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
21 -------------------------------------------------------------------------------
24 04-26-2008 Sun Microsystems, Inc.
26 * Support for networks:
27 socklnd - any kernel supported by Lustre,
28 qswlnd - Qsnet kernel modules 5.20 and later,
29 openiblnd - IbGold 1.8.2,
30 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
31 viblnd - Voltaire ibhost 3.4.5 and later,
32 ciblnd - Topspin 3.2.0,
33 iiblnd - Infiniserv 3.3 + PathBits patch,
34 gmlnd - GM 2.1.22 and later,
35 mxlnd - MX 1.2.1 or later,
36 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
40 Description: ksocknal_create_conn() hit ASSERTION during connection race
41 Details : ksocknal_create_conn() hit ASSERTION during connection race
45 Description: ksocknal_send_hello() hit ASSERTION while connecting race
46 Details : ksocknal_send_hello() hit ASSERTION while connecting race
50 Description: o2iblnd/ptllnd credit deadlock in a routed config.
51 Details : o2iblnd/ptllnd credit deadlock in a routed config.
55 Description: High load after starting lnet
56 Details : gmlnd should sleep in rx thread in interruptible way. Otherwise,
57 uptime utility reports high load that looks confusingly.
61 Description: ksocklnd fails to establish connection if accept_port is high
62 Details : PID remapping must not be done for active (outgoing) connections
64 --------------------------------------------------------------------------------
66 2008-01-11 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 and 1.2.0, 1.2.5
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
81 Description: liblustre network error
82 Details : liblustre clients should understand LNET_ACCEPT_PORT environment
83 variable even if they don't start lnet acceptor.
87 Description: Strange message from lnet (Ignoring prediction from the future)
88 Details : Incorrect calculation of peer's last_alive value in ksocklnd
90 --------------------------------------------------------------------------------
92 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
94 * Support for networks:
95 socklnd - any kernel supported by Lustre,
96 qswlnd - Qsnet kernel modules 5.20 and later,
97 openiblnd - IbGold 1.8.2,
98 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5.
99 viblnd - Voltaire ibhost 3.4.5 and later,
100 ciblnd - Topspin 3.2.0,
101 iiblnd - Infiniserv 3.3 + PathBits patch,
102 gmlnd - GM 2.1.22 and later,
103 mxlnd - MX 1.2.1 or later,
104 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
108 Description: ASSERTION(me == md->md_me) failed in lnet_match_md()
112 Description: increase send queue size for ciblnd/openiblnd
116 Description: new userspace socklnd
117 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
118 with new one - usocklnd.
120 Severity : enhancement
122 Description: Console message flood
123 Details : Make cdls ratelimiting more tunable by adding several tunable in
124 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
125 /proc/sys/lnet/console_backoff.
127 --------------------------------------------------------------------------------
129 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
131 * Support for networks:
132 socklnd - any kernel supported by Lustre,
133 qswlnd - Qsnet kernel modules 5.20 and later,
134 openiblnd - IbGold 1.8.2,
135 o2iblnd - OFED 1.1 and 1.2,
136 viblnd - Voltaire ibhost 3.4.5 and later,
137 ciblnd - Topspin 3.2.0,
138 iiblnd - Infiniserv 3.3 + PathBits patch,
139 gmlnd - GM 2.1.22 and later,
140 mxlnd - MX 1.2.1 or later,
141 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
145 Description: /proc/sys/lnet has non-sysctl entries
146 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
150 Description: TOE Kernel panic by ksocklnd
151 Details : offloaded sockets provide their own implementation of sendpage,
152 can't call tcp_sendpage() directly
156 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
157 Details : races between lnd_shutdown and peer creation prevent
158 lnd_shutdown from finishing.
162 Description: open files rlimit 1024 reached while liblustre testing
163 Details : ulnds/socklnd must close open socket after unsuccessful
168 Description: build error
169 Details : fix typos in gmlnd, ptllnd and viblnd
171 ------------------------------------------------------------------------------
173 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
175 * Support for networks:
176 socklnd - kernels up to 2.6.16,
177 qswlnd - Qsnet kernel modules 5.20 and later,
178 openiblnd - IbGold 1.8.2,
179 o2iblnd - OFED 1.1 and 1.2
180 viblnd - Voltaire ibhost 3.4.5 and later,
181 ciblnd - Topspin 3.2.0,
182 iiblnd - Infiniserv 3.3 + PathBits patch,
183 gmlnd - GM 2.1.22 and later,
184 mxlnd - MX 1.2.1 or later,
185 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
187 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
189 * Support for networks:
190 socklnd - kernels up to 2.6.16,
191 qswlnd - Qsnet kernel modules 5.20 and later,
192 openiblnd - IbGold 1.8.2,
194 viblnd - Voltaire ibhost 3.4.5 and later,
195 ciblnd - Topspin 3.2.0,
196 iiblnd - Infiniserv 3.3 + PathBits patch,
197 gmlnd - GM 2.1.22 and later,
198 mxlnd - MX 1.2.1 or later,
199 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
203 Description: Initialize cpumask before use
207 Description: ASSERTION failures when upgrading to the patchless zero-copy
209 Details : This bug affects "rolling upgrades", causing an inconsistent
210 protocol version negotiation and subsequent assertion failure
211 during rolling upgrades after the first wave of upgrades.
215 Details : Change "dropped message" CERRORs to D_NETERROR so they are
216 logged instead of creating "console chatter" when a lustre
217 timeout races with normal RPC completion.
220 Details : lnet_clear_peer_table can wait forever if user forgets to
224 Details : libcfs_id2str should check pid against LNET_PID_ANY.
228 Description: added LNET self test
229 Details : landing b_self_test
234 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
236 Details : do_div() macro is used incorrectly.
238 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
242 Description: make panic on lbug configurable
246 Description: Add OFED1.2 support to o2iblnd
247 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
248 are installed (other than kernel's in-tree infiniband), there
249 could be some problem while insmod o2iblnd (mismatch CRC of
251 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
252 this link provides solution:
253 https://bugs.openfabrics.org/show_bug.cgi?id=355
254 if extra Module.symvers is not supported in kernel, we will
255 have to run the script in bug 12316 to update
256 $LINUX/module.symvers before building o2iblnd.
257 More details about this are in bug 12316.
259 ------------------------------------------------------------------------------
261 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
262 * version 1.4.10 / 1.6.0
263 * Support for networks:
264 socklnd - kernels up to 2.6.16,
265 qswlnd - Qsnet kernel modules 5.20 and later,
266 openiblnd - IbGold 1.8.2,
268 viblnd - Voltaire ibhost 3.4.5 and later,
269 ciblnd - Topspin 3.2.0,
270 iiblnd - Infiniserv 3.3 + PathBits patch,
271 gmlnd - GM 2.1.22 and later,
272 mxlnd - MX 1.2.1 or later,
273 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
277 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
278 possibly accessed in kptllnd_shutdown. Ptllnd should init
279 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
283 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
287 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
291 Description: the_lnet.ln_finalizing was not set when the current thread is
292 about to complete messages. It only affects multi-threaded
298 Description: Changed the default kqswlnd ntxmsg=512
303 Description: Assertion failure in kernel ptllnd caused by posting passive
304 bulk buffers before connection establishment complete.
309 Description: A race in kernel ptllnd between deleting a peer and posting
310 new communications for it could hang communications -
311 manifesting as "Unexpectedly long timeout" messages.
316 Description: Kernel ptllnd lock ordering issue could hang a node.
321 Description: node crash on socket teardown race
324 Frequency : 'lctl peer_list' issued on a mx net
326 Description: Enable lctl's peer_list for MXLND
329 Frequency : after Ptllnd timeouts and portals congestion
331 Description: Credit overflows
332 Details : This was a bug in ptllnd connection establishment. The fix
333 implements better peer stamps to disambiguate connection
334 establishment and ensure both peers enter the credit flow
335 state machine consistently.
340 Description: kptllnd didn't propagate some network errors up to LNET
341 Details : This bug was spotted while investigating 11394. The fix
342 ensures network errors on sends and bulk transfers are
343 propagated to LNET/lustre correctly.
345 Severity : enhancement
347 Description: Fixed console chatter in case of -ETIMEDOUT.
349 Severity : enhancement
351 Description: Added D_NETTRACE for recording network packet history
352 (initially only for ptllnd). Also a separate userspace
353 ptllnd facility to gather history which should really be
354 covered by D_NETTRACE too, if only CDEBUG recorded history in
360 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
361 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
362 callback can occur before a connection has actually been
363 established. This caused an assertion failure previously.
365 Severity : enhancement
367 Description: Multiple instances for o2iblnd
368 Details : Allow multiple instances of o2iblnd to enable networking over
369 multiple HCAs and routing between them.
373 Description: lnet deadlock in router_checker
374 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
375 into BH locks to eliminate potential deadlock caused by
376 ksocknal_data_ready() preempting code holding these locks.
380 Description: Millions of failed socklnd connection attempts cause a very slow FS
381 Details : added a new route flag ksnr_scheduled to distinguish from
382 ksnr_connecting, so that a peer connection request is only turned
383 down for race concerns when an active connection to the same peer
384 is under progress (instead of just being scheduled).
386 ------------------------------------------------------------------------------
388 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
390 * Support for networks:
391 socklnd - kernels up to 2.6.16
392 qswlnd - Qsnet kernel modules 5.20 and later
393 openiblnd - IbGold 1.8.2
395 viblnd - Voltaire ibhost 3.4.5 and later
396 ciblnd - Topspin 3.2.0
397 iiblnd - Infiniserv 3.3 + PathBits patch
398 gmlnd - GM 2.1.22 and later
399 mxlnd - MX 1.2.1 or later
400 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
403 Severity : major on XT3
405 Description: libcfs overwrites /proc/sys/portals
406 Details : libcfs created a symlink from /proc/sys/portals to
407 /proc/sys/lnet for backwards compatibility. This is no
408 longer required and makes the Cray portals /proc variables
413 Description: OFED FMR API change
414 Details : This changes parameter usage to reflect a change in
415 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
416 that FMR support is only used in experimental versions of the
417 o2iblnd - this change does not affect standard usage at all.
419 Severity : enhancement
421 Description: new ko2iblnd module parameter: ib_mtu
422 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
423 HCAs. You can avoid this problem by setting the MTU to 1024
424 using this module parameter.
426 Severity : enhancement
427 Bugzilla : 11118/11620
428 Description: ptllnd small request message buffer alignment fix
429 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
430 Round up small message size on sends in case this option
431 is not supported. 11620 was a defect in the initial
432 implementation which effectively asserted all peers had to be
433 running the correct protocol version which was fixed by always
434 NAK-ing such requests and handling any misalignments they
439 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
440 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
441 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
443 Severity : enhancement
445 Description: Patchless ZC(zero copy) socklnd
446 Details : New protocol for socklnd, socklnd can support zero copy without
447 kernel patch, it's compatible with old socklnd. Checksum is
448 moved from tunables to modparams.
452 Description: When ksocknal_del_peer() is called upon a peer whose
453 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
454 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
457 Frequency : when ptlrpc is under heavy use and runs out of request buffer
459 Description: In lnet_match_blocked_msg(), md can be used without holding a
463 Frequency : very rarely
465 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
466 If connd connects a route which has been closed by
467 ksocknal_shutdown(), ksocknal_create_routes() may create new
468 routes which hold references on the peer, causing shutdown
469 process to wait for peer to disappear forever.
471 Severity : enhancement
473 Description: Dump XT3 portals traces on kptllnd timeout
474 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
475 dump Cray portals debug traces to a file. The kptllnd module
476 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
477 is the basename of the dump file.
480 Frequency : infrequent
482 Description: kernel ptllnd fix bug in connection re-establishment
483 Details : Kernel ptllnd could produce protocol errors e.g. illegal
484 matchbits and/or violate the credit flow protocol when trying
485 to re-establish a connection with a peer after an error or
488 Severity : enhancement
490 Description: Allow /proc/sys/lnet/debug to be set symbolically
491 Details : Allow debug and subsystem debug values to be read/set by name
492 in addition to numerically, for ease of use.
495 Frequency : only in configurations with LNET routers
497 Description: routes automatically marked down and recovered
498 Details : In configurations with LNET routers if a router fails routers
499 now actively try to recover routes that are down, unless they
500 are marked down by an administrator.
502 ------------------------------------------------------------------------------
504 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
507 Frequency : very rarely, in configurations with LNET routers and TCP
509 Description: incorrect data written to files on OSTs
510 Details : In certain high-load conditions incorrect data may be written
511 to files on the OST when using TCP networks.
513 ------------------------------------------------------------------------------
515 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
517 - rework CDEBUG messages rate-limiting mechanism b=10375
518 - add per-socket tunables for socklnd if the kernel is patched b=10327
520 ------------------------------------------------------------------------------
522 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
524 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
525 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
526 - quiet LNET startup LNI message for liblustre b=10128
527 - Better console error messages if 'ip2nets' can't match an IP address
528 - Fixed overflow/use-before-set bugs in linux-time.h
529 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
530 - LNET teardown failed an assertion about the route table being empty
531 - Fixed a crash in LNetEQPoll(<invalid handle>)
532 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
533 - improve debug message for liblustre/Catamount nodes (b=10116)
535 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
536 * Configuration change for the XT3
537 The PTLLND is now used to run Lustre over Portals on the XT3.
538 The configure option(s) --with-cray-portals are no longer
539 used. Rather --with-portals=<path-to-portals-includes> is
540 used to enable building on the XT3. In addition to enable
541 XT3 specific features the option --enable-cray-xt3 must be
544 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
545 * Portals has been removed, replaced by LNET.
546 LNET is new networking infrastructure for Lustre, it includes a
547 reorganized network configuration mode (see the user
548 documentation for full details) as well as support for routing
549 between different network fabrics. Lustre Networking Devices
550 (LNDS) for the supported network fabrics have also been created
551 for this new infrastructure.
553 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
558 Frequency : rare (large Voltaire clusters only)
560 Description: the default number of reserved transmit descriptors was too low
561 for some large clusters
562 Details : As a workaround, the number was increased. A proper fix includes
565 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
570 Frequency : occasional (large-scale events, cluster reboot, network failure)
572 Description: too many error messages on console obscure actual problem and
573 can slow down/panic server, or cause recovery to fail repeatedly
574 Details : enable rate-limiting of console error messages, and some messages
575 that were console errors now only go to the kernel log
577 Severity : enhancement
579 Description: add /proc/sys/portals/catastrophe entry which will report if
580 that node has previously LBUGged
582 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
584 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
586 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
588 - handle error return code in kranal_check_fma_rx() (5915,6054)
590 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
592 - update vibnal (Voltaire IB NAL)
593 - update gmnal (Myrinet NAL), gmnalid
595 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
597 * Landed portals:b_port_step as follows...
599 - removed CFS_DECL_SPIN*
600 just use 'spinlock_t' and initialise with spin_lock_init()
602 - removed CFS_DECL_MUTEX*
603 just use 'struct semaphore' and initialise with init_mutex()
605 - removed CFS_DECL_RWSEM*
606 just use 'struct rw_semaphore' and initialise with init_rwsem()
608 - renamed cfs_sleep_chan -> cfs_waitq
609 cfs_sleep_link -> cfs_waitlink
611 - fixed race in linux version of arch-independent socknal
612 (the ENOMEM/EAGAIN decision).
614 - Didn't fix problems in Darwin version of arch-independent socknal
615 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
617 - removed libcfs types from non-socknal header files (only some types
618 in the header files had been changed; the .c files hadn't been