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 1.97.2.1.2.2 retrieving revision 1.140 diff -u -p -r1.97.2.1.2.2 -r1.140 --- src/sys/kern/uipc_socket.c 2005/09/13 23:23:56 1.97.2.1.2.2 +++ src/sys/kern/uipc_socket.c 2007/05/02 20:40:23 1.140 @@ -1,7 +1,7 @@ -/* $NetBSD: uipc_socket.c,v 1.97.2.1.2.2 2005/09/13 23:23:56 riz Exp $ */ +/* $NetBSD: uipc_socket.c,v 1.140 2007/05/02 20:40:23 dyoung 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 1.97.2.1.2.2 2005/09/13 23:23:56 riz Exp $"); +__KERNEL_RCSID(0, "$NetBSD: uipc_socket.c,v 1.140 2007/05/02 20:40:23 dyoung Exp $"); #include "opt_sock_counters.h" #include "opt_sosend_loan.h" @@ -91,10 +91,14 @@ __KERNEL_RCSID(0, "$NetBSD: uipc_socket. #include #include #include +#include +#include +#include #include -struct pool socket_pool; +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"); @@ -105,73 +109,56 @@ int somaxconn = SOMAXCONN; #ifdef SOSEND_COUNTERS #include -struct evcnt sosend_loan_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, +static struct evcnt sosend_loan_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, NULL, "sosend", "loan big"); -struct evcnt sosend_copy_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, +static struct evcnt sosend_copy_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, NULL, "sosend", "copy big"); -struct evcnt sosend_copy_small = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, +static struct evcnt sosend_copy_small = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, NULL, "sosend", "copy small"); -struct evcnt sosend_kvalimit = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, +static struct evcnt sosend_kvalimit = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, NULL, "sosend", "kva limit"); #define SOSEND_COUNTER_INCR(ev) (ev)->ev_count++ +EVCNT_ATTACH_STATIC(sosend_loan_big); +EVCNT_ATTACH_STATIC(sosend_copy_big); +EVCNT_ATTACH_STATIC(sosend_copy_small); +EVCNT_ATTACH_STATIC(sosend_kvalimit); #else #define SOSEND_COUNTER_INCR(ev) /* nothing */ #endif /* SOSEND_COUNTERS */ -void -soinit(void) -{ - - /* Set the initial adjusted socket buffer size. */ - if (sb_max_set(sb_max)) - panic("bad initial sb_max value: %lu\n", sb_max); - - pool_init(&socket_pool, sizeof(struct socket), 0, 0, 0, - "sockpl", NULL); - -#ifdef SOSEND_COUNTERS - evcnt_attach_static(&sosend_loan_big); - evcnt_attach_static(&sosend_copy_big); - evcnt_attach_static(&sosend_copy_small); - evcnt_attach_static(&sosend_kvalimit); -#endif /* SOSEND_COUNTERS */ -} +static struct callback_entry sokva_reclaimerentry; #ifdef SOSEND_NO_LOAN -int use_sosend_loan = 0; +int sock_loan_thresh = -1; #else -int use_sosend_loan = 1; +int sock_loan_thresh = 4096; #endif -struct simplelock so_pendfree_slock = SIMPLELOCK_INITIALIZER; -struct mbuf *so_pendfree; +static kmutex_t so_pendfree_lock; +static struct mbuf *so_pendfree; #ifndef SOMAXKVA #define SOMAXKVA (16 * 1024 * 1024) #endif int somaxkva = SOMAXKVA; -int socurkva; -int sokvawaiters; +static int socurkva; +static kcondvar_t socurkva_cv; -#define SOCK_LOAN_THRESH 4096 #define SOCK_LOAN_CHUNK 65536 -static size_t sodopendfree(struct socket *); -static size_t sodopendfreel(struct socket *); -static __inline void sokvareserve(struct socket *, vsize_t); -static __inline void sokvaunreserve(vsize_t); +static size_t sodopendfree(void); +static size_t sodopendfreel(void); -static __inline void +static vsize_t 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; @@ -179,7 +166,7 @@ sokvareserve(struct socket *so, vsize_t * try to do pendfree. */ - freed = sodopendfreel(so); + freed = sodopendfreel(); /* * if some kva was freed, try again. @@ -189,27 +176,25 @@ sokvareserve(struct socket *so, vsize_t continue; SOSEND_COUNTER_INCR(&sosend_kvalimit); - sokvawaiters++; - (void) ltsleep(&socurkva, PVM, "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 __inline void +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); } /* @@ -225,13 +210,14 @@ sokvaalloc(vsize_t len, struct socket *s * reserve kva. */ - sokvareserve(so, len); + if (sokvareserve(so, len) == 0) + return 0; /* * allocate kva. */ - lva = uvm_km_valloc_wait(kernel_map, len); + lva = uvm_km_alloc(kernel_map, len, 0, UVM_KMF_VAONLY | UVM_KMF_WAITVA); if (lva == 0) { sokvaunreserve(len); return (0); @@ -252,7 +238,7 @@ sokvafree(vaddr_t sva, vsize_t len) * free kva. */ - uvm_km_free(kernel_map, sva, len); + uvm_km_free(kernel_map, sva, len, UVM_KMF_VAONLY); /* * unreserve kva. @@ -262,7 +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; @@ -278,7 +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); } @@ -291,45 +277,37 @@ sodoloanfree(struct vm_page **pgs, caddr } static size_t -sodopendfree(struct socket *so) +sodopendfree() { - int s; size_t rv; - s = splvm(); - simple_lock(&so_pendfree_slock); - rv = sodopendfreel(so); - simple_unlock(&so_pendfree_slock); - splx(s); + mutex_enter(&so_pendfree_lock); + rv = sodopendfreel(); + 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 socket *so) +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; @@ -338,20 +316,20 @@ sodopendfreel(struct socket *so) 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) +soloanfree(struct mbuf *m, void *buf, size_t size, void *arg) { - int s; if (m == NULL) { @@ -370,14 +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 @@ -389,7 +364,7 @@ sosend_loan(struct socket *so, struct ui vaddr_t lva, va; int npgs, i, error; - if (uio->uio_segflg != UIO_USERSPACE) + if (VMSPACE_IS_KERNEL_P(uio->uio_vmspace)) return (0); if (iov->iov_len < (size_t) space) @@ -409,7 +384,7 @@ sosend_loan(struct socket *so, struct ui if (lva == 0) return 0; - error = uvm_loan(&uio->uio_procp->p_vmspace->vm_map, sva, len, + error = uvm_loan(&uio->uio_vmspace->vm_map, sva, len, m->m_ext.ext_pgs, UVM_LOAN_TOPAGE); if (error) { sokvafree(lva, len); @@ -423,12 +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++; @@ -438,6 +413,35 @@ sosend_loan(struct socket *so, struct ui return (space); } +static int +sokva_reclaim_callback(struct callback_entry *ce, void *obj, void *arg) +{ + + KASSERT(ce == &sokva_reclaimerentry); + KASSERT(obj == NULL); + + sodopendfree(); + if (!vm_map_starved_p(kernel_map)) { + return CALLBACK_CHAIN_ABORT; + } + return CALLBACK_CHAIN_CONTINUE; +} + +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); + + callback_register(&vm_map_to_kernel(kernel_map)->vmk_reclaim_callback, + &sokva_reclaimerentry, NULL, sokva_reclaim_callback); +} + /* * Socket operation routines. * These routines are called by the routines in @@ -447,25 +451,39 @@ sosend_loan(struct socket *so, struct ui */ /*ARGSUSED*/ int -socreate(int dom, struct socket **aso, int type, int proto) +socreate(int dom, struct socket **aso, int type, int proto, struct lwp *l) { - struct proc *p; - struct protosw *prp; + const struct protosw *prp; struct socket *so; + uid_t uid; int error, s; - p = curproc; /* XXX */ + 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 || prp->pr_usrreq == 0) - return (EPROTONOSUPPORT); + if (prp == NULL) { + /* no support for domain */ + if (pffinddomain(dom) == 0) + return EAFNOSUPPORT; + /* no support for socket type */ + if (proto == 0 && type != 0) + return EPROTOTYPE; + 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; @@ -477,31 +495,30 @@ 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 (p != 0) - so->so_uid = p->p_ucred->cr_uid; - error = (*prp->pr_usrreq)(so, PRU_ATTACH, (struct mbuf *)0, - (struct mbuf *)(long)proto, (struct mbuf *)0, p); - if (error) { + uid = kauth_cred_geteuid(l->l_cred); + so->so_uidinfo = uid_find(uid); + 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 -sobind(struct socket *so, struct mbuf *nam, struct proc *p) +sobind(struct socket *so, struct mbuf *nam, struct lwp *l) { int s, error; s = splsoftnet(); - error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, (struct mbuf *)0, - nam, (struct mbuf *)0, p); + error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, NULL, nam, NULL, l); splx(s); - return (error); + return error; } int @@ -510,11 +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 proc *)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; @@ -522,7 +539,7 @@ solisten(struct socket *so, int backlog) backlog = 0; so->so_qlimit = min(backlog, somaxconn); splx(s); - return (0); + return 0; } void @@ -540,7 +557,13 @@ sofree(struct socket *so) if (!soqremque(so, 0)) return; } - sbrelease(&so->so_snd); + if (so->so_rcv.sb_hiwat) + (void)chgsbsize(so->so_uidinfo, &so->so_rcv.sb_hiwat, 0, + RLIM_INFINITY); + if (so->so_snd.sb_hiwat) + (void)chgsbsize(so->so_uidinfo, &so->so_snd.sb_hiwat, 0, + RLIM_INFINITY); + sbrelease(&so->so_snd, so); sorflush(so); pool_put(&socket_pool, so); } @@ -581,7 +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) @@ -592,8 +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 proc *)0); + NULL, NULL, NULL, NULL); if (error == 0) error = error2; } @@ -612,9 +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 proc *)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 @@ -630,7 +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 proc *)0); + NULL, nam, NULL, NULL); else error = ECONNABORTED; @@ -639,12 +667,10 @@ soaccept(struct socket *so, struct mbuf } int -soconnect(struct socket *so, struct mbuf *nam) +soconnect(struct socket *so, struct mbuf *nam, struct lwp *l) { - struct proc *p; int s, error; - p = curproc; /* XXX */ if (so->so_options & SO_ACCEPTCONN) return (EOPNOTSUPP); s = splsoftnet(); @@ -660,7 +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, p); + NULL, nam, NULL, l); splx(s); return (error); } @@ -672,8 +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 proc *)0); + NULL, (struct mbuf *)so2, NULL, NULL); splx(s); return (error); } @@ -693,11 +718,10 @@ 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 proc *)0); + NULL, NULL, NULL, NULL); bad: splx(s); - sodopendfree(so); + sodopendfree(); return (error); } @@ -721,16 +745,16 @@ sodisconnect(struct socket *so) */ int sosend(struct socket *so, struct mbuf *addr, struct uio *uio, struct mbuf *top, - struct mbuf *control, int flags) + struct mbuf *control, int flags, struct lwp *l) { - struct proc *p; struct mbuf **mp, *m; + struct proc *p; long space, len, resid, clen, mlen; int error, s, dontroute, atomic; - sodopendfree(so); + p = l->l_proc; + sodopendfree(); - p = curproc; /* XXX */ clen = 0; atomic = sosendallatonce(so) || top; if (uio) @@ -751,7 +775,8 @@ sosend(struct socket *so, struct mbuf *a dontroute = (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 && (so->so_proto->pr_flags & PR_ATOMIC); - p->p_stats->p_ru.ru_msgsnd++; + if (p) + p->p_stats->p_ru.ru_msgsnd++; if (control) clen = control->m_len; #define snderr(errno) { error = errno; splx(s); goto release; } @@ -810,15 +835,15 @@ 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; } MCLAIM(m, so->so_snd.sb_mowner); - if (use_sosend_loan && - uio->uio_iov->iov_len >= SOCK_LOAN_THRESH && - space >= SOCK_LOAN_THRESH && + if (sock_loan_thresh >= 0 && + uio->uio_iov->iov_len >= sock_loan_thresh && + space >= sock_loan_thresh && (len = sosend_loan(so, uio, m, space)) != 0) { SOSEND_COUNTER_INCR(&sosend_loan_big); @@ -850,7 +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; @@ -866,7 +891,7 @@ sosend(struct socket *so, struct mbuf *a break; } } while (space > 0 && atomic); - + s = splsoftnet(); if (so->so_state & SS_CANTSENDMORE) @@ -878,7 +903,7 @@ sosend(struct socket *so, struct mbuf *a so->so_state |= SS_MORETOCOME; error = (*so->so_proto->pr_usrreq)(so, (flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND, - top, addr, control, p); + top, addr, control, curlwp); /* XXX */ if (dontroute) so->so_options &= ~SO_DONTROUTE; if (resid > 0) @@ -924,9 +949,10 @@ int soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio, struct mbuf **mp0, struct mbuf **controlp, int *flagsp) { + struct lwp *l = curlwp; struct mbuf *m, **mp; int flags, len, error, s, offset, moff, type, orig_resid; - struct protosw *pr; + const struct protosw *pr; struct mbuf *nextrecord; int mbuf_removed = 0; @@ -934,6 +960,7 @@ soreceive(struct socket *so, struct mbuf mp = mp0; type = 0; orig_resid = uio->uio_resid; + if (paddr) *paddr = 0; if (controlp) @@ -944,17 +971,16 @@ soreceive(struct socket *so, struct mbuf flags = 0; if ((flags & MSG_DONTWAIT) == 0) - sodopendfree(so); + sodopendfree(); 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, - (struct proc *)0); + (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); @@ -964,10 +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, (struct proc *)0); + (*pr->pr_usrreq)(so, PRU_RCVD, NULL, NULL, NULL, l); restart: if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0) @@ -1040,10 +1065,8 @@ soreceive(struct socket *so, struct mbuf * While we process the initial mbufs containing address and control * info, we save a copy of m->m_nextpkt into nextrecord. */ -#ifdef notyet /* XXXX */ - if (uio->uio_procp) - uio->uio_procp->p_stats->p_ru.ru_msgrcv++; -#endif + if (l) + l->l_proc->p_stats->p_ru.ru_msgrcv++; KASSERT(m == so->so_rcv.sb_mb); SBLASTRECORDCHK(&so->so_rcv, "soreceive 1"); SBLASTMBUFCHK(&so->so_rcv, "soreceive 1"); @@ -1081,15 +1104,23 @@ soreceive(struct socket *so, struct mbuf sbfree(&so->so_rcv, m); mbuf_removed = 1; if (controlp) { - if (pr->pr_domain->dom_externalize && + struct domain *dom = pr->pr_domain; + if (dom->dom_externalize && l && mtod(m, struct cmsghdr *)->cmsg_type == SCM_RIGHTS) - error = (*pr->pr_domain->dom_externalize)(m); + error = (*dom->dom_externalize)(m, l); *controlp = m; so->so_rcv.sb_mb = m->m_next; m->m_next = 0; m = so->so_rcv.sb_mb; } else { + /* + * Dispose of any SCM_RIGHTS message that went + * through the read path rather than recv. + */ + if (pr->pr_domain->dom_dispose && + mtod(m, struct cmsghdr *)->cmsg_type == SCM_RIGHTS) + (*pr->pr_domain->dom_dispose)(m); MFREE(m, so->so_rcv.sb_mb); m = so->so_rcv.sb_mb; } @@ -1162,7 +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) { /* @@ -1197,7 +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; @@ -1268,10 +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, - (struct proc *)0); + 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); @@ -1307,9 +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, - (struct proc *)0); + (*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) { @@ -1317,7 +1344,7 @@ soreceive(struct socket *so, struct mbuf splx(s); goto restart; } - + if (flagsp) *flagsp |= flags; release: @@ -1329,7 +1356,7 @@ soreceive(struct socket *so, struct mbuf int soshutdown(struct socket *so, int how) { - struct protosw *pr; + const struct protosw *pr; pr = so->so_proto; if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR)) @@ -1338,8 +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 proc *)0); + return (*pr->pr_usrreq)(so, PRU_SHUTDOWN, NULL, + NULL, NULL, NULL); return (0); } @@ -1347,7 +1374,7 @@ void sorflush(struct socket *so) { struct sockbuf *sb, asb; - struct protosw *pr; + const struct protosw *pr; int s; sb = &so->so_rcv; @@ -1367,7 +1394,7 @@ sorflush(struct socket *so) splx(s); if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose) (*pr->pr_domain->dom_dispose)(asb.sb_mb); - sbrelease(&asb); + sbrelease(&asb, so); } int @@ -1375,6 +1402,7 @@ sosetopt(struct socket *so, int level, i { int error; struct mbuf *m; + struct linger *l; error = 0; m = m0; @@ -1391,8 +1419,18 @@ sosetopt(struct socket *so, int level, i error = EINVAL; goto bad; } - so->so_linger = mtod(m, struct linger *)->l_linger; - /* fall thru... */ + 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 = 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: @@ -1441,7 +1479,7 @@ sosetopt(struct socket *so, int level, i case SO_RCVBUF: if (sbreserve(optname == SO_SNDBUF ? &so->so_snd : &so->so_rcv, - (u_long) optval) == 0) { + (u_long) optval, so) == 0) { error = ENOBUFS; goto bad; } @@ -1469,14 +1507,14 @@ sosetopt(struct socket *so, int level, i case SO_RCVTIMEO: { struct timeval *tv; - short val; + int val; if (m == NULL || m->m_len < sizeof(*tv)) { error = EINVAL; goto bad; } tv = mtod(m, struct timeval *); - if (tv->tv_sec > (SHRT_MAX - tv->tv_usec / tick) / hz) { + if (tv->tv_sec > (INT_MAX - tv->tv_usec / tick) / hz) { error = EDOM; goto bad; } @@ -1532,7 +1570,7 @@ 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; @@ -1545,7 +1583,7 @@ 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: @@ -1586,6 +1624,10 @@ sogetopt(struct socket *so, int level, i break; } + case SO_OVERFLOWED: + *mtod(m, int *) = so->so_rcv.sb_overflowed; + break; + default: (void)m_free(m); return (ENOPROTOOPT); @@ -1622,7 +1664,7 @@ filt_soread(struct knote *kn, long hint) so = (struct socket *)kn->kn_fp->f_data; kn->kn_data = so->so_rcv.sb_cc; if (so->so_state & SS_CANTRCVMORE) { - kn->kn_flags |= EV_EOF; + kn->kn_flags |= EV_EOF; kn->kn_fflags = so->so_error; return (1); } @@ -1653,7 +1695,7 @@ filt_sowrite(struct knote *kn, long hint so = (struct socket *)kn->kn_fp->f_data; kn->kn_data = sbspace(&so->so_snd); if (so->so_state & SS_CANTSENDMORE) { - kn->kn_flags |= EV_EOF; + kn->kn_flags |= EV_EOF; kn->kn_fflags = so->so_error; return (1); } @@ -1678,7 +1720,7 @@ filt_solisten(struct knote *kn, long hin /* * Set kn_data to number of incoming connections, not * counting partial (incomplete) connections. - */ + */ kn->kn_data = so->so_qlen; return (kn->kn_data > 0); } @@ -1731,7 +1773,6 @@ sysctl_kern_somaxkva(SYSCTLFN_ARGS) { int error, new_somaxkva; struct sysctlnode node; - int s; new_somaxkva = somaxkva; node = *rnode; @@ -1743,12 +1784,10 @@ 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); }