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, 1.4.1, and 1.4.2
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
16 -------------------------------------------------------------------------------
17 2010-01-29 Sun Microsystems, Inc.
19 * Support for networks:
20 socklnd - any kernel supported by Lustre,
21 qswlnd - Qsnet kernel modules 5.20 and later,
22 openiblnd - IbGold 1.8.2,
23 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, 1.3, 1.4.1, and 1.4.2
24 viblnd - Voltaire ibhost 3.4.5 and later,
25 ciblnd - Topspin 3.2.0,
26 iiblnd - Infiniserv 3.3 + PathBits patch,
27 gmlnd - GM 2.1.22 and later,
28 mxlnd - MX 1.2.10 or later,
29 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
33 Description: should update lp_alive for non-router peers.
35 Severity : enhancement
37 Description: LNet router shuffler.
39 Severity : enhancement
41 Description: LNet fine grain routing support.
45 Description: router checker stops working when system wall clock goes backward
46 Details : use monotonic timing source instead of system wall clock time.
48 Severity : enhancement
50 Description: avoid asymmetrical router failures
52 Severity : enhancement
54 Description: multiple-instance support for kptllnd
58 Description: ksocknal_close_conn_locked connection race
59 Details : A race was possible when ksocknal_create_conn calls
60 ksocknal_close_conn_locked for already closed conn.
62 Severity : enhancement
64 Description: port router pinger to userspace
68 Description: kptllnd HELLO protocol deadlock
69 Details : kptllnd HELLO protocol doesn't run to completion in finite time
73 Description: LNet selftest fixes and enhancements
75 Severity : enhancement
77 Description: allow a test node to be a member of multiple test groups
79 Severity : enhancement
81 Description: MXLND: eliminate hosts file, use arp for peer nic_id resolution
82 Details : an update from the upstream developer Scott Atchley.
85 -------------------------------------------------------------------------------
86 2009-07-31 Sun Microsystems, Inc.
88 * Support for networks:
89 socklnd - any kernel supported by Lustre,
90 qswlnd - Qsnet kernel modules 5.20 and later,
91 openiblnd - IbGold 1.8.2,
92 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, 1.3, and 1.4.1
93 viblnd - Voltaire ibhost 3.4.5 and later,
94 ciblnd - Topspin 3.2.0,
95 iiblnd - Infiniserv 3.3 + PathBits patch,
96 gmlnd - GM 2.1.22 and later,
97 mxlnd - MX 1.2.1 or later,
98 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
102 Description: router_proc.c is rewritten to use sysctl-interface for parameters
103 residing in /proc/sys/lnet
105 Severity : enhancement
107 Description: add a new LND optiion to control peer buffer credits on routers
111 Description: Fixing deadlock in usocklnd
112 Details : A deadlock was possible in usocklnd due to race condition while
113 tearing connection down. The problem resulted from erroneous
114 assumption that lnet_finalize() could have been called holding
115 some lnd-level locks.
118 Bugzilla : 13621, 15983
119 Description: Protocol V2 of o2iblnd
120 Details : o2iblnd V2 has several new features:
121 . map-on-demand: map-on-demand is disabled by default, it can
122 be enabled by using modparam "map_on_demand=@value@", @value@
123 should >= 0 and < 256, 0 will disable map-on-demand, any other
124 valid value will enable map-on-demand.
125 Oi2blnd will create FMR or physical MR for RDMA if fragments of
127 Enable map-on-demand will take less memory for new connection,
128 but a little more CPU for RDMA.
129 . iWARP : to support iWARP, please enable map-on-demand, 32 and 64
130 are recommanded value. iWARP will probably fail for value >=128.
131 . OOB NOOP message: to resolve deadlock on router.
132 . tunable peer_credits_hiw: (high water to return credits),
133 default value of peer_credits_hiw equals to (peer_credits -1),
134 user can change it between peer_credits/2 and (peer_credits - 1).
135 Lower value is recommended for high latency network.
136 . tunable message queue size: it always equals to peer_credits,
137 higher value is recommended for high latency network.
138 . It's compatible with earlier version of o2iblnd
142 Description: Fixing 'running out of ports' issue
143 Details : Add a delay before next reconnect attempt in ksocklnd in
144 the case of lost race. Limit the frequency of query-requests
145 in lnet. Improved handling of 'dead peer' notifications in
150 Description: Change ptllnd timeout and watchdog timers
151 Details : Add ptltrace_on_nal_failed and bump ptllnd timeout to match
152 Portals wire timeout.
156 Description: One down Lustre FS hangs ALL mounted Lustre filesystems
157 Details : Shared routing enhancements - peer health detection.
161 Description: IB path MTU mistakenly set to 1st path MTU when ib_mtu is off
162 Details : See comment 46 in bug 11245 for details - it's indeed a bug
163 introduced by the original 11245 fix.
167 Description: uptllnd credit overflow fix
168 Details : kptl_msg_t::ptlm_credits could be overflown by uptllnd since
173 Description: socklnd protocol version 3
174 Details : With current protocol V2, connections on router can be
175 blocked and can't receive any incoming messages when there is no
176 more router buffer, so ZC-ACK can't be handled (LNet message
177 can't be finalized) and will cause deadlock on router.
178 Protocol V3 has a dedicated connection for emergency messages
179 like ZC-ACK to router, messages on this dedicated connection
180 don't need any credit so will never be blocked. Also, V3 can send
181 keepalive ping in specified period for router healthy checking.
183 -------------------------------------------------------------------------------
184 12-31-2008 Sun Microsystems, Inc.
186 * Support for networks:
187 socklnd - any kernel supported by Lustre,
188 qswlnd - Qsnet kernel modules 5.20 and later,
189 openiblnd - IbGold 1.8.2,
190 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
191 viblnd - Voltaire ibhost 3.4.5 and later,
192 ciblnd - Topspin 3.2.0,
193 iiblnd - Infiniserv 3.3 + PathBits patch,
194 gmlnd - GM 2.1.22 and later,
195 mxlnd - MX 1.2.1 or later,
196 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
205 Description: workaround for OOM from o2iblnd
206 Details : OFED needs allocate big chunk of memory for QP while creating
207 connection for o2iblnd, OOM can happen if no such a contiguous
209 QP size is decided by concurrent_sends and max_fragments of
210 o2iblnd, now we permit user to specify smaller value for
211 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
212 will decrease memory block size required by creating QP.
216 Description: Support Zerocopy receive of Chelsio device
217 Details : Chelsio driver can support zerocopy for iov[1] if it's
218 contiguous and large enough.
222 Description: fix credit flow deadlock in uptllnd
226 Description: finalize network operation in reasonable time
227 Details : conf-sanity test_32a couldn't stop ost and mds because it
228 tried to access non-existent peer and tcp connect took
229 quite long before timing out.
233 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
234 Details : Lost reference on conn prevents peer from being destroyed, which
235 could prevent new peer creation if peer count has reached upper
240 Description: LNET Selftest results in Soft lockup on OSS CPU
241 Details : only hits when 8 or more o2ib clients involved and a session is
242 torn down with 'lst end_session' without preceeding 'lst stop'.
246 Description: concurrent_sends in IB LNDs should not be changeable at run time
247 Details : concurrent_sends in IB LNDs should not be changeable at run time
251 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
252 Details : only hits under out-of-memory situations
255 -------------------------------------------------------------------------------
257 2009-02-07 Sun Microsystems, Inc.
259 * Support for networks:
260 socklnd - any kernel supported by Lustre,
261 qswlnd - Qsnet kernel modules 5.20 and later,
262 openiblnd - IbGold 1.8.2,
263 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
264 viblnd - Voltaire ibhost 3.4.5 and later,
265 ciblnd - Topspin 3.2.0,
266 iiblnd - Infiniserv 3.3 + PathBits patch,
267 gmlnd - GM 2.1.22 and later,
268 mxlnd - MX 1.2.1 or later,
269 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
273 Description: workaround for OOM from o2iblnd
274 Details : OFED needs allocate big chunk of memory for QP while creating
275 connection for o2iblnd, OOM can happen if no such a contiguous
277 QP size is decided by concurrent_sends and max_fragments of
278 o2iblnd, now we permit user to specify smaller value for
279 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
280 will decrease memory block size required by creating QP.
284 Description: Support Zerocopy receive of Chelsio device
285 Details : Chelsio driver can support zerocopy for iov[1] if it's
286 contiguous and large enough.
290 Description: fix credit flow deadlock in uptllnd
294 Description: finalize network operation in reasonable time
295 Details : conf-sanity test_32a couldn't stop ost and mds because it
296 tried to access non-existent peer and tcp connect took
297 quite long before timing out.
301 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
302 Details : Lost reference on conn prevents peer from being destroyed, which
303 could prevent new peer creation if peer count has reached upper
308 Description: LNET Selftest results in Soft lockup on OSS CPU
309 Details : only hits when 8 or more o2ib clients involved and a session is
310 torn down with 'lst end_session' without preceeding 'lst stop'.
314 Description: concurrent_sends in IB LNDs should not be changeable at run time
315 Details : concurrent_sends in IB LNDs should not be changeable at run time
317 -------------------------------------------------------------------------------
319 11-03-2008 Sun Microsystems, Inc.
321 * Support for networks:
322 socklnd - any kernel supported by Lustre,
323 qswlnd - Qsnet kernel modules 5.20 and later,
324 openiblnd - IbGold 1.8.2,
325 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
326 viblnd - Voltaire ibhost 3.4.5 and later,
327 ciblnd - Topspin 3.2.0,
328 iiblnd - Infiniserv 3.3 + PathBits patch,
329 gmlnd - GM 2.1.22 and later,
330 mxlnd - MX 1.2.1 or later,
331 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
335 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
336 Details : only hits under out-of-memory situations
338 -------------------------------------------------------------------------------
341 04-26-2008 Sun Microsystems, Inc.
343 * Support for networks:
344 socklnd - any kernel supported by Lustre,
345 qswlnd - Qsnet kernel modules 5.20 and later,
346 openiblnd - IbGold 1.8.2,
347 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
348 viblnd - Voltaire ibhost 3.4.5 and later,
349 ciblnd - Topspin 3.2.0,
350 iiblnd - Infiniserv 3.3 + PathBits patch,
351 gmlnd - GM 2.1.22 and later,
352 mxlnd - MX 1.2.1 or later,
353 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
357 Description: excessive debug information removed
358 Details : excessive debug information removed
362 Description: ksocknal_create_conn() hit ASSERTION during connection race
363 Details : ksocknal_create_conn() hit ASSERTION during connection race
367 Description: ksocknal_send_hello() hit ASSERTION while connecting race
368 Details : ksocknal_send_hello() hit ASSERTION while connecting race
372 Description: o2iblnd/ptllnd credit deadlock in a routed config.
373 Details : o2iblnd/ptllnd credit deadlock in a routed config.
377 Description: High load after starting lnet
378 Details : gmlnd should sleep in rx thread in interruptible way. Otherwise,
379 uptime utility reports high load that looks confusingly.
383 Description: ksocklnd fails to establish connection if accept_port is high
384 Details : PID remapping must not be done for active (outgoing) connections
386 --------------------------------------------------------------------------------
388 2008-01-11 Sun Microsystems, Inc.
390 * Support for networks:
391 socklnd - any kernel supported by Lustre,
392 qswlnd - Qsnet kernel modules 5.20 and later,
393 openiblnd - IbGold 1.8.2,
394 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
395 viblnd - Voltaire ibhost 3.4.5 and later,
396 ciblnd - Topspin 3.2.0,
397 iiblnd - Infiniserv 3.3 + PathBits patch,
398 gmlnd - GM 2.1.22 and later,
399 mxlnd - MX 1.2.1 or later,
400 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
403 Description: liblustre network error
404 Details : liblustre clients should understand LNET_ACCEPT_PORT environment
405 variable even if they don't start lnet acceptor.
409 Description: Strange message from lnet (Ignoring prediction from the future)
410 Details : Incorrect calculation of peer's last_alive value in ksocklnd
412 --------------------------------------------------------------------------------
414 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
416 * Support for networks:
417 socklnd - any kernel supported by Lustre,
418 qswlnd - Qsnet kernel modules 5.20 and later,
419 openiblnd - IbGold 1.8.2,
420 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5.
421 viblnd - Voltaire ibhost 3.4.5 and later,
422 ciblnd - Topspin 3.2.0,
423 iiblnd - Infiniserv 3.3 + PathBits patch,
424 gmlnd - GM 2.1.22 and later,
425 mxlnd - MX 1.2.1 or later,
426 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
430 Description: ASSERTION(me == md->md_me) failed in lnet_match_md()
434 Description: increase send queue size for ciblnd/openiblnd
438 Description: new userspace socklnd
439 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
440 with new one - usocklnd.
442 Severity : enhancement
444 Description: Console message flood
445 Details : Make cdls ratelimiting more tunable by adding several tunable in
446 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
447 /proc/sys/lnet/console_backoff.
449 --------------------------------------------------------------------------------
451 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
453 * Support for networks:
454 socklnd - any kernel supported by Lustre,
455 qswlnd - Qsnet kernel modules 5.20 and later,
456 openiblnd - IbGold 1.8.2,
457 o2iblnd - OFED 1.1 and 1.2,
458 viblnd - Voltaire ibhost 3.4.5 and later,
459 ciblnd - Topspin 3.2.0,
460 iiblnd - Infiniserv 3.3 + PathBits patch,
461 gmlnd - GM 2.1.22 and later,
462 mxlnd - MX 1.2.1 or later,
463 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
467 Description: /proc/sys/lnet has non-sysctl entries
468 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
472 Description: TOE Kernel panic by ksocklnd
473 Details : offloaded sockets provide their own implementation of sendpage,
474 can't call tcp_sendpage() directly
478 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
479 Details : races between lnd_shutdown and peer creation prevent
480 lnd_shutdown from finishing.
484 Description: open files rlimit 1024 reached while liblustre testing
485 Details : ulnds/socklnd must close open socket after unsuccessful
490 Description: build error
491 Details : fix typos in gmlnd, ptllnd and viblnd
493 ------------------------------------------------------------------------------
495 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
497 * Support for networks:
498 socklnd - kernels up to 2.6.16,
499 qswlnd - Qsnet kernel modules 5.20 and later,
500 openiblnd - IbGold 1.8.2,
501 o2iblnd - OFED 1.1 and 1.2
502 viblnd - Voltaire ibhost 3.4.5 and later,
503 ciblnd - Topspin 3.2.0,
504 iiblnd - Infiniserv 3.3 + PathBits patch,
505 gmlnd - GM 2.1.22 and later,
506 mxlnd - MX 1.2.1 or later,
507 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
509 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
511 * Support for networks:
512 socklnd - kernels up to 2.6.16,
513 qswlnd - Qsnet kernel modules 5.20 and later,
514 openiblnd - IbGold 1.8.2,
516 viblnd - Voltaire ibhost 3.4.5 and later,
517 ciblnd - Topspin 3.2.0,
518 iiblnd - Infiniserv 3.3 + PathBits patch,
519 gmlnd - GM 2.1.22 and later,
520 mxlnd - MX 1.2.1 or later,
521 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
525 Description: Initialize cpumask before use
529 Description: ASSERTION failures when upgrading to the patchless zero-copy
531 Details : This bug affects "rolling upgrades", causing an inconsistent
532 protocol version negotiation and subsequent assertion failure
533 during rolling upgrades after the first wave of upgrades.
537 Details : Change "dropped message" CERRORs to D_NETERROR so they are
538 logged instead of creating "console chatter" when a lustre
539 timeout races with normal RPC completion.
542 Details : lnet_clear_peer_table can wait forever if user forgets to
546 Details : libcfs_id2str should check pid against LNET_PID_ANY.
550 Description: added LNET self test
551 Details : landing b_self_test
556 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
558 Details : do_div() macro is used incorrectly.
560 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
564 Description: make panic on lbug configurable
568 Description: Add OFED1.2 support to o2iblnd
569 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
570 are installed (other than kernel's in-tree infiniband), there
571 could be some problem while insmod o2iblnd (mismatch CRC of
573 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
574 this link provides solution:
575 https://bugs.openfabrics.org/show_bug.cgi?id=355
576 if extra Module.symvers is not supported in kernel, we will
577 have to run the script in bug 12316 to update
578 $LINUX/module.symvers before building o2iblnd.
579 More details about this are in bug 12316.
581 ------------------------------------------------------------------------------
583 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
584 * version 1.4.10 / 1.6.0
585 * Support for networks:
586 socklnd - kernels up to 2.6.16,
587 qswlnd - Qsnet kernel modules 5.20 and later,
588 openiblnd - IbGold 1.8.2,
590 viblnd - Voltaire ibhost 3.4.5 and later,
591 ciblnd - Topspin 3.2.0,
592 iiblnd - Infiniserv 3.3 + PathBits patch,
593 gmlnd - GM 2.1.22 and later,
594 mxlnd - MX 1.2.1 or later,
595 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
599 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
600 possibly accessed in kptllnd_shutdown. Ptllnd should init
601 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
605 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
609 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
613 Description: the_lnet.ln_finalizing was not set when the current thread is
614 about to complete messages. It only affects multi-threaded
620 Description: Changed the default kqswlnd ntxmsg=512
625 Description: Assertion failure in kernel ptllnd caused by posting passive
626 bulk buffers before connection establishment complete.
631 Description: A race in kernel ptllnd between deleting a peer and posting
632 new communications for it could hang communications -
633 manifesting as "Unexpectedly long timeout" messages.
638 Description: Kernel ptllnd lock ordering issue could hang a node.
643 Description: node crash on socket teardown race
646 Frequency : 'lctl peer_list' issued on a mx net
648 Description: Enable lctl's peer_list for MXLND
651 Frequency : after Ptllnd timeouts and portals congestion
653 Description: Credit overflows
654 Details : This was a bug in ptllnd connection establishment. The fix
655 implements better peer stamps to disambiguate connection
656 establishment and ensure both peers enter the credit flow
657 state machine consistently.
662 Description: kptllnd didn't propagate some network errors up to LNET
663 Details : This bug was spotted while investigating 11394. The fix
664 ensures network errors on sends and bulk transfers are
665 propagated to LNET/lustre correctly.
667 Severity : enhancement
669 Description: Fixed console chatter in case of -ETIMEDOUT.
671 Severity : enhancement
673 Description: Added D_NETTRACE for recording network packet history
674 (initially only for ptllnd). Also a separate userspace
675 ptllnd facility to gather history which should really be
676 covered by D_NETTRACE too, if only CDEBUG recorded history in
682 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
683 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
684 callback can occur before a connection has actually been
685 established. This caused an assertion failure previously.
687 Severity : enhancement
689 Description: Multiple instances for o2iblnd
690 Details : Allow multiple instances of o2iblnd to enable networking over
691 multiple HCAs and routing between them.
695 Description: lnet deadlock in router_checker
696 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
697 into BH locks to eliminate potential deadlock caused by
698 ksocknal_data_ready() preempting code holding these locks.
702 Description: Millions of failed socklnd connection attempts cause a very slow FS
703 Details : added a new route flag ksnr_scheduled to distinguish from
704 ksnr_connecting, so that a peer connection request is only turned
705 down for race concerns when an active connection to the same peer
706 is under progress (instead of just being scheduled).
708 ------------------------------------------------------------------------------
710 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
712 * Support for networks:
713 socklnd - kernels up to 2.6.16
714 qswlnd - Qsnet kernel modules 5.20 and later
715 openiblnd - IbGold 1.8.2
717 viblnd - Voltaire ibhost 3.4.5 and later
718 ciblnd - Topspin 3.2.0
719 iiblnd - Infiniserv 3.3 + PathBits patch
720 gmlnd - GM 2.1.22 and later
721 mxlnd - MX 1.2.1 or later
722 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
725 Severity : major on XT3
727 Description: libcfs overwrites /proc/sys/portals
728 Details : libcfs created a symlink from /proc/sys/portals to
729 /proc/sys/lnet for backwards compatibility. This is no
730 longer required and makes the Cray portals /proc variables
735 Description: OFED FMR API change
736 Details : This changes parameter usage to reflect a change in
737 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
738 that FMR support is only used in experimental versions of the
739 o2iblnd - this change does not affect standard usage at all.
741 Severity : enhancement
743 Description: new ko2iblnd module parameter: ib_mtu
744 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
745 HCAs. You can avoid this problem by setting the MTU to 1024
746 using this module parameter.
748 Severity : enhancement
749 Bugzilla : 11118/11620
750 Description: ptllnd small request message buffer alignment fix
751 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
752 Round up small message size on sends in case this option
753 is not supported. 11620 was a defect in the initial
754 implementation which effectively asserted all peers had to be
755 running the correct protocol version which was fixed by always
756 NAK-ing such requests and handling any misalignments they
761 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
762 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
763 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
765 Severity : enhancement
767 Description: Patchless ZC(zero copy) socklnd
768 Details : New protocol for socklnd, socklnd can support zero copy without
769 kernel patch, it's compatible with old socklnd. Checksum is
770 moved from tunables to modparams.
774 Description: When ksocknal_del_peer() is called upon a peer whose
775 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
776 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
779 Frequency : when ptlrpc is under heavy use and runs out of request buffer
781 Description: In lnet_match_blocked_msg(), md can be used without holding a
785 Frequency : very rarely
787 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
788 If connd connects a route which has been closed by
789 ksocknal_shutdown(), ksocknal_create_routes() may create new
790 routes which hold references on the peer, causing shutdown
791 process to wait for peer to disappear forever.
793 Severity : enhancement
795 Description: Dump XT3 portals traces on kptllnd timeout
796 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
797 dump Cray portals debug traces to a file. The kptllnd module
798 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
799 is the basename of the dump file.
802 Frequency : infrequent
804 Description: kernel ptllnd fix bug in connection re-establishment
805 Details : Kernel ptllnd could produce protocol errors e.g. illegal
806 matchbits and/or violate the credit flow protocol when trying
807 to re-establish a connection with a peer after an error or
810 Severity : enhancement
812 Description: Allow /proc/sys/lnet/debug to be set symbolically
813 Details : Allow debug and subsystem debug values to be read/set by name
814 in addition to numerically, for ease of use.
817 Frequency : only in configurations with LNET routers
819 Description: routes automatically marked down and recovered
820 Details : In configurations with LNET routers if a router fails routers
821 now actively try to recover routes that are down, unless they
822 are marked down by an administrator.
824 ------------------------------------------------------------------------------
826 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
829 Frequency : very rarely, in configurations with LNET routers and TCP
831 Description: incorrect data written to files on OSTs
832 Details : In certain high-load conditions incorrect data may be written
833 to files on the OST when using TCP networks.
835 ------------------------------------------------------------------------------
837 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
839 - rework CDEBUG messages rate-limiting mechanism b=10375
840 - add per-socket tunables for socklnd if the kernel is patched b=10327
842 ------------------------------------------------------------------------------
844 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
846 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
847 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
848 - quiet LNET startup LNI message for liblustre b=10128
849 - Better console error messages if 'ip2nets' can't match an IP address
850 - Fixed overflow/use-before-set bugs in linux-time.h
851 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
852 - LNET teardown failed an assertion about the route table being empty
853 - Fixed a crash in LNetEQPoll(<invalid handle>)
854 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
855 - improve debug message for liblustre/Catamount nodes (b=10116)
857 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
858 * Configuration change for the XT3
859 The PTLLND is now used to run Lustre over Portals on the XT3.
860 The configure option(s) --with-cray-portals are no longer
861 used. Rather --with-portals=<path-to-portals-includes> is
862 used to enable building on the XT3. In addition to enable
863 XT3 specific features the option --enable-cray-xt3 must be
866 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
867 * Portals has been removed, replaced by LNET.
868 LNET is new networking infrastructure for Lustre, it includes a
869 reorganized network configuration mode (see the user
870 documentation for full details) as well as support for routing
871 between different network fabrics. Lustre Networking Devices
872 (LNDS) for the supported network fabrics have also been created
873 for this new infrastructure.
875 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
880 Frequency : rare (large Voltaire clusters only)
882 Description: the default number of reserved transmit descriptors was too low
883 for some large clusters
884 Details : As a workaround, the number was increased. A proper fix includes
887 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
892 Frequency : occasional (large-scale events, cluster reboot, network failure)
894 Description: too many error messages on console obscure actual problem and
895 can slow down/panic server, or cause recovery to fail repeatedly
896 Details : enable rate-limiting of console error messages, and some messages
897 that were console errors now only go to the kernel log
899 Severity : enhancement
901 Description: add /proc/sys/portals/catastrophe entry which will report if
902 that node has previously LBUGged
904 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
906 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
908 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
910 - handle error return code in kranal_check_fma_rx() (5915,6054)
912 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
914 - update vibnal (Voltaire IB NAL)
915 - update gmnal (Myrinet NAL), gmnalid
917 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
919 * Landed portals:b_port_step as follows...
921 - removed CFS_DECL_SPIN*
922 just use 'spinlock_t' and initialise with spin_lock_init()
924 - removed CFS_DECL_MUTEX*
925 just use 'struct semaphore' and initialise with init_mutex()
927 - removed CFS_DECL_RWSEM*
928 just use 'struct rw_semaphore' and initialise with init_rwsem()
930 - renamed cfs_sleep_chan -> cfs_waitq
931 cfs_sleep_link -> cfs_waitlink
933 - fixed race in linux version of arch-independent socknal
934 (the ENOMEM/EAGAIN decision).
936 - Didn't fix problems in Darwin version of arch-independent socknal
937 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
939 - removed libcfs types from non-socknal header files (only some types
940 in the header files had been changed; the .c files hadn't been