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 and 1.2.0, 1.2.5
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_send_hello() hit ASSERTION while connecting race
41 Details : ksocknal_send_hello() hit ASSERTION while connecting race
45 Description: o2iblnd/ptllnd credit deadlock in a routed config.
46 Details : o2iblnd/ptllnd credit deadlock in a routed config.
50 Description: High load after starting lnet
51 Details : gmlnd should sleep in rx thread in interruptible way. Otherwise,
52 uptime utility reports high load that looks confusingly.
56 Description: ksocklnd fails to establish connection if accept_port is high
57 Details : PID remapping must not be done for active (outgoing) connections
59 --------------------------------------------------------------------------------
61 2008-01-11 Sun Microsystems, Inc.
63 * Support for networks:
64 socklnd - any kernel supported by Lustre,
65 qswlnd - Qsnet kernel modules 5.20 and later,
66 openiblnd - IbGold 1.8.2,
67 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
68 viblnd - Voltaire ibhost 3.4.5 and later,
69 ciblnd - Topspin 3.2.0,
70 iiblnd - Infiniserv 3.3 + PathBits patch,
71 gmlnd - GM 2.1.22 and later,
72 mxlnd - MX 1.2.1 or later,
73 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
76 Description: liblustre network error
77 Details : liblustre clients should understand LNET_ACCEPT_PORT environment
78 variable even if they don't start lnet acceptor.
82 Description: Strange message from lnet (Ignoring prediction from the future)
83 Details : Incorrect calculation of peer's last_alive value in ksocklnd
85 --------------------------------------------------------------------------------
87 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
89 * Support for networks:
90 socklnd - any kernel supported by Lustre,
91 qswlnd - Qsnet kernel modules 5.20 and later,
92 openiblnd - IbGold 1.8.2,
93 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5.
94 viblnd - Voltaire ibhost 3.4.5 and later,
95 ciblnd - Topspin 3.2.0,
96 iiblnd - Infiniserv 3.3 + PathBits patch,
97 gmlnd - GM 2.1.22 and later,
98 mxlnd - MX 1.2.1 or later,
99 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
103 Description: ASSERTION(me == md->md_me) failed in lnet_match_md()
107 Description: increase send queue size for ciblnd/openiblnd
111 Description: new userspace socklnd
112 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
113 with new one - usocklnd.
115 Severity : enhancement
117 Description: Console message flood
118 Details : Make cdls ratelimiting more tunable by adding several tunable in
119 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
120 /proc/sys/lnet/console_backoff.
122 --------------------------------------------------------------------------------
124 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
126 * Support for networks:
127 socklnd - any kernel supported by Lustre,
128 qswlnd - Qsnet kernel modules 5.20 and later,
129 openiblnd - IbGold 1.8.2,
130 o2iblnd - OFED 1.1 and 1.2,
131 viblnd - Voltaire ibhost 3.4.5 and later,
132 ciblnd - Topspin 3.2.0,
133 iiblnd - Infiniserv 3.3 + PathBits patch,
134 gmlnd - GM 2.1.22 and later,
135 mxlnd - MX 1.2.1 or later,
136 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
140 Description: /proc/sys/lnet has non-sysctl entries
141 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
145 Description: TOE Kernel panic by ksocklnd
146 Details : offloaded sockets provide their own implementation of sendpage,
147 can't call tcp_sendpage() directly
151 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
152 Details : races between lnd_shutdown and peer creation prevent
153 lnd_shutdown from finishing.
157 Description: open files rlimit 1024 reached while liblustre testing
158 Details : ulnds/socklnd must close open socket after unsuccessful
163 Description: build error
164 Details : fix typos in gmlnd, ptllnd and viblnd
166 ------------------------------------------------------------------------------
168 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
170 * Support for networks:
171 socklnd - kernels up to 2.6.16,
172 qswlnd - Qsnet kernel modules 5.20 and later,
173 openiblnd - IbGold 1.8.2,
174 o2iblnd - OFED 1.1 and 1.2
175 viblnd - Voltaire ibhost 3.4.5 and later,
176 ciblnd - Topspin 3.2.0,
177 iiblnd - Infiniserv 3.3 + PathBits patch,
178 gmlnd - GM 2.1.22 and later,
179 mxlnd - MX 1.2.1 or later,
180 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
182 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
184 * Support for networks:
185 socklnd - kernels up to 2.6.16,
186 qswlnd - Qsnet kernel modules 5.20 and later,
187 openiblnd - IbGold 1.8.2,
189 viblnd - Voltaire ibhost 3.4.5 and later,
190 ciblnd - Topspin 3.2.0,
191 iiblnd - Infiniserv 3.3 + PathBits patch,
192 gmlnd - GM 2.1.22 and later,
193 mxlnd - MX 1.2.1 or later,
194 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
198 Description: Initialize cpumask before use
202 Description: ASSERTION failures when upgrading to the patchless zero-copy
204 Details : This bug affects "rolling upgrades", causing an inconsistent
205 protocol version negotiation and subsequent assertion failure
206 during rolling upgrades after the first wave of upgrades.
210 Details : Change "dropped message" CERRORs to D_NETERROR so they are
211 logged instead of creating "console chatter" when a lustre
212 timeout races with normal RPC completion.
215 Details : lnet_clear_peer_table can wait forever if user forgets to
219 Details : libcfs_id2str should check pid against LNET_PID_ANY.
223 Description: added LNET self test
224 Details : landing b_self_test
229 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
231 Details : do_div() macro is used incorrectly.
233 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
237 Description: make panic on lbug configurable
241 Description: Add OFED1.2 support to o2iblnd
242 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
243 are installed (other than kernel's in-tree infiniband), there
244 could be some problem while insmod o2iblnd (mismatch CRC of
246 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
247 this link provides solution:
248 https://bugs.openfabrics.org/show_bug.cgi?id=355
249 if extra Module.symvers is not supported in kernel, we will
250 have to run the script in bug 12316 to update
251 $LINUX/module.symvers before building o2iblnd.
252 More details about this are in bug 12316.
254 ------------------------------------------------------------------------------
256 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
257 * version 1.4.10 / 1.6.0
258 * Support for networks:
259 socklnd - kernels up to 2.6.16,
260 qswlnd - Qsnet kernel modules 5.20 and later,
261 openiblnd - IbGold 1.8.2,
263 viblnd - Voltaire ibhost 3.4.5 and later,
264 ciblnd - Topspin 3.2.0,
265 iiblnd - Infiniserv 3.3 + PathBits patch,
266 gmlnd - GM 2.1.22 and later,
267 mxlnd - MX 1.2.1 or later,
268 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
272 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
273 possibly accessed in kptllnd_shutdown. Ptllnd should init
274 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
278 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
282 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
286 Description: the_lnet.ln_finalizing was not set when the current thread is
287 about to complete messages. It only affects multi-threaded
293 Description: Changed the default kqswlnd ntxmsg=512
298 Description: Assertion failure in kernel ptllnd caused by posting passive
299 bulk buffers before connection establishment complete.
304 Description: A race in kernel ptllnd between deleting a peer and posting
305 new communications for it could hang communications -
306 manifesting as "Unexpectedly long timeout" messages.
311 Description: Kernel ptllnd lock ordering issue could hang a node.
316 Description: node crash on socket teardown race
319 Frequency : 'lctl peer_list' issued on a mx net
321 Description: Enable lctl's peer_list for MXLND
324 Frequency : after Ptllnd timeouts and portals congestion
326 Description: Credit overflows
327 Details : This was a bug in ptllnd connection establishment. The fix
328 implements better peer stamps to disambiguate connection
329 establishment and ensure both peers enter the credit flow
330 state machine consistently.
335 Description: kptllnd didn't propagate some network errors up to LNET
336 Details : This bug was spotted while investigating 11394. The fix
337 ensures network errors on sends and bulk transfers are
338 propagated to LNET/lustre correctly.
340 Severity : enhancement
342 Description: Fixed console chatter in case of -ETIMEDOUT.
344 Severity : enhancement
346 Description: Added D_NETTRACE for recording network packet history
347 (initially only for ptllnd). Also a separate userspace
348 ptllnd facility to gather history which should really be
349 covered by D_NETTRACE too, if only CDEBUG recorded history in
355 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
356 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
357 callback can occur before a connection has actually been
358 established. This caused an assertion failure previously.
360 Severity : enhancement
362 Description: Multiple instances for o2iblnd
363 Details : Allow multiple instances of o2iblnd to enable networking over
364 multiple HCAs and routing between them.
368 Description: lnet deadlock in router_checker
369 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
370 into BH locks to eliminate potential deadlock caused by
371 ksocknal_data_ready() preempting code holding these locks.
375 Description: Millions of failed socklnd connection attempts cause a very slow FS
376 Details : added a new route flag ksnr_scheduled to distinguish from
377 ksnr_connecting, so that a peer connection request is only turned
378 down for race concerns when an active connection to the same peer
379 is under progress (instead of just being scheduled).
381 ------------------------------------------------------------------------------
383 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
385 * Support for networks:
386 socklnd - kernels up to 2.6.16
387 qswlnd - Qsnet kernel modules 5.20 and later
388 openiblnd - IbGold 1.8.2
390 viblnd - Voltaire ibhost 3.4.5 and later
391 ciblnd - Topspin 3.2.0
392 iiblnd - Infiniserv 3.3 + PathBits patch
393 gmlnd - GM 2.1.22 and later
394 mxlnd - MX 1.2.1 or later
395 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
398 Severity : major on XT3
400 Description: libcfs overwrites /proc/sys/portals
401 Details : libcfs created a symlink from /proc/sys/portals to
402 /proc/sys/lnet for backwards compatibility. This is no
403 longer required and makes the Cray portals /proc variables
408 Description: OFED FMR API change
409 Details : This changes parameter usage to reflect a change in
410 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
411 that FMR support is only used in experimental versions of the
412 o2iblnd - this change does not affect standard usage at all.
414 Severity : enhancement
416 Description: new ko2iblnd module parameter: ib_mtu
417 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
418 HCAs. You can avoid this problem by setting the MTU to 1024
419 using this module parameter.
421 Severity : enhancement
422 Bugzilla : 11118/11620
423 Description: ptllnd small request message buffer alignment fix
424 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
425 Round up small message size on sends in case this option
426 is not supported. 11620 was a defect in the initial
427 implementation which effectively asserted all peers had to be
428 running the correct protocol version which was fixed by always
429 NAK-ing such requests and handling any misalignments they
434 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
435 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
436 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
438 Severity : enhancement
440 Description: Patchless ZC(zero copy) socklnd
441 Details : New protocol for socklnd, socklnd can support zero copy without
442 kernel patch, it's compatible with old socklnd. Checksum is
443 moved from tunables to modparams.
447 Description: When ksocknal_del_peer() is called upon a peer whose
448 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
449 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
452 Frequency : when ptlrpc is under heavy use and runs out of request buffer
454 Description: In lnet_match_blocked_msg(), md can be used without holding a
458 Frequency : very rarely
460 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
461 If connd connects a route which has been closed by
462 ksocknal_shutdown(), ksocknal_create_routes() may create new
463 routes which hold references on the peer, causing shutdown
464 process to wait for peer to disappear forever.
466 Severity : enhancement
468 Description: Dump XT3 portals traces on kptllnd timeout
469 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
470 dump Cray portals debug traces to a file. The kptllnd module
471 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
472 is the basename of the dump file.
475 Frequency : infrequent
477 Description: kernel ptllnd fix bug in connection re-establishment
478 Details : Kernel ptllnd could produce protocol errors e.g. illegal
479 matchbits and/or violate the credit flow protocol when trying
480 to re-establish a connection with a peer after an error or
483 Severity : enhancement
485 Description: Allow /proc/sys/lnet/debug to be set symbolically
486 Details : Allow debug and subsystem debug values to be read/set by name
487 in addition to numerically, for ease of use.
490 Frequency : only in configurations with LNET routers
492 Description: routes automatically marked down and recovered
493 Details : In configurations with LNET routers if a router fails routers
494 now actively try to recover routes that are down, unless they
495 are marked down by an administrator.
497 ------------------------------------------------------------------------------
499 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
502 Frequency : very rarely, in configurations with LNET routers and TCP
504 Description: incorrect data written to files on OSTs
505 Details : In certain high-load conditions incorrect data may be written
506 to files on the OST when using TCP networks.
508 ------------------------------------------------------------------------------
510 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
512 - rework CDEBUG messages rate-limiting mechanism b=10375
513 - add per-socket tunables for socklnd if the kernel is patched b=10327
515 ------------------------------------------------------------------------------
517 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
519 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
520 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
521 - quiet LNET startup LNI message for liblustre b=10128
522 - Better console error messages if 'ip2nets' can't match an IP address
523 - Fixed overflow/use-before-set bugs in linux-time.h
524 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
525 - LNET teardown failed an assertion about the route table being empty
526 - Fixed a crash in LNetEQPoll(<invalid handle>)
527 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
528 - improve debug message for liblustre/Catamount nodes (b=10116)
530 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
531 * Configuration change for the XT3
532 The PTLLND is now used to run Lustre over Portals on the XT3.
533 The configure option(s) --with-cray-portals are no longer
534 used. Rather --with-portals=<path-to-portals-includes> is
535 used to enable building on the XT3. In addition to enable
536 XT3 specific features the option --enable-cray-xt3 must be
539 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
540 * Portals has been removed, replaced by LNET.
541 LNET is new networking infrastructure for Lustre, it includes a
542 reorganized network configuration mode (see the user
543 documentation for full details) as well as support for routing
544 between different network fabrics. Lustre Networking Devices
545 (LNDS) for the supported network fabrics have also been created
546 for this new infrastructure.
548 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
553 Frequency : rare (large Voltaire clusters only)
555 Description: the default number of reserved transmit descriptors was too low
556 for some large clusters
557 Details : As a workaround, the number was increased. A proper fix includes
560 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
565 Frequency : occasional (large-scale events, cluster reboot, network failure)
567 Description: too many error messages on console obscure actual problem and
568 can slow down/panic server, or cause recovery to fail repeatedly
569 Details : enable rate-limiting of console error messages, and some messages
570 that were console errors now only go to the kernel log
572 Severity : enhancement
574 Description: add /proc/sys/portals/catastrophe entry which will report if
575 that node has previously LBUGged
577 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
579 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
581 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
583 - handle error return code in kranal_check_fma_rx() (5915,6054)
585 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
587 - update vibnal (Voltaire IB NAL)
588 - update gmnal (Myrinet NAL), gmnalid
590 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
592 * Landed portals:b_port_step as follows...
594 - removed CFS_DECL_SPIN*
595 just use 'spinlock_t' and initialise with spin_lock_init()
597 - removed CFS_DECL_MUTEX*
598 just use 'struct semaphore' and initialise with init_mutex()
600 - removed CFS_DECL_RWSEM*
601 just use 'struct rw_semaphore' and initialise with init_rwsem()
603 - renamed cfs_sleep_chan -> cfs_waitq
604 cfs_sleep_link -> cfs_waitlink
606 - fixed race in linux version of arch-independent socknal
607 (the ENOMEM/EAGAIN decision).
609 - Didn't fix problems in Darwin version of arch-independent socknal
610 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
612 - removed libcfs types from non-socknal header files (only some types
613 in the header files had been changed; the .c files hadn't been