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: Remove portals compatibility
23 Details : Remove portals compatibility, not interoperable with releases
28 Description: Continuous recovery on 33 of 413 nodes after lustre oss failure
29 Details : Lost reference on conn prevents peer from being destroyed, which
30 could prevent new peer creation if peer count has reached upper
35 Description: LNET Selftest results in Soft lockup on OSS CPU
36 Details : only hits when 8 or more o2ib clients involved and a session is
37 torn down with 'lst end_session' without preceeding 'lst stop'.
41 Description: concurrent_sends in IB LNDs should not be changeable at run time
42 Details : concurrent_sends in IB LNDs should not be changeable at run time
46 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
47 Details : only hits under out-of-memory situations
50 -------------------------------------------------------------------------------
53 04-26-2008 Sun Microsystems, Inc.
55 * Support for networks:
56 socklnd - any kernel supported by Lustre,
57 qswlnd - Qsnet kernel modules 5.20 and later,
58 openiblnd - IbGold 1.8.2,
59 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
60 viblnd - Voltaire ibhost 3.4.5 and later,
61 ciblnd - Topspin 3.2.0,
62 iiblnd - Infiniserv 3.3 + PathBits patch,
63 gmlnd - GM 2.1.22 and later,
64 mxlnd - MX 1.2.1 or later,
65 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
69 Description: excessive debug information removed
70 Details : excessive debug information removed
74 Description: ksocknal_create_conn() hit ASSERTION during connection race
75 Details : ksocknal_create_conn() hit ASSERTION during connection race
79 Description: ksocknal_send_hello() hit ASSERTION while connecting race
80 Details : ksocknal_send_hello() hit ASSERTION while connecting race
84 Description: o2iblnd/ptllnd credit deadlock in a routed config.
85 Details : o2iblnd/ptllnd credit deadlock in a routed config.
89 Description: High load after starting lnet
90 Details : gmlnd should sleep in rx thread in interruptible way. Otherwise,
91 uptime utility reports high load that looks confusingly.
95 Description: ksocklnd fails to establish connection if accept_port is high
96 Details : PID remapping must not be done for active (outgoing) connections
98 --------------------------------------------------------------------------------
100 2008-01-11 Sun Microsystems, Inc.
102 * Support for networks:
103 socklnd - any kernel supported by Lustre,
104 qswlnd - Qsnet kernel modules 5.20 and later,
105 openiblnd - IbGold 1.8.2,
106 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
107 viblnd - Voltaire ibhost 3.4.5 and later,
108 ciblnd - Topspin 3.2.0,
109 iiblnd - Infiniserv 3.3 + PathBits patch,
110 gmlnd - GM 2.1.22 and later,
111 mxlnd - MX 1.2.1 or later,
112 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
115 Description: liblustre network error
116 Details : liblustre clients should understand LNET_ACCEPT_PORT environment
117 variable even if they don't start lnet acceptor.
121 Description: Strange message from lnet (Ignoring prediction from the future)
122 Details : Incorrect calculation of peer's last_alive value in ksocklnd
124 --------------------------------------------------------------------------------
126 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
128 * Support for networks:
129 socklnd - any kernel supported by Lustre,
130 qswlnd - Qsnet kernel modules 5.20 and later,
131 openiblnd - IbGold 1.8.2,
132 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5.
133 viblnd - Voltaire ibhost 3.4.5 and later,
134 ciblnd - Topspin 3.2.0,
135 iiblnd - Infiniserv 3.3 + PathBits patch,
136 gmlnd - GM 2.1.22 and later,
137 mxlnd - MX 1.2.1 or later,
138 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
142 Description: ASSERTION(me == md->md_me) failed in lnet_match_md()
146 Description: increase send queue size for ciblnd/openiblnd
150 Description: new userspace socklnd
151 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
152 with new one - usocklnd.
154 Severity : enhancement
156 Description: Console message flood
157 Details : Make cdls ratelimiting more tunable by adding several tunable in
158 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
159 /proc/sys/lnet/console_backoff.
161 --------------------------------------------------------------------------------
163 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
165 * Support for networks:
166 socklnd - any kernel supported by Lustre,
167 qswlnd - Qsnet kernel modules 5.20 and later,
168 openiblnd - IbGold 1.8.2,
169 o2iblnd - OFED 1.1 and 1.2,
170 viblnd - Voltaire ibhost 3.4.5 and later,
171 ciblnd - Topspin 3.2.0,
172 iiblnd - Infiniserv 3.3 + PathBits patch,
173 gmlnd - GM 2.1.22 and later,
174 mxlnd - MX 1.2.1 or later,
175 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
179 Description: /proc/sys/lnet has non-sysctl entries
180 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
184 Description: TOE Kernel panic by ksocklnd
185 Details : offloaded sockets provide their own implementation of sendpage,
186 can't call tcp_sendpage() directly
190 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
191 Details : races between lnd_shutdown and peer creation prevent
192 lnd_shutdown from finishing.
196 Description: open files rlimit 1024 reached while liblustre testing
197 Details : ulnds/socklnd must close open socket after unsuccessful
202 Description: build error
203 Details : fix typos in gmlnd, ptllnd and viblnd
205 ------------------------------------------------------------------------------
207 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
209 * Support for networks:
210 socklnd - kernels up to 2.6.16,
211 qswlnd - Qsnet kernel modules 5.20 and later,
212 openiblnd - IbGold 1.8.2,
213 o2iblnd - OFED 1.1 and 1.2
214 viblnd - Voltaire ibhost 3.4.5 and later,
215 ciblnd - Topspin 3.2.0,
216 iiblnd - Infiniserv 3.3 + PathBits patch,
217 gmlnd - GM 2.1.22 and later,
218 mxlnd - MX 1.2.1 or later,
219 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
221 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
223 * Support for networks:
224 socklnd - kernels up to 2.6.16,
225 qswlnd - Qsnet kernel modules 5.20 and later,
226 openiblnd - IbGold 1.8.2,
228 viblnd - Voltaire ibhost 3.4.5 and later,
229 ciblnd - Topspin 3.2.0,
230 iiblnd - Infiniserv 3.3 + PathBits patch,
231 gmlnd - GM 2.1.22 and later,
232 mxlnd - MX 1.2.1 or later,
233 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
237 Description: Initialize cpumask before use
241 Description: ASSERTION failures when upgrading to the patchless zero-copy
243 Details : This bug affects "rolling upgrades", causing an inconsistent
244 protocol version negotiation and subsequent assertion failure
245 during rolling upgrades after the first wave of upgrades.
249 Details : Change "dropped message" CERRORs to D_NETERROR so they are
250 logged instead of creating "console chatter" when a lustre
251 timeout races with normal RPC completion.
254 Details : lnet_clear_peer_table can wait forever if user forgets to
258 Details : libcfs_id2str should check pid against LNET_PID_ANY.
262 Description: added LNET self test
263 Details : landing b_self_test
268 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
270 Details : do_div() macro is used incorrectly.
272 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
276 Description: make panic on lbug configurable
280 Description: Add OFED1.2 support to o2iblnd
281 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
282 are installed (other than kernel's in-tree infiniband), there
283 could be some problem while insmod o2iblnd (mismatch CRC of
285 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
286 this link provides solution:
287 https://bugs.openfabrics.org/show_bug.cgi?id=355
288 if extra Module.symvers is not supported in kernel, we will
289 have to run the script in bug 12316 to update
290 $LINUX/module.symvers before building o2iblnd.
291 More details about this are in bug 12316.
293 ------------------------------------------------------------------------------
295 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
296 * version 1.4.10 / 1.6.0
297 * Support for networks:
298 socklnd - kernels up to 2.6.16,
299 qswlnd - Qsnet kernel modules 5.20 and later,
300 openiblnd - IbGold 1.8.2,
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: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
312 possibly accessed in kptllnd_shutdown. Ptllnd should init
313 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
317 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
321 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
325 Description: the_lnet.ln_finalizing was not set when the current thread is
326 about to complete messages. It only affects multi-threaded
332 Description: Changed the default kqswlnd ntxmsg=512
337 Description: Assertion failure in kernel ptllnd caused by posting passive
338 bulk buffers before connection establishment complete.
343 Description: A race in kernel ptllnd between deleting a peer and posting
344 new communications for it could hang communications -
345 manifesting as "Unexpectedly long timeout" messages.
350 Description: Kernel ptllnd lock ordering issue could hang a node.
355 Description: node crash on socket teardown race
358 Frequency : 'lctl peer_list' issued on a mx net
360 Description: Enable lctl's peer_list for MXLND
363 Frequency : after Ptllnd timeouts and portals congestion
365 Description: Credit overflows
366 Details : This was a bug in ptllnd connection establishment. The fix
367 implements better peer stamps to disambiguate connection
368 establishment and ensure both peers enter the credit flow
369 state machine consistently.
374 Description: kptllnd didn't propagate some network errors up to LNET
375 Details : This bug was spotted while investigating 11394. The fix
376 ensures network errors on sends and bulk transfers are
377 propagated to LNET/lustre correctly.
379 Severity : enhancement
381 Description: Fixed console chatter in case of -ETIMEDOUT.
383 Severity : enhancement
385 Description: Added D_NETTRACE for recording network packet history
386 (initially only for ptllnd). Also a separate userspace
387 ptllnd facility to gather history which should really be
388 covered by D_NETTRACE too, if only CDEBUG recorded history in
394 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
395 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
396 callback can occur before a connection has actually been
397 established. This caused an assertion failure previously.
399 Severity : enhancement
401 Description: Multiple instances for o2iblnd
402 Details : Allow multiple instances of o2iblnd to enable networking over
403 multiple HCAs and routing between them.
407 Description: lnet deadlock in router_checker
408 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
409 into BH locks to eliminate potential deadlock caused by
410 ksocknal_data_ready() preempting code holding these locks.
414 Description: Millions of failed socklnd connection attempts cause a very slow FS
415 Details : added a new route flag ksnr_scheduled to distinguish from
416 ksnr_connecting, so that a peer connection request is only turned
417 down for race concerns when an active connection to the same peer
418 is under progress (instead of just being scheduled).
420 ------------------------------------------------------------------------------
422 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
424 * Support for networks:
425 socklnd - kernels up to 2.6.16
426 qswlnd - Qsnet kernel modules 5.20 and later
427 openiblnd - IbGold 1.8.2
429 viblnd - Voltaire ibhost 3.4.5 and later
430 ciblnd - Topspin 3.2.0
431 iiblnd - Infiniserv 3.3 + PathBits patch
432 gmlnd - GM 2.1.22 and later
433 mxlnd - MX 1.2.1 or later
434 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
437 Severity : major on XT3
439 Description: libcfs overwrites /proc/sys/portals
440 Details : libcfs created a symlink from /proc/sys/portals to
441 /proc/sys/lnet for backwards compatibility. This is no
442 longer required and makes the Cray portals /proc variables
447 Description: OFED FMR API change
448 Details : This changes parameter usage to reflect a change in
449 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
450 that FMR support is only used in experimental versions of the
451 o2iblnd - this change does not affect standard usage at all.
453 Severity : enhancement
455 Description: new ko2iblnd module parameter: ib_mtu
456 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
457 HCAs. You can avoid this problem by setting the MTU to 1024
458 using this module parameter.
460 Severity : enhancement
461 Bugzilla : 11118/11620
462 Description: ptllnd small request message buffer alignment fix
463 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
464 Round up small message size on sends in case this option
465 is not supported. 11620 was a defect in the initial
466 implementation which effectively asserted all peers had to be
467 running the correct protocol version which was fixed by always
468 NAK-ing such requests and handling any misalignments they
473 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
474 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
475 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
477 Severity : enhancement
479 Description: Patchless ZC(zero copy) socklnd
480 Details : New protocol for socklnd, socklnd can support zero copy without
481 kernel patch, it's compatible with old socklnd. Checksum is
482 moved from tunables to modparams.
486 Description: When ksocknal_del_peer() is called upon a peer whose
487 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
488 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
491 Frequency : when ptlrpc is under heavy use and runs out of request buffer
493 Description: In lnet_match_blocked_msg(), md can be used without holding a
497 Frequency : very rarely
499 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
500 If connd connects a route which has been closed by
501 ksocknal_shutdown(), ksocknal_create_routes() may create new
502 routes which hold references on the peer, causing shutdown
503 process to wait for peer to disappear forever.
505 Severity : enhancement
507 Description: Dump XT3 portals traces on kptllnd timeout
508 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
509 dump Cray portals debug traces to a file. The kptllnd module
510 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
511 is the basename of the dump file.
514 Frequency : infrequent
516 Description: kernel ptllnd fix bug in connection re-establishment
517 Details : Kernel ptllnd could produce protocol errors e.g. illegal
518 matchbits and/or violate the credit flow protocol when trying
519 to re-establish a connection with a peer after an error or
522 Severity : enhancement
524 Description: Allow /proc/sys/lnet/debug to be set symbolically
525 Details : Allow debug and subsystem debug values to be read/set by name
526 in addition to numerically, for ease of use.
529 Frequency : only in configurations with LNET routers
531 Description: routes automatically marked down and recovered
532 Details : In configurations with LNET routers if a router fails routers
533 now actively try to recover routes that are down, unless they
534 are marked down by an administrator.
536 ------------------------------------------------------------------------------
538 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
541 Frequency : very rarely, in configurations with LNET routers and TCP
543 Description: incorrect data written to files on OSTs
544 Details : In certain high-load conditions incorrect data may be written
545 to files on the OST when using TCP networks.
547 ------------------------------------------------------------------------------
549 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
551 - rework CDEBUG messages rate-limiting mechanism b=10375
552 - add per-socket tunables for socklnd if the kernel is patched b=10327
554 ------------------------------------------------------------------------------
556 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
558 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
559 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
560 - quiet LNET startup LNI message for liblustre b=10128
561 - Better console error messages if 'ip2nets' can't match an IP address
562 - Fixed overflow/use-before-set bugs in linux-time.h
563 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
564 - LNET teardown failed an assertion about the route table being empty
565 - Fixed a crash in LNetEQPoll(<invalid handle>)
566 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
567 - improve debug message for liblustre/Catamount nodes (b=10116)
569 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
570 * Configuration change for the XT3
571 The PTLLND is now used to run Lustre over Portals on the XT3.
572 The configure option(s) --with-cray-portals are no longer
573 used. Rather --with-portals=<path-to-portals-includes> is
574 used to enable building on the XT3. In addition to enable
575 XT3 specific features the option --enable-cray-xt3 must be
578 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
579 * Portals has been removed, replaced by LNET.
580 LNET is new networking infrastructure for Lustre, it includes a
581 reorganized network configuration mode (see the user
582 documentation for full details) as well as support for routing
583 between different network fabrics. Lustre Networking Devices
584 (LNDS) for the supported network fabrics have also been created
585 for this new infrastructure.
587 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
592 Frequency : rare (large Voltaire clusters only)
594 Description: the default number of reserved transmit descriptors was too low
595 for some large clusters
596 Details : As a workaround, the number was increased. A proper fix includes
599 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
604 Frequency : occasional (large-scale events, cluster reboot, network failure)
606 Description: too many error messages on console obscure actual problem and
607 can slow down/panic server, or cause recovery to fail repeatedly
608 Details : enable rate-limiting of console error messages, and some messages
609 that were console errors now only go to the kernel log
611 Severity : enhancement
613 Description: add /proc/sys/portals/catastrophe entry which will report if
614 that node has previously LBUGged
616 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
618 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
620 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
622 - handle error return code in kranal_check_fma_rx() (5915,6054)
624 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
626 - update vibnal (Voltaire IB NAL)
627 - update gmnal (Myrinet NAL), gmnalid
629 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
631 * Landed portals:b_port_step as follows...
633 - removed CFS_DECL_SPIN*
634 just use 'spinlock_t' and initialise with spin_lock_init()
636 - removed CFS_DECL_MUTEX*
637 just use 'struct semaphore' and initialise with init_mutex()
639 - removed CFS_DECL_RWSEM*
640 just use 'struct rw_semaphore' and initialise with init_rwsem()
642 - renamed cfs_sleep_chan -> cfs_waitq
643 cfs_sleep_link -> cfs_waitlink
645 - fixed race in linux version of arch-independent socknal
646 (the ENOMEM/EAGAIN decision).
648 - Didn't fix problems in Darwin version of arch-independent socknal
649 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
651 - removed libcfs types from non-socknal header files (only some types
652 in the header files had been changed; the .c files hadn't been