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: o2iblnd/ptllnd credit deadlock in a routed config.
18 Details : o2iblnd/ptllnd credit deadlock in a routed config.
22 Description: High load after starting lnet
23 Details : gmlnd should sleep in rx thread in interruptible way. Otherwise,
24 uptime utility reports high load that looks confusingly.
28 Description: ksocklnd fails to establish connection if accept_port is high
29 Details : PID remapping must not be done for active (outgoing) connections
31 --------------------------------------------------------------------------------
33 2008-01-11 Sun Microsystems, Inc.
35 * Support for networks:
36 socklnd - any kernel supported by Lustre,
37 qswlnd - Qsnet kernel modules 5.20 and later,
38 openiblnd - IbGold 1.8.2,
39 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
40 viblnd - Voltaire ibhost 3.4.5 and later,
41 ciblnd - Topspin 3.2.0,
42 iiblnd - Infiniserv 3.3 + PathBits patch,
43 gmlnd - GM 2.1.22 and later,
44 mxlnd - MX 1.2.1 or later,
45 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
48 Description: liblustre network error
49 Details : liblustre clients should understand LNET_ACCEPT_PORT environment
50 variable even if they don't start lnet acceptor.
54 Description: Strange message from lnet (Ignoring prediction from the future)
55 Details : Incorrect calculation of peer's last_alive value in ksocklnd
57 --------------------------------------------------------------------------------
59 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
61 * Support for networks:
62 socklnd - any kernel supported by Lustre,
63 qswlnd - Qsnet kernel modules 5.20 and later,
64 openiblnd - IbGold 1.8.2,
65 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5.
66 viblnd - Voltaire ibhost 3.4.5 and later,
67 ciblnd - Topspin 3.2.0,
68 iiblnd - Infiniserv 3.3 + PathBits patch,
69 gmlnd - GM 2.1.22 and later,
70 mxlnd - MX 1.2.1 or later,
71 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
75 Description: ASSERTION(me == md->md_me) failed in lnet_match_md()
79 Description: increase send queue size for ciblnd/openiblnd
83 Description: new userspace socklnd
84 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
85 with new one - usocklnd.
87 Severity : enhancement
89 Description: Console message flood
90 Details : Make cdls ratelimiting more tunable by adding several tunable in
91 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
92 /proc/sys/lnet/console_backoff.
94 --------------------------------------------------------------------------------
96 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
98 * Support for networks:
99 socklnd - any kernel supported by Lustre,
100 qswlnd - Qsnet kernel modules 5.20 and later,
101 openiblnd - IbGold 1.8.2,
102 o2iblnd - OFED 1.1 and 1.2,
103 viblnd - Voltaire ibhost 3.4.5 and later,
104 ciblnd - Topspin 3.2.0,
105 iiblnd - Infiniserv 3.3 + PathBits patch,
106 gmlnd - GM 2.1.22 and later,
107 mxlnd - MX 1.2.1 or later,
108 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
112 Description: /proc/sys/lnet has non-sysctl entries
113 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
117 Description: TOE Kernel panic by ksocklnd
118 Details : offloaded sockets provide their own implementation of sendpage,
119 can't call tcp_sendpage() directly
123 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
124 Details : races between lnd_shutdown and peer creation prevent
125 lnd_shutdown from finishing.
129 Description: open files rlimit 1024 reached while liblustre testing
130 Details : ulnds/socklnd must close open socket after unsuccessful
135 Description: build error
136 Details : fix typos in gmlnd, ptllnd and viblnd
138 ------------------------------------------------------------------------------
140 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
142 * Support for networks:
143 socklnd - kernels up to 2.6.16,
144 qswlnd - Qsnet kernel modules 5.20 and later,
145 openiblnd - IbGold 1.8.2,
146 o2iblnd - OFED 1.1 and 1.2
147 viblnd - Voltaire ibhost 3.4.5 and later,
148 ciblnd - Topspin 3.2.0,
149 iiblnd - Infiniserv 3.3 + PathBits patch,
150 gmlnd - GM 2.1.22 and later,
151 mxlnd - MX 1.2.1 or later,
152 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
154 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
156 * Support for networks:
157 socklnd - kernels up to 2.6.16,
158 qswlnd - Qsnet kernel modules 5.20 and later,
159 openiblnd - IbGold 1.8.2,
161 viblnd - Voltaire ibhost 3.4.5 and later,
162 ciblnd - Topspin 3.2.0,
163 iiblnd - Infiniserv 3.3 + PathBits patch,
164 gmlnd - GM 2.1.22 and later,
165 mxlnd - MX 1.2.1 or later,
166 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
170 Description: Initialize cpumask before use
174 Description: ASSERTION failures when upgrading to the patchless zero-copy
176 Details : This bug affects "rolling upgrades", causing an inconsistent
177 protocol version negotiation and subsequent assertion failure
178 during rolling upgrades after the first wave of upgrades.
182 Details : Change "dropped message" CERRORs to D_NETERROR so they are
183 logged instead of creating "console chatter" when a lustre
184 timeout races with normal RPC completion.
187 Details : lnet_clear_peer_table can wait forever if user forgets to
191 Details : libcfs_id2str should check pid against LNET_PID_ANY.
195 Description: added LNET self test
196 Details : landing b_self_test
201 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
203 Details : do_div() macro is used incorrectly.
205 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
209 Description: make panic on lbug configurable
213 Description: Add OFED1.2 support to o2iblnd
214 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
215 are installed (other than kernel's in-tree infiniband), there
216 could be some problem while insmod o2iblnd (mismatch CRC of
218 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
219 this link provides solution:
220 https://bugs.openfabrics.org/show_bug.cgi?id=355
221 if extra Module.symvers is not supported in kernel, we will
222 have to run the script in bug 12316 to update
223 $LINUX/module.symvers before building o2iblnd.
224 More details about this are in bug 12316.
226 ------------------------------------------------------------------------------
228 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
229 * version 1.4.10 / 1.6.0
230 * Support for networks:
231 socklnd - kernels up to 2.6.16,
232 qswlnd - Qsnet kernel modules 5.20 and later,
233 openiblnd - IbGold 1.8.2,
235 viblnd - Voltaire ibhost 3.4.5 and later,
236 ciblnd - Topspin 3.2.0,
237 iiblnd - Infiniserv 3.3 + PathBits patch,
238 gmlnd - GM 2.1.22 and later,
239 mxlnd - MX 1.2.1 or later,
240 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
244 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
245 possibly accessed in kptllnd_shutdown. Ptllnd should init
246 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
250 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
254 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
258 Description: the_lnet.ln_finalizing was not set when the current thread is
259 about to complete messages. It only affects multi-threaded
265 Description: Changed the default kqswlnd ntxmsg=512
270 Description: Assertion failure in kernel ptllnd caused by posting passive
271 bulk buffers before connection establishment complete.
276 Description: A race in kernel ptllnd between deleting a peer and posting
277 new communications for it could hang communications -
278 manifesting as "Unexpectedly long timeout" messages.
283 Description: Kernel ptllnd lock ordering issue could hang a node.
288 Description: node crash on socket teardown race
291 Frequency : 'lctl peer_list' issued on a mx net
293 Description: Enable lctl's peer_list for MXLND
296 Frequency : after Ptllnd timeouts and portals congestion
298 Description: Credit overflows
299 Details : This was a bug in ptllnd connection establishment. The fix
300 implements better peer stamps to disambiguate connection
301 establishment and ensure both peers enter the credit flow
302 state machine consistently.
307 Description: kptllnd didn't propagate some network errors up to LNET
308 Details : This bug was spotted while investigating 11394. The fix
309 ensures network errors on sends and bulk transfers are
310 propagated to LNET/lustre correctly.
312 Severity : enhancement
314 Description: Fixed console chatter in case of -ETIMEDOUT.
316 Severity : enhancement
318 Description: Added D_NETTRACE for recording network packet history
319 (initially only for ptllnd). Also a separate userspace
320 ptllnd facility to gather history which should really be
321 covered by D_NETTRACE too, if only CDEBUG recorded history in
327 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
328 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
329 callback can occur before a connection has actually been
330 established. This caused an assertion failure previously.
332 Severity : enhancement
334 Description: Multiple instances for o2iblnd
335 Details : Allow multiple instances of o2iblnd to enable networking over
336 multiple HCAs and routing between them.
340 Description: lnet deadlock in router_checker
341 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
342 into BH locks to eliminate potential deadlock caused by
343 ksocknal_data_ready() preempting code holding these locks.
347 Description: Millions of failed socklnd connection attempts cause a very slow FS
348 Details : added a new route flag ksnr_scheduled to distinguish from
349 ksnr_connecting, so that a peer connection request is only turned
350 down for race concerns when an active connection to the same peer
351 is under progress (instead of just being scheduled).
353 ------------------------------------------------------------------------------
355 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
357 * Support for networks:
358 socklnd - kernels up to 2.6.16
359 qswlnd - Qsnet kernel modules 5.20 and later
360 openiblnd - IbGold 1.8.2
362 viblnd - Voltaire ibhost 3.4.5 and later
363 ciblnd - Topspin 3.2.0
364 iiblnd - Infiniserv 3.3 + PathBits patch
365 gmlnd - GM 2.1.22 and later
366 mxlnd - MX 1.2.1 or later
367 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
370 Severity : major on XT3
372 Description: libcfs overwrites /proc/sys/portals
373 Details : libcfs created a symlink from /proc/sys/portals to
374 /proc/sys/lnet for backwards compatibility. This is no
375 longer required and makes the Cray portals /proc variables
380 Description: OFED FMR API change
381 Details : This changes parameter usage to reflect a change in
382 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
383 that FMR support is only used in experimental versions of the
384 o2iblnd - this change does not affect standard usage at all.
386 Severity : enhancement
388 Description: new ko2iblnd module parameter: ib_mtu
389 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
390 HCAs. You can avoid this problem by setting the MTU to 1024
391 using this module parameter.
393 Severity : enhancement
394 Bugzilla : 11118/11620
395 Description: ptllnd small request message buffer alignment fix
396 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
397 Round up small message size on sends in case this option
398 is not supported. 11620 was a defect in the initial
399 implementation which effectively asserted all peers had to be
400 running the correct protocol version which was fixed by always
401 NAK-ing such requests and handling any misalignments they
406 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
407 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
408 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
410 Severity : enhancement
412 Description: Patchless ZC(zero copy) socklnd
413 Details : New protocol for socklnd, socklnd can support zero copy without
414 kernel patch, it's compatible with old socklnd. Checksum is
415 moved from tunables to modparams.
419 Description: When ksocknal_del_peer() is called upon a peer whose
420 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
421 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
424 Frequency : when ptlrpc is under heavy use and runs out of request buffer
426 Description: In lnet_match_blocked_msg(), md can be used without holding a
430 Frequency : very rarely
432 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
433 If connd connects a route which has been closed by
434 ksocknal_shutdown(), ksocknal_create_routes() may create new
435 routes which hold references on the peer, causing shutdown
436 process to wait for peer to disappear forever.
438 Severity : enhancement
440 Description: Dump XT3 portals traces on kptllnd timeout
441 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
442 dump Cray portals debug traces to a file. The kptllnd module
443 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
444 is the basename of the dump file.
447 Frequency : infrequent
449 Description: kernel ptllnd fix bug in connection re-establishment
450 Details : Kernel ptllnd could produce protocol errors e.g. illegal
451 matchbits and/or violate the credit flow protocol when trying
452 to re-establish a connection with a peer after an error or
455 Severity : enhancement
457 Description: Allow /proc/sys/lnet/debug to be set symbolically
458 Details : Allow debug and subsystem debug values to be read/set by name
459 in addition to numerically, for ease of use.
462 Frequency : only in configurations with LNET routers
464 Description: routes automatically marked down and recovered
465 Details : In configurations with LNET routers if a router fails routers
466 now actively try to recover routes that are down, unless they
467 are marked down by an administrator.
469 ------------------------------------------------------------------------------
471 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
474 Frequency : very rarely, in configurations with LNET routers and TCP
476 Description: incorrect data written to files on OSTs
477 Details : In certain high-load conditions incorrect data may be written
478 to files on the OST when using TCP networks.
480 ------------------------------------------------------------------------------
482 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
484 - rework CDEBUG messages rate-limiting mechanism b=10375
485 - add per-socket tunables for socklnd if the kernel is patched b=10327
487 ------------------------------------------------------------------------------
489 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
491 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
492 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
493 - quiet LNET startup LNI message for liblustre b=10128
494 - Better console error messages if 'ip2nets' can't match an IP address
495 - Fixed overflow/use-before-set bugs in linux-time.h
496 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
497 - LNET teardown failed an assertion about the route table being empty
498 - Fixed a crash in LNetEQPoll(<invalid handle>)
499 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
500 - improve debug message for liblustre/Catamount nodes (b=10116)
502 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
503 * Configuration change for the XT3
504 The PTLLND is now used to run Lustre over Portals on the XT3.
505 The configure option(s) --with-cray-portals are no longer
506 used. Rather --with-portals=<path-to-portals-includes> is
507 used to enable building on the XT3. In addition to enable
508 XT3 specific features the option --enable-cray-xt3 must be
511 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
512 * Portals has been removed, replaced by LNET.
513 LNET is new networking infrastructure for Lustre, it includes a
514 reorganized network configuration mode (see the user
515 documentation for full details) as well as support for routing
516 between different network fabrics. Lustre Networking Devices
517 (LNDS) for the supported network fabrics have also been created
518 for this new infrastructure.
520 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
525 Frequency : rare (large Voltaire clusters only)
527 Description: the default number of reserved transmit descriptors was too low
528 for some large clusters
529 Details : As a workaround, the number was increased. A proper fix includes
532 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
537 Frequency : occasional (large-scale events, cluster reboot, network failure)
539 Description: too many error messages on console obscure actual problem and
540 can slow down/panic server, or cause recovery to fail repeatedly
541 Details : enable rate-limiting of console error messages, and some messages
542 that were console errors now only go to the kernel log
544 Severity : enhancement
546 Description: add /proc/sys/portals/catastrophe entry which will report if
547 that node has previously LBUGged
549 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
551 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
553 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
555 - handle error return code in kranal_check_fma_rx() (5915,6054)
557 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
559 - update vibnal (Voltaire IB NAL)
560 - update gmnal (Myrinet NAL), gmnalid
562 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
564 * Landed portals:b_port_step as follows...
566 - removed CFS_DECL_SPIN*
567 just use 'spinlock_t' and initialise with spin_lock_init()
569 - removed CFS_DECL_MUTEX*
570 just use 'struct semaphore' and initialise with init_mutex()
572 - removed CFS_DECL_RWSEM*
573 just use 'struct rw_semaphore' and initialise with init_rwsem()
575 - renamed cfs_sleep_chan -> cfs_waitq
576 cfs_sleep_link -> cfs_waitlink
578 - fixed race in linux version of arch-independent socknal
579 (the ENOMEM/EAGAIN decision).
581 - Didn't fix problems in Darwin version of arch-independent socknal
582 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
584 - removed libcfs types from non-socknal header files (only some types
585 in the header files had been changed; the .c files hadn't been