1994-05-27 12:33:43 +00:00
|
|
|
/*
|
|
|
|
* Copyright (c) 1983, 1988, 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.
|
|
|
|
*/
|
|
|
|
|
1997-07-29 06:51:41 +00:00
|
|
|
#if 0
|
2004-07-26 20:18:11 +00:00
|
|
|
#ifndef lint
|
1994-05-27 12:33:43 +00:00
|
|
|
static char sccsid[] = "@(#)mbuf.c 8.1 (Berkeley) 6/6/93";
|
|
|
|
#endif /* not lint */
|
2004-07-26 20:18:11 +00:00
|
|
|
#endif
|
|
|
|
|
|
|
|
#include <sys/cdefs.h>
|
|
|
|
__FBSDID("$FreeBSD$");
|
1994-05-27 12:33:43 +00:00
|
|
|
|
|
|
|
#include <sys/param.h>
|
1997-02-24 20:40:40 +00:00
|
|
|
#include <sys/mbuf.h>
|
1994-05-27 12:33:43 +00:00
|
|
|
#include <sys/protosw.h>
|
|
|
|
#include <sys/socket.h>
|
1997-02-24 20:40:40 +00:00
|
|
|
#include <sys/sysctl.h>
|
1994-05-27 12:33:43 +00:00
|
|
|
|
1997-07-29 06:51:41 +00:00
|
|
|
#include <err.h>
|
1994-05-27 12:33:43 +00:00
|
|
|
#include <stdio.h>
|
2000-07-15 06:02:48 +00:00
|
|
|
#include <stdlib.h>
|
2001-06-22 06:35:32 +00:00
|
|
|
#include <string.h>
|
1994-05-27 12:33:43 +00:00
|
|
|
#include "netstat.h"
|
|
|
|
|
|
|
|
#define YES 1
|
|
|
|
typedef int bool;
|
|
|
|
|
2000-07-15 06:02:48 +00:00
|
|
|
static struct mbtypenames {
|
2001-09-30 01:58:39 +00:00
|
|
|
short mt_type;
|
2002-09-05 17:06:51 +00:00
|
|
|
const char *mt_name;
|
2000-07-15 06:02:48 +00:00
|
|
|
} mbtypenames[] = {
|
1994-05-27 12:33:43 +00:00
|
|
|
{ MT_DATA, "data" },
|
|
|
|
{ MT_OOBDATA, "oob data" },
|
|
|
|
{ MT_CONTROL, "ancillary data" },
|
|
|
|
{ MT_HEADER, "packet headers" },
|
1997-08-17 09:12:43 +00:00
|
|
|
#ifdef MT_SOCKET
|
1994-05-27 12:33:43 +00:00
|
|
|
{ MT_SOCKET, "socket structures" }, /* XXX */
|
1997-08-17 09:12:43 +00:00
|
|
|
#endif
|
|
|
|
#ifdef MT_PCB
|
1994-05-27 12:33:43 +00:00
|
|
|
{ MT_PCB, "protocol control blocks" }, /* XXX */
|
1997-08-17 09:12:43 +00:00
|
|
|
#endif
|
1999-01-18 12:40:44 +00:00
|
|
|
#ifdef MT_RTABLE
|
1994-05-27 12:33:43 +00:00
|
|
|
{ MT_RTABLE, "routing table entries" }, /* XXX */
|
1999-01-18 12:40:44 +00:00
|
|
|
#endif
|
1997-08-17 09:12:43 +00:00
|
|
|
#ifdef MT_HTABLE
|
1994-05-27 12:33:43 +00:00
|
|
|
{ MT_HTABLE, "IMP host table entries" }, /* XXX */
|
1997-08-17 09:12:43 +00:00
|
|
|
#endif
|
|
|
|
#ifdef MT_ATABLE
|
1994-05-27 12:33:43 +00:00
|
|
|
{ MT_ATABLE, "address resolution tables" },
|
1997-08-17 09:12:43 +00:00
|
|
|
#endif
|
1994-05-27 12:33:43 +00:00
|
|
|
{ MT_FTABLE, "fragment reassembly queue headers" }, /* XXX */
|
|
|
|
{ MT_SONAME, "socket names and addresses" },
|
1998-08-23 15:23:11 +00:00
|
|
|
#ifdef MT_SOOPTS
|
1994-05-27 12:33:43 +00:00
|
|
|
{ MT_SOOPTS, "socket options" },
|
1998-08-23 15:23:11 +00:00
|
|
|
#endif
|
1997-08-17 09:12:43 +00:00
|
|
|
#ifdef MT_RIGHTS
|
1994-05-27 12:33:43 +00:00
|
|
|
{ MT_RIGHTS, "access rights" },
|
1997-08-17 09:12:43 +00:00
|
|
|
#endif
|
|
|
|
#ifdef MT_IFADDR
|
1994-05-27 12:33:43 +00:00
|
|
|
{ MT_IFADDR, "interface addresses" }, /* XXX */
|
1997-08-17 09:12:43 +00:00
|
|
|
#endif
|
1994-05-27 12:33:43 +00:00
|
|
|
{ 0, 0 }
|
|
|
|
};
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Print mbuf statistics.
|
|
|
|
*/
|
|
|
|
void
|
2002-09-05 17:06:51 +00:00
|
|
|
mbpr(u_long mbaddr, u_long mbtaddr __unused, u_long nmbcaddr, u_long nmbufaddr,
|
o Allow "buckets" in mb_alloc to be differently sized (according to
compile-time constants). That is, a "bucket" now is not necessarily
a page-worth of mbufs or clusters, but it is MBUF_BUCK_SZ, CLUS_BUCK_SZ
worth of mbufs, clusters.
o Rename {mbuf,clust}_limit to {mbuf,clust}_hiwm and introduce
{mbuf,clust}_lowm, which currently has no effect but will be used
to set the low watermarks.
o Fix netstat so that it can deal with the differently-sized buckets
and teach it about the low watermarks too.
o Make sure the per-cpu stats for an absent CPU has mb_active set to 0,
explicitly.
o Get rid of the allocate refcounts from mbuf map mess. Instead,
just malloc() the refcounts in one shot from mbuf_init()
o Clean up / update comments in subr_mbuf.c
2003-02-20 04:26:58 +00:00
|
|
|
u_long mbhiaddr, u_long clhiaddr, u_long mbloaddr, u_long clloaddr,
|
|
|
|
u_long cpusaddr __unused, u_long pgsaddr, u_long mbpaddr)
|
1994-05-27 12:33:43 +00:00
|
|
|
{
|
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
|
|
|
int i, nmbclusters;
|
2003-12-27 09:02:31 +00:00
|
|
|
int nsfbufs, nsfbufspeak, nsfbufsused;
|
2001-09-30 01:58:39 +00:00
|
|
|
short nmbtypes;
|
2001-06-22 06:35:32 +00:00
|
|
|
size_t mlen;
|
2001-09-30 01:58:39 +00:00
|
|
|
long *mbtypes = NULL;
|
2001-06-22 06:35:32 +00:00
|
|
|
struct mbstat *mbstat = NULL;
|
|
|
|
struct mbtypenames *mp;
|
|
|
|
bool *seen = NULL;
|
1994-05-27 12:33:43 +00:00
|
|
|
|
2001-06-23 17:04:17 +00:00
|
|
|
mlen = sizeof *mbstat;
|
2001-06-22 06:35:32 +00:00
|
|
|
if ((mbstat = malloc(mlen)) == NULL) {
|
|
|
|
warn("malloc: cannot allocate memory for mbstat");
|
|
|
|
goto err;
|
|
|
|
}
|
|
|
|
|
2001-06-14 15:45:09 +00:00
|
|
|
if (mbaddr) {
|
2001-06-22 06:35:32 +00:00
|
|
|
if (kread(mbaddr, (char *)mbstat, sizeof mbstat))
|
2001-06-14 15:45:09 +00:00
|
|
|
goto err;
|
|
|
|
if (kread(nmbcaddr, (char *)&nmbclusters, sizeof(int)))
|
|
|
|
goto err;
|
|
|
|
} else {
|
2001-06-23 17:04:17 +00:00
|
|
|
mlen = sizeof *mbstat;
|
2001-06-22 06:35:32 +00:00
|
|
|
if (sysctlbyname("kern.ipc.mbstat", mbstat, &mlen, NULL, 0)
|
|
|
|
< 0) {
|
2001-06-14 15:45:09 +00:00
|
|
|
warn("sysctl: retrieving mbstat");
|
|
|
|
goto err;
|
|
|
|
}
|
2001-06-22 06:35:32 +00:00
|
|
|
mlen = sizeof(int);
|
|
|
|
if (sysctlbyname("kern.ipc.nmbclusters", &nmbclusters, &mlen,
|
|
|
|
NULL, 0) < 0) {
|
2001-06-14 15:45:09 +00:00
|
|
|
warn("sysctl: retrieving nmbclusters");
|
|
|
|
goto err;
|
|
|
|
}
|
1999-12-28 06:38:37 +00:00
|
|
|
}
|
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 (mbstat->m_mbufs < 0) mbstat->m_mbufs = 0; /* XXX */
|
|
|
|
if (mbstat->m_mclusts < 0) mbstat->m_mclusts = 0; /* XXX */
|
1999-12-28 06:38:37 +00:00
|
|
|
|
2001-09-30 01:58:39 +00:00
|
|
|
nmbtypes = mbstat->m_numtypes;
|
|
|
|
if ((seen = calloc(nmbtypes, sizeof(*seen))) == NULL) {
|
|
|
|
warn("calloc: cannot allocate memory for mbtypes seen flag");
|
|
|
|
goto err;
|
|
|
|
}
|
|
|
|
if ((mbtypes = calloc(nmbtypes, sizeof(long *))) == NULL) {
|
|
|
|
warn("calloc: cannot allocate memory for mbtypes");
|
|
|
|
goto err;
|
|
|
|
}
|
|
|
|
|
1997-02-24 20:40:40 +00:00
|
|
|
#undef MSIZE
|
2001-06-22 06:35:32 +00:00
|
|
|
#define MSIZE (mbstat->m_msize)
|
1997-02-24 20:40:40 +00:00
|
|
|
#undef MCLBYTES
|
2001-06-22 06:35:32 +00:00
|
|
|
#define MCLBYTES (mbstat->m_mclbytes)
|
|
|
|
|
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
|
|
|
printf("%lu mbufs in use\n", mbstat->m_mbufs);
|
2003-12-23 13:24:03 +00:00
|
|
|
|
2001-09-30 01:58:39 +00:00
|
|
|
for (mp = mbtypenames; mp->mt_name; mp++) {
|
|
|
|
if (mbtypes[mp->mt_type]) {
|
|
|
|
seen[mp->mt_type] = YES;
|
|
|
|
printf("\t %lu mbufs allocated to %s\n",
|
|
|
|
mbtypes[mp->mt_type], mp->mt_name);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
for (i = 1; i < nmbtypes; i++) {
|
|
|
|
if (!seen[i] && mbtypes[i])
|
|
|
|
printf("\t %lu mbufs allocated to <mbuf type: %d>\n",
|
|
|
|
mbtypes[i], i);
|
|
|
|
}
|
2001-06-22 06:35:32 +00:00
|
|
|
|
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
|
|
|
printf("%lu/%d mbuf clusters in use (current/max)\n",
|
|
|
|
mbstat->m_mclusts, nmbclusters);
|
|
|
|
|
2003-12-27 07:58:31 +00:00
|
|
|
mlen = sizeof(nsfbufs);
|
2004-03-11 10:19:45 +00:00
|
|
|
if (!sysctlbyname("kern.ipc.nsfbufs", &nsfbufs, &mlen, NULL, 0) &&
|
2003-12-27 07:58:31 +00:00
|
|
|
!sysctlbyname("kern.ipc.nsfbufsused", &nsfbufsused, &mlen, NULL,
|
2004-03-11 10:19:45 +00:00
|
|
|
0) &&
|
2003-12-27 07:58:31 +00:00
|
|
|
!sysctlbyname("kern.ipc.nsfbufspeak", &nsfbufspeak, &mlen, NULL,
|
2004-03-11 10:19:45 +00:00
|
|
|
0)) {
|
2003-12-27 09:02:31 +00:00
|
|
|
printf("%d/%d/%d sfbufs in use (current/peak/max)\n",
|
|
|
|
nsfbufsused, nsfbufspeak, nsfbufs);
|
2003-12-27 07:58:31 +00:00
|
|
|
}
|
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
|
|
|
printf("%lu KBytes allocated to network\n", (mbstat->m_mbufs * MSIZE +
|
|
|
|
mbstat->m_mclusts * MCLBYTES) / 1024);
|
2003-12-28 08:59:41 +00:00
|
|
|
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",
|
|
|
|
mbstat->sf_iocnt);
|
2001-06-22 06:35:32 +00:00
|
|
|
printf("%lu calls to protocol drain routines\n", mbstat->m_drain);
|
2000-07-15 06:02:48 +00:00
|
|
|
|
|
|
|
err:
|
|
|
|
if (mbtypes != NULL)
|
|
|
|
free(mbtypes);
|
|
|
|
if (seen != NULL)
|
|
|
|
free(seen);
|
2001-06-22 06:35:32 +00:00
|
|
|
if (mbstat != NULL)
|
|
|
|
free(mbstat);
|
1994-05-27 12:33:43 +00:00
|
|
|
}
|