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
20 Severity : enhancement
22 Description: port router pinger to userspace
26 Description: kptllnd HELLO protocol deadlock
27 Details : kptllnd HELLO protocol doesn't run to completion in finite time
31 Description: LNet selftest fixes and enhancements
33 Severity : enhancement
35 Description: allow a test node to be a member of multiple test groups
37 Severity : enhancement
39 Description: MXLND: eliminate hosts file, use arp for peer nic_id resolution
40 Details : an update from the upstream developer Scott Atchley.
42 Severity : enhancement
44 Description: add a new LND optiion to control peer buffer credits on routers
48 Description: Fixing deadlock in usocklnd
49 Details : A deadlock was possible in usocklnd due to race condition while
50 tearing connection down. The problem resulted from erroneous
51 assumption that lnet_finalize() could have been called holding
55 Bugzilla : 13621, 15983
56 Description: Protocol V2 of o2iblnd
57 Details : o2iblnd V2 has several new features:
58 . map-on-demand: map-on-demand is disabled by default, it can
59 be enabled by using modparam "map_on_demand=@value@", @value@
60 should >= 0 and < 256, 0 will disable map-on-demand, any other
61 valid value will enable map-on-demand.
62 Oi2blnd will create FMR or physical MR for RDMA if fragments of
64 Enable map-on-demand will take less memory for new connection,
65 but a little more CPU for RDMA.
66 . iWARP : to support iWARP, please enable map-on-demand, 32 and 64
67 are recommanded value. iWARP will probably fail for value >=128.
68 . OOB NOOP message: to resolve deadlock on router.
69 . tunable peer_credits_hiw: (high water to return credits),
70 default value of peer_credits_hiw equals to (peer_credits -1),
71 user can change it between peer_credits/2 and (peer_credits - 1).
72 Lower value is recommended for high latency network.
73 . tunable message queue size: it always equals to peer_credits,
74 higher value is recommended for high latency network.
75 . It's compatible with earlier version of o2iblnd
79 Description: Fixing 'running out of ports' issue
80 Details : Add a delay before next reconnect attempt in ksocklnd in
81 the case of lost race. Limit the frequency of query-requests
82 in lnet. Improved handling of 'dead peer' notifications in
87 Description: Change ptllnd timeout and watchdog timers
88 Details : Add ptltrace_on_nal_failed and bump ptllnd timeout to match
93 Description: One down Lustre FS hangs ALL mounted Lustre filesystems
94 Details : Shared routing enhancements - peer health detection.
96 Severity : enhancement
98 Description: acceptor.c cleanup
99 Details : Code duplication in acceptor.c for the cases of kernel and
100 user-space removed. User-space libcfs tcpip primitives
101 uniformed to have prototypes similar to kernel ones. Minor
102 cosmetic changes in usocklnd to use cfs_socket_t as
103 representation of socket.
107 Description: IB path MTU mistakenly set to 1st path MTU when ib_mtu is off
108 Details : See comment 46 in bug 11245 for details - it's indeed a bug
109 introduced by the original 11245 fix.
113 Description: uptllnd credit overflow fix
114 Details : kptl_msg_t::ptlm_credits could be overflown by uptllnd since
119 Description: socklnd protocol version 3
120 Details : With current protocol V2, connections on router can be
121 blocked and can't receive any incoming messages when there is no
122 more router buffer, so ZC-ACK can't be handled (LNet message
123 can't be finalized) and will cause deadlock on router.
124 Protocol V3 has a dedicated connection for emergency messages
125 like ZC-ACK to router, messages on this dedicated connection
126 don't need any credit so will never be blocked. Also, V3 can send
127 keepalive ping in specified period for router healthy checking.
129 -------------------------------------------------------------------------------
131 12-31-2008 Sun Microsystems, Inc.
133 * Support for networks:
134 socklnd - any kernel supported by Lustre,
135 qswlnd - Qsnet kernel modules 5.20 and later,
136 openiblnd - IbGold 1.8.2,
137 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
138 viblnd - Voltaire ibhost 3.4.5 and later,
139 ciblnd - Topspin 3.2.0,
140 iiblnd - Infiniserv 3.3 + PathBits patch,
141 gmlnd - GM 2.1.22 and later,
142 mxlnd - MX 1.2.1 or later,
143 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
147 Description: workaround for OOM from o2iblnd
148 Details : OFED needs allocate big chunk of memory for QP while creating
149 connection for o2iblnd, OOM can happen if no such a contiguous
151 QP size is decided by concurrent_sends and max_fragments of
152 o2iblnd, now we permit user to specify smaller value for
153 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
154 will decrease memory block size required by creating QP.
158 Description: Support Zerocopy receive of Chelsio device
159 Details : Chelsio driver can support zerocopy for iov[1] if it's
160 contiguous and large enough.
164 Description: fix credit flow deadlock in uptllnd
168 Description: finalize network operation in reasonable time
169 Details : conf-sanity test_32a couldn't stop ost and mds because it
170 tried to access non-existent peer and tcp connect took
171 quite long before timing out.
175 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
176 Details : Lost reference on conn prevents peer from being destroyed, which
177 could prevent new peer creation if peer count has reached upper
182 Description: LNET Selftest results in Soft lockup on OSS CPU
183 Details : only hits when 8 or more o2ib clients involved and a session is
184 torn down with 'lst end_session' without preceeding 'lst stop'.
188 Description: concurrent_sends in IB LNDs should not be changeable at run time
189 Details : concurrent_sends in IB LNDs should not be changeable at run time
193 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
194 Details : only hits under out-of-memory situations
197 -------------------------------------------------------------------------------
199 2009-02-07 Sun Microsystems, Inc.
201 * Support for networks:
202 socklnd - any kernel supported by Lustre,
203 qswlnd - Qsnet kernel modules 5.20 and later,
204 openiblnd - IbGold 1.8.2,
205 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
206 viblnd - Voltaire ibhost 3.4.5 and later,
207 ciblnd - Topspin 3.2.0,
208 iiblnd - Infiniserv 3.3 + PathBits patch,
209 gmlnd - GM 2.1.22 and later,
210 mxlnd - MX 1.2.1 or later,
211 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
214 Description: workaround for OOM from o2iblnd
215 Details : OFED needs allocate big chunk of memory for QP while creating
216 connection for o2iblnd, OOM can happen if no such a contiguous
218 QP size is decided by concurrent_sends and max_fragments of
219 o2iblnd, now we permit user to specify smaller value for
220 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
221 will decrease memory block size required by creating QP.
225 Description: Support Zerocopy receive of Chelsio device
226 Details : Chelsio driver can support zerocopy for iov[1] if it's
227 contiguous and large enough.
230 Description: fix credit flow deadlock in uptllnd
234 Description: finalize network operation in reasonable time
235 Details : conf-sanity test_32a couldn't stop ost and mds because it
236 tried to access non-existent peer and tcp connect took
237 quite long before timing out.
241 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
242 Details : Lost reference on conn prevents peer from being destroyed, which
243 could prevent new peer creation if peer count has reached upper
248 Description: LNET Selftest results in Soft lockup on OSS CPU
249 Details : only hits when 8 or more o2ib clients involved and a session is
250 torn down with 'lst end_session' without preceeding 'lst stop'.
254 Description: concurrent_sends in IB LNDs should not be changeable at run time
255 Details : concurrent_sends in IB LNDs should not be changeable at run time
257 -------------------------------------------------------------------------------
259 11-03-2008 Sun Microsystems, Inc.
261 * Support for networks:
262 socklnd - any kernel supported by Lustre,
263 qswlnd - Qsnet kernel modules 5.20 and later,
264 openiblnd - IbGold 1.8.2,
265 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
266 viblnd - Voltaire ibhost 3.4.5 and later,
267 ciblnd - Topspin 3.2.0,
268 iiblnd - Infiniserv 3.3 + PathBits patch,
269 gmlnd - GM 2.1.22 and later,
270 mxlnd - MX 1.2.1 or later,
271 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
275 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
276 Details : only hits under out-of-memory situations
278 -------------------------------------------------------------------------------
280 04-26-2008 Sun Microsystems, Inc.
282 * Support for networks:
283 socklnd - any kernel supported by Lustre,
284 qswlnd - Qsnet kernel modules 5.20 and later,
285 openiblnd - IbGold 1.8.2,
286 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
287 viblnd - Voltaire ibhost 3.4.5 and later,
288 ciblnd - Topspin 3.2.0,
289 iiblnd - Infiniserv 3.3 + PathBits patch,
290 gmlnd - GM 2.1.22 and later,
291 mxlnd - MX 1.2.1 or later,
292 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
296 Description: excessive debug information removed
297 Details : excessive debug information removed
301 Description: ksocknal_create_conn() hit ASSERTION during connection race
302 Details : ksocknal_create_conn() hit ASSERTION during connection race
306 Description: ksocknal_send_hello() hit ASSERTION while connecting race
307 Details : ksocknal_send_hello() hit ASSERTION while connecting race
311 Description: o2iblnd/ptllnd credit deadlock in a routed config.
312 Details : o2iblnd/ptllnd credit deadlock in a routed config.
316 Description: High load after starting lnet
317 Details : gmlnd should sleep in rx thread in interruptible way. Otherwise,
318 uptime utility reports high load that looks confusingly.
322 Description: ksocklnd fails to establish connection if accept_port is high
323 Details : PID remapping must not be done for active (outgoing) connections
326 --------------------------------------------------------------------------------
328 2008-01-11 Sun Microsystems, Inc.
330 * Support for networks:
331 socklnd - any kernel supported by Lustre,
332 qswlnd - Qsnet kernel modules 5.20 and later,
333 openiblnd - IbGold 1.8.2,
334 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
335 viblnd - Voltaire ibhost 3.4.5 and later,
336 ciblnd - Topspin 3.2.0,
337 iiblnd - Infiniserv 3.3 + PathBits patch,
338 gmlnd - GM 2.1.22 and later,
339 mxlnd - MX 1.2.1 or later,
340 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
344 Description: liblustre network error
345 Details : liblustre clients should understand LNET_ACCEPT_PORT environment
346 variable even if they don't start lnet acceptor.
350 Description: Strange message from lnet (Ignoring prediction from the future)
351 Details : Incorrect calculation of peer's last_alive value in ksocklnd
353 --------------------------------------------------------------------------------
355 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
357 * Support for networks:
358 socklnd - any kernel supported by Lustre,
359 qswlnd - Qsnet kernel modules 5.20 and later,
360 openiblnd - IbGold 1.8.2,
361 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5.
362 viblnd - Voltaire ibhost 3.4.5 and later,
363 ciblnd - Topspin 3.2.0,
364 iiblnd - Infiniserv 3.3 + PathBits patch,
365 gmlnd - GM 2.1.22 and later,
366 mxlnd - MX 1.2.1 or later,
367 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
371 Description: ASSERTION(me == md->md_me) failed in lnet_match_md()
375 Description: increase send queue size for ciblnd/openiblnd
379 Description: new userspace socklnd
380 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
381 with new one - usocklnd.
383 Severity : enhancement
385 Description: Console message flood
386 Details : Make cdls ratelimiting more tunable by adding several tunable in
387 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
388 /proc/sys/lnet/console_backoff.
390 --------------------------------------------------------------------------------
392 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
394 * Support for networks:
395 socklnd - any kernel supported by Lustre,
396 qswlnd - Qsnet kernel modules 5.20 and later,
397 openiblnd - IbGold 1.8.2,
398 o2iblnd - OFED 1.1 and 1.2,
399 viblnd - Voltaire ibhost 3.4.5 and later,
400 ciblnd - Topspin 3.2.0,
401 iiblnd - Infiniserv 3.3 + PathBits patch,
402 gmlnd - GM 2.1.22 and later,
403 mxlnd - MX 1.2.1 or later,
404 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
408 Description: /proc/sys/lnet has non-sysctl entries
409 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
413 Description: TOE Kernel panic by ksocklnd
414 Details : offloaded sockets provide their own implementation of sendpage,
415 can't call tcp_sendpage() directly
419 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
420 Details : races between lnd_shutdown and peer creation prevent
421 lnd_shutdown from finishing.
425 Description: open files rlimit 1024 reached while liblustre testing
426 Details : ulnds/socklnd must close open socket after unsuccessful
431 Description: build error
432 Details : fix typos in gmlnd, ptllnd and viblnd
434 --------------------------------------------------------------------------------
436 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
438 * Support for networks:
439 socklnd - kernels up to 2.6.16,
440 qswlnd - Qsnet kernel modules 5.20 and later,
441 openiblnd - IbGold 1.8.2,
442 o2iblnd - OFED 1.1 and 1.2
443 viblnd - Voltaire ibhost 3.4.5 and later,
444 ciblnd - Topspin 3.2.0,
445 iiblnd - Infiniserv 3.3 + PathBits patch,
446 gmlnd - GM 2.1.22 and later,
447 mxlnd - MX 1.2.1 or later,
448 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
450 --------------------------------------------------------------------------------
452 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
454 * Support for networks:
455 socklnd - kernels up to 2.6.16,
456 qswlnd - Qsnet kernel modules 5.20 and later,
457 openiblnd - IbGold 1.8.2,
459 viblnd - Voltaire ibhost 3.4.5 and later,
460 ciblnd - Topspin 3.2.0,
461 iiblnd - Infiniserv 3.3 + PathBits patch,
462 gmlnd - GM 2.1.22 and later,
463 mxlnd - MX 1.2.1 or later,
464 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
468 Description: Initialize cpumask before use
472 Description: ASSERTION failures when upgrading to the patchless zero-copy
474 Details : This bug affects "rolling upgrades", causing an inconsistent
475 protocol version negotiation and subsequent assertion failure
476 during rolling upgrades after the first wave of upgrades.
480 Details : Change "dropped message" CERRORs to D_NETERROR so they are
481 logged instead of creating "console chatter" when a lustre
482 timeout races with normal RPC completion.
485 Details : lnet_clear_peer_table can wait forever if user forgets to
489 Details : libcfs_id2str should check pid against LNET_PID_ANY.
493 Description: added LNET self test
494 Details : landing b_self_test
499 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
501 Details : do_div() macro is used incorrectly.
503 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
507 Description: make panic on lbug configurable
511 Description: Add OFED1.2 support to o2iblnd
512 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
513 are installed (other than kernel's in-tree infiniband), there
514 could be some problem while insmod o2iblnd (mismatch CRC of
516 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
517 this link provides solution:
518 https://bugs.openfabrics.org/show_bug.cgi?id=355
519 if extra Module.symvers is not supported in kernel, we will
520 have to run the script in bug 12316 to update
521 $LINUX/module.symvers before building o2iblnd.
522 More details about this are in bug 12316.
524 ------------------------------------------------------------------------------
526 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
527 * version 1.4.10 / 1.6.0
528 * Support for networks:
529 socklnd - kernels up to 2.6.16,
530 qswlnd - Qsnet kernel modules 5.20 and later,
531 openiblnd - IbGold 1.8.2,
533 viblnd - Voltaire ibhost 3.4.5 and later,
534 ciblnd - Topspin 3.2.0,
535 iiblnd - Infiniserv 3.3 + PathBits patch,
536 gmlnd - GM 2.1.22 and later,
537 mxlnd - MX 1.2.1 or later,
538 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
542 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
543 possibly accessed in kptllnd_shutdown. Ptllnd should init
544 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
548 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
552 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
556 Description: the_lnet.ln_finalizing was not set when the current thread is
557 about to complete messages. It only affects multi-threaded
563 Description: Changed the default kqswlnd ntxmsg=512
568 Description: Assertion failure in kernel ptllnd caused by posting passive
569 bulk buffers before connection establishment complete.
574 Description: A race in kernel ptllnd between deleting a peer and posting
575 new communications for it could hang communications -
576 manifesting as "Unexpectedly long timeout" messages.
581 Description: Kernel ptllnd lock ordering issue could hang a node.
586 Description: node crash on socket teardown race
589 Frequency : 'lctl peer_list' issued on a mx net
591 Description: Enable lctl's peer_list for MXLND
594 Frequency : after Ptllnd timeouts and portals congestion
596 Description: Credit overflows
597 Details : This was a bug in ptllnd connection establishment. The fix
598 implements better peer stamps to disambiguate connection
599 establishment and ensure both peers enter the credit flow
600 state machine consistently.
605 Description: kptllnd didn't propagate some network errors up to LNET
606 Details : This bug was spotted while investigating 11394. The fix
607 ensures network errors on sends and bulk transfers are
608 propagated to LNET/lustre correctly.
610 Severity : enhancement
612 Description: Fixed console chatter in case of -ETIMEDOUT.
614 Severity : enhancement
616 Description: Added D_NETTRACE for recording network packet history
617 (initially only for ptllnd). Also a separate userspace
618 ptllnd facility to gather history which should really be
619 covered by D_NETTRACE too, if only CDEBUG recorded history in
625 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
626 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
627 callback can occur before a connection has actually been
628 established. This caused an assertion failure previously.
630 Severity : enhancement
632 Description: Multiple instances for o2iblnd
633 Details : Allow multiple instances of o2iblnd to enable networking over
634 multiple HCAs and routing between them.
638 Description: lnet deadlock in router_checker
639 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
640 into BH locks to eliminate potential deadlock caused by
641 ksocknal_data_ready() preempting code holding these locks.
645 Description: Millions of failed socklnd connection attempts cause a very slow FS
646 Details : added a new route flag ksnr_scheduled to distinguish from
647 ksnr_connecting, so that a peer connection request is only turned
648 down for race concerns when an active connection to the same peer
649 is under progress (instead of just being scheduled).
651 ------------------------------------------------------------------------------
653 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
655 * Support for networks:
656 socklnd - kernels up to 2.6.16
657 qswlnd - Qsnet kernel modules 5.20 and later
658 openiblnd - IbGold 1.8.2
660 viblnd - Voltaire ibhost 3.4.5 and later
661 ciblnd - Topspin 3.2.0
662 iiblnd - Infiniserv 3.3 + PathBits patch
663 gmlnd - GM 2.1.22 and later
664 mxlnd - MX 1.2.1 or later
665 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
668 Severity : major on XT3
670 Description: libcfs overwrites /proc/sys/portals
671 Details : libcfs created a symlink from /proc/sys/portals to
672 /proc/sys/lnet for backwards compatibility. This is no
673 longer required and makes the Cray portals /proc variables
678 Description: OFED FMR API change
679 Details : This changes parameter usage to reflect a change in
680 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
681 that FMR support is only used in experimental versions of the
682 o2iblnd - this change does not affect standard usage at all.
684 Severity : enhancement
686 Description: new ko2iblnd module parameter: ib_mtu
687 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
688 HCAs. You can avoid this problem by setting the MTU to 1024
689 using this module parameter.
691 Severity : enhancement
692 Bugzilla : 11118/11620
693 Description: ptllnd small request message buffer alignment fix
694 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
695 Round up small message size on sends in case this option
696 is not supported. 11620 was a defect in the initial
697 implementation which effectively asserted all peers had to be
698 running the correct protocol version which was fixed by always
699 NAK-ing such requests and handling any misalignments they
704 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
705 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
706 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
708 Severity : enhancement
710 Description: Patchless ZC(zero copy) socklnd
711 Details : New protocol for socklnd, socklnd can support zero copy without
712 kernel patch, it's compatible with old socklnd. Checksum is
713 moved from tunables to modparams.
717 Description: When ksocknal_del_peer() is called upon a peer whose
718 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
719 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
722 Frequency : when ptlrpc is under heavy use and runs out of request buffer
724 Description: In lnet_match_blocked_msg(), md can be used without holding a
728 Frequency : very rarely
730 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
731 If connd connects a route which has been closed by
732 ksocknal_shutdown(), ksocknal_create_routes() may create new
733 routes which hold references on the peer, causing shutdown
734 process to wait for peer to disappear forever.
736 Severity : enhancement
738 Description: Dump XT3 portals traces on kptllnd timeout
739 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
740 dump Cray portals debug traces to a file. The kptllnd module
741 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
742 is the basename of the dump file.
745 Frequency : infrequent
747 Description: kernel ptllnd fix bug in connection re-establishment
748 Details : Kernel ptllnd could produce protocol errors e.g. illegal
749 matchbits and/or violate the credit flow protocol when trying
750 to re-establish a connection with a peer after an error or
753 Severity : enhancement
755 Description: Allow /proc/sys/lnet/debug to be set symbolically
756 Details : Allow debug and subsystem debug values to be read/set by name
757 in addition to numerically, for ease of use.
760 Frequency : only in configurations with LNET routers
762 Description: routes automatically marked down and recovered
763 Details : In configurations with LNET routers if a router fails routers
764 now actively try to recover routes that are down, unless they
765 are marked down by an administrator.
767 ------------------------------------------------------------------------------
769 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
772 Frequency : very rarely, in configurations with LNET routers and TCP
774 Description: incorrect data written to files on OSTs
775 Details : In certain high-load conditions incorrect data may be written
776 to files on the OST when using TCP networks.
778 ------------------------------------------------------------------------------
780 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
782 - rework CDEBUG messages rate-limiting mechanism b=10375
783 - add per-socket tunables for socklnd if the kernel is patched b=10327
785 ------------------------------------------------------------------------------
787 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
789 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
790 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
791 - quiet LNET startup LNI message for liblustre b=10128
792 - Better console error messages if 'ip2nets' can't match an IP address
793 - Fixed overflow/use-before-set bugs in linux-time.h
794 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
795 - LNET teardown failed an assertion about the route table being empty
796 - Fixed a crash in LNetEQPoll(<invalid handle>)
797 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
798 - improve debug message for liblustre/Catamount nodes (b=10116)
800 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
801 * Configuration change for the XT3
802 The PTLLND is now used to run Lustre over Portals on the XT3.
803 The configure option(s) --with-cray-portals are no longer
804 used. Rather --with-portals=<path-to-portals-includes> is
805 used to enable building on the XT3. In addition to enable
806 XT3 specific features the option --enable-cray-xt3 must be
809 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
810 * Portals has been removed, replaced by LNET.
811 LNET is new networking infrastructure for Lustre, it includes a
812 reorganized network configuration mode (see the user
813 documentation for full details) as well as support for routing
814 between different network fabrics. Lustre Networking Devices
815 (LNDS) for the supported network fabrics have also been created
816 for this new infrastructure.
818 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
823 Frequency : rare (large Voltaire clusters only)
825 Description: the default number of reserved transmit descriptors was too low
826 for some large clusters
827 Details : As a workaround, the number was increased. A proper fix includes
830 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
835 Frequency : occasional (large-scale events, cluster reboot, network failure)
837 Description: too many error messages on console obscure actual problem and
838 can slow down/panic server, or cause recovery to fail repeatedly
839 Details : enable rate-limiting of console error messages, and some messages
840 that were console errors now only go to the kernel log
842 Severity : enhancement
844 Description: add /proc/sys/portals/catastrophe entry which will report if
845 that node has previously LBUGged
847 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
849 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
851 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
853 - handle error return code in kranal_check_fma_rx() (5915,6054)
855 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
857 - update vibnal (Voltaire IB NAL)
858 - update gmnal (Myrinet NAL), gmnalid
860 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
862 * Landed portals:b_port_step as follows...
864 - removed CFS_DECL_SPIN*
865 just use 'spinlock_t' and initialise with spin_lock_init()
867 - removed CFS_DECL_MUTEX*
868 just use 'struct semaphore' and initialise with init_mutex()
870 - removed CFS_DECL_RWSEM*
871 just use 'struct rw_semaphore' and initialise with init_rwsem()
873 - renamed cfs_sleep_chan -> cfs_waitq
874 cfs_sleep_link -> cfs_waitlink
876 - fixed race in linux version of arch-independent socknal
877 (the ENOMEM/EAGAIN decision).
879 - Didn't fix problems in Darwin version of arch-independent socknal
880 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
882 - removed libcfs types from non-socknal header files (only some types
883 in the header files had been changed; the .c files hadn't been