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,
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: 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
39 Description: Initialize cpumask before use
43 Description: ASSERTION failures when upgrading to the patchless zero-copy
45 Details : This bug affects "rolling upgrades", causing an inconsistent
46 protocol version negotiation and subsequent assertion failure
47 during rolling upgrades after the first wave of upgrades.
51 Details : Change "dropped message" CERRORs to D_NETERROR so they are
52 logged instead of creating "console chatter" when a lustre
53 timeout races with normal RPC completion.
56 Details : lnet_clear_peer_table can wait forever if user forgets to
60 Details : libcfs_id2str should check pid against LNET_PID_ANY.
64 Description: added LNET self test
65 Details : landing b_self_test
70 Description: cfs_duration_{u,n}sec() wrongly calculate nanosecond part of
72 Details : do_div() macro is used incorrectly.
74 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
78 Description: make panic on lbug configurable
82 Description: Add OFED1.2 support to o2iblnd
83 Details : o2iblnd depends on OFED's modules, if out-tree OFED's modules
84 are installed (other than kernel's in-tree infiniband), there
85 could be some problem while insmod o2iblnd (mismatch CRC of
87 If extra Module.symvers is supported in kernel (i.e, 2.6.17),
88 this link provides solution:
89 https://bugs.openfabrics.org/show_bug.cgi?id=355
90 if extra Module.symvers is not supported in kernel, we will
91 have to run the script in bug 12316 to update
92 $LINUX/module.symvers before building o2iblnd.
93 More details about this are in bug 12316.
95 ------------------------------------------------------------------------------
97 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
98 * version 1.4.10 / 1.6.0
99 * Support for networks:
100 socklnd - kernels up to 2.6.16,
101 qswlnd - Qsnet kernel modules 5.20 and later,
102 openiblnd - IbGold 1.8.2,
104 viblnd - Voltaire ibhost 3.4.5 and later,
105 ciblnd - Topspin 3.2.0,
106 iiblnd - Infiniserv 3.3 + PathBits patch,
107 gmlnd - GM 2.1.22 and later,
108 mxlnd - MX 1.2.1 or later,
109 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
113 Description: Ptllnd didn't init kptllnd_data.kptl_idle_txs before it could be
114 possibly accessed in kptllnd_shutdown. Ptllnd should init
115 kptllnd_data.kptl_ptlid2str_lock before calling kptllnd_ptlid2str.
119 Description: gmlnd ignored some transmit errors when finalizing lnet messages.
123 Description: ptllnd logs a piece of incorrect debug info in kptllnd_peer_handle_hello.
127 Description: the_lnet.ln_finalizing was not set when the current thread is
128 about to complete messages. It only affects multi-threaded
134 Description: Changed the default kqswlnd ntxmsg=512
139 Description: Assertion failure in kernel ptllnd caused by posting passive
140 bulk buffers before connection establishment complete.
145 Description: A race in kernel ptllnd between deleting a peer and posting
146 new communications for it could hang communications -
147 manifesting as "Unexpectedly long timeout" messages.
152 Description: Kernel ptllnd lock ordering issue could hang a node.
157 Description: node crash on socket teardown race
160 Frequency : 'lctl peer_list' issued on a mx net
162 Description: Enable lctl's peer_list for MXLND
165 Frequency : after Ptllnd timeouts and portals congestion
167 Description: Credit overflows
168 Details : This was a bug in ptllnd connection establishment. The fix
169 implements better peer stamps to disambiguate connection
170 establishment and ensure both peers enter the credit flow
171 state machine consistently.
176 Description: kptllnd didn't propagate some network errors up to LNET
177 Details : This bug was spotted while investigating 11394. The fix
178 ensures network errors on sends and bulk transfers are
179 propagated to LNET/lustre correctly.
181 Severity : enhancement
183 Description: Fixed console chatter in case of -ETIMEDOUT.
185 Severity : enhancement
187 Description: Added D_NETTRACE for recording network packet history
188 (initially only for ptllnd). Also a separate userspace
189 ptllnd facility to gather history which should really be
190 covered by D_NETTRACE too, if only CDEBUG recorded history in
196 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
197 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
198 callback can occur before a connection has actually been
199 established. This caused an assertion failure previously.
201 Severity : enhancement
203 Description: Multiple instances for o2iblnd
204 Details : Allow multiple instances of o2iblnd to enable networking over
205 multiple HCAs and routing between them.
209 Description: lnet deadlock in router_checker
210 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
211 into BH locks to eliminate potential deadlock caused by
212 ksocknal_data_ready() preempting code holding these locks.
216 Description: Millions of failed socklnd connection attempts cause a very slow FS
217 Details : added a new route flag ksnr_scheduled to distinguish from
218 ksnr_connecting, so that a peer connection request is only turned
219 down for race concerns when an active connection to the same peer
220 is under progress (instead of just being scheduled).
222 ------------------------------------------------------------------------------
224 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
226 * Support for networks:
227 socklnd - kernels up to 2.6.16
228 qswlnd - Qsnet kernel modules 5.20 and later
229 openiblnd - IbGold 1.8.2
231 viblnd - Voltaire ibhost 3.4.5 and later
232 ciblnd - Topspin 3.2.0
233 iiblnd - Infiniserv 3.3 + PathBits patch
234 gmlnd - GM 2.1.22 and later
235 mxlnd - MX 1.2.1 or later
236 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
239 Severity : major on XT3
241 Description: libcfs overwrites /proc/sys/portals
242 Details : libcfs created a symlink from /proc/sys/portals to
243 /proc/sys/lnet for backwards compatibility. This is no
244 longer required and makes the Cray portals /proc variables
249 Description: OFED FMR API change
250 Details : This changes parameter usage to reflect a change in
251 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
252 that FMR support is only used in experimental versions of the
253 o2iblnd - this change does not affect standard usage at all.
255 Severity : enhancement
257 Description: new ko2iblnd module parameter: ib_mtu
258 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
259 HCAs. You can avoid this problem by setting the MTU to 1024
260 using this module parameter.
262 Severity : enhancement
263 Bugzilla : 11118/11620
264 Description: ptllnd small request message buffer alignment fix
265 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
266 Round up small message size on sends in case this option
267 is not supported. 11620 was a defect in the initial
268 implementation which effectively asserted all peers had to be
269 running the correct protocol version which was fixed by always
270 NAK-ing such requests and handling any misalignments they
275 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
276 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
277 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
279 Severity : enhancement
281 Description: Patchless ZC(zero copy) socklnd
282 Details : New protocol for socklnd, socklnd can support zero copy without
283 kernel patch, it's compatible with old socklnd. Checksum is
284 moved from tunables to modparams.
288 Description: When ksocknal_del_peer() is called upon a peer whose
289 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
290 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
293 Frequency : when ptlrpc is under heavy use and runs out of request buffer
295 Description: In lnet_match_blocked_msg(), md can be used without holding a
299 Frequency : very rarely
301 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
302 If connd connects a route which has been closed by
303 ksocknal_shutdown(), ksocknal_create_routes() may create new
304 routes which hold references on the peer, causing shutdown
305 process to wait for peer to disappear forever.
307 Severity : enhancement
309 Description: Dump XT3 portals traces on kptllnd timeout
310 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
311 dump Cray portals debug traces to a file. The kptllnd module
312 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
313 is the basename of the dump file.
316 Frequency : infrequent
318 Description: kernel ptllnd fix bug in connection re-establishment
319 Details : Kernel ptllnd could produce protocol errors e.g. illegal
320 matchbits and/or violate the credit flow protocol when trying
321 to re-establish a connection with a peer after an error or
324 Severity : enhancement
326 Description: Allow /proc/sys/lnet/debug to be set symbolically
327 Details : Allow debug and subsystem debug values to be read/set by name
328 in addition to numerically, for ease of use.
331 Frequency : only in configurations with LNET routers
333 Description: routes automatically marked down and recovered
334 Details : In configurations with LNET routers if a router fails routers
335 now actively try to recover routes that are down, unless they
336 are marked down by an administrator.
338 ------------------------------------------------------------------------------
340 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
343 Frequency : very rarely, in configurations with LNET routers and TCP
345 Description: incorrect data written to files on OSTs
346 Details : In certain high-load conditions incorrect data may be written
347 to files on the OST when using TCP networks.
349 ------------------------------------------------------------------------------
351 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
353 - rework CDEBUG messages rate-limiting mechanism b=10375
354 - add per-socket tunables for socklnd if the kernel is patched b=10327
356 ------------------------------------------------------------------------------
358 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
360 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
361 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
362 - quiet LNET startup LNI message for liblustre b=10128
363 - Better console error messages if 'ip2nets' can't match an IP address
364 - Fixed overflow/use-before-set bugs in linux-time.h
365 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
366 - LNET teardown failed an assertion about the route table being empty
367 - Fixed a crash in LNetEQPoll(<invalid handle>)
368 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
369 - improve debug message for liblustre/Catamount nodes (b=10116)
371 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
372 * Configuration change for the XT3
373 The PTLLND is now used to run Lustre over Portals on the XT3.
374 The configure option(s) --with-cray-portals are no longer
375 used. Rather --with-portals=<path-to-portals-includes> is
376 used to enable building on the XT3. In addition to enable
377 XT3 specific features the option --enable-cray-xt3 must be
380 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
381 * Portals has been removed, replaced by LNET.
382 LNET is new networking infrastructure for Lustre, it includes a
383 reorganized network configuration mode (see the user
384 documentation for full details) as well as support for routing
385 between different network fabrics. Lustre Networking Devices
386 (LNDS) for the supported network fabrics have also been created
387 for this new infrastructure.
389 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
394 Frequency : rare (large Voltaire clusters only)
396 Description: the default number of reserved transmit descriptors was too low
397 for some large clusters
398 Details : As a workaround, the number was increased. A proper fix includes
401 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
406 Frequency : occasional (large-scale events, cluster reboot, network failure)
408 Description: too many error messages on console obscure actual problem and
409 can slow down/panic server, or cause recovery to fail repeatedly
410 Details : enable rate-limiting of console error messages, and some messages
411 that were console errors now only go to the kernel log
413 Severity : enhancement
415 Description: add /proc/sys/portals/catastrophe entry which will report if
416 that node has previously LBUGged
418 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
420 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
422 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
424 - handle error return code in kranal_check_fma_rx() (5915,6054)
426 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
428 - update vibnal (Voltaire IB NAL)
429 - update gmnal (Myrinet NAL), gmnalid
431 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
433 * Landed portals:b_port_step as follows...
435 - removed CFS_DECL_SPIN*
436 just use 'spinlock_t' and initialise with spin_lock_init()
438 - removed CFS_DECL_MUTEX*
439 just use 'struct semaphore' and initialise with init_mutex()
441 - removed CFS_DECL_RWSEM*
442 just use 'struct rw_semaphore' and initialise with init_rwsem()
444 - renamed cfs_sleep_chan -> cfs_waitq
445 cfs_sleep_link -> cfs_waitlink
447 - fixed race in linux version of arch-independent socknal
448 (the ENOMEM/EAGAIN decision).
450 - Didn't fix problems in Darwin version of arch-independent socknal
451 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
453 - removed libcfs types from non-socknal header files (only some types
454 in the header files had been changed; the .c files hadn't been