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, and 1.3
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: finalize network operation in reasonable time
23 Details : conf-sanity test_32a couldn't stop ost and mds because it
24 tried to access non-existent peer and tcp connect took
25 quite long before timing out.
29 Description: Remove portals compatibility
30 Details : Remove portals compatibility, not interoperable with releases
35 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
36 Details : Lost reference on conn prevents peer from being destroyed, which
37 could prevent new peer creation if peer count has reached upper
42 Description: LNET Selftest results in Soft lockup on OSS CPU
43 Details : only hits when 8 or more o2ib clients involved and a session is
44 torn down with 'lst end_session' without preceeding 'lst stop'.
48 Description: concurrent_sends in IB LNDs should not be changeable at run time
49 Details : concurrent_sends in IB LNDs should not be changeable at run time
53 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
54 Details : only hits under out-of-memory situations
57 -------------------------------------------------------------------------------
60 04-26-2008 Sun Microsystems, Inc.
62 * Support for networks:
63 socklnd - any kernel supported by Lustre,
64 qswlnd - Qsnet kernel modules 5.20 and later,
65 openiblnd - IbGold 1.8.2,
66 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
67 viblnd - Voltaire ibhost 3.4.5 and later,
68 ciblnd - Topspin 3.2.0,
69 iiblnd - Infiniserv 3.3 + PathBits patch,
70 gmlnd - GM 2.1.22 and later,
71 mxlnd - MX 1.2.1 or later,
72 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
76 Description: excessive debug information removed
77 Details : excessive debug information removed
81 Description: ksocknal_create_conn() hit ASSERTION during connection race
82 Details : ksocknal_create_conn() hit ASSERTION during connection race
86 Description: ksocknal_send_hello() hit ASSERTION while connecting race
87 Details : ksocknal_send_hello() hit ASSERTION while connecting race
91 Description: o2iblnd/ptllnd credit deadlock in a routed config.
92 Details : o2iblnd/ptllnd credit deadlock in a routed config.
96 Description: High load after starting lnet
97 Details : gmlnd should sleep in rx thread in interruptible way. Otherwise,
98 uptime utility reports high load that looks confusingly.
102 Description: ksocklnd fails to establish connection if accept_port is high
103 Details : PID remapping must not be done for active (outgoing) connections
105 --------------------------------------------------------------------------------
107 2008-01-11 Sun Microsystems, Inc.
109 * Support for networks:
110 socklnd - any kernel supported by Lustre,
111 qswlnd - Qsnet kernel modules 5.20 and later,
112 openiblnd - IbGold 1.8.2,
113 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
114 viblnd - Voltaire ibhost 3.4.5 and later,
115 ciblnd - Topspin 3.2.0,
116 iiblnd - Infiniserv 3.3 + PathBits patch,
117 gmlnd - GM 2.1.22 and later,
118 mxlnd - MX 1.2.1 or later,
119 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
122 Description: liblustre network error
123 Details : liblustre clients should understand LNET_ACCEPT_PORT environment
124 variable even if they don't start lnet acceptor.
128 Description: Strange message from lnet (Ignoring prediction from the future)
129 Details : Incorrect calculation of peer's last_alive value in ksocklnd
131 --------------------------------------------------------------------------------
133 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
135 * Support for networks:
136 socklnd - any kernel supported by Lustre,
137 qswlnd - Qsnet kernel modules 5.20 and later,
138 openiblnd - IbGold 1.8.2,
139 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5.
140 viblnd - Voltaire ibhost 3.4.5 and later,
141 ciblnd - Topspin 3.2.0,
142 iiblnd - Infiniserv 3.3 + PathBits patch,
143 gmlnd - GM 2.1.22 and later,
144 mxlnd - MX 1.2.1 or later,
145 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
149 Description: ASSERTION(me == md->md_me) failed in lnet_match_md()
153 Description: increase send queue size for ciblnd/openiblnd
157 Description: new userspace socklnd
158 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
159 with new one - usocklnd.
161 Severity : enhancement
163 Description: Console message flood
164 Details : Make cdls ratelimiting more tunable by adding several tunable in
165 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
166 /proc/sys/lnet/console_backoff.
168 --------------------------------------------------------------------------------
170 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
172 * Support for networks:
173 socklnd - any kernel supported by Lustre,
174 qswlnd - Qsnet kernel modules 5.20 and later,
175 openiblnd - IbGold 1.8.2,
176 o2iblnd - OFED 1.1 and 1.2,
177 viblnd - Voltaire ibhost 3.4.5 and later,
178 ciblnd - Topspin 3.2.0,
179 iiblnd - Infiniserv 3.3 + PathBits patch,
180 gmlnd - GM 2.1.22 and later,
181 mxlnd - MX 1.2.1 or later,
182 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
186 Description: /proc/sys/lnet has non-sysctl entries
187 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
191 Description: TOE Kernel panic by ksocklnd
192 Details : offloaded sockets provide their own implementation of sendpage,
193 can't call tcp_sendpage() directly
197 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
198 Details : races between lnd_shutdown and peer creation prevent
199 lnd_shutdown from finishing.
203 Description: open files rlimit 1024 reached while liblustre testing
204 Details : ulnds/socklnd must close open socket after unsuccessful
209 Description: build error
210 Details : fix typos in gmlnd, ptllnd and viblnd
212 ------------------------------------------------------------------------------
214 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
216 * Support for networks:
217 socklnd - kernels up to 2.6.16,
218 qswlnd - Qsnet kernel modules 5.20 and later,
219 openiblnd - IbGold 1.8.2,
220 o2iblnd - OFED 1.1 and 1.2
221 viblnd - Voltaire ibhost 3.4.5 and later,
222 ciblnd - Topspin 3.2.0,
223 iiblnd - Infiniserv 3.3 + PathBits patch,
224 gmlnd - GM 2.1.22 and later,
225 mxlnd - MX 1.2.1 or later,
226 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
228 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
230 * Support for networks:
231 socklnd - kernels up to 2.6.16,
232 qswlnd - Qsnet kernel modules 5.20 and later,
233 openiblnd - IbGold 1.8.2,
235 viblnd - Voltaire ibhost 3.4.5 and later,
236 ciblnd - Topspin 3.2.0,
237 iiblnd - Infiniserv 3.3 + PathBits patch,
238 gmlnd - GM 2.1.22 and later,
239 mxlnd - MX 1.2.1 or later,
240 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
244 Description: Initialize cpumask before use
248 Description: ASSERTION failures when upgrading to the patchless zero-copy
250 Details : This bug affects "rolling upgrades", causing an inconsistent
251 protocol version negotiation and subsequent assertion failure
252 during rolling upgrades after the first wave of upgrades.
256 Details : Change "dropped message" CERRORs to D_NETERROR so they are
257 logged instead of creating "console chatter" when a lustre
258 timeout races with normal RPC completion.
261 Details : lnet_clear_peer_table can wait forever if user forgets to
265 Details : libcfs_id2str should check pid against LNET_PID_ANY.
269 Description: added LNET self test
270 Details : landing b_self_test
275 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
277 Details : do_div() macro is used incorrectly.
279 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
283 Description: make panic on lbug configurable
287 Description: Add OFED1.2 support to o2iblnd
288 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
289 are installed (other than kernel's in-tree infiniband), there
290 could be some problem while insmod o2iblnd (mismatch CRC of
292 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
293 this link provides solution:
294 https://bugs.openfabrics.org/show_bug.cgi?id=355
295 if extra Module.symvers is not supported in kernel, we will
296 have to run the script in bug 12316 to update
297 $LINUX/module.symvers before building o2iblnd.
298 More details about this are in bug 12316.
300 ------------------------------------------------------------------------------
302 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
303 * version 1.4.10 / 1.6.0
304 * Support for networks:
305 socklnd - kernels up to 2.6.16,
306 qswlnd - Qsnet kernel modules 5.20 and later,
307 openiblnd - IbGold 1.8.2,
309 viblnd - Voltaire ibhost 3.4.5 and later,
310 ciblnd - Topspin 3.2.0,
311 iiblnd - Infiniserv 3.3 + PathBits patch,
312 gmlnd - GM 2.1.22 and later,
313 mxlnd - MX 1.2.1 or later,
314 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
318 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
319 possibly accessed in kptllnd_shutdown. Ptllnd should init
320 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
324 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
328 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
332 Description: the_lnet.ln_finalizing was not set when the current thread is
333 about to complete messages. It only affects multi-threaded
339 Description: Changed the default kqswlnd ntxmsg=512
344 Description: Assertion failure in kernel ptllnd caused by posting passive
345 bulk buffers before connection establishment complete.
350 Description: A race in kernel ptllnd between deleting a peer and posting
351 new communications for it could hang communications -
352 manifesting as "Unexpectedly long timeout" messages.
357 Description: Kernel ptllnd lock ordering issue could hang a node.
362 Description: node crash on socket teardown race
365 Frequency : 'lctl peer_list' issued on a mx net
367 Description: Enable lctl's peer_list for MXLND
370 Frequency : after Ptllnd timeouts and portals congestion
372 Description: Credit overflows
373 Details : This was a bug in ptllnd connection establishment. The fix
374 implements better peer stamps to disambiguate connection
375 establishment and ensure both peers enter the credit flow
376 state machine consistently.
381 Description: kptllnd didn't propagate some network errors up to LNET
382 Details : This bug was spotted while investigating 11394. The fix
383 ensures network errors on sends and bulk transfers are
384 propagated to LNET/lustre correctly.
386 Severity : enhancement
388 Description: Fixed console chatter in case of -ETIMEDOUT.
390 Severity : enhancement
392 Description: Added D_NETTRACE for recording network packet history
393 (initially only for ptllnd). Also a separate userspace
394 ptllnd facility to gather history which should really be
395 covered by D_NETTRACE too, if only CDEBUG recorded history in
401 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
402 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
403 callback can occur before a connection has actually been
404 established. This caused an assertion failure previously.
406 Severity : enhancement
408 Description: Multiple instances for o2iblnd
409 Details : Allow multiple instances of o2iblnd to enable networking over
410 multiple HCAs and routing between them.
414 Description: lnet deadlock in router_checker
415 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
416 into BH locks to eliminate potential deadlock caused by
417 ksocknal_data_ready() preempting code holding these locks.
421 Description: Millions of failed socklnd connection attempts cause a very slow FS
422 Details : added a new route flag ksnr_scheduled to distinguish from
423 ksnr_connecting, so that a peer connection request is only turned
424 down for race concerns when an active connection to the same peer
425 is under progress (instead of just being scheduled).
427 ------------------------------------------------------------------------------
429 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
431 * Support for networks:
432 socklnd - kernels up to 2.6.16
433 qswlnd - Qsnet kernel modules 5.20 and later
434 openiblnd - IbGold 1.8.2
436 viblnd - Voltaire ibhost 3.4.5 and later
437 ciblnd - Topspin 3.2.0
438 iiblnd - Infiniserv 3.3 + PathBits patch
439 gmlnd - GM 2.1.22 and later
440 mxlnd - MX 1.2.1 or later
441 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
444 Severity : major on XT3
446 Description: libcfs overwrites /proc/sys/portals
447 Details : libcfs created a symlink from /proc/sys/portals to
448 /proc/sys/lnet for backwards compatibility. This is no
449 longer required and makes the Cray portals /proc variables
454 Description: OFED FMR API change
455 Details : This changes parameter usage to reflect a change in
456 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
457 that FMR support is only used in experimental versions of the
458 o2iblnd - this change does not affect standard usage at all.
460 Severity : enhancement
462 Description: new ko2iblnd module parameter: ib_mtu
463 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
464 HCAs. You can avoid this problem by setting the MTU to 1024
465 using this module parameter.
467 Severity : enhancement
468 Bugzilla : 11118/11620
469 Description: ptllnd small request message buffer alignment fix
470 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
471 Round up small message size on sends in case this option
472 is not supported. 11620 was a defect in the initial
473 implementation which effectively asserted all peers had to be
474 running the correct protocol version which was fixed by always
475 NAK-ing such requests and handling any misalignments they
480 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
481 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
482 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
484 Severity : enhancement
486 Description: Patchless ZC(zero copy) socklnd
487 Details : New protocol for socklnd, socklnd can support zero copy without
488 kernel patch, it's compatible with old socklnd. Checksum is
489 moved from tunables to modparams.
493 Description: When ksocknal_del_peer() is called upon a peer whose
494 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
495 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
498 Frequency : when ptlrpc is under heavy use and runs out of request buffer
500 Description: In lnet_match_blocked_msg(), md can be used without holding a
504 Frequency : very rarely
506 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
507 If connd connects a route which has been closed by
508 ksocknal_shutdown(), ksocknal_create_routes() may create new
509 routes which hold references on the peer, causing shutdown
510 process to wait for peer to disappear forever.
512 Severity : enhancement
514 Description: Dump XT3 portals traces on kptllnd timeout
515 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
516 dump Cray portals debug traces to a file. The kptllnd module
517 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
518 is the basename of the dump file.
521 Frequency : infrequent
523 Description: kernel ptllnd fix bug in connection re-establishment
524 Details : Kernel ptllnd could produce protocol errors e.g. illegal
525 matchbits and/or violate the credit flow protocol when trying
526 to re-establish a connection with a peer after an error or
529 Severity : enhancement
531 Description: Allow /proc/sys/lnet/debug to be set symbolically
532 Details : Allow debug and subsystem debug values to be read/set by name
533 in addition to numerically, for ease of use.
536 Frequency : only in configurations with LNET routers
538 Description: routes automatically marked down and recovered
539 Details : In configurations with LNET routers if a router fails routers
540 now actively try to recover routes that are down, unless they
541 are marked down by an administrator.
543 ------------------------------------------------------------------------------
545 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
548 Frequency : very rarely, in configurations with LNET routers and TCP
550 Description: incorrect data written to files on OSTs
551 Details : In certain high-load conditions incorrect data may be written
552 to files on the OST when using TCP networks.
554 ------------------------------------------------------------------------------
556 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
558 - rework CDEBUG messages rate-limiting mechanism b=10375
559 - add per-socket tunables for socklnd if the kernel is patched b=10327
561 ------------------------------------------------------------------------------
563 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
565 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
566 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
567 - quiet LNET startup LNI message for liblustre b=10128
568 - Better console error messages if 'ip2nets' can't match an IP address
569 - Fixed overflow/use-before-set bugs in linux-time.h
570 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
571 - LNET teardown failed an assertion about the route table being empty
572 - Fixed a crash in LNetEQPoll(<invalid handle>)
573 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
574 - improve debug message for liblustre/Catamount nodes (b=10116)
576 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
577 * Configuration change for the XT3
578 The PTLLND is now used to run Lustre over Portals on the XT3.
579 The configure option(s) --with-cray-portals are no longer
580 used. Rather --with-portals=<path-to-portals-includes> is
581 used to enable building on the XT3. In addition to enable
582 XT3 specific features the option --enable-cray-xt3 must be
585 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
586 * Portals has been removed, replaced by LNET.
587 LNET is new networking infrastructure for Lustre, it includes a
588 reorganized network configuration mode (see the user
589 documentation for full details) as well as support for routing
590 between different network fabrics. Lustre Networking Devices
591 (LNDS) for the supported network fabrics have also been created
592 for this new infrastructure.
594 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
599 Frequency : rare (large Voltaire clusters only)
601 Description: the default number of reserved transmit descriptors was too low
602 for some large clusters
603 Details : As a workaround, the number was increased. A proper fix includes
606 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
611 Frequency : occasional (large-scale events, cluster reboot, network failure)
613 Description: too many error messages on console obscure actual problem and
614 can slow down/panic server, or cause recovery to fail repeatedly
615 Details : enable rate-limiting of console error messages, and some messages
616 that were console errors now only go to the kernel log
618 Severity : enhancement
620 Description: add /proc/sys/portals/catastrophe entry which will report if
621 that node has previously LBUGged
623 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
625 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
627 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
629 - handle error return code in kranal_check_fma_rx() (5915,6054)
631 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
633 - update vibnal (Voltaire IB NAL)
634 - update gmnal (Myrinet NAL), gmnalid
636 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
638 * Landed portals:b_port_step as follows...
640 - removed CFS_DECL_SPIN*
641 just use 'spinlock_t' and initialise with spin_lock_init()
643 - removed CFS_DECL_MUTEX*
644 just use 'struct semaphore' and initialise with init_mutex()
646 - removed CFS_DECL_RWSEM*
647 just use 'struct rw_semaphore' and initialise with init_rwsem()
649 - renamed cfs_sleep_chan -> cfs_waitq
650 cfs_sleep_link -> cfs_waitlink
652 - fixed race in linux version of arch-independent socknal
653 (the ENOMEM/EAGAIN decision).
655 - Didn't fix problems in Darwin version of arch-independent socknal
656 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
658 - removed libcfs types from non-socknal header files (only some types
659 in the header files had been changed; the .c files hadn't been