freebsd-dev/sys/kern/uipc_sockbuf.c

1138 lines
28 KiB
C
Raw Normal View History

1994-05-24 10:09:53 +00:00
/*
* Copyright (c) 1982, 1986, 1988, 1990, 1993
* The Regents of the University of California. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 4. Neither the name of the University 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 REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* @(#)uipc_socket2.c 8.1 (Berkeley) 6/10/93
*/
2003-06-11 00:56:59 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_mac.h"
#include "opt_param.h"
1994-05-24 10:09:53 +00:00
#include <sys/param.h>
#include <sys/aio.h> /* for aio_swake proto */
#include <sys/domain.h>
#include <sys/event.h>
#include <sys/file.h> /* for maxfiles */
#include <sys/kernel.h>
#include <sys/lock.h>
#include <sys/mac.h>
#include <sys/malloc.h>
1994-05-24 10:09:53 +00:00
#include <sys/mbuf.h>
#include <sys/mutex.h>
#include <sys/proc.h>
1994-05-24 10:09:53 +00:00
#include <sys/protosw.h>
#include <sys/resourcevar.h>
#include <sys/signalvar.h>
1994-05-24 10:09:53 +00:00
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <sys/stat.h>
#include <sys/sysctl.h>
#include <sys/systm.h>
int maxsockets;
void (*aio_swake)(struct socket *, struct sockbuf *);
1994-05-24 10:09:53 +00:00
/*
* Primitive routines for operating on sockets and socket buffers
*/
u_long sb_max = SB_MAX;
static u_long sb_max_adj =
SB_MAX * MCLBYTES / (MSIZE + MCLBYTES); /* adjusted sb_max */
1994-05-24 10:09:53 +00:00
static u_long sb_efficiency = 8; /* parameter for sbreserve() */
1994-05-24 10:09:53 +00:00
/*
* Procedures to manipulate state flags of socket
* and do appropriate wakeups. Normal sequence from the
* active (originating) side is that soisconnecting() is
* called during processing of connect() call,
* resulting in an eventual call to soisconnected() if/when the
* connection is established. When the connection is torn down
* soisdisconnecting() is called during processing of disconnect() call,
* and soisdisconnected() is called when the connection to the peer
* is totally severed. The semantics of these routines are such that
* connectionless protocols can call soisconnected() and soisdisconnected()
* only, bypassing the in-progress calls when setting up a ``connection''
* takes no time.
*
* From the passive side, a socket is created with
* two queues of sockets: so_incomp for connections in progress
* and so_comp for connections already made and awaiting user acceptance.
1994-05-24 10:09:53 +00:00
* As a protocol is preparing incoming connections, it creates a socket
* structure queued on so_incomp by calling sonewconn(). When the connection
1994-05-24 10:09:53 +00:00
* is established, soisconnected() is called, and transfers the
* socket structure to so_comp, making it available to accept().
1995-05-30 08:16:23 +00:00
*
1994-05-24 10:09:53 +00:00
* If a socket is closed with sockets on either
* so_incomp or so_comp, these sockets are dropped.
1994-05-24 10:09:53 +00:00
*
* If higher level protocols are implemented in
* the kernel, the wakeups done here will sometimes
* cause software-interrupt process scheduling.
*/
void
1994-05-24 10:09:53 +00:00
soisconnecting(so)
register struct socket *so;
{
so->so_state &= ~(SS_ISCONNECTED|SS_ISDISCONNECTING);
so->so_state |= SS_ISCONNECTING;
}
void
soisconnected(so)
struct socket *so;
{
struct socket *head = so->so_head;
so->so_state &= ~(SS_ISCONNECTING|SS_ISDISCONNECTING|SS_ISCONFIRMING);
so->so_state |= SS_ISCONNECTED;
if (head && (so->so_state & SS_INCOMP)) {
if ((so->so_options & SO_ACCEPTFILTER) != 0) {
so->so_upcall = head->so_accf->so_accept_filter->accf_callback;
so->so_upcallarg = head->so_accf->so_accept_filter_arg;
so->so_rcv.sb_flags |= SB_UPCALL;
so->so_options &= ~SO_ACCEPTFILTER;
so->so_upcall(so, so->so_upcallarg, M_TRYWAIT);
return;
}
TAILQ_REMOVE(&head->so_incomp, so, so_list);
head->so_incqlen--;
so->so_state &= ~SS_INCOMP;
TAILQ_INSERT_TAIL(&head->so_comp, so, so_list);
head->so_qlen++;
so->so_state |= SS_COMP;
sorwakeup(head);
wakeup_one(&head->so_timeo);
1994-05-24 10:09:53 +00:00
} else {
wakeup(&so->so_timeo);
sorwakeup(so);
sowwakeup(so);
1994-05-24 10:09:53 +00:00
}
}
void
1994-05-24 10:09:53 +00:00
soisdisconnecting(so)
register struct socket *so;
{
so->so_state &= ~SS_ISCONNECTING;
so->so_state |= (SS_ISDISCONNECTING|SS_CANTRCVMORE|SS_CANTSENDMORE);
wakeup(&so->so_timeo);
sowwakeup(so);
sorwakeup(so);
1994-05-24 10:09:53 +00:00
}
void
soisdisconnected(so)
1994-05-24 10:09:53 +00:00
register struct socket *so;
{
so->so_state &= ~(SS_ISCONNECTING|SS_ISCONNECTED|SS_ISDISCONNECTING);
so->so_state |= (SS_CANTRCVMORE|SS_CANTSENDMORE|SS_ISDISCONNECTED);
wakeup(&so->so_timeo);
sbdrop(&so->so_snd, so->so_snd.sb_cc);
sowwakeup(so);
sorwakeup(so);
1994-05-24 10:09:53 +00:00
}
/*
* When an attempt at a new connection is noted on a socket
* which accepts connections, sonewconn is called. If the
* connection is possible (subject to space constraints, etc.)
* then we allocate a new structure, propoerly linked into the
* data structure of the original socket, and return this.
* Connstatus may be 0, or SO_ISCONFIRMING, or SO_ISCONNECTED.
*
* note: the ref count on the socket is 0 on return
1994-05-24 10:09:53 +00:00
*/
struct socket *
sonewconn(head, connstatus)
1994-05-24 10:09:53 +00:00
register struct socket *head;
int connstatus;
{
register struct socket *so;
if (head->so_qlen > 3 * head->so_qlimit / 2)
1994-05-24 10:09:53 +00:00
return ((struct socket *)0);
so = soalloc(M_NOWAIT);
1995-05-30 08:16:23 +00:00
if (so == NULL)
1994-05-24 10:09:53 +00:00
return ((struct socket *)0);
if ((head->so_options & SO_ACCEPTFILTER) != 0)
connstatus = 0;
so->so_head = head;
1994-05-24 10:09:53 +00:00
so->so_type = head->so_type;
so->so_options = head->so_options &~ SO_ACCEPTCONN;
so->so_linger = head->so_linger;
so->so_state = head->so_state | SS_NOFDREF;
so->so_proto = head->so_proto;
so->so_timeo = head->so_timeo;
so->so_cred = crhold(head->so_cred);
#ifdef MAC
mac_create_socket_from_socket(head, so);
#endif
if (soreserve(so, head->so_snd.sb_hiwat, head->so_rcv.sb_hiwat) ||
(*so->so_proto->pr_usrreqs->pru_attach)(so, 0, NULL)) {
sodealloc(so);
return ((struct socket *)0);
}
if (connstatus) {
TAILQ_INSERT_TAIL(&head->so_comp, so, so_list);
so->so_state |= SS_COMP;
head->so_qlen++;
} else {
if (head->so_incqlen > head->so_qlimit) {
struct socket *sp;
sp = TAILQ_FIRST(&head->so_incomp);
(void) soabort(sp);
}
TAILQ_INSERT_TAIL(&head->so_incomp, so, so_list);
so->so_state |= SS_INCOMP;
head->so_incqlen++;
}
1994-05-24 10:09:53 +00:00
if (connstatus) {
sorwakeup(head);
wakeup_one(&head->so_timeo);
so->so_state |= connstatus;
1994-05-24 10:09:53 +00:00
}
return (so);
}
/*
* Socantsendmore indicates that no more data will be sent on the
* socket; it would normally be applied to a socket when the user
* informs the system that no more data is to be sent, by the protocol
* code (in case PRU_SHUTDOWN). Socantrcvmore indicates that no more data
* will be received, and will normally be applied to the socket by a
* protocol when it detects that the peer will send no more data.
* Data queued for reading in the socket may yet be read.
*/
void
1994-05-24 10:09:53 +00:00
socantsendmore(so)
struct socket *so;
{
so->so_state |= SS_CANTSENDMORE;
sowwakeup(so);
1994-05-24 10:09:53 +00:00
}
void
1994-05-24 10:09:53 +00:00
socantrcvmore(so)
struct socket *so;
{
so->so_state |= SS_CANTRCVMORE;
sorwakeup(so);
1994-05-24 10:09:53 +00:00
}
/*
* Wait for data to arrive at/drain from a socket buffer.
*/
int
1994-05-24 10:09:53 +00:00
sbwait(sb)
struct sockbuf *sb;
{
sb->sb_flags |= SB_WAIT;
return (tsleep(&sb->sb_cc,
(sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK | PCATCH, "sbwait",
1994-05-24 10:09:53 +00:00
sb->sb_timeo));
}
1995-05-30 08:16:23 +00:00
/*
1994-05-24 10:09:53 +00:00
* Lock a sockbuf already known to be locked;
* return any error returned from sleep (EINTR).
*/
int
1994-05-24 10:09:53 +00:00
sb_lock(sb)
register struct sockbuf *sb;
{
int error;
while (sb->sb_flags & SB_LOCK) {
sb->sb_flags |= SB_WANT;
error = tsleep(&sb->sb_flags,
1994-05-24 10:09:53 +00:00
(sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK|PCATCH,
"sblock", 0);
if (error)
1994-05-24 10:09:53 +00:00
return (error);
}
sb->sb_flags |= SB_LOCK;
return (0);
}
/*
* Wakeup processes waiting on a socket buffer.
* Do asynchronous notification via SIGIO
* if the socket has the SS_ASYNC flag set.
*/
void
1994-05-24 10:09:53 +00:00
sowakeup(so, sb)
register struct socket *so;
register struct sockbuf *sb;
{
selwakeuppri(&sb->sb_sel, PSOCK);
1994-05-24 10:09:53 +00:00
sb->sb_flags &= ~SB_SEL;
if (sb->sb_flags & SB_WAIT) {
sb->sb_flags &= ~SB_WAIT;
wakeup(&sb->sb_cc);
1994-05-24 10:09:53 +00:00
}
if ((so->so_state & SS_ASYNC) && so->so_sigio != NULL)
pgsigio(&so->so_sigio, SIGIO, 0);
if (sb->sb_flags & SB_UPCALL)
(*so->so_upcall)(so, so->so_upcallarg, M_DONTWAIT);
if (sb->sb_flags & SB_AIO)
aio_swake(so, sb);
KNOTE(&sb->sb_sel.si_note, 0);
1994-05-24 10:09:53 +00:00
}
/*
* Socket buffer (struct sockbuf) utility routines.
*
* Each socket contains two socket buffers: one for sending data and
* one for receiving data. Each buffer contains a queue of mbufs,
* information about the number of mbufs and amount of data in the
* queue, and other fields allowing select() statements and notification
* on data availability to be implemented.
*
* Data stored in a socket buffer is maintained as a list of records.
* Each record is a list of mbufs chained together with the m_next
* field. Records are chained together with the m_nextpkt field. The upper
* level routine soreceive() expects the following conventions to be
* observed when placing information in the receive buffer:
*
* 1. If the protocol requires each message be preceded by the sender's
* name, then a record containing that name must be present before
* any associated data (mbuf's must be of type MT_SONAME).
* 2. If the protocol supports the exchange of ``access rights'' (really
* just additional data associated with the message), and there are
* ``rights'' to be received, then a record containing this data
* should be present (mbuf's must be of type MT_RIGHTS).
* 3. If a name or rights record exists, then it must be followed by
* a data record, perhaps of zero length.
*
* Before using a new socket structure it is first necessary to reserve
* buffer space to the socket, by calling sbreserve(). This should commit
* some of the available buffer space in the system buffer pool for the
* socket (currently, it does nothing but enforce limits). The space
* should be released by calling sbrelease() when the socket is destroyed.
*/
int
1994-05-24 10:09:53 +00:00
soreserve(so, sndcc, rcvcc)
register struct socket *so;
u_long sndcc, rcvcc;
{
struct thread *td = curthread;
1994-05-24 10:09:53 +00:00
if (sbreserve(&so->so_snd, sndcc, so, td) == 0)
1994-05-24 10:09:53 +00:00
goto bad;
if (sbreserve(&so->so_rcv, rcvcc, so, td) == 0)
1994-05-24 10:09:53 +00:00
goto bad2;
if (so->so_rcv.sb_lowat == 0)
so->so_rcv.sb_lowat = 1;
if (so->so_snd.sb_lowat == 0)
so->so_snd.sb_lowat = MCLBYTES;
if (so->so_snd.sb_lowat > so->so_snd.sb_hiwat)
so->so_snd.sb_lowat = so->so_snd.sb_hiwat;
return (0);
bad2:
sbrelease(&so->so_snd, so);
1994-05-24 10:09:53 +00:00
bad:
return (ENOBUFS);
}
static int
sysctl_handle_sb_max(SYSCTL_HANDLER_ARGS)
{
int error = 0;
u_long old_sb_max = sb_max;
error = SYSCTL_OUT(req, arg1, sizeof(u_long));
if (error || !req->newptr)
return (error);
error = SYSCTL_IN(req, arg1, sizeof(u_long));
if (error)
return (error);
if (sb_max < MSIZE + MCLBYTES) {
sb_max = old_sb_max;
return (EINVAL);
}
sb_max_adj = (u_quad_t)sb_max * MCLBYTES / (MSIZE + MCLBYTES);
return (0);
}
1994-05-24 10:09:53 +00:00
/*
* Allot mbufs to a sockbuf.
* Attempt to scale mbmax so that mbcnt doesn't become limiting
* if buffering efficiency is near the normal case.
*/
int
sbreserve(sb, cc, so, td)
1994-05-24 10:09:53 +00:00
struct sockbuf *sb;
u_long cc;
struct socket *so;
struct thread *td;
1994-05-24 10:09:53 +00:00
{
Locking for the per-process resource limits structure. - struct plimit includes a mutex to protect a reference count. The plimit structure is treated similarly to struct ucred in that is is always copy on write, so having a reference to a structure is sufficient to read from it without needing a further lock. - The proc lock protects the p_limit pointer and must be held while reading limits from a process to keep the limit structure from changing out from under you while reading from it. - Various global limits that are ints are not protected by a lock since int writes are atomic on all the archs we support and thus a lock wouldn't buy us anything. - All accesses to individual resource limits from a process are abstracted behind a simple lim_rlimit(), lim_max(), and lim_cur() API that return either an rlimit, or the current or max individual limit of the specified resource from a process. - dosetrlimit() was renamed to kern_setrlimit() to match existing style of other similar syscall helper functions. - The alpha OSF/1 compat layer no longer calls getrlimit() and setrlimit() (it didn't used the stackgap when it should have) but uses lim_rlimit() and kern_setrlimit() instead. - The svr4 compat no longer uses the stackgap for resource limits calls, but uses lim_rlimit() and kern_setrlimit() instead. - The ibcs2 compat no longer uses the stackgap for resource limits. It also no longer uses the stackgap for accessing sysctl's for the ibcs2_sysconf() syscall but uses kernel_sysctl() instead. As a result, ibcs2_sysconf() no longer needs Giant. - The p_rlimit macro no longer exists. Submitted by: mtm (mostly, I only did a few cleanups and catchups) Tested on: i386 Compiled on: alpha, amd64
2004-02-04 21:52:57 +00:00
rlim_t sbsize_limit;
/*
* td will only be NULL when we're in an interrupt
* (e.g. in tcp_input())
*/
if (cc > sb_max_adj)
1994-05-24 10:09:53 +00:00
return (0);
Locking for the per-process resource limits structure. - struct plimit includes a mutex to protect a reference count. The plimit structure is treated similarly to struct ucred in that is is always copy on write, so having a reference to a structure is sufficient to read from it without needing a further lock. - The proc lock protects the p_limit pointer and must be held while reading limits from a process to keep the limit structure from changing out from under you while reading from it. - Various global limits that are ints are not protected by a lock since int writes are atomic on all the archs we support and thus a lock wouldn't buy us anything. - All accesses to individual resource limits from a process are abstracted behind a simple lim_rlimit(), lim_max(), and lim_cur() API that return either an rlimit, or the current or max individual limit of the specified resource from a process. - dosetrlimit() was renamed to kern_setrlimit() to match existing style of other similar syscall helper functions. - The alpha OSF/1 compat layer no longer calls getrlimit() and setrlimit() (it didn't used the stackgap when it should have) but uses lim_rlimit() and kern_setrlimit() instead. - The svr4 compat no longer uses the stackgap for resource limits calls, but uses lim_rlimit() and kern_setrlimit() instead. - The ibcs2 compat no longer uses the stackgap for resource limits. It also no longer uses the stackgap for accessing sysctl's for the ibcs2_sysconf() syscall but uses kernel_sysctl() instead. As a result, ibcs2_sysconf() no longer needs Giant. - The p_rlimit macro no longer exists. Submitted by: mtm (mostly, I only did a few cleanups and catchups) Tested on: i386 Compiled on: alpha, amd64
2004-02-04 21:52:57 +00:00
if (td != NULL) {
PROC_LOCK(td->td_proc);
sbsize_limit = lim_cur(td->td_proc, RLIMIT_SBSIZE);
PROC_UNLOCK(td->td_proc);
} else
sbsize_limit = RLIM_INFINITY;
if (!chgsbsize(so->so_cred->cr_uidinfo, &sb->sb_hiwat, cc,
Locking for the per-process resource limits structure. - struct plimit includes a mutex to protect a reference count. The plimit structure is treated similarly to struct ucred in that is is always copy on write, so having a reference to a structure is sufficient to read from it without needing a further lock. - The proc lock protects the p_limit pointer and must be held while reading limits from a process to keep the limit structure from changing out from under you while reading from it. - Various global limits that are ints are not protected by a lock since int writes are atomic on all the archs we support and thus a lock wouldn't buy us anything. - All accesses to individual resource limits from a process are abstracted behind a simple lim_rlimit(), lim_max(), and lim_cur() API that return either an rlimit, or the current or max individual limit of the specified resource from a process. - dosetrlimit() was renamed to kern_setrlimit() to match existing style of other similar syscall helper functions. - The alpha OSF/1 compat layer no longer calls getrlimit() and setrlimit() (it didn't used the stackgap when it should have) but uses lim_rlimit() and kern_setrlimit() instead. - The svr4 compat no longer uses the stackgap for resource limits calls, but uses lim_rlimit() and kern_setrlimit() instead. - The ibcs2 compat no longer uses the stackgap for resource limits. It also no longer uses the stackgap for accessing sysctl's for the ibcs2_sysconf() syscall but uses kernel_sysctl() instead. As a result, ibcs2_sysconf() no longer needs Giant. - The p_rlimit macro no longer exists. Submitted by: mtm (mostly, I only did a few cleanups and catchups) Tested on: i386 Compiled on: alpha, amd64
2004-02-04 21:52:57 +00:00
sbsize_limit))
return (0);
sb->sb_mbmax = min(cc * sb_efficiency, sb_max);
1994-05-24 10:09:53 +00:00
if (sb->sb_lowat > sb->sb_hiwat)
sb->sb_lowat = sb->sb_hiwat;
return (1);
}
/*
* Free mbufs held by a socket, and reserved mbuf space.
*/
void
sbrelease(sb, so)
1994-05-24 10:09:53 +00:00
struct sockbuf *sb;
struct socket *so;
1994-05-24 10:09:53 +00:00
{
sbflush(sb);
(void)chgsbsize(so->so_cred->cr_uidinfo, &sb->sb_hiwat, 0,
RLIM_INFINITY);
sb->sb_mbmax = 0;
1994-05-24 10:09:53 +00:00
}
/*
* Routines to add and remove
* data from an mbuf queue.
*
* The routines sbappend() or sbappendrecord() are normally called to
* append new mbufs to a socket buffer, after checking that adequate
* space is available, comparing the function sbspace() with the amount
* of data to be added. sbappendrecord() differs from sbappend() in
* that data supplied is treated as the beginning of a new record.
* To place a sender's address, optional access rights, and data in a
* socket receive buffer, sbappendaddr() should be used. To place
* access rights and data in a socket receive buffer, sbappendrights()
* should be used. In either case, the new data begins a new record.
* Note that unlike sbappend() and sbappendrecord(), these routines check
* for the caller that there will be enough space to store the data.
* Each fails if there is not enough space, or if it cannot find mbufs
* to store additional information in.
*
* Reliable protocols may use the socket send buffer to hold data
* awaiting acknowledgement. Data is normally copied from a socket
* send buffer in a protocol with m_copy for output to a peer,
* and then removing the data from the socket buffer with sbdrop()
* or sbdroprecord() when the data is acknowledged by the peer.
*/
#ifdef SOCKBUF_DEBUG
void
sblastrecordchk(struct sockbuf *sb, const char *file, int line)
{
struct mbuf *m = sb->sb_mb;
while (m && m->m_nextpkt)
m = m->m_nextpkt;
if (m != sb->sb_lastrecord) {
printf("%s: sb_mb %p sb_lastrecord %p last %p\n",
__func__, sb->sb_mb, sb->sb_lastrecord, m);
printf("packet chain:\n");
for (m = sb->sb_mb; m != NULL; m = m->m_nextpkt)
printf("\t%p\n", m);
panic("%s from %s:%u", __func__, file, line);
}
}
void
sblastmbufchk(struct sockbuf *sb, const char *file, int line)
{
struct mbuf *m = sb->sb_mb;
struct mbuf *n;
while (m && m->m_nextpkt)
m = m->m_nextpkt;
while (m && m->m_next)
m = m->m_next;
if (m != sb->sb_mbtail) {
printf("%s: sb_mb %p sb_mbtail %p last %p\n",
__func__, sb->sb_mb, sb->sb_mbtail, m);
printf("packet tree:\n");
for (m = sb->sb_mb; m != NULL; m = m->m_nextpkt) {
printf("\t");
for (n = m; n != NULL; n = n->m_next)
printf("%p ", n);
printf("\n");
}
panic("%s from %s:%u", __func__, file, line);
}
}
#endif /* SOCKBUF_DEBUG */
#define SBLINKRECORD(sb, m0) do { \
if ((sb)->sb_lastrecord != NULL) \
(sb)->sb_lastrecord->m_nextpkt = (m0); \
else \
(sb)->sb_mb = (m0); \
(sb)->sb_lastrecord = (m0); \
} while (/*CONSTCOND*/0)
1994-05-24 10:09:53 +00:00
/*
* Append mbuf chain m to the last record in the
* socket buffer sb. The additional space associated
* the mbuf chain is recorded in sb. Empty mbufs are
* discarded and mbufs are compacted where possible.
*/
void
1994-05-24 10:09:53 +00:00
sbappend(sb, m)
struct sockbuf *sb;
struct mbuf *m;
{
register struct mbuf *n;
if (m == 0)
return;
SBLASTRECORDCHK(sb);
n = sb->sb_mb;
if (n) {
1994-05-24 10:09:53 +00:00
while (n->m_nextpkt)
n = n->m_nextpkt;
do {
if (n->m_flags & M_EOR) {
sbappendrecord(sb, m); /* XXXXXX!!!! */
return;
}
} while (n->m_next && (n = n->m_next));
} else {
/*
* XXX Would like to simply use sb_mbtail here, but
* XXX I need to verify that I won't miss an EOR that
* XXX way.
*/
if ((n = sb->sb_lastrecord) != NULL) {
do {
if (n->m_flags & M_EOR) {
sbappendrecord(sb, m); /* XXXXXX!!!! */
return;
}
} while (n->m_next && (n = n->m_next));
} else {
/*
* If this is the first record in the socket buffer,
* it's also the last record.
*/
sb->sb_lastrecord = m;
}
1994-05-24 10:09:53 +00:00
}
sbcompress(sb, m, n);
SBLASTRECORDCHK(sb);
}
/*
* This version of sbappend() should only be used when the caller
* absolutely knows that there will never be more than one record
* in the socket buffer, that is, a stream protocol (such as TCP).
*/
void
sbappendstream(struct sockbuf *sb, struct mbuf *m)
{
KASSERT(m->m_nextpkt == NULL,("sbappendstream 0"));
KASSERT(sb->sb_mb == sb->sb_lastrecord,("sbappendstream 1"));
SBLASTMBUFCHK(sb);
sbcompress(sb, m, sb->sb_mbtail);
sb->sb_lastrecord = sb->sb_mb;
SBLASTRECORDCHK(sb);
1994-05-24 10:09:53 +00:00
}
#ifdef SOCKBUF_DEBUG
void
1994-05-24 10:09:53 +00:00
sbcheck(sb)
struct sockbuf *sb;
1994-05-24 10:09:53 +00:00
{
struct mbuf *m;
struct mbuf *n = 0;
u_long len = 0, mbcnt = 0;
1994-05-24 10:09:53 +00:00
for (m = sb->sb_mb; m; m = n) {
n = m->m_nextpkt;
for (; m; m = m->m_next) {
1994-05-24 10:09:53 +00:00
len += m->m_len;
mbcnt += MSIZE;
if (m->m_flags & M_EXT) /*XXX*/ /* pretty sure this is bogus */
1994-05-24 10:09:53 +00:00
mbcnt += m->m_ext.ext_size;
}
1994-05-24 10:09:53 +00:00
}
if (len != sb->sb_cc || mbcnt != sb->sb_mbcnt) {
printf("cc %ld != %u || mbcnt %ld != %u\n", len, sb->sb_cc,
1994-05-24 10:09:53 +00:00
mbcnt, sb->sb_mbcnt);
panic("sbcheck");
}
}
#endif
/*
* As above, except the mbuf chain
* begins a new record.
*/
void
1994-05-24 10:09:53 +00:00
sbappendrecord(sb, m0)
register struct sockbuf *sb;
register struct mbuf *m0;
{
register struct mbuf *m;
if (m0 == 0)
return;
m = sb->sb_mb;
if (m)
1994-05-24 10:09:53 +00:00
while (m->m_nextpkt)
m = m->m_nextpkt;
/*
* Put the first mbuf on the queue.
* Note this permits zero length records.
*/
sballoc(sb, m0);
SBLASTRECORDCHK(sb);
SBLINKRECORD(sb, m0);
1994-05-24 10:09:53 +00:00
if (m)
m->m_nextpkt = m0;
else
sb->sb_mb = m0;
m = m0->m_next;
m0->m_next = 0;
if (m && (m0->m_flags & M_EOR)) {
m0->m_flags &= ~M_EOR;
m->m_flags |= M_EOR;
}
sbcompress(sb, m, m0);
}
/*
* As above except that OOB data
* is inserted at the beginning of the sockbuf,
* but after any other OOB data.
*/
void
1994-05-24 10:09:53 +00:00
sbinsertoob(sb, m0)
register struct sockbuf *sb;
register struct mbuf *m0;
{
register struct mbuf *m;
register struct mbuf **mp;
if (m0 == 0)
return;
for (mp = &sb->sb_mb; *mp ; mp = &((*mp)->m_nextpkt)) {
m = *mp;
1994-05-24 10:09:53 +00:00
again:
switch (m->m_type) {
case MT_OOBDATA:
continue; /* WANT next train */
case MT_CONTROL:
m = m->m_next;
if (m)
1994-05-24 10:09:53 +00:00
goto again; /* inspect THIS train further */
}
break;
}
/*
* Put the first mbuf on the queue.
* Note this permits zero length records.
*/
sballoc(sb, m0);
m0->m_nextpkt = *mp;
*mp = m0;
m = m0->m_next;
m0->m_next = 0;
if (m && (m0->m_flags & M_EOR)) {
m0->m_flags &= ~M_EOR;
m->m_flags |= M_EOR;
}
sbcompress(sb, m, m0);
}
/*
* Append address and data, and optionally, control (ancillary) data
* to the receive queue of a socket. If present,
* m0 must include a packet header with total length.
* Returns 0 if no space in sockbuf or insufficient mbufs.
*/
int
1994-05-24 10:09:53 +00:00
sbappendaddr(sb, asa, m0, control)
struct sockbuf *sb;
1994-05-24 10:09:53 +00:00
struct sockaddr *asa;
struct mbuf *m0, *control;
{
struct mbuf *m, *n, *nlast;
1994-05-24 10:09:53 +00:00
int space = asa->sa_len;
2001-06-29 04:01:38 +00:00
if (m0 && (m0->m_flags & M_PKTHDR) == 0)
panic("sbappendaddr");
1994-05-24 10:09:53 +00:00
if (m0)
space += m0->m_pkthdr.len;
space += m_length(control, &n);
1994-05-24 10:09:53 +00:00
if (space > sbspace(sb))
return (0);
#if MSIZE <= 256
1994-05-24 10:09:53 +00:00
if (asa->sa_len > MLEN)
return (0);
#endif
MGET(m, M_DONTWAIT, MT_SONAME);
1994-05-24 10:09:53 +00:00
if (m == 0)
return (0);
m->m_len = asa->sa_len;
bcopy(asa, mtod(m, caddr_t), asa->sa_len);
1994-05-24 10:09:53 +00:00
if (n)
n->m_next = m0; /* concatenate data to control */
else
control = m0;
m->m_next = control;
for (n = m; n->m_next != NULL; n = n->m_next)
1994-05-24 10:09:53 +00:00
sballoc(sb, n);
sballoc(sb, n);
nlast = n;
SBLINKRECORD(sb, m);
sb->sb_mbtail = nlast;
SBLASTMBUFCHK(sb);
SBLASTRECORDCHK(sb);
1994-05-24 10:09:53 +00:00
return (1);
}
int
1994-05-24 10:09:53 +00:00
sbappendcontrol(sb, m0, control)
struct sockbuf *sb;
struct mbuf *control, *m0;
{
struct mbuf *m, *n, *mlast;
int space;
1994-05-24 10:09:53 +00:00
if (control == 0)
panic("sbappendcontrol");
space = m_length(control, &n) + m_length(m0, NULL);
1994-05-24 10:09:53 +00:00
if (space > sbspace(sb))
return (0);
n->m_next = m0; /* concatenate data to control */
SBLASTRECORDCHK(sb);
for (m = control; m->m_next; m = m->m_next)
1994-05-24 10:09:53 +00:00
sballoc(sb, m);
sballoc(sb, m);
mlast = m;
SBLINKRECORD(sb, control);
sb->sb_mbtail = mlast;
SBLASTMBUFCHK(sb);
SBLASTRECORDCHK(sb);
1994-05-24 10:09:53 +00:00
return (1);
}
/*
* Compress mbuf chain m into the socket
* buffer sb following mbuf n. If n
* is null, the buffer is presumed empty.
*/
void
1994-05-24 10:09:53 +00:00
sbcompress(sb, m, n)
register struct sockbuf *sb;
register struct mbuf *m, *n;
{
register int eor = 0;
register struct mbuf *o;
while (m) {
eor |= m->m_flags & M_EOR;
if (m->m_len == 0 &&
(eor == 0 ||
(((o = m->m_next) || (o = n)) &&
o->m_type == m->m_type))) {
if (sb->sb_lastrecord == m)
sb->sb_lastrecord = m->m_next;
1994-05-24 10:09:53 +00:00
m = m_free(m);
continue;
}
if (n && (n->m_flags & M_EOR) == 0 &&
M_WRITABLE(n) &&
m->m_len <= MCLBYTES / 4 && /* XXX: Don't copy too much */
m->m_len <= M_TRAILINGSPACE(n) &&
1994-05-24 10:09:53 +00:00
n->m_type == m->m_type) {
bcopy(mtod(m, caddr_t), mtod(n, caddr_t) + n->m_len,
(unsigned)m->m_len);
n->m_len += m->m_len;
sb->sb_cc += m->m_len;
if (m->m_type != MT_DATA && m->m_type != MT_HEADER &&
m->m_type != MT_OOBDATA)
/* XXX: Probably don't need.*/
sb->sb_ctl += m->m_len;
1994-05-24 10:09:53 +00:00
m = m_free(m);
continue;
}
if (n)
n->m_next = m;
else
sb->sb_mb = m;
sb->sb_mbtail = m;
1994-05-24 10:09:53 +00:00
sballoc(sb, m);
n = m;
m->m_flags &= ~M_EOR;
m = m->m_next;
n->m_next = 0;
}
if (eor) {
if (n)
n->m_flags |= eor;
else
printf("semi-panic: sbcompress\n");
}
SBLASTMBUFCHK(sb);
1994-05-24 10:09:53 +00:00
}
/*
* Free all mbufs in a sockbuf.
* Check that all resources are reclaimed.
*/
void
1994-05-24 10:09:53 +00:00
sbflush(sb)
register struct sockbuf *sb;
{
if (sb->sb_flags & SB_LOCK)
panic("sbflush: locked");
while (sb->sb_mbcnt) {
/*
* Don't call sbdrop(sb, 0) if the leading mbuf is non-empty:
* we would loop forever. Panic instead.
*/
if (!sb->sb_cc && (sb->sb_mb == NULL || sb->sb_mb->m_len))
break;
1994-05-24 10:09:53 +00:00
sbdrop(sb, (int)sb->sb_cc);
}
if (sb->sb_cc || sb->sb_mb || sb->sb_mbcnt)
panic("sbflush: cc %u || mb %p || mbcnt %u", sb->sb_cc, (void *)sb->sb_mb, sb->sb_mbcnt);
1994-05-24 10:09:53 +00:00
}
/*
* Drop data from (the front of) a sockbuf.
*/
void
1994-05-24 10:09:53 +00:00
sbdrop(sb, len)
register struct sockbuf *sb;
register int len;
{
register struct mbuf *m;
1994-05-24 10:09:53 +00:00
struct mbuf *next;
next = (m = sb->sb_mb) ? m->m_nextpkt : 0;
while (len > 0) {
if (m == 0) {
if (next == 0)
panic("sbdrop");
m = next;
next = m->m_nextpkt;
continue;
}
if (m->m_len > len) {
m->m_len -= len;
m->m_data += len;
sb->sb_cc -= len;
if (m->m_type != MT_DATA && m->m_type != MT_HEADER &&
m->m_type != MT_OOBDATA)
sb->sb_ctl -= len;
1994-05-24 10:09:53 +00:00
break;
}
len -= m->m_len;
sbfree(sb, m);
m = m_free(m);
1994-05-24 10:09:53 +00:00
}
while (m && m->m_len == 0) {
sbfree(sb, m);
m = m_free(m);
1994-05-24 10:09:53 +00:00
}
if (m) {
sb->sb_mb = m;
m->m_nextpkt = next;
} else
sb->sb_mb = next;
/*
* First part is an inline SB_EMPTY_FIXUP(). Second part
* makes sure sb_lastrecord is up-to-date if we dropped
* part of the last record.
*/
m = sb->sb_mb;
if (m == NULL) {
sb->sb_mbtail = NULL;
sb->sb_lastrecord = NULL;
} else if (m->m_nextpkt == NULL) {
sb->sb_lastrecord = m;
}
1994-05-24 10:09:53 +00:00
}
/*
* Drop a record off the front of a sockbuf
* and move the next record to the front.
*/
void
1994-05-24 10:09:53 +00:00
sbdroprecord(sb)
register struct sockbuf *sb;
{
register struct mbuf *m;
1994-05-24 10:09:53 +00:00
m = sb->sb_mb;
if (m) {
sb->sb_mb = m->m_nextpkt;
do {
sbfree(sb, m);
m = m_free(m);
} while (m);
1994-05-24 10:09:53 +00:00
}
SB_EMPTY_FIXUP(sb);
1994-05-24 10:09:53 +00:00
}
/*
* Create a "control" mbuf containing the specified data
* with the specified type for presentation on a socket buffer.
*/
struct mbuf *
sbcreatecontrol(p, size, type, level)
caddr_t p;
register int size;
int type, level;
{
register struct cmsghdr *cp;
struct mbuf *m;
if (CMSG_SPACE((u_int)size) > MCLBYTES)
return ((struct mbuf *) NULL);
Bring in mbuma to replace mballoc. mbuma is an Mbuf & Cluster allocator built on top of a number of extensions to the UMA framework, all included herein. Extensions to UMA worth noting: - Better layering between slab <-> zone caches; introduce Keg structure which splits off slab cache away from the zone structure and allows multiple zones to be stacked on top of a single Keg (single type of slab cache); perhaps we should look into defining a subset API on top of the Keg for special use by malloc(9), for example. - UMA_ZONE_REFCNT zones can now be added, and reference counters automagically allocated for them within the end of the associated slab structures. uma_find_refcnt() does a kextract to fetch the slab struct reference from the underlying page, and lookup the corresponding refcnt. mbuma things worth noting: - integrates mbuf & cluster allocations with extended UMA and provides caches for commonly-allocated items; defines several zones (two primary, one secondary) and two kegs. - change up certain code paths that always used to do: m_get() + m_clget() to instead just use m_getcl() and try to take advantage of the newly defined secondary Packet zone. - netstat(1) and systat(1) quickly hacked up to do basic stat reporting but additional stats work needs to be done once some other details within UMA have been taken care of and it becomes clearer to how stats will work within the modified framework. From the user perspective, one implication is that the NMBCLUSTERS compile-time option is no longer used. The maximum number of clusters is still capped off according to maxusers, but it can be made unlimited by setting the kern.ipc.nmbclusters boot-time tunable to zero. Work should be done to write an appropriate sysctl handler allowing dynamic tuning of kern.ipc.nmbclusters at runtime. Additional things worth noting/known issues (READ): - One report of 'ips' (ServeRAID) driver acting really slow in conjunction with mbuma. Need more data. Latest report is that ips is equally sucking with and without mbuma. - Giant leak in NFS code sometimes occurs, can't reproduce but currently analyzing; brueffer is able to reproduce but THIS IS NOT an mbuma-specific problem and currently occurs even WITHOUT mbuma. - Issues in network locking: there is at least one code path in the rip code where one or more locks are acquired and we end up in m_prepend() with M_WAITOK, which causes WITNESS to whine from within UMA. Current temporary solution: force all UMA allocations to be M_NOWAIT from within UMA for now to avoid deadlocks unless WITNESS is defined and we can determine with certainty that we're not holding any locks when we're M_WAITOK. - I've seen at least one weird socketbuffer empty-but- mbuf-still-attached panic. I don't believe this to be related to mbuma but please keep your eyes open, turn on debugging, and capture crash dumps. This change removes more code than it adds. A paper is available detailing the change and considering various performance issues, it was presented at BSDCan2004: http://www.unixdaemons.com/~bmilekic/netbuf_bmilekic.pdf Please read the paper for Future Work and implementation details, as well as credits. Testing and Debugging: rwatson, brueffer, Ketrien I. Saihr-Kesenchedra, ... Reviewed by: Lots of people (for different parts)
2004-05-31 21:46:06 +00:00
if (CMSG_SPACE((u_int)size > MLEN))
m = m_getcl(M_DONTWAIT, MT_CONTROL, 0);
else
m = m_get(M_DONTWAIT, MT_CONTROL);
if (m == NULL)
return ((struct mbuf *) NULL);
cp = mtod(m, struct cmsghdr *);
m->m_len = 0;
KASSERT(CMSG_SPACE((u_int)size) <= M_TRAILINGSPACE(m),
("sbcreatecontrol: short mbuf"));
if (p != NULL)
(void)memcpy(CMSG_DATA(cp), p, size);
m->m_len = CMSG_SPACE(size);
cp->cmsg_len = CMSG_LEN(size);
cp->cmsg_level = level;
cp->cmsg_type = type;
return (m);
}
/*
* Some routines that return EOPNOTSUPP for entry points that are not
* supported by a protocol. Fill in as needed.
*/
int
pru_accept_notsupp(struct socket *so, struct sockaddr **nam)
{
return EOPNOTSUPP;
}
int
pru_connect_notsupp(struct socket *so, struct sockaddr *nam, struct thread *td)
{
return EOPNOTSUPP;
}
int
pru_connect2_notsupp(struct socket *so1, struct socket *so2)
{
return EOPNOTSUPP;
}
int
pru_control_notsupp(struct socket *so, u_long cmd, caddr_t data,
struct ifnet *ifp, struct thread *td)
{
return EOPNOTSUPP;
}
int
pru_listen_notsupp(struct socket *so, struct thread *td)
{
return EOPNOTSUPP;
}
int
pru_rcvd_notsupp(struct socket *so, int flags)
{
return EOPNOTSUPP;
}
int
pru_rcvoob_notsupp(struct socket *so, struct mbuf *m, int flags)
{
return EOPNOTSUPP;
}
/*
* This isn't really a ``null'' operation, but it's the default one
* and doesn't do anything destructive.
*/
int
pru_sense_null(struct socket *so, struct stat *sb)
{
sb->st_blksize = so->so_snd.sb_hiwat;
return 0;
}
/*
* For protocol types that don't keep cached copies of labels in their
* pcbs, provide a null sosetlabel that does a NOOP.
*/
void
pru_sosetlabel_null(struct socket *so)
{
}
/*
* Make a copy of a sockaddr in a malloced buffer of type M_SONAME.
*/
struct sockaddr *
sodupsockaddr(const struct sockaddr *sa, int mflags)
{
struct sockaddr *sa2;
sa2 = malloc(sa->sa_len, M_SONAME, mflags);
if (sa2)
bcopy(sa, sa2, sa->sa_len);
return sa2;
}
/*
* Create an external-format (``xsocket'') structure using the information
* in the kernel-format socket structure pointed to by so. This is done
* to reduce the spew of irrelevant information over this interface,
* to isolate user code from changes in the kernel structure, and
* potentially to provide information-hiding if we decide that
* some of this information should be hidden from users.
*/
void
sotoxsocket(struct socket *so, struct xsocket *xso)
{
xso->xso_len = sizeof *xso;
xso->xso_so = so;
xso->so_type = so->so_type;
xso->so_options = so->so_options;
xso->so_linger = so->so_linger;
xso->so_state = so->so_state;
xso->so_pcb = so->so_pcb;
xso->xso_protocol = so->so_proto->pr_protocol;
xso->xso_family = so->so_proto->pr_domain->dom_family;
xso->so_qlen = so->so_qlen;
xso->so_incqlen = so->so_incqlen;
xso->so_qlimit = so->so_qlimit;
xso->so_timeo = so->so_timeo;
xso->so_error = so->so_error;
xso->so_pgid = so->so_sigio ? so->so_sigio->sio_pgid : 0;
xso->so_oobmark = so->so_oobmark;
sbtoxsockbuf(&so->so_snd, &xso->so_snd);
sbtoxsockbuf(&so->so_rcv, &xso->so_rcv);
xso->so_uid = so->so_cred->cr_uid;
}
/*
* This does the same for sockbufs. Note that the xsockbuf structure,
* since it is always embedded in a socket, does not include a self
* pointer nor a length. We make this entry point public in case
* some other mechanism needs it.
*/
void
sbtoxsockbuf(struct sockbuf *sb, struct xsockbuf *xsb)
{
xsb->sb_cc = sb->sb_cc;
xsb->sb_hiwat = sb->sb_hiwat;
xsb->sb_mbcnt = sb->sb_mbcnt;
xsb->sb_mbmax = sb->sb_mbmax;
xsb->sb_lowat = sb->sb_lowat;
xsb->sb_flags = sb->sb_flags;
xsb->sb_timeo = sb->sb_timeo;
}
/*
* Here is the definition of some of the basic objects in the kern.ipc
* branch of the MIB.
*/
SYSCTL_NODE(_kern, KERN_IPC, ipc, CTLFLAG_RW, 0, "IPC");
/* This takes the place of kern.maxsockbuf, which moved to kern.ipc. */
static int dummy;
SYSCTL_INT(_kern, KERN_DUMMY, dummy, CTLFLAG_RW, &dummy, 0, "");
SYSCTL_OID(_kern_ipc, KIPC_MAXSOCKBUF, maxsockbuf, CTLTYPE_ULONG|CTLFLAG_RW,
&sb_max, 0, sysctl_handle_sb_max, "LU", "Maximum socket buffer size");
SYSCTL_INT(_kern_ipc, OID_AUTO, maxsockets, CTLFLAG_RDTUN,
&maxsockets, 0, "Maximum number of sockets avaliable");
SYSCTL_ULONG(_kern_ipc, KIPC_SOCKBUF_WASTE, sockbuf_waste_factor, CTLFLAG_RW,
&sb_efficiency, 0, "");
/*
* Initialise maxsockets
*/
static void init_maxsockets(void *ignored)
{
TUNABLE_INT_FETCH("kern.ipc.maxsockets", &maxsockets);
maxsockets = imax(maxsockets, imax(maxfiles, nmbclusters));
}
SYSINIT(param, SI_SUB_TUNABLES, SI_ORDER_ANY, init_maxsockets, NULL);