CHAPTER VAX PAGESWAPPER - April 1986 - Volume 7 Number 9 Security Hole in VMS 4.2 . . . . . . . . . . . . VAX-3 Editor's Workfile . . . . . . . . . . . . . . . VAX-4 A Concern about Security and DZ Ports . . . . . VAX-5 A User's Experience with Terminal Servers . . . VAX-8 European DECUS Question and Answer Session . . VAX-13 VAX System SIG Committee List . . . . . . . . VAX-61 INPUT/OUTPUT . . . . . . . . . . . . . . . . . VAX-64 Forms at the End INPUT/OUTPUT Submission Form . . . . . . . . . VAX-65 System Improvement Request Submission Form . . VAX-67 VAX Systems SIG Spring 1986 SIR Ballot . . . . VAX-69 PAGESWAPPER - April 1986 - Volume 7 Number 9 General material for publication in the Pageswapper should be sent (US mail only -- no "express" services please) to: Larry Kilgallen, PAGESWAPPER Editor Box 81, MIT Station Cambridge, MA 02139-0901 USA Preference is given to material submitted as machine-readable text (best is Runoff source). Line length should not exceed 64 characters. Please do not submit program source, as that is better distributed on the VAX SIG tape. Change of address, reports of non-receipt, and other circulation correspondence should be sent to: DECUS U.S. Chapter Attention: Publications Department 249 Northboro Road (BPO2) Marlborough, MA 01752 USA Only if discrepancies of the mailing system are reported can they be analyzed and corrected. VAX-2 PAGESWAPPER - April 1986 - Volume 7 Number 9 Security Hole in VMS 4.2 Security Hole in VMS 4.2 Art McClinton Mitre Corporation McLean, Va Every feature of an operating system that is designed to allow a user access to data that he would not otherwise have access to has the potential of leading to pitfalls for the system manager. Normally this would only come into play if you take over a system that was installed by someone else. It is very easy to install a few Access Control List (ACL) entries that can be used at a later date to perform extensive damage. The first thing a new system manager should do is analyze all ACL's in the system for potential side effects. Additionally, VMS 4.2 has a bug that allows non-privileged users to obtain SYSNAM privilege. The hole may be easily plugged by the system manager. The remainder of this brief article will describe the hole, its impact, and how to fix your system so that users cannot get in through the hole. Normally I would not write about security holes but this one is so widely publicized at this point that not to write about it would be the worse of the two evils. ACL's may be placed on files, devices or logical name tables. In the case of the logical name table, the system default is to allow any user the privilege of placing an ACL on any logical name table. This enables a non-privileged user the ability to set an ACL on the LNM$SYSTEM table to allow read+write+control. This enables the same user to modify entries in the table. This is fully equivalent to SYSNAM privilege. It does not take much imagination to write a trojan horse if you have SYSNAM privilege. The details will be left to the hacker. Like most security holes, it is very easy to close it once it has been discovered. The following code should be added to the system startup command file to close the hole: $SET ACL/OBJ=LOGICAL/ACL=(ID=[*,*],ACCESS=READ) LNM$SYSTEM__TABLE $SET ACL/OBJ=LOGICAL/ACL=(ID=[*,*],ACCESS=READ) LNM$SYSTEM__DIRECTORY Similar lines of code should be added for any other site specific system logical name tables that you have created. Digital is aware of the problem and plans to fix it in a future release of VMS. They have installed a tech tip at the Telephone Support Center. This information has also appeared on ARPANET. With such wide exposure, I feel that we need to communicate the problem and solution to as many system managers as possible so VAX-3 PAGESWAPPER - April 1986 - Volume 7 Number 9 Security Hole in VMS 4.2 that they may protect their systems. Editor's Workfile by Larry Kilgallen, Pageswapper Editor Don't Forget to Vote - The System Improvement Request list is in the issue of two months ago, and the ballot is reprinted at the end of this issue. Photocopies of the form are quite acceptable; it is your membership number which validates the ballot as being yours. Isn't it nice - that a European DECUS member put in the effort to share their VAX Q & A with everyone who did not attend their symposium? Isn't it too bad it has been so long since anyone from the US has put in the effort to transcribe the similar sessions which are held at US Symposia? VAX-4 PAGESWAPPER - April 1986 - Volume 7 Number 9 A Concern about Security and DZ Ports A Concern about Security and DZ Ports by Jean Pollock and Jim Haskett Bloomington Academic Computing Services Indiana University Bloomington, IN 47405 The Problem DZ11 terminal ports are polled for dial-up and hang-up events, as a result of which VMS connects users to or disconnects them from interactive terminal sessions. When this polling window is too large, a new user can slip into an existing session following a non-standard disconnect by a previous user. Such a disconnect can occur for users connected to a VAX via a terminal network. When the window is too small, there is a significant increase in CPU overhead. The width of this window is determined by a VMS operating system parameter called TTY_SCANDELTA. The polling interval is currently 1/20 sec. on our VAXs. VMS requires TTY_SCANDELTA to be larger than 1/100 sec. The default value is 1 second. The Benchmark An attempt was made to determine the optimal setting for the polling interval which would keep the CPU overhead and the security window at a minimum. To this end, benchmarks were run under VMS 3.7 (on Node Orange) and VMS 4.1 (on Node White). Both are VAX 11/780s. All unnecessary processes were terminated while the benchmarks were run. A benchmark program was written in non-standard FORTRAN to start an interrupt timer and then increment a counter until the timer expired. The value of the counter was recorded as an indicator of the amount of overhead lost to polling. Thus, smaller values of the counter are expected for frequent polling (i.e. smaller polling windows). Different values were assigned to TTY_SCANDELTA for each run. Generally, for each value of TTY_SCANDELTA, the program was run three times for one minute each. The largest variation in the results obtained for each parameter value was approximately 0.1%. VAX-5 PAGESWAPPER - April 1986 - Volume 7 Number 9 A Concern about Security and DZ Ports The Recommendation The results of these benchmarks follow. As a result of these benchmarks, we have recommended reducing the polling time from .05 sec. (i.e. TTY_SCANDELTA = 500,000) to 0.03 sec. (TTY_SCANDELTA = 300,000) on our machines. This will result in an increase in CPU overhead of approximately 0.6% We would like to thank Dave Schwab for his AST code and John Stockton for his help in executing the benchmark. The Results Polling TTY_SCANDELTA Counter Value Per Cent Change Interval Orange in Counter Value VMS 3.7 .01 100,000 29,114,516 - .02 200,000 - .03 300,000 29,572,458 1.55 .04 400,000 - .05 500,000 29,659,082 0.29 0.10 sec. 1,000,000 29,734,726 0.25 Polling TTY_SCANDELTA Counter Value Per Cent Change Interval White in Counter Value VMS 4.1 .01 100,000 29,106,293 - .02 200,000 29,438,817 1.13 .03 300,000 29,559,069 0.41 .04 400,000 29,609,928 0.17 .05 500,000 29,634,821 0.08 0.10 sec. 1,000,000 - - VAX-6 PAGESWAPPER - April 1986 - Volume 7 Number 9 A Concern about Security and DZ Ports TTY_SCANDELTA Benchmark 29750000 + * | | 29700000 + | | 29650000 + * | # | 29600000 + # | | * 29550000 + # | | 29500000 + | C | o 29450000 + u | # n | t 29400000 + * Orange VMS 3.7 e | # White VMS 4.1 r | 29350000 + V | a | l 29300000 + u | e | 29250000 + | | 29200000 + | | 29150000 + | | * 29100000 + # | | 29050000 + | | 29000000 +---+---+---+---+---+---+---+---+---+---+ 0 1 2 3 4 5 6 7 8 9 10 TTY_SCAN_DELTA (X 100,000) VAX-7 PAGESWAPPER - April 1986 - Volume 7 Number 9 A User's Experience with Terminal Servers A User's Experience with Terminal Servers Stanley M. Rose Vice-President, Distributed Processing Technical Support Bankers Trust Company New York, New York The purpose of this article is to outline the experience of Bankers Trust with terminal server technology. This article is not intended to be a tutorial on Terminal Servers as many excellent articles have been written on that subject. After evaluating the DECSA, DECserver-100, and non-DEC units, the Bank chose the DECserver-100 and the decision has been substantiated by our subsequent experience. __________ Background Bankers Trust is a large user of DEC equipment. At present we have 18 PDP-11/70s (RSX11M-PLUS), and approximately 40 VAX systems, ranging from 3 MicroVAX-II to 7 VAX-8600. These systems have a terminal population of over 2500 terminals . Because of the criticality of the systems to our business most applications have a backup processor, which has, in turn, resulted in a good deal of terminal switching between primary and backup processors. In some cases there are secondary levels of backup, making the switching all the more complicated. The net result is that we had developed a very complex, and costly, investment in matrix switches, patch panels, etc. The complexity of the switching had become quite labor intensive as well. Compounding costs was the expense to install each point to point line. Due to the high costs of labor in New York it was not uncommon to have the costs of the line exceed the costs of the terminal, sometimes by a factor of two. To address these issues the Bank embraced the concept of Terminal Server based communications. After a detailed evaluation of several alternatives, which included installing and using foreign vendor equipment, as well as participating in Field Tests of the DECSA terminal server software and DECserver-100 products, the decision was made to use the DEC technologies. Subsequent work has further refined that decision to our usage of the DECserver product. VAX-8 PAGESWAPPER - April 1986 - Volume 7 Number 9 A User's Experience with Terminal Servers We felt that although, at times, other vendors might have had products that better handled certain niche situations, in the long run a better strategic decision was to use the DEC technology. We have found that the DEC products act as an extension of the system in a symbiotic relationship with VMS. Some of the advantages of the DEC products over the non-DEC units are: > No modifications to VMS, and no problems with lagging support for operating system upgrades. > Simple connection through standard DEUNA. > Easy installation of units. > Very easy network maintenance and configuration control. > Knowledge of VMS (now RSX), with the result that a server does automatic connection to the backup processor without the necessity of issuing any commands. > The software is centrally maintained on the VAX, rather than at each individual server. Some of the key features of both DEC units, that are now being used include: > Load balancing by automatic connection to least loaded processor. > Automatic connection to backup processor. > Preferred Service > Autoconnect > Multiple sessions Some specifics of each unit follow. _____ DECSA The first server that we received was the DECSA (or PLUTO) box based Ethernet terminal server. This unit supports up to 32 lines with modem control and reverse LAT. VAX-9 PAGESWAPPER - April 1986 - Volume 7 Number 9 A User's Experience with Terminal Servers Reverse LAT is the ability to use the server with a CPU that does not have native mode LAT support (such as from another vendor). Each port of the server is connected to individual asynchronous lines on the CPU. The terminal is connected to another port on the same or different server and can issue a connect to a group of ports which are assigned to a series of physical terminal server ports. Since we were only using the servers with VAX systems at the time, this was not a feature that we needed. For security reasons we have not installed dial-up access on the Ethernet, and therefore do not presently need the modem control feature either. Functionally, the DECSA supported all the features we needed, but was rather large for locating in a user area. The cost per port was higher than the cost on the DECserver-100, and since we did not need its added functionality we made the decision to standardize on the DECserver-100 instead. _____________ DECserver-100 As with the DECSA, the DECserver-100 has satisfied virtually all of our functional requirements. Installation has been very easy and takes no more than a few moments to set up the characteristics. The multi-system access has significantly reduced the overhead and labor intensive activities involved in system switching. One design feature that we find appropriate for our use is the 8 line size of the DECserver-100 . We have found this size to be a good trade off between economy of scale, system overhead, and minimization of single points of failure. The unit has been extremely reliable, and with close to 100 installed units we have experienced no more than 2 or 3 failures. Our maintenance strategy has been to purchase several spare units, and the minimum level of field service coverage. In the event of a failure, we simply swap in a spare unit and have field service replace the spare. Ethernet overhead is quite low, and with as many as 50 servers and 350 terminals on one cable, we have seen no perceptible impact on response time. The physical size and lack of special environmental requirements has made it easy to locate the DECserver-100 in office areas. Where we have had many units to co-locate, we have removed the outer plastic shell from around the inner metal case and shelf mounted them in standard EIA electronic racks along with DELNIs for concentration onto the Ethernet. VAX-10 PAGESWAPPER - April 1986 - Volume 7 Number 9 A User's Experience with Terminal Servers The Field Test of the DECserver based line printer support has progressed very smoothly and has provided an additional functionality that will be used extensively. All-in-all, we have been very happy with the units. ____ ____ Wish List No one is ever completely satisfied with what is available - give us a hand and we want an arm. There are several features we would like to see implemented in the future; some may be just software enhancements, while others will require new devices. The most important new software feature we need is host identification of the physical access port. This is a vital security feature to allow applications to be able to determine which physical terminal, at which server and port, is accessing the system. This has been the one feature we have lost in moving away from hard wired terminals. Another software feature we would like to have implemented is LAT support of multiple Ethernets to avoid a single cable failure affecting large numbers of terminals. We have developed a non-standard and, of course, unsupported patch to handle multiple LATs on several DEUNAs, but would like to have a supported method. A hardware feature that would complement the multiple Ethernet support would be a server that connected to two Ethernets, and did hot fail-over in the event of a cable failure. Note that we are addressing here the problems of cable failure, not server failure. Not actually a terminal server future, but one needed to support additional servers, is a high speed bridge between facilities. We have a need to locate the processors in one building, and the terminals in another - beyond the accessibility of fiber optic repeaters. As the number of devices increases, our need for more bandwidth grows with it. A device allowing us to connect remote Ethernets at T1 speeds is becoming more necessary. A capability that would off-load work from the central processors would be the implementation of the terminal driver in the server itself, especially on input. The READ QIO is actually a quite deterministic process, namely getting a set of characters until some condition is reached, and this could be moved out into the server. Once the terminal driver is implemented in the server, the various forms control packages, FMS and TDMS, should be implemented in the server as well. TDMS seems to lend itself quite well to this capability since it is full-screen oriented. VAX-11 PAGESWAPPER - April 1986 - Volume 7 Number 9 A User's Experience with Terminal Servers And, of course, we may find a use for reverse LAT and modem control and would like to have those implemented as well in the DECserver. _______ Summary We have installed approximately 100 DECserver-100s in the last year and have been extremely happy with the unit. We are continuing to migrate terminals from point-to-point line interfaces to DECserver-100s and expect to double the number of installed units in the next year. VAX-12 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session European DECUS Question and Answer Session Transcribed by Tony Arnold VMS SIG Chairman, DECUS UK, Ireland and Middle East University of Manchester, Computer Graphics Unit Oxford Road, Manchester M13 9PL, England Transcript of VAX SIG Q and A session held at the DECUS European Symposium at Cannes, France. September 1985. Chairman Good evening ladies and gentleman. My name is Alan Silverman and I foolishly volunteered last year to the System Improvement Request in Europe. This time last year I stood up and asked why we had sent out 2,000 forms last year and only had 76 replies. I complained, and said I thought you could do better. Well I should have kept my mouth shut because this year we got 766 replies and it took me a long time to count them! so, next year you will be asked to vote on the top 5, not the top 10. Okay. We did count all of them, and sent the top 5 to Digital. Andy Goldstein is here tonight to present the reply to the top 5. We will do them one at a time. The top one was number 30 on the list, which was split screen editing. This got 2402 votes including 53 first choices and was voted for by 370 people. Over to Andy. Andy I am pleased to report that we actually managed to get one right. Split screen editing is available in the new VMS editor VAX TPU and that is available in VMS 4.2 TPU is a completely programmable editor and it is intended to be used not so much as an editor by itself, because in fact it looks very much like a programming language rather than what you would view as an editor. Rather it is meant to form the basis of other editors. In fact, we already have several editors that are built on TPU and there are more coming. We provide 2 general purpose editing interfaces with TPU. There is an EDT emulator which is our way of eventually phasing out the existing EDT. We will provide the equivalent compatible EDT function through TPU. There is a new editor called EVE which was developed with extensive consulting with our human factors group and the EVE editor does support multiple window editing. It also supports arbitrary extensibility through TPU callouts and all kings of good stuff like that. EVE was designed specifically to be used in either a single editing or a multiple window editing environment. Any questions or comments? VAX-13 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session Question from audience. Presumably this is all going to be supported on VT100s and VT200s and all that good stuff. How are you coping with the fact that the VT100 is not really all that good at split screens and scrolling when you have got them mapped together? Andy Okay, well I haven't been directly involved with EVE, but basically, the VT100 does support split screen scrolling regions and about the best you can do is swap them around and just set up the area that you want to scroll for scrolling. I haven't actually used it myself, so I can't comment personally on the performance. Chairman The second most popular item was number 50, which was one that we have seen before, and that is on line disc compression. Andy SIR number 50 relates to on line disc compression and observes that the increasing use of Winchester discs, in particular single disc systems makes it increasingly impractical to reorganise your storage space with Backup, and of course that is also a problem in that you have to take down your system or at least your application to do that, and also as the discs get larger, again using tape as the intermediary storage becomes less practical because of the tape volume and simply because of the length of time it takes. So the request is for a utility that allows disc compression and free block consolidation and preferably on line as concurrently as possible with system activity. We agree that's a fine idea. We obviously haven't done it yet! This is a relatively new SIR request. This has not come up in past United States SIR ballots. We certainly intend to take a good hard look at this problem. I have some experience with it. I actually worked for RSX back in the old days when we had a gadget called DCU which some of you may remember. I think DCU was a very good demonstration to us all that this is not an easy problem. One of the interesting problems other than simply interpreting the file structure correctly is coordinating the compression with ongoing file activity, and of course clusters add a unique twist to that. Also the performance problem is quite difficult because inplace disc compression is to the best of my understanding an inherently n squared problem and so with some very large discs you run into some very serious performance problems. Obviously you can do intelligent things to make it run faster and we are going to have to spend some considerable time and attention in getting reasonable performance out of it. Any comments or questions? VAX-14 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session Chairman The next one was a request for the log out of inactive terminals. Andy The reason I got to do this session, by the way, was because I guess I ended up writing 4 out of 6 of the SIR responses. This has been a feature that has been requested for a long time. We have been thinking about it. One of the reasons we have not done it is because it is not quite as easy as it looks. At the same time it hasn't bubbled up as high on the priority list because it is the kind of thing that can be supplied by the installation. It is simple to build an extra watcher process that performs this function. Obviously we understand that you would rather see something like this provided by DEC and be general enough to handle everyones problems but that is exactly where the rub lies. The difficulty is coming up with a correct decision that a terminal is actually inactive. You might start out by saying that if there are no IOs over a certain period of time then that terminal is inactive, then you raise the question, what if there is some kind of long running job on the terminal in question? you can follow these through several levels. You can argue that maybe there is a server process connected to the terminal and in fact all of the active computation is being done in sub processes and you can dig yourself as deep as you like. I think one of the thing that we are looking for in terms of guidance, is what level of intelligence people feel is really necessary in such a facility, and I am certainly willing to entertain suggestions from the floor. Long silence. This is going to be a very boring session unless people get up and start saying something! What more can I say about this capability except we have had it in mind and as soon as we get out of the way what we consider some of the more pressing security features, we are going to look at it. It has appeared on the bottom end of some of our project plans in the past and so far it has always ended up below the cut off line. The fact that it has been specifically requested by a couple of SIR ballots should influence our decision on that. G. S. Bal, University of Liverpool VMS 4.1 We have the problem that the architecture of VMS does not have any timeout feature built into it. We have a network controlled by Gandalf PACX and the contention device. At times it is possible to switch the controller off or the terminal off. That switches the terminal port off on the contention device and leaves the whole port in. Anybody who comes in next gets into VAX-15 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session somebody else's session. Andy I would answer that by suggesting that the terminal lines that you have connected to the terminal switch are incorrectly configured in that you do not have modem control enabled. When a line has modem control enabled and goes into any kind of terminal switching state, VMS should treat that as a modem hang up and either log out the job or disconnect the process depending upon whether you have disconnectable terminals enabled or not. If we leave processes floating that someone else can accidentally reconnect to that is the result of the modem signals not being correctly handled. D. Bal This system did work on 3.7? Andy This is true on 3.7. Obviously 3.7 did not have disconnectable terminals but a properly configured terminal set up in 3.7 should disconnect the lines to the best of my knowledge. D. Bal DMF 32 interfaces have got modem signals on both 0 and 1 not on the 6 lines on the same board. How do I put the modem signals on those 6 lines? Andy The DMF supports modem control on port 0 and 1 only. If you use the other ports on a switched connection you are running at your own risk. D. Bal The risk is that this is an undergraduate teaching service. You know the challenge of the undergraduates to explore the system and wreck it if they can possibly do it, and it hampers our resources time and time again to sort the things out. Hence the requirement. Chairman The next one is batch checkpointing with 785 votes. Andy VAX-16 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session This SIR is a request to be able to stop and resume long running batch jobs in case the system needs to be shut down for whatever purpose and the considerations are obvious. When you have a batch job that runs for a number of hours or a number of days, perhaps even weeks, you cannot afford to lose that investment in time on that particular job if the system either crashes unexpectedly or needs to be shut down for any reason. We have been hearing about the desire for this ever since we announced the future existence of the VMS checkpointing facility which I think must be close to two years ago now. That's part of the data integrity features that we are working on for future VMS releases. We have had the checkpointing facility, that is what we have implemented on VMS, pretty wel ready to go for a while and it was a casualty of the deferment of the journaling facility. It turns out that checkpointing makes use of recovery units which in turn makes use of journals and as a result of that set of dependancies we have not been able to ship checkpointing yet. We do intend to ship checkpointing some time in the near future and it will be followed shortly thereafter by the checkpointing facility. Now I need to point out that the checkpointing facility that we intend to provide is not going to be as fully general as people would like. The reason is that saving and restoring the state of a job is a very difficult problem and the more transparent you try to be about it the more difficult it gets. We started with the level of dealing with assigned channels and open files and the address spece and things like that, that we felt were doable. However, when you get to the level of saving the complete DCL context when you have a job that has IOs that might be currently in progress and other side effects of that sort, things start to become extremely difficult and we found it necessary to defer that set of problems to another time. So there are a number of restrictions that will be in effect on the initial checkpointing capacity. The job must declare checkpoints by itself, you cannot from an operator terminal enter a command that causes the job to be checkpointed, the job must be checkpointed itself. It must request the checkpoint voluntarily. Secondly the job cannot run in the full DCL environment; it must be run as a separate process by itself. It is obviously not the best that people would like but it is the best that we felt we were able to build at this point and we are going to continue to look at this. Comments or questions? Chairman The last of our top five was the request for disc quotas by groups. VAX-17 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session Andy This SIR requests that quotas be allocatable by groups of users, that is, by the group field of UIC as well as by the individual UIC. The primary rationale of this appears to be that people want to allow disc space to be charged to, say, a project as a whole rather than to the individual members of that project, since presumably the file space in question is owned by the project at large rather than by the individual people. Now in V4 we have provided the identifier mechanism that allows the assignment of arbitrary project groups and arbitrary project membership, membership in multiple projects by one person and related features. One of the features that goes along with that mechanism is the ability to charge disc space to a project identifier. This means that you can assign an identifier to a person with resource priviledge and that allows that person to create files owned by that project identifier. To make use of that, then, what you would do is set up a project library directory where the directory file is owned by the project identifier. Access control lists make that directory and the files in it accessible to the project members and the file ownership defaulting rules in the file system will cause files that are created in that project directory to be owned by the project identifier, provided that the person creating the file holds the project identifier with the resource priviledge. I guess my question to you is: Did we do it right? Is this what you wanted? Questioner (no name). We are using this feature with partial satisfaction. It is fine as far as it goes, but if you have somebody using one of these identifiers with resource attribute, each individual user still needs a disc quota of their own. It is required by some of the utilities like the sort merge and certain Dataretrieve functions. So irrespective of the fact that you can have this group accounting, each individual still requires their own disc quota. Andy That's correct for things like temporary files, files that in effect belong to the individual user. So you are suggesting that you would also like just for simplicity of authorisation, to hand out the quota to the group rather than to the individual persons. Thank you, we'll take that back. New questioner. Yes, we made use of this to try and set up a project account. It had several side effects with some of which it didn't actually work. In the book, I can't remember which guide it was, it sets up the file ownership with the file ACL with "no VAX-18 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session propagate" so if you then edit your file it doesn't propagate the access control list and you lose touch with it. Andy Let me just explain for those of you who are not quite so familiar with the mechanism. In a circumstance like this, where the file system creates a file whose ownership is not the UIC of the person creating it, the file system automatically adds an access control list entry to the file that gives the creator access to the file. That is based on the principle that you really ought to have access to what you just created. That access control list entry is marked with a no propagate attribute so that it is not propagated to future versions of the file. I believe that is actually the right way because every time a new version of that file is created the access control list entry that grants access to the creator will in fact be added to it. What that means is that for the general project accessibility to the file you have to make sure that the file's access control list otherwise grants access to the project but if we did not mark the ACEs no propagate, what would happen is that as different individuals created different versions of that file, the ACEs would pile up. Every time somebody created a new version, we would pile up a new access control list entry that granted access to that new person and we felt that was just incorrect behaviour. So rather you get full access to the particular file instance that you created and you rely on the project access controls for the general access for the project. Questioner So you are actually saying that you can't edit the file. It means the user that created that file can't successfully edit that file. Andy No, that should not be. Questioner As he edits that file he creates the higher version of course. And it is that higher version that does not propagate the ACL. Andy That's right. It does not propagate the access control list entry that we put on it. Rather, it gets a new one, which is appropriate to the person that has created that new version. Questioner VAX-19 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session Then maybe there is a bug hanging around in 4.1 because it certainly does not progogate anything. It has the project identifier on the ACL but not the owner. Andy Okay, yes, something is wrong. There are some other wrinkles. If you are running with something like sysprv then also we don't put it on. The rules are somewhat complicated but they are complicated to work right in the greatest number of cases. New questioner I have installed this facility, too, and I have had a lot of problems. You have to be very careful if you edit this feature for files or directories which no longer exist, because if the users are allowed to edit under this new identifier, they have the files and if they first change the owner, they can do nothing with the file. So they first have to add the ACEs and then have to set the owner and if they don't do it this way they always come to me as system manager and say, "please give me access to my file". The other difficulty we had: the normal user who is allowed to access disc space via such an identifier cannot see the used disc quota of the identifier. So if the quota command says "no privilege" for a certain operation, the only way is to give all users access to quota so every user can see this quota for every other user. Andy That is an oversight which we should fix. New questioner I was going to ask the same point because I have found the same problem: you cannot find out the quota. But in addition, the previous comment about the propagating ownership. When the file is edited by different users with a version limit on the disc, say version limit of 5, when the sixth edit is done you can't get out of the edit because you can't delete the sixth file. Andy Okay yes, because you don't have permission to delete the file and .. I guess my answer to that is that the project access then should be giving you permission to delete the file. It simply means that you have to set up your project access control lists correctly for that to work. New questioner VAX-20 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session Is there any way of stopping that behaviour of the new access control list? Is there any way of preventing the access control list entry being added on creation of the file. Because it also gives control to the creator. We have applications where we would like a user to be able to create data files but not be able to delete them or change any attributes of the file. Is there any way of stopping them doing this? Andy At present, no, there isn't. We will have to take that back as a suggestion. I would suggest that if you want protection behaviour different from the default that you make use of the VMS services to make the file protection what you want it to be. There are any number of services available to you so that you can set the protection to what you feel is right and override whatever the file system does. Questioner It's a long way round, though, isn't it. Chairman Thank you Alan To all those of you who sent in responses to the SIRs, thank you very much. If you have ideas for new SIRs please send them to your chairman or to Peter, and we will try and do the same thing next year. I will send for publication the top 5 of the replies and list all of the votes to the Pageswapper so you should see that in a few months time. Thank you. Over to Peter. Peter Before you go Alan, we would like to thank you very much for your hard work. Now. The busy part of the evening. I am going to call out the names of the questioners. Christiane Letertre, CERN. VAX VMS 4.1 We have a problem with quorum disk in a cluster. Each time the system was rebooted we got the disc improperly dismounted. Is that normal or did we do something wrong? Peter This is something which a number of us have had. I get it with RM05s. I think it must be normal. VAX-21 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session Andy Do you have any kind of error messages that might indicate that the rebuild on the disc in question is failing? Christiane We do not. Apparently the disc is good. Andy Obviously the disc will get rebuilt and will require rebuild any time a node crashes in the cluster. Are you also seeing this when you reboot after a perfectly normal shutdown? Christiane When we reboot after a normal shutdown. Andy And you are running 4.1. It does not sound completely normal. I can't really say very much more about. It does sound as if something is out of phase in the volume storage control block. That is where we track the information about whether the volume needs to be rebuilt or not. We do that by keeping track of the count of the number of nodes that currently have the volume mounted and that is supposed to be reset by the rebuild function. If the rebuild function is not properly resetting that value then the volume will rebuild every time you mount it. About the best I can say is that we will have to take that back to see if we can identify any problems. Mr Dramer We have had the same problem, but we made a backup and cleaned the disc. We put 5 backed discs with BACKUP/IMAGE and we have not had any problems up to now. Andy That certainly suggests to me that there is something in the storage control block that is not properly being cleaned out by the rebuild. When you copy the disc with Backup the storage control block does get completely rebuilt, and so I guess I would suggest taking that approach and seeing if that clears the problem. That also gives us some additional information as to what might be wrong. Questioner VAX-22 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session I would like to add that it has nothing to do with clusters because we had the same thing on a perfectly normal RM03. New questioner Could I say that in the 4.2 documentation it indicates that at least some instances of this have been corrected. New questioner As a note I have found that if your quorum disc is not your system disc a normal system shutdown does not dismount it. That could be the cause of the problem. Andy Ordinarily simply marking the disc for dismount should be sufficient to clear the bits in the storage control block but again I think we will have to look at it. No 2 What could be recommended in between the time between now and the eventual standard utility that could become available to reorganise discs without backup-restore actions. Peter At present what we recommend is shutting down the use of that disc and doing an image backup, meaning either a disc to disc image copy or if you don't have a second disc to do an image save to tape and then stand-alone backup image restore from the tape back onto the disc. Questioner And if you want to provide a 24 hour service? Peter I have to admit that is not possible and we don't have the ability to do that at present. No. 3 Anders Lyngarth, TSL data, Sweden. MONITOR, SHOW DEBUG. VMS 4.1 I have a question about how to implement foreign terminals on the VAX VMS 4.1. I have tried it by way of using the module in sys examples called SCRFT.MAR which is the same module as the VMS 3.x. The module doesn't work as is stated. Where could one find information on implementing foreign terminals, if SCRFT.MAR is obsolete? Or is there an error in SCRSHR? It could be said, too, that the SMG$ routines don't work as the documentation VAX-23 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session suggests in defined foreign terminals. Peter The old SCRFT was always an undocumented unsupported interface and although in V4 there is an attempt to make it upward compatible I have heard at least a couple of reports that it is not 100% upward compatible. We do recommend that with V4 you use the SMG foreign terminal package. Anders I do that, but DEBUG doesn't use the SMG. Peter. Correct. It is going to take VMS a while to get its utilities using SMG. It will take us a while. We have had that requirement from US DEC users as well. You mention here that there are problems with the documentation. We are aware of several of those problems and there are a couple of bugs in 4.0 and possibly 4.1 that are being cleaned up in 4.2. Anders May I mention one thing. If you have another terminal with the line drawing character sets consisting of two characters you define the line drawing characters in the SMG document as strings which only yields the first character and that is not a string. What have you changed between V3 and V4 in the SCR interface? Peter. In the SCRFT? The old screen package, the SCR entry points still exist in V4 but the code that is behind them is completely different. Although the intent was that they be replacement routines, it was a total rewrite from MACRO to BLISS so things like counting on R6 having the right data structure pointing to it and things like that which people did do is not true when it was rewritten in BLISS and that was what gave most people problems. There was a data structure which was around most of the time which was what people used. Anders Okay, so I will close the openings for foreign terminals to work as display terminals on VAX for the moment. Until you come up with the utilities.. Peter VAX-24 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session Correct. Anders Do you know how long this will take? Peter No, it probably won't happen at the same time, we will probably move our applications to SMG as time allows. One thing that has been suggested is that an ambitious customer could replace SCRSHR with a SCRSHR which called SMG and via that route you could get things that called SCR really calling SMG then you would get the foreign terminal support. Another voice I believe DEBUG for 4.2 is now using SMG. You could talk to the people downstairs at the software house tomorrow morning to make that sure but I believe DEBUG for 4.2 has been converted. Anders I thought it a little bit funny because you said it was totally new but it didn't use SMG. Wasn't the DEBUG totally rewritten for V4. Answer Yes it was, but it was considerably rewritten before the SMG package existed. It was a timing problem. No 5 Harneit Jeus How can I have an SYLOGIN that is effective for users that log in both under the DCL CLI and MCR? If I am using a system wide log in procedure I cannot write it only for DCL. Andy What we normally recommend is that you define SYS$SYLOGIN to point to SYS$MANAGER:SYLOGIN and leave off the file type. What will then happen is the CLI will supply its own default file type for opening the command file, DCL will use .COM and MCR will use .CMD, and that you allows you to have separate login files for each one, and then you can either make them functionally the same, one in DCL, one in MCR, or you might have completely different ones, even an empty case if you didn't want it. No 7. A Marriott. Contraves Ag. Zurich VAX-25 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session How may a program at run time define a process permanent message file to supplement the existing system messages? That is to achieve the equivalent of the DCL "set message" command. TAPE ENDED HERE - ANSWER NOT ON THE TAPE. No. 8 Cino Crepaldi, Germany Under RSX there is a command $SET TERM/SLAVE. We have a couple of Perifold devices connected via terminal to VMS and we would like to break that protocol between username and user authorisation failure when you reboot the system. And you added index file support to DCL, however how can I delete a record in these index files? Peter The answer to the first problem is SET TERMINAL/NOTYPEAHEAD, one of the more intuitive DCL commands. The action of turning off typeahead disables the login. The second question was how can I delete an ISAM record from DCL. I don't remember the syntax. There is a way to do it. It may be something like a qualifier on the read command or something like, .comment from audience.. Peter It is WRITE/DELETE. I know it is there because we use it on some of our own DCL hacks and so it is a matter of perusing the manuals. The READ and WRITE commands are the only ones that DCL operates on at that level so it has got to be a qualifier to one of those two. No. 9 Drozak, Belgium. VMS 4.1 When you use the shut down command it asks you if you want to spin down the discs. If I answer yes to the question all my discs spin down except the system disc. Peter We can't spin down the system disc. The problem is a kind of "you can't get there from here" situation. The problem is that the system disc remains in use until very, very late in the shut down process. The very last phase of the shut down process involves making sure that all of the modified pages on the modified page list are flushed back to the backing files on the disc. This is particularly important when you have had user processes that have had files open as global sections or whatever. Obviously you want to make sure that the global sections are properly written back to the files. That is done in the very last phase of shut down by the OPCRASH utility VAX-26 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session program. Immediately after that OPCRASH does a bug check. We write out the system dump file and again you might argue that it is not necessary to write out a system dump file for an operator shutdown. On the other hand there is one piece of information that must be written out which is done using that mechanism and those are the error log buffers. Once we are into bug check we no longer have the regular system disc driver available. In fact, the bug check is even written out using the boot driver, which is what we booted the system with and the boot driver does not have the capability to spin down the disc, so - to make a long story short - by the time it is possible to spin down the system disc, we no longer have the ability to do so. Drozak Well can anybody help? Because I must stop my machine when there is nobody in the room and I am afraid to stop the power on the machine when the system disc is still running. No 10 (no name) One of my users used EDT to edit a directory file and EDT got confused and the directory got corrupted so I have a directory file that I cannot delete. Peter Despite the documentation, this is the purpose of the SET FILE/NODIRECTORY command. This command was specifically put in place to allow you to delete a corrupted directory and you issue that command naming the bad directory file. That clears the bit in the file header that identifies that file as a directory. Subsequently you can delete it and then reclaim the files under the directory with ANALYZE/DISC. The documentation for SET FILE/NODIRECTORY has been fixed in 4.2. No 13 W P Ingenegeren. Utrecht. VAX/VMS 4.0 This has happened to me today, I demonstrated it. Some people are told they have new mail when they log on but there is no mail. It looked almost as though it was counting pages and decrementing files. I have been receiving these questions, mainly SIRs, from the VAX info file and I have been getting... when I finally finished I think I had six more mail messages but no mail. Andy I can't point to the specific event that might have caused this problem. This problem can result because the new mail count is not stored in the mail file. In VMS V3 it was stored as a field VAX-27 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session in the authorisation record. In VMS V4 there is a separate file called VMSMAIL.DAT in SYS$SYSTEM which is MAIL's general purpose database file. The new mail counts for all of the users are stored in that file. Obviously if that file gets out of sync, somehow with an individual's mail file (which might happen, for example, if somebody changes system discs or boots the system off a different system root at some point) then the mail count will get out of phase. The way that you reset the mail count is to do a READ/NEW when there are no new messages and that will force the mail count back to zero and get it back in phase. NO 12 Mats Jansson, Sweden I have tried to make possible to print out in both 6 and 8 lines per inch through DEFINE/FORM with set up file that does the necessary things with the printer. I have found that it was not possible to define the reset module to be printed out at the end of the printing at the form. It was necessary to do this at the queue. I think it is not what you always want, a blank page at the end of every file printed. I want to have a switch on the DEFINE/FORM to make it possible to just after this file with this form to reset the printer. Is this planned or must I reset after every file in the future? Peter You are sending a setup module? Mats Yes. On V4.1. Peter And what is the content of this setup module? Is there an escape sequence in here? Mats Yes, defining a number of lines on the printer. Peter Is there any text in that set up module? Mats No. Peter VAX-28 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session Just an escape sequence? Mats Yes. Peter I believe this problem is fixed in 4.2. We have a problem in which the symbiont should recognize the fact that it is receiving strictly a setup sequence and there is no text to be displayed on the paper and it is still saying, "I have received more than zero characters therefore I need to issue a form feed to get to the top of the form". No. 13 Sally Antill, Vickers Design. U.K. VMS 4.0 Do DEC really believe that it is acceptable to remove commands and system services in a new version? And to change commands such that command files, etc., uddenly give errors? Andy I would like to see a more documented example of what the problem is. Voice She mentioned this in her session. One of the examples was in the MAIL utility where the EXTRACT command replaced the FILE command. She used the FILE to extract mail now it is EXTRACT to file. Andy Yes, we changed the command in mail. Certainly in DCL in system services I believe we have been as careful as we can possibly be to preserve compatibility between releases. Comment With the exception of the INITIALIZE/QUEUE command where you change the /PRIORITY to a /BASE_PRIORITY. It makes no sense. Andy The fellow who did that is no longer with us... (LAUGHTER) Peter With regard to /PRIORITY and /BASE_PRIORITY I think the rationale for making that change was that in V3 the /PRIORITY qualifier was used to mean two distinct things: the relative VAX-29 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session schedule and priority of a job within a queue and also the execution priority of that job. We got a lot of SPRs and questions at DECUS sessions where people were getting that concept confused, and I think that was one of the motivations for making that change. I can't read the guy's mind.. Andy Every once in a while we do make a change in behaviour. MAIL for example: I use the excuse that we really were presenting a significantly different MAIL version than we had in V3. You can either accept that or not as an excuse. In other circumstances, we have made some changes and they have often been by substantial popular request. One of the ones that I recall was changing the default setting of /REWIND in Backup. I think from the general level of feedback that we got, I believe that the change was right. The level of complaints when we made the change was actually relatively low. I was pleasantly surprised. Comment May I give a comment on the MAIL thing we just discussed? In V3 VMS when you use the FILE command you create a text file which is in fact a mail file with a form feed at the beginning. So if you wish to print it you just print it. But if you use the file command with the same name it in fact creates a folder and a lot of people using V3 do not know this. So if you type "FILE", then message 1, 2 and 3 with the same file name, you will notice you don't have the message file created because all these messages are added one after the other and in fact the file command creates a folder. With V4 VMS it also creates a folder. Andy Frankly my comment is that I wish the person who implemented the MAIL facility were here to speak for it. Bill tells me that he (Ben Schreiber) will be here tomorrow morning. Comment All the layered product people - they had a Q&A already today - if questions come up during the VMS Q&A for the layered product people, of which Ben Schreiber is now one, they can be directed tomorrow on the demo floor. No 14 Qar 357 Since we run a fairly open shop (being a large international scientific lab), we decided not to implement password lifetimes, so these are set to zero in the UAF. Recently, however, a neighbouring system had a malicious break-in on an account where username=password. So we decided to issue all future accounts with pre-expired passwords. We found that indeed all new VAX-30 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session accounts were marked as pre-expired but VMS did not care. No warning message on login, no requirement to change the password, nothing. Only when we made PWDLIFE non zero (we chose 3650 days) did pre-expiry do what the documentation says it should do. Either this is a bug or the documentation should be changed to point out that the pre-expiry flag depends on finding a non-zero PWDLIFE. Andy That is a cute little wrinkle that had not occurred to us and we will take it back. NO. L5 Eva Larsson, University of Stockholm VMS 4.1 The problem I seem to have with my VAX 785 might be very site specific but may be interesting if someone here has come across it before. As I said I have a 785 which is equipped with one Unibus, 8 DZ11 and 96 Emulex DMF32. This worked fine under V3.7 but now on V4.1 there seems to be a problem with hanging. If there are more than 35 users on the system, eventually the processes will hang one after the other until the entire system is hung. You can't get through the Emulexes, nor even the console. The only way to manage this is with control-P and rebooting the system. This is very irritating as it happens something like 3 times a day. I have heard at a computer conference system that there is a machine in Germany and another in U.S. which seem to have the same problem. At least they have similar symptoms. Peter I think every system in the U.S. that had either Emulex or Able ports started hanging when V4 came out. In V4 we enabled a feature that DMF had but we had never turned on in V3 called auto xon/xoff which allows the hardware to do the xon handling. DMF apparently had had the bugs fixed by the time we got to use it, and Emulex and Able had never had that feature turned on and hangs occur with that situation. It is my understanding that both of the companies have been very good about giving out new microcode to people who had those boards and you should probably contact your company. Another work-around that we don't have a whole lot of faith in, but seems to work (we don't have Emulex boards, that is why I seem to hesitate), is there is a special SYSGEN parameter, TTYDEFPORT, not DEFPROT which is default protection but DEFPORT which is a port control word. If you set the low bit of that... Eva We were aware of this before installing V4.1 so that is not a problem. VAX-31 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session Peter You still get hangs. I guess take a crash and send us dump. Eva I have spoken to the entire system department in DEC Stockholm and they seem just as puzzled as I am. The problem is new even to you? Andy In the problem description you said you find a long XQP queue, meaning which queue? Are you talking about a request queue inside a particular process? Eva Yes. I am not very good at VMS internals but if you look at the system dump analyser at the crash dump you can see that the request IO queue is empty and the request XQP queue is normally empty as well but when you force the crash you get quite an amount of requests queued up in the XQP queue. Andy When I read the problem description I started thinking of something altogether different. We have noticed that a good way to get the entire system bottled up is to have the disc driver subsist and lose an IO on you. When this happens, chances are very good that it is a file system IO. What happens then is that the process who is executing the file function that is responsible for that IO obviously hangs in the middle of whatever they are doing. They are presumably holding some set of file system synchronisation locks. Any other process which subsequently comes through a similar path in the file system will then get hung behind that lock and, for example, if that is the volume allocation lock you are guaranteed to bottle up the system fairly rapidly. Now the way you would backtrace that is to look at all of the processes making no progress. Do a SHOW PROCESS/LOCK on a couple of those to identify what they are waiting for. The file system locks are all identified by a resource name beginning with F11B. Take one of those locks, do a SHOW RESOURCE/LOCK=, giving the lock id. That will tell you who is holding the lock. Then you go to the process who is holding the lock, again do a SHOW PROCESS/LOCK on that, to establish what that process is waiting for. You may have to trace through several levels of locking dependencies. Ultimately you will find a process who has no lock dependencies who is waiting for some other reasons. One of the things to look for here is a busy IO channel. See if there is an IO outstanding, then look at the device to which the busy channel points and see if you can find either the stray or stopped IO VAX-32 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session packet. If you want to go to that level of chasing, that is what you can do. Alternatively if there is a stray undiagnosable hang, it should come to us with an SPR. Peter When your system is in this state do your terminals echo control O, output on, output off? Does that work? Eva No they don't. Peter That happens at driver IPL. Sounds like hardware to me, sounds like it is the Unibus. If the system is hung in the way that Andy is describing it is going to be looping at a lower IPL and the terminals will be able to do output on, output off. If they can't do that your system is sick. Comment I have had the same situation. I switched off the Unibus and it didn't help anything. I have Emulex only on one Unibus and I switched it off. Bill Travers told me it would help but it didn't. It must be something different. Comment We also have this beast called Camak. It causes us many problems. This is a SYSGEN configuration map of the Unibus of an 8600 with Camak on it and DMZ lines. What we don't understand is that at TXB it jumps up from vector 320 to 450. Then it carries on up to 600 where it overflows onto the Camak and gives us a hardware problem. What I would like to know is: How does autoconfigure work and could anyone explain why it suddenly jumps from 320 to 450? Peter Your question is: Why do the interrupt vectors go from 320 to 450. Well you got me. I don't know (much laughter). I wrote the code (more laughter). If you configure your system like that, does it ever configure? Questioner This is autoconfigure. Peter VAX-33 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session Is that a show configuration or is that the result of...? Questioner That's AUTOCONFIGURE ALL. I looked at the drivers manual and there were some very strange devices which had very peculiar names with vectors in that area so I did exclude these devices and it made no difference. Where can I look? Do you test the vectors? Peter Everybody in VMS is on SYSGEN some time or another. We test CSRs. Vectors just kind of follow along merrily. The computing of the interrupt vectors is really quite simple. It's just is there a device here...no... Questioner That could be a clue because this Camak thing has a large number of CSRs. Peter That's the little CAA device down at the bottom. Get rid of it! (much laughter) There are a couple of things you can do when you are debugging these configurations. There is a SHOW/UNIBUS command. Do it on hard copy because it is going to generate a lot of output. It gives a complete map of what AUTOCONFIGURE sees and how it makes its decisions. Also when you do AUTOCONFIGURE, you can do AUTOCONFIGURE/LOG and if you see any strange devices pop up..5 line printers, its devices found that if it doesn't find a driver, if it is not a supported driver it doesn't complain. Questioner We actually figured that out and we do see 5 line printers so we exclude line printer. Peter You have got things set up in such a way that you are confusing autoconfigure. Questioner What devices have you got in that space, between 320 and 450. VAX-34 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session Peter It is not the vectors that are doing that. Andy I think there is a jump in the CSR space there between the two. What AUTOCONFIGURE is doing is seeing CSRs there and thinking there is some line printer device, and it is also allocating vector space as it is allocating those devices. It goes to load the driver, finds there is no driver, this is an unsupported device, just keep going. Questioner So I actually have to configure by hand every device. Tony A follow up to this very problem. We have a KMS11-PX, the single line version of the KMS 11 and if you do an AUTOCONFIGURE ALL it very intelligently loads the XS driver whereas PSI requires the XN driver. So I did what I did on V3 which had the same problem and I did AUTOCONFIGURE ALL/EXCLUDE=XS. It then loaded the XN driver. At that point I gave up excluding devices and I had to do an AUTOCONFIGURE ALL and select each individual device which seems to defeat the purpose of AUTOCONFIGURE ALL. Andy The algorithm that SHOW/UNIBUS uses, it sniffs around through Unibus IO space and does a test word on a particular location and if it finds something there, if memory responds, then it says ah ha, you're a DZ, because it knew what it was looking for at a particular location. If your special device is in a range of CSRs that AUTOCONFIGURE looks at then it is going to confuse AUTOCONFIGURE. There are user reserved CSR locations which are up above the line printer locations. No 17 Jan Schoubo, Denmark I should like to modify the print symbiont with the new PSM$PRINT routines and write an input field in Pascal. I have not yet met anyone who has actually done it although Larry Kilgallen says he has some people working on it now. Are there any reasons why it should not work, writing this program in Pascal, rather than the example in the manual which is in Macro and doesn't work. Answer VAX-35 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session I know of customers that have written a modified symbiont in Basic and FORTRAN but I don't know what the problems might be in Pascal. Jan Could this end with the suggestion at least that a better example might be in high level language next time. Answer Actually that is on our work program to do that, to give a high level language example. You are right that the Macro example does not work and we are trying to get that out in the next update. Jan Has anyone in the audience modified symbiont? No 18 Now that the VAX has moved into the mainframe league with clusters and 8600s which require somebody to do a system manager function, it ought to be possible for this system manager to manage mag tape resource allocation. The currently available tools - REQUEST, REPLY, MOUNT and ALLOCATE - do not provide this sort of function in the way that most people want to do in a large tape shop. Andy We are well aware of the shortcomings and are in the process of working up a design for a tape management facility and the like, so we intend to provide that kind of capability in future versions of VMS. I can't make any promises yet as to how long it is going to take us. NO 19 Niall Mansfield, Heidelberg VMS 4.0 In some applications and in languages like Pascal and C, but especially in C, you would like to write a primitive that deals with relative files, you would like to say, "Get me block N", or something like that. The SYS$SPACE primitive gets you a block relative to where you are now and the question is is there any way to get a relative block from the beginning of a file without closing the file and reopening it? In other words you don't want to have some global datastructure telling you where you were because you are going to call this primitive from all over your program. VAX-36 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session Andy You want to simply present a record number to RMS and read that file. If I remember my RMS correctly, you specify key to access and give the relative record number as the key. The capability is definitely there. Try and catch me tomorrow and we can go through the manual. When I use RMS I usually go to the manual. No 20. Mats Jansson, Sweden We now have the restart feature on batch jobs. Would it be possible to write a program in a way to make it restartable if you have a very large computation? Peter We agree that is a good idea and we will take that back with us. DCL does communicate that information, translates the restart value call into a send to JBC system service call so you should be able to use that supported interface. N0 21 Sally Anthill, Vickers WPS/+ utility Why are all the keys in different places to KED & EDT? Are there any plans to fix this? If not, why not? Peter The KED and EDT were built with different targets in mind and really what we are promoting is TPU as the new editor and TPU has the capability of remapping the keys. So does EDT for that matter, so they could set up EDT to look like KED. Is the person who asked the question in the audience? No 22. Mike Waters Rutherford Lab, UK Identifiers again. I have set up lots of identifiers as they come out of authorisation, two in particular: one called AC750E which is set to [1,*] and identifier 750E which doesn't have the AC in front of it, set to [1,45], or [45,45] and then trying to RUN/UIC=[AC750E,750E] and it fails and CREATE/OWNER=[AC750E,750E] also fails trying to parse that string. If you use that string on something like DISKQUOTA it works fine. It seems to be things with an identifier starting with a numeric character. It is not clear whether it doesn't always fail. Andy Well we will have to take that back. I think I can tell you at least where the problem lies. The problem is that the identifiers and the symbolic UIC expressions are parsed in a couple of different places. The place that is accepting it VAX-37 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session properly, that is allowing you to get it through DISKQUOTA, is the $ASCTOID security service. The place that is not working is the command line parsing in DCL which does not apparently accept all of the identifiers that it should. I will take that back. No 23. G P Kristjansson, Iceland, RMS utility My programmers claim that there is no way to scroll or step backwards in a keyed RMS file. They make a mirror key, they take keys and convert them so when they want to scroll backwards the program always looks at the mirror key. But it is no solution for my programers - they are very heavily loaded. Andy The answer to that is that we are working on that feature. In case some of you didn't hear the requested feature is to be able to read in opposite key order in an RMS ISAM file. We are working on that feature and expect to have that available in a future VMS release. It is not in 4.2. No 24. Peter Holderman Written software can make use of the system communication service for a high speed task to task communication inside the VAX cluster. Andy At some point in the future it may be possible. At present it isn't, for a couple of reasons. First of all we do not provide driver access to the system communications layer. Secondly we do not even document it at present and the reason is that the SCS protocol at the moment is a very new design. We do not yet feel comfortable at the moment with its stability. We suspect that the SCS protocol is going to change at some point in the future as we discover new requirements. In addition there are some interesting security problems. The SCS protocol is a system internal protocol and as such has no security controls on it. Obviously that may not be a concern for a real time application. Questioner This is not a problem for a communication class driver. For instance for a standard conventional driver there is no security at all. Andy The point is that certainly we would feel uncomfortable at providing user QIO access to the SCS protocol. The difference between that and writing a user driver is of course that the VAX-38 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session user driver is of course installed by the system manager. That is somewhat different than our just providing access to this internal layer of the system. We have had a number of requests to make SCS public and so we understand there is a lot of desire for it. At present you might consider seeing what you can get out of the CI DECnet, whether its performance would be adequate for your needs. No 25 Daniele Cuzin, Schulberger EPS, VMS 4.1, 4.2, MSCP utility We have a problem on a cluster of four machines with local discs. Two of the machines still have local discs and we have not found the right way to boot those discs. In fact when we boot all the cluster the local discs boot on their own machine and all the other machines but if one of the machines which has a local disc on crashes or is stopped then when we boot that machine the local discs on that machine are not recognised by the other machines. This is a problem because we have users on both machines on those discs and the only solution we have is to mount the discs manually. Is there another way? Andy I do not completely understand the problem but from my experience with clusters I would guess that what is happening to you is that on the other systems that are making use of this local disc that is served by the system that has gone down, is that those other machines are timing out mount verification. Mount verificiation is the process that holds IOs to a disc base temporarily unavailable until that disc becomes available again because, if disc IOs were held, any lock dependencies that potentially pile up behind those can also cause other processes to hang and, in general, if they are not resolved in some way can end up with significant parts of the system bottled up. We do have a time out on mount verification and I forget exactly what the default is but it is somewhere between 5 and 15 minutes. If that time out period is shorter than the time it takes for the system serving those discs to reboot and get those discs back in service, then mount verification will time out, the IOs will be failed and the only way to recover the disc is in fact to dismount and remount it. So to prevent that problem what you have to do is set the sysgen parameter MVTIMEOUT to be greater than the length of time it takes one of the systems in the cluster to reboot - greater plus some safety margin. No 26 I have got a distribution medium from decatel with a recoverable read error on it but VMS install aborted installation. Is it necessary to abort installation if it is a recoverable read error? VAX-39 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session Andy I would call that a bug in VMSINSTAL and I would take that back. VMSINSTAL intercepts the messages that come out of Backup in the process of restoring the product save sets. It does that so that it can detect certain specific conditions and from what you say it sounds to me like it is not properly handling those messages coming out of Backup. I agree that if the error is recoverable it should allow you to install the kit. No 27 h. Kennedy, Oadweald Ltd VMS 4.1 I was looking at some of the code on security services data structures and on the terminal driver came up with the name CIA$. That was compound intrusion analysis. I am curious to find out where KGB$ comes from and where NSA$ comes from. Andy You are wondering what possible justification did we have for picking those acronyms. KGB is the structure prefix name used for the rights database records and it stands for "key grant block". NSA stands for "non discretionary security audit". The best one of all, which you may or may not have stumbled across is the break-in detection list that is maintained by login, those blocks are called CIA which stands for "compound intrusion analysis". Questioner Thank you very much, at least someones got a sense of humour! Andy Thank you! We haven't quite figured out yet what to do with OSS and NKVD. No 28 Kurt Jensberg, Switzerland We used on VMS V3 the utility "wat". I know that is not from DEC, but from DECUS library probably. Is there anything similar available on V4. Andy To the best of my knowledge there are versions of WAT floating around the dec engineering net that has been pretty thoroughly upgraded to V4 so my guess is that the fellow that wrote it has it running on V4 and just needs to be motivated to release the new version to DECUS. The guys name is Stanley Rabinowitz. Perhaps write to Stan through DECUS and see if you can persuade him to ship a new version. It is a DECUS item and not a supported part of VMS. VAX-40 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session Kurt Are there are plans to integrate that into VMS and make it available to everybody. We found it very useful, especially in real time application and when you want to look at different parameters from different processes it helped a lot and we had problems for example with ast limits that were too low and processes just hung and with this utility at least we found the problems. Andy We have no plans at the moment to incorporate into VMS, certainly it is worth taking back as a suggestion, either that directly or similar capabilities. We would want to look at what that is providing that some of our existing facilities are not. Bob Real, U.S. Air Force in Germany I can answer that for you. DEC Munich has taken the old V3 of WAT and completely modernised it to completely support on V4. It is offered through Franfurt free of charge, you can get in Franfurt or Munich. It has complete window encapability, along with supporting of just about every function you have under the normal SYSUAF. You can display interactively on the new WAT. No 29 Daniel Cuzin TA78 utility on cluster We have upgraded a TU78 to a TA78 and we have discovered that some IBM sysmic tape cannot be read on the TA but could be read on the TU and apparently the application program which reads the tape uses a code on the TU and we do not have any documentation about the TA. Comment I think we can help you, we have the same problems. Some active drivers put some garbage on the beginning of the tape. You don't discover it, using Digital tapes but if you write the tape on a Digital mag tape station and go to the IBM, then the IBM says thats garbage and skips it. Comment There's another possibility and that's the famous byte swap problem. The TU78 drives post byte swap, the TA controller and TMSCP protocol do not support byte swap. Daniele So that means we have to write a program to do the byte swap. Why is there not a program. VAX-41 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session Answer As for the documentation I cannot answer that. The TMSCP protocol which is supported on HSC controllers, does not support byte swap, I mean, why not. Peter We can take that back for documentation. Daniele I have another question. Is it possible to dual port a TA78 on 2 HSC50. Andy At present VMS will support static dual porting of the TA78 meaning that you can hook it up dual ported to the 2 HSCs but we will handle the tape drive correctly only if you have only one of the ports enabled at any one time. VMS does not yet support the proper coordination of the two access paths, so if you are willing to use it in a static sense and just change the port switches manually, should one HSC go down and use it on the other, then you can use it that way. Do not use it with the ports set up in both port positions because in effect VMS will then see two ta78s in its IO data base, not recognize that they are the same and you will have resulting confusion. Daniele Do you plan to enable VMS to be able to support two. Andy It has been suggested, it is one of the things we are considering in the future. Voice ?chairman Two things. First of all we have a TA driver dual ported to 2 HSCs and it doesn't seem to cause problems. But it is static. The other thing was, you said you had taken this back to documentation. We would prefer you took it back to your developers and changed TMSCP to support byte swapping. We have a requirement for that. No 31 Johan Hamaker, SRZM, Netherlands VMS 4.1 I frequently use spawn/nowait/notify/output as some log file to do for instance little compilation jobs and that sort of thing. Since we have started with V4 I find these jobs die as if they have had a heart attack! At unpredictable moments. When you VAX-42 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session look at the log file you sometimes see that the compilation has been done but then after compilation has been stopped, another time you will see that it has done a compilation and a library replacement. Then it has stopped. Sometimes it will run right through, sometimes it won't do anything at all. Now somebody suggested that there might be a problem there with pooled file quota or something like that. That does not sound unreasonable except in that case I would expect to find an error message in the log file and I don't. Andy You are saying that there is absolutely no message in the log file at all; it just arbitrarily stops. We have not made any changes in the way, for example, file quotas or log quotas or anything like that are handled. On the other hand with some of the changes in RMS the same file operations in V4 require a greater enqueue limit. It is conceivable that having gone to V4 your job is running out on enqueue limit or something like that and for whatever obscure reason the error is not being reported... Voice Could he be running into the control Y problem being delivered into the subprocess and then they go away and if you don't know that happens you have to say input=LAN0, otherwise you spawn and you are doing something else and you completely forget this if you are used to V3 and you hit control y and the sub process is blown away and you never know it? Or am I misremembering something? Andy I wasn't sufficiently familiar enough with spawn to realise that was even a problem. But that is an interesting point. Johan But wouldn't that problem have appeared on V3 as well. Answer No, we wrote the way control-Y asts are delivered to avoid that problem actually. There were a good deal of trouble in V3 with out-of-band characters and spawn processes and we did a good deal of work to clean that up. It may be that there is a hole there. I suspect if there is that the DCL person has gotten complaints and by 4.2 it will be fixed. If you still see this in 4.2, try and get a little more data together to send us. VAX-43 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session Johan I would have been willing to send an SPR but I would have no idea what supporting documentation to send. No 33 Jaume Sole, G.D.S., Barcelona VMS 4.1 It's a dual system, 2 VAXes 2 750s connected with Ethernet. System hangs for 2 to 3 minutes. After doing nothing, activity continues normally. During the hang there is no control c, no control y response, no echoes prompts, no user name prompts, no spool activity, typeahead does not work meaning the characters are lost. Andy The fact that the typeahead characters are being lost means that the system is going into a loop at rather high IPL. Jake suggests that it might be a false power fail problem, something to that effect. It is conceivable that the Ethernet interface is misbehaving and is causing the driver to become sick which in turn might cause a lot of execution at driver IPL which is the kind of thing which would block this activity. To analyse this sort of problem, we really have to catch the system in the act, force a crash and then analyse the crash dump. The other thing to look at is, if possible, does the problem persist while the net is shut down, something like that, see if you can isolate the cause by isolating components. No 34 (no name) In VMS 4.1 there was a patch to COBRTL.EXE which I believe is part of VMS not COBOL that fixed an apparent buffer overflow problem in display verb. The fix increased the buffer allocation from approximately 1k to 64k bytes on the user stack. I have an application that multithreads with a p user stack stuffed in a global section. The problem is that my TSC say that this fix is not a problem. I find 64k of stack for an output buffer is a little bit excessive. I think it's ECO no 1. Incidentally perhaps the guy who wrote the patch should have started out on PDP11 then he would appreciate what 64k bytes were. Andy Unfortunately we have no representatives here from the organisation responsible for either COBOL or the run time library, so I think the best we can do is to take your comment and basically ask them what's all this then. VAX-44 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session No 34 Just as a side point on the same matter, is there any other thing that any of the panel know of that is COBOL from a user program which will allocate an excessive amount of stack. Andy There is none that I know of but then I am not familiar with most portions of the run time library in general. The code in the base VMS system and most of the base library modules allocate stack that pretty much amounts to what they really need. No 35 (no name) A machine having a UDA50 and a Versatec 121 DMA interface on the same Unibus regularly clashes with all sorts of different bugchecks. Field service say this can be caused by an incorrect delay setting on the UDA. Have we got a hardware man that can tell us what a delay setting could be doing. Panel member My first answer! Depending on the memory band within the buffering characteristics of the device, if the device under question does not have sufficient internal buffering it is possible for UDA to dominate the IO bandwidth of Unibus. There is a setting on the UDA which is hardware settable to delay the time that the UDA will vie for arbitration of the Unibus. By default I believe it is set to 6.2 microseconds which means that it will pause for that length of time before arbitrating again to gain control of the bus. It could be that the other optional settings are zero microseconds and 10 microseconds. I think it is worthwhile for the hardware service representative to have a look at that and make sure there is sufficient bandwidth left for the other devices on the Unibus. No 36 Peter Humble, Leicester University From time to time we get problems with an interactive user running a job which really ought to be a batch job and is just sitting there in compute for hours and hours on end, somewhat hogging the machine. With the exception of restricting the CPU time available to anybody, in the SYSUAF for example, or spending some time whenever we see it has happened to try and find out who has done it and sort something out personally, is there any mechanism to make it difficult for them to do this sort of thing, and if not, would it be a good idea if something was in there, for example some sort of dynamic lowering of base priority for people who didn't do any io to a terminal from an interactive job. VAX-45 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session Peter Well the system does have some dynamic lowering of the priority in that when you do IO you in fact get a priority boost. Now obviously whether this is a serious problem or not depends on how you have your batch keys configured. If you are running your batch queues at a lower priority, let us say priority 3, then clearly doing heavy compute down work for an interactive session would constitute abuse of the system because it would seriously interfere with the batch jobs. I think the real solution for this requires a class scheduler which is something that would actually allow you to allocate CPU time in a proportional way between the service classes. That would in fact allow you to make sure that an interactive user could not get more than a certain share of the CPU. We are looking at that kind of scheduler for future VMS versions. At present what you can do is put a CPU time limit into the users UAF record which limits the amount which he can accomplish in one login session and then override that with a larger CPU limit on the batch queue, so that jobs requiring larger amounts of CPU must then be run through the batch queue. Of course the other thing that you can do is filter the accounting records, looking for interactive logouts that have accumulated a large amount of CPU time which is what you referred to in terms of chasing after the matter personally. Comment It has been noted to me that some people put ACLs on certain images so that they only can be run from batch so if it is running the compiler for example, and you don't want compilers run interactively, you could put an ACL on the image for the compiler. Questioner Unfortunately it is nearly always the users own application. No 37 Ugor Ergun, Turkey My company is the representative of DEC in Turkey. One of our customers wants to use C for their transaction processing. Is it easy (as in COBOL) to handle RMS organisations in C. How does C compare with COBOL. Andy It is not that difficult to do RMS from C and the documentation is pretty good. C will work in the transaction processing product that we support acms so I think all that is possible and not that difficult. VAX-46 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session No 38 (no name) How is it possible to make printers (connected via serial tty lines) available from other VAX systems connected via Ethernet? Answer I think all I can say on this question is that we have some work going on to potentially extend the capability of the lab server to support remote printers, such that a connection could be established from that host but that support is not here currently. No 39 Terse Grevle, Norway VMS 4 When will the function 'delete to end of line' be implemented (or is it already)? Peter Probably never. It was not a matter of being so much code or something like that but the work that we did with our line editing staff was the result of a human factor study done within Digital for what novice users need to get by with the system editing functions. All of the things that we did were based on that study. Although it is not as powerful as EDT there is not as much code there. We think we have a good balance between usability and the functions that we provide so we do not plan at this time to add any more functions. Andy One of the things that I would like to add is that one of the results that came out of that human factor study is that one of the important aspects of the command line editing was that it be simple and therefore it was important that we only provide the minimum set of functions necessary and not clutter up the keyboad with a lot of fancy editing functions that people would not use very often. No 40 This may be a hardware problem but I will go anyway. I have an VT200 which I use in VT100 mode because thats all my iden will support and I have to login into some DEC systems which do a SET TERMINAL/INQUIRE. And SET TERMINAL/INQUIRE on a VT200 in VT100 mode sets the terminal to VT200 and my terminal starts speaking me in Swedish. I have enough problems in English, I can't answer in Swedish. If the VT200 is set to VT52 mode and you do SET TERMINAL/INQUIRE it stays in VT52 mode. Where is the bug. VAX-47 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session Peter The VT200 tells us even when it is in VT100 mode that it is a VT200 and so VMS says ok, you are and if it is in VT100 mode becomes a VT200. We do that because it is the only way that the characters that SET TERMINAL/INQUIRE sets up can be truthful, it tells us that it has soft characters and various things and unless we force it to VT200 mode we can't really guarantee applications that those characteristics are there. We complained to terminals engineering about the design and they didn't listen to me and the VMS group. The customers complain and they listen to the customers, so you can get a v2 of your VT200 microcode which will allow you to change the set up to answer back VT100 when it is in VT100 mode. So contact your local office. Comment Do not change our terminals because they turn into English! (Swedish gentleman). No 41 Jean Yves Le Poec Have you got any plan to support Ethernet on stand-alone Backup. If you crash the MicroVAX you have to reload up the 50 floppies and Ethernet stand-alone Backup you would only have to load stand-alone Backup and then take it over the Ethernet. Andy We do not have any plans to do that. The real problem with that is the amount of mechanism necessary to get the network up. To use the Ethernet in stand-alone Backup we need a lot more than just the NI driver, we also need all of DECnet because that is how Backup accesses its files over the net. I think at that point we also need RMS which is not in stand-alone Backup. We end up with a large number of system components that are currently not there and I think all told we would probably about triple the size of the stand-alone Backup kit and if you think booting it off tu58s is bad now... What it boils down to is that stand-alone Backup runs in a very restricted environment and any changes of that sort require very substantial extensions to the environment which are of a magnitude as to be impractical. In terms of MicroVAXs on Ethernets failing and getting the software serviced we are looking at some other options. I think it is premature to talk about some of the things we are thinking about but there is work going on in that area. No 42 Hans Magneson, Sweden VMS 4.1 VAX-48 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session On our system we have set up the printer with the flags feed and flag page and when users want to get rid of this flag page then they qualify no flag but nothing happens on the printer, it still has a flag page. Peter Then you initialize a queue there is a qualifier called /SEPARATE which takes a key word list and it can take a flag burst trailer type keywords and this represents the flag burst and trailer pages which will be put on the job. On a job wide basis. There is a separate qualifier on the queue commands /DEFAULT which takes the same keywords and this pertains to the file level flag separation and trailer pages. Now the print command also has these qualifiers as distinct qualifiers including feed and it is able to override the setting specified with the /DEFAULT qualifier on the INIT QUEUE. That is, your print command can override the flag on the file level separation page policy but it can not modify or change override that which you have set up for the job. End of tape, part of speech missing. Continues in middle of speech. - because of the buffer, I am guessing but that is probably what it is. The host system where you set host to is probably in the same range, it is the QIO that the application does which it is going to have to do anyway and then a DEC IO every now and then but that is not even a QIO, thats an internal IRP mechanism which is very efficient. I don't think that the overhead is that severe, especially with V4 so I wouldnt worry about it. Comment Just a couple of additional comments with file transfer. The overhead on our micro7 70 for doing a sustained file transfer operation, I am talking about say a thousand block file where you are not worried about the overhead for opening and closing and connection and all that stuff is really not too bad. I am trying to remember this. On 2 780s with Ethernet connection where there was nothing else really going on, I was able to get sustained file tranfer rate of approximately half a megabit. The maximum that you can get with task to task file transfer is a little over 1 megabit over Ethernet cable. You are really limited by the transceiver. Now there is a special case, and that is with MicroVAX in that there is a end to end CRC check done between RMS and the remote file and on the MicroVAX, MicroVAX 1 in particular, where the CRC structuring is emulated it can soak up quite a few cycles on a systeme file transfer for that MicroVAX. But then again you are probably in a single user mode there and it is probably not that important for you. VAX-49 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session Also, just on file transfer is that there is a parameter, SET RMS DEFAULT command that you can use to specify a parameter called network block count I believe and that is a parameter that RMS uses to determine how big a QIO it should issue when doing a block mode file transfer and I think we set that to 8 blocks by default which is quite sufficient. No 45 G S Bal, University of Liverpool VMS 4.1 The terminals which used to log out and drop the lines on the V3 do not work on V4. Earlier a mention was made of system parameter TTYDEFPORT and if I heard it correctly, you set the lower bit value to 1. If I were to do that can I have an assurance that it would restore the behaviour of the terminal handler to the same level as it was on v 3. Andy It restores the behaviour with respect to the class driver, it doesn't tell the port driver to turn on auto xrx. Is that what you think your problem is. G. Bal My problem is that if the terminal is switched off and the lds box switch is switched on and the contention device my session stays on but I would like to terminate my session. It used to work on the 3.7, if you could hold the line independently for 25 to 30 seconds the system used to kill the process, which I could control from the contention device, not activate that device for the next 30 seconds. Andy Okay, basically you are saying that the gandalf hangs up but the VMS doesn't. There are no known problems in this area in VMS and haven't been since V2 I believe. Anything that is a hang up problem is generally traced to a hardware or wiring problem. There is one thing you could modify, that is tty dial type. If you set the low bit of that it shortens the amount of time that it actually cancels the terminal drivers debouncing of carrier signal, so that if it sees carrier go at all it will drop the line. Larry There is a problem associated with the upgrade to V4 in that it destroys your default modem settings in your default hang up settings and everything if you had them in the default characteristics word for the tty. It could just be in the transition that those things need to be set up again, in your default terminal projections. VAX-50 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session G bal All the lines have been set up with modem signals and even we tried the other day with a /DISABLE, and that didn't work either. Andy This is a thing where you really have to get a breakout box and play with the signals and get a field service tech out there to check your wiring and that sort of thing. No 46 N. K. Mooljee Edinburgh University Some computer systems that I know and very often in a university environment keep a record of terminal dialogue between user and computer in a physic file if the user so chooses. Is DEC seriously considering this question or the provision of such a facility for interactive users and is there a way of getting round this problem at the moment. Andy SET HOST/LOG 0, which will log you in on the local node and keep a log file of the events that happened while you were logged in. Nerris And this is a user facility which any user can use. Andy As long as you have DECnet running. That is a part of DECnet you can use without having the DECnet licence. SET HOST/LOG 0, no name, just gets you back to the same node and that will allow you to do that. It turned out easier to do it there than in the terminal driver so that was the quick fix. No 47 P V Diennen ITT, Netherlands I think it is question pertaining to the REPLY/LOG. We try to print out the operator log at daily intervals and we do that by a batch job that runs daily. If the operator logging is active you can't print the file so we have to re open the file using REPLY/LOG. This is not working from the batch. It returns successful but it does not do anything. In a desperate moment I turned on all privileges but it doesn't help. Andy I don't see anything on the face of it why this shouldn't work. We will have to take this back. Shouted comment, couldnt hear it. VAX-51 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session Andy The comments from the floor are that the REPLY/LOG command must be entered from an operator terminal, that is a terminal that has been enabled as a operator terminal and obviously a batch job is not enabled as an operator terminal. The work around for that is that the reply command uses the logical name SYS$COMMAND to direct what the operator terminal is that it is working against so by doing a logical name definition, say, define SYS$COMMAND OPA0: I believe you can make that work. For example thats how we enable OPA0: as an operator terminal from the system start up file. Because the system start up file is in effect a batch job, we do a define SYS$COMMAND OPA0: REPLY/ENABLE so I am not going to guarantee it but it sounds as if that is the sort of thing that should solve the problem. No 48 no name When we reboot 8600 after a crash or a shutdown/reboot we get the message "micro-code not up to level required for VMS" we watch it and it continues to boot and seems to run okay without any problem. So we called in our engineer and he said it's okay it's not a problem, but we had him check it out and he came back and said the CI microcode is too advanced for VMS. So we changed it for an old version of CI microcode and in the next boot the message was still there. We did notice that when you do a boot from the console in console mode the message does not appear. We would like to know where the message is coming from and should we worry about it. Trudy Matthews It has nothing to do with the CI microcode. Actually VMS during bootstrap will enquire of the CPU (in the 8600) that the console keeps the microcode version and so we enquire of the console what the current microcode version we expect. However there is a bug in the 8600 console microcode which will be corrected with the next update of the 8600 console microcode in that at times, and I can't even remember under which circumstances, it will give us back a wrong value for the microcode. But it is really safe to ignore that message. Again, end of tape, part of conference missing. Continues: Andy How much of a program is this that is required to reproduce this. VAX-52 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session Questioner (unknown) Well unfortunately some of the tools were written some time ago and we don't have the original source code for it. Program is about 700,000 lines of FORTRAN. Andy Boy you got me. The only idea I have is get it in that state and take a crash dump and send us the details of what the application is trying to do in terms of qios and so forth, the general philosophy behind the program maybe. And also include which DZ line was hung so we don't have to go all through the dump trying to figure out which terminal line was hung, that is a big help, and we will take a look at it, but it doesn't sound like the kind of thing we will be able to track down. Questioner Our operators tend when they shut down the machine, to start using carriage returns, and when you are in a cluster environment and you do that, if you don't type in "REMOVENODE" then you end up with a hang in that cluster until they all decide that that node is indeed shutting down. I recommend that you put a little check in the shtudown and if the node you are shutting down is a cluster member, make the remove node the default. Comment I have a comment on this gentleman's question. I have some problems with a DMF32 port, the first two ports on it have the full modem control and I have also a small problem in the source code, but it won't run on the first two ports of DMF32 because of some changes in the driver. I have had spoken to the field service engineer and he was pretty sure that the changes in the driver were the cause of my problem. I have changed to another port. Andy Do you know offhand what the rev level of your DMF is. Commenter No. Andy There are no problems known with the current VMS V4. VAX-53 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session Commenter Actually I can fix it in my program, quite easily, I have to do something which I didn't have to do earlier, so it was just an inaccurate program at first and so now I have to improve it a little to make it work on the first two ports. It may be that the problems look alike. Andy It is my understanding that very old versions of DMF like rev C or rev d of the microcode in the DMF did have the same XON/XOFF problem but if you are on a field service contract you should be way beyond that at this point. Another comment (no name) I have a similar problem with a DZ11 port hooked up to a 240. Under certain circumstances, especially when it is in VT52 mode the VT driver crashes the VAX, and it does it consistently. It works for a while and then all of a sudden several pages of garbage are sent to the terminal and it crashes. Andy We fixed some strange conditions in the DZ driver, I think it was in 4.1 but maybe it's 4.2. No 58 (no name) In a cluster when one member dies for any reason the other members are stopped for a while. What can be done to decrease that dead time. Andy In this case are you actually getting a crash, or is it stopping and rebooting? No 58 I booted one voluntarily to see what happened on the other one and the other one stopped doing anything for 50 to 60 seconds. Andy When you simply stop and reboot a node, one of the problems that you can run into is that the CI port hardware has a 99 second timeout on its virtual circuit. Usually an unjam command on a 780 will break that virtual circuit. However, lets say the operating system simply hangs or something like that, then there is a 99 second timeout in the CI port, after which the CI port finally gives up and decides that its processor is dead. That VAX-54 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session time out can not be modified... it is built into the CI microcode. So the best you can do is to try to avoid hitting that time out. For example when you reboot a machine, in particular, give it an unjam command which immediately shuts down the CI port and eliminates that time out interval. Beyond that you are right, the next time out is the reconnection interval so you can shorten that up and then that will shorten the time it takes to fail the machine out of the cluster. No 58 Is it a problem, to shorten down this value? Andy The only problem that you can run into is that you increase your exposure to things like transient power failures or potentially transient communications failures, lets say, something goes wrong for a few seconds with the CI communications or lets say that the machine looses power for a very short period of time. Normally we ship it with reconnection intervals set up so that a very brief power outage will allow the machine to remain in the cluster. Of course the problem is that while that machine is out the rest of the cluster has to wait for it to come back. If you take the other option of shortening the reconnection interval, then if the machine takes a power failure, it will immediately leave the cluster, but it cannot re-enter. It means that the machine must then be rebooted, even it managed to survive the power fail. That is really only the significant negative effect of shortening re connection interval. No 61 Bojan Dremelj, Yugoslavia, VAX/VMS 4.1 Why does the error command not work as it is described in the help facility? $help errors...for a more detailed explanation of the error and suggestions for error recovery, type error followed by the ident portion of the error message... $error ident...%dcl-w-iwerb, unrecognised command.../error/ Andy That certainly sounds like we left some loose ends in our help file. The help files are prepared by our documentation group, we will have to take that back to the documentation group and ask them what is going on. I think what has happened is that you have been led astray by a very minor error in the help text. Clearly the way it should work to find help on a particular error message is that you should say.."help error" and then the error identification rather than simply error identification. VAX-55 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session That doesn't work either. Trudy Matthews According to the UK TSC it never did work and the entry pointing you up to it is removed at V4.2. No 63 W Kegel, Dynaflow Software Systems, Holland I have processes having sysprv privileges for various reasons and up to V3.7 those processes were creating files specifying uic [0,0], they would be created on the user's account. As of V4 they are created on the owner of the directory instead. My question is this a temporary process or is it going to go away. Andy This is an intentional change. This is part of the new ownership and protection defaulting rules in the file system. In general, we felt that it made more sense to default the file ownership to be the ownership of the directory that you are creating the file in, provided that the person creating the file has the privileges to charge space to that UIC. For example what that does, is it allows the system manager to create a file in some user's directory and have that file be owned by the user. This is the sort of rationale, we felt it was a more consistent way for things to work. We apologise for the inconvenience. We made the change feeling that a greater number of users would benefit from it. W. Kegel Then I would suggest that you do something about your documentation because your documentation just says that you need sysprv to create a file on another account on your own but it doesn't state that if you have the privilege it will be done. Comment The documentation was fixed in 4.2. No 65 Larry Kilgallen VMS 4.2 What is the policy regarding distribution of VMS 4.2 source kit updates? Andy There will be V4.2 source kit. The last I heard, the 4.2 project leader was still banging his head against the wall, getting the thing to build properly. VAX-56 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session Larry That's a technical question Andy. The real question has to do with whenever it happens. Andy It is our intention in the future to provide source kits for the functional updates. Peter (I think) For 4.2 in particular, I believe it is going to be an update to the 4.0 source kit as opposed to a remastered source kit. Larry What I am really trying to do is dance around the forbidden issue of prices. People who already bought a 4.0: Will they get this... Peter We talked to our product management about that, we being the people worried about trying to put the kits together. Larry I have the greatest faith that they will get the kit together, it is just getting it out to the field that I am questioning. Peter It is under discussion from what we can see and there will be resolution shortly. Andy It turns out, Larry, that I was just recently looking in the price book about source kits anyway and noticed that even for 4.0, the upgrade price from V3 is considerably less than the price of a new source kit, so without knowing any of the details, never mind mentioning them, I would guess that we would do something similar. Larry To repeat something that has been said before, regardless of how much money you hope to make out of source kits, it would be a lot nicer on our end if it were a monthly charge that we pay forever, because we can sell that to management. We never know when VMS is going to come out and having to, all of a sudden, when the stuff comes out, run in and get a purchase VAX-57 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session authorisation is a lot more cumbersome. Certainly with doc sets we pay a certain amount per month, never knowing how many pages we are going to get per dollar how often, and if a similar policy could be instituted for source kits it might be nice. Peter That is a very good suggestion and we are looking at some of the licencing policies and that is one of the alternatives. Larry Someone else in the organisation might like an even revenue flow as well, regardless of whether software is produced, and that is one way to do it. No 66 (no name) We occasionally have processes having an outstanding read on the mail box and we then ask them to terminate by means of a $FORCEX system service. That worked fine until 3.7 but then all of a sudden those processes didn't go away any more. Instead they returned an IO error which, after the usual fiddling around, trying to convert a FORTRAN IO error to a system service error, we found to be "activity precludes operation" and the exit handlers are never activated any more. Can you explain why? Andy You are using RMS to read the mail boxes? No 66 Implicitly. It is a FORTRAN read. Andy I don't remember the details, I know we went through some mayhem in getting RMS rundown to work correctly and I know it has been changed several times. It has to do with RMS keeping track of whether or not there is IO on a particular RMS open file at the time you are doing rundown. The problem is, you enter your exit handler and you are now attempting to do IO on an RMS channel that already has IO pending. I believe what is likely to have happened in 3.7 is that they tightened up some of those checks, where you used to get away with things that maybe you were lucky with, but in other circumstances can crash RMS and I know we have been tightening up the checks on that sort of thing recently. No 66 VAX-58 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session I am pretty sure that it doesn't enter my user written exit handlers. All those processes have one, I force one to be there by doing LIB$INITIALIZE. Andy I think we are going to have to have an SPR and some documentation on that. I think it is too complicated for me to deal with here. No 66 I did a work around by using a $DELPRC instead which worked fine. No 67 Bengt Eriksson, Stockholm County Council, Utility: Backup Normally when a user makes a mount request that request is of course spread round to all the terminals that have been enabled as operator terminals but when I perform backup, and especially backup onto a tape station which requires many reels of tapes, Backup issues the message %BACKUP-I-READYWRITE, please mount next volume or something like that, but that message is not broadcast round to other terminals. That has the consequence that I as system manager have to be in the vicinity of the terminal or the console from where the backup is invoked. It would be very handy if I could sit in my own room and just go into the VAX every now and change tape reads. Is that possible? Andy That is a worthwhile suggestion. Backup in fact has the logic in it to use the operator request for tape mounts and at present it is coded so it only uses that when you run it in a batch job. I think that is a worthwhile suggestion, we have heard it in the past, that there at least be a qualifier to cause it to use the operator messages for tape mounts. It was done the way it was done originally, based on the assumption that things like multitape backups would most likely be done by operators who would be logged in near the tape drive anyway. That assumption holds true in many cases but obviously not all. No 69 David Roberts, Univ. of Leicester, VMS 4.1 mag tapes. I have no need to ask the question, because the question was answered in a different form earlier on. No 88 (no name) VAX-59 PAGESWAPPER - April 1986 - Volume 7 Number 9 European DECUS Question and Answer Session This one is so simple that I am sure it must have been asked before in some context but I did not hear the answer. Is there ever going to be in SYS$ACTIVATEIMAGE service, which will do just the image activation, nothing about creating processes or spawning etc? Andy Obviously that service exists and so what it really boils down is when are we going to get round to be willing to document the thing. The answer is as soon as we feel that we have the interface to that service frozen. We have in fact changed the interface to the image activation service a couple of times as we have added features to the image activator. Because we feel that the interface is not fully stable yet is why we have not documented it. Obviously documenting it means committing to that service. We do plan to stabilise that at some point in the not too distant future so that we can document it. No 88 Would it possible to map images and call them as subprograms but we couldn't recover the space? Andy What you want to do is recover the space after the image. That turns out to be very very difficult. At least from the point of view of the system, because once you have activated a sharable image or whatever and called into it, there is really no way for VMS to tell that that image has become truely idle. It is possible for it to have any number of ASTs outstanding for example, because of terminal driver events or whatever. No 88 Is that true if it is a subprogram? Andy Well no, thats not true because subprograms can have side effects that last way beyond their invocation and about the only thing we can recommend is that, in the case where you have done that, you know the piece of address space that that has been loaded into. If you want to get rid of it do a $DELETVA on that address space and then it is your responsibility. Then something has gone wrong. In fact you deleted the address space when the thing that you were using was not fully idle and that is exactly the kind of problem that we feel unable to solve. VAX-60 PAGESWAPPER - April 1986 - Volume 7 Number 9 VAX System SIG Committee List VAX System SIG Committee List As of January 8, 1985 Osman K. Ahmad - Large Systems Integration Working Group Association of American Railroads Technical Center, Research and Test Department 3140 South Federal Street Chicago, IL 60616 Joe Angelico - Assistant Symposium Coordinator US Coast Guard CCGD8(DT) Hale Boggs Federal Building 500 Camp Street, New Orleans, LA 70130 Elizabeth Bailey - Volunteer Coordinator 222 CEB Tennessee Valley Authority Muscle Shoals, AL 35660 June Baker - Advisor Computer Sciences Corporation 6565 Arlington Boulevard Falls Church, VA 22046 Joe L. Bingham - Librarian Mantech International 2320 Mill Road Alexandria, VA 22314 Bob Boyd - Commercial Working Group GE Microelectronics Center MS 2P-04 Post Office Box 13409 Research Triangle Park, NC 27709 C. Douglas Brown - Security Sandia Labs Division 2644 P.O. Box 5800 Albuquerque, NM 87185 Jim Caddick - VAXcluster General Datacom Strait Turnpike Middlebury, CT 06762-1299 Jack Cundiff - Symposium Coordinator Horry-Georgetown Post Office Box 1966 Conway, SC 29526 VAX-61 PAGESWAPPER - April 1986 - Volume 7 Number 9 VAX System SIG Committee List Tom Danforth - Handout Editor Woods Hole Oceanographic Institute Woods Hole, MA 02543 Jim Downward - Migration and Host Development, VAXintosh Working Group KMS Fusion Incorporated 3941 Research Park Drive Ann Arbor MI 48106 Jane Furze - Campground 3830 West Cochise Phoenix, AZ 85064 Dennis Frayne - Real Time/Process Control Working Group McDonnell Douglas 5301 Bolsa Avenue Huntington Beach, CA 92646 Carl E. Friedberg - Internals Working Group In House Systems 165 William Street New York, NY 10038 Don Golden - Communications Committee Representative c/o Shell Oil Company Westhollow Research Center Post Office Box 1380, Room D2132 Houston, TX 77001 Gary Grebus - System Improvement Request Battelle Columbis Labs Room 11-6011 505 King Avenue Columbus, OH 43201-2693 B. Hancock - Network Working Group Dimension Data Systems, Incorporated 2510 Limestone Lane Garland, TX 75040 (214) 495-7353 Jeffrey S. Jalbert - Historian J C C Post Office Box 381 Granville, OH 43023 614-587-0157 Ray Kaplan - MicroVAX Working Group Pivotal Incorporated 6892 East Dorado Court Tucson, AZ 85715 VAX-62 PAGESWAPPER - April 1986 - Volume 7 Number 9 VAX System SIG Committee List Lawrence J. Kilgallen - Newsletter Editor Box 81, MIT Station Cambridge, MA 02139-0901 Margaret Knox - Chair Computation Center University of Texas Austin, Texas 78712 Art McClinton - Advisor MITRE 1820 Dolley Madison Boulevard McLean, VA 22102 Ross W. Miller - Vice Chair and Working Group Coordinator Online Data Processing, Inc. N 637 Hamilton Spokane, WA 99202 Eugene Pal - Multiprocessor Working Group US Army CAORA (ATOR-CAT-C) Fort Leavenworth, KA Susan Rehse - System Management Working Group Lockheed Missiles 3251 Hanover Street Palo Alto, CA 94301-1187 Bob Robbins - Advisor Array Computer Consultants 5364 Woodvale Drive Sarasota, FL 33582 Larry Robertson - Real Time/Process Control Working Group Bear Computer Systems Inc. 5651 Case Avenue North Hollywood, CA David Schmidt - LUG Coordinator, Hardware Working Group Management Sciences Associates 5100 Centre Avenue Pittsburgh, PA 15232 Al Siegel - Advisor Battelle Memorial Institute 505 King Avenue Columbus, OH 43201-2693 D. Slater - Artificial Intelligence Working Group Institute for Defense Analysis 1801 North Beavregard Street Alexandria, VA 22314 VAX-63 PAGESWAPPER - April 1986 - Volume 7 Number 9 INPUT/OUTPUT INPUT/OUTPUT A SIG Information Interchange A form for INPUT/OUTPUT submissions is available at the back of the issue. INPUT/OUTPUT 486 Caption: TeX DVI post-processor for terminals Message: We are looking for a way to direct TeX output to a terminal, rather than to a hardcopy device. Ideally, this could be routed to a high-resolution graphics terminal (i.e., Tektronix 4107), but support for ANSI (VT100) terminals would suffice (not supporting special features, just the text). Does anybody have this capability or has anybody heard of it? Contact: Mr. Brian R. Shamblin Medical Sciences Building Mayo Foundation 200 First Street Southwest Rochester, MN 55905 Telephone (507) 284-4069 Date: February 12, 1986 VAX-64 PAGESWAPPER - April 1986 - Volume 7 Number 9 INPUT/OUTPUT Submission Form INPUT/OUTPUT Submission Form A SIG Information Interchange Please reprint in the next issue of the Pageswapper If this is a reply to a previous I/O, which number? ________ Caption: ______________________________________________________ Message: ______________________________________________________ ________________________________________________________________ ________________________________________________________________ ________________________________________________________________ ________________________________________________________________ ________________________________________________________________ ________________________________________________________________ Contact: Name _______________________________________________________ Address ____________________________________________________ ____________________________________________________________ ____________________________________________________________ ____________________________________________________________ Telephone ____________________________ Signature _____________________________ Date ________________ Mail this form to: Larry Kilgallen, PAGESWAPPER Editor Box 81, MIT Station, Cambridge, MA 02139-0901, USA PAGESWAPPER - April 1986 - Volume 7 Number 9 INPUT/OUTPUT Submission Form Tear out or photocopy reverse to submit an I/O item Larry Kilgallen, PAGESWAPPER Editor Box 81, MIT Station Cambridge, MA 02139-0901 USA PAGESWAPPER - April 1986 - Volume 7 Number 9 System Improvement Request Submission Form System Improvement Request Submission Form Page 1 of _____ ________________________________________________________________ Submittor: Firm: Address: Phone: ________________________________________________________________ How to write an SIR: Describe the capability you would like to see available on VAX systems. Be as specific as possible. Please don't assume we know how it's done on the XYZ system. Justify why the capability would be useful and give an example of its use. If you wish, suggest a possible implementation of your request. ________________________________________________________________ Abstract (Please limit to four lines): ________________________________________________________________ Description and examples (use additional pages if required) PAGESWAPPER - April 1986 - Volume 7 Number 9 System Improvement Request Submission Form Tear out or photocopy reverse to submit an SIR Gary L. Grebus Battelle Columbus Division Room 11-6011 505 King Avenue Columbus, Ohio 43201-2693 USA PAGESWAPPER - April 1986 - Volume 7 Number 9 VAX Systems SIG Spring 1986 SIR Ballot VAX Systems SIG Spring 1986 SIR Ballot DECUS membership number __________________ (six digits) Our site uses the following VAX models (check all that apply) 8600 ____ 11/782 ____ 11/780,11/785 ____ 11/750 ____ 11/730,11/725 ____ MicroVAX ____ We use VAX's in the following applications (Check all that apply) Business EDP ____ Software Development ____ Education ____ Computer Science Research ____ Data Acquisition/Control____ CAD/CAM ____ Service Bureau ____ Hardware Development ____ Scientific/Engineering ____ Office Automation ____ Telecommunications _____ Other ____________________________ I support the following as the most important System Improvement Requests. (List from zero to fifteen SIR's): SIR Number: ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- I oppose the following SIR's as detrimental. (List from zero to five SIR's): SIR Number: ---------- ---------- ---------- ---------- ---------- Mail to: Gary L. Grebus Battelle Columbus Division Room 11-6011 505 King Avenue Columbus, OH 43201 To be counted, you ballot must be received by April 1. PAGESWAPPER - April 1986 - Volume 7 Number 9 VAX Systems SIG Spring 1986 SIR Ballot Tear out or photocopy reverse to vote on SIRs Gary L. Grebus Battelle Columbus Division Room 11-6011 505 King Avenue Columbus, Ohio 43201-2693 USA