1 2010-07-15 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 1.2.10 or later,
13 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
17 Description: should update lp_alive for non-router peers
19 Severity : enhancement
21 Description: LNet router shuffler.
23 Severity : enhancement
25 Description: LNet fine grain routing support.
29 Description: router checker stops working when system wall clock goes backward
30 Details : use monotonic timing source instead of system wall clock time.
32 Severity : enhancement
34 Description: avoid asymmetrical router failures
36 Severity : enhancement
38 Description: multiple-instance support for kptllnd
42 Description: ksocknal_close_conn_locked connection race
43 Details : A race was possible when ksocknal_create_conn calls
44 ksocknal_close_conn_locked for already closed conn.
48 Description: router_proc.c is rewritten to use sysctl-interface for parameters
49 residing in /proc/sys/lnet
51 Severity : enhancement
53 Description: port router pinger to userspace
57 Description: kptllnd HELLO protocol deadlock
58 Details : kptllnd HELLO protocol doesn't run to completion in finite time
62 Description: LNet selftest fixes and enhancements
64 Severity : enhancement
66 Description: allow a test node to be a member of multiple test groups
68 Severity : enhancement
70 Description: MXLND: eliminate hosts file, use arp for peer nic_id resolution
71 Details : an update from the upstream developer Scott Atchley.
73 Severity : enhancement
75 Description: add a new LND optiion to control peer buffer credits on routers
79 Description: Fixing deadlock in usocklnd
80 Details : A deadlock was possible in usocklnd due to race condition while
81 tearing connection down. The problem resulted from erroneous
82 assumption that lnet_finalize() could have been called holding
86 Bugzilla : 13621, 15983
87 Description: Protocol V2 of o2iblnd
88 Details : o2iblnd V2 has several new features:
89 . map-on-demand: map-on-demand is disabled by default, it can
90 be enabled by using modparam "map_on_demand=@value@", @value@
91 should >= 0 and < 256, 0 will disable map-on-demand, any other
92 valid value will enable map-on-demand.
93 Oi2blnd will create FMR or physical MR for RDMA if fragments of
95 Enable map-on-demand will take less memory for new connection,
96 but a little more CPU for RDMA.
97 . iWARP : to support iWARP, please enable map-on-demand, 32 and 64
98 are recommanded value. iWARP will probably fail for value >=128.
99 . OOB NOOP message: to resolve deadlock on router.
100 . tunable peer_credits_hiw: (high water to return credits),
101 default value of peer_credits_hiw equals to (peer_credits -1),
102 user can change it between peer_credits/2 and (peer_credits - 1).
103 Lower value is recommended for high latency network.
104 . tunable message queue size: it always equals to peer_credits,
105 higher value is recommended for high latency network.
106 . It's compatible with earlier version of o2iblnd
110 Description: Fixing 'running out of ports' issue
111 Details : Add a delay before next reconnect attempt in ksocklnd in
112 the case of lost race. Limit the frequency of query-requests
113 in lnet. Improved handling of 'dead peer' notifications in
118 Description: Change ptllnd timeout and watchdog timers
119 Details : Add ptltrace_on_nal_failed and bump ptllnd timeout to match
120 Portals wire timeout.
124 Description: One down Lustre FS hangs ALL mounted Lustre filesystems
125 Details : Shared routing enhancements - peer health detection.
127 Severity : enhancement
129 Description: acceptor.c cleanup
130 Details : Code duplication in acceptor.c for the cases of kernel and
131 user-space removed. User-space libcfs tcpip primitives
132 uniformed to have prototypes similar to kernel ones. Minor
133 cosmetic changes in usocklnd to use cfs_socket_t as
134 representation of socket.
138 Description: IB path MTU mistakenly set to 1st path MTU when ib_mtu is off
139 Details : See comment 46 in bug 11245 for details - it's indeed a bug
140 introduced by the original 11245 fix.
144 Description: uptllnd credit overflow fix
145 Details : kptl_msg_t::ptlm_credits could be overflown by uptllnd since
150 Description: socklnd protocol version 3
151 Details : With current protocol V2, connections on router can be
152 blocked and can't receive any incoming messages when there is no
153 more router buffer, so ZC-ACK can't be handled (LNet message
154 can't be finalized) and will cause deadlock on router.
155 Protocol V3 has a dedicated connection for emergency messages
156 like ZC-ACK to router, messages on this dedicated connection
157 don't need any credit so will never be blocked. Also, V3 can send
158 keepalive ping in specified period for router healthy checking.
160 -------------------------------------------------------------------------------
162 12-31-2008 Sun Microsystems, Inc.
164 * Support for networks:
165 socklnd - any kernel supported by Lustre,
166 qswlnd - Qsnet kernel modules 5.20 and later,
167 openiblnd - IbGold 1.8.2,
168 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
169 viblnd - Voltaire ibhost 3.4.5 and later,
170 ciblnd - Topspin 3.2.0,
171 iiblnd - Infiniserv 3.3 + PathBits patch,
172 gmlnd - GM 2.1.22 and later,
173 mxlnd - MX 1.2.1 or later,
174 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
178 Description: workaround for OOM from o2iblnd
179 Details : OFED needs allocate big chunk of memory for QP while creating
180 connection for o2iblnd, OOM can happen if no such a contiguous
182 QP size is decided by concurrent_sends and max_fragments of
183 o2iblnd, now we permit user to specify smaller value for
184 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
185 will decrease memory block size required by creating QP.
189 Description: Support Zerocopy receive of Chelsio device
190 Details : Chelsio driver can support zerocopy for iov[1] if it's
191 contiguous and large enough.
195 Description: fix credit flow deadlock in uptllnd
199 Description: finalize network operation in reasonable time
200 Details : conf-sanity test_32a couldn't stop ost and mds because it
201 tried to access non-existent peer and tcp connect took
202 quite long before timing out.
206 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
207 Details : Lost reference on conn prevents peer from being destroyed, which
208 could prevent new peer creation if peer count has reached upper
213 Description: LNET Selftest results in Soft lockup on OSS CPU
214 Details : only hits when 8 or more o2ib clients involved and a session is
215 torn down with 'lst end_session' without preceeding 'lst stop'.
219 Description: concurrent_sends in IB LNDs should not be changeable at run time
220 Details : concurrent_sends in IB LNDs should not be changeable at run time
224 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
225 Details : only hits under out-of-memory situations
228 -------------------------------------------------------------------------------
230 2009-02-07 Sun Microsystems, Inc.
232 * Support for networks:
233 socklnd - any kernel supported by Lustre,
234 qswlnd - Qsnet kernel modules 5.20 and later,
235 openiblnd - IbGold 1.8.2,
236 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
237 viblnd - Voltaire ibhost 3.4.5 and later,
238 ciblnd - Topspin 3.2.0,
239 iiblnd - Infiniserv 3.3 + PathBits patch,
240 gmlnd - GM 2.1.22 and later,
241 mxlnd - MX 1.2.1 or later,
242 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
245 Description: workaround for OOM from o2iblnd
246 Details : OFED needs allocate big chunk of memory for QP while creating
247 connection for o2iblnd, OOM can happen if no such a contiguous
249 QP size is decided by concurrent_sends and max_fragments of
250 o2iblnd, now we permit user to specify smaller value for
251 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
252 will decrease memory block size required by creating QP.
256 Description: Support Zerocopy receive of Chelsio device
257 Details : Chelsio driver can support zerocopy for iov[1] if it's
258 contiguous and large enough.
261 Description: fix credit flow deadlock in uptllnd
265 Description: finalize network operation in reasonable time
266 Details : conf-sanity test_32a couldn't stop ost and mds because it
267 tried to access non-existent peer and tcp connect took
268 quite long before timing out.
272 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
273 Details : Lost reference on conn prevents peer from being destroyed, which
274 could prevent new peer creation if peer count has reached upper
279 Description: LNET Selftest results in Soft lockup on OSS CPU
280 Details : only hits when 8 or more o2ib clients involved and a session is
281 torn down with 'lst end_session' without preceeding 'lst stop'.
285 Description: concurrent_sends in IB LNDs should not be changeable at run time
286 Details : concurrent_sends in IB LNDs should not be changeable at run time
288 -------------------------------------------------------------------------------
290 11-03-2008 Sun Microsystems, Inc.
292 * Support for networks:
293 socklnd - any kernel supported by Lustre,
294 qswlnd - Qsnet kernel modules 5.20 and later,
295 openiblnd - IbGold 1.8.2,
296 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
297 viblnd - Voltaire ibhost 3.4.5 and later,
298 ciblnd - Topspin 3.2.0,
299 iiblnd - Infiniserv 3.3 + PathBits patch,
300 gmlnd - GM 2.1.22 and later,
301 mxlnd - MX 1.2.1 or later,
302 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
306 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
307 Details : only hits under out-of-memory situations
309 -------------------------------------------------------------------------------
311 04-26-2008 Sun Microsystems, Inc.
313 * Support for networks:
314 socklnd - any kernel supported by Lustre,
315 qswlnd - Qsnet kernel modules 5.20 and later,
316 openiblnd - IbGold 1.8.2,
317 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
318 viblnd - Voltaire ibhost 3.4.5 and later,
319 ciblnd - Topspin 3.2.0,
320 iiblnd - Infiniserv 3.3 + PathBits patch,
321 gmlnd - GM 2.1.22 and later,
322 mxlnd - MX 1.2.1 or later,
323 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
327 Description: excessive debug information removed
328 Details : excessive debug information removed
332 Description: ksocknal_create_conn() hit ASSERTION during connection race
333 Details : ksocknal_create_conn() hit ASSERTION during connection race
337 Description: ksocknal_send_hello() hit ASSERTION while connecting race
338 Details : ksocknal_send_hello() hit ASSERTION while connecting race
342 Description: o2iblnd/ptllnd credit deadlock in a routed config.
343 Details : o2iblnd/ptllnd credit deadlock in a routed config.
347 Description: High load after starting lnet
348 Details : gmlnd should sleep in rx thread in interruptible way. Otherwise,
349 uptime utility reports high load that looks confusingly.
353 Description: ksocklnd fails to establish connection if accept_port is high
354 Details : PID remapping must not be done for active (outgoing) connections
357 --------------------------------------------------------------------------------
359 2008-01-11 Sun Microsystems, Inc.
361 * Support for networks:
362 socklnd - any kernel supported by Lustre,
363 qswlnd - Qsnet kernel modules 5.20 and later,
364 openiblnd - IbGold 1.8.2,
365 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
366 viblnd - Voltaire ibhost 3.4.5 and later,
367 ciblnd - Topspin 3.2.0,
368 iiblnd - Infiniserv 3.3 + PathBits patch,
369 gmlnd - GM 2.1.22 and later,
370 mxlnd - MX 1.2.1 or later,
371 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
375 Description: liblustre network error
376 Details : liblustre clients should understand LNET_ACCEPT_PORT environment
377 variable even if they don't start lnet acceptor.
381 Description: Strange message from lnet (Ignoring prediction from the future)
382 Details : Incorrect calculation of peer's last_alive value in ksocklnd
384 --------------------------------------------------------------------------------
386 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
388 * Support for networks:
389 socklnd - any kernel supported by Lustre,
390 qswlnd - Qsnet kernel modules 5.20 and later,
391 openiblnd - IbGold 1.8.2,
392 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5.
393 viblnd - Voltaire ibhost 3.4.5 and later,
394 ciblnd - Topspin 3.2.0,
395 iiblnd - Infiniserv 3.3 + PathBits patch,
396 gmlnd - GM 2.1.22 and later,
397 mxlnd - MX 1.2.1 or later,
398 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
402 Description: ASSERTION(me == md->md_me) failed in lnet_match_md()
406 Description: increase send queue size for ciblnd/openiblnd
410 Description: new userspace socklnd
411 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
412 with new one - usocklnd.
414 Severity : enhancement
416 Description: Console message flood
417 Details : Make cdls ratelimiting more tunable by adding several tunable in
418 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
419 /proc/sys/lnet/console_backoff.
421 --------------------------------------------------------------------------------
423 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
425 * Support for networks:
426 socklnd - any kernel supported by Lustre,
427 qswlnd - Qsnet kernel modules 5.20 and later,
428 openiblnd - IbGold 1.8.2,
429 o2iblnd - OFED 1.1 and 1.2,
430 viblnd - Voltaire ibhost 3.4.5 and later,
431 ciblnd - Topspin 3.2.0,
432 iiblnd - Infiniserv 3.3 + PathBits patch,
433 gmlnd - GM 2.1.22 and later,
434 mxlnd - MX 1.2.1 or later,
435 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
439 Description: /proc/sys/lnet has non-sysctl entries
440 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
444 Description: TOE Kernel panic by ksocklnd
445 Details : offloaded sockets provide their own implementation of sendpage,
446 can't call tcp_sendpage() directly
450 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
451 Details : races between lnd_shutdown and peer creation prevent
452 lnd_shutdown from finishing.
456 Description: open files rlimit 1024 reached while liblustre testing
457 Details : ulnds/socklnd must close open socket after unsuccessful
462 Description: build error
463 Details : fix typos in gmlnd, ptllnd and viblnd
465 --------------------------------------------------------------------------------
467 2007-07-30 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,
473 o2iblnd - OFED 1.1 and 1.2
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
481 --------------------------------------------------------------------------------
483 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
485 * Support for networks:
486 socklnd - kernels up to 2.6.16,
487 qswlnd - Qsnet kernel modules 5.20 and later,
488 openiblnd - IbGold 1.8.2,
490 viblnd - Voltaire ibhost 3.4.5 and later,
491 ciblnd - Topspin 3.2.0,
492 iiblnd - Infiniserv 3.3 + PathBits patch,
493 gmlnd - GM 2.1.22 and later,
494 mxlnd - MX 1.2.1 or later,
495 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
499 Description: Initialize cpumask before use
503 Description: ASSERTION failures when upgrading to the patchless zero-copy
505 Details : This bug affects "rolling upgrades", causing an inconsistent
506 protocol version negotiation and subsequent assertion failure
507 during rolling upgrades after the first wave of upgrades.
511 Details : Change "dropped message" CERRORs to D_NETERROR so they are
512 logged instead of creating "console chatter" when a lustre
513 timeout races with normal RPC completion.
516 Details : lnet_clear_peer_table can wait forever if user forgets to
520 Details : libcfs_id2str should check pid against LNET_PID_ANY.
524 Description: added LNET self test
525 Details : landing b_self_test
530 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
532 Details : do_div() macro is used incorrectly.
534 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
538 Description: make panic on lbug configurable
542 Description: Add OFED1.2 support to o2iblnd
543 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
544 are installed (other than kernel's in-tree infiniband), there
545 could be some problem while insmod o2iblnd (mismatch CRC of
547 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
548 this link provides solution:
549 https://bugs.openfabrics.org/show_bug.cgi?id=355
550 if extra Module.symvers is not supported in kernel, we will
551 have to run the script in bug 12316 to update
552 $LINUX/module.symvers before building o2iblnd.
553 More details about this are in bug 12316.
555 ------------------------------------------------------------------------------
557 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
558 * version 1.4.10 / 1.6.0
559 * Support for networks:
560 socklnd - kernels up to 2.6.16,
561 qswlnd - Qsnet kernel modules 5.20 and later,
562 openiblnd - IbGold 1.8.2,
564 viblnd - Voltaire ibhost 3.4.5 and later,
565 ciblnd - Topspin 3.2.0,
566 iiblnd - Infiniserv 3.3 + PathBits patch,
567 gmlnd - GM 2.1.22 and later,
568 mxlnd - MX 1.2.1 or later,
569 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
573 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
574 possibly accessed in kptllnd_shutdown. Ptllnd should init
575 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
579 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
583 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
587 Description: the_lnet.ln_finalizing was not set when the current thread is
588 about to complete messages. It only affects multi-threaded
594 Description: Changed the default kqswlnd ntxmsg=512
599 Description: Assertion failure in kernel ptllnd caused by posting passive
600 bulk buffers before connection establishment complete.
605 Description: A race in kernel ptllnd between deleting a peer and posting
606 new communications for it could hang communications -
607 manifesting as "Unexpectedly long timeout" messages.
612 Description: Kernel ptllnd lock ordering issue could hang a node.
617 Description: node crash on socket teardown race
620 Frequency : 'lctl peer_list' issued on a mx net
622 Description: Enable lctl's peer_list for MXLND
625 Frequency : after Ptllnd timeouts and portals congestion
627 Description: Credit overflows
628 Details : This was a bug in ptllnd connection establishment. The fix
629 implements better peer stamps to disambiguate connection
630 establishment and ensure both peers enter the credit flow
631 state machine consistently.
636 Description: kptllnd didn't propagate some network errors up to LNET
637 Details : This bug was spotted while investigating 11394. The fix
638 ensures network errors on sends and bulk transfers are
639 propagated to LNET/lustre correctly.
641 Severity : enhancement
643 Description: Fixed console chatter in case of -ETIMEDOUT.
645 Severity : enhancement
647 Description: Added D_NETTRACE for recording network packet history
648 (initially only for ptllnd). Also a separate userspace
649 ptllnd facility to gather history which should really be
650 covered by D_NETTRACE too, if only CDEBUG recorded history in
656 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
657 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
658 callback can occur before a connection has actually been
659 established. This caused an assertion failure previously.
661 Severity : enhancement
663 Description: Multiple instances for o2iblnd
664 Details : Allow multiple instances of o2iblnd to enable networking over
665 multiple HCAs and routing between them.
669 Description: lnet deadlock in router_checker
670 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
671 into BH locks to eliminate potential deadlock caused by
672 ksocknal_data_ready() preempting code holding these locks.
676 Description: Millions of failed socklnd connection attempts cause a very slow FS
677 Details : added a new route flag ksnr_scheduled to distinguish from
678 ksnr_connecting, so that a peer connection request is only turned
679 down for race concerns when an active connection to the same peer
680 is under progress (instead of just being scheduled).
682 ------------------------------------------------------------------------------
684 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
686 * Support for networks:
687 socklnd - kernels up to 2.6.16
688 qswlnd - Qsnet kernel modules 5.20 and later
689 openiblnd - IbGold 1.8.2
691 viblnd - Voltaire ibhost 3.4.5 and later
692 ciblnd - Topspin 3.2.0
693 iiblnd - Infiniserv 3.3 + PathBits patch
694 gmlnd - GM 2.1.22 and later
695 mxlnd - MX 1.2.1 or later
696 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
699 Severity : major on XT3
701 Description: libcfs overwrites /proc/sys/portals
702 Details : libcfs created a symlink from /proc/sys/portals to
703 /proc/sys/lnet for backwards compatibility. This is no
704 longer required and makes the Cray portals /proc variables
709 Description: OFED FMR API change
710 Details : This changes parameter usage to reflect a change in
711 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
712 that FMR support is only used in experimental versions of the
713 o2iblnd - this change does not affect standard usage at all.
715 Severity : enhancement
717 Description: new ko2iblnd module parameter: ib_mtu
718 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
719 HCAs. You can avoid this problem by setting the MTU to 1024
720 using this module parameter.
722 Severity : enhancement
723 Bugzilla : 11118/11620
724 Description: ptllnd small request message buffer alignment fix
725 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
726 Round up small message size on sends in case this option
727 is not supported. 11620 was a defect in the initial
728 implementation which effectively asserted all peers had to be
729 running the correct protocol version which was fixed by always
730 NAK-ing such requests and handling any misalignments they
735 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
736 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
737 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
739 Severity : enhancement
741 Description: Patchless ZC(zero copy) socklnd
742 Details : New protocol for socklnd, socklnd can support zero copy without
743 kernel patch, it's compatible with old socklnd. Checksum is
744 moved from tunables to modparams.
748 Description: When ksocknal_del_peer() is called upon a peer whose
749 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
750 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
753 Frequency : when ptlrpc is under heavy use and runs out of request buffer
755 Description: In lnet_match_blocked_msg(), md can be used without holding a
759 Frequency : very rarely
761 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
762 If connd connects a route which has been closed by
763 ksocknal_shutdown(), ksocknal_create_routes() may create new
764 routes which hold references on the peer, causing shutdown
765 process to wait for peer to disappear forever.
767 Severity : enhancement
769 Description: Dump XT3 portals traces on kptllnd timeout
770 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
771 dump Cray portals debug traces to a file. The kptllnd module
772 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
773 is the basename of the dump file.
776 Frequency : infrequent
778 Description: kernel ptllnd fix bug in connection re-establishment
779 Details : Kernel ptllnd could produce protocol errors e.g. illegal
780 matchbits and/or violate the credit flow protocol when trying
781 to re-establish a connection with a peer after an error or
784 Severity : enhancement
786 Description: Allow /proc/sys/lnet/debug to be set symbolically
787 Details : Allow debug and subsystem debug values to be read/set by name
788 in addition to numerically, for ease of use.
791 Frequency : only in configurations with LNET routers
793 Description: routes automatically marked down and recovered
794 Details : In configurations with LNET routers if a router fails routers
795 now actively try to recover routes that are down, unless they
796 are marked down by an administrator.
798 ------------------------------------------------------------------------------
800 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
803 Frequency : very rarely, in configurations with LNET routers and TCP
805 Description: incorrect data written to files on OSTs
806 Details : In certain high-load conditions incorrect data may be written
807 to files on the OST when using TCP networks.
809 ------------------------------------------------------------------------------
811 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
813 - rework CDEBUG messages rate-limiting mechanism b=10375
814 - add per-socket tunables for socklnd if the kernel is patched b=10327
816 ------------------------------------------------------------------------------
818 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
820 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
821 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
822 - quiet LNET startup LNI message for liblustre b=10128
823 - Better console error messages if 'ip2nets' can't match an IP address
824 - Fixed overflow/use-before-set bugs in linux-time.h
825 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
826 - LNET teardown failed an assertion about the route table being empty
827 - Fixed a crash in LNetEQPoll(<invalid handle>)
828 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
829 - improve debug message for liblustre/Catamount nodes (b=10116)
831 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
832 * Configuration change for the XT3
833 The PTLLND is now used to run Lustre over Portals on the XT3.
834 The configure option(s) --with-cray-portals are no longer
835 used. Rather --with-portals=<path-to-portals-includes> is
836 used to enable building on the XT3. In addition to enable
837 XT3 specific features the option --enable-cray-xt3 must be
840 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
841 * Portals has been removed, replaced by LNET.
842 LNET is new networking infrastructure for Lustre, it includes a
843 reorganized network configuration mode (see the user
844 documentation for full details) as well as support for routing
845 between different network fabrics. Lustre Networking Devices
846 (LNDS) for the supported network fabrics have also been created
847 for this new infrastructure.
849 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
854 Frequency : rare (large Voltaire clusters only)
856 Description: the default number of reserved transmit descriptors was too low
857 for some large clusters
858 Details : As a workaround, the number was increased. A proper fix includes
861 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
866 Frequency : occasional (large-scale events, cluster reboot, network failure)
868 Description: too many error messages on console obscure actual problem and
869 can slow down/panic server, or cause recovery to fail repeatedly
870 Details : enable rate-limiting of console error messages, and some messages
871 that were console errors now only go to the kernel log
873 Severity : enhancement
875 Description: add /proc/sys/portals/catastrophe entry which will report if
876 that node has previously LBUGged
878 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
880 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
882 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
884 - handle error return code in kranal_check_fma_rx() (5915,6054)
886 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
888 - update vibnal (Voltaire IB NAL)
889 - update gmnal (Myrinet NAL), gmnalid
891 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
893 * Landed portals:b_port_step as follows...
895 - removed CFS_DECL_SPIN*
896 just use 'spinlock_t' and initialise with spin_lock_init()
898 - removed CFS_DECL_MUTEX*
899 just use 'struct semaphore' and initialise with init_mutex()
901 - removed CFS_DECL_RWSEM*
902 just use 'struct rw_semaphore' and initialise with init_rwsem()
904 - renamed cfs_sleep_chan -> cfs_waitq
905 cfs_sleep_link -> cfs_waitlink
907 - fixed race in linux version of arch-independent socknal
908 (the ENOMEM/EAGAIN decision).
910 - Didn't fix problems in Darwin version of arch-independent socknal
911 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
913 - removed libcfs types from non-socknal header files (only some types
914 in the header files had been changed; the .c files hadn't been