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 2007-12-07 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: ASSERTION(me == md->md_me) failed in lnet_match_md()
37 Description: increase send queue size for ciblnd/openiblnd
41 Description: new userspace socklnd
42 Details : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
43 with new one - usocklnd.
45 Severity : enhancement
47 Description: Console message flood
48 Details : Make cdls ratelimiting more tunable by adding several tunable in
49 procfs /proc/sys/lnet/console_{min,max}_delay_centisecs and
50 /proc/sys/lnet/console_backoff.
52 --------------------------------------------------------------------------------
54 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
56 * Support for networks:
57 socklnd - any kernel supported by Lustre,
58 qswlnd - Qsnet kernel modules 5.20 and later,
59 openiblnd - IbGold 1.8.2,
60 o2iblnd - OFED 1.1 and 1.2,
61 viblnd - Voltaire ibhost 3.4.5 and later,
62 ciblnd - Topspin 3.2.0,
63 iiblnd - Infiniserv 3.3 + PathBits patch,
64 gmlnd - GM 2.1.22 and later,
65 mxlnd - MX 1.2.1 or later,
66 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
70 Description: /proc/sys/lnet has non-sysctl entries
71 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
75 Description: TOE Kernel panic by ksocklnd
76 Details : offloaded sockets provide their own implementation of sendpage,
77 can't call tcp_sendpage() directly
81 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
82 Details : races between lnd_shutdown and peer creation prevent
83 lnd_shutdown from finishing.
87 Description: open files rlimit 1024 reached while liblustre testing
88 Details : ulnds/socklnd must close open socket after unsuccessful
93 Description: build error
94 Details : fix typos in gmlnd, ptllnd and viblnd
96 ------------------------------------------------------------------------------
98 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
100 * Support for networks:
101 socklnd - kernels up to 2.6.16,
102 qswlnd - Qsnet kernel modules 5.20 and later,
103 openiblnd - IbGold 1.8.2,
104 o2iblnd - OFED 1.1 and 1.2
105 viblnd - Voltaire ibhost 3.4.5 and later,
106 ciblnd - Topspin 3.2.0,
107 iiblnd - Infiniserv 3.3 + PathBits patch,
108 gmlnd - GM 2.1.22 and later,
109 mxlnd - MX 1.2.1 or later,
110 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
112 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
114 * Support for networks:
115 socklnd - kernels up to 2.6.16,
116 qswlnd - Qsnet kernel modules 5.20 and later,
117 openiblnd - IbGold 1.8.2,
119 viblnd - Voltaire ibhost 3.4.5 and later,
120 ciblnd - Topspin 3.2.0,
121 iiblnd - Infiniserv 3.3 + PathBits patch,
122 gmlnd - GM 2.1.22 and later,
123 mxlnd - MX 1.2.1 or later,
124 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
128 Description: Initialize cpumask before use
132 Description: ASSERTION failures when upgrading to the patchless zero-copy
134 Details : This bug affects "rolling upgrades", causing an inconsistent
135 protocol version negotiation and subsequent assertion failure
136 during rolling upgrades after the first wave of upgrades.
140 Details : Change "dropped message" CERRORs to D_NETERROR so they are
141 logged instead of creating "console chatter" when a lustre
142 timeout races with normal RPC completion.
145 Details : lnet_clear_peer_table can wait forever if user forgets to
149 Details : libcfs_id2str should check pid against LNET_PID_ANY.
153 Description: added LNET self test
154 Details : landing b_self_test
159 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
161 Details : do_div() macro is used incorrectly.
163 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
167 Description: make panic on lbug configurable
171 Description: Add OFED1.2 support to o2iblnd
172 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
173 are installed (other than kernel's in-tree infiniband), there
174 could be some problem while insmod o2iblnd (mismatch CRC of
176 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
177 this link provides solution:
178 https://bugs.openfabrics.org/show_bug.cgi?id=355
179 if extra Module.symvers is not supported in kernel, we will
180 have to run the script in bug 12316 to update
181 $LINUX/module.symvers before building o2iblnd.
182 More details about this are in bug 12316.
184 ------------------------------------------------------------------------------
186 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
187 * version 1.4.10 / 1.6.0
188 * Support for networks:
189 socklnd - kernels up to 2.6.16,
190 qswlnd - Qsnet kernel modules 5.20 and later,
191 openiblnd - IbGold 1.8.2,
193 viblnd - Voltaire ibhost 3.4.5 and later,
194 ciblnd - Topspin 3.2.0,
195 iiblnd - Infiniserv 3.3 + PathBits patch,
196 gmlnd - GM 2.1.22 and later,
197 mxlnd - MX 1.2.1 or later,
198 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
202 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
203 possibly accessed in kptllnd_shutdown. Ptllnd should init
204 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
208 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
212 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
216 Description: the_lnet.ln_finalizing was not set when the current thread is
217 about to complete messages. It only affects multi-threaded
223 Description: Changed the default kqswlnd ntxmsg=512
228 Description: Assertion failure in kernel ptllnd caused by posting passive
229 bulk buffers before connection establishment complete.
234 Description: A race in kernel ptllnd between deleting a peer and posting
235 new communications for it could hang communications -
236 manifesting as "Unexpectedly long timeout" messages.
241 Description: Kernel ptllnd lock ordering issue could hang a node.
246 Description: node crash on socket teardown race
249 Frequency : 'lctl peer_list' issued on a mx net
251 Description: Enable lctl's peer_list for MXLND
254 Frequency : after Ptllnd timeouts and portals congestion
256 Description: Credit overflows
257 Details : This was a bug in ptllnd connection establishment. The fix
258 implements better peer stamps to disambiguate connection
259 establishment and ensure both peers enter the credit flow
260 state machine consistently.
265 Description: kptllnd didn't propagate some network errors up to LNET
266 Details : This bug was spotted while investigating 11394. The fix
267 ensures network errors on sends and bulk transfers are
268 propagated to LNET/lustre correctly.
270 Severity : enhancement
272 Description: Fixed console chatter in case of -ETIMEDOUT.
274 Severity : enhancement
276 Description: Added D_NETTRACE for recording network packet history
277 (initially only for ptllnd). Also a separate userspace
278 ptllnd facility to gather history which should really be
279 covered by D_NETTRACE too, if only CDEBUG recorded history in
285 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
286 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
287 callback can occur before a connection has actually been
288 established. This caused an assertion failure previously.
290 Severity : enhancement
292 Description: Multiple instances for o2iblnd
293 Details : Allow multiple instances of o2iblnd to enable networking over
294 multiple HCAs and routing between them.
298 Description: lnet deadlock in router_checker
299 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
300 into BH locks to eliminate potential deadlock caused by
301 ksocknal_data_ready() preempting code holding these locks.
305 Description: Millions of failed socklnd connection attempts cause a very slow FS
306 Details : added a new route flag ksnr_scheduled to distinguish from
307 ksnr_connecting, so that a peer connection request is only turned
308 down for race concerns when an active connection to the same peer
309 is under progress (instead of just being scheduled).
311 ------------------------------------------------------------------------------
313 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
315 * Support for networks:
316 socklnd - kernels up to 2.6.16
317 qswlnd - Qsnet kernel modules 5.20 and later
318 openiblnd - IbGold 1.8.2
320 viblnd - Voltaire ibhost 3.4.5 and later
321 ciblnd - Topspin 3.2.0
322 iiblnd - Infiniserv 3.3 + PathBits patch
323 gmlnd - GM 2.1.22 and later
324 mxlnd - MX 1.2.1 or later
325 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
328 Severity : major on XT3
330 Description: libcfs overwrites /proc/sys/portals
331 Details : libcfs created a symlink from /proc/sys/portals to
332 /proc/sys/lnet for backwards compatibility. This is no
333 longer required and makes the Cray portals /proc variables
338 Description: OFED FMR API change
339 Details : This changes parameter usage to reflect a change in
340 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
341 that FMR support is only used in experimental versions of the
342 o2iblnd - this change does not affect standard usage at all.
344 Severity : enhancement
346 Description: new ko2iblnd module parameter: ib_mtu
347 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
348 HCAs. You can avoid this problem by setting the MTU to 1024
349 using this module parameter.
351 Severity : enhancement
352 Bugzilla : 11118/11620
353 Description: ptllnd small request message buffer alignment fix
354 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
355 Round up small message size on sends in case this option
356 is not supported. 11620 was a defect in the initial
357 implementation which effectively asserted all peers had to be
358 running the correct protocol version which was fixed by always
359 NAK-ing such requests and handling any misalignments they
364 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
365 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
366 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
368 Severity : enhancement
370 Description: Patchless ZC(zero copy) socklnd
371 Details : New protocol for socklnd, socklnd can support zero copy without
372 kernel patch, it's compatible with old socklnd. Checksum is
373 moved from tunables to modparams.
377 Description: When ksocknal_del_peer() is called upon a peer whose
378 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
379 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
382 Frequency : when ptlrpc is under heavy use and runs out of request buffer
384 Description: In lnet_match_blocked_msg(), md can be used without holding a
388 Frequency : very rarely
390 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
391 If connd connects a route which has been closed by
392 ksocknal_shutdown(), ksocknal_create_routes() may create new
393 routes which hold references on the peer, causing shutdown
394 process to wait for peer to disappear forever.
396 Severity : enhancement
398 Description: Dump XT3 portals traces on kptllnd timeout
399 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
400 dump Cray portals debug traces to a file. The kptllnd module
401 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
402 is the basename of the dump file.
405 Frequency : infrequent
407 Description: kernel ptllnd fix bug in connection re-establishment
408 Details : Kernel ptllnd could produce protocol errors e.g. illegal
409 matchbits and/or violate the credit flow protocol when trying
410 to re-establish a connection with a peer after an error or
413 Severity : enhancement
415 Description: Allow /proc/sys/lnet/debug to be set symbolically
416 Details : Allow debug and subsystem debug values to be read/set by name
417 in addition to numerically, for ease of use.
420 Frequency : only in configurations with LNET routers
422 Description: routes automatically marked down and recovered
423 Details : In configurations with LNET routers if a router fails routers
424 now actively try to recover routes that are down, unless they
425 are marked down by an administrator.
427 ------------------------------------------------------------------------------
429 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
432 Frequency : very rarely, in configurations with LNET routers and TCP
434 Description: incorrect data written to files on OSTs
435 Details : In certain high-load conditions incorrect data may be written
436 to files on the OST when using TCP networks.
438 ------------------------------------------------------------------------------
440 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
442 - rework CDEBUG messages rate-limiting mechanism b=10375
443 - add per-socket tunables for socklnd if the kernel is patched b=10327
445 ------------------------------------------------------------------------------
447 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
449 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
450 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
451 - quiet LNET startup LNI message for liblustre b=10128
452 - Better console error messages if 'ip2nets' can't match an IP address
453 - Fixed overflow/use-before-set bugs in linux-time.h
454 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
455 - LNET teardown failed an assertion about the route table being empty
456 - Fixed a crash in LNetEQPoll(<invalid handle>)
457 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
458 - improve debug message for liblustre/Catamount nodes (b=10116)
460 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
461 * Configuration change for the XT3
462 The PTLLND is now used to run Lustre over Portals on the XT3.
463 The configure option(s) --with-cray-portals are no longer
464 used. Rather --with-portals=<path-to-portals-includes> is
465 used to enable building on the XT3. In addition to enable
466 XT3 specific features the option --enable-cray-xt3 must be
469 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
470 * Portals has been removed, replaced by LNET.
471 LNET is new networking infrastructure for Lustre, it includes a
472 reorganized network configuration mode (see the user
473 documentation for full details) as well as support for routing
474 between different network fabrics. Lustre Networking Devices
475 (LNDS) for the supported network fabrics have also been created
476 for this new infrastructure.
478 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
483 Frequency : rare (large Voltaire clusters only)
485 Description: the default number of reserved transmit descriptors was too low
486 for some large clusters
487 Details : As a workaround, the number was increased. A proper fix includes
490 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
495 Frequency : occasional (large-scale events, cluster reboot, network failure)
497 Description: too many error messages on console obscure actual problem and
498 can slow down/panic server, or cause recovery to fail repeatedly
499 Details : enable rate-limiting of console error messages, and some messages
500 that were console errors now only go to the kernel log
502 Severity : enhancement
504 Description: add /proc/sys/portals/catastrophe entry which will report if
505 that node has previously LBUGged
507 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
509 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
511 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
513 - handle error return code in kranal_check_fma_rx() (5915,6054)
515 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
517 - update vibnal (Voltaire IB NAL)
518 - update gmnal (Myrinet NAL), gmnalid
520 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
522 * Landed portals:b_port_step as follows...
524 - removed CFS_DECL_SPIN*
525 just use 'spinlock_t' and initialise with spin_lock_init()
527 - removed CFS_DECL_MUTEX*
528 just use 'struct semaphore' and initialise with init_mutex()
530 - removed CFS_DECL_RWSEM*
531 just use 'struct rw_semaphore' and initialise with init_rwsem()
533 - renamed cfs_sleep_chan -> cfs_waitq
534 cfs_sleep_link -> cfs_waitlink
536 - fixed race in linux version of arch-independent socknal
537 (the ENOMEM/EAGAIN decision).
539 - Didn't fix problems in Darwin version of arch-independent socknal
540 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
542 - removed libcfs types from non-socknal header files (only some types
543 in the header files had been changed; the .c files hadn't been