X-Git-Url: https://git.subgeniuskitty.com/unix-history/.git/blobdiff_plain/39d536e67f78a1d5cb735af8573e874d3b4aae66..01932c99d1800e02c98500a66b1c9d62d0f820c8:/usr/src/sys/kern/uipc_mbuf.c

diff --git a/usr/src/sys/kern/uipc_mbuf.c b/usr/src/sys/kern/uipc_mbuf.c
index cedc50038a..7959e40778 100644
--- a/usr/src/sys/kern/uipc_mbuf.c
+++ b/usr/src/sys/kern/uipc_mbuf.c
@@ -1,98 +1,126 @@
-/*	uipc_mbuf.c	1.39	82/10/21	*/
-
-#include "../h/param.h"
-#include "../h/dir.h"
-#include "../h/user.h"
-#include "../h/proc.h"
-#include "../h/pte.h"
-#include "../h/cmap.h"
-#include "../h/map.h"
-#include "../h/mbuf.h"
-#include "../h/vm.h"
-#include "../h/kernel.h"
+/*
+ * All rights reserved.
+ *
+ * %sccs.include.redist.c%
+ *
+ *	@(#)uipc_mbuf.c	7.22 (Berkeley) %G%
+ */
+
+#include "param.h"
+#include "proc.h"
+#include "malloc.h"
+#include "map.h"
+#define MBTYPES
+#include "mbuf.h"
+#include "kernel.h"
+#include "syslog.h"
+#include "domain.h"
+#include "protosw.h"
+#include "vm/vm.h"
+
+extern	vm_map_t mb_map;
+struct	mbuf *mbutl;
+char	*mclrefcnt;
 
 mbinit()
 {
+	int s;
 
-	if (m_clalloc(4, MPG_MBUFS) == 0)
-		goto bad;
-	if (m_clalloc(32, MPG_CLUSTERS) == 0)
+#if CLBYTES < 4096
+#define NCL_INIT	(4096/CLBYTES)
+#else
+#define NCL_INIT	1
+#endif
+	s = splimp();
+	if (m_clalloc(NCL_INIT, M_DONTWAIT) == 0)
 		goto bad;
+	splx(s);
 	return;
 bad:
 	panic("mbinit");
 }
 
-caddr_t
-m_clalloc(ncl, how)
+/*
+ * Allocate some number of mbuf clusters
+ * and place on cluster free list.
+ * Must be called at splimp.
+ */
+/* ARGSUSED */
+m_clalloc(ncl, nowait)
 	register int ncl;
-	int how;
+	int nowait;
 {
 	int npg, mbx;
-	register struct mbuf *m;
+	register caddr_t p;
 	register int i;
-	int s;
+	static int logged;
 
 	npg = ncl * CLSIZE;
-	s = splimp();		/* careful: rmalloc isn't reentrant */
-	mbx = rmalloc(mbmap, npg);
-	splx(s);
-	if (mbx == 0)
-		return (0);
-	m = cltom(mbx / CLSIZE);
-	if (memall(&Mbmap[mbx], npg, proc, CSYS) == 0)
-		return (0);
-	vmaccess(&Mbmap[mbx], (caddr_t)m, npg);
-	switch (how) {
-
-	case MPG_CLUSTERS:
-		s = splimp();
-		for (i = 0; i < ncl; i++) {
-			m->m_off = 0;
-			m->m_next = mclfree;
-			mclfree = m;
-			m += CLBYTES / sizeof (*m);
-			mbstat.m_clfree++;
-		}
-		mbstat.m_clusters += ncl;
-		splx(s);
-		break;
-
-	case MPG_MBUFS:
-		for (i = ncl * CLBYTES / sizeof (*m); i > 0; i--) {
-			m->m_off = 0;
-			m->m_free = 0;
-			mbstat.m_mbufs++;
-			(void) m_free(m);
-			m++;
+	p = (caddr_t)kmem_malloc(mb_map, ctob(npg), !nowait);
+	if (p == NULL) {
+		if (logged == 0) {
+			logged++;
+			log(LOG_ERR, "mb_map full\n");
 		}
-		break;
+		return (0);
 	}
-	return ((caddr_t)m);
+	ncl = ncl * CLBYTES / MCLBYTES;
+	for (i = 0; i < ncl; i++) {
+		((union mcluster *)p)->mcl_next = mclfree;
+		mclfree = (union mcluster *)p;
+		p += MCLBYTES;
+		mbstat.m_clfree++;
+	}
+	mbstat.m_clusters += ncl;
+	return (1);
 }
 
-m_pgfree(addr, n)
-	caddr_t addr;
-	int n;
+/*
+ * When MGET failes, ask protocols to free space when short of memory,
+ * then re-attempt to allocate an mbuf.
+ */
+struct mbuf *
+m_retry(i, t)
+	int i, t;
 {
+	register struct mbuf *m;
 
-#ifdef lint
-	addr = addr; n = n;
-#endif
+	m_reclaim();
+#define m_retry(i, t)	(struct mbuf *)0
+	MGET(m, i, t);
+#undef m_retry
+	return (m);
 }
 
-m_expand()
+/*
+ * As above; retry an MGETHDR.
+ */
+struct mbuf *
+m_retryhdr(i, t)
+	int i, t;
 {
+	register struct mbuf *m;
 
-	if (m_clalloc(1, MPG_MBUFS) == 0)
-		goto steal;
-	return (1);
-steal:
-	/* should ask protocols to free code */
-	return (0);
+	m_reclaim();
+#define m_retryhdr(i, t) (struct mbuf *)0
+	MGETHDR(m, i, t);
+#undef m_retryhdr
+	return (m);
 }
 
-/* NEED SOME WAY TO RELEASE SPACE */
+m_reclaim()
+{
+	register struct domain *dp;
+	register struct protosw *pr;
+	int s = splimp();
+
+	for (dp = domains; dp; dp = dp->dom_next)
+		for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++)
+			if (pr->pr_drain)
+				(*pr->pr_drain)();
+	splx(s);
+	mbstat.m_drain++;
+}
 
 /*
  * Space allocation routines.
@@ -100,22 +128,32 @@ steal:
  * for critical paths.
  */
 struct mbuf *
-m_get(canwait)
-	int canwait;
+m_get(nowait, type)
+	int nowait, type;
+{
+	register struct mbuf *m;
+
+	MGET(m, nowait, type);
+	return (m);
+}
+
+struct mbuf *
+m_gethdr(nowait, type)
+	int nowait, type;
 {
 	register struct mbuf *m;
 
-	MGET(m, canwait);
+	MGETHDR(m, nowait, type);
 	return (m);
 }
 
 struct mbuf *
-m_getclr(canwait)
-	int canwait;
+m_getclr(nowait, type)
+	int nowait, type;
 {
 	register struct mbuf *m;
 
-	m = m_get(canwait);
+	MGET(m, nowait, type);
 	if (m == 0)
 		return (0);
 	bzero(mtod(m, caddr_t), MLEN);
@@ -132,57 +170,77 @@ m_free(m)
 	return (n);
 }
 
-/*ARGSUSED*/
-struct mbuf *
-m_more(type)
-	int type;
-{
-	register struct mbuf *m;
-
-	if (!m_expand()) {
-		mbstat.m_drops++;
-		return (NULL);
-	}
-#define m_more(x) (panic("m_more"), (struct mbuf *)0)
-	MGET(m, type);
-#undef m_more
-	return (m);
-}
-
+void
 m_freem(m)
 	register struct mbuf *m;
 {
 	register struct mbuf *n;
-	register int s;
 
 	if (m == NULL)
 		return;
-	s = splimp();
 	do {
 		MFREE(m, n);
 	} while (m = n);
-	splx(s);
 }
 
 /*
  * Mbuffer utility routines.
  */
+
+/*
+ * Lesser-used path for M_PREPEND:
+ * allocate new mbuf to prepend to chain,
+ * copy junk along.
+ */
 struct mbuf *
-m_copy(m, off, len)
+m_prepend(m, len, how)
 	register struct mbuf *m;
-	int off;
+	int len, how;
+{
+	struct mbuf *mn;
+
+	MGET(mn, how, m->m_type);
+	if (mn == (struct mbuf *)NULL) {
+		m_freem(m);
+		return ((struct mbuf *)NULL);
+	}
+	if (m->m_flags & M_PKTHDR) {
+		M_COPY_PKTHDR(mn, m);
+		m->m_flags &= ~M_PKTHDR;
+	}
+	mn->m_next = m;
+	m = mn;
+	if (len < MHLEN)
+		MH_ALIGN(m, len);
+	m->m_len = len;
+	return (m);
+}
+
+/*
+ * Make a copy of an mbuf chain starting "off0" bytes from the beginning,
+ * continuing for "len" bytes.  If len is M_COPYALL, copy to end of mbuf.
+ * The wait parameter is a choice of M_WAIT/M_DONTWAIT from caller.
+ */
+int MCFail;
+
+struct mbuf *
+m_copym(m, off0, len, wait)
+	register struct mbuf *m;
+	int off0, wait;
 	register int len;
 {
 	register struct mbuf *n, **np;
-	struct mbuf *top, *p;
+	register int off = off0;
+	struct mbuf *top;
+	int copyhdr = 0;
 
-	if (len == 0)
-		return (0);
 	if (off < 0 || len < 0)
-		panic("m_copy");
+		panic("m_copym");
+	if (off == 0 && m->m_flags & M_PKTHDR)
+		copyhdr = 1;
 	while (off > 0) {
 		if (m == 0)
-			panic("m_copy");
+			panic("m_copym");
 		if (off < m->m_len)
 			break;
 		off -= m->m_len;
@@ -193,18 +251,27 @@ m_copy(m, off, len)
 	while (len > 0) {
 		if (m == 0) {
 			if (len != M_COPYALL)
-				panic("m_copy");
+				panic("m_copym");
 			break;
 		}
-		MGET(n, 1);
+		MGET(n, wait, m->m_type);
 		*np = n;
 		if (n == 0)
 			goto nospace;
+		if (copyhdr) {
+			M_COPY_PKTHDR(n, m);
+			if (len == M_COPYALL)
+				n->m_pkthdr.len -= off0;
+			else
+				n->m_pkthdr.len = len;
+			copyhdr = 0;
+		}
 		n->m_len = MIN(len, m->m_len - off);
-		if (m->m_off > MMAXOFF) {
-			p = mtod(m, struct mbuf *);
-			n->m_off = ((int)p - (int)n) + off;
-			mclrefcnt[mtocl(p)]++;
+		if (m->m_flags & M_EXT) {
+			n->m_data = m->m_data + off;
+			mclrefcnt[mtocl(m->m_ext.ext_buf)]++;
+			n->m_ext = m->m_ext;
+			n->m_flags |= M_EXT;
 		} else
 			bcopy(mtod(m, caddr_t)+off, mtod(n, caddr_t),
 			    (unsigned)n->m_len);
@@ -214,20 +281,62 @@ m_copy(m, off, len)
 		m = m->m_next;
 		np = &n->m_next;
 	}
+	if (top == 0)
+		MCFail++;
 	return (top);
 nospace:
 	m_freem(top);
+	MCFail++;
 	return (0);
 }
 
+/*
+ * Copy data from an mbuf chain starting "off" bytes from the beginning,
+ * continuing for "len" bytes, into the indicated buffer.
+ */
+m_copydata(m, off, len, cp)
+	register struct mbuf *m;
+	register int off;
+	register int len;
+	caddr_t cp;
+{
+	register unsigned count;
+
+	if (off < 0 || len < 0)
+		panic("m_copydata");
+	while (off > 0) {
+		if (m == 0)
+			panic("m_copydata");
+		if (off < m->m_len)
+			break;
+		off -= m->m_len;
+		m = m->m_next;
+	}
+	while (len > 0) {
+		if (m == 0)
+			panic("m_copydata");
+		count = MIN(m->m_len - off, len);
+		bcopy(mtod(m, caddr_t) + off, cp, count);
+		len -= count;
+		cp += count;
+		off = 0;
+		m = m->m_next;
+	}
+}
+
+/*
+ * Concatenate mbuf chain n to m.
+ * Both chains must be of the same type (e.g. MT_DATA).
+ * Any m_pkthdr is not updated.
+ */
 m_cat(m, n)
 	register struct mbuf *m, *n;
 {
 	while (m->m_next)
 		m = m->m_next;
 	while (n) {
-		if (m->m_off >= MMAXOFF ||
-		    m->m_off + m->m_len + n->m_len > MMAXOFF) {
+		if (m->m_flags & M_EXT ||
+		    m->m_data + m->m_len + n->m_len >= &m->m_dat[MLEN]) {
 			/* just join the two chains */
 			m->m_next = n;
 			return;
@@ -240,15 +349,20 @@ m_cat(m, n)
 	}
 }
 
-m_adj(mp, len)
+m_adj(mp, req_len)
 	struct mbuf *mp;
-	register int len;
+	int req_len;
 {
-	register struct mbuf *m, *n;
+	register int len = req_len;
+	register struct mbuf *m;
+	register count;
 
 	if ((m = mp) == NULL)
 		return;
 	if (len >= 0) {
+		/*
+		 * Trim from head.
+		 */
 		while (m != NULL && len > 0) {
 			if (m->m_len <= len) {
 				len -= m->m_len;
@@ -256,60 +370,116 @@ m_adj(mp, len)
 				m = m->m_next;
 			} else {
 				m->m_len -= len;
-				m->m_off += len;
-				break;
+				m->m_data += len;
+				len = 0;
 			}
 		}
+		m = mp;
+		if (mp->m_flags & M_PKTHDR)
+			m->m_pkthdr.len -= (req_len - len);
 	} else {
-		/* a 2 pass algorithm might be better */
+		/*
+		 * Trim from tail.  Scan the mbuf chain,
+		 * calculating its length and finding the last mbuf.
+		 * If the adjustment only affects this mbuf, then just
+		 * adjust and return.  Otherwise, rescan and truncate
+		 * after the remaining size.
+		 */
 		len = -len;
-		while (len > 0 && m->m_len != 0) {
-			while (m != NULL && m->m_len != 0) {
-				n = m;
-				m = m->m_next;
-			}
-			if (n->m_len <= len) {
-				len -= n->m_len;
-				n->m_len = 0;
-				m = mp;
-			} else {
-				n->m_len -= len;
+		count = 0;
+		for (;;) {
+			count += m->m_len;
+			if (m->m_next == (struct mbuf *)0)
+				break;
+			m = m->m_next;
+		}
+		if (m->m_len >= len) {
+			m->m_len -= len;
+			if ((mp = m)->m_flags & M_PKTHDR)
+				m->m_pkthdr.len -= len;
+			return;
+		}
+		count -= len;
+		if (count < 0)
+			count = 0;
+		/*
+		 * Correct length for chain is "count".
+		 * Find the mbuf with last data, adjust its length,
+		 * and toss data from remaining mbufs on chain.
+		 */
+		m = mp;
+		if (m->m_flags & M_PKTHDR)
+			m->m_pkthdr.len = count;
+		for (; m; m = m->m_next) {
+			if (m->m_len >= count) {
+				m->m_len = count;
 				break;
 			}
+			count -= m->m_len;
 		}
+		while (m = m->m_next)
+			m->m_len = 0;
 	}
 }
 
+/*
+ * Rearange an mbuf chain so that len bytes are contiguous
+ * and in the data area of an mbuf (so that mtod and dtom
+ * will work for a structure of size len).  Returns the resulting
+ * mbuf chain on success, frees it and returns null on failure.
+ * If there is room, it will add up to max_protohdr-len extra bytes to the
+ * contiguous region in an attempt to avoid being called next time.
+ */
+int MPFail;
+
 struct mbuf *
-m_pullup(m0, len)
-	struct mbuf *m0;
+m_pullup(n, len)
+	register struct mbuf *n;
 	int len;
 {
-	register struct mbuf *m, *n;
-	int count;
+	register struct mbuf *m;
+	register int count;
+	int space;
 
-	n = m0;
-	if (len > MLEN)
-		goto bad;
-	MGET(m, M_DONTWAIT);
-	if (m == 0)
-		goto bad;
-	m->m_len = 0;
+	/*
+	 * If first mbuf has no cluster, and has room for len bytes
+	 * without shifting current data, pullup into it,
+	 * otherwise allocate a new mbuf to prepend to the chain.
+	 */
+	if ((n->m_flags & M_EXT) == 0 &&
+	    n->m_data + len < &n->m_dat[MLEN] && n->m_next) {
+		if (n->m_len >= len)
+			return (n);
+		m = n;
+		n = n->m_next;
+		len -= m->m_len;
+	} else {
+		if (len > MHLEN)
+			goto bad;
+		MGET(m, M_DONTWAIT, n->m_type);
+		if (m == 0)
+			goto bad;
+		m->m_len = 0;
+		if (n->m_flags & M_PKTHDR) {
+			M_COPY_PKTHDR(m, n);
+			n->m_flags &= ~M_PKTHDR;
+		}
+	}
+	space = &m->m_dat[MLEN] - (m->m_data + m->m_len);
 	do {
-		count = MIN(MLEN - m->m_len, len);
-		if (count > n->m_len)
-			count = n->m_len;
-		bcopy(mtod(n, caddr_t), mtod(m, caddr_t)+m->m_len,
+		count = min(min(max(len, max_protohdr), space), n->m_len);
+		bcopy(mtod(n, caddr_t), mtod(m, caddr_t) + m->m_len,
 		  (unsigned)count);
 		len -= count;
 		m->m_len += count;
-		n->m_off += count;
 		n->m_len -= count;
+		space -= count;
 		if (n->m_len)
-			break;
-		n = m_free(n);
-	} while (n);
-	if (len) {
+			n->m_data += count;
+		else
+			n = m_free(n);
+	} while (len > 0 && n);
+	if (len > 0) {
 		(void) m_free(m);
 		goto bad;
 	}
@@ -317,5 +487,140 @@ m_pullup(m0, len)
 	return (m);
 bad:
 	m_freem(n);
+	MPFail++;
 	return (0);
 }
+
+/*
+ * Partition an mbuf chain in two pieces, returning the tail --
+ * all but the first len0 bytes.  In case of failure, it returns NULL and
+ * attempts to restore the chain to its original state.
+ */
+struct mbuf *
+m_split(m0, len0, wait)
+	register struct mbuf *m0;
+	int len0, wait;
+{
+	register struct mbuf *m, *n;
+	unsigned len = len0, remain;
+
+	for (m = m0; m && len > m->m_len; m = m->m_next)
+		len -= m->m_len;
+	if (m == 0)
+		return (0);
+	remain = m->m_len - len;
+	if (m0->m_flags & M_PKTHDR) {
+		MGETHDR(n, wait, m0->m_type);
+		if (n == 0)
+			return (0);
+		n->m_pkthdr.rcvif = m0->m_pkthdr.rcvif;
+		n->m_pkthdr.len = m0->m_pkthdr.len - len0;
+		m0->m_pkthdr.len = len0;
+		if (m->m_flags & M_EXT)
+			goto extpacket;
+		if (remain > MHLEN) {
+			/* m can't be the lead packet */
+			MH_ALIGN(n, 0);
+			n->m_next = m_split(m, len, wait);
+			if (n->m_next == 0) {
+				(void) m_free(n);
+				return (0);
+			} else
+				return (n);
+		} else
+			MH_ALIGN(n, remain);
+	} else if (remain == 0) {
+		n = m->m_next;
+		m->m_next = 0;
+		return (n);
+	} else {
+		MGET(n, wait, m->m_type);
+		if (n == 0)
+			return (0);
+		M_ALIGN(n, remain);
+	}
+extpacket:
+	if (m->m_flags & M_EXT) {
+		n->m_flags |= M_EXT;
+		n->m_ext = m->m_ext;
+		mclrefcnt[mtocl(m->m_ext.ext_buf)]++;
+		m->m_ext.ext_size = 0; /* For Accounting XXXXXX danger */
+		n->m_data = m->m_data + len;
+	} else {
+		bcopy(mtod(m, caddr_t) + len, mtod(n, caddr_t), remain);
+	}
+	n->m_len = remain;
+	m->m_len = len;
+	n->m_next = m->m_next;
+	m->m_next = 0;
+	return (n);
+}
+/*
+ * Routine to copy from device local memory into mbufs.
+ */
+struct mbuf *
+m_devget(buf, totlen, off0, ifp, copy)
+	char *buf;
+	int totlen, off0;
+	struct ifnet *ifp;
+	void (*copy)();
+{
+	register struct mbuf *m;
+	struct mbuf *top = 0, **mp = &top;
+	register int off = off0, len;
+	register char *cp;
+	char *epkt;
+
+	cp = buf;
+	epkt = cp + totlen;
+	if (off) {
+		cp += off + 2 * sizeof(u_short);
+		totlen -= 2 * sizeof(u_short);
+	}
+	MGETHDR(m, M_DONTWAIT, MT_DATA);
+	if (m == 0)
+		return (0);
+	m->m_pkthdr.rcvif = ifp;
+	m->m_pkthdr.len = totlen;
+	m->m_len = MHLEN;
+
+	while (totlen > 0) {
+		if (top) {
+			MGET(m, M_DONTWAIT, MT_DATA);
+			if (m == 0) {
+				m_freem(top);
+				return (0);
+			}
+			m->m_len = MLEN;
+		}
+		len = min(totlen, epkt - cp);
+		if (len >= MINCLSIZE) {
+			MCLGET(m, M_DONTWAIT);
+			if (m->m_flags & M_EXT)
+				m->m_len = len = min(len, MCLBYTES);
+			else
+				len = m->m_len;
+		} else {
+			/*
+			 * Place initial small packet/header at end of mbuf.
+			 */
+			if (len < m->m_len) {
+				if (top == 0 && len + max_linkhdr <= m->m_len)
+					m->m_data += max_linkhdr;
+				m->m_len = len;
+			} else
+				len = m->m_len;
+		}
+		if (copy)
+			copy(cp, mtod(m, caddr_t), (unsigned)len);
+		else
+			bcopy(cp, mtod(m, caddr_t), (unsigned)len);
+		cp += len;
+		*mp = m;
+		mp = &m->m_next;
+		totlen -= len;
+		if (cp == epkt)
+			cp = buf;
+	}
+	return (top);
+}