seenu@isieng.UUCP (Seenu Banda) (05/26/88)
I am posting this for a friend of mine.
-------------------
I have a question regarding Apollo and Sun NFS. We have a network
of Sun workstations (4.3BSD + NFS) and an apollo (4.2 Domain/IX). We are
porting some software from Sun environment to Apollo and in doing so had to
duplicate source on both (Sun and Apollo) due to lack of NFS connection
between Sun and Apollo. This resulted in disk space problems on the Apollo.
So, now the problem:
* How can we integrate apollo into the Sun NFS environment?
* Does apollo follow standard NFS protocol. (No mention of NFS was
made in any of the manuals).
* First of all, can it be done with minor fixes/patches to existing code?
* If so, could you please give me pointers as to where to start?
* If not, is there any public/private domain software which does this?
Sorry for asking so many questions in one breath. Any ideas, theories,
anything at all will be greatly appreciated.
--------
I will gladly pass along any information to the person concerned.
Seenu Banda
..!pyramid!isieng!seenu
--
Seenu
..pyramid!isieng!seenu jec@iuvax.cs.indiana.edu (James E. Conley) (05/27/88)
NFS is available for Apollos from Apollo. It does have some problems
though. The major problem we have is that you cannot mount a non-Apollo
NFS partition and expect everything to work as if it were an Apollo disk.
The reason for this (I believe) is problems with file types. Our major
gripe with this is that we cannot execute programs that are on an NFS
partition on our VAX.
III Usenet: iuvax!jec
UUU I UUU ARPANet: jec@iuvax.cs.indiana.edu
U I U Phone: (812) 335-7729
U I U U.S. Mail: Indiana University
U I U Dept. of Computer Science
UUUIUUU 021-E Lindley Hall
I Bloomington, IN. 47405
III (Home of Bob Knight and the Indiana Hoosiers)karish@denali.stanford.edu (Chuck Karish) (05/27/88)
In article <9145@iuvax.cs.indiana.edu> jec@iuvax.UUCP (James E. Conley) writes: >NFS is available for Apollos from Apollo. It does have some problems >though. The major problem we have is that you cannot mount a non-Apollo >NFS partition and expect everything to work as if it were an Apollo disk. >The reason for this (I believe) is problems with file types. Our major >gripe with this is that we cannot execute programs that are on an NFS ^^ ^^^^^^ ^^^^^^^ ^^^^^^^^ ^^^^ ^^^ ^^ ^^ ^^^ >partition on our VAX. ^^^^^^^^^ ^^ ^^^ ^^^ Are you trying to run VAX binaries on an Apollo? Are you surprised that it doesn't work? As part of a discussion of standards in another newsgroup, an employee of Apollo (speaking on a personal basis, not as a representative of Apollo) said that there may be some real incompatibilities between Apollo NFS and Sun NFS. He said that Apollo wrote their NFS implementation from the specifications given in the standard published by Sun, and indicated that Sun's implementation doesn't match their own standard precisely. I don't know what problems this caused, or what has been done to solve them. Chuck Karish ARPA: karish@denali.stanford.edu BITNET: karish%denali@forsythe.stanford.edu UUCP: {decvax,hplabs!hpda}!mindcrf!karish USPS: 1825 California St. #5 Mountain View, CA 94041
krowitz@RICHTER.MIT.EDU (David Krowitz) (05/27/88)
Actually, the problem with executing files via NFS is not a problem with the NFS implementation itself. The problem is that the Apollo loader (so I was told) does not use streams (IOS) calls to open the executable file. It used the map segment calls which bypass the IOS type managers on which NFS is built. -- David Krowitz krowitz@richter.mit.edu (18.83.0.109) krowitz%richter@eddie.mit.edu krowitz%richter@athena.mit.edu krowitz%richter.mit.edu@mitvma.bitnet (in order of decreasing preference)
crgabb@sdrc.UUCP (Rob Gabbard) (05/28/88)
> In article <9145@iuvax.cs.indiana.edu> jec@iuvax.UUCP (James E. Conley) writes: > >gripe with this is that we cannot execute programs that are on an NFS > >partition on our VAX. > ^^^^^^^^^ ^^ ^^^ ^^^ > In article <21859@labrea.STANFORD.EDU>, karish@denali.stanford.edu (Chuck Karish) writes: > Are you trying to run VAX binaries on an Apollo? Are you surprised > that it doesn't work? > > said that there may be some real incompatibilities between Apollo NFS > and Sun NFS. He said that Apollo wrote their NFS implementation from > the specifications given in the standard published by Sun, and indicated > that Sun's implementation doesn't match their own standard precisely. Apollo's NFS will not allow you to run an Apollo binary from an NFS disk. Also, although the compilers can read source from an NFS disk, they cannot write to an NFS disk - they must write to an Apollo filesystem. Apollo cites the reason for this as being that their file system is a single level store allowing for the direct mapping of binaries into address space instead of a page area. The glitch in this reasoning is that Sun can do this with NFS. *Sigh* - wouldn't it have been nice it Apollo had made DOMAIN public domain before NFS came along ? :-) -- =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= Rob Gabbard | UUNET: uunet!sdrc!crgabb Workstation Systems Programmer | PHONE: (513)576-2600 Structural Dynamics Research Corporation | =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
benoni@ssc-vax.UUCP (Charles L Ditzel) (06/01/88)
in article <21859@labrea.STANFORD.EDU>, karish@denali.stanford.edu (Chuck Karish) says:
+ As part of a discussion of standards in another newsgroup, an employee of
+ Apollo (speaking on a personal basis, not as a representative of Apollo)
+ said that there may be some real incompatibilities between Apollo NFS
+ and Sun NFS. He said that Apollo wrote their NFS implementation from
+ the specifications given in the standard published by Sun, and indicated
+ that Sun's implementation doesn't match their own standard precisely.
Why is that other vendors are capable of writing NFS implementation correctly?
Has anyone tried file serving PCs off of Apollo NFS as is possible on Suns?
Given that we may get an Iris anyday now (which has NFS) has anyone had
any experiences with Apollo-Iris NFS?wesommer@athena.mit.edu (William Sommerfeld) (06/02/88)
In article <1955@ssc-vax.UUCP> benoni@ssc-vax.UUCP (Charles L Ditzel) writes: >Why is that other vendors are capable of writing NFS implementation correctly? Because they don't write it from scratch; instead, they plug Sun's code into their (BSD4.2-derived) kernel. Of course, since it's in the kernel, bugs in NFS client or server code (typically induced by various malformed "poison packets") will crash or wedge your server. I have heard that NFS implementations for "different" operating systems (such as the one written in LISP for Symbolics Lisp Machines) don't work very well, either. - Bill
guy@gorodish.Sun.COM (Guy Harris) (06/02/88)
> Because they don't write it from scratch; instead, they plug Sun's > code into their (BSD4.2-derived) kernel. Or plug Lachman's code into their S5-derived kernel. The basic point still stands, however; it's easier to drop existing NFS server code into a UNIX-derived system than it is to write a server from scratch.
benoni@ssc-vax.UUCP (Charles L Ditzel) (06/02/88)
in article <285@sdrc.UUCP>, crgabb@sdrc.UUCP (Rob Gabbard) says: > *Sigh* - wouldn't it have been nice it Apollo had made DOMAIN public domain > before NFS came along ? :-) Wouldn't it be nice if apollo would *sell* a *working* version of NFS????? Is this a clue of what type of stuff will come out of OSF??? Seriously, is Apollo *fixing* their bugs with NFS or are they simply selling a broken product with the idea of discouraging its use...seeing as how I am going to have to use this product ... I have more than a vested interest... ----------------- my opinions are my own.
paul@DELRIO.CC.UMICH.EDU ('da Kingfish) (06/09/88)
>Has anyone tried file serving PCs off of Apollo NFS as is possible on Suns?
We will make that attempt in a couple days. Dunno if people in the engin
college here are doing that or not. Also will try something called CAP, so
Macs can use Apollo disk.
--paulkrowitz@RICHTER.MIT.EDU (David Krowitz) (06/10/88)
We had a PC with the SUN PC-NFS running on it working with both Suns, Alliants, and some Apollos when we were beta-testing the Apollo NFS implementation. I assume it still works. -- David Krowitz krowitz@richter.mit.edu (18.83.0.109) krowitz%richter@eddie.mit.edu krowitz%richter@athena.mit.edu krowitz%richter.mit.edu@mitvma.bitnet (in order of decreasing preference)
benoni@ssc-vax.UUCP (Charles L Ditzel) (06/13/88)
in article <8806091828.AA06185@richter.mit.edu>, krowitz@RICHTER.MIT.EDU (David Krowitz) says: > We had a PC with the SUN PC-NFS running on it working with both > Suns, Alliants, and some Apollos when we were beta-testing the > Apollo NFS implementation. I assume it still works. Again maybe it's my fault for not defining file-serving. Can an Apollo be used as a file server for *both* MS-DOS executables and ASCII. The word "working" is again fuzzy. I don't just mean reading an ASCII file, but can the PC write to the Apollo disk and read from it (both ASCII and DOS executables).
crgabb@sdrc.UUCP (Rob Gabbard) (06/14/88)
In article <2001@ssc-vax.UUCP>, benoni@ssc-vax.UUCP (Charles L Ditzel) writes: > Again maybe it's my fault for not defining file-serving. Can an > Apollo be used as a file server for *both* MS-DOS executables and > ASCII. The word "working" is again fuzzy. I don't just mean I'm not sure about PC's but you can copy a SUN or HP executable to an Apollo disk via NFS and execute it. My test was simple however - % mount -o soft apollo:// /mnt % cp /bin/date /mnt/apollo/sys/node_data/tmp % /mnt/apollo/sys/node_data/tmp/date Tue Jun 14 10:01:21 EDT 1988 -- =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= Rob Gabbard | UUNET: uunet!sdrc!crgabb Workstation Systems Programmer | PHONE: (513)576-2600 Structural Dynamics Research Corporation | =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
krowitz@RICHTER.MIT.EDU (David Krowitz) (06/15/88)
Good question ... we never tried using executable files ... by the way, the SUN PC-NFS implementation comes with its own PCNFSD server which you compile with your C compiler and run as the server for the PC. You don't use the normal NFSD for the PC (even on a Sun running as the server for the PC). We're now using the DPCI-RING software to connect our PC's to the Apollos, so I can't really test it out (we changed our ethernet cabling scheme from thin ethernet to the real stuff and don't have transceivers for the PC's). If I remember correctly, the PCNFSD server does no file conversion as the file is transfered from the PC to the server (Apollo, Sun, or whatever), unlike the DPCI software which will convert ascii files from the MS-DOS format (CR-LF at the end of each line) to the Apollo/Unix format (LF at the end of each line). If this is correct, then the PC should have no trouble storing executable files on the server. -- David Krowitz krowitz@richter.mit.edu (18.83.0.109) krowitz%richter@eddie.mit.edu krowitz%richter@athena.mit.edu krowitz%richter.mit.edu@mitvma.bitnet (in order of decreasing preference)
grindef@caen.engin.umich.edu (Wesley Craig) (06/15/88)
In article <8806090336.AA20691@delrio.cc.umich.edu>, paul@DELRIO.CC.UMICH.EDU ('da Kingfish) writes: > > >Has anyone tried file serving PCs off of Apollo NFS as is possible on Suns? > > We will make that attempt in a couple days. Dunno if people in the engin > college here are doing that or not. Also will try something called CAP, so > Macs can use Apollo disk. > > --paul The pcs seem to work just fine. Wes
guy@gorodish.Sun.COM (Guy Harris) (06/16/88)
> You don't use the normal NFSD for the PC (even on a Sun running > as the server for the PC). If by "the normal NFSD" you mean the process started by running "/usr/etc/nfsd", you most definitely *do* use it for the PC; as I pointed out before, PC-NFS uses the NFS protocol for file access, which means a PC client talks to the exact same NFS server as all other NFS clients do. "pcnfsd" provides some sort of record-locking service, not raw file access. > If I remember correctly, the PCNFSD server does no file conversion as the > file is transfered from the PC to the server (Apollo, Sun, or whatever), Since the "pcnfsd" server doesn't do file access, this is trivially correct; "nfsd" doesn't do file conversion either.
frank@CAEN.ENGIN.UMICH.EDU (Randy Frank) (06/16/88)
pcnfnd does **NOT** do record locking. Pcnfsd is there to provide authentication services for PCs, which obviously do not have normal login services and therefore can't be trusted to provide a valid uid to nfs. Pcnfsd is totally out of the loop once authentication has occurred, and therefore PC/NFS has exactly the same set of file conversion services as ordinary Unix NFS, inother words, none.
benoni@ssc-vax.UUCP (Charles L Ditzel) (06/18/88)
in article <8806141958.AA08619@richter.mit.edu>, krowitz@RICHTER.MIT.EDU (David Krowitz) says: > > Good question ... we never tried using executable files ... by the ... > transfered from the PC to the server (Apollo, Sun, or whatever), unlike > the DPCI software which will convert ascii files from the MS-DOS format > (CR-LF at the end of each line) to the Apollo/Unix format (LF at the > end of each line). If this is correct, then the PC should have no > trouble storing executable files on the server. We *ARE* currently using DPCI and that is probably motivating our switch. We are interesting in a general solution to the problem rather than buy DPCI, UShare and 50 other solutions for other machines... NFS fits nicely (or so we think) because it is a GENERAL solution...that is virtually all machines have NFS (PC, Macs, Silicon Graphics, Suns, ... even IBM :-0 ,...something like 200+ licenses to vendors exist).