X-Git-Url: https://git.saurik.com/apple/xnu.git/blobdiff_plain/43866e378188c25dd1e2208016ab3cbeb086ae6c..935ed37a5c468c8a1c07408573c08b8b7ef80e8b:/bsd/kern/uipc_usrreq.c diff --git a/bsd/kern/uipc_usrreq.c b/bsd/kern/uipc_usrreq.c index d8022230c..9b568d24c 100644 --- a/bsd/kern/uipc_usrreq.c +++ b/bsd/kern/uipc_usrreq.c @@ -1,16 +1,19 @@ /* - * Copyright (c) 2000-2001 Apple Computer, Inc. All rights reserved. + * Copyright (c) 2000-2007 Apple Inc. All rights reserved. * - * @APPLE_LICENSE_HEADER_START@ - * - * Copyright (c) 1999-2003 Apple Computer, 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 - * compliance with the License. Please obtain a copy of the License at - * http://www.opensource.apple.com/apsl/ and read it before using this - * file. + * compliance with the License. The rights granted to you under the License + * may not be used to create, or enable the creation or redistribution of, + * 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 @@ -20,7 +23,7 @@ * Please see the License for the specific language governing rights and * limitations under the License. * - * @APPLE_LICENSE_HEADER_END@ + * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ */ /* * Copyright (c) 1982, 1986, 1989, 1991, 1993 @@ -56,6 +59,12 @@ * * From: @(#)uipc_usrreq.c 8.3 (Berkeley) 1/4/94 */ +/* + * NOTICE: This file was modified by SPARTA, Inc. in 2005 to introduce + * support for mandatory and extensible security protections. This notice + * is included in support of clause 2.2 (b) of the Apple Public License, + * Version 2.0. + */ #include #include @@ -63,12 +72,13 @@ #include #include #include /* XXX must be before */ -#include +#include #include #include #include #include -#include +#include +#include #include #include #include @@ -76,14 +86,31 @@ #include #include #include -#include +#include +#include #include +#include + +#if CONFIG_MACF_SOCKET +#include +#endif /* MAC_SOCKET */ +#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 struct zone *unp_zone; static unp_gen_t unp_gencnt; static u_int unp_count; +static lck_attr_t *unp_mtx_attr; +static lck_grp_t *unp_mtx_grp; +static lck_grp_attr_t *unp_mtx_grp_attr; +static lck_rw_t *unp_list_mtx; + +extern lck_mtx_t *uipc_lock; static struct unp_head unp_shead, unp_dhead; /* @@ -95,22 +122,26 @@ static struct unp_head unp_shead, unp_dhead; * need a proper out-of-band * lock pushdown */ -static struct sockaddr sun_noname = { sizeof(sun_noname), AF_LOCAL }; +static struct sockaddr sun_noname = { sizeof (sun_noname), AF_LOCAL, { 0 } }; static ino_t unp_ino; /* prototype for fake inode numbers */ -static int unp_attach __P((struct socket *)); -static void unp_detach __P((struct unpcb *)); -static int unp_bind __P((struct unpcb *,struct sockaddr *, struct proc *)); -static int unp_connect __P((struct socket *,struct sockaddr *, - struct proc *)); -static void unp_disconnect __P((struct unpcb *)); -static void unp_shutdown __P((struct unpcb *)); -static void unp_drop __P((struct unpcb *, int)); -static void unp_gc __P((void)); -static void unp_scan __P((struct mbuf *, void (*)(struct file *))); -static void unp_mark __P((struct file *)); -static void unp_discard __P((struct file *)); -static int unp_internalize __P((struct mbuf *, struct proc *)); +static int unp_attach(struct socket *); +static void unp_detach(struct unpcb *); +static int unp_bind(struct unpcb *, struct sockaddr *, proc_t); +static int unp_connect(struct socket *, struct sockaddr *, proc_t); +static void unp_disconnect(struct unpcb *); +static void unp_shutdown(struct unpcb *); +static void unp_drop(struct unpcb *, int); +static void unp_gc(void); +static void unp_scan(struct mbuf *, void (*)(struct fileglob *)); +static void unp_mark(struct fileglob *); +static void unp_discard(struct fileglob *); +static void unp_discard_fdlocked(struct fileglob *, proc_t); +static int unp_internalize(struct mbuf *, proc_t); +static int unp_listen(struct unpcb *, proc_t); + +/* TODO: this should be in header file */ +extern int fdgetf_noref(proc_t, int, struct fileproc **); static int uipc_abort(struct socket *so) @@ -118,9 +149,11 @@ uipc_abort(struct socket *so) struct unpcb *unp = sotounpcb(so); if (unp == 0) - return EINVAL; + return (EINVAL); unp_drop(unp, ECONNABORTED); - return 0; + unp_detach(unp); + sofree(so); + return (0); } static int @@ -129,7 +162,7 @@ uipc_accept(struct socket *so, struct sockaddr **nam) struct unpcb *unp = sotounpcb(so); if (unp == 0) - return EINVAL; + return (EINVAL); /* * Pass back name of connected socket, @@ -137,54 +170,70 @@ uipc_accept(struct socket *so, struct sockaddr **nam) * (our peer may have closed already!). */ if (unp->unp_conn && unp->unp_conn->unp_addr) { - *nam = dup_sockaddr((struct sockaddr *)unp->unp_conn->unp_addr, - 1); + *nam = dup_sockaddr((struct sockaddr *) + unp->unp_conn->unp_addr, 1); } else { *nam = dup_sockaddr((struct sockaddr *)&sun_noname, 1); } - return 0; + return (0); } +/* + * Returns: 0 Success + * EISCONN + * unp_attach: + */ static int -uipc_attach(struct socket *so, int proto, struct proc *p) +uipc_attach(struct socket *so, __unused int proto, __unused proc_t p) { struct unpcb *unp = sotounpcb(so); if (unp != 0) - return EISCONN; - return unp_attach(so); + return (EISCONN); + return (unp_attach(so)); } static int -uipc_bind(struct socket *so, struct sockaddr *nam, struct proc *p) +uipc_bind(struct socket *so, struct sockaddr *nam, proc_t p) { struct unpcb *unp = sotounpcb(so); if (unp == 0) - return EINVAL; + return (EINVAL); - return unp_bind(unp, nam, p); + return (unp_bind(unp, nam, p)); } +/* + * Returns: 0 Success + * EINVAL + * unp_connect:??? [See elsewhere in this file] + */ static int -uipc_connect(struct socket *so, struct sockaddr *nam, struct proc *p) +uipc_connect(struct socket *so, struct sockaddr *nam, proc_t p) { struct unpcb *unp = sotounpcb(so); if (unp == 0) - return EINVAL; - return unp_connect(so, nam, p); + return (EINVAL); + return (unp_connect(so, nam, p)); } +/* + * Returns: 0 Success + * EINVAL + * unp_connect2:EPROTOTYPE Protocol wrong type for socket + * unp_connect2:EINVAL Invalid argument + */ static int uipc_connect2(struct socket *so1, struct socket *so2) { struct unpcb *unp = sotounpcb(so1); if (unp == 0) - return EINVAL; + return (EINVAL); - return unp_connect2(so1, so2); + return (unp_connect2(so1, so2)); } /* control is EOPNOTSUPP */ @@ -195,10 +244,10 @@ uipc_detach(struct socket *so) struct unpcb *unp = sotounpcb(so); if (unp == 0) - return EINVAL; + return (EINVAL); unp_detach(unp); - return 0; + return (0); } static int @@ -207,19 +256,23 @@ uipc_disconnect(struct socket *so) struct unpcb *unp = sotounpcb(so); if (unp == 0) - return EINVAL; + return (EINVAL); unp_disconnect(unp); - return 0; + return (0); } +/* + * Returns: 0 Success + * EINVAL + */ static int -uipc_listen(struct socket *so, struct proc *p) +uipc_listen(struct socket *so, __unused proc_t p) { struct unpcb *unp = sotounpcb(so); if (unp == 0 || unp->unp_vnode == 0) - return EINVAL; - return 0; + return (EINVAL); + return (unp_listen(unp, p)); } static int @@ -227,22 +280,25 @@ uipc_peeraddr(struct socket *so, struct sockaddr **nam) { struct unpcb *unp = sotounpcb(so); - if (unp == 0) - return EINVAL; - if (unp->unp_conn && unp->unp_conn->unp_addr) - *nam = dup_sockaddr((struct sockaddr *)unp->unp_conn->unp_addr, - 1); - return 0; + if (unp == NULL) + return (EINVAL); + if (unp->unp_conn != NULL && unp->unp_conn->unp_addr != NULL) { + *nam = dup_sockaddr((struct sockaddr *) + unp->unp_conn->unp_addr, 1); + } else { + *nam = dup_sockaddr((struct sockaddr *)&sun_noname, 1); + } + return (0); } static int -uipc_rcvd(struct socket *so, int flags) +uipc_rcvd(struct socket *so, __unused int flags) { struct unpcb *unp = sotounpcb(so); struct socket *so2; if (unp == 0) - return EINVAL; + return (EINVAL); switch (so->so_type) { case SOCK_DGRAM: panic("uipc_rcvd DGRAM?"); @@ -250,7 +306,7 @@ uipc_rcvd(struct socket *so, int flags) case SOCK_STREAM: #define rcv (&so->so_rcv) -#define snd (&so2->so_snd) +#define snd (&so2->so_snd) if (unp->unp_conn == 0) break; so2 = unp->unp_conn->unp_socket; @@ -270,14 +326,33 @@ uipc_rcvd(struct socket *so, int flags) default: panic("uipc_rcvd unknown socktype"); } - return 0; + return (0); } /* pru_rcvoob is EOPNOTSUPP */ +/* + * Returns: 0 Success + * EINVAL + * EOPNOTSUPP + * EPIPE + * ENOTCONN + * EISCONN + * unp_internalize:EINVAL + * unp_internalize:EBADF + * unp_connect:EAFNOSUPPORT Address family not supported + * unp_connect:EINVAL Invalid argument + * unp_connect:ENOTSOCK Not a socket + * unp_connect:ECONNREFUSED Connection refused + * unp_connect:EISCONN Socket is connected + * unp_connect:EPROTOTYPE Protocol wrong type for socket + * unp_connect:??? + * sbappendaddr:ENOBUFS [5th argument, contents modified] + * sbappendaddr:??? [whatever a filter author chooses] + */ static int uipc_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, - struct mbuf *control, struct proc *p) + struct mbuf *control, proc_t p) { int error = 0; struct unpcb *unp = sotounpcb(so); @@ -292,11 +367,17 @@ uipc_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, goto release; } - if (control && (error = unp_internalize(control, p))) - goto release; + if (control) { + /* release global lock to avoid deadlock (4436174) */ + socket_unlock(so, 0); + error = unp_internalize(control, p); + socket_lock(so, 0); + if (error) + goto release; + } switch (so->so_type) { - case SOCK_DGRAM: + case SOCK_DGRAM: { struct sockaddr *from; @@ -319,18 +400,29 @@ uipc_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, from = (struct sockaddr *)unp->unp_addr; else from = &sun_noname; - if (sbappendaddr(&so2->so_rcv, from, m, control)) { + /* + * sbappendaddr() will fail when the receiver runs out of + * space; in contrast to SOCK_STREAM, we will lose messages + * for the SOCK_DGRAM case when the receiver's queue overflows. + * SB_UNIX on the socket buffer implies that the callee will + * not free the control message, if any, because we would need + * to call unp_dispose() on it. + */ + if (sbappendaddr(&so2->so_rcv, from, m, control, &error)) { + control = NULL; sorwakeup(so2); - m = 0; - control = 0; - } else - error = ENOBUFS; + } else if (control != NULL && error == 0) { + /* A socket filter took control; don't touch it */ + control = NULL; + } + m = NULL; if (nam) unp_disconnect(unp); break; } - case SOCK_STREAM: + case SOCK_STREAM: { + int didreceive = 0; #define rcv (&so2->so_rcv) #define snd (&so->so_snd) /* Connect if not connected yet. */ @@ -357,24 +449,32 @@ uipc_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, panic("uipc_send connected but no connection?"); so2 = unp->unp_conn->unp_socket; /* - * Send to paired receive port, and then reduce - * send buffer hiwater marks to maintain backpressure. - * Wake up readers. + * Send to paired receive port, and then reduce send buffer + * hiwater marks to maintain backpressure. Wake up readers. + * SB_UNIX flag will allow new record to be appended to the + * receiver's queue even when it is already full. It is + * possible, however, that append might fail. In that case, + * we will need to call unp_dispose() on the control message; + * the callee will not free it since SB_UNIX is set. */ - if (control) { - if (sbappendcontrol(rcv, m, control)) - control = 0; - } else - sbappend(rcv, m); - snd->sb_mbmax -= - rcv->sb_mbcnt - unp->unp_conn->unp_mbcnt; + didreceive = control ? + sbappendcontrol(rcv, m, control, &error) : sbappend(rcv, m); + + snd->sb_mbmax -= rcv->sb_mbcnt - unp->unp_conn->unp_mbcnt; unp->unp_conn->unp_mbcnt = rcv->sb_mbcnt; snd->sb_hiwat -= rcv->sb_cc - unp->unp_conn->unp_cc; unp->unp_conn->unp_cc = rcv->sb_cc; - sorwakeup(so2); - m = 0; + if (didreceive) { + control = NULL; + sorwakeup(so2); + } else if (control != NULL && error == 0) { + /* A socket filter took control; don't touch it */ + control = NULL; + } + m = NULL; #undef snd #undef rcv + } break; default: @@ -390,56 +490,95 @@ uipc_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, unp_shutdown(unp); } + if (control && error != 0) { + socket_unlock(so, 0); + unp_dispose(control); + socket_lock(so, 0); + } + release: if (control) m_freem(control); if (m) m_freem(m); - return error; + return (error); } static int -uipc_sense(struct socket *so, struct stat *sb) +uipc_sense(struct socket *so, void *ub, int isstat64) { struct unpcb *unp = sotounpcb(so); struct socket *so2; + blksize_t blksize; if (unp == 0) - return EINVAL; - sb->st_blksize = so->so_snd.sb_hiwat; + return (EINVAL); + + blksize = so->so_snd.sb_hiwat; if (so->so_type == SOCK_STREAM && unp->unp_conn != 0) { so2 = unp->unp_conn->unp_socket; - sb->st_blksize += so2->so_rcv.sb_cc; + blksize += so2->so_rcv.sb_cc; } - sb->st_dev = NODEV; if (unp->unp_ino == 0) unp->unp_ino = unp_ino++; - sb->st_ino = unp->unp_ino; + + if (isstat64 != 0) { + struct stat64 *sb64; + + sb64 = (struct stat64 *)ub; + sb64->st_blksize = blksize; + sb64->st_dev = NODEV; + sb64->st_ino = (ino64_t)unp->unp_ino; + } else { + struct stat *sb; + + sb = (struct stat *)ub; + sb->st_blksize = blksize; + sb->st_dev = NODEV; + sb->st_ino = (ino_t)unp->unp_ino; + } + return (0); } +/* + * Returns: 0 Success + * EINVAL + * + * Notes: This is not strictly correct, as unp_shutdown() also calls + * socantrcvmore(). These should maybe both be conditionalized + * on the 'how' argument in soshutdown() as called from the + * shutdown() system call. + */ static int uipc_shutdown(struct socket *so) { struct unpcb *unp = sotounpcb(so); if (unp == 0) - return EINVAL; + return (EINVAL); socantsendmore(so); unp_shutdown(unp); - return 0; + return (0); } +/* + * Returns: 0 Success + * EINVAL Invalid argument + */ static int uipc_sockaddr(struct socket *so, struct sockaddr **nam) { struct unpcb *unp = sotounpcb(so); - if (unp == 0) - return EINVAL; - if (unp->unp_addr) + if (unp == NULL) + return (EINVAL); + if (unp->unp_addr != NULL) { *nam = dup_sockaddr((struct sockaddr *)unp->unp_addr, 1); - return 0; + } else { + *nam = dup_sockaddr((struct sockaddr *)&sun_noname, 1); + } + return (0); } struct pr_usrreqs uipc_usrreqs = { @@ -447,9 +586,42 @@ struct pr_usrreqs uipc_usrreqs = { uipc_connect2, pru_control_notsupp, uipc_detach, uipc_disconnect, uipc_listen, uipc_peeraddr, uipc_rcvd, pru_rcvoob_notsupp, uipc_send, uipc_sense, uipc_shutdown, uipc_sockaddr, - sosend, soreceive, sopoll + sosend, soreceive, pru_sopoll_notsupp }; - + +int +uipc_ctloutput(struct socket *so, struct sockopt *sopt) +{ + struct unpcb *unp = sotounpcb(so); + int error; + + switch (sopt->sopt_dir) { + case SOPT_GET: + switch (sopt->sopt_name) { + case LOCAL_PEERCRED: + if (unp->unp_flags & UNP_HAVEPC) { + error = sooptcopyout(sopt, &unp->unp_peercred, + sizeof (unp->unp_peercred)); + } else { + if (so->so_type == SOCK_STREAM) + error = ENOTCONN; + else + error = EINVAL; + } + break; + default: + error = EOPNOTSUPP; + break; + } + break; + case SOPT_SET: + default: + error = EOPNOTSUPP; + break; + } + return (error); +} + /* * Both send and receive buffers are allocated PIPSIZ bytes of buffering * for stream sockets, although the total for sender and receiver is @@ -467,26 +639,31 @@ static u_long unpdg_sendspace = 2*1024; /* really max datagram size */ static u_long unpdg_recvspace = 4*1024; static int unp_rights; /* file descriptors in flight */ +static int unp_disposed; /* discarded file descriptors */ SYSCTL_DECL(_net_local_stream); -SYSCTL_INT(_net_local_stream, OID_AUTO, sendspace, CTLFLAG_RW, - &unpst_sendspace, 0, ""); +SYSCTL_INT(_net_local_stream, OID_AUTO, sendspace, CTLFLAG_RW, + &unpst_sendspace, 0, ""); SYSCTL_INT(_net_local_stream, OID_AUTO, recvspace, CTLFLAG_RW, - &unpst_recvspace, 0, ""); + &unpst_recvspace, 0, ""); SYSCTL_DECL(_net_local_dgram); SYSCTL_INT(_net_local_dgram, OID_AUTO, maxdgram, CTLFLAG_RW, - &unpdg_sendspace, 0, ""); + &unpdg_sendspace, 0, ""); SYSCTL_INT(_net_local_dgram, OID_AUTO, recvspace, CTLFLAG_RW, - &unpdg_recvspace, 0, ""); + &unpdg_recvspace, 0, ""); SYSCTL_DECL(_net_local); SYSCTL_INT(_net_local, OID_AUTO, inflight, CTLFLAG_RD, &unp_rights, 0, ""); +/* + * Returns: 0 Success + * ENOBUFS + * soreserve:ENOBUFS + */ static int -unp_attach(so) - struct socket *so; +unp_attach(struct socket *so) { - register struct unpcb *unp; - int error; + struct unpcb *unp; + int error = 0; if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { switch (so->so_type) { @@ -505,41 +682,64 @@ unp_attach(so) if (error) return (error); } - unp = (struct unpcb*)zalloc(unp_zone); + unp = (struct unpcb *)zalloc(unp_zone); if (unp == NULL) return (ENOBUFS); - bzero(unp, sizeof *unp); - unp->unp_gencnt = ++unp_gencnt; - unp_count++; + bzero(unp, sizeof (*unp)); + lck_rw_lock_exclusive(unp_list_mtx); LIST_INIT(&unp->unp_refs); unp->unp_socket = so; - LIST_INSERT_HEAD(so->so_type == SOCK_DGRAM ? &unp_dhead - : &unp_shead, unp, unp_link); + unp->unp_gencnt = ++unp_gencnt; + unp_count++; + LIST_INSERT_HEAD(so->so_type == SOCK_DGRAM ? + &unp_dhead : &unp_shead, unp, unp_link); so->so_pcb = (caddr_t)unp; + /* + * Mark AF_UNIX socket buffers accordingly so that: + * + * a. In the SOCK_STREAM case, socket buffer append won't fail due to + * the lack of space; this essentially loosens the sbspace() check, + * since there is disconnect between sosend() and uipc_send() with + * respect to flow control that might result in our dropping the + * data in uipc_send(). By setting this, we allow for slightly + * more records to be appended to the receiving socket to avoid + * losing data (which we can't afford in the SOCK_STREAM case). + * Flow control still takes place since we adjust the sender's + * hiwat during each send. This doesn't affect the SOCK_DGRAM + * case and append would still fail when the queue overflows. + * + * b. In the presence of control messages containing internalized + * file descriptors, the append routines will not free them since + * we'd need to undo the work first via unp_dispose(). + */ + so->so_rcv.sb_flags |= SB_UNIX; + so->so_snd.sb_flags |= SB_UNIX; + lck_rw_done(unp_list_mtx); return (0); } static void -unp_detach(unp) - register struct unpcb *unp; +unp_detach(struct unpcb *unp) { + lck_rw_lock_exclusive(unp_list_mtx); LIST_REMOVE(unp, unp_link); unp->unp_gencnt = ++unp_gencnt; + lck_rw_done(unp_list_mtx); --unp_count; if (unp->unp_vnode) { struct vnode *tvp = unp->unp_vnode; - unp->unp_vnode->v_socket = 0; - unp->unp_vnode = 0; - thread_funnel_switch(NETWORK_FUNNEL, KERNEL_FUNNEL); - vrele(tvp); - thread_funnel_switch(KERNEL_FUNNEL, NETWORK_FUNNEL); + unp->unp_vnode->v_socket = NULL; + unp->unp_vnode = NULL; + vnode_rele(tvp); /* drop the usecount */ } if (unp->unp_conn) unp_disconnect(unp); while (unp->unp_refs.lh_first) unp_drop(unp->unp_refs.lh_first, ECONNRESET); soisdisconnected(unp->unp_socket); - unp->unp_socket->so_pcb = 0; + /* makes sure we're getting dealloced */ + unp->unp_socket->so_flags |= SOF_PCBCLEARING; + unp->unp_socket->so_pcb = NULL; if (unp_rights) { /* * Normally the receive buffer is flushed later, @@ -553,146 +753,249 @@ unp_detach(unp) } if (unp->unp_addr) FREE(unp->unp_addr, M_SONAME); - zfree(unp_zone, (vm_offset_t)unp); + zfree(unp_zone, unp); } +/* + * Returns: 0 Success + * EAFNOSUPPORT + * EINVAL + * EADDRINUSE + * namei:??? [anything namei can return] + * vnode_authorize:??? [anything vnode_authorize can return] + * + * Notes: p at this point is the current process, as this function is + * only called by sobind(). + */ static int -unp_bind(unp, nam, p) - struct unpcb *unp; - struct sockaddr *nam; - struct proc *p; +unp_bind( + struct unpcb *unp, + struct sockaddr *nam, + proc_t p) { struct sockaddr_un *soun = (struct sockaddr_un *)nam; - register struct vnode *vp; - struct vattr vattr; + struct vnode *vp, *dvp; + struct vnode_attr va; + vfs_context_t ctx = vfs_context_current(); int error, namelen; struct nameidata nd; char buf[SOCK_MAXADDRLEN]; + if (nam->sa_family != 0 && nam->sa_family != AF_UNIX) { + return (EAFNOSUPPORT); + } + if (unp->unp_vnode != NULL) return (EINVAL); namelen = soun->sun_len - offsetof(struct sockaddr_un, sun_path); if (namelen <= 0) - return EINVAL; - strncpy(buf, soun->sun_path, namelen); - buf[namelen] = 0; /* null-terminate the string */ - thread_funnel_switch(NETWORK_FUNNEL, KERNEL_FUNNEL); - NDINIT(&nd, CREATE, FOLLOW | LOCKPARENT, UIO_SYSSPACE, - buf, p); -/* SHOULD BE ABLE TO ADOPT EXISTING AND wakeup() ALA FIFO's */ + return (EINVAL); + + strlcpy(buf, soun->sun_path, namelen+1); + NDINIT(&nd, CREATE, FOLLOW | LOCKPARENT, UIO_SYSSPACE32, + CAST_USER_ADDR_T(buf), ctx); + /* SHOULD BE ABLE TO ADOPT EXISTING AND wakeup() ALA FIFO's */ error = namei(&nd); if (error) { - thread_funnel_switch(KERNEL_FUNNEL, NETWORK_FUNNEL); return (error); } + dvp = nd.ni_dvp; vp = nd.ni_vp; + if (vp != NULL) { - VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); - if (nd.ni_dvp == vp) - vrele(nd.ni_dvp); - else - vput(nd.ni_dvp); - vrele(vp); - thread_funnel_switch(KERNEL_FUNNEL, NETWORK_FUNNEL); + /* + * need to do this before the vnode_put of dvp + * since we may have to release an fs_nodelock + */ + nameidone(&nd); + + vnode_put(dvp); + vnode_put(vp); + return (EADDRINUSE); } - VATTR_NULL(&vattr); - vattr.va_type = VSOCK; - vattr.va_mode = (ACCESSPERMS & ~p->p_fd->fd_cmask); - VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); - error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); -#if 0 - /* In FreeBSD create leave s parent held ; not here */ - vput(nd.ni_dvp); -#endif + + VATTR_INIT(&va); + VATTR_SET(&va, va_type, VSOCK); + VATTR_SET(&va, va_mode, (ACCESSPERMS & ~p->p_fd->fd_cmask)); + +#if CONFIG_MACF_SOCKET + /* + * This is #if MAC_SOCKET, because it affects the connection rate + * of Unix domain dockets that is critical for server performance + */ + error = mac_vnode_check_create(ctx, + nd.ni_dvp, &nd.ni_cnd, &va); + + if (error == 0) +#endif /* MAC_SOCKET */ + /* authorize before creating */ + error = vnode_authorize(dvp, NULL, KAUTH_VNODE_ADD_FILE, ctx); + + if (!error) { + /* create the socket */ + error = vn_create(dvp, &vp, &nd.ni_cnd, &va, 0, ctx); + } + + nameidone(&nd); + vnode_put(dvp); + if (error) { - thread_funnel_switch(KERNEL_FUNNEL, NETWORK_FUNNEL); return (error); } - vp = nd.ni_vp; + vnode_ref(vp); /* gain a longterm reference */ vp->v_socket = unp->unp_socket; unp->unp_vnode = vp; unp->unp_addr = (struct sockaddr_un *)dup_sockaddr(nam, 1); - VOP_UNLOCK(vp, 0, p); - thread_funnel_switch(KERNEL_FUNNEL, NETWORK_FUNNEL); + vnode_put(vp); /* drop the iocount */ + return (0); } + +/* + * Returns: 0 Success + * EAFNOSUPPORT Address family not supported + * EINVAL Invalid argument + * ENOTSOCK Not a socket + * ECONNREFUSED Connection refused + * EPROTOTYPE Protocol wrong type for socket + * EISCONN Socket is connected + * unp_connect2:EPROTOTYPE Protocol wrong type for socket + * unp_connect2:EINVAL Invalid argument + * namei:??? [anything namei can return] + * vnode_authorize:???? [anything vnode_authorize can return] + * + * Notes: p at this point is the current process, as this function is + * only called by sosend(), sendfile(), and soconnectlock(). + */ static int -unp_connect(so, nam, p) - struct socket *so; - struct sockaddr *nam; - struct proc *p; +unp_connect(struct socket *so, struct sockaddr *nam, __unused proc_t p) { - register struct sockaddr_un *soun = (struct sockaddr_un *)nam; - register struct vnode *vp; - register struct socket *so2, *so3; - struct unpcb *unp2, *unp3; + struct sockaddr_un *soun = (struct sockaddr_un *)nam; + struct vnode *vp; + struct socket *so2, *so3; + struct unpcb *unp, *unp2, *unp3; + vfs_context_t ctx = vfs_context_current(); int error, len; struct nameidata nd; char buf[SOCK_MAXADDRLEN]; + if (nam->sa_family != 0 && nam->sa_family != AF_UNIX) { + return (EAFNOSUPPORT); + } + + so2 = so3 = NULL; + len = nam->sa_len - offsetof(struct sockaddr_un, sun_path); if (len <= 0) - return EINVAL; - strncpy(buf, soun->sun_path, len); - buf[len] = 0; + return (EINVAL); + + strlcpy(buf, soun->sun_path, len+1); - thread_funnel_switch(NETWORK_FUNNEL, KERNEL_FUNNEL); - NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, buf, p); + NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE32, + CAST_USER_ADDR_T(buf), ctx); error = namei(&nd); if (error) { - thread_funnel_switch(KERNEL_FUNNEL, NETWORK_FUNNEL); return (error); } + nameidone(&nd); vp = nd.ni_vp; if (vp->v_type != VSOCK) { error = ENOTSOCK; goto bad; } - error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p); + + error = vnode_authorize(vp, NULL, KAUTH_VNODE_WRITE_DATA, ctx); if (error) goto bad; so2 = vp->v_socket; - if (so2 == 0) { + if (so2 == 0 || so2->so_pcb == NULL) { error = ECONNREFUSED; goto bad; } + + /* make sure the socket can't go away while we're connecting */ + so2->so_usecount++; + if (so->so_type != so2->so_type) { error = EPROTOTYPE; goto bad; } - thread_funnel_switch(KERNEL_FUNNEL, NETWORK_FUNNEL); + + /* + * Check if socket was connected while we were trying to + * acquire the funnel. + * XXX - probably shouldn't return an error for SOCK_DGRAM + */ + if ((so->so_state & SS_ISCONNECTED) != 0) { + error = EISCONN; + goto bad; + } + if (so->so_proto->pr_flags & PR_CONNREQUIRED) { if ((so2->so_options & SO_ACCEPTCONN) == 0 || - (so3 = sonewconn(so2, 0)) == 0) { + (so3 = sonewconn(so2, 0, nam)) == 0) { error = ECONNREFUSED; - thread_funnel_switch(NETWORK_FUNNEL, KERNEL_FUNNEL); goto bad; } + unp = sotounpcb(so); unp2 = sotounpcb(so2); unp3 = sotounpcb(so3); if (unp2->unp_addr) unp3->unp_addr = (struct sockaddr_un *) - dup_sockaddr((struct sockaddr *) - unp2->unp_addr, 1); + dup_sockaddr((struct sockaddr *)unp2->unp_addr, 1); + + /* + * unp_peercred management: + * + * The connecter's (client's) credentials are copied + * from its process structure at the time of connect() + * (which is now). + */ + cru2x(vfs_context_ucred(ctx), &unp3->unp_peercred); + unp3->unp_flags |= UNP_HAVEPC; + /* + * The receiver's (server's) credentials are copied + * from the unp_peercred member of socket on which the + * former called listen(); unp_listen() cached that + * process's credentials at that time so we can use + * them now. + */ + KASSERT(unp2->unp_flags & UNP_HAVEPCCACHED, + ("unp_connect: listener without cached peercred")); + memcpy(&unp->unp_peercred, &unp2->unp_peercred, + sizeof (unp->unp_peercred)); + unp->unp_flags |= UNP_HAVEPC; + +#if CONFIG_MACF_SOCKET + /* XXXMAC: recursive lock: SOCK_LOCK(so); */ + mac_socketpeer_label_associate_socket(so, so3); + mac_socketpeer_label_associate_socket(so3, so); + /* XXXMAC: SOCK_UNLOCK(so); */ +#endif /* MAC_SOCKET */ + so2->so_usecount--; /* drop reference taken on so2 */ so2 = so3; + so3->so_usecount++; /* make sure we keep it around */ } error = unp_connect2(so, so2); - thread_funnel_switch(NETWORK_FUNNEL, KERNEL_FUNNEL); bad: - vput(vp); - thread_funnel_switch(KERNEL_FUNNEL, NETWORK_FUNNEL); + if (so2 != NULL) + so2->so_usecount--; /* release count on socket */ + vnode_put(vp); return (error); } +/* + * Returns: 0 Success + * EPROTOTYPE Protocol wrong type for socket + * EINVAL Invalid argument + */ int -unp_connect2(so, so2) - register struct socket *so; - register struct socket *so2; +unp_connect2(struct socket *so, struct socket *so2) { - register struct unpcb *unp = sotounpcb(so); - register struct unpcb *unp2; + struct unpcb *unp = sotounpcb(so); + struct unpcb *unp2; if (so2->so_type != so->so_type) return (EPROTOTYPE); @@ -711,6 +1014,15 @@ unp_connect2(so, so2) break; case SOCK_STREAM: + /* This takes care of socketpair */ + if (!(unp->unp_flags & UNP_HAVEPC) && + !(unp2->unp_flags & UNP_HAVEPC)) { + cru2x(kauth_cred_get(), &unp->unp_peercred); + unp->unp_flags |= UNP_HAVEPC; + + cru2x(kauth_cred_get(), &unp2->unp_peercred); + unp2->unp_flags |= UNP_HAVEPC; + } unp2->unp_conn = unp; soisconnected(so); soisconnected(so2); @@ -723,24 +1035,25 @@ unp_connect2(so, so2) } static void -unp_disconnect(unp) - struct unpcb *unp; +unp_disconnect(struct unpcb *unp) { - register struct unpcb *unp2 = unp->unp_conn; + struct unpcb *unp2 = unp->unp_conn; if (unp2 == 0) return; - unp->unp_conn = 0; + unp->unp_conn = NULL; switch (unp->unp_socket->so_type) { case SOCK_DGRAM: + lck_rw_lock_exclusive(unp_list_mtx); LIST_REMOVE(unp, unp_reflink); + lck_rw_done(unp_list_mtx); unp->unp_socket->so_state &= ~SS_ISCONNECTED; break; case SOCK_STREAM: soisdisconnected(unp->unp_socket); - unp2->unp_conn = 0; + unp2->unp_conn = NULL; soisdisconnected(unp2->unp_socket); break; } @@ -748,8 +1061,7 @@ unp_disconnect(unp) #ifdef notdef void -unp_abort(unp) - struct unpcb *unp; +unp_abort(struct unpcb *unp) { unp_detach(unp); @@ -759,27 +1071,32 @@ unp_abort(unp) static int unp_pcblist SYSCTL_HANDLER_ARGS { +#pragma unused(oidp,arg2) int error, i, n; struct unpcb *unp, **unp_list; unp_gen_t gencnt; struct xunpgen xug; struct unp_head *head; + lck_rw_lock_shared(unp_list_mtx); head = ((intptr_t)arg1 == SOCK_DGRAM ? &unp_dhead : &unp_shead); /* * The process of preparing the PCB list is too time-consuming and * resource-intensive to repeat twice on every request. */ - if (req->oldptr == 0) { + if (req->oldptr == USER_ADDR_NULL) { n = unp_count; - req->oldidx = 2 * (sizeof xug) - + (n + n/8) * sizeof(struct xunpcb); - return 0; + req->oldidx = 2 * sizeof (xug) + (n + n / 8) * + sizeof (struct xunpcb); + lck_rw_done(unp_list_mtx); + return (0); } - if (req->newptr != 0) - return EPERM; + if (req->newptr != USER_ADDR_NULL) { + lck_rw_done(unp_list_mtx); + return (EPERM); + } /* * OK, now we're committed to doing something. @@ -787,26 +1104,34 @@ unp_pcblist SYSCTL_HANDLER_ARGS gencnt = unp_gencnt; n = unp_count; - xug.xug_len = sizeof xug; + bzero(&xug, sizeof (xug)); + xug.xug_len = sizeof (xug); xug.xug_count = n; xug.xug_gen = gencnt; xug.xug_sogen = so_gencnt; - error = SYSCTL_OUT(req, &xug, sizeof xug); - if (error) - return error; + error = SYSCTL_OUT(req, &xug, sizeof (xug)); + if (error) { + lck_rw_done(unp_list_mtx); + return (error); + } /* * We are done if there is no pcb */ - if (n == 0) - return 0; + if (n == 0) { + lck_rw_done(unp_list_mtx); + return (0); + } + + MALLOC(unp_list, struct unpcb **, n * sizeof (*unp_list), + M_TEMP, M_WAITOK); + if (unp_list == 0) { + lck_rw_done(unp_list_mtx); + return (ENOMEM); + } - unp_list = _MALLOC(n * sizeof *unp_list, M_TEMP, M_WAITOK); - if (unp_list == 0) - return ENOMEM; - for (unp = head->lh_first, i = 0; unp && i < n; - unp = unp->unp_link.le_next) { + unp = unp->unp_link.le_next) { if (unp->unp_gencnt <= gencnt) unp_list[i++] = unp; } @@ -817,22 +1142,24 @@ unp_pcblist SYSCTL_HANDLER_ARGS unp = unp_list[i]; if (unp->unp_gencnt <= gencnt) { struct xunpcb xu; - xu.xu_len = sizeof xu; - xu.xu_unpp = unp; + + bzero(&xu, sizeof (xu)); + xu.xu_len = sizeof (xu); + xu.xu_unpp = (struct unpcb_compat *)unp; /* * XXX - need more locking here to protect against * connect/disconnect races for SMP. */ if (unp->unp_addr) - bcopy(unp->unp_addr, &xu.xu_addr, - unp->unp_addr->sun_len); + bcopy(unp->unp_addr, &xu.xu_addr, + unp->unp_addr->sun_len); if (unp->unp_conn && unp->unp_conn->unp_addr) bcopy(unp->unp_conn->unp_addr, - &xu.xu_caddr, - unp->unp_conn->unp_addr->sun_len); - bcopy(unp, &xu.xu_unp, sizeof *unp); + &xu.xu_caddr, + unp->unp_conn->unp_addr->sun_len); + bcopy(unp, &xu.xu_unp, sizeof (xu.xu_unp)); sotoxsocket(unp->unp_socket, &xu.xu_socket); - error = SYSCTL_OUT(req, &xu, sizeof xu); + error = SYSCTL_OUT(req, &xu, sizeof (xu)); } } if (!error) { @@ -843,25 +1170,27 @@ unp_pcblist SYSCTL_HANDLER_ARGS * while we were processing this request, and it * might be necessary to retry. */ + bzero(&xug, sizeof (xug)); + xug.xug_len = sizeof (xug); xug.xug_gen = unp_gencnt; xug.xug_sogen = so_gencnt; xug.xug_count = unp_count; - error = SYSCTL_OUT(req, &xug, sizeof xug); + error = SYSCTL_OUT(req, &xug, sizeof (xug)); } FREE(unp_list, M_TEMP); - return error; + lck_rw_done(unp_list_mtx); + return (error); } -SYSCTL_PROC(_net_local_dgram, OID_AUTO, pcblist, CTLFLAG_RD, +SYSCTL_PROC(_net_local_dgram, OID_AUTO, pcblist, CTLFLAG_RD, (caddr_t)(long)SOCK_DGRAM, 0, unp_pcblist, "S,xunpcb", "List of active local datagram sockets"); -SYSCTL_PROC(_net_local_stream, OID_AUTO, pcblist, CTLFLAG_RD, +SYSCTL_PROC(_net_local_stream, OID_AUTO, pcblist, CTLFLAG_RD, (caddr_t)(long)SOCK_STREAM, 0, unp_pcblist, "S,xunpcb", "List of active local stream sockets"); static void -unp_shutdown(unp) - struct unpcb *unp; +unp_shutdown(struct unpcb *unp) { struct socket *so; @@ -871,24 +1200,12 @@ unp_shutdown(unp) } static void -unp_drop(unp, errno) - struct unpcb *unp; - int errno; +unp_drop(struct unpcb *unp, int errno) { struct socket *so = unp->unp_socket; so->so_error = errno; unp_disconnect(unp); - if (so->so_head) { - LIST_REMOVE(unp, unp_link); - unp->unp_gencnt = ++unp_gencnt; - unp_count--; - so->so_pcb = (caddr_t) 0; - if (unp->unp_addr) - FREE(unp->unp_addr, M_SONAME); - zfree(unp_zone, (vm_offset_t)unp); - sofree(so); - } } #ifdef notdef @@ -899,180 +1216,253 @@ unp_drain() } #endif +/* + * Returns: 0 Success + * EMSGSIZE The new fd's will not fit + * ENOBUFS Cannot alloc struct fileproc + */ int -unp_externalize(rights) - struct mbuf *rights; +unp_externalize(struct mbuf *rights) { - struct proc *p = current_proc(); /* XXX */ - register int i; - register struct cmsghdr *cm = mtod(rights, struct cmsghdr *); - register struct file **rp = (struct file **)(cm + 1); - register struct file *fp; - int newfds = (cm->cmsg_len - sizeof(*cm)) / sizeof (int); + proc_t p = current_proc(); /* XXX */ + int i; + struct cmsghdr *cm = mtod(rights, struct cmsghdr *); + struct fileglob **rp = (struct fileglob **)(cm + 1); + struct fileproc *fp; + struct fileglob *fg; + int newfds = (cm->cmsg_len - sizeof (*cm)) / sizeof (int); int f; - - thread_funnel_switch(NETWORK_FUNNEL, KERNEL_FUNNEL); + proc_fdlock(p); /* * if the new FD's will not fit, then we free them all */ if (!fdavail(p, newfds)) { for (i = 0; i < newfds; i++) { - fp = *rp; - unp_discard(fp); - *rp++ = 0; + fg = *rp; + unp_discard_fdlocked(fg, p); + *rp++ = NULL; } + proc_fdunlock(p); - thread_funnel_switch(KERNEL_FUNNEL, NETWORK_FUNNEL); return (EMSGSIZE); } /* - * now change each pointer to an fd in the global table to + * now change each pointer to an fd in the global table to * an integer that is the index to the local fd table entry * that we set up to point to the global one we are transferring. - * XXX this assumes a pointer and int are the same size...! + * XXX (1) this assumes a pointer and int are the same size...! + * XXX (2) allocation failures should be non-fatal */ for (i = 0; i < newfds; i++) { +#if CONFIG_MACF_SOCKET + /* + * If receive access is denied, don't pass along + * and error message, just discard the descriptor. + */ + if (mac_file_check_receive(kauth_cred_get(), *rp)) { + fg = *rp; + *rp++ = 0; + unp_discard_fdlocked(fg, p); + continue; + } +#endif if (fdalloc(p, 0, &f)) - panic("unp_externalize"); - fp = *rp; - p->p_fd->fd_ofiles[f] = fp; - *fdflags(p, f) &= ~UF_RESERVED; - fp->f_msgcount--; - unp_rights--; + panic("unp_externalize:fdalloc"); + fg = *rp; + MALLOC_ZONE(fp, struct fileproc *, sizeof (struct fileproc), + M_FILEPROC, M_WAITOK); + if (fp == NULL) + panic("unp_externalize: MALLOC_ZONE"); + bzero(fp, sizeof (struct fileproc)); + fp->f_iocount = 0; + fp->f_fglob = fg; + fg_removeuipc(fg); + procfdtbl_releasefd(p, f, fp); + (void) OSAddAtomic(-1, (volatile SInt32 *)&unp_rights); *(int *)rp++ = f; } + proc_fdunlock(p); - thread_funnel_switch(KERNEL_FUNNEL, NETWORK_FUNNEL); return (0); } void unp_init(void) { - unp_zone = zinit(sizeof(struct unpcb), - (nmbclusters * sizeof(struct unpcb)), - 4096, "unpzone"); + unp_zone = zinit(sizeof (struct unpcb), + (nmbclusters * sizeof (struct unpcb)), 4096, "unpzone"); + if (unp_zone == 0) panic("unp_init"); LIST_INIT(&unp_dhead); LIST_INIT(&unp_shead); + + /* + * allocate lock group attribute and group for udp pcb mutexes + */ + unp_mtx_grp_attr = lck_grp_attr_alloc_init(); + + unp_mtx_grp = lck_grp_alloc_init("unp_list", unp_mtx_grp_attr); + + unp_mtx_attr = lck_attr_alloc_init(); + + if ((unp_list_mtx = lck_rw_alloc_init(unp_mtx_grp, + unp_mtx_attr)) == NULL) + return; /* pretty much dead if this fails... */ + } #ifndef MIN -#define MIN(a,b) (((a)<(b))?(a):(b)) +#define MIN(a, b) (((a) < (b)) ? (a) : (b)) #endif +/* + * Returns: 0 Success + * EINVAL + * fdgetf_noref:EBADF + */ static int -unp_internalize(control, p) - struct mbuf *control; - struct proc *p; +unp_internalize(struct mbuf *control, proc_t p) { - register struct cmsghdr *cm = mtod(control, struct cmsghdr *); - register struct file **rp; - struct file *fp; - register int i, error; + struct cmsghdr *cm = mtod(control, struct cmsghdr *); + struct fileglob **rp; + struct fileproc *fp; + int i, error; int oldfds; - thread_funnel_switch(NETWORK_FUNNEL, KERNEL_FUNNEL); + /* 64bit: cmsg_len is 'uint32_t', m_len is 'long' */ if (cm->cmsg_type != SCM_RIGHTS || cm->cmsg_level != SOL_SOCKET || - cm->cmsg_len != control->m_len) { - thread_funnel_switch(KERNEL_FUNNEL, NETWORK_FUNNEL); - return (EINVAL); + (unsigned long)cm->cmsg_len != (unsigned long)control->m_len) { + return (EINVAL); } - oldfds = (cm->cmsg_len - sizeof (*cm)) / sizeof (int); - rp = (struct file **)(cm + 1); - for (i = 0; i < oldfds; i++) - if (error = fdgetf(p, *(int *)rp++, 0)) { - thread_funnel_switch(KERNEL_FUNNEL, NETWORK_FUNNEL); - return (error); - } + proc_fdlock(p); + rp = (struct fileglob **)(cm + 1); - rp = (struct file **)(cm + 1); for (i = 0; i < oldfds; i++) { - (void) fdgetf(p, *(int *)rp, &fp); - *rp++ = fp; - fref(fp); - fp->f_msgcount++; - unp_rights++; + if ((error = fdgetf_noref(p, *(int *)rp++, NULL)) != 0) { + proc_fdunlock(p); + return (error); + } } + rp = (struct fileglob **)(cm + 1); + + for (i = 0; i < oldfds; i++) { + (void) fdgetf_noref(p, *(int *)rp, &fp); + fg_insertuipc(fp->f_fglob); + *rp++ = fp->f_fglob; + (void) OSAddAtomic(1, (volatile SInt32 *)&unp_rights); + } + proc_fdunlock(p); - thread_funnel_switch(KERNEL_FUNNEL, NETWORK_FUNNEL); return (0); } -static int unp_defer, unp_gcing; +static int unp_defer, unp_gcing, unp_gcwait; + +/* always called under uipc_lock */ +void +unp_gc_wait(void) +{ + while (unp_gcing != 0) { + unp_gcwait = 1; + msleep(&unp_gcing, uipc_lock, 0 , "unp_gc_wait", NULL); + } +} + static void -unp_gc() +unp_gc(void) { - register struct file *fp, *nextfp; - register struct socket *so; - struct file **extra_ref, **fpp; + struct fileglob *fg, *nextfg; + struct socket *so; + struct fileglob **extra_ref, **fpp; int nunref, i; + int need_gcwakeup = 0; - if (unp_gcing) + lck_mtx_lock(uipc_lock); + if (unp_gcing) { + lck_mtx_unlock(uipc_lock); return; + } unp_gcing = 1; unp_defer = 0; - /* - * before going through all this, set all FDs to + lck_mtx_unlock(uipc_lock); + /* + * before going through all this, set all FDs to * be NOT defered and NOT externally accessible */ - thread_funnel_switch(NETWORK_FUNNEL, KERNEL_FUNNEL); - for (fp = filehead.lh_first; fp != 0; fp = fp->f_list.le_next) - fp->f_flag &= ~(FMARK|FDEFER); + for (fg = fmsghead.lh_first; fg != 0; fg = fg->f_msglist.le_next) { + lck_mtx_lock(&fg->fg_lock); + fg->fg_flag &= ~(FMARK|FDEFER); + lck_mtx_unlock(&fg->fg_lock); + } do { - for (fp = filehead.lh_first; fp != 0; fp = fp->f_list.le_next) { + for (fg = fmsghead.lh_first; fg != 0; + fg = fg->f_msglist.le_next) { + lck_mtx_lock(&fg->fg_lock); /* * If the file is not open, skip it */ - if (fcount(fp) == 0) + if (fg->fg_count == 0) { + lck_mtx_unlock(&fg->fg_lock); continue; + } /* * If we already marked it as 'defer' in a * previous pass, then try process it this time * and un-mark it */ - if (fp->f_flag & FDEFER) { - fp->f_flag &= ~FDEFER; + if (fg->fg_flag & FDEFER) { + fg->fg_flag &= ~FDEFER; unp_defer--; } else { /* * if it's not defered, then check if it's * already marked.. if so skip it */ - if (fp->f_flag & FMARK) + if (fg->fg_flag & FMARK) { + lck_mtx_unlock(&fg->fg_lock); continue; - /* + } + /* * If all references are from messages - * in transit, then skip it. it's not + * in transit, then skip it. it's not * externally accessible. - */ - if (fcount(fp) == fp->f_msgcount) + */ + if (fg->fg_count == fg->fg_msgcount) { + lck_mtx_unlock(&fg->fg_lock); continue; - /* + } + /* * If it got this far then it must be * externally accessible. */ - fp->f_flag |= FMARK; + fg->fg_flag |= FMARK; } /* - * either it was defered, or it is externally + * either it was defered, or it is externally * accessible and not already marked so. * Now check if it is possibly one of OUR sockets. - */ - if (fp->f_type != DTYPE_SOCKET || - (so = (struct socket *)fp->f_data) == 0) + */ + if (fg->fg_type != DTYPE_SOCKET || + (so = (struct socket *)fg->fg_data) == 0) { + lck_mtx_unlock(&fg->fg_lock); continue; + } if (so->so_proto->pr_domain != &localdomain || - (so->so_proto->pr_flags&PR_RIGHTS) == 0) + (so->so_proto->pr_flags&PR_RIGHTS) == 0) { + lck_mtx_unlock(&fg->fg_lock); continue; + } #ifdef notdef - /* if this code is enabled need to run under network funnel */ + /* + * if this code is enabled need to run + * under network funnel + */ if (so->so_rcv.sb_flags & SB_LOCK) { /* * This is problematical; it's not clear @@ -1092,10 +1482,11 @@ unp_gc() * So, Ok, it's one of our sockets and it IS externally * accessible (or was defered). Now we look * to see if we hold any file descriptors in its - * message buffers. Follow those links and mark them + * message buffers. Follow those links and mark them * as accessible too. */ unp_scan(so->so_rcv.sb_mb, unp_mark); + lck_mtx_unlock(&fg->fg_lock); } } while (unp_defer); /* @@ -1137,83 +1528,117 @@ unp_gc() * * 91/09/19, bsy@cs.cmu.edu */ - extra_ref = _MALLOC(nfiles * sizeof(struct file *), M_FILE, M_WAITOK); - for (nunref = 0, fp = filehead.lh_first, fpp = extra_ref; fp != 0; - fp = nextfp) { - nextfp = fp->f_list.le_next; - /* + extra_ref = _MALLOC(nfiles * sizeof (struct fileglob *), + M_FILEGLOB, M_WAITOK); + for (nunref = 0, fg = fmsghead.lh_first, fpp = extra_ref; fg != 0; + fg = nextfg) { + lck_mtx_lock(&fg->fg_lock); + + nextfg = fg->f_msglist.le_next; + /* * If it's not open, skip it */ - if (fcount(fp) == 0) + if (fg->fg_count == 0) { + lck_mtx_unlock(&fg->fg_lock); continue; - /* + } + /* * If all refs are from msgs, and it's not marked accessible * then it must be referenced from some unreachable cycle * of (shut-down) FDs, so include it in our * list of FDs to remove */ - if (fcount(fp) == fp->f_msgcount && !(fp->f_flag & FMARK)) { - *fpp++ = fp; + if (fg->fg_count == fg->fg_msgcount && !(fg->fg_flag & FMARK)) { + fg->fg_count++; + *fpp++ = fg; nunref++; - fref(fp); } + lck_mtx_unlock(&fg->fg_lock); } - /* + /* * for each FD on our hit list, do the following two things */ for (i = nunref, fpp = extra_ref; --i >= 0; ++fpp) { - struct file *tfp = *fpp; - if (tfp->f_type == DTYPE_SOCKET && tfp->f_data != NULL) { - thread_funnel_switch(KERNEL_FUNNEL, NETWORK_FUNNEL); - sorflush((struct socket *)(tfp->f_data)); - thread_funnel_switch(NETWORK_FUNNEL, KERNEL_FUNNEL); - } - } + struct fileglob *tfg; + + tfg = *fpp; + + if (tfg->fg_type == DTYPE_SOCKET && tfg->fg_data != NULL) { + int locked = 0; + + so = (struct socket *)(tfg->fg_data); + /* XXXX */ + /* Assume local sockets use a global lock */ + if (so->so_proto->pr_domain->dom_family != PF_LOCAL) { + socket_lock(so, 0); + locked = 1; + } + sorflush(so); + if (locked) + socket_unlock(so, 0); + } + } for (i = nunref, fpp = extra_ref; --i >= 0; ++fpp) - closef(*fpp, (struct proc *) NULL); - FREE((caddr_t)extra_ref, M_FILE); - thread_funnel_switch(KERNEL_FUNNEL, NETWORK_FUNNEL); + closef_locked((struct fileproc *)0, *fpp, (proc_t)NULL); + lck_mtx_lock(uipc_lock); unp_gcing = 0; + + if (unp_gcwait != 0) { + unp_gcwait = 0; + need_gcwakeup = 1; + } + lck_mtx_unlock(uipc_lock); + + if (need_gcwakeup != 0) + wakeup(&unp_gcing); + FREE((caddr_t)extra_ref, M_FILEGLOB); } void -unp_dispose(m) - struct mbuf *m; +unp_dispose(struct mbuf *m) { - if (m) { - thread_funnel_switch(NETWORK_FUNNEL, KERNEL_FUNNEL); unp_scan(m, unp_discard); - thread_funnel_switch(KERNEL_FUNNEL, NETWORK_FUNNEL); } } +/* + * Returns: 0 Success + */ +static int +unp_listen(struct unpcb *unp, proc_t p) +{ + kauth_cred_t safecred = kauth_cred_proc_ref(p); + cru2x(safecred, &unp->unp_peercred); + kauth_cred_unref(&safecred); + unp->unp_flags |= UNP_HAVEPCCACHED; + return (0); +} + /* should run under kernel funnel */ static void -unp_scan(m0, op) - register struct mbuf *m0; - void (*op) __P((struct file *)); +unp_scan(struct mbuf *m0, void (*op)(struct fileglob *)) { - register struct mbuf *m; - register struct file **rp; - register struct cmsghdr *cm; - register int i; + struct mbuf *m; + struct fileglob **rp; + struct cmsghdr *cm; + int i; int qfds; while (m0) { for (m = m0; m; m = m->m_next) if (m->m_type == MT_CONTROL && - m->m_len >= sizeof(*cm)) { + (size_t)m->m_len >= sizeof (*cm)) { cm = mtod(m, struct cmsghdr *); if (cm->cmsg_level != SOL_SOCKET || cm->cmsg_type != SCM_RIGHTS) continue; - qfds = (cm->cmsg_len - sizeof *cm) - / sizeof (struct file *); - rp = (struct file **)(cm + 1); + qfds = (cm->cmsg_len - sizeof (*cm)) / + sizeof (struct fileglob *); + rp = (struct fileglob **)(cm + 1); for (i = 0; i < qfds; i++) (*op)(*rp++); break; /* XXX, but saves time */ @@ -1224,23 +1649,38 @@ unp_scan(m0, op) /* should run under kernel funnel */ static void -unp_mark(fp) - struct file *fp; +unp_mark(struct fileglob *fg) { + lck_mtx_lock(&fg->fg_lock); - if (fp->f_flag & FMARK) + if (fg->fg_flag & FMARK) { + lck_mtx_unlock(&fg->fg_lock); return; + } + fg->fg_flag |= (FMARK|FDEFER); + + lck_mtx_unlock(&fg->fg_lock); + unp_defer++; - fp->f_flag |= (FMARK|FDEFER); } /* should run under kernel funnel */ static void -unp_discard(fp) - struct file *fp; +unp_discard(struct fileglob *fg) +{ + proc_t p = current_proc(); /* XXX */ + + (void) OSAddAtomic(1, (volatile SInt32 *)&unp_disposed); + + proc_fdlock(p); + unp_discard_fdlocked(fg, p); + proc_fdunlock(p); +} +static void +unp_discard_fdlocked(struct fileglob *fg, proc_t p) { + fg_removeuipc(fg); - fp->f_msgcount--; - unp_rights--; - (void) closef(fp, (struct proc *)NULL); + (void) OSAddAtomic(-1, (volatile SInt32 *)&unp_rights); + (void) closef_locked((struct fileproc *)0, fg, p); }