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 --------------------------------------------------------------------------------
17 2007-09-27 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,
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: /proc/sys/lnet has non-sysctl entries
34 Details : Updating dump_kernel/daemon_file/debug_mb to use sysctl variables
38 Description: TOE Kernel panic by ksocklnd
39 Details : offloaded sockets provide their own implementation of sendpage,
40 can't call tcp_sendpage() directly
44 Description: kibnal_shutdown() doesn't finish; lconf --cleanup hangs
45 Details : races between lnd_shutdown and peer creation prevent
46 lnd_shutdown from finishing.
50 Description: open files rlimit 1024 reached while liblustre testing
51 Details : ulnds/socklnd must close open socket after unsuccessful
54 ------------------------------------------------------------------------------
56 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
58 * Support for networks:
59 socklnd - kernels up to 2.6.16,
60 qswlnd - Qsnet kernel modules 5.20 and later,
61 openiblnd - IbGold 1.8.2,
62 o2iblnd - OFED 1.1 and 1.2
63 viblnd - Voltaire ibhost 3.4.5 and later,
64 ciblnd - Topspin 3.2.0,
65 iiblnd - Infiniserv 3.3 + PathBits patch,
66 gmlnd - GM 2.1.22 and later,
67 mxlnd - MX 1.2.1 or later,
68 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
70 2007-06-21 Cluster File Systems, Inc. <info@clusterfs.com>
72 * Support for networks:
73 socklnd - kernels up to 2.6.16,
74 qswlnd - Qsnet kernel modules 5.20 and later,
75 openiblnd - IbGold 1.8.2,
77 viblnd - Voltaire ibhost 3.4.5 and later,
78 ciblnd - Topspin 3.2.0,
79 iiblnd - Infiniserv 3.3 + PathBits patch,
80 gmlnd - GM 2.1.22 and later,
81 mxlnd - MX 1.2.1 or later,
82 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
86 Description: Initialize cpumask before use
90 Description: ASSERTION failures when upgrading to the patchless zero-copy
92 Details : This bug affects "rolling upgrades", causing an inconsistent
93 protocol version negotiation and subsequent assertion failure
94 during rolling upgrades after the first wave of upgrades.
98 Details : Change "dropped message" CERRORs to D_NETERROR so they are
99 logged instead of creating "console chatter" when a lustre
100 timeout races with normal RPC completion.
103 Details : lnet_clear_peer_table can wait forever if user forgets to
107 Details : libcfs_id2str should check pid against LNET_PID_ANY.
111 Description: added LNET self test
112 Details : landing b_self_test
117 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
119 Details : do_div() macro is used incorrectly.
121 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
125 Description: make panic on lbug configurable
129 Description: Add OFED1.2 support to o2iblnd
130 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
131 are installed (other than kernel's in-tree infiniband), there
132 could be some problem while insmod o2iblnd (mismatch CRC of
134 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
135 this link provides solution:
136 https://bugs.openfabrics.org/show_bug.cgi?id=355
137 if extra Module.symvers is not supported in kernel, we will
138 have to run the script in bug 12316 to update
139 $LINUX/module.symvers before building o2iblnd.
140 More details about this are in bug 12316.
142 ------------------------------------------------------------------------------
144 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
145 * version 1.4.10 / 1.6.0
146 * Support for networks:
147 socklnd - kernels up to 2.6.16,
148 qswlnd - Qsnet kernel modules 5.20 and later,
149 openiblnd - IbGold 1.8.2,
151 viblnd - Voltaire ibhost 3.4.5 and later,
152 ciblnd - Topspin 3.2.0,
153 iiblnd - Infiniserv 3.3 + PathBits patch,
154 gmlnd - GM 2.1.22 and later,
155 mxlnd - MX 1.2.1 or later,
156 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
160 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
161 possibly accessed in kptllnd_shutdown. Ptllnd should init
162 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
166 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
170 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
174 Description: the_lnet.ln_finalizing was not set when the current thread is
175 about to complete messages. It only affects multi-threaded
181 Description: Changed the default kqswlnd ntxmsg=512
186 Description: Assertion failure in kernel ptllnd caused by posting passive
187 bulk buffers before connection establishment complete.
192 Description: A race in kernel ptllnd between deleting a peer and posting
193 new communications for it could hang communications -
194 manifesting as "Unexpectedly long timeout" messages.
199 Description: Kernel ptllnd lock ordering issue could hang a node.
204 Description: node crash on socket teardown race
207 Frequency : 'lctl peer_list' issued on a mx net
209 Description: Enable lctl's peer_list for MXLND
212 Frequency : after Ptllnd timeouts and portals congestion
214 Description: Credit overflows
215 Details : This was a bug in ptllnd connection establishment. The fix
216 implements better peer stamps to disambiguate connection
217 establishment and ensure both peers enter the credit flow
218 state machine consistently.
223 Description: kptllnd didn't propagate some network errors up to LNET
224 Details : This bug was spotted while investigating 11394. The fix
225 ensures network errors on sends and bulk transfers are
226 propagated to LNET/lustre correctly.
228 Severity : enhancement
230 Description: Fixed console chatter in case of -ETIMEDOUT.
232 Severity : enhancement
234 Description: Added D_NETTRACE for recording network packet history
235 (initially only for ptllnd). Also a separate userspace
236 ptllnd facility to gather history which should really be
237 covered by D_NETTRACE too, if only CDEBUG recorded history in
243 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
244 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
245 callback can occur before a connection has actually been
246 established. This caused an assertion failure previously.
248 Severity : enhancement
250 Description: Multiple instances for o2iblnd
251 Details : Allow multiple instances of o2iblnd to enable networking over
252 multiple HCAs and routing between them.
256 Description: lnet deadlock in router_checker
257 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
258 into BH locks to eliminate potential deadlock caused by
259 ksocknal_data_ready() preempting code holding these locks.
263 Description: Millions of failed socklnd connection attempts cause a very slow FS
264 Details : added a new route flag ksnr_scheduled to distinguish from
265 ksnr_connecting, so that a peer connection request is only turned
266 down for race concerns when an active connection to the same peer
267 is under progress (instead of just being scheduled).
269 ------------------------------------------------------------------------------
271 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
273 * Support for networks:
274 socklnd - kernels up to 2.6.16
275 qswlnd - Qsnet kernel modules 5.20 and later
276 openiblnd - IbGold 1.8.2
278 viblnd - Voltaire ibhost 3.4.5 and later
279 ciblnd - Topspin 3.2.0
280 iiblnd - Infiniserv 3.3 + PathBits patch
281 gmlnd - GM 2.1.22 and later
282 mxlnd - MX 1.2.1 or later
283 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
286 Severity : major on XT3
288 Description: libcfs overwrites /proc/sys/portals
289 Details : libcfs created a symlink from /proc/sys/portals to
290 /proc/sys/lnet for backwards compatibility. This is no
291 longer required and makes the Cray portals /proc variables
296 Description: OFED FMR API change
297 Details : This changes parameter usage to reflect a change in
298 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
299 that FMR support is only used in experimental versions of the
300 o2iblnd - this change does not affect standard usage at all.
302 Severity : enhancement
304 Description: new ko2iblnd module parameter: ib_mtu
305 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
306 HCAs. You can avoid this problem by setting the MTU to 1024
307 using this module parameter.
309 Severity : enhancement
310 Bugzilla : 11118/11620
311 Description: ptllnd small request message buffer alignment fix
312 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
313 Round up small message size on sends in case this option
314 is not supported. 11620 was a defect in the initial
315 implementation which effectively asserted all peers had to be
316 running the correct protocol version which was fixed by always
317 NAK-ing such requests and handling any misalignments they
322 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
323 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
324 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
326 Severity : enhancement
328 Description: Patchless ZC(zero copy) socklnd
329 Details : New protocol for socklnd, socklnd can support zero copy without
330 kernel patch, it's compatible with old socklnd. Checksum is
331 moved from tunables to modparams.
335 Description: When ksocknal_del_peer() is called upon a peer whose
336 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
337 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
340 Frequency : when ptlrpc is under heavy use and runs out of request buffer
342 Description: In lnet_match_blocked_msg(), md can be used without holding a
346 Frequency : very rarely
348 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
349 If connd connects a route which has been closed by
350 ksocknal_shutdown(), ksocknal_create_routes() may create new
351 routes which hold references on the peer, causing shutdown
352 process to wait for peer to disappear forever.
354 Severity : enhancement
356 Description: Dump XT3 portals traces on kptllnd timeout
357 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
358 dump Cray portals debug traces to a file. The kptllnd module
359 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
360 is the basename of the dump file.
363 Frequency : infrequent
365 Description: kernel ptllnd fix bug in connection re-establishment
366 Details : Kernel ptllnd could produce protocol errors e.g. illegal
367 matchbits and/or violate the credit flow protocol when trying
368 to re-establish a connection with a peer after an error or
371 Severity : enhancement
373 Description: Allow /proc/sys/lnet/debug to be set symbolically
374 Details : Allow debug and subsystem debug values to be read/set by name
375 in addition to numerically, for ease of use.
378 Frequency : only in configurations with LNET routers
380 Description: routes automatically marked down and recovered
381 Details : In configurations with LNET routers if a router fails routers
382 now actively try to recover routes that are down, unless they
383 are marked down by an administrator.
385 ------------------------------------------------------------------------------
387 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
390 Frequency : very rarely, in configurations with LNET routers and TCP
392 Description: incorrect data written to files on OSTs
393 Details : In certain high-load conditions incorrect data may be written
394 to files on the OST when using TCP networks.
396 ------------------------------------------------------------------------------
398 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
400 - rework CDEBUG messages rate-limiting mechanism b=10375
401 - add per-socket tunables for socklnd if the kernel is patched b=10327
403 ------------------------------------------------------------------------------
405 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
407 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
408 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
409 - quiet LNET startup LNI message for liblustre b=10128
410 - Better console error messages if 'ip2nets' can't match an IP address
411 - Fixed overflow/use-before-set bugs in linux-time.h
412 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
413 - LNET teardown failed an assertion about the route table being empty
414 - Fixed a crash in LNetEQPoll(<invalid handle>)
415 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
416 - improve debug message for liblustre/Catamount nodes (b=10116)
418 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
419 * Configuration change for the XT3
420 The PTLLND is now used to run Lustre over Portals on the XT3.
421 The configure option(s) --with-cray-portals are no longer
422 used. Rather --with-portals=<path-to-portals-includes> is
423 used to enable building on the XT3. In addition to enable
424 XT3 specific features the option --enable-cray-xt3 must be
427 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
428 * Portals has been removed, replaced by LNET.
429 LNET is new networking infrastructure for Lustre, it includes a
430 reorganized network configuration mode (see the user
431 documentation for full details) as well as support for routing
432 between different network fabrics. Lustre Networking Devices
433 (LNDS) for the supported network fabrics have also been created
434 for this new infrastructure.
436 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
441 Frequency : rare (large Voltaire clusters only)
443 Description: the default number of reserved transmit descriptors was too low
444 for some large clusters
445 Details : As a workaround, the number was increased. A proper fix includes
448 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
453 Frequency : occasional (large-scale events, cluster reboot, network failure)
455 Description: too many error messages on console obscure actual problem and
456 can slow down/panic server, or cause recovery to fail repeatedly
457 Details : enable rate-limiting of console error messages, and some messages
458 that were console errors now only go to the kernel log
460 Severity : enhancement
462 Description: add /proc/sys/portals/catastrophe entry which will report if
463 that node has previously LBUGged
465 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
467 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
469 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
471 - handle error return code in kranal_check_fma_rx() (5915,6054)
473 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
475 - update vibnal (Voltaire IB NAL)
476 - update gmnal (Myrinet NAL), gmnalid
478 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
480 * Landed portals:b_port_step as follows...
482 - removed CFS_DECL_SPIN*
483 just use 'spinlock_t' and initialise with spin_lock_init()
485 - removed CFS_DECL_MUTEX*
486 just use 'struct semaphore' and initialise with init_mutex()
488 - removed CFS_DECL_RWSEM*
489 just use 'struct rw_semaphore' and initialise with init_rwsem()
491 - renamed cfs_sleep_chan -> cfs_waitq
492 cfs_sleep_link -> cfs_waitlink
494 - fixed race in linux version of arch-independent socknal
495 (the ENOMEM/EAGAIN decision).
497 - Didn't fix problems in Darwin version of arch-independent socknal
498 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
500 - removed libcfs types from non-socknal header files (only some types
501 in the header files had been changed; the .c files hadn't been