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
22 Description: Fixing deadlock in usocklnd
23 Details : A deadlock was possible in usocklnd due to race condition while
24 tearing connection down. The problem resulted from erroneous
25 assumption that lnet_finalize() could have been called holding
29 Bugzilla : 13621, 15983
30 Description: Protocol V2 of o2iblnd
31 Details : o2iblnd V2 has several new features:
32 . map-on-demand: map-on-demand is disabled by default, it can
33 be enabled by using modparam "map_on_demand=@value@", @value@
34 should >= 0 and < 256, 0 will disable map-on-demand, any other
35 valid value will enable map-on-demand.
36 Oi2blnd will create FMR or physical MR for RDMA if fragments of
38 Enable map-on-demand will take less memory for new connection,
39 but a little more CPU for RDMA.
40 . iWARP : to support iWARP, please enable map-on-demand, 32 and 64
41 are recommanded value. iWARP will probably fail for value >=128.
42 . OOB NOOP message: to resolve deadlock on router.
43 . tunable peer_credits_hiw: (high water to return credits),
44 default value of peer_credits_hiw equals to (peer_credits -1),
45 user can change it between peer_credits/2 and (peer_credits - 1).
46 Lower value is recommended for high latency network.
47 . tunable message queue size: it always equals to peer_credits,
48 higher value is recommended for high latency network.
49 . It's compatible with earlier version of o2iblnd
53 Description: Fixing 'running out of ports' issue
54 Details : Add a delay before next reconnect attempt in ksocklnd in
55 the case of lost race. Limit the frequency of query-requests
56 in lnet. Improved handling of 'dead peer' notifications in
61 Description: Change ptllnd timeout and watchdog timers
62 Details : Add ptltrace_on_nal_failed and bump ptllnd timeout to match
67 Description: One down Lustre FS hangs ALL mounted Lustre filesystems
68 Details : Shared routing enhancements - peer health detection.
70 Severity : enhancement
72 Description: acceptor.c cleanup
73 Details : Code duplication in acceptor.c for the cases of kernel and
74 user-space removed. User-space libcfs tcpip primitives
75 uniformed to have prototypes similar to kernel ones. Minor
76 cosmetic changes in usocklnd to use cfs_socket_t as
77 representation of socket.
81 Description: IB path MTU mistakenly set to 1st path MTU when ib_mtu is off
82 Details : See comment 46 in bug 11245 for details - it's indeed a bug
83 introduced by the original 11245 fix.
87 Description: uptllnd credit overflow fix
88 Details : kptl_msg_t::ptlm_credits could be overflown by uptllnd since
93 Description: socklnd protocol version 3
94 Details : With current protocol V2, connections on router can be
95 blocked and can't receive any incoming messages when there is no
96 more router buffer, so ZC-ACK can't be handled (LNet message
97 can't be finalized) and will cause deadlock on router.
98 Protocol V3 has a dedicated connection for emergency messages
99 like ZC-ACK to router, messages on this dedicated connection
100 don't need any credit so will never be blocked. Also, V3 can send
101 keepalive ping in specified period for router healthy checking.
103 -------------------------------------------------------------------------------
105 12-31-2008 Sun Microsystems, Inc.
107 * Support for networks:
108 socklnd - any kernel supported by Lustre,
109 qswlnd - Qsnet kernel modules 5.20 and later,
110 openiblnd - IbGold 1.8.2,
111 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
112 viblnd - Voltaire ibhost 3.4.5 and later,
113 ciblnd - Topspin 3.2.0,
114 iiblnd - Infiniserv 3.3 + PathBits patch,
115 gmlnd - GM 2.1.22 and later,
116 mxlnd - MX 1.2.1 or later,
117 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
121 Description: workaround for OOM from o2iblnd
122 Details : OFED needs allocate big chunk of memory for QP while creating
123 connection for o2iblnd, OOM can happen if no such a contiguous
125 QP size is decided by concurrent_sends and max_fragments of
126 o2iblnd, now we permit user to specify smaller value for
127 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
128 will decrease memory block size required by creating QP.
132 Description: Support Zerocopy receive of Chelsio device
133 Details : Chelsio driver can support zerocopy for iov[1] if it's
134 contiguous and large enough.
138 Description: fix credit flow deadlock in uptllnd
142 Description: finalize network operation in reasonable time
143 Details : conf-sanity test_32a couldn't stop ost and mds because it
144 tried to access non-existent peer and tcp connect took
145 quite long before timing out.
149 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
150 Details : Lost reference on conn prevents peer from being destroyed, which
151 could prevent new peer creation if peer count has reached upper
156 Description: LNET Selftest results in Soft lockup on OSS CPU
157 Details : only hits when 8 or more o2ib clients involved and a session is
158 torn down with 'lst end_session' without preceeding 'lst stop'.
162 Description: concurrent_sends in IB LNDs should not be changeable at run time
163 Details : concurrent_sends in IB LNDs should not be changeable at run time
167 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
168 Details : only hits under out-of-memory situations
171 -------------------------------------------------------------------------------
173 2009-02-07 Sun Microsystems, Inc.
175 * Support for networks:
176 socklnd - any kernel supported by Lustre,
177 qswlnd - Qsnet kernel modules 5.20 and later,
178 openiblnd - IbGold 1.8.2,
179 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
180 viblnd - Voltaire ibhost 3.4.5 and later,
181 ciblnd - Topspin 3.2.0,
182 iiblnd - Infiniserv 3.3 + PathBits patch,
183 gmlnd - GM 2.1.22 and later,
184 mxlnd - MX 1.2.1 or later,
185 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
188 Description: workaround for OOM from o2iblnd
189 Details : OFED needs allocate big chunk of memory for QP while creating
190 connection for o2iblnd, OOM can happen if no such a contiguous
192 QP size is decided by concurrent_sends and max_fragments of
193 o2iblnd, now we permit user to specify smaller value for
194 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
195 will decrease memory block size required by creating QP.
199 Description: Support Zerocopy receive of Chelsio device
200 Details : Chelsio driver can support zerocopy for iov[1] if it's
201 contiguous and large enough.
204 Description: fix credit flow deadlock in uptllnd
208 Description: finalize network operation in reasonable time
209 Details : conf-sanity test_32a couldn't stop ost and mds because it
210 tried to access non-existent peer and tcp connect took
211 quite long before timing out.
215 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
216 Details : Lost reference on conn prevents peer from being destroyed, which
217 could prevent new peer creation if peer count has reached upper
222 Description: LNET Selftest results in Soft lockup on OSS CPU
223 Details : only hits when 8 or more o2ib clients involved and a session is
224 torn down with 'lst end_session' without preceeding 'lst stop'.
228 Description: concurrent_sends in IB LNDs should not be changeable at run time
229 Details : concurrent_sends in IB LNDs should not be changeable at run time
231 -------------------------------------------------------------------------------
233 11-03-2008 Sun Microsystems, Inc.
235 * Support for networks:
236 socklnd - any kernel supported by Lustre,
237 qswlnd - Qsnet kernel modules 5.20 and later,
238 openiblnd - IbGold 1.8.2,
239 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
240 viblnd - Voltaire ibhost 3.4.5 and later,
241 ciblnd - Topspin 3.2.0,
242 iiblnd - Infiniserv 3.3 + PathBits patch,
243 gmlnd - GM 2.1.22 and later,
244 mxlnd - MX 1.2.1 or later,
245 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
249 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
250 Details : only hits under out-of-memory situations
252 -------------------------------------------------------------------------------
254 04-26-2008 Sun Microsystems, Inc.
256 * Support for networks:
257 socklnd - any kernel supported by Lustre,
258 qswlnd - Qsnet kernel modules 5.20 and later,
259 openiblnd - IbGold 1.8.2,
260 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
261 viblnd - Voltaire ibhost 3.4.5 and later,
262 ciblnd - Topspin 3.2.0,
263 iiblnd - Infiniserv 3.3 + PathBits patch,
264 gmlnd - GM 2.1.22 and later,
265 mxlnd - MX 1.2.1 or later,
266 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
270 Description: excessive debug information removed
271 Details : excessive debug information removed
275 Description: ksocknal_create_conn() hit ASSERTION during connection race
276 Details : ksocknal_create_conn() hit ASSERTION during connection race
280 Description: ksocknal_send_hello() hit ASSERTION while connecting race
281 Details : ksocknal_send_hello() hit ASSERTION while connecting race
285 Description: o2iblnd/ptllnd credit deadlock in a routed config.
286 Details : o2iblnd/ptllnd credit deadlock in a routed config.
290 Description: High load after starting lnet
291 Details : gmlnd should sleep in rx thread in interruptible way. Otherwise,
292 uptime utility reports high load that looks confusingly.
296 Description: ksocklnd fails to establish connection if accept_port is high
297 Details : PID remapping must not be done for active (outgoing) connections
300 --------------------------------------------------------------------------------
302 2008-01-11 Sun Microsystems, Inc.
304 * Support for networks:
305 socklnd - any kernel supported by Lustre,
306 qswlnd - Qsnet kernel modules 5.20 and later,
307 openiblnd - IbGold 1.8.2,
308 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
309 viblnd - Voltaire ibhost 3.4.5 and later,
310 ciblnd - Topspin 3.2.0,
311 iiblnd - Infiniserv 3.3 + PathBits patch,
312 gmlnd - GM 2.1.22 and later,
313 mxlnd - MX 1.2.1 or later,
314 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
318 Description: liblustre network error
319 Details : liblustre clients should understand LNET_ACCEPT_PORT environment
320 variable even if they don't start lnet acceptor.
324 Description: Strange message from lnet (Ignoring prediction from the future)
325 Details : Incorrect calculation of peer's last_alive value in ksocklnd
327 --------------------------------------------------------------------------------
329 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
331 * Support for networks:
332 socklnd - any kernel supported by Lustre,
333 qswlnd - Qsnet kernel modules 5.20 and later,
334 openiblnd - IbGold 1.8.2,
335 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5.
336 viblnd - Voltaire ibhost 3.4.5 and later,
337 ciblnd - Topspin 3.2.0,
338 iiblnd - Infiniserv 3.3 + PathBits patch,
339 gmlnd - GM 2.1.22 and later,
340 mxlnd - MX 1.2.1 or later,
341 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
345 Description: ASSERTION(me == md->md_me) failed in lnet_match_md()
349 Description: increase send queue size for ciblnd/openiblnd
353 Description: new userspace socklnd
354 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
355 with new one - usocklnd.
357 Severity : enhancement
359 Description: Console message flood
360 Details : Make cdls ratelimiting more tunable by adding several tunable in
361 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
362 /proc/sys/lnet/console_backoff.
364 --------------------------------------------------------------------------------
366 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
368 * Support for networks:
369 socklnd - any kernel supported by Lustre,
370 qswlnd - Qsnet kernel modules 5.20 and later,
371 openiblnd - IbGold 1.8.2,
372 o2iblnd - OFED 1.1 and 1.2,
373 viblnd - Voltaire ibhost 3.4.5 and later,
374 ciblnd - Topspin 3.2.0,
375 iiblnd - Infiniserv 3.3 + PathBits patch,
376 gmlnd - GM 2.1.22 and later,
377 mxlnd - MX 1.2.1 or later,
378 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
382 Description: /proc/sys/lnet has non-sysctl entries
383 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
387 Description: TOE Kernel panic by ksocklnd
388 Details : offloaded sockets provide their own implementation of sendpage,
389 can't call tcp_sendpage() directly
393 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
394 Details : races between lnd_shutdown and peer creation prevent
395 lnd_shutdown from finishing.
399 Description: open files rlimit 1024 reached while liblustre testing
400 Details : ulnds/socklnd must close open socket after unsuccessful
405 Description: build error
406 Details : fix typos in gmlnd, ptllnd and viblnd
408 --------------------------------------------------------------------------------
410 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
412 * Support for networks:
413 socklnd - kernels up to 2.6.16,
414 qswlnd - Qsnet kernel modules 5.20 and later,
415 openiblnd - IbGold 1.8.2,
416 o2iblnd - OFED 1.1 and 1.2
417 viblnd - Voltaire ibhost 3.4.5 and later,
418 ciblnd - Topspin 3.2.0,
419 iiblnd - Infiniserv 3.3 + PathBits patch,
420 gmlnd - GM 2.1.22 and later,
421 mxlnd - MX 1.2.1 or later,
422 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
424 --------------------------------------------------------------------------------
426 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
428 * Support for networks:
429 socklnd - kernels up to 2.6.16,
430 qswlnd - Qsnet kernel modules 5.20 and later,
431 openiblnd - IbGold 1.8.2,
433 viblnd - Voltaire ibhost 3.4.5 and later,
434 ciblnd - Topspin 3.2.0,
435 iiblnd - Infiniserv 3.3 + PathBits patch,
436 gmlnd - GM 2.1.22 and later,
437 mxlnd - MX 1.2.1 or later,
438 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
442 Description: Initialize cpumask before use
446 Description: ASSERTION failures when upgrading to the patchless zero-copy
448 Details : This bug affects "rolling upgrades", causing an inconsistent
449 protocol version negotiation and subsequent assertion failure
450 during rolling upgrades after the first wave of upgrades.
454 Details : Change "dropped message" CERRORs to D_NETERROR so they are
455 logged instead of creating "console chatter" when a lustre
456 timeout races with normal RPC completion.
459 Details : lnet_clear_peer_table can wait forever if user forgets to
463 Details : libcfs_id2str should check pid against LNET_PID_ANY.
467 Description: added LNET self test
468 Details : landing b_self_test
473 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
475 Details : do_div() macro is used incorrectly.
477 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
481 Description: make panic on lbug configurable
485 Description: Add OFED1.2 support to o2iblnd
486 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
487 are installed (other than kernel's in-tree infiniband), there
488 could be some problem while insmod o2iblnd (mismatch CRC of
490 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
491 this link provides solution:
492 https://bugs.openfabrics.org/show_bug.cgi?id=355
493 if extra Module.symvers is not supported in kernel, we will
494 have to run the script in bug 12316 to update
495 $LINUX/module.symvers before building o2iblnd.
496 More details about this are in bug 12316.
498 ------------------------------------------------------------------------------
500 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
501 * version 1.4.10 / 1.6.0
502 * Support for networks:
503 socklnd - kernels up to 2.6.16,
504 qswlnd - Qsnet kernel modules 5.20 and later,
505 openiblnd - IbGold 1.8.2,
507 viblnd - Voltaire ibhost 3.4.5 and later,
508 ciblnd - Topspin 3.2.0,
509 iiblnd - Infiniserv 3.3 + PathBits patch,
510 gmlnd - GM 2.1.22 and later,
511 mxlnd - MX 1.2.1 or later,
512 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
516 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
517 possibly accessed in kptllnd_shutdown. Ptllnd should init
518 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
522 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
526 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
530 Description: the_lnet.ln_finalizing was not set when the current thread is
531 about to complete messages. It only affects multi-threaded
537 Description: Changed the default kqswlnd ntxmsg=512
542 Description: Assertion failure in kernel ptllnd caused by posting passive
543 bulk buffers before connection establishment complete.
548 Description: A race in kernel ptllnd between deleting a peer and posting
549 new communications for it could hang communications -
550 manifesting as "Unexpectedly long timeout" messages.
555 Description: Kernel ptllnd lock ordering issue could hang a node.
560 Description: node crash on socket teardown race
563 Frequency : 'lctl peer_list' issued on a mx net
565 Description: Enable lctl's peer_list for MXLND
568 Frequency : after Ptllnd timeouts and portals congestion
570 Description: Credit overflows
571 Details : This was a bug in ptllnd connection establishment. The fix
572 implements better peer stamps to disambiguate connection
573 establishment and ensure both peers enter the credit flow
574 state machine consistently.
579 Description: kptllnd didn't propagate some network errors up to LNET
580 Details : This bug was spotted while investigating 11394. The fix
581 ensures network errors on sends and bulk transfers are
582 propagated to LNET/lustre correctly.
584 Severity : enhancement
586 Description: Fixed console chatter in case of -ETIMEDOUT.
588 Severity : enhancement
590 Description: Added D_NETTRACE for recording network packet history
591 (initially only for ptllnd). Also a separate userspace
592 ptllnd facility to gather history which should really be
593 covered by D_NETTRACE too, if only CDEBUG recorded history in
599 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
600 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
601 callback can occur before a connection has actually been
602 established. This caused an assertion failure previously.
604 Severity : enhancement
606 Description: Multiple instances for o2iblnd
607 Details : Allow multiple instances of o2iblnd to enable networking over
608 multiple HCAs and routing between them.
612 Description: lnet deadlock in router_checker
613 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
614 into BH locks to eliminate potential deadlock caused by
615 ksocknal_data_ready() preempting code holding these locks.
619 Description: Millions of failed socklnd connection attempts cause a very slow FS
620 Details : added a new route flag ksnr_scheduled to distinguish from
621 ksnr_connecting, so that a peer connection request is only turned
622 down for race concerns when an active connection to the same peer
623 is under progress (instead of just being scheduled).
625 ------------------------------------------------------------------------------
627 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
629 * Support for networks:
630 socklnd - kernels up to 2.6.16
631 qswlnd - Qsnet kernel modules 5.20 and later
632 openiblnd - IbGold 1.8.2
634 viblnd - Voltaire ibhost 3.4.5 and later
635 ciblnd - Topspin 3.2.0
636 iiblnd - Infiniserv 3.3 + PathBits patch
637 gmlnd - GM 2.1.22 and later
638 mxlnd - MX 1.2.1 or later
639 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
642 Severity : major on XT3
644 Description: libcfs overwrites /proc/sys/portals
645 Details : libcfs created a symlink from /proc/sys/portals to
646 /proc/sys/lnet for backwards compatibility. This is no
647 longer required and makes the Cray portals /proc variables
652 Description: OFED FMR API change
653 Details : This changes parameter usage to reflect a change in
654 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
655 that FMR support is only used in experimental versions of the
656 o2iblnd - this change does not affect standard usage at all.
658 Severity : enhancement
660 Description: new ko2iblnd module parameter: ib_mtu
661 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
662 HCAs. You can avoid this problem by setting the MTU to 1024
663 using this module parameter.
665 Severity : enhancement
666 Bugzilla : 11118/11620
667 Description: ptllnd small request message buffer alignment fix
668 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
669 Round up small message size on sends in case this option
670 is not supported. 11620 was a defect in the initial
671 implementation which effectively asserted all peers had to be
672 running the correct protocol version which was fixed by always
673 NAK-ing such requests and handling any misalignments they
678 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
679 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
680 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
682 Severity : enhancement
684 Description: Patchless ZC(zero copy) socklnd
685 Details : New protocol for socklnd, socklnd can support zero copy without
686 kernel patch, it's compatible with old socklnd. Checksum is
687 moved from tunables to modparams.
691 Description: When ksocknal_del_peer() is called upon a peer whose
692 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
693 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
696 Frequency : when ptlrpc is under heavy use and runs out of request buffer
698 Description: In lnet_match_blocked_msg(), md can be used without holding a
702 Frequency : very rarely
704 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
705 If connd connects a route which has been closed by
706 ksocknal_shutdown(), ksocknal_create_routes() may create new
707 routes which hold references on the peer, causing shutdown
708 process to wait for peer to disappear forever.
710 Severity : enhancement
712 Description: Dump XT3 portals traces on kptllnd timeout
713 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
714 dump Cray portals debug traces to a file. The kptllnd module
715 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
716 is the basename of the dump file.
719 Frequency : infrequent
721 Description: kernel ptllnd fix bug in connection re-establishment
722 Details : Kernel ptllnd could produce protocol errors e.g. illegal
723 matchbits and/or violate the credit flow protocol when trying
724 to re-establish a connection with a peer after an error or
727 Severity : enhancement
729 Description: Allow /proc/sys/lnet/debug to be set symbolically
730 Details : Allow debug and subsystem debug values to be read/set by name
731 in addition to numerically, for ease of use.
734 Frequency : only in configurations with LNET routers
736 Description: routes automatically marked down and recovered
737 Details : In configurations with LNET routers if a router fails routers
738 now actively try to recover routes that are down, unless they
739 are marked down by an administrator.
741 ------------------------------------------------------------------------------
743 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
746 Frequency : very rarely, in configurations with LNET routers and TCP
748 Description: incorrect data written to files on OSTs
749 Details : In certain high-load conditions incorrect data may be written
750 to files on the OST when using TCP networks.
752 ------------------------------------------------------------------------------
754 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
756 - rework CDEBUG messages rate-limiting mechanism b=10375
757 - add per-socket tunables for socklnd if the kernel is patched b=10327
759 ------------------------------------------------------------------------------
761 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
763 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
764 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
765 - quiet LNET startup LNI message for liblustre b=10128
766 - Better console error messages if 'ip2nets' can't match an IP address
767 - Fixed overflow/use-before-set bugs in linux-time.h
768 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
769 - LNET teardown failed an assertion about the route table being empty
770 - Fixed a crash in LNetEQPoll(<invalid handle>)
771 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
772 - improve debug message for liblustre/Catamount nodes (b=10116)
774 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
775 * Configuration change for the XT3
776 The PTLLND is now used to run Lustre over Portals on the XT3.
777 The configure option(s) --with-cray-portals are no longer
778 used. Rather --with-portals=<path-to-portals-includes> is
779 used to enable building on the XT3. In addition to enable
780 XT3 specific features the option --enable-cray-xt3 must be
783 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
784 * Portals has been removed, replaced by LNET.
785 LNET is new networking infrastructure for Lustre, it includes a
786 reorganized network configuration mode (see the user
787 documentation for full details) as well as support for routing
788 between different network fabrics. Lustre Networking Devices
789 (LNDS) for the supported network fabrics have also been created
790 for this new infrastructure.
792 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
797 Frequency : rare (large Voltaire clusters only)
799 Description: the default number of reserved transmit descriptors was too low
800 for some large clusters
801 Details : As a workaround, the number was increased. A proper fix includes
804 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
809 Frequency : occasional (large-scale events, cluster reboot, network failure)
811 Description: too many error messages on console obscure actual problem and
812 can slow down/panic server, or cause recovery to fail repeatedly
813 Details : enable rate-limiting of console error messages, and some messages
814 that were console errors now only go to the kernel log
816 Severity : enhancement
818 Description: add /proc/sys/portals/catastrophe entry which will report if
819 that node has previously LBUGged
821 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
823 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
825 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
827 - handle error return code in kranal_check_fma_rx() (5915,6054)
829 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
831 - update vibnal (Voltaire IB NAL)
832 - update gmnal (Myrinet NAL), gmnalid
834 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
836 * Landed portals:b_port_step as follows...
838 - removed CFS_DECL_SPIN*
839 just use 'spinlock_t' and initialise with spin_lock_init()
841 - removed CFS_DECL_MUTEX*
842 just use 'struct semaphore' and initialise with init_mutex()
844 - removed CFS_DECL_RWSEM*
845 just use 'struct rw_semaphore' and initialise with init_rwsem()
847 - renamed cfs_sleep_chan -> cfs_waitq
848 cfs_sleep_link -> cfs_waitlink
850 - fixed race in linux version of arch-independent socknal
851 (the ENOMEM/EAGAIN decision).
853 - Didn't fix problems in Darwin version of arch-independent socknal
854 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
856 - removed libcfs types from non-socknal header files (only some types
857 in the header files had been changed; the .c files hadn't been