3 * Support for networks:
4 socklnd - any kernel supported by Lustre,
5 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, 1.3, and 1.4.1
6 mxlnd - MX 1.2.10 or later
7 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
9 -------------------------------------------------------------------------------
11 09-30-2011 Whamcloud, Inc.
13 * Support for networks:
14 socklnd - any kernel supported by Lustre,
15 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, 1.3, and 1.4.1
16 * Available but unsupported:
17 mxlnd - MX 1.2.10 or later
18 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
20 -------------------------------------------------------------------------------
22 2010-07-15 Oracle, Inc.
24 * Support for networks:
25 socklnd - any kernel supported by Lustre,
26 qswlnd - Qsnet kernel modules 5.20 and later,
27 openiblnd - IbGold 1.8.2,
28 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, 1.3, and 1.4.1
29 viblnd - Voltaire ibhost 3.4.5 and later,
30 ciblnd - Topspin 3.2.0,
31 iiblnd - Infiniserv 3.3 + PathBits patch,
32 gmlnd - GM 2.1.22 and later,
33 mxlnd - MX 1.2.10 or later,
34 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
38 Description: should update lp_alive for non-router peers
40 Severity : enhancement
42 Description: LNet router shuffler.
44 Severity : enhancement
46 Description: LNet fine grain routing support.
50 Description: router checker stops working when system wall clock goes backward
51 Details : use monotonic timing source instead of system wall clock time.
53 Severity : enhancement
55 Description: avoid asymmetrical router failures
57 Severity : enhancement
59 Description: multiple-instance support for kptllnd
63 Description: ksocknal_close_conn_locked connection race
64 Details : A race was possible when ksocknal_create_conn calls
65 ksocknal_close_conn_locked for already closed conn.
69 Description: router_proc.c is rewritten to use sysctl-interface for parameters
70 residing in /proc/sys/lnet
72 Severity : enhancement
74 Description: port router pinger to userspace
78 Description: kptllnd HELLO protocol deadlock
79 Details : kptllnd HELLO protocol doesn't run to completion in finite time
83 Description: LNet selftest fixes and enhancements
85 Severity : enhancement
87 Description: allow a test node to be a member of multiple test groups
89 Severity : enhancement
91 Description: MXLND: eliminate hosts file, use arp for peer nic_id resolution
92 Details : an update from the upstream developer Scott Atchley.
94 Severity : enhancement
96 Description: add a new LND optiion to control peer buffer credits on routers
100 Description: Fixing deadlock in usocklnd
101 Details : A deadlock was possible in usocklnd due to race condition while
102 tearing connection down. The problem resulted from erroneous
103 assumption that lnet_finalize() could have been called holding
104 some lnd-level locks.
107 Bugzilla : 13621, 15983
108 Description: Protocol V2 of o2iblnd
109 Details : o2iblnd V2 has several new features:
110 . map-on-demand: map-on-demand is disabled by default, it can
111 be enabled by using modparam "map_on_demand=@value@", @value@
112 should >= 0 and < 256, 0 will disable map-on-demand, any other
113 valid value will enable map-on-demand.
114 Oi2blnd will create FMR or physical MR for RDMA if fragments of
116 Enable map-on-demand will take less memory for new connection,
117 but a little more CPU for RDMA.
118 . iWARP : to support iWARP, please enable map-on-demand, 32 and 64
119 are recommanded value. iWARP will probably fail for value >=128.
120 . OOB NOOP message: to resolve deadlock on router.
121 . tunable peer_credits_hiw: (high water to return credits),
122 default value of peer_credits_hiw equals to (peer_credits -1),
123 user can change it between peer_credits/2 and (peer_credits - 1).
124 Lower value is recommended for high latency network.
125 . tunable message queue size: it always equals to peer_credits,
126 higher value is recommended for high latency network.
127 . It's compatible with earlier version of o2iblnd
131 Description: Fixing 'running out of ports' issue
132 Details : Add a delay before next reconnect attempt in ksocklnd in
133 the case of lost race. Limit the frequency of query-requests
134 in lnet. Improved handling of 'dead peer' notifications in
139 Description: Change ptllnd timeout and watchdog timers
140 Details : Add ptltrace_on_nal_failed and bump ptllnd timeout to match
141 Portals wire timeout.
145 Description: One down Lustre FS hangs ALL mounted Lustre filesystems
146 Details : Shared routing enhancements - peer health detection.
148 Severity : enhancement
150 Description: acceptor.c cleanup
151 Details : Code duplication in acceptor.c for the cases of kernel and
152 user-space removed. User-space libcfs tcpip primitives
153 uniformed to have prototypes similar to kernel ones. Minor
154 cosmetic changes in usocklnd to use cfs_socket_t as
155 representation of socket.
159 Description: IB path MTU mistakenly set to 1st path MTU when ib_mtu is off
160 Details : See comment 46 in bug 11245 for details - it's indeed a bug
161 introduced by the original 11245 fix.
165 Description: uptllnd credit overflow fix
166 Details : kptl_msg_t::ptlm_credits could be overflown by uptllnd since
171 Description: socklnd protocol version 3
172 Details : With current protocol V2, connections on router can be
173 blocked and can't receive any incoming messages when there is no
174 more router buffer, so ZC-ACK can't be handled (LNet message
175 can't be finalized) and will cause deadlock on router.
176 Protocol V3 has a dedicated connection for emergency messages
177 like ZC-ACK to router, messages on this dedicated connection
178 don't need any credit so will never be blocked. Also, V3 can send
179 keepalive ping in specified period for router healthy checking.
181 -------------------------------------------------------------------------------
183 12-31-2008 Sun Microsystems, Inc.
185 * Support for networks:
186 socklnd - any kernel supported by Lustre,
187 qswlnd - Qsnet kernel modules 5.20 and later,
188 openiblnd - IbGold 1.8.2,
189 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
190 viblnd - Voltaire ibhost 3.4.5 and later,
191 ciblnd - Topspin 3.2.0,
192 iiblnd - Infiniserv 3.3 + PathBits patch,
193 gmlnd - GM 2.1.22 and later,
194 mxlnd - MX 1.2.1 or later,
195 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
199 Description: workaround for OOM from o2iblnd
200 Details : OFED needs allocate big chunk of memory for QP while creating
201 connection for o2iblnd, OOM can happen if no such a contiguous
203 QP size is decided by concurrent_sends and max_fragments of
204 o2iblnd, now we permit user to specify smaller value for
205 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
206 will decrease memory block size required by creating QP.
210 Description: Support Zerocopy receive of Chelsio device
211 Details : Chelsio driver can support zerocopy for iov[1] if it's
212 contiguous and large enough.
216 Description: fix credit flow deadlock in uptllnd
220 Description: finalize network operation in reasonable time
221 Details : conf-sanity test_32a couldn't stop ost and mds because it
222 tried to access non-existent peer and tcp connect took
223 quite long before timing out.
227 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
228 Details : Lost reference on conn prevents peer from being destroyed, which
229 could prevent new peer creation if peer count has reached upper
234 Description: LNET Selftest results in Soft lockup on OSS CPU
235 Details : only hits when 8 or more o2ib clients involved and a session is
236 torn down with 'lst end_session' without preceeding 'lst stop'.
240 Description: concurrent_sends in IB LNDs should not be changeable at run time
241 Details : concurrent_sends in IB LNDs should not be changeable at run time
245 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
246 Details : only hits under out-of-memory situations
249 -------------------------------------------------------------------------------
251 2009-02-07 Sun Microsystems, Inc.
253 * Support for networks:
254 socklnd - any kernel supported by Lustre,
255 qswlnd - Qsnet kernel modules 5.20 and later,
256 openiblnd - IbGold 1.8.2,
257 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
258 viblnd - Voltaire ibhost 3.4.5 and later,
259 ciblnd - Topspin 3.2.0,
260 iiblnd - Infiniserv 3.3 + PathBits patch,
261 gmlnd - GM 2.1.22 and later,
262 mxlnd - MX 1.2.1 or later,
263 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
266 Description: workaround for OOM from o2iblnd
267 Details : OFED needs allocate big chunk of memory for QP while creating
268 connection for o2iblnd, OOM can happen if no such a contiguous
270 QP size is decided by concurrent_sends and max_fragments of
271 o2iblnd, now we permit user to specify smaller value for
272 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
273 will decrease memory block size required by creating QP.
277 Description: Support Zerocopy receive of Chelsio device
278 Details : Chelsio driver can support zerocopy for iov[1] if it's
279 contiguous and large enough.
282 Description: fix credit flow deadlock in uptllnd
286 Description: finalize network operation in reasonable time
287 Details : conf-sanity test_32a couldn't stop ost and mds because it
288 tried to access non-existent peer and tcp connect took
289 quite long before timing out.
293 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
294 Details : Lost reference on conn prevents peer from being destroyed, which
295 could prevent new peer creation if peer count has reached upper
300 Description: LNET Selftest results in Soft lockup on OSS CPU
301 Details : only hits when 8 or more o2ib clients involved and a session is
302 torn down with 'lst end_session' without preceeding 'lst stop'.
306 Description: concurrent_sends in IB LNDs should not be changeable at run time
307 Details : concurrent_sends in IB LNDs should not be changeable at run time
309 -------------------------------------------------------------------------------
311 11-03-2008 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, 1.2.0, 1.2.5, and 1.3
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: ptl_send_rpc hits LASSERT when ptl_send_buf fails
328 Details : only hits under out-of-memory situations
330 -------------------------------------------------------------------------------
332 04-26-2008 Sun Microsystems, Inc.
334 * Support for networks:
335 socklnd - any kernel supported by Lustre,
336 qswlnd - Qsnet kernel modules 5.20 and later,
337 openiblnd - IbGold 1.8.2,
338 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
339 viblnd - Voltaire ibhost 3.4.5 and later,
340 ciblnd - Topspin 3.2.0,
341 iiblnd - Infiniserv 3.3 + PathBits patch,
342 gmlnd - GM 2.1.22 and later,
343 mxlnd - MX 1.2.1 or later,
344 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
348 Description: excessive debug information removed
349 Details : excessive debug information removed
353 Description: ksocknal_create_conn() hit ASSERTION during connection race
354 Details : ksocknal_create_conn() hit ASSERTION during connection race
358 Description: ksocknal_send_hello() hit ASSERTION while connecting race
359 Details : ksocknal_send_hello() hit ASSERTION while connecting race
363 Description: o2iblnd/ptllnd credit deadlock in a routed config.
364 Details : o2iblnd/ptllnd credit deadlock in a routed config.
368 Description: High load after starting lnet
369 Details : gmlnd should sleep in rx thread in interruptible way. Otherwise,
370 uptime utility reports high load that looks confusingly.
374 Description: ksocklnd fails to establish connection if accept_port is high
375 Details : PID remapping must not be done for active (outgoing) connections
378 --------------------------------------------------------------------------------
380 2008-01-11 Sun Microsystems, Inc.
382 * Support for networks:
383 socklnd - any kernel supported by Lustre,
384 qswlnd - Qsnet kernel modules 5.20 and later,
385 openiblnd - IbGold 1.8.2,
386 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
387 viblnd - Voltaire ibhost 3.4.5 and later,
388 ciblnd - Topspin 3.2.0,
389 iiblnd - Infiniserv 3.3 + PathBits patch,
390 gmlnd - GM 2.1.22 and later,
391 mxlnd - MX 1.2.1 or later,
392 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
396 Description: liblustre network error
397 Details : liblustre clients should understand LNET_ACCEPT_PORT environment
398 variable even if they don't start lnet acceptor.
402 Description: Strange message from lnet (Ignoring prediction from the future)
403 Details : Incorrect calculation of peer's last_alive value in ksocklnd
405 --------------------------------------------------------------------------------
407 2007-12-07 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 1.2.0, 1.2.5.
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: ASSERTION(me == md->md_me) failed in lnet_match_md()
427 Description: increase send queue size for ciblnd/openiblnd
431 Description: new userspace socklnd
432 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
433 with new one - usocklnd.
435 Severity : enhancement
437 Description: Console message flood
438 Details : Make cdls ratelimiting more tunable by adding several tunable in
439 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
440 /proc/sys/lnet/console_backoff.
442 --------------------------------------------------------------------------------
444 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
446 * Support for networks:
447 socklnd - any kernel supported by Lustre,
448 qswlnd - Qsnet kernel modules 5.20 and later,
449 openiblnd - IbGold 1.8.2,
450 o2iblnd - OFED 1.1 and 1.2,
451 viblnd - Voltaire ibhost 3.4.5 and later,
452 ciblnd - Topspin 3.2.0,
453 iiblnd - Infiniserv 3.3 + PathBits patch,
454 gmlnd - GM 2.1.22 and later,
455 mxlnd - MX 1.2.1 or later,
456 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
460 Description: /proc/sys/lnet has non-sysctl entries
461 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
465 Description: TOE Kernel panic by ksocklnd
466 Details : offloaded sockets provide their own implementation of sendpage,
467 can't call tcp_sendpage() directly
471 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
472 Details : races between lnd_shutdown and peer creation prevent
473 lnd_shutdown from finishing.
477 Description: open files rlimit 1024 reached while liblustre testing
478 Details : ulnds/socklnd must close open socket after unsuccessful
483 Description: build error
484 Details : fix typos in gmlnd, ptllnd and viblnd
486 --------------------------------------------------------------------------------
488 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
490 * Support for networks:
491 socklnd - kernels up to 2.6.16,
492 qswlnd - Qsnet kernel modules 5.20 and later,
493 openiblnd - IbGold 1.8.2,
494 o2iblnd - OFED 1.1 and 1.2
495 viblnd - Voltaire ibhost 3.4.5 and later,
496 ciblnd - Topspin 3.2.0,
497 iiblnd - Infiniserv 3.3 + PathBits patch,
498 gmlnd - GM 2.1.22 and later,
499 mxlnd - MX 1.2.1 or later,
500 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
502 --------------------------------------------------------------------------------
504 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
506 * Support for networks:
507 socklnd - kernels up to 2.6.16,
508 qswlnd - Qsnet kernel modules 5.20 and later,
509 openiblnd - IbGold 1.8.2,
511 viblnd - Voltaire ibhost 3.4.5 and later,
512 ciblnd - Topspin 3.2.0,
513 iiblnd - Infiniserv 3.3 + PathBits patch,
514 gmlnd - GM 2.1.22 and later,
515 mxlnd - MX 1.2.1 or later,
516 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
520 Description: Initialize cpumask before use
524 Description: ASSERTION failures when upgrading to the patchless zero-copy
526 Details : This bug affects "rolling upgrades", causing an inconsistent
527 protocol version negotiation and subsequent assertion failure
528 during rolling upgrades after the first wave of upgrades.
532 Details : Change "dropped message" CERRORs to D_NETERROR so they are
533 logged instead of creating "console chatter" when a lustre
534 timeout races with normal RPC completion.
537 Details : lnet_clear_peer_table can wait forever if user forgets to
541 Details : libcfs_id2str should check pid against LNET_PID_ANY.
545 Description: added LNET self test
546 Details : landing b_self_test
551 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
553 Details : do_div() macro is used incorrectly.
555 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
559 Description: make panic on lbug configurable
563 Description: Add OFED1.2 support to o2iblnd
564 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
565 are installed (other than kernel's in-tree infiniband), there
566 could be some problem while insmod o2iblnd (mismatch CRC of
568 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
569 this link provides solution:
570 https://bugs.openfabrics.org/show_bug.cgi?id=355
571 if extra Module.symvers is not supported in kernel, we will
572 have to run the script in bug 12316 to update
573 $LINUX/module.symvers before building o2iblnd.
574 More details about this are in bug 12316.
576 ------------------------------------------------------------------------------
578 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
579 * version 1.4.10 / 1.6.0
580 * Support for networks:
581 socklnd - kernels up to 2.6.16,
582 qswlnd - Qsnet kernel modules 5.20 and later,
583 openiblnd - IbGold 1.8.2,
585 viblnd - Voltaire ibhost 3.4.5 and later,
586 ciblnd - Topspin 3.2.0,
587 iiblnd - Infiniserv 3.3 + PathBits patch,
588 gmlnd - GM 2.1.22 and later,
589 mxlnd - MX 1.2.1 or later,
590 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
594 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
595 possibly accessed in kptllnd_shutdown. Ptllnd should init
596 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
600 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
604 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
608 Description: the_lnet.ln_finalizing was not set when the current thread is
609 about to complete messages. It only affects multi-threaded
615 Description: Changed the default kqswlnd ntxmsg=512
620 Description: Assertion failure in kernel ptllnd caused by posting passive
621 bulk buffers before connection establishment complete.
626 Description: A race in kernel ptllnd between deleting a peer and posting
627 new communications for it could hang communications -
628 manifesting as "Unexpectedly long timeout" messages.
633 Description: Kernel ptllnd lock ordering issue could hang a node.
638 Description: node crash on socket teardown race
641 Frequency : 'lctl peer_list' issued on a mx net
643 Description: Enable lctl's peer_list for MXLND
646 Frequency : after Ptllnd timeouts and portals congestion
648 Description: Credit overflows
649 Details : This was a bug in ptllnd connection establishment. The fix
650 implements better peer stamps to disambiguate connection
651 establishment and ensure both peers enter the credit flow
652 state machine consistently.
657 Description: kptllnd didn't propagate some network errors up to LNET
658 Details : This bug was spotted while investigating 11394. The fix
659 ensures network errors on sends and bulk transfers are
660 propagated to LNET/lustre correctly.
662 Severity : enhancement
664 Description: Fixed console chatter in case of -ETIMEDOUT.
666 Severity : enhancement
668 Description: Added D_NETTRACE for recording network packet history
669 (initially only for ptllnd). Also a separate userspace
670 ptllnd facility to gather history which should really be
671 covered by D_NETTRACE too, if only CDEBUG recorded history in
677 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
678 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
679 callback can occur before a connection has actually been
680 established. This caused an assertion failure previously.
682 Severity : enhancement
684 Description: Multiple instances for o2iblnd
685 Details : Allow multiple instances of o2iblnd to enable networking over
686 multiple HCAs and routing between them.
690 Description: lnet deadlock in router_checker
691 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
692 into BH locks to eliminate potential deadlock caused by
693 ksocknal_data_ready() preempting code holding these locks.
697 Description: Millions of failed socklnd connection attempts cause a very slow FS
698 Details : added a new route flag ksnr_scheduled to distinguish from
699 ksnr_connecting, so that a peer connection request is only turned
700 down for race concerns when an active connection to the same peer
701 is under progress (instead of just being scheduled).
703 ------------------------------------------------------------------------------
705 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
707 * Support for networks:
708 socklnd - kernels up to 2.6.16
709 qswlnd - Qsnet kernel modules 5.20 and later
710 openiblnd - IbGold 1.8.2
712 viblnd - Voltaire ibhost 3.4.5 and later
713 ciblnd - Topspin 3.2.0
714 iiblnd - Infiniserv 3.3 + PathBits patch
715 gmlnd - GM 2.1.22 and later
716 mxlnd - MX 1.2.1 or later
717 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
720 Severity : major on XT3
722 Description: libcfs overwrites /proc/sys/portals
723 Details : libcfs created a symlink from /proc/sys/portals to
724 /proc/sys/lnet for backwards compatibility. This is no
725 longer required and makes the Cray portals /proc variables
730 Description: OFED FMR API change
731 Details : This changes parameter usage to reflect a change in
732 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
733 that FMR support is only used in experimental versions of the
734 o2iblnd - this change does not affect standard usage at all.
736 Severity : enhancement
738 Description: new ko2iblnd module parameter: ib_mtu
739 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
740 HCAs. You can avoid this problem by setting the MTU to 1024
741 using this module parameter.
743 Severity : enhancement
744 Bugzilla : 11118/11620
745 Description: ptllnd small request message buffer alignment fix
746 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
747 Round up small message size on sends in case this option
748 is not supported. 11620 was a defect in the initial
749 implementation which effectively asserted all peers had to be
750 running the correct protocol version which was fixed by always
751 NAK-ing such requests and handling any misalignments they
756 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
757 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
758 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
760 Severity : enhancement
762 Description: Patchless ZC(zero copy) socklnd
763 Details : New protocol for socklnd, socklnd can support zero copy without
764 kernel patch, it's compatible with old socklnd. Checksum is
765 moved from tunables to modparams.
769 Description: When ksocknal_del_peer() is called upon a peer whose
770 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
771 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
774 Frequency : when ptlrpc is under heavy use and runs out of request buffer
776 Description: In lnet_match_blocked_msg(), md can be used without holding a
780 Frequency : very rarely
782 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
783 If connd connects a route which has been closed by
784 ksocknal_shutdown(), ksocknal_create_routes() may create new
785 routes which hold references on the peer, causing shutdown
786 process to wait for peer to disappear forever.
788 Severity : enhancement
790 Description: Dump XT3 portals traces on kptllnd timeout
791 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
792 dump Cray portals debug traces to a file. The kptllnd module
793 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
794 is the basename of the dump file.
797 Frequency : infrequent
799 Description: kernel ptllnd fix bug in connection re-establishment
800 Details : Kernel ptllnd could produce protocol errors e.g. illegal
801 matchbits and/or violate the credit flow protocol when trying
802 to re-establish a connection with a peer after an error or
805 Severity : enhancement
807 Description: Allow /proc/sys/lnet/debug to be set symbolically
808 Details : Allow debug and subsystem debug values to be read/set by name
809 in addition to numerically, for ease of use.
812 Frequency : only in configurations with LNET routers
814 Description: routes automatically marked down and recovered
815 Details : In configurations with LNET routers if a router fails routers
816 now actively try to recover routes that are down, unless they
817 are marked down by an administrator.
819 ------------------------------------------------------------------------------
821 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
824 Frequency : very rarely, in configurations with LNET routers and TCP
826 Description: incorrect data written to files on OSTs
827 Details : In certain high-load conditions incorrect data may be written
828 to files on the OST when using TCP networks.
830 ------------------------------------------------------------------------------
832 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
834 - rework CDEBUG messages rate-limiting mechanism b=10375
835 - add per-socket tunables for socklnd if the kernel is patched b=10327
837 ------------------------------------------------------------------------------
839 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
841 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
842 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
843 - quiet LNET startup LNI message for liblustre b=10128
844 - Better console error messages if 'ip2nets' can't match an IP address
845 - Fixed overflow/use-before-set bugs in linux-time.h
846 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
847 - LNET teardown failed an assertion about the route table being empty
848 - Fixed a crash in LNetEQPoll(<invalid handle>)
849 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
850 - improve debug message for liblustre/Catamount nodes (b=10116)
852 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
853 * Configuration change for the XT3
854 The PTLLND is now used to run Lustre over Portals on the XT3.
855 The configure option(s) --with-cray-portals are no longer
856 used. Rather --with-portals=<path-to-portals-includes> is
857 used to enable building on the XT3. In addition to enable
858 XT3 specific features the option --enable-cray-xt3 must be
861 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
862 * Portals has been removed, replaced by LNET.
863 LNET is new networking infrastructure for Lustre, it includes a
864 reorganized network configuration mode (see the user
865 documentation for full details) as well as support for routing
866 between different network fabrics. Lustre Networking Devices
867 (LNDS) for the supported network fabrics have also been created
868 for this new infrastructure.
870 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
875 Frequency : rare (large Voltaire clusters only)
877 Description: the default number of reserved transmit descriptors was too low
878 for some large clusters
879 Details : As a workaround, the number was increased. A proper fix includes
882 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
887 Frequency : occasional (large-scale events, cluster reboot, network failure)
889 Description: too many error messages on console obscure actual problem and
890 can slow down/panic server, or cause recovery to fail repeatedly
891 Details : enable rate-limiting of console error messages, and some messages
892 that were console errors now only go to the kernel log
894 Severity : enhancement
896 Description: add /proc/sys/portals/catastrophe entry which will report if
897 that node has previously LBUGged
899 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
901 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
903 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
905 - handle error return code in kranal_check_fma_rx() (5915,6054)
907 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
909 - update vibnal (Voltaire IB NAL)
910 - update gmnal (Myrinet NAL), gmnalid
912 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
914 * Landed portals:b_port_step as follows...
916 - removed CFS_DECL_SPIN*
917 just use 'spinlock_t' and initialise with spin_lock_init()
919 - removed CFS_DECL_MUTEX*
920 just use 'struct semaphore' and initialise with init_mutex()
922 - removed CFS_DECL_RWSEM*
923 just use 'struct rw_semaphore' and initialise with init_rwsem()
925 - renamed cfs_sleep_chan -> cfs_waitq
926 cfs_sleep_link -> cfs_waitlink
928 - fixed race in linux version of arch-independent socknal
929 (the ENOMEM/EAGAIN decision).
931 - Didn't fix problems in Darwin version of arch-independent socknal
932 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
934 - removed libcfs types from non-socknal header files (only some types
935 in the header files had been changed; the .c files hadn't been