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, and 1.3
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: add a new LND optiion to control peer buffer credits on routers
26 Description: Fixing deadlock in usocklnd
27 Details : A deadlock was possible in usocklnd due to race condition while
28 tearing connection down. The problem resulted from erroneous
29 assumption that lnet_finalize() could have been called holding
33 Bugzilla : 13621, 15983
34 Description: Protocol V2 of o2iblnd
35 Details : o2iblnd V2 has several new features:
36 . map-on-demand: map-on-demand is disabled by default, it can
37 be enabled by using modparam "map_on_demand=@value@", @value@
38 should >= 0 and < 256, 0 will disable map-on-demand, any other
39 valid value will enable map-on-demand.
40 Oi2blnd will create FMR or physical MR for RDMA if fragments of
42 Enable map-on-demand will take less memory for new connection,
43 but a little more CPU for RDMA.
44 . iWARP : to support iWARP, please enable map-on-demand, 32 and 64
45 are recommanded value. iWARP will probably fail for value >=128.
46 . OOB NOOP message: to resolve deadlock on router.
47 . tunable peer_credits_hiw: (high water to return credits),
48 default value of peer_credits_hiw equals to (peer_credits -1),
49 user can change it between peer_credits/2 and (peer_credits - 1).
50 Lower value is recommended for high latency network.
51 . tunable message queue size: it always equals to peer_credits,
52 higher value is recommended for high latency network.
53 . It's compatible with earlier version of o2iblnd
57 Description: Fixing 'running out of ports' issue
58 Details : Add a delay before next reconnect attempt in ksocklnd in
59 the case of lost race. Limit the frequency of query-requests
60 in lnet. Improved handling of 'dead peer' notifications in
65 Description: Change ptllnd timeout and watchdog timers
66 Details : Add ptltrace_on_nal_failed and bump ptllnd timeout to match
71 Description: One down Lustre FS hangs ALL mounted Lustre filesystems
72 Details : Shared routing enhancements - peer health detection.
74 Severity : enhancement
76 Description: acceptor.c cleanup
77 Details : Code duplication in acceptor.c for the cases of kernel and
78 user-space removed. User-space libcfs tcpip primitives
79 uniformed to have prototypes similar to kernel ones. Minor
80 cosmetic changes in usocklnd to use cfs_socket_t as
81 representation of socket.
85 Description: IB path MTU mistakenly set to 1st path MTU when ib_mtu is off
86 Details : See comment 46 in bug 11245 for details - it's indeed a bug
87 introduced by the original 11245 fix.
91 Description: uptllnd credit overflow fix
92 Details : kptl_msg_t::ptlm_credits could be overflown by uptllnd since
97 Description: socklnd protocol version 3
98 Details : With current protocol V2, connections on router can be
99 blocked and can't receive any incoming messages when there is no
100 more router buffer, so ZC-ACK can't be handled (LNet message
101 can't be finalized) and will cause deadlock on router.
102 Protocol V3 has a dedicated connection for emergency messages
103 like ZC-ACK to router, messages on this dedicated connection
104 don't need any credit so will never be blocked. Also, V3 can send
105 keepalive ping in specified period for router healthy checking.
107 -------------------------------------------------------------------------------
109 12-31-2008 Sun Microsystems, Inc.
111 * Support for networks:
112 socklnd - any kernel supported by Lustre,
113 qswlnd - Qsnet kernel modules 5.20 and later,
114 openiblnd - IbGold 1.8.2,
115 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
116 viblnd - Voltaire ibhost 3.4.5 and later,
117 ciblnd - Topspin 3.2.0,
118 iiblnd - Infiniserv 3.3 + PathBits patch,
119 gmlnd - GM 2.1.22 and later,
120 mxlnd - MX 1.2.1 or later,
121 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
125 Description: workaround for OOM from o2iblnd
126 Details : OFED needs allocate big chunk of memory for QP while creating
127 connection for o2iblnd, OOM can happen if no such a contiguous
129 QP size is decided by concurrent_sends and max_fragments of
130 o2iblnd, now we permit user to specify smaller value for
131 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
132 will decrease memory block size required by creating QP.
136 Description: Support Zerocopy receive of Chelsio device
137 Details : Chelsio driver can support zerocopy for iov[1] if it's
138 contiguous and large enough.
142 Description: fix credit flow deadlock in uptllnd
146 Description: finalize network operation in reasonable time
147 Details : conf-sanity test_32a couldn't stop ost and mds because it
148 tried to access non-existent peer and tcp connect took
149 quite long before timing out.
153 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
154 Details : Lost reference on conn prevents peer from being destroyed, which
155 could prevent new peer creation if peer count has reached upper
160 Description: LNET Selftest results in Soft lockup on OSS CPU
161 Details : only hits when 8 or more o2ib clients involved and a session is
162 torn down with 'lst end_session' without preceeding 'lst stop'.
166 Description: concurrent_sends in IB LNDs should not be changeable at run time
167 Details : concurrent_sends in IB LNDs should not be changeable at run time
171 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
172 Details : only hits under out-of-memory situations
175 -------------------------------------------------------------------------------
177 2009-02-07 Sun Microsystems, Inc.
179 * Support for networks:
180 socklnd - any kernel supported by Lustre,
181 qswlnd - Qsnet kernel modules 5.20 and later,
182 openiblnd - IbGold 1.8.2,
183 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
184 viblnd - Voltaire ibhost 3.4.5 and later,
185 ciblnd - Topspin 3.2.0,
186 iiblnd - Infiniserv 3.3 + PathBits patch,
187 gmlnd - GM 2.1.22 and later,
188 mxlnd - MX 1.2.1 or later,
189 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
192 Description: workaround for OOM from o2iblnd
193 Details : OFED needs allocate big chunk of memory for QP while creating
194 connection for o2iblnd, OOM can happen if no such a contiguous
196 QP size is decided by concurrent_sends and max_fragments of
197 o2iblnd, now we permit user to specify smaller value for
198 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
199 will decrease memory block size required by creating QP.
203 Description: Support Zerocopy receive of Chelsio device
204 Details : Chelsio driver can support zerocopy for iov[1] if it's
205 contiguous and large enough.
208 Description: fix credit flow deadlock in uptllnd
212 Description: finalize network operation in reasonable time
213 Details : conf-sanity test_32a couldn't stop ost and mds because it
214 tried to access non-existent peer and tcp connect took
215 quite long before timing out.
219 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
220 Details : Lost reference on conn prevents peer from being destroyed, which
221 could prevent new peer creation if peer count has reached upper
226 Description: LNET Selftest results in Soft lockup on OSS CPU
227 Details : only hits when 8 or more o2ib clients involved and a session is
228 torn down with 'lst end_session' without preceeding 'lst stop'.
232 Description: concurrent_sends in IB LNDs should not be changeable at run time
233 Details : concurrent_sends in IB LNDs should not be changeable at run time
235 -------------------------------------------------------------------------------
237 11-03-2008 Sun Microsystems, Inc.
239 * Support for networks:
240 socklnd - any kernel supported by Lustre,
241 qswlnd - Qsnet kernel modules 5.20 and later,
242 openiblnd - IbGold 1.8.2,
243 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
244 viblnd - Voltaire ibhost 3.4.5 and later,
245 ciblnd - Topspin 3.2.0,
246 iiblnd - Infiniserv 3.3 + PathBits patch,
247 gmlnd - GM 2.1.22 and later,
248 mxlnd - MX 1.2.1 or later,
249 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
253 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
254 Details : only hits under out-of-memory situations
256 -------------------------------------------------------------------------------
258 04-26-2008 Sun Microsystems, Inc.
260 * Support for networks:
261 socklnd - any kernel supported by Lustre,
262 qswlnd - Qsnet kernel modules 5.20 and later,
263 openiblnd - IbGold 1.8.2,
264 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
265 viblnd - Voltaire ibhost 3.4.5 and later,
266 ciblnd - Topspin 3.2.0,
267 iiblnd - Infiniserv 3.3 + PathBits patch,
268 gmlnd - GM 2.1.22 and later,
269 mxlnd - MX 1.2.1 or later,
270 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
274 Description: excessive debug information removed
275 Details : excessive debug information removed
279 Description: ksocknal_create_conn() hit ASSERTION during connection race
280 Details : ksocknal_create_conn() hit ASSERTION during connection race
284 Description: ksocknal_send_hello() hit ASSERTION while connecting race
285 Details : ksocknal_send_hello() hit ASSERTION while connecting race
289 Description: o2iblnd/ptllnd credit deadlock in a routed config.
290 Details : o2iblnd/ptllnd credit deadlock in a routed config.
294 Description: High load after starting lnet
295 Details : gmlnd should sleep in rx thread in interruptible way. Otherwise,
296 uptime utility reports high load that looks confusingly.
300 Description: ksocklnd fails to establish connection if accept_port is high
301 Details : PID remapping must not be done for active (outgoing) connections
304 --------------------------------------------------------------------------------
306 2008-01-11 Sun Microsystems, Inc.
308 * Support for networks:
309 socklnd - any kernel supported by Lustre,
310 qswlnd - Qsnet kernel modules 5.20 and later,
311 openiblnd - IbGold 1.8.2,
312 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
313 viblnd - Voltaire ibhost 3.4.5 and later,
314 ciblnd - Topspin 3.2.0,
315 iiblnd - Infiniserv 3.3 + PathBits patch,
316 gmlnd - GM 2.1.22 and later,
317 mxlnd - MX 1.2.1 or later,
318 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
322 Description: liblustre network error
323 Details : liblustre clients should understand LNET_ACCEPT_PORT environment
324 variable even if they don't start lnet acceptor.
328 Description: Strange message from lnet (Ignoring prediction from the future)
329 Details : Incorrect calculation of peer's last_alive value in ksocklnd
331 --------------------------------------------------------------------------------
333 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
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: ASSERTION(me == md->md_me) failed in lnet_match_md()
353 Description: increase send queue size for ciblnd/openiblnd
357 Description: new userspace socklnd
358 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
359 with new one - usocklnd.
361 Severity : enhancement
363 Description: Console message flood
364 Details : Make cdls ratelimiting more tunable by adding several tunable in
365 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
366 /proc/sys/lnet/console_backoff.
368 --------------------------------------------------------------------------------
370 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
372 * Support for networks:
373 socklnd - any kernel supported by Lustre,
374 qswlnd - Qsnet kernel modules 5.20 and later,
375 openiblnd - IbGold 1.8.2,
376 o2iblnd - OFED 1.1 and 1.2,
377 viblnd - Voltaire ibhost 3.4.5 and later,
378 ciblnd - Topspin 3.2.0,
379 iiblnd - Infiniserv 3.3 + PathBits patch,
380 gmlnd - GM 2.1.22 and later,
381 mxlnd - MX 1.2.1 or later,
382 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
386 Description: /proc/sys/lnet has non-sysctl entries
387 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
391 Description: TOE Kernel panic by ksocklnd
392 Details : offloaded sockets provide their own implementation of sendpage,
393 can't call tcp_sendpage() directly
397 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
398 Details : races between lnd_shutdown and peer creation prevent
399 lnd_shutdown from finishing.
403 Description: open files rlimit 1024 reached while liblustre testing
404 Details : ulnds/socklnd must close open socket after unsuccessful
409 Description: build error
410 Details : fix typos in gmlnd, ptllnd and viblnd
412 --------------------------------------------------------------------------------
414 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
416 * Support for networks:
417 socklnd - kernels up to 2.6.16,
418 qswlnd - Qsnet kernel modules 5.20 and later,
419 openiblnd - IbGold 1.8.2,
420 o2iblnd - OFED 1.1 and 1.2
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
428 --------------------------------------------------------------------------------
430 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
432 * Support for networks:
433 socklnd - kernels up to 2.6.16,
434 qswlnd - Qsnet kernel modules 5.20 and later,
435 openiblnd - IbGold 1.8.2,
437 viblnd - Voltaire ibhost 3.4.5 and later,
438 ciblnd - Topspin 3.2.0,
439 iiblnd - Infiniserv 3.3 + PathBits patch,
440 gmlnd - GM 2.1.22 and later,
441 mxlnd - MX 1.2.1 or later,
442 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
446 Description: Initialize cpumask before use
450 Description: ASSERTION failures when upgrading to the patchless zero-copy
452 Details : This bug affects "rolling upgrades", causing an inconsistent
453 protocol version negotiation and subsequent assertion failure
454 during rolling upgrades after the first wave of upgrades.
458 Details : Change "dropped message" CERRORs to D_NETERROR so they are
459 logged instead of creating "console chatter" when a lustre
460 timeout races with normal RPC completion.
463 Details : lnet_clear_peer_table can wait forever if user forgets to
467 Details : libcfs_id2str should check pid against LNET_PID_ANY.
471 Description: added LNET self test
472 Details : landing b_self_test
477 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
479 Details : do_div() macro is used incorrectly.
481 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
485 Description: make panic on lbug configurable
489 Description: Add OFED1.2 support to o2iblnd
490 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
491 are installed (other than kernel's in-tree infiniband), there
492 could be some problem while insmod o2iblnd (mismatch CRC of
494 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
495 this link provides solution:
496 https://bugs.openfabrics.org/show_bug.cgi?id=355
497 if extra Module.symvers is not supported in kernel, we will
498 have to run the script in bug 12316 to update
499 $LINUX/module.symvers before building o2iblnd.
500 More details about this are in bug 12316.
502 ------------------------------------------------------------------------------
504 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
505 * version 1.4.10 / 1.6.0
506 * Support for networks:
507 socklnd - kernels up to 2.6.16,
508 qswlnd - Qsnet kernel modules 5.20 and later,
509 openiblnd - IbGold 1.8.2,
511 viblnd - Voltaire ibhost 3.4.5 and later,
512 ciblnd - Topspin 3.2.0,
513 iiblnd - Infiniserv 3.3 + PathBits patch,
514 gmlnd - GM 2.1.22 and later,
515 mxlnd - MX 1.2.1 or later,
516 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
520 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
521 possibly accessed in kptllnd_shutdown. Ptllnd should init
522 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
526 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
530 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
534 Description: the_lnet.ln_finalizing was not set when the current thread is
535 about to complete messages. It only affects multi-threaded
541 Description: Changed the default kqswlnd ntxmsg=512
546 Description: Assertion failure in kernel ptllnd caused by posting passive
547 bulk buffers before connection establishment complete.
552 Description: A race in kernel ptllnd between deleting a peer and posting
553 new communications for it could hang communications -
554 manifesting as "Unexpectedly long timeout" messages.
559 Description: Kernel ptllnd lock ordering issue could hang a node.
564 Description: node crash on socket teardown race
567 Frequency : 'lctl peer_list' issued on a mx net
569 Description: Enable lctl's peer_list for MXLND
572 Frequency : after Ptllnd timeouts and portals congestion
574 Description: Credit overflows
575 Details : This was a bug in ptllnd connection establishment. The fix
576 implements better peer stamps to disambiguate connection
577 establishment and ensure both peers enter the credit flow
578 state machine consistently.
583 Description: kptllnd didn't propagate some network errors up to LNET
584 Details : This bug was spotted while investigating 11394. The fix
585 ensures network errors on sends and bulk transfers are
586 propagated to LNET/lustre correctly.
588 Severity : enhancement
590 Description: Fixed console chatter in case of -ETIMEDOUT.
592 Severity : enhancement
594 Description: Added D_NETTRACE for recording network packet history
595 (initially only for ptllnd). Also a separate userspace
596 ptllnd facility to gather history which should really be
597 covered by D_NETTRACE too, if only CDEBUG recorded history in
603 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
604 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
605 callback can occur before a connection has actually been
606 established. This caused an assertion failure previously.
608 Severity : enhancement
610 Description: Multiple instances for o2iblnd
611 Details : Allow multiple instances of o2iblnd to enable networking over
612 multiple HCAs and routing between them.
616 Description: lnet deadlock in router_checker
617 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
618 into BH locks to eliminate potential deadlock caused by
619 ksocknal_data_ready() preempting code holding these locks.
623 Description: Millions of failed socklnd connection attempts cause a very slow FS
624 Details : added a new route flag ksnr_scheduled to distinguish from
625 ksnr_connecting, so that a peer connection request is only turned
626 down for race concerns when an active connection to the same peer
627 is under progress (instead of just being scheduled).
629 ------------------------------------------------------------------------------
631 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
633 * Support for networks:
634 socklnd - kernels up to 2.6.16
635 qswlnd - Qsnet kernel modules 5.20 and later
636 openiblnd - IbGold 1.8.2
638 viblnd - Voltaire ibhost 3.4.5 and later
639 ciblnd - Topspin 3.2.0
640 iiblnd - Infiniserv 3.3 + PathBits patch
641 gmlnd - GM 2.1.22 and later
642 mxlnd - MX 1.2.1 or later
643 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
646 Severity : major on XT3
648 Description: libcfs overwrites /proc/sys/portals
649 Details : libcfs created a symlink from /proc/sys/portals to
650 /proc/sys/lnet for backwards compatibility. This is no
651 longer required and makes the Cray portals /proc variables
656 Description: OFED FMR API change
657 Details : This changes parameter usage to reflect a change in
658 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
659 that FMR support is only used in experimental versions of the
660 o2iblnd - this change does not affect standard usage at all.
662 Severity : enhancement
664 Description: new ko2iblnd module parameter: ib_mtu
665 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
666 HCAs. You can avoid this problem by setting the MTU to 1024
667 using this module parameter.
669 Severity : enhancement
670 Bugzilla : 11118/11620
671 Description: ptllnd small request message buffer alignment fix
672 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
673 Round up small message size on sends in case this option
674 is not supported. 11620 was a defect in the initial
675 implementation which effectively asserted all peers had to be
676 running the correct protocol version which was fixed by always
677 NAK-ing such requests and handling any misalignments they
682 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
683 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
684 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
686 Severity : enhancement
688 Description: Patchless ZC(zero copy) socklnd
689 Details : New protocol for socklnd, socklnd can support zero copy without
690 kernel patch, it's compatible with old socklnd. Checksum is
691 moved from tunables to modparams.
695 Description: When ksocknal_del_peer() is called upon a peer whose
696 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
697 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
700 Frequency : when ptlrpc is under heavy use and runs out of request buffer
702 Description: In lnet_match_blocked_msg(), md can be used without holding a
706 Frequency : very rarely
708 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
709 If connd connects a route which has been closed by
710 ksocknal_shutdown(), ksocknal_create_routes() may create new
711 routes which hold references on the peer, causing shutdown
712 process to wait for peer to disappear forever.
714 Severity : enhancement
716 Description: Dump XT3 portals traces on kptllnd timeout
717 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
718 dump Cray portals debug traces to a file. The kptllnd module
719 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
720 is the basename of the dump file.
723 Frequency : infrequent
725 Description: kernel ptllnd fix bug in connection re-establishment
726 Details : Kernel ptllnd could produce protocol errors e.g. illegal
727 matchbits and/or violate the credit flow protocol when trying
728 to re-establish a connection with a peer after an error or
731 Severity : enhancement
733 Description: Allow /proc/sys/lnet/debug to be set symbolically
734 Details : Allow debug and subsystem debug values to be read/set by name
735 in addition to numerically, for ease of use.
738 Frequency : only in configurations with LNET routers
740 Description: routes automatically marked down and recovered
741 Details : In configurations with LNET routers if a router fails routers
742 now actively try to recover routes that are down, unless they
743 are marked down by an administrator.
745 ------------------------------------------------------------------------------
747 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
750 Frequency : very rarely, in configurations with LNET routers and TCP
752 Description: incorrect data written to files on OSTs
753 Details : In certain high-load conditions incorrect data may be written
754 to files on the OST when using TCP networks.
756 ------------------------------------------------------------------------------
758 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
760 - rework CDEBUG messages rate-limiting mechanism b=10375
761 - add per-socket tunables for socklnd if the kernel is patched b=10327
763 ------------------------------------------------------------------------------
765 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
767 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
768 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
769 - quiet LNET startup LNI message for liblustre b=10128
770 - Better console error messages if 'ip2nets' can't match an IP address
771 - Fixed overflow/use-before-set bugs in linux-time.h
772 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
773 - LNET teardown failed an assertion about the route table being empty
774 - Fixed a crash in LNetEQPoll(<invalid handle>)
775 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
776 - improve debug message for liblustre/Catamount nodes (b=10116)
778 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
779 * Configuration change for the XT3
780 The PTLLND is now used to run Lustre over Portals on the XT3.
781 The configure option(s) --with-cray-portals are no longer
782 used. Rather --with-portals=<path-to-portals-includes> is
783 used to enable building on the XT3. In addition to enable
784 XT3 specific features the option --enable-cray-xt3 must be
787 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
788 * Portals has been removed, replaced by LNET.
789 LNET is new networking infrastructure for Lustre, it includes a
790 reorganized network configuration mode (see the user
791 documentation for full details) as well as support for routing
792 between different network fabrics. Lustre Networking Devices
793 (LNDS) for the supported network fabrics have also been created
794 for this new infrastructure.
796 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
801 Frequency : rare (large Voltaire clusters only)
803 Description: the default number of reserved transmit descriptors was too low
804 for some large clusters
805 Details : As a workaround, the number was increased. A proper fix includes
808 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
813 Frequency : occasional (large-scale events, cluster reboot, network failure)
815 Description: too many error messages on console obscure actual problem and
816 can slow down/panic server, or cause recovery to fail repeatedly
817 Details : enable rate-limiting of console error messages, and some messages
818 that were console errors now only go to the kernel log
820 Severity : enhancement
822 Description: add /proc/sys/portals/catastrophe entry which will report if
823 that node has previously LBUGged
825 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
827 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
829 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
831 - handle error return code in kranal_check_fma_rx() (5915,6054)
833 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
835 - update vibnal (Voltaire IB NAL)
836 - update gmnal (Myrinet NAL), gmnalid
838 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
840 * Landed portals:b_port_step as follows...
842 - removed CFS_DECL_SPIN*
843 just use 'spinlock_t' and initialise with spin_lock_init()
845 - removed CFS_DECL_MUTEX*
846 just use 'struct semaphore' and initialise with init_mutex()
848 - removed CFS_DECL_RWSEM*
849 just use 'struct rw_semaphore' and initialise with init_rwsem()
851 - renamed cfs_sleep_chan -> cfs_waitq
852 cfs_sleep_link -> cfs_waitlink
854 - fixed race in linux version of arch-independent socknal
855 (the ENOMEM/EAGAIN decision).
857 - Didn't fix problems in Darwin version of arch-independent socknal
858 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
860 - removed libcfs types from non-socknal header files (only some types
861 in the header files had been changed; the .c files hadn't been