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: Continuous recovery on 33 of 413 nodes after lustre oss failure
23 Details : Lost reference on conn prevents peer from being destroyed, which
24 could prevent new peer creation if peer count has reached upper
29 Description: LNET Selftest results in Soft lockup on OSS CPU
30 Details : only hits when 8 or more o2ib clients involved and a session is
31 torn down with 'lst end_session' without preceeding 'lst stop'.
35 Description: concurrent_sends in IB LNDs should not be changeable at run time
36 Details : concurrent_sends in IB LNDs should not be changeable at run time
40 Description: ptl_send_rpc hits LASSERT when ptl_send_buf fails
41 Details : only hits under out-of-memory situations
44 -------------------------------------------------------------------------------
47 04-26-2008 Sun Microsystems, Inc.
49 * Support for networks:
50 socklnd - any kernel supported by Lustre,
51 qswlnd - Qsnet kernel modules 5.20 and later,
52 openiblnd - IbGold 1.8.2,
53 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
54 viblnd - Voltaire ibhost 3.4.5 and later,
55 ciblnd - Topspin 3.2.0,
56 iiblnd - Infiniserv 3.3 + PathBits patch,
57 gmlnd - GM 2.1.22 and later,
58 mxlnd - MX 1.2.1 or later,
59 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
63 Description: excessive debug information removed
64 Details : excessive debug information removed
68 Description: ksocknal_create_conn() hit ASSERTION during connection race
69 Details : ksocknal_create_conn() hit ASSERTION during connection race
73 Description: ksocknal_send_hello() hit ASSERTION while connecting race
74 Details : ksocknal_send_hello() hit ASSERTION while connecting race
78 Description: o2iblnd/ptllnd credit deadlock in a routed config.
79 Details : o2iblnd/ptllnd credit deadlock in a routed config.
83 Description: High load after starting lnet
84 Details : gmlnd should sleep in rx thread in interruptible way. Otherwise,
85 uptime utility reports high load that looks confusingly.
89 Description: ksocklnd fails to establish connection if accept_port is high
90 Details : PID remapping must not be done for active (outgoing) connections
92 --------------------------------------------------------------------------------
94 2008-01-11 Sun Microsystems, Inc.
96 * Support for networks:
97 socklnd - any kernel supported by Lustre,
98 qswlnd - Qsnet kernel modules 5.20 and later,
99 openiblnd - IbGold 1.8.2,
100 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
101 viblnd - Voltaire ibhost 3.4.5 and later,
102 ciblnd - Topspin 3.2.0,
103 iiblnd - Infiniserv 3.3 + PathBits patch,
104 gmlnd - GM 2.1.22 and later,
105 mxlnd - MX 1.2.1 or later,
106 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
109 Description: liblustre network error
110 Details : liblustre clients should understand LNET_ACCEPT_PORT environment
111 variable even if they don't start lnet acceptor.
115 Description: Strange message from lnet (Ignoring prediction from the future)
116 Details : Incorrect calculation of peer's last_alive value in ksocklnd
118 --------------------------------------------------------------------------------
120 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
122 * Support for networks:
123 socklnd - any kernel supported by Lustre,
124 qswlnd - Qsnet kernel modules 5.20 and later,
125 openiblnd - IbGold 1.8.2,
126 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5.
127 viblnd - Voltaire ibhost 3.4.5 and later,
128 ciblnd - Topspin 3.2.0,
129 iiblnd - Infiniserv 3.3 + PathBits patch,
130 gmlnd - GM 2.1.22 and later,
131 mxlnd - MX 1.2.1 or later,
132 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
136 Description: ASSERTION(me == md->md_me) failed in lnet_match_md()
140 Description: increase send queue size for ciblnd/openiblnd
144 Description: new userspace socklnd
145 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
146 with new one - usocklnd.
148 Severity : enhancement
150 Description: Console message flood
151 Details : Make cdls ratelimiting more tunable by adding several tunable in
152 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
153 /proc/sys/lnet/console_backoff.
155 --------------------------------------------------------------------------------
157 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
159 * Support for networks:
160 socklnd - any kernel supported by Lustre,
161 qswlnd - Qsnet kernel modules 5.20 and later,
162 openiblnd - IbGold 1.8.2,
163 o2iblnd - OFED 1.1 and 1.2,
164 viblnd - Voltaire ibhost 3.4.5 and later,
165 ciblnd - Topspin 3.2.0,
166 iiblnd - Infiniserv 3.3 + PathBits patch,
167 gmlnd - GM 2.1.22 and later,
168 mxlnd - MX 1.2.1 or later,
169 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
173 Description: /proc/sys/lnet has non-sysctl entries
174 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
178 Description: TOE Kernel panic by ksocklnd
179 Details : offloaded sockets provide their own implementation of sendpage,
180 can't call tcp_sendpage() directly
184 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
185 Details : races between lnd_shutdown and peer creation prevent
186 lnd_shutdown from finishing.
190 Description: open files rlimit 1024 reached while liblustre testing
191 Details : ulnds/socklnd must close open socket after unsuccessful
196 Description: build error
197 Details : fix typos in gmlnd, ptllnd and viblnd
199 ------------------------------------------------------------------------------
201 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
203 * Support for networks:
204 socklnd - kernels up to 2.6.16,
205 qswlnd - Qsnet kernel modules 5.20 and later,
206 openiblnd - IbGold 1.8.2,
207 o2iblnd - OFED 1.1 and 1.2
208 viblnd - Voltaire ibhost 3.4.5 and later,
209 ciblnd - Topspin 3.2.0,
210 iiblnd - Infiniserv 3.3 + PathBits patch,
211 gmlnd - GM 2.1.22 and later,
212 mxlnd - MX 1.2.1 or later,
213 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
215 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
217 * Support for networks:
218 socklnd - kernels up to 2.6.16,
219 qswlnd - Qsnet kernel modules 5.20 and later,
220 openiblnd - IbGold 1.8.2,
222 viblnd - Voltaire ibhost 3.4.5 and later,
223 ciblnd - Topspin 3.2.0,
224 iiblnd - Infiniserv 3.3 + PathBits patch,
225 gmlnd - GM 2.1.22 and later,
226 mxlnd - MX 1.2.1 or later,
227 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
231 Description: Initialize cpumask before use
235 Description: ASSERTION failures when upgrading to the patchless zero-copy
237 Details : This bug affects "rolling upgrades", causing an inconsistent
238 protocol version negotiation and subsequent assertion failure
239 during rolling upgrades after the first wave of upgrades.
243 Details : Change "dropped message" CERRORs to D_NETERROR so they are
244 logged instead of creating "console chatter" when a lustre
245 timeout races with normal RPC completion.
248 Details : lnet_clear_peer_table can wait forever if user forgets to
252 Details : libcfs_id2str should check pid against LNET_PID_ANY.
256 Description: added LNET self test
257 Details : landing b_self_test
262 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
264 Details : do_div() macro is used incorrectly.
266 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
270 Description: make panic on lbug configurable
274 Description: Add OFED1.2 support to o2iblnd
275 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
276 are installed (other than kernel's in-tree infiniband), there
277 could be some problem while insmod o2iblnd (mismatch CRC of
279 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
280 this link provides solution:
281 https://bugs.openfabrics.org/show_bug.cgi?id=355
282 if extra Module.symvers is not supported in kernel, we will
283 have to run the script in bug 12316 to update
284 $LINUX/module.symvers before building o2iblnd.
285 More details about this are in bug 12316.
287 ------------------------------------------------------------------------------
289 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
290 * version 1.4.10 / 1.6.0
291 * Support for networks:
292 socklnd - kernels up to 2.6.16,
293 qswlnd - Qsnet kernel modules 5.20 and later,
294 openiblnd - IbGold 1.8.2,
296 viblnd - Voltaire ibhost 3.4.5 and later,
297 ciblnd - Topspin 3.2.0,
298 iiblnd - Infiniserv 3.3 + PathBits patch,
299 gmlnd - GM 2.1.22 and later,
300 mxlnd - MX 1.2.1 or later,
301 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
305 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
306 possibly accessed in kptllnd_shutdown. Ptllnd should init
307 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
311 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
315 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
319 Description: the_lnet.ln_finalizing was not set when the current thread is
320 about to complete messages. It only affects multi-threaded
326 Description: Changed the default kqswlnd ntxmsg=512
331 Description: Assertion failure in kernel ptllnd caused by posting passive
332 bulk buffers before connection establishment complete.
337 Description: A race in kernel ptllnd between deleting a peer and posting
338 new communications for it could hang communications -
339 manifesting as "Unexpectedly long timeout" messages.
344 Description: Kernel ptllnd lock ordering issue could hang a node.
349 Description: node crash on socket teardown race
352 Frequency : 'lctl peer_list' issued on a mx net
354 Description: Enable lctl's peer_list for MXLND
357 Frequency : after Ptllnd timeouts and portals congestion
359 Description: Credit overflows
360 Details : This was a bug in ptllnd connection establishment. The fix
361 implements better peer stamps to disambiguate connection
362 establishment and ensure both peers enter the credit flow
363 state machine consistently.
368 Description: kptllnd didn't propagate some network errors up to LNET
369 Details : This bug was spotted while investigating 11394. The fix
370 ensures network errors on sends and bulk transfers are
371 propagated to LNET/lustre correctly.
373 Severity : enhancement
375 Description: Fixed console chatter in case of -ETIMEDOUT.
377 Severity : enhancement
379 Description: Added D_NETTRACE for recording network packet history
380 (initially only for ptllnd). Also a separate userspace
381 ptllnd facility to gather history which should really be
382 covered by D_NETTRACE too, if only CDEBUG recorded history in
388 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
389 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
390 callback can occur before a connection has actually been
391 established. This caused an assertion failure previously.
393 Severity : enhancement
395 Description: Multiple instances for o2iblnd
396 Details : Allow multiple instances of o2iblnd to enable networking over
397 multiple HCAs and routing between them.
401 Description: lnet deadlock in router_checker
402 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
403 into BH locks to eliminate potential deadlock caused by
404 ksocknal_data_ready() preempting code holding these locks.
408 Description: Millions of failed socklnd connection attempts cause a very slow FS
409 Details : added a new route flag ksnr_scheduled to distinguish from
410 ksnr_connecting, so that a peer connection request is only turned
411 down for race concerns when an active connection to the same peer
412 is under progress (instead of just being scheduled).
414 ------------------------------------------------------------------------------
416 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
418 * Support for networks:
419 socklnd - kernels up to 2.6.16
420 qswlnd - Qsnet kernel modules 5.20 and later
421 openiblnd - IbGold 1.8.2
423 viblnd - Voltaire ibhost 3.4.5 and later
424 ciblnd - Topspin 3.2.0
425 iiblnd - Infiniserv 3.3 + PathBits patch
426 gmlnd - GM 2.1.22 and later
427 mxlnd - MX 1.2.1 or later
428 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
431 Severity : major on XT3
433 Description: libcfs overwrites /proc/sys/portals
434 Details : libcfs created a symlink from /proc/sys/portals to
435 /proc/sys/lnet for backwards compatibility. This is no
436 longer required and makes the Cray portals /proc variables
441 Description: OFED FMR API change
442 Details : This changes parameter usage to reflect a change in
443 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
444 that FMR support is only used in experimental versions of the
445 o2iblnd - this change does not affect standard usage at all.
447 Severity : enhancement
449 Description: new ko2iblnd module parameter: ib_mtu
450 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
451 HCAs. You can avoid this problem by setting the MTU to 1024
452 using this module parameter.
454 Severity : enhancement
455 Bugzilla : 11118/11620
456 Description: ptllnd small request message buffer alignment fix
457 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
458 Round up small message size on sends in case this option
459 is not supported. 11620 was a defect in the initial
460 implementation which effectively asserted all peers had to be
461 running the correct protocol version which was fixed by always
462 NAK-ing such requests and handling any misalignments they
467 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
468 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
469 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
471 Severity : enhancement
473 Description: Patchless ZC(zero copy) socklnd
474 Details : New protocol for socklnd, socklnd can support zero copy without
475 kernel patch, it's compatible with old socklnd. Checksum is
476 moved from tunables to modparams.
480 Description: When ksocknal_del_peer() is called upon a peer whose
481 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
482 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
485 Frequency : when ptlrpc is under heavy use and runs out of request buffer
487 Description: In lnet_match_blocked_msg(), md can be used without holding a
491 Frequency : very rarely
493 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
494 If connd connects a route which has been closed by
495 ksocknal_shutdown(), ksocknal_create_routes() may create new
496 routes which hold references on the peer, causing shutdown
497 process to wait for peer to disappear forever.
499 Severity : enhancement
501 Description: Dump XT3 portals traces on kptllnd timeout
502 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
503 dump Cray portals debug traces to a file. The kptllnd module
504 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
505 is the basename of the dump file.
508 Frequency : infrequent
510 Description: kernel ptllnd fix bug in connection re-establishment
511 Details : Kernel ptllnd could produce protocol errors e.g. illegal
512 matchbits and/or violate the credit flow protocol when trying
513 to re-establish a connection with a peer after an error or
516 Severity : enhancement
518 Description: Allow /proc/sys/lnet/debug to be set symbolically
519 Details : Allow debug and subsystem debug values to be read/set by name
520 in addition to numerically, for ease of use.
523 Frequency : only in configurations with LNET routers
525 Description: routes automatically marked down and recovered
526 Details : In configurations with LNET routers if a router fails routers
527 now actively try to recover routes that are down, unless they
528 are marked down by an administrator.
530 ------------------------------------------------------------------------------
532 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
535 Frequency : very rarely, in configurations with LNET routers and TCP
537 Description: incorrect data written to files on OSTs
538 Details : In certain high-load conditions incorrect data may be written
539 to files on the OST when using TCP networks.
541 ------------------------------------------------------------------------------
543 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
545 - rework CDEBUG messages rate-limiting mechanism b=10375
546 - add per-socket tunables for socklnd if the kernel is patched b=10327
548 ------------------------------------------------------------------------------
550 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
552 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
553 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
554 - quiet LNET startup LNI message for liblustre b=10128
555 - Better console error messages if 'ip2nets' can't match an IP address
556 - Fixed overflow/use-before-set bugs in linux-time.h
557 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
558 - LNET teardown failed an assertion about the route table being empty
559 - Fixed a crash in LNetEQPoll(<invalid handle>)
560 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
561 - improve debug message for liblustre/Catamount nodes (b=10116)
563 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
564 * Configuration change for the XT3
565 The PTLLND is now used to run Lustre over Portals on the XT3.
566 The configure option(s) --with-cray-portals are no longer
567 used. Rather --with-portals=<path-to-portals-includes> is
568 used to enable building on the XT3. In addition to enable
569 XT3 specific features the option --enable-cray-xt3 must be
572 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
573 * Portals has been removed, replaced by LNET.
574 LNET is new networking infrastructure for Lustre, it includes a
575 reorganized network configuration mode (see the user
576 documentation for full details) as well as support for routing
577 between different network fabrics. Lustre Networking Devices
578 (LNDS) for the supported network fabrics have also been created
579 for this new infrastructure.
581 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
586 Frequency : rare (large Voltaire clusters only)
588 Description: the default number of reserved transmit descriptors was too low
589 for some large clusters
590 Details : As a workaround, the number was increased. A proper fix includes
593 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
598 Frequency : occasional (large-scale events, cluster reboot, network failure)
600 Description: too many error messages on console obscure actual problem and
601 can slow down/panic server, or cause recovery to fail repeatedly
602 Details : enable rate-limiting of console error messages, and some messages
603 that were console errors now only go to the kernel log
605 Severity : enhancement
607 Description: add /proc/sys/portals/catastrophe entry which will report if
608 that node has previously LBUGged
610 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
612 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
614 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
616 - handle error return code in kranal_check_fma_rx() (5915,6054)
618 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
620 - update vibnal (Voltaire IB NAL)
621 - update gmnal (Myrinet NAL), gmnalid
623 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
625 * Landed portals:b_port_step as follows...
627 - removed CFS_DECL_SPIN*
628 just use 'spinlock_t' and initialise with spin_lock_init()
630 - removed CFS_DECL_MUTEX*
631 just use 'struct semaphore' and initialise with init_mutex()
633 - removed CFS_DECL_RWSEM*
634 just use 'struct rw_semaphore' and initialise with init_rwsem()
636 - renamed cfs_sleep_chan -> cfs_waitq
637 cfs_sleep_link -> cfs_waitlink
639 - fixed race in linux version of arch-independent socknal
640 (the ENOMEM/EAGAIN decision).
642 - Didn't fix problems in Darwin version of arch-independent socknal
643 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
645 - removed libcfs types from non-socknal header files (only some types
646 in the header files had been changed; the .c files hadn't been