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 1.2.10 or later,
13 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
22 Description: should update lp_alive for non-router peers
24 Severity : enhancement
26 Description: LNet router shuffler.
28 Severity : enhancement
30 Description: LNet fine grain routing support.
34 Description: router checker stops working when system wall clock goes backward
35 Details : use monotonic timing source instead of system wall clock time.
37 Severity : enhancement
39 Description: avoid asymmetrical router failures
41 Severity : enhancement
43 Description: multiple-instance support for kptllnd
47 Description: ksocknal_close_conn_locked connection race
48 Details : A race was possible when ksocknal_create_conn calls
49 ksocknal_close_conn_locked for already closed conn.
53 Description: router_proc.c is rewritten to use sysctl-interface for parameters
54 residing in /proc/sys/lnet
56 Severity : enhancement
58 Description: port router pinger to userspace
62 Description: kptllnd HELLO protocol deadlock
63 Details : kptllnd HELLO protocol doesn't run to completion in finite time
67 Description: LNet selftest fixes and enhancements
69 Severity : enhancement
71 Description: allow a test node to be a member of multiple test groups
73 Severity : enhancement
75 Description: MXLND: eliminate hosts file, use arp for peer nic_id resolution
76 Details : an update from the upstream developer Scott Atchley.
78 Severity : enhancement
80 Description: add a new LND optiion to control peer buffer credits on routers
84 Description: Fixing deadlock in usocklnd
85 Details : A deadlock was possible in usocklnd due to race condition while
86 tearing connection down. The problem resulted from erroneous
87 assumption that lnet_finalize() could have been called holding
91 Bugzilla : 13621, 15983
92 Description: Protocol V2 of o2iblnd
93 Details : o2iblnd V2 has several new features:
94 . map-on-demand: map-on-demand is disabled by default, it can
95 be enabled by using modparam "map_on_demand=@value@", @value@
96 should >= 0 and < 256, 0 will disable map-on-demand, any other
97 valid value will enable map-on-demand.
98 Oi2blnd will create FMR or physical MR for RDMA if fragments of
100 Enable map-on-demand will take less memory for new connection,
101 but a little more CPU for RDMA.
102 . iWARP : to support iWARP, please enable map-on-demand, 32 and 64
103 are recommanded value. iWARP will probably fail for value >=128.
104 . OOB NOOP message: to resolve deadlock on router.
105 . tunable peer_credits_hiw: (high water to return credits),
106 default value of peer_credits_hiw equals to (peer_credits -1),
107 user can change it between peer_credits/2 and (peer_credits - 1).
108 Lower value is recommended for high latency network.
109 . tunable message queue size: it always equals to peer_credits,
110 higher value is recommended for high latency network.
111 . It's compatible with earlier version of o2iblnd
115 Description: Fixing 'running out of ports' issue
116 Details : Add a delay before next reconnect attempt in ksocklnd in
117 the case of lost race. Limit the frequency of query-requests
118 in lnet. Improved handling of 'dead peer' notifications in
123 Description: Change ptllnd timeout and watchdog timers
124 Details : Add ptltrace_on_nal_failed and bump ptllnd timeout to match
125 Portals wire timeout.
129 Description: One down Lustre FS hangs ALL mounted Lustre filesystems
130 Details : Shared routing enhancements - peer health detection.
132 Severity : enhancement
134 Description: acceptor.c cleanup
135 Details : Code duplication in acceptor.c for the cases of kernel and
136 user-space removed. User-space libcfs tcpip primitives
137 uniformed to have prototypes similar to kernel ones. Minor
138 cosmetic changes in usocklnd to use cfs_socket_t as
139 representation of socket.
143 Description: IB path MTU mistakenly set to 1st path MTU when ib_mtu is off
144 Details : See comment 46 in bug 11245 for details - it's indeed a bug
145 introduced by the original 11245 fix.
149 Description: uptllnd credit overflow fix
150 Details : kptl_msg_t::ptlm_credits could be overflown by uptllnd since
155 Description: socklnd protocol version 3
156 Details : With current protocol V2, connections on router can be
157 blocked and can't receive any incoming messages when there is no
158 more router buffer, so ZC-ACK can't be handled (LNet message
159 can't be finalized) and will cause deadlock on router.
160 Protocol V3 has a dedicated connection for emergency messages
161 like ZC-ACK to router, messages on this dedicated connection
162 don't need any credit so will never be blocked. Also, V3 can send
163 keepalive ping in specified period for router healthy checking.
165 -------------------------------------------------------------------------------
167 12-31-2008 Sun Microsystems, Inc.
169 * Support for networks:
170 socklnd - any kernel supported by Lustre,
171 qswlnd - Qsnet kernel modules 5.20 and later,
172 openiblnd - IbGold 1.8.2,
173 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
174 viblnd - Voltaire ibhost 3.4.5 and later,
175 ciblnd - Topspin 3.2.0,
176 iiblnd - Infiniserv 3.3 + PathBits patch,
177 gmlnd - GM 2.1.22 and later,
178 mxlnd - MX 1.2.1 or later,
179 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
183 Description: workaround for OOM from o2iblnd
184 Details : OFED needs allocate big chunk of memory for QP while creating
185 connection for o2iblnd, OOM can happen if no such a contiguous
187 QP size is decided by concurrent_sends and max_fragments of
188 o2iblnd, now we permit user to specify smaller value for
189 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
190 will decrease memory block size required by creating QP.
194 Description: Support Zerocopy receive of Chelsio device
195 Details : Chelsio driver can support zerocopy for iov[1] if it's
196 contiguous and large enough.
200 Description: fix credit flow deadlock in uptllnd
204 Description: finalize network operation in reasonable time
205 Details : conf-sanity test_32a couldn't stop ost and mds because it
206 tried to access non-existent peer and tcp connect took
207 quite long before timing out.
211 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
212 Details : Lost reference on conn prevents peer from being destroyed, which
213 could prevent new peer creation if peer count has reached upper
218 Description: LNET Selftest results in Soft lockup on OSS CPU
219 Details : only hits when 8 or more o2ib clients involved and a session is
220 torn down with 'lst end_session' without preceeding 'lst stop'.
224 Description: concurrent_sends in IB LNDs should not be changeable at run time
225 Details : concurrent_sends in IB LNDs should not be changeable at run time
229 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
230 Details : only hits under out-of-memory situations
233 -------------------------------------------------------------------------------
235 2009-02-07 Sun Microsystems, Inc.
237 * Support for networks:
238 socklnd - any kernel supported by Lustre,
239 qswlnd - Qsnet kernel modules 5.20 and later,
240 openiblnd - IbGold 1.8.2,
241 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
242 viblnd - Voltaire ibhost 3.4.5 and later,
243 ciblnd - Topspin 3.2.0,
244 iiblnd - Infiniserv 3.3 + PathBits patch,
245 gmlnd - GM 2.1.22 and later,
246 mxlnd - MX 1.2.1 or later,
247 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
250 Description: workaround for OOM from o2iblnd
251 Details : OFED needs allocate big chunk of memory for QP while creating
252 connection for o2iblnd, OOM can happen if no such a contiguous
254 QP size is decided by concurrent_sends and max_fragments of
255 o2iblnd, now we permit user to specify smaller value for
256 concurrent_sends of o2iblnd(i.e: concurrent_sends=7), which
257 will decrease memory block size required by creating QP.
261 Description: Support Zerocopy receive of Chelsio device
262 Details : Chelsio driver can support zerocopy for iov[1] if it's
263 contiguous and large enough.
266 Description: fix credit flow deadlock in uptllnd
270 Description: finalize network operation in reasonable time
271 Details : conf-sanity test_32a couldn't stop ost and mds because it
272 tried to access non-existent peer and tcp connect took
273 quite long before timing out.
277 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
278 Details : Lost reference on conn prevents peer from being destroyed, which
279 could prevent new peer creation if peer count has reached upper
284 Description: LNET Selftest results in Soft lockup on OSS CPU
285 Details : only hits when 8 or more o2ib clients involved and a session is
286 torn down with 'lst end_session' without preceeding 'lst stop'.
290 Description: concurrent_sends in IB LNDs should not be changeable at run time
291 Details : concurrent_sends in IB LNDs should not be changeable at run time
293 -------------------------------------------------------------------------------
295 11-03-2008 Sun Microsystems, Inc.
297 * Support for networks:
298 socklnd - any kernel supported by Lustre,
299 qswlnd - Qsnet kernel modules 5.20 and later,
300 openiblnd - IbGold 1.8.2,
301 o2iblnd - OFED 1.1, 1.2.0, 1.2.5, and 1.3
302 viblnd - Voltaire ibhost 3.4.5 and later,
303 ciblnd - Topspin 3.2.0,
304 iiblnd - Infiniserv 3.3 + PathBits patch,
305 gmlnd - GM 2.1.22 and later,
306 mxlnd - MX 1.2.1 or later,
307 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
311 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
312 Details : only hits under out-of-memory situations
314 -------------------------------------------------------------------------------
316 04-26-2008 Sun Microsystems, Inc.
318 * Support for networks:
319 socklnd - any kernel supported by Lustre,
320 qswlnd - Qsnet kernel modules 5.20 and later,
321 openiblnd - IbGold 1.8.2,
322 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
323 viblnd - Voltaire ibhost 3.4.5 and later,
324 ciblnd - Topspin 3.2.0,
325 iiblnd - Infiniserv 3.3 + PathBits patch,
326 gmlnd - GM 2.1.22 and later,
327 mxlnd - MX 1.2.1 or later,
328 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
332 Description: excessive debug information removed
333 Details : excessive debug information removed
337 Description: ksocknal_create_conn() hit ASSERTION during connection race
338 Details : ksocknal_create_conn() hit ASSERTION during connection race
342 Description: ksocknal_send_hello() hit ASSERTION while connecting race
343 Details : ksocknal_send_hello() hit ASSERTION while connecting race
347 Description: o2iblnd/ptllnd credit deadlock in a routed config.
348 Details : o2iblnd/ptllnd credit deadlock in a routed config.
352 Description: High load after starting lnet
353 Details : gmlnd should sleep in rx thread in interruptible way. Otherwise,
354 uptime utility reports high load that looks confusingly.
358 Description: ksocklnd fails to establish connection if accept_port is high
359 Details : PID remapping must not be done for active (outgoing) connections
362 --------------------------------------------------------------------------------
364 2008-01-11 Sun Microsystems, Inc.
366 * Support for networks:
367 socklnd - any kernel supported by Lustre,
368 qswlnd - Qsnet kernel modules 5.20 and later,
369 openiblnd - IbGold 1.8.2,
370 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
371 viblnd - Voltaire ibhost 3.4.5 and later,
372 ciblnd - Topspin 3.2.0,
373 iiblnd - Infiniserv 3.3 + PathBits patch,
374 gmlnd - GM 2.1.22 and later,
375 mxlnd - MX 1.2.1 or later,
376 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
380 Description: liblustre network error
381 Details : liblustre clients should understand LNET_ACCEPT_PORT environment
382 variable even if they don't start lnet acceptor.
386 Description: Strange message from lnet (Ignoring prediction from the future)
387 Details : Incorrect calculation of peer's last_alive value in ksocklnd
389 --------------------------------------------------------------------------------
391 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
393 * Support for networks:
394 socklnd - any kernel supported by Lustre,
395 qswlnd - Qsnet kernel modules 5.20 and later,
396 openiblnd - IbGold 1.8.2,
397 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5.
398 viblnd - Voltaire ibhost 3.4.5 and later,
399 ciblnd - Topspin 3.2.0,
400 iiblnd - Infiniserv 3.3 + PathBits patch,
401 gmlnd - GM 2.1.22 and later,
402 mxlnd - MX 1.2.1 or later,
403 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
407 Description: ASSERTION(me == md->md_me) failed in lnet_match_md()
411 Description: increase send queue size for ciblnd/openiblnd
415 Description: new userspace socklnd
416 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
417 with new one - usocklnd.
419 Severity : enhancement
421 Description: Console message flood
422 Details : Make cdls ratelimiting more tunable by adding several tunable in
423 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
424 /proc/sys/lnet/console_backoff.
426 --------------------------------------------------------------------------------
428 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
430 * Support for networks:
431 socklnd - any kernel supported by Lustre,
432 qswlnd - Qsnet kernel modules 5.20 and later,
433 openiblnd - IbGold 1.8.2,
434 o2iblnd - OFED 1.1 and 1.2,
435 viblnd - Voltaire ibhost 3.4.5 and later,
436 ciblnd - Topspin 3.2.0,
437 iiblnd - Infiniserv 3.3 + PathBits patch,
438 gmlnd - GM 2.1.22 and later,
439 mxlnd - MX 1.2.1 or later,
440 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
444 Description: /proc/sys/lnet has non-sysctl entries
445 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
449 Description: TOE Kernel panic by ksocklnd
450 Details : offloaded sockets provide their own implementation of sendpage,
451 can't call tcp_sendpage() directly
455 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
456 Details : races between lnd_shutdown and peer creation prevent
457 lnd_shutdown from finishing.
461 Description: open files rlimit 1024 reached while liblustre testing
462 Details : ulnds/socklnd must close open socket after unsuccessful
467 Description: build error
468 Details : fix typos in gmlnd, ptllnd and viblnd
470 --------------------------------------------------------------------------------
472 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
474 * Support for networks:
475 socklnd - kernels up to 2.6.16,
476 qswlnd - Qsnet kernel modules 5.20 and later,
477 openiblnd - IbGold 1.8.2,
478 o2iblnd - OFED 1.1 and 1.2
479 viblnd - Voltaire ibhost 3.4.5 and later,
480 ciblnd - Topspin 3.2.0,
481 iiblnd - Infiniserv 3.3 + PathBits patch,
482 gmlnd - GM 2.1.22 and later,
483 mxlnd - MX 1.2.1 or later,
484 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
486 --------------------------------------------------------------------------------
488 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
490 * Support for networks:
491 socklnd - kernels up to 2.6.16,
492 qswlnd - Qsnet kernel modules 5.20 and later,
493 openiblnd - IbGold 1.8.2,
495 viblnd - Voltaire ibhost 3.4.5 and later,
496 ciblnd - Topspin 3.2.0,
497 iiblnd - Infiniserv 3.3 + PathBits patch,
498 gmlnd - GM 2.1.22 and later,
499 mxlnd - MX 1.2.1 or later,
500 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
504 Description: Initialize cpumask before use
508 Description: ASSERTION failures when upgrading to the patchless zero-copy
510 Details : This bug affects "rolling upgrades", causing an inconsistent
511 protocol version negotiation and subsequent assertion failure
512 during rolling upgrades after the first wave of upgrades.
516 Details : Change "dropped message" CERRORs to D_NETERROR so they are
517 logged instead of creating "console chatter" when a lustre
518 timeout races with normal RPC completion.
521 Details : lnet_clear_peer_table can wait forever if user forgets to
525 Details : libcfs_id2str should check pid against LNET_PID_ANY.
529 Description: added LNET self test
530 Details : landing b_self_test
535 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
537 Details : do_div() macro is used incorrectly.
539 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
543 Description: make panic on lbug configurable
547 Description: Add OFED1.2 support to o2iblnd
548 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
549 are installed (other than kernel's in-tree infiniband), there
550 could be some problem while insmod o2iblnd (mismatch CRC of
552 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
553 this link provides solution:
554 https://bugs.openfabrics.org/show_bug.cgi?id=355
555 if extra Module.symvers is not supported in kernel, we will
556 have to run the script in bug 12316 to update
557 $LINUX/module.symvers before building o2iblnd.
558 More details about this are in bug 12316.
560 ------------------------------------------------------------------------------
562 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
563 * version 1.4.10 / 1.6.0
564 * Support for networks:
565 socklnd - kernels up to 2.6.16,
566 qswlnd - Qsnet kernel modules 5.20 and later,
567 openiblnd - IbGold 1.8.2,
569 viblnd - Voltaire ibhost 3.4.5 and later,
570 ciblnd - Topspin 3.2.0,
571 iiblnd - Infiniserv 3.3 + PathBits patch,
572 gmlnd - GM 2.1.22 and later,
573 mxlnd - MX 1.2.1 or later,
574 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
578 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
579 possibly accessed in kptllnd_shutdown. Ptllnd should init
580 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
584 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
588 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
592 Description: the_lnet.ln_finalizing was not set when the current thread is
593 about to complete messages. It only affects multi-threaded
599 Description: Changed the default kqswlnd ntxmsg=512
604 Description: Assertion failure in kernel ptllnd caused by posting passive
605 bulk buffers before connection establishment complete.
610 Description: A race in kernel ptllnd between deleting a peer and posting
611 new communications for it could hang communications -
612 manifesting as "Unexpectedly long timeout" messages.
617 Description: Kernel ptllnd lock ordering issue could hang a node.
622 Description: node crash on socket teardown race
625 Frequency : 'lctl peer_list' issued on a mx net
627 Description: Enable lctl's peer_list for MXLND
630 Frequency : after Ptllnd timeouts and portals congestion
632 Description: Credit overflows
633 Details : This was a bug in ptllnd connection establishment. The fix
634 implements better peer stamps to disambiguate connection
635 establishment and ensure both peers enter the credit flow
636 state machine consistently.
641 Description: kptllnd didn't propagate some network errors up to LNET
642 Details : This bug was spotted while investigating 11394. The fix
643 ensures network errors on sends and bulk transfers are
644 propagated to LNET/lustre correctly.
646 Severity : enhancement
648 Description: Fixed console chatter in case of -ETIMEDOUT.
650 Severity : enhancement
652 Description: Added D_NETTRACE for recording network packet history
653 (initially only for ptllnd). Also a separate userspace
654 ptllnd facility to gather history which should really be
655 covered by D_NETTRACE too, if only CDEBUG recorded history in
661 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
662 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
663 callback can occur before a connection has actually been
664 established. This caused an assertion failure previously.
666 Severity : enhancement
668 Description: Multiple instances for o2iblnd
669 Details : Allow multiple instances of o2iblnd to enable networking over
670 multiple HCAs and routing between them.
674 Description: lnet deadlock in router_checker
675 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
676 into BH locks to eliminate potential deadlock caused by
677 ksocknal_data_ready() preempting code holding these locks.
681 Description: Millions of failed socklnd connection attempts cause a very slow FS
682 Details : added a new route flag ksnr_scheduled to distinguish from
683 ksnr_connecting, so that a peer connection request is only turned
684 down for race concerns when an active connection to the same peer
685 is under progress (instead of just being scheduled).
687 ------------------------------------------------------------------------------
689 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
691 * Support for networks:
692 socklnd - kernels up to 2.6.16
693 qswlnd - Qsnet kernel modules 5.20 and later
694 openiblnd - IbGold 1.8.2
696 viblnd - Voltaire ibhost 3.4.5 and later
697 ciblnd - Topspin 3.2.0
698 iiblnd - Infiniserv 3.3 + PathBits patch
699 gmlnd - GM 2.1.22 and later
700 mxlnd - MX 1.2.1 or later
701 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
704 Severity : major on XT3
706 Description: libcfs overwrites /proc/sys/portals
707 Details : libcfs created a symlink from /proc/sys/portals to
708 /proc/sys/lnet for backwards compatibility. This is no
709 longer required and makes the Cray portals /proc variables
714 Description: OFED FMR API change
715 Details : This changes parameter usage to reflect a change in
716 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
717 that FMR support is only used in experimental versions of the
718 o2iblnd - this change does not affect standard usage at all.
720 Severity : enhancement
722 Description: new ko2iblnd module parameter: ib_mtu
723 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
724 HCAs. You can avoid this problem by setting the MTU to 1024
725 using this module parameter.
727 Severity : enhancement
728 Bugzilla : 11118/11620
729 Description: ptllnd small request message buffer alignment fix
730 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
731 Round up small message size on sends in case this option
732 is not supported. 11620 was a defect in the initial
733 implementation which effectively asserted all peers had to be
734 running the correct protocol version which was fixed by always
735 NAK-ing such requests and handling any misalignments they
740 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
741 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
742 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
744 Severity : enhancement
746 Description: Patchless ZC(zero copy) socklnd
747 Details : New protocol for socklnd, socklnd can support zero copy without
748 kernel patch, it's compatible with old socklnd. Checksum is
749 moved from tunables to modparams.
753 Description: When ksocknal_del_peer() is called upon a peer whose
754 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
755 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
758 Frequency : when ptlrpc is under heavy use and runs out of request buffer
760 Description: In lnet_match_blocked_msg(), md can be used without holding a
764 Frequency : very rarely
766 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
767 If connd connects a route which has been closed by
768 ksocknal_shutdown(), ksocknal_create_routes() may create new
769 routes which hold references on the peer, causing shutdown
770 process to wait for peer to disappear forever.
772 Severity : enhancement
774 Description: Dump XT3 portals traces on kptllnd timeout
775 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
776 dump Cray portals debug traces to a file. The kptllnd module
777 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
778 is the basename of the dump file.
781 Frequency : infrequent
783 Description: kernel ptllnd fix bug in connection re-establishment
784 Details : Kernel ptllnd could produce protocol errors e.g. illegal
785 matchbits and/or violate the credit flow protocol when trying
786 to re-establish a connection with a peer after an error or
789 Severity : enhancement
791 Description: Allow /proc/sys/lnet/debug to be set symbolically
792 Details : Allow debug and subsystem debug values to be read/set by name
793 in addition to numerically, for ease of use.
796 Frequency : only in configurations with LNET routers
798 Description: routes automatically marked down and recovered
799 Details : In configurations with LNET routers if a router fails routers
800 now actively try to recover routes that are down, unless they
801 are marked down by an administrator.
803 ------------------------------------------------------------------------------
805 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
808 Frequency : very rarely, in configurations with LNET routers and TCP
810 Description: incorrect data written to files on OSTs
811 Details : In certain high-load conditions incorrect data may be written
812 to files on the OST when using TCP networks.
814 ------------------------------------------------------------------------------
816 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
818 - rework CDEBUG messages rate-limiting mechanism b=10375
819 - add per-socket tunables for socklnd if the kernel is patched b=10327
821 ------------------------------------------------------------------------------
823 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
825 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
826 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
827 - quiet LNET startup LNI message for liblustre b=10128
828 - Better console error messages if 'ip2nets' can't match an IP address
829 - Fixed overflow/use-before-set bugs in linux-time.h
830 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
831 - LNET teardown failed an assertion about the route table being empty
832 - Fixed a crash in LNetEQPoll(<invalid handle>)
833 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
834 - improve debug message for liblustre/Catamount nodes (b=10116)
836 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
837 * Configuration change for the XT3
838 The PTLLND is now used to run Lustre over Portals on the XT3.
839 The configure option(s) --with-cray-portals are no longer
840 used. Rather --with-portals=<path-to-portals-includes> is
841 used to enable building on the XT3. In addition to enable
842 XT3 specific features the option --enable-cray-xt3 must be
845 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
846 * Portals has been removed, replaced by LNET.
847 LNET is new networking infrastructure for Lustre, it includes a
848 reorganized network configuration mode (see the user
849 documentation for full details) as well as support for routing
850 between different network fabrics. Lustre Networking Devices
851 (LNDS) for the supported network fabrics have also been created
852 for this new infrastructure.
854 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
859 Frequency : rare (large Voltaire clusters only)
861 Description: the default number of reserved transmit descriptors was too low
862 for some large clusters
863 Details : As a workaround, the number was increased. A proper fix includes
866 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
871 Frequency : occasional (large-scale events, cluster reboot, network failure)
873 Description: too many error messages on console obscure actual problem and
874 can slow down/panic server, or cause recovery to fail repeatedly
875 Details : enable rate-limiting of console error messages, and some messages
876 that were console errors now only go to the kernel log
878 Severity : enhancement
880 Description: add /proc/sys/portals/catastrophe entry which will report if
881 that node has previously LBUGged
883 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
885 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
887 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
889 - handle error return code in kranal_check_fma_rx() (5915,6054)
891 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
893 - update vibnal (Voltaire IB NAL)
894 - update gmnal (Myrinet NAL), gmnalid
896 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
898 * Landed portals:b_port_step as follows...
900 - removed CFS_DECL_SPIN*
901 just use 'spinlock_t' and initialise with spin_lock_init()
903 - removed CFS_DECL_MUTEX*
904 just use 'struct semaphore' and initialise with init_mutex()
906 - removed CFS_DECL_RWSEM*
907 just use 'struct rw_semaphore' and initialise with init_rwsem()
909 - renamed cfs_sleep_chan -> cfs_waitq
910 cfs_sleep_link -> cfs_waitlink
912 - fixed race in linux version of arch-independent socknal
913 (the ENOMEM/EAGAIN decision).
915 - Didn't fix problems in Darwin version of arch-independent socknal
916 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
918 - removed libcfs types from non-socknal header files (only some types
919 in the header files had been changed; the .c files hadn't been