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
17 Description: ksocknal_send_hello() hit ASSERTION while connecting race
18 Details : ksocknal_send_hello() hit ASSERTION while connecting race
22 Description: o2iblnd/ptllnd credit deadlock in a routed config.
23 Details : o2iblnd/ptllnd credit deadlock in a routed config.
27 Description: High load after starting lnet
28 Details : gmlnd should sleep in rx thread in interruptible way. Otherwise,
29 uptime utility reports high load that looks confusingly.
33 Description: ksocklnd fails to establish connection if accept_port is high
34 Details : PID remapping must not be done for active (outgoing) connections
36 --------------------------------------------------------------------------------
38 2008-01-11 Sun Microsystems, Inc.
40 * Support for networks:
41 socklnd - any kernel supported by Lustre,
42 qswlnd - Qsnet kernel modules 5.20 and later,
43 openiblnd - IbGold 1.8.2,
44 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
45 viblnd - Voltaire ibhost 3.4.5 and later,
46 ciblnd - Topspin 3.2.0,
47 iiblnd - Infiniserv 3.3 + PathBits patch,
48 gmlnd - GM 2.1.22 and later,
49 mxlnd - MX 1.2.1 or later,
50 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
53 Description: liblustre network error
54 Details : liblustre clients should understand LNET_ACCEPT_PORT environment
55 variable even if they don't start lnet acceptor.
59 Description: Strange message from lnet (Ignoring prediction from the future)
60 Details : Incorrect calculation of peer's last_alive value in ksocklnd
62 --------------------------------------------------------------------------------
64 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
66 * Support for networks:
67 socklnd - any kernel supported by Lustre,
68 qswlnd - Qsnet kernel modules 5.20 and later,
69 openiblnd - IbGold 1.8.2,
70 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5.
71 viblnd - Voltaire ibhost 3.4.5 and later,
72 ciblnd - Topspin 3.2.0,
73 iiblnd - Infiniserv 3.3 + PathBits patch,
74 gmlnd - GM 2.1.22 and later,
75 mxlnd - MX 1.2.1 or later,
76 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
80 Description: ASSERTION(me == md->md_me) failed in lnet_match_md()
84 Description: increase send queue size for ciblnd/openiblnd
88 Description: new userspace socklnd
89 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
90 with new one - usocklnd.
92 Severity : enhancement
94 Description: Console message flood
95 Details : Make cdls ratelimiting more tunable by adding several tunable in
96 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
97 /proc/sys/lnet/console_backoff.
99 --------------------------------------------------------------------------------
101 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
103 * Support for networks:
104 socklnd - any kernel supported by Lustre,
105 qswlnd - Qsnet kernel modules 5.20 and later,
106 openiblnd - IbGold 1.8.2,
107 o2iblnd - OFED 1.1 and 1.2,
108 viblnd - Voltaire ibhost 3.4.5 and later,
109 ciblnd - Topspin 3.2.0,
110 iiblnd - Infiniserv 3.3 + PathBits patch,
111 gmlnd - GM 2.1.22 and later,
112 mxlnd - MX 1.2.1 or later,
113 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
117 Description: /proc/sys/lnet has non-sysctl entries
118 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
122 Description: TOE Kernel panic by ksocklnd
123 Details : offloaded sockets provide their own implementation of sendpage,
124 can't call tcp_sendpage() directly
128 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
129 Details : races between lnd_shutdown and peer creation prevent
130 lnd_shutdown from finishing.
134 Description: open files rlimit 1024 reached while liblustre testing
135 Details : ulnds/socklnd must close open socket after unsuccessful
140 Description: build error
141 Details : fix typos in gmlnd, ptllnd and viblnd
143 ------------------------------------------------------------------------------
145 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
147 * Support for networks:
148 socklnd - kernels up to 2.6.16,
149 qswlnd - Qsnet kernel modules 5.20 and later,
150 openiblnd - IbGold 1.8.2,
151 o2iblnd - OFED 1.1 and 1.2
152 viblnd - Voltaire ibhost 3.4.5 and later,
153 ciblnd - Topspin 3.2.0,
154 iiblnd - Infiniserv 3.3 + PathBits patch,
155 gmlnd - GM 2.1.22 and later,
156 mxlnd - MX 1.2.1 or later,
157 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
159 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
161 * Support for networks:
162 socklnd - kernels up to 2.6.16,
163 qswlnd - Qsnet kernel modules 5.20 and later,
164 openiblnd - IbGold 1.8.2,
166 viblnd - Voltaire ibhost 3.4.5 and later,
167 ciblnd - Topspin 3.2.0,
168 iiblnd - Infiniserv 3.3 + PathBits patch,
169 gmlnd - GM 2.1.22 and later,
170 mxlnd - MX 1.2.1 or later,
171 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
175 Description: Initialize cpumask before use
179 Description: ASSERTION failures when upgrading to the patchless zero-copy
181 Details : This bug affects "rolling upgrades", causing an inconsistent
182 protocol version negotiation and subsequent assertion failure
183 during rolling upgrades after the first wave of upgrades.
187 Details : Change "dropped message" CERRORs to D_NETERROR so they are
188 logged instead of creating "console chatter" when a lustre
189 timeout races with normal RPC completion.
192 Details : lnet_clear_peer_table can wait forever if user forgets to
196 Details : libcfs_id2str should check pid against LNET_PID_ANY.
200 Description: added LNET self test
201 Details : landing b_self_test
206 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
208 Details : do_div() macro is used incorrectly.
210 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
214 Description: make panic on lbug configurable
218 Description: Add OFED1.2 support to o2iblnd
219 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
220 are installed (other than kernel's in-tree infiniband), there
221 could be some problem while insmod o2iblnd (mismatch CRC of
223 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
224 this link provides solution:
225 https://bugs.openfabrics.org/show_bug.cgi?id=355
226 if extra Module.symvers is not supported in kernel, we will
227 have to run the script in bug 12316 to update
228 $LINUX/module.symvers before building o2iblnd.
229 More details about this are in bug 12316.
231 ------------------------------------------------------------------------------
233 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
234 * version 1.4.10 / 1.6.0
235 * Support for networks:
236 socklnd - kernels up to 2.6.16,
237 qswlnd - Qsnet kernel modules 5.20 and later,
238 openiblnd - IbGold 1.8.2,
240 viblnd - Voltaire ibhost 3.4.5 and later,
241 ciblnd - Topspin 3.2.0,
242 iiblnd - Infiniserv 3.3 + PathBits patch,
243 gmlnd - GM 2.1.22 and later,
244 mxlnd - MX 1.2.1 or later,
245 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
249 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
250 possibly accessed in kptllnd_shutdown. Ptllnd should init
251 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
255 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
259 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
263 Description: the_lnet.ln_finalizing was not set when the current thread is
264 about to complete messages. It only affects multi-threaded
270 Description: Changed the default kqswlnd ntxmsg=512
275 Description: Assertion failure in kernel ptllnd caused by posting passive
276 bulk buffers before connection establishment complete.
281 Description: A race in kernel ptllnd between deleting a peer and posting
282 new communications for it could hang communications -
283 manifesting as "Unexpectedly long timeout" messages.
288 Description: Kernel ptllnd lock ordering issue could hang a node.
293 Description: node crash on socket teardown race
296 Frequency : 'lctl peer_list' issued on a mx net
298 Description: Enable lctl's peer_list for MXLND
301 Frequency : after Ptllnd timeouts and portals congestion
303 Description: Credit overflows
304 Details : This was a bug in ptllnd connection establishment. The fix
305 implements better peer stamps to disambiguate connection
306 establishment and ensure both peers enter the credit flow
307 state machine consistently.
312 Description: kptllnd didn't propagate some network errors up to LNET
313 Details : This bug was spotted while investigating 11394. The fix
314 ensures network errors on sends and bulk transfers are
315 propagated to LNET/lustre correctly.
317 Severity : enhancement
319 Description: Fixed console chatter in case of -ETIMEDOUT.
321 Severity : enhancement
323 Description: Added D_NETTRACE for recording network packet history
324 (initially only for ptllnd). Also a separate userspace
325 ptllnd facility to gather history which should really be
326 covered by D_NETTRACE too, if only CDEBUG recorded history in
332 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
333 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
334 callback can occur before a connection has actually been
335 established. This caused an assertion failure previously.
337 Severity : enhancement
339 Description: Multiple instances for o2iblnd
340 Details : Allow multiple instances of o2iblnd to enable networking over
341 multiple HCAs and routing between them.
345 Description: lnet deadlock in router_checker
346 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
347 into BH locks to eliminate potential deadlock caused by
348 ksocknal_data_ready() preempting code holding these locks.
352 Description: Millions of failed socklnd connection attempts cause a very slow FS
353 Details : added a new route flag ksnr_scheduled to distinguish from
354 ksnr_connecting, so that a peer connection request is only turned
355 down for race concerns when an active connection to the same peer
356 is under progress (instead of just being scheduled).
358 ------------------------------------------------------------------------------
360 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
362 * Support for networks:
363 socklnd - kernels up to 2.6.16
364 qswlnd - Qsnet kernel modules 5.20 and later
365 openiblnd - IbGold 1.8.2
367 viblnd - Voltaire ibhost 3.4.5 and later
368 ciblnd - Topspin 3.2.0
369 iiblnd - Infiniserv 3.3 + PathBits patch
370 gmlnd - GM 2.1.22 and later
371 mxlnd - MX 1.2.1 or later
372 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
375 Severity : major on XT3
377 Description: libcfs overwrites /proc/sys/portals
378 Details : libcfs created a symlink from /proc/sys/portals to
379 /proc/sys/lnet for backwards compatibility. This is no
380 longer required and makes the Cray portals /proc variables
385 Description: OFED FMR API change
386 Details : This changes parameter usage to reflect a change in
387 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
388 that FMR support is only used in experimental versions of the
389 o2iblnd - this change does not affect standard usage at all.
391 Severity : enhancement
393 Description: new ko2iblnd module parameter: ib_mtu
394 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
395 HCAs. You can avoid this problem by setting the MTU to 1024
396 using this module parameter.
398 Severity : enhancement
399 Bugzilla : 11118/11620
400 Description: ptllnd small request message buffer alignment fix
401 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
402 Round up small message size on sends in case this option
403 is not supported. 11620 was a defect in the initial
404 implementation which effectively asserted all peers had to be
405 running the correct protocol version which was fixed by always
406 NAK-ing such requests and handling any misalignments they
411 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
412 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
413 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
415 Severity : enhancement
417 Description: Patchless ZC(zero copy) socklnd
418 Details : New protocol for socklnd, socklnd can support zero copy without
419 kernel patch, it's compatible with old socklnd. Checksum is
420 moved from tunables to modparams.
424 Description: When ksocknal_del_peer() is called upon a peer whose
425 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
426 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
429 Frequency : when ptlrpc is under heavy use and runs out of request buffer
431 Description: In lnet_match_blocked_msg(), md can be used without holding a
435 Frequency : very rarely
437 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
438 If connd connects a route which has been closed by
439 ksocknal_shutdown(), ksocknal_create_routes() may create new
440 routes which hold references on the peer, causing shutdown
441 process to wait for peer to disappear forever.
443 Severity : enhancement
445 Description: Dump XT3 portals traces on kptllnd timeout
446 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
447 dump Cray portals debug traces to a file. The kptllnd module
448 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
449 is the basename of the dump file.
452 Frequency : infrequent
454 Description: kernel ptllnd fix bug in connection re-establishment
455 Details : Kernel ptllnd could produce protocol errors e.g. illegal
456 matchbits and/or violate the credit flow protocol when trying
457 to re-establish a connection with a peer after an error or
460 Severity : enhancement
462 Description: Allow /proc/sys/lnet/debug to be set symbolically
463 Details : Allow debug and subsystem debug values to be read/set by name
464 in addition to numerically, for ease of use.
467 Frequency : only in configurations with LNET routers
469 Description: routes automatically marked down and recovered
470 Details : In configurations with LNET routers if a router fails routers
471 now actively try to recover routes that are down, unless they
472 are marked down by an administrator.
474 ------------------------------------------------------------------------------
476 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
479 Frequency : very rarely, in configurations with LNET routers and TCP
481 Description: incorrect data written to files on OSTs
482 Details : In certain high-load conditions incorrect data may be written
483 to files on the OST when using TCP networks.
485 ------------------------------------------------------------------------------
487 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
489 - rework CDEBUG messages rate-limiting mechanism b=10375
490 - add per-socket tunables for socklnd if the kernel is patched b=10327
492 ------------------------------------------------------------------------------
494 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
496 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
497 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
498 - quiet LNET startup LNI message for liblustre b=10128
499 - Better console error messages if 'ip2nets' can't match an IP address
500 - Fixed overflow/use-before-set bugs in linux-time.h
501 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
502 - LNET teardown failed an assertion about the route table being empty
503 - Fixed a crash in LNetEQPoll(<invalid handle>)
504 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
505 - improve debug message for liblustre/Catamount nodes (b=10116)
507 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
508 * Configuration change for the XT3
509 The PTLLND is now used to run Lustre over Portals on the XT3.
510 The configure option(s) --with-cray-portals are no longer
511 used. Rather --with-portals=<path-to-portals-includes> is
512 used to enable building on the XT3. In addition to enable
513 XT3 specific features the option --enable-cray-xt3 must be
516 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
517 * Portals has been removed, replaced by LNET.
518 LNET is new networking infrastructure for Lustre, it includes a
519 reorganized network configuration mode (see the user
520 documentation for full details) as well as support for routing
521 between different network fabrics. Lustre Networking Devices
522 (LNDS) for the supported network fabrics have also been created
523 for this new infrastructure.
525 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
530 Frequency : rare (large Voltaire clusters only)
532 Description: the default number of reserved transmit descriptors was too low
533 for some large clusters
534 Details : As a workaround, the number was increased. A proper fix includes
537 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
542 Frequency : occasional (large-scale events, cluster reboot, network failure)
544 Description: too many error messages on console obscure actual problem and
545 can slow down/panic server, or cause recovery to fail repeatedly
546 Details : enable rate-limiting of console error messages, and some messages
547 that were console errors now only go to the kernel log
549 Severity : enhancement
551 Description: add /proc/sys/portals/catastrophe entry which will report if
552 that node has previously LBUGged
554 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
556 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
558 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
560 - handle error return code in kranal_check_fma_rx() (5915,6054)
562 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
564 - update vibnal (Voltaire IB NAL)
565 - update gmnal (Myrinet NAL), gmnalid
567 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
569 * Landed portals:b_port_step as follows...
571 - removed CFS_DECL_SPIN*
572 just use 'spinlock_t' and initialise with spin_lock_init()
574 - removed CFS_DECL_MUTEX*
575 just use 'struct semaphore' and initialise with init_mutex()
577 - removed CFS_DECL_RWSEM*
578 just use 'struct rw_semaphore' and initialise with init_rwsem()
580 - renamed cfs_sleep_chan -> cfs_waitq
581 cfs_sleep_link -> cfs_waitlink
583 - fixed race in linux version of arch-independent socknal
584 (the ENOMEM/EAGAIN decision).
586 - Didn't fix problems in Darwin version of arch-independent socknal
587 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
589 - removed libcfs types from non-socknal header files (only some types
590 in the header files had been changed; the .c files hadn't been