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 1.111.2.2 diff -u -p -r1.97.2.1.2.2 -r1.111.2.2 --- src/sys/kern/uipc_socket.c 2005/09/13 23:23:56 1.97.2.1.2.2 +++ src/sys/kern/uipc_socket.c 2005/07/07 12:42:25 1.111.2.2 @@ -1,4 +1,4 @@ -/* $NetBSD: uipc_socket.c,v 1.97.2.1.2.2 2005/09/13 23:23:56 riz Exp $ */ +/* $NetBSD: uipc_socket.c,v 1.111.2.2 2005/07/07 12:42:25 yamt Exp $ */ /*- * Copyright (c) 2002 The NetBSD Foundation, Inc. @@ -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 1.111.2.2 2005/07/07 12:42:25 yamt Exp $"); #include "opt_sock_counters.h" #include "opt_sosend_loan.h" @@ -94,7 +94,7 @@ __KERNEL_RCSID(0, "$NetBSD: uipc_socket. #include -struct pool socket_pool; +POOL_INIT(socket_pool, sizeof(struct socket), 0, 0, 0, "sockpl", NULL); MALLOC_DEFINE(M_SOOPTS, "soopts", "socket options"); MALLOC_DEFINE(M_SONAME, "soname", "socket name"); @@ -116,6 +116,10 @@ struct evcnt sosend_kvalimit = EVCNT_INI #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 */ @@ -128,17 +132,8 @@ 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); + panic("bad initial sb_max value: %lu", 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 */ } #ifdef SOSEND_NO_LOAN @@ -162,13 +157,14 @@ int sokvawaiters; static size_t sodopendfree(struct socket *); static size_t sodopendfreel(struct socket *); -static __inline void sokvareserve(struct socket *, vsize_t); +static __inline vsize_t sokvareserve(struct socket *, vsize_t); static __inline void sokvaunreserve(vsize_t); -static __inline void +static __inline vsize_t sokvareserve(struct socket *so, vsize_t len) { int s; + int error; s = splvm(); simple_lock(&so_pendfree_slock); @@ -190,12 +186,18 @@ sokvareserve(struct socket *so, vsize_t SOSEND_COUNTER_INCR(&sosend_kvalimit); sokvawaiters++; - (void) ltsleep(&socurkva, PVM, "sokva", 0, &so_pendfree_slock); + error = ltsleep(&socurkva, PVM | PCATCH, "sokva", 0, + &so_pendfree_slock); sokvawaiters--; + if (error) { + len = 0; + break; + } } socurkva += len; simple_unlock(&so_pendfree_slock); splx(s); + return len; } static __inline void @@ -225,13 +227,14 @@ 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 +255,7 @@ 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. @@ -262,30 +265,23 @@ 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, caddr_t buf, size_t size, boolean_t mapped) { - vaddr_t va, sva, eva; + vaddr_t sva, eva; vsize_t len; - paddr_t pa; - int i, npgs; + int npgs; + + KASSERT(pgs != NULL); eva = round_page((vaddr_t) buf + size); sva = trunc_page((vaddr_t) buf); len = eva - sva; npgs = len >> PAGE_SHIFT; - if (__predict_false(pgs == NULL)) { - pgs = alloca(npgs * sizeof(*pgs)); - - for (i = 0, va = sva; va < eva; i++, va += PAGE_SIZE) { - if (pmap_extract(pmap_kernel(), va, &pa) == FALSE) - panic("sodoloanfree: va 0x%lx not mapped", va); - pgs[i] = PHYS_TO_VM_PAGE(pa); - } + if (mapped) { + pmap_kremove(sva, len); + pmap_update(pmap_kernel()); } - - pmap_kremove(sva, len); - pmap_update(pmap_kernel()); uvm_unloan(pgs, npgs, UVM_LOAN_TOPAGE); sokvafree(sva, len); } @@ -333,11 +329,13 @@ sodopendfreel(struct socket *so) for (; m != NULL; m = next) { next = m->m_next; + KASSERT((~m->m_flags & (M_EXT|M_EXT_PAGES)) == 0); + KASSERT(!MCLISREFERENCED(m)); rv += m->m_ext.ext_size; - sodoloanfree((m->m_flags & M_EXT_PAGES) ? - m->m_ext.ext_pgs : NULL, m->m_ext.ext_buf, - m->m_ext.ext_size); + sodoloanfree(m->m_ext.ext_pgs, m->m_ext.ext_buf, + m->m_ext.ext_size, + (m->m_ext.ext_flags & M_EXT_LAZY) == 0); pool_cache_put(&mbpool_cache, m); } @@ -353,15 +351,7 @@ soloanfree(struct mbuf *m, caddr_t buf, { int s; - if (m == NULL) { - - /* - * called from MEXTREMOVE. - */ - - sodoloanfree(NULL, buf, size); - return; - } + KASSERT(m != NULL); /* * postpone freeing mbuf. @@ -386,8 +376,12 @@ sosend_loan(struct socket *so, struct ui struct iovec *iov = uio->uio_iov; vaddr_t sva, eva; vsize_t len; - vaddr_t lva, va; - int npgs, i, error; + vaddr_t lva; + int npgs, error; +#if !defined(__HAVE_LAZY_MBUF) + vaddr_t va; + int i; +#endif /* !defined(__HAVE_LAZY_MBUF) */ if (uio->uio_segflg != UIO_USERSPACE) return (0); @@ -404,6 +398,7 @@ sosend_loan(struct socket *so, struct ui /* XXX KDASSERT */ KASSERT(npgs <= M_EXT_MAXPAGES); + KASSERT(uio->uio_procp != NULL); lva = sokvaalloc(len, so); if (lva == 0) @@ -416,16 +411,23 @@ sosend_loan(struct socket *so, struct ui return (0); } +#if !defined(__HAVE_LAZY_MBUF) for (i = 0, va = lva; i < npgs; i++, va += PAGE_SIZE) pmap_kenter_pa(va, VM_PAGE_TO_PHYS(m->m_ext.ext_pgs[i]), VM_PROT_READ); pmap_update(pmap_kernel()); +#endif /* !defined(__HAVE_LAZY_MBUF) */ lva += (vaddr_t) iov->iov_base & PAGE_MASK; MEXTADD(m, (caddr_t) lva, space, M_MBUF, soloanfree, so); m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP; +#if defined(__HAVE_LAZY_MBUF) + m->m_flags |= M_EXT_LAZY; + m->m_ext.ext_flags |= M_EXT_LAZY; +#endif /* defined(__HAVE_LAZY_MBUF) */ + 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; @@ -447,14 +449,12 @@ 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 proc *p) { - struct proc *p; - struct protosw *prp; + const struct protosw *prp; struct socket *so; int error, s; - p = curproc; /* XXX */ if (proto) prp = pffindproto(dom, proto, type); else @@ -478,7 +478,9 @@ socreate(int dom, struct socket **aso, i so->so_mowner = &prp->pr_domain->dom_mowner; #endif if (p != 0) - so->so_uid = p->p_ucred->cr_uid; + so->so_uidinfo = uid_find(p->p_ucred->cr_uid); + else + so->so_uidinfo = uid_find(0); error = (*prp->pr_usrreq)(so, PRU_ATTACH, (struct mbuf *)0, (struct mbuf *)(long)proto, (struct mbuf *)0, p); if (error) { @@ -540,7 +542,13 @@ 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); } @@ -639,12 +647,10 @@ soaccept(struct socket *so, struct mbuf } int -soconnect(struct socket *so, struct mbuf *nam) +soconnect(struct socket *so, struct mbuf *nam, struct proc *p) { - struct proc *p; int s, error; - p = curproc; /* XXX */ if (so->so_options & SO_ACCEPTCONN) return (EOPNOTSUPP); s = splsoftnet(); @@ -721,16 +727,14 @@ 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 proc *p) { - struct proc *p; struct mbuf **mp, *m; long space, len, resid, clen, mlen; int error, s, dontroute, atomic; sodopendfree(so); - p = curproc; /* XXX */ clen = 0; atomic = sosendallatonce(so) || top; if (uio) @@ -751,7 +755,8 @@ 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; } @@ -866,7 +871,7 @@ sosend(struct socket *so, struct mbuf *a break; } } while (space > 0 && atomic); - + s = splsoftnet(); if (so->so_state & SS_CANTSENDMORE) @@ -924,9 +929,10 @@ int soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio, struct mbuf **mp0, struct mbuf **controlp, int *flagsp) { + struct proc * p; 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 +940,8 @@ soreceive(struct socket *so, struct mbuf mp = mp0; type = 0; orig_resid = uio->uio_resid; + p = uio->uio_procp; + if (paddr) *paddr = 0; if (controlp) @@ -949,8 +957,8 @@ 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, - (struct proc *)0); + (struct mbuf *)(long)(flags & MSG_PEEK), + (struct mbuf *)0, p); if (error) goto bad; do { @@ -967,7 +975,7 @@ soreceive(struct socket *so, struct mbuf *mp = (struct mbuf *)0; 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); + (struct mbuf *)0, (struct mbuf *)0, p); restart: if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0) @@ -1040,10 +1048,8 @@ 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 (p) + p->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 +1087,23 @@ 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 && p && mtod(m, struct cmsghdr *)->cmsg_type == SCM_RIGHTS) - error = (*pr->pr_domain->dom_externalize)(m); + error = (*dom->dom_externalize)(m, p); *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; } @@ -1270,8 +1284,7 @@ soreceive(struct socket *so, struct mbuf (*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0, (struct mbuf *)(long)flags, - (struct mbuf *)0, - (struct proc *)0); + (struct mbuf *)0, p); SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 2"); SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 2"); error = sbwait(&so->so_rcv); @@ -1308,8 +1321,7 @@ soreceive(struct socket *so, struct mbuf 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); + (struct mbuf *)(long)flags, (struct mbuf *)0, p); } if (orig_resid == uio->uio_resid && orig_resid && (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) { @@ -1317,7 +1329,7 @@ soreceive(struct socket *so, struct mbuf splx(s); goto restart; } - + if (flagsp) *flagsp |= flags; release: @@ -1329,7 +1341,7 @@ 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)) @@ -1347,7 +1359,7 @@ void sorflush(struct socket *so) { struct sockbuf *sb, asb; - struct protosw *pr; + const struct protosw *pr; int s; sb = &so->so_rcv; @@ -1367,7 +1379,7 @@ 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 @@ -1441,7 +1453,7 @@ 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 +1481,14 @@ 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; } @@ -1586,6 +1598,10 @@ 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 +1638,7 @@ 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 +1669,7 @@ 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 +1694,7 @@ 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); }