Bug fix release.
- Let build system handle missing libite GIT submodule
- Issue #61: Debian packaging moved to https://github.com/bobek/pkg-pimd
- Issue #53: Build problem with Clang on FreeBSD
- Issue #55: Default config uses
/etcpimd.confinstead of/etc/pimd.conf. Slashes added and nowpimd -hlists the default path instead of a hard coded string. - Issue #60: Fix minor spelling errors.
The significant new features in this release would not have been possible if not for the hard work of Markus Veranen mailto:markus.veranen@gmail.com
Tested on Ubuntu 14.04 (GLIBC/Linux 3.13), Debian 8.1 (GLIBC/Linux 3.16), FreeBSD, NetBSD, and OpenBSD.
- Support for PIM-SSM and IGMPv3, by Markus Veranen
- IGMPv3 is now default, use
phyint ifname igmpv2for old behaviour - Default IGMP query interval has changed from 125 sec to 12 sec
In
pimd.conf: igmp-query-interval <SEC> - Default IGMP querier timeout has changed from 255 sec to 42 sec
In
pimd.conf: igmp-querier-timeout <SEC> - The built-in IGMP robustness value changed from 2 to 3
- Support for changing the PIM Hello interval, by Markus Veranen
In
pimd.conf: hello-interval <SEC> - Support for multiple multicast routing tables, and running multiple pimd instances, by Markus Veranen. (Only supported on Linux atm.)
- Support for advertising, and acting upon changes to, Generation ID in PIM Hello messages, by Markus Veranen
- Support for advertising DR Priority option in PIM Hello messages.
If all routers on a LAN send this option this value is used in the
DR election rather than the IP address. The priority is configured
per
phyint. This closes the long-standing issue #5. - Distribution archive format changed from XZ to Gzip, for the benefit of OpenBSD that only ships Gzip in the base system.
The pimd.conf syntax has been changed in this release. Mainly, the
configuration file now use dashes - instead of underscore _ as word
separators. However several settings have also been renamed to be more
familiar to commands used by major router vendors:
bsr-candidate- replaces
cand_bootstrap_router rp-candidate- replaces
cand_rp group-prefix- replaces
group_prefix rp-address- replaces
rp_address spt-threshold- replaces the two deprecated
switch_register_thresholdandswitch_data_thresholdsettings hello-interval- replaces
hello_period default-route-distance- replaces
default_source_preference default-route-metric- replaces
default-source-metric
Also, for phyint the preference sub-option has been replaced with
the less confusing distance and ttl-threshold replaces threshold.
See the README or the man page for more information on the metric
preference and admin distance confusion.
*Note:* The =pimd.conf= parser remains backwards compatible with the old syntax!
The following are new features that must be enabled at compile time,
using the configure script, to take effect. For details, see
./configure --help
--prefix=PATH- Standard prefix to be used at installation,
default
/usr/local --sysconfdir=PATH- Prefix path to be used for
pimd.conf, default/etc, unless--prefixis given. --embedded-libc- Enable uClib or musl libc build, on Linux.
--disable-exit-on-error- Allow pimd to continue running despite encountering errors.
--disable-pim-genid- Disable advertisement of PIM Hello GenID, use for compatibility problems with older versions of pimd.
--with-max-vifs=MAXVIFS- Raise max number of VIFs to MAXVIFS.
Note: this requires raising MAXVIFS in the kernel as well! Most kernels cannot handle >255, if this is a problem, try using multiple multicast routing tables instead.
--disable-masklen-check- Allow tunctl VIFs with masklen 32.
- Fix issue #40: FTBS with
./configure --enable-scoped-acls - Properly support cross compiling. It is now possible to actually
define the
$CROSSenvironment variable when callingmaketo allow cross compiling pimd. Should work with both GCC and Clang. Tested on Ubuntu, Debian and FreeBSD.
- Fix another problem with issue #22 (reopened), as laid out in
issue #37. This time the crash is induced when there is a link down
event. Lot of help debugging the propblem by @mfspeer, who also
suggested the fix – to call
pim_proto.c:delete_pim_nbr()invif.c:stop_vif()instead of just calling free. - Fix issue with not checking return value of
open()in daemonizing code inmain(), found by Coverity Scan. - Fix issue with scoped
phyintinconfig.c, found by Coverity Scan. Themasklenmay not be zero, config file problem, alert the user.
- Support for IP fragmentation of PIM register messages, by Michael Fine, Cumulus Networks
- Support
/LENsyntax inphyintto complementmasklen LEN, issue #12 - Add support for /31 networks, point-to-point, thanks to Apollon Oikonomopoulos
- Remove old broken SNMP support
- OpenBSD inspired cleanup (deregister)
- General code cleanup, shorten local variable names, func decl. etc.
- Support for router alert IP option in IGMP queries
- Support for reading IGMPv3 membership reports
- Update IGMP code to support FreeBSD >= 8.x
- Retry read of routing tables on FreeBSD
- Fix join/leve of ALL PIM Routers for FreeBSD and other UNIX kernels
- Tested on FreeBSD, NetBSD and OpenBSD
- Add very simple homegrown configure script
- Update and document support for
rp_address,cand_rp, andcand_bootstrap_router - Add new
spt_thresholdto replace existingswitch_register_thresholdandswitch_data_threshold settings. Cisco-like and easier to understand
- Fix to avoid infinite loop during unicast send failure, by Alex Tessmer
- Fix bug in bootstrap when configured as candidate RP, issue #15
- Fix segfault in
accept_igmp(), issue #29 - Fix default source preference, should be 101 (not 1024!)
- Fix
ip_lenhandling on older BSD’s, thanks to Olivier Cochard-Labbé, issue #23 - Fix default prefix len in static RP example in
pimd.conf, should be /4 - Fix issue #31: Make IGMP query interval and querier timeout configurable
- Fix issue #33: pimd does not work in background under FreeBSD
- Fix issue #35: support for timing out other queriers from mrouted
- Hopefully fix issue #22: Crash in (S,G) state when neighbor is lost
- Misc. bug fixes thanks to Coverity Scan, static code analysis tool https://scan.coverity.com/projects/3319
- Update docs of static Rendez-Vous Point,
rp_address, configuration in man page and examplepimd.conf. Thanks to Andriy Senkovych mailto:andriysenkovych@gmail.com and YAMAMOTO Shigeru mailto:shigeru@iij.ad.jp - Replaced
malloc()withcalloc()to mitigate risk of accessing junk data and ease debugging. Thanks to YAMAMOTO Shigeru mailto:shigeru@iij.ad.jp - Extend .conf file
rp_addressoption withpriorityfield. Code changes and documentation updates by YAMAMOTO Shigeru mailto:shigeru@iij.ad.jp
- A serious bug in
pim_proto.c:receive_pim_register()was found and fixed by Jean-Pascal Billaud. In essence, the RP check was broken since the code only looked atmy_cand_rp_address, which is not set when using therp_addressconfig. Everything works fine with auto-RP mode though. This issue completely breaks the register path since the JOIN(S,G) is never sent back … - Fix FTBFS issues reported from Debian. Later GCC versions trigger unused variable warnings. Patches and cleanup Antonin Kral mailto:a.kral@bobek.cz
- The previous move of runtime dump files to
/var/lib/mischave been changed to/var/run/pimdinstead. This to accomodate *BSD systems that do not have the/var/libtree, and also recommended in the Filesystem Hierarchy Standard, http://www.pathname.com/fhs/pub/fhs-2.3.html#VARRUNRUNTIMEVARIABLEDATA
- Debian package now conflicts with
smcroute, in addition tomrouted. It is only possible to run one multicast routing daemon at a time, kernel limitation. - The location of the dump file(s) have been moved from
/var/tmpto/var/lib/miscdue to the insecure nature of/var/tmp. See more below.
kern.c:k_del_vif(): Fix build error on GNU/kFreeBSD- CVE-2011-0007: Insecure file creation in
/var/tmp. “On USR1, pimd will write to/var/tmp/pimd.dumpa dump of the multicast route table. Since/var/tmpis writable by any user, a user can create a symlink to any file he wants to destroy with the content of the multicast routing table.”
- Improved error messages in kern.c
- Renamed CHANGES to ChangeLog
- Import mrouted fix: on GNU/Linux systems (only!) the call to
kern.c:k_del_vif()fails with:setsockopt MRT_DEL_VIF on vif 3: Invalid argument. This is due to differences in the Linux and *BSDMRT_DEL_VIFAPI. The Linux kernel expects to receive astruct vifctlassociated with the VIF to be deleted, *BSD systems on the other hand expect to receive the index of that VIF.Bug reported and fixed on mrouted by Dan Kruchinin mailto:dkruchinin@acm.org
- Updates for support on Debian GNU/kFreeBSD, FreeBSD kernel with GNU userland.
- Lior Dotan mailto:liodot@gmail.com reports that pimd 2.1.2 and
2.1.3 are severely broken w.r.t. uninformed systematic replace of
bcopy()withmemcpy()API.
debug.c:syslog(): Removed GNU:ism %m, usestrerror(errno)instead.- Cleanup and ansification of a couple of files: rp.c, mrt.c, vif.c, route.c
- Initialize stack variables to silence overzealous GCC on PowerPC and S/390. Debian bug 595584, this closes pimd issue #3 on GitHub.
- Merge bug fix for static-rp configurations from Kame’s pim6sd route.c r1.28
- Close TODO item by merging in relevant changes from Kame’s pim6sd
vif.c r1.3 - Tried fixing
debug.c:logit()build failure on Sparc due to mixup in headers fortv_usectype.
- License change on mrouted code from OpenBSD team => pimd fully free under the simlified 3-clause BSD license! This was also covered in v2.1.0-alpha29.17, but now all files have been updated, including LICENSE.mrouted.
- Code cleanup and ansification.
- Simplified Makefile so that it works seamlessly on GNU Make and BSD PMake.
- Replaced all calls to
bzero()andbcopy()withmemset()andmemcpy(). - Use
getopt_long()for argument parsing. - Add, and improve, -h,–help output.
- Add -f,–foreground option.
- Add -v,–version option.
- Add -l,–reload-config which sends SIGHUP to a running daemon.
- Add -r,–show-routes which sends SIGUSR1 to a running daemon.
- Add -q,–quit-daemon which sends SIGTERM to a running daemon.
- Make it possible to call pimd as a regular user, for –help and –version.
- Man page cleaned up, a lot, and updated with new options.
- Replaced dangerous old string functions with safer
snprintf()andstrlcpy() - Added checks for
malloc()return values, all over the code base. - Fixed issues reported by Sparse (CC=cgcc).
- Make sure to retry syscalls
recvfrom()andsendto()on signal (SIGINT). - Fix build issues on OpenBSD 4.7 and FreeBSD 8.1 thanks to Guillaume Sellier.
- Kernel include issues on Ubuntu 8.04, Linux <= 2.6.25, by Nikola Knežević
- Fix build issues on NetBSD
- Bumping version again to celebrate the changes and make it easier for distributions to handle the upgrade.
Merged all patches from http://lintrack.org.
001-debian-6.diff: Already merged, no-op - only documenting in case anyone wonders about it.002-better-rp_address.diff: Support multicast group address in static Rendez-Vous Point .conf option.003-ltfixes.diff: Various bug fixes and error handling improvements.004-disableall.diff: Add -N option to pimd.005-vifenable.diff: Add enable keyword to phyint .conf option.006-dot19.diff: The lost alpha29.18 and alpha29.19 fixes by Pavlin Radoslavov.
- Integrated the latest Debian patches from
pimd_2.1.0-alpha29.17-9.diff.gz - Fixed the new file include/linux/netinet/in-my.h (Debian) so that the #else fallback uses the system netinet/in.h, which seems to work now.
- Bumped version number, this code has been available for a while now.
- Don’t ignore PIM Null Register messages if the IP version of the inner header is not valid.
- Add a missing bracket inside rsrr.c (a bug report and a fix by mailto:seyon@oullim.co.kr)
- Compilation fix for Solaris 8. Though, no guarantee pimd still works on that platform.
- Define
BYTE_ORDERif missing. - Update include/netinet/pim.h file with its lastest version
- Update the copyright message of
include/netinet/pim_var.h
- The mrouted license, LICENSE.mrouted, updated with BSD-like license!! Thanks to
the OpenBSD folks for the 2 years of hard work to make this happen:
http://www.openbsd.org/cgi-bin/cvsweb/src/usr.sbin/mrouted/LICENSE
- Moved the pimd contact email address upfront in README. Let me repeat that here: If you have any questions, suggestions, bug reports, etc., do NOT send them to the PIM IETF Working Group mailing list! Instead, use the contact email address specified in README.
- Compilation bugfix for Linux. Bug report by Serdar Uezuemcue mailto:serdar@eikon.tum.de
- Routing socket descriptor leak. Bug report and fix by SUZUKI Shinsuke mailto:suz@crl.hitachi.co.jp; incorporated back from pim6sd.
- PIM join does not go upstream. Bug report and fix by SUZUKI Shinsuke mailto:suz@crl.hitachi.co.jp; incorporated back from pim6sd.
[problem]
PIM join does not go upstream in the following topology, because oif-list
is NULL after subtracting iif from oif-list.
receiver---rtr1---|
rtr2---|---rtr3----sender
rtr1's nexthop to sender = rtr2
rtr2's nexthop to sender = rtr3
[reason]
Owing to a difference between RFC2362 and the new pim-sm draft.
[solution]
Prunes iif from oiflist when installing it into kernel, instead of
PIM route calculation time.
- Bugfix in calculating the netmask for POINTOPOINT interface in config.c. Bug report by J.W. (Bill) Atwood mailto:bill@cs.concordia.ca
rp.c:rp_grp_match(): SERIOUS bugfix in calculating the RP per group when there are a number of group prefixes in the Cand-RP set. Bug report by Eva Pless mailto:eva.pless@imk.fraunhofer.de
- Bugfix in rp.c
bootstrap_initial_delay()in calculating BSR election delay. Fix by SAKAI Hiroaki mailto:sakai.hiroaki@finet.fujitsu.com
- Increase size of send buffers in the kernel. Bug report by Andrea Gambirasio mailto:andrea.gambirasio@softsolutions.it
Bug reports and fixes by SAKAI Hiroaki mailto:sakai.hiroaki@finet.fujitsu.com
init_routesock(): Bugfix: initializing a forgotten variable. The particular code related to that variable is commented-out by default, but a bug is a bug.main.c:restart(): Bugfix: close theudp_socketonly when it is is different fromigmp_socket.main.c:main(): if SIGHUP signal is received, reconstruct readers and nfds- Three serious bug fixes thanks to Jiahao Wang mailto:jiahaow@yahoo.com.cn and
Bo Cheng mailto:bobobocheng@hotmail.com:
pim_proto.c:receive_pim_join_prune(): two bugfixes related to the processing of (*,*,RP)pim_proto.c:add_jp_entry(): Bugfix regarding adding prune entries
- Remove the FTP URL from the various README files, and replace it with an HTTP URL, because the FTP server on catarina.usc.edu is not operational anymore.
- Widen the space for “Subnet” addresses printed under “Virtual Interface Table”
- Added (commented-out code) to enable different interfaces to belong to overlapping subnets. See around line 200 in config.c
- Bugfix in handling of Join/Prune messages when there is one join and one prune for the same group. Thanks to Xiaofeng Liu mailto:liu_xiao_feng@yahoo.com.
First three entries contributed by Hiroyuki Komatsu mailto:komatsu@taiyaki.org
- Print line number if there is conf file error.
- If there is an error in the conf file, pimd won’t start.
- GRE configuration examples added to README.config.
- New file README.debug (still very short though).
- Increase the config line buffer size to 1024. Bug fix by Hiroyuki Komatsu mailto:komatsu@taiyaki.org
- Better log messages for point-to-point links in config.c. Thanks to Hitoshi Asaeda mailto:asaeda@yamato.ibm.com
- Added “phyint altnet” (see pimd.conf for usage) for allowing some senders look like directly connected to a local subnet. Implemented by Marian Stagarescu mailto:marian@bile.cidera.com
- Added “phyint scoped” (see pimd.conf for usage) for administartively disabling the forwarding of multicast groups. Implemented by Marian Stagarescu mailto:marian@bile.cidera.com
- The License has changed from the original USC to the more familiar BSD-like (the KAME+OpenBSD guys brought to my attention that the original working in the USC license “…and without fee…” is ambiguous and makes it sound that noone can distribute pimd as part of some other software distribution and charge for that distribution.
- RSRR disabled by default in Makefile
- Memory leaks bugs fixed in rp.c, thanks to Sri V mailto:vallepal@yahoo.com
- Compilation problems for RedHat-7.1 fixed. Bug report by Philip Ho mailto:cbho@ie.cuhk.edu.hk
- PID computation fixed (it should be recomputed after a child
fork()). Thanks to Marian Stagarescu mailto:marian@bile.cidera.com find_route()-related bug fixes (always explicitly check for NULL return). Bug report by Marian Stagarescu mailto:marian@bile.cidera.com- Bug fix re. adding a local member with older ciscos (in
add_leaf()). Bug report by Marian Stagarescu mailto:marian@bile.cidera.com - Added explicit check whether
BYTE_ORDERin pimd.h is defined. Bug report by mailto:mistkhan@indiatimes.com
- Bug fixes in processing Join/Prune messages. Thanks to Sri V mailto:vallepal@yahoo.com
VIFM_FORWARDER()macro renamed toVIFM_LASTHOP_ROUTER.- Mini-FAQ entries added to README.
- When there is a new member,
add_leaf()is called by IGMP code for any router, not only for a DR. The reason is because not only the DR must know about local members, but the last-hop router as well (so eventually it will initiate a SPT switch). Similar fixes toadd_leaf()inside route.c as well. Problem reported by Hitoshi Asaeda mailto:asaeda@yamato.ibm.com. XXX: Note the lenghty comment in the beginning ofadd_leaf()about a pimd desing problem that may result in SPT switch not initiated immediately by the last-hop router. - DR entry timer bug fix in timer.c: When (*,G)’s iif and (S,G)’s iif are not same, (S,G)’s timer for the DR doesn’t increase. Reported indirectly by mailto:toshiaki.nakatsu@fujixerox.co.jp
- README cleanup + Mini-FAQ added
igmp_proto.c: printf argument cleanup (courtesy KAME)main.c:restart(): forgotten printf argument added (courtesy KAME)
kern.c:k_stop_pim(): Fix the ordering ofMRT_PIMandMRT_DONE, thanks to Hitoshi Asaeda mailto:asaeda@yamato.ibm.co.jp.route.c:add_leaf(): mrtentry creation logic bug fix. If the router is not a DR, a mrtentry is never created. Tanks to Hitoshi Asaeda mailto:asaeda@yamato.ibm.co.jp & (indirectly) mailto:toshiaki.nakatsu@fujixerox.co.jppim_proto.c: Two critical bug fixes. J/P prune suppression related message and J/P message with (*,*,RP) entry inside. Thanks to Azzurra Pantella mailto:s198804@studenti.ing.unipi.it and Nicola Dicosmo from University of Pisapim_proto.c:receive_pim_bootstrap(): BSR-related fix from Kame’s pim6sd. Even when the BSR changes, just schedule an immediate advertisemnet of C-RP-ADV, instead of sending message, in order to avoid sending the advertisement to the old BSR. In response to comment from mailto:toshiaki.nakatsu@fujixerox.co.jp
ADVANCE()bug fix in routesock.c (if your system doesn’t haveSA_LEN) thanks to Eric S. Johnson mailto:esj@cs.fiu.edu
NB: THIS pimd VERSION WON’T WORK WITH OLDER PIM-SM KERNEL PATCHES (kernel patches released prior to this version)!
- The daemon that the kernel will prepare completely the inner multicast packet for PIM register messages that the kernel is supposed to encapsulate and send to the RP.
- Now pimd compiles on OpenBSD-2.7. PIM control messages exchange test passed. Ddon’t have the infrastructure to perform more complete testing.
main.c:cleanup(): SendPIM_HELLOwith holdtime of ‘0’ if pimd is going away, thanks to JINMEI Tatuya mailto:jinmei@isl.rdc.toshiba.co.jpinclude/netinet/pim.hupdated- pimd code adapted to the new
struct pimdefinition. - Added
PIM_OLD_KERNELandBROKEN_CISCO_CHECKSUMentries in the Makefile. - Don’t ignore kernel signals if any of src or dst are NULL.
- Don’t touch
ip_idon a PIM register message - README cleanup: kernel patches location, obsoleted systems clarification, etc.
k_stop_pim()added tocleanup()inmain.c(courtesy Kame)
RANDOM()-related bug fix re.jp_valuecalculation inpim_proto.c, thanks to JINMEI Tatuya mailto:jinmei@isl.rdc.toshiba.co.jprealloc()related memory leak bug inconfig_vifs_from_kernel()in config.c courtesy Kame’s pim6sd code.- Solaris-8 fixes thanks to Eric S. Johnson mailto:esj@cs.fiu.edu
BROKEN_CISCO_CHECKSUMbug fix thanks to Eric S. Johnson mailto:esj@cs.fiu.edu and Hitoshi Asaeda.main.c: 1000000 usec -> 1 sec 0 usec. Fix courtesy of the Kame projectmain.c:restart()fixup courtesy of the Kame project- various min. message length check for the received control messages courtesy of the Kame project. XXX: the pimd check is not enough!
- VIF name string comparison fix in
routesock.c:getmsg()courtesy of the Kame project - missing brackets added inside
age_routes()(a bug that will show up only ifKERNEL_MFC_WC_Gwas defined); courtesy of the Kame project
- added #ifdef
BROKEN_CISCO_CHECKSUM(disabled by default) to make cisco RPs happy (read the comments in pim.c) - added #ifdef
PIM_TYPEVERS_DECLin netinet/pim.h as a workaround that ANSI-C doesn’t guarantee that bit-fields are tightly packed together (although all modern C compilers should not create a problem).
- Fixes to enable point-to-point interfaces being added correctly, thanks to Roger Venning mailto:Roger.Venning@corpmail.telstra.com.au
- A number of minor bug fixes
NB: this release may the the last one from 2.1.0. The next release will be 2.2.0 and there will be lots of changes inside.
- Bug fix in
rp.c:add_grp_mask()andrp.c:delete_grp_mask(): in some cases if the RPs are configured with nested multicast prefixes, the add/delete may fail. Thanks to Hitoshi Asaeda and the KAME team for pointing out this one.
- Bug fix in
receive_pim_register()inpim_proto.c:ntohl()was missing insideIN_MULTICAST(). Thanks to Fred Griffoul mailto:griffoul@ccrle.nec.de - Bug report and fix by Hitoshi Asaeda mailto:asaeda@yamato.ibm.co.jp in
pim_proto.c:receive_pim_cand_rp_adv()(if a router is not a BSR). Another bug inrp.c:delete_grp_mask_entry(): an entry not in the head of the list was not deleted propertly. - Some
VIFF_TUNNELchecks added or deleted in various places. Slowly preparing pimd to be able to work with GRE tunnels…
Bug reports and fixes by Hitoshi Asaeda mailto:asaeda@yamato.ibm.co.jp inside
parse_reg_threshold() and parse_data_threshold() in config.c
- Successfully added multicast prefixes configured in pimd.conf are displayed at startup
- Use
include/freebsdas FreeBSD-3.x include files andinclude/freebsd2for FreeBSD-2.x.
- Test is performed whether a
PIM_REGISTERhas invalid source and/or group address of the internal packet.
PIM_DEFAULT_CAND_RP_ADV_PERIODdefinition set to 60, but default ‘time’ value for inter Cand-RP messages is set in pimd.conf to 30 sec.PIM_REGISTERchecksum verification inreceive_pim_register()relaxed for compatibility with some older routers. The checksum has to be computed only over the first 8 bytes of the PIM Register (i.e. only over the header), but some older routers might compute it over the whole packet. Hence, the checksum verification is over the first 8 bytes first, and if if it fails, then over the whole packet. Thus, pimd that is RP should still work with older routers that act as DR, but if an older router is the RP, then pimd cannot be the DR. Sorry, don’t know which particular routers and models create the checksum over the whole PIM Register (if there are still any left).
- Finally pimd works under Linux (probably 2.1.126, 2.2.x and 2.3.x). However, a
small fix in the kernel
linux/net/ipv4/ipmr.cis necessary. In functionpim_rcv(), remove the call toip_compute_csum():
--- linux/net/ipv4/ipmr.c.org Thu Mar 25 09:23:34 1999
+++ linux/net/ipv4/ipmr.c Mon May 24 15:42:45 1999
@@ -1342,8 +1342,7 @@
if (len < sizeof(*pim) + sizeof(*encap) ||
pim->type != ((PIM_VERSION<<4)|(PIM_REGISTER)) ||
(pim->flags&PIM_NULL_REGISTER) ||
- reg_dev == NULL ||
- ip_compute_csum((void *)pim, len)) {
+ reg_dev == NULL) {
kfree_skb(skb);
return -EINVAL;
}- in pimd.conf “phyint” can be specified not only by IP address, but by name too (e.g. “phyint de1 disable”)
- in pimd.conf ‘preference’ and ‘metric’ can be specified per “phyint” Note that these ‘preference’ and ‘metric’ are like per iif.
MRT_PIMused (again) instead ofMRT_ASSERTin kern.c. The problem is that Linux has bothMRT_ASSERTandMRT_PIM, while *BSD has onlyMRT_ASSERT.
#ifndef MRT_PIM
#define MRT_PIM MRT_ASSERT
#endif- Rely on
__bsdi__, which is defined by the OS, instead of -DBSDI in Makefile, change by Hitoshi Asaeda. Similarly, use__FreeBSD__instead of -DFreeBSD - Linux patches by Fred Griffoul mailto:griffoul@ccrle.nec.de including
a
netlink.cinstead ofroutesock.c vif.c:zero_vif(): New function
All bug reports thanks to Kaifu Wu mailto:kaifu@3com.com
- Linux-related bug fixes regarding raw IP packets byte ordering
- Join/Prune message bug fixed if the message contains several groups joined/pruned
Bug reports by Jonathan Day mailto:jd9812@my-dejanews.com
- Bug fixes to compile under newer Linux kernel (linux-2.1.127) To compile for
older kernels ( ver < ???), add
-Dold_Linuxto the Makefile - For convenience, the
include/linux/netinet/{in.h,mroute.h}files are added, with few modifications applied.
pim_proto.c:join_or_prune(): Bug fixes in case of (S,G) overlapping with (*,G). Bug report by Dirk Ooms mailto:Dirk.Ooms@alcatel.beroute.c:change_interfaces(): Join/Prune (*,G), (*,*,RP) fire timer optimization/fix.
- (Almost) all timers manipulation now use macros
pim.handpim_var.hare in separate common directory- Added BSDI definition to
pim_var.h, thanks to Hitoshi Asaeda.
- fix TIMEOUT definitions in difs.h (bug report by Nidhi Bhaskar) (originally, if timer value less than 5 seconds, it won’t become 0) It is HIGHLY recommended to apply that fix, so here it is:
-------------BEGIN BUG FIX-------------------
1) Add the following lines to defs.h (after #define FALSE):
#ifndef MAX
#define MAX(a,b) (((a) >= (b))? (a) : (b))
#define MIN(a,b) (((a) <= (b))? (a) : (b))
#endif /* MAX & MIN */
2) Change the listed below TIMEOUT macros to:
#define IF_TIMEOUT(timer) \
if (!((timer) -= (MIN(timer, TIMER_INTERVAL))))
#define IF_NOT_TIMEOUT(timer) \
if ((timer) -= (MIN(timer, TIMER_INTERVAL)))
#define TIMEOUT(timer) \
(!((timer) -= (MIN(timer, TIMER_INTERVAL))))
#define NOT_TIMEOUT(timer) \
((timer) -= (MIN(timer, TIMER_INTERVAL)))
---------------END BUG FIX-------Both bug reports by Chirayu Shah mailto:shahzad@torrentnet.com-
- bug fix in
find_route()when searching for (*,*,RP) - bug fix in
move_kernel_cache(): no need to domove_kernel_cache()from (*,*,R) to (*,G) first when we callmove_kernel_cache()for (S,G)
- Now compiles under Linux (haven’t checked whether the PIMv2 kernel support in linux-2.1.103 works)
parse_default_source*()bug fix (bug reports by Nidhi Bhaskar)- allpimrouters deleted from igmp.c (already defined in pim.c)
- igmpmsg defined for IRIX
- (*,G) MFC kernel support completed and verified. Compile with
KERNEL_MFC_WC_Gdefined in Makefile, but then must use it only with a kernel that supports (*,G), e.g.pimkern-PATCH_7. Currently, kernel patches available for FreeBSD and SunOS only.
MRTF_MFC_CLONE_SGflag set afterdelete_single_kernel_cache()is called
- PIM registers kernel encapsulation support. Build with
PIM_REG_KERNEL_ENCAPdefined in Makefile. - (*,G) MFC support. Build with
KERNEL_MFC_WC_Gdefined in Makefile. However,MFC_WC_Gis still not supported withpimkern-PATCH_6, must disable it for now. mrt.c:delete_single_kernel_cache_addr(): New function, uses source, group to specify an MFC to be deleted
- Another few bug fixes related to NetBSD definitions thanks to Heiko W.Rupp mailto:hwr@pilhuhn.de
- A few bug fixes related to NetBSD definitions thanks to Heiko W.Rupp mailto:hwr@pilhuhn.de
- If the RP changes, the necessary actions are taken to pass the new RP address to
the kernel. To be used for kernel register encap. support. Wnat needs to be done
is: (a) add
rp_addrentry to the mfcctl structure, and then just set it inkern.c:k_chf_mfc(). Obviously, the kernel needs to support the register encapsulation (instead of sending WHOLEPKT to the user level). In the near few days will make the necessary kernel changes. change_interfaces(): Added “flags” argument. The only valid flag isMFC_UPDATE_FORCE, used for forcing kernel call when only the RP changes.k_chg_mfc()has a new argument: rp_addr. To be used for kernel register encapsulation supportMRT_PIMcompletely replaced byMRT_ASSERTmove_kernel_cache(): ArgumentMFC_MOVE_FORCEis a flag instead of TRUE/FALSEprocess_cache_miss(): removed unneeded piece of code
- Use the cleaned up
netinet/pim.h - Remove the no needed anymore pim header definition in
pimd.h - Don’t use
MRT_PIMin in kern.c anymore, replaced back withMRT_ASSERT. added default_source_metricanddefault_source_preference(1024) because the kernel’s unicast routing table is not a good source of info; configurable in pimd.conf- Can now compile under NetBSD-1.3, thanks to Heiko W.Rupp mailto:hwr@pilhuhn.de
- Incorrect setup of the borderBit and nullRegisterBit (different for big and
little endian machines) fixed;
*_BORDER_BITand*NULL_REGISTER_BITredefined - don’t send
pim_asserton tunnels or register vifs (if for whatever reason we receive on such interface) - ignore
WRONGVIFmessages for register and tunnel vifs (the cleaned up kernel mods dont send such signal, but the older (before May 9 ‘98) pimd mods that signaling was enabled
vif.c:find_vif_direct_local(): New function, used inroutesock.c,igmp_proto.c- Use
MFC_MOVE_FORCE/MFC_MOVE_DONT_FORCEflag inmrt.c,route.c,pim_proto.c, when need to move the kernel cache entries between (*,*,RP), (*,G), (S,G) - new timer related macros:
SET_TIMER(),FIRE_TIMER(),IF_TIMER_SET(),IF_TIMER_NOT_SET()
timer.c:age_routes(): bunch of fixes regarding J/P message fragmentationroute.c:process_wrong_iif(): (S,G) SPT switch bug fix: ANDedMRTF_RPfixed toMRTF_RPpim_proto.c&timer.c: (S,G) Prune now is sent toward RP, when iif toward S and iif toward RP are differentpim_proto.c:join_or_prune()bug fixespim_proto.c: (S,G)Prune entry’s timer now set to J/P message holdtimepim_proto.c:receive_pim_join_prune(): Ensure pruned interfaces are correctly reestablishedtimer.c:age_routes(): now (S,G) entry with local members (inherited from (*,G)) is timeout propertlytimer.c:age_routes(): (S,G) J/P timer restarted propertlytimer.c:age_routes(): check also the (S,G)RPbit entries in the forwarders and RP and eventually switch to the shortest path if data rate too highroute.c:process_wrong_vif()fire J/P timerroute.c:switch_shortest_path(): reset the iif toward S if there is already (S,G)RPbit entry
Temp. non-public release.
- `interval` can be applied for data rate check. The statement in
pimd.confthat only the default value will be used is not true anymore. - The RP-initiated and the forwarder-initiated (S,G) switch threshold rate can be different.
pim_proto.c:receive_pim_register(): check if I am the RP for that group, and if “no”, sendPIM_REGISTER_STOP(XXX: not in the spec, but should be!)pim_proto.c:receive_pim_register_stop(): check if thePIM_REGISTER_STOPoriginator is really the RP, before suppressing the sending of the PIM registers. (XXX: not in the spec but should be there)rp.c:check_mrtentry_rp(): new function added to check whether the RP address is the corresponding one for the given mrtentrydebug.c:dump_mrt()timer values addedroute.c:add_leaf(),process_cache_miss(),process_wrong_iif()no routing entries created for the LAN scoped addressesDEBUG_DVMRP_DETAILandDEBUG_PIM_DETAILadded
mrt.c:add_kernel_cache(): no kernel cache duplicatesmrt.c:move_kernel_cache(): if the iif of the (*,*,R) (or (*,G)) and (S,G) are different, dont move the cache entry “UP”timer.c:age_routes(): (S,G)add_jp_entry()flag fixed, SPT switch related.kern.c:k_get_sg_cnt(): modified to compensate for the kernel’s return code bug for getting (S,G) byte count (SIOCGETSGCNT)pim_proto.c:receive_pim_register(): if the (S,G) oif is NULL, now checks whether the iif isregister_vif
- “non-commersial” statement deleted from the copyright message
- mrinfo support added
- mtrace support added (not completed and not enough tested)
- if invalid local address for
cand_rporcand_bootstrap_routerinpimd.conf, automatically will use the largest local multicast enabled address - “include” directory for FreeBSD and SunOS added, so now pimd can be compiled without having the necesary “include” files added to your system. Probably a bad idea and may remove it later.
- some default values for the IP headers of the IGMP and PIM packets are fixed
VIFF_PIM_NBRandVIFF_DVMRP_NBRflags addedVIFF_REGISTERnow included in the RSRR vifs reportfind_route()debug messages removed- #ifdef for
HAVE_SA_LENcorrected debug.c: small fixes
- BSDI related bug fix in defs.h
- small changes in Makefile
- RSRR support for (*,G) completed
- BSDI 3.0/3.1 support by Hitoshi Asaeda mailto:asaeda@yamato.ibm.co.jp (the kernel patches will be available soon)
- Improved debug messages format (thanks to Hitoshi Asaeda)
- A new function
netname()for network IP address print instead ofinet_fmts(), thanks to Hitoshi Asaeda. pimd.conf: format changed
- Remove the inherited leaves from (S,G) when a receiver drops membership
- some parameters when calling
change_interface()fixed - use
send_pim_null_register+ take the appropriate action when the register suppression timer expires - bug fix related to choosing the largest local IP address for little endian machines.
main.c:main(): startup message fixtimer.c:age_routes(): bug fix in debug code
- Minor changes, so pimd now compiles for SunOS 4.1.3 (cc, gcc)
pim_proto.csend_periodic_pim_join_prune(): bug fix thanks to SunOS cc warning(!), only affects the (*,*,RP) stuff.pimd.conf: two errors, related to the rate limit fixed
Makefile: cleanupdefs.h: cleanuproutesock.c: cleanup
igmp_proto.c:accept_group_report(): bug fixespim_proto.c:receive_pim_hello(): bug fixesroute.c:change_interfaces(): bug fixesrp.c: bug fixes ininit_rp_and_bsr(),add_cand_rp(), andcreate_pim_bootstrap_message()
Makefile: “make diff” code addeddebug.c: debug output slightly changed
defs.h:*TIMEOUT(): definitions fixedroute.c: bugs fixed inchange_interface()andswitch_shortest_path()timer.c:age_routes(): number of bugs fixed
First alpha version of the “new, up to date” pimd. RSRR support + Solaris support added. Many functions rewritten and/or modified.