scatter
* mbufs that point to the file pages.
*/
socket_lock(so, 1);
- error = sblock(&so->so_snd, M_WAIT);
+ error = sblock(&so->so_snd, SBL_WAIT);
if (error) {
socket_unlock(so, 1);
goto done2;
}
for (off = uap->offset; ; off += xfsize, sbytes += xfsize) {
mbuf_t m0 = NULL, m;
- unsigned int nbufs = sendfileuiobufs, i;
+ unsigned int nbufs = SFUIOBUFS, i;
uio_t auio;
- char uio_buf[UIO_SIZEOF(sendfileuiobufs)]; /* 1 KB !!! */
+ char uio_buf[UIO_SIZEOF(SFUIOBUFS)]; /* 1 KB !!! */
size_t uiolen;
user_ssize_t rlen;
off_t pgoff;
socket_unlock(so, 0);
alloc_sendpkt(M_WAIT, xfsize, &nbufs, &m0, jumbocl);
- pktlen = mbuf_pkt_maxlen(m0);
+ pktlen = mbuf_pkthdr_maxlen(m0);
if (pktlen < (size_t)xfsize)
xfsize = pktlen;
auio = uio_createwithbuffer(nbufs, off, UIO_SYSSPACE,
UIO_READ, &uio_buf[0], sizeof (uio_buf));
if (auio == NULL) {
- //printf("sendfile: uio_createwithbuffer failed\n");
+ printf("sendfile failed. nbufs = %d. %s", nbufs,
+ "File a radar related to rdar://10146739.\n");
mbuf_freem(m0);
error = ENXIO;
socket_lock(so, 0);
(unsigned int)(xfsize & 0x0ffffffff), 0, 0);
if (xfsize == 0) {
- //printf("sendfile: fo_read 0 bytes, EOF\n");
+ // printf("sendfile: fo_read 0 bytes, EOF\n");
break;
}
if (xfsize + off > file_size)
}
goto retry_space;
}
-
+
struct mbuf *control = NULL;
{
/*
goto done3;
}
}
- sbunlock(&so->so_snd, 0); /* will unlock socket */
+ sbunlock(&so->so_snd, FALSE); /* will unlock socket */
/*
* Send trailers. Wimp out and use writev(2).
*/
nuap.iovp = user_hdtr.trailers;
nuap.iovcnt = user_hdtr.trl_cnt;
error = writev_nocancel(p, &nuap, &writev_retval);
- if (error)
+ if (error) {
goto done2;
+ }
sbytes += writev_retval;
}
done2:
(unsigned int)(sbytes & 0x0ffffffff), error, 0);
return (error);
done3:
- sbunlock(&so->so_snd, 0); /* will unlock socket */
+ sbunlock(&so->so_snd, FALSE); /* will unlock socket */
goto done2;
}
projects / apple / xnu.git / blobdiff