19 Jun
2006
19 Jun
'06
3:15 a.m.
Probably this is documented somewhere, but I really need a pointer or a
brief tutorial on the major/minor device numbers for mknod() and the device
names for MSCP drives in 2.11bsd.
If I have a really simple PDP with an RQDXn and one RDxx disk, then the
device name is conventionally /dev/ra0x and the first partition, ra0a is
(5,0), the second, ra0b, is (5,1), etc. Pretty easy.
If I have two drives on my single RQDXn, then the second hard disk is
/dev/ra1 and ra1a is (5,8), ra1b is (5,9), etc. I guess the offset of 8
must be the maximum number of partitions on a drive - OK, I'm still with
you.
But what if I have a second MSCP controller? Assuming that I've built the
kernel to handle it and modified dtab to autoconfigure it, that is. What
are the usual names and mknod() numbers for the drives on the second
controller?
Worse, what if the MSCP controller isn't a RQDX but is a real UDA/QDA ?
Now the drives have their own MSCP unit numbers that can be anything from 0
to 250 - where does this figure in?
Same question for TMSCP - what if I have more than one tape controller?
This case is easier, though, since TMSCP controllers normally have only one
drive associated with them.
Thanks,
Bob Armstrong
19 Jun
19 Jun
6:57 a.m.
On Sun, 18 Jun 2006 20:15:52 -0700
"Robert Armstrong" <bob(a)jfcl.com> wrote:
But what if I have a second MSCP controller?
/dev/ra1a is still (5,8). The disks are numbered in the same order as
they are found. (Modulo devices that are "nailed down" in the kernel
config file.) This is independent of the controler they are connected
to. Some years ago I used a MSCP ESDI disk controller and in addition a
MSCP SMD disk controller to connect two disks to my PDP-11/73 and IIRC
the SMD disk showd up as /dev/ra1[a-h].
At least this is "the generic BSD way". 2.11BSD may be a bit different
due to its age.
--
tschüß,
Jochen
Homepage: http://www.unixag-kl.fh-kl.de/~jkunz/
2:51 p.m.
New subject: [pups] Major/minor device numbers and device names for MSCP drives?
Jochen Kunz wrote:
The disks are numbered in the same order as they are found.
Is it safe to assume that drives are always discovered in ascending unit
number order, starting with the first controller and continuing with the
second ?
Is there any utility that will examine the running system and tell you
which drives and units were actually discovered?
init will say something like "ra 0 at ...." and "ra 1 at ...", but
that's
talking about controllers. AFAIK it says nothing about the drives
discovered.
Sorry to complain, but it seems like it can be a little bit ambiguous as
to whether BSD actually discovered the drives you think it should have.
This is especially true if you have something like a SCSI controller where
it may not be immediately obviously which drives are online or what their
MSCP unit numbers are.
And it's important to know which actual disk drive you're writing on :-)
Bob
3:08 p.m.
On Sun, 18 Jun 2006, Robert Armstrong wrote:
Worse, what if the MSCP controller isn't a RQDX
but is a real UDA/QDA ?
Now the drives have their own MSCP unit numbers that can be anything from 0
to 250 - where does this figure in?
Note that this isn't really anything special.
You can both have a RQDX with an extender card which allows you to set
other unit numbers on RD disks, and if you have two UDA controllers, they
could both have disks starting from number 0.
The other response about automatic numbering of disks is true of NetBSD,
as well as Ultrix. However, I'm not sure that BSD2 do this.
Unfortunately I can't remember for sure.
But don't you actually tell the unit numbers in the configuration file,
along with the controller? Or do BSD2 also do a full autodetect and
connect of MSCP disks?
If not, it could work in several ways, but I would suspect that disks on
the second controller would start with minor # 64 (8 disks * 8 partitions
per disk is I believe the default). But that assumes that the first disk
found will be "0", no matter what the physical unit number is set to.
Unless of course, this also is set in the configuration file. :-)
Same question for TMSCP - what if I have more than
one tape controller?
This case is easier, though, since TMSCP controllers normally have only one
drive associated with them.
Well, DEC only have TMSCP controllers with a single unit for the PDP-11.
Third party controllers can have several units...
Johnny
9:43 p.m.
New subject: [pups] Major/minor device numbers and device names for MSCP drives?
Robert Armstrong wrote:
Probably this is documented somewhere, but I really
need a pointer
or a brief tutorial on the major/minor device numbers for mknod() and
the device names for MSCP drives in 2.11bsd.
This is what I remember about it all
when I struggled with it a couple
of years ago.
The 8 bit minor number has 3 parts: the upper two bits are the
controller (4 controllers max), the middle 3 bits are unit number (8
drives per controller) and the lower 3 bits are the partition (a-h).
The assignment of device names is static (or nearly static). The
discovery code for the controllers checks them in the order listed in
/dev/dtab and assigns a controller number to each that is discovered and
in that order (the nearly static part). Note also that DEC has a
standard order for controller addresses too (which might be different).
After that, drives are accessed using the 3 bit unit number which MUST
correspond to the MSCP unit number (a catch follows because of this).
Partitions are accessed with the 3 bit partition number.
The catch is that MSCP unit numbers are supposed to be global cross all
controllers and the microPDP-11/83 boot code assumes this. The boot code
can only see the first unit n that is encountered and 2.11BSD can only
see the drives with MSCP unit numbers less than 8.
If you only have one controller there is no problem. When you add
multiple controllers, things get more complex because you can only boot
the first 8 MSCP units if you want to be able to access them from 2.11BSD.
-chuck
20 Jun
20 Jun
3:45 p.m.
New subject: [pups] Major/minor device numbers and device names for MSCP drives?
C. H. Dickman writes:
The 8 bit minor number has 3 parts: ...
Thanks, Chuck - this is exactly what I wanted to know. Now I've got to
try it out and see if I can get my SCSI ZIP drive talking to BSD.
Thanks again,
Bob
19 Jun
19 Jun
9:53 p.m.
New subject: [pups] Major/minor device numbers and device names for MSCP drives?
Robert Armstrong wrote:
Probably this is documented somewhere, but I really
need a pointer
or a brief tutorial on the major/minor device numbers for mknod() and
the device names for MSCP drives in 2.11bsd.
If I have a really simple PDP with an RQDXn and one RDxx disk, then
the device name is conventionally /dev/ra0x and the first partition,
ra0a is (5,0), the second, ra0b, is (5,1), etc. Pretty easy.
If I have two drives on my single RQDXn, then the second hard disk
is /dev/ra1 and ra1a is (5,8), ra1b is (5,9), etc. I guess the offset
of 8 must be the maximum number of partitions on a drive - OK, I'm
still with you.
But what if I have a second MSCP controller? Assuming that I've
built the kernel to handle it and modified dtab to autoconfigure it,
that is. What are the usual names and mknod() numbers for the drives
on the second controller?
The second controller starts at ra8a (5, 8), and ra8a
must be MSCP unit 0.
Worse, what if the MSCP controller isn't a RQDX
but is a real
UDA/QDA ? Now the drives have their own MSCP unit numbers that can be
anything from 0 to 250 - where does this figure in?
An MSCP unit number greater
than 8 cannot be accessed from 2.11BSD.
Same question for TMSCP - what if I have more than
one tape
controller? This case is easier, though, since TMSCP controllers
normally have only one drive associated with them.
Don't know a think about
it...
Thanks,
Bob Armstrong
-chuck
6519
days inactive
6520
days old
6 comments
4 participants
participants (4)
-
C. H. Dickman -
Jochen Kunz -
Johnny Billquist -
Robert Armstrong