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: multiple-instance support for kptllnd
26 Description: ksocknal_close_conn_locked connection race
27 Details : A race was possible when ksocknal_create_conn calls
28 ksocknal_close_conn_locked for already closed conn.
32 Description: router_proc.c is rewritten to use sysctl-interface for parameters
33 residing in /proc/sys/lnet
35 Severity : enhancement
37 Description: port router pinger to userspace
41 Description: kptllnd HELLO protocol deadlock
42 Details : kptllnd HELLO protocol doesn't run to completion in finite time
46 Description: LNet selftest fixes and enhancements
48 Severity : enhancement
50 Description: allow a test node to be a member of multiple test groups
52 Severity : enhancement
54 Description: MXLND: eliminate hosts file, use arp for peer nic_id resolution
55 Details : an update from the upstream developer Scott Atchley.
57 Severity : enhancement
59 Description: add a new LND optiion to control peer buffer credits on routers
63 Description: Fixing deadlock in usocklnd
64 Details : A deadlock was possible in usocklnd due to race condition while
65 tearing connection down. The problem resulted from erroneous
66 assumption that lnet_finalize() could have been called holding
70 Bugzilla : 13621, 15983
71 Description: Protocol V2 of o2iblnd
72 Details : o2iblnd V2 has several new features:
73 . map-on-demand: map-on-demand is disabled by default, it can
74 be enabled by using modparam "map_on_demand=@value@", @value@
75 should >= 0 and < 256, 0 will disable map-on-demand, any other
76 valid value will enable map-on-demand.
77 Oi2blnd will create FMR or physical MR for RDMA if fragments of
79 Enable map-on-demand will take less memory for new connection,
80 but a little more CPU for RDMA.
81 . iWARP : to support iWARP, please enable map-on-demand, 32 and 64
82 are recommanded value. iWARP will probably fail for value >=128.
83 . OOB NOOP message: to resolve deadlock on router.
84 . tunable peer_credits_hiw: (high water to return credits),
85 default value of peer_credits_hiw equals to (peer_credits -1),
86 user can change it between peer_credits/2 and (peer_credits - 1).
87 Lower value is recommended for high latency network.
88 . tunable message queue size: it always equals to peer_credits,
89 higher value is recommended for high latency network.
90 . It's compatible with earlier version of o2iblnd
94 Description: Fixing 'running out of ports' issue
95 Details : Add a delay before next reconnect attempt in ksocklnd in
96 the case of lost race. Limit the frequency of query-requests
97 in lnet. Improved handling of 'dead peer' notifications in
102 Description: Change ptllnd timeout and watchdog timers
103 Details : Add ptltrace_on_nal_failed and bump ptllnd timeout to match
104 Portals wire timeout.
108 Description: One down Lustre FS hangs ALL mounted Lustre filesystems
109 Details : Shared routing enhancements - peer health detection.
111 Severity : enhancement
113 Description: acceptor.c cleanup
114 Details : Code duplication in acceptor.c for the cases of kernel and
115 user-space removed. User-space libcfs tcpip primitives
116 uniformed to have prototypes similar to kernel ones. Minor
117 cosmetic changes in usocklnd to use cfs_socket_t as
118 representation of socket.
122 Description: IB path MTU mistakenly set to 1st path MTU when ib_mtu is off
123 Details : See comment 46 in bug 11245 for details - it's indeed a bug
124 introduced by the original 11245 fix.
128 Description: uptllnd credit overflow fix
129 Details : kptl_msg_t::ptlm_credits could be overflown by uptllnd since
134 Description: socklnd protocol version 3
135 Details : With current protocol V2, connections on router can be
136 blocked and can't receive any incoming messages when there is no
137 more router buffer, so ZC-ACK can't be handled (LNet message
138 can't be finalized) and will cause deadlock on router.
139 Protocol V3 has a dedicated connection for emergency messages
140 like ZC-ACK to router, messages on this dedicated connection
141 don't need any credit so will never be blocked. Also, V3 can send
142 keepalive ping in specified period for router healthy checking.
144 -------------------------------------------------------------------------------
146 12-31-2008 Sun Microsystems, Inc.
148 * Support for networks:
149 socklnd - any kernel supported by Lustre,
150 qswlnd - Qsnet kernel modules 5.20 and later,
151 openiblnd - IbGold 1.8.2,
152 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
153 viblnd - Voltaire ibhost 3.4.5 and later,
154 ciblnd - Topspin 3.2.0,
155 iiblnd - Infiniserv 3.3 + PathBits patch,
156 gmlnd - GM 2.1.22 and later,
157 mxlnd - MX 1.2.1 or later,
158 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
162 Description: workaround for OOM from o2iblnd
163 Details : OFED needs allocate big chunk of memory for QP while creating
164 connection for o2iblnd, OOM can happen if no such a contiguous
166 QP size is decided by concurrent_sends and max_fragments of
167 o2iblnd, now we permit user to specify smaller value for
168 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
169 will decrease memory block size required by creating QP.
173 Description: Support Zerocopy receive of Chelsio device
174 Details : Chelsio driver can support zerocopy for iov[1] if it's
175 contiguous and large enough.
179 Description: fix credit flow deadlock in uptllnd
183 Description: finalize network operation in reasonable time
184 Details : conf-sanity test_32a couldn't stop ost and mds because it
185 tried to access non-existent peer and tcp connect took
186 quite long before timing out.
190 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
191 Details : Lost reference on conn prevents peer from being destroyed, which
192 could prevent new peer creation if peer count has reached upper
197 Description: LNET Selftest results in Soft lockup on OSS CPU
198 Details : only hits when 8 or more o2ib clients involved and a session is
199 torn down with 'lst end_session' without preceeding 'lst stop'.
203 Description: concurrent_sends in IB LNDs should not be changeable at run time
204 Details : concurrent_sends in IB LNDs should not be changeable at run time
208 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
209 Details : only hits under out-of-memory situations
212 -------------------------------------------------------------------------------
214 2009-02-07 Sun Microsystems, Inc.
216 * Support for networks:
217 socklnd - any kernel supported by Lustre,
218 qswlnd - Qsnet kernel modules 5.20 and later,
219 openiblnd - IbGold 1.8.2,
220 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
221 viblnd - Voltaire ibhost 3.4.5 and later,
222 ciblnd - Topspin 3.2.0,
223 iiblnd - Infiniserv 3.3 + PathBits patch,
224 gmlnd - GM 2.1.22 and later,
225 mxlnd - MX 1.2.1 or later,
226 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
229 Description: workaround for OOM from o2iblnd
230 Details : OFED needs allocate big chunk of memory for QP while creating
231 connection for o2iblnd, OOM can happen if no such a contiguous
233 QP size is decided by concurrent_sends and max_fragments of
234 o2iblnd, now we permit user to specify smaller value for
235 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
236 will decrease memory block size required by creating QP.
240 Description: Support Zerocopy receive of Chelsio device
241 Details : Chelsio driver can support zerocopy for iov[1] if it's
242 contiguous and large enough.
245 Description: fix credit flow deadlock in uptllnd
249 Description: finalize network operation in reasonable time
250 Details : conf-sanity test_32a couldn't stop ost and mds because it
251 tried to access non-existent peer and tcp connect took
252 quite long before timing out.
256 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
257 Details : Lost reference on conn prevents peer from being destroyed, which
258 could prevent new peer creation if peer count has reached upper
263 Description: LNET Selftest results in Soft lockup on OSS CPU
264 Details : only hits when 8 or more o2ib clients involved and a session is
265 torn down with 'lst end_session' without preceeding 'lst stop'.
269 Description: concurrent_sends in IB LNDs should not be changeable at run time
270 Details : concurrent_sends in IB LNDs should not be changeable at run time
272 -------------------------------------------------------------------------------
274 11-03-2008 Sun Microsystems, Inc.
276 * Support for networks:
277 socklnd - any kernel supported by Lustre,
278 qswlnd - Qsnet kernel modules 5.20 and later,
279 openiblnd - IbGold 1.8.2,
280 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
281 viblnd - Voltaire ibhost 3.4.5 and later,
282 ciblnd - Topspin 3.2.0,
283 iiblnd - Infiniserv 3.3 + PathBits patch,
284 gmlnd - GM 2.1.22 and later,
285 mxlnd - MX 1.2.1 or later,
286 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
290 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
291 Details : only hits under out-of-memory situations
293 -------------------------------------------------------------------------------
295 04-26-2008 Sun Microsystems, Inc.
297 * Support for networks:
298 socklnd - any kernel supported by Lustre,
299 qswlnd - Qsnet kernel modules 5.20 and later,
300 openiblnd - IbGold 1.8.2,
301 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
302 viblnd - Voltaire ibhost 3.4.5 and later,
303 ciblnd - Topspin 3.2.0,
304 iiblnd - Infiniserv 3.3 + PathBits patch,
305 gmlnd - GM 2.1.22 and later,
306 mxlnd - MX 1.2.1 or later,
307 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
311 Description: excessive debug information removed
312 Details : excessive debug information removed
316 Description: ksocknal_create_conn() hit ASSERTION during connection race
317 Details : ksocknal_create_conn() hit ASSERTION during connection race
321 Description: ksocknal_send_hello() hit ASSERTION while connecting race
322 Details : ksocknal_send_hello() hit ASSERTION while connecting race
326 Description: o2iblnd/ptllnd credit deadlock in a routed config.
327 Details : o2iblnd/ptllnd credit deadlock in a routed config.
331 Description: High load after starting lnet
332 Details : gmlnd should sleep in rx thread in interruptible way. Otherwise,
333 uptime utility reports high load that looks confusingly.
337 Description: ksocklnd fails to establish connection if accept_port is high
338 Details : PID remapping must not be done for active (outgoing) connections
341 --------------------------------------------------------------------------------
343 2008-01-11 Sun Microsystems, Inc.
345 * Support for networks:
346 socklnd - any kernel supported by Lustre,
347 qswlnd - Qsnet kernel modules 5.20 and later,
348 openiblnd - IbGold 1.8.2,
349 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
350 viblnd - Voltaire ibhost 3.4.5 and later,
351 ciblnd - Topspin 3.2.0,
352 iiblnd - Infiniserv 3.3 + PathBits patch,
353 gmlnd - GM 2.1.22 and later,
354 mxlnd - MX 1.2.1 or later,
355 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
359 Description: liblustre network error
360 Details : liblustre clients should understand LNET_ACCEPT_PORT environment
361 variable even if they don't start lnet acceptor.
365 Description: Strange message from lnet (Ignoring prediction from the future)
366 Details : Incorrect calculation of peer's last_alive value in ksocklnd
368 --------------------------------------------------------------------------------
370 2007-12-07 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 1.2.0, 1.2.5.
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: ASSERTION(me == md->md_me) failed in lnet_match_md()
390 Description: increase send queue size for ciblnd/openiblnd
394 Description: new userspace socklnd
395 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
396 with new one - usocklnd.
398 Severity : enhancement
400 Description: Console message flood
401 Details : Make cdls ratelimiting more tunable by adding several tunable in
402 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
403 /proc/sys/lnet/console_backoff.
405 --------------------------------------------------------------------------------
407 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
409 * Support for networks:
410 socklnd - any kernel supported by Lustre,
411 qswlnd - Qsnet kernel modules 5.20 and later,
412 openiblnd - IbGold 1.8.2,
413 o2iblnd - OFED 1.1 and 1.2,
414 viblnd - Voltaire ibhost 3.4.5 and later,
415 ciblnd - Topspin 3.2.0,
416 iiblnd - Infiniserv 3.3 + PathBits patch,
417 gmlnd - GM 2.1.22 and later,
418 mxlnd - MX 1.2.1 or later,
419 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
423 Description: /proc/sys/lnet has non-sysctl entries
424 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
428 Description: TOE Kernel panic by ksocklnd
429 Details : offloaded sockets provide their own implementation of sendpage,
430 can't call tcp_sendpage() directly
434 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
435 Details : races between lnd_shutdown and peer creation prevent
436 lnd_shutdown from finishing.
440 Description: open files rlimit 1024 reached while liblustre testing
441 Details : ulnds/socklnd must close open socket after unsuccessful
446 Description: build error
447 Details : fix typos in gmlnd, ptllnd and viblnd
449 --------------------------------------------------------------------------------
451 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
453 * Support for networks:
454 socklnd - kernels up to 2.6.16,
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
465 --------------------------------------------------------------------------------
467 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
469 * Support for networks:
470 socklnd - kernels up to 2.6.16,
471 qswlnd - Qsnet kernel modules 5.20 and later,
472 openiblnd - IbGold 1.8.2,
474 viblnd - Voltaire ibhost 3.4.5 and later,
475 ciblnd - Topspin 3.2.0,
476 iiblnd - Infiniserv 3.3 + PathBits patch,
477 gmlnd - GM 2.1.22 and later,
478 mxlnd - MX 1.2.1 or later,
479 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
483 Description: Initialize cpumask before use
487 Description: ASSERTION failures when upgrading to the patchless zero-copy
489 Details : This bug affects "rolling upgrades", causing an inconsistent
490 protocol version negotiation and subsequent assertion failure
491 during rolling upgrades after the first wave of upgrades.
495 Details : Change "dropped message" CERRORs to D_NETERROR so they are
496 logged instead of creating "console chatter" when a lustre
497 timeout races with normal RPC completion.
500 Details : lnet_clear_peer_table can wait forever if user forgets to
504 Details : libcfs_id2str should check pid against LNET_PID_ANY.
508 Description: added LNET self test
509 Details : landing b_self_test
514 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
516 Details : do_div() macro is used incorrectly.
518 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
522 Description: make panic on lbug configurable
526 Description: Add OFED1.2 support to o2iblnd
527 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
528 are installed (other than kernel's in-tree infiniband), there
529 could be some problem while insmod o2iblnd (mismatch CRC of
531 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
532 this link provides solution:
533 https://bugs.openfabrics.org/show_bug.cgi?id=355
534 if extra Module.symvers is not supported in kernel, we will
535 have to run the script in bug 12316 to update
536 $LINUX/module.symvers before building o2iblnd.
537 More details about this are in bug 12316.
539 ------------------------------------------------------------------------------
541 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
542 * version 1.4.10 / 1.6.0
543 * Support for networks:
544 socklnd - kernels up to 2.6.16,
545 qswlnd - Qsnet kernel modules 5.20 and later,
546 openiblnd - IbGold 1.8.2,
548 viblnd - Voltaire ibhost 3.4.5 and later,
549 ciblnd - Topspin 3.2.0,
550 iiblnd - Infiniserv 3.3 + PathBits patch,
551 gmlnd - GM 2.1.22 and later,
552 mxlnd - MX 1.2.1 or later,
553 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
557 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
558 possibly accessed in kptllnd_shutdown. Ptllnd should init
559 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
563 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
567 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
571 Description: the_lnet.ln_finalizing was not set when the current thread is
572 about to complete messages. It only affects multi-threaded
578 Description: Changed the default kqswlnd ntxmsg=512
583 Description: Assertion failure in kernel ptllnd caused by posting passive
584 bulk buffers before connection establishment complete.
589 Description: A race in kernel ptllnd between deleting a peer and posting
590 new communications for it could hang communications -
591 manifesting as "Unexpectedly long timeout" messages.
596 Description: Kernel ptllnd lock ordering issue could hang a node.
601 Description: node crash on socket teardown race
604 Frequency : 'lctl peer_list' issued on a mx net
606 Description: Enable lctl's peer_list for MXLND
609 Frequency : after Ptllnd timeouts and portals congestion
611 Description: Credit overflows
612 Details : This was a bug in ptllnd connection establishment. The fix
613 implements better peer stamps to disambiguate connection
614 establishment and ensure both peers enter the credit flow
615 state machine consistently.
620 Description: kptllnd didn't propagate some network errors up to LNET
621 Details : This bug was spotted while investigating 11394. The fix
622 ensures network errors on sends and bulk transfers are
623 propagated to LNET/lustre correctly.
625 Severity : enhancement
627 Description: Fixed console chatter in case of -ETIMEDOUT.
629 Severity : enhancement
631 Description: Added D_NETTRACE for recording network packet history
632 (initially only for ptllnd). Also a separate userspace
633 ptllnd facility to gather history which should really be
634 covered by D_NETTRACE too, if only CDEBUG recorded history in
640 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
641 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
642 callback can occur before a connection has actually been
643 established. This caused an assertion failure previously.
645 Severity : enhancement
647 Description: Multiple instances for o2iblnd
648 Details : Allow multiple instances of o2iblnd to enable networking over
649 multiple HCAs and routing between them.
653 Description: lnet deadlock in router_checker
654 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
655 into BH locks to eliminate potential deadlock caused by
656 ksocknal_data_ready() preempting code holding these locks.
660 Description: Millions of failed socklnd connection attempts cause a very slow FS
661 Details : added a new route flag ksnr_scheduled to distinguish from
662 ksnr_connecting, so that a peer connection request is only turned
663 down for race concerns when an active connection to the same peer
664 is under progress (instead of just being scheduled).
666 ------------------------------------------------------------------------------
668 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
670 * Support for networks:
671 socklnd - kernels up to 2.6.16
672 qswlnd - Qsnet kernel modules 5.20 and later
673 openiblnd - IbGold 1.8.2
675 viblnd - Voltaire ibhost 3.4.5 and later
676 ciblnd - Topspin 3.2.0
677 iiblnd - Infiniserv 3.3 + PathBits patch
678 gmlnd - GM 2.1.22 and later
679 mxlnd - MX 1.2.1 or later
680 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
683 Severity : major on XT3
685 Description: libcfs overwrites /proc/sys/portals
686 Details : libcfs created a symlink from /proc/sys/portals to
687 /proc/sys/lnet for backwards compatibility. This is no
688 longer required and makes the Cray portals /proc variables
693 Description: OFED FMR API change
694 Details : This changes parameter usage to reflect a change in
695 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
696 that FMR support is only used in experimental versions of the
697 o2iblnd - this change does not affect standard usage at all.
699 Severity : enhancement
701 Description: new ko2iblnd module parameter: ib_mtu
702 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
703 HCAs. You can avoid this problem by setting the MTU to 1024
704 using this module parameter.
706 Severity : enhancement
707 Bugzilla : 11118/11620
708 Description: ptllnd small request message buffer alignment fix
709 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
710 Round up small message size on sends in case this option
711 is not supported. 11620 was a defect in the initial
712 implementation which effectively asserted all peers had to be
713 running the correct protocol version which was fixed by always
714 NAK-ing such requests and handling any misalignments they
719 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
720 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
721 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
723 Severity : enhancement
725 Description: Patchless ZC(zero copy) socklnd
726 Details : New protocol for socklnd, socklnd can support zero copy without
727 kernel patch, it's compatible with old socklnd. Checksum is
728 moved from tunables to modparams.
732 Description: When ksocknal_del_peer() is called upon a peer whose
733 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
734 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
737 Frequency : when ptlrpc is under heavy use and runs out of request buffer
739 Description: In lnet_match_blocked_msg(), md can be used without holding a
743 Frequency : very rarely
745 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
746 If connd connects a route which has been closed by
747 ksocknal_shutdown(), ksocknal_create_routes() may create new
748 routes which hold references on the peer, causing shutdown
749 process to wait for peer to disappear forever.
751 Severity : enhancement
753 Description: Dump XT3 portals traces on kptllnd timeout
754 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
755 dump Cray portals debug traces to a file. The kptllnd module
756 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
757 is the basename of the dump file.
760 Frequency : infrequent
762 Description: kernel ptllnd fix bug in connection re-establishment
763 Details : Kernel ptllnd could produce protocol errors e.g. illegal
764 matchbits and/or violate the credit flow protocol when trying
765 to re-establish a connection with a peer after an error or
768 Severity : enhancement
770 Description: Allow /proc/sys/lnet/debug to be set symbolically
771 Details : Allow debug and subsystem debug values to be read/set by name
772 in addition to numerically, for ease of use.
775 Frequency : only in configurations with LNET routers
777 Description: routes automatically marked down and recovered
778 Details : In configurations with LNET routers if a router fails routers
779 now actively try to recover routes that are down, unless they
780 are marked down by an administrator.
782 ------------------------------------------------------------------------------
784 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
787 Frequency : very rarely, in configurations with LNET routers and TCP
789 Description: incorrect data written to files on OSTs
790 Details : In certain high-load conditions incorrect data may be written
791 to files on the OST when using TCP networks.
793 ------------------------------------------------------------------------------
795 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
797 - rework CDEBUG messages rate-limiting mechanism b=10375
798 - add per-socket tunables for socklnd if the kernel is patched b=10327
800 ------------------------------------------------------------------------------
802 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
804 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
805 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
806 - quiet LNET startup LNI message for liblustre b=10128
807 - Better console error messages if 'ip2nets' can't match an IP address
808 - Fixed overflow/use-before-set bugs in linux-time.h
809 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
810 - LNET teardown failed an assertion about the route table being empty
811 - Fixed a crash in LNetEQPoll(<invalid handle>)
812 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
813 - improve debug message for liblustre/Catamount nodes (b=10116)
815 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
816 * Configuration change for the XT3
817 The PTLLND is now used to run Lustre over Portals on the XT3.
818 The configure option(s) --with-cray-portals are no longer
819 used. Rather --with-portals=<path-to-portals-includes> is
820 used to enable building on the XT3. In addition to enable
821 XT3 specific features the option --enable-cray-xt3 must be
824 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
825 * Portals has been removed, replaced by LNET.
826 LNET is new networking infrastructure for Lustre, it includes a
827 reorganized network configuration mode (see the user
828 documentation for full details) as well as support for routing
829 between different network fabrics. Lustre Networking Devices
830 (LNDS) for the supported network fabrics have also been created
831 for this new infrastructure.
833 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
838 Frequency : rare (large Voltaire clusters only)
840 Description: the default number of reserved transmit descriptors was too low
841 for some large clusters
842 Details : As a workaround, the number was increased. A proper fix includes
845 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
850 Frequency : occasional (large-scale events, cluster reboot, network failure)
852 Description: too many error messages on console obscure actual problem and
853 can slow down/panic server, or cause recovery to fail repeatedly
854 Details : enable rate-limiting of console error messages, and some messages
855 that were console errors now only go to the kernel log
857 Severity : enhancement
859 Description: add /proc/sys/portals/catastrophe entry which will report if
860 that node has previously LBUGged
862 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
864 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
866 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
868 - handle error return code in kranal_check_fma_rx() (5915,6054)
870 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
872 - update vibnal (Voltaire IB NAL)
873 - update gmnal (Myrinet NAL), gmnalid
875 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
877 * Landed portals:b_port_step as follows...
879 - removed CFS_DECL_SPIN*
880 just use 'spinlock_t' and initialise with spin_lock_init()
882 - removed CFS_DECL_MUTEX*
883 just use 'struct semaphore' and initialise with init_mutex()
885 - removed CFS_DECL_RWSEM*
886 just use 'struct rw_semaphore' and initialise with init_rwsem()
888 - renamed cfs_sleep_chan -> cfs_waitq
889 cfs_sleep_link -> cfs_waitlink
891 - fixed race in linux version of arch-independent socknal
892 (the ENOMEM/EAGAIN decision).
894 - Didn't fix problems in Darwin version of arch-independent socknal
895 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
897 - removed libcfs types from non-socknal header files (only some types
898 in the header files had been changed; the .c files hadn't been