1 2007-09-24 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
17 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
18 Details : races between lnd_shutdown and peer creation prevent
19 lnd_shutdown from finishing.
23 Description: open files rlimit 1024 reached while liblustre testing
24 Details : ulnds/socklnd must close open socket after unsuccessful
27 ------------------------------------------------------------------------------
29 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
31 * Support for networks:
32 socklnd - kernels up to 2.6.16,
33 qswlnd - Qsnet kernel modules 5.20 and later,
34 openiblnd - IbGold 1.8.2,
35 o2iblnd - OFED 1.1 and 1.2
36 viblnd - Voltaire ibhost 3.4.5 and later,
37 ciblnd - Topspin 3.2.0,
38 iiblnd - Infiniserv 3.3 + PathBits patch,
39 gmlnd - GM 2.1.22 and later,
40 mxlnd - MX 1.2.1 or later,
41 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
43 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
45 * Support for networks:
46 socklnd - kernels up to 2.6.16,
47 qswlnd - Qsnet kernel modules 5.20 and later,
48 openiblnd - IbGold 1.8.2,
50 viblnd - Voltaire ibhost 3.4.5 and later,
51 ciblnd - Topspin 3.2.0,
52 iiblnd - Infiniserv 3.3 + PathBits patch,
53 gmlnd - GM 2.1.22 and later,
54 mxlnd - MX 1.2.1 or later,
55 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
59 Description: Initialize cpumask before use
63 Description: ASSERTION failures when upgrading to the patchless zero-copy
65 Details : This bug affects "rolling upgrades", causing an inconsistent
66 protocol version negotiation and subsequent assertion failure
67 during rolling upgrades after the first wave of upgrades.
71 Details : Change "dropped message" CERRORs to D_NETERROR so they are
72 logged instead of creating "console chatter" when a lustre
73 timeout races with normal RPC completion.
76 Details : lnet_clear_peer_table can wait forever if user forgets to
80 Details : libcfs_id2str should check pid against LNET_PID_ANY.
84 Description: added LNET self test
85 Details : landing b_self_test
90 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
92 Details : do_div() macro is used incorrectly.
94 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
98 Description: make panic on lbug configurable
102 Description: Add OFED1.2 support to o2iblnd
103 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
104 are installed (other than kernel's in-tree infiniband), there
105 could be some problem while insmod o2iblnd (mismatch CRC of
107 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
108 this link provides solution:
109 https://bugs.openfabrics.org/show_bug.cgi?id=355
110 if extra Module.symvers is not supported in kernel, we will
111 have to run the script in bug 12316 to update
112 $LINUX/module.symvers before building o2iblnd.
113 More details about this are in bug 12316.
115 ------------------------------------------------------------------------------
117 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
118 * version 1.4.10 / 1.6.0
119 * Support for networks:
120 socklnd - kernels up to 2.6.16,
121 qswlnd - Qsnet kernel modules 5.20 and later,
122 openiblnd - IbGold 1.8.2,
124 viblnd - Voltaire ibhost 3.4.5 and later,
125 ciblnd - Topspin 3.2.0,
126 iiblnd - Infiniserv 3.3 + PathBits patch,
127 gmlnd - GM 2.1.22 and later,
128 mxlnd - MX 1.2.1 or later,
129 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
133 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
134 possibly accessed in kptllnd_shutdown. Ptllnd should init
135 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
139 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
143 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
147 Description: the_lnet.ln_finalizing was not set when the current thread is
148 about to complete messages. It only affects multi-threaded
154 Description: Changed the default kqswlnd ntxmsg=512
159 Description: Assertion failure in kernel ptllnd caused by posting passive
160 bulk buffers before connection establishment complete.
165 Description: A race in kernel ptllnd between deleting a peer and posting
166 new communications for it could hang communications -
167 manifesting as "Unexpectedly long timeout" messages.
172 Description: Kernel ptllnd lock ordering issue could hang a node.
177 Description: node crash on socket teardown race
180 Frequency : 'lctl peer_list' issued on a mx net
182 Description: Enable lctl's peer_list for MXLND
185 Frequency : after Ptllnd timeouts and portals congestion
187 Description: Credit overflows
188 Details : This was a bug in ptllnd connection establishment. The fix
189 implements better peer stamps to disambiguate connection
190 establishment and ensure both peers enter the credit flow
191 state machine consistently.
196 Description: kptllnd didn't propagate some network errors up to LNET
197 Details : This bug was spotted while investigating 11394. The fix
198 ensures network errors on sends and bulk transfers are
199 propagated to LNET/lustre correctly.
201 Severity : enhancement
203 Description: Fixed console chatter in case of -ETIMEDOUT.
205 Severity : enhancement
207 Description: Added D_NETTRACE for recording network packet history
208 (initially only for ptllnd). Also a separate userspace
209 ptllnd facility to gather history which should really be
210 covered by D_NETTRACE too, if only CDEBUG recorded history in
216 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
217 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
218 callback can occur before a connection has actually been
219 established. This caused an assertion failure previously.
221 Severity : enhancement
223 Description: Multiple instances for o2iblnd
224 Details : Allow multiple instances of o2iblnd to enable networking over
225 multiple HCAs and routing between them.
229 Description: lnet deadlock in router_checker
230 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
231 into BH locks to eliminate potential deadlock caused by
232 ksocknal_data_ready() preempting code holding these locks.
236 Description: Millions of failed socklnd connection attempts cause a very slow FS
237 Details : added a new route flag ksnr_scheduled to distinguish from
238 ksnr_connecting, so that a peer connection request is only turned
239 down for race concerns when an active connection to the same peer
240 is under progress (instead of just being scheduled).
242 ------------------------------------------------------------------------------
244 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
246 * Support for networks:
247 socklnd - kernels up to 2.6.16
248 qswlnd - Qsnet kernel modules 5.20 and later
249 openiblnd - IbGold 1.8.2
251 viblnd - Voltaire ibhost 3.4.5 and later
252 ciblnd - Topspin 3.2.0
253 iiblnd - Infiniserv 3.3 + PathBits patch
254 gmlnd - GM 2.1.22 and later
255 mxlnd - MX 1.2.1 or later
256 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
259 Severity : major on XT3
261 Description: libcfs overwrites /proc/sys/portals
262 Details : libcfs created a symlink from /proc/sys/portals to
263 /proc/sys/lnet for backwards compatibility. This is no
264 longer required and makes the Cray portals /proc variables
269 Description: OFED FMR API change
270 Details : This changes parameter usage to reflect a change in
271 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
272 that FMR support is only used in experimental versions of the
273 o2iblnd - this change does not affect standard usage at all.
275 Severity : enhancement
277 Description: new ko2iblnd module parameter: ib_mtu
278 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
279 HCAs. You can avoid this problem by setting the MTU to 1024
280 using this module parameter.
282 Severity : enhancement
283 Bugzilla : 11118/11620
284 Description: ptllnd small request message buffer alignment fix
285 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
286 Round up small message size on sends in case this option
287 is not supported. 11620 was a defect in the initial
288 implementation which effectively asserted all peers had to be
289 running the correct protocol version which was fixed by always
290 NAK-ing such requests and handling any misalignments they
295 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
296 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
297 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
299 Severity : enhancement
301 Description: Patchless ZC(zero copy) socklnd
302 Details : New protocol for socklnd, socklnd can support zero copy without
303 kernel patch, it's compatible with old socklnd. Checksum is
304 moved from tunables to modparams.
308 Description: When ksocknal_del_peer() is called upon a peer whose
309 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
310 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
313 Frequency : when ptlrpc is under heavy use and runs out of request buffer
315 Description: In lnet_match_blocked_msg(), md can be used without holding a
319 Frequency : very rarely
321 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
322 If connd connects a route which has been closed by
323 ksocknal_shutdown(), ksocknal_create_routes() may create new
324 routes which hold references on the peer, causing shutdown
325 process to wait for peer to disappear forever.
327 Severity : enhancement
329 Description: Dump XT3 portals traces on kptllnd timeout
330 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
331 dump Cray portals debug traces to a file. The kptllnd module
332 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
333 is the basename of the dump file.
336 Frequency : infrequent
338 Description: kernel ptllnd fix bug in connection re-establishment
339 Details : Kernel ptllnd could produce protocol errors e.g. illegal
340 matchbits and/or violate the credit flow protocol when trying
341 to re-establish a connection with a peer after an error or
344 Severity : enhancement
346 Description: Allow /proc/sys/lnet/debug to be set symbolically
347 Details : Allow debug and subsystem debug values to be read/set by name
348 in addition to numerically, for ease of use.
351 Frequency : only in configurations with LNET routers
353 Description: routes automatically marked down and recovered
354 Details : In configurations with LNET routers if a router fails routers
355 now actively try to recover routes that are down, unless they
356 are marked down by an administrator.
358 ------------------------------------------------------------------------------
360 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
363 Frequency : very rarely, in configurations with LNET routers and TCP
365 Description: incorrect data written to files on OSTs
366 Details : In certain high-load conditions incorrect data may be written
367 to files on the OST when using TCP networks.
369 ------------------------------------------------------------------------------
371 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
373 - rework CDEBUG messages rate-limiting mechanism b=10375
374 - add per-socket tunables for socklnd if the kernel is patched b=10327
376 ------------------------------------------------------------------------------
378 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
380 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
381 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
382 - quiet LNET startup LNI message for liblustre b=10128
383 - Better console error messages if 'ip2nets' can't match an IP address
384 - Fixed overflow/use-before-set bugs in linux-time.h
385 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
386 - LNET teardown failed an assertion about the route table being empty
387 - Fixed a crash in LNetEQPoll(<invalid handle>)
388 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
389 - improve debug message for liblustre/Catamount nodes (b=10116)
391 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
392 * Configuration change for the XT3
393 The PTLLND is now used to run Lustre over Portals on the XT3.
394 The configure option(s) --with-cray-portals are no longer
395 used. Rather --with-portals=<path-to-portals-includes> is
396 used to enable building on the XT3. In addition to enable
397 XT3 specific features the option --enable-cray-xt3 must be
400 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
401 * Portals has been removed, replaced by LNET.
402 LNET is new networking infrastructure for Lustre, it includes a
403 reorganized network configuration mode (see the user
404 documentation for full details) as well as support for routing
405 between different network fabrics. Lustre Networking Devices
406 (LNDS) for the supported network fabrics have also been created
407 for this new infrastructure.
409 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
414 Frequency : rare (large Voltaire clusters only)
416 Description: the default number of reserved transmit descriptors was too low
417 for some large clusters
418 Details : As a workaround, the number was increased. A proper fix includes
421 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
426 Frequency : occasional (large-scale events, cluster reboot, network failure)
428 Description: too many error messages on console obscure actual problem and
429 can slow down/panic server, or cause recovery to fail repeatedly
430 Details : enable rate-limiting of console error messages, and some messages
431 that were console errors now only go to the kernel log
433 Severity : enhancement
435 Description: add /proc/sys/portals/catastrophe entry which will report if
436 that node has previously LBUGged
438 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
440 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
442 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
444 - handle error return code in kranal_check_fma_rx() (5915,6054)
446 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
448 - update vibnal (Voltaire IB NAL)
449 - update gmnal (Myrinet NAL), gmnalid
451 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
453 * Landed portals:b_port_step as follows...
455 - removed CFS_DECL_SPIN*
456 just use 'spinlock_t' and initialise with spin_lock_init()
458 - removed CFS_DECL_MUTEX*
459 just use 'struct semaphore' and initialise with init_mutex()
461 - removed CFS_DECL_RWSEM*
462 just use 'struct rw_semaphore' and initialise with init_rwsem()
464 - renamed cfs_sleep_chan -> cfs_waitq
465 cfs_sleep_link -> cfs_waitlink
467 - fixed race in linux version of arch-independent socknal
468 (the ENOMEM/EAGAIN decision).
470 - Didn't fix problems in Darwin version of arch-independent socknal
471 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
473 - removed libcfs types from non-socknal header files (only some types
474 in the header files had been changed; the .c files hadn't been