version 1.111.2.15, 2008/03/24 09:39:02 |
version 1.209.2.1.4.3, 2017/07/21 03:55:56 |
|
|
/* $NetBSD$ */ |
/* $NetBSD$ */ |
|
|
/*- |
/*- |
* Copyright (c) 2002, 2007, 2008 The NetBSD Foundation, Inc. |
* Copyright (c) 2002, 2007, 2008, 2009 The NetBSD Foundation, Inc. |
* All rights reserved. |
* All rights reserved. |
* |
* |
* This code is derived from software contributed to The NetBSD Foundation |
* This code is derived from software contributed to The NetBSD Foundation |
* by Jason R. Thorpe of Wasabi Systems, Inc. |
* by Jason R. Thorpe of Wasabi Systems, Inc, and by Andrew Doran. |
* |
* |
* Redistribution and use in source and binary forms, with or without |
* Redistribution and use in source and binary forms, with or without |
* modification, are permitted provided that the following conditions |
* modification, are permitted provided that the following conditions |
|
|
* 2. Redistributions in binary form must reproduce the above copyright |
* 2. Redistributions in binary form must reproduce the above copyright |
* notice, this list of conditions and the following disclaimer in the |
* notice, this list of conditions and the following disclaimer in the |
* documentation and/or other materials provided with the distribution. |
* documentation and/or other materials provided with the distribution. |
* 3. All advertising materials mentioning features or use of this software |
|
* must display the following acknowledgement: |
|
* This product includes software developed by the NetBSD |
|
* Foundation, Inc. and its contributors. |
|
* 4. Neither the name of The NetBSD Foundation nor the names of its |
|
* contributors may be used to endorse or promote products derived |
|
* from this software without specific prior written permission. |
|
* |
* |
* THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS |
* THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS |
* ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED |
* ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED |
|
|
*/ |
*/ |
|
|
/* |
/* |
|
* Copyright (c) 2004 The FreeBSD Foundation |
|
* Copyright (c) 2004 Robert Watson |
* Copyright (c) 1982, 1986, 1988, 1990, 1993 |
* Copyright (c) 1982, 1986, 1988, 1990, 1993 |
* The Regents of the University of California. All rights reserved. |
* The Regents of the University of California. All rights reserved. |
* |
* |
|
|
#include <sys/cdefs.h> |
#include <sys/cdefs.h> |
__KERNEL_RCSID(0, "$NetBSD$"); |
__KERNEL_RCSID(0, "$NetBSD$"); |
|
|
|
#include "opt_compat_netbsd.h" |
#include "opt_sock_counters.h" |
#include "opt_sock_counters.h" |
#include "opt_sosend_loan.h" |
#include "opt_sosend_loan.h" |
#include "opt_mbuftrace.h" |
#include "opt_mbuftrace.h" |
#include "opt_somaxkva.h" |
#include "opt_somaxkva.h" |
|
#include "opt_multiprocessor.h" /* XXX */ |
|
|
#include <sys/param.h> |
#include <sys/param.h> |
#include <sys/systm.h> |
#include <sys/systm.h> |
#include <sys/proc.h> |
#include <sys/proc.h> |
#include <sys/file.h> |
#include <sys/file.h> |
#include <sys/filedesc.h> |
#include <sys/filedesc.h> |
#include <sys/malloc.h> |
#include <sys/kmem.h> |
#include <sys/mbuf.h> |
#include <sys/mbuf.h> |
#include <sys/domain.h> |
#include <sys/domain.h> |
#include <sys/kernel.h> |
#include <sys/kernel.h> |
Line 89 __KERNEL_RCSID(0, "$NetBSD$"); |
|
Line 86 __KERNEL_RCSID(0, "$NetBSD$"); |
|
#include <sys/socketvar.h> |
#include <sys/socketvar.h> |
#include <sys/signalvar.h> |
#include <sys/signalvar.h> |
#include <sys/resourcevar.h> |
#include <sys/resourcevar.h> |
#include <sys/pool.h> |
#include <sys/uidinfo.h> |
#include <sys/event.h> |
#include <sys/event.h> |
#include <sys/poll.h> |
#include <sys/poll.h> |
#include <sys/kauth.h> |
#include <sys/kauth.h> |
#include <sys/mutex.h> |
#include <sys/mutex.h> |
#include <sys/condvar.h> |
#include <sys/condvar.h> |
|
#include <sys/kthread.h> |
|
|
#include <uvm/uvm.h> |
#ifdef COMPAT_50 |
|
#include <compat/sys/time.h> |
|
#include <compat/sys/socket.h> |
|
#endif |
|
|
POOL_INIT(socket_pool, sizeof(struct socket), 0, 0, 0, "sockpl", NULL, |
#include <uvm/uvm_extern.h> |
IPL_SOFTNET); |
#include <uvm/uvm_loan.h> |
|
#include <uvm/uvm_page.h> |
|
|
MALLOC_DEFINE(M_SOOPTS, "soopts", "socket options"); |
MALLOC_DEFINE(M_SOOPTS, "soopts", "socket options"); |
MALLOC_DEFINE(M_SONAME, "soname", "socket name"); |
MALLOC_DEFINE(M_SONAME, "soname", "socket name"); |
Line 108 extern const struct fileops socketops; |
|
Line 110 extern const struct fileops socketops; |
|
|
|
extern int somaxconn; /* patchable (XXX sysctl) */ |
extern int somaxconn; /* patchable (XXX sysctl) */ |
int somaxconn = SOMAXCONN; |
int somaxconn = SOMAXCONN; |
|
kmutex_t *softnet_lock; |
|
|
#ifdef SOSEND_COUNTERS |
#ifdef SOSEND_COUNTERS |
#include <sys/device.h> |
#include <sys/device.h> |
Line 133 EVCNT_ATTACH_STATIC(sosend_kvalimit); |
|
Line 136 EVCNT_ATTACH_STATIC(sosend_kvalimit); |
|
|
|
#endif /* SOSEND_COUNTERS */ |
#endif /* SOSEND_COUNTERS */ |
|
|
static struct callback_entry sokva_reclaimerentry; |
#if defined(SOSEND_NO_LOAN) || defined(MULTIPROCESSOR) |
|
|
#ifdef SOSEND_NO_LOAN |
|
int sock_loan_thresh = -1; |
int sock_loan_thresh = -1; |
#else |
#else |
int sock_loan_thresh = 4096; |
int sock_loan_thresh = 4096; |
#endif |
#endif |
|
|
static kmutex_t so_pendfree_lock; |
static kmutex_t so_pendfree_lock; |
static struct mbuf *so_pendfree; |
static struct mbuf *so_pendfree = NULL; |
|
|
#ifndef SOMAXKVA |
#ifndef SOMAXKVA |
#define SOMAXKVA (16 * 1024 * 1024) |
#define SOMAXKVA (16 * 1024 * 1024) |
Line 151 int somaxkva = SOMAXKVA; |
|
Line 152 int somaxkva = SOMAXKVA; |
|
static int socurkva; |
static int socurkva; |
static kcondvar_t socurkva_cv; |
static kcondvar_t socurkva_cv; |
|
|
|
static kauth_listener_t socket_listener; |
|
|
#define SOCK_LOAN_CHUNK 65536 |
#define SOCK_LOAN_CHUNK 65536 |
|
|
static size_t sodopendfree(void); |
static void sopendfree_thread(void *); |
static size_t sodopendfreel(void); |
static kcondvar_t pendfree_thread_cv; |
|
static lwp_t *sopendfree_lwp; |
|
|
|
static void sysctl_kern_somaxkva_setup(void); |
|
static struct sysctllog *socket_sysctllog; |
|
|
static vsize_t |
static vsize_t |
sokvareserve(struct socket *so, vsize_t len) |
sokvareserve(struct socket *so, vsize_t len) |
Line 163 sokvareserve(struct socket *so, vsize_t |
|
Line 170 sokvareserve(struct socket *so, vsize_t |
|
|
|
mutex_enter(&so_pendfree_lock); |
mutex_enter(&so_pendfree_lock); |
while (socurkva + len > somaxkva) { |
while (socurkva + len > somaxkva) { |
size_t freed; |
|
|
|
/* |
|
* try to do pendfree. |
|
*/ |
|
|
|
freed = sodopendfreel(); |
|
|
|
/* |
|
* if some kva was freed, try again. |
|
*/ |
|
|
|
if (freed) |
|
continue; |
|
|
|
SOSEND_COUNTER_INCR(&sosend_kvalimit); |
SOSEND_COUNTER_INCR(&sosend_kvalimit); |
error = cv_wait_sig(&socurkva_cv, &so_pendfree_lock); |
error = cv_wait_sig(&socurkva_cv, &so_pendfree_lock); |
if (error) { |
if (error) { |
Line 205 sokvaunreserve(vsize_t len) |
|
Line 197 sokvaunreserve(vsize_t len) |
|
*/ |
*/ |
|
|
vaddr_t |
vaddr_t |
sokvaalloc(vsize_t len, struct socket *so) |
sokvaalloc(vaddr_t sva, vsize_t len, struct socket *so) |
{ |
{ |
vaddr_t lva; |
vaddr_t lva; |
|
|
Line 220 sokvaalloc(vsize_t len, struct socket *s |
|
Line 212 sokvaalloc(vsize_t len, struct socket *s |
|
* allocate kva. |
* allocate kva. |
*/ |
*/ |
|
|
lva = uvm_km_alloc(kernel_map, len, 0, UVM_KMF_VAONLY | UVM_KMF_WAITVA); |
lva = uvm_km_alloc(kernel_map, len, atop(sva) & uvmexp.colormask, |
|
UVM_KMF_COLORMATCH | UVM_KMF_VAONLY | UVM_KMF_WAITVA); |
if (lva == 0) { |
if (lva == 0) { |
sokvaunreserve(len); |
sokvaunreserve(len); |
return (0); |
return (0); |
Line 270 sodoloanfree(struct vm_page **pgs, void |
|
Line 263 sodoloanfree(struct vm_page **pgs, void |
|
sokvafree(sva, len); |
sokvafree(sva, len); |
} |
} |
|
|
static size_t |
|
sodopendfree(void) |
|
{ |
|
size_t rv; |
|
|
|
mutex_enter(&so_pendfree_lock); |
|
rv = sodopendfreel(); |
|
mutex_exit(&so_pendfree_lock); |
|
|
|
return rv; |
|
} |
|
|
|
/* |
/* |
* sodopendfreel: free mbufs on "pendfree" list. |
* sopendfree_thread: free mbufs on "pendfree" list. |
* unlock and relock so_pendfree_lock when freeing mbufs. |
* unlock and relock so_pendfree_lock when freeing mbufs. |
* |
|
* => called with so_pendfree_lock held. |
|
*/ |
*/ |
|
|
static size_t |
static void |
sodopendfreel(void) |
sopendfree_thread(void *v) |
{ |
{ |
struct mbuf *m, *next; |
struct mbuf *m, *next; |
size_t rv = 0; |
size_t rv; |
|
|
KASSERT(mutex_owned(&so_pendfree_lock)); |
|
|
|
while (so_pendfree != NULL) { |
mutex_enter(&so_pendfree_lock); |
m = so_pendfree; |
|
so_pendfree = NULL; |
|
mutex_exit(&so_pendfree_lock); |
|
|
|
for (; m != NULL; m = next) { |
for (;;) { |
next = m->m_next; |
rv = 0; |
KASSERT((~m->m_flags & (M_EXT|M_EXT_PAGES)) == 0); |
while (so_pendfree != NULL) { |
KASSERT(m->m_ext.ext_refcnt == 0); |
m = so_pendfree; |
|
so_pendfree = NULL; |
|
mutex_exit(&so_pendfree_lock); |
|
|
|
for (; m != NULL; m = next) { |
|
next = m->m_next; |
|
KASSERT((~m->m_flags & (M_EXT|M_EXT_PAGES)) == 0); |
|
KASSERT(m->m_ext.ext_refcnt == 0); |
|
|
|
rv += m->m_ext.ext_size; |
|
sodoloanfree(m->m_ext.ext_pgs, m->m_ext.ext_buf, |
|
m->m_ext.ext_size); |
|
pool_cache_put(mb_cache, m); |
|
} |
|
|
rv += m->m_ext.ext_size; |
mutex_enter(&so_pendfree_lock); |
sodoloanfree(m->m_ext.ext_pgs, m->m_ext.ext_buf, |
|
m->m_ext.ext_size); |
|
pool_cache_put(mb_cache, m); |
|
} |
} |
|
if (rv) |
mutex_enter(&so_pendfree_lock); |
cv_broadcast(&socurkva_cv); |
|
cv_wait(&pendfree_thread_cv, &so_pendfree_lock); |
} |
} |
|
panic("sopendfree_thread"); |
return (rv); |
/* NOTREACHED */ |
} |
} |
|
|
void |
void |
Line 335 soloanfree(struct mbuf *m, void *buf, si |
|
Line 320 soloanfree(struct mbuf *m, void *buf, si |
|
mutex_enter(&so_pendfree_lock); |
mutex_enter(&so_pendfree_lock); |
m->m_next = so_pendfree; |
m->m_next = so_pendfree; |
so_pendfree = m; |
so_pendfree = m; |
cv_broadcast(&socurkva_cv); |
cv_signal(&pendfree_thread_cv); |
mutex_exit(&so_pendfree_lock); |
mutex_exit(&so_pendfree_lock); |
} |
} |
|
|
Line 363 sosend_loan(struct socket *so, struct ui |
|
Line 348 sosend_loan(struct socket *so, struct ui |
|
len = eva - sva; |
len = eva - sva; |
npgs = len >> PAGE_SHIFT; |
npgs = len >> PAGE_SHIFT; |
|
|
/* XXX KDASSERT */ |
|
KASSERT(npgs <= M_EXT_MAXPAGES); |
KASSERT(npgs <= M_EXT_MAXPAGES); |
|
|
lva = sokvaalloc(len, so); |
lva = sokvaalloc(sva, len, so); |
if (lva == 0) |
if (lva == 0) |
return 0; |
return 0; |
|
|
Line 379 sosend_loan(struct socket *so, struct ui |
|
Line 363 sosend_loan(struct socket *so, struct ui |
|
|
|
for (i = 0, va = lva; i < npgs; i++, va += PAGE_SIZE) |
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]), |
pmap_kenter_pa(va, VM_PAGE_TO_PHYS(m->m_ext.ext_pgs[i]), |
VM_PROT_READ); |
VM_PROT_READ, 0); |
pmap_update(pmap_kernel()); |
pmap_update(pmap_kernel()); |
|
|
lva += (vaddr_t) iov->iov_base & PAGE_MASK; |
lva += (vaddr_t) iov->iov_base & PAGE_MASK; |
Line 399 sosend_loan(struct socket *so, struct ui |
|
Line 383 sosend_loan(struct socket *so, struct ui |
|
return (space); |
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; |
|
} |
|
|
|
struct mbuf * |
struct mbuf * |
getsombuf(struct socket *so, int type) |
getsombuf(struct socket *so, int type) |
{ |
{ |
Line 423 getsombuf(struct socket *so, int type) |
|
Line 393 getsombuf(struct socket *so, int type) |
|
return m; |
return m; |
} |
} |
|
|
struct mbuf * |
static int |
m_intopt(struct socket *so, int val) |
socket_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, |
|
void *arg0, void *arg1, void *arg2, void *arg3) |
{ |
{ |
struct mbuf *m; |
int result; |
|
enum kauth_network_req req; |
|
|
m = getsombuf(so, MT_SOOPTS); |
result = KAUTH_RESULT_DEFER; |
m->m_len = sizeof(int); |
req = (enum kauth_network_req)arg0; |
*mtod(m, int *) = val; |
|
return m; |
if ((action != KAUTH_NETWORK_SOCKET) && |
|
(action != KAUTH_NETWORK_BIND)) |
|
return result; |
|
|
|
switch (req) { |
|
case KAUTH_REQ_NETWORK_BIND_PORT: |
|
result = KAUTH_RESULT_ALLOW; |
|
break; |
|
|
|
case KAUTH_REQ_NETWORK_SOCKET_DROP: { |
|
/* Normal users can only drop their own connections. */ |
|
struct socket *so = (struct socket *)arg1; |
|
|
|
if (so->so_cred && proc_uidmatch(cred, so->so_cred) == 0) |
|
result = KAUTH_RESULT_ALLOW; |
|
|
|
break; |
|
} |
|
|
|
case KAUTH_REQ_NETWORK_SOCKET_OPEN: |
|
/* We allow "raw" routing/bluetooth sockets to anyone. */ |
|
if ((u_long)arg1 == PF_ROUTE || (u_long)arg1 == PF_OROUTE |
|
|| (u_long)arg1 == PF_BLUETOOTH) { |
|
result = KAUTH_RESULT_ALLOW; |
|
} else { |
|
/* Privileged, let secmodel handle this. */ |
|
if ((u_long)arg2 == SOCK_RAW) |
|
break; |
|
} |
|
|
|
result = KAUTH_RESULT_ALLOW; |
|
|
|
break; |
|
|
|
case KAUTH_REQ_NETWORK_SOCKET_CANSEE: |
|
result = KAUTH_RESULT_ALLOW; |
|
|
|
break; |
|
|
|
default: |
|
break; |
|
} |
|
|
|
return result; |
} |
} |
|
|
void |
void |
soinit(void) |
soinit(void) |
{ |
{ |
|
|
|
sysctl_kern_somaxkva_setup(); |
|
|
mutex_init(&so_pendfree_lock, MUTEX_DEFAULT, IPL_VM); |
mutex_init(&so_pendfree_lock, MUTEX_DEFAULT, IPL_VM); |
|
softnet_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); |
cv_init(&socurkva_cv, "sokva"); |
cv_init(&socurkva_cv, "sokva"); |
|
cv_init(&pendfree_thread_cv, "sopendfr"); |
|
soinit2(); |
|
|
/* Set the initial adjusted socket buffer size. */ |
/* Set the initial adjusted socket buffer size. */ |
if (sb_max_set(sb_max)) |
if (sb_max_set(sb_max)) |
panic("bad initial sb_max value: %lu", sb_max); |
panic("bad initial sb_max value: %lu", sb_max); |
|
|
callback_register(&vm_map_to_kernel(kernel_map)->vmk_reclaim_callback, |
socket_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK, |
&sokva_reclaimerentry, NULL, sokva_reclaim_callback); |
socket_listener_cb, NULL); |
|
} |
|
|
|
void |
|
soinit1(void) |
|
{ |
|
int error = kthread_create(PRI_NONE, KTHREAD_MPSAFE, NULL, |
|
sopendfree_thread, NULL, &sopendfree_lwp, "sopendfree"); |
|
if (error) |
|
panic("soinit1 %d", error); |
} |
} |
|
|
/* |
/* |
|
|
*/ |
*/ |
/*ARGSUSED*/ |
/*ARGSUSED*/ |
int |
int |
socreate(int dom, struct socket **aso, int type, int proto, struct lwp *l) |
socreate(int dom, struct socket **aso, int type, int proto, struct lwp *l, |
|
struct socket *lockso) |
{ |
{ |
const struct protosw *prp; |
const struct protosw *prp; |
struct socket *so; |
struct socket *so; |
uid_t uid; |
uid_t uid; |
int error, s; |
int error; |
|
kmutex_t *lock; |
|
|
error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET, |
error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET, |
KAUTH_REQ_NETWORK_SOCKET_OPEN, KAUTH_ARG(dom), KAUTH_ARG(type), |
KAUTH_REQ_NETWORK_SOCKET_OPEN, KAUTH_ARG(dom), KAUTH_ARG(type), |
Line 488 socreate(int dom, struct socket **aso, i |
|
Line 519 socreate(int dom, struct socket **aso, i |
|
return EPROTONOSUPPORT; |
return EPROTONOSUPPORT; |
if (prp->pr_type != type) |
if (prp->pr_type != type) |
return EPROTOTYPE; |
return EPROTOTYPE; |
s = splsoftnet(); |
|
so = pool_get(&socket_pool, PR_WAITOK); |
so = soget(true); |
memset(so, 0, sizeof(*so)); |
|
TAILQ_INIT(&so->so_q0); |
|
TAILQ_INIT(&so->so_q); |
|
so->so_type = type; |
so->so_type = type; |
so->so_proto = prp; |
so->so_proto = prp; |
so->so_send = sosend; |
so->so_send = sosend; |
Line 502 socreate(int dom, struct socket **aso, i |
|
Line 530 socreate(int dom, struct socket **aso, i |
|
so->so_snd.sb_mowner = &prp->pr_domain->dom_mowner; |
so->so_snd.sb_mowner = &prp->pr_domain->dom_mowner; |
so->so_mowner = &prp->pr_domain->dom_mowner; |
so->so_mowner = &prp->pr_domain->dom_mowner; |
#endif |
#endif |
selinit(&so->so_rcv.sb_sel); |
|
selinit(&so->so_snd.sb_sel); |
|
uid = kauth_cred_geteuid(l->l_cred); |
uid = kauth_cred_geteuid(l->l_cred); |
so->so_uidinfo = uid_find(uid); |
so->so_uidinfo = uid_find(uid); |
|
so->so_cpid = l->l_proc->p_pid; |
|
if (lockso != NULL) { |
|
/* Caller wants us to share a lock. */ |
|
lock = lockso->so_lock; |
|
so->so_lock = lock; |
|
mutex_obj_hold(lock); |
|
mutex_enter(lock); |
|
} else { |
|
/* Lock assigned and taken during PRU_ATTACH. */ |
|
} |
error = (*prp->pr_usrreq)(so, PRU_ATTACH, NULL, |
error = (*prp->pr_usrreq)(so, PRU_ATTACH, NULL, |
(struct mbuf *)(long)proto, NULL, l); |
(struct mbuf *)(long)proto, NULL, l); |
|
KASSERT(solocked(so)); |
if (error != 0) { |
if (error != 0) { |
so->so_state |= SS_NOFDREF; |
so->so_state |= SS_NOFDREF; |
sofree(so); |
sofree(so); |
splx(s); |
|
return error; |
return error; |
} |
} |
splx(s); |
so->so_cred = kauth_cred_dup(l->l_cred); |
|
sounlock(so); |
*aso = so; |
*aso = so; |
return 0; |
return 0; |
} |
} |
Line 529 fsocreate(int domain, struct socket **so |
|
Line 566 fsocreate(int domain, struct socket **so |
|
struct socket *so; |
struct socket *so; |
struct file *fp; |
struct file *fp; |
int fd, error; |
int fd, error; |
|
int flags = type & SOCK_FLAGS_MASK; |
|
|
|
type &= ~SOCK_FLAGS_MASK; |
if ((error = fd_allocfile(&fp, &fd)) != 0) |
if ((error = fd_allocfile(&fp, &fd)) != 0) |
return (error); |
return error; |
fp->f_flag = FREAD|FWRITE; |
fd_set_exclose(l, fd, (flags & SOCK_CLOEXEC) != 0); |
|
fp->f_flag = FREAD|FWRITE|((flags & SOCK_NONBLOCK) ? FNONBLOCK : 0)| |
|
((flags & SOCK_NOSIGPIPE) ? FNOSIGPIPE : 0); |
fp->f_type = DTYPE_SOCKET; |
fp->f_type = DTYPE_SOCKET; |
fp->f_ops = &socketops; |
fp->f_ops = &socketops; |
error = socreate(domain, &so, type, protocol, l); |
error = socreate(domain, &so, type, protocol, l, NULL); |
if (error != 0) { |
if (error != 0) { |
fd_abort(curproc, fp, fd); |
fd_abort(curproc, fp, fd); |
} else { |
} else { |
Line 544 fsocreate(int domain, struct socket **so |
|
Line 585 fsocreate(int domain, struct socket **so |
|
fp->f_data = so; |
fp->f_data = so; |
fd_affix(curproc, fp, fd); |
fd_affix(curproc, fp, fd); |
*fdout = fd; |
*fdout = fd; |
|
if (flags & SOCK_NONBLOCK) |
|
so->so_state |= SS_NBIO; |
} |
} |
return error; |
return error; |
} |
} |
|
|
int |
int |
|
sofamily(const struct socket *so) |
|
{ |
|
const struct protosw *pr; |
|
const struct domain *dom; |
|
|
|
if ((pr = so->so_proto) == NULL) |
|
return AF_UNSPEC; |
|
if ((dom = pr->pr_domain) == NULL) |
|
return AF_UNSPEC; |
|
return dom->dom_family; |
|
} |
|
|
|
int |
sobind(struct socket *so, struct mbuf *nam, struct lwp *l) |
sobind(struct socket *so, struct mbuf *nam, struct lwp *l) |
{ |
{ |
int s, error; |
int error; |
|
|
s = splsoftnet(); |
solock(so); |
error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, NULL, nam, NULL, l); |
error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, NULL, nam, NULL, l); |
splx(s); |
sounlock(so); |
return error; |
return error; |
} |
} |
|
|
int |
int |
solisten(struct socket *so, int backlog, struct lwp *l) |
solisten(struct socket *so, int backlog, struct lwp *l) |
{ |
{ |
int s, error; |
int error; |
|
|
s = splsoftnet(); |
solock(so); |
|
if ((so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING | |
|
SS_ISDISCONNECTING)) != 0) { |
|
sounlock(so); |
|
return (EOPNOTSUPP); |
|
} |
error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, NULL, |
error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, NULL, |
NULL, NULL, l); |
NULL, NULL, l); |
if (error != 0) { |
if (error != 0) { |
splx(s); |
sounlock(so); |
return error; |
return error; |
} |
} |
if (TAILQ_EMPTY(&so->so_q)) |
if (TAILQ_EMPTY(&so->so_q)) |
Line 576 solisten(struct socket *so, int backlog, |
|
Line 637 solisten(struct socket *so, int backlog, |
|
if (backlog < 0) |
if (backlog < 0) |
backlog = 0; |
backlog = 0; |
so->so_qlimit = min(backlog, somaxconn); |
so->so_qlimit = min(backlog, somaxconn); |
splx(s); |
sounlock(so); |
return 0; |
return 0; |
} |
} |
|
|
void |
void |
sofree(struct socket *so) |
sofree(struct socket *so) |
{ |
{ |
|
u_int refs; |
|
|
|
KASSERT(solocked(so)); |
|
|
if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0) |
if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0) { |
|
sounlock(so); |
return; |
return; |
|
} |
if (so->so_head) { |
if (so->so_head) { |
/* |
/* |
* We must not decommission a socket that's on the accept(2) |
* We must not decommission a socket that's on the accept(2) |
* queue. If we do, then accept(2) may hang after select(2) |
* queue. If we do, then accept(2) may hang after select(2) |
* indicated that the listening socket was ready. |
* indicated that the listening socket was ready. |
*/ |
*/ |
if (!soqremque(so, 0)) |
if (!soqremque(so, 0)) { |
|
sounlock(so); |
return; |
return; |
|
} |
} |
} |
if (so->so_rcv.sb_hiwat) |
if (so->so_rcv.sb_hiwat) |
(void)chgsbsize(so->so_uidinfo, &so->so_rcv.sb_hiwat, 0, |
(void)chgsbsize(so->so_uidinfo, &so->so_rcv.sb_hiwat, 0, |
Line 602 sofree(struct socket *so) |
|
Line 670 sofree(struct socket *so) |
|
(void)chgsbsize(so->so_uidinfo, &so->so_snd.sb_hiwat, 0, |
(void)chgsbsize(so->so_uidinfo, &so->so_snd.sb_hiwat, 0, |
RLIM_INFINITY); |
RLIM_INFINITY); |
sbrelease(&so->so_snd, so); |
sbrelease(&so->so_snd, so); |
|
KASSERT(!cv_has_waiters(&so->so_cv)); |
|
KASSERT(!cv_has_waiters(&so->so_rcv.sb_cv)); |
|
KASSERT(!cv_has_waiters(&so->so_snd.sb_cv)); |
sorflush(so); |
sorflush(so); |
seldestroy(&so->so_rcv.sb_sel); |
refs = so->so_aborting; /* XXX */ |
seldestroy(&so->so_snd.sb_sel); |
/* Remove acccept filter if one is present. */ |
pool_put(&socket_pool, so); |
if (so->so_accf != NULL) |
|
(void)accept_filt_clear(so); |
|
sounlock(so); |
|
if (refs == 0) /* XXX */ |
|
soput(so); |
} |
} |
|
|
/* |
/* |
|
|
soclose(struct socket *so) |
soclose(struct socket *so) |
{ |
{ |
struct socket *so2; |
struct socket *so2; |
int s, error; |
int error; |
|
int error2; |
|
|
error = 0; |
error = 0; |
s = splsoftnet(); /* conservative */ |
solock(so); |
if (so->so_options & SO_ACCEPTCONN) { |
if (so->so_options & SO_ACCEPTCONN) { |
while ((so2 = TAILQ_FIRST(&so->so_q0)) != 0) { |
for (;;) { |
(void) soqremque(so2, 0); |
if ((so2 = TAILQ_FIRST(&so->so_q0)) != 0) { |
(void) soabort(so2); |
KASSERT(solocked2(so, so2)); |
} |
(void) soqremque(so2, 0); |
while ((so2 = TAILQ_FIRST(&so->so_q)) != 0) { |
/* soabort drops the lock. */ |
(void) soqremque(so2, 1); |
(void) soabort(so2); |
(void) soabort(so2); |
solock(so); |
|
continue; |
|
} |
|
if ((so2 = TAILQ_FIRST(&so->so_q)) != 0) { |
|
KASSERT(solocked2(so, so2)); |
|
(void) soqremque(so2, 1); |
|
/* soabort drops the lock. */ |
|
(void) soabort(so2); |
|
solock(so); |
|
continue; |
|
} |
|
break; |
} |
} |
} |
} |
if (so->so_pcb == 0) |
if (so->so_pcb == 0) |
Line 640 soclose(struct socket *so) |
|
Line 727 soclose(struct socket *so) |
|
goto drop; |
goto drop; |
} |
} |
if (so->so_options & SO_LINGER) { |
if (so->so_options & SO_LINGER) { |
if ((so->so_state & SS_ISDISCONNECTING) && so->so_nbio) |
if ((so->so_state & (SS_ISDISCONNECTING|SS_NBIO)) == |
|
(SS_ISDISCONNECTING|SS_NBIO)) |
goto drop; |
goto drop; |
while (so->so_state & SS_ISCONNECTED) { |
while (so->so_state & SS_ISCONNECTED) { |
error = tsleep((void *)&so->so_timeo, |
error = sowait(so, true, so->so_linger * hz); |
PSOCK | PCATCH, netcls, |
|
so->so_linger * hz); |
|
if (error) |
if (error) |
break; |
break; |
} |
} |
Line 653 soclose(struct socket *so) |
|
Line 739 soclose(struct socket *so) |
|
} |
} |
drop: |
drop: |
if (so->so_pcb) { |
if (so->so_pcb) { |
int error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH, |
error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH, |
NULL, NULL, NULL, NULL); |
NULL, NULL, NULL, NULL); |
if (error == 0) |
if (error == 0) |
error = error2; |
error = error2; |
Line 661 soclose(struct socket *so) |
|
Line 747 soclose(struct socket *so) |
|
discard: |
discard: |
if (so->so_state & SS_NOFDREF) |
if (so->so_state & SS_NOFDREF) |
panic("soclose: NOFDREF"); |
panic("soclose: NOFDREF"); |
|
kauth_cred_free(so->so_cred); |
so->so_state |= SS_NOFDREF; |
so->so_state |= SS_NOFDREF; |
sofree(so); |
sofree(so); |
splx(s); |
|
return (error); |
return (error); |
} |
} |
|
|
/* |
/* |
* Must be called at splsoftnet... |
* Must be called with the socket locked.. Will return with it unlocked. |
*/ |
*/ |
int |
int |
soabort(struct socket *so) |
soabort(struct socket *so) |
{ |
{ |
|
u_int refs; |
int error; |
int error; |
|
|
|
KASSERT(solocked(so)); |
KASSERT(so->so_head == NULL); |
KASSERT(so->so_head == NULL); |
|
|
|
so->so_aborting++; /* XXX */ |
error = (*so->so_proto->pr_usrreq)(so, PRU_ABORT, NULL, |
error = (*so->so_proto->pr_usrreq)(so, PRU_ABORT, NULL, |
NULL, NULL, NULL); |
NULL, NULL, NULL); |
if (error) { |
refs = --so->so_aborting; /* XXX */ |
|
if (error || (refs == 0)) { |
sofree(so); |
sofree(so); |
|
} else { |
|
sounlock(so); |
} |
} |
return error; |
return error; |
} |
} |
Line 687 soabort(struct socket *so) |
|
Line 780 soabort(struct socket *so) |
|
int |
int |
soaccept(struct socket *so, struct mbuf *nam) |
soaccept(struct socket *so, struct mbuf *nam) |
{ |
{ |
int s, error; |
int error; |
|
|
|
KASSERT(solocked(so)); |
|
|
error = 0; |
error = 0; |
s = splsoftnet(); |
|
if ((so->so_state & SS_NOFDREF) == 0) |
if ((so->so_state & SS_NOFDREF) == 0) |
panic("soaccept: !NOFDREF"); |
panic("soaccept: !NOFDREF"); |
so->so_state &= ~SS_NOFDREF; |
so->so_state &= ~SS_NOFDREF; |
Line 701 soaccept(struct socket *so, struct mbuf |
|
Line 795 soaccept(struct socket *so, struct mbuf |
|
else |
else |
error = ECONNABORTED; |
error = ECONNABORTED; |
|
|
splx(s); |
|
return (error); |
return (error); |
} |
} |
|
|
int |
int |
soconnect(struct socket *so, struct mbuf *nam, struct lwp *l) |
soconnect(struct socket *so, struct mbuf *nam, struct lwp *l) |
{ |
{ |
int s, error; |
int error; |
|
|
|
KASSERT(solocked(so)); |
|
|
if (so->so_options & SO_ACCEPTCONN) |
if (so->so_options & SO_ACCEPTCONN) |
return (EOPNOTSUPP); |
return (EOPNOTSUPP); |
s = splsoftnet(); |
|
/* |
/* |
* If protocol is connection-based, can only connect once. |
* If protocol is connection-based, can only connect once. |
* Otherwise, if connected, try to disconnect first. |
* Otherwise, if connected, try to disconnect first. |
Line 726 soconnect(struct socket *so, struct mbuf |
|
Line 820 soconnect(struct socket *so, struct mbuf |
|
else |
else |
error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT, |
error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT, |
NULL, nam, NULL, l); |
NULL, nam, NULL, l); |
splx(s); |
|
return (error); |
return (error); |
} |
} |
|
|
int |
int |
soconnect2(struct socket *so1, struct socket *so2) |
soconnect2(struct socket *so1, struct socket *so2) |
{ |
{ |
int s, error; |
int error; |
|
|
|
KASSERT(solocked2(so1, so2)); |
|
|
s = splsoftnet(); |
|
error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2, |
error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2, |
NULL, (struct mbuf *)so2, NULL, NULL); |
NULL, (struct mbuf *)so2, NULL, NULL); |
splx(s); |
|
return (error); |
return (error); |
} |
} |
|
|
int |
int |
sodisconnect(struct socket *so) |
sodisconnect(struct socket *so) |
{ |
{ |
int s, error; |
int error; |
|
|
|
KASSERT(solocked(so)); |
|
|
s = splsoftnet(); |
|
if ((so->so_state & SS_ISCONNECTED) == 0) { |
if ((so->so_state & SS_ISCONNECTED) == 0) { |
error = ENOTCONN; |
error = ENOTCONN; |
goto bad; |
} else if (so->so_state & SS_ISDISCONNECTING) { |
} |
|
if (so->so_state & SS_ISDISCONNECTING) { |
|
error = EALREADY; |
error = EALREADY; |
goto bad; |
} else { |
|
error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT, |
|
NULL, NULL, NULL, NULL); |
} |
} |
error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT, |
|
NULL, NULL, NULL, NULL); |
|
bad: |
|
splx(s); |
|
sodopendfree(); |
|
return (error); |
return (error); |
} |
} |
|
|
Line 790 sosend(struct socket *so, struct mbuf *a |
|
Line 879 sosend(struct socket *so, struct mbuf *a |
|
struct proc *p; |
struct proc *p; |
long space, len, resid, clen, mlen; |
long space, len, resid, clen, mlen; |
int error, s, dontroute, atomic; |
int error, s, dontroute, atomic; |
|
short wakeup_state = 0; |
|
|
p = l->l_proc; |
p = l->l_proc; |
sodopendfree(); |
|
|
|
clen = 0; |
clen = 0; |
|
|
|
/* |
|
* solock() provides atomicity of access. splsoftnet() prevents |
|
* protocol processing soft interrupts from interrupting us and |
|
* blocking (expensive). |
|
*/ |
|
s = splsoftnet(); |
|
solock(so); |
atomic = sosendallatonce(so) || top; |
atomic = sosendallatonce(so) || top; |
if (uio) |
if (uio) |
resid = uio->uio_resid; |
resid = uio->uio_resid; |
Line 814 sosend(struct socket *so, struct mbuf *a |
|
Line 910 sosend(struct socket *so, struct mbuf *a |
|
dontroute = |
dontroute = |
(flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 && |
(flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 && |
(so->so_proto->pr_flags & PR_ATOMIC); |
(so->so_proto->pr_flags & PR_ATOMIC); |
if (p) |
l->l_ru.ru_msgsnd++; |
p->p_stats->p_ru.ru_msgsnd++; |
|
if (control) |
if (control) |
clen = control->m_len; |
clen = control->m_len; |
#define snderr(errno) { error = errno; splx(s); goto release; } |
|
|
|
restart: |
restart: |
if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0) |
if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0) |
goto out; |
goto out; |
do { |
do { |
s = splsoftnet(); |
if (so->so_state & SS_CANTSENDMORE) { |
if (so->so_state & SS_CANTSENDMORE) |
error = EPIPE; |
snderr(EPIPE); |
goto release; |
|
} |
if (so->so_error) { |
if (so->so_error) { |
error = so->so_error; |
error = so->so_error; |
so->so_error = 0; |
so->so_error = 0; |
splx(s); |
|
goto release; |
goto release; |
} |
} |
if ((so->so_state & SS_ISCONNECTED) == 0) { |
if ((so->so_state & SS_ISCONNECTED) == 0) { |
if (so->so_proto->pr_flags & PR_CONNREQUIRED) { |
if (so->so_proto->pr_flags & PR_CONNREQUIRED) { |
if ((so->so_state & SS_ISCONFIRMING) == 0 && |
if ((so->so_state & SS_ISCONFIRMING) == 0 && |
!(resid == 0 && clen != 0)) |
!(resid == 0 && clen != 0)) { |
snderr(ENOTCONN); |
error = ENOTCONN; |
} else if (addr == 0) |
goto release; |
snderr(EDESTADDRREQ); |
} |
|
} else if (addr == 0) { |
|
error = EDESTADDRREQ; |
|
goto release; |
|
} |
} |
} |
space = sbspace(&so->so_snd); |
space = sbspace(&so->so_snd); |
if (flags & MSG_OOB) |
if (flags & MSG_OOB) |
space += 1024; |
space += 1024; |
if ((atomic && resid > so->so_snd.sb_hiwat) || |
if ((atomic && resid > so->so_snd.sb_hiwat) || |
clen > so->so_snd.sb_hiwat) |
clen > so->so_snd.sb_hiwat) { |
snderr(EMSGSIZE); |
error = EMSGSIZE; |
|
goto release; |
|
} |
if (space < resid + clen && |
if (space < resid + clen && |
(atomic || space < so->so_snd.sb_lowat || space < clen)) { |
(atomic || space < so->so_snd.sb_lowat || space < clen)) { |
if (so->so_nbio) |
if ((so->so_state & SS_NBIO) || (flags & MSG_NBIO)) { |
snderr(EWOULDBLOCK); |
error = EWOULDBLOCK; |
|
goto release; |
|
} |
sbunlock(&so->so_snd); |
sbunlock(&so->so_snd); |
|
if (wakeup_state & SS_RESTARTSYS) { |
|
error = ERESTART; |
|
goto out; |
|
} |
error = sbwait(&so->so_snd); |
error = sbwait(&so->so_snd); |
splx(s); |
|
if (error) |
if (error) |
goto out; |
goto out; |
|
wakeup_state = so->so_state; |
goto restart; |
goto restart; |
} |
} |
splx(s); |
wakeup_state = 0; |
mp = ⊤ |
mp = ⊤ |
space -= clen; |
space -= clen; |
do { |
do { |
Line 870 sosend(struct socket *so, struct mbuf *a |
|
Line 975 sosend(struct socket *so, struct mbuf *a |
|
if (flags & MSG_EOR) |
if (flags & MSG_EOR) |
top->m_flags |= M_EOR; |
top->m_flags |= M_EOR; |
} else do { |
} else do { |
|
sounlock(so); |
|
splx(s); |
if (top == NULL) { |
if (top == NULL) { |
m = m_gethdr(M_WAIT, MT_DATA); |
m = m_gethdr(M_WAIT, MT_DATA); |
mlen = MHLEN; |
mlen = MHLEN; |
Line 891 sosend(struct socket *so, struct mbuf *a |
|
Line 998 sosend(struct socket *so, struct mbuf *a |
|
} |
} |
if (resid >= MINCLSIZE && space >= MCLBYTES) { |
if (resid >= MINCLSIZE && space >= MCLBYTES) { |
SOSEND_COUNTER_INCR(&sosend_copy_big); |
SOSEND_COUNTER_INCR(&sosend_copy_big); |
m_clget(m, M_WAIT); |
m_clget(m, M_DONTWAIT); |
if ((m->m_flags & M_EXT) == 0) |
if ((m->m_flags & M_EXT) == 0) |
goto nopages; |
goto nopages; |
mlen = MCLBYTES; |
mlen = MCLBYTES; |
Line 920 sosend(struct socket *so, struct mbuf *a |
|
Line 1027 sosend(struct socket *so, struct mbuf *a |
|
m->m_len = len; |
m->m_len = len; |
*mp = m; |
*mp = m; |
top->m_pkthdr.len += len; |
top->m_pkthdr.len += len; |
|
s = splsoftnet(); |
|
solock(so); |
if (error != 0) |
if (error != 0) |
goto release; |
goto release; |
mp = &m->m_next; |
mp = &m->m_next; |
Line 930 sosend(struct socket *so, struct mbuf *a |
|
Line 1039 sosend(struct socket *so, struct mbuf *a |
|
} |
} |
} while (space > 0 && atomic); |
} while (space > 0 && atomic); |
|
|
s = splsoftnet(); |
if (so->so_state & SS_CANTSENDMORE) { |
|
error = EPIPE; |
if (so->so_state & SS_CANTSENDMORE) |
goto release; |
snderr(EPIPE); |
} |
|
|
if (dontroute) |
if (dontroute) |
so->so_options |= SO_DONTROUTE; |
so->so_options |= SO_DONTROUTE; |
if (resid > 0) |
if (resid > 0) |
so->so_state |= SS_MORETOCOME; |
so->so_state |= SS_MORETOCOME; |
error = (*so->so_proto->pr_usrreq)(so, |
error = (*so->so_proto->pr_usrreq)(so, |
(flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND, |
(flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND, |
top, addr, control, curlwp); /* XXX */ |
top, addr, control, curlwp); |
if (dontroute) |
if (dontroute) |
so->so_options &= ~SO_DONTROUTE; |
so->so_options &= ~SO_DONTROUTE; |
if (resid > 0) |
if (resid > 0) |
so->so_state &= ~SS_MORETOCOME; |
so->so_state &= ~SS_MORETOCOME; |
splx(s); |
|
|
|
clen = 0; |
clen = 0; |
control = NULL; |
control = NULL; |
top = NULL; |
top = NULL; |
Line 960 sosend(struct socket *so, struct mbuf *a |
|
Line 1066 sosend(struct socket *so, struct mbuf *a |
|
release: |
release: |
sbunlock(&so->so_snd); |
sbunlock(&so->so_snd); |
out: |
out: |
|
sounlock(so); |
|
splx(s); |
if (top) |
if (top) |
m_freem(top); |
m_freem(top); |
if (control) |
if (control) |
Line 968 sosend(struct socket *so, struct mbuf *a |
|
Line 1076 sosend(struct socket *so, struct mbuf *a |
|
} |
} |
|
|
/* |
/* |
|
* Following replacement or removal of the first mbuf on the first |
|
* mbuf chain of a socket buffer, push necessary state changes back |
|
* into the socket buffer so that other consumers see the values |
|
* consistently. 'nextrecord' is the callers locally stored value of |
|
* the original value of sb->sb_mb->m_nextpkt which must be restored |
|
* when the lead mbuf changes. NOTE: 'nextrecord' may be NULL. |
|
*/ |
|
static void |
|
sbsync(struct sockbuf *sb, struct mbuf *nextrecord) |
|
{ |
|
|
|
KASSERT(solocked(sb->sb_so)); |
|
|
|
/* |
|
* First, update for the new value of nextrecord. If necessary, |
|
* make it the first record. |
|
*/ |
|
if (sb->sb_mb != NULL) |
|
sb->sb_mb->m_nextpkt = nextrecord; |
|
else |
|
sb->sb_mb = nextrecord; |
|
|
|
/* |
|
* Now update any dependent socket buffer fields to reflect |
|
* the new state. This is an inline of SB_EMPTY_FIXUP, with |
|
* the addition of a second clause that takes care of the |
|
* case where sb_mb has been updated, but remains the last |
|
* record. |
|
*/ |
|
if (sb->sb_mb == NULL) { |
|
sb->sb_mbtail = NULL; |
|
sb->sb_lastrecord = NULL; |
|
} else if (sb->sb_mb->m_nextpkt == NULL) |
|
sb->sb_lastrecord = sb->sb_mb; |
|
} |
|
|
|
/* |
* Implement receive operations on a socket. |
* Implement receive operations on a socket. |
* We depend on the way that records are added to the sockbuf |
* We depend on the way that records are added to the sockbuf |
* by sbappend*. In particular, each record (mbufs linked through m_next) |
* by sbappend*. In particular, each record (mbufs linked through m_next) |
Line 988 soreceive(struct socket *so, struct mbuf |
|
Line 1133 soreceive(struct socket *so, struct mbuf |
|
struct mbuf **mp0, struct mbuf **controlp, int *flagsp) |
struct mbuf **mp0, struct mbuf **controlp, int *flagsp) |
{ |
{ |
struct lwp *l = curlwp; |
struct lwp *l = curlwp; |
struct mbuf *m, **mp; |
struct mbuf *m, **mp, *mt; |
int atomic, flags, len, error, s, offset, moff, type, orig_resid; |
size_t len, offset, moff, orig_resid; |
|
int atomic, flags, error, s, type; |
const struct protosw *pr; |
const struct protosw *pr; |
struct mbuf *nextrecord; |
struct mbuf *nextrecord; |
int mbuf_removed = 0; |
int mbuf_removed = 0; |
const struct domain *dom; |
const struct domain *dom; |
|
short wakeup_state = 0; |
|
|
pr = so->so_proto; |
pr = so->so_proto; |
atomic = pr->pr_flags & PR_ATOMIC; |
atomic = pr->pr_flags & PR_ATOMIC; |
Line 1011 soreceive(struct socket *so, struct mbuf |
|
Line 1158 soreceive(struct socket *so, struct mbuf |
|
else |
else |
flags = 0; |
flags = 0; |
|
|
if ((flags & MSG_DONTWAIT) == 0) |
|
sodopendfree(); |
|
|
|
if (flags & MSG_OOB) { |
if (flags & MSG_OOB) { |
m = m_get(M_WAIT, MT_DATA); |
m = m_get(M_WAIT, MT_DATA); |
|
solock(so); |
error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m, |
error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m, |
(struct mbuf *)(long)(flags & MSG_PEEK), NULL, l); |
(struct mbuf *)(long)(flags & MSG_PEEK), NULL, l); |
|
sounlock(so); |
if (error) |
if (error) |
goto bad; |
goto bad; |
do { |
do { |
error = uiomove(mtod(m, void *), |
error = uiomove(mtod(m, void *), |
(int) min(uio->uio_resid, m->m_len), uio); |
MIN(uio->uio_resid, m->m_len), uio); |
m = m_free(m); |
m = m_free(m); |
} while (uio->uio_resid > 0 && error == 0 && m); |
} while (uio->uio_resid > 0 && error == 0 && m); |
bad: |
bad: |
Line 1032 soreceive(struct socket *so, struct mbuf |
|
Line 1178 soreceive(struct socket *so, struct mbuf |
|
} |
} |
if (mp != NULL) |
if (mp != NULL) |
*mp = NULL; |
*mp = NULL; |
|
|
|
/* |
|
* solock() provides atomicity of access. splsoftnet() prevents |
|
* protocol processing soft interrupts from interrupting us and |
|
* blocking (expensive). |
|
*/ |
|
s = splsoftnet(); |
|
solock(so); |
if (so->so_state & SS_ISCONFIRMING && uio->uio_resid) |
if (so->so_state & SS_ISCONFIRMING && uio->uio_resid) |
(*pr->pr_usrreq)(so, PRU_RCVD, NULL, NULL, NULL, l); |
(*pr->pr_usrreq)(so, PRU_RCVD, NULL, NULL, NULL, l); |
|
|
restart: |
restart: |
if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0) |
if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0) { |
|
sounlock(so); |
|
splx(s); |
return error; |
return error; |
s = splsoftnet(); |
} |
|
|
m = so->so_rcv.sb_mb; |
m = so->so_rcv.sb_mb; |
/* |
/* |
Line 1089 soreceive(struct socket *so, struct mbuf |
|
Line 1245 soreceive(struct socket *so, struct mbuf |
|
} |
} |
if (uio->uio_resid == 0) |
if (uio->uio_resid == 0) |
goto release; |
goto release; |
if (so->so_nbio || (flags & MSG_DONTWAIT)) { |
if ((so->so_state & SS_NBIO) || |
|
(flags & (MSG_DONTWAIT|MSG_NBIO))) { |
error = EWOULDBLOCK; |
error = EWOULDBLOCK; |
goto release; |
goto release; |
} |
} |
SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 1"); |
SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 1"); |
SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 1"); |
SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 1"); |
sbunlock(&so->so_rcv); |
sbunlock(&so->so_rcv); |
error = sbwait(&so->so_rcv); |
if (wakeup_state & SS_RESTARTSYS) |
splx(s); |
error = ERESTART; |
if (error != 0) |
else |
|
error = sbwait(&so->so_rcv); |
|
if (error != 0) { |
|
sounlock(so); |
|
splx(s); |
return error; |
return error; |
|
} |
|
wakeup_state = so->so_state; |
goto restart; |
goto restart; |
} |
} |
dontblock: |
dontblock: |
/* |
/* |
* On entry here, m points to the first record of the socket buffer. |
* On entry here, m points to the first record of the socket buffer. |
* While we process the initial mbufs containing address and control |
* From this point onward, we maintain 'nextrecord' as a cache of the |
* info, we save a copy of m->m_nextpkt into nextrecord. |
* pointer to the next record in the socket buffer. We must keep the |
|
* various socket buffer pointers and local stack versions of the |
|
* pointers in sync, pushing out modifications before dropping the |
|
* socket lock, and re-reading them when picking it up. |
|
* |
|
* Otherwise, we will race with the network stack appending new data |
|
* or records onto the socket buffer by using inconsistent/stale |
|
* versions of the field, possibly resulting in socket buffer |
|
* corruption. |
|
* |
|
* By holding the high-level sblock(), we prevent simultaneous |
|
* readers from pulling off the front of the socket buffer. |
*/ |
*/ |
if (l != NULL) |
if (l != NULL) |
l->l_proc->p_stats->p_ru.ru_msgrcv++; |
l->l_ru.ru_msgrcv++; |
KASSERT(m == so->so_rcv.sb_mb); |
KASSERT(m == so->so_rcv.sb_mb); |
SBLASTRECORDCHK(&so->so_rcv, "soreceive 1"); |
SBLASTRECORDCHK(&so->so_rcv, "soreceive 1"); |
SBLASTMBUFCHK(&so->so_rcv, "soreceive 1"); |
SBLASTMBUFCHK(&so->so_rcv, "soreceive 1"); |
Line 1136 soreceive(struct socket *so, struct mbuf |
|
Line 1310 soreceive(struct socket *so, struct mbuf |
|
MFREE(m, so->so_rcv.sb_mb); |
MFREE(m, so->so_rcv.sb_mb); |
m = so->so_rcv.sb_mb; |
m = so->so_rcv.sb_mb; |
} |
} |
|
sbsync(&so->so_rcv, nextrecord); |
} |
} |
} |
} |
while (m != NULL && m->m_type == MT_CONTROL && error == 0) { |
|
if (flags & MSG_PEEK) { |
/* |
if (controlp != NULL) |
* Process one or more MT_CONTROL mbufs present before any data mbufs |
*controlp = m_copy(m, 0, m->m_len); |
* in the first mbuf chain on the socket buffer. If MSG_PEEK, we |
m = m->m_next; |
* just copy the data; if !MSG_PEEK, we call into the protocol to |
} else { |
* perform externalization (or freeing if controlp == NULL). |
sbfree(&so->so_rcv, m); |
*/ |
mbuf_removed = 1; |
if (__predict_false(m != NULL && m->m_type == MT_CONTROL)) { |
if (controlp != NULL) { |
struct mbuf *cm = NULL, *cmn; |
if (dom->dom_externalize && l && |
struct mbuf **cme = &cm; |
mtod(m, struct cmsghdr *)->cmsg_type == |
|
SCM_RIGHTS) |
do { |
error = (*dom->dom_externalize)(m, l); |
if (flags & MSG_PEEK) { |
*controlp = m; |
if (controlp != NULL) { |
|
*controlp = m_copy(m, 0, m->m_len); |
|
controlp = &(*controlp)->m_next; |
|
} |
|
m = m->m_next; |
|
} else { |
|
sbfree(&so->so_rcv, m); |
so->so_rcv.sb_mb = m->m_next; |
so->so_rcv.sb_mb = m->m_next; |
m->m_next = NULL; |
m->m_next = NULL; |
|
*cme = m; |
|
cme = &(*cme)->m_next; |
m = so->so_rcv.sb_mb; |
m = so->so_rcv.sb_mb; |
|
} |
|
} while (m != NULL && m->m_type == MT_CONTROL); |
|
if ((flags & MSG_PEEK) == 0) |
|
sbsync(&so->so_rcv, nextrecord); |
|
for (; cm != NULL; cm = cmn) { |
|
cmn = cm->m_next; |
|
cm->m_next = NULL; |
|
type = mtod(cm, struct cmsghdr *)->cmsg_type; |
|
if (controlp != NULL) { |
|
if (dom->dom_externalize != NULL && |
|
type == SCM_RIGHTS) { |
|
sounlock(so); |
|
splx(s); |
|
error = (*dom->dom_externalize)(cm, l, |
|
(flags & MSG_CMSG_CLOEXEC) ? |
|
O_CLOEXEC : 0); |
|
s = splsoftnet(); |
|
solock(so); |
|
} |
|
*controlp = cm; |
|
while (*controlp != NULL) |
|
controlp = &(*controlp)->m_next; |
} else { |
} else { |
/* |
/* |
* Dispose of any SCM_RIGHTS message that went |
* Dispose of any SCM_RIGHTS message that went |
* through the read path rather than recv. |
* through the read path rather than recv. |
*/ |
*/ |
if (dom->dom_dispose && |
if (dom->dom_dispose != NULL && |
mtod(m, struct cmsghdr *)->cmsg_type == SCM_RIGHTS) |
type == SCM_RIGHTS) { |
(*dom->dom_dispose)(m); |
sounlock(so); |
MFREE(m, so->so_rcv.sb_mb); |
(*dom->dom_dispose)(cm); |
m = so->so_rcv.sb_mb; |
solock(so); |
|
} |
|
m_freem(cm); |
} |
} |
} |
} |
if (controlp != NULL) { |
if (m != NULL) |
orig_resid = 0; |
nextrecord = so->so_rcv.sb_mb->m_nextpkt; |
controlp = &(*controlp)->m_next; |
else |
} |
nextrecord = so->so_rcv.sb_mb; |
|
orig_resid = 0; |
} |
} |
|
|
/* |
/* If m is non-NULL, we have some data to read. */ |
* If m is non-NULL, we have some data to read. From now on, |
if (__predict_true(m != NULL)) { |
* make sure to keep sb_lastrecord consistent when working on |
|
* the last packet on the chain (nextrecord == NULL) and we |
|
* change m->m_nextpkt. |
|
*/ |
|
if (m != NULL) { |
|
if ((flags & MSG_PEEK) == 0) { |
|
m->m_nextpkt = nextrecord; |
|
/* |
|
* If nextrecord == NULL (this is a single chain), |
|
* then sb_lastrecord may not be valid here if m |
|
* was changed earlier. |
|
*/ |
|
if (nextrecord == NULL) { |
|
KASSERT(so->so_rcv.sb_mb == m); |
|
so->so_rcv.sb_lastrecord = m; |
|
} |
|
} |
|
type = m->m_type; |
type = m->m_type; |
if (type == MT_OOBDATA) |
if (type == MT_OOBDATA) |
flags |= MSG_OOB; |
flags |= MSG_OOB; |
} else { |
|
if ((flags & MSG_PEEK) == 0) { |
|
KASSERT(so->so_rcv.sb_mb == m); |
|
so->so_rcv.sb_mb = nextrecord; |
|
SB_EMPTY_FIXUP(&so->so_rcv); |
|
} |
|
} |
} |
SBLASTRECORDCHK(&so->so_rcv, "soreceive 2"); |
SBLASTRECORDCHK(&so->so_rcv, "soreceive 2"); |
SBLASTMBUFCHK(&so->so_rcv, "soreceive 2"); |
SBLASTMBUFCHK(&so->so_rcv, "soreceive 2"); |
Line 1218 soreceive(struct socket *so, struct mbuf |
|
Line 1403 soreceive(struct socket *so, struct mbuf |
|
panic("receive 3"); |
panic("receive 3"); |
#endif |
#endif |
so->so_state &= ~SS_RCVATMARK; |
so->so_state &= ~SS_RCVATMARK; |
|
wakeup_state = 0; |
len = uio->uio_resid; |
len = uio->uio_resid; |
if (so->so_oobmark && len > so->so_oobmark - offset) |
if (so->so_oobmark && len > so->so_oobmark - offset) |
len = so->so_oobmark - offset; |
len = so->so_oobmark - offset; |
Line 1234 soreceive(struct socket *so, struct mbuf |
|
Line 1420 soreceive(struct socket *so, struct mbuf |
|
if (mp == NULL) { |
if (mp == NULL) { |
SBLASTRECORDCHK(&so->so_rcv, "soreceive uiomove"); |
SBLASTRECORDCHK(&so->so_rcv, "soreceive uiomove"); |
SBLASTMBUFCHK(&so->so_rcv, "soreceive uiomove"); |
SBLASTMBUFCHK(&so->so_rcv, "soreceive uiomove"); |
|
sounlock(so); |
splx(s); |
splx(s); |
error = uiomove(mtod(m, char *) + moff, (int)len, uio); |
error = uiomove(mtod(m, char *) + moff, len, uio); |
s = splsoftnet(); |
s = splsoftnet(); |
|
solock(so); |
if (error != 0) { |
if (error != 0) { |
/* |
/* |
* If any part of the record has been removed |
* If any part of the record has been removed |
Line 1293 soreceive(struct socket *so, struct mbuf |
|
Line 1481 soreceive(struct socket *so, struct mbuf |
|
} else if (flags & MSG_PEEK) |
} else if (flags & MSG_PEEK) |
moff += len; |
moff += len; |
else { |
else { |
if (mp != NULL) |
if (mp != NULL) { |
*mp = m_copym(m, 0, len, M_WAIT); |
mt = m_copym(m, 0, len, M_NOWAIT); |
|
if (__predict_false(mt == NULL)) { |
|
sounlock(so); |
|
mt = m_copym(m, 0, len, M_WAIT); |
|
solock(so); |
|
} |
|
*mp = mt; |
|
} |
m->m_data += len; |
m->m_data += len; |
m->m_len -= len; |
m->m_len -= len; |
so->so_rcv.sb_cc -= len; |
so->so_rcv.sb_cc -= len; |
Line 1341 soreceive(struct socket *so, struct mbuf |
|
Line 1536 soreceive(struct socket *so, struct mbuf |
|
NULL, (struct mbuf *)(long)flags, NULL, l); |
NULL, (struct mbuf *)(long)flags, NULL, l); |
SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 2"); |
SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 2"); |
SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 2"); |
SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 2"); |
error = sbwait(&so->so_rcv); |
if (wakeup_state & SS_RESTARTSYS) |
|
error = ERESTART; |
|
else |
|
error = sbwait(&so->so_rcv); |
if (error != 0) { |
if (error != 0) { |
sbunlock(&so->so_rcv); |
sbunlock(&so->so_rcv); |
|
sounlock(so); |
splx(s); |
splx(s); |
return 0; |
return 0; |
} |
} |
if ((m = so->so_rcv.sb_mb) != NULL) |
if ((m = so->so_rcv.sb_mb) != NULL) |
nextrecord = m->m_nextpkt; |
nextrecord = m->m_nextpkt; |
|
wakeup_state = so->so_state; |
} |
} |
} |
} |
|
|
Line 1380 soreceive(struct socket *so, struct mbuf |
|
Line 1580 soreceive(struct socket *so, struct mbuf |
|
if (orig_resid == uio->uio_resid && orig_resid && |
if (orig_resid == uio->uio_resid && orig_resid && |
(flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) { |
(flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) { |
sbunlock(&so->so_rcv); |
sbunlock(&so->so_rcv); |
splx(s); |
|
goto restart; |
goto restart; |
} |
} |
|
|
Line 1388 soreceive(struct socket *so, struct mbuf |
|
Line 1587 soreceive(struct socket *so, struct mbuf |
|
*flagsp |= flags; |
*flagsp |= flags; |
release: |
release: |
sbunlock(&so->so_rcv); |
sbunlock(&so->so_rcv); |
|
sounlock(so); |
splx(s); |
splx(s); |
return error; |
return error; |
} |
} |
|
|
soshutdown(struct socket *so, int how) |
soshutdown(struct socket *so, int how) |
{ |
{ |
const struct protosw *pr; |
const struct protosw *pr; |
|
int error; |
|
|
|
KASSERT(solocked(so)); |
|
|
pr = so->so_proto; |
pr = so->so_proto; |
if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR)) |
if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR)) |
return (EINVAL); |
return (EINVAL); |
|
|
if (how == SHUT_RD || how == SHUT_RDWR) |
if (how == SHUT_RD || how == SHUT_RDWR) { |
sorflush(so); |
sorflush(so); |
|
error = 0; |
|
} |
if (how == SHUT_WR || how == SHUT_RDWR) |
if (how == SHUT_WR || how == SHUT_RDWR) |
return (*pr->pr_usrreq)(so, PRU_SHUTDOWN, NULL, |
error = (*pr->pr_usrreq)(so, PRU_SHUTDOWN, NULL, |
NULL, NULL, NULL); |
NULL, NULL, NULL); |
return 0; |
|
|
return error; |
|
} |
|
|
|
void |
|
sorestart(struct socket *so) |
|
{ |
|
/* |
|
* An application has called close() on an fd on which another |
|
* of its threads has called a socket system call. |
|
* Mark this and wake everyone up, and code that would block again |
|
* instead returns ERESTART. |
|
* On system call re-entry the fd is validated and EBADF returned. |
|
* Any other fd will block again on the 2nd syscall. |
|
*/ |
|
solock(so); |
|
so->so_state |= SS_RESTARTSYS; |
|
cv_broadcast(&so->so_cv); |
|
cv_broadcast(&so->so_snd.sb_cv); |
|
cv_broadcast(&so->so_rcv.sb_cv); |
|
sounlock(so); |
} |
} |
|
|
void |
void |
Line 1414 sorflush(struct socket *so) |
|
Line 1639 sorflush(struct socket *so) |
|
{ |
{ |
struct sockbuf *sb, asb; |
struct sockbuf *sb, asb; |
const struct protosw *pr; |
const struct protosw *pr; |
int s; |
|
|
KASSERT(solocked(so)); |
|
|
sb = &so->so_rcv; |
sb = &so->so_rcv; |
pr = so->so_proto; |
pr = so->so_proto; |
sb->sb_flags |= SB_NOINTR; |
|
(void) sblock(sb, M_WAITOK); |
|
s = splnet(); |
|
socantrcvmore(so); |
socantrcvmore(so); |
|
sb->sb_flags |= SB_NOINTR; |
|
(void )sblock(sb, M_WAITOK); |
sbunlock(sb); |
sbunlock(sb); |
asb = *sb; |
asb = *sb; |
/* |
/* |
Line 1430 sorflush(struct socket *so) |
|
Line 1655 sorflush(struct socket *so) |
|
*/ |
*/ |
memset(&sb->sb_startzero, 0, |
memset(&sb->sb_startzero, 0, |
sizeof(*sb) - offsetof(struct sockbuf, sb_startzero)); |
sizeof(*sb) - offsetof(struct sockbuf, sb_startzero)); |
splx(s); |
if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose) { |
if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose) |
sounlock(so); |
(*pr->pr_domain->dom_dispose)(asb.sb_mb); |
(*pr->pr_domain->dom_dispose)(asb.sb_mb); |
|
solock(so); |
|
} |
sbrelease(&asb, so); |
sbrelease(&asb, so); |
} |
} |
|
|
|
/* |
|
* internal set SOL_SOCKET options |
|
*/ |
static int |
static int |
sosetopt1(struct socket *so, int level, int optname, struct mbuf *m) |
sosetopt1(struct socket *so, const struct sockopt *sopt) |
{ |
{ |
int optval, val; |
int error = EINVAL, optval, opt; |
struct linger *l; |
struct linger l; |
struct sockbuf *sb; |
struct timeval tv; |
struct timeval *tv; |
|
|
|
switch (optname) { |
switch ((opt = sopt->sopt_name)) { |
|
|
case SO_LINGER: |
case SO_ACCEPTFILTER: |
if (m == NULL || m->m_len != sizeof(struct linger)) |
error = accept_filt_setopt(so, sopt); |
return EINVAL; |
KASSERT(solocked(so)); |
l = mtod(m, struct linger *); |
|
if (l->l_linger < 0 || l->l_linger > USHRT_MAX || |
|
l->l_linger > (INT_MAX / hz)) |
|
return EDOM; |
|
so->so_linger = l->l_linger; |
|
if (l->l_onoff) |
|
so->so_options |= SO_LINGER; |
|
else |
|
so->so_options &= ~SO_LINGER; |
|
break; |
break; |
|
|
|
case SO_LINGER: |
|
error = sockopt_get(sopt, &l, sizeof(l)); |
|
solock(so); |
|
if (error) |
|
break; |
|
if (l.l_linger < 0 || l.l_linger > USHRT_MAX || |
|
l.l_linger > (INT_MAX / hz)) { |
|
error = EDOM; |
|
break; |
|
} |
|
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_DEBUG: |
case SO_KEEPALIVE: |
case SO_KEEPALIVE: |
case SO_DONTROUTE: |
case SO_DONTROUTE: |
Line 1469 sosetopt1(struct socket *so, int level, |
|
Line 1706 sosetopt1(struct socket *so, int level, |
|
case SO_REUSEPORT: |
case SO_REUSEPORT: |
case SO_OOBINLINE: |
case SO_OOBINLINE: |
case SO_TIMESTAMP: |
case SO_TIMESTAMP: |
if (m == NULL || m->m_len < sizeof(int)) |
case SO_NOSIGPIPE: |
return EINVAL; |
#ifdef SO_OTIMESTAMP |
if (*mtod(m, int *)) |
case SO_OTIMESTAMP: |
so->so_options |= optname; |
#endif |
|
error = sockopt_getint(sopt, &optval); |
|
solock(so); |
|
if (error) |
|
break; |
|
if (optval) |
|
so->so_options |= opt; |
else |
else |
so->so_options &= ~optname; |
so->so_options &= ~opt; |
break; |
break; |
|
|
case SO_SNDBUF: |
case SO_SNDBUF: |
case SO_RCVBUF: |
case SO_RCVBUF: |
case SO_SNDLOWAT: |
case SO_SNDLOWAT: |
case SO_RCVLOWAT: |
case SO_RCVLOWAT: |
if (m == NULL || m->m_len < sizeof(int)) |
error = sockopt_getint(sopt, &optval); |
return EINVAL; |
solock(so); |
|
if (error) |
|
break; |
|
|
/* |
/* |
* Values < 1 make no sense for any of these |
* Values < 1 make no sense for any of these |
* options, so disallow them. |
* options, so disallow them. |
*/ |
*/ |
optval = *mtod(m, int *); |
if (optval < 1) { |
if (optval < 1) |
error = EINVAL; |
return EINVAL; |
break; |
|
} |
switch (optname) { |
|
|
|
|
switch (opt) { |
case SO_SNDBUF: |
case SO_SNDBUF: |
|
if (sbreserve(&so->so_snd, (u_long)optval, so) == 0) { |
|
error = ENOBUFS; |
|
break; |
|
} |
|
so->so_snd.sb_flags &= ~SB_AUTOSIZE; |
|
break; |
|
|
case SO_RCVBUF: |
case SO_RCVBUF: |
sb = (optname == SO_SNDBUF) ? |
if (sbreserve(&so->so_rcv, (u_long)optval, so) == 0) { |
&so->so_snd : &so->so_rcv; |
error = ENOBUFS; |
if (sbreserve(sb, (u_long)optval, so) == 0) |
break; |
return ENOBUFS; |
} |
sb->sb_flags &= ~SB_AUTOSIZE; |
so->so_rcv.sb_flags &= ~SB_AUTOSIZE; |
break; |
break; |
|
|
/* |
/* |
Line 1508 sosetopt1(struct socket *so, int level, |
|
Line 1760 sosetopt1(struct socket *so, int level, |
|
* the high-water. |
* the high-water. |
*/ |
*/ |
case SO_SNDLOWAT: |
case SO_SNDLOWAT: |
so->so_snd.sb_lowat = |
if (optval > so->so_snd.sb_hiwat) |
(optval > so->so_snd.sb_hiwat) ? |
optval = so->so_snd.sb_hiwat; |
so->so_snd.sb_hiwat : optval; |
|
|
so->so_snd.sb_lowat = optval; |
break; |
break; |
|
|
case SO_RCVLOWAT: |
case SO_RCVLOWAT: |
so->so_rcv.sb_lowat = |
if (optval > so->so_rcv.sb_hiwat) |
(optval > so->so_rcv.sb_hiwat) ? |
optval = so->so_rcv.sb_hiwat; |
so->so_rcv.sb_hiwat : optval; |
|
|
so->so_rcv.sb_lowat = optval; |
break; |
break; |
} |
} |
break; |
break; |
|
|
|
#ifdef COMPAT_50 |
|
case SO_OSNDTIMEO: |
|
case SO_ORCVTIMEO: { |
|
struct timeval50 otv; |
|
error = sockopt_get(sopt, &otv, sizeof(otv)); |
|
if (error) { |
|
solock(so); |
|
break; |
|
} |
|
timeval50_to_timeval(&otv, &tv); |
|
opt = opt == SO_OSNDTIMEO ? SO_SNDTIMEO : SO_RCVTIMEO; |
|
error = 0; |
|
/*FALLTHROUGH*/ |
|
} |
|
#endif /* COMPAT_50 */ |
|
|
case SO_SNDTIMEO: |
case SO_SNDTIMEO: |
case SO_RCVTIMEO: |
case SO_RCVTIMEO: |
if (m == NULL || m->m_len < sizeof(*tv)) |
if (error) |
return EINVAL; |
error = sockopt_get(sopt, &tv, sizeof(tv)); |
tv = mtod(m, struct timeval *); |
solock(so); |
if (tv->tv_sec > (INT_MAX - tv->tv_usec / tick) / hz) |
if (error) |
return EDOM; |
break; |
val = tv->tv_sec * hz + tv->tv_usec / tick; |
|
if (val == 0 && tv->tv_usec != 0) |
|
val = 1; |
|
|
|
switch (optname) { |
if (tv.tv_sec > (INT_MAX - tv.tv_usec / tick) / hz) { |
|
error = EDOM; |
|
break; |
|
} |
|
|
|
optval = tv.tv_sec * hz + tv.tv_usec / tick; |
|
if (optval == 0 && tv.tv_usec != 0) |
|
optval = 1; |
|
|
|
switch (opt) { |
case SO_SNDTIMEO: |
case SO_SNDTIMEO: |
so->so_snd.sb_timeo = val; |
so->so_snd.sb_timeo = optval; |
break; |
break; |
case SO_RCVTIMEO: |
case SO_RCVTIMEO: |
so->so_rcv.sb_timeo = val; |
so->so_rcv.sb_timeo = optval; |
break; |
break; |
} |
} |
break; |
break; |
|
|
default: |
default: |
return ENOPROTOOPT; |
solock(so); |
|
error = ENOPROTOOPT; |
|
break; |
} |
} |
return 0; |
KASSERT(solocked(so)); |
|
return error; |
} |
} |
|
|
int |
int |
sosetopt(struct socket *so, int level, int optname, struct mbuf *m) |
sosetopt(struct socket *so, struct sockopt *sopt) |
{ |
{ |
int error, prerr; |
int error, prerr; |
|
|
if (level == SOL_SOCKET) |
if (sopt->sopt_level == SOL_SOCKET) { |
error = sosetopt1(so, level, optname, m); |
error = sosetopt1(so, sopt); |
else |
KASSERT(solocked(so)); |
|
} else { |
error = ENOPROTOOPT; |
error = ENOPROTOOPT; |
|
solock(so); |
|
} |
|
|
if ((error == 0 || error == ENOPROTOOPT) && |
if ((error == 0 || error == ENOPROTOOPT) && |
so->so_proto != NULL && so->so_proto->pr_ctloutput != NULL) { |
so->so_proto != NULL && so->so_proto->pr_ctloutput != NULL) { |
/* give the protocol stack a shot */ |
/* give the protocol stack a shot */ |
prerr = (*so->so_proto->pr_ctloutput)(PRCO_SETOPT, so, level, |
prerr = (*so->so_proto->pr_ctloutput)(PRCO_SETOPT, so, sopt); |
optname, &m); |
|
if (prerr == 0) |
if (prerr == 0) |
error = 0; |
error = 0; |
else if (prerr != ENOPROTOOPT) |
else if (prerr != ENOPROTOOPT) |
error = prerr; |
error = prerr; |
} else if (m != NULL) |
} |
(void)m_free(m); |
sounlock(so); |
return error; |
return error; |
} |
} |
|
|
|
/* |
|
* so_setsockopt() is a wrapper providing a sockopt structure for sosetopt() |
|
*/ |
int |
int |
sogetopt(struct socket *so, int level, int optname, struct mbuf **mp) |
so_setsockopt(struct lwp *l, struct socket *so, int level, int name, |
|
const void *val, size_t valsize) |
|
{ |
|
struct sockopt sopt; |
|
int error; |
|
|
|
KASSERT(valsize == 0 || val != NULL); |
|
|
|
sockopt_init(&sopt, level, name, valsize); |
|
sockopt_set(&sopt, val, valsize); |
|
|
|
error = sosetopt(so, &sopt); |
|
|
|
sockopt_destroy(&sopt); |
|
|
|
return error; |
|
} |
|
|
|
/* |
|
* internal get SOL_SOCKET options |
|
*/ |
|
static int |
|
sogetopt1(struct socket *so, struct sockopt *sopt) |
{ |
{ |
struct mbuf *m; |
int error, optval, opt; |
|
struct linger l; |
|
struct timeval tv; |
|
|
if (level != SOL_SOCKET) { |
switch ((opt = sopt->sopt_name)) { |
|
|
|
case SO_ACCEPTFILTER: |
|
error = accept_filt_getopt(so, sopt); |
|
break; |
|
|
|
case SO_LINGER: |
|
l.l_onoff = (so->so_options & SO_LINGER) ? 1 : 0; |
|
l.l_linger = so->so_linger; |
|
|
|
error = sockopt_set(sopt, &l, sizeof(l)); |
|
break; |
|
|
|
case SO_USELOOPBACK: |
|
case SO_DONTROUTE: |
|
case SO_DEBUG: |
|
case SO_KEEPALIVE: |
|
case SO_REUSEADDR: |
|
case SO_REUSEPORT: |
|
case SO_BROADCAST: |
|
case SO_OOBINLINE: |
|
case SO_TIMESTAMP: |
|
case SO_NOSIGPIPE: |
|
#ifdef SO_OTIMESTAMP |
|
case SO_OTIMESTAMP: |
|
#endif |
|
error = sockopt_setint(sopt, (so->so_options & opt) ? 1 : 0); |
|
break; |
|
|
|
case SO_TYPE: |
|
error = sockopt_setint(sopt, so->so_type); |
|
break; |
|
|
|
case SO_ERROR: |
|
error = sockopt_setint(sopt, so->so_error); |
|
so->so_error = 0; |
|
break; |
|
|
|
case SO_SNDBUF: |
|
error = sockopt_setint(sopt, so->so_snd.sb_hiwat); |
|
break; |
|
|
|
case SO_RCVBUF: |
|
error = sockopt_setint(sopt, so->so_rcv.sb_hiwat); |
|
break; |
|
|
|
case SO_SNDLOWAT: |
|
error = sockopt_setint(sopt, so->so_snd.sb_lowat); |
|
break; |
|
|
|
case SO_RCVLOWAT: |
|
error = sockopt_setint(sopt, so->so_rcv.sb_lowat); |
|
break; |
|
|
|
#ifdef COMPAT_50 |
|
case SO_OSNDTIMEO: |
|
case SO_ORCVTIMEO: { |
|
struct timeval50 otv; |
|
|
|
optval = (opt == SO_OSNDTIMEO ? |
|
so->so_snd.sb_timeo : so->so_rcv.sb_timeo); |
|
|
|
otv.tv_sec = optval / hz; |
|
otv.tv_usec = (optval % hz) * tick; |
|
|
|
error = sockopt_set(sopt, &otv, sizeof(otv)); |
|
break; |
|
} |
|
#endif /* COMPAT_50 */ |
|
|
|
case SO_SNDTIMEO: |
|
case SO_RCVTIMEO: |
|
optval = (opt == SO_SNDTIMEO ? |
|
so->so_snd.sb_timeo : so->so_rcv.sb_timeo); |
|
|
|
tv.tv_sec = optval / hz; |
|
tv.tv_usec = (optval % hz) * tick; |
|
|
|
error = sockopt_set(sopt, &tv, sizeof(tv)); |
|
break; |
|
|
|
case SO_OVERFLOWED: |
|
error = sockopt_setint(sopt, so->so_rcv.sb_overflowed); |
|
break; |
|
|
|
default: |
|
error = ENOPROTOOPT; |
|
break; |
|
} |
|
|
|
return (error); |
|
} |
|
|
|
int |
|
sogetopt(struct socket *so, struct sockopt *sopt) |
|
{ |
|
int error; |
|
|
|
solock(so); |
|
if (sopt->sopt_level != SOL_SOCKET) { |
if (so->so_proto && so->so_proto->pr_ctloutput) { |
if (so->so_proto && so->so_proto->pr_ctloutput) { |
return ((*so->so_proto->pr_ctloutput) |
error = ((*so->so_proto->pr_ctloutput) |
(PRCO_GETOPT, so, level, optname, mp)); |
(PRCO_GETOPT, so, sopt)); |
} else |
} else |
return (ENOPROTOOPT); |
error = (ENOPROTOOPT); |
} else { |
} else { |
m = m_get(M_WAIT, MT_SOOPTS); |
error = sogetopt1(so, sopt); |
m->m_len = sizeof(int); |
} |
|
sounlock(so); |
|
return (error); |
|
} |
|
|
switch (optname) { |
/* |
|
* alloc sockopt data buffer buffer |
|
* - will be released at destroy |
|
*/ |
|
static int |
|
sockopt_alloc(struct sockopt *sopt, size_t len, km_flag_t kmflag) |
|
{ |
|
|
case SO_LINGER: |
KASSERT(sopt->sopt_size == 0); |
m->m_len = sizeof(struct linger); |
|
mtod(m, struct linger *)->l_onoff = |
|
(so->so_options & SO_LINGER) ? 1 : 0; |
|
mtod(m, struct linger *)->l_linger = so->so_linger; |
|
break; |
|
|
|
case SO_USELOOPBACK: |
if (len > sizeof(sopt->sopt_buf)) { |
case SO_DONTROUTE: |
sopt->sopt_data = kmem_zalloc(len, kmflag); |
case SO_DEBUG: |
if (sopt->sopt_data == NULL) |
case SO_KEEPALIVE: |
return ENOMEM; |
case SO_REUSEADDR: |
} else |
case SO_REUSEPORT: |
sopt->sopt_data = sopt->sopt_buf; |
case SO_BROADCAST: |
|
case SO_OOBINLINE: |
|
case SO_TIMESTAMP: |
|
*mtod(m, int *) = (so->so_options & optname) ? 1 : 0; |
|
break; |
|
|
|
case SO_TYPE: |
sopt->sopt_size = len; |
*mtod(m, int *) = so->so_type; |
return 0; |
break; |
} |
|
|
case SO_ERROR: |
/* |
*mtod(m, int *) = so->so_error; |
* initialise sockopt storage |
so->so_error = 0; |
* - MAY sleep during allocation |
break; |
*/ |
|
void |
|
sockopt_init(struct sockopt *sopt, int level, int name, size_t size) |
|
{ |
|
|
case SO_SNDBUF: |
memset(sopt, 0, sizeof(*sopt)); |
*mtod(m, int *) = so->so_snd.sb_hiwat; |
|
break; |
|
|
|
case SO_RCVBUF: |
sopt->sopt_level = level; |
*mtod(m, int *) = so->so_rcv.sb_hiwat; |
sopt->sopt_name = name; |
break; |
(void)sockopt_alloc(sopt, size, KM_SLEEP); |
|
} |
|
|
case SO_SNDLOWAT: |
/* |
*mtod(m, int *) = so->so_snd.sb_lowat; |
* destroy sockopt storage |
break; |
* - will release any held memory references |
|
*/ |
|
void |
|
sockopt_destroy(struct sockopt *sopt) |
|
{ |
|
|
case SO_RCVLOWAT: |
if (sopt->sopt_data != sopt->sopt_buf) |
*mtod(m, int *) = so->so_rcv.sb_lowat; |
kmem_free(sopt->sopt_data, sopt->sopt_size); |
break; |
|
|
|
case SO_SNDTIMEO: |
memset(sopt, 0, sizeof(*sopt)); |
case SO_RCVTIMEO: |
} |
{ |
|
int val = (optname == SO_SNDTIMEO ? |
|
so->so_snd.sb_timeo : so->so_rcv.sb_timeo); |
|
|
|
m->m_len = sizeof(struct timeval); |
|
mtod(m, struct timeval *)->tv_sec = val / hz; |
|
mtod(m, struct timeval *)->tv_usec = |
|
(val % hz) * tick; |
|
break; |
|
} |
|
|
|
case SO_OVERFLOWED: |
/* |
*mtod(m, int *) = so->so_rcv.sb_overflowed; |
* set sockopt value |
break; |
* - value is copied into sockopt |
|
* - memory is allocated when necessary, will not sleep |
|
*/ |
|
int |
|
sockopt_set(struct sockopt *sopt, const void *buf, size_t len) |
|
{ |
|
int error; |
|
|
default: |
if (sopt->sopt_size == 0) { |
(void)m_free(m); |
error = sockopt_alloc(sopt, len, KM_NOSLEEP); |
return (ENOPROTOOPT); |
if (error) |
|
return error; |
|
} |
|
|
|
KASSERT(sopt->sopt_size == len); |
|
memcpy(sopt->sopt_data, buf, len); |
|
return 0; |
|
} |
|
|
|
/* |
|
* common case of set sockopt integer value |
|
*/ |
|
int |
|
sockopt_setint(struct sockopt *sopt, int val) |
|
{ |
|
|
|
return sockopt_set(sopt, &val, sizeof(int)); |
|
} |
|
|
|
/* |
|
* get sockopt value |
|
* - correct size must be given |
|
*/ |
|
int |
|
sockopt_get(const struct sockopt *sopt, void *buf, size_t len) |
|
{ |
|
|
|
if (sopt->sopt_size != len) |
|
return EINVAL; |
|
|
|
memcpy(buf, sopt->sopt_data, len); |
|
return 0; |
|
} |
|
|
|
/* |
|
* common case of get sockopt integer value |
|
*/ |
|
int |
|
sockopt_getint(const struct sockopt *sopt, int *valp) |
|
{ |
|
|
|
return sockopt_get(sopt, valp, sizeof(int)); |
|
} |
|
|
|
/* |
|
* set sockopt value from mbuf |
|
* - ONLY for legacy code |
|
* - mbuf is released by sockopt |
|
* - will not sleep |
|
*/ |
|
int |
|
sockopt_setmbuf(struct sockopt *sopt, struct mbuf *m) |
|
{ |
|
size_t len; |
|
int error; |
|
|
|
len = m_length(m); |
|
|
|
if (sopt->sopt_size == 0) { |
|
error = sockopt_alloc(sopt, len, KM_NOSLEEP); |
|
if (error) |
|
return error; |
|
} |
|
|
|
KASSERT(sopt->sopt_size == len); |
|
m_copydata(m, 0, len, sopt->sopt_data); |
|
m_freem(m); |
|
|
|
return 0; |
|
} |
|
|
|
/* |
|
* get sockopt value into mbuf |
|
* - ONLY for legacy code |
|
* - mbuf to be released by the caller |
|
* - will not sleep |
|
*/ |
|
struct mbuf * |
|
sockopt_getmbuf(const struct sockopt *sopt) |
|
{ |
|
struct mbuf *m; |
|
|
|
if (sopt->sopt_size > MCLBYTES) |
|
return NULL; |
|
|
|
m = m_get(M_DONTWAIT, MT_SOOPTS); |
|
if (m == NULL) |
|
return NULL; |
|
|
|
if (sopt->sopt_size > MLEN) { |
|
MCLGET(m, M_DONTWAIT); |
|
if ((m->m_flags & M_EXT) == 0) { |
|
m_free(m); |
|
return NULL; |
} |
} |
*mp = m; |
|
return (0); |
|
} |
} |
|
|
|
memcpy(mtod(m, void *), sopt->sopt_data, sopt->sopt_size); |
|
m->m_len = sopt->sopt_size; |
|
|
|
return m; |
} |
} |
|
|
void |
void |
Line 1664 sohasoutofband(struct socket *so) |
|
Line 2167 sohasoutofband(struct socket *so) |
|
{ |
{ |
|
|
fownsignal(so->so_pgid, SIGURG, POLL_PRI, POLLPRI|POLLRDBAND, so); |
fownsignal(so->so_pgid, SIGURG, POLL_PRI, POLLPRI|POLLRDBAND, so); |
selnotify(&so->so_rcv.sb_sel, POLLPRI | POLLRDBAND, 0); |
selnotify(&so->so_rcv.sb_sel, POLLPRI | POLLRDBAND, NOTE_SUBMIT); |
} |
} |
|
|
static void |
static void |
Line 1673 filt_sordetach(struct knote *kn) |
|
Line 2176 filt_sordetach(struct knote *kn) |
|
struct socket *so; |
struct socket *so; |
|
|
so = ((file_t *)kn->kn_obj)->f_data; |
so = ((file_t *)kn->kn_obj)->f_data; |
|
solock(so); |
SLIST_REMOVE(&so->so_rcv.sb_sel.sel_klist, kn, knote, kn_selnext); |
SLIST_REMOVE(&so->so_rcv.sb_sel.sel_klist, kn, knote, kn_selnext); |
if (SLIST_EMPTY(&so->so_rcv.sb_sel.sel_klist)) |
if (SLIST_EMPTY(&so->so_rcv.sb_sel.sel_klist)) |
so->so_rcv.sb_flags &= ~SB_KNOTE; |
so->so_rcv.sb_flags &= ~SB_KNOTE; |
|
sounlock(so); |
} |
} |
|
|
/*ARGSUSED*/ |
/*ARGSUSED*/ |
|
|
filt_soread(struct knote *kn, long hint) |
filt_soread(struct knote *kn, long hint) |
{ |
{ |
struct socket *so; |
struct socket *so; |
|
int rv; |
|
|
so = ((file_t *)kn->kn_obj)->f_data; |
so = ((file_t *)kn->kn_obj)->f_data; |
|
if (hint != NOTE_SUBMIT) |
|
solock(so); |
kn->kn_data = so->so_rcv.sb_cc; |
kn->kn_data = so->so_rcv.sb_cc; |
if (so->so_state & SS_CANTRCVMORE) { |
if (so->so_state & SS_CANTRCVMORE) { |
kn->kn_flags |= EV_EOF; |
kn->kn_flags |= EV_EOF; |
kn->kn_fflags = so->so_error; |
kn->kn_fflags = so->so_error; |
return (1); |
rv = 1; |
} |
} else if (so->so_error) /* temporary udp error */ |
if (so->so_error) /* temporary udp error */ |
rv = 1; |
return (1); |
else if (kn->kn_sfflags & NOTE_LOWAT) |
if (kn->kn_sfflags & NOTE_LOWAT) |
rv = (kn->kn_data >= kn->kn_sdata); |
return (kn->kn_data >= kn->kn_sdata); |
else |
return (kn->kn_data >= so->so_rcv.sb_lowat); |
rv = (kn->kn_data >= so->so_rcv.sb_lowat); |
|
if (hint != NOTE_SUBMIT) |
|
sounlock(so); |
|
return rv; |
} |
} |
|
|
static void |
static void |
Line 1704 filt_sowdetach(struct knote *kn) |
|
Line 2215 filt_sowdetach(struct knote *kn) |
|
struct socket *so; |
struct socket *so; |
|
|
so = ((file_t *)kn->kn_obj)->f_data; |
so = ((file_t *)kn->kn_obj)->f_data; |
|
solock(so); |
SLIST_REMOVE(&so->so_snd.sb_sel.sel_klist, kn, knote, kn_selnext); |
SLIST_REMOVE(&so->so_snd.sb_sel.sel_klist, kn, knote, kn_selnext); |
if (SLIST_EMPTY(&so->so_snd.sb_sel.sel_klist)) |
if (SLIST_EMPTY(&so->so_snd.sb_sel.sel_klist)) |
so->so_snd.sb_flags &= ~SB_KNOTE; |
so->so_snd.sb_flags &= ~SB_KNOTE; |
|
sounlock(so); |
} |
} |
|
|
/*ARGSUSED*/ |
/*ARGSUSED*/ |
|
|
filt_sowrite(struct knote *kn, long hint) |
filt_sowrite(struct knote *kn, long hint) |
{ |
{ |
struct socket *so; |
struct socket *so; |
|
int rv; |
|
|
so = ((file_t *)kn->kn_obj)->f_data; |
so = ((file_t *)kn->kn_obj)->f_data; |
|
if (hint != NOTE_SUBMIT) |
|
solock(so); |
kn->kn_data = sbspace(&so->so_snd); |
kn->kn_data = sbspace(&so->so_snd); |
if (so->so_state & SS_CANTSENDMORE) { |
if (so->so_state & SS_CANTSENDMORE) { |
kn->kn_flags |= EV_EOF; |
kn->kn_flags |= EV_EOF; |
kn->kn_fflags = so->so_error; |
kn->kn_fflags = so->so_error; |
return (1); |
rv = 1; |
} |
} else if (so->so_error) /* temporary udp error */ |
if (so->so_error) /* temporary udp error */ |
rv = 1; |
return (1); |
else if (((so->so_state & SS_ISCONNECTED) == 0) && |
if (((so->so_state & SS_ISCONNECTED) == 0) && |
|
(so->so_proto->pr_flags & PR_CONNREQUIRED)) |
(so->so_proto->pr_flags & PR_CONNREQUIRED)) |
return (0); |
rv = 0; |
if (kn->kn_sfflags & NOTE_LOWAT) |
else if (kn->kn_sfflags & NOTE_LOWAT) |
return (kn->kn_data >= kn->kn_sdata); |
rv = (kn->kn_data >= kn->kn_sdata); |
return (kn->kn_data >= so->so_snd.sb_lowat); |
else |
|
rv = (kn->kn_data >= so->so_snd.sb_lowat); |
|
if (hint != NOTE_SUBMIT) |
|
sounlock(so); |
|
return rv; |
} |
} |
|
|
/*ARGSUSED*/ |
/*ARGSUSED*/ |
|
|
filt_solisten(struct knote *kn, long hint) |
filt_solisten(struct knote *kn, long hint) |
{ |
{ |
struct socket *so; |
struct socket *so; |
|
int rv; |
|
|
so = ((file_t *)kn->kn_obj)->f_data; |
so = ((file_t *)kn->kn_obj)->f_data; |
|
|
Line 1744 filt_solisten(struct knote *kn, long hin |
|
Line 2264 filt_solisten(struct knote *kn, long hin |
|
* Set kn_data to number of incoming connections, not |
* Set kn_data to number of incoming connections, not |
* counting partial (incomplete) connections. |
* counting partial (incomplete) connections. |
*/ |
*/ |
|
if (hint != NOTE_SUBMIT) |
|
solock(so); |
kn->kn_data = so->so_qlen; |
kn->kn_data = so->so_qlen; |
return (kn->kn_data > 0); |
rv = (kn->kn_data > 0); |
|
if (hint != NOTE_SUBMIT) |
|
sounlock(so); |
|
return rv; |
} |
} |
|
|
static const struct filterops solisten_filtops = |
static const struct filterops solisten_filtops = |
Line 1762 soo_kqfilter(struct file *fp, struct kno |
|
Line 2287 soo_kqfilter(struct file *fp, struct kno |
|
struct sockbuf *sb; |
struct sockbuf *sb; |
|
|
so = ((file_t *)kn->kn_obj)->f_data; |
so = ((file_t *)kn->kn_obj)->f_data; |
|
solock(so); |
switch (kn->kn_filter) { |
switch (kn->kn_filter) { |
case EVFILT_READ: |
case EVFILT_READ: |
if (so->so_options & SO_ACCEPTCONN) |
if (so->so_options & SO_ACCEPTCONN) |
Line 1775 soo_kqfilter(struct file *fp, struct kno |
|
Line 2301 soo_kqfilter(struct file *fp, struct kno |
|
sb = &so->so_snd; |
sb = &so->so_snd; |
break; |
break; |
default: |
default: |
|
sounlock(so); |
return (EINVAL); |
return (EINVAL); |
} |
} |
SLIST_INSERT_HEAD(&sb->sb_sel.sel_klist, kn, kn_selnext); |
SLIST_INSERT_HEAD(&sb->sb_sel.sel_klist, kn, kn_selnext); |
sb->sb_flags |= SB_KNOTE; |
sb->sb_flags |= SB_KNOTE; |
|
sounlock(so); |
return (0); |
return (0); |
} |
} |
|
|
|
|
sopoll(struct socket *so, int events) |
sopoll(struct socket *so, int events) |
{ |
{ |
int revents = 0; |
int revents = 0; |
int s; |
|
|
|
|
#ifndef DIAGNOSTIC |
|
/* |
|
* Do a quick, unlocked check in expectation that the socket |
|
* will be ready for I/O. Don't do this check if DIAGNOSTIC, |
|
* as the solocked() assertions will fail. |
|
*/ |
if ((revents = sodopoll(so, events)) != 0) |
if ((revents = sodopoll(so, events)) != 0) |
return revents; |
return revents; |
|
#endif |
|
|
KERNEL_LOCK(1, curlwp); |
solock(so); |
s = splsoftnet(); |
|
|
|
if ((revents = sodopoll(so, events)) == 0) { |
if ((revents = sodopoll(so, events)) == 0) { |
if (events & (POLLIN | POLLPRI | POLLRDNORM | POLLRDBAND)) { |
if (events & (POLLIN | POLLPRI | POLLRDNORM | POLLRDBAND)) { |
selrecord(curlwp, &so->so_rcv.sb_sel); |
selrecord(curlwp, &so->so_rcv.sb_sel); |
so->so_rcv.sb_flags |= SB_SEL; |
so->so_rcv.sb_flags |= SB_NOTIFY; |
} |
} |
|
|
if (events & (POLLOUT | POLLWRNORM)) { |
if (events & (POLLOUT | POLLWRNORM)) { |
selrecord(curlwp, &so->so_snd.sb_sel); |
selrecord(curlwp, &so->so_snd.sb_sel); |
so->so_snd.sb_flags |= SB_SEL; |
so->so_snd.sb_flags |= SB_NOTIFY; |
} |
} |
} |
} |
|
sounlock(so); |
splx(s); |
|
KERNEL_UNLOCK_ONE(curlwp); |
|
|
|
return revents; |
return revents; |
} |
} |
Line 1868 sysctl_kern_somaxkva(SYSCTLFN_ARGS) |
|
Line 2398 sysctl_kern_somaxkva(SYSCTLFN_ARGS) |
|
return (error); |
return (error); |
} |
} |
|
|
SYSCTL_SETUP(sysctl_kern_somaxkva_setup, "sysctl kern.somaxkva setup") |
static void |
|
sysctl_kern_somaxkva_setup(void) |
{ |
{ |
|
|
sysctl_createv(clog, 0, NULL, NULL, |
KASSERT(socket_sysctllog == NULL); |
|
sysctl_createv(&socket_sysctllog, 0, NULL, NULL, |
CTLFLAG_PERMANENT, |
CTLFLAG_PERMANENT, |
CTLTYPE_NODE, "kern", NULL, |
CTLTYPE_NODE, "kern", NULL, |
NULL, 0, NULL, 0, |
NULL, 0, NULL, 0, |
CTL_KERN, CTL_EOL); |
CTL_KERN, CTL_EOL); |
|
|
sysctl_createv(clog, 0, NULL, NULL, |
sysctl_createv(&socket_sysctllog, 0, NULL, NULL, |
CTLFLAG_PERMANENT|CTLFLAG_READWRITE, |
CTLFLAG_PERMANENT|CTLFLAG_READWRITE, |
CTLTYPE_INT, "somaxkva", |
CTLTYPE_INT, "somaxkva", |
SYSCTL_DESCR("Maximum amount of kernel memory to be " |
SYSCTL_DESCR("Maximum amount of kernel memory to be " |