[Info-vax] Interesting NFS bug...

Tue Jan 5 00:53:50 EST 2010

"John Santos" <john at egh.com> wrote in message 
news:MPG.25a9c6173f832149989690 at news.giganews.com...
> In article <udCdnVnI6cY2C6jWnZ2dnUVZ_judnZ2d at dls.net>, john at nfw-
> invalid.cibtrikker.com says...>
>> "JF Mezei" <jfmezei.spamnot at vaxination.ca> wrote in message
>> news:00a6591e$0$1587$c3e8da3 at news.astraweb.com...
>> > OS-X mounted  disk via NFS on VMS (VMS acts as NFS server).
>> >
>> > ditto used to copy files tro the OS-X disk.
>> >
>> > the last bit of index.html from VMS:
>> >
>> >> 22534D56 20796220 64657265 776F5022 "Powered by VMS" 0000C0
>> >>  72742F3C 0005003E 64742F3C 0005003E >...</td>...</tr 0000D0
>> >>  3E64743C 3E72743C 00150000 0000003E >.......<tr><td> 0000E0
>> >>  0000003E 72742F3C 3E64742F 3C3E703C <p></td></tr>... 0000F0
>> >>  2F3C0007 00003E65 6C626174 2F3C0008 ..</table>....</ 000100
>> >>  003E6C6D 74682F3C 0007003E 79646F62 body>...</html>. 000110
>> >>  00000000 00000000 00000000 0000FFFF ................ 000120
>> >>  00000000 00000000 00000000 00000000 ................ 000130
>> >>  00000000 00000000 00000000 00000000 ................ 000140
>> >>  00000000 00000000 00000000 00000000 ................ 000150
>> >>  00000000 00000000 00000000 00000000 ................ 000160
>> >>  00000000 00000000 00000000 00000000 ................ 000170
>> >>  00000000 00000000 00000000 00000000 ................ 000180
>> >>  00000000 00000000 00000000 00000000 ................ 000190
>> >>  00000000 00000000 00000000 00000000 ................ 0001A0
>> >>  00000000 00000000 00000000 00000000 ................ 0001B0
>> >>  00000000 00000000 00000000 00000000 ................ 0001C0
>> >>  00000000 00000000 00000000 00000000 ................ 0001D0
>> >>  00000000 00000000 00000000 00000000 ................ 0001E0
>> >>  00000000 00000000 00000000 00000000 ................ 0001F0
>> >
>> >
>> > After copy to OS-X:
>> >
>> >
>> >> 000011c0  74 3d 22 50 6f 77 65 72  65 64 20 62 79 20 56 4d 
>> >> |t="Powered
>> >> by VM|
>> >> 000011d0  53 22 3e 0a 3c 2f 74 64  3e 0a 3c 2f 74 72 3e 0a
>> >> |S">.</td>.</tr>.|
>> >> 000011e0  0a 0a 3c 74 72 3e 3c 74  64 3e 3c 70 3e 3c 2f 74
>> >> |..<tr><td><p></t|
>> >> 000011f0  64 3e 3c 2f 74 72 3e 0a  0a 3c 2f 74 61 62 6c 65
>> >> |d></tr>..</table|
>> >> 00001200  3e 0a 0a 3c 2f 62 6f 64  79 3e 0a 3c 2f 68 74 6d
>> >> |>..</body>.</htm|
>> >> 00001210  6c 3e 0a 00 3f 00 a0 00  39 00 0b 01 05 00 20 20
>> >> |l>..?...9.....  |
>> >> 00001220  20 31 35 00 38 00 04 00  3f 00 a0 00 00 40 00 00  |
>> >> 15.8...?.... at ..|
>> >> 00001230  08 00 a0 00 38 00 04 00  3f 00 a0 00 00 40 00 00
>> >> |....8...?.... at ..|
>> >> 00001240  3b 00 a0 00 4a 00 08 00  82 00 a0 00 3f 00 a0 00
>> >> |;...J.......?...|
>> >> 00001250  6a 00 0b 01 0b 00 32 31  2d 4e 4f 56 2d 32 30 30
>> >> |j.....21-NOV-200|
>> >> 00001260  39 00 00 00 38 00 04 00  3f 00 a0 00 00 40 00 00
>> >> |9...8...?.... at ..|
>> >> 00001270  08 00 a0 00 38 00 04 00  3f 00 a0 00 00 40 00 00
>> >> |....8...?.... at ..|
>> >> 00001280  3b 00 a0 00 4a 00 0a 00  82 00 a0 00 3f 00 a0 00
>> >> |;...J.......?...|
>> >> 00001290  c5 00 0b 01 14 00 53 4d  54 50 25 22 64 6a 66 30
>> >> |......SMTP%"xxxx|
>> >> 000012a0  31 40 62 69 6b 65 6f 64  79 73 00 00 38 00 04 00
>> >> |x at xxxxxxxx..8...|
>> >> 000012b0  3f 00 a0 00 00 40 00 00  08 00 a0 00 38 00 04 00
>> >> |?.... at ......8...|
>> >> 000012c0  3f 00 a0 00 00 40 00 00  3b 00 a0 00 4a 00 0c 00
>> >> |?.... at ..;...J...|
>> >> 000012d0  82 00 a0 00 3f 00 a0 00  5f 01 0b 01 1e 00 52 65
>> >> |....?..._.....Re|
>> >> 000012e0  3a 20 4c 61 73 74 20 57  65 65 6b 73 20 54 72 61  |: Last 
>> >> Weeks
>> >> Tra|
>> >> 000012f0  69 6e 20 41 64 76 65 6e  74 75 72 65 38 00 04 00  |in
>> >> Adventure8...|
>> >> 00001300  3f 00 a0 00 00 40 00 00  08 00 a0 00 82 00 c0 00
>> >> |?.... at ..........|
>> >> 00001310  82 00 c0 00 40 00 c0 00  00 00 00 01 00 00 00 00
>> >> |.... at ...........|
>> >> 00001320
>> >
>> >
>> > In other words, the NFS server on VMS filled data beyond the end of 
>> > file
>> > with text from another file. So much for VMS file system security.
>> >
>> >
>> > Performing the copy from VMS ($COPY on VMS with Unix acting as NFS
>> > server) did not add content to the end of file.
>> >
>> > However, one cannot copy a whole directory tree this way. Backup 
>> > creates
>> > files on the other system with file extensions as part of file name. so
>> > instead of "myimage.jpg" you have "myimage.jpg;3"
>> >
>> > Looks like FTP is needed.
>> >
>>
>> Interesting.
>>
>> A few facts about the VMS NFS server:
>>
>> 1) I believe that we (they, I don't work there anymore) recommend that 
>> high
>> water marking is disabled. This is an XQP feature that causes awful
>> performance issues for NFS.
>
> Interesting...  I know of a bug in *TCPWare's* NFS server where the
> workaround is to enable high-water marking.  Don't know if this is
> fixed in the latest version
>
>
>
>>
>> 2) When a non-stream file is accessed, the file is scanned to:
>>     a) determine the actual length of the file in its stream equivalent 
>> form
>> (as it will be returned on read), and
>>     b) create an offset translation map for on-the-fly conversion to 
>> stream
>> format.
>>
>> If I were to be debugging this problem, I would be looking for a length
>> mismatch and perhaps even a confusion between file-handles (they have
>> embedded options). But then, while I do work on NFS today, I no longer do 
>> so
>> for HP.
>>
>> John
>
>
>
> -- 
> John Santos
> Evans Griffiths & Hart, Inc.

That probably addresses Unix sparse files (which VMS did not support).

If you leave it to the XQP, it is quite expensive as it can tie up a thread 
to
fill all of that space with nulls using smaller IO. I chose to have the NFS
server zero fill the area itself with massive asynchronous IO. 