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: LNet selftest fixes and enhancements
24 Severity : enhancement
26 Description: allow a test node to be a member of multiple test groups
28 Severity : enhancement
30 Description: MXLND: eliminate hosts file, use arp for peer nic_id resolution
31 Details : an update from the upstream developer Scott Atchley.
33 Severity : enhancement
35 Description: add a new LND optiion to control peer buffer credits on routers
39 Description: Fixing deadlock in usocklnd
40 Details : A deadlock was possible in usocklnd due to race condition while
41 tearing connection down. The problem resulted from erroneous
42 assumption that lnet_finalize() could have been called holding
46 Bugzilla : 13621, 15983
47 Description: Protocol V2 of o2iblnd
48 Details : o2iblnd V2 has several new features:
49 . map-on-demand: map-on-demand is disabled by default, it can
50 be enabled by using modparam "map_on_demand=@value@", @value@
51 should >= 0 and < 256, 0 will disable map-on-demand, any other
52 valid value will enable map-on-demand.
53 Oi2blnd will create FMR or physical MR for RDMA if fragments of
55 Enable map-on-demand will take less memory for new connection,
56 but a little more CPU for RDMA.
57 . iWARP : to support iWARP, please enable map-on-demand, 32 and 64
58 are recommanded value. iWARP will probably fail for value >=128.
59 . OOB NOOP message: to resolve deadlock on router.
60 . tunable peer_credits_hiw: (high water to return credits),
61 default value of peer_credits_hiw equals to (peer_credits -1),
62 user can change it between peer_credits/2 and (peer_credits - 1).
63 Lower value is recommended for high latency network.
64 . tunable message queue size: it always equals to peer_credits,
65 higher value is recommended for high latency network.
66 . It's compatible with earlier version of o2iblnd
70 Description: Fixing 'running out of ports' issue
71 Details : Add a delay before next reconnect attempt in ksocklnd in
72 the case of lost race. Limit the frequency of query-requests
73 in lnet. Improved handling of 'dead peer' notifications in
78 Description: Change ptllnd timeout and watchdog timers
79 Details : Add ptltrace_on_nal_failed and bump ptllnd timeout to match
84 Description: One down Lustre FS hangs ALL mounted Lustre filesystems
85 Details : Shared routing enhancements - peer health detection.
87 Severity : enhancement
89 Description: acceptor.c cleanup
90 Details : Code duplication in acceptor.c for the cases of kernel and
91 user-space removed. User-space libcfs tcpip primitives
92 uniformed to have prototypes similar to kernel ones. Minor
93 cosmetic changes in usocklnd to use cfs_socket_t as
94 representation of socket.
98 Description: IB path MTU mistakenly set to 1st path MTU when ib_mtu is off
99 Details : See comment 46 in bug 11245 for details - it's indeed a bug
100 introduced by the original 11245 fix.
104 Description: uptllnd credit overflow fix
105 Details : kptl_msg_t::ptlm_credits could be overflown by uptllnd since
110 Description: socklnd protocol version 3
111 Details : With current protocol V2, connections on router can be
112 blocked and can't receive any incoming messages when there is no
113 more router buffer, so ZC-ACK can't be handled (LNet message
114 can't be finalized) and will cause deadlock on router.
115 Protocol V3 has a dedicated connection for emergency messages
116 like ZC-ACK to router, messages on this dedicated connection
117 don't need any credit so will never be blocked. Also, V3 can send
118 keepalive ping in specified period for router healthy checking.
120 -------------------------------------------------------------------------------
122 12-31-2008 Sun Microsystems, Inc.
124 * Support for networks:
125 socklnd - any kernel supported by Lustre,
126 qswlnd - Qsnet kernel modules 5.20 and later,
127 openiblnd - IbGold 1.8.2,
128 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
129 viblnd - Voltaire ibhost 3.4.5 and later,
130 ciblnd - Topspin 3.2.0,
131 iiblnd - Infiniserv 3.3 + PathBits patch,
132 gmlnd - GM 2.1.22 and later,
133 mxlnd - MX 1.2.1 or later,
134 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
138 Description: workaround for OOM from o2iblnd
139 Details : OFED needs allocate big chunk of memory for QP while creating
140 connection for o2iblnd, OOM can happen if no such a contiguous
142 QP size is decided by concurrent_sends and max_fragments of
143 o2iblnd, now we permit user to specify smaller value for
144 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
145 will decrease memory block size required by creating QP.
149 Description: Support Zerocopy receive of Chelsio device
150 Details : Chelsio driver can support zerocopy for iov[1] if it's
151 contiguous and large enough.
155 Description: fix credit flow deadlock in uptllnd
159 Description: finalize network operation in reasonable time
160 Details : conf-sanity test_32a couldn't stop ost and mds because it
161 tried to access non-existent peer and tcp connect took
162 quite long before timing out.
166 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
167 Details : Lost reference on conn prevents peer from being destroyed, which
168 could prevent new peer creation if peer count has reached upper
173 Description: LNET Selftest results in Soft lockup on OSS CPU
174 Details : only hits when 8 or more o2ib clients involved and a session is
175 torn down with 'lst end_session' without preceeding 'lst stop'.
179 Description: concurrent_sends in IB LNDs should not be changeable at run time
180 Details : concurrent_sends in IB LNDs should not be changeable at run time
184 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
185 Details : only hits under out-of-memory situations
188 -------------------------------------------------------------------------------
190 2009-02-07 Sun Microsystems, Inc.
192 * Support for networks:
193 socklnd - any kernel supported by Lustre,
194 qswlnd - Qsnet kernel modules 5.20 and later,
195 openiblnd - IbGold 1.8.2,
196 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
197 viblnd - Voltaire ibhost 3.4.5 and later,
198 ciblnd - Topspin 3.2.0,
199 iiblnd - Infiniserv 3.3 + PathBits patch,
200 gmlnd - GM 2.1.22 and later,
201 mxlnd - MX 1.2.1 or later,
202 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
205 Description: workaround for OOM from o2iblnd
206 Details : OFED needs allocate big chunk of memory for QP while creating
207 connection for o2iblnd, OOM can happen if no such a contiguous
209 QP size is decided by concurrent_sends and max_fragments of
210 o2iblnd, now we permit user to specify smaller value for
211 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
212 will decrease memory block size required by creating QP.
216 Description: Support Zerocopy receive of Chelsio device
217 Details : Chelsio driver can support zerocopy for iov[1] if it's
218 contiguous and large enough.
221 Description: fix credit flow deadlock in uptllnd
225 Description: finalize network operation in reasonable time
226 Details : conf-sanity test_32a couldn't stop ost and mds because it
227 tried to access non-existent peer and tcp connect took
228 quite long before timing out.
232 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
233 Details : Lost reference on conn prevents peer from being destroyed, which
234 could prevent new peer creation if peer count has reached upper
239 Description: LNET Selftest results in Soft lockup on OSS CPU
240 Details : only hits when 8 or more o2ib clients involved and a session is
241 torn down with 'lst end_session' without preceeding 'lst stop'.
245 Description: concurrent_sends in IB LNDs should not be changeable at run time
246 Details : concurrent_sends in IB LNDs should not be changeable at run time
248 -------------------------------------------------------------------------------
250 11-03-2008 Sun Microsystems, Inc.
252 * Support for networks:
253 socklnd - any kernel supported by Lustre,
254 qswlnd - Qsnet kernel modules 5.20 and later,
255 openiblnd - IbGold 1.8.2,
256 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
257 viblnd - Voltaire ibhost 3.4.5 and later,
258 ciblnd - Topspin 3.2.0,
259 iiblnd - Infiniserv 3.3 + PathBits patch,
260 gmlnd - GM 2.1.22 and later,
261 mxlnd - MX 1.2.1 or later,
262 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
266 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
267 Details : only hits under out-of-memory situations
269 -------------------------------------------------------------------------------
271 04-26-2008 Sun Microsystems, Inc.
273 * Support for networks:
274 socklnd - any kernel supported by Lustre,
275 qswlnd - Qsnet kernel modules 5.20 and later,
276 openiblnd - IbGold 1.8.2,
277 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
278 viblnd - Voltaire ibhost 3.4.5 and later,
279 ciblnd - Topspin 3.2.0,
280 iiblnd - Infiniserv 3.3 + PathBits patch,
281 gmlnd - GM 2.1.22 and later,
282 mxlnd - MX 1.2.1 or later,
283 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
287 Description: excessive debug information removed
288 Details : excessive debug information removed
292 Description: ksocknal_create_conn() hit ASSERTION during connection race
293 Details : ksocknal_create_conn() hit ASSERTION during connection race
297 Description: ksocknal_send_hello() hit ASSERTION while connecting race
298 Details : ksocknal_send_hello() hit ASSERTION while connecting race
302 Description: o2iblnd/ptllnd credit deadlock in a routed config.
303 Details : o2iblnd/ptllnd credit deadlock in a routed config.
307 Description: High load after starting lnet
308 Details : gmlnd should sleep in rx thread in interruptible way. Otherwise,
309 uptime utility reports high load that looks confusingly.
313 Description: ksocklnd fails to establish connection if accept_port is high
314 Details : PID remapping must not be done for active (outgoing) connections
317 --------------------------------------------------------------------------------
319 2008-01-11 Sun Microsystems, Inc.
321 * Support for networks:
322 socklnd - any kernel supported by Lustre,
323 qswlnd - Qsnet kernel modules 5.20 and later,
324 openiblnd - IbGold 1.8.2,
325 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
326 viblnd - Voltaire ibhost 3.4.5 and later,
327 ciblnd - Topspin 3.2.0,
328 iiblnd - Infiniserv 3.3 + PathBits patch,
329 gmlnd - GM 2.1.22 and later,
330 mxlnd - MX 1.2.1 or later,
331 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
335 Description: liblustre network error
336 Details : liblustre clients should understand LNET_ACCEPT_PORT environment
337 variable even if they don't start lnet acceptor.
341 Description: Strange message from lnet (Ignoring prediction from the future)
342 Details : Incorrect calculation of peer's last_alive value in ksocklnd
344 --------------------------------------------------------------------------------
346 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
348 * Support for networks:
349 socklnd - any kernel supported by Lustre,
350 qswlnd - Qsnet kernel modules 5.20 and later,
351 openiblnd - IbGold 1.8.2,
352 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5.
353 viblnd - Voltaire ibhost 3.4.5 and later,
354 ciblnd - Topspin 3.2.0,
355 iiblnd - Infiniserv 3.3 + PathBits patch,
356 gmlnd - GM 2.1.22 and later,
357 mxlnd - MX 1.2.1 or later,
358 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
362 Description: ASSERTION(me == md->md_me) failed in lnet_match_md()
366 Description: increase send queue size for ciblnd/openiblnd
370 Description: new userspace socklnd
371 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
372 with new one - usocklnd.
374 Severity : enhancement
376 Description: Console message flood
377 Details : Make cdls ratelimiting more tunable by adding several tunable in
378 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
379 /proc/sys/lnet/console_backoff.
381 --------------------------------------------------------------------------------
383 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
385 * Support for networks:
386 socklnd - any kernel supported by Lustre,
387 qswlnd - Qsnet kernel modules 5.20 and later,
388 openiblnd - IbGold 1.8.2,
389 o2iblnd - OFED 1.1 and 1.2,
390 viblnd - Voltaire ibhost 3.4.5 and later,
391 ciblnd - Topspin 3.2.0,
392 iiblnd - Infiniserv 3.3 + PathBits patch,
393 gmlnd - GM 2.1.22 and later,
394 mxlnd - MX 1.2.1 or later,
395 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
399 Description: /proc/sys/lnet has non-sysctl entries
400 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
404 Description: TOE Kernel panic by ksocklnd
405 Details : offloaded sockets provide their own implementation of sendpage,
406 can't call tcp_sendpage() directly
410 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
411 Details : races between lnd_shutdown and peer creation prevent
412 lnd_shutdown from finishing.
416 Description: open files rlimit 1024 reached while liblustre testing
417 Details : ulnds/socklnd must close open socket after unsuccessful
422 Description: build error
423 Details : fix typos in gmlnd, ptllnd and viblnd
425 --------------------------------------------------------------------------------
427 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
429 * Support for networks:
430 socklnd - kernels up to 2.6.16,
431 qswlnd - Qsnet kernel modules 5.20 and later,
432 openiblnd - IbGold 1.8.2,
433 o2iblnd - OFED 1.1 and 1.2
434 viblnd - Voltaire ibhost 3.4.5 and later,
435 ciblnd - Topspin 3.2.0,
436 iiblnd - Infiniserv 3.3 + PathBits patch,
437 gmlnd - GM 2.1.22 and later,
438 mxlnd - MX 1.2.1 or later,
439 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
441 --------------------------------------------------------------------------------
443 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
445 * Support for networks:
446 socklnd - kernels up to 2.6.16,
447 qswlnd - Qsnet kernel modules 5.20 and later,
448 openiblnd - IbGold 1.8.2,
450 viblnd - Voltaire ibhost 3.4.5 and later,
451 ciblnd - Topspin 3.2.0,
452 iiblnd - Infiniserv 3.3 + PathBits patch,
453 gmlnd - GM 2.1.22 and later,
454 mxlnd - MX 1.2.1 or later,
455 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
459 Description: Initialize cpumask before use
463 Description: ASSERTION failures when upgrading to the patchless zero-copy
465 Details : This bug affects "rolling upgrades", causing an inconsistent
466 protocol version negotiation and subsequent assertion failure
467 during rolling upgrades after the first wave of upgrades.
471 Details : Change "dropped message" CERRORs to D_NETERROR so they are
472 logged instead of creating "console chatter" when a lustre
473 timeout races with normal RPC completion.
476 Details : lnet_clear_peer_table can wait forever if user forgets to
480 Details : libcfs_id2str should check pid against LNET_PID_ANY.
484 Description: added LNET self test
485 Details : landing b_self_test
490 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
492 Details : do_div() macro is used incorrectly.
494 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
498 Description: make panic on lbug configurable
502 Description: Add OFED1.2 support to o2iblnd
503 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
504 are installed (other than kernel's in-tree infiniband), there
505 could be some problem while insmod o2iblnd (mismatch CRC of
507 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
508 this link provides solution:
509 https://bugs.openfabrics.org/show_bug.cgi?id=355
510 if extra Module.symvers is not supported in kernel, we will
511 have to run the script in bug 12316 to update
512 $LINUX/module.symvers before building o2iblnd.
513 More details about this are in bug 12316.
515 ------------------------------------------------------------------------------
517 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
518 * version 1.4.10 / 1.6.0
519 * Support for networks:
520 socklnd - kernels up to 2.6.16,
521 qswlnd - Qsnet kernel modules 5.20 and later,
522 openiblnd - IbGold 1.8.2,
524 viblnd - Voltaire ibhost 3.4.5 and later,
525 ciblnd - Topspin 3.2.0,
526 iiblnd - Infiniserv 3.3 + PathBits patch,
527 gmlnd - GM 2.1.22 and later,
528 mxlnd - MX 1.2.1 or later,
529 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
533 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
534 possibly accessed in kptllnd_shutdown. Ptllnd should init
535 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
539 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
543 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
547 Description: the_lnet.ln_finalizing was not set when the current thread is
548 about to complete messages. It only affects multi-threaded
554 Description: Changed the default kqswlnd ntxmsg=512
559 Description: Assertion failure in kernel ptllnd caused by posting passive
560 bulk buffers before connection establishment complete.
565 Description: A race in kernel ptllnd between deleting a peer and posting
566 new communications for it could hang communications -
567 manifesting as "Unexpectedly long timeout" messages.
572 Description: Kernel ptllnd lock ordering issue could hang a node.
577 Description: node crash on socket teardown race
580 Frequency : 'lctl peer_list' issued on a mx net
582 Description: Enable lctl's peer_list for MXLND
585 Frequency : after Ptllnd timeouts and portals congestion
587 Description: Credit overflows
588 Details : This was a bug in ptllnd connection establishment. The fix
589 implements better peer stamps to disambiguate connection
590 establishment and ensure both peers enter the credit flow
591 state machine consistently.
596 Description: kptllnd didn't propagate some network errors up to LNET
597 Details : This bug was spotted while investigating 11394. The fix
598 ensures network errors on sends and bulk transfers are
599 propagated to LNET/lustre correctly.
601 Severity : enhancement
603 Description: Fixed console chatter in case of -ETIMEDOUT.
605 Severity : enhancement
607 Description: Added D_NETTRACE for recording network packet history
608 (initially only for ptllnd). Also a separate userspace
609 ptllnd facility to gather history which should really be
610 covered by D_NETTRACE too, if only CDEBUG recorded history in
616 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
617 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
618 callback can occur before a connection has actually been
619 established. This caused an assertion failure previously.
621 Severity : enhancement
623 Description: Multiple instances for o2iblnd
624 Details : Allow multiple instances of o2iblnd to enable networking over
625 multiple HCAs and routing between them.
629 Description: lnet deadlock in router_checker
630 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
631 into BH locks to eliminate potential deadlock caused by
632 ksocknal_data_ready() preempting code holding these locks.
636 Description: Millions of failed socklnd connection attempts cause a very slow FS
637 Details : added a new route flag ksnr_scheduled to distinguish from
638 ksnr_connecting, so that a peer connection request is only turned
639 down for race concerns when an active connection to the same peer
640 is under progress (instead of just being scheduled).
642 ------------------------------------------------------------------------------
644 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
646 * Support for networks:
647 socklnd - kernels up to 2.6.16
648 qswlnd - Qsnet kernel modules 5.20 and later
649 openiblnd - IbGold 1.8.2
651 viblnd - Voltaire ibhost 3.4.5 and later
652 ciblnd - Topspin 3.2.0
653 iiblnd - Infiniserv 3.3 + PathBits patch
654 gmlnd - GM 2.1.22 and later
655 mxlnd - MX 1.2.1 or later
656 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
659 Severity : major on XT3
661 Description: libcfs overwrites /proc/sys/portals
662 Details : libcfs created a symlink from /proc/sys/portals to
663 /proc/sys/lnet for backwards compatibility. This is no
664 longer required and makes the Cray portals /proc variables
669 Description: OFED FMR API change
670 Details : This changes parameter usage to reflect a change in
671 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
672 that FMR support is only used in experimental versions of the
673 o2iblnd - this change does not affect standard usage at all.
675 Severity : enhancement
677 Description: new ko2iblnd module parameter: ib_mtu
678 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
679 HCAs. You can avoid this problem by setting the MTU to 1024
680 using this module parameter.
682 Severity : enhancement
683 Bugzilla : 11118/11620
684 Description: ptllnd small request message buffer alignment fix
685 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
686 Round up small message size on sends in case this option
687 is not supported. 11620 was a defect in the initial
688 implementation which effectively asserted all peers had to be
689 running the correct protocol version which was fixed by always
690 NAK-ing such requests and handling any misalignments they
695 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
696 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
697 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
699 Severity : enhancement
701 Description: Patchless ZC(zero copy) socklnd
702 Details : New protocol for socklnd, socklnd can support zero copy without
703 kernel patch, it's compatible with old socklnd. Checksum is
704 moved from tunables to modparams.
708 Description: When ksocknal_del_peer() is called upon a peer whose
709 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
710 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
713 Frequency : when ptlrpc is under heavy use and runs out of request buffer
715 Description: In lnet_match_blocked_msg(), md can be used without holding a
719 Frequency : very rarely
721 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
722 If connd connects a route which has been closed by
723 ksocknal_shutdown(), ksocknal_create_routes() may create new
724 routes which hold references on the peer, causing shutdown
725 process to wait for peer to disappear forever.
727 Severity : enhancement
729 Description: Dump XT3 portals traces on kptllnd timeout
730 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
731 dump Cray portals debug traces to a file. The kptllnd module
732 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
733 is the basename of the dump file.
736 Frequency : infrequent
738 Description: kernel ptllnd fix bug in connection re-establishment
739 Details : Kernel ptllnd could produce protocol errors e.g. illegal
740 matchbits and/or violate the credit flow protocol when trying
741 to re-establish a connection with a peer after an error or
744 Severity : enhancement
746 Description: Allow /proc/sys/lnet/debug to be set symbolically
747 Details : Allow debug and subsystem debug values to be read/set by name
748 in addition to numerically, for ease of use.
751 Frequency : only in configurations with LNET routers
753 Description: routes automatically marked down and recovered
754 Details : In configurations with LNET routers if a router fails routers
755 now actively try to recover routes that are down, unless they
756 are marked down by an administrator.
758 ------------------------------------------------------------------------------
760 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
763 Frequency : very rarely, in configurations with LNET routers and TCP
765 Description: incorrect data written to files on OSTs
766 Details : In certain high-load conditions incorrect data may be written
767 to files on the OST when using TCP networks.
769 ------------------------------------------------------------------------------
771 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
773 - rework CDEBUG messages rate-limiting mechanism b=10375
774 - add per-socket tunables for socklnd if the kernel is patched b=10327
776 ------------------------------------------------------------------------------
778 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
780 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
781 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
782 - quiet LNET startup LNI message for liblustre b=10128
783 - Better console error messages if 'ip2nets' can't match an IP address
784 - Fixed overflow/use-before-set bugs in linux-time.h
785 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
786 - LNET teardown failed an assertion about the route table being empty
787 - Fixed a crash in LNetEQPoll(<invalid handle>)
788 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
789 - improve debug message for liblustre/Catamount nodes (b=10116)
791 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
792 * Configuration change for the XT3
793 The PTLLND is now used to run Lustre over Portals on the XT3.
794 The configure option(s) --with-cray-portals are no longer
795 used. Rather --with-portals=<path-to-portals-includes> is
796 used to enable building on the XT3. In addition to enable
797 XT3 specific features the option --enable-cray-xt3 must be
800 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
801 * Portals has been removed, replaced by LNET.
802 LNET is new networking infrastructure for Lustre, it includes a
803 reorganized network configuration mode (see the user
804 documentation for full details) as well as support for routing
805 between different network fabrics. Lustre Networking Devices
806 (LNDS) for the supported network fabrics have also been created
807 for this new infrastructure.
809 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
814 Frequency : rare (large Voltaire clusters only)
816 Description: the default number of reserved transmit descriptors was too low
817 for some large clusters
818 Details : As a workaround, the number was increased. A proper fix includes
821 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
826 Frequency : occasional (large-scale events, cluster reboot, network failure)
828 Description: too many error messages on console obscure actual problem and
829 can slow down/panic server, or cause recovery to fail repeatedly
830 Details : enable rate-limiting of console error messages, and some messages
831 that were console errors now only go to the kernel log
833 Severity : enhancement
835 Description: add /proc/sys/portals/catastrophe entry which will report if
836 that node has previously LBUGged
838 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
840 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
842 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
844 - handle error return code in kranal_check_fma_rx() (5915,6054)
846 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
848 - update vibnal (Voltaire IB NAL)
849 - update gmnal (Myrinet NAL), gmnalid
851 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
853 * Landed portals:b_port_step as follows...
855 - removed CFS_DECL_SPIN*
856 just use 'spinlock_t' and initialise with spin_lock_init()
858 - removed CFS_DECL_MUTEX*
859 just use 'struct semaphore' and initialise with init_mutex()
861 - removed CFS_DECL_RWSEM*
862 just use 'struct rw_semaphore' and initialise with init_rwsem()
864 - renamed cfs_sleep_chan -> cfs_waitq
865 cfs_sleep_link -> cfs_waitlink
867 - fixed race in linux version of arch-independent socknal
868 (the ENOMEM/EAGAIN decision).
870 - Didn't fix problems in Darwin version of arch-independent socknal
871 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
873 - removed libcfs types from non-socknal header files (only some types
874 in the header files had been changed; the .c files hadn't been