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
15 Severity : enhancement
17 Description: LNet router shuffler.
19 Severity : enhancement
21 Description: LNet fine grain routing support.
25 Description: router checker stops working when system wall clock goes backward
26 Details : use monotonic timing source instead of system wall clock time.
28 Severity : enhancement
30 Description: avoid asymmetrical router failures
32 Severity : enhancement
34 Description: multiple-instance support for kptllnd
38 Description: ksocknal_close_conn_locked connection race
39 Details : A race was possible when ksocknal_create_conn calls
40 ksocknal_close_conn_locked for already closed conn.
42 Severity : enhancement
44 Description: port router pinger to userspace
48 Description: kptllnd HELLO protocol deadlock
49 Details : kptllnd HELLO protocol doesn't run to completion in finite time
53 Description: LNet selftest fixes and enhancements
55 Severity : enhancement
57 Description: allow a test node to be a member of multiple test groups
59 Severity : enhancement
61 Description: MXLND: eliminate hosts file, use arp for peer nic_id resolution
62 Details : an update from the upstream developer Scott Atchley.
65 -------------------------------------------------------------------------------
66 2009-07-31 Sun Microsystems, Inc.
68 * Support for networks:
69 socklnd - any kernel supported by Lustre,
70 qswlnd - Qsnet kernel modules 5.20 and later,
71 openiblnd - IbGold 1.8.2,
72 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, 1.3, and 1.4.1
73 viblnd - Voltaire ibhost 3.4.5 and later,
74 ciblnd - Topspin 3.2.0,
75 iiblnd - Infiniserv 3.3 + PathBits patch,
76 gmlnd - GM 2.1.22 and later,
77 mxlnd - MX 1.2.1 or later,
78 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
82 Description: router_proc.c is rewritten to use sysctl-interface for parameters
83 residing in /proc/sys/lnet
85 Severity : enhancement
87 Description: add a new LND optiion to control peer buffer credits on routers
91 Description: Fixing deadlock in usocklnd
92 Details : A deadlock was possible in usocklnd due to race condition while
93 tearing connection down. The problem resulted from erroneous
94 assumption that lnet_finalize() could have been called holding
98 Bugzilla : 13621, 15983
99 Description: Protocol V2 of o2iblnd
100 Details : o2iblnd V2 has several new features:
101 . map-on-demand: map-on-demand is disabled by default, it can
102 be enabled by using modparam "map_on_demand=@value@", @value@
103 should >= 0 and < 256, 0 will disable map-on-demand, any other
104 valid value will enable map-on-demand.
105 Oi2blnd will create FMR or physical MR for RDMA if fragments of
107 Enable map-on-demand will take less memory for new connection,
108 but a little more CPU for RDMA.
109 . iWARP : to support iWARP, please enable map-on-demand, 32 and 64
110 are recommanded value. iWARP will probably fail for value >=128.
111 . OOB NOOP message: to resolve deadlock on router.
112 . tunable peer_credits_hiw: (high water to return credits),
113 default value of peer_credits_hiw equals to (peer_credits -1),
114 user can change it between peer_credits/2 and (peer_credits - 1).
115 Lower value is recommended for high latency network.
116 . tunable message queue size: it always equals to peer_credits,
117 higher value is recommended for high latency network.
118 . It's compatible with earlier version of o2iblnd
122 Description: Fixing 'running out of ports' issue
123 Details : Add a delay before next reconnect attempt in ksocklnd in
124 the case of lost race. Limit the frequency of query-requests
125 in lnet. Improved handling of 'dead peer' notifications in
130 Description: Change ptllnd timeout and watchdog timers
131 Details : Add ptltrace_on_nal_failed and bump ptllnd timeout to match
132 Portals wire timeout.
136 Description: One down Lustre FS hangs ALL mounted Lustre filesystems
137 Details : Shared routing enhancements - peer health detection.
141 Description: IB path MTU mistakenly set to 1st path MTU when ib_mtu is off
142 Details : See comment 46 in bug 11245 for details - it's indeed a bug
143 introduced by the original 11245 fix.
147 Description: uptllnd credit overflow fix
148 Details : kptl_msg_t::ptlm_credits could be overflown by uptllnd since
153 Description: socklnd protocol version 3
154 Details : With current protocol V2, connections on router can be
155 blocked and can't receive any incoming messages when there is no
156 more router buffer, so ZC-ACK can't be handled (LNet message
157 can't be finalized) and will cause deadlock on router.
158 Protocol V3 has a dedicated connection for emergency messages
159 like ZC-ACK to router, messages on this dedicated connection
160 don't need any credit so will never be blocked. Also, V3 can send
161 keepalive ping in specified period for router healthy checking.
163 -------------------------------------------------------------------------------
164 12-31-2008 Sun Microsystems, Inc.
166 * Support for networks:
167 socklnd - any kernel supported by Lustre,
168 qswlnd - Qsnet kernel modules 5.20 and later,
169 openiblnd - IbGold 1.8.2,
170 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
171 viblnd - Voltaire ibhost 3.4.5 and later,
172 ciblnd - Topspin 3.2.0,
173 iiblnd - Infiniserv 3.3 + PathBits patch,
174 gmlnd - GM 2.1.22 and later,
175 mxlnd - MX 1.2.1 or later,
176 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
185 Description: workaround for OOM from o2iblnd
186 Details : OFED needs allocate big chunk of memory for QP while creating
187 connection for o2iblnd, OOM can happen if no such a contiguous
189 QP size is decided by concurrent_sends and max_fragments of
190 o2iblnd, now we permit user to specify smaller value for
191 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
192 will decrease memory block size required by creating QP.
196 Description: Support Zerocopy receive of Chelsio device
197 Details : Chelsio driver can support zerocopy for iov[1] if it's
198 contiguous and large enough.
202 Description: fix credit flow deadlock in uptllnd
206 Description: finalize network operation in reasonable time
207 Details : conf-sanity test_32a couldn't stop ost and mds because it
208 tried to access non-existent peer and tcp connect took
209 quite long before timing out.
213 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
214 Details : Lost reference on conn prevents peer from being destroyed, which
215 could prevent new peer creation if peer count has reached upper
220 Description: LNET Selftest results in Soft lockup on OSS CPU
221 Details : only hits when 8 or more o2ib clients involved and a session is
222 torn down with 'lst end_session' without preceeding 'lst stop'.
226 Description: concurrent_sends in IB LNDs should not be changeable at run time
227 Details : concurrent_sends in IB LNDs should not be changeable at run time
231 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
232 Details : only hits under out-of-memory situations
235 -------------------------------------------------------------------------------
237 2009-02-07 Sun Microsystems, Inc.
239 * Support for networks:
240 socklnd - any kernel supported by Lustre,
241 qswlnd - Qsnet kernel modules 5.20 and later,
242 openiblnd - IbGold 1.8.2,
243 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
244 viblnd - Voltaire ibhost 3.4.5 and later,
245 ciblnd - Topspin 3.2.0,
246 iiblnd - Infiniserv 3.3 + PathBits patch,
247 gmlnd - GM 2.1.22 and later,
248 mxlnd - MX 1.2.1 or later,
249 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
253 Description: workaround for OOM from o2iblnd
254 Details : OFED needs allocate big chunk of memory for QP while creating
255 connection for o2iblnd, OOM can happen if no such a contiguous
257 QP size is decided by concurrent_sends and max_fragments of
258 o2iblnd, now we permit user to specify smaller value for
259 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
260 will decrease memory block size required by creating QP.
264 Description: Support Zerocopy receive of Chelsio device
265 Details : Chelsio driver can support zerocopy for iov[1] if it's
266 contiguous and large enough.
270 Description: fix credit flow deadlock in uptllnd
274 Description: finalize network operation in reasonable time
275 Details : conf-sanity test_32a couldn't stop ost and mds because it
276 tried to access non-existent peer and tcp connect took
277 quite long before timing out.
281 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
282 Details : Lost reference on conn prevents peer from being destroyed, which
283 could prevent new peer creation if peer count has reached upper
288 Description: LNET Selftest results in Soft lockup on OSS CPU
289 Details : only hits when 8 or more o2ib clients involved and a session is
290 torn down with 'lst end_session' without preceeding 'lst stop'.
294 Description: concurrent_sends in IB LNDs should not be changeable at run time
295 Details : concurrent_sends in IB LNDs should not be changeable at run time
297 -------------------------------------------------------------------------------
299 11-03-2008 Sun Microsystems, Inc.
301 * Support for networks:
302 socklnd - any kernel supported by Lustre,
303 qswlnd - Qsnet kernel modules 5.20 and later,
304 openiblnd - IbGold 1.8.2,
305 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
306 viblnd - Voltaire ibhost 3.4.5 and later,
307 ciblnd - Topspin 3.2.0,
308 iiblnd - Infiniserv 3.3 + PathBits patch,
309 gmlnd - GM 2.1.22 and later,
310 mxlnd - MX 1.2.1 or later,
311 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
315 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
316 Details : only hits under out-of-memory situations
318 -------------------------------------------------------------------------------
321 04-26-2008 Sun Microsystems, Inc.
323 * Support for networks:
324 socklnd - any kernel supported by Lustre,
325 qswlnd - Qsnet kernel modules 5.20 and later,
326 openiblnd - IbGold 1.8.2,
327 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
328 viblnd - Voltaire ibhost 3.4.5 and later,
329 ciblnd - Topspin 3.2.0,
330 iiblnd - Infiniserv 3.3 + PathBits patch,
331 gmlnd - GM 2.1.22 and later,
332 mxlnd - MX 1.2.1 or later,
333 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
337 Description: excessive debug information removed
338 Details : excessive debug information removed
342 Description: ksocknal_create_conn() hit ASSERTION during connection race
343 Details : ksocknal_create_conn() hit ASSERTION during connection race
347 Description: ksocknal_send_hello() hit ASSERTION while connecting race
348 Details : ksocknal_send_hello() hit ASSERTION while connecting race
352 Description: o2iblnd/ptllnd credit deadlock in a routed config.
353 Details : o2iblnd/ptllnd credit deadlock in a routed config.
357 Description: High load after starting lnet
358 Details : gmlnd should sleep in rx thread in interruptible way. Otherwise,
359 uptime utility reports high load that looks confusingly.
363 Description: ksocklnd fails to establish connection if accept_port is high
364 Details : PID remapping must not be done for active (outgoing) connections
366 --------------------------------------------------------------------------------
368 2008-01-11 Sun Microsystems, Inc.
370 * Support for networks:
371 socklnd - any kernel supported by Lustre,
372 qswlnd - Qsnet kernel modules 5.20 and later,
373 openiblnd - IbGold 1.8.2,
374 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
375 viblnd - Voltaire ibhost 3.4.5 and later,
376 ciblnd - Topspin 3.2.0,
377 iiblnd - Infiniserv 3.3 + PathBits patch,
378 gmlnd - GM 2.1.22 and later,
379 mxlnd - MX 1.2.1 or later,
380 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
383 Description: liblustre network error
384 Details : liblustre clients should understand LNET_ACCEPT_PORT environment
385 variable even if they don't start lnet acceptor.
389 Description: Strange message from lnet (Ignoring prediction from the future)
390 Details : Incorrect calculation of peer's last_alive value in ksocklnd
392 --------------------------------------------------------------------------------
394 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
396 * Support for networks:
397 socklnd - any kernel supported by Lustre,
398 qswlnd - Qsnet kernel modules 5.20 and later,
399 openiblnd - IbGold 1.8.2,
400 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5.
401 viblnd - Voltaire ibhost 3.4.5 and later,
402 ciblnd - Topspin 3.2.0,
403 iiblnd - Infiniserv 3.3 + PathBits patch,
404 gmlnd - GM 2.1.22 and later,
405 mxlnd - MX 1.2.1 or later,
406 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
410 Description: ASSERTION(me == md->md_me) failed in lnet_match_md()
414 Description: increase send queue size for ciblnd/openiblnd
418 Description: new userspace socklnd
419 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
420 with new one - usocklnd.
422 Severity : enhancement
424 Description: Console message flood
425 Details : Make cdls ratelimiting more tunable by adding several tunable in
426 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
427 /proc/sys/lnet/console_backoff.
429 --------------------------------------------------------------------------------
431 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
433 * Support for networks:
434 socklnd - any kernel supported by Lustre,
435 qswlnd - Qsnet kernel modules 5.20 and later,
436 openiblnd - IbGold 1.8.2,
437 o2iblnd - OFED 1.1 and 1.2,
438 viblnd - Voltaire ibhost 3.4.5 and later,
439 ciblnd - Topspin 3.2.0,
440 iiblnd - Infiniserv 3.3 + PathBits patch,
441 gmlnd - GM 2.1.22 and later,
442 mxlnd - MX 1.2.1 or later,
443 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
447 Description: /proc/sys/lnet has non-sysctl entries
448 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
452 Description: TOE Kernel panic by ksocklnd
453 Details : offloaded sockets provide their own implementation of sendpage,
454 can't call tcp_sendpage() directly
458 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
459 Details : races between lnd_shutdown and peer creation prevent
460 lnd_shutdown from finishing.
464 Description: open files rlimit 1024 reached while liblustre testing
465 Details : ulnds/socklnd must close open socket after unsuccessful
470 Description: build error
471 Details : fix typos in gmlnd, ptllnd and viblnd
473 ------------------------------------------------------------------------------
475 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
477 * Support for networks:
478 socklnd - kernels up to 2.6.16,
479 qswlnd - Qsnet kernel modules 5.20 and later,
480 openiblnd - IbGold 1.8.2,
481 o2iblnd - OFED 1.1 and 1.2
482 viblnd - Voltaire ibhost 3.4.5 and later,
483 ciblnd - Topspin 3.2.0,
484 iiblnd - Infiniserv 3.3 + PathBits patch,
485 gmlnd - GM 2.1.22 and later,
486 mxlnd - MX 1.2.1 or later,
487 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
489 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
491 * Support for networks:
492 socklnd - kernels up to 2.6.16,
493 qswlnd - Qsnet kernel modules 5.20 and later,
494 openiblnd - IbGold 1.8.2,
496 viblnd - Voltaire ibhost 3.4.5 and later,
497 ciblnd - Topspin 3.2.0,
498 iiblnd - Infiniserv 3.3 + PathBits patch,
499 gmlnd - GM 2.1.22 and later,
500 mxlnd - MX 1.2.1 or later,
501 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
505 Description: Initialize cpumask before use
509 Description: ASSERTION failures when upgrading to the patchless zero-copy
511 Details : This bug affects "rolling upgrades", causing an inconsistent
512 protocol version negotiation and subsequent assertion failure
513 during rolling upgrades after the first wave of upgrades.
517 Details : Change "dropped message" CERRORs to D_NETERROR so they are
518 logged instead of creating "console chatter" when a lustre
519 timeout races with normal RPC completion.
522 Details : lnet_clear_peer_table can wait forever if user forgets to
526 Details : libcfs_id2str should check pid against LNET_PID_ANY.
530 Description: added LNET self test
531 Details : landing b_self_test
536 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
538 Details : do_div() macro is used incorrectly.
540 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
544 Description: make panic on lbug configurable
548 Description: Add OFED1.2 support to o2iblnd
549 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
550 are installed (other than kernel's in-tree infiniband), there
551 could be some problem while insmod o2iblnd (mismatch CRC of
553 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
554 this link provides solution:
555 https://bugs.openfabrics.org/show_bug.cgi?id=355
556 if extra Module.symvers is not supported in kernel, we will
557 have to run the script in bug 12316 to update
558 $LINUX/module.symvers before building o2iblnd.
559 More details about this are in bug 12316.
561 ------------------------------------------------------------------------------
563 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
564 * version 1.4.10 / 1.6.0
565 * Support for networks:
566 socklnd - kernels up to 2.6.16,
567 qswlnd - Qsnet kernel modules 5.20 and later,
568 openiblnd - IbGold 1.8.2,
570 viblnd - Voltaire ibhost 3.4.5 and later,
571 ciblnd - Topspin 3.2.0,
572 iiblnd - Infiniserv 3.3 + PathBits patch,
573 gmlnd - GM 2.1.22 and later,
574 mxlnd - MX 1.2.1 or later,
575 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
579 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
580 possibly accessed in kptllnd_shutdown. Ptllnd should init
581 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
585 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
589 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
593 Description: the_lnet.ln_finalizing was not set when the current thread is
594 about to complete messages. It only affects multi-threaded
600 Description: Changed the default kqswlnd ntxmsg=512
605 Description: Assertion failure in kernel ptllnd caused by posting passive
606 bulk buffers before connection establishment complete.
611 Description: A race in kernel ptllnd between deleting a peer and posting
612 new communications for it could hang communications -
613 manifesting as "Unexpectedly long timeout" messages.
618 Description: Kernel ptllnd lock ordering issue could hang a node.
623 Description: node crash on socket teardown race
626 Frequency : 'lctl peer_list' issued on a mx net
628 Description: Enable lctl's peer_list for MXLND
631 Frequency : after Ptllnd timeouts and portals congestion
633 Description: Credit overflows
634 Details : This was a bug in ptllnd connection establishment. The fix
635 implements better peer stamps to disambiguate connection
636 establishment and ensure both peers enter the credit flow
637 state machine consistently.
642 Description: kptllnd didn't propagate some network errors up to LNET
643 Details : This bug was spotted while investigating 11394. The fix
644 ensures network errors on sends and bulk transfers are
645 propagated to LNET/lustre correctly.
647 Severity : enhancement
649 Description: Fixed console chatter in case of -ETIMEDOUT.
651 Severity : enhancement
653 Description: Added D_NETTRACE for recording network packet history
654 (initially only for ptllnd). Also a separate userspace
655 ptllnd facility to gather history which should really be
656 covered by D_NETTRACE too, if only CDEBUG recorded history in
662 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
663 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
664 callback can occur before a connection has actually been
665 established. This caused an assertion failure previously.
667 Severity : enhancement
669 Description: Multiple instances for o2iblnd
670 Details : Allow multiple instances of o2iblnd to enable networking over
671 multiple HCAs and routing between them.
675 Description: lnet deadlock in router_checker
676 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
677 into BH locks to eliminate potential deadlock caused by
678 ksocknal_data_ready() preempting code holding these locks.
682 Description: Millions of failed socklnd connection attempts cause a very slow FS
683 Details : added a new route flag ksnr_scheduled to distinguish from
684 ksnr_connecting, so that a peer connection request is only turned
685 down for race concerns when an active connection to the same peer
686 is under progress (instead of just being scheduled).
688 ------------------------------------------------------------------------------
690 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
692 * Support for networks:
693 socklnd - kernels up to 2.6.16
694 qswlnd - Qsnet kernel modules 5.20 and later
695 openiblnd - IbGold 1.8.2
697 viblnd - Voltaire ibhost 3.4.5 and later
698 ciblnd - Topspin 3.2.0
699 iiblnd - Infiniserv 3.3 + PathBits patch
700 gmlnd - GM 2.1.22 and later
701 mxlnd - MX 1.2.1 or later
702 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
705 Severity : major on XT3
707 Description: libcfs overwrites /proc/sys/portals
708 Details : libcfs created a symlink from /proc/sys/portals to
709 /proc/sys/lnet for backwards compatibility. This is no
710 longer required and makes the Cray portals /proc variables
715 Description: OFED FMR API change
716 Details : This changes parameter usage to reflect a change in
717 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
718 that FMR support is only used in experimental versions of the
719 o2iblnd - this change does not affect standard usage at all.
721 Severity : enhancement
723 Description: new ko2iblnd module parameter: ib_mtu
724 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
725 HCAs. You can avoid this problem by setting the MTU to 1024
726 using this module parameter.
728 Severity : enhancement
729 Bugzilla : 11118/11620
730 Description: ptllnd small request message buffer alignment fix
731 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
732 Round up small message size on sends in case this option
733 is not supported. 11620 was a defect in the initial
734 implementation which effectively asserted all peers had to be
735 running the correct protocol version which was fixed by always
736 NAK-ing such requests and handling any misalignments they
741 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
742 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
743 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
745 Severity : enhancement
747 Description: Patchless ZC(zero copy) socklnd
748 Details : New protocol for socklnd, socklnd can support zero copy without
749 kernel patch, it's compatible with old socklnd. Checksum is
750 moved from tunables to modparams.
754 Description: When ksocknal_del_peer() is called upon a peer whose
755 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
756 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
759 Frequency : when ptlrpc is under heavy use and runs out of request buffer
761 Description: In lnet_match_blocked_msg(), md can be used without holding a
765 Frequency : very rarely
767 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
768 If connd connects a route which has been closed by
769 ksocknal_shutdown(), ksocknal_create_routes() may create new
770 routes which hold references on the peer, causing shutdown
771 process to wait for peer to disappear forever.
773 Severity : enhancement
775 Description: Dump XT3 portals traces on kptllnd timeout
776 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
777 dump Cray portals debug traces to a file. The kptllnd module
778 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
779 is the basename of the dump file.
782 Frequency : infrequent
784 Description: kernel ptllnd fix bug in connection re-establishment
785 Details : Kernel ptllnd could produce protocol errors e.g. illegal
786 matchbits and/or violate the credit flow protocol when trying
787 to re-establish a connection with a peer after an error or
790 Severity : enhancement
792 Description: Allow /proc/sys/lnet/debug to be set symbolically
793 Details : Allow debug and subsystem debug values to be read/set by name
794 in addition to numerically, for ease of use.
797 Frequency : only in configurations with LNET routers
799 Description: routes automatically marked down and recovered
800 Details : In configurations with LNET routers if a router fails routers
801 now actively try to recover routes that are down, unless they
802 are marked down by an administrator.
804 ------------------------------------------------------------------------------
806 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
809 Frequency : very rarely, in configurations with LNET routers and TCP
811 Description: incorrect data written to files on OSTs
812 Details : In certain high-load conditions incorrect data may be written
813 to files on the OST when using TCP networks.
815 ------------------------------------------------------------------------------
817 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
819 - rework CDEBUG messages rate-limiting mechanism b=10375
820 - add per-socket tunables for socklnd if the kernel is patched b=10327
822 ------------------------------------------------------------------------------
824 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
826 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
827 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
828 - quiet LNET startup LNI message for liblustre b=10128
829 - Better console error messages if 'ip2nets' can't match an IP address
830 - Fixed overflow/use-before-set bugs in linux-time.h
831 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
832 - LNET teardown failed an assertion about the route table being empty
833 - Fixed a crash in LNetEQPoll(<invalid handle>)
834 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
835 - improve debug message for liblustre/Catamount nodes (b=10116)
837 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
838 * Configuration change for the XT3
839 The PTLLND is now used to run Lustre over Portals on the XT3.
840 The configure option(s) --with-cray-portals are no longer
841 used. Rather --with-portals=<path-to-portals-includes> is
842 used to enable building on the XT3. In addition to enable
843 XT3 specific features the option --enable-cray-xt3 must be
846 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
847 * Portals has been removed, replaced by LNET.
848 LNET is new networking infrastructure for Lustre, it includes a
849 reorganized network configuration mode (see the user
850 documentation for full details) as well as support for routing
851 between different network fabrics. Lustre Networking Devices
852 (LNDS) for the supported network fabrics have also been created
853 for this new infrastructure.
855 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
860 Frequency : rare (large Voltaire clusters only)
862 Description: the default number of reserved transmit descriptors was too low
863 for some large clusters
864 Details : As a workaround, the number was increased. A proper fix includes
867 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
872 Frequency : occasional (large-scale events, cluster reboot, network failure)
874 Description: too many error messages on console obscure actual problem and
875 can slow down/panic server, or cause recovery to fail repeatedly
876 Details : enable rate-limiting of console error messages, and some messages
877 that were console errors now only go to the kernel log
879 Severity : enhancement
881 Description: add /proc/sys/portals/catastrophe entry which will report if
882 that node has previously LBUGged
884 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
886 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
888 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
890 - handle error return code in kranal_check_fma_rx() (5915,6054)
892 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
894 - update vibnal (Voltaire IB NAL)
895 - update gmnal (Myrinet NAL), gmnalid
897 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
899 * Landed portals:b_port_step as follows...
901 - removed CFS_DECL_SPIN*
902 just use 'spinlock_t' and initialise with spin_lock_init()
904 - removed CFS_DECL_MUTEX*
905 just use 'struct semaphore' and initialise with init_mutex()
907 - removed CFS_DECL_RWSEM*
908 just use 'struct rw_semaphore' and initialise with init_rwsem()
910 - renamed cfs_sleep_chan -> cfs_waitq
911 cfs_sleep_link -> cfs_waitlink
913 - fixed race in linux version of arch-independent socknal
914 (the ENOMEM/EAGAIN decision).
916 - Didn't fix problems in Darwin version of arch-independent socknal
917 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
919 - removed libcfs types from non-socknal header files (only some types
920 in the header files had been changed; the .c files hadn't been