Paul Jakma [Sat, 9 Jan 2010 16:15:00 +0000 (16:15 +0000)]
lib: Add a command to clear the thread CPU history data
* (general) this can be useful when investigating thread latency problems,
when you don't want to have to restart a daemon between tests.
* thread.c: (cpu_record_(hash_)clear) wipe the stored thread cpu history
data, according to the filter, similar to the vty print code.
(clear_thread_cpu_cmd) new command to clear data.
* thread.h: export new command
* command.c: install it
Paul Jakma [Mon, 11 Jan 2010 16:33:07 +0000 (16:33 +0000)]
lib: Thread scheduler should be fair and not let events starve I/O and timers
* thread.c: (thread_fetch) the current scheduler will service events
indefinitely, ignoring I/O and timers, so long as there are events.
In other words, events can crowd out I/O and timers. In theory this
shouldn't be a huge problem as events are generated only by timers
and I/O, however in practice it means normal-load behaviour is not as
useful a predictor of high-load behaviour as it should be.
Fix this by considering all the kinds of threads, in every run of the
scheduler. For any given run, we prioritise events, however across
runs the scheduler should be fair.
This has been observed to give more stable inter-packet times in
testing of ospfd (i.e. lower std-dev).
(thread_process) new heler to queue all the given threads onto the ready
list
Paul Jakma [Mon, 11 Jan 2010 16:30:45 +0000 (16:30 +0000)]
lib: thread history funcname shouldn't be constant, it's freed
* thread.h: (struct cpu_thread_history.funcname) malloc gets called on this
so it should not be const
* thread.c: (cpu_record_print) cast the (const char *) to (char *), this
function knows what it's doing (free wont be called on it).
Greg Troxel [Wed, 3 Nov 2010 11:37:23 +0000 (07:37 -0400)]
infrastructure: Express preference for published git repos
* HACKING: Express notion that a published git repository is
preferred. Fold request for commit message into patch section.
Express desire for comments in code explaining correctness of
post-commit state, and for commit message to explain correctness of
the change.
Dmitry Tejblum [Mon, 18 Oct 2010 15:05:39 +0000 (19:05 +0400)]
zclient: fix router-id calculation for IPv6 (#595)
If router-id is not specified in ospf6d.conf, ospf6d will get it from
the zebra daemon. But ospf6d originates Link LSAs before the router-id
is returned by zebra, thus this router's Link LSAs will be flooded
with AdvRouter set to 0.
* zclient.c: zclient_start(): send ZEBRA_INTERFACE_ADD message after
ZEBRA_ROUTER_ID_ADD, not before
Michael Lambert [Thu, 22 Jul 2010 17:20:55 +0000 (13:20 -0400)]
bgpd, lib: adopt afi_t and safi_t in several places
* bgpd/bgp_attr.c, bgpd/bgp_open.h, bgpd/bgp_route.c, lib/prefix.c,
lib/prefix.h: Various integer types were being used where, if we
had strict type checking, afi_t and safi_t would be required.
bgpd/bgp_packet.c:bgp_update_packet(): When extracting the peer, don't
fail to extract it because "binfo->extra" is NULL. While one should
certainly avoid dereferencing binfo->extra, that's not a good reason
not to use binfo->peer.
Fixes https://bugzilla.quagga.net/show_bug.cgi?id=497.
Patch by Eric Sobocinksi.
Chris Hall [Mon, 9 Aug 2010 18:31:37 +0000 (22:31 +0400)]
bgpd: fix handling of AS path data
* bgpd/bgp_aspath.c
* assegments_parse(): add handling of AS4_PATH input, update bounds
checks, add check for AS segment type
* aspath_parse(): add handling of AS4_PATH input, expect
assegments_parse() to do length checking
* aspath_empty(): update for the new function prototype
* bgpd/bgp_aspath.h: ditto
* tests/aspath_test.c: ditto
* bgpd/bgp_attr.c
* bgp_attr_aspath(): add handling of AS4_PATH input, update flags
checks, change returned type
* bgp_attr_as4_path(): discard, superseded by bgp_attr_aspath()
* bgp_attr_parse(): update respectively
Chris Hall [Fri, 14 May 2010 12:38:39 +0000 (16:38 +0400)]
bgpd: tighten bounds checking in RR ORF msg reader
* bgp_packet.c: (bgp_route_refresh_receive) add validation of
"Length" (RFC5292) field value, check input stream bounds
each time bytes are pulled from it
David Lamparter [Tue, 2 Feb 2010 20:49:35 +0000 (21:49 +0100)]
isisd: change ISIS_METHOD to use C preprocessor
this fixes warnings from vtysh extract.pl by making sure the isis method
files always 'work'. (previously, extract.pl would grab unselected isis
method sources and then complain about missing headers)
David Ward [Tue, 5 Jan 2010 02:45:39 +0000 (02:45 +0000)]
ospf6d: Fix crash when '[no] ipv6 ospf6 advertise prefix-list' is in startup-config
* ospf6_interface.c: When '[no] ipv6 ospf6 advertise prefix-list'
appears in the startup configuration for ospf6d, a crash occurs,
because ospf6d attempts to schedule LSAs when the 'oi->area'
structure has not yet been initialized.
Now, when the command above is issued (either in the startup
configuration or at runtime), ospf6d will only schedule LSAs if
the 'oi->area' structure has been initalized. A similar test is
already used when handling the commands 'ipv6 ospf6 priority'
and 'ipv6 ospf6 cost'.
* ospf_zebra.c: (ospf_distribute_list_update_timer) forces a
refresh of default route each time it finds a default prefix.
This is suboptimal, just record that it needs to be done and
do it once.
ospfd: Make sure ospf_distribute_list_update_timer() eventually runs.
* ospf_zebra.c: (ospf_distribute_list_update_timer)
If there are updates to the distribute list every 5 second or less,
ospf_distribute_list_update_timer() will never run as the timer gets
rearmed for each update. This fixes it by never rearming an active
distribute list timer.
When adding a connected route (using vtysh, without restart) to the
redistribution access list of ospfd, while static routes already exist,
the update timer ospf_distribute_list_update_timer() is being run for
static routes only. That way, the connected route never appears in the
OSPF database, until quagga is completely restarted.
The update timer for connected routes is cancelled in
ospfd/ospfd_zebra.c:ospf_distribute_list_update():976, were a new timer
is scheduled for static routes, caused by the loop in ospf_filter_update().
* ospf_zebra.c: (ospf_distribute_list_update_timer) make it
refresh all external routes. This fixes the problem
reported by Roman.
Stephen:
A recent change breaks build Quagga on Debian Lenny with amd64 (but not
i386). Not sure whether this is cause by ld, gcc, or libtool; but
whatever it just won't work for a large number of users.
Mathias:
I would guess it's a problem of libtool because it adds "-fPIC -DPIC" to
the compiler flags but doesn't remove -fPIE. That wouldn't be a problem
if the compiler would ignore the -fPIE in favor to the latter -fPIC, but
obviously it does not. [...] Those objects were actually compiled with
-fPIC but also -fPIE which seems to confuse gcc.
Mathias Krause [Wed, 20 Jan 2010 14:39:13 +0000 (17:39 +0300)]
lib: fix PIE build flags
Because the final executables are built as position independent
executables (PIE) -- when configure has detected compiler supported for
PIE -- the objects in the library archive must be built in that way,
too. Otherwise the runtime linker has to do unneccesary relocation for
each start of the program. Even worse, the programs won't even be able
to start on a (hardened) kernel that doesn't allow those relocations to
happen by preventing making the .text segment writable (PaX's MPROTECT
feature comes to mind). The attached patch fixes this issue by adding
the appropriate flags to the Makefile.am.
BGP uses time() to get system time of day; but that value
fluctuates with time adjustments from NTP. This can cause premature
flapping of peer sessions and other failures.
Use the system monotonic clock supported by Quagga thread library
to avoid issue.
* bgpd/bgp_fsm.c
* bgp_uptime_reset(): dismiss function
* bgpd/bgpd.c
* bgp_clock(): new function
* bgpd/bgp_damp.c
* bgp_reuse_timer(): employ bgp_clock() instead of time(NULL)
* bgp_damp_withdraw(): idem
* bgp_damp_update(): idem
* bgp_damp_scan(): idem
* bgp_damp_info_vty(): idem
* bgp_damp_reuse_time_vty(): idem
* bgpd/bgp_fsm.c
* bgp_routeadv_timer(): idem
* bgp_stop(): idem
* bgp_establish(): idem
* bgpd/bgp_packet.c
* bgp_update_receive(): idem
* bgpd/bgp_route.c
* bgp_update_rsclient(): idem
* bgp_update_main(): idem
* bgp_static_update_rsclient(): idem
* bgp_static_update_main(): idem
* bgp_static_update_vpnv4(): idem
* bgp_aggregate_route(): idem
* bgp_aggregate_add(): idem
* bgp_redistribute_add(): idem
* bgpd/bgp_snmp.c
* bgpPeerTable(): idem
* bgpTrapEstablished(): idem
* bgpTrapBackwardTransition(): idem
* bgpd/bgpd.c
* peer_create(): idem
* peer_uptime(): idem
* bgp_master_init(): idem
Paul Jakma [Thu, 14 Jan 2010 13:19:40 +0000 (16:19 +0300)]
ospfd: remove unneeded memset from a very hot function
* ospf_lsdb.c: (lsdb_prefix_set) memset is unneeded, as all fields are
initialised explicitly, and this function can be in the top-3 of a
profile when there are a lot of LSAs.
David Ward [Wed, 13 Jan 2010 17:10:56 +0000 (20:10 +0300)]
lib: fix memory logging
* lib/memory.h
* mtype_zcalloc(): correct function prototype
* XFREE(): make both definitions consistent in setting
the pointer to NULL after freeing the memory
These changes will only have an effect if MEMORY_LOG is defined
(it is not by default).
The router-id table looks like is supposed to be sorted in current
quagga code, but the nodes are not added with the sorting
function.
The sorting function is host byte order dependent.
The values need to converted before comparison.
Fixing this causes Zebra to choose the largest IP address
as router-id, rather than the last address. This probably will
surprise some users. The other option would be to just remove the
comparison function and keep the existing LIFO behavior.
Lastly, simple subtraction works well for comparing.
* isisd/isis_pdu.c
* maskbit: this pre-initialized array is not modified
* lib/prefix.c
* maskbit: idem
* lib/command.c
* facility_map: idem
* itoa64: idem
* default_motd: make local var static
* facility_name(): update local var accordingly
* facility_match(): idem
* bgpd/bgp_aspath.c
* ashash: only used in one file, make static
* aspath_count_numas(): dead code, sayonara
* bgpd/bgpd.c
* peer_nsf_stop(): only used in one file, make static
* bgpd/bgp_packet.h
* bgp_capability_receive(): add missing prototype for a
global function
* ripd/rip_interface.c
* rip_request_neighbor(): comment out, unused
* rip_request_neighbor_all(): idem
* rip_interface_up(): Cast flags otherwise compiler complains
about %lld not matching uint64_t on 64 bit x86. Print in hex
since flags are bit field.
* rip_interface_add(): idem
* rip_interface_delete(): idem
* ripd/rip_zebra.c
* rip_redistribute_set(): comment out, unused
* ripd/ripd.h
* rip_redistribute_check(): move prototype here so compiler
can check function against prototype
* ripd/ripd.c
* rip_update_default_metric(): comment out, unused
* zebra/irdp.h
* irdp_send_thread(): move prototype to common header file
* irdp_advert_off(): idem
* process_solicit(): idem
* irdp_read_raw(): idem
* send_packet(): idem
* zebra/irdp_interface.c
* inet_2a(): move function to where it is used
* in_cksum(): lib/checksum.h already provides prototype
* irdp_send_thread(): prototype moved away
* inet_2a(): idem
* irdp_advert_off(): idem
* b1, b2, b3, b4: get rid of global buffers for inet_2a()
* if_group(): add local buffer for inet_2a()
* if_add_group(): idem
* if_drop_group(): idem
* irdp_config_write(): idem
* Adv_new(): make static
* zebra/irdp_main.c
* irdp_read_raw(): prototype moved away
* send_packet(): idem
* in_cksum(): lib/checksum.h already provides prototype
* inet_2a(): function moved to irdp_interface.c
* zebra/irdp_packet.c
* b1, b2, b3, b4: get rid of global buffers for inet_2a()
* in_cksum(): lib/checksum.h already provides prototype
* process_solicit(): prototype moved away
* irdp_read_raw(): fix uninitialized variable
* bgpd/bgp_fsm.c
* bgp_clearing_completed(): only used in one file, can be static
* bgpd/bgp_packet.c
* afi2str(): sayonara
* safi2str(): sayonara
* bgpd/bgp_route.c
* bgp_distance_reset(): sayonara
* bgpd/bgp_zebra.c
* bgp_ifindex_by_nexthop(): sayonara
Make one version of check prefix bit, and put it inline
with proper prototype. This gets rid of some macro's and also some
assert() that can never happen on a non-broken compiler.
* bgpd/bgp_table.c
* CHECK_BIT(): sayonara
* check_bit(): sayonara
* SET_LINK(): sayonara
* set_link(): make use of prefix_bit() instead of check_bit()
* bgp_node_match(): idem
* bgp_node_lookup(): idem
* bgp_node_get(): idem
* lib/prefix.h
* prefix_bit(): new inline version of check_bit()
* lib/table.c
* CHECK_BIT(): sayonara
* check_bit(): sayonara
* SET_LINK(): sayonara
* set_link(): make use of prefix_bit() instead of check_bit()
* route_node_match(): idem
* route_node_lookup(): idem
* route_node_get(): idem
* ospf6d/ospf6_lsdb.c
* CHECK_BIT(): sayonara
* ospf6_lsdb_lookup_next(): make use of prefix_bit() instead of
CHECK_BIT()
* ospf6_lsdb_type_router_head(): idem
* ospf6_lsdb_type_head(): idem
* ospf6d/ospf6_route.c
* CHECK_BIT(): sayonara
* ospf6_route_match_head() make use of prefix_bit() instead of
* CHECK_BIT()
* md5.c
* md5_calc(): If building on little endian then X
will be 'const uint32_t' which causes compiler
warning in code that never gets used. Just move
the endif to make sure code is not checked.
* zebra/rtadv.c
* rtadv_free(): remove unused function
* rtadv_recv_packet(): break up cast/dereference to avoid
compiler type pun warning
* rtadv_read(): initialize ifindex, because compiler can't
figure out that it is okay
* zebra/zebra_rib.c
* nexthop_active_update(): make local int vars unsigned
* nexthop_active_check(): return unsigned for consistency
* rib_dump(): cast time in printf format
* vrf_free(): remove unused function
* vrf_lookup_by_name(): idem
* rib_if_up(): idem
* rib_if_down(): idem
* table.c: general type safety and compiler help:
* maskbit[]: become const
* route_node_match(): take const args
* route_node_match_ipv4(): idem
* route_node_match_ipv6(): idem
* check_bit(): idem, plus adjust local vars typing
Ondrej Zajicek [Mon, 7 Dec 2009 09:33:20 +0000 (12:33 +0300)]
ospf6d: review LSA sequence number comparison
It seems that there is a bug in ospf6d in ospf6_lsa_compare(): If LSA A
has sequence number smaller than 0x80000000 and LSA B has sequence
number larger than 0x80000000, ospf6_lsa_compare() returns that B is
more recent than A, although RFC says that sequence numbers should be
compared as signed numbers (0x8000001 smallest and 0x7FFFFFFF largest).
In ospfd, the function ospf_lsa_more_recent() has it right.
The problem appears when Quagga is used together with OSPFv3 in
development version of BIRD daemon ( http://bird.network.cz/ ),
which creates LSAs with maximum sequence number (0x7FFFFFFF)
as a part of flushing/premature aging LSA from OSPF area.
Because both daemons has different idea of which LSA instance
is more recent, it would lead to LSA storm.
David Ward [Thu, 3 Dec 2009 18:43:11 +0000 (21:43 +0300)]
zebra: handle RTF_CLONING removal from FreeBSD 8.0
The RTF_CLONING flag has been removed in FreeBSD 8.0 with arp-v2 (see
http://svn.freebsd.org/viewvc/base/stable/8/UPDATING?view=markup),
since it no longer has any meaning. This patch checks if RTF_CLONING
exists before using it.
This code is only used one place and can be made local.
Gcc is smart enough to inline local functions if it wants to.
The function also has a big chunk of compatiablity code that
is no longer used; since quagga is now in a version control system
the source does not need to be used as a historical reference.
Tom Goff [Tue, 1 Dec 2009 18:12:38 +0000 (21:12 +0300)]
ospf6d: fix LSA locking in ospf6_new_ls_id()
* ospf6_lsdb.c: (ospf6_new_ls_id) Unlock the current LSA when breaking
out of the ospf6_lsdb_*_head() / ospf6_lsdb_*_next() loop early. No
explicit unlocking is needed when all LSAs are looped through
because ospf6_lsdb_*_next() manages everything in that case.
* bgp_network.c: (bgp_md5_set) Missing piece from earlier listener change
did not get ported from Vyatta code into upstream. The list
listener_sockets changed from (int *) to (struct bgp_listener *).
ospfd: Tighten up the connected check for redistribution
* ospf_zebra.c: (ospf_distribute_check_connected) check to make filter out
routes matching connected routes was matching against OSPF networks, which
can be far more general than the actual connected interfaces. Fix.
* bgp_network.c: (bgp_accept) The code in current git will crash as part of
the revised listener code is missing. The new listener thread code passes
a pointer to a bgp_listener structure, not the bgp pointer. The old code
always got a NULL for bgp pointer, so that is now hard coded.
ospfd: fix performance problem with redistribute delete
Doing redistribute delete with full BGP table was taking
30 minutes, this drops it down to less than a second.
* ospf_lsa.c: (ospf_lsa_maxage) When flushing lots of entries the
performance is terrible because it looks up each LSA entry through
ospf_lsa_maxage_exist before deleting causing O(N^2) performance. Use a
new OSPF_LSA_MAXAGE flag instead of scan - and maintain it.
(ospf_lsa_maxage_exist) removed
(ospf_lsa_maxage_delete) maintain OSPF_LSA_MAXAGE flag
In some cases ospfd does not recalc the route table. This
happens when ospfd receives an old LSA which will trigger
recalc but the this recalc will fail because all interfaces
isn't up yet. Next LSA that is originated matches the old one
so no recalc will be performed. This problem has been observed
when there are only 2 ppp I/Fs in an area, both go down at the
same time, then they come up again with a few seconds apart.
* ospf_lsa.c: (ospf_{router,network}_lsa_install) avoid a needless scheduling
of SPF.
(ospf_lsa_different) fix bug in LSA comparison that would lead to the
described failure to schedule SPF.
Paul Jakma [Tue, 11 Aug 2009 11:25:42 +0000 (12:25 +0100)]
ospfd: Make "Packet ... received on wrong link" conditional on debug
* ospf_packet.c: make this message conditional on 'debug ospf event', as it
be easily triggered with, e.g., multiple subnets sharing same physical
network. E.g, see bug #532.
Paul Jakma [Wed, 5 Aug 2009 15:25:16 +0000 (16:25 +0100)]
bgpd: Fix mistakes in applying 'allow inbound connections to non-default view'
* bgpd.c: (peer_lookup_with_open) Bodged application of previous patch
meant the second loop around bgp->peer wasn't included in the loop
around bm->bgp as it was supposed to be. Fix..
Paul Jakma [Mon, 3 Aug 2009 15:34:16 +0000 (16:34 +0100)]
ospfd: neuter some compiler warnings
* ospf_lsa.c: (link_info_set) Use %zd for size_t - C99 is old enough now.
Lots of similar warnings all over the code.
(ospf_lsa_translated_nssa_compare) Unused func - delete.
* global: In struct ospf_path, change struct ospf_interface *oi to int
ifindex. It is unsafe to reference *oi as an ospf interface can be
deleted under your feet. Use a weak reference instead.
ospfd: Discriminate better when selecting links between vertices in SPF
* ospf_spf.c: (ospf_get_next_link) One must check the vertex type, Router or
Network, to select type link to match against. Link type 1 has neighbour
router ID in link_id and link type 2 has IP address of DR. Since router
id may have same value as an existing IP address one risks matching a
router ID against a DR.
Michael Lambert [Tue, 28 Jul 2009 15:26:14 +0000 (11:26 -0400)]
bgpd: Workaround for invalid MBGP next hop
* bgp_attr.c: (bgp_mp_reach_parse) There are some interoperability issues
for MBGP (particularly IPv4 multicast NLRI) between different
implementations. In order to get some next hops to install correctly in
the BGP tables, it appears to be necessary to copy the multiprotocol next
hop into the base next hop field. This is related to differences in RFC
2283 and RFC 2858.
ospfd: Make ospf_if_lookup_recv_if() find the right unnumbered i/f
This function will return the interface for the first matching
remote address for PtP i/f's. That won't work for multiple
unnumbered i/f's as these may all have the same address.
Pass in the struct interface pointer, ifp, to find the
correct set of oi's to search in. This also reduces the
size of the search list, making it faster.
* ospfd/ospf_interface.c: Add struct interface * param to
ospf_if_lookup_recv_if() to select the right list to search in.
* ospfd/ospf_interface.h: ditto.
* ospfd/ospf_packet.c: Pass new ifp argument to ospf_if_lookup_recv_if()
ospfd: export ospf_if_table_lookup() and use it in ospf_network_run_interface()
Makes it possible to run OSPF on multiple PtP interfaces
with the same remote address.
* ospfd/ospf_interface.c: Export ospf_if_table_lookup().
* ospfd/ospf_interface.h: ditto.
* ospfd/ospfd.c: (ospf_network_run_interface) Use ospf_if_table_lookup() to
determine whether OSPF is already configured for a subnet and interface.