freebsd-dev/usr.bin/netstat/mbuf.c

360 lines
11 KiB
C
Raw Normal View History

1994-05-27 12:33:43 +00:00
/*
* Copyright (c) 1983, 1988, 1993
* The Regents of the University of California.
* Copyright (c) 2005 Robert N. M. Watson
* All rights reserved.
1994-05-27 12:33:43 +00:00
*
* 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.
*/
#if 0
#ifndef lint
1994-05-27 12:33:43 +00:00
static char sccsid[] = "@(#)mbuf.c 8.1 (Berkeley) 6/6/93";
#endif /* not lint */
#endif
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
1994-05-27 12:33:43 +00:00
#include <sys/param.h>
#include <sys/mbuf.h>
1994-05-27 12:33:43 +00:00
#include <sys/protosw.h>
#include <sys/socket.h>
#include <sys/sysctl.h>
1994-05-27 12:33:43 +00:00
#include <err.h>
#include <memstat.h>
1994-05-27 12:33:43 +00:00
#include <stdio.h>
#include <stdlib.h>
Introduce numerous SMP friendly changes to the mbuf allocator. Namely, introduce a modified allocation mechanism for mbufs and mbuf clusters; one which can scale under SMP and which offers the possibility of resource reclamation to be implemented in the future. Notable advantages: o Reduce contention for SMP by offering per-CPU pools and locks. o Better use of data cache due to per-CPU pools. o Much less code cache pollution due to excessively large allocation macros. o Framework for `grouping' objects from same page together so as to be able to possibly free wired-down pages back to the system if they are no longer needed by the network stacks. Additional things changed with this addition: - Moved some mbuf specific declarations and initializations from sys/conf/param.c into mbuf-specific code where they belong. - m_getclr() has been renamed to m_get_clrd() because the old name is really confusing. m_getclr() HAS been preserved though and is defined to the new name. No tree sweep has been done "to change the interface," as the old name will continue to be supported and is not depracated. The change was merely done because m_getclr() sounds too much like "m_get a cluster." - TEMPORARILY disabled mbtypes statistics displaying in netstat(1) and systat(1) (see TODO below). - Fixed systat(1) to display number of "free mbufs" based on new per-CPU stat structures. - Fixed netstat(1) to display new per-CPU stats based on sysctl-exported per-CPU stat structures. All infos are fetched via sysctl. TODO (in order of priority): - Re-enable mbtypes statistics in both netstat(1) and systat(1) after introducing an SMP friendly way to collect the mbtypes stats under the already introduced per-CPU locks (i.e. hopefully don't use atomic() - it seems too costly for a mere stat update, especially when other locks are already present). - Optionally have systat(1) display not only "total free mbufs" but also "total free mbufs per CPU pool." - Fix minor length-fetching issues in netstat(1) related to recently re-enabled option to read mbuf stats from a core file. - Move reference counters at least for mbuf clusters into an unused portion of the cluster itself, to save space and need to allocate a counter. - Look into introducing resource freeing possibly from a kproc. Reviewed by (in parts): jlemon, jake, silby, terry Tested by: jlemon (Intel & Alpha), mjacob (Intel & Alpha) Preliminary performance measurements: jlemon (and me, obviously) URL: http://people.freebsd.org/~bmilekic/mb_alloc/
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;
static struct mbtypenames {
short mt_type;
const char *mt_name;
} mbtypenames[] = {
1994-05-27 12:33:43 +00:00
{ MT_DATA, "data" },
{ MT_OOBDATA, "oob data" },
{ MT_CONTROL, "ancillary data" },
{ MT_HEADER, "packet headers" },
#ifdef MT_SOCKET
1994-05-27 12:33:43 +00:00
{ MT_SOCKET, "socket structures" }, /* XXX */
#endif
#ifdef MT_PCB
1994-05-27 12:33:43 +00:00
{ MT_PCB, "protocol control blocks" }, /* XXX */
#endif
#ifdef MT_RTABLE
1994-05-27 12:33:43 +00:00
{ MT_RTABLE, "routing table entries" }, /* XXX */
#endif
#ifdef MT_HTABLE
1994-05-27 12:33:43 +00:00
{ MT_HTABLE, "IMP host table entries" }, /* XXX */
#endif
#ifdef MT_ATABLE
1994-05-27 12:33:43 +00:00
{ MT_ATABLE, "address resolution tables" },
#endif
1994-05-27 12:33:43 +00:00
{ MT_FTABLE, "fragment reassembly queue headers" }, /* XXX */
{ MT_SONAME, "socket names and addresses" },
#ifdef MT_SOOPTS
1994-05-27 12:33:43 +00:00
{ MT_SOOPTS, "socket options" },
#endif
#ifdef MT_RIGHTS
1994-05-27 12:33:43 +00:00
{ MT_RIGHTS, "access rights" },
#endif
#ifdef MT_IFADDR
1994-05-27 12:33:43 +00:00
{ MT_IFADDR, "interface addresses" }, /* XXX */
#endif
1994-05-27 12:33:43 +00:00
{ 0, 0 }
};
/*
* Print mbuf statistics extracted from kmem.
1994-05-27 12:33:43 +00:00
*/
void
mbpr_kmem(u_long mbaddr, u_long nmbcaddr, u_long nmbufaddr, u_long mbhiaddr,
u_long clhiaddr, u_long mbloaddr, u_long clloaddr, 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;
short nmbtypes;
Introduce numerous SMP friendly changes to the mbuf allocator. Namely, introduce a modified allocation mechanism for mbufs and mbuf clusters; one which can scale under SMP and which offers the possibility of resource reclamation to be implemented in the future. Notable advantages: o Reduce contention for SMP by offering per-CPU pools and locks. o Better use of data cache due to per-CPU pools. o Much less code cache pollution due to excessively large allocation macros. o Framework for `grouping' objects from same page together so as to be able to possibly free wired-down pages back to the system if they are no longer needed by the network stacks. Additional things changed with this addition: - Moved some mbuf specific declarations and initializations from sys/conf/param.c into mbuf-specific code where they belong. - m_getclr() has been renamed to m_get_clrd() because the old name is really confusing. m_getclr() HAS been preserved though and is defined to the new name. No tree sweep has been done "to change the interface," as the old name will continue to be supported and is not depracated. The change was merely done because m_getclr() sounds too much like "m_get a cluster." - TEMPORARILY disabled mbtypes statistics displaying in netstat(1) and systat(1) (see TODO below). - Fixed systat(1) to display number of "free mbufs" based on new per-CPU stat structures. - Fixed netstat(1) to display new per-CPU stats based on sysctl-exported per-CPU stat structures. All infos are fetched via sysctl. TODO (in order of priority): - Re-enable mbtypes statistics in both netstat(1) and systat(1) after introducing an SMP friendly way to collect the mbtypes stats under the already introduced per-CPU locks (i.e. hopefully don't use atomic() - it seems too costly for a mere stat update, especially when other locks are already present). - Optionally have systat(1) display not only "total free mbufs" but also "total free mbufs per CPU pool." - Fix minor length-fetching issues in netstat(1) related to recently re-enabled option to read mbuf stats from a core file. - Move reference counters at least for mbuf clusters into an unused portion of the cluster itself, to save space and need to allocate a counter. - Look into introducing resource freeing possibly from a kproc. Reviewed by (in parts): jlemon, jake, silby, terry Tested by: jlemon (Intel & Alpha), mjacob (Intel & Alpha) Preliminary performance measurements: jlemon (and me, obviously) URL: http://people.freebsd.org/~bmilekic/mb_alloc/
2001-06-22 06:35:32 +00:00
size_t mlen;
long *mbtypes = NULL;
Introduce numerous SMP friendly changes to the mbuf allocator. Namely, introduce a modified allocation mechanism for mbufs and mbuf clusters; one which can scale under SMP and which offers the possibility of resource reclamation to be implemented in the future. Notable advantages: o Reduce contention for SMP by offering per-CPU pools and locks. o Better use of data cache due to per-CPU pools. o Much less code cache pollution due to excessively large allocation macros. o Framework for `grouping' objects from same page together so as to be able to possibly free wired-down pages back to the system if they are no longer needed by the network stacks. Additional things changed with this addition: - Moved some mbuf specific declarations and initializations from sys/conf/param.c into mbuf-specific code where they belong. - m_getclr() has been renamed to m_get_clrd() because the old name is really confusing. m_getclr() HAS been preserved though and is defined to the new name. No tree sweep has been done "to change the interface," as the old name will continue to be supported and is not depracated. The change was merely done because m_getclr() sounds too much like "m_get a cluster." - TEMPORARILY disabled mbtypes statistics displaying in netstat(1) and systat(1) (see TODO below). - Fixed systat(1) to display number of "free mbufs" based on new per-CPU stat structures. - Fixed netstat(1) to display new per-CPU stats based on sysctl-exported per-CPU stat structures. All infos are fetched via sysctl. TODO (in order of priority): - Re-enable mbtypes statistics in both netstat(1) and systat(1) after introducing an SMP friendly way to collect the mbtypes stats under the already introduced per-CPU locks (i.e. hopefully don't use atomic() - it seems too costly for a mere stat update, especially when other locks are already present). - Optionally have systat(1) display not only "total free mbufs" but also "total free mbufs per CPU pool." - Fix minor length-fetching issues in netstat(1) related to recently re-enabled option to read mbuf stats from a core file. - Move reference counters at least for mbuf clusters into an unused portion of the cluster itself, to save space and need to allocate a counter. - Look into introducing resource freeing possibly from a kproc. Reviewed by (in parts): jlemon, jake, silby, terry Tested by: jlemon (Intel & Alpha), mjacob (Intel & Alpha) Preliminary performance measurements: jlemon (and me, obviously) URL: http://people.freebsd.org/~bmilekic/mb_alloc/
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
mlen = sizeof *mbstat;
Introduce numerous SMP friendly changes to the mbuf allocator. Namely, introduce a modified allocation mechanism for mbufs and mbuf clusters; one which can scale under SMP and which offers the possibility of resource reclamation to be implemented in the future. Notable advantages: o Reduce contention for SMP by offering per-CPU pools and locks. o Better use of data cache due to per-CPU pools. o Much less code cache pollution due to excessively large allocation macros. o Framework for `grouping' objects from same page together so as to be able to possibly free wired-down pages back to the system if they are no longer needed by the network stacks. Additional things changed with this addition: - Moved some mbuf specific declarations and initializations from sys/conf/param.c into mbuf-specific code where they belong. - m_getclr() has been renamed to m_get_clrd() because the old name is really confusing. m_getclr() HAS been preserved though and is defined to the new name. No tree sweep has been done "to change the interface," as the old name will continue to be supported and is not depracated. The change was merely done because m_getclr() sounds too much like "m_get a cluster." - TEMPORARILY disabled mbtypes statistics displaying in netstat(1) and systat(1) (see TODO below). - Fixed systat(1) to display number of "free mbufs" based on new per-CPU stat structures. - Fixed netstat(1) to display new per-CPU stats based on sysctl-exported per-CPU stat structures. All infos are fetched via sysctl. TODO (in order of priority): - Re-enable mbtypes statistics in both netstat(1) and systat(1) after introducing an SMP friendly way to collect the mbtypes stats under the already introduced per-CPU locks (i.e. hopefully don't use atomic() - it seems too costly for a mere stat update, especially when other locks are already present). - Optionally have systat(1) display not only "total free mbufs" but also "total free mbufs per CPU pool." - Fix minor length-fetching issues in netstat(1) related to recently re-enabled option to read mbuf stats from a core file. - Move reference counters at least for mbuf clusters into an unused portion of the cluster itself, to save space and need to allocate a counter. - Look into introducing resource freeing possibly from a kproc. Reviewed by (in parts): jlemon, jake, silby, terry Tested by: jlemon (Intel & Alpha), mjacob (Intel & Alpha) Preliminary performance measurements: jlemon (and me, obviously) URL: http://people.freebsd.org/~bmilekic/mb_alloc/
2001-06-22 06:35:32 +00:00
if ((mbstat = malloc(mlen)) == NULL) {
warn("malloc: cannot allocate memory for mbstat");
goto err;
}
if (kread(mbaddr, (char *)mbstat, sizeof mbstat))
goto err;
if (kread(nmbcaddr, (char *)&nmbclusters, sizeof(int)))
goto err;
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 */
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;
}
#undef MSIZE
Introduce numerous SMP friendly changes to the mbuf allocator. Namely, introduce a modified allocation mechanism for mbufs and mbuf clusters; one which can scale under SMP and which offers the possibility of resource reclamation to be implemented in the future. Notable advantages: o Reduce contention for SMP by offering per-CPU pools and locks. o Better use of data cache due to per-CPU pools. o Much less code cache pollution due to excessively large allocation macros. o Framework for `grouping' objects from same page together so as to be able to possibly free wired-down pages back to the system if they are no longer needed by the network stacks. Additional things changed with this addition: - Moved some mbuf specific declarations and initializations from sys/conf/param.c into mbuf-specific code where they belong. - m_getclr() has been renamed to m_get_clrd() because the old name is really confusing. m_getclr() HAS been preserved though and is defined to the new name. No tree sweep has been done "to change the interface," as the old name will continue to be supported and is not depracated. The change was merely done because m_getclr() sounds too much like "m_get a cluster." - TEMPORARILY disabled mbtypes statistics displaying in netstat(1) and systat(1) (see TODO below). - Fixed systat(1) to display number of "free mbufs" based on new per-CPU stat structures. - Fixed netstat(1) to display new per-CPU stats based on sysctl-exported per-CPU stat structures. All infos are fetched via sysctl. TODO (in order of priority): - Re-enable mbtypes statistics in both netstat(1) and systat(1) after introducing an SMP friendly way to collect the mbtypes stats under the already introduced per-CPU locks (i.e. hopefully don't use atomic() - it seems too costly for a mere stat update, especially when other locks are already present). - Optionally have systat(1) display not only "total free mbufs" but also "total free mbufs per CPU pool." - Fix minor length-fetching issues in netstat(1) related to recently re-enabled option to read mbuf stats from a core file. - Move reference counters at least for mbuf clusters into an unused portion of the cluster itself, to save space and need to allocate a counter. - Look into introducing resource freeing possibly from a kproc. Reviewed by (in parts): jlemon, jake, silby, terry Tested by: jlemon (Intel & Alpha), mjacob (Intel & Alpha) Preliminary performance measurements: jlemon (and me, obviously) URL: http://people.freebsd.org/~bmilekic/mb_alloc/
2001-06-22 06:35:32 +00:00
#define MSIZE (mbstat->m_msize)
#undef MCLBYTES
Introduce numerous SMP friendly changes to the mbuf allocator. Namely, introduce a modified allocation mechanism for mbufs and mbuf clusters; one which can scale under SMP and which offers the possibility of resource reclamation to be implemented in the future. Notable advantages: o Reduce contention for SMP by offering per-CPU pools and locks. o Better use of data cache due to per-CPU pools. o Much less code cache pollution due to excessively large allocation macros. o Framework for `grouping' objects from same page together so as to be able to possibly free wired-down pages back to the system if they are no longer needed by the network stacks. Additional things changed with this addition: - Moved some mbuf specific declarations and initializations from sys/conf/param.c into mbuf-specific code where they belong. - m_getclr() has been renamed to m_get_clrd() because the old name is really confusing. m_getclr() HAS been preserved though and is defined to the new name. No tree sweep has been done "to change the interface," as the old name will continue to be supported and is not depracated. The change was merely done because m_getclr() sounds too much like "m_get a cluster." - TEMPORARILY disabled mbtypes statistics displaying in netstat(1) and systat(1) (see TODO below). - Fixed systat(1) to display number of "free mbufs" based on new per-CPU stat structures. - Fixed netstat(1) to display new per-CPU stats based on sysctl-exported per-CPU stat structures. All infos are fetched via sysctl. TODO (in order of priority): - Re-enable mbtypes statistics in both netstat(1) and systat(1) after introducing an SMP friendly way to collect the mbtypes stats under the already introduced per-CPU locks (i.e. hopefully don't use atomic() - it seems too costly for a mere stat update, especially when other locks are already present). - Optionally have systat(1) display not only "total free mbufs" but also "total free mbufs per CPU pool." - Fix minor length-fetching issues in netstat(1) related to recently re-enabled option to read mbuf stats from a core file. - Move reference counters at least for mbuf clusters into an unused portion of the cluster itself, to save space and need to allocate a counter. - Look into introducing resource freeing possibly from a kproc. Reviewed by (in parts): jlemon, jake, silby, terry Tested by: jlemon (Intel & Alpha), mjacob (Intel & Alpha) Preliminary performance measurements: jlemon (and me, obviously) URL: http://people.freebsd.org/~bmilekic/mb_alloc/
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);
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);
}
Introduce numerous SMP friendly changes to the mbuf allocator. Namely, introduce a modified allocation mechanism for mbufs and mbuf clusters; one which can scale under SMP and which offers the possibility of resource reclamation to be implemented in the future. Notable advantages: o Reduce contention for SMP by offering per-CPU pools and locks. o Better use of data cache due to per-CPU pools. o Much less code cache pollution due to excessively large allocation macros. o Framework for `grouping' objects from same page together so as to be able to possibly free wired-down pages back to the system if they are no longer needed by the network stacks. Additional things changed with this addition: - Moved some mbuf specific declarations and initializations from sys/conf/param.c into mbuf-specific code where they belong. - m_getclr() has been renamed to m_get_clrd() because the old name is really confusing. m_getclr() HAS been preserved though and is defined to the new name. No tree sweep has been done "to change the interface," as the old name will continue to be supported and is not depracated. The change was merely done because m_getclr() sounds too much like "m_get a cluster." - TEMPORARILY disabled mbtypes statistics displaying in netstat(1) and systat(1) (see TODO below). - Fixed systat(1) to display number of "free mbufs" based on new per-CPU stat structures. - Fixed netstat(1) to display new per-CPU stats based on sysctl-exported per-CPU stat structures. All infos are fetched via sysctl. TODO (in order of priority): - Re-enable mbtypes statistics in both netstat(1) and systat(1) after introducing an SMP friendly way to collect the mbtypes stats under the already introduced per-CPU locks (i.e. hopefully don't use atomic() - it seems too costly for a mere stat update, especially when other locks are already present). - Optionally have systat(1) display not only "total free mbufs" but also "total free mbufs per CPU pool." - Fix minor length-fetching issues in netstat(1) related to recently re-enabled option to read mbuf stats from a core file. - Move reference counters at least for mbuf clusters into an unused portion of the cluster itself, to save space and need to allocate a counter. - Look into introducing resource freeing possibly from a kproc. Reviewed by (in parts): jlemon, jake, silby, terry Tested by: jlemon (Intel & Alpha), mjacob (Intel & Alpha) Preliminary performance measurements: jlemon (and me, obviously) URL: http://people.freebsd.org/~bmilekic/mb_alloc/
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);
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",
mbstat->sf_iocnt);
Introduce numerous SMP friendly changes to the mbuf allocator. Namely, introduce a modified allocation mechanism for mbufs and mbuf clusters; one which can scale under SMP and which offers the possibility of resource reclamation to be implemented in the future. Notable advantages: o Reduce contention for SMP by offering per-CPU pools and locks. o Better use of data cache due to per-CPU pools. o Much less code cache pollution due to excessively large allocation macros. o Framework for `grouping' objects from same page together so as to be able to possibly free wired-down pages back to the system if they are no longer needed by the network stacks. Additional things changed with this addition: - Moved some mbuf specific declarations and initializations from sys/conf/param.c into mbuf-specific code where they belong. - m_getclr() has been renamed to m_get_clrd() because the old name is really confusing. m_getclr() HAS been preserved though and is defined to the new name. No tree sweep has been done "to change the interface," as the old name will continue to be supported and is not depracated. The change was merely done because m_getclr() sounds too much like "m_get a cluster." - TEMPORARILY disabled mbtypes statistics displaying in netstat(1) and systat(1) (see TODO below). - Fixed systat(1) to display number of "free mbufs" based on new per-CPU stat structures. - Fixed netstat(1) to display new per-CPU stats based on sysctl-exported per-CPU stat structures. All infos are fetched via sysctl. TODO (in order of priority): - Re-enable mbtypes statistics in both netstat(1) and systat(1) after introducing an SMP friendly way to collect the mbtypes stats under the already introduced per-CPU locks (i.e. hopefully don't use atomic() - it seems too costly for a mere stat update, especially when other locks are already present). - Optionally have systat(1) display not only "total free mbufs" but also "total free mbufs per CPU pool." - Fix minor length-fetching issues in netstat(1) related to recently re-enabled option to read mbuf stats from a core file. - Move reference counters at least for mbuf clusters into an unused portion of the cluster itself, to save space and need to allocate a counter. - Look into introducing resource freeing possibly from a kproc. Reviewed by (in parts): jlemon, jake, silby, terry Tested by: jlemon (Intel & Alpha), mjacob (Intel & Alpha) Preliminary performance measurements: jlemon (and me, obviously) URL: http://people.freebsd.org/~bmilekic/mb_alloc/
2001-06-22 06:35:32 +00:00
printf("%lu calls to protocol drain routines\n", mbstat->m_drain);
err:
if (mbtypes != NULL)
free(mbtypes);
if (seen != NULL)
free(seen);
Introduce numerous SMP friendly changes to the mbuf allocator. Namely, introduce a modified allocation mechanism for mbufs and mbuf clusters; one which can scale under SMP and which offers the possibility of resource reclamation to be implemented in the future. Notable advantages: o Reduce contention for SMP by offering per-CPU pools and locks. o Better use of data cache due to per-CPU pools. o Much less code cache pollution due to excessively large allocation macros. o Framework for `grouping' objects from same page together so as to be able to possibly free wired-down pages back to the system if they are no longer needed by the network stacks. Additional things changed with this addition: - Moved some mbuf specific declarations and initializations from sys/conf/param.c into mbuf-specific code where they belong. - m_getclr() has been renamed to m_get_clrd() because the old name is really confusing. m_getclr() HAS been preserved though and is defined to the new name. No tree sweep has been done "to change the interface," as the old name will continue to be supported and is not depracated. The change was merely done because m_getclr() sounds too much like "m_get a cluster." - TEMPORARILY disabled mbtypes statistics displaying in netstat(1) and systat(1) (see TODO below). - Fixed systat(1) to display number of "free mbufs" based on new per-CPU stat structures. - Fixed netstat(1) to display new per-CPU stats based on sysctl-exported per-CPU stat structures. All infos are fetched via sysctl. TODO (in order of priority): - Re-enable mbtypes statistics in both netstat(1) and systat(1) after introducing an SMP friendly way to collect the mbtypes stats under the already introduced per-CPU locks (i.e. hopefully don't use atomic() - it seems too costly for a mere stat update, especially when other locks are already present). - Optionally have systat(1) display not only "total free mbufs" but also "total free mbufs per CPU pool." - Fix minor length-fetching issues in netstat(1) related to recently re-enabled option to read mbuf stats from a core file. - Move reference counters at least for mbuf clusters into an unused portion of the cluster itself, to save space and need to allocate a counter. - Look into introducing resource freeing possibly from a kproc. Reviewed by (in parts): jlemon, jake, silby, terry Tested by: jlemon (Intel & Alpha), mjacob (Intel & Alpha) Preliminary performance measurements: jlemon (and me, obviously) URL: http://people.freebsd.org/~bmilekic/mb_alloc/
2001-06-22 06:35:32 +00:00
if (mbstat != NULL)
free(mbstat);
1994-05-27 12:33:43 +00:00
}
/*
* If running on a live kernel, directly query the zone allocator for stats
* from the four mbuf-related zones/types.
*/
static void
mbpr_sysctl(void)
{
struct memory_type_list *mtlp;
struct memory_type *mtp;
u_int64_t mbuf_count, mbuf_bytes, mbuf_free, mbuf_failures, mbuf_size;
u_int64_t cluster_count, cluster_bytes, cluster_limit, cluster_free;
u_int64_t cluster_failures, cluster_size;
u_int64_t packet_count, packet_bytes, packet_free, packet_failures;
u_int64_t tag_count, tag_bytes;
u_int64_t bytes_inuse, bytes_incache, bytes_total;
int nsfbufs, nsfbufspeak, nsfbufsused;
struct mbstat mbstat;
size_t mlen;
mtlp = memstat_mtl_alloc();
if (mtlp == NULL) {
warn("memstat_mtl_alloc");
return;
}
/*
* Use memstat_sysctl_all() because some mbuf-related memory is in
* uma(9), and some malloc(9).
*/
if (memstat_sysctl_all(mtlp, 0) < 0) {
warnx("memstat_sysctl_all: %s",
memstat_strerror(memstat_mtl_geterror(mtlp)));
goto out;
}
mtp = memstat_mtl_find(mtlp, ALLOCATOR_UMA, MBUF_MEM_NAME);
if (mtp == NULL) {
warnx("memstat_mtl_find: zone %s not found", MBUF_MEM_NAME);
goto out;
}
mbuf_count = memstat_get_count(mtp);
mbuf_bytes = memstat_get_bytes(mtp);
mbuf_free = memstat_get_free(mtp);
mbuf_failures = memstat_get_failures(mtp);
mbuf_size = memstat_get_size(mtp);
mtp = memstat_mtl_find(mtlp, ALLOCATOR_UMA, MBUF_PACKET_MEM_NAME);
if (mtp == NULL) {
warnx("memstat_mtl_find: zone %s not found",
MBUF_PACKET_MEM_NAME);
goto out;
}
packet_count = memstat_get_count(mtp);
packet_bytes = memstat_get_bytes(mtp);
packet_free = memstat_get_free(mtp);
packet_failures = memstat_get_failures(mtp);
mtp = memstat_mtl_find(mtlp, ALLOCATOR_UMA, MBUF_CLUSTER_MEM_NAME);
if (mtp == NULL) {
warnx("memstat_mtl_find: zone %s not found",
MBUF_CLUSTER_MEM_NAME);
goto out;
}
cluster_count = memstat_get_count(mtp);
cluster_bytes = memstat_get_bytes(mtp);
cluster_limit = memstat_get_countlimit(mtp);
cluster_free = memstat_get_free(mtp);
cluster_failures = memstat_get_failures(mtp);
cluster_size = memstat_get_size(mtp);
mtp = memstat_mtl_find(mtlp, ALLOCATOR_MALLOC, MBUF_TAG_MEM_NAME);
if (mtp == NULL) {
warnx("memstat_mtl_find: malloc type %s not found",
MBUF_TAG_MEM_NAME);
goto out;
}
tag_count = memstat_get_count(mtp);
tag_bytes = memstat_get_bytes(mtp);
printf("%llu/%llu/%llu mbufs in use (current/cache/total)\n",
mbuf_count + packet_count, mbuf_free + packet_free,
mbuf_count + packet_count + mbuf_free + packet_free);
printf("%llu/%llu/%llu/%llu mbuf clusters in use "
"(current/cache/total/max)\n",
cluster_count - packet_free, cluster_free + packet_free,
cluster_count + cluster_free, cluster_limit);
#if 0
printf("%llu mbuf tags in use\n", tag_count);
#endif
mlen = sizeof(nsfbufs);
if (!sysctlbyname("kern.ipc.nsfbufs", &nsfbufs, &mlen, NULL, 0) &&
!sysctlbyname("kern.ipc.nsfbufsused", &nsfbufsused, &mlen, NULL,
0) &&
!sysctlbyname("kern.ipc.nsfbufspeak", &nsfbufspeak, &mlen, NULL,
0)) {
printf("%d/%d/%d sfbufs in use (current/peak/max)\n",
nsfbufsused, nsfbufspeak, nsfbufs);
}
/*-
* Calculate in-use bytes as:
* - straight mbuf memory
* - mbuf memory in packets
* - the clusters attached to packets
* - and the rest of the non-packet-attached clusters.
* - m_tag memory
* This avoids counting the clusters attached to packets in the cache.
* This currently excludes sf_buf space.
*/
bytes_inuse =
mbuf_bytes + /* straight mbuf memory */
packet_bytes + /* mbufs in packets */
(packet_count * cluster_size) + /* clusters in packets */
/* other clusters */
((cluster_count - packet_count - packet_free) * cluster_size) +
tag_bytes;
/*
* Calculate in-cache bytes as:
* - cached straught mbufs
* - cached packet mbufs
* - cached packet clusters
* - cached straight clusters
* This currently excludes sf_buf space.
*/
bytes_incache =
(mbuf_free * mbuf_size) + /* straight free mbufs */
(packet_free * mbuf_size) + /* mbufs in free packets */
(packet_free * cluster_size) + /* clusters in free packets */
(cluster_free * cluster_size); /* free clusters */
/*
* Total is bytes in use + bytes in cache. This doesn't take into
* account various other misc data structures, overhead, etc, but
* gives the user something useful despite that.
*/
bytes_total = bytes_inuse + bytes_incache;
printf("%lluK/%lluK/%lluK bytes allocated to network "
"(current/cache/total)\n", bytes_inuse / 1024,
bytes_incache / 1024, bytes_total / 1024);
#if 0
printf("%llu/%llu/%llu requests for mbufs denied (mbufs/clusters/"
"mbuf+clusters)\n", mbuf_failures, cluster_failures,
packet_failures);
#endif
mlen = sizeof(mbstat);
if (!sysctlbyname("kern.ipc.mbstat", &mbstat, &mlen, NULL, 0)) {
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);
printf("%lu calls to protocol drain routines\n",
mbstat.m_drain);
}
out:
memstat_mtl_free(mtlp);
}
/*
* Print mbuf statistics.
*/
void
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)
{
if (mbaddr != 0)
mbpr_kmem(mbaddr, nmbcaddr, nmbufaddr, mbhiaddr, clhiaddr,
mbloaddr, clloaddr, pgsaddr, mbpaddr);
else
mbpr_sysctl();
}