Please note that diffs are not public domain; they are subject to the copyright notices on the relevant files. =================================================================== RCS file: /ftp/cvs/cvsroot/src/sys/kern/uipc_socket.c,v rcsdiff: /ftp/cvs/cvsroot/src/sys/kern/uipc_socket.c,v: warning: Unknown phrases like `commitid ...;' are present. retrieving revision 1.122.6.1 retrieving revision 1.140.6.1 diff -u -p -r1.122.6.1 -r1.140.6.1 --- src/sys/kern/uipc_socket.c 2006/10/22 06:07:11 1.122.6.1 +++ src/sys/kern/uipc_socket.c 2007/08/09 02:37:21 1.140.6.1 @@ -1,7 +1,7 @@ -/* $NetBSD: uipc_socket.c,v 1.122.6.1 2006/10/22 06:07:11 yamt Exp $ */ +/* $NetBSD: uipc_socket.c,v 1.140.6.1 2007/08/09 02:37:21 jmcneill Exp $ */ /*- - * Copyright (c) 2002 The NetBSD Foundation, Inc. + * Copyright (c) 2002, 2007 The NetBSD Foundation, Inc. * All rights reserved. * * This code is derived from software contributed to The NetBSD Foundation @@ -68,7 +68,7 @@ */ #include -__KERNEL_RCSID(0, "$NetBSD: uipc_socket.c,v 1.122.6.1 2006/10/22 06:07:11 yamt Exp $"); +__KERNEL_RCSID(0, "$NetBSD: uipc_socket.c,v 1.140.6.1 2007/08/09 02:37:21 jmcneill Exp $"); #include "opt_sock_counters.h" #include "opt_sosend_loan.h" @@ -92,10 +92,13 @@ __KERNEL_RCSID(0, "$NetBSD: uipc_socket. #include #include #include +#include +#include #include -POOL_INIT(socket_pool, sizeof(struct socket), 0, 0, 0, "sockpl", NULL); +POOL_INIT(socket_pool, sizeof(struct socket), 0, 0, 0, "sockpl", NULL, + IPL_SOFTNET); MALLOC_DEFINE(M_SOOPTS, "soopts", "socket options"); MALLOC_DEFINE(M_SONAME, "soname", "socket name"); @@ -135,7 +138,7 @@ int sock_loan_thresh = -1; int sock_loan_thresh = 4096; #endif -static struct simplelock so_pendfree_slock = SIMPLELOCK_INITIALIZER; +static kmutex_t so_pendfree_lock; static struct mbuf *so_pendfree; #ifndef SOMAXKVA @@ -143,7 +146,7 @@ static struct mbuf *so_pendfree; #endif int somaxkva = SOMAXKVA; static int socurkva; -static int sokvawaiters; +static kcondvar_t socurkva_cv; #define SOCK_LOAN_CHUNK 65536 @@ -151,13 +154,11 @@ static size_t sodopendfree(void); static size_t sodopendfreel(void); static vsize_t -sokvareserve(struct socket *so __unused, vsize_t len) +sokvareserve(struct socket *so, vsize_t len) { - int s; int error; - s = splvm(); - simple_lock(&so_pendfree_slock); + mutex_enter(&so_pendfree_lock); while (socurkva + len > somaxkva) { size_t freed; @@ -175,33 +176,25 @@ sokvareserve(struct socket *so __unused, continue; SOSEND_COUNTER_INCR(&sosend_kvalimit); - sokvawaiters++; - error = ltsleep(&socurkva, PVM | PCATCH, "sokva", 0, - &so_pendfree_slock); - sokvawaiters--; + error = cv_wait_sig(&socurkva_cv, &so_pendfree_lock); if (error) { len = 0; break; } } socurkva += len; - simple_unlock(&so_pendfree_slock); - splx(s); + mutex_exit(&so_pendfree_lock); return len; } static void sokvaunreserve(vsize_t len) { - int s; - s = splvm(); - simple_lock(&so_pendfree_slock); + mutex_enter(&so_pendfree_lock); socurkva -= len; - if (sokvawaiters) - wakeup(&socurkva); - simple_unlock(&so_pendfree_slock); - splx(s); + cv_broadcast(&socurkva_cv); + mutex_exit(&so_pendfree_lock); } /* @@ -255,7 +248,7 @@ sokvafree(vaddr_t sva, vsize_t len) } static void -sodoloanfree(struct vm_page **pgs, caddr_t buf, size_t size) +sodoloanfree(struct vm_page **pgs, void *buf, size_t size) { vaddr_t va, sva, eva; vsize_t len; @@ -271,7 +264,7 @@ sodoloanfree(struct vm_page **pgs, caddr pgs = alloca(npgs * sizeof(*pgs)); for (i = 0, va = sva; va < eva; i++, va += PAGE_SIZE) { - if (pmap_extract(pmap_kernel(), va, &pa) == FALSE) + if (pmap_extract(pmap_kernel(), va, &pa) == false) panic("sodoloanfree: va 0x%lx not mapped", va); pgs[i] = PHYS_TO_VM_PAGE(pa); } @@ -286,43 +279,35 @@ sodoloanfree(struct vm_page **pgs, caddr static size_t sodopendfree() { - int s; size_t rv; - s = splvm(); - simple_lock(&so_pendfree_slock); + mutex_enter(&so_pendfree_lock); rv = sodopendfreel(); - simple_unlock(&so_pendfree_slock); - splx(s); + mutex_exit(&so_pendfree_lock); return rv; } /* * sodopendfreel: free mbufs on "pendfree" list. - * unlock and relock so_pendfree_slock when freeing mbufs. + * unlock and relock so_pendfree_lock when freeing mbufs. * - * => called with so_pendfree_slock held. - * => called at splvm. + * => called with so_pendfree_lock held. */ static size_t sodopendfreel() { + struct mbuf *m, *next; size_t rv = 0; + int s; - LOCK_ASSERT(simple_lock_held(&so_pendfree_slock)); - - for (;;) { - struct mbuf *m; - struct mbuf *next; + KASSERT(mutex_owned(&so_pendfree_lock)); + while (so_pendfree != NULL) { m = so_pendfree; - if (m == NULL) - break; so_pendfree = NULL; - simple_unlock(&so_pendfree_slock); - /* XXX splx */ + mutex_exit(&so_pendfree_lock); for (; m != NULL; m = next) { next = m->m_next; @@ -331,20 +316,20 @@ sodopendfreel() sodoloanfree((m->m_flags & M_EXT_PAGES) ? m->m_ext.ext_pgs : NULL, m->m_ext.ext_buf, m->m_ext.ext_size); + s = splvm(); pool_cache_put(&mbpool_cache, m); + splx(s); } - /* XXX splvm */ - simple_lock(&so_pendfree_slock); + mutex_enter(&so_pendfree_lock); } return (rv); } void -soloanfree(struct mbuf *m, caddr_t buf, size_t size, void *arg __unused) +soloanfree(struct mbuf *m, void *buf, size_t size, void *arg) { - int s; if (m == NULL) { @@ -363,14 +348,11 @@ soloanfree(struct mbuf *m, caddr_t buf, * because we need to put kva back to kernel_map. */ - s = splvm(); - simple_lock(&so_pendfree_slock); + mutex_enter(&so_pendfree_lock); m->m_next = so_pendfree; so_pendfree = m; - if (sokvawaiters) - wakeup(&socurkva); - simple_unlock(&so_pendfree_slock); - splx(s); + cv_broadcast(&socurkva_cv); + mutex_exit(&so_pendfree_lock); } static long @@ -416,12 +398,12 @@ sosend_loan(struct socket *so, struct ui lva += (vaddr_t) iov->iov_base & PAGE_MASK; - MEXTADD(m, (caddr_t) lva, space, M_MBUF, soloanfree, so); + MEXTADD(m, (void *) lva, space, M_MBUF, soloanfree, so); m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP; uio->uio_resid -= space; /* uio_offset not updated, not set/used for write(2) */ - uio->uio_iov->iov_base = (caddr_t) uio->uio_iov->iov_base + space; + uio->uio_iov->iov_base = (char *)uio->uio_iov->iov_base + space; uio->uio_iov->iov_len -= space; if (uio->uio_iov->iov_len == 0) { uio->uio_iov++; @@ -432,8 +414,7 @@ sosend_loan(struct socket *so, struct ui } static int -sokva_reclaim_callback(struct callback_entry *ce __unused, void *obj __unused, - void *arg __unused) +sokva_reclaim_callback(struct callback_entry *ce, void *obj, void *arg) { KASSERT(ce == &sokva_reclaimerentry); @@ -450,6 +431,9 @@ void soinit(void) { + mutex_init(&so_pendfree_lock, MUTEX_DRIVER, IPL_VM); + cv_init(&socurkva_cv, "sokva"); + /* Set the initial adjusted socket buffer size. */ if (sb_max_set(sb_max)) panic("bad initial sb_max value: %lu", sb_max); @@ -474,26 +458,32 @@ socreate(int dom, struct socket **aso, i uid_t uid; int error, s; + error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET, + KAUTH_REQ_NETWORK_SOCKET_OPEN, KAUTH_ARG(dom), KAUTH_ARG(type), + KAUTH_ARG(proto)); + if (error != 0) + return error; + if (proto) prp = pffindproto(dom, proto, type); else prp = pffindtype(dom, type); - if (prp == 0) { + if (prp == NULL) { /* no support for domain */ if (pffinddomain(dom) == 0) - return (EAFNOSUPPORT); + return EAFNOSUPPORT; /* no support for socket type */ if (proto == 0 && type != 0) - return (EPROTOTYPE); - return (EPROTONOSUPPORT); + return EPROTOTYPE; + return EPROTONOSUPPORT; } - if (prp->pr_usrreq == 0) - return (EPROTONOSUPPORT); + if (prp->pr_usrreq == NULL) + return EPROTONOSUPPORT; if (prp->pr_type != type) - return (EPROTOTYPE); + return EPROTOTYPE; s = splsoftnet(); so = pool_get(&socket_pool, PR_WAITOK); - memset((caddr_t)so, 0, sizeof(*so)); + memset(so, 0, sizeof(*so)); TAILQ_INIT(&so->so_q0); TAILQ_INIT(&so->so_q); so->so_type = type; @@ -505,23 +495,19 @@ socreate(int dom, struct socket **aso, i so->so_snd.sb_mowner = &prp->pr_domain->dom_mowner; so->so_mowner = &prp->pr_domain->dom_mowner; #endif - if (l != NULL) { - uid = kauth_cred_geteuid(l->l_cred); - } else { - uid = 0; - } + uid = kauth_cred_geteuid(l->l_cred); so->so_uidinfo = uid_find(uid); - error = (*prp->pr_usrreq)(so, PRU_ATTACH, (struct mbuf *)0, - (struct mbuf *)(long)proto, (struct mbuf *)0, l); - if (error) { + error = (*prp->pr_usrreq)(so, PRU_ATTACH, NULL, + (struct mbuf *)(long)proto, NULL, l); + if (error != 0) { so->so_state |= SS_NOFDREF; sofree(so); splx(s); - return (error); + return error; } splx(s); *aso = so; - return (0); + return 0; } int @@ -530,10 +516,9 @@ sobind(struct socket *so, struct mbuf *n int s, error; s = splsoftnet(); - error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, (struct mbuf *)0, - nam, (struct mbuf *)0, l); + error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, NULL, nam, NULL, l); splx(s); - return (error); + return error; } int @@ -542,11 +527,11 @@ solisten(struct socket *so, int backlog) int s, error; s = splsoftnet(); - error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, (struct mbuf *)0, - (struct mbuf *)0, (struct mbuf *)0, (struct lwp *)0); - if (error) { + error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, NULL, + NULL, NULL, NULL); + if (error != 0) { splx(s); - return (error); + return error; } if (TAILQ_EMPTY(&so->so_q)) so->so_options |= SO_ACCEPTCONN; @@ -554,7 +539,7 @@ solisten(struct socket *so, int backlog) backlog = 0; so->so_qlimit = min(backlog, somaxconn); splx(s); - return (0); + return 0; } void @@ -619,7 +604,7 @@ soclose(struct socket *so) (so->so_state & SS_NBIO)) goto drop; while (so->so_state & SS_ISCONNECTED) { - error = tsleep((caddr_t)&so->so_timeo, + error = tsleep((void *)&so->so_timeo, PSOCK | PCATCH, netcls, so->so_linger * hz); if (error) @@ -630,8 +615,7 @@ soclose(struct socket *so) drop: if (so->so_pcb) { int error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH, - (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0, - (struct lwp *)0); + NULL, NULL, NULL, NULL); if (error == 0) error = error2; } @@ -650,9 +634,15 @@ soclose(struct socket *so) int soabort(struct socket *so) { + int error; - return (*so->so_proto->pr_usrreq)(so, PRU_ABORT, (struct mbuf *)0, - (struct mbuf *)0, (struct mbuf *)0, (struct lwp *)0); + KASSERT(so->so_head == NULL); + error = (*so->so_proto->pr_usrreq)(so, PRU_ABORT, NULL, + NULL, NULL, NULL); + if (error) { + sofree(so); + } + return error; } int @@ -668,7 +658,7 @@ soaccept(struct socket *so, struct mbuf if ((so->so_state & SS_ISDISCONNECTED) == 0 || (so->so_proto->pr_flags & PR_ABRTACPTDIS) == 0) error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT, - (struct mbuf *)0, nam, (struct mbuf *)0, (struct lwp *)0); + NULL, nam, NULL, NULL); else error = ECONNABORTED; @@ -696,7 +686,7 @@ soconnect(struct socket *so, struct mbuf error = EISCONN; else error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT, - (struct mbuf *)0, nam, (struct mbuf *)0, l); + NULL, nam, NULL, l); splx(s); return (error); } @@ -708,8 +698,7 @@ soconnect2(struct socket *so1, struct so s = splsoftnet(); error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2, - (struct mbuf *)0, (struct mbuf *)so2, (struct mbuf *)0, - (struct lwp *)0); + NULL, (struct mbuf *)so2, NULL, NULL); splx(s); return (error); } @@ -729,8 +718,7 @@ sodisconnect(struct socket *so) goto bad; } error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT, - (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0, - (struct lwp *)0); + NULL, NULL, NULL, NULL); bad: splx(s); sodopendfree(); @@ -847,7 +835,7 @@ sosend(struct socket *so, struct mbuf *a m = m_gethdr(M_WAIT, MT_DATA); mlen = MHLEN; m->m_pkthdr.len = 0; - m->m_pkthdr.rcvif = (struct ifnet *)0; + m->m_pkthdr.rcvif = NULL; } else { m = m_get(M_WAIT, MT_DATA); mlen = MLEN; @@ -887,7 +875,7 @@ sosend(struct socket *so, struct mbuf *a if (atomic && top == 0 && len < mlen) MH_ALIGN(m, len); } - error = uiomove(mtod(m, caddr_t), (int)len, + error = uiomove(mtod(m, void *), (int)len, uio); have_data: resid = uio->uio_resid; @@ -988,12 +976,11 @@ soreceive(struct socket *so, struct mbuf if (flags & MSG_OOB) { m = m_get(M_WAIT, MT_DATA); error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m, - (struct mbuf *)(long)(flags & MSG_PEEK), - (struct mbuf *)0, l); + (struct mbuf *)(long)(flags & MSG_PEEK), NULL, l); if (error) goto bad; do { - error = uiomove(mtod(m, caddr_t), + error = uiomove(mtod(m, void *), (int) min(uio->uio_resid, m->m_len), uio); m = m_free(m); } while (uio->uio_resid && error == 0 && m); @@ -1003,10 +990,9 @@ soreceive(struct socket *so, struct mbuf return (error); } if (mp) - *mp = (struct mbuf *)0; + *mp = NULL; if (so->so_state & SS_ISCONFIRMING && uio->uio_resid) - (*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0, - (struct mbuf *)0, (struct mbuf *)0, l); + (*pr->pr_usrreq)(so, PRU_RCVD, NULL, NULL, NULL, l); restart: if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0) @@ -1207,7 +1193,7 @@ soreceive(struct socket *so, struct mbuf SBLASTRECORDCHK(&so->so_rcv, "soreceive uiomove"); SBLASTMBUFCHK(&so->so_rcv, "soreceive uiomove"); splx(s); - error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio); + error = uiomove(mtod(m, char *) + moff, (int)len, uio); s = splsoftnet(); if (error) { /* @@ -1242,7 +1228,7 @@ soreceive(struct socket *so, struct mbuf *mp = m; mp = &m->m_next; so->so_rcv.sb_mb = m = m->m_next; - *mp = (struct mbuf *)0; + *mp = NULL; } else { MFREE(m, so->so_rcv.sb_mb); m = so->so_rcv.sb_mb; @@ -1313,9 +1299,7 @@ soreceive(struct socket *so, struct mbuf */ if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb) (*pr->pr_usrreq)(so, PRU_RCVD, - (struct mbuf *)0, - (struct mbuf *)(long)flags, - (struct mbuf *)0, l); + NULL, (struct mbuf *)(long)flags, NULL, l); SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 2"); SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 2"); error = sbwait(&so->so_rcv); @@ -1351,8 +1335,8 @@ soreceive(struct socket *so, struct mbuf SBLASTRECORDCHK(&so->so_rcv, "soreceive 4"); SBLASTMBUFCHK(&so->so_rcv, "soreceive 4"); if (pr->pr_flags & PR_WANTRCVD && so->so_pcb) - (*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0, - (struct mbuf *)(long)flags, (struct mbuf *)0, l); + (*pr->pr_usrreq)(so, PRU_RCVD, NULL, + (struct mbuf *)(long)flags, NULL, l); } if (orig_resid == uio->uio_resid && orig_resid && (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) { @@ -1381,8 +1365,8 @@ soshutdown(struct socket *so, int how) if (how == SHUT_RD || how == SHUT_RDWR) sorflush(so); if (how == SHUT_WR || how == SHUT_RDWR) - return (*pr->pr_usrreq)(so, PRU_SHUTDOWN, (struct mbuf *)0, - (struct mbuf *)0, (struct mbuf *)0, (struct lwp *)0); + return (*pr->pr_usrreq)(so, PRU_SHUTDOWN, NULL, + NULL, NULL, NULL); return (0); } @@ -1418,6 +1402,8 @@ sosetopt(struct socket *so, int level, i { int error; struct mbuf *m; + struct linger *l; + struct sockbuf *sb; error = 0; m = m0; @@ -1434,13 +1420,18 @@ sosetopt(struct socket *so, int level, i error = EINVAL; goto bad; } - if (mtod(m, struct linger *)->l_linger < 0 || - mtod(m, struct linger *)->l_linger > (INT_MAX / hz)) { + l = mtod(m, struct linger *); + if (l->l_linger < 0 || l->l_linger > USHRT_MAX || + l->l_linger > (INT_MAX / hz)) { error = EDOM; goto bad; } - so->so_linger = mtod(m, struct linger *)->l_linger; - /* fall thru... */ + so->so_linger = l->l_linger; + if (l->l_onoff) + so->so_options |= SO_LINGER; + else + so->so_options &= ~SO_LINGER; + break; case SO_DEBUG: case SO_KEEPALIVE: @@ -1487,12 +1478,13 @@ sosetopt(struct socket *so, int level, i case SO_SNDBUF: case SO_RCVBUF: - if (sbreserve(optname == SO_SNDBUF ? - &so->so_snd : &so->so_rcv, - (u_long) optval, so) == 0) { + sb = (optname == SO_SNDBUF) ? + &so->so_snd : &so->so_rcv; + if (sbreserve(sb, (u_long)optval, so) == 0) { error = ENOBUFS; goto bad; } + sb->sb_flags &= ~SB_AUTOSIZE; break; /* @@ -1580,7 +1572,7 @@ sogetopt(struct socket *so, int level, i case SO_LINGER: m->m_len = sizeof(struct linger); mtod(m, struct linger *)->l_onoff = - so->so_options & SO_LINGER; + (so->so_options & SO_LINGER) ? 1 : 0; mtod(m, struct linger *)->l_linger = so->so_linger; break; @@ -1593,7 +1585,7 @@ sogetopt(struct socket *so, int level, i case SO_BROADCAST: case SO_OOBINLINE: case SO_TIMESTAMP: - *mtod(m, int *) = so->so_options & optname; + *mtod(m, int *) = (so->so_options & optname) ? 1 : 0; break; case SO_TYPE: @@ -1667,7 +1659,7 @@ filt_sordetach(struct knote *kn) /*ARGSUSED*/ static int -filt_soread(struct knote *kn, long hint __unused) +filt_soread(struct knote *kn, long hint) { struct socket *so; @@ -1698,7 +1690,7 @@ filt_sowdetach(struct knote *kn) /*ARGSUSED*/ static int -filt_sowrite(struct knote *kn, long hint __unused) +filt_sowrite(struct knote *kn, long hint) { struct socket *so; @@ -1721,7 +1713,7 @@ filt_sowrite(struct knote *kn, long hint /*ARGSUSED*/ static int -filt_solisten(struct knote *kn, long hint __unused) +filt_solisten(struct knote *kn, long hint) { struct socket *so; @@ -1743,7 +1735,7 @@ static const struct filterops sowrite_fi { 1, NULL, filt_sowdetach, filt_sowrite }; int -soo_kqfilter(struct file *fp __unused, struct knote *kn) +soo_kqfilter(struct file *fp, struct knote *kn) { struct socket *so; struct sockbuf *sb; @@ -1783,7 +1775,6 @@ sysctl_kern_somaxkva(SYSCTLFN_ARGS) { int error, new_somaxkva; struct sysctlnode node; - int s; new_somaxkva = somaxkva; node = *rnode; @@ -1795,12 +1786,10 @@ sysctl_kern_somaxkva(SYSCTLFN_ARGS) if (new_somaxkva < (16 * 1024 * 1024)) /* sanity */ return (EINVAL); - s = splvm(); - simple_lock(&so_pendfree_slock); + mutex_enter(&so_pendfree_lock); somaxkva = new_somaxkva; - wakeup(&socurkva); - simple_unlock(&so_pendfree_slock); - splx(s); + cv_broadcast(&socurkva_cv); + mutex_exit(&so_pendfree_lock); return (error); }