1 tbd Cluster File Systems, Inc. <info@clusterfs.com>
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: liblustre network error
18 Details : liblustre clients should understand LNET_ACCEPT_PORT environment
19 variable even if they don't start lnet acceptor.
23 Description: Strange message from lnet (Ignoring prediction from the future)
24 Details : Incorrect calculation of peer's last_alive value in ksocklnd
26 --------------------------------------------------------------------------------
28 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
30 * Support for networks:
31 socklnd - any kernel supported by Lustre,
32 qswlnd - Qsnet kernel modules 5.20 and later,
33 openiblnd - IbGold 1.8.2,
34 o2iblnd - OFED 1.1 and 1.2.0, 1.2.5.
35 viblnd - Voltaire ibhost 3.4.5 and later,
36 ciblnd - Topspin 3.2.0,
37 iiblnd - Infiniserv 3.3 + PathBits patch,
38 gmlnd - GM 2.1.22 and later,
39 mxlnd - MX 1.2.1 or later,
40 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
44 Description: ASSERTION(me == md->md_me) failed in lnet_match_md()
48 Description: increase send queue size for ciblnd/openiblnd
52 Description: new userspace socklnd
53 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
54 with new one - usocklnd.
56 Severity : enhancement
58 Description: Console message flood
59 Details : Make cdls ratelimiting more tunable by adding several tunable in
60 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
61 /proc/sys/lnet/console_backoff.
63 --------------------------------------------------------------------------------
65 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
67 * Support for networks:
68 socklnd - any kernel supported by Lustre,
69 qswlnd - Qsnet kernel modules 5.20 and later,
70 openiblnd - IbGold 1.8.2,
71 o2iblnd - OFED 1.1 and 1.2,
72 viblnd - Voltaire ibhost 3.4.5 and later,
73 ciblnd - Topspin 3.2.0,
74 iiblnd - Infiniserv 3.3 + PathBits patch,
75 gmlnd - GM 2.1.22 and later,
76 mxlnd - MX 1.2.1 or later,
77 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
81 Description: /proc/sys/lnet has non-sysctl entries
82 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
86 Description: TOE Kernel panic by ksocklnd
87 Details : offloaded sockets provide their own implementation of sendpage,
88 can't call tcp_sendpage() directly
92 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
93 Details : races between lnd_shutdown and peer creation prevent
94 lnd_shutdown from finishing.
98 Description: open files rlimit 1024 reached while liblustre testing
99 Details : ulnds/socklnd must close open socket after unsuccessful
104 Description: build error
105 Details : fix typos in gmlnd, ptllnd and viblnd
107 ------------------------------------------------------------------------------
109 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
111 * Support for networks:
112 socklnd - kernels up to 2.6.16,
113 qswlnd - Qsnet kernel modules 5.20 and later,
114 openiblnd - IbGold 1.8.2,
115 o2iblnd - OFED 1.1 and 1.2
116 viblnd - Voltaire ibhost 3.4.5 and later,
117 ciblnd - Topspin 3.2.0,
118 iiblnd - Infiniserv 3.3 + PathBits patch,
119 gmlnd - GM 2.1.22 and later,
120 mxlnd - MX 1.2.1 or later,
121 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
123 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
125 * Support for networks:
126 socklnd - kernels up to 2.6.16,
127 qswlnd - Qsnet kernel modules 5.20 and later,
128 openiblnd - IbGold 1.8.2,
130 viblnd - Voltaire ibhost 3.4.5 and later,
131 ciblnd - Topspin 3.2.0,
132 iiblnd - Infiniserv 3.3 + PathBits patch,
133 gmlnd - GM 2.1.22 and later,
134 mxlnd - MX 1.2.1 or later,
135 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
139 Description: Initialize cpumask before use
143 Description: ASSERTION failures when upgrading to the patchless zero-copy
145 Details : This bug affects "rolling upgrades", causing an inconsistent
146 protocol version negotiation and subsequent assertion failure
147 during rolling upgrades after the first wave of upgrades.
151 Details : Change "dropped message" CERRORs to D_NETERROR so they are
152 logged instead of creating "console chatter" when a lustre
153 timeout races with normal RPC completion.
156 Details : lnet_clear_peer_table can wait forever if user forgets to
160 Details : libcfs_id2str should check pid against LNET_PID_ANY.
164 Description: added LNET self test
165 Details : landing b_self_test
170 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
172 Details : do_div() macro is used incorrectly.
174 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
178 Description: make panic on lbug configurable
182 Description: Add OFED1.2 support to o2iblnd
183 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
184 are installed (other than kernel's in-tree infiniband), there
185 could be some problem while insmod o2iblnd (mismatch CRC of
187 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
188 this link provides solution:
189 https://bugs.openfabrics.org/show_bug.cgi?id=355
190 if extra Module.symvers is not supported in kernel, we will
191 have to run the script in bug 12316 to update
192 $LINUX/module.symvers before building o2iblnd.
193 More details about this are in bug 12316.
195 ------------------------------------------------------------------------------
197 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
198 * version 1.4.10 / 1.6.0
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: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
214 possibly accessed in kptllnd_shutdown. Ptllnd should init
215 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
219 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
223 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
227 Description: the_lnet.ln_finalizing was not set when the current thread is
228 about to complete messages. It only affects multi-threaded
234 Description: Changed the default kqswlnd ntxmsg=512
239 Description: Assertion failure in kernel ptllnd caused by posting passive
240 bulk buffers before connection establishment complete.
245 Description: A race in kernel ptllnd between deleting a peer and posting
246 new communications for it could hang communications -
247 manifesting as "Unexpectedly long timeout" messages.
252 Description: Kernel ptllnd lock ordering issue could hang a node.
257 Description: node crash on socket teardown race
260 Frequency : 'lctl peer_list' issued on a mx net
262 Description: Enable lctl's peer_list for MXLND
265 Frequency : after Ptllnd timeouts and portals congestion
267 Description: Credit overflows
268 Details : This was a bug in ptllnd connection establishment. The fix
269 implements better peer stamps to disambiguate connection
270 establishment and ensure both peers enter the credit flow
271 state machine consistently.
276 Description: kptllnd didn't propagate some network errors up to LNET
277 Details : This bug was spotted while investigating 11394. The fix
278 ensures network errors on sends and bulk transfers are
279 propagated to LNET/lustre correctly.
281 Severity : enhancement
283 Description: Fixed console chatter in case of -ETIMEDOUT.
285 Severity : enhancement
287 Description: Added D_NETTRACE for recording network packet history
288 (initially only for ptllnd). Also a separate userspace
289 ptllnd facility to gather history which should really be
290 covered by D_NETTRACE too, if only CDEBUG recorded history in
296 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
297 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
298 callback can occur before a connection has actually been
299 established. This caused an assertion failure previously.
301 Severity : enhancement
303 Description: Multiple instances for o2iblnd
304 Details : Allow multiple instances of o2iblnd to enable networking over
305 multiple HCAs and routing between them.
309 Description: lnet deadlock in router_checker
310 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
311 into BH locks to eliminate potential deadlock caused by
312 ksocknal_data_ready() preempting code holding these locks.
316 Description: Millions of failed socklnd connection attempts cause a very slow FS
317 Details : added a new route flag ksnr_scheduled to distinguish from
318 ksnr_connecting, so that a peer connection request is only turned
319 down for race concerns when an active connection to the same peer
320 is under progress (instead of just being scheduled).
322 ------------------------------------------------------------------------------
324 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
326 * Support for networks:
327 socklnd - kernels up to 2.6.16
328 qswlnd - Qsnet kernel modules 5.20 and later
329 openiblnd - IbGold 1.8.2
331 viblnd - Voltaire ibhost 3.4.5 and later
332 ciblnd - Topspin 3.2.0
333 iiblnd - Infiniserv 3.3 + PathBits patch
334 gmlnd - GM 2.1.22 and later
335 mxlnd - MX 1.2.1 or later
336 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
339 Severity : major on XT3
341 Description: libcfs overwrites /proc/sys/portals
342 Details : libcfs created a symlink from /proc/sys/portals to
343 /proc/sys/lnet for backwards compatibility. This is no
344 longer required and makes the Cray portals /proc variables
349 Description: OFED FMR API change
350 Details : This changes parameter usage to reflect a change in
351 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
352 that FMR support is only used in experimental versions of the
353 o2iblnd - this change does not affect standard usage at all.
355 Severity : enhancement
357 Description: new ko2iblnd module parameter: ib_mtu
358 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
359 HCAs. You can avoid this problem by setting the MTU to 1024
360 using this module parameter.
362 Severity : enhancement
363 Bugzilla : 11118/11620
364 Description: ptllnd small request message buffer alignment fix
365 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
366 Round up small message size on sends in case this option
367 is not supported. 11620 was a defect in the initial
368 implementation which effectively asserted all peers had to be
369 running the correct protocol version which was fixed by always
370 NAK-ing such requests and handling any misalignments they
375 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
376 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
377 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
379 Severity : enhancement
381 Description: Patchless ZC(zero copy) socklnd
382 Details : New protocol for socklnd, socklnd can support zero copy without
383 kernel patch, it's compatible with old socklnd. Checksum is
384 moved from tunables to modparams.
388 Description: When ksocknal_del_peer() is called upon a peer whose
389 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
390 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
393 Frequency : when ptlrpc is under heavy use and runs out of request buffer
395 Description: In lnet_match_blocked_msg(), md can be used without holding a
399 Frequency : very rarely
401 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
402 If connd connects a route which has been closed by
403 ksocknal_shutdown(), ksocknal_create_routes() may create new
404 routes which hold references on the peer, causing shutdown
405 process to wait for peer to disappear forever.
407 Severity : enhancement
409 Description: Dump XT3 portals traces on kptllnd timeout
410 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
411 dump Cray portals debug traces to a file. The kptllnd module
412 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
413 is the basename of the dump file.
416 Frequency : infrequent
418 Description: kernel ptllnd fix bug in connection re-establishment
419 Details : Kernel ptllnd could produce protocol errors e.g. illegal
420 matchbits and/or violate the credit flow protocol when trying
421 to re-establish a connection with a peer after an error or
424 Severity : enhancement
426 Description: Allow /proc/sys/lnet/debug to be set symbolically
427 Details : Allow debug and subsystem debug values to be read/set by name
428 in addition to numerically, for ease of use.
431 Frequency : only in configurations with LNET routers
433 Description: routes automatically marked down and recovered
434 Details : In configurations with LNET routers if a router fails routers
435 now actively try to recover routes that are down, unless they
436 are marked down by an administrator.
438 ------------------------------------------------------------------------------
440 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
443 Frequency : very rarely, in configurations with LNET routers and TCP
445 Description: incorrect data written to files on OSTs
446 Details : In certain high-load conditions incorrect data may be written
447 to files on the OST when using TCP networks.
449 ------------------------------------------------------------------------------
451 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
453 - rework CDEBUG messages rate-limiting mechanism b=10375
454 - add per-socket tunables for socklnd if the kernel is patched b=10327
456 ------------------------------------------------------------------------------
458 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
460 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
461 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
462 - quiet LNET startup LNI message for liblustre b=10128
463 - Better console error messages if 'ip2nets' can't match an IP address
464 - Fixed overflow/use-before-set bugs in linux-time.h
465 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
466 - LNET teardown failed an assertion about the route table being empty
467 - Fixed a crash in LNetEQPoll(<invalid handle>)
468 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
469 - improve debug message for liblustre/Catamount nodes (b=10116)
471 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
472 * Configuration change for the XT3
473 The PTLLND is now used to run Lustre over Portals on the XT3.
474 The configure option(s) --with-cray-portals are no longer
475 used. Rather --with-portals=<path-to-portals-includes> is
476 used to enable building on the XT3. In addition to enable
477 XT3 specific features the option --enable-cray-xt3 must be
480 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
481 * Portals has been removed, replaced by LNET.
482 LNET is new networking infrastructure for Lustre, it includes a
483 reorganized network configuration mode (see the user
484 documentation for full details) as well as support for routing
485 between different network fabrics. Lustre Networking Devices
486 (LNDS) for the supported network fabrics have also been created
487 for this new infrastructure.
489 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
494 Frequency : rare (large Voltaire clusters only)
496 Description: the default number of reserved transmit descriptors was too low
497 for some large clusters
498 Details : As a workaround, the number was increased. A proper fix includes
501 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
506 Frequency : occasional (large-scale events, cluster reboot, network failure)
508 Description: too many error messages on console obscure actual problem and
509 can slow down/panic server, or cause recovery to fail repeatedly
510 Details : enable rate-limiting of console error messages, and some messages
511 that were console errors now only go to the kernel log
513 Severity : enhancement
515 Description: add /proc/sys/portals/catastrophe entry which will report if
516 that node has previously LBUGged
518 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
520 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
522 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
524 - handle error return code in kranal_check_fma_rx() (5915,6054)
526 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
528 - update vibnal (Voltaire IB NAL)
529 - update gmnal (Myrinet NAL), gmnalid
531 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
533 * Landed portals:b_port_step as follows...
535 - removed CFS_DECL_SPIN*
536 just use 'spinlock_t' and initialise with spin_lock_init()
538 - removed CFS_DECL_MUTEX*
539 just use 'struct semaphore' and initialise with init_mutex()
541 - removed CFS_DECL_RWSEM*
542 just use 'struct rw_semaphore' and initialise with init_rwsem()
544 - renamed cfs_sleep_chan -> cfs_waitq
545 cfs_sleep_link -> cfs_waitlink
547 - fixed race in linux version of arch-independent socknal
548 (the ENOMEM/EAGAIN decision).
550 - Didn't fix problems in Darwin version of arch-independent socknal
551 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
553 - removed libcfs types from non-socknal header files (only some types
554 in the header files had been changed; the .c files hadn't been