NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/sys/amiga/misc-part1/5429

Internet Message Format | 1996-08-05 | 5.8 KB

Path: news-feed.iguide.com!usenet
From: [e-mail] (Joanne Dow)
Newsgroups: comp.sys.amiga.misc
Subject: Re: 9GB SCSI II Controller Card for the A3000?
Date: 21 Feb 1996 08:48:44 GMT
Organization: News Corporation
Message-ID: <4gem9c$[e-mail]>
References: <[e-mail]> <[e-mail]> <[e-mail]>
Reply-To: [e-mail] (Joanne Dow)
NNTP-Posting-Host: bix.com
X-Newsreader: NetBix 1.250
X-Newsreader-Author: [e-mail] (Loren Wilton)
    This message has been posted with development version software.
    If there are any problems with the message format, send me mail.

In article <[e-mail]>,
   [e-mail] (Mike) wrote:
>
>Hi, James Cooper , on 09-Feb-96 14:45:59 you scribbled....
>>In article <[e-mail]>,
>>[e-mail] (Dale Frameli) writes:
>>>Anyone know if there's a SCSI II controller board that's both A3000
>>>compatable AND recognizes, partitions and formats a 9GB hard drive?  I'd
>>>really like to get an 060 accelerator with a SCSI II option that would
>>>support such a drive...
>>> Maybe CyberStorm or DKB will build one.
>
>>The A3000 internal SCSI interface will handle 9GB (and larger) drives
>>with no problems at all.
>
>>The problem is that neither dos.library nor the filesystem can easily
>>deal with anything over 2GB.  It doesn't matter *what* SCSI interface
>>you use, until the OS is updated, you're stuck.
>
>does this mean you can't use disks >2GB or just that you can't have partitions
>over 2GB ??

OK - I see it is time to sit you guys on mommy's knee and tell you all the facts
of life as pertains Amigas and large disks - yet again. (<sigh> This info is
STILL up on BIX and currently available to anyone who joins and looks for it.)

1) No device driver can accept a "Read()" or "Write()" to an offset that is
beyond 2^32 bytes. The device driver interface is a (signed) 32 bit integer.
(The signed aspect of it generally does not matter on good device drivers. I
suppose there might be one out there that performs signed arithmetic
internally, though.) Hence the absolute size limit for any filesystem that does
NOT use SCSI direct access is 4 gigabytes. In point of fact here in Wizardess'
Castle we have a 4 gigabyte device perking along just fine. (Its top 100
megabytes or so are not readable, though, as the device's real size is that much
more than 2^32 bytes.) This limitation affects both absolute maximum partition
size and absolute maximum disk size (without kludges.)

2) If SCSI direct access is used the limitation on device access becomes some
2^24 BLOCKS. Block size can be virtually anything (on some drives) from 180
bytes through <very large>. (I have seen 16k blocks and heard of larger.) This
leads to potentially VERY large devices being serviceable through SCSI direct.
Do note that all Amiga filesystems with which I am familiar speak in powers of 2
times 512 bytes. (eg 512, 1024, 2048, ... bytes.) I have heard that AFS will
speak SCSI direct for its professional versions. Whether this "will" is as in
present tense "if you want it to" or is as in "someday it will" is something I
do not know. Again this affects absolute maximum partition and disk sizes.

3) The file access model involves a signed 32 bit integer offset and read size.
As it turns out due to standard C library "features" (or so I understand) the
"signed" part of this integer is taken seriously. Hence absolute maximum FILE
size for using C library function access is 2 gigabytes. I am, to tell you the
truth, unsure how this involves the OS Read() and Write() functions. I have
heard that they also take the "signed integer" concept seriously. (Please do not
ask me what a negative size read means. This concept still has me a tad
bemused.)

4) Practical effects may also come to bear on maximum partition size. For
example presume you have a 2 megabyte Amiga and a full 4 gigabyte partition with
an average filesize of 10k bytes. This means you have a grand total of about
400,000 files. Suppose the disk becomes unvalidated and will not normally
revalidate. Hence you have to use DiskSalv to rebuild the disk with those
400,000 files on it. Suppose Disksalv uses only 10 bytes per file for
reconstructing the disk, which seems to me to be most optimistic, your poor 2
meg Amiga will run out of memory half way through. Hence you need a smaller
partition size to handle this practical issue. A second misfeature of this large
a partition is the time it takes to scan all these directory entries to perform
a normal revalidation. Even 450 megabyte partitions on a Maxtor Magic drive can
take long enough that you learn anew about stomache acid generated by nerves.

5) It is possible to build a "psuedo-device" that can sit on top of the actual
SCSI device and give the "Read()" and "Write()" access points to the hardware
that are needed for normal filesystems to work. One such beastie is called
something like "giga.device". It may work fine if you really need it. The
technology certainly should be easy enough to generate so it is easy to believe
that giga.device works fine. As it happens I have not had reason to play with
this thing yet so I cannot say one way or the other from personal experience. I
figure on making my own kludge if and when I need it.

I hope this sheds some light on things....

>Does AFS solve either of these btw?

See above - it may or may not depending on the interpretation of the "will" that
I have read here on UseNet a few times.

>Mike
>
>--
>---------------------------------------------------------------------------
>Mike Redrobe - [e-mail]         MikeRR on #Amiga
>---------------------------------------------------------------------------
>

{^_^}     Joanne Dow, Amiga Exchange Editor on BIX, aka The Wizardess
          [e-mail], [e-mail]