OpenSM Release Notes

                          ======================

Version: OpenFabric Enterprise Distribution (OFED) 1.0

Repo:    https://openib.org/svn/gen2/branches/1.0/src/userspace/management/osm

Version: 7992

Date:    Jun 2006

1 Overview

----------

This document describes the contents of the OpenSM OFED 1.0 release.

OpenSM is an InfiniBand compliant Subnet Manager and Administrator,

and runs on top of OpenIB. The OpenSM version for this release

is openib-1.2.1

This document includes the following sections:

1 This Overview section (describing new features and software

  dependencies)

2 Known Issues And Limitations

3 Unsupported IB compliance statements

4 Major Bug Fixes

5 Main Verification Flows

6 Qualified software stacks and devices

1.1 New Features

* SA GuidInfoRecord support

* Default for maxsmps changed:

  Control the number of SMP sent in parallel and thus shorten the

  fabric initialization time.

* osmtest/osmt_slvl_vl_arb.c:

  Output file name changed from vl_arbs.txt to qos.txt

* Support new IBTA Errata IsPortInfoCapMaskMatchSupported:

  This new capability of the SA enables matching of individual port

  capability bits dramatically reducing the query size for agents like

  the SRP initiator query for finding SRP targets.

* Honor guid2lid when coming out of standby:

  This change adds an option to the opensm that forces it to honor the

  guid2lid file given when it comes out of Standby state. Currently,

  when opensm comes out of Standby state, it ignores the guid2lid file

  it read, and honors only the lids defined on the ports themselves.

* Add guid to opensm opts

  This enables the port on which to run the SM to be defined through

  the configuration file as well as through the command line.

* PPC support:

  No PPC QA was performed.

1.2 Software Dependencies

OpenSM depends on the installation of either OFED 1.0,

OpenIB gen2 (e.g. IBG2 distribution), OpenIB gen1 (e.g. IBGD

distribution) or Mellanox VAPI stacks. The qualified driver versions

are provided in Table 2, "Qualified IB Stacks".

1.4 Supported Devices Firmware

The main task of OpenSM is to initialize InfiniBand devices. The

qualified devices and their corresponding firmware versions

are listed in Table 3.

2 Known Issues And Limitations

------------------------------

* No Partition/Pkey policy support:

  OpenSM does not provide means to set partitions.

* No Service / Key associations:

  There is no way to manage Service access by Keys.

* No SM to SM SMDB synchronization:

  Puts the burden of re-registering services, multicast groups, and

  inform-info on the client application (or IB access layer core).

* No "port down" event handling:

  Changing the switch port through which OpenSM connects to the IB

  fabric may cause incorrect operation. Please restart OpenSM whenever

  such a connectivity change is made.

* Changing connections during SM operation:

  Under some conditions the SM can get confused by a change in

  cabling (moving a cable from one switch port to the other) and

  momentarily see this as having the same GUID appear connected

  to two different IB ports. Under some conditions, when the SM fails to

  get the corresponding change event it might mistakenly report this case

  as a "duplicated GUID" case and abort. It is advisable to double-check

  the syslog after each such change in connectivity and restart

  OpenSM if it has exited.

* No QoS support:

  No SL2VL and VLArbitration setting is performed by the SM.

3 Unsupported IB Compliance Statements

--------------------------------------

The following section lists all the IB compliance statements which

OpenSM does not support. Please refer to the IB specification for detailed

information regarding each compliance statement.

* C14-22 (Authentication):

  M_Key M_KeyProtectBits and M_KeyLeasePeriod shall be set in one

  SubnSet method. As a work-around, an OpenSM option is provided for

  defining the protect bits.

* C14-67 (Authentication):

  On SubnGet(SMInfo) and SubnSet(SMInfo) - if M_Key is not zero then

  the SM shall generate a SubnGetResp if the M_Key matches, or

  silently drop the packet if M_Key does not match.

* C15-0.1.23.4 (Authentication):

  InformInfoRecords shall always be provided with the QPN set to 0,

  except for the case of a trusted request, in which case the actual

  subscriber QPN shall be returned.

* o13-17.1.2 (Event-FWD):

  If no permission to forward, the subscription should be removed and

  no further forwarding should occur.

* C14-37.1.2 (Handover):

  Priority should be kept in non-volatile memory.

* C14-24.1.1.5 and C14-62.1.1.22 (Initialization):

  GUIDInfo - SM should enable assigning Port GUIDInfo.

* C14-44 (Initialization):

  If the SM discovers that it is missing an M_Key to update CA/RT/SW,

  it should notify the higher level.

* C14-62.1.1.11 (Initialization):

  PortInfo:VLHighLimit should match the configured VLArb on the port.

* C14-62.1.1.12 (Initialization):

  PortInfo:M_Key - Set the M_Key to a node based random value.

* C14-62.1.1.13 (Initialization):

  PortInfo:P_KeyProtectBits - set according to an optional policy.

* C14-62.1.1.24 (Initialization):

  SwitchInfo:DefaultPort - should be configured for random FDB.

* C14-62.1.1.32 (Initialization):

  RandomForwardingTable should be configured.

* o15-0.1.12 (Multicast):

  If the JoinState is SendOnlyNonMember = 1 (only), then the endport

  should join as sender only.

* o15-0.1.13 (Multicast):

  If a Join request using unrealistic parameters is received, return

  ERR_REQ_INVALID.

* o15-0.1.8 (Multicast):

  If a request for creating an MCG with fields that cannot be met,

  return ERR_REQ_INVALID (currently ignoring SL and FlowLabelTclass).

* C15-0.1.11 (SA-Query):

  Query response should use only base LIDs (as the feature has not

  been qualified yet).

* C15-0.1.19 (SA-Query):

  Respond to SubnGetMulti(MultiPathRec)

* C15-0.1.8.6 (SA-Query):

  Respond to SubnAdmGetTraceTable - this is an optional attribute.

* C15-0.1.8.7 (SA-Query):

  SubnAdmGetMulti SubnAdmGetMultiResp - Only in case of a MultiPath.

* C15-0.1.13 Services:

  Reject ServiceRecord create, modify or delete if the given

  ServiceP_Key does not match the one included in the ServiceGID port

  and the port that sent the request.

* C15-0.1.14 (Services):

  Provide means to associate service name and ServiceKeys.

4 Major Bug Fixes

-----------------

The following is a list of bugs that were fixed. Note that other less critical

or visible bugs were also fixed.

* Eliminate error on active -> active port state transition

  SM may transition port from armed to active but in the meantime, due

  to passing a data packet with active enable set, the port may

  already have transitioned to active. Active -> active port state

  transition is indicated as an error but it isn't really an error so

  don't indicate error in the osm log.

* Routing not set for the first LID in the last LFT block:

  Fix: osm_switch.c: In osm_switch_get_fwd_tbl_block last block calculation

* Corrupted guid2lid file causes OpenSM exit

  Fix: exit only if exit_on_fatal option is set (the default)

* OpenSM was causing Client-Re-Registration continuously:

  The SM was storing the response PortInfo.ClientReRegstration

  bit and using it during next Set(PortInfo). Fix: clear the bit on

  receive.

* Multicast Query Selectors MTU, rate, and PacketLifeTime were not exact

* Try not to recognize port change as duplicated GUID

  This fix solves the issue of a port move during heavy sweep

  being recognized as a duplicated guid. Fix: If the SM sees what

  seems to be a duplicated guid, but it also received an indication

  for immediately forcing another heavy sweep (for example, as a

  result of receiving trap 128) then it shouldn't issue a duplicated

  guid error (and possibly exit), but should just ignore this and

  continue. This means that only if the SM recognizes such a

  duplication in a stable subnet that it'll issue the error (and

  possibly exit).

* Set PKey table on switch ports not supporting it:

  OpenSM attempts to set pkey table entries on external switch ports

  even if the switch declares a PartitionEnforcementCap of zero.  The

  consequence is ERR 4108. Fix: Observe PartitionEnforcementCap of zero.

* Incorrect MCMemberRecord Get/GetTable in trusted mode:

  This change fixes the retrieval of the MCMember records according to

  Errata MGTWG3280. This fix provide means to obtain all the group

  members by issuing a trusted GetTable query.

* Trusted MCMemberRecord query was not recognized as trusted:

  Trust is checked by comparing the request SM_Key field to the SM

  SM_Key. The bug was in looking up the SM_Key from the response not

  the request.

* Port left in down state after setting MTU or OpVLs on its neighbor:

  In case of a difference between the MTU of two ports, only the port

  with the higher MTU was set to down. Its remote port was written in

  the DB as in the ACTIVE state although its real status was INIT.

  Because of this, the SM didn't try to move the remote port to

  ACTIVE.

* Atomic counters used throughout the code were broken:

  A new implementation has been provided.

* MC Group creation with "less than" MTU ignores the requester MTU:

  When requesting to create an MC group with MTU(rate) selector 1

  (meaning less than rate specified), the MC group is created with

  MTU(rate) requested - 1. This is without checking the MTU(rate) of

  the port requesting the creation of the multicast group. This means

  that if, for example, port with MTU=2 sends a request for MC group

  creation with MTU selector=1 and MTU=5, Opensm will try to create a

  MC group with MTU=4, and fail, since the port capabilities are not

  realizable. Fix: creation of the MC group with MTU(rate) also takes

  into account the MTU(rate) of the port requesting the creation.

* MC Group join does not validate that the joining port's capabilities

  match those of the MC.  Fix: Add verification of endport physical

  capability to join MC group.

* ClientReRegistration not sent to ports discovered after first sweep:

  PortInfo sent with ClientReRegistration bit turned on only during

  the first sweep after becoming Master. This doesn't cover all cases

  where ClientReRegistration should be turned on. Fix: turn on this

  bit also on new ports it discovers (in cases of subnet merging, for

  example).

* segfault during a report on deleted multicast group:

  osm_mcast_mgr.c, executing the line of code:

  osm_mgrp_send_delete_notice( p_mgr->p_subn, p_mgr->p_log, p_mgrp );

  caused segmentation fault since the handle p_mgrp was already

  deleted while the function was called. Fix: inserted the line above

  into the protected section.

* segfault in osm_get_gid_by_mad_addr:

  The affected flows are ports and multicast joins.

* segfault in LID manager:

  Handle NULL p_rem_physp can validly be NULL when the remote SMA is

  not responding (but physical link is up).

* segfault in Up/Down routing engine

5 Main Verification Flows

-------------------------

OpenSM verification is run using the following activities:

* osmtest - a stand-alone program

* ibmgtsim (IB management simulator) based - a set of flows that

  simulate clusters, inject errors and verify OpenSM capability to

  respond and bring up the network correctly.

* small cluster regression testing - where the SM is used on back to

  back or single switch configurations. The regression includes

  multiple OpenSM dedicated tests.

* cluster testing - when we run OpenSM to setup a large cluster, perform

  hand-off, reboots and reconnects, verify routing correctness and SA

  responsiveness at the ULP level (IPoIB and SDP).

5.1 osmtest

osmtest is an automated verification tool used for OpenSM

testing. Its verification flows are described by list below.

* Inventory File: Obtain and verify all port info, node info, and path

  records parameters.

* Service Record:

   - Register new service

   - Register another service (with a lease period)

   - Register another service (with service p_key set to zero)

   - Get all services by name

   - Delete the first service

   - Delete the third service.

   - Added bad flows of get/delete  non valid service

   - Add / Get same service with different data

   - Add / Get / Delete by different component  mask values (services

     by Name & Key / Name & Data / Name & Id / Id only )

* Multicast Member Record:

   - Query of existing Groups (IPoIB)

   - BAD Join with insufficient comp mask (o15.0.1.3)

   - Create given MGID=0 (o15.0.1.4)

   - Create given MGID=0xFF12A01C,FE800000,00000000,12345678 (o15.0.1.4)

   - Create BAD MGID=0xFA. (o15.0.1.6)

   - Create BAD MGID=0xFF12A01B w/ link-local not set (o15.0.1.6)

   - New MGID with invalid join state (o15.0.1.9)

   - Retry of existing MGID - See JoinState update (o15.0.1.11)

   - BAD RATE when connecting to existing MGID (o15.0.1.13)

   - Partial JoinState delete request - removing FullMember (o15.0.1.14)

   - Full Delete of a group (o15.0.1.14)

   - Verify Delete by trying to Join deleted group (o15.0.1.14)

   - BAD Delete of IPoIB membership (no prev join) (o15.0.1.15)

* GUIDInfo Record:

   - All GUIDInfoRecords in subnet are obtained

* Event Forwarding: Register for trap forwarding using reports

   - Send a trap and wait for report

   - Unregister non-existing

* Trap 64/65 Flow: Register to Trap 64-65, create traps (by

  disconnecting/connecting ports) and wait for report, then unregister.

* Stress Test: send PortInfoRecord queries, both single and RMPP and

  check for the rate of responses as well as their validity.

5.2 IB Management Simulator OpenSM Test Flows:

The simulator provides ability to simulate the SM handling of virtual

topologies that are not limited to actual lab equipment availability.

OpenSM was simulated to bring up clusters of up to 10,000 nodes. Daily

regressions use smaller (16 and 128 nodes clusters).

The following test flows are run on the IB management simulator:

* Stability:

  Up to 12 links from the fabric are randomly selected to drop packets

  at drop rates up to 90%. The SM is required to succeed in bringing the

  fabric up. The resulting routing is verified to be correct, too.

* LID Manager:

  Using LMC = 2 the fabric is initialized with LIDs. Faults such as

  zero LID, Duplicated LID, non-aligned (to LMC) LIDs are

  randomly assigned to various nodes and other errors are randomly

  output to the guid2lid cache file. The SM sweep is run 5 times and

  after each iteration a complete verification is made to ensure that all

  LIDs that could possibly be maintained are kept, as well as that all nodes

  were assigned a legal LID range.

* Multicast Routing:

  Nodes randomly join the 0xc000 group and eventually the

  resulting routing is verified for completeness and adherence to

  Up/Down routing rules.

* osmtest:

  The complete osmtest flow as described in the previous table is run on

  the simulated fabrics.

* Stress Test:

  This flow merges fabric, LID and stability issues with continuous

  PathRecord, ServiceRecord and Multicast Join/Leave activity to

  stress the SM/SA during continuous sweeps.

5.3 OpenSM Regression

Using a back-to-back or single switch connection, the following set of

tests is run nightly on the stacks described in table 2. The included

tests are:

* Stress Testing: Flood the SA with queries from multiple channel

  adapters to check the robustness of the entire stack up to the SA.

* Dynamic Changes: Dynamic Topology changes, through randomly

  dropping SMP packets, used to test OpenSM adaptation to an unstable

  network & verify DB correctness.

* Trap Injection: This flow injects traps to the SM and verifies that it

  handles them gracefully.

* SA Query Test: This test exhaustively checks the SA responses to all

  possible single component mask. To do that the test examines the

  entire set of records the SA can provide, classifies them by their

  field values and then selects every field (using component mask and a

  value) and verifies that the response matches the expected set of records.

  A random selection using multiple component mask bits is also performed.

5.4 Cluster testing:

Cluster testing is usually run before a distribution release. It

involves real hardware setups of 16 to 32 nodes (or more if a beta site

is available). Each test is validated by running all-to-all ping through the IB

interface. The test procedure includes:

* Cluster bringup

* Hand-off between 2 or 3 SM's while performing

  - Node reboots

  - Switch power cycles (disconnecting the SM's)

* Unresponsive port detection and recovery

* osmtest from multiple nodes

* Trap injection and recovery

6 Qualification

----------------

Table 2 - Qualified IB Stacks

=============================

Stack                                    | Version

-----------------------------------------|--------------------------

OFED                                     |   1.0

OpenIB Gen2 (IBG2 distribution)          |   1.0

OpenIB Gen1 (IBGD distribution)          |   1.8.0

VAPI (Mellanox InfiniBand HCA Driver)    |   3.2 and later

Table 3 - Qualified Devices and Corresponding Firmware

======================================================

Mellanox

Device  |   FW versions

--------|-----------------------------------------------------------

MT43132 |   InfiniScale - fw-43132  5.2.0 (and later)

MT47396 |   InfiniScale III - fw-47396 0.5.0 (and later)

MT23108 |   InfiniHost - fw-23108   3.3.2

MT25204 |   InfiniHost III Lx - fw-25204  1.0.1

MT25208 |   InfiniHost III Ex (InfiniHost Mode) - fw-25208  4.6.2 (and later)

MT25208 |   InfiniHost III Ex (MemFree Mode) - fw-25218  5.0.1 (and later)

QLogic/PathScale

Device  |   Note

--------|-----------------------------------------------------------

iPath   | QHT6040 (PathScale InfiniPath HT-460)

iPath   | QHT6140 (PathScale InfiniPath HT-465)

iPath   | QLE6140 (PathScale InfiniPath PE-880)

Note: OpenSM does not run on an IBM Galaxy (eHCA) as it does not expose

QP0 and QP1. However, it does support it as a device on the subnet.