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
15 -------------------------------------------------------------------------------
17 2010-07-15 Sun Microsystems, Inc.
19 * Support for networks:
20 socklnd - any kernel supported by Lustre,
21 qswlnd - Qsnet kernel modules 5.20 and later,
22 openiblnd - IbGold 1.8.2,
23 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, 1.3, and 1.4.1
24 viblnd - Voltaire ibhost 3.4.5 and later,
25 ciblnd - Topspin 3.2.0,
26 iiblnd - Infiniserv 3.3 + PathBits patch,
27 gmlnd - GM 2.1.22 and later,
28 mxlnd - MX 1.2.10 or later,
29 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
33 Description: should update lp_alive for non-router peers
35 Severity : enhancement
37 Description: LNet router shuffler.
39 Severity : enhancement
41 Description: LNet fine grain routing support.
45 Description: router checker stops working when system wall clock goes backward
46 Details : use monotonic timing source instead of system wall clock time.
48 Severity : enhancement
50 Description: avoid asymmetrical router failures
52 Severity : enhancement
54 Description: multiple-instance support for kptllnd
58 Description: ksocknal_close_conn_locked connection race
59 Details : A race was possible when ksocknal_create_conn calls
60 ksocknal_close_conn_locked for already closed conn.
64 Description: router_proc.c is rewritten to use sysctl-interface for parameters
65 residing in /proc/sys/lnet
67 Severity : enhancement
69 Description: port router pinger to userspace
73 Description: kptllnd HELLO protocol deadlock
74 Details : kptllnd HELLO protocol doesn't run to completion in finite time
78 Description: LNet selftest fixes and enhancements
80 Severity : enhancement
82 Description: allow a test node to be a member of multiple test groups
84 Severity : enhancement
86 Description: MXLND: eliminate hosts file, use arp for peer nic_id resolution
87 Details : an update from the upstream developer Scott Atchley.
89 Severity : enhancement
91 Description: add a new LND optiion to control peer buffer credits on routers
95 Description: Fixing deadlock in usocklnd
96 Details : A deadlock was possible in usocklnd due to race condition while
97 tearing connection down. The problem resulted from erroneous
98 assumption that lnet_finalize() could have been called holding
102 Bugzilla : 13621, 15983
103 Description: Protocol V2 of o2iblnd
104 Details : o2iblnd V2 has several new features:
105 . map-on-demand: map-on-demand is disabled by default, it can
106 be enabled by using modparam "map_on_demand=@value@", @value@
107 should >= 0 and < 256, 0 will disable map-on-demand, any other
108 valid value will enable map-on-demand.
109 Oi2blnd will create FMR or physical MR for RDMA if fragments of
111 Enable map-on-demand will take less memory for new connection,
112 but a little more CPU for RDMA.
113 . iWARP : to support iWARP, please enable map-on-demand, 32 and 64
114 are recommanded value. iWARP will probably fail for value >=128.
115 . OOB NOOP message: to resolve deadlock on router.
116 . tunable peer_credits_hiw: (high water to return credits),
117 default value of peer_credits_hiw equals to (peer_credits -1),
118 user can change it between peer_credits/2 and (peer_credits - 1).
119 Lower value is recommended for high latency network.
120 . tunable message queue size: it always equals to peer_credits,
121 higher value is recommended for high latency network.
122 . It's compatible with earlier version of o2iblnd
126 Description: Fixing 'running out of ports' issue
127 Details : Add a delay before next reconnect attempt in ksocklnd in
128 the case of lost race. Limit the frequency of query-requests
129 in lnet. Improved handling of 'dead peer' notifications in
134 Description: Change ptllnd timeout and watchdog timers
135 Details : Add ptltrace_on_nal_failed and bump ptllnd timeout to match
136 Portals wire timeout.
140 Description: One down Lustre FS hangs ALL mounted Lustre filesystems
141 Details : Shared routing enhancements - peer health detection.
143 Severity : enhancement
145 Description: acceptor.c cleanup
146 Details : Code duplication in acceptor.c for the cases of kernel and
147 user-space removed. User-space libcfs tcpip primitives
148 uniformed to have prototypes similar to kernel ones. Minor
149 cosmetic changes in usocklnd to use cfs_socket_t as
150 representation of socket.
154 Description: IB path MTU mistakenly set to 1st path MTU when ib_mtu is off
155 Details : See comment 46 in bug 11245 for details - it's indeed a bug
156 introduced by the original 11245 fix.
160 Description: uptllnd credit overflow fix
161 Details : kptl_msg_t::ptlm_credits could be overflown by uptllnd since
166 Description: socklnd protocol version 3
167 Details : With current protocol V2, connections on router can be
168 blocked and can't receive any incoming messages when there is no
169 more router buffer, so ZC-ACK can't be handled (LNet message
170 can't be finalized) and will cause deadlock on router.
171 Protocol V3 has a dedicated connection for emergency messages
172 like ZC-ACK to router, messages on this dedicated connection
173 don't need any credit so will never be blocked. Also, V3 can send
174 keepalive ping in specified period for router healthy checking.
176 -------------------------------------------------------------------------------
178 12-31-2008 Sun Microsystems, Inc.
180 * Support for networks:
181 socklnd - any kernel supported by Lustre,
182 qswlnd - Qsnet kernel modules 5.20 and later,
183 openiblnd - IbGold 1.8.2,
184 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
185 viblnd - Voltaire ibhost 3.4.5 and later,
186 ciblnd - Topspin 3.2.0,
187 iiblnd - Infiniserv 3.3 + PathBits patch,
188 gmlnd - GM 2.1.22 and later,
189 mxlnd - MX 1.2.1 or later,
190 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
194 Description: workaround for OOM from o2iblnd
195 Details : OFED needs allocate big chunk of memory for QP while creating
196 connection for o2iblnd, OOM can happen if no such a contiguous
198 QP size is decided by concurrent_sends and max_fragments of
199 o2iblnd, now we permit user to specify smaller value for
200 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
201 will decrease memory block size required by creating QP.
205 Description: Support Zerocopy receive of Chelsio device
206 Details : Chelsio driver can support zerocopy for iov[1] if it's
207 contiguous and large enough.
211 Description: fix credit flow deadlock in uptllnd
215 Description: finalize network operation in reasonable time
216 Details : conf-sanity test_32a couldn't stop ost and mds because it
217 tried to access non-existent peer and tcp connect took
218 quite long before timing out.
222 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
223 Details : Lost reference on conn prevents peer from being destroyed, which
224 could prevent new peer creation if peer count has reached upper
229 Description: LNET Selftest results in Soft lockup on OSS CPU
230 Details : only hits when 8 or more o2ib clients involved and a session is
231 torn down with 'lst end_session' without preceeding 'lst stop'.
235 Description: concurrent_sends in IB LNDs should not be changeable at run time
236 Details : concurrent_sends in IB LNDs should not be changeable at run time
240 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
241 Details : only hits under out-of-memory situations
244 -------------------------------------------------------------------------------
246 2009-02-07 Sun Microsystems, Inc.
248 * Support for networks:
249 socklnd - any kernel supported by Lustre,
250 qswlnd - Qsnet kernel modules 5.20 and later,
251 openiblnd - IbGold 1.8.2,
252 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
253 viblnd - Voltaire ibhost 3.4.5 and later,
254 ciblnd - Topspin 3.2.0,
255 iiblnd - Infiniserv 3.3 + PathBits patch,
256 gmlnd - GM 2.1.22 and later,
257 mxlnd - MX 1.2.1 or later,
258 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
261 Description: workaround for OOM from o2iblnd
262 Details : OFED needs allocate big chunk of memory for QP while creating
263 connection for o2iblnd, OOM can happen if no such a contiguous
265 QP size is decided by concurrent_sends and max_fragments of
266 o2iblnd, now we permit user to specify smaller value for
267 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
268 will decrease memory block size required by creating QP.
272 Description: Support Zerocopy receive of Chelsio device
273 Details : Chelsio driver can support zerocopy for iov[1] if it's
274 contiguous and large enough.
277 Description: fix credit flow deadlock in uptllnd
281 Description: finalize network operation in reasonable time
282 Details : conf-sanity test_32a couldn't stop ost and mds because it
283 tried to access non-existent peer and tcp connect took
284 quite long before timing out.
288 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
289 Details : Lost reference on conn prevents peer from being destroyed, which
290 could prevent new peer creation if peer count has reached upper
295 Description: LNET Selftest results in Soft lockup on OSS CPU
296 Details : only hits when 8 or more o2ib clients involved and a session is
297 torn down with 'lst end_session' without preceeding 'lst stop'.
301 Description: concurrent_sends in IB LNDs should not be changeable at run time
302 Details : concurrent_sends in IB LNDs should not be changeable at run time
304 -------------------------------------------------------------------------------
306 11-03-2008 Sun Microsystems, Inc.
308 * Support for networks:
309 socklnd - any kernel supported by Lustre,
310 qswlnd - Qsnet kernel modules 5.20 and later,
311 openiblnd - IbGold 1.8.2,
312 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
313 viblnd - Voltaire ibhost 3.4.5 and later,
314 ciblnd - Topspin 3.2.0,
315 iiblnd - Infiniserv 3.3 + PathBits patch,
316 gmlnd - GM 2.1.22 and later,
317 mxlnd - MX 1.2.1 or later,
318 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
322 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
323 Details : only hits under out-of-memory situations
325 -------------------------------------------------------------------------------
327 04-26-2008 Sun Microsystems, Inc.
329 * Support for networks:
330 socklnd - any kernel supported by Lustre,
331 qswlnd - Qsnet kernel modules 5.20 and later,
332 openiblnd - IbGold 1.8.2,
333 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
334 viblnd - Voltaire ibhost 3.4.5 and later,
335 ciblnd - Topspin 3.2.0,
336 iiblnd - Infiniserv 3.3 + PathBits patch,
337 gmlnd - GM 2.1.22 and later,
338 mxlnd - MX 1.2.1 or later,
339 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
343 Description: excessive debug information removed
344 Details : excessive debug information removed
348 Description: ksocknal_create_conn() hit ASSERTION during connection race
349 Details : ksocknal_create_conn() hit ASSERTION during connection race
353 Description: ksocknal_send_hello() hit ASSERTION while connecting race
354 Details : ksocknal_send_hello() hit ASSERTION while connecting race
358 Description: o2iblnd/ptllnd credit deadlock in a routed config.
359 Details : o2iblnd/ptllnd credit deadlock in a routed config.
363 Description: High load after starting lnet
364 Details : gmlnd should sleep in rx thread in interruptible way. Otherwise,
365 uptime utility reports high load that looks confusingly.
369 Description: ksocklnd fails to establish connection if accept_port is high
370 Details : PID remapping must not be done for active (outgoing) connections
373 --------------------------------------------------------------------------------
375 2008-01-11 Sun Microsystems, Inc.
377 * Support for networks:
378 socklnd - any kernel supported by Lustre,
379 qswlnd - Qsnet kernel modules 5.20 and later,
380 openiblnd - IbGold 1.8.2,
381 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
382 viblnd - Voltaire ibhost 3.4.5 and later,
383 ciblnd - Topspin 3.2.0,
384 iiblnd - Infiniserv 3.3 + PathBits patch,
385 gmlnd - GM 2.1.22 and later,
386 mxlnd - MX 1.2.1 or later,
387 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
391 Description: liblustre network error
392 Details : liblustre clients should understand LNET_ACCEPT_PORT environment
393 variable even if they don't start lnet acceptor.
397 Description: Strange message from lnet (Ignoring prediction from the future)
398 Details : Incorrect calculation of peer's last_alive value in ksocklnd
400 --------------------------------------------------------------------------------
402 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
404 * Support for networks:
405 socklnd - any kernel supported by Lustre,
406 qswlnd - Qsnet kernel modules 5.20 and later,
407 openiblnd - IbGold 1.8.2,
408 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5.
409 viblnd - Voltaire ibhost 3.4.5 and later,
410 ciblnd - Topspin 3.2.0,
411 iiblnd - Infiniserv 3.3 + PathBits patch,
412 gmlnd - GM 2.1.22 and later,
413 mxlnd - MX 1.2.1 or later,
414 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
418 Description: ASSERTION(me == md->md_me) failed in lnet_match_md()
422 Description: increase send queue size for ciblnd/openiblnd
426 Description: new userspace socklnd
427 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
428 with new one - usocklnd.
430 Severity : enhancement
432 Description: Console message flood
433 Details : Make cdls ratelimiting more tunable by adding several tunable in
434 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
435 /proc/sys/lnet/console_backoff.
437 --------------------------------------------------------------------------------
439 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
441 * Support for networks:
442 socklnd - any kernel supported by Lustre,
443 qswlnd - Qsnet kernel modules 5.20 and later,
444 openiblnd - IbGold 1.8.2,
445 o2iblnd - OFED 1.1 and 1.2,
446 viblnd - Voltaire ibhost 3.4.5 and later,
447 ciblnd - Topspin 3.2.0,
448 iiblnd - Infiniserv 3.3 + PathBits patch,
449 gmlnd - GM 2.1.22 and later,
450 mxlnd - MX 1.2.1 or later,
451 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
455 Description: /proc/sys/lnet has non-sysctl entries
456 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
460 Description: TOE Kernel panic by ksocklnd
461 Details : offloaded sockets provide their own implementation of sendpage,
462 can't call tcp_sendpage() directly
466 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
467 Details : races between lnd_shutdown and peer creation prevent
468 lnd_shutdown from finishing.
472 Description: open files rlimit 1024 reached while liblustre testing
473 Details : ulnds/socklnd must close open socket after unsuccessful
478 Description: build error
479 Details : fix typos in gmlnd, ptllnd and viblnd
481 --------------------------------------------------------------------------------
483 2007-07-30 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,
489 o2iblnd - OFED 1.1 and 1.2
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
497 --------------------------------------------------------------------------------
499 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
501 * Support for networks:
502 socklnd - kernels up to 2.6.16,
503 qswlnd - Qsnet kernel modules 5.20 and later,
504 openiblnd - IbGold 1.8.2,
506 viblnd - Voltaire ibhost 3.4.5 and later,
507 ciblnd - Topspin 3.2.0,
508 iiblnd - Infiniserv 3.3 + PathBits patch,
509 gmlnd - GM 2.1.22 and later,
510 mxlnd - MX 1.2.1 or later,
511 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
515 Description: Initialize cpumask before use
519 Description: ASSERTION failures when upgrading to the patchless zero-copy
521 Details : This bug affects "rolling upgrades", causing an inconsistent
522 protocol version negotiation and subsequent assertion failure
523 during rolling upgrades after the first wave of upgrades.
527 Details : Change "dropped message" CERRORs to D_NETERROR so they are
528 logged instead of creating "console chatter" when a lustre
529 timeout races with normal RPC completion.
532 Details : lnet_clear_peer_table can wait forever if user forgets to
536 Details : libcfs_id2str should check pid against LNET_PID_ANY.
540 Description: added LNET self test
541 Details : landing b_self_test
546 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
548 Details : do_div() macro is used incorrectly.
550 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
554 Description: make panic on lbug configurable
558 Description: Add OFED1.2 support to o2iblnd
559 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
560 are installed (other than kernel's in-tree infiniband), there
561 could be some problem while insmod o2iblnd (mismatch CRC of
563 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
564 this link provides solution:
565 https://bugs.openfabrics.org/show_bug.cgi?id=355
566 if extra Module.symvers is not supported in kernel, we will
567 have to run the script in bug 12316 to update
568 $LINUX/module.symvers before building o2iblnd.
569 More details about this are in bug 12316.
571 ------------------------------------------------------------------------------
573 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
574 * version 1.4.10 / 1.6.0
575 * Support for networks:
576 socklnd - kernels up to 2.6.16,
577 qswlnd - Qsnet kernel modules 5.20 and later,
578 openiblnd - IbGold 1.8.2,
580 viblnd - Voltaire ibhost 3.4.5 and later,
581 ciblnd - Topspin 3.2.0,
582 iiblnd - Infiniserv 3.3 + PathBits patch,
583 gmlnd - GM 2.1.22 and later,
584 mxlnd - MX 1.2.1 or later,
585 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
589 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
590 possibly accessed in kptllnd_shutdown. Ptllnd should init
591 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
595 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
599 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
603 Description: the_lnet.ln_finalizing was not set when the current thread is
604 about to complete messages. It only affects multi-threaded
610 Description: Changed the default kqswlnd ntxmsg=512
615 Description: Assertion failure in kernel ptllnd caused by posting passive
616 bulk buffers before connection establishment complete.
621 Description: A race in kernel ptllnd between deleting a peer and posting
622 new communications for it could hang communications -
623 manifesting as "Unexpectedly long timeout" messages.
628 Description: Kernel ptllnd lock ordering issue could hang a node.
633 Description: node crash on socket teardown race
636 Frequency : 'lctl peer_list' issued on a mx net
638 Description: Enable lctl's peer_list for MXLND
641 Frequency : after Ptllnd timeouts and portals congestion
643 Description: Credit overflows
644 Details : This was a bug in ptllnd connection establishment. The fix
645 implements better peer stamps to disambiguate connection
646 establishment and ensure both peers enter the credit flow
647 state machine consistently.
652 Description: kptllnd didn't propagate some network errors up to LNET
653 Details : This bug was spotted while investigating 11394. The fix
654 ensures network errors on sends and bulk transfers are
655 propagated to LNET/lustre correctly.
657 Severity : enhancement
659 Description: Fixed console chatter in case of -ETIMEDOUT.
661 Severity : enhancement
663 Description: Added D_NETTRACE for recording network packet history
664 (initially only for ptllnd). Also a separate userspace
665 ptllnd facility to gather history which should really be
666 covered by D_NETTRACE too, if only CDEBUG recorded history in
672 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
673 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
674 callback can occur before a connection has actually been
675 established. This caused an assertion failure previously.
677 Severity : enhancement
679 Description: Multiple instances for o2iblnd
680 Details : Allow multiple instances of o2iblnd to enable networking over
681 multiple HCAs and routing between them.
685 Description: lnet deadlock in router_checker
686 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
687 into BH locks to eliminate potential deadlock caused by
688 ksocknal_data_ready() preempting code holding these locks.
692 Description: Millions of failed socklnd connection attempts cause a very slow FS
693 Details : added a new route flag ksnr_scheduled to distinguish from
694 ksnr_connecting, so that a peer connection request is only turned
695 down for race concerns when an active connection to the same peer
696 is under progress (instead of just being scheduled).
698 ------------------------------------------------------------------------------
700 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
702 * Support for networks:
703 socklnd - kernels up to 2.6.16
704 qswlnd - Qsnet kernel modules 5.20 and later
705 openiblnd - IbGold 1.8.2
707 viblnd - Voltaire ibhost 3.4.5 and later
708 ciblnd - Topspin 3.2.0
709 iiblnd - Infiniserv 3.3 + PathBits patch
710 gmlnd - GM 2.1.22 and later
711 mxlnd - MX 1.2.1 or later
712 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
715 Severity : major on XT3
717 Description: libcfs overwrites /proc/sys/portals
718 Details : libcfs created a symlink from /proc/sys/portals to
719 /proc/sys/lnet for backwards compatibility. This is no
720 longer required and makes the Cray portals /proc variables
725 Description: OFED FMR API change
726 Details : This changes parameter usage to reflect a change in
727 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
728 that FMR support is only used in experimental versions of the
729 o2iblnd - this change does not affect standard usage at all.
731 Severity : enhancement
733 Description: new ko2iblnd module parameter: ib_mtu
734 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
735 HCAs. You can avoid this problem by setting the MTU to 1024
736 using this module parameter.
738 Severity : enhancement
739 Bugzilla : 11118/11620
740 Description: ptllnd small request message buffer alignment fix
741 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
742 Round up small message size on sends in case this option
743 is not supported. 11620 was a defect in the initial
744 implementation which effectively asserted all peers had to be
745 running the correct protocol version which was fixed by always
746 NAK-ing such requests and handling any misalignments they
751 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
752 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
753 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
755 Severity : enhancement
757 Description: Patchless ZC(zero copy) socklnd
758 Details : New protocol for socklnd, socklnd can support zero copy without
759 kernel patch, it's compatible with old socklnd. Checksum is
760 moved from tunables to modparams.
764 Description: When ksocknal_del_peer() is called upon a peer whose
765 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
766 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
769 Frequency : when ptlrpc is under heavy use and runs out of request buffer
771 Description: In lnet_match_blocked_msg(), md can be used without holding a
775 Frequency : very rarely
777 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
778 If connd connects a route which has been closed by
779 ksocknal_shutdown(), ksocknal_create_routes() may create new
780 routes which hold references on the peer, causing shutdown
781 process to wait for peer to disappear forever.
783 Severity : enhancement
785 Description: Dump XT3 portals traces on kptllnd timeout
786 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
787 dump Cray portals debug traces to a file. The kptllnd module
788 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
789 is the basename of the dump file.
792 Frequency : infrequent
794 Description: kernel ptllnd fix bug in connection re-establishment
795 Details : Kernel ptllnd could produce protocol errors e.g. illegal
796 matchbits and/or violate the credit flow protocol when trying
797 to re-establish a connection with a peer after an error or
800 Severity : enhancement
802 Description: Allow /proc/sys/lnet/debug to be set symbolically
803 Details : Allow debug and subsystem debug values to be read/set by name
804 in addition to numerically, for ease of use.
807 Frequency : only in configurations with LNET routers
809 Description: routes automatically marked down and recovered
810 Details : In configurations with LNET routers if a router fails routers
811 now actively try to recover routes that are down, unless they
812 are marked down by an administrator.
814 ------------------------------------------------------------------------------
816 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
819 Frequency : very rarely, in configurations with LNET routers and TCP
821 Description: incorrect data written to files on OSTs
822 Details : In certain high-load conditions incorrect data may be written
823 to files on the OST when using TCP networks.
825 ------------------------------------------------------------------------------
827 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
829 - rework CDEBUG messages rate-limiting mechanism b=10375
830 - add per-socket tunables for socklnd if the kernel is patched b=10327
832 ------------------------------------------------------------------------------
834 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
836 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
837 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
838 - quiet LNET startup LNI message for liblustre b=10128
839 - Better console error messages if 'ip2nets' can't match an IP address
840 - Fixed overflow/use-before-set bugs in linux-time.h
841 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
842 - LNET teardown failed an assertion about the route table being empty
843 - Fixed a crash in LNetEQPoll(<invalid handle>)
844 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
845 - improve debug message for liblustre/Catamount nodes (b=10116)
847 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
848 * Configuration change for the XT3
849 The PTLLND is now used to run Lustre over Portals on the XT3.
850 The configure option(s) --with-cray-portals are no longer
851 used. Rather --with-portals=<path-to-portals-includes> is
852 used to enable building on the XT3. In addition to enable
853 XT3 specific features the option --enable-cray-xt3 must be
856 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
857 * Portals has been removed, replaced by LNET.
858 LNET is new networking infrastructure for Lustre, it includes a
859 reorganized network configuration mode (see the user
860 documentation for full details) as well as support for routing
861 between different network fabrics. Lustre Networking Devices
862 (LNDS) for the supported network fabrics have also been created
863 for this new infrastructure.
865 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
870 Frequency : rare (large Voltaire clusters only)
872 Description: the default number of reserved transmit descriptors was too low
873 for some large clusters
874 Details : As a workaround, the number was increased. A proper fix includes
877 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
882 Frequency : occasional (large-scale events, cluster reboot, network failure)
884 Description: too many error messages on console obscure actual problem and
885 can slow down/panic server, or cause recovery to fail repeatedly
886 Details : enable rate-limiting of console error messages, and some messages
887 that were console errors now only go to the kernel log
889 Severity : enhancement
891 Description: add /proc/sys/portals/catastrophe entry which will report if
892 that node has previously LBUGged
894 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
896 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
898 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
900 - handle error return code in kranal_check_fma_rx() (5915,6054)
902 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
904 - update vibnal (Voltaire IB NAL)
905 - update gmnal (Myrinet NAL), gmnalid
907 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
909 * Landed portals:b_port_step as follows...
911 - removed CFS_DECL_SPIN*
912 just use 'spinlock_t' and initialise with spin_lock_init()
914 - removed CFS_DECL_MUTEX*
915 just use 'struct semaphore' and initialise with init_mutex()
917 - removed CFS_DECL_RWSEM*
918 just use 'struct rw_semaphore' and initialise with init_rwsem()
920 - renamed cfs_sleep_chan -> cfs_waitq
921 cfs_sleep_link -> cfs_waitlink
923 - fixed race in linux version of arch-independent socknal
924 (the ENOMEM/EAGAIN decision).
926 - Didn't fix problems in Darwin version of arch-independent socknal
927 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
929 - removed libcfs types from non-socknal header files (only some types
930 in the header files had been changed; the .c files hadn't been