1 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
3 * Support for networks:
4 socklnd - kernels up to 2.6.16
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
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: open files rlimit 1024 reached while liblustre testing
18 Details : ulnds/socklnd must close open socket after unsuccessful
21 ------------------------------------------------------------------------------
23 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
24 * version 1.4.11 / 1.6.1
25 * Support for networks:
26 socklnd - kernels up to 2.6.16
27 qswlnd - Qsnet kernel modules 5.20 and later
28 openiblnd - IbGold 1.8.2
29 o2iblnd - OFED 1.1 and 1.2
30 viblnd - Voltaire ibhost 3.4.5 and later
31 ciblnd - Topspin 3.2.0
32 iiblnd - Infiniserv 3.3 + PathBits patch
33 gmlnd - GM 2.1.22 and later
34 mxlnd - MX 1.2.1 or later
35 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
40 Description: Initialize cpumask before use
44 Description: ASSERTION failures when upgrading to the patchless zero-copy
46 Details : This bug affects "rolling upgrades", causing an inconsistent
47 protocol version negotiation and subsequent assertion failure
48 during rolling upgrades after the first wave of upgrades.
52 Details : Change "dropped message" CERRORs to D_NETERROR so they are
53 logged instead of creating "console chatter" when a lustre
54 timeout races with normal RPC completion.
57 Details : lnet_clear_peer_table can wait forever if user forgets to
61 Details : libcfs_id2str should check pid against LNET_PID_ANY.
65 Description: added LNET self test
66 Details : landing b_self_test
71 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
73 Details : do_div() macro is used incorrectly.
75 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
79 Description: make panic on lbug configurable
83 Description: Add OFED1.2 support to o2iblnd
84 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
85 are installed (other than kernel's in-tree infiniband), there
86 could be some problem while insmod o2iblnd (mismatch CRC of
88 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
89 this link provides solution:
90 https://bugs.openfabrics.org/show_bug.cgi?id=355
91 if extra Module.symvers is not supported in kernel, we will
92 have to run the script in bug 12316 to update
93 $LINUX/module.symvers before building o2iblnd.
94 More details about this are in bug 12316.
96 ------------------------------------------------------------------------------
98 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
99 * version 1.4.10 / 1.6.0
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
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
115 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
116 possibly accessed in kptllnd_shutdown. Ptllnd should init
117 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
121 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
125 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
129 Description: the_lnet.ln_finalizing was not set when the current thread is
130 about to complete messages. It only affects multi-threaded
136 Description: Changed the default kqswlnd ntxmsg=512
141 Description: Assertion failure in kernel ptllnd caused by posting passive
142 bulk buffers before connection establishment complete.
147 Description: A race in kernel ptllnd between deleting a peer and posting
148 new communications for it could hang communications -
149 manifesting as "Unexpectedly long timeout" messages.
154 Description: Kernel ptllnd lock ordering issue could hang a node.
159 Description: node crash on socket teardown race
162 Frequency : 'lctl peer_list' issued on a mx net
164 Description: Enable lctl's peer_list for MXLND
167 Frequency : after Ptllnd timeouts and portals congestion
169 Description: Credit overflows
170 Details : This was a bug in ptllnd connection establishment. The fix
171 implements better peer stamps to disambiguate connection
172 establishment and ensure both peers enter the credit flow
173 state machine consistently.
178 Description: kptllnd didn't propagate some network errors up to LNET
179 Details : This bug was spotted while investigating 11394. The fix
180 ensures network errors on sends and bulk transfers are
181 propagated to LNET/lustre correctly.
183 Severity : enhancement
185 Description: Fixed console chatter in case of -ETIMEDOUT.
187 Severity : enhancement
189 Description: Added D_NETTRACE for recording network packet history
190 (initially only for ptllnd). Also a separate userspace
191 ptllnd facility to gather history which should really be
192 covered by D_NETTRACE too, if only CDEBUG recorded history in
198 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
199 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
200 callback can occur before a connection has actually been
201 established. This caused an assertion failure previously.
203 Severity : enhancement
205 Description: Multiple instances for o2iblnd
206 Details : Allow multiple instances of o2iblnd to enable networking over
207 multiple HCAs and routing between them.
211 Description: lnet deadlock in router_checker
212 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
213 into BH locks to eliminate potential deadlock caused by
214 ksocknal_data_ready() preempting code holding these locks.
218 Description: Millions of failed socklnd connection attempts cause a very slow FS
219 Details : added a new route flag ksnr_scheduled to distinguish from
220 ksnr_connecting, so that a peer connection request is only turned
221 down for race concerns when an active connection to the same peer
222 is under progress (instead of just being scheduled).
224 ------------------------------------------------------------------------------
226 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
228 * Support for networks:
229 socklnd - kernels up to 2.6.16
230 qswlnd - Qsnet kernel modules 5.20 and later
231 openiblnd - IbGold 1.8.2
233 viblnd - Voltaire ibhost 3.4.5 and later
234 ciblnd - Topspin 3.2.0
235 iiblnd - Infiniserv 3.3 + PathBits patch
236 gmlnd - GM 2.1.22 and later
237 mxlnd - MX 1.2.1 or later
238 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
241 Severity : major on XT3
243 Description: libcfs overwrites /proc/sys/portals
244 Details : libcfs created a symlink from /proc/sys/portals to
245 /proc/sys/lnet for backwards compatibility. This is no
246 longer required and makes the Cray portals /proc variables
251 Description: OFED FMR API change
252 Details : This changes parameter usage to reflect a change in
253 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
254 that FMR support is only used in experimental versions of the
255 o2iblnd - this change does not affect standard usage at all.
257 Severity : enhancement
259 Description: new ko2iblnd module parameter: ib_mtu
260 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
261 HCAs. You can avoid this problem by setting the MTU to 1024
262 using this module parameter.
264 Severity : enhancement
265 Bugzilla : 11118/11620
266 Description: ptllnd small request message buffer alignment fix
267 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
268 Round up small message size on sends in case this option
269 is not supported. 11620 was a defect in the initial
270 implementation which effectively asserted all peers had to be
271 running the correct protocol version which was fixed by always
272 NAK-ing such requests and handling any misalignments they
277 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
278 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
279 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
281 Severity : enhancement
283 Description: Patchless ZC(zero copy) socklnd
284 Details : New protocol for socklnd, socklnd can support zero copy without
285 kernel patch, it's compatible with old socklnd. Checksum is
286 moved from tunables to modparams.
290 Description: When ksocknal_del_peer() is called upon a peer whose
291 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
292 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
295 Frequency : when ptlrpc is under heavy use and runs out of request buffer
297 Description: In lnet_match_blocked_msg(), md can be used without holding a
301 Frequency : very rarely
303 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
304 If connd connects a route which has been closed by
305 ksocknal_shutdown(), ksocknal_create_routes() may create new
306 routes which hold references on the peer, causing shutdown
307 process to wait for peer to disappear forever.
309 Severity : enhancement
311 Description: Dump XT3 portals traces on kptllnd timeout
312 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
313 dump Cray portals debug traces to a file. The kptllnd module
314 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
315 is the basename of the dump file.
318 Frequency : infrequent
320 Description: kernel ptllnd fix bug in connection re-establishment
321 Details : Kernel ptllnd could produce protocol errors e.g. illegal
322 matchbits and/or violate the credit flow protocol when trying
323 to re-establish a connection with a peer after an error or
326 Severity : enhancement
328 Description: Allow /proc/sys/lnet/debug to be set symbolically
329 Details : Allow debug and subsystem debug values to be read/set by name
330 in addition to numerically, for ease of use.
333 Frequency : only in configurations with LNET routers
335 Description: routes automatically marked down and recovered
336 Details : In configurations with LNET routers if a router fails routers
337 now actively try to recover routes that are down, unless they
338 are marked down by an administrator.
340 ------------------------------------------------------------------------------
342 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
345 Frequency : very rarely, in configurations with LNET routers and TCP
347 Description: incorrect data written to files on OSTs
348 Details : In certain high-load conditions incorrect data may be written
349 to files on the OST when using TCP networks.
351 ------------------------------------------------------------------------------
353 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
355 - rework CDEBUG messages rate-limiting mechanism b=10375
356 - add per-socket tunables for socklnd if the kernel is patched b=10327
358 ------------------------------------------------------------------------------
360 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
362 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
363 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
364 - quiet LNET startup LNI message for liblustre b=10128
365 - Better console error messages if 'ip2nets' can't match an IP address
366 - Fixed overflow/use-before-set bugs in linux-time.h
367 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
368 - LNET teardown failed an assertion about the route table being empty
369 - Fixed a crash in LNetEQPoll(<invalid handle>)
370 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
371 - improve debug message for liblustre/Catamount nodes (b=10116)
373 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
374 * Configuration change for the XT3
375 The PTLLND is now used to run Lustre over Portals on the XT3.
376 The configure option(s) --with-cray-portals are no longer
377 used. Rather --with-portals=<path-to-portals-includes> is
378 used to enable building on the XT3. In addition to enable
379 XT3 specific features the option --enable-cray-xt3 must be
382 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
383 * Portals has been removed, replaced by LNET.
384 LNET is new networking infrastructure for Lustre, it includes a
385 reorganized network configuration mode (see the user
386 documentation for full details) as well as support for routing
387 between different network fabrics. Lustre Networking Devices
388 (LNDS) for the supported network fabrics have also been created
389 for this new infrastructure.
391 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
396 Frequency : rare (large Voltaire clusters only)
398 Description: the default number of reserved transmit descriptors was too low
399 for some large clusters
400 Details : As a workaround, the number was increased. A proper fix includes
403 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
408 Frequency : occasional (large-scale events, cluster reboot, network failure)
410 Description: too many error messages on console obscure actual problem and
411 can slow down/panic server, or cause recovery to fail repeatedly
412 Details : enable rate-limiting of console error messages, and some messages
413 that were console errors now only go to the kernel log
415 Severity : enhancement
417 Description: add /proc/sys/portals/catastrophe entry which will report if
418 that node has previously LBUGged
420 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
422 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
424 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
426 - handle error return code in kranal_check_fma_rx() (5915,6054)
428 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
430 - update vibnal (Voltaire IB NAL)
431 - update gmnal (Myrinet NAL), gmnalid
433 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
435 * Landed portals:b_port_step as follows...
437 - removed CFS_DECL_SPIN*
438 just use 'spinlock_t' and initialise with spin_lock_init()
440 - removed CFS_DECL_MUTEX*
441 just use 'struct semaphore' and initialise with init_mutex()
443 - removed CFS_DECL_RWSEM*
444 just use 'struct rw_semaphore' and initialise with init_rwsem()
446 - renamed cfs_sleep_chan -> cfs_waitq
447 cfs_sleep_link -> cfs_waitlink
449 - fixed race in linux version of arch-independent socknal
450 (the ENOMEM/EAGAIN decision).
452 - Didn't fix problems in Darwin version of arch-independent socknal
453 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
455 - removed libcfs types from non-socknal header files (only some types
456 in the header files had been changed; the .c files hadn't been