1 tbd Sun Microsystems, Inc.
3 * Support for networks:
4 socklnd - any kernel supported by Lustre,
5 qswlnd - Qsnet kernel modules 5.20 and later,
6 openiblnd - IbGold 1.8.2,
7 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, 1.3, and 1.4.1
8 viblnd - Voltaire ibhost 3.4.5 and later,
9 ciblnd - Topspin 3.2.0,
10 iiblnd - Infiniserv 3.3 + PathBits patch,
11 gmlnd - GM 2.1.22 and later,
12 mxlnd - MX 1.2.1 or later,
13 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
22 Description: router_proc.c is rewritten to use sysctl-interface for parameters
23 residing in /proc/sys/lnet
25 Severity : enhancement
27 Description: port router pinger to userspace
31 Description: kptllnd HELLO protocol deadlock
32 Details : kptllnd HELLO protocol doesn't run to completion in finite time
36 Description: LNet selftest fixes and enhancements
38 Severity : enhancement
40 Description: allow a test node to be a member of multiple test groups
42 Severity : enhancement
44 Description: MXLND: eliminate hosts file, use arp for peer nic_id resolution
45 Details : an update from the upstream developer Scott Atchley.
47 Severity : enhancement
49 Description: add a new LND optiion to control peer buffer credits on routers
53 Description: Fixing deadlock in usocklnd
54 Details : A deadlock was possible in usocklnd due to race condition while
55 tearing connection down. The problem resulted from erroneous
56 assumption that lnet_finalize() could have been called holding
60 Bugzilla : 13621, 15983
61 Description: Protocol V2 of o2iblnd
62 Details : o2iblnd V2 has several new features:
63 . map-on-demand: map-on-demand is disabled by default, it can
64 be enabled by using modparam "map_on_demand=@value@", @value@
65 should >= 0 and < 256, 0 will disable map-on-demand, any other
66 valid value will enable map-on-demand.
67 Oi2blnd will create FMR or physical MR for RDMA if fragments of
69 Enable map-on-demand will take less memory for new connection,
70 but a little more CPU for RDMA.
71 . iWARP : to support iWARP, please enable map-on-demand, 32 and 64
72 are recommanded value. iWARP will probably fail for value >=128.
73 . OOB NOOP message: to resolve deadlock on router.
74 . tunable peer_credits_hiw: (high water to return credits),
75 default value of peer_credits_hiw equals to (peer_credits -1),
76 user can change it between peer_credits/2 and (peer_credits - 1).
77 Lower value is recommended for high latency network.
78 . tunable message queue size: it always equals to peer_credits,
79 higher value is recommended for high latency network.
80 . It's compatible with earlier version of o2iblnd
84 Description: Fixing 'running out of ports' issue
85 Details : Add a delay before next reconnect attempt in ksocklnd in
86 the case of lost race. Limit the frequency of query-requests
87 in lnet. Improved handling of 'dead peer' notifications in
92 Description: Change ptllnd timeout and watchdog timers
93 Details : Add ptltrace_on_nal_failed and bump ptllnd timeout to match
98 Description: One down Lustre FS hangs ALL mounted Lustre filesystems
99 Details : Shared routing enhancements - peer health detection.
101 Severity : enhancement
103 Description: acceptor.c cleanup
104 Details : Code duplication in acceptor.c for the cases of kernel and
105 user-space removed. User-space libcfs tcpip primitives
106 uniformed to have prototypes similar to kernel ones. Minor
107 cosmetic changes in usocklnd to use cfs_socket_t as
108 representation of socket.
112 Description: IB path MTU mistakenly set to 1st path MTU when ib_mtu is off
113 Details : See comment 46 in bug 11245 for details - it's indeed a bug
114 introduced by the original 11245 fix.
118 Description: uptllnd credit overflow fix
119 Details : kptl_msg_t::ptlm_credits could be overflown by uptllnd since
124 Description: socklnd protocol version 3
125 Details : With current protocol V2, connections on router can be
126 blocked and can't receive any incoming messages when there is no
127 more router buffer, so ZC-ACK can't be handled (LNet message
128 can't be finalized) and will cause deadlock on router.
129 Protocol V3 has a dedicated connection for emergency messages
130 like ZC-ACK to router, messages on this dedicated connection
131 don't need any credit so will never be blocked. Also, V3 can send
132 keepalive ping in specified period for router healthy checking.
134 -------------------------------------------------------------------------------
136 12-31-2008 Sun Microsystems, Inc.
138 * Support for networks:
139 socklnd - any kernel supported by Lustre,
140 qswlnd - Qsnet kernel modules 5.20 and later,
141 openiblnd - IbGold 1.8.2,
142 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
143 viblnd - Voltaire ibhost 3.4.5 and later,
144 ciblnd - Topspin 3.2.0,
145 iiblnd - Infiniserv 3.3 + PathBits patch,
146 gmlnd - GM 2.1.22 and later,
147 mxlnd - MX 1.2.1 or later,
148 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
152 Description: workaround for OOM from o2iblnd
153 Details : OFED needs allocate big chunk of memory for QP while creating
154 connection for o2iblnd, OOM can happen if no such a contiguous
156 QP size is decided by concurrent_sends and max_fragments of
157 o2iblnd, now we permit user to specify smaller value for
158 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
159 will decrease memory block size required by creating QP.
163 Description: Support Zerocopy receive of Chelsio device
164 Details : Chelsio driver can support zerocopy for iov[1] if it's
165 contiguous and large enough.
169 Description: fix credit flow deadlock in uptllnd
173 Description: finalize network operation in reasonable time
174 Details : conf-sanity test_32a couldn't stop ost and mds because it
175 tried to access non-existent peer and tcp connect took
176 quite long before timing out.
180 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
181 Details : Lost reference on conn prevents peer from being destroyed, which
182 could prevent new peer creation if peer count has reached upper
187 Description: LNET Selftest results in Soft lockup on OSS CPU
188 Details : only hits when 8 or more o2ib clients involved and a session is
189 torn down with 'lst end_session' without preceeding 'lst stop'.
193 Description: concurrent_sends in IB LNDs should not be changeable at run time
194 Details : concurrent_sends in IB LNDs should not be changeable at run time
198 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
199 Details : only hits under out-of-memory situations
202 -------------------------------------------------------------------------------
204 2009-02-07 Sun Microsystems, Inc.
206 * Support for networks:
207 socklnd - any kernel supported by Lustre,
208 qswlnd - Qsnet kernel modules 5.20 and later,
209 openiblnd - IbGold 1.8.2,
210 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
211 viblnd - Voltaire ibhost 3.4.5 and later,
212 ciblnd - Topspin 3.2.0,
213 iiblnd - Infiniserv 3.3 + PathBits patch,
214 gmlnd - GM 2.1.22 and later,
215 mxlnd - MX 1.2.1 or later,
216 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
219 Description: workaround for OOM from o2iblnd
220 Details : OFED needs allocate big chunk of memory for QP while creating
221 connection for o2iblnd, OOM can happen if no such a contiguous
223 QP size is decided by concurrent_sends and max_fragments of
224 o2iblnd, now we permit user to specify smaller value for
225 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
226 will decrease memory block size required by creating QP.
230 Description: Support Zerocopy receive of Chelsio device
231 Details : Chelsio driver can support zerocopy for iov[1] if it's
232 contiguous and large enough.
235 Description: fix credit flow deadlock in uptllnd
239 Description: finalize network operation in reasonable time
240 Details : conf-sanity test_32a couldn't stop ost and mds because it
241 tried to access non-existent peer and tcp connect took
242 quite long before timing out.
246 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
247 Details : Lost reference on conn prevents peer from being destroyed, which
248 could prevent new peer creation if peer count has reached upper
253 Description: LNET Selftest results in Soft lockup on OSS CPU
254 Details : only hits when 8 or more o2ib clients involved and a session is
255 torn down with 'lst end_session' without preceeding 'lst stop'.
259 Description: concurrent_sends in IB LNDs should not be changeable at run time
260 Details : concurrent_sends in IB LNDs should not be changeable at run time
262 -------------------------------------------------------------------------------
264 11-03-2008 Sun Microsystems, Inc.
266 * Support for networks:
267 socklnd - any kernel supported by Lustre,
268 qswlnd - Qsnet kernel modules 5.20 and later,
269 openiblnd - IbGold 1.8.2,
270 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
271 viblnd - Voltaire ibhost 3.4.5 and later,
272 ciblnd - Topspin 3.2.0,
273 iiblnd - Infiniserv 3.3 + PathBits patch,
274 gmlnd - GM 2.1.22 and later,
275 mxlnd - MX 1.2.1 or later,
276 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
280 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
281 Details : only hits under out-of-memory situations
283 -------------------------------------------------------------------------------
285 04-26-2008 Sun Microsystems, Inc.
287 * Support for networks:
288 socklnd - any kernel supported by Lustre,
289 qswlnd - Qsnet kernel modules 5.20 and later,
290 openiblnd - IbGold 1.8.2,
291 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
292 viblnd - Voltaire ibhost 3.4.5 and later,
293 ciblnd - Topspin 3.2.0,
294 iiblnd - Infiniserv 3.3 + PathBits patch,
295 gmlnd - GM 2.1.22 and later,
296 mxlnd - MX 1.2.1 or later,
297 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
301 Description: excessive debug information removed
302 Details : excessive debug information removed
306 Description: ksocknal_create_conn() hit ASSERTION during connection race
307 Details : ksocknal_create_conn() hit ASSERTION during connection race
311 Description: ksocknal_send_hello() hit ASSERTION while connecting race
312 Details : ksocknal_send_hello() hit ASSERTION while connecting race
316 Description: o2iblnd/ptllnd credit deadlock in a routed config.
317 Details : o2iblnd/ptllnd credit deadlock in a routed config.
321 Description: High load after starting lnet
322 Details : gmlnd should sleep in rx thread in interruptible way. Otherwise,
323 uptime utility reports high load that looks confusingly.
327 Description: ksocklnd fails to establish connection if accept_port is high
328 Details : PID remapping must not be done for active (outgoing) connections
331 --------------------------------------------------------------------------------
333 2008-01-11 Sun Microsystems, Inc.
335 * Support for networks:
336 socklnd - any kernel supported by Lustre,
337 qswlnd - Qsnet kernel modules 5.20 and later,
338 openiblnd - IbGold 1.8.2,
339 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
340 viblnd - Voltaire ibhost 3.4.5 and later,
341 ciblnd - Topspin 3.2.0,
342 iiblnd - Infiniserv 3.3 + PathBits patch,
343 gmlnd - GM 2.1.22 and later,
344 mxlnd - MX 1.2.1 or later,
345 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
349 Description: liblustre network error
350 Details : liblustre clients should understand LNET_ACCEPT_PORT environment
351 variable even if they don't start lnet acceptor.
355 Description: Strange message from lnet (Ignoring prediction from the future)
356 Details : Incorrect calculation of peer's last_alive value in ksocklnd
358 --------------------------------------------------------------------------------
360 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
362 * Support for networks:
363 socklnd - any kernel supported by Lustre,
364 qswlnd - Qsnet kernel modules 5.20 and later,
365 openiblnd - IbGold 1.8.2,
366 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5.
367 viblnd - Voltaire ibhost 3.4.5 and later,
368 ciblnd - Topspin 3.2.0,
369 iiblnd - Infiniserv 3.3 + PathBits patch,
370 gmlnd - GM 2.1.22 and later,
371 mxlnd - MX 1.2.1 or later,
372 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
376 Description: ASSERTION(me == md->md_me) failed in lnet_match_md()
380 Description: increase send queue size for ciblnd/openiblnd
384 Description: new userspace socklnd
385 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
386 with new one - usocklnd.
388 Severity : enhancement
390 Description: Console message flood
391 Details : Make cdls ratelimiting more tunable by adding several tunable in
392 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
393 /proc/sys/lnet/console_backoff.
395 --------------------------------------------------------------------------------
397 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
399 * Support for networks:
400 socklnd - any kernel supported by Lustre,
401 qswlnd - Qsnet kernel modules 5.20 and later,
402 openiblnd - IbGold 1.8.2,
403 o2iblnd - OFED 1.1 and 1.2,
404 viblnd - Voltaire ibhost 3.4.5 and later,
405 ciblnd - Topspin 3.2.0,
406 iiblnd - Infiniserv 3.3 + PathBits patch,
407 gmlnd - GM 2.1.22 and later,
408 mxlnd - MX 1.2.1 or later,
409 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
413 Description: /proc/sys/lnet has non-sysctl entries
414 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
418 Description: TOE Kernel panic by ksocklnd
419 Details : offloaded sockets provide their own implementation of sendpage,
420 can't call tcp_sendpage() directly
424 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
425 Details : races between lnd_shutdown and peer creation prevent
426 lnd_shutdown from finishing.
430 Description: open files rlimit 1024 reached while liblustre testing
431 Details : ulnds/socklnd must close open socket after unsuccessful
436 Description: build error
437 Details : fix typos in gmlnd, ptllnd and viblnd
439 --------------------------------------------------------------------------------
441 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
443 * Support for networks:
444 socklnd - kernels up to 2.6.16,
445 qswlnd - Qsnet kernel modules 5.20 and later,
446 openiblnd - IbGold 1.8.2,
447 o2iblnd - OFED 1.1 and 1.2
448 viblnd - Voltaire ibhost 3.4.5 and later,
449 ciblnd - Topspin 3.2.0,
450 iiblnd - Infiniserv 3.3 + PathBits patch,
451 gmlnd - GM 2.1.22 and later,
452 mxlnd - MX 1.2.1 or later,
453 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
455 --------------------------------------------------------------------------------
457 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
459 * Support for networks:
460 socklnd - kernels up to 2.6.16,
461 qswlnd - Qsnet kernel modules 5.20 and later,
462 openiblnd - IbGold 1.8.2,
464 viblnd - Voltaire ibhost 3.4.5 and later,
465 ciblnd - Topspin 3.2.0,
466 iiblnd - Infiniserv 3.3 + PathBits patch,
467 gmlnd - GM 2.1.22 and later,
468 mxlnd - MX 1.2.1 or later,
469 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
473 Description: Initialize cpumask before use
477 Description: ASSERTION failures when upgrading to the patchless zero-copy
479 Details : This bug affects "rolling upgrades", causing an inconsistent
480 protocol version negotiation and subsequent assertion failure
481 during rolling upgrades after the first wave of upgrades.
485 Details : Change "dropped message" CERRORs to D_NETERROR so they are
486 logged instead of creating "console chatter" when a lustre
487 timeout races with normal RPC completion.
490 Details : lnet_clear_peer_table can wait forever if user forgets to
494 Details : libcfs_id2str should check pid against LNET_PID_ANY.
498 Description: added LNET self test
499 Details : landing b_self_test
504 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
506 Details : do_div() macro is used incorrectly.
508 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
512 Description: make panic on lbug configurable
516 Description: Add OFED1.2 support to o2iblnd
517 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
518 are installed (other than kernel's in-tree infiniband), there
519 could be some problem while insmod o2iblnd (mismatch CRC of
521 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
522 this link provides solution:
523 https://bugs.openfabrics.org/show_bug.cgi?id=355
524 if extra Module.symvers is not supported in kernel, we will
525 have to run the script in bug 12316 to update
526 $LINUX/module.symvers before building o2iblnd.
527 More details about this are in bug 12316.
529 ------------------------------------------------------------------------------
531 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
532 * version 1.4.10 / 1.6.0
533 * Support for networks:
534 socklnd - kernels up to 2.6.16,
535 qswlnd - Qsnet kernel modules 5.20 and later,
536 openiblnd - IbGold 1.8.2,
538 viblnd - Voltaire ibhost 3.4.5 and later,
539 ciblnd - Topspin 3.2.0,
540 iiblnd - Infiniserv 3.3 + PathBits patch,
541 gmlnd - GM 2.1.22 and later,
542 mxlnd - MX 1.2.1 or later,
543 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
547 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
548 possibly accessed in kptllnd_shutdown. Ptllnd should init
549 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
553 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
557 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
561 Description: the_lnet.ln_finalizing was not set when the current thread is
562 about to complete messages. It only affects multi-threaded
568 Description: Changed the default kqswlnd ntxmsg=512
573 Description: Assertion failure in kernel ptllnd caused by posting passive
574 bulk buffers before connection establishment complete.
579 Description: A race in kernel ptllnd between deleting a peer and posting
580 new communications for it could hang communications -
581 manifesting as "Unexpectedly long timeout" messages.
586 Description: Kernel ptllnd lock ordering issue could hang a node.
591 Description: node crash on socket teardown race
594 Frequency : 'lctl peer_list' issued on a mx net
596 Description: Enable lctl's peer_list for MXLND
599 Frequency : after Ptllnd timeouts and portals congestion
601 Description: Credit overflows
602 Details : This was a bug in ptllnd connection establishment. The fix
603 implements better peer stamps to disambiguate connection
604 establishment and ensure both peers enter the credit flow
605 state machine consistently.
610 Description: kptllnd didn't propagate some network errors up to LNET
611 Details : This bug was spotted while investigating 11394. The fix
612 ensures network errors on sends and bulk transfers are
613 propagated to LNET/lustre correctly.
615 Severity : enhancement
617 Description: Fixed console chatter in case of -ETIMEDOUT.
619 Severity : enhancement
621 Description: Added D_NETTRACE for recording network packet history
622 (initially only for ptllnd). Also a separate userspace
623 ptllnd facility to gather history which should really be
624 covered by D_NETTRACE too, if only CDEBUG recorded history in
630 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
631 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
632 callback can occur before a connection has actually been
633 established. This caused an assertion failure previously.
635 Severity : enhancement
637 Description: Multiple instances for o2iblnd
638 Details : Allow multiple instances of o2iblnd to enable networking over
639 multiple HCAs and routing between them.
643 Description: lnet deadlock in router_checker
644 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
645 into BH locks to eliminate potential deadlock caused by
646 ksocknal_data_ready() preempting code holding these locks.
650 Description: Millions of failed socklnd connection attempts cause a very slow FS
651 Details : added a new route flag ksnr_scheduled to distinguish from
652 ksnr_connecting, so that a peer connection request is only turned
653 down for race concerns when an active connection to the same peer
654 is under progress (instead of just being scheduled).
656 ------------------------------------------------------------------------------
658 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
660 * Support for networks:
661 socklnd - kernels up to 2.6.16
662 qswlnd - Qsnet kernel modules 5.20 and later
663 openiblnd - IbGold 1.8.2
665 viblnd - Voltaire ibhost 3.4.5 and later
666 ciblnd - Topspin 3.2.0
667 iiblnd - Infiniserv 3.3 + PathBits patch
668 gmlnd - GM 2.1.22 and later
669 mxlnd - MX 1.2.1 or later
670 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
673 Severity : major on XT3
675 Description: libcfs overwrites /proc/sys/portals
676 Details : libcfs created a symlink from /proc/sys/portals to
677 /proc/sys/lnet for backwards compatibility. This is no
678 longer required and makes the Cray portals /proc variables
683 Description: OFED FMR API change
684 Details : This changes parameter usage to reflect a change in
685 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
686 that FMR support is only used in experimental versions of the
687 o2iblnd - this change does not affect standard usage at all.
689 Severity : enhancement
691 Description: new ko2iblnd module parameter: ib_mtu
692 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
693 HCAs. You can avoid this problem by setting the MTU to 1024
694 using this module parameter.
696 Severity : enhancement
697 Bugzilla : 11118/11620
698 Description: ptllnd small request message buffer alignment fix
699 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
700 Round up small message size on sends in case this option
701 is not supported. 11620 was a defect in the initial
702 implementation which effectively asserted all peers had to be
703 running the correct protocol version which was fixed by always
704 NAK-ing such requests and handling any misalignments they
709 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
710 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
711 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
713 Severity : enhancement
715 Description: Patchless ZC(zero copy) socklnd
716 Details : New protocol for socklnd, socklnd can support zero copy without
717 kernel patch, it's compatible with old socklnd. Checksum is
718 moved from tunables to modparams.
722 Description: When ksocknal_del_peer() is called upon a peer whose
723 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
724 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
727 Frequency : when ptlrpc is under heavy use and runs out of request buffer
729 Description: In lnet_match_blocked_msg(), md can be used without holding a
733 Frequency : very rarely
735 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
736 If connd connects a route which has been closed by
737 ksocknal_shutdown(), ksocknal_create_routes() may create new
738 routes which hold references on the peer, causing shutdown
739 process to wait for peer to disappear forever.
741 Severity : enhancement
743 Description: Dump XT3 portals traces on kptllnd timeout
744 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
745 dump Cray portals debug traces to a file. The kptllnd module
746 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
747 is the basename of the dump file.
750 Frequency : infrequent
752 Description: kernel ptllnd fix bug in connection re-establishment
753 Details : Kernel ptllnd could produce protocol errors e.g. illegal
754 matchbits and/or violate the credit flow protocol when trying
755 to re-establish a connection with a peer after an error or
758 Severity : enhancement
760 Description: Allow /proc/sys/lnet/debug to be set symbolically
761 Details : Allow debug and subsystem debug values to be read/set by name
762 in addition to numerically, for ease of use.
765 Frequency : only in configurations with LNET routers
767 Description: routes automatically marked down and recovered
768 Details : In configurations with LNET routers if a router fails routers
769 now actively try to recover routes that are down, unless they
770 are marked down by an administrator.
772 ------------------------------------------------------------------------------
774 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
777 Frequency : very rarely, in configurations with LNET routers and TCP
779 Description: incorrect data written to files on OSTs
780 Details : In certain high-load conditions incorrect data may be written
781 to files on the OST when using TCP networks.
783 ------------------------------------------------------------------------------
785 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
787 - rework CDEBUG messages rate-limiting mechanism b=10375
788 - add per-socket tunables for socklnd if the kernel is patched b=10327
790 ------------------------------------------------------------------------------
792 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
794 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
795 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
796 - quiet LNET startup LNI message for liblustre b=10128
797 - Better console error messages if 'ip2nets' can't match an IP address
798 - Fixed overflow/use-before-set bugs in linux-time.h
799 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
800 - LNET teardown failed an assertion about the route table being empty
801 - Fixed a crash in LNetEQPoll(<invalid handle>)
802 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
803 - improve debug message for liblustre/Catamount nodes (b=10116)
805 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
806 * Configuration change for the XT3
807 The PTLLND is now used to run Lustre over Portals on the XT3.
808 The configure option(s) --with-cray-portals are no longer
809 used. Rather --with-portals=<path-to-portals-includes> is
810 used to enable building on the XT3. In addition to enable
811 XT3 specific features the option --enable-cray-xt3 must be
814 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
815 * Portals has been removed, replaced by LNET.
816 LNET is new networking infrastructure for Lustre, it includes a
817 reorganized network configuration mode (see the user
818 documentation for full details) as well as support for routing
819 between different network fabrics. Lustre Networking Devices
820 (LNDS) for the supported network fabrics have also been created
821 for this new infrastructure.
823 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
828 Frequency : rare (large Voltaire clusters only)
830 Description: the default number of reserved transmit descriptors was too low
831 for some large clusters
832 Details : As a workaround, the number was increased. A proper fix includes
835 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
840 Frequency : occasional (large-scale events, cluster reboot, network failure)
842 Description: too many error messages on console obscure actual problem and
843 can slow down/panic server, or cause recovery to fail repeatedly
844 Details : enable rate-limiting of console error messages, and some messages
845 that were console errors now only go to the kernel log
847 Severity : enhancement
849 Description: add /proc/sys/portals/catastrophe entry which will report if
850 that node has previously LBUGged
852 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
854 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
856 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
858 - handle error return code in kranal_check_fma_rx() (5915,6054)
860 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
862 - update vibnal (Voltaire IB NAL)
863 - update gmnal (Myrinet NAL), gmnalid
865 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
867 * Landed portals:b_port_step as follows...
869 - removed CFS_DECL_SPIN*
870 just use 'spinlock_t' and initialise with spin_lock_init()
872 - removed CFS_DECL_MUTEX*
873 just use 'struct semaphore' and initialise with init_mutex()
875 - removed CFS_DECL_RWSEM*
876 just use 'struct rw_semaphore' and initialise with init_rwsem()
878 - renamed cfs_sleep_chan -> cfs_waitq
879 cfs_sleep_link -> cfs_waitlink
881 - fixed race in linux version of arch-independent socknal
882 (the ENOMEM/EAGAIN decision).
884 - Didn't fix problems in Darwin version of arch-independent socknal
885 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
887 - removed libcfs types from non-socknal header files (only some types
888 in the header files had been changed; the .c files hadn't been