1 2007-04-23 Cluster File Systems, Inc. <info@clusterfs.com>
2 * version 1.4.11 / 1.6.1
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
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
16 ------------------------------------------------------------------------------
18 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
19 * version 1.4.10 / 1.6.0
20 * Support for networks:
21 socklnd - kernels up to 2.6.16
22 qswlnd - Qsnet kernel modules 5.20 and later
23 openiblnd - IbGold 1.8.2
25 viblnd - Voltaire ibhost 3.4.5 and later
26 ciblnd - Topspin 3.2.0
27 iiblnd - Infiniserv 3.3 + PathBits patch
28 gmlnd - GM 2.1.22 and later
29 mxlnd - MX 1.2.1 or later
30 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
34 Frequency : 'lctl peer_list' issued on a mx net
36 Description: Enable lctl's peer_list for MXLND
39 Frequency : after Ptllnd timeouts and portals congestion
41 Description: Credit overflows
42 Details : This was a bug in ptllnd connection establishment. The fix
43 implements better peer stamps to disambiguate connection
44 establishment and ensure both peers enter the credit flow
45 state machine consistently.
50 Description: kptllnd didn't propagate some network errors up to LNET
51 Details : This bug was spotted while investigating 11394. The fix
52 ensures network errors on sends and bulk transfers are
53 propagated to LNET/lustre correctly.
55 Severity : enhancement
57 Description: Fixed console chatter in case of -ETIMEDOUT.
59 Severity : enhancement
61 Description: Added D_NETTRACE for recording network packet history
62 (initially only for ptllnd). Also a separate userspace
63 ptllnd facility to gather history which should really be
64 covered by D_NETTRACE too, if only CDEBUG recorded history in
70 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
71 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
72 callback can occur before a connection has actually been
73 established. This caused an assertion failure previously.
75 Severity : enhancement
77 Description: Multiple instances for o2iblnd
78 Details : Allow multiple instances of o2iblnd to enable networking over
79 multiple HCAs and routing between them.
83 Description: lnet deadlock in router_checker
84 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
85 into BH locks to eliminate potential deadlock caused by
86 ksocknal_data_ready() preempting code holding these locks.
90 Description: Millions of failed socklnd connection attempts cause a very slow FS
91 Details : added a new route flag ksnr_scheduled to distinguish from
92 ksnr_connecting, so that a peer connection request is only turned
93 down for race concerns when an active connection to the same peer
94 is under progress (instead of just being scheduled).
96 ------------------------------------------------------------------------------
98 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
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
113 Severity : major on XT3
115 Description: libcfs overwrites /proc/sys/portals
116 Details : libcfs created a symlink from /proc/sys/portals to
117 /proc/sys/lnet for backwards compatibility. This is no
118 longer required and makes the Cray portals /proc variables
123 Description: OFED FMR API change
124 Details : This changes parameter usage to reflect a change in
125 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
126 that FMR support is only used in experimental versions of the
127 o2iblnd - this change does not affect standard usage at all.
129 Severity : enhancement
131 Description: new ko2iblnd module parameter: ib_mtu
132 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
133 HCAs. You can avoid this problem by setting the MTU to 1024
134 using this module parameter.
136 Severity : enhancement
137 Bugzilla : 11118/11620
138 Description: ptllnd small request message buffer alignment fix
139 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
140 Round up small message size on sends in case this option
141 is not supported. 11620 was a defect in the initial
142 implementation which effectively asserted all peers had to be
143 running the correct protocol version which was fixed by always
144 NAK-ing such requests and handling any misalignments they
149 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
150 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
151 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
153 Severity : enhancement
155 Description: Patchless ZC(zero copy) socklnd
156 Details : New protocol for socklnd, socklnd can support zero copy without
157 kernel patch, it's compatible with old socklnd. Checksum is
158 moved from tunables to modparams.
162 Description: When ksocknal_del_peer() is called upon a peer whose
163 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
164 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
167 Frequency : when ptlrpc is under heavy use and runs out of request buffer
169 Description: In lnet_match_blocked_msg(), md can be used without holding a
173 Frequency : very rarely
175 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
176 If connd connects a route which has been closed by
177 ksocknal_shutdown(), ksocknal_create_routes() may create new
178 routes which hold references on the peer, causing shutdown
179 process to wait for peer to disappear forever.
181 Severity : enhancement
183 Description: Dump XT3 portals traces on kptllnd timeout
184 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
185 dump Cray portals debug traces to a file. The kptllnd module
186 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
187 is the basename of the dump file.
190 Frequency : infrequent
192 Description: kernel ptllnd fix bug in connection re-establishment
193 Details : Kernel ptllnd could produce protocol errors e.g. illegal
194 matchbits and/or violate the credit flow protocol when trying
195 to re-establish a connection with a peer after an error or
198 Severity : enhancement
200 Description: Allow /proc/sys/lnet/debug to be set symbolically
201 Details : Allow debug and subsystem debug values to be read/set by name
202 in addition to numerically, for ease of use.
205 Frequency : only in configurations with LNET routers
207 Description: routes automatically marked down and recovered
208 Details : In configurations with LNET routers if a router fails routers
209 now actively try to recover routes that are down, unless they
210 are marked down by an administrator.
212 ------------------------------------------------------------------------------
214 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
217 Frequency : very rarely, in configurations with LNET routers and TCP
219 Description: incorrect data written to files on OSTs
220 Details : In certain high-load conditions incorrect data may be written
221 to files on the OST when using TCP networks.
223 ------------------------------------------------------------------------------
225 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
227 - rework CDEBUG messages rate-limiting mechanism b=10375
228 - add per-socket tunables for socklnd if the kernel is patched b=10327
230 ------------------------------------------------------------------------------
232 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
234 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
235 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
236 - quiet LNET startup LNI message for liblustre b=10128
237 - Better console error messages if 'ip2nets' can't match an IP address
238 - Fixed overflow/use-before-set bugs in linux-time.h
239 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
240 - LNET teardown failed an assertion about the route table being empty
241 - Fixed a crash in LNetEQPoll(<invalid handle>)
242 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
243 - improve debug message for liblustre/Catamount nodes (b=10116)
245 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
246 * Configuration change for the XT3
247 The PTLLND is now used to run Lustre over Portals on the XT3.
248 The configure option(s) --with-cray-portals are no longer
249 used. Rather --with-portals=<path-to-portals-includes> is
250 used to enable building on the XT3. In addition to enable
251 XT3 specific features the option --enable-cray-xt3 must be
254 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
255 * Portals has been removed, replaced by LNET.
256 LNET is new networking infrastructure for Lustre, it includes a
257 reorganized network configuration mode (see the user
258 documentation for full details) as well as support for routing
259 between different network fabrics. Lustre Networking Devices
260 (LNDS) for the supported network fabrics have also been created
261 for this new infrastructure.
263 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
268 Frequency : rare (large Voltaire clusters only)
270 Description: the default number of reserved transmit descriptors was too low
271 for some large clusters
272 Details : As a workaround, the number was increased. A proper fix includes
275 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
280 Frequency : occasional (large-scale events, cluster reboot, network failure)
282 Description: too many error messages on console obscure actual problem and
283 can slow down/panic server, or cause recovery to fail repeatedly
284 Details : enable rate-limiting of console error messages, and some messages
285 that were console errors now only go to the kernel log
287 Severity : enhancement
289 Description: add /proc/sys/portals/catastrophe entry which will report if
290 that node has previously LBUGged
292 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
294 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
296 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
298 - handle error return code in kranal_check_fma_rx() (5915,6054)
300 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
302 - update vibnal (Voltaire IB NAL)
303 - update gmnal (Myrinet NAL), gmnalid
305 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
307 * Landed portals:b_port_step as follows...
309 - removed CFS_DECL_SPIN*
310 just use 'spinlock_t' and initialise with spin_lock_init()
312 - removed CFS_DECL_MUTEX*
313 just use 'struct semaphore' and initialise with init_mutex()
315 - removed CFS_DECL_RWSEM*
316 just use 'struct rw_semaphore' and initialise with init_rwsem()
318 - renamed cfs_sleep_chan -> cfs_waitq
319 cfs_sleep_link -> cfs_waitlink
321 - fixed race in linux version of arch-independent socknal
322 (the ENOMEM/EAGAIN decision).
324 - Didn't fix problems in Darwin version of arch-independent socknal
325 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
327 - removed libcfs types from non-socknal header files (only some types
328 in the header files had been changed; the .c files hadn't been