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
20 Severity : enhancement
22 Description: allow a test node to be a member of multiple test groups
24 Severity : enhancement
26 Description: MXLND: eliminate hosts file, use arp for peer nic_id resolution
27 Details : an update from the upstream developer Scott Atchley.
29 Severity : enhancement
31 Description: add a new LND optiion to control peer buffer credits on routers
35 Description: Fixing deadlock in usocklnd
36 Details : A deadlock was possible in usocklnd due to race condition while
37 tearing connection down. The problem resulted from erroneous
38 assumption that lnet_finalize() could have been called holding
42 Bugzilla : 13621, 15983
43 Description: Protocol V2 of o2iblnd
44 Details : o2iblnd V2 has several new features:
45 . map-on-demand: map-on-demand is disabled by default, it can
46 be enabled by using modparam "map_on_demand=@value@", @value@
47 should >= 0 and < 256, 0 will disable map-on-demand, any other
48 valid value will enable map-on-demand.
49 Oi2blnd will create FMR or physical MR for RDMA if fragments of
51 Enable map-on-demand will take less memory for new connection,
52 but a little more CPU for RDMA.
53 . iWARP : to support iWARP, please enable map-on-demand, 32 and 64
54 are recommanded value. iWARP will probably fail for value >=128.
55 . OOB NOOP message: to resolve deadlock on router.
56 . tunable peer_credits_hiw: (high water to return credits),
57 default value of peer_credits_hiw equals to (peer_credits -1),
58 user can change it between peer_credits/2 and (peer_credits - 1).
59 Lower value is recommended for high latency network.
60 . tunable message queue size: it always equals to peer_credits,
61 higher value is recommended for high latency network.
62 . It's compatible with earlier version of o2iblnd
66 Description: Fixing 'running out of ports' issue
67 Details : Add a delay before next reconnect attempt in ksocklnd in
68 the case of lost race. Limit the frequency of query-requests
69 in lnet. Improved handling of 'dead peer' notifications in
74 Description: Change ptllnd timeout and watchdog timers
75 Details : Add ptltrace_on_nal_failed and bump ptllnd timeout to match
80 Description: One down Lustre FS hangs ALL mounted Lustre filesystems
81 Details : Shared routing enhancements - peer health detection.
83 Severity : enhancement
85 Description: acceptor.c cleanup
86 Details : Code duplication in acceptor.c for the cases of kernel and
87 user-space removed. User-space libcfs tcpip primitives
88 uniformed to have prototypes similar to kernel ones. Minor
89 cosmetic changes in usocklnd to use cfs_socket_t as
90 representation of socket.
94 Description: IB path MTU mistakenly set to 1st path MTU when ib_mtu is off
95 Details : See comment 46 in bug 11245 for details - it's indeed a bug
96 introduced by the original 11245 fix.
100 Description: uptllnd credit overflow fix
101 Details : kptl_msg_t::ptlm_credits could be overflown by uptllnd since
106 Description: socklnd protocol version 3
107 Details : With current protocol V2, connections on router can be
108 blocked and can't receive any incoming messages when there is no
109 more router buffer, so ZC-ACK can't be handled (LNet message
110 can't be finalized) and will cause deadlock on router.
111 Protocol V3 has a dedicated connection for emergency messages
112 like ZC-ACK to router, messages on this dedicated connection
113 don't need any credit so will never be blocked. Also, V3 can send
114 keepalive ping in specified period for router healthy checking.
116 -------------------------------------------------------------------------------
118 12-31-2008 Sun Microsystems, Inc.
120 * Support for networks:
121 socklnd - any kernel supported by Lustre,
122 qswlnd - Qsnet kernel modules 5.20 and later,
123 openiblnd - IbGold 1.8.2,
124 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
125 viblnd - Voltaire ibhost 3.4.5 and later,
126 ciblnd - Topspin 3.2.0,
127 iiblnd - Infiniserv 3.3 + PathBits patch,
128 gmlnd - GM 2.1.22 and later,
129 mxlnd - MX 1.2.1 or later,
130 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
134 Description: workaround for OOM from o2iblnd
135 Details : OFED needs allocate big chunk of memory for QP while creating
136 connection for o2iblnd, OOM can happen if no such a contiguous
138 QP size is decided by concurrent_sends and max_fragments of
139 o2iblnd, now we permit user to specify smaller value for
140 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
141 will decrease memory block size required by creating QP.
145 Description: Support Zerocopy receive of Chelsio device
146 Details : Chelsio driver can support zerocopy for iov[1] if it's
147 contiguous and large enough.
151 Description: fix credit flow deadlock in uptllnd
155 Description: finalize network operation in reasonable time
156 Details : conf-sanity test_32a couldn't stop ost and mds because it
157 tried to access non-existent peer and tcp connect took
158 quite long before timing out.
162 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
163 Details : Lost reference on conn prevents peer from being destroyed, which
164 could prevent new peer creation if peer count has reached upper
169 Description: LNET Selftest results in Soft lockup on OSS CPU
170 Details : only hits when 8 or more o2ib clients involved and a session is
171 torn down with 'lst end_session' without preceeding 'lst stop'.
175 Description: concurrent_sends in IB LNDs should not be changeable at run time
176 Details : concurrent_sends in IB LNDs should not be changeable at run time
180 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
181 Details : only hits under out-of-memory situations
184 -------------------------------------------------------------------------------
186 2009-02-07 Sun Microsystems, Inc.
188 * Support for networks:
189 socklnd - any kernel supported by Lustre,
190 qswlnd - Qsnet kernel modules 5.20 and later,
191 openiblnd - IbGold 1.8.2,
192 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
193 viblnd - Voltaire ibhost 3.4.5 and later,
194 ciblnd - Topspin 3.2.0,
195 iiblnd - Infiniserv 3.3 + PathBits patch,
196 gmlnd - GM 2.1.22 and later,
197 mxlnd - MX 1.2.1 or later,
198 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
201 Description: workaround for OOM from o2iblnd
202 Details : OFED needs allocate big chunk of memory for QP while creating
203 connection for o2iblnd, OOM can happen if no such a contiguous
205 QP size is decided by concurrent_sends and max_fragments of
206 o2iblnd, now we permit user to specify smaller value for
207 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
208 will decrease memory block size required by creating QP.
212 Description: Support Zerocopy receive of Chelsio device
213 Details : Chelsio driver can support zerocopy for iov[1] if it's
214 contiguous and large enough.
217 Description: fix credit flow deadlock in uptllnd
221 Description: finalize network operation in reasonable time
222 Details : conf-sanity test_32a couldn't stop ost and mds because it
223 tried to access non-existent peer and tcp connect took
224 quite long before timing out.
228 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
229 Details : Lost reference on conn prevents peer from being destroyed, which
230 could prevent new peer creation if peer count has reached upper
235 Description: LNET Selftest results in Soft lockup on OSS CPU
236 Details : only hits when 8 or more o2ib clients involved and a session is
237 torn down with 'lst end_session' without preceeding 'lst stop'.
241 Description: concurrent_sends in IB LNDs should not be changeable at run time
242 Details : concurrent_sends in IB LNDs should not be changeable at run time
244 -------------------------------------------------------------------------------
246 11-03-2008 Sun Microsystems, Inc.
248 * Support for networks:
249 socklnd - any kernel supported by Lustre,
250 qswlnd - Qsnet kernel modules 5.20 and later,
251 openiblnd - IbGold 1.8.2,
252 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
253 viblnd - Voltaire ibhost 3.4.5 and later,
254 ciblnd - Topspin 3.2.0,
255 iiblnd - Infiniserv 3.3 + PathBits patch,
256 gmlnd - GM 2.1.22 and later,
257 mxlnd - MX 1.2.1 or later,
258 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
262 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
263 Details : only hits under out-of-memory situations
265 -------------------------------------------------------------------------------
267 04-26-2008 Sun Microsystems, Inc.
269 * Support for networks:
270 socklnd - any kernel supported by Lustre,
271 qswlnd - Qsnet kernel modules 5.20 and later,
272 openiblnd - IbGold 1.8.2,
273 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
274 viblnd - Voltaire ibhost 3.4.5 and later,
275 ciblnd - Topspin 3.2.0,
276 iiblnd - Infiniserv 3.3 + PathBits patch,
277 gmlnd - GM 2.1.22 and later,
278 mxlnd - MX 1.2.1 or later,
279 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
283 Description: excessive debug information removed
284 Details : excessive debug information removed
288 Description: ksocknal_create_conn() hit ASSERTION during connection race
289 Details : ksocknal_create_conn() hit ASSERTION during connection race
293 Description: ksocknal_send_hello() hit ASSERTION while connecting race
294 Details : ksocknal_send_hello() hit ASSERTION while connecting race
298 Description: o2iblnd/ptllnd credit deadlock in a routed config.
299 Details : o2iblnd/ptllnd credit deadlock in a routed config.
303 Description: High load after starting lnet
304 Details : gmlnd should sleep in rx thread in interruptible way. Otherwise,
305 uptime utility reports high load that looks confusingly.
309 Description: ksocklnd fails to establish connection if accept_port is high
310 Details : PID remapping must not be done for active (outgoing) connections
313 --------------------------------------------------------------------------------
315 2008-01-11 Sun Microsystems, Inc.
317 * Support for networks:
318 socklnd - any kernel supported by Lustre,
319 qswlnd - Qsnet kernel modules 5.20 and later,
320 openiblnd - IbGold 1.8.2,
321 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
322 viblnd - Voltaire ibhost 3.4.5 and later,
323 ciblnd - Topspin 3.2.0,
324 iiblnd - Infiniserv 3.3 + PathBits patch,
325 gmlnd - GM 2.1.22 and later,
326 mxlnd - MX 1.2.1 or later,
327 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
331 Description: liblustre network error
332 Details : liblustre clients should understand LNET_ACCEPT_PORT environment
333 variable even if they don't start lnet acceptor.
337 Description: Strange message from lnet (Ignoring prediction from the future)
338 Details : Incorrect calculation of peer's last_alive value in ksocklnd
340 --------------------------------------------------------------------------------
342 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
344 * Support for networks:
345 socklnd - any kernel supported by Lustre,
346 qswlnd - Qsnet kernel modules 5.20 and later,
347 openiblnd - IbGold 1.8.2,
348 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5.
349 viblnd - Voltaire ibhost 3.4.5 and later,
350 ciblnd - Topspin 3.2.0,
351 iiblnd - Infiniserv 3.3 + PathBits patch,
352 gmlnd - GM 2.1.22 and later,
353 mxlnd - MX 1.2.1 or later,
354 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
358 Description: ASSERTION(me == md->md_me) failed in lnet_match_md()
362 Description: increase send queue size for ciblnd/openiblnd
366 Description: new userspace socklnd
367 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
368 with new one - usocklnd.
370 Severity : enhancement
372 Description: Console message flood
373 Details : Make cdls ratelimiting more tunable by adding several tunable in
374 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
375 /proc/sys/lnet/console_backoff.
377 --------------------------------------------------------------------------------
379 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
381 * Support for networks:
382 socklnd - any kernel supported by Lustre,
383 qswlnd - Qsnet kernel modules 5.20 and later,
384 openiblnd - IbGold 1.8.2,
385 o2iblnd - OFED 1.1 and 1.2,
386 viblnd - Voltaire ibhost 3.4.5 and later,
387 ciblnd - Topspin 3.2.0,
388 iiblnd - Infiniserv 3.3 + PathBits patch,
389 gmlnd - GM 2.1.22 and later,
390 mxlnd - MX 1.2.1 or later,
391 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
395 Description: /proc/sys/lnet has non-sysctl entries
396 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
400 Description: TOE Kernel panic by ksocklnd
401 Details : offloaded sockets provide their own implementation of sendpage,
402 can't call tcp_sendpage() directly
406 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
407 Details : races between lnd_shutdown and peer creation prevent
408 lnd_shutdown from finishing.
412 Description: open files rlimit 1024 reached while liblustre testing
413 Details : ulnds/socklnd must close open socket after unsuccessful
418 Description: build error
419 Details : fix typos in gmlnd, ptllnd and viblnd
421 --------------------------------------------------------------------------------
423 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
425 * Support for networks:
426 socklnd - kernels up to 2.6.16,
427 qswlnd - Qsnet kernel modules 5.20 and later,
428 openiblnd - IbGold 1.8.2,
429 o2iblnd - OFED 1.1 and 1.2
430 viblnd - Voltaire ibhost 3.4.5 and later,
431 ciblnd - Topspin 3.2.0,
432 iiblnd - Infiniserv 3.3 + PathBits patch,
433 gmlnd - GM 2.1.22 and later,
434 mxlnd - MX 1.2.1 or later,
435 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
437 --------------------------------------------------------------------------------
439 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
441 * Support for networks:
442 socklnd - kernels up to 2.6.16,
443 qswlnd - Qsnet kernel modules 5.20 and later,
444 openiblnd - IbGold 1.8.2,
446 viblnd - Voltaire ibhost 3.4.5 and later,
447 ciblnd - Topspin 3.2.0,
448 iiblnd - Infiniserv 3.3 + PathBits patch,
449 gmlnd - GM 2.1.22 and later,
450 mxlnd - MX 1.2.1 or later,
451 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
455 Description: Initialize cpumask before use
459 Description: ASSERTION failures when upgrading to the patchless zero-copy
461 Details : This bug affects "rolling upgrades", causing an inconsistent
462 protocol version negotiation and subsequent assertion failure
463 during rolling upgrades after the first wave of upgrades.
467 Details : Change "dropped message" CERRORs to D_NETERROR so they are
468 logged instead of creating "console chatter" when a lustre
469 timeout races with normal RPC completion.
472 Details : lnet_clear_peer_table can wait forever if user forgets to
476 Details : libcfs_id2str should check pid against LNET_PID_ANY.
480 Description: added LNET self test
481 Details : landing b_self_test
486 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
488 Details : do_div() macro is used incorrectly.
490 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
494 Description: make panic on lbug configurable
498 Description: Add OFED1.2 support to o2iblnd
499 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
500 are installed (other than kernel's in-tree infiniband), there
501 could be some problem while insmod o2iblnd (mismatch CRC of
503 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
504 this link provides solution:
505 https://bugs.openfabrics.org/show_bug.cgi?id=355
506 if extra Module.symvers is not supported in kernel, we will
507 have to run the script in bug 12316 to update
508 $LINUX/module.symvers before building o2iblnd.
509 More details about this are in bug 12316.
511 ------------------------------------------------------------------------------
513 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
514 * version 1.4.10 / 1.6.0
515 * Support for networks:
516 socklnd - kernels up to 2.6.16,
517 qswlnd - Qsnet kernel modules 5.20 and later,
518 openiblnd - IbGold 1.8.2,
520 viblnd - Voltaire ibhost 3.4.5 and later,
521 ciblnd - Topspin 3.2.0,
522 iiblnd - Infiniserv 3.3 + PathBits patch,
523 gmlnd - GM 2.1.22 and later,
524 mxlnd - MX 1.2.1 or later,
525 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
529 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
530 possibly accessed in kptllnd_shutdown. Ptllnd should init
531 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
535 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
539 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
543 Description: the_lnet.ln_finalizing was not set when the current thread is
544 about to complete messages. It only affects multi-threaded
550 Description: Changed the default kqswlnd ntxmsg=512
555 Description: Assertion failure in kernel ptllnd caused by posting passive
556 bulk buffers before connection establishment complete.
561 Description: A race in kernel ptllnd between deleting a peer and posting
562 new communications for it could hang communications -
563 manifesting as "Unexpectedly long timeout" messages.
568 Description: Kernel ptllnd lock ordering issue could hang a node.
573 Description: node crash on socket teardown race
576 Frequency : 'lctl peer_list' issued on a mx net
578 Description: Enable lctl's peer_list for MXLND
581 Frequency : after Ptllnd timeouts and portals congestion
583 Description: Credit overflows
584 Details : This was a bug in ptllnd connection establishment. The fix
585 implements better peer stamps to disambiguate connection
586 establishment and ensure both peers enter the credit flow
587 state machine consistently.
592 Description: kptllnd didn't propagate some network errors up to LNET
593 Details : This bug was spotted while investigating 11394. The fix
594 ensures network errors on sends and bulk transfers are
595 propagated to LNET/lustre correctly.
597 Severity : enhancement
599 Description: Fixed console chatter in case of -ETIMEDOUT.
601 Severity : enhancement
603 Description: Added D_NETTRACE for recording network packet history
604 (initially only for ptllnd). Also a separate userspace
605 ptllnd facility to gather history which should really be
606 covered by D_NETTRACE too, if only CDEBUG recorded history in
612 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
613 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
614 callback can occur before a connection has actually been
615 established. This caused an assertion failure previously.
617 Severity : enhancement
619 Description: Multiple instances for o2iblnd
620 Details : Allow multiple instances of o2iblnd to enable networking over
621 multiple HCAs and routing between them.
625 Description: lnet deadlock in router_checker
626 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
627 into BH locks to eliminate potential deadlock caused by
628 ksocknal_data_ready() preempting code holding these locks.
632 Description: Millions of failed socklnd connection attempts cause a very slow FS
633 Details : added a new route flag ksnr_scheduled to distinguish from
634 ksnr_connecting, so that a peer connection request is only turned
635 down for race concerns when an active connection to the same peer
636 is under progress (instead of just being scheduled).
638 ------------------------------------------------------------------------------
640 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
642 * Support for networks:
643 socklnd - kernels up to 2.6.16
644 qswlnd - Qsnet kernel modules 5.20 and later
645 openiblnd - IbGold 1.8.2
647 viblnd - Voltaire ibhost 3.4.5 and later
648 ciblnd - Topspin 3.2.0
649 iiblnd - Infiniserv 3.3 + PathBits patch
650 gmlnd - GM 2.1.22 and later
651 mxlnd - MX 1.2.1 or later
652 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
655 Severity : major on XT3
657 Description: libcfs overwrites /proc/sys/portals
658 Details : libcfs created a symlink from /proc/sys/portals to
659 /proc/sys/lnet for backwards compatibility. This is no
660 longer required and makes the Cray portals /proc variables
665 Description: OFED FMR API change
666 Details : This changes parameter usage to reflect a change in
667 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
668 that FMR support is only used in experimental versions of the
669 o2iblnd - this change does not affect standard usage at all.
671 Severity : enhancement
673 Description: new ko2iblnd module parameter: ib_mtu
674 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
675 HCAs. You can avoid this problem by setting the MTU to 1024
676 using this module parameter.
678 Severity : enhancement
679 Bugzilla : 11118/11620
680 Description: ptllnd small request message buffer alignment fix
681 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
682 Round up small message size on sends in case this option
683 is not supported. 11620 was a defect in the initial
684 implementation which effectively asserted all peers had to be
685 running the correct protocol version which was fixed by always
686 NAK-ing such requests and handling any misalignments they
691 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
692 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
693 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
695 Severity : enhancement
697 Description: Patchless ZC(zero copy) socklnd
698 Details : New protocol for socklnd, socklnd can support zero copy without
699 kernel patch, it's compatible with old socklnd. Checksum is
700 moved from tunables to modparams.
704 Description: When ksocknal_del_peer() is called upon a peer whose
705 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
706 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
709 Frequency : when ptlrpc is under heavy use and runs out of request buffer
711 Description: In lnet_match_blocked_msg(), md can be used without holding a
715 Frequency : very rarely
717 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
718 If connd connects a route which has been closed by
719 ksocknal_shutdown(), ksocknal_create_routes() may create new
720 routes which hold references on the peer, causing shutdown
721 process to wait for peer to disappear forever.
723 Severity : enhancement
725 Description: Dump XT3 portals traces on kptllnd timeout
726 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
727 dump Cray portals debug traces to a file. The kptllnd module
728 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
729 is the basename of the dump file.
732 Frequency : infrequent
734 Description: kernel ptllnd fix bug in connection re-establishment
735 Details : Kernel ptllnd could produce protocol errors e.g. illegal
736 matchbits and/or violate the credit flow protocol when trying
737 to re-establish a connection with a peer after an error or
740 Severity : enhancement
742 Description: Allow /proc/sys/lnet/debug to be set symbolically
743 Details : Allow debug and subsystem debug values to be read/set by name
744 in addition to numerically, for ease of use.
747 Frequency : only in configurations with LNET routers
749 Description: routes automatically marked down and recovered
750 Details : In configurations with LNET routers if a router fails routers
751 now actively try to recover routes that are down, unless they
752 are marked down by an administrator.
754 ------------------------------------------------------------------------------
756 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
759 Frequency : very rarely, in configurations with LNET routers and TCP
761 Description: incorrect data written to files on OSTs
762 Details : In certain high-load conditions incorrect data may be written
763 to files on the OST when using TCP networks.
765 ------------------------------------------------------------------------------
767 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
769 - rework CDEBUG messages rate-limiting mechanism b=10375
770 - add per-socket tunables for socklnd if the kernel is patched b=10327
772 ------------------------------------------------------------------------------
774 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
776 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
777 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
778 - quiet LNET startup LNI message for liblustre b=10128
779 - Better console error messages if 'ip2nets' can't match an IP address
780 - Fixed overflow/use-before-set bugs in linux-time.h
781 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
782 - LNET teardown failed an assertion about the route table being empty
783 - Fixed a crash in LNetEQPoll(<invalid handle>)
784 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
785 - improve debug message for liblustre/Catamount nodes (b=10116)
787 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
788 * Configuration change for the XT3
789 The PTLLND is now used to run Lustre over Portals on the XT3.
790 The configure option(s) --with-cray-portals are no longer
791 used. Rather --with-portals=<path-to-portals-includes> is
792 used to enable building on the XT3. In addition to enable
793 XT3 specific features the option --enable-cray-xt3 must be
796 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
797 * Portals has been removed, replaced by LNET.
798 LNET is new networking infrastructure for Lustre, it includes a
799 reorganized network configuration mode (see the user
800 documentation for full details) as well as support for routing
801 between different network fabrics. Lustre Networking Devices
802 (LNDS) for the supported network fabrics have also been created
803 for this new infrastructure.
805 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
810 Frequency : rare (large Voltaire clusters only)
812 Description: the default number of reserved transmit descriptors was too low
813 for some large clusters
814 Details : As a workaround, the number was increased. A proper fix includes
817 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
822 Frequency : occasional (large-scale events, cluster reboot, network failure)
824 Description: too many error messages on console obscure actual problem and
825 can slow down/panic server, or cause recovery to fail repeatedly
826 Details : enable rate-limiting of console error messages, and some messages
827 that were console errors now only go to the kernel log
829 Severity : enhancement
831 Description: add /proc/sys/portals/catastrophe entry which will report if
832 that node has previously LBUGged
834 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
836 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
838 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
840 - handle error return code in kranal_check_fma_rx() (5915,6054)
842 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
844 - update vibnal (Voltaire IB NAL)
845 - update gmnal (Myrinet NAL), gmnalid
847 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
849 * Landed portals:b_port_step as follows...
851 - removed CFS_DECL_SPIN*
852 just use 'spinlock_t' and initialise with spin_lock_init()
854 - removed CFS_DECL_MUTEX*
855 just use 'struct semaphore' and initialise with init_mutex()
857 - removed CFS_DECL_RWSEM*
858 just use 'struct rw_semaphore' and initialise with init_rwsem()
860 - renamed cfs_sleep_chan -> cfs_waitq
861 cfs_sleep_link -> cfs_waitlink
863 - fixed race in linux version of arch-independent socknal
864 (the ENOMEM/EAGAIN decision).
866 - Didn't fix problems in Darwin version of arch-independent socknal
867 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
869 - removed libcfs types from non-socknal header files (only some types
870 in the header files had been changed; the .c files hadn't been