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)

4 files changed, 9 insertions, 200 deletions

diff --git a/usr.bin/netstat/main.c b/usr.bin/netstat/main.c
index ac9dd26..8992599 100644
--- a/usr.bin/netstat/main.c
+++ b/usr.bin/netstat/main.c
@@ -256,7 +256,6 @@ static char *nlistf = NULL, *memf = NULL;
 int	Aflag;		/* show addresses of protocol control block */
 int	aflag;		/* show all sockets (including servers) */
 int	bflag;		/* show i/f total bytes in/out */
-int	cflag;		/* show mbuf cache information */
 int	dflag;		/* show i/f dropped packets */
 int	gflag;		/* show group (multicast) routing or stats */
 int	iflag;		/* show interfaces */
@@ -297,9 +296,6 @@ main(int argc, char *argv[])
 		case 'b':
 			bflag = 1;
 			break;
-		case 'c':
-			cflag = 1;
-			break;
 		case 'd':
 			dflag = 1;
 			break;
@@ -425,10 +421,6 @@ main(int argc, char *argv[])
 	if (nlistf != NULL || memf != NULL)
 		setgid(getgid());
 
-	if (cflag && !mflag) {
-		(void)fprintf(stderr, "-c only valid with -m\n");
-		usage();
-	}
 	if (mflag) {
 		if (memf != NULL) {
 			if (kread(0, 0, 0) == 0)
diff --git a/usr.bin/netstat/mbuf.c b/usr.bin/netstat/mbuf.c
index aa6a8d2..98546c4 100644
--- a/usr.bin/netstat/mbuf.c
+++ b/usr.bin/netstat/mbuf.c
@@ -99,17 +99,12 @@ mbpr(u_long mbaddr, u_long mbtaddr __unused, u_long nmbcaddr, u_long nmbufaddr,
     u_long mbhiaddr, u_long clhiaddr, u_long mbloaddr, u_long clloaddr,
     u_long cpusaddr __unused, u_long pgsaddr, u_long mbpaddr)
 {
-	int i, j, nmbufs, nmbclusters, page_size, num_objs;
+	int i, nmbclusters;
 	int nsfbufs, nsfbufspeak, nsfbufsused;
-	u_int mbuf_hiwm, clust_hiwm, mbuf_lowm, clust_lowm;
-	u_long totspace[2], totused[2];
-	u_long gentotnum, gentotfree, totnum, totfree;
-	u_long totmem, totmemalloced, totmemused;
 	short nmbtypes;
 	size_t mlen;
 	long *mbtypes = NULL;
 	struct mbstat *mbstat = NULL;
-	struct mbpstat **mbpstat = NULL;
 	struct mbtypenames *mp;
 	bool *seen = NULL;
 
@@ -119,50 +114,12 @@ mbpr(u_long mbaddr, u_long mbtaddr __unused, u_long nmbcaddr, u_long nmbufaddr,
 		goto err;
 	}
 
-	/*
-	 * XXX: Unfortunately, for the time being, we have to fetch
-	 * the total length of the per-CPU stats area via sysctl
-	 * (regardless of whether we're looking at a core or not.
-	 */
-	if (sysctlbyname("kern.ipc.mb_statpcpu", NULL, &mlen, NULL, 0) < 0) {
-		warn("sysctl: retrieving mb_statpcpu len");
-		goto err;
-	} 
-	num_objs = (int)(mlen / sizeof(struct mbpstat));
-	if ((mbpstat = calloc(num_objs, sizeof(struct mbpstat *))) == NULL) {
-		warn("calloc: cannot allocate memory for mbpstats pointers");
-		goto err;
-	}
-	if ((mbpstat[0] = calloc(num_objs, sizeof(struct mbpstat))) == NULL) {
-		warn("calloc: cannot allocate memory for mbpstats");
-		goto err;
-	}
-
 	if (mbaddr) {
-		if (kread(mbpaddr, (char *)mbpstat[0], mlen))
-			goto err; 
 		if (kread(mbaddr, (char *)mbstat, sizeof mbstat))
 			goto err;
 		if (kread(nmbcaddr, (char *)&nmbclusters, sizeof(int)))
 			goto err;
-		if (kread(nmbufaddr, (char *)&nmbufs, sizeof(int)))
-			goto err;
-		if (kread(mbhiaddr, (char *)&mbuf_hiwm, sizeof(u_int)))
-			goto err;
-		if (kread(clhiaddr, (char *)&clust_hiwm, sizeof(u_int)))
-			goto err;
-		if (kread(mbloaddr, (char *)&mbuf_lowm, sizeof(u_int)))
-			goto err;
-		if (kread(clloaddr, (char *)&clust_lowm, sizeof(u_int)))
-			goto err;
-		if (kread(pgsaddr, (char *)&page_size, sizeof(int)))
-			goto err;
 	} else {
-		if (sysctlbyname("kern.ipc.mb_statpcpu", mbpstat[0], &mlen,
-		    NULL, 0) < 0) {
-			warn("sysctl: retrieving mb_statpcpu");
-			goto err;
-		}
 		mlen = sizeof *mbstat;
 		if (sysctlbyname("kern.ipc.mbstat", mbstat, &mlen, NULL, 0)
 		    < 0) {
@@ -175,43 +132,9 @@ mbpr(u_long mbaddr, u_long mbtaddr __unused, u_long nmbcaddr, u_long nmbufaddr,
 			warn("sysctl: retrieving nmbclusters");
 			goto err;
 		}
-		mlen = sizeof(int);
-		if (sysctlbyname("kern.ipc.nmbufs", &nmbufs, &mlen, NULL, 0)
-		    < 0) {
-			warn("sysctl: retrieving nmbufs");
-			goto err;
-		}
-		mlen = sizeof(u_int);
-		if (sysctlbyname("kern.ipc.mbuf_hiwm", &mbuf_hiwm, &mlen,
-		    NULL, 0) < 0) {
-			warn("sysctl: retrieving mbuf_hiwm");
-			goto err;
-		}
-		mlen = sizeof(u_int);
-		if (sysctlbyname("kern.ipc.clust_hiwm", &clust_hiwm, &mlen,
-		    NULL, 0) < 0) {
-			warn("sysctl: retrieving clust_hiwm");
-			goto err;
-		}
-		mlen = sizeof(u_int);
-		if (sysctlbyname("kern.ipc.mbuf_lowm", &mbuf_lowm, &mlen,
-		    NULL, 0) < 0) {
-			warn("sysctl: retrieving mbuf_lowm");
-			goto err;
-		}
-		mlen = sizeof(u_int);
-		if (sysctlbyname("kern.ipc.clust_lowm", &clust_lowm, &mlen,
-		    NULL, 0) < 0) {
-			warn("sysctl: retrieving clust_lowm");
-			goto err;
-		}
-		mlen = sizeof(int);
-		if (sysctlbyname("hw.pagesize", &page_size, &mlen, NULL, 0)
-		    < 0) {
-			warn("sysctl: retrieving hw.pagesize");
-			goto err;
-		}
 	}
+	if (mbstat->m_mbufs < 0) mbstat->m_mbufs = 0;		/* XXX */
+	if (mbstat->m_mclusts < 0) mbstat->m_mclusts = 0;	/* XXX */
 
 	nmbtypes = mbstat->m_numtypes;
 	if ((seen = calloc(nmbtypes, sizeof(*seen))) == NULL) {
@@ -223,59 +146,13 @@ mbpr(u_long mbaddr, u_long mbtaddr __unused, u_long nmbcaddr, u_long nmbufaddr,
 		goto err;
 	}
 
-	for (i = 0; i < num_objs; i++)
-		mbpstat[i] = mbpstat[0] + i;
-
 #undef MSIZE
 #define MSIZE		(mbstat->m_msize)
 #undef MCLBYTES
 #define	MCLBYTES	(mbstat->m_mclbytes)
-#define	GENLST		(num_objs - 1)
 
-	totnum = mbpstat[GENLST]->mb_mbbucks * mbstat->m_mbperbuck;
-	totfree = mbpstat[GENLST]->mb_mbfree;
-	for (j = 1; j < nmbtypes; j++)
-		mbtypes[j] += mbpstat[GENLST]->mb_mbtypes[j];
-	totspace[0] = mbpstat[GENLST]->mb_mbbucks * mbstat->m_mbperbuck * MSIZE;
-	for (i = 0; i < (num_objs - 1); i++) {
-		if (mbpstat[i]->mb_active == 0)
-			continue;
-		totspace[0] += mbpstat[i]->mb_mbbucks*mbstat->m_mbperbuck*MSIZE;
-		totnum += mbpstat[i]->mb_mbbucks * mbstat->m_mbperbuck;
-		totfree += mbpstat[i]->mb_mbfree;
-		for (j = 1; j < nmbtypes; j++)
-			mbtypes[j] += mbpstat[i]->mb_mbtypes[j]; 
-	}
-	totused[0] = totnum - totfree;
-	if (cflag) {
-		printf("mbuf usage:\n"
-		    "\tTotal:\t\t%lu/%lu/%d (in use/in pool/max)\n",
-		    totused[0], totnum, nmbufs);
-		gentotnum = mbpstat[GENLST]->mb_mbbucks * mbstat->m_mbperbuck;
-		gentotfree = mbpstat[GENLST]->mb_mbfree;
-		printf("\tGEN cache:\t%lu/%lu (in use/in pool)\n",
-		    gentotnum - gentotfree, gentotnum);
-	} else {
-		/* XXX: peak is now wrong. */
-		printf("%lu/%lu/%d mbufs in use (current/peak/max):\n",
-		    totused[0], totnum, nmbufs);
-	}
+	printf("%lu mbufs in use\n", mbstat->m_mbufs);
 
-	for (i = 0; cflag && i < (num_objs - 1); i++) {
-		if (mbpstat[i]->mb_active == 0)
-			continue;
-		printf("\tCPU #%d cache:\t%lu/%lu (in use/in pool)\n",
-		    i,
-		    (mbpstat[i]->mb_mbbucks * mbstat->m_mbperbuck -
-		     mbpstat[i]->mb_mbfree),
-		    (mbpstat[i]->mb_mbbucks * mbstat->m_mbperbuck));
-	}
-	if (cflag) {
-		printf("\tMbuf cache high watermark: %d\n", mbuf_hiwm);
-#ifdef NOTYET
-		printf("\tMbuf cache low watermark: %d\n", mbuf_lowm);
-#endif
-	}
 	for (mp = mbtypenames; mp->mt_name; mp++) {
 		if (mbtypes[mp->mt_type]) {
 			seen[mp->mt_type] = YES;
@@ -288,53 +165,10 @@ mbpr(u_long mbaddr, u_long mbtaddr __unused, u_long nmbcaddr, u_long nmbufaddr,
 			printf("\t  %lu mbufs allocated to <mbuf type: %d>\n",
 			    mbtypes[i], i);
 	}
-	if (cflag)
-		printf("\t%.1f%% of mbuf map consumed\n",
-		    totspace[0] * 100.0 / (nmbufs * MSIZE));
 
-	totnum = mbpstat[GENLST]->mb_clbucks * mbstat->m_clperbuck;
-	totfree = mbpstat[GENLST]->mb_clfree;
-	totspace[1] = mbpstat[GENLST]->mb_clbucks*mbstat->m_clperbuck*MCLBYTES;
-	for (i = 0; i < (num_objs - 1); i++) {
-		if (mbpstat[i]->mb_active == 0)
-			continue;
-		totspace[1] += mbpstat[i]->mb_clbucks * mbstat->m_clperbuck
-		    * MCLBYTES;
-		totnum += mbpstat[i]->mb_clbucks * mbstat->m_clperbuck;
-		totfree += mbpstat[i]->mb_clfree;
-	}
-	totused[1] = totnum - totfree;
-	if (cflag) {
-		printf("mbuf cluster usage:\n"
-		    "\tTotal:\t\t%lu/%lu/%d (in use/in pool/max)\n",
-		    totused[1], totnum, nmbclusters);
-		gentotnum = mbpstat[GENLST]->mb_clbucks * mbstat->m_clperbuck;
-		gentotfree = mbpstat[GENLST]->mb_clfree;
-		printf("\tGEN cache:\t%lu/%lu (in use/in pool)\n",
-		    gentotnum - gentotfree, gentotnum);
-	} else {
-		/* XXX: peak is now wrong. */
-		printf("%lu/%lu/%d mbuf clusters in use (current/peak/max)\n",
-		    totused[1], totnum, nmbclusters);
-	}
-	for (i = 0; cflag && i < (num_objs - 1); i++) {
-		if (mbpstat[i]->mb_active == 0)
-			continue;
-		printf("\tCPU #%d cache:\t%lu/%lu (in use/in pool)\n",
-		    i,
-		    (mbpstat[i]->mb_clbucks * mbstat->m_clperbuck -
-		     mbpstat[i]->mb_clfree),
-		    (mbpstat[i]->mb_clbucks * mbstat->m_clperbuck));
-	}
-	if (cflag) {
-		printf("\tCluster cache high watermark: %d\n", clust_hiwm);
-#ifdef NOTYET
-		printf("\tCluster cache low watermark: %d\n", clust_lowm);
-#endif
-	}
-	if (cflag)
-		printf("\t%.1f%% of cluster map consumed\n",
-		    totspace[1] * 100.0 / (nmbclusters * MCLBYTES));
+	printf("%lu/%d mbuf clusters in use (current/max)\n",
+	    mbstat->m_mclusts, nmbclusters);
+
 	mlen = sizeof(nsfbufs);
 	if (!sysctlbyname("kern.ipc.nsfbufs", &nsfbufs, &mlen, NULL, 0) &&
 	    !sysctlbyname("kern.ipc.nsfbufsused", &nsfbufsused, &mlen, NULL,
@@ -344,15 +178,8 @@ mbpr(u_long mbaddr, u_long mbtaddr __unused, u_long nmbcaddr, u_long nmbufaddr,
 		printf("%d/%d/%d sfbufs in use (current/peak/max)\n",
 		    nsfbufsused, nsfbufspeak, nsfbufs);
 	}
-	totmem = nmbufs * MSIZE + nmbclusters * MCLBYTES;
-	totmemalloced = totspace[0] + totspace[1];
-	totmemused = totused[0] * MSIZE + totused[1] * MCLBYTES;
-	printf(
-	    "%lu KBytes allocated to network (%.1f%% in use, %.1f%% wired)\n",
-	    totmem / 1024, totmemused * 100.0 / totmem,
-	    totmemalloced * 100.0 / totmem);
-	printf("%lu requests for memory denied\n", mbstat->m_drops);
-	printf("%lu requests for memory delayed\n", mbstat->m_wait);
+	printf("%lu KBytes allocated to network\n", (mbstat->m_mbufs * MSIZE +
+	    mbstat->m_mclusts * MCLBYTES) / 1024);
 	printf("%lu requests for sfbufs denied\n", mbstat->sf_allocfail);
 	printf("%lu requests for sfbufs delayed\n", mbstat->sf_allocwait);
 	printf("%lu requests for I/O initiated by sendfile\n",
@@ -366,9 +193,4 @@ err:
 		free(seen);
 	if (mbstat != NULL)
 		free(mbstat);
-	if (mbpstat != NULL) {
-		if (mbpstat[0] != NULL)
-			free(mbpstat[0]);
-		free(mbpstat);
-	}
 }
diff --git a/usr.bin/netstat/netstat.1 b/usr.bin/netstat/netstat.1
index 45023fe..32edfec 100644
--- a/usr.bin/netstat/netstat.1
+++ b/usr.bin/netstat/netstat.1
@@ -181,7 +181,6 @@ or for a single
 .Bk -words
 .Nm
 .Fl m
-.Op Fl c
 .Op Fl M Ar core
 .Op Fl N Ar system
 .Ek
@@ -189,9 +188,6 @@ or for a single
 Show statistics recorded by the memory management routines
 .Pq Xr mbuf 9 .
 The network manages a private pool of memory buffers.
-The
-.Fl c
-option shows per-CPU statistics for caching.
 .It Xo
 .Bk -words
 .Nm
diff --git a/usr.bin/netstat/netstat.h b/usr.bin/netstat/netstat.h
index c59b7e8..e2b3f29 100644
--- a/usr.bin/netstat/netstat.h
+++ b/usr.bin/netstat/netstat.h
@@ -39,7 +39,6 @@
 extern int	Aflag;	/* show addresses of protocol control block */
 extern int	aflag;	/* show all sockets (including servers) */
 extern int	bflag;	/* show i/f total bytes in/out */
-extern int	cflag;	/* show mbuf cache information */
 extern int	dflag;	/* show i/f dropped packets */
 extern int	gflag;	/* show group (multicast) routing or stats */
 extern int	iflag;	/* show interfaces */