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,
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 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
17 * Support for networks:
18 socklnd - any kernel supported by Lustre,
19 qswlnd - Qsnet kernel modules 5.20 and later,
20 openiblnd - IbGold 1.8.2,
21 o2iblnd - OFED 1.1 and 1.2,
22 viblnd - Voltaire ibhost 3.4.5 and later,
23 ciblnd - Topspin 3.2.0,
24 iiblnd - Infiniserv 3.3 + PathBits patch,
25 gmlnd - GM 2.1.22 and later,
26 mxlnd - MX 1.2.1 or later,
27 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
31 Description: TOE Kernel panic by ksocklnd
32 Details : offloaded sockets provide their own implementation of sendpage,
33 can't call tcp_sendpage() directly
37 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
38 Details : races between lnd_shutdown and peer creation prevent
39 lnd_shutdown from finishing.
43 Description: open files rlimit 1024 reached while liblustre testing
44 Details : ulnds/socklnd must close open socket after unsuccessful
47 ------------------------------------------------------------------------------
49 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
51 * Support for networks:
52 socklnd - kernels up to 2.6.16,
53 qswlnd - Qsnet kernel modules 5.20 and later,
54 openiblnd - IbGold 1.8.2,
55 o2iblnd - OFED 1.1 and 1.2
56 viblnd - Voltaire ibhost 3.4.5 and later,
57 ciblnd - Topspin 3.2.0,
58 iiblnd - Infiniserv 3.3 + PathBits patch,
59 gmlnd - GM 2.1.22 and later,
60 mxlnd - MX 1.2.1 or later,
61 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
63 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
65 * Support for networks:
66 socklnd - kernels up to 2.6.16,
67 qswlnd - Qsnet kernel modules 5.20 and later,
68 openiblnd - IbGold 1.8.2,
70 viblnd - Voltaire ibhost 3.4.5 and later,
71 ciblnd - Topspin 3.2.0,
72 iiblnd - Infiniserv 3.3 + PathBits patch,
73 gmlnd - GM 2.1.22 and later,
74 mxlnd - MX 1.2.1 or later,
75 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
79 Description: Initialize cpumask before use
83 Description: ASSERTION failures when upgrading to the patchless zero-copy
85 Details : This bug affects "rolling upgrades", causing an inconsistent
86 protocol version negotiation and subsequent assertion failure
87 during rolling upgrades after the first wave of upgrades.
91 Details : Change "dropped message" CERRORs to D_NETERROR so they are
92 logged instead of creating "console chatter" when a lustre
93 timeout races with normal RPC completion.
96 Details : lnet_clear_peer_table can wait forever if user forgets to
100 Details : libcfs_id2str should check pid against LNET_PID_ANY.
104 Description: added LNET self test
105 Details : landing b_self_test
110 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
112 Details : do_div() macro is used incorrectly.
114 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
118 Description: make panic on lbug configurable
122 Description: Add OFED1.2 support to o2iblnd
123 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
124 are installed (other than kernel's in-tree infiniband), there
125 could be some problem while insmod o2iblnd (mismatch CRC of
127 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
128 this link provides solution:
129 https://bugs.openfabrics.org/show_bug.cgi?id=355
130 if extra Module.symvers is not supported in kernel, we will
131 have to run the script in bug 12316 to update
132 $LINUX/module.symvers before building o2iblnd.
133 More details about this are in bug 12316.
135 ------------------------------------------------------------------------------
137 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
138 * version 1.4.10 / 1.6.0
139 * Support for networks:
140 socklnd - kernels up to 2.6.16,
141 qswlnd - Qsnet kernel modules 5.20 and later,
142 openiblnd - IbGold 1.8.2,
144 viblnd - Voltaire ibhost 3.4.5 and later,
145 ciblnd - Topspin 3.2.0,
146 iiblnd - Infiniserv 3.3 + PathBits patch,
147 gmlnd - GM 2.1.22 and later,
148 mxlnd - MX 1.2.1 or later,
149 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
153 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
154 possibly accessed in kptllnd_shutdown. Ptllnd should init
155 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
159 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
163 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
167 Description: the_lnet.ln_finalizing was not set when the current thread is
168 about to complete messages. It only affects multi-threaded
174 Description: Changed the default kqswlnd ntxmsg=512
179 Description: Assertion failure in kernel ptllnd caused by posting passive
180 bulk buffers before connection establishment complete.
185 Description: A race in kernel ptllnd between deleting a peer and posting
186 new communications for it could hang communications -
187 manifesting as "Unexpectedly long timeout" messages.
192 Description: Kernel ptllnd lock ordering issue could hang a node.
197 Description: node crash on socket teardown race
200 Frequency : 'lctl peer_list' issued on a mx net
202 Description: Enable lctl's peer_list for MXLND
205 Frequency : after Ptllnd timeouts and portals congestion
207 Description: Credit overflows
208 Details : This was a bug in ptllnd connection establishment. The fix
209 implements better peer stamps to disambiguate connection
210 establishment and ensure both peers enter the credit flow
211 state machine consistently.
216 Description: kptllnd didn't propagate some network errors up to LNET
217 Details : This bug was spotted while investigating 11394. The fix
218 ensures network errors on sends and bulk transfers are
219 propagated to LNET/lustre correctly.
221 Severity : enhancement
223 Description: Fixed console chatter in case of -ETIMEDOUT.
225 Severity : enhancement
227 Description: Added D_NETTRACE for recording network packet history
228 (initially only for ptllnd). Also a separate userspace
229 ptllnd facility to gather history which should really be
230 covered by D_NETTRACE too, if only CDEBUG recorded history in
236 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
237 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
238 callback can occur before a connection has actually been
239 established. This caused an assertion failure previously.
241 Severity : enhancement
243 Description: Multiple instances for o2iblnd
244 Details : Allow multiple instances of o2iblnd to enable networking over
245 multiple HCAs and routing between them.
249 Description: lnet deadlock in router_checker
250 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
251 into BH locks to eliminate potential deadlock caused by
252 ksocknal_data_ready() preempting code holding these locks.
256 Description: Millions of failed socklnd connection attempts cause a very slow FS
257 Details : added a new route flag ksnr_scheduled to distinguish from
258 ksnr_connecting, so that a peer connection request is only turned
259 down for race concerns when an active connection to the same peer
260 is under progress (instead of just being scheduled).
262 ------------------------------------------------------------------------------
264 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
266 * Support for networks:
267 socklnd - kernels up to 2.6.16
268 qswlnd - Qsnet kernel modules 5.20 and later
269 openiblnd - IbGold 1.8.2
271 viblnd - Voltaire ibhost 3.4.5 and later
272 ciblnd - Topspin 3.2.0
273 iiblnd - Infiniserv 3.3 + PathBits patch
274 gmlnd - GM 2.1.22 and later
275 mxlnd - MX 1.2.1 or later
276 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
279 Severity : major on XT3
281 Description: libcfs overwrites /proc/sys/portals
282 Details : libcfs created a symlink from /proc/sys/portals to
283 /proc/sys/lnet for backwards compatibility. This is no
284 longer required and makes the Cray portals /proc variables
289 Description: OFED FMR API change
290 Details : This changes parameter usage to reflect a change in
291 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
292 that FMR support is only used in experimental versions of the
293 o2iblnd - this change does not affect standard usage at all.
295 Severity : enhancement
297 Description: new ko2iblnd module parameter: ib_mtu
298 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
299 HCAs. You can avoid this problem by setting the MTU to 1024
300 using this module parameter.
302 Severity : enhancement
303 Bugzilla : 11118/11620
304 Description: ptllnd small request message buffer alignment fix
305 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
306 Round up small message size on sends in case this option
307 is not supported. 11620 was a defect in the initial
308 implementation which effectively asserted all peers had to be
309 running the correct protocol version which was fixed by always
310 NAK-ing such requests and handling any misalignments they
315 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
316 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
317 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
319 Severity : enhancement
321 Description: Patchless ZC(zero copy) socklnd
322 Details : New protocol for socklnd, socklnd can support zero copy without
323 kernel patch, it's compatible with old socklnd. Checksum is
324 moved from tunables to modparams.
328 Description: When ksocknal_del_peer() is called upon a peer whose
329 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
330 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
333 Frequency : when ptlrpc is under heavy use and runs out of request buffer
335 Description: In lnet_match_blocked_msg(), md can be used without holding a
339 Frequency : very rarely
341 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
342 If connd connects a route which has been closed by
343 ksocknal_shutdown(), ksocknal_create_routes() may create new
344 routes which hold references on the peer, causing shutdown
345 process to wait for peer to disappear forever.
347 Severity : enhancement
349 Description: Dump XT3 portals traces on kptllnd timeout
350 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
351 dump Cray portals debug traces to a file. The kptllnd module
352 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
353 is the basename of the dump file.
356 Frequency : infrequent
358 Description: kernel ptllnd fix bug in connection re-establishment
359 Details : Kernel ptllnd could produce protocol errors e.g. illegal
360 matchbits and/or violate the credit flow protocol when trying
361 to re-establish a connection with a peer after an error or
364 Severity : enhancement
366 Description: Allow /proc/sys/lnet/debug to be set symbolically
367 Details : Allow debug and subsystem debug values to be read/set by name
368 in addition to numerically, for ease of use.
371 Frequency : only in configurations with LNET routers
373 Description: routes automatically marked down and recovered
374 Details : In configurations with LNET routers if a router fails routers
375 now actively try to recover routes that are down, unless they
376 are marked down by an administrator.
378 ------------------------------------------------------------------------------
380 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
383 Frequency : very rarely, in configurations with LNET routers and TCP
385 Description: incorrect data written to files on OSTs
386 Details : In certain high-load conditions incorrect data may be written
387 to files on the OST when using TCP networks.
389 ------------------------------------------------------------------------------
391 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
393 - rework CDEBUG messages rate-limiting mechanism b=10375
394 - add per-socket tunables for socklnd if the kernel is patched b=10327
396 ------------------------------------------------------------------------------
398 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
400 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
401 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
402 - quiet LNET startup LNI message for liblustre b=10128
403 - Better console error messages if 'ip2nets' can't match an IP address
404 - Fixed overflow/use-before-set bugs in linux-time.h
405 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
406 - LNET teardown failed an assertion about the route table being empty
407 - Fixed a crash in LNetEQPoll(<invalid handle>)
408 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
409 - improve debug message for liblustre/Catamount nodes (b=10116)
411 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
412 * Configuration change for the XT3
413 The PTLLND is now used to run Lustre over Portals on the XT3.
414 The configure option(s) --with-cray-portals are no longer
415 used. Rather --with-portals=<path-to-portals-includes> is
416 used to enable building on the XT3. In addition to enable
417 XT3 specific features the option --enable-cray-xt3 must be
420 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
421 * Portals has been removed, replaced by LNET.
422 LNET is new networking infrastructure for Lustre, it includes a
423 reorganized network configuration mode (see the user
424 documentation for full details) as well as support for routing
425 between different network fabrics. Lustre Networking Devices
426 (LNDS) for the supported network fabrics have also been created
427 for this new infrastructure.
429 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
434 Frequency : rare (large Voltaire clusters only)
436 Description: the default number of reserved transmit descriptors was too low
437 for some large clusters
438 Details : As a workaround, the number was increased. A proper fix includes
441 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
446 Frequency : occasional (large-scale events, cluster reboot, network failure)
448 Description: too many error messages on console obscure actual problem and
449 can slow down/panic server, or cause recovery to fail repeatedly
450 Details : enable rate-limiting of console error messages, and some messages
451 that were console errors now only go to the kernel log
453 Severity : enhancement
455 Description: add /proc/sys/portals/catastrophe entry which will report if
456 that node has previously LBUGged
458 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
460 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
462 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
464 - handle error return code in kranal_check_fma_rx() (5915,6054)
466 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
468 - update vibnal (Voltaire IB NAL)
469 - update gmnal (Myrinet NAL), gmnalid
471 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
473 * Landed portals:b_port_step as follows...
475 - removed CFS_DECL_SPIN*
476 just use 'spinlock_t' and initialise with spin_lock_init()
478 - removed CFS_DECL_MUTEX*
479 just use 'struct semaphore' and initialise with init_mutex()
481 - removed CFS_DECL_RWSEM*
482 just use 'struct rw_semaphore' and initialise with init_rwsem()
484 - renamed cfs_sleep_chan -> cfs_waitq
485 cfs_sleep_link -> cfs_waitlink
487 - fixed race in linux version of arch-independent socknal
488 (the ENOMEM/EAGAIN decision).
490 - Didn't fix problems in Darwin version of arch-independent socknal
491 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
493 - removed libcfs types from non-socknal header files (only some types
494 in the header files had been changed; the .c files hadn't been