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
36 Description: node crash on socket teardown race
39 Frequency : 'lctl peer_list' issued on a mx net
41 Description: Enable lctl's peer_list for MXLND
44 Frequency : after Ptllnd timeouts and portals congestion
46 Description: Credit overflows
47 Details : This was a bug in ptllnd connection establishment. The fix
48 implements better peer stamps to disambiguate connection
49 establishment and ensure both peers enter the credit flow
50 state machine consistently.
55 Description: kptllnd didn't propagate some network errors up to LNET
56 Details : This bug was spotted while investigating 11394. The fix
57 ensures network errors on sends and bulk transfers are
58 propagated to LNET/lustre correctly.
60 Severity : enhancement
62 Description: Fixed console chatter in case of -ETIMEDOUT.
64 Severity : enhancement
66 Description: Added D_NETTRACE for recording network packet history
67 (initially only for ptllnd). Also a separate userspace
68 ptllnd facility to gather history which should really be
69 covered by D_NETTRACE too, if only CDEBUG recorded history in
75 Description: o2iblnd handle early RDMA_CM_EVENT_DISCONNECTED.
76 Details : If the fabric is lossy, an RDMA_CM_EVENT_DISCONNECTED
77 callback can occur before a connection has actually been
78 established. This caused an assertion failure previously.
80 Severity : enhancement
82 Description: Multiple instances for o2iblnd
83 Details : Allow multiple instances of o2iblnd to enable networking over
84 multiple HCAs and routing between them.
88 Description: lnet deadlock in router_checker
89 Details : turned ksnd_connd_lock, ksnd_reaper_lock, and ksock_net_t:ksnd_lock
90 into BH locks to eliminate potential deadlock caused by
91 ksocknal_data_ready() preempting code holding these locks.
95 Description: Millions of failed socklnd connection attempts cause a very slow FS
96 Details : added a new route flag ksnr_scheduled to distinguish from
97 ksnr_connecting, so that a peer connection request is only turned
98 down for race concerns when an active connection to the same peer
99 is under progress (instead of just being scheduled).
101 ------------------------------------------------------------------------------
103 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
105 * Support for networks:
106 socklnd - kernels up to 2.6.16
107 qswlnd - Qsnet kernel modules 5.20 and later
108 openiblnd - IbGold 1.8.2
110 viblnd - Voltaire ibhost 3.4.5 and later
111 ciblnd - Topspin 3.2.0
112 iiblnd - Infiniserv 3.3 + PathBits patch
113 gmlnd - GM 2.1.22 and later
114 mxlnd - MX 1.2.1 or later
115 ptllnd - Portals 3.3 / UNICOS/lc 1.5.x, 2.0.x
118 Severity : major on XT3
120 Description: libcfs overwrites /proc/sys/portals
121 Details : libcfs created a symlink from /proc/sys/portals to
122 /proc/sys/lnet for backwards compatibility. This is no
123 longer required and makes the Cray portals /proc variables
128 Description: OFED FMR API change
129 Details : This changes parameter usage to reflect a change in
130 ib_fmr_pool_map_phys() between OFED 1.0 and OFED 1.1. Note
131 that FMR support is only used in experimental versions of the
132 o2iblnd - this change does not affect standard usage at all.
134 Severity : enhancement
136 Description: new ko2iblnd module parameter: ib_mtu
137 Details : the default IB MTU of 2048 performs badly on 23108 Tavor
138 HCAs. You can avoid this problem by setting the MTU to 1024
139 using this module parameter.
141 Severity : enhancement
142 Bugzilla : 11118/11620
143 Description: ptllnd small request message buffer alignment fix
144 Details : Set the PTL_MD_LOCAL_ALIGN8 option on small message receives.
145 Round up small message size on sends in case this option
146 is not supported. 11620 was a defect in the initial
147 implementation which effectively asserted all peers had to be
148 running the correct protocol version which was fixed by always
149 NAK-ing such requests and handling any misalignments they
154 Description: When kib(nal|lnd)_del_peer() is called upon a peer whose
155 ibp_tx_queue is not empty, kib(nal|lnd)_destroy_peer()'s
156 'LASSERT(list_empty(&peer->ibp_tx_queue))' will fail.
158 Severity : enhancement
160 Description: Patchless ZC(zero copy) socklnd
161 Details : New protocol for socklnd, socklnd can support zero copy without
162 kernel patch, it's compatible with old socklnd. Checksum is
163 moved from tunables to modparams.
167 Description: When ksocknal_del_peer() is called upon a peer whose
168 ksnp_tx_queue is not empty, ksocknal_destroy_peer()'s
169 'LASSERT(list_empty(&peer->ksnp_tx_queue))' will fail.
172 Frequency : when ptlrpc is under heavy use and runs out of request buffer
174 Description: In lnet_match_blocked_msg(), md can be used without holding a
178 Frequency : very rarely
180 Description: If ksocknal_lib_setup_sock() fails, a ref on peer is lost.
181 If connd connects a route which has been closed by
182 ksocknal_shutdown(), ksocknal_create_routes() may create new
183 routes which hold references on the peer, causing shutdown
184 process to wait for peer to disappear forever.
186 Severity : enhancement
188 Description: Dump XT3 portals traces on kptllnd timeout
189 Details : Set the kptllnd module parameter "ptltrace_on_timeout=1" to
190 dump Cray portals debug traces to a file. The kptllnd module
191 parameter "ptltrace_basename", default "/tmp/lnet-ptltrace",
192 is the basename of the dump file.
195 Frequency : infrequent
197 Description: kernel ptllnd fix bug in connection re-establishment
198 Details : Kernel ptllnd could produce protocol errors e.g. illegal
199 matchbits and/or violate the credit flow protocol when trying
200 to re-establish a connection with a peer after an error or
203 Severity : enhancement
205 Description: Allow /proc/sys/lnet/debug to be set symbolically
206 Details : Allow debug and subsystem debug values to be read/set by name
207 in addition to numerically, for ease of use.
210 Frequency : only in configurations with LNET routers
212 Description: routes automatically marked down and recovered
213 Details : In configurations with LNET routers if a router fails routers
214 now actively try to recover routes that are down, unless they
215 are marked down by an administrator.
217 ------------------------------------------------------------------------------
219 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
222 Frequency : very rarely, in configurations with LNET routers and TCP
224 Description: incorrect data written to files on OSTs
225 Details : In certain high-load conditions incorrect data may be written
226 to files on the OST when using TCP networks.
228 ------------------------------------------------------------------------------
230 2006-07-31 Cluster File Systems, Inc. <info@clusterfs.com>
232 - rework CDEBUG messages rate-limiting mechanism b=10375
233 - add per-socket tunables for socklnd if the kernel is patched b=10327
235 ------------------------------------------------------------------------------
237 2006-02-15 Cluster File Systems, Inc. <info@clusterfs.com>
239 - fix use of portals/lnet pid to avoid dropping RPCs b=10074
240 - iiblnd wasn't mapping all memory, resulting in comms errors b=9776
241 - quiet LNET startup LNI message for liblustre b=10128
242 - Better console error messages if 'ip2nets' can't match an IP address
243 - Fixed overflow/use-before-set bugs in linux-time.h
244 - Fixed ptllnd bug that wasn't initialising rx descriptors completely
245 - LNET teardown failed an assertion about the route table being empty
246 - Fixed a crash in LNetEQPoll(<invalid handle>)
247 - Future protocol compatibility work (b_rls146_lnetprotovrsn)
248 - improve debug message for liblustre/Catamount nodes (b=10116)
250 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
251 * Configuration change for the XT3
252 The PTLLND is now used to run Lustre over Portals on the XT3.
253 The configure option(s) --with-cray-portals are no longer
254 used. Rather --with-portals=<path-to-portals-includes> is
255 used to enable building on the XT3. In addition to enable
256 XT3 specific features the option --enable-cray-xt3 must be
259 2005-10-10 Cluster File Systems, Inc. <info@clusterfs.com>
260 * Portals has been removed, replaced by LNET.
261 LNET is new networking infrastructure for Lustre, it includes a
262 reorganized network configuration mode (see the user
263 documentation for full details) as well as support for routing
264 between different network fabrics. Lustre Networking Devices
265 (LNDS) for the supported network fabrics have also been created
266 for this new infrastructure.
268 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
273 Frequency : rare (large Voltaire clusters only)
275 Description: the default number of reserved transmit descriptors was too low
276 for some large clusters
277 Details : As a workaround, the number was increased. A proper fix includes
280 2005-06-02 Cluster File Systems, Inc. <info@clusterfs.com>
285 Frequency : occasional (large-scale events, cluster reboot, network failure)
287 Description: too many error messages on console obscure actual problem and
288 can slow down/panic server, or cause recovery to fail repeatedly
289 Details : enable rate-limiting of console error messages, and some messages
290 that were console errors now only go to the kernel log
292 Severity : enhancement
294 Description: add /proc/sys/portals/catastrophe entry which will report if
295 that node has previously LBUGged
297 2005-04-06 Cluster File Systems, Inc. <info@clusterfs.com>
299 - update gmnal to use PTL_MTU, fix module refcounting (b=5786)
301 2005-04-04 Cluster File Systems, Inc. <info@clusterfs.com>
303 - handle error return code in kranal_check_fma_rx() (5915,6054)
305 2005-02-04 Cluster File Systems, Inc. <info@clusterfs.com>
307 - update vibnal (Voltaire IB NAL)
308 - update gmnal (Myrinet NAL), gmnalid
310 2005-02-04 Eric Barton <eeb@bartonsoftware.com>
312 * Landed portals:b_port_step as follows...
314 - removed CFS_DECL_SPIN*
315 just use 'spinlock_t' and initialise with spin_lock_init()
317 - removed CFS_DECL_MUTEX*
318 just use 'struct semaphore' and initialise with init_mutex()
320 - removed CFS_DECL_RWSEM*
321 just use 'struct rw_semaphore' and initialise with init_rwsem()
323 - renamed cfs_sleep_chan -> cfs_waitq
324 cfs_sleep_link -> cfs_waitlink
326 - fixed race in linux version of arch-independent socknal
327 (the ENOMEM/EAGAIN decision).
329 - Didn't fix problems in Darwin version of arch-independent socknal
330 (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)
332 - removed libcfs types from non-socknal header files (only some types
333 in the header files had been changed; the .c files hadn't been