U From:	STAR::STOCKDALE    "DTN 381-2751 ZK3-4/W23 Office 4U22" 10-MAR-1993 07:00:41.83 2 To:	GAGNE,THOMAS,ORGOVAN,L_LEAHY,PILSMAKER,LBENSON CC:	HAYAKAWA,HEATH,STOCKDALE" Subj:	XXDRIVER Enhancement Project  D Pursuant to the attached mail from Howard, please review the plannedF enhancements for XXDRIVER.  Send me any comments or suggested changes.   What to review:    (1) Completeness of the test.   B     Does the data integrity test cover the functionality needed by'     applications (clusters, UCX, etc.)?   2 (2) Matching of the schedule to CORAL and EPSILON.  
 (3) Staffing.   G     The person doing the work is myself.  The work listed here consists H     of about 3 days code/debug mixed in with about a month of code/debugD     for other planned enhancements to XXDRIVER.  Some of these otherJ     changes are intended to 'productize' XXDRIVER such that it can be usedL     more easily by a qual group so that future testing can include XXDRIVER.  . (4) Who will run the test on an ongoing basis.  K     The LAN qual effort done for each driver and adapter to be shipped in a I     VMS release will continue to include XXDRIVER.  XXDRIVER will also be K     given to appropriate test groups to include in the suite of tests being J     run (it is already happening with DELTA, with the ALPHA qual group and/     with COBRA and LASER Python qualification).   ( XXDRIVER Enhancement Project, the Design( ----------------------------------------  D What XXDRIVER currently does with respect to data integrity testing:  H   (1) The packet header consists of the packet header plus some XXDRIVERJ       defined fields (16 bytes) to contain packet context.  So the minimum&       packet size is 16+packet header.  M   (2) Given VCRPs allocated by the LAN ALLOCATE_VCRP routine (which means the I       alignment and physically contiguous requirements for the particular        driver are met).  G   (3) Receive all packet sizes from header size to maximum packet size.   N   (4) Transmit non-chained all packet sizes from header size to maximum packet       size.   P   (5) Transmit chained (1 additional segment, DCBE) all packet sizes from headerQ       size to maximum packet with size of the first segment varying from 32 bytes Q       to the MIN(200,full packet size) (ie. includes zero-byte chained segments). '       Buffers are virtually contiguous.   O   (6) DCBE segments are described alternately by an SVA or a SVAPTE/BOFF, where O       the BOFF is given as 0..255*pagesize.  The SVAPTE list contains PTEs with O       the valid bit set (the SVAPTE is the system virtual address of the PTE in        the SPT).   *   (7) Transmit with pipeline from 1 to 32.  L What functionality will be added to XXDRIVER in time for EPSILON / CORAL LANL qualification (numbers in parentheses refer to which item above is changed):  M   - Enhance (5) the operation of transmit chaining to separate the additional 5     segment(s) from the first segment (VCRP section).   K   - Enhance the building of transmit VCRPs (4) to vary the buffer offset to J     cover all byte offsets within a 512-byte page.  (The varying offset isI     already being done for the chain segments).  This includes crossing a K     page boundary on occasion on the Alpha side (it would already for VAX).   M   - Enhance transmit chaining (5) to allow multiple chain segments (up to 8).   I   - Enhance the working of the SVAPTE/BOFF in the DCBE (6) to specify the      following forms of PTEs:  (     PTE<V,TYP1,TYP0> = 1,x,x - valid PFN(                        0,0,0 - valid PFNM                        0,0,1 - GPT index, for which GPT + GPTX*PTESIZE is the M                                SVA of a second PTE which has PTE<V,TYP1,TYP0> 1                                of 1,x,x or 0,0,0.   H   - Enhance the working of the SVAPTE/BOFF in the DCBE (6) to supply the     PTEs in backwards form.   I   - Enhance (1) to specify multiple data patterns in the header (and give J     the different data pattern in the remainder of the buffer) so that the!     data pattern is less regular.   / --------------- Attachment --------------------   ; From:	STAR::HAYAKAWA     "OAtmOS"    1-MAR-1993 11:04:54.19   To:	THOMAS,STOCKDALE,GAGNE,HEATH CC:	ORGOVAN,HAYAKAWA# Subj:	FDDI test enhancement project   G Please schedule review of this proposal so I can send this note forward E to a wide audience.   It would be a big help if you can get this done E this week so the world doesn't get nervous but  if people's calendar  B doesn't permit a meeting this week, let me know when the review isA scheduled so I can get a progress report out.    I think you have A all items covered including timing and only need the formality of  having partners agree.  E I'd like you to explicitly address the following items in your review J and the report.   With this information, your partners and people outside . VMS should feel comfortable with your plan.      	1. completeness of the tests 3 	2. matching of your schedule to CORAL and EPSILON   		and staffing.  	3. staffing commitment - 	4. who will run the test on an ongoing basis    Project reviewers:  , 	CORAL release mgr, Linda Benson or designee# 	EPSILON release mgr, Vince Orgovan $ 	ALPHA I/O technical ldr, Ben Thomas* 	VAX & ALPHA LAN technical ldr, Dave Gagne% 	VAX test,   Linda Benson or designee ' 	ALPHA test, Jeff Pilsmaker or designee  	VAX & ALPHA Lavc, Lee Leahy  J From:	STAR::STOCKDALE    "DTN 381-2751 ZK3-4/W23 Office 4U22" 27-FEB-1993  07:10:25.03  To:	STAR::HAYAKAWA# CC:	GAGNE,HEATH,STOCKDALE,STOCKDALE L Subj:	RE: Regression test, re: Boeing FXdriver issue - add minor corrections  D The regression testing needed to validate the areas of functionalityA recently found deficient at Boeing and CVG consists of two items:   K   - Validate global page handling for SVAPTE/BOFF form of transmit chaining 3   - Validate zero byte count DCBE chained transmits   L The regression test is part of XXDRIVER.  XXDRIVER is a LAN qual test driverL that was originally done for VAX for performance testing.  For ALPHA V1.0 itH was rewritten to improve the performance testing capability and add dataI integrity testing of transmit and receive paths for the LAN drivers.  For H ALPHA V1.5, better handling of FDDI has been added and more capabilities" with respect to transmit chaining.  H The functionality currently present in XXDRIVER with respect to transmit	 chaining:   >   - Validates all forms and permutations of transmit chaining,?     including multi-page buffer offsets, byte alignment, packet :     sizes and arrangement between chained and non-chained.3   - Validate zero byte count DCBE chained transmits    What is not done is:     - Global page handling@   - Chained buffer arrangements where odd forms of PTE lists are=     present (ie. presenting buffers in backwards order, etc.) >   - Multiple DCBE chaining for FDDI (may be used by UCX in the.     future, not currently used by anyone else)  N This work has been planned for EPSILON/CORAL qual tool development for severalP months.  The work consists of a month of design and debug and then validation ofB the existing drivers on VAX and ALPHA, spread over several months.        Planned   Completion date    Task    ---------------    ----     1-May-93          Design   D    1-Jun-93          Rework XX.MAR and XXDRIVER.MAR for code changes: 		     (this includes other planned XXDRIVER enhancements)  #    1-Jul-93          Debug complete   <    1-Aug-93          Validate existing VAX and ALPHA drivers  C The data integrity portion of XXDRIVER will be code reviewed (which E would include a look at the design as well).  This will be done after : the current code reviews of the FDDI drivers are complete.  G LAN code reviews are done every Thursday morning.  Anyone interested in F attending is welcome.  Attendance for the XXDRIVER piece by Lee Leahy C would help ensure we didn't miss anything with respect to what sort  of requests PEDRIVER gives us.  7 Any other suggestions for LAN qual are welcome anytime.   