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 1.83.2.9 retrieving revision 1.132 diff -u -p -r1.83.2.9 -r1.132 --- src/sys/kern/uipc_socket.c 2005/04/01 14:30:57 1.83.2.9 +++ src/sys/kern/uipc_socket.c 2007/01/17 12:21:34 1.132 @@ -1,4 +1,4 @@ -/* $NetBSD: uipc_socket.c,v 1.83.2.9 2005/04/01 14:30:57 skrll Exp $ */ +/* $NetBSD: uipc_socket.c,v 1.132 2007/01/17 12:21:34 elad Exp $ */ /*- * Copyright (c) 2002 The NetBSD Foundation, Inc. @@ -68,7 +68,7 @@ */ #include -__KERNEL_RCSID(0, "$NetBSD: uipc_socket.c,v 1.83.2.9 2005/04/01 14:30:57 skrll Exp $"); +__KERNEL_RCSID(0, "$NetBSD: uipc_socket.c,v 1.132 2007/01/17 12:21:34 elad Exp $"); #include "opt_sock_counters.h" #include "opt_sosend_loan.h" @@ -91,6 +91,7 @@ __KERNEL_RCSID(0, "$NetBSD: uipc_socket. #include #include #include +#include #include @@ -105,13 +106,13 @@ 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++ @@ -126,41 +127,30 @@ EVCNT_ATTACH_STATIC(sosend_kvalimit); #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); - -} +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 struct simplelock so_pendfree_slock = SIMPLELOCK_INITIALIZER; +static struct mbuf *so_pendfree; #ifndef SOMAXKVA #define SOMAXKVA (16 * 1024 * 1024) #endif int somaxkva = SOMAXKVA; -int socurkva; -int sokvawaiters; +static int socurkva; +static int sokvawaiters; -#define SOCK_LOAN_THRESH 4096 #define SOCK_LOAN_CHUNK 65536 -static size_t sodopendfree(struct socket *); -static size_t sodopendfreel(struct socket *); -static __inline vsize_t sokvareserve(struct socket *, vsize_t); -static __inline void sokvaunreserve(vsize_t); +static size_t sodopendfree(void); +static size_t sodopendfreel(void); -static __inline vsize_t +static vsize_t sokvareserve(struct socket *so, vsize_t len) { int s; @@ -175,7 +165,7 @@ sokvareserve(struct socket *so, vsize_t * try to do pendfree. */ - freed = sodopendfreel(so); + freed = sodopendfreel(); /* * if some kva was freed, try again. @@ -200,7 +190,7 @@ sokvareserve(struct socket *so, vsize_t return len; } -static __inline void +static void sokvaunreserve(vsize_t len) { int s; @@ -294,14 +284,14 @@ 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); + rv = sodopendfreel(); simple_unlock(&so_pendfree_slock); splx(s); @@ -317,7 +307,7 @@ sodopendfree(struct socket *so) */ static size_t -sodopendfreel(struct socket *so) +sodopendfreel() { size_t rv = 0; @@ -392,7 +382,7 @@ 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) @@ -407,13 +397,12 @@ sosend_loan(struct socket *so, struct ui /* XXX KDASSERT */ KASSERT(npgs <= M_EXT_MAXPAGES); - KASSERT(uio->uio_lwp != NULL); lva = sokvaalloc(len, so); if (lva == 0) return 0; - error = uvm_loan(&uio->uio_lwp->l_proc->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); @@ -442,6 +431,32 @@ 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) +{ + + /* 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 @@ -455,15 +470,29 @@ socreate(int dom, struct socket **aso, i { const struct protosw *prp; struct socket *so; - struct proc *p; + uid_t uid; int error, s; - p = l->l_proc; + 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) + return (error); + if (proto) prp = pffindproto(dom, proto, type); else prp = pffindtype(dom, type); - if (prp == 0 || prp->pr_usrreq == 0) + if (prp == 0) { + /* 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 == 0) return (EPROTONOSUPPORT); if (prp->pr_type != type) return (EPROTOTYPE); @@ -481,10 +510,12 @@ 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; - else - so->so_uid = UID_MAX; + if (l != NULL) { + uid = kauth_cred_geteuid(l->l_cred); + } else { + uid = 0; + } + 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) { @@ -547,10 +578,10 @@ sofree(struct socket *so) return; } if (so->so_rcv.sb_hiwat) - (void)chgsbsize(so->so_uid, &so->so_rcv.sb_hiwat, 0, + (void)chgsbsize(so->so_uidinfo, &so->so_rcv.sb_hiwat, 0, RLIM_INFINITY); if (so->so_snd.sb_hiwat) - (void)chgsbsize(so->so_uid, &so->so_snd.sb_hiwat, 0, + (void)chgsbsize(so->so_uidinfo, &so->so_snd.sb_hiwat, 0, RLIM_INFINITY); sbrelease(&so->so_snd, so); sorflush(so); @@ -707,7 +738,7 @@ sodisconnect(struct socket *so) (struct lwp *)0); bad: splx(s); - sodopendfree(so); + sodopendfree(); return (error); } @@ -739,7 +770,7 @@ sosend(struct socket *so, struct mbuf *a int error, s, dontroute, atomic; p = l->l_proc; - sodopendfree(so); + sodopendfree(); clen = 0; atomic = sosendallatonce(so) || top; @@ -827,9 +858,9 @@ sosend(struct socket *so, struct mbuf *a 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); @@ -935,7 +966,7 @@ int soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio, struct mbuf **mp0, struct mbuf **controlp, int *flagsp) { - struct lwp *l; + struct lwp *l = curlwp; struct mbuf *m, **mp; int flags, len, error, s, offset, moff, type, orig_resid; const struct protosw *pr; @@ -946,7 +977,6 @@ soreceive(struct socket *so, struct mbuf mp = mp0; type = 0; orig_resid = uio->uio_resid; - l = uio->uio_lwp; if (paddr) *paddr = 0; @@ -958,7 +988,7 @@ 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); @@ -1393,6 +1423,7 @@ sosetopt(struct socket *so, int level, i { int error; struct mbuf *m; + struct linger *l; error = 0; m = m0; @@ -1409,8 +1440,18 @@ 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: @@ -1550,7 +1591,7 @@ 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; @@ -1563,7 +1604,7 @@ 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: