freebsd-dev/sys/kern/uipc_socket2.c
Garrett Wollman 4b29bc4f8a Eliminate the dramatic TCP performance decrease observed for writes in
the range [210:260] by sweeping the problem under the rug.  This change
has the following effects:

1) A new MIB variable in the kern branch is defined to allow modification
of the socket buffer layer's ``wastage factor'' (which determines how
much unused-but-allocated space in mbufs and mbuf clusters is allowed
in a socket buffer).

2) The default value of the wastage factor is changed from 2 to 8.
The original value was chosen when MINCLSIZE was 7*MLEN (!), and is not
appropriate for an environment where MINCLSIZE is much less.

The real solution to this problem is to scrap both mbufs and sockbufs
and completely redesign the buffering mechanism used at both levels.
1996-01-05 21:41:54 +00:00

793 lines
20 KiB
C

/*
* 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.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by the University of
* California, Berkeley and its contributors.
* 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
* $Id: uipc_socket2.c,v 1.7 1995/12/14 22:51:02 bde Exp $
*/
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/proc.h>
#include <sys/file.h>
#include <sys/buf.h>
#include <sys/malloc.h>
#include <sys/mbuf.h>
#include <sys/protosw.h>
#include <sys/stat.h>
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <sys/signalvar.h>
#include <sys/sysctl.h>
/*
* Primitive routines for operating on sockets and socket buffers
*/
u_long sb_max = SB_MAX; /* XXX should be static */
SYSCTL_INT(_kern, KERN_MAXSOCKBUF, maxsockbuf, CTLFLAG_RW, &sb_max, 0, "")
static u_long sb_efficiency = 8; /* parameter for sbreserve() */
SYSCTL_INT(_kern, OID_AUTO, sockbuf_waste_factor, CTLFLAG_RW, &sb_efficiency,
0, "");
/*
* 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_q0 for connections in progress
* and so_q for connections already made and awaiting user acceptance.
* As a protocol is preparing incoming connections, it creates a socket
* structure queued on so_q0 by calling sonewconn(). When the connection
* is established, soisconnected() is called, and transfers the
* socket structure to so_q, making it available to accept().
*
* If a socket is closed with sockets on either
* so_q0 or so_q, these sockets are dropped.
*
* If higher level protocols are implemented in
* the kernel, the wakeups done here will sometimes
* cause software-interrupt process scheduling.
*/
void
soisconnecting(so)
register struct socket *so;
{
so->so_state &= ~(SS_ISCONNECTED|SS_ISDISCONNECTING);
so->so_state |= SS_ISCONNECTING;
}
void
soisconnected(so)
register struct socket *so;
{
register struct socket *head = so->so_head;
so->so_state &= ~(SS_ISCONNECTING|SS_ISDISCONNECTING|SS_ISCONFIRMING);
so->so_state |= SS_ISCONNECTED;
if (head && soqremque(so, 0)) {
soqinsque(head, so, 1);
sorwakeup(head);
wakeup((caddr_t)&head->so_timeo);
} else {
wakeup((caddr_t)&so->so_timeo);
sorwakeup(so);
sowwakeup(so);
}
}
void
soisdisconnecting(so)
register struct socket *so;
{
so->so_state &= ~SS_ISCONNECTING;
so->so_state |= (SS_ISDISCONNECTING|SS_CANTRCVMORE|SS_CANTSENDMORE);
wakeup((caddr_t)&so->so_timeo);
sowwakeup(so);
sorwakeup(so);
}
void
soisdisconnected(so)
register struct socket *so;
{
so->so_state &= ~(SS_ISCONNECTING|SS_ISCONNECTED|SS_ISDISCONNECTING);
so->so_state |= (SS_CANTRCVMORE|SS_CANTSENDMORE);
wakeup((caddr_t)&so->so_timeo);
sowwakeup(so);
sorwakeup(so);
}
/*
* 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.
*
* Currently, sonewconn() is defined as sonewconn1() in socketvar.h
* to catch calls that are missing the (new) second parameter.
*/
struct socket *
sonewconn1(head, connstatus)
register struct socket *head;
int connstatus;
{
register struct socket *so;
int soqueue = connstatus ? 1 : 0;
if (head->so_qlen + head->so_q0len > 3 * head->so_qlimit / 2)
return ((struct socket *)0);
MALLOC(so, struct socket *, sizeof(*so), M_SOCKET, M_DONTWAIT);
if (so == NULL)
return ((struct socket *)0);
bzero((caddr_t)so, sizeof(*so));
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_pgid = head->so_pgid;
(void) soreserve(so, head->so_snd.sb_hiwat, head->so_rcv.sb_hiwat);
soqinsque(head, so, soqueue);
if ((*so->so_proto->pr_usrreq)(so, PRU_ATTACH,
(struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0)) {
(void) soqremque(so, soqueue);
(void) free((caddr_t)so, M_SOCKET);
return ((struct socket *)0);
}
if (connstatus) {
sorwakeup(head);
wakeup((caddr_t)&head->so_timeo);
so->so_state |= connstatus;
}
return (so);
}
void
soqinsque(head, so, q)
register struct socket *head, *so;
int q;
{
register struct socket **prev;
so->so_head = head;
if (q == 0) {
head->so_q0len++;
so->so_q0 = 0;
for (prev = &(head->so_q0); *prev; )
prev = &((*prev)->so_q0);
} else {
head->so_qlen++;
so->so_q = 0;
for (prev = &(head->so_q); *prev; )
prev = &((*prev)->so_q);
}
*prev = so;
}
int
soqremque(so, q)
register struct socket *so;
int q;
{
register struct socket *head, *prev, *next;
head = so->so_head;
prev = head;
for (;;) {
next = q ? prev->so_q : prev->so_q0;
if (next == so)
break;
if (next == 0)
return (0);
prev = next;
}
if (q == 0) {
prev->so_q0 = next->so_q0;
head->so_q0len--;
} else {
prev->so_q = next->so_q;
head->so_qlen--;
}
next->so_q0 = next->so_q = 0;
next->so_head = 0;
return (1);
}
/*
* 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
socantsendmore(so)
struct socket *so;
{
so->so_state |= SS_CANTSENDMORE;
sowwakeup(so);
}
void
socantrcvmore(so)
struct socket *so;
{
so->so_state |= SS_CANTRCVMORE;
sorwakeup(so);
}
/*
* Wait for data to arrive at/drain from a socket buffer.
*/
int
sbwait(sb)
struct sockbuf *sb;
{
sb->sb_flags |= SB_WAIT;
return (tsleep((caddr_t)&sb->sb_cc,
(sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK | PCATCH, "sbwait",
sb->sb_timeo));
}
/*
* Lock a sockbuf already known to be locked;
* return any error returned from sleep (EINTR).
*/
int
sb_lock(sb)
register struct sockbuf *sb;
{
int error;
while (sb->sb_flags & SB_LOCK) {
sb->sb_flags |= SB_WANT;
error = tsleep((caddr_t)&sb->sb_flags,
(sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK|PCATCH,
"sblock", 0);
if (error)
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
sowakeup(so, sb)
register struct socket *so;
register struct sockbuf *sb;
{
struct proc *p;
selwakeup(&sb->sb_sel);
sb->sb_flags &= ~SB_SEL;
if (sb->sb_flags & SB_WAIT) {
sb->sb_flags &= ~SB_WAIT;
wakeup((caddr_t)&sb->sb_cc);
}
if (so->so_state & SS_ASYNC) {
if (so->so_pgid < 0)
gsignal(-so->so_pgid, SIGIO);
else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0)
psignal(p, SIGIO);
}
}
/*
* 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
soreserve(so, sndcc, rcvcc)
register struct socket *so;
u_long sndcc, rcvcc;
{
if (sbreserve(&so->so_snd, sndcc) == 0)
goto bad;
if (sbreserve(&so->so_rcv, rcvcc) == 0)
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);
bad:
return (ENOBUFS);
}
/*
* 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)
struct sockbuf *sb;
u_long cc;
{
if (cc > sb_max * MCLBYTES / (MSIZE + MCLBYTES))
return (0);
sb->sb_hiwat = cc;
sb->sb_mbmax = min(cc * sb_efficiency, sb_max);
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)
struct sockbuf *sb;
{
sbflush(sb);
sb->sb_hiwat = sb->sb_mbmax = 0;
}
/*
* 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.
*/
/*
* 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
sbappend(sb, m)
struct sockbuf *sb;
struct mbuf *m;
{
register struct mbuf *n;
if (m == 0)
return;
n = sb->sb_mb;
if (n) {
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));
}
sbcompress(sb, m, n);
}
#ifdef SOCKBUF_DEBUG
void
sbcheck(sb)
register struct sockbuf *sb;
{
register struct mbuf *m;
register int len = 0, mbcnt = 0;
for (m = sb->sb_mb; m; m = m->m_next) {
len += m->m_len;
mbcnt += MSIZE;
if (m->m_flags & M_EXT)
mbcnt += m->m_ext.ext_size;
if (m->m_nextpkt)
panic("sbcheck nextpkt");
}
if (len != sb->sb_cc || mbcnt != sb->sb_mbcnt) {
printf("cc %d != %d || mbcnt %d != %d\n", len, sb->sb_cc,
mbcnt, sb->sb_mbcnt);
panic("sbcheck");
}
}
#endif
/*
* As above, except the mbuf chain
* begins a new record.
*/
void
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)
while (m->m_nextpkt)
m = m->m_nextpkt;
/*
* Put the first mbuf on the queue.
* Note this permits zero length records.
*/
sballoc(sb, m0);
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
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;
again:
switch (m->m_type) {
case MT_OOBDATA:
continue; /* WANT next train */
case MT_CONTROL:
m = m->m_next;
if (m)
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
sbappendaddr(sb, asa, m0, control)
register struct sockbuf *sb;
struct sockaddr *asa;
struct mbuf *m0, *control;
{
register struct mbuf *m, *n;
int space = asa->sa_len;
if (m0 && (m0->m_flags & M_PKTHDR) == 0)
panic("sbappendaddr");
if (m0)
space += m0->m_pkthdr.len;
for (n = control; n; n = n->m_next) {
space += n->m_len;
if (n->m_next == 0) /* keep pointer to last control buf */
break;
}
if (space > sbspace(sb))
return (0);
if (asa->sa_len > MLEN)
return (0);
MGET(m, M_DONTWAIT, MT_SONAME);
if (m == 0)
return (0);
m->m_len = asa->sa_len;
bcopy((caddr_t)asa, mtod(m, caddr_t), asa->sa_len);
if (n)
n->m_next = m0; /* concatenate data to control */
else
control = m0;
m->m_next = control;
for (n = m; n; n = n->m_next)
sballoc(sb, n);
n = sb->sb_mb;
if (n) {
while (n->m_nextpkt)
n = n->m_nextpkt;
n->m_nextpkt = m;
} else
sb->sb_mb = m;
return (1);
}
int
sbappendcontrol(sb, m0, control)
struct sockbuf *sb;
struct mbuf *control, *m0;
{
register struct mbuf *m, *n;
int space = 0;
if (control == 0)
panic("sbappendcontrol");
for (m = control; ; m = m->m_next) {
space += m->m_len;
if (m->m_next == 0)
break;
}
n = m; /* save pointer to last control buffer */
for (m = m0; m; m = m->m_next)
space += m->m_len;
if (space > sbspace(sb))
return (0);
n->m_next = m0; /* concatenate data to control */
for (m = control; m; m = m->m_next)
sballoc(sb, m);
n = sb->sb_mb;
if (n) {
while (n->m_nextpkt)
n = n->m_nextpkt;
n->m_nextpkt = control;
} else
sb->sb_mb = control;
return (1);
}
/*
* Compress mbuf chain m into the socket
* buffer sb following mbuf n. If n
* is null, the buffer is presumed empty.
*/
void
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))) {
m = m_free(m);
continue;
}
if (n && (n->m_flags & (M_EXT | M_EOR)) == 0 &&
(n->m_data + n->m_len + m->m_len) < &n->m_dat[MLEN] &&
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;
m = m_free(m);
continue;
}
if (n)
n->m_next = m;
else
sb->sb_mb = m;
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");
}
}
/*
* Free all mbufs in a sockbuf.
* Check that all resources are reclaimed.
*/
void
sbflush(sb)
register struct sockbuf *sb;
{
if (sb->sb_flags & SB_LOCK)
panic("sbflush");
while (sb->sb_mbcnt)
sbdrop(sb, (int)sb->sb_cc);
if (sb->sb_cc || sb->sb_mb)
panic("sbflush 2");
}
/*
* Drop data from (the front of) a sockbuf.
*/
void
sbdrop(sb, len)
register struct sockbuf *sb;
register int len;
{
register struct mbuf *m, *mn;
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;
break;
}
len -= m->m_len;
sbfree(sb, m);
MFREE(m, mn);
m = mn;
}
while (m && m->m_len == 0) {
sbfree(sb, m);
MFREE(m, mn);
m = mn;
}
if (m) {
sb->sb_mb = m;
m->m_nextpkt = next;
} else
sb->sb_mb = next;
}
/*
* Drop a record off the front of a sockbuf
* and move the next record to the front.
*/
void
sbdroprecord(sb)
register struct sockbuf *sb;
{
register struct mbuf *m, *mn;
m = sb->sb_mb;
if (m) {
sb->sb_mb = m->m_nextpkt;
do {
sbfree(sb, m);
MFREE(m, mn);
m = mn;
} while (m);
}
}