Minutes from the 10/27/04 LCLS SLC IOC meetings.

Next Meeting: 11/03/04, 9:30am, in B5, second floor.

(1) MSG/DB design review moved to 11/10 due to down.
    Steph - intro, Diane - message service, Debbie - database service
    We will probably need more than one meeting for this.  We will have
    a rough draft of what we'll present/discuss by next meeting.  We will
    include functional requirements, design concepts and assumptions, and
    system data flow diagrams.  Not necessary to go into detail on each 
    individual routine but we should list utility routines available to
    the BPM and MGNT apps.

(2) PRIMARY.MAP Data File and Dictionary - Ken will combine the two files
    into one.  Debbie will enter all unique slc "primary names" and 
    "primary secondary" names in the dictionary for name validity error 
    checking. All primary/secondary file entries that match the ioc ST0 
    database downloaded will be stored as "primary unit secondary" in 
    the dictionary. In addition, each of these matching 
    "primary unit secondary" entries will be added to a linked list, 
    managed by the associated "primary" dictionary entry, for use with 
    ALL* commands.  A dbput/dbget/dblist of either a non-existent primary or 
    unit (except ALL*) returns an error.  Currently on the existing micros, 
    a dbget/dblist of an invalid secondary will return good status but no 
    update to the db list.  We will make this an error for IOCs since
    PV names have to be mapped to prim/unit/secn and we have to expect
    typos in PV names.

(3) DB IO Routines - Debbie is done with porting dblist and dbunits from 
    assembly to C using ST0 offset as the dblist pointer.  She can now test
    and debug her ST0 endian swap work.  She currently uses linear
    progression for ALL* units.  She will now work on the dictionary that
    combines ST0 and the other database info and hashing. Because of the
    down, design review, and her other work, she won't finish the DB IO
    routines until later in Nov.

(4) SLC IOC "IPL" - Sequence of events when triggered by the SCP (note that
    prototype doesn't yet do all this):
    (1) SCP IPL sets the IOC online in the micr_active_mask, sends the 
        "BOOT" message, and waits for the reply.  Any SCP can now attempt
	to send messages to the IOC even though it may not be ready to 
	accept messages in which case the SCP will time out waiting
	for a reply.
    (2) IOC msg tasks receive the BOOT message and reply.  
    (3) SCP IPL gets the reply or times out waiting for it.  For now the SCP
        will ignore absence of reply.  Once the reply is implemented, 
	if the reply is absent, step (4) may be skipped and the SCP action 
	finishes.
    (4) SCP IPL waits for IOC to copy CSTR:VTIM into CSTR:MTIM with a timeout 
        of >=90 seconds, while DBEX is responding to IOC's dbdownload request.
    (5) IOC msgHdlr tells the slcExec to restart which then 
        stops and restarts ALL tasks in the proper order.
    (6) IOC slcExec sends a message to the IOC dbSend to download which 
        forwards it to DBEX.
    (7) DBEX sets the IOC online in the micr_active_mask if it is not already
        set.  
    (8) DBEX and the IOC dbSend and dbRecv work to get the database downloaded.
    (9) While the database is downloaded, the IOC msg tasks will process any 
        MSG messages (ie, from PARANOIA which polls "online" IOCs to find 
	those that are no longer responsive).  The msg tasks do not need to
	access the database.  Any non-MSG messages received during
	this time will be thrown away and a message logged (with func code) 
	for each discarded message.  We will want to review this action later -
	we may want to allow some messages to queue up for a short time
	before discarding.
    (10)Any IOC task that needs access to the SLC database will need to wait
	for the database to be ready.
    (11)If IOC dbRecv finishes the database download successfully, it then 
        sends a message to dbHdlr to convert ST0 data to new format.
	dbHdlr reads additional database information from the PRIMARY.MAP 
	file on the NFS server and creates a dictionary with ST0 and the 
	PRIMARY.MAP information.
    (12)If IOC dbRecv or dbHdlr have any error, a specific message is logged 
        and no further action is taken.  The SCP IPL will time out.  Another 
	IPL is required.
    (13)IOC dbHdlr then allows any task waiting on the database to proceed.
    (14)IOC dbHdlr copies CSTR VTIM to CSTR MTIM and sends a message to 
        dbCtrl to send CSTR MTIM to DBEX as a ST3 update.
    (15)DBEX updates CSTR MTIM.
    (16)SCP IPL sees the update and the action finishes.

(5) SLC IOC restart when not triggered by the SCP - only steps (5) to (15)
    apply.

(6) DBEX Up and Down Messages - the IOC will store a flag that DBEX
    is up or down in some global area.  It will also store the database
    major/minor version that comes with the DBEX up message.  We may use 
    the flag to decide whether or not to attempt sending database updates.  
    We may use the DBEX-up message to trigger an update of all data that's 
    changed since the last successful update.

(7) SLC Task Restart - Resource Allocation and Freeing
    Diane has studied this and written a discussion of issues and 
    recommendations.  We'll get this posted on the web and discuss it 
    in the next meeting.  Note that most EPICS utilities call cantproceed 
    (which will log a message and suspend the task) when there is a fatal 
    error allocating a resource.  This will force an IOC reboot which is a 
    reasonable action for this kind of error.  Note that the error that is 
    logged is from errlogPrintf (so won't have all the tags that come with 
    iocCmlogLogMsg).  Diane and Debbie both plan to use cantproceed in their 
    own code to suspend the task on a fatal error.

(8) EPICS vs SLC Typedefs - Diane has noticed a lot of similarity between
    EPICS and SLC typedefs in particular with data types (int2u (SLC) is
    epicsUInt16 (EPICS)).  We've agreed that Diane and Debbie should be
    using EPICS typedefs whenever possible.  This may mean fewer include files
    to ftp from the VMS machines.

(9) IPLing from EPICS and IOC Health - Diane has a subroutine record ready to 
    test.  It allows restart or stop using CA (in case the user has no SCP or 
    cannot access iocsh).  Diane is working on the EDM display.  Note that 
    this display is the beginnings of the SLC IOC "health" display which will 
    eventually show things like CPU, file descriptor, memory usage from the 
    non-SLC side of the IOC.  And crate status if available.  It can also show 
    data from CSTR.  And have a button to reboot the whole IOC.  We can also 
    add records to display diagnostics from the SLC tasks (ie, any information 
    that may be stored in global structures, etc).

(10)CMLOG - James still hasn't been able to start iocCmlog.  RonM needs to
    check that the VMS errlog report will work properly for SLC IOCs. 

(11)Magnet Job Functional Requirements - Dayle is working on understanding 
    what happens for each function code.  She is shooting to be done by Dec 1.