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
15 --------------------------------------------------------------------------------
17 2007-10-26 Cluster File Systems, Inc. <info@clusterfs.com>
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
33 Description: increase send queue size for ciblnd/openiblnd
37 Description: new userspace socklnd
38 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
39 with new one - usocklnd.
41 Severity : enhancement
43 Description: Console message flood
44 Details : Make cdls ratelimiting more tunable by adding several tunable in
45 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
46 /proc/sys/lnet/console_backoff.
48 --------------------------------------------------------------------------------
50 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
52 * Support for networks:
53 socklnd - any kernel supported by Lustre,
54 qswlnd - Qsnet kernel modules 5.20 and later,
55 openiblnd - IbGold 1.8.2,
56 o2iblnd - OFED 1.1 and 1.2,
57 viblnd - Voltaire ibhost 3.4.5 and later,
58 ciblnd - Topspin 3.2.0,
59 iiblnd - Infiniserv 3.3 + PathBits patch,
60 gmlnd - GM 2.1.22 and later,
61 mxlnd - MX 1.2.1 or later,
62 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
66 Description: /proc/sys/lnet has non-sysctl entries
67 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
71 Description: TOE Kernel panic by ksocklnd
72 Details : offloaded sockets provide their own implementation of sendpage,
73 can't call tcp_sendpage() directly
77 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
78 Details : races between lnd_shutdown and peer creation prevent
79 lnd_shutdown from finishing.
83 Description: open files rlimit 1024 reached while liblustre testing
84 Details : ulnds/socklnd must close open socket after unsuccessful
89 Description: build error
90 Details : fix typos in gmlnd, ptllnd and viblnd
92 ------------------------------------------------------------------------------
94 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
96 * Support for networks:
97 socklnd - kernels up to 2.6.16,
98 qswlnd - Qsnet kernel modules 5.20 and later,
99 openiblnd - IbGold 1.8.2,
100 o2iblnd - OFED 1.1 and 1.2
101 viblnd - Voltaire ibhost 3.4.5 and later,
102 ciblnd - Topspin 3.2.0,
103 iiblnd - Infiniserv 3.3 + PathBits patch,
104 gmlnd - GM 2.1.22 and later,
105 mxlnd - MX 1.2.1 or later,
106 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
108 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
110 * Support for networks:
111 socklnd - kernels up to 2.6.16,
112 qswlnd - Qsnet kernel modules 5.20 and later,
113 openiblnd - IbGold 1.8.2,
115 viblnd - Voltaire ibhost 3.4.5 and later,
116 ciblnd - Topspin 3.2.0,
117 iiblnd - Infiniserv 3.3 + PathBits patch,
118 gmlnd - GM 2.1.22 and later,
119 mxlnd - MX 1.2.1 or later,
120 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
124 Description: Initialize cpumask before use
128 Description: ASSERTION failures when upgrading to the patchless zero-copy
130 Details : This bug affects "rolling upgrades", causing an inconsistent
131 protocol version negotiation and subsequent assertion failure
132 during rolling upgrades after the first wave of upgrades.
136 Details : Change "dropped message" CERRORs to D_NETERROR so they are
137 logged instead of creating "console chatter" when a lustre
138 timeout races with normal RPC completion.
141 Details : lnet_clear_peer_table can wait forever if user forgets to
145 Details : libcfs_id2str should check pid against LNET_PID_ANY.
149 Description: added LNET self test
150 Details : landing b_self_test
155 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
157 Details : do_div() macro is used incorrectly.
159 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
163 Description: make panic on lbug configurable
167 Description: Add OFED1.2 support to o2iblnd
168 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
169 are installed (other than kernel's in-tree infiniband), there
170 could be some problem while insmod o2iblnd (mismatch CRC of
172 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
173 this link provides solution:
174 https://bugs.openfabrics.org/show_bug.cgi?id=355
175 if extra Module.symvers is not supported in kernel, we will
176 have to run the script in bug 12316 to update
177 $LINUX/module.symvers before building o2iblnd.
178 More details about this are in bug 12316.
180 ------------------------------------------------------------------------------
182 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
183 * version 1.4.10 / 1.6.0
184 * Support for networks:
185 socklnd - kernels up to 2.6.16,
186 qswlnd - Qsnet kernel modules 5.20 and later,
187 openiblnd - IbGold 1.8.2,
189 viblnd - Voltaire ibhost 3.4.5 and later,
190 ciblnd - Topspin 3.2.0,
191 iiblnd - Infiniserv 3.3 + PathBits patch,
192 gmlnd - GM 2.1.22 and later,
193 mxlnd - MX 1.2.1 or later,
194 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
198 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
199 possibly accessed in kptllnd_shutdown. Ptllnd should init
200 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
204 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
208 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
212 Description: the_lnet.ln_finalizing was not set when the current thread is
213 about to complete messages. It only affects multi-threaded
219 Description: Changed the default kqswlnd ntxmsg=512
224 Description: Assertion failure in kernel ptllnd caused by posting passive
225 bulk buffers before connection establishment complete.
230 Description: A race in kernel ptllnd between deleting a peer and posting
231 new communications for it could hang communications -
232 manifesting as "Unexpectedly long timeout" messages.
237 Description: Kernel ptllnd lock ordering issue could hang a node.
242 Description: node crash on socket teardown race
245 Frequency : 'lctl peer_list' issued on a mx net
247 Description: Enable lctl's peer_list for MXLND
250 Frequency : after Ptllnd timeouts and portals congestion
252 Description: Credit overflows
253 Details : This was a bug in ptllnd connection establishment. The fix
254 implements better peer stamps to disambiguate connection
255 establishment and ensure both peers enter the credit flow
256 state machine consistently.
261 Description: kptllnd didn't propagate some network errors up to LNET
262 Details : This bug was spotted while investigating 11394. The fix
263 ensures network errors on sends and bulk transfers are
264 propagated to LNET/lustre correctly.
266 Severity : enhancement
268 Description: Fixed console chatter in case of -ETIMEDOUT.
270 Severity : enhancement
272 Description: Added D_NETTRACE for recording network packet history
273 (initially only for ptllnd). Also a separate userspace
274 ptllnd facility to gather history which should really be
275 covered by D_NETTRACE too, if only CDEBUG recorded history in
281 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
282 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
283 callback can occur before a connection has actually been
284 established. This caused an assertion failure previously.
286 Severity : enhancement
288 Description: Multiple instances for o2iblnd
289 Details : Allow multiple instances of o2iblnd to enable networking over
290 multiple HCAs and routing between them.
294 Description: lnet deadlock in router_checker
295 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
296 into BH locks to eliminate potential deadlock caused by
297 ksocknal_data_ready() preempting code holding these locks.
301 Description: Millions of failed socklnd connection attempts cause a very slow FS
302 Details : added a new route flag ksnr_scheduled to distinguish from
303 ksnr_connecting, so that a peer connection request is only turned
304 down for race concerns when an active connection to the same peer
305 is under progress (instead of just being scheduled).
307 ------------------------------------------------------------------------------
309 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
311 * Support for networks:
312 socklnd - kernels up to 2.6.16
313 qswlnd - Qsnet kernel modules 5.20 and later
314 openiblnd - IbGold 1.8.2
316 viblnd - Voltaire ibhost 3.4.5 and later
317 ciblnd - Topspin 3.2.0
318 iiblnd - Infiniserv 3.3 + PathBits patch
319 gmlnd - GM 2.1.22 and later
320 mxlnd - MX 1.2.1 or later
321 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
324 Severity : major on XT3
326 Description: libcfs overwrites /proc/sys/portals
327 Details : libcfs created a symlink from /proc/sys/portals to
328 /proc/sys/lnet for backwards compatibility. This is no
329 longer required and makes the Cray portals /proc variables
334 Description: OFED FMR API change
335 Details : This changes parameter usage to reflect a change in
336 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
337 that FMR support is only used in experimental versions of the
338 o2iblnd - this change does not affect standard usage at all.
340 Severity : enhancement
342 Description: new ko2iblnd module parameter: ib_mtu
343 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
344 HCAs. You can avoid this problem by setting the MTU to 1024
345 using this module parameter.
347 Severity : enhancement
348 Bugzilla : 11118/11620
349 Description: ptllnd small request message buffer alignment fix
350 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
351 Round up small message size on sends in case this option
352 is not supported. 11620 was a defect in the initial
353 implementation which effectively asserted all peers had to be
354 running the correct protocol version which was fixed by always
355 NAK-ing such requests and handling any misalignments they
360 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
361 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
362 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
364 Severity : enhancement
366 Description: Patchless ZC(zero copy) socklnd
367 Details : New protocol for socklnd, socklnd can support zero copy without
368 kernel patch, it's compatible with old socklnd. Checksum is
369 moved from tunables to modparams.
373 Description: When ksocknal_del_peer() is called upon a peer whose
374 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
375 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
378 Frequency : when ptlrpc is under heavy use and runs out of request buffer
380 Description: In lnet_match_blocked_msg(), md can be used without holding a
384 Frequency : very rarely
386 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
387 If connd connects a route which has been closed by
388 ksocknal_shutdown(), ksocknal_create_routes() may create new
389 routes which hold references on the peer, causing shutdown
390 process to wait for peer to disappear forever.
392 Severity : enhancement
394 Description: Dump XT3 portals traces on kptllnd timeout
395 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
396 dump Cray portals debug traces to a file. The kptllnd module
397 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
398 is the basename of the dump file.
401 Frequency : infrequent
403 Description: kernel ptllnd fix bug in connection re-establishment
404 Details : Kernel ptllnd could produce protocol errors e.g. illegal
405 matchbits and/or violate the credit flow protocol when trying
406 to re-establish a connection with a peer after an error or
409 Severity : enhancement
411 Description: Allow /proc/sys/lnet/debug to be set symbolically
412 Details : Allow debug and subsystem debug values to be read/set by name
413 in addition to numerically, for ease of use.
416 Frequency : only in configurations with LNET routers
418 Description: routes automatically marked down and recovered
419 Details : In configurations with LNET routers if a router fails routers
420 now actively try to recover routes that are down, unless they
421 are marked down by an administrator.
423 ------------------------------------------------------------------------------
425 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
428 Frequency : very rarely, in configurations with LNET routers and TCP
430 Description: incorrect data written to files on OSTs
431 Details : In certain high-load conditions incorrect data may be written
432 to files on the OST when using TCP networks.
434 ------------------------------------------------------------------------------
436 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
438 - rework CDEBUG messages rate-limiting mechanism b=10375
439 - add per-socket tunables for socklnd if the kernel is patched b=10327
441 ------------------------------------------------------------------------------
443 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
445 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
446 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
447 - quiet LNET startup LNI message for liblustre b=10128
448 - Better console error messages if 'ip2nets' can't match an IP address
449 - Fixed overflow/use-before-set bugs in linux-time.h
450 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
451 - LNET teardown failed an assertion about the route table being empty
452 - Fixed a crash in LNetEQPoll(<invalid handle>)
453 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
454 - improve debug message for liblustre/Catamount nodes (b=10116)
456 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
457 * Configuration change for the XT3
458 The PTLLND is now used to run Lustre over Portals on the XT3.
459 The configure option(s) --with-cray-portals are no longer
460 used. Rather --with-portals=<path-to-portals-includes> is
461 used to enable building on the XT3. In addition to enable
462 XT3 specific features the option --enable-cray-xt3 must be
465 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
466 * Portals has been removed, replaced by LNET.
467 LNET is new networking infrastructure for Lustre, it includes a
468 reorganized network configuration mode (see the user
469 documentation for full details) as well as support for routing
470 between different network fabrics. Lustre Networking Devices
471 (LNDS) for the supported network fabrics have also been created
472 for this new infrastructure.
474 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
479 Frequency : rare (large Voltaire clusters only)
481 Description: the default number of reserved transmit descriptors was too low
482 for some large clusters
483 Details : As a workaround, the number was increased. A proper fix includes
486 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
491 Frequency : occasional (large-scale events, cluster reboot, network failure)
493 Description: too many error messages on console obscure actual problem and
494 can slow down/panic server, or cause recovery to fail repeatedly
495 Details : enable rate-limiting of console error messages, and some messages
496 that were console errors now only go to the kernel log
498 Severity : enhancement
500 Description: add /proc/sys/portals/catastrophe entry which will report if
501 that node has previously LBUGged
503 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
505 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
507 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
509 - handle error return code in kranal_check_fma_rx() (5915,6054)
511 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
513 - update vibnal (Voltaire IB NAL)
514 - update gmnal (Myrinet NAL), gmnalid
516 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
518 * Landed portals:b_port_step as follows...
520 - removed CFS_DECL_SPIN*
521 just use 'spinlock_t' and initialise with spin_lock_init()
523 - removed CFS_DECL_MUTEX*
524 just use 'struct semaphore' and initialise with init_mutex()
526 - removed CFS_DECL_RWSEM*
527 just use 'struct rw_semaphore' and initialise with init_rwsem()
529 - renamed cfs_sleep_chan -> cfs_waitq
530 cfs_sleep_link -> cfs_waitlink
532 - fixed race in linux version of arch-independent socknal
533 (the ENOMEM/EAGAIN decision).
535 - Didn't fix problems in Darwin version of arch-independent socknal
536 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
538 - removed libcfs types from non-socknal header files (only some types
539 in the header files had been changed; the .c files hadn't been