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: excessive debug information removed
46 Details : excessive debug information removed
50 Description: ksocknal_create_conn() hit ASSERTION during connection race
51 Details : ksocknal_create_conn() hit ASSERTION during connection race
55 Description: ksocknal_send_hello() hit ASSERTION while connecting race
56 Details : ksocknal_send_hello() hit ASSERTION while connecting race
60 Description: o2iblnd/ptllnd credit deadlock in a routed config.
61 Details : o2iblnd/ptllnd credit deadlock in a routed config.
65 Description: High load after starting lnet
66 Details : gmlnd should sleep in rx thread in interruptible way. Otherwise,
67 uptime utility reports high load that looks confusingly.
71 Description: ksocklnd fails to establish connection if accept_port is high
72 Details : PID remapping must not be done for active (outgoing) connections
74 --------------------------------------------------------------------------------
76 2008-01-11 Sun Microsystems, Inc.
78 * Support for networks:
79 socklnd - any kernel supported by Lustre,
80 qswlnd - Qsnet kernel modules 5.20 and later,
81 openiblnd - IbGold 1.8.2,
82 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
83 viblnd - Voltaire ibhost 3.4.5 and later,
84 ciblnd - Topspin 3.2.0,
85 iiblnd - Infiniserv 3.3 + PathBits patch,
86 gmlnd - GM 2.1.22 and later,
87 mxlnd - MX 1.2.1 or later,
88 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
91 Description: liblustre network error
92 Details : liblustre clients should understand LNET_ACCEPT_PORT environment
93 variable even if they don't start lnet acceptor.
97 Description: Strange message from lnet (Ignoring prediction from the future)
98 Details : Incorrect calculation of peer's last_alive value in ksocklnd
100 --------------------------------------------------------------------------------
102 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
104 * Support for networks:
105 socklnd - any kernel supported by Lustre,
106 qswlnd - Qsnet kernel modules 5.20 and later,
107 openiblnd - IbGold 1.8.2,
108 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5.
109 viblnd - Voltaire ibhost 3.4.5 and later,
110 ciblnd - Topspin 3.2.0,
111 iiblnd - Infiniserv 3.3 + PathBits patch,
112 gmlnd - GM 2.1.22 and later,
113 mxlnd - MX 1.2.1 or later,
114 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
118 Description: ASSERTION(me == md->md_me) failed in lnet_match_md()
122 Description: increase send queue size for ciblnd/openiblnd
126 Description: new userspace socklnd
127 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
128 with new one - usocklnd.
130 Severity : enhancement
132 Description: Console message flood
133 Details : Make cdls ratelimiting more tunable by adding several tunable in
134 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
135 /proc/sys/lnet/console_backoff.
137 --------------------------------------------------------------------------------
139 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
141 * Support for networks:
142 socklnd - any kernel supported by Lustre,
143 qswlnd - Qsnet kernel modules 5.20 and later,
144 openiblnd - IbGold 1.8.2,
145 o2iblnd - OFED 1.1 and 1.2,
146 viblnd - Voltaire ibhost 3.4.5 and later,
147 ciblnd - Topspin 3.2.0,
148 iiblnd - Infiniserv 3.3 + PathBits patch,
149 gmlnd - GM 2.1.22 and later,
150 mxlnd - MX 1.2.1 or later,
151 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
155 Description: /proc/sys/lnet has non-sysctl entries
156 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
160 Description: TOE Kernel panic by ksocklnd
161 Details : offloaded sockets provide their own implementation of sendpage,
162 can't call tcp_sendpage() directly
166 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
167 Details : races between lnd_shutdown and peer creation prevent
168 lnd_shutdown from finishing.
172 Description: open files rlimit 1024 reached while liblustre testing
173 Details : ulnds/socklnd must close open socket after unsuccessful
178 Description: build error
179 Details : fix typos in gmlnd, ptllnd and viblnd
181 ------------------------------------------------------------------------------
183 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
185 * Support for networks:
186 socklnd - kernels up to 2.6.16,
187 qswlnd - Qsnet kernel modules 5.20 and later,
188 openiblnd - IbGold 1.8.2,
189 o2iblnd - OFED 1.1 and 1.2
190 viblnd - Voltaire ibhost 3.4.5 and later,
191 ciblnd - Topspin 3.2.0,
192 iiblnd - Infiniserv 3.3 + PathBits patch,
193 gmlnd - GM 2.1.22 and later,
194 mxlnd - MX 1.2.1 or later,
195 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
197 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
199 * Support for networks:
200 socklnd - kernels up to 2.6.16,
201 qswlnd - Qsnet kernel modules 5.20 and later,
202 openiblnd - IbGold 1.8.2,
204 viblnd - Voltaire ibhost 3.4.5 and later,
205 ciblnd - Topspin 3.2.0,
206 iiblnd - Infiniserv 3.3 + PathBits patch,
207 gmlnd - GM 2.1.22 and later,
208 mxlnd - MX 1.2.1 or later,
209 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
213 Description: Initialize cpumask before use
217 Description: ASSERTION failures when upgrading to the patchless zero-copy
219 Details : This bug affects "rolling upgrades", causing an inconsistent
220 protocol version negotiation and subsequent assertion failure
221 during rolling upgrades after the first wave of upgrades.
225 Details : Change "dropped message" CERRORs to D_NETERROR so they are
226 logged instead of creating "console chatter" when a lustre
227 timeout races with normal RPC completion.
230 Details : lnet_clear_peer_table can wait forever if user forgets to
234 Details : libcfs_id2str should check pid against LNET_PID_ANY.
238 Description: added LNET self test
239 Details : landing b_self_test
244 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
246 Details : do_div() macro is used incorrectly.
248 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
252 Description: make panic on lbug configurable
256 Description: Add OFED1.2 support to o2iblnd
257 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
258 are installed (other than kernel's in-tree infiniband), there
259 could be some problem while insmod o2iblnd (mismatch CRC of
261 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
262 this link provides solution:
263 https://bugs.openfabrics.org/show_bug.cgi?id=355
264 if extra Module.symvers is not supported in kernel, we will
265 have to run the script in bug 12316 to update
266 $LINUX/module.symvers before building o2iblnd.
267 More details about this are in bug 12316.
269 ------------------------------------------------------------------------------
271 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
272 * version 1.4.10 / 1.6.0
273 * Support for networks:
274 socklnd - kernels up to 2.6.16,
275 qswlnd - Qsnet kernel modules 5.20 and later,
276 openiblnd - IbGold 1.8.2,
278 viblnd - Voltaire ibhost 3.4.5 and later,
279 ciblnd - Topspin 3.2.0,
280 iiblnd - Infiniserv 3.3 + PathBits patch,
281 gmlnd - GM 2.1.22 and later,
282 mxlnd - MX 1.2.1 or later,
283 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
287 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
288 possibly accessed in kptllnd_shutdown. Ptllnd should init
289 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
293 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
297 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
301 Description: the_lnet.ln_finalizing was not set when the current thread is
302 about to complete messages. It only affects multi-threaded
308 Description: Changed the default kqswlnd ntxmsg=512
313 Description: Assertion failure in kernel ptllnd caused by posting passive
314 bulk buffers before connection establishment complete.
319 Description: A race in kernel ptllnd between deleting a peer and posting
320 new communications for it could hang communications -
321 manifesting as "Unexpectedly long timeout" messages.
326 Description: Kernel ptllnd lock ordering issue could hang a node.
331 Description: node crash on socket teardown race
334 Frequency : 'lctl peer_list' issued on a mx net
336 Description: Enable lctl's peer_list for MXLND
339 Frequency : after Ptllnd timeouts and portals congestion
341 Description: Credit overflows
342 Details : This was a bug in ptllnd connection establishment. The fix
343 implements better peer stamps to disambiguate connection
344 establishment and ensure both peers enter the credit flow
345 state machine consistently.
350 Description: kptllnd didn't propagate some network errors up to LNET
351 Details : This bug was spotted while investigating 11394. The fix
352 ensures network errors on sends and bulk transfers are
353 propagated to LNET/lustre correctly.
355 Severity : enhancement
357 Description: Fixed console chatter in case of -ETIMEDOUT.
359 Severity : enhancement
361 Description: Added D_NETTRACE for recording network packet history
362 (initially only for ptllnd). Also a separate userspace
363 ptllnd facility to gather history which should really be
364 covered by D_NETTRACE too, if only CDEBUG recorded history in
370 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
371 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
372 callback can occur before a connection has actually been
373 established. This caused an assertion failure previously.
375 Severity : enhancement
377 Description: Multiple instances for o2iblnd
378 Details : Allow multiple instances of o2iblnd to enable networking over
379 multiple HCAs and routing between them.
383 Description: lnet deadlock in router_checker
384 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
385 into BH locks to eliminate potential deadlock caused by
386 ksocknal_data_ready() preempting code holding these locks.
390 Description: Millions of failed socklnd connection attempts cause a very slow FS
391 Details : added a new route flag ksnr_scheduled to distinguish from
392 ksnr_connecting, so that a peer connection request is only turned
393 down for race concerns when an active connection to the same peer
394 is under progress (instead of just being scheduled).
396 ------------------------------------------------------------------------------
398 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
400 * Support for networks:
401 socklnd - kernels up to 2.6.16
402 qswlnd - Qsnet kernel modules 5.20 and later
403 openiblnd - IbGold 1.8.2
405 viblnd - Voltaire ibhost 3.4.5 and later
406 ciblnd - Topspin 3.2.0
407 iiblnd - Infiniserv 3.3 + PathBits patch
408 gmlnd - GM 2.1.22 and later
409 mxlnd - MX 1.2.1 or later
410 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
413 Severity : major on XT3
415 Description: libcfs overwrites /proc/sys/portals
416 Details : libcfs created a symlink from /proc/sys/portals to
417 /proc/sys/lnet for backwards compatibility. This is no
418 longer required and makes the Cray portals /proc variables
423 Description: OFED FMR API change
424 Details : This changes parameter usage to reflect a change in
425 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
426 that FMR support is only used in experimental versions of the
427 o2iblnd - this change does not affect standard usage at all.
429 Severity : enhancement
431 Description: new ko2iblnd module parameter: ib_mtu
432 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
433 HCAs. You can avoid this problem by setting the MTU to 1024
434 using this module parameter.
436 Severity : enhancement
437 Bugzilla : 11118/11620
438 Description: ptllnd small request message buffer alignment fix
439 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
440 Round up small message size on sends in case this option
441 is not supported. 11620 was a defect in the initial
442 implementation which effectively asserted all peers had to be
443 running the correct protocol version which was fixed by always
444 NAK-ing such requests and handling any misalignments they
449 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
450 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
451 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
453 Severity : enhancement
455 Description: Patchless ZC(zero copy) socklnd
456 Details : New protocol for socklnd, socklnd can support zero copy without
457 kernel patch, it's compatible with old socklnd. Checksum is
458 moved from tunables to modparams.
462 Description: When ksocknal_del_peer() is called upon a peer whose
463 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
464 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
467 Frequency : when ptlrpc is under heavy use and runs out of request buffer
469 Description: In lnet_match_blocked_msg(), md can be used without holding a
473 Frequency : very rarely
475 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
476 If connd connects a route which has been closed by
477 ksocknal_shutdown(), ksocknal_create_routes() may create new
478 routes which hold references on the peer, causing shutdown
479 process to wait for peer to disappear forever.
481 Severity : enhancement
483 Description: Dump XT3 portals traces on kptllnd timeout
484 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
485 dump Cray portals debug traces to a file. The kptllnd module
486 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
487 is the basename of the dump file.
490 Frequency : infrequent
492 Description: kernel ptllnd fix bug in connection re-establishment
493 Details : Kernel ptllnd could produce protocol errors e.g. illegal
494 matchbits and/or violate the credit flow protocol when trying
495 to re-establish a connection with a peer after an error or
498 Severity : enhancement
500 Description: Allow /proc/sys/lnet/debug to be set symbolically
501 Details : Allow debug and subsystem debug values to be read/set by name
502 in addition to numerically, for ease of use.
505 Frequency : only in configurations with LNET routers
507 Description: routes automatically marked down and recovered
508 Details : In configurations with LNET routers if a router fails routers
509 now actively try to recover routes that are down, unless they
510 are marked down by an administrator.
512 ------------------------------------------------------------------------------
514 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
517 Frequency : very rarely, in configurations with LNET routers and TCP
519 Description: incorrect data written to files on OSTs
520 Details : In certain high-load conditions incorrect data may be written
521 to files on the OST when using TCP networks.
523 ------------------------------------------------------------------------------
525 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
527 - rework CDEBUG messages rate-limiting mechanism b=10375
528 - add per-socket tunables for socklnd if the kernel is patched b=10327
530 ------------------------------------------------------------------------------
532 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
534 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
535 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
536 - quiet LNET startup LNI message for liblustre b=10128
537 - Better console error messages if 'ip2nets' can't match an IP address
538 - Fixed overflow/use-before-set bugs in linux-time.h
539 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
540 - LNET teardown failed an assertion about the route table being empty
541 - Fixed a crash in LNetEQPoll(<invalid handle>)
542 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
543 - improve debug message for liblustre/Catamount nodes (b=10116)
545 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
546 * Configuration change for the XT3
547 The PTLLND is now used to run Lustre over Portals on the XT3.
548 The configure option(s) --with-cray-portals are no longer
549 used. Rather --with-portals=<path-to-portals-includes> is
550 used to enable building on the XT3. In addition to enable
551 XT3 specific features the option --enable-cray-xt3 must be
554 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
555 * Portals has been removed, replaced by LNET.
556 LNET is new networking infrastructure for Lustre, it includes a
557 reorganized network configuration mode (see the user
558 documentation for full details) as well as support for routing
559 between different network fabrics. Lustre Networking Devices
560 (LNDS) for the supported network fabrics have also been created
561 for this new infrastructure.
563 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
568 Frequency : rare (large Voltaire clusters only)
570 Description: the default number of reserved transmit descriptors was too low
571 for some large clusters
572 Details : As a workaround, the number was increased. A proper fix includes
575 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
580 Frequency : occasional (large-scale events, cluster reboot, network failure)
582 Description: too many error messages on console obscure actual problem and
583 can slow down/panic server, or cause recovery to fail repeatedly
584 Details : enable rate-limiting of console error messages, and some messages
585 that were console errors now only go to the kernel log
587 Severity : enhancement
589 Description: add /proc/sys/portals/catastrophe entry which will report if
590 that node has previously LBUGged
592 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
594 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
596 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
598 - handle error return code in kranal_check_fma_rx() (5915,6054)
600 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
602 - update vibnal (Voltaire IB NAL)
603 - update gmnal (Myrinet NAL), gmnalid
605 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
607 * Landed portals:b_port_step as follows...
609 - removed CFS_DECL_SPIN*
610 just use 'spinlock_t' and initialise with spin_lock_init()
612 - removed CFS_DECL_MUTEX*
613 just use 'struct semaphore' and initialise with init_mutex()
615 - removed CFS_DECL_RWSEM*
616 just use 'struct rw_semaphore' and initialise with init_rwsem()
618 - renamed cfs_sleep_chan -> cfs_waitq
619 cfs_sleep_link -> cfs_waitlink
621 - fixed race in linux version of arch-independent socknal
622 (the ENOMEM/EAGAIN decision).
624 - Didn't fix problems in Darwin version of arch-independent socknal
625 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
627 - removed libcfs types from non-socknal header files (only some types
628 in the header files had been changed; the .c files hadn't been