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 1.3, 1.4.1, 1.4.2 and 1.5.1
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 1.2.10 or later,
13 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
15 Severity : enhancement
17 Description: Quiet some LNET messages
19 Severity : enhancement
21 Description: Add OFED 1.5.1 support
23 Severity : enhancement
25 Description: The peer health code lacked some important debugging info in
26 lnd_query code paths. We've added necessary debug prints,
27 not just for bug 21678, but also for future troubleshooting.
29 -------------------------------------------------------------------------------
31 2010-04-30 Oracle, Inc.
33 * Support for networks:
34 socklnd - any kernel supported by Lustre,
35 qswlnd - Qsnet kernel modules 5.20 and later,
36 openiblnd - IbGold 1.8.2,
37 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, 1.3, 1.4.1, and 1.4.2
38 viblnd - Voltaire ibhost 3.4.5 and later,
39 ciblnd - Topspin 3.2.0,
40 iiblnd - Infiniserv 3.3 + PathBits patch,
41 gmlnd - GM 2.1.22 and later,
42 mxlnd - MX 1.2.10 or later,
43 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
47 Description: lst: check # of remained RPCs before aborting
48 Details : lstcon_rpc_trans_postwait() calls lstcon_rpc_trans_abort() only
49 when the transaction is timeout, so if we got "end_session" to
50 interrupt waiting on transaction, then we can hit the assertion
51 failure ASSERTION(crpc->crp_stamp != 0)
55 Description: print more debug info for timedout ZC-req
56 Details : Print more information for timedout ZC-req and partial
57 received connection. Close connection for timedout ZC-req
58 Always send ZC_ACK on non-blocking connection(BULK_IN)
62 Description: Adding WIRE_ATTR attribute to LNET types
63 Details : LST nodes on different platforms might not communicate well
64 due to the lack of WIRE_ATTR attribute in some LNET structures
65 traversing network. The patch fixes the problem by adding
66 WIRE_ATTR where needed.
70 Description: hash MEs on RDMA portal
71 Details : RDMA portal can have very long ME list on client side, which
72 will trigger soft lockup because of long searching on list.
73 Hash MEs on RDMA portal can resolve this problem.
77 Description: fix for double release of ibc_lock in o2iblnd
78 Details : Re-acquire ibc_lock in kiblnd_post_tx_locked(). Add extra
79 reference to conn before calling kiblnd_post_tx_locked()
80 to avoid scenario when conn disappears inside
81 kiblnd_post_tx_locked().
83 -------------------------------------------------------------------------------
84 2010-01-29 Sun Microsystems, Inc.
86 * Support for networks:
87 socklnd - any kernel supported by Lustre,
88 qswlnd - Qsnet kernel modules 5.20 and later,
89 openiblnd - IbGold 1.8.2,
90 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, 1.3, 1.4.1, and 1.4.2
91 viblnd - Voltaire ibhost 3.4.5 and later,
92 ciblnd - Topspin 3.2.0,
93 iiblnd - Infiniserv 3.3 + PathBits patch,
94 gmlnd - GM 2.1.22 and later,
95 mxlnd - MX 1.2.10 or later,
96 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
100 Description: should update lp_alive for non-router peers.
102 Severity : enhancement
104 Description: LNet router shuffler.
106 Severity : enhancement
108 Description: LNet fine grain routing support.
112 Description: router checker stops working when system wall clock goes backward
113 Details : use monotonic timing source instead of system wall clock time.
115 Severity : enhancement
117 Description: avoid asymmetrical router failures
119 Severity : enhancement
121 Description: multiple-instance support for kptllnd
125 Description: ksocknal_close_conn_locked connection race
126 Details : A race was possible when ksocknal_create_conn calls
127 ksocknal_close_conn_locked for already closed conn.
129 Severity : enhancement
131 Description: port router pinger to userspace
135 Description: kptllnd HELLO protocol deadlock
136 Details : kptllnd HELLO protocol doesn't run to completion in finite time
140 Description: LNet selftest fixes and enhancements
142 Severity : enhancement
144 Description: allow a test node to be a member of multiple test groups
146 Severity : enhancement
148 Description: MXLND: eliminate hosts file, use arp for peer nic_id resolution
149 Details : an update from the upstream developer Scott Atchley.
152 -------------------------------------------------------------------------------
153 2009-07-31 Sun Microsystems, Inc.
155 * Support for networks:
156 socklnd - any kernel supported by Lustre,
157 qswlnd - Qsnet kernel modules 5.20 and later,
158 openiblnd - IbGold 1.8.2,
159 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, 1.3, and 1.4.1
160 viblnd - Voltaire ibhost 3.4.5 and later,
161 ciblnd - Topspin 3.2.0,
162 iiblnd - Infiniserv 3.3 + PathBits patch,
163 gmlnd - GM 2.1.22 and later,
164 mxlnd - MX 1.2.1 or later,
165 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
169 Description: router_proc.c is rewritten to use sysctl-interface for parameters
170 residing in /proc/sys/lnet
172 Severity : enhancement
174 Description: add a new LND optiion to control peer buffer credits on routers
178 Description: Fixing deadlock in usocklnd
179 Details : A deadlock was possible in usocklnd due to race condition while
180 tearing connection down. The problem resulted from erroneous
181 assumption that lnet_finalize() could have been called holding
182 some lnd-level locks.
185 Bugzilla : 13621, 15983
186 Description: Protocol V2 of o2iblnd
187 Details : o2iblnd V2 has several new features:
188 . map-on-demand: map-on-demand is disabled by default, it can
189 be enabled by using modparam "map_on_demand=@value@", @value@
190 should >= 0 and < 256, 0 will disable map-on-demand, any other
191 valid value will enable map-on-demand.
192 Oi2blnd will create FMR or physical MR for RDMA if fragments of
194 Enable map-on-demand will take less memory for new connection,
195 but a little more CPU for RDMA.
196 . iWARP : to support iWARP, please enable map-on-demand, 32 and 64
197 are recommanded value. iWARP will probably fail for value >=128.
198 . OOB NOOP message: to resolve deadlock on router.
199 . tunable peer_credits_hiw: (high water to return credits),
200 default value of peer_credits_hiw equals to (peer_credits -1),
201 user can change it between peer_credits/2 and (peer_credits - 1).
202 Lower value is recommended for high latency network.
203 . tunable message queue size: it always equals to peer_credits,
204 higher value is recommended for high latency network.
205 . It's compatible with earlier version of o2iblnd
209 Description: Fixing 'running out of ports' issue
210 Details : Add a delay before next reconnect attempt in ksocklnd in
211 the case of lost race. Limit the frequency of query-requests
212 in lnet. Improved handling of 'dead peer' notifications in
217 Description: Change ptllnd timeout and watchdog timers
218 Details : Add ptltrace_on_nal_failed and bump ptllnd timeout to match
219 Portals wire timeout.
223 Description: One down Lustre FS hangs ALL mounted Lustre filesystems
224 Details : Shared routing enhancements - peer health detection.
228 Description: IB path MTU mistakenly set to 1st path MTU when ib_mtu is off
229 Details : See comment 46 in bug 11245 for details - it's indeed a bug
230 introduced by the original 11245 fix.
234 Description: uptllnd credit overflow fix
235 Details : kptl_msg_t::ptlm_credits could be overflown by uptllnd since
240 Description: socklnd protocol version 3
241 Details : With current protocol V2, connections on router can be
242 blocked and can't receive any incoming messages when there is no
243 more router buffer, so ZC-ACK can't be handled (LNet message
244 can't be finalized) and will cause deadlock on router.
245 Protocol V3 has a dedicated connection for emergency messages
246 like ZC-ACK to router, messages on this dedicated connection
247 don't need any credit so will never be blocked. Also, V3 can send
248 keepalive ping in specified period for router healthy checking.
250 -------------------------------------------------------------------------------
251 12-31-2008 Sun Microsystems, Inc.
253 * Support for networks:
254 socklnd - any kernel supported by Lustre,
255 qswlnd - Qsnet kernel modules 5.20 and later,
256 openiblnd - IbGold 1.8.2,
257 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
258 viblnd - Voltaire ibhost 3.4.5 and later,
259 ciblnd - Topspin 3.2.0,
260 iiblnd - Infiniserv 3.3 + PathBits patch,
261 gmlnd - GM 2.1.22 and later,
262 mxlnd - MX 1.2.1 or later,
263 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
272 Description: workaround for OOM from o2iblnd
273 Details : OFED needs allocate big chunk of memory for QP while creating
274 connection for o2iblnd, OOM can happen if no such a contiguous
276 QP size is decided by concurrent_sends and max_fragments of
277 o2iblnd, now we permit user to specify smaller value for
278 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
279 will decrease memory block size required by creating QP.
283 Description: Support Zerocopy receive of Chelsio device
284 Details : Chelsio driver can support zerocopy for iov[1] if it's
285 contiguous and large enough.
289 Description: fix credit flow deadlock in uptllnd
293 Description: finalize network operation in reasonable time
294 Details : conf-sanity test_32a couldn't stop ost and mds because it
295 tried to access non-existent peer and tcp connect took
296 quite long before timing out.
300 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
301 Details : Lost reference on conn prevents peer from being destroyed, which
302 could prevent new peer creation if peer count has reached upper
307 Description: LNET Selftest results in Soft lockup on OSS CPU
308 Details : only hits when 8 or more o2ib clients involved and a session is
309 torn down with 'lst end_session' without preceeding 'lst stop'.
313 Description: concurrent_sends in IB LNDs should not be changeable at run time
314 Details : concurrent_sends in IB LNDs should not be changeable at run time
318 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
319 Details : only hits under out-of-memory situations
322 -------------------------------------------------------------------------------
324 2009-02-07 Sun Microsystems, Inc.
326 * Support for networks:
327 socklnd - any kernel supported by Lustre,
328 qswlnd - Qsnet kernel modules 5.20 and later,
329 openiblnd - IbGold 1.8.2,
330 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
331 viblnd - Voltaire ibhost 3.4.5 and later,
332 ciblnd - Topspin 3.2.0,
333 iiblnd - Infiniserv 3.3 + PathBits patch,
334 gmlnd - GM 2.1.22 and later,
335 mxlnd - MX 1.2.1 or later,
336 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
340 Description: workaround for OOM from o2iblnd
341 Details : OFED needs allocate big chunk of memory for QP while creating
342 connection for o2iblnd, OOM can happen if no such a contiguous
344 QP size is decided by concurrent_sends and max_fragments of
345 o2iblnd, now we permit user to specify smaller value for
346 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
347 will decrease memory block size required by creating QP.
351 Description: Support Zerocopy receive of Chelsio device
352 Details : Chelsio driver can support zerocopy for iov[1] if it's
353 contiguous and large enough.
357 Description: fix credit flow deadlock in uptllnd
361 Description: finalize network operation in reasonable time
362 Details : conf-sanity test_32a couldn't stop ost and mds because it
363 tried to access non-existent peer and tcp connect took
364 quite long before timing out.
368 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
369 Details : Lost reference on conn prevents peer from being destroyed, which
370 could prevent new peer creation if peer count has reached upper
375 Description: LNET Selftest results in Soft lockup on OSS CPU
376 Details : only hits when 8 or more o2ib clients involved and a session is
377 torn down with 'lst end_session' without preceeding 'lst stop'.
381 Description: concurrent_sends in IB LNDs should not be changeable at run time
382 Details : concurrent_sends in IB LNDs should not be changeable at run time
384 -------------------------------------------------------------------------------
386 11-03-2008 Sun Microsystems, Inc.
388 * Support for networks:
389 socklnd - any kernel supported by Lustre,
390 qswlnd - Qsnet kernel modules 5.20 and later,
391 openiblnd - IbGold 1.8.2,
392 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
393 viblnd - Voltaire ibhost 3.4.5 and later,
394 ciblnd - Topspin 3.2.0,
395 iiblnd - Infiniserv 3.3 + PathBits patch,
396 gmlnd - GM 2.1.22 and later,
397 mxlnd - MX 1.2.1 or later,
398 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
402 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
403 Details : only hits under out-of-memory situations
405 -------------------------------------------------------------------------------
408 04-26-2008 Sun Microsystems, Inc.
410 * Support for networks:
411 socklnd - any kernel supported by Lustre,
412 qswlnd - Qsnet kernel modules 5.20 and later,
413 openiblnd - IbGold 1.8.2,
414 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
415 viblnd - Voltaire ibhost 3.4.5 and later,
416 ciblnd - Topspin 3.2.0,
417 iiblnd - Infiniserv 3.3 + PathBits patch,
418 gmlnd - GM 2.1.22 and later,
419 mxlnd - MX 1.2.1 or later,
420 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
424 Description: excessive debug information removed
425 Details : excessive debug information removed
429 Description: ksocknal_create_conn() hit ASSERTION during connection race
430 Details : ksocknal_create_conn() hit ASSERTION during connection race
434 Description: ksocknal_send_hello() hit ASSERTION while connecting race
435 Details : ksocknal_send_hello() hit ASSERTION while connecting race
439 Description: o2iblnd/ptllnd credit deadlock in a routed config.
440 Details : o2iblnd/ptllnd credit deadlock in a routed config.
444 Description: High load after starting lnet
445 Details : gmlnd should sleep in rx thread in interruptible way. Otherwise,
446 uptime utility reports high load that looks confusingly.
450 Description: ksocklnd fails to establish connection if accept_port is high
451 Details : PID remapping must not be done for active (outgoing) connections
453 --------------------------------------------------------------------------------
455 2008-01-11 Sun Microsystems, Inc.
457 * Support for networks:
458 socklnd - any kernel supported by Lustre,
459 qswlnd - Qsnet kernel modules 5.20 and later,
460 openiblnd - IbGold 1.8.2,
461 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
462 viblnd - Voltaire ibhost 3.4.5 and later,
463 ciblnd - Topspin 3.2.0,
464 iiblnd - Infiniserv 3.3 + PathBits patch,
465 gmlnd - GM 2.1.22 and later,
466 mxlnd - MX 1.2.1 or later,
467 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
470 Description: liblustre network error
471 Details : liblustre clients should understand LNET_ACCEPT_PORT environment
472 variable even if they don't start lnet acceptor.
476 Description: Strange message from lnet (Ignoring prediction from the future)
477 Details : Incorrect calculation of peer's last_alive value in ksocklnd
479 --------------------------------------------------------------------------------
481 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
483 * Support for networks:
484 socklnd - any kernel supported by Lustre,
485 qswlnd - Qsnet kernel modules 5.20 and later,
486 openiblnd - IbGold 1.8.2,
487 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5.
488 viblnd - Voltaire ibhost 3.4.5 and later,
489 ciblnd - Topspin 3.2.0,
490 iiblnd - Infiniserv 3.3 + PathBits patch,
491 gmlnd - GM 2.1.22 and later,
492 mxlnd - MX 1.2.1 or later,
493 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
497 Description: ASSERTION(me == md->md_me) failed in lnet_match_md()
501 Description: increase send queue size for ciblnd/openiblnd
505 Description: new userspace socklnd
506 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
507 with new one - usocklnd.
509 Severity : enhancement
511 Description: Console message flood
512 Details : Make cdls ratelimiting more tunable by adding several tunable in
513 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
514 /proc/sys/lnet/console_backoff.
516 --------------------------------------------------------------------------------
518 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
520 * Support for networks:
521 socklnd - any kernel supported by Lustre,
522 qswlnd - Qsnet kernel modules 5.20 and later,
523 openiblnd - IbGold 1.8.2,
524 o2iblnd - OFED 1.1 and 1.2,
525 viblnd - Voltaire ibhost 3.4.5 and later,
526 ciblnd - Topspin 3.2.0,
527 iiblnd - Infiniserv 3.3 + PathBits patch,
528 gmlnd - GM 2.1.22 and later,
529 mxlnd - MX 1.2.1 or later,
530 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
534 Description: /proc/sys/lnet has non-sysctl entries
535 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
539 Description: TOE Kernel panic by ksocklnd
540 Details : offloaded sockets provide their own implementation of sendpage,
541 can't call tcp_sendpage() directly
545 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
546 Details : races between lnd_shutdown and peer creation prevent
547 lnd_shutdown from finishing.
551 Description: open files rlimit 1024 reached while liblustre testing
552 Details : ulnds/socklnd must close open socket after unsuccessful
557 Description: build error
558 Details : fix typos in gmlnd, ptllnd and viblnd
560 ------------------------------------------------------------------------------
562 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
564 * Support for networks:
565 socklnd - kernels up to 2.6.16,
566 qswlnd - Qsnet kernel modules 5.20 and later,
567 openiblnd - IbGold 1.8.2,
568 o2iblnd - OFED 1.1 and 1.2
569 viblnd - Voltaire ibhost 3.4.5 and later,
570 ciblnd - Topspin 3.2.0,
571 iiblnd - Infiniserv 3.3 + PathBits patch,
572 gmlnd - GM 2.1.22 and later,
573 mxlnd - MX 1.2.1 or later,
574 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
576 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
578 * Support for networks:
579 socklnd - kernels up to 2.6.16,
580 qswlnd - Qsnet kernel modules 5.20 and later,
581 openiblnd - IbGold 1.8.2,
583 viblnd - Voltaire ibhost 3.4.5 and later,
584 ciblnd - Topspin 3.2.0,
585 iiblnd - Infiniserv 3.3 + PathBits patch,
586 gmlnd - GM 2.1.22 and later,
587 mxlnd - MX 1.2.1 or later,
588 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
592 Description: Initialize cpumask before use
596 Description: ASSERTION failures when upgrading to the patchless zero-copy
598 Details : This bug affects "rolling upgrades", causing an inconsistent
599 protocol version negotiation and subsequent assertion failure
600 during rolling upgrades after the first wave of upgrades.
604 Details : Change "dropped message" CERRORs to D_NETERROR so they are
605 logged instead of creating "console chatter" when a lustre
606 timeout races with normal RPC completion.
609 Details : lnet_clear_peer_table can wait forever if user forgets to
613 Details : libcfs_id2str should check pid against LNET_PID_ANY.
617 Description: added LNET self test
618 Details : landing b_self_test
623 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
625 Details : do_div() macro is used incorrectly.
627 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
631 Description: make panic on lbug configurable
635 Description: Add OFED1.2 support to o2iblnd
636 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
637 are installed (other than kernel's in-tree infiniband), there
638 could be some problem while insmod o2iblnd (mismatch CRC of
640 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
641 this link provides solution:
642 https://bugs.openfabrics.org/show_bug.cgi?id=355
643 if extra Module.symvers is not supported in kernel, we will
644 have to run the script in bug 12316 to update
645 $LINUX/module.symvers before building o2iblnd.
646 More details about this are in bug 12316.
648 ------------------------------------------------------------------------------
650 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
651 * version 1.4.10 / 1.6.0
652 * Support for networks:
653 socklnd - kernels up to 2.6.16,
654 qswlnd - Qsnet kernel modules 5.20 and later,
655 openiblnd - IbGold 1.8.2,
657 viblnd - Voltaire ibhost 3.4.5 and later,
658 ciblnd - Topspin 3.2.0,
659 iiblnd - Infiniserv 3.3 + PathBits patch,
660 gmlnd - GM 2.1.22 and later,
661 mxlnd - MX 1.2.1 or later,
662 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
666 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
667 possibly accessed in kptllnd_shutdown. Ptllnd should init
668 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
672 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
676 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
680 Description: the_lnet.ln_finalizing was not set when the current thread is
681 about to complete messages. It only affects multi-threaded
687 Description: Changed the default kqswlnd ntxmsg=512
692 Description: Assertion failure in kernel ptllnd caused by posting passive
693 bulk buffers before connection establishment complete.
698 Description: A race in kernel ptllnd between deleting a peer and posting
699 new communications for it could hang communications -
700 manifesting as "Unexpectedly long timeout" messages.
705 Description: Kernel ptllnd lock ordering issue could hang a node.
710 Description: node crash on socket teardown race
713 Frequency : 'lctl peer_list' issued on a mx net
715 Description: Enable lctl's peer_list for MXLND
718 Frequency : after Ptllnd timeouts and portals congestion
720 Description: Credit overflows
721 Details : This was a bug in ptllnd connection establishment. The fix
722 implements better peer stamps to disambiguate connection
723 establishment and ensure both peers enter the credit flow
724 state machine consistently.
729 Description: kptllnd didn't propagate some network errors up to LNET
730 Details : This bug was spotted while investigating 11394. The fix
731 ensures network errors on sends and bulk transfers are
732 propagated to LNET/lustre correctly.
734 Severity : enhancement
736 Description: Fixed console chatter in case of -ETIMEDOUT.
738 Severity : enhancement
740 Description: Added D_NETTRACE for recording network packet history
741 (initially only for ptllnd). Also a separate userspace
742 ptllnd facility to gather history which should really be
743 covered by D_NETTRACE too, if only CDEBUG recorded history in
749 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
750 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
751 callback can occur before a connection has actually been
752 established. This caused an assertion failure previously.
754 Severity : enhancement
756 Description: Multiple instances for o2iblnd
757 Details : Allow multiple instances of o2iblnd to enable networking over
758 multiple HCAs and routing between them.
762 Description: lnet deadlock in router_checker
763 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
764 into BH locks to eliminate potential deadlock caused by
765 ksocknal_data_ready() preempting code holding these locks.
769 Description: Millions of failed socklnd connection attempts cause a very slow FS
770 Details : added a new route flag ksnr_scheduled to distinguish from
771 ksnr_connecting, so that a peer connection request is only turned
772 down for race concerns when an active connection to the same peer
773 is under progress (instead of just being scheduled).
775 ------------------------------------------------------------------------------
777 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
779 * Support for networks:
780 socklnd - kernels up to 2.6.16
781 qswlnd - Qsnet kernel modules 5.20 and later
782 openiblnd - IbGold 1.8.2
784 viblnd - Voltaire ibhost 3.4.5 and later
785 ciblnd - Topspin 3.2.0
786 iiblnd - Infiniserv 3.3 + PathBits patch
787 gmlnd - GM 2.1.22 and later
788 mxlnd - MX 1.2.1 or later
789 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
792 Severity : major on XT3
794 Description: libcfs overwrites /proc/sys/portals
795 Details : libcfs created a symlink from /proc/sys/portals to
796 /proc/sys/lnet for backwards compatibility. This is no
797 longer required and makes the Cray portals /proc variables
802 Description: OFED FMR API change
803 Details : This changes parameter usage to reflect a change in
804 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
805 that FMR support is only used in experimental versions of the
806 o2iblnd - this change does not affect standard usage at all.
808 Severity : enhancement
810 Description: new ko2iblnd module parameter: ib_mtu
811 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
812 HCAs. You can avoid this problem by setting the MTU to 1024
813 using this module parameter.
815 Severity : enhancement
816 Bugzilla : 11118/11620
817 Description: ptllnd small request message buffer alignment fix
818 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
819 Round up small message size on sends in case this option
820 is not supported. 11620 was a defect in the initial
821 implementation which effectively asserted all peers had to be
822 running the correct protocol version which was fixed by always
823 NAK-ing such requests and handling any misalignments they
828 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
829 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
830 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
832 Severity : enhancement
834 Description: Patchless ZC(zero copy) socklnd
835 Details : New protocol for socklnd, socklnd can support zero copy without
836 kernel patch, it's compatible with old socklnd. Checksum is
837 moved from tunables to modparams.
841 Description: When ksocknal_del_peer() is called upon a peer whose
842 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
843 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
846 Frequency : when ptlrpc is under heavy use and runs out of request buffer
848 Description: In lnet_match_blocked_msg(), md can be used without holding a
852 Frequency : very rarely
854 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
855 If connd connects a route which has been closed by
856 ksocknal_shutdown(), ksocknal_create_routes() may create new
857 routes which hold references on the peer, causing shutdown
858 process to wait for peer to disappear forever.
860 Severity : enhancement
862 Description: Dump XT3 portals traces on kptllnd timeout
863 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
864 dump Cray portals debug traces to a file. The kptllnd module
865 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
866 is the basename of the dump file.
869 Frequency : infrequent
871 Description: kernel ptllnd fix bug in connection re-establishment
872 Details : Kernel ptllnd could produce protocol errors e.g. illegal
873 matchbits and/or violate the credit flow protocol when trying
874 to re-establish a connection with a peer after an error or
877 Severity : enhancement
879 Description: Allow /proc/sys/lnet/debug to be set symbolically
880 Details : Allow debug and subsystem debug values to be read/set by name
881 in addition to numerically, for ease of use.
884 Frequency : only in configurations with LNET routers
886 Description: routes automatically marked down and recovered
887 Details : In configurations with LNET routers if a router fails routers
888 now actively try to recover routes that are down, unless they
889 are marked down by an administrator.
891 ------------------------------------------------------------------------------
893 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
896 Frequency : very rarely, in configurations with LNET routers and TCP
898 Description: incorrect data written to files on OSTs
899 Details : In certain high-load conditions incorrect data may be written
900 to files on the OST when using TCP networks.
902 ------------------------------------------------------------------------------
904 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
906 - rework CDEBUG messages rate-limiting mechanism b=10375
907 - add per-socket tunables for socklnd if the kernel is patched b=10327
909 ------------------------------------------------------------------------------
911 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
913 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
914 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
915 - quiet LNET startup LNI message for liblustre b=10128
916 - Better console error messages if 'ip2nets' can't match an IP address
917 - Fixed overflow/use-before-set bugs in linux-time.h
918 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
919 - LNET teardown failed an assertion about the route table being empty
920 - Fixed a crash in LNetEQPoll(<invalid handle>)
921 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
922 - improve debug message for liblustre/Catamount nodes (b=10116)
924 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
925 * Configuration change for the XT3
926 The PTLLND is now used to run Lustre over Portals on the XT3.
927 The configure option(s) --with-cray-portals are no longer
928 used. Rather --with-portals=<path-to-portals-includes> is
929 used to enable building on the XT3. In addition to enable
930 XT3 specific features the option --enable-cray-xt3 must be
933 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
934 * Portals has been removed, replaced by LNET.
935 LNET is new networking infrastructure for Lustre, it includes a
936 reorganized network configuration mode (see the user
937 documentation for full details) as well as support for routing
938 between different network fabrics. Lustre Networking Devices
939 (LNDS) for the supported network fabrics have also been created
940 for this new infrastructure.
942 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
947 Frequency : rare (large Voltaire clusters only)
949 Description: the default number of reserved transmit descriptors was too low
950 for some large clusters
951 Details : As a workaround, the number was increased. A proper fix includes
954 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
959 Frequency : occasional (large-scale events, cluster reboot, network failure)
961 Description: too many error messages on console obscure actual problem and
962 can slow down/panic server, or cause recovery to fail repeatedly
963 Details : enable rate-limiting of console error messages, and some messages
964 that were console errors now only go to the kernel log
966 Severity : enhancement
968 Description: add /proc/sys/portals/catastrophe entry which will report if
969 that node has previously LBUGged
971 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
973 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
975 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
977 - handle error return code in kranal_check_fma_rx() (5915,6054)
979 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
981 - update vibnal (Voltaire IB NAL)
982 - update gmnal (Myrinet NAL), gmnalid
984 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
986 * Landed portals:b_port_step as follows...
988 - removed CFS_DECL_SPIN*
989 just use 'spinlock_t' and initialise with spin_lock_init()
991 - removed CFS_DECL_MUTEX*
992 just use 'struct semaphore' and initialise with init_mutex()
994 - removed CFS_DECL_RWSEM*
995 just use 'struct rw_semaphore' and initialise with init_rwsem()
997 - renamed cfs_sleep_chan -> cfs_waitq
998 cfs_sleep_link -> cfs_waitlink
1000 - fixed race in linux version of arch-independent socknal
1001 (the ENOMEM/EAGAIN decision).
1003 - Didn't fix problems in Darwin version of arch-independent socknal
1004 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
1006 - removed libcfs types from non-socknal header files (only some types
1007 in the header files had been changed; the .c files hadn't been