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
22 Description: router checker stops working when system wall clock goes backward
23 Details : use monotonic timing source instead of system wall clock time.
25 Severity : enhancement
27 Description: avoid asymmetrical router failures
29 Severity : enhancement
31 Description: multiple-instance support for kptllnd
35 Description: ksocknal_close_conn_locked connection race
36 Details : A race was possible when ksocknal_create_conn calls
37 ksocknal_close_conn_locked for already closed conn.
41 Description: router_proc.c is rewritten to use sysctl-interface for parameters
42 residing in /proc/sys/lnet
44 Severity : enhancement
46 Description: port router pinger to userspace
50 Description: kptllnd HELLO protocol deadlock
51 Details : kptllnd HELLO protocol doesn't run to completion in finite time
55 Description: LNet selftest fixes and enhancements
57 Severity : enhancement
59 Description: allow a test node to be a member of multiple test groups
61 Severity : enhancement
63 Description: MXLND: eliminate hosts file, use arp for peer nic_id resolution
64 Details : an update from the upstream developer Scott Atchley.
66 Severity : enhancement
68 Description: add a new LND optiion to control peer buffer credits on routers
72 Description: Fixing deadlock in usocklnd
73 Details : A deadlock was possible in usocklnd due to race condition while
74 tearing connection down. The problem resulted from erroneous
75 assumption that lnet_finalize() could have been called holding
79 Bugzilla : 13621, 15983
80 Description: Protocol V2 of o2iblnd
81 Details : o2iblnd V2 has several new features:
82 . map-on-demand: map-on-demand is disabled by default, it can
83 be enabled by using modparam "map_on_demand=@value@", @value@
84 should >= 0 and < 256, 0 will disable map-on-demand, any other
85 valid value will enable map-on-demand.
86 Oi2blnd will create FMR or physical MR for RDMA if fragments of
88 Enable map-on-demand will take less memory for new connection,
89 but a little more CPU for RDMA.
90 . iWARP : to support iWARP, please enable map-on-demand, 32 and 64
91 are recommanded value. iWARP will probably fail for value >=128.
92 . OOB NOOP message: to resolve deadlock on router.
93 . tunable peer_credits_hiw: (high water to return credits),
94 default value of peer_credits_hiw equals to (peer_credits -1),
95 user can change it between peer_credits/2 and (peer_credits - 1).
96 Lower value is recommended for high latency network.
97 . tunable message queue size: it always equals to peer_credits,
98 higher value is recommended for high latency network.
99 . It's compatible with earlier version of o2iblnd
103 Description: Fixing 'running out of ports' issue
104 Details : Add a delay before next reconnect attempt in ksocklnd in
105 the case of lost race. Limit the frequency of query-requests
106 in lnet. Improved handling of 'dead peer' notifications in
111 Description: Change ptllnd timeout and watchdog timers
112 Details : Add ptltrace_on_nal_failed and bump ptllnd timeout to match
113 Portals wire timeout.
117 Description: One down Lustre FS hangs ALL mounted Lustre filesystems
118 Details : Shared routing enhancements - peer health detection.
120 Severity : enhancement
122 Description: acceptor.c cleanup
123 Details : Code duplication in acceptor.c for the cases of kernel and
124 user-space removed. User-space libcfs tcpip primitives
125 uniformed to have prototypes similar to kernel ones. Minor
126 cosmetic changes in usocklnd to use cfs_socket_t as
127 representation of socket.
131 Description: IB path MTU mistakenly set to 1st path MTU when ib_mtu is off
132 Details : See comment 46 in bug 11245 for details - it's indeed a bug
133 introduced by the original 11245 fix.
137 Description: uptllnd credit overflow fix
138 Details : kptl_msg_t::ptlm_credits could be overflown by uptllnd since
143 Description: socklnd protocol version 3
144 Details : With current protocol V2, connections on router can be
145 blocked and can't receive any incoming messages when there is no
146 more router buffer, so ZC-ACK can't be handled (LNet message
147 can't be finalized) and will cause deadlock on router.
148 Protocol V3 has a dedicated connection for emergency messages
149 like ZC-ACK to router, messages on this dedicated connection
150 don't need any credit so will never be blocked. Also, V3 can send
151 keepalive ping in specified period for router healthy checking.
153 -------------------------------------------------------------------------------
155 12-31-2008 Sun Microsystems, Inc.
157 * Support for networks:
158 socklnd - any kernel supported by Lustre,
159 qswlnd - Qsnet kernel modules 5.20 and later,
160 openiblnd - IbGold 1.8.2,
161 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
162 viblnd - Voltaire ibhost 3.4.5 and later,
163 ciblnd - Topspin 3.2.0,
164 iiblnd - Infiniserv 3.3 + PathBits patch,
165 gmlnd - GM 2.1.22 and later,
166 mxlnd - MX 1.2.1 or later,
167 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
171 Description: workaround for OOM from o2iblnd
172 Details : OFED needs allocate big chunk of memory for QP while creating
173 connection for o2iblnd, OOM can happen if no such a contiguous
175 QP size is decided by concurrent_sends and max_fragments of
176 o2iblnd, now we permit user to specify smaller value for
177 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
178 will decrease memory block size required by creating QP.
182 Description: Support Zerocopy receive of Chelsio device
183 Details : Chelsio driver can support zerocopy for iov[1] if it's
184 contiguous and large enough.
188 Description: fix credit flow deadlock in uptllnd
192 Description: finalize network operation in reasonable time
193 Details : conf-sanity test_32a couldn't stop ost and mds because it
194 tried to access non-existent peer and tcp connect took
195 quite long before timing out.
199 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
200 Details : Lost reference on conn prevents peer from being destroyed, which
201 could prevent new peer creation if peer count has reached upper
206 Description: LNET Selftest results in Soft lockup on OSS CPU
207 Details : only hits when 8 or more o2ib clients involved and a session is
208 torn down with 'lst end_session' without preceeding 'lst stop'.
212 Description: concurrent_sends in IB LNDs should not be changeable at run time
213 Details : concurrent_sends in IB LNDs should not be changeable at run time
217 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
218 Details : only hits under out-of-memory situations
221 -------------------------------------------------------------------------------
223 2009-02-07 Sun Microsystems, Inc.
225 * Support for networks:
226 socklnd - any kernel supported by Lustre,
227 qswlnd - Qsnet kernel modules 5.20 and later,
228 openiblnd - IbGold 1.8.2,
229 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
230 viblnd - Voltaire ibhost 3.4.5 and later,
231 ciblnd - Topspin 3.2.0,
232 iiblnd - Infiniserv 3.3 + PathBits patch,
233 gmlnd - GM 2.1.22 and later,
234 mxlnd - MX 1.2.1 or later,
235 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
238 Description: workaround for OOM from o2iblnd
239 Details : OFED needs allocate big chunk of memory for QP while creating
240 connection for o2iblnd, OOM can happen if no such a contiguous
242 QP size is decided by concurrent_sends and max_fragments of
243 o2iblnd, now we permit user to specify smaller value for
244 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
245 will decrease memory block size required by creating QP.
249 Description: Support Zerocopy receive of Chelsio device
250 Details : Chelsio driver can support zerocopy for iov[1] if it's
251 contiguous and large enough.
254 Description: fix credit flow deadlock in uptllnd
258 Description: finalize network operation in reasonable time
259 Details : conf-sanity test_32a couldn't stop ost and mds because it
260 tried to access non-existent peer and tcp connect took
261 quite long before timing out.
265 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
266 Details : Lost reference on conn prevents peer from being destroyed, which
267 could prevent new peer creation if peer count has reached upper
272 Description: LNET Selftest results in Soft lockup on OSS CPU
273 Details : only hits when 8 or more o2ib clients involved and a session is
274 torn down with 'lst end_session' without preceeding 'lst stop'.
278 Description: concurrent_sends in IB LNDs should not be changeable at run time
279 Details : concurrent_sends in IB LNDs should not be changeable at run time
281 -------------------------------------------------------------------------------
283 11-03-2008 Sun Microsystems, Inc.
285 * Support for networks:
286 socklnd - any kernel supported by Lustre,
287 qswlnd - Qsnet kernel modules 5.20 and later,
288 openiblnd - IbGold 1.8.2,
289 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
290 viblnd - Voltaire ibhost 3.4.5 and later,
291 ciblnd - Topspin 3.2.0,
292 iiblnd - Infiniserv 3.3 + PathBits patch,
293 gmlnd - GM 2.1.22 and later,
294 mxlnd - MX 1.2.1 or later,
295 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
299 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
300 Details : only hits under out-of-memory situations
302 -------------------------------------------------------------------------------
304 04-26-2008 Sun Microsystems, Inc.
306 * Support for networks:
307 socklnd - any kernel supported by Lustre,
308 qswlnd - Qsnet kernel modules 5.20 and later,
309 openiblnd - IbGold 1.8.2,
310 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
311 viblnd - Voltaire ibhost 3.4.5 and later,
312 ciblnd - Topspin 3.2.0,
313 iiblnd - Infiniserv 3.3 + PathBits patch,
314 gmlnd - GM 2.1.22 and later,
315 mxlnd - MX 1.2.1 or later,
316 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
320 Description: excessive debug information removed
321 Details : excessive debug information removed
325 Description: ksocknal_create_conn() hit ASSERTION during connection race
326 Details : ksocknal_create_conn() hit ASSERTION during connection race
330 Description: ksocknal_send_hello() hit ASSERTION while connecting race
331 Details : ksocknal_send_hello() hit ASSERTION while connecting race
335 Description: o2iblnd/ptllnd credit deadlock in a routed config.
336 Details : o2iblnd/ptllnd credit deadlock in a routed config.
340 Description: High load after starting lnet
341 Details : gmlnd should sleep in rx thread in interruptible way. Otherwise,
342 uptime utility reports high load that looks confusingly.
346 Description: ksocklnd fails to establish connection if accept_port is high
347 Details : PID remapping must not be done for active (outgoing) connections
350 --------------------------------------------------------------------------------
352 2008-01-11 Sun Microsystems, Inc.
354 * Support for networks:
355 socklnd - any kernel supported by Lustre,
356 qswlnd - Qsnet kernel modules 5.20 and later,
357 openiblnd - IbGold 1.8.2,
358 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
359 viblnd - Voltaire ibhost 3.4.5 and later,
360 ciblnd - Topspin 3.2.0,
361 iiblnd - Infiniserv 3.3 + PathBits patch,
362 gmlnd - GM 2.1.22 and later,
363 mxlnd - MX 1.2.1 or later,
364 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
368 Description: liblustre network error
369 Details : liblustre clients should understand LNET_ACCEPT_PORT environment
370 variable even if they don't start lnet acceptor.
374 Description: Strange message from lnet (Ignoring prediction from the future)
375 Details : Incorrect calculation of peer's last_alive value in ksocklnd
377 --------------------------------------------------------------------------------
379 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
381 * Support for networks:
382 socklnd - any kernel supported by Lustre,
383 qswlnd - Qsnet kernel modules 5.20 and later,
384 openiblnd - IbGold 1.8.2,
385 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5.
386 viblnd - Voltaire ibhost 3.4.5 and later,
387 ciblnd - Topspin 3.2.0,
388 iiblnd - Infiniserv 3.3 + PathBits patch,
389 gmlnd - GM 2.1.22 and later,
390 mxlnd - MX 1.2.1 or later,
391 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
395 Description: ASSERTION(me == md->md_me) failed in lnet_match_md()
399 Description: increase send queue size for ciblnd/openiblnd
403 Description: new userspace socklnd
404 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
405 with new one - usocklnd.
407 Severity : enhancement
409 Description: Console message flood
410 Details : Make cdls ratelimiting more tunable by adding several tunable in
411 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
412 /proc/sys/lnet/console_backoff.
414 --------------------------------------------------------------------------------
416 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
418 * Support for networks:
419 socklnd - any kernel supported by Lustre,
420 qswlnd - Qsnet kernel modules 5.20 and later,
421 openiblnd - IbGold 1.8.2,
422 o2iblnd - OFED 1.1 and 1.2,
423 viblnd - Voltaire ibhost 3.4.5 and later,
424 ciblnd - Topspin 3.2.0,
425 iiblnd - Infiniserv 3.3 + PathBits patch,
426 gmlnd - GM 2.1.22 and later,
427 mxlnd - MX 1.2.1 or later,
428 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
432 Description: /proc/sys/lnet has non-sysctl entries
433 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
437 Description: TOE Kernel panic by ksocklnd
438 Details : offloaded sockets provide their own implementation of sendpage,
439 can't call tcp_sendpage() directly
443 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
444 Details : races between lnd_shutdown and peer creation prevent
445 lnd_shutdown from finishing.
449 Description: open files rlimit 1024 reached while liblustre testing
450 Details : ulnds/socklnd must close open socket after unsuccessful
455 Description: build error
456 Details : fix typos in gmlnd, ptllnd and viblnd
458 --------------------------------------------------------------------------------
460 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
462 * Support for networks:
463 socklnd - kernels up to 2.6.16,
464 qswlnd - Qsnet kernel modules 5.20 and later,
465 openiblnd - IbGold 1.8.2,
466 o2iblnd - OFED 1.1 and 1.2
467 viblnd - Voltaire ibhost 3.4.5 and later,
468 ciblnd - Topspin 3.2.0,
469 iiblnd - Infiniserv 3.3 + PathBits patch,
470 gmlnd - GM 2.1.22 and later,
471 mxlnd - MX 1.2.1 or later,
472 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
474 --------------------------------------------------------------------------------
476 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
478 * Support for networks:
479 socklnd - kernels up to 2.6.16,
480 qswlnd - Qsnet kernel modules 5.20 and later,
481 openiblnd - IbGold 1.8.2,
483 viblnd - Voltaire ibhost 3.4.5 and later,
484 ciblnd - Topspin 3.2.0,
485 iiblnd - Infiniserv 3.3 + PathBits patch,
486 gmlnd - GM 2.1.22 and later,
487 mxlnd - MX 1.2.1 or later,
488 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
492 Description: Initialize cpumask before use
496 Description: ASSERTION failures when upgrading to the patchless zero-copy
498 Details : This bug affects "rolling upgrades", causing an inconsistent
499 protocol version negotiation and subsequent assertion failure
500 during rolling upgrades after the first wave of upgrades.
504 Details : Change "dropped message" CERRORs to D_NETERROR so they are
505 logged instead of creating "console chatter" when a lustre
506 timeout races with normal RPC completion.
509 Details : lnet_clear_peer_table can wait forever if user forgets to
513 Details : libcfs_id2str should check pid against LNET_PID_ANY.
517 Description: added LNET self test
518 Details : landing b_self_test
523 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
525 Details : do_div() macro is used incorrectly.
527 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
531 Description: make panic on lbug configurable
535 Description: Add OFED1.2 support to o2iblnd
536 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
537 are installed (other than kernel's in-tree infiniband), there
538 could be some problem while insmod o2iblnd (mismatch CRC of
540 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
541 this link provides solution:
542 https://bugs.openfabrics.org/show_bug.cgi?id=355
543 if extra Module.symvers is not supported in kernel, we will
544 have to run the script in bug 12316 to update
545 $LINUX/module.symvers before building o2iblnd.
546 More details about this are in bug 12316.
548 ------------------------------------------------------------------------------
550 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
551 * version 1.4.10 / 1.6.0
552 * Support for networks:
553 socklnd - kernels up to 2.6.16,
554 qswlnd - Qsnet kernel modules 5.20 and later,
555 openiblnd - IbGold 1.8.2,
557 viblnd - Voltaire ibhost 3.4.5 and later,
558 ciblnd - Topspin 3.2.0,
559 iiblnd - Infiniserv 3.3 + PathBits patch,
560 gmlnd - GM 2.1.22 and later,
561 mxlnd - MX 1.2.1 or later,
562 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
566 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
567 possibly accessed in kptllnd_shutdown. Ptllnd should init
568 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
572 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
576 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
580 Description: the_lnet.ln_finalizing was not set when the current thread is
581 about to complete messages. It only affects multi-threaded
587 Description: Changed the default kqswlnd ntxmsg=512
592 Description: Assertion failure in kernel ptllnd caused by posting passive
593 bulk buffers before connection establishment complete.
598 Description: A race in kernel ptllnd between deleting a peer and posting
599 new communications for it could hang communications -
600 manifesting as "Unexpectedly long timeout" messages.
605 Description: Kernel ptllnd lock ordering issue could hang a node.
610 Description: node crash on socket teardown race
613 Frequency : 'lctl peer_list' issued on a mx net
615 Description: Enable lctl's peer_list for MXLND
618 Frequency : after Ptllnd timeouts and portals congestion
620 Description: Credit overflows
621 Details : This was a bug in ptllnd connection establishment. The fix
622 implements better peer stamps to disambiguate connection
623 establishment and ensure both peers enter the credit flow
624 state machine consistently.
629 Description: kptllnd didn't propagate some network errors up to LNET
630 Details : This bug was spotted while investigating 11394. The fix
631 ensures network errors on sends and bulk transfers are
632 propagated to LNET/lustre correctly.
634 Severity : enhancement
636 Description: Fixed console chatter in case of -ETIMEDOUT.
638 Severity : enhancement
640 Description: Added D_NETTRACE for recording network packet history
641 (initially only for ptllnd). Also a separate userspace
642 ptllnd facility to gather history which should really be
643 covered by D_NETTRACE too, if only CDEBUG recorded history in
649 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
650 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
651 callback can occur before a connection has actually been
652 established. This caused an assertion failure previously.
654 Severity : enhancement
656 Description: Multiple instances for o2iblnd
657 Details : Allow multiple instances of o2iblnd to enable networking over
658 multiple HCAs and routing between them.
662 Description: lnet deadlock in router_checker
663 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
664 into BH locks to eliminate potential deadlock caused by
665 ksocknal_data_ready() preempting code holding these locks.
669 Description: Millions of failed socklnd connection attempts cause a very slow FS
670 Details : added a new route flag ksnr_scheduled to distinguish from
671 ksnr_connecting, so that a peer connection request is only turned
672 down for race concerns when an active connection to the same peer
673 is under progress (instead of just being scheduled).
675 ------------------------------------------------------------------------------
677 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
679 * Support for networks:
680 socklnd - kernels up to 2.6.16
681 qswlnd - Qsnet kernel modules 5.20 and later
682 openiblnd - IbGold 1.8.2
684 viblnd - Voltaire ibhost 3.4.5 and later
685 ciblnd - Topspin 3.2.0
686 iiblnd - Infiniserv 3.3 + PathBits patch
687 gmlnd - GM 2.1.22 and later
688 mxlnd - MX 1.2.1 or later
689 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
692 Severity : major on XT3
694 Description: libcfs overwrites /proc/sys/portals
695 Details : libcfs created a symlink from /proc/sys/portals to
696 /proc/sys/lnet for backwards compatibility. This is no
697 longer required and makes the Cray portals /proc variables
702 Description: OFED FMR API change
703 Details : This changes parameter usage to reflect a change in
704 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
705 that FMR support is only used in experimental versions of the
706 o2iblnd - this change does not affect standard usage at all.
708 Severity : enhancement
710 Description: new ko2iblnd module parameter: ib_mtu
711 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
712 HCAs. You can avoid this problem by setting the MTU to 1024
713 using this module parameter.
715 Severity : enhancement
716 Bugzilla : 11118/11620
717 Description: ptllnd small request message buffer alignment fix
718 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
719 Round up small message size on sends in case this option
720 is not supported. 11620 was a defect in the initial
721 implementation which effectively asserted all peers had to be
722 running the correct protocol version which was fixed by always
723 NAK-ing such requests and handling any misalignments they
728 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
729 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
730 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
732 Severity : enhancement
734 Description: Patchless ZC(zero copy) socklnd
735 Details : New protocol for socklnd, socklnd can support zero copy without
736 kernel patch, it's compatible with old socklnd. Checksum is
737 moved from tunables to modparams.
741 Description: When ksocknal_del_peer() is called upon a peer whose
742 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
743 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
746 Frequency : when ptlrpc is under heavy use and runs out of request buffer
748 Description: In lnet_match_blocked_msg(), md can be used without holding a
752 Frequency : very rarely
754 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
755 If connd connects a route which has been closed by
756 ksocknal_shutdown(), ksocknal_create_routes() may create new
757 routes which hold references on the peer, causing shutdown
758 process to wait for peer to disappear forever.
760 Severity : enhancement
762 Description: Dump XT3 portals traces on kptllnd timeout
763 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
764 dump Cray portals debug traces to a file. The kptllnd module
765 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
766 is the basename of the dump file.
769 Frequency : infrequent
771 Description: kernel ptllnd fix bug in connection re-establishment
772 Details : Kernel ptllnd could produce protocol errors e.g. illegal
773 matchbits and/or violate the credit flow protocol when trying
774 to re-establish a connection with a peer after an error or
777 Severity : enhancement
779 Description: Allow /proc/sys/lnet/debug to be set symbolically
780 Details : Allow debug and subsystem debug values to be read/set by name
781 in addition to numerically, for ease of use.
784 Frequency : only in configurations with LNET routers
786 Description: routes automatically marked down and recovered
787 Details : In configurations with LNET routers if a router fails routers
788 now actively try to recover routes that are down, unless they
789 are marked down by an administrator.
791 ------------------------------------------------------------------------------
793 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
796 Frequency : very rarely, in configurations with LNET routers and TCP
798 Description: incorrect data written to files on OSTs
799 Details : In certain high-load conditions incorrect data may be written
800 to files on the OST when using TCP networks.
802 ------------------------------------------------------------------------------
804 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
806 - rework CDEBUG messages rate-limiting mechanism b=10375
807 - add per-socket tunables for socklnd if the kernel is patched b=10327
809 ------------------------------------------------------------------------------
811 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
813 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
814 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
815 - quiet LNET startup LNI message for liblustre b=10128
816 - Better console error messages if 'ip2nets' can't match an IP address
817 - Fixed overflow/use-before-set bugs in linux-time.h
818 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
819 - LNET teardown failed an assertion about the route table being empty
820 - Fixed a crash in LNetEQPoll(<invalid handle>)
821 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
822 - improve debug message for liblustre/Catamount nodes (b=10116)
824 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
825 * Configuration change for the XT3
826 The PTLLND is now used to run Lustre over Portals on the XT3.
827 The configure option(s) --with-cray-portals are no longer
828 used. Rather --with-portals=<path-to-portals-includes> is
829 used to enable building on the XT3. In addition to enable
830 XT3 specific features the option --enable-cray-xt3 must be
833 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
834 * Portals has been removed, replaced by LNET.
835 LNET is new networking infrastructure for Lustre, it includes a
836 reorganized network configuration mode (see the user
837 documentation for full details) as well as support for routing
838 between different network fabrics. Lustre Networking Devices
839 (LNDS) for the supported network fabrics have also been created
840 for this new infrastructure.
842 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
847 Frequency : rare (large Voltaire clusters only)
849 Description: the default number of reserved transmit descriptors was too low
850 for some large clusters
851 Details : As a workaround, the number was increased. A proper fix includes
854 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
859 Frequency : occasional (large-scale events, cluster reboot, network failure)
861 Description: too many error messages on console obscure actual problem and
862 can slow down/panic server, or cause recovery to fail repeatedly
863 Details : enable rate-limiting of console error messages, and some messages
864 that were console errors now only go to the kernel log
866 Severity : enhancement
868 Description: add /proc/sys/portals/catastrophe entry which will report if
869 that node has previously LBUGged
871 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
873 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
875 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
877 - handle error return code in kranal_check_fma_rx() (5915,6054)
879 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
881 - update vibnal (Voltaire IB NAL)
882 - update gmnal (Myrinet NAL), gmnalid
884 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
886 * Landed portals:b_port_step as follows...
888 - removed CFS_DECL_SPIN*
889 just use 'spinlock_t' and initialise with spin_lock_init()
891 - removed CFS_DECL_MUTEX*
892 just use 'struct semaphore' and initialise with init_mutex()
894 - removed CFS_DECL_RWSEM*
895 just use 'struct rw_semaphore' and initialise with init_rwsem()
897 - renamed cfs_sleep_chan -> cfs_waitq
898 cfs_sleep_link -> cfs_waitlink
900 - fixed race in linux version of arch-independent socknal
901 (the ENOMEM/EAGAIN decision).
903 - Didn't fix problems in Darwin version of arch-independent socknal
904 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
906 - removed libcfs types from non-socknal header files (only some types
907 in the header files had been changed; the .c files hadn't been