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
15 --------------------------------------------------------------------------------
17 2008-01-11 Sun Microsystems, Inc.
19 * Support for networks:
20 socklnd - any kernel supported by Lustre,
21 qswlnd - Qsnet kernel modules 5.20 and later,
22 openiblnd - IbGold 1.8.2,
23 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5
24 viblnd - Voltaire ibhost 3.4.5 and later,
25 ciblnd - Topspin 3.2.0,
26 iiblnd - Infiniserv 3.3 + PathBits patch,
27 gmlnd - GM 2.1.22 and later,
28 mxlnd - MX 1.2.1 or later,
29 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
32 Description: liblustre network error
33 Details : liblustre clients should understand LNET_ACCEPT_PORT environment
34 variable even if they don't start lnet acceptor.
38 Description: Strange message from lnet (Ignoring prediction from the future)
39 Details : Incorrect calculation of peer's last_alive value in ksocklnd
41 --------------------------------------------------------------------------------
43 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
45 * Support for networks:
46 socklnd - any kernel supported by Lustre,
47 qswlnd - Qsnet kernel modules 5.20 and later,
48 openiblnd - IbGold 1.8.2,
49 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5.
50 viblnd - Voltaire ibhost 3.4.5 and later,
51 ciblnd - Topspin 3.2.0,
52 iiblnd - Infiniserv 3.3 + PathBits patch,
53 gmlnd - GM 2.1.22 and later,
54 mxlnd - MX 1.2.1 or later,
55 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
59 Description: ASSERTION(me == md->md_me) failed in lnet_match_md()
63 Description: increase send queue size for ciblnd/openiblnd
67 Description: new userspace socklnd
68 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
69 with new one - usocklnd.
71 Severity : enhancement
73 Description: Console message flood
74 Details : Make cdls ratelimiting more tunable by adding several tunable in
75 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
76 /proc/sys/lnet/console_backoff.
78 --------------------------------------------------------------------------------
80 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
82 * Support for networks:
83 socklnd - any kernel supported by Lustre,
84 qswlnd - Qsnet kernel modules 5.20 and later,
85 openiblnd - IbGold 1.8.2,
86 o2iblnd - OFED 1.1 and 1.2,
87 viblnd - Voltaire ibhost 3.4.5 and later,
88 ciblnd - Topspin 3.2.0,
89 iiblnd - Infiniserv 3.3 + PathBits patch,
90 gmlnd - GM 2.1.22 and later,
91 mxlnd - MX 1.2.1 or later,
92 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
96 Description: /proc/sys/lnet has non-sysctl entries
97 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
101 Description: TOE Kernel panic by ksocklnd
102 Details : offloaded sockets provide their own implementation of sendpage,
103 can't call tcp_sendpage() directly
107 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
108 Details : races between lnd_shutdown and peer creation prevent
109 lnd_shutdown from finishing.
113 Description: open files rlimit 1024 reached while liblustre testing
114 Details : ulnds/socklnd must close open socket after unsuccessful
119 Description: build error
120 Details : fix typos in gmlnd, ptllnd and viblnd
122 ------------------------------------------------------------------------------
124 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
126 * Support for networks:
127 socklnd - kernels up to 2.6.16,
128 qswlnd - Qsnet kernel modules 5.20 and later,
129 openiblnd - IbGold 1.8.2,
130 o2iblnd - OFED 1.1 and 1.2
131 viblnd - Voltaire ibhost 3.4.5 and later,
132 ciblnd - Topspin 3.2.0,
133 iiblnd - Infiniserv 3.3 + PathBits patch,
134 gmlnd - GM 2.1.22 and later,
135 mxlnd - MX 1.2.1 or later,
136 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
138 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
140 * Support for networks:
141 socklnd - kernels up to 2.6.16,
142 qswlnd - Qsnet kernel modules 5.20 and later,
143 openiblnd - IbGold 1.8.2,
145 viblnd - Voltaire ibhost 3.4.5 and later,
146 ciblnd - Topspin 3.2.0,
147 iiblnd - Infiniserv 3.3 + PathBits patch,
148 gmlnd - GM 2.1.22 and later,
149 mxlnd - MX 1.2.1 or later,
150 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
154 Description: Initialize cpumask before use
158 Description: ASSERTION failures when upgrading to the patchless zero-copy
160 Details : This bug affects "rolling upgrades", causing an inconsistent
161 protocol version negotiation and subsequent assertion failure
162 during rolling upgrades after the first wave of upgrades.
166 Details : Change "dropped message" CERRORs to D_NETERROR so they are
167 logged instead of creating "console chatter" when a lustre
168 timeout races with normal RPC completion.
171 Details : lnet_clear_peer_table can wait forever if user forgets to
175 Details : libcfs_id2str should check pid against LNET_PID_ANY.
179 Description: added LNET self test
180 Details : landing b_self_test
185 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
187 Details : do_div() macro is used incorrectly.
189 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
193 Description: make panic on lbug configurable
197 Description: Add OFED1.2 support to o2iblnd
198 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
199 are installed (other than kernel's in-tree infiniband), there
200 could be some problem while insmod o2iblnd (mismatch CRC of
202 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
203 this link provides solution:
204 https://bugs.openfabrics.org/show_bug.cgi?id=355
205 if extra Module.symvers is not supported in kernel, we will
206 have to run the script in bug 12316 to update
207 $LINUX/module.symvers before building o2iblnd.
208 More details about this are in bug 12316.
210 ------------------------------------------------------------------------------
212 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
213 * version 1.4.10 / 1.6.0
214 * Support for networks:
215 socklnd - kernels up to 2.6.16,
216 qswlnd - Qsnet kernel modules 5.20 and later,
217 openiblnd - IbGold 1.8.2,
219 viblnd - Voltaire ibhost 3.4.5 and later,
220 ciblnd - Topspin 3.2.0,
221 iiblnd - Infiniserv 3.3 + PathBits patch,
222 gmlnd - GM 2.1.22 and later,
223 mxlnd - MX 1.2.1 or later,
224 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
228 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
229 possibly accessed in kptllnd_shutdown. Ptllnd should init
230 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
234 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
238 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
242 Description: the_lnet.ln_finalizing was not set when the current thread is
243 about to complete messages. It only affects multi-threaded
249 Description: Changed the default kqswlnd ntxmsg=512
254 Description: Assertion failure in kernel ptllnd caused by posting passive
255 bulk buffers before connection establishment complete.
260 Description: A race in kernel ptllnd between deleting a peer and posting
261 new communications for it could hang communications -
262 manifesting as "Unexpectedly long timeout" messages.
267 Description: Kernel ptllnd lock ordering issue could hang a node.
272 Description: node crash on socket teardown race
275 Frequency : 'lctl peer_list' issued on a mx net
277 Description: Enable lctl's peer_list for MXLND
280 Frequency : after Ptllnd timeouts and portals congestion
282 Description: Credit overflows
283 Details : This was a bug in ptllnd connection establishment. The fix
284 implements better peer stamps to disambiguate connection
285 establishment and ensure both peers enter the credit flow
286 state machine consistently.
291 Description: kptllnd didn't propagate some network errors up to LNET
292 Details : This bug was spotted while investigating 11394. The fix
293 ensures network errors on sends and bulk transfers are
294 propagated to LNET/lustre correctly.
296 Severity : enhancement
298 Description: Fixed console chatter in case of -ETIMEDOUT.
300 Severity : enhancement
302 Description: Added D_NETTRACE for recording network packet history
303 (initially only for ptllnd). Also a separate userspace
304 ptllnd facility to gather history which should really be
305 covered by D_NETTRACE too, if only CDEBUG recorded history in
311 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
312 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
313 callback can occur before a connection has actually been
314 established. This caused an assertion failure previously.
316 Severity : enhancement
318 Description: Multiple instances for o2iblnd
319 Details : Allow multiple instances of o2iblnd to enable networking over
320 multiple HCAs and routing between them.
324 Description: lnet deadlock in router_checker
325 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
326 into BH locks to eliminate potential deadlock caused by
327 ksocknal_data_ready() preempting code holding these locks.
331 Description: Millions of failed socklnd connection attempts cause a very slow FS
332 Details : added a new route flag ksnr_scheduled to distinguish from
333 ksnr_connecting, so that a peer connection request is only turned
334 down for race concerns when an active connection to the same peer
335 is under progress (instead of just being scheduled).
337 ------------------------------------------------------------------------------
339 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
341 * Support for networks:
342 socklnd - kernels up to 2.6.16
343 qswlnd - Qsnet kernel modules 5.20 and later
344 openiblnd - IbGold 1.8.2
346 viblnd - Voltaire ibhost 3.4.5 and later
347 ciblnd - Topspin 3.2.0
348 iiblnd - Infiniserv 3.3 + PathBits patch
349 gmlnd - GM 2.1.22 and later
350 mxlnd - MX 1.2.1 or later
351 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
354 Severity : major on XT3
356 Description: libcfs overwrites /proc/sys/portals
357 Details : libcfs created a symlink from /proc/sys/portals to
358 /proc/sys/lnet for backwards compatibility. This is no
359 longer required and makes the Cray portals /proc variables
364 Description: OFED FMR API change
365 Details : This changes parameter usage to reflect a change in
366 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
367 that FMR support is only used in experimental versions of the
368 o2iblnd - this change does not affect standard usage at all.
370 Severity : enhancement
372 Description: new ko2iblnd module parameter: ib_mtu
373 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
374 HCAs. You can avoid this problem by setting the MTU to 1024
375 using this module parameter.
377 Severity : enhancement
378 Bugzilla : 11118/11620
379 Description: ptllnd small request message buffer alignment fix
380 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
381 Round up small message size on sends in case this option
382 is not supported. 11620 was a defect in the initial
383 implementation which effectively asserted all peers had to be
384 running the correct protocol version which was fixed by always
385 NAK-ing such requests and handling any misalignments they
390 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
391 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
392 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
394 Severity : enhancement
396 Description: Patchless ZC(zero copy) socklnd
397 Details : New protocol for socklnd, socklnd can support zero copy without
398 kernel patch, it's compatible with old socklnd. Checksum is
399 moved from tunables to modparams.
403 Description: When ksocknal_del_peer() is called upon a peer whose
404 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
405 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
408 Frequency : when ptlrpc is under heavy use and runs out of request buffer
410 Description: In lnet_match_blocked_msg(), md can be used without holding a
414 Frequency : very rarely
416 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
417 If connd connects a route which has been closed by
418 ksocknal_shutdown(), ksocknal_create_routes() may create new
419 routes which hold references on the peer, causing shutdown
420 process to wait for peer to disappear forever.
422 Severity : enhancement
424 Description: Dump XT3 portals traces on kptllnd timeout
425 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
426 dump Cray portals debug traces to a file. The kptllnd module
427 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
428 is the basename of the dump file.
431 Frequency : infrequent
433 Description: kernel ptllnd fix bug in connection re-establishment
434 Details : Kernel ptllnd could produce protocol errors e.g. illegal
435 matchbits and/or violate the credit flow protocol when trying
436 to re-establish a connection with a peer after an error or
439 Severity : enhancement
441 Description: Allow /proc/sys/lnet/debug to be set symbolically
442 Details : Allow debug and subsystem debug values to be read/set by name
443 in addition to numerically, for ease of use.
446 Frequency : only in configurations with LNET routers
448 Description: routes automatically marked down and recovered
449 Details : In configurations with LNET routers if a router fails routers
450 now actively try to recover routes that are down, unless they
451 are marked down by an administrator.
453 ------------------------------------------------------------------------------
455 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
458 Frequency : very rarely, in configurations with LNET routers and TCP
460 Description: incorrect data written to files on OSTs
461 Details : In certain high-load conditions incorrect data may be written
462 to files on the OST when using TCP networks.
464 ------------------------------------------------------------------------------
466 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
468 - rework CDEBUG messages rate-limiting mechanism b=10375
469 - add per-socket tunables for socklnd if the kernel is patched b=10327
471 ------------------------------------------------------------------------------
473 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
475 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
476 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
477 - quiet LNET startup LNI message for liblustre b=10128
478 - Better console error messages if 'ip2nets' can't match an IP address
479 - Fixed overflow/use-before-set bugs in linux-time.h
480 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
481 - LNET teardown failed an assertion about the route table being empty
482 - Fixed a crash in LNetEQPoll(<invalid handle>)
483 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
484 - improve debug message for liblustre/Catamount nodes (b=10116)
486 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
487 * Configuration change for the XT3
488 The PTLLND is now used to run Lustre over Portals on the XT3.
489 The configure option(s) --with-cray-portals are no longer
490 used. Rather --with-portals=<path-to-portals-includes> is
491 used to enable building on the XT3. In addition to enable
492 XT3 specific features the option --enable-cray-xt3 must be
495 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
496 * Portals has been removed, replaced by LNET.
497 LNET is new networking infrastructure for Lustre, it includes a
498 reorganized network configuration mode (see the user
499 documentation for full details) as well as support for routing
500 between different network fabrics. Lustre Networking Devices
501 (LNDS) for the supported network fabrics have also been created
502 for this new infrastructure.
504 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
509 Frequency : rare (large Voltaire clusters only)
511 Description: the default number of reserved transmit descriptors was too low
512 for some large clusters
513 Details : As a workaround, the number was increased. A proper fix includes
516 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
521 Frequency : occasional (large-scale events, cluster reboot, network failure)
523 Description: too many error messages on console obscure actual problem and
524 can slow down/panic server, or cause recovery to fail repeatedly
525 Details : enable rate-limiting of console error messages, and some messages
526 that were console errors now only go to the kernel log
528 Severity : enhancement
530 Description: add /proc/sys/portals/catastrophe entry which will report if
531 that node has previously LBUGged
533 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
535 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
537 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
539 - handle error return code in kranal_check_fma_rx() (5915,6054)
541 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
543 - update vibnal (Voltaire IB NAL)
544 - update gmnal (Myrinet NAL), gmnalid
546 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
548 * Landed portals:b_port_step as follows...
550 - removed CFS_DECL_SPIN*
551 just use 'spinlock_t' and initialise with spin_lock_init()
553 - removed CFS_DECL_MUTEX*
554 just use 'struct semaphore' and initialise with init_mutex()
556 - removed CFS_DECL_RWSEM*
557 just use 'struct rw_semaphore' and initialise with init_rwsem()
559 - renamed cfs_sleep_chan -> cfs_waitq
560 cfs_sleep_link -> cfs_waitlink
562 - fixed race in linux version of arch-independent socknal
563 (the ENOMEM/EAGAIN decision).
565 - Didn't fix problems in Darwin version of arch-independent socknal
566 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
568 - removed libcfs types from non-socknal header files (only some types
569 in the header files had been changed; the .c files hadn't been