no more opt parameter to usrreq

[unix-history] / usr / src / sys / kern / uipc_socket2.c

diff --git a/usr/src/sys/kern/uipc_socket2.c b/usr/src/sys/kern/uipc_socket2.c
index b8ae8e3..6eaf141 100644 (file)
--- a/usr/src/sys/kern/uipc_socket2.c
+++ b/usr/src/sys/kern/uipc_socket2.c
@@ -1,4 +1,4 @@
-/*     uipc_socket2.c  4.8     81/11/22        */
+/*     uipc_socket2.c  4.35    83/01/13        */

 
 #include "../h/param.h"
 #include "../h/systm.h"
 
 #include "../h/param.h"
 #include "../h/systm.h"


@@ -12,8 +12,6 @@
 #include "../h/protosw.h"
 #include "../h/socket.h"
 #include "../h/socketvar.h"
 #include "../h/protosw.h"
 #include "../h/socket.h"
 #include "../h/socketvar.h"

-#include "../net/inet.h"
-#include "../net/inet_systm.h"

 
 /*
  * Primitive routines for operating on sockets and socket buffers
 
 /*
  * Primitive routines for operating on sockets and socket buffers


@@ -21,8 +19,34 @@
 
 /*
  * Procedures to manipulate state flags of socket
 
 /*
  * Procedures to manipulate state flags of socket

- * and do appropriate wakeups.
+ * 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 SO_ACCEPTCONN
+ * creating 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 SO_ACCEPTCONN socket is closed with sockets on either
+ * so_q0 or so_q, these sockets are dropped.
+ *
+ * If and when higher level protocols are implemented in
+ * the kernel, the wakeups done here will sometimes
+ * be implemented as software-interrupt process scheduling.

  */
  */

+

 soisconnecting(so)
        struct socket *so;
 {
 soisconnecting(so)
        struct socket *so;
 {


@@ -35,19 +59,30 @@ soisconnecting(so)
 soisconnected(so)
        struct socket *so;
 {
 soisconnected(so)
        struct socket *so;
 {

+       register struct socket *head = so->so_head;

 
 

+       if (head) {
+               if (soqremque(so, 0) == 0)
+                       panic("soisconnected");
+               soqinsque(head, so, 1);
+               wakeup((caddr_t)&head->so_timeo);
+       }

        so->so_state &= ~(SS_ISCONNECTING|SS_ISDISCONNECTING);
        so->so_state |= SS_ISCONNECTED;
        wakeup((caddr_t)&so->so_timeo);
        so->so_state &= ~(SS_ISCONNECTING|SS_ISDISCONNECTING);
        so->so_state |= SS_ISCONNECTED;
        wakeup((caddr_t)&so->so_timeo);

+       sorwakeup(so);
+       sowwakeup(so);

 }
 
 soisdisconnecting(so)
        struct socket *so;
 {
 
 }
 
 soisdisconnecting(so)
        struct socket *so;
 {
 

-       so->so_state &= ~(SS_ISCONNECTED|SS_ISCONNECTING);
+       so->so_state &= ~SS_ISCONNECTING;

        so->so_state |= (SS_ISDISCONNECTING|SS_CANTRCVMORE|SS_CANTSENDMORE);
        wakeup((caddr_t)&so->so_timeo);
        so->so_state |= (SS_ISDISCONNECTING|SS_CANTRCVMORE|SS_CANTSENDMORE);
        wakeup((caddr_t)&so->so_timeo);

+       sowwakeup(so);
+       sorwakeup(so);

 }
 
 soisdisconnected(so)
 }
 
 soisdisconnected(so)


@@ -61,6 +96,100 @@ soisdisconnected(so)
        sorwakeup(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.
+ */
+struct socket *
+sonewconn(head)
+       register struct socket *head;
+{
+       register struct socket *so;
+       struct mbuf *m;
+
+       if (head->so_qlen + head->so_q0len > 3 * head->so_qlimit / 2)
+               goto bad;
+       m = m_getclr(M_DONTWAIT, MT_SOCKET);
+       if (m == NULL)
+               goto bad;
+       so = mtod(m, struct socket *);
+       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_pgrp = head->so_pgrp;
+       soqinsque(head, so, 0);
+       if ((*so->so_proto->pr_usrreq)(so, PRU_ATTACH, (struct mbuf *)0,
+         (struct mbuf *)0)) {
+               (void) soqremque(so, 0);
+               (void) m_free(m);
+               goto bad;
+       }
+       return (so);
+bad:
+       return ((struct socket *)0);
+}
+
+soqinsque(head, so, q)
+       register struct socket *head, *so;
+       int q;
+{
+
+       so->so_head = head;
+       if (q == 0) {
+               head->so_q0len++;
+               so->so_q0 = head->so_q0;
+               head->so_q0 = so;
+       } else {
+               head->so_qlen++;
+               so->so_q = head->so_q;
+               head->so_q = so;
+       }
+}
+
+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 == head)
+                       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.
+ */
+

 socantsendmore(so)
        struct socket *so;
 {
 socantsendmore(so)
        struct socket *so;
 {


@@ -78,23 +207,38 @@ socantrcvmore(so)
 }
 
 /*
 }
 
 /*

- * Select a socket.
+ * Socket select/wakeup routines.
+ */
+
+/*
+ * Interface routine to select() system
+ * call for sockets.

  */
  */

-soselect(so, flag)
+soselect(so, rw)

        register struct socket *so;
        register struct socket *so;

-       int flag;
+       int rw;

 {
 {

+       int s = splnet();

 
 

-       if (flag & FREAD) {
-               if (soreadable(so))
+       switch (rw) {
+
+       case FREAD:
+               if (soreadable(so)) {
+                       splx(s);

                        return (1);
                        return (1);

+               }

                sbselqueue(&so->so_rcv);
                sbselqueue(&so->so_rcv);

-       }
-       if (flag & FWRITE) {
-               if (sowriteable(so))
+               break;
+
+       case FWRITE:
+               if (sowriteable(so)) {
+                       splx(s);

                        return (1);
                        return (1);

+               }

                sbselqueue(&so->so_snd);
                sbselqueue(&so->so_snd);

+               break;

        }
        }

+       splx(s);

        return (0);
 }
 
        return (0);
 }
 


@@ -131,7 +275,6 @@ sbwakeup(sb)
 {
 
        if (sb->sb_sel) {
 {
 
        if (sb->sb_sel) {

-               printf("sb %x\n", sb);

                selwakeup(sb->sb_sel, sb->sb_flags & SB_COLL);
                sb->sb_sel = 0;
                sb->sb_flags &= ~SB_COLL;
                selwakeup(sb->sb_sel, sb->sb_flags & SB_COLL);
                sb->sb_sel = 0;
                sb->sb_flags &= ~SB_COLL;


@@ -142,6 +285,60 @@ sbwakeup(sb)
        }
 }
 
        }
 }
 

+/*
+ * 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.
+ *
+ * Before using a new socket structure it is first necessary to reserve
+ * buffer space to the socket, by calling sbreserve.  This commits
+ * some of the available buffer space in the system buffer pool for the
+ * socket.  The space should be released by calling sbrelease when the
+ * socket is destroyed.
+ *
+ * The routine sbappend() is normally called to append new mbufs
+ * to a socket buffer, after checking that adequate space is available
+ * comparing the function spspace() with the amount of data to be added.
+ * Data is normally removed from a socket buffer in a protocol by
+ * first calling m_copy on the socket buffer mbuf chain and sending this
+ * to a peer, and then removing the data from the socket buffer with
+ * sbdrop when the data is acknowledged by the peer (or immediately
+ * in the case of unreliable protocols.)
+ *
+ * Protocols which do not require connections place both source address
+ * and data information in socket buffer queues.  The source addresses
+ * are stored in single mbufs after each data item, and are easily found
+ * as the data items are all marked with end of record markers.  The
+ * sbappendaddr() routine stores a datum and associated address in
+ * a socket buffer.  Note that, unlike sbappend(), this routine checks
+ * for the caller that there will be enough space to store the data.
+ * It fails if there is not enough space, or if it cannot find
+ * a mbuf to store the address in.
+ *
+ * The higher-level routines sosend and soreceive (in socket.c)
+ * also add data to, and remove data from socket buffers repectively.
+ */
+
+soreserve(so, sndcc, rcvcc)
+       struct socket *so;
+       int sndcc, rcvcc;
+{
+
+       if (sbreserve(&so->so_snd, sndcc) == 0)
+               goto bad;
+       if (sbreserve(&so->so_rcv, rcvcc) == 0)
+               goto bad2;
+       return (0);
+bad2:
+       sbrelease(&so->so_snd);
+bad:
+       return (ENOBUFS);
+}
+

 /*
  * Allot mbufs to a sockbuf.
  */
 /*
  * Allot mbufs to a sockbuf.
  */


@@ -149,9 +346,9 @@ sbreserve(sb, cc)
        struct sockbuf *sb;
 {
 
        struct sockbuf *sb;
 {
 

-       if (m_reserve((cc*2)/MSIZE) == 0)
-               return (0);
+       /* someday maybe this routine will fail... */

        sb->sb_hiwat = cc;
        sb->sb_hiwat = cc;

+       /* the 2 implies names can be no more than 1 mbuf each */

        sb->sb_mbmax = cc*2;
        return (1);
 }
        sb->sb_mbmax = cc*2;
        return (1);
 }


@@ -164,7 +361,6 @@ sbrelease(sb)
 {
 
        sbflush(sb);
 {
 
        sbflush(sb);

-       m_release(sb->sb_mbmax/MSIZE);

        sb->sb_hiwat = sb->sb_mbmax = 0;
 }
 
        sb->sb_hiwat = sb->sb_mbmax = 0;
 }
 


@@ -180,16 +376,17 @@ sbappend(sb, m)
        register struct mbuf *m;
        register struct sockbuf *sb;
 {
        register struct mbuf *m;
        register struct sockbuf *sb;
 {

-       register struct mbuf **np, *n;
+       register struct mbuf *n;

 
 

-       np = &sb->sb_mb;
-       n = 0;
-       while (*np) {
-               n = *np;
-               np = &n->m_next;
-       }
+       n = sb->sb_mb;
+       if (n)
+               while (n->m_next)
+                       n = n->m_next;

        while (m) {
        while (m) {

-/*
+               if (m->m_len == 0 && (int)m->m_act == 0) {
+                       m = m_free(m);
+                       continue;
+               }

                if (n && n->m_off <= MMAXOFF && m->m_off <= MMAXOFF &&
                   (int)n->m_act == 0 && (int)m->m_act == 0 &&
                   (n->m_off + n->m_len + m->m_len) <= MMAXOFF) {
                if (n && n->m_off <= MMAXOFF && m->m_off <= MMAXOFF &&
                   (int)n->m_act == 0 && (int)m->m_act == 0 &&
                   (n->m_off + n->m_len + m->m_len) <= MMAXOFF) {


@@ -200,16 +397,22 @@ sbappend(sb, m)
                        m = m_free(m);
                        continue;
                }
                        m = m_free(m);
                        continue;
                }

-*/

                sballoc(sb, m);
                sballoc(sb, m);

-               *np = m;
+               if (n == 0)
+                       sb->sb_mb = m;
+               else
+                       n->m_next = m;

                n = m;
                n = m;

-               np = &n->m_next;

                m = m->m_next;
                m = m->m_next;

+               n->m_next = 0;

        }
        }

-       nullchk("sbappend", sb->sb_mb);

 }
 
 }
 

+/*
+ * Append data and address.
+ * Return 0 if no space in sockbuf or if
+ * can't get mbuf to stuff address in.
+ */

 sbappendaddr(sb, asa, m0)
        struct sockbuf *sb;
        struct sockaddr *asa;
 sbappendaddr(sb, asa, m0)
        struct sockbuf *sb;
        struct sockaddr *asa;


@@ -219,7 +422,6 @@ sbappendaddr(sb, asa, m0)
        register struct mbuf *m;
        register int len = sizeof (struct sockaddr);
 
        register struct mbuf *m;
        register int len = sizeof (struct sockaddr);
 

-       printf("sbappendaddr %x asa %x ", sb, asa);

        m = m0;
        if (m == 0)
                panic("sbappendaddr");
        m = m0;
        if (m == 0)
                panic("sbappendaddr");


@@ -231,29 +433,42 @@ sbappendaddr(sb, asa, m0)
                }
                m = m->m_next;
        }
                }
                m = m->m_next;
        }

-       printf(": sb_cc %d sb_hiwat %d sb_mbcnt %d sb_mbmax %d sbspace %d: ",
-           sb->sb_cc, sb->sb_hiwat, sb->sb_mbcnt, sb->sb_mbmax,
-           sbspace(sb));
-       if (len > sbspace(sb)) {
-               printf("no space\n");
+       if (len > sbspace(sb))

                return (0);
                return (0);

-       }
-       m = m_get(0);
-       if (m == 0) {
-               printf("no mbufs\n");
+       m = m_get(M_DONTWAIT, MT_SONAME);
+       if (m == 0)

                return (0);
                return (0);

-       }
-       m->m_off = MMINOFF;

        m->m_len = sizeof (struct sockaddr);
        msa = mtod(m, struct sockaddr *);
        *msa = *asa;
        m->m_act = (struct mbuf *)1;
        sbappend(sb, m);
        sbappend(sb, m0);
        m->m_len = sizeof (struct sockaddr);
        msa = mtod(m, struct sockaddr *);
        *msa = *asa;
        m->m_act = (struct mbuf *)1;
        sbappend(sb, m);
        sbappend(sb, m0);

-       printf("inserted: sb_cc now %d\n", sb->sb_cc);

        return (1);
 }
 
        return (1);
 }
 

+#ifdef notdef
+SBCHECK(sb, str)
+       struct sockbuf *sb;
+       char *str;
+{
+       register int cnt = sb->sb_cc;
+       register int mbcnt = sb->sb_mbcnt;
+       register struct mbuf *m;
+
+       for (m = sb->sb_mb; m; m = m->m_next) {
+               cnt -= m->m_len;
+               mbcnt -= MSIZE;
+               if (m->m_off > MMAXOFF)
+                       mbcnt -= CLBYTES;
+       }
+       if (cnt || mbcnt) {
+               printf("cnt %d mbcnt %d\n", cnt, mbcnt);
+               panic(str);
+       }
+}
+#endif
+

 /*
  * Free all mbufs on a sockbuf mbuf chain.
  * Check that resource allocations return to 0.
 /*
  * Free all mbufs on a sockbuf mbuf chain.
  * Check that resource allocations return to 0.


@@ -264,7 +479,8 @@ sbflush(sb)
 
        if (sb->sb_flags & SB_LOCK)
                panic("sbflush");
 
        if (sb->sb_flags & SB_LOCK)
                panic("sbflush");

-       sbdrop(sb, sb->sb_cc);
+       if (sb->sb_cc)
+               sbdrop(sb, sb->sb_cc);

        if (sb->sb_cc || sb->sb_mbcnt || sb->sb_mb)
                panic("sbflush 2");
 }
        if (sb->sb_cc || sb->sb_mbcnt || sb->sb_mb)
                panic("sbflush 2");
 }


@@ -281,18 +497,16 @@ sbdrop(sb, len)
        while (len > 0) {
                if (m == 0)
                        panic("sbdrop");
        while (len > 0) {
                if (m == 0)
                        panic("sbdrop");

-               if (m->m_len <= len) {
-                       len -= m->m_len;
-                       sbfree(sb, m);
-                       MFREE(m, mn);
-                       m = mn;
-               } else {
+               if (m->m_len > len) {

                        m->m_len -= len;
                        m->m_off += len;
                        sb->sb_cc -= len;
                        break;
                }
                        m->m_len -= len;
                        m->m_off += len;
                        sb->sb_cc -= len;
                        break;
                }

+               len -= m->m_len;
+               sbfree(sb, m);
+               MFREE(m, mn);
+               m = mn;

        }
        sb->sb_mb = m;
        }
        sb->sb_mb = m;

-       nullchk("sbdrop", sb->sb_mb);

 }
 }