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: LNet router shuffler.
24 Severity : enhancement
26 Description: LNet fine grain routing support.
30 Description: router checker stops working when system wall clock goes backward
31 Details : use monotonic timing source instead of system wall clock time.
33 Severity : enhancement
35 Description: avoid asymmetrical router failures
37 Severity : enhancement
39 Description: multiple-instance support for kptllnd
43 Description: ksocknal_close_conn_locked connection race
44 Details : A race was possible when ksocknal_create_conn calls
45 ksocknal_close_conn_locked for already closed conn.
49 Description: router_proc.c is rewritten to use sysctl-interface for parameters
50 residing in /proc/sys/lnet
52 Severity : enhancement
54 Description: port router pinger to userspace
58 Description: kptllnd HELLO protocol deadlock
59 Details : kptllnd HELLO protocol doesn't run to completion in finite time
63 Description: LNet selftest fixes and enhancements
65 Severity : enhancement
67 Description: allow a test node to be a member of multiple test groups
69 Severity : enhancement
71 Description: MXLND: eliminate hosts file, use arp for peer nic_id resolution
72 Details : an update from the upstream developer Scott Atchley.
74 Severity : enhancement
76 Description: add a new LND optiion to control peer buffer credits on routers
80 Description: Fixing deadlock in usocklnd
81 Details : A deadlock was possible in usocklnd due to race condition while
82 tearing connection down. The problem resulted from erroneous
83 assumption that lnet_finalize() could have been called holding
87 Bugzilla : 13621, 15983
88 Description: Protocol V2 of o2iblnd
89 Details : o2iblnd V2 has several new features:
90 . map-on-demand: map-on-demand is disabled by default, it can
91 be enabled by using modparam "map_on_demand=@value@", @value@
92 should >= 0 and < 256, 0 will disable map-on-demand, any other
93 valid value will enable map-on-demand.
94 Oi2blnd will create FMR or physical MR for RDMA if fragments of
96 Enable map-on-demand will take less memory for new connection,
97 but a little more CPU for RDMA.
98 . iWARP : to support iWARP, please enable map-on-demand, 32 and 64
99 are recommanded value. iWARP will probably fail for value >=128.
100 . OOB NOOP message: to resolve deadlock on router.
101 . tunable peer_credits_hiw: (high water to return credits),
102 default value of peer_credits_hiw equals to (peer_credits -1),
103 user can change it between peer_credits/2 and (peer_credits - 1).
104 Lower value is recommended for high latency network.
105 . tunable message queue size: it always equals to peer_credits,
106 higher value is recommended for high latency network.
107 . It's compatible with earlier version of o2iblnd
111 Description: Fixing 'running out of ports' issue
112 Details : Add a delay before next reconnect attempt in ksocklnd in
113 the case of lost race. Limit the frequency of query-requests
114 in lnet. Improved handling of 'dead peer' notifications in
119 Description: Change ptllnd timeout and watchdog timers
120 Details : Add ptltrace_on_nal_failed and bump ptllnd timeout to match
121 Portals wire timeout.
125 Description: One down Lustre FS hangs ALL mounted Lustre filesystems
126 Details : Shared routing enhancements - peer health detection.
128 Severity : enhancement
130 Description: acceptor.c cleanup
131 Details : Code duplication in acceptor.c for the cases of kernel and
132 user-space removed. User-space libcfs tcpip primitives
133 uniformed to have prototypes similar to kernel ones. Minor
134 cosmetic changes in usocklnd to use cfs_socket_t as
135 representation of socket.
139 Description: IB path MTU mistakenly set to 1st path MTU when ib_mtu is off
140 Details : See comment 46 in bug 11245 for details - it's indeed a bug
141 introduced by the original 11245 fix.
145 Description: uptllnd credit overflow fix
146 Details : kptl_msg_t::ptlm_credits could be overflown by uptllnd since
151 Description: socklnd protocol version 3
152 Details : With current protocol V2, connections on router can be
153 blocked and can't receive any incoming messages when there is no
154 more router buffer, so ZC-ACK can't be handled (LNet message
155 can't be finalized) and will cause deadlock on router.
156 Protocol V3 has a dedicated connection for emergency messages
157 like ZC-ACK to router, messages on this dedicated connection
158 don't need any credit so will never be blocked. Also, V3 can send
159 keepalive ping in specified period for router healthy checking.
161 -------------------------------------------------------------------------------
163 12-31-2008 Sun Microsystems, Inc.
165 * Support for networks:
166 socklnd - any kernel supported by Lustre,
167 qswlnd - Qsnet kernel modules 5.20 and later,
168 openiblnd - IbGold 1.8.2,
169 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
170 viblnd - Voltaire ibhost 3.4.5 and later,
171 ciblnd - Topspin 3.2.0,
172 iiblnd - Infiniserv 3.3 + PathBits patch,
173 gmlnd - GM 2.1.22 and later,
174 mxlnd - MX 1.2.1 or later,
175 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
179 Description: workaround for OOM from o2iblnd
180 Details : OFED needs allocate big chunk of memory for QP while creating
181 connection for o2iblnd, OOM can happen if no such a contiguous
183 QP size is decided by concurrent_sends and max_fragments of
184 o2iblnd, now we permit user to specify smaller value for
185 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
186 will decrease memory block size required by creating QP.
190 Description: Support Zerocopy receive of Chelsio device
191 Details : Chelsio driver can support zerocopy for iov[1] if it's
192 contiguous and large enough.
196 Description: fix credit flow deadlock in uptllnd
200 Description: finalize network operation in reasonable time
201 Details : conf-sanity test_32a couldn't stop ost and mds because it
202 tried to access non-existent peer and tcp connect took
203 quite long before timing out.
207 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
208 Details : Lost reference on conn prevents peer from being destroyed, which
209 could prevent new peer creation if peer count has reached upper
214 Description: LNET Selftest results in Soft lockup on OSS CPU
215 Details : only hits when 8 or more o2ib clients involved and a session is
216 torn down with 'lst end_session' without preceeding 'lst stop'.
220 Description: concurrent_sends in IB LNDs should not be changeable at run time
221 Details : concurrent_sends in IB LNDs should not be changeable at run time
225 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
226 Details : only hits under out-of-memory situations
229 -------------------------------------------------------------------------------
231 2009-02-07 Sun Microsystems, Inc.
233 * Support for networks:
234 socklnd - any kernel supported by Lustre,
235 qswlnd - Qsnet kernel modules 5.20 and later,
236 openiblnd - IbGold 1.8.2,
237 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
238 viblnd - Voltaire ibhost 3.4.5 and later,
239 ciblnd - Topspin 3.2.0,
240 iiblnd - Infiniserv 3.3 + PathBits patch,
241 gmlnd - GM 2.1.22 and later,
242 mxlnd - MX 1.2.1 or later,
243 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
246 Description: workaround for OOM from o2iblnd
247 Details : OFED needs allocate big chunk of memory for QP while creating
248 connection for o2iblnd, OOM can happen if no such a contiguous
250 QP size is decided by concurrent_sends and max_fragments of
251 o2iblnd, now we permit user to specify smaller value for
252 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
253 will decrease memory block size required by creating QP.
257 Description: Support Zerocopy receive of Chelsio device
258 Details : Chelsio driver can support zerocopy for iov[1] if it's
259 contiguous and large enough.
262 Description: fix credit flow deadlock in uptllnd
266 Description: finalize network operation in reasonable time
267 Details : conf-sanity test_32a couldn't stop ost and mds because it
268 tried to access non-existent peer and tcp connect took
269 quite long before timing out.
273 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
274 Details : Lost reference on conn prevents peer from being destroyed, which
275 could prevent new peer creation if peer count has reached upper
280 Description: LNET Selftest results in Soft lockup on OSS CPU
281 Details : only hits when 8 or more o2ib clients involved and a session is
282 torn down with 'lst end_session' without preceeding 'lst stop'.
286 Description: concurrent_sends in IB LNDs should not be changeable at run time
287 Details : concurrent_sends in IB LNDs should not be changeable at run time
289 -------------------------------------------------------------------------------
291 11-03-2008 Sun Microsystems, Inc.
293 * Support for networks:
294 socklnd - any kernel supported by Lustre,
295 qswlnd - Qsnet kernel modules 5.20 and later,
296 openiblnd - IbGold 1.8.2,
297 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
298 viblnd - Voltaire ibhost 3.4.5 and later,
299 ciblnd - Topspin 3.2.0,
300 iiblnd - Infiniserv 3.3 + PathBits patch,
301 gmlnd - GM 2.1.22 and later,
302 mxlnd - MX 1.2.1 or later,
303 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
307 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
308 Details : only hits under out-of-memory situations
310 -------------------------------------------------------------------------------
312 04-26-2008 Sun Microsystems, Inc.
314 * Support for networks:
315 socklnd - any kernel supported by Lustre,
316 qswlnd - Qsnet kernel modules 5.20 and later,
317 openiblnd - IbGold 1.8.2,
318 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
319 viblnd - Voltaire ibhost 3.4.5 and later,
320 ciblnd - Topspin 3.2.0,
321 iiblnd - Infiniserv 3.3 + PathBits patch,
322 gmlnd - GM 2.1.22 and later,
323 mxlnd - MX 1.2.1 or later,
324 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
328 Description: excessive debug information removed
329 Details : excessive debug information removed
333 Description: ksocknal_create_conn() hit ASSERTION during connection race
334 Details : ksocknal_create_conn() hit ASSERTION during connection race
338 Description: ksocknal_send_hello() hit ASSERTION while connecting race
339 Details : ksocknal_send_hello() hit ASSERTION while connecting race
343 Description: o2iblnd/ptllnd credit deadlock in a routed config.
344 Details : o2iblnd/ptllnd credit deadlock in a routed config.
348 Description: High load after starting lnet
349 Details : gmlnd should sleep in rx thread in interruptible way. Otherwise,
350 uptime utility reports high load that looks confusingly.
354 Description: ksocklnd fails to establish connection if accept_port is high
355 Details : PID remapping must not be done for active (outgoing) connections
358 --------------------------------------------------------------------------------
360 2008-01-11 Sun Microsystems, Inc.
362 * Support for networks:
363 socklnd - any kernel supported by Lustre,
364 qswlnd - Qsnet kernel modules 5.20 and later,
365 openiblnd - IbGold 1.8.2,
366 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
367 viblnd - Voltaire ibhost 3.4.5 and later,
368 ciblnd - Topspin 3.2.0,
369 iiblnd - Infiniserv 3.3 + PathBits patch,
370 gmlnd - GM 2.1.22 and later,
371 mxlnd - MX 1.2.1 or later,
372 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
376 Description: liblustre network error
377 Details : liblustre clients should understand LNET_ACCEPT_PORT environment
378 variable even if they don't start lnet acceptor.
382 Description: Strange message from lnet (Ignoring prediction from the future)
383 Details : Incorrect calculation of peer's last_alive value in ksocklnd
385 --------------------------------------------------------------------------------
387 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
389 * Support for networks:
390 socklnd - any kernel supported by Lustre,
391 qswlnd - Qsnet kernel modules 5.20 and later,
392 openiblnd - IbGold 1.8.2,
393 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5.
394 viblnd - Voltaire ibhost 3.4.5 and later,
395 ciblnd - Topspin 3.2.0,
396 iiblnd - Infiniserv 3.3 + PathBits patch,
397 gmlnd - GM 2.1.22 and later,
398 mxlnd - MX 1.2.1 or later,
399 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
403 Description: ASSERTION(me == md->md_me) failed in lnet_match_md()
407 Description: increase send queue size for ciblnd/openiblnd
411 Description: new userspace socklnd
412 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
413 with new one - usocklnd.
415 Severity : enhancement
417 Description: Console message flood
418 Details : Make cdls ratelimiting more tunable by adding several tunable in
419 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
420 /proc/sys/lnet/console_backoff.
422 --------------------------------------------------------------------------------
424 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
426 * Support for networks:
427 socklnd - any kernel supported by Lustre,
428 qswlnd - Qsnet kernel modules 5.20 and later,
429 openiblnd - IbGold 1.8.2,
430 o2iblnd - OFED 1.1 and 1.2,
431 viblnd - Voltaire ibhost 3.4.5 and later,
432 ciblnd - Topspin 3.2.0,
433 iiblnd - Infiniserv 3.3 + PathBits patch,
434 gmlnd - GM 2.1.22 and later,
435 mxlnd - MX 1.2.1 or later,
436 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
440 Description: /proc/sys/lnet has non-sysctl entries
441 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
445 Description: TOE Kernel panic by ksocklnd
446 Details : offloaded sockets provide their own implementation of sendpage,
447 can't call tcp_sendpage() directly
451 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
452 Details : races between lnd_shutdown and peer creation prevent
453 lnd_shutdown from finishing.
457 Description: open files rlimit 1024 reached while liblustre testing
458 Details : ulnds/socklnd must close open socket after unsuccessful
463 Description: build error
464 Details : fix typos in gmlnd, ptllnd and viblnd
466 --------------------------------------------------------------------------------
468 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
470 * Support for networks:
471 socklnd - kernels up to 2.6.16,
472 qswlnd - Qsnet kernel modules 5.20 and later,
473 openiblnd - IbGold 1.8.2,
474 o2iblnd - OFED 1.1 and 1.2
475 viblnd - Voltaire ibhost 3.4.5 and later,
476 ciblnd - Topspin 3.2.0,
477 iiblnd - Infiniserv 3.3 + PathBits patch,
478 gmlnd - GM 2.1.22 and later,
479 mxlnd - MX 1.2.1 or later,
480 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
482 --------------------------------------------------------------------------------
484 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
486 * Support for networks:
487 socklnd - kernels up to 2.6.16,
488 qswlnd - Qsnet kernel modules 5.20 and later,
489 openiblnd - IbGold 1.8.2,
491 viblnd - Voltaire ibhost 3.4.5 and later,
492 ciblnd - Topspin 3.2.0,
493 iiblnd - Infiniserv 3.3 + PathBits patch,
494 gmlnd - GM 2.1.22 and later,
495 mxlnd - MX 1.2.1 or later,
496 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
500 Description: Initialize cpumask before use
504 Description: ASSERTION failures when upgrading to the patchless zero-copy
506 Details : This bug affects "rolling upgrades", causing an inconsistent
507 protocol version negotiation and subsequent assertion failure
508 during rolling upgrades after the first wave of upgrades.
512 Details : Change "dropped message" CERRORs to D_NETERROR so they are
513 logged instead of creating "console chatter" when a lustre
514 timeout races with normal RPC completion.
517 Details : lnet_clear_peer_table can wait forever if user forgets to
521 Details : libcfs_id2str should check pid against LNET_PID_ANY.
525 Description: added LNET self test
526 Details : landing b_self_test
531 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
533 Details : do_div() macro is used incorrectly.
535 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
539 Description: make panic on lbug configurable
543 Description: Add OFED1.2 support to o2iblnd
544 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
545 are installed (other than kernel's in-tree infiniband), there
546 could be some problem while insmod o2iblnd (mismatch CRC of
548 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
549 this link provides solution:
550 https://bugs.openfabrics.org/show_bug.cgi?id=355
551 if extra Module.symvers is not supported in kernel, we will
552 have to run the script in bug 12316 to update
553 $LINUX/module.symvers before building o2iblnd.
554 More details about this are in bug 12316.
556 ------------------------------------------------------------------------------
558 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
559 * version 1.4.10 / 1.6.0
560 * Support for networks:
561 socklnd - kernels up to 2.6.16,
562 qswlnd - Qsnet kernel modules 5.20 and later,
563 openiblnd - IbGold 1.8.2,
565 viblnd - Voltaire ibhost 3.4.5 and later,
566 ciblnd - Topspin 3.2.0,
567 iiblnd - Infiniserv 3.3 + PathBits patch,
568 gmlnd - GM 2.1.22 and later,
569 mxlnd - MX 1.2.1 or later,
570 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
574 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
575 possibly accessed in kptllnd_shutdown. Ptllnd should init
576 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
580 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
584 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
588 Description: the_lnet.ln_finalizing was not set when the current thread is
589 about to complete messages. It only affects multi-threaded
595 Description: Changed the default kqswlnd ntxmsg=512
600 Description: Assertion failure in kernel ptllnd caused by posting passive
601 bulk buffers before connection establishment complete.
606 Description: A race in kernel ptllnd between deleting a peer and posting
607 new communications for it could hang communications -
608 manifesting as "Unexpectedly long timeout" messages.
613 Description: Kernel ptllnd lock ordering issue could hang a node.
618 Description: node crash on socket teardown race
621 Frequency : 'lctl peer_list' issued on a mx net
623 Description: Enable lctl's peer_list for MXLND
626 Frequency : after Ptllnd timeouts and portals congestion
628 Description: Credit overflows
629 Details : This was a bug in ptllnd connection establishment. The fix
630 implements better peer stamps to disambiguate connection
631 establishment and ensure both peers enter the credit flow
632 state machine consistently.
637 Description: kptllnd didn't propagate some network errors up to LNET
638 Details : This bug was spotted while investigating 11394. The fix
639 ensures network errors on sends and bulk transfers are
640 propagated to LNET/lustre correctly.
642 Severity : enhancement
644 Description: Fixed console chatter in case of -ETIMEDOUT.
646 Severity : enhancement
648 Description: Added D_NETTRACE for recording network packet history
649 (initially only for ptllnd). Also a separate userspace
650 ptllnd facility to gather history which should really be
651 covered by D_NETTRACE too, if only CDEBUG recorded history in
657 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
658 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
659 callback can occur before a connection has actually been
660 established. This caused an assertion failure previously.
662 Severity : enhancement
664 Description: Multiple instances for o2iblnd
665 Details : Allow multiple instances of o2iblnd to enable networking over
666 multiple HCAs and routing between them.
670 Description: lnet deadlock in router_checker
671 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
672 into BH locks to eliminate potential deadlock caused by
673 ksocknal_data_ready() preempting code holding these locks.
677 Description: Millions of failed socklnd connection attempts cause a very slow FS
678 Details : added a new route flag ksnr_scheduled to distinguish from
679 ksnr_connecting, so that a peer connection request is only turned
680 down for race concerns when an active connection to the same peer
681 is under progress (instead of just being scheduled).
683 ------------------------------------------------------------------------------
685 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
687 * Support for networks:
688 socklnd - kernels up to 2.6.16
689 qswlnd - Qsnet kernel modules 5.20 and later
690 openiblnd - IbGold 1.8.2
692 viblnd - Voltaire ibhost 3.4.5 and later
693 ciblnd - Topspin 3.2.0
694 iiblnd - Infiniserv 3.3 + PathBits patch
695 gmlnd - GM 2.1.22 and later
696 mxlnd - MX 1.2.1 or later
697 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
700 Severity : major on XT3
702 Description: libcfs overwrites /proc/sys/portals
703 Details : libcfs created a symlink from /proc/sys/portals to
704 /proc/sys/lnet for backwards compatibility. This is no
705 longer required and makes the Cray portals /proc variables
710 Description: OFED FMR API change
711 Details : This changes parameter usage to reflect a change in
712 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
713 that FMR support is only used in experimental versions of the
714 o2iblnd - this change does not affect standard usage at all.
716 Severity : enhancement
718 Description: new ko2iblnd module parameter: ib_mtu
719 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
720 HCAs. You can avoid this problem by setting the MTU to 1024
721 using this module parameter.
723 Severity : enhancement
724 Bugzilla : 11118/11620
725 Description: ptllnd small request message buffer alignment fix
726 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
727 Round up small message size on sends in case this option
728 is not supported. 11620 was a defect in the initial
729 implementation which effectively asserted all peers had to be
730 running the correct protocol version which was fixed by always
731 NAK-ing such requests and handling any misalignments they
736 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
737 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
738 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
740 Severity : enhancement
742 Description: Patchless ZC(zero copy) socklnd
743 Details : New protocol for socklnd, socklnd can support zero copy without
744 kernel patch, it's compatible with old socklnd. Checksum is
745 moved from tunables to modparams.
749 Description: When ksocknal_del_peer() is called upon a peer whose
750 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
751 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
754 Frequency : when ptlrpc is under heavy use and runs out of request buffer
756 Description: In lnet_match_blocked_msg(), md can be used without holding a
760 Frequency : very rarely
762 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
763 If connd connects a route which has been closed by
764 ksocknal_shutdown(), ksocknal_create_routes() may create new
765 routes which hold references on the peer, causing shutdown
766 process to wait for peer to disappear forever.
768 Severity : enhancement
770 Description: Dump XT3 portals traces on kptllnd timeout
771 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
772 dump Cray portals debug traces to a file. The kptllnd module
773 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
774 is the basename of the dump file.
777 Frequency : infrequent
779 Description: kernel ptllnd fix bug in connection re-establishment
780 Details : Kernel ptllnd could produce protocol errors e.g. illegal
781 matchbits and/or violate the credit flow protocol when trying
782 to re-establish a connection with a peer after an error or
785 Severity : enhancement
787 Description: Allow /proc/sys/lnet/debug to be set symbolically
788 Details : Allow debug and subsystem debug values to be read/set by name
789 in addition to numerically, for ease of use.
792 Frequency : only in configurations with LNET routers
794 Description: routes automatically marked down and recovered
795 Details : In configurations with LNET routers if a router fails routers
796 now actively try to recover routes that are down, unless they
797 are marked down by an administrator.
799 ------------------------------------------------------------------------------
801 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
804 Frequency : very rarely, in configurations with LNET routers and TCP
806 Description: incorrect data written to files on OSTs
807 Details : In certain high-load conditions incorrect data may be written
808 to files on the OST when using TCP networks.
810 ------------------------------------------------------------------------------
812 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
814 - rework CDEBUG messages rate-limiting mechanism b=10375
815 - add per-socket tunables for socklnd if the kernel is patched b=10327
817 ------------------------------------------------------------------------------
819 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
821 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
822 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
823 - quiet LNET startup LNI message for liblustre b=10128
824 - Better console error messages if 'ip2nets' can't match an IP address
825 - Fixed overflow/use-before-set bugs in linux-time.h
826 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
827 - LNET teardown failed an assertion about the route table being empty
828 - Fixed a crash in LNetEQPoll(<invalid handle>)
829 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
830 - improve debug message for liblustre/Catamount nodes (b=10116)
832 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
833 * Configuration change for the XT3
834 The PTLLND is now used to run Lustre over Portals on the XT3.
835 The configure option(s) --with-cray-portals are no longer
836 used. Rather --with-portals=<path-to-portals-includes> is
837 used to enable building on the XT3. In addition to enable
838 XT3 specific features the option --enable-cray-xt3 must be
841 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
842 * Portals has been removed, replaced by LNET.
843 LNET is new networking infrastructure for Lustre, it includes a
844 reorganized network configuration mode (see the user
845 documentation for full details) as well as support for routing
846 between different network fabrics. Lustre Networking Devices
847 (LNDS) for the supported network fabrics have also been created
848 for this new infrastructure.
850 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
855 Frequency : rare (large Voltaire clusters only)
857 Description: the default number of reserved transmit descriptors was too low
858 for some large clusters
859 Details : As a workaround, the number was increased. A proper fix includes
862 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
867 Frequency : occasional (large-scale events, cluster reboot, network failure)
869 Description: too many error messages on console obscure actual problem and
870 can slow down/panic server, or cause recovery to fail repeatedly
871 Details : enable rate-limiting of console error messages, and some messages
872 that were console errors now only go to the kernel log
874 Severity : enhancement
876 Description: add /proc/sys/portals/catastrophe entry which will report if
877 that node has previously LBUGged
879 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
881 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
883 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
885 - handle error return code in kranal_check_fma_rx() (5915,6054)
887 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
889 - update vibnal (Voltaire IB NAL)
890 - update gmnal (Myrinet NAL), gmnalid
892 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
894 * Landed portals:b_port_step as follows...
896 - removed CFS_DECL_SPIN*
897 just use 'spinlock_t' and initialise with spin_lock_init()
899 - removed CFS_DECL_MUTEX*
900 just use 'struct semaphore' and initialise with init_mutex()
902 - removed CFS_DECL_RWSEM*
903 just use 'struct rw_semaphore' and initialise with init_rwsem()
905 - renamed cfs_sleep_chan -> cfs_waitq
906 cfs_sleep_link -> cfs_waitlink
908 - fixed race in linux version of arch-independent socknal
909 (the ENOMEM/EAGAIN decision).
911 - Didn't fix problems in Darwin version of arch-independent socknal
912 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
914 - removed libcfs types from non-socknal header files (only some types
915 in the header files had been changed; the .c files hadn't been