I am having trouble understanding the difference between a passthrough device and a named device and when you would use one or the other to access equipment.

As an example, we have a tape library and giving the command
"camcontrol devlist" gives the following output:

akx[22]# camcontrol devlist
<SPECTRA 215 1014> at scbus3 target 3 lun 0 (pass4,ch0)
<SONY SDX-300C 04c7> at scbus3 target 8 lun 0 (pass5,sa0)

The SPECTRA 215 is the tape library the SONY SDX is the tape drive in the library.
To address the tape library a command such as "mtx -f /dev/pass4 command" is used, however to address the tape unit then /dev/sa0 would be used as in
"/sbin/dump -0uaf /dev/sa0 /dev/da4s1e".

Could someone give me a clue on when you address a passthrough device and when you address a named device, or at point me to some documentation that might clear this up?

Thanks
Thumper

They are both the same type of objects. I don't know what os you're using so I talk talk specificly to that. But imagine a version of unix with device files like /dev/mt0 to talk to a tape drive. Any device file has a major and a minor number. The major number identifies which driver is being used. The minor number is passed to the driver for its use. It the case of a typical mt driver, it ids the particular tape drive. So you can open /dev/mt0 and read or write to the tape drive via it. You can also invoke the ioctl system call for special purposes like rewinding the tape. This is the typical type a tape driver that existed for years and it is what you call a "named device".

Now someone invents a new type of tape drive that can hold 5 tapes. How to control that? The mt ioctl doesn't handle stuff like that. Well one answer is to redefine the rewind or the unload command to now mean "go to the next tape". A lot of juke box style drives do just that and call it stacker mode. But we want something more... we want to just jump to tape number 4 (as an example) regardless of where we were. The tape drive can do that but the mt driver doesn't have a way to send the right command. All we need is some quick a dirty way to send a particular scsi command to the tape drive. This is where the pass-though driver comes in. It does not know it is controlling a tape drive. But you can give it a scsi command and it can send it to the device. And it can even return a status code. But that is it. It just passes simple commands to a device. Now a real smart program figures out what scsi command is needed to jump to tape number 4, it uses the pass though driver to send it, and it gets a status code back. So now we are using a secondary driver to access extra device features that the primary driver can't control.

But the pass through device is a special file too with a major and minor number. You don't use it to tranfer large blocks of data though. Just special commands.

In this case both special files referred to the same device. But often there will be a collection of tape drives in a juke box with a single pass-though device for the box itself.

We would use the mt driver where we can. And we use the pass-though driver only when we must. Drivers like mt have a man page that describes which ioctl command it can do. Pass though drivers usually have sparse man pages because you can't document all possible scsi commands.