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: ptl_send_rpc hits LASSERT when ptl_send_buf fails
23 Details : only hits under out-of-memory situations
26 -------------------------------------------------------------------------------
29 04-26-2008 Sun Microsystems, Inc.
31 * Support for networks:
32 socklnd - any kernel supported by Lustre,
33 qswlnd - Qsnet kernel modules 5.20 and later,
34 openiblnd - IbGold 1.8.2,
35 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
36 viblnd - Voltaire ibhost 3.4.5 and later,
37 ciblnd - Topspin 3.2.0,
38 iiblnd - Infiniserv 3.3 + PathBits patch,
39 gmlnd - GM 2.1.22 and later,
40 mxlnd - MX 1.2.1 or later,
41 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
45 Description: ksocknal_create_conn() hit ASSERTION during connection race
46 Details : ksocknal_create_conn() hit ASSERTION during connection race
50 Description: ksocknal_send_hello() hit ASSERTION while connecting race
51 Details : ksocknal_send_hello() hit ASSERTION while connecting race
55 Description: o2iblnd/ptllnd credit deadlock in a routed config.
56 Details : o2iblnd/ptllnd credit deadlock in a routed config.
60 Description: High load after starting lnet
61 Details : gmlnd should sleep in rx thread in interruptible way. Otherwise,
62 uptime utility reports high load that looks confusingly.
66 Description: ksocklnd fails to establish connection if accept_port is high
67 Details : PID remapping must not be done for active (outgoing) connections
69 --------------------------------------------------------------------------------
71 2008-01-11 Sun Microsystems, Inc.
73 * Support for networks:
74 socklnd - any kernel supported by Lustre,
75 qswlnd - Qsnet kernel modules 5.20 and later,
76 openiblnd - IbGold 1.8.2,
77 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
78 viblnd - Voltaire ibhost 3.4.5 and later,
79 ciblnd - Topspin 3.2.0,
80 iiblnd - Infiniserv 3.3 + PathBits patch,
81 gmlnd - GM 2.1.22 and later,
82 mxlnd - MX 1.2.1 or later,
83 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
86 Description: liblustre network error
87 Details : liblustre clients should understand LNET_ACCEPT_PORT environment
88 variable even if they don't start lnet acceptor.
92 Description: Strange message from lnet (Ignoring prediction from the future)
93 Details : Incorrect calculation of peer's last_alive value in ksocklnd
95 --------------------------------------------------------------------------------
97 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
99 * Support for networks:
100 socklnd - any kernel supported by Lustre,
101 qswlnd - Qsnet kernel modules 5.20 and later,
102 openiblnd - IbGold 1.8.2,
103 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5.
104 viblnd - Voltaire ibhost 3.4.5 and later,
105 ciblnd - Topspin 3.2.0,
106 iiblnd - Infiniserv 3.3 + PathBits patch,
107 gmlnd - GM 2.1.22 and later,
108 mxlnd - MX 1.2.1 or later,
109 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
113 Description: ASSERTION(me == md->md_me) failed in lnet_match_md()
117 Description: increase send queue size for ciblnd/openiblnd
121 Description: new userspace socklnd
122 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
123 with new one - usocklnd.
125 Severity : enhancement
127 Description: Console message flood
128 Details : Make cdls ratelimiting more tunable by adding several tunable in
129 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
130 /proc/sys/lnet/console_backoff.
132 --------------------------------------------------------------------------------
134 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
136 * Support for networks:
137 socklnd - any kernel supported by Lustre,
138 qswlnd - Qsnet kernel modules 5.20 and later,
139 openiblnd - IbGold 1.8.2,
140 o2iblnd - OFED 1.1 and 1.2,
141 viblnd - Voltaire ibhost 3.4.5 and later,
142 ciblnd - Topspin 3.2.0,
143 iiblnd - Infiniserv 3.3 + PathBits patch,
144 gmlnd - GM 2.1.22 and later,
145 mxlnd - MX 1.2.1 or later,
146 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
150 Description: /proc/sys/lnet has non-sysctl entries
151 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
155 Description: TOE Kernel panic by ksocklnd
156 Details : offloaded sockets provide their own implementation of sendpage,
157 can't call tcp_sendpage() directly
161 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
162 Details : races between lnd_shutdown and peer creation prevent
163 lnd_shutdown from finishing.
167 Description: open files rlimit 1024 reached while liblustre testing
168 Details : ulnds/socklnd must close open socket after unsuccessful
173 Description: build error
174 Details : fix typos in gmlnd, ptllnd and viblnd
176 ------------------------------------------------------------------------------
178 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
180 * Support for networks:
181 socklnd - kernels up to 2.6.16,
182 qswlnd - Qsnet kernel modules 5.20 and later,
183 openiblnd - IbGold 1.8.2,
184 o2iblnd - OFED 1.1 and 1.2
185 viblnd - Voltaire ibhost 3.4.5 and later,
186 ciblnd - Topspin 3.2.0,
187 iiblnd - Infiniserv 3.3 + PathBits patch,
188 gmlnd - GM 2.1.22 and later,
189 mxlnd - MX 1.2.1 or later,
190 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
192 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
194 * Support for networks:
195 socklnd - kernels up to 2.6.16,
196 qswlnd - Qsnet kernel modules 5.20 and later,
197 openiblnd - IbGold 1.8.2,
199 viblnd - Voltaire ibhost 3.4.5 and later,
200 ciblnd - Topspin 3.2.0,
201 iiblnd - Infiniserv 3.3 + PathBits patch,
202 gmlnd - GM 2.1.22 and later,
203 mxlnd - MX 1.2.1 or later,
204 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
208 Description: Initialize cpumask before use
212 Description: ASSERTION failures when upgrading to the patchless zero-copy
214 Details : This bug affects "rolling upgrades", causing an inconsistent
215 protocol version negotiation and subsequent assertion failure
216 during rolling upgrades after the first wave of upgrades.
220 Details : Change "dropped message" CERRORs to D_NETERROR so they are
221 logged instead of creating "console chatter" when a lustre
222 timeout races with normal RPC completion.
225 Details : lnet_clear_peer_table can wait forever if user forgets to
229 Details : libcfs_id2str should check pid against LNET_PID_ANY.
233 Description: added LNET self test
234 Details : landing b_self_test
239 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
241 Details : do_div() macro is used incorrectly.
243 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
247 Description: make panic on lbug configurable
251 Description: Add OFED1.2 support to o2iblnd
252 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
253 are installed (other than kernel's in-tree infiniband), there
254 could be some problem while insmod o2iblnd (mismatch CRC of
256 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
257 this link provides solution:
258 https://bugs.openfabrics.org/show_bug.cgi?id=355
259 if extra Module.symvers is not supported in kernel, we will
260 have to run the script in bug 12316 to update
261 $LINUX/module.symvers before building o2iblnd.
262 More details about this are in bug 12316.
264 ------------------------------------------------------------------------------
266 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
267 * version 1.4.10 / 1.6.0
268 * Support for networks:
269 socklnd - kernels up to 2.6.16,
270 qswlnd - Qsnet kernel modules 5.20 and later,
271 openiblnd - IbGold 1.8.2,
273 viblnd - Voltaire ibhost 3.4.5 and later,
274 ciblnd - Topspin 3.2.0,
275 iiblnd - Infiniserv 3.3 + PathBits patch,
276 gmlnd - GM 2.1.22 and later,
277 mxlnd - MX 1.2.1 or later,
278 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
282 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
283 possibly accessed in kptllnd_shutdown. Ptllnd should init
284 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
288 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
292 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
296 Description: the_lnet.ln_finalizing was not set when the current thread is
297 about to complete messages. It only affects multi-threaded
303 Description: Changed the default kqswlnd ntxmsg=512
308 Description: Assertion failure in kernel ptllnd caused by posting passive
309 bulk buffers before connection establishment complete.
314 Description: A race in kernel ptllnd between deleting a peer and posting
315 new communications for it could hang communications -
316 manifesting as "Unexpectedly long timeout" messages.
321 Description: Kernel ptllnd lock ordering issue could hang a node.
326 Description: node crash on socket teardown race
329 Frequency : 'lctl peer_list' issued on a mx net
331 Description: Enable lctl's peer_list for MXLND
334 Frequency : after Ptllnd timeouts and portals congestion
336 Description: Credit overflows
337 Details : This was a bug in ptllnd connection establishment. The fix
338 implements better peer stamps to disambiguate connection
339 establishment and ensure both peers enter the credit flow
340 state machine consistently.
345 Description: kptllnd didn't propagate some network errors up to LNET
346 Details : This bug was spotted while investigating 11394. The fix
347 ensures network errors on sends and bulk transfers are
348 propagated to LNET/lustre correctly.
350 Severity : enhancement
352 Description: Fixed console chatter in case of -ETIMEDOUT.
354 Severity : enhancement
356 Description: Added D_NETTRACE for recording network packet history
357 (initially only for ptllnd). Also a separate userspace
358 ptllnd facility to gather history which should really be
359 covered by D_NETTRACE too, if only CDEBUG recorded history in
365 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
366 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
367 callback can occur before a connection has actually been
368 established. This caused an assertion failure previously.
370 Severity : enhancement
372 Description: Multiple instances for o2iblnd
373 Details : Allow multiple instances of o2iblnd to enable networking over
374 multiple HCAs and routing between them.
378 Description: lnet deadlock in router_checker
379 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
380 into BH locks to eliminate potential deadlock caused by
381 ksocknal_data_ready() preempting code holding these locks.
385 Description: Millions of failed socklnd connection attempts cause a very slow FS
386 Details : added a new route flag ksnr_scheduled to distinguish from
387 ksnr_connecting, so that a peer connection request is only turned
388 down for race concerns when an active connection to the same peer
389 is under progress (instead of just being scheduled).
391 ------------------------------------------------------------------------------
393 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
395 * Support for networks:
396 socklnd - kernels up to 2.6.16
397 qswlnd - Qsnet kernel modules 5.20 and later
398 openiblnd - IbGold 1.8.2
400 viblnd - Voltaire ibhost 3.4.5 and later
401 ciblnd - Topspin 3.2.0
402 iiblnd - Infiniserv 3.3 + PathBits patch
403 gmlnd - GM 2.1.22 and later
404 mxlnd - MX 1.2.1 or later
405 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
408 Severity : major on XT3
410 Description: libcfs overwrites /proc/sys/portals
411 Details : libcfs created a symlink from /proc/sys/portals to
412 /proc/sys/lnet for backwards compatibility. This is no
413 longer required and makes the Cray portals /proc variables
418 Description: OFED FMR API change
419 Details : This changes parameter usage to reflect a change in
420 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
421 that FMR support is only used in experimental versions of the
422 o2iblnd - this change does not affect standard usage at all.
424 Severity : enhancement
426 Description: new ko2iblnd module parameter: ib_mtu
427 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
428 HCAs. You can avoid this problem by setting the MTU to 1024
429 using this module parameter.
431 Severity : enhancement
432 Bugzilla : 11118/11620
433 Description: ptllnd small request message buffer alignment fix
434 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
435 Round up small message size on sends in case this option
436 is not supported. 11620 was a defect in the initial
437 implementation which effectively asserted all peers had to be
438 running the correct protocol version which was fixed by always
439 NAK-ing such requests and handling any misalignments they
444 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
445 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
446 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
448 Severity : enhancement
450 Description: Patchless ZC(zero copy) socklnd
451 Details : New protocol for socklnd, socklnd can support zero copy without
452 kernel patch, it's compatible with old socklnd. Checksum is
453 moved from tunables to modparams.
457 Description: When ksocknal_del_peer() is called upon a peer whose
458 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
459 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
462 Frequency : when ptlrpc is under heavy use and runs out of request buffer
464 Description: In lnet_match_blocked_msg(), md can be used without holding a
468 Frequency : very rarely
470 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
471 If connd connects a route which has been closed by
472 ksocknal_shutdown(), ksocknal_create_routes() may create new
473 routes which hold references on the peer, causing shutdown
474 process to wait for peer to disappear forever.
476 Severity : enhancement
478 Description: Dump XT3 portals traces on kptllnd timeout
479 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
480 dump Cray portals debug traces to a file. The kptllnd module
481 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
482 is the basename of the dump file.
485 Frequency : infrequent
487 Description: kernel ptllnd fix bug in connection re-establishment
488 Details : Kernel ptllnd could produce protocol errors e.g. illegal
489 matchbits and/or violate the credit flow protocol when trying
490 to re-establish a connection with a peer after an error or
493 Severity : enhancement
495 Description: Allow /proc/sys/lnet/debug to be set symbolically
496 Details : Allow debug and subsystem debug values to be read/set by name
497 in addition to numerically, for ease of use.
500 Frequency : only in configurations with LNET routers
502 Description: routes automatically marked down and recovered
503 Details : In configurations with LNET routers if a router fails routers
504 now actively try to recover routes that are down, unless they
505 are marked down by an administrator.
507 ------------------------------------------------------------------------------
509 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
512 Frequency : very rarely, in configurations with LNET routers and TCP
514 Description: incorrect data written to files on OSTs
515 Details : In certain high-load conditions incorrect data may be written
516 to files on the OST when using TCP networks.
518 ------------------------------------------------------------------------------
520 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
522 - rework CDEBUG messages rate-limiting mechanism b=10375
523 - add per-socket tunables for socklnd if the kernel is patched b=10327
525 ------------------------------------------------------------------------------
527 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
529 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
530 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
531 - quiet LNET startup LNI message for liblustre b=10128
532 - Better console error messages if 'ip2nets' can't match an IP address
533 - Fixed overflow/use-before-set bugs in linux-time.h
534 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
535 - LNET teardown failed an assertion about the route table being empty
536 - Fixed a crash in LNetEQPoll(<invalid handle>)
537 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
538 - improve debug message for liblustre/Catamount nodes (b=10116)
540 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
541 * Configuration change for the XT3
542 The PTLLND is now used to run Lustre over Portals on the XT3.
543 The configure option(s) --with-cray-portals are no longer
544 used. Rather --with-portals=<path-to-portals-includes> is
545 used to enable building on the XT3. In addition to enable
546 XT3 specific features the option --enable-cray-xt3 must be
549 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
550 * Portals has been removed, replaced by LNET.
551 LNET is new networking infrastructure for Lustre, it includes a
552 reorganized network configuration mode (see the user
553 documentation for full details) as well as support for routing
554 between different network fabrics. Lustre Networking Devices
555 (LNDS) for the supported network fabrics have also been created
556 for this new infrastructure.
558 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
563 Frequency : rare (large Voltaire clusters only)
565 Description: the default number of reserved transmit descriptors was too low
566 for some large clusters
567 Details : As a workaround, the number was increased. A proper fix includes
570 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
575 Frequency : occasional (large-scale events, cluster reboot, network failure)
577 Description: too many error messages on console obscure actual problem and
578 can slow down/panic server, or cause recovery to fail repeatedly
579 Details : enable rate-limiting of console error messages, and some messages
580 that were console errors now only go to the kernel log
582 Severity : enhancement
584 Description: add /proc/sys/portals/catastrophe entry which will report if
585 that node has previously LBUGged
587 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
589 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
591 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
593 - handle error return code in kranal_check_fma_rx() (5915,6054)
595 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
597 - update vibnal (Voltaire IB NAL)
598 - update gmnal (Myrinet NAL), gmnalid
600 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
602 * Landed portals:b_port_step as follows...
604 - removed CFS_DECL_SPIN*
605 just use 'spinlock_t' and initialise with spin_lock_init()
607 - removed CFS_DECL_MUTEX*
608 just use 'struct semaphore' and initialise with init_mutex()
610 - removed CFS_DECL_RWSEM*
611 just use 'struct rw_semaphore' and initialise with init_rwsem()
613 - renamed cfs_sleep_chan -> cfs_waitq
614 cfs_sleep_link -> cfs_waitlink
616 - fixed race in linux version of arch-independent socknal
617 (the ENOMEM/EAGAIN decision).
619 - Didn't fix problems in Darwin version of arch-independent socknal
620 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
622 - removed libcfs types from non-socknal header files (only some types
623 in the header files had been changed; the .c files hadn't been