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: new userspace socklnd
34 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
35 with new one - usocklnd.
37 Severity : enhancement
39 Description: Console message flood
40 Details : Make cdls ratelimiting more tunable by adding several tunable in
41 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
42 /proc/sys/lnet/console_backoff.
44 --------------------------------------------------------------------------------
46 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
48 * Support for networks:
49 socklnd - any kernel supported by Lustre,
50 qswlnd - Qsnet kernel modules 5.20 and later,
51 openiblnd - IbGold 1.8.2,
52 o2iblnd - OFED 1.1 and 1.2,
53 viblnd - Voltaire ibhost 3.4.5 and later,
54 ciblnd - Topspin 3.2.0,
55 iiblnd - Infiniserv 3.3 + PathBits patch,
56 gmlnd - GM 2.1.22 and later,
57 mxlnd - MX 1.2.1 or later,
58 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
62 Description: /proc/sys/lnet has non-sysctl entries
63 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
67 Description: TOE Kernel panic by ksocklnd
68 Details : offloaded sockets provide their own implementation of sendpage,
69 can't call tcp_sendpage() directly
73 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
74 Details : races between lnd_shutdown and peer creation prevent
75 lnd_shutdown from finishing.
79 Description: open files rlimit 1024 reached while liblustre testing
80 Details : ulnds/socklnd must close open socket after unsuccessful
85 Description: build error
86 Details : fix typos in gmlnd, ptllnd and viblnd
88 ------------------------------------------------------------------------------
90 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
92 * Support for networks:
93 socklnd - kernels up to 2.6.16,
94 qswlnd - Qsnet kernel modules 5.20 and later,
95 openiblnd - IbGold 1.8.2,
96 o2iblnd - OFED 1.1 and 1.2
97 viblnd - Voltaire ibhost 3.4.5 and later,
98 ciblnd - Topspin 3.2.0,
99 iiblnd - Infiniserv 3.3 + PathBits patch,
100 gmlnd - GM 2.1.22 and later,
101 mxlnd - MX 1.2.1 or later,
102 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
104 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
106 * Support for networks:
107 socklnd - kernels up to 2.6.16,
108 qswlnd - Qsnet kernel modules 5.20 and later,
109 openiblnd - IbGold 1.8.2,
111 viblnd - Voltaire ibhost 3.4.5 and later,
112 ciblnd - Topspin 3.2.0,
113 iiblnd - Infiniserv 3.3 + PathBits patch,
114 gmlnd - GM 2.1.22 and later,
115 mxlnd - MX 1.2.1 or later,
116 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
120 Description: Initialize cpumask before use
124 Description: ASSERTION failures when upgrading to the patchless zero-copy
126 Details : This bug affects "rolling upgrades", causing an inconsistent
127 protocol version negotiation and subsequent assertion failure
128 during rolling upgrades after the first wave of upgrades.
132 Details : Change "dropped message" CERRORs to D_NETERROR so they are
133 logged instead of creating "console chatter" when a lustre
134 timeout races with normal RPC completion.
137 Details : lnet_clear_peer_table can wait forever if user forgets to
141 Details : libcfs_id2str should check pid against LNET_PID_ANY.
145 Description: added LNET self test
146 Details : landing b_self_test
151 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
153 Details : do_div() macro is used incorrectly.
155 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
159 Description: make panic on lbug configurable
163 Description: Add OFED1.2 support to o2iblnd
164 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
165 are installed (other than kernel's in-tree infiniband), there
166 could be some problem while insmod o2iblnd (mismatch CRC of
168 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
169 this link provides solution:
170 https://bugs.openfabrics.org/show_bug.cgi?id=355
171 if extra Module.symvers is not supported in kernel, we will
172 have to run the script in bug 12316 to update
173 $LINUX/module.symvers before building o2iblnd.
174 More details about this are in bug 12316.
176 ------------------------------------------------------------------------------
178 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
179 * version 1.4.10 / 1.6.0
180 * Support for networks:
181 socklnd - kernels up to 2.6.16,
182 qswlnd - Qsnet kernel modules 5.20 and later,
183 openiblnd - IbGold 1.8.2,
185 viblnd - Voltaire ibhost 3.4.5 and later,
186 ciblnd - Topspin 3.2.0,
187 iiblnd - Infiniserv 3.3 + PathBits patch,
188 gmlnd - GM 2.1.22 and later,
189 mxlnd - MX 1.2.1 or later,
190 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
194 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
195 possibly accessed in kptllnd_shutdown. Ptllnd should init
196 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
200 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
204 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
208 Description: the_lnet.ln_finalizing was not set when the current thread is
209 about to complete messages. It only affects multi-threaded
215 Description: Changed the default kqswlnd ntxmsg=512
220 Description: Assertion failure in kernel ptllnd caused by posting passive
221 bulk buffers before connection establishment complete.
226 Description: A race in kernel ptllnd between deleting a peer and posting
227 new communications for it could hang communications -
228 manifesting as "Unexpectedly long timeout" messages.
233 Description: Kernel ptllnd lock ordering issue could hang a node.
238 Description: node crash on socket teardown race
241 Frequency : 'lctl peer_list' issued on a mx net
243 Description: Enable lctl's peer_list for MXLND
246 Frequency : after Ptllnd timeouts and portals congestion
248 Description: Credit overflows
249 Details : This was a bug in ptllnd connection establishment. The fix
250 implements better peer stamps to disambiguate connection
251 establishment and ensure both peers enter the credit flow
252 state machine consistently.
257 Description: kptllnd didn't propagate some network errors up to LNET
258 Details : This bug was spotted while investigating 11394. The fix
259 ensures network errors on sends and bulk transfers are
260 propagated to LNET/lustre correctly.
262 Severity : enhancement
264 Description: Fixed console chatter in case of -ETIMEDOUT.
266 Severity : enhancement
268 Description: Added D_NETTRACE for recording network packet history
269 (initially only for ptllnd). Also a separate userspace
270 ptllnd facility to gather history which should really be
271 covered by D_NETTRACE too, if only CDEBUG recorded history in
277 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
278 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
279 callback can occur before a connection has actually been
280 established. This caused an assertion failure previously.
282 Severity : enhancement
284 Description: Multiple instances for o2iblnd
285 Details : Allow multiple instances of o2iblnd to enable networking over
286 multiple HCAs and routing between them.
290 Description: lnet deadlock in router_checker
291 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
292 into BH locks to eliminate potential deadlock caused by
293 ksocknal_data_ready() preempting code holding these locks.
297 Description: Millions of failed socklnd connection attempts cause a very slow FS
298 Details : added a new route flag ksnr_scheduled to distinguish from
299 ksnr_connecting, so that a peer connection request is only turned
300 down for race concerns when an active connection to the same peer
301 is under progress (instead of just being scheduled).
303 ------------------------------------------------------------------------------
305 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
307 * Support for networks:
308 socklnd - kernels up to 2.6.16
309 qswlnd - Qsnet kernel modules 5.20 and later
310 openiblnd - IbGold 1.8.2
312 viblnd - Voltaire ibhost 3.4.5 and later
313 ciblnd - Topspin 3.2.0
314 iiblnd - Infiniserv 3.3 + PathBits patch
315 gmlnd - GM 2.1.22 and later
316 mxlnd - MX 1.2.1 or later
317 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
320 Severity : major on XT3
322 Description: libcfs overwrites /proc/sys/portals
323 Details : libcfs created a symlink from /proc/sys/portals to
324 /proc/sys/lnet for backwards compatibility. This is no
325 longer required and makes the Cray portals /proc variables
330 Description: OFED FMR API change
331 Details : This changes parameter usage to reflect a change in
332 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
333 that FMR support is only used in experimental versions of the
334 o2iblnd - this change does not affect standard usage at all.
336 Severity : enhancement
338 Description: new ko2iblnd module parameter: ib_mtu
339 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
340 HCAs. You can avoid this problem by setting the MTU to 1024
341 using this module parameter.
343 Severity : enhancement
344 Bugzilla : 11118/11620
345 Description: ptllnd small request message buffer alignment fix
346 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
347 Round up small message size on sends in case this option
348 is not supported. 11620 was a defect in the initial
349 implementation which effectively asserted all peers had to be
350 running the correct protocol version which was fixed by always
351 NAK-ing such requests and handling any misalignments they
356 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
357 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
358 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
360 Severity : enhancement
362 Description: Patchless ZC(zero copy) socklnd
363 Details : New protocol for socklnd, socklnd can support zero copy without
364 kernel patch, it's compatible with old socklnd. Checksum is
365 moved from tunables to modparams.
369 Description: When ksocknal_del_peer() is called upon a peer whose
370 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
371 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
374 Frequency : when ptlrpc is under heavy use and runs out of request buffer
376 Description: In lnet_match_blocked_msg(), md can be used without holding a
380 Frequency : very rarely
382 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
383 If connd connects a route which has been closed by
384 ksocknal_shutdown(), ksocknal_create_routes() may create new
385 routes which hold references on the peer, causing shutdown
386 process to wait for peer to disappear forever.
388 Severity : enhancement
390 Description: Dump XT3 portals traces on kptllnd timeout
391 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
392 dump Cray portals debug traces to a file. The kptllnd module
393 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
394 is the basename of the dump file.
397 Frequency : infrequent
399 Description: kernel ptllnd fix bug in connection re-establishment
400 Details : Kernel ptllnd could produce protocol errors e.g. illegal
401 matchbits and/or violate the credit flow protocol when trying
402 to re-establish a connection with a peer after an error or
405 Severity : enhancement
407 Description: Allow /proc/sys/lnet/debug to be set symbolically
408 Details : Allow debug and subsystem debug values to be read/set by name
409 in addition to numerically, for ease of use.
412 Frequency : only in configurations with LNET routers
414 Description: routes automatically marked down and recovered
415 Details : In configurations with LNET routers if a router fails routers
416 now actively try to recover routes that are down, unless they
417 are marked down by an administrator.
419 ------------------------------------------------------------------------------
421 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
424 Frequency : very rarely, in configurations with LNET routers and TCP
426 Description: incorrect data written to files on OSTs
427 Details : In certain high-load conditions incorrect data may be written
428 to files on the OST when using TCP networks.
430 ------------------------------------------------------------------------------
432 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
434 - rework CDEBUG messages rate-limiting mechanism b=10375
435 - add per-socket tunables for socklnd if the kernel is patched b=10327
437 ------------------------------------------------------------------------------
439 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
441 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
442 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
443 - quiet LNET startup LNI message for liblustre b=10128
444 - Better console error messages if 'ip2nets' can't match an IP address
445 - Fixed overflow/use-before-set bugs in linux-time.h
446 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
447 - LNET teardown failed an assertion about the route table being empty
448 - Fixed a crash in LNetEQPoll(<invalid handle>)
449 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
450 - improve debug message for liblustre/Catamount nodes (b=10116)
452 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
453 * Configuration change for the XT3
454 The PTLLND is now used to run Lustre over Portals on the XT3.
455 The configure option(s) --with-cray-portals are no longer
456 used. Rather --with-portals=<path-to-portals-includes> is
457 used to enable building on the XT3. In addition to enable
458 XT3 specific features the option --enable-cray-xt3 must be
461 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
462 * Portals has been removed, replaced by LNET.
463 LNET is new networking infrastructure for Lustre, it includes a
464 reorganized network configuration mode (see the user
465 documentation for full details) as well as support for routing
466 between different network fabrics. Lustre Networking Devices
467 (LNDS) for the supported network fabrics have also been created
468 for this new infrastructure.
470 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
475 Frequency : rare (large Voltaire clusters only)
477 Description: the default number of reserved transmit descriptors was too low
478 for some large clusters
479 Details : As a workaround, the number was increased. A proper fix includes
482 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
487 Frequency : occasional (large-scale events, cluster reboot, network failure)
489 Description: too many error messages on console obscure actual problem and
490 can slow down/panic server, or cause recovery to fail repeatedly
491 Details : enable rate-limiting of console error messages, and some messages
492 that were console errors now only go to the kernel log
494 Severity : enhancement
496 Description: add /proc/sys/portals/catastrophe entry which will report if
497 that node has previously LBUGged
499 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
501 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
503 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
505 - handle error return code in kranal_check_fma_rx() (5915,6054)
507 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
509 - update vibnal (Voltaire IB NAL)
510 - update gmnal (Myrinet NAL), gmnalid
512 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
514 * Landed portals:b_port_step as follows...
516 - removed CFS_DECL_SPIN*
517 just use 'spinlock_t' and initialise with spin_lock_init()
519 - removed CFS_DECL_MUTEX*
520 just use 'struct semaphore' and initialise with init_mutex()
522 - removed CFS_DECL_RWSEM*
523 just use 'struct rw_semaphore' and initialise with init_rwsem()
525 - renamed cfs_sleep_chan -> cfs_waitq
526 cfs_sleep_link -> cfs_waitlink
528 - fixed race in linux version of arch-independent socknal
529 (the ENOMEM/EAGAIN decision).
531 - Didn't fix problems in Darwin version of arch-independent socknal
532 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
534 - removed libcfs types from non-socknal header files (only some types
535 in the header files had been changed; the .c files hadn't been