[Date Prev][Date Next] [Thread Prev][Thread Next]
[Thread Index]
[Date Index]
[Author Index]
adding Tekram 390U3W as second SCSI adapter
- From: Michal Szymanski <msz astrouw edu pl>
- To: Enigma List <enigma-list redhat com>
- Subject: adding Tekram 390U3W as second SCSI adapter
- Date: Wed, 30 Jan 2002 06:34:35 +0100
Hello,
I've just added Tekram 390U3W SCSI controller to a dual PIII/850 system
with ASUS P2B-DS MoBo. The first controller is an on-board Adaptec.
Two questions:
1. The Tekram card was recognized by kudzu at boot time and configured
as Symbios with sym53c8xx module. It seems to work fine. Is it worth
trying to rebuild the kernel with the Tekram driver provided on their
site?
2. As for now, the only device attached to to the Tekram card is an
Ultrium tape. Even if it is switched ON at boot time, it is not
automatically configured by the system (as it was before, being attached
to the primary controller, but then with some disks also, so I guess
that was the purpose the aic7xxx module was loaded automatically and was
able to recognize the tape drive also).
What is the most elegant way to add such a tape device to the system.
I can use "modprobe sym53c8xx" but this requires root priviledges.
I could put this command into /etc/rc.d/rc.local but if the drive
is not ON at boot time, it will not work.
My understanding (maybe wrong :-) of the idea of scsi_hostadapter
aliases in /etc/modules.conf and of the modules themselves was always
that it should allow loading needed modules on user's (I mean ordinary
user, not root) demand. So I would expect that any attempt to use the
device at scsi no N, channel C, id I and lun L, should load the
appropriate (modules.conf!) module if not already loaded.
I found extremely unfortunate the idea that Linux seems to follow, that
SCSI devices (/dev/stXX tape, /dev/sdX disks) are assigned dynamically
to real SCSI devices every time the module is loaded, that is, at every
boot at least. This really makes life hard sometimes. The two obvious
examples could be:
- adding new SCSI disk thathas (for any possible reason) the ID number
*less* than any of the disks already present in the system.
The new one will be named "sda" and all others will be shifted one
letter up, crashing immediately /etc/fstab assignments, RAID definitions
in /etc/raidtab etc.
- booting the system equipped with two or more tape drives (this is my
case - I have Ultrium, DLT and DAT drives) with any of the drives OFF.
All the /dev/stN and any possible symbolic links one could have
(as e.g. /dev/ultrium -> /dev/st1) are also messed up at once.
And, having a looong experience with tape usage, I can assure you that
it is not a smart idea to keep the tape drive ON for a long time when
not in use. Unless one has a sterile, completely dust-free environment.
It would be much better (IMHO) to follow rather the Sun's idea present
in Solaris (and earlier in SunOS) so that the devices are created
(if not present) only at special "reconfiguration" boot. And ever since
then they are available. As for the disks, they are always addressed
using explicit SCSI numbers for controller, channel, id and lun.
So the disk jumpered as ID=4 on 1st controller and 1st channel will
always be there, independently of other disks configuration.
any comments would be welcome,
regards, Michal.
--
Michal Szymanski (msz astrouw edu pl)
Warsaw University Observatory, Warszawa, POLAND
[Date Prev][Date Next] [Thread Prev][Thread Next]
[Thread Index]
[Date Index]
[Author Index]