dan@contex.UUCP (Daniel Easly) (07/31/90)
We are experiencing a problem with the "tpsc" driver supplied by SGI. It seems that any long tape command (rewind, fsf 10, ...) the driver holds the bus until the operation has finished. This is causing significant delays for other users as all users are effected. This driver was supplied in source form with Release 3.2. Is there a new asynch version? Was this an oversight? Any help on resolving this would be greatly appreciated!!! --------------------------- Daniel Easley Xyvision Design Systems contex!dan@uunet.uu.net WORK : 617-245-9004 (x5573) -- --------------------------- Daniel Easley Xyvision Design Systems contex!dan@uunet.uu.net WORK : 617-245-9004 (x5573)
olson@anchor.esd.sgi.com (Dave Olson) (07/31/90)
In <1363@contex.UUCP> dan@contex.UUCP (Daniel Easly) writes: | We are experiencing a problem with the "tpsc" driver supplied by | SGI. It seems that any long tape command (rewind, fsf 10, ...) the | driver holds the bus until the operation has finished. This is | causing significant delays for other users as all users are effected. I think you are mistaking the driver blocking a process for the device disconnecting from the SCSI bus. For historical reasons, ALL tpsc commands are non-interruptible. Programs doing i/o or using the tpsc ioctls will block until the drive completes the request. This driver has 'always' been like this, and is not likely to change in the near future. This is not unlike most disk drivers, except that tape drives tend to have very long lived single operations (as long as 2 hours in the case of an 'mt erase' on an 8mm tape). All of the SGI qualified SCSI tape drives disconnect from the SCSI bus on long operations. The only exception is if you change the variable 'scsi_enable_disconnect' in master.d/scsi to support some device that does not work correctly with disconnect supported; if you do so, NO SCSI devices will disconnect on any command. SGI ships NO devices requiring this; it was done to enable a number of customers to use drives like the Sony C501 optical drive which did not work correctly (at least with several revs of firmware) in an environment where disconnect is supported. | This driver was supplied in source form with Release 3.2. Only if you signed a special contract with SGI. I am unaware of any SGI device driver that has been generally released in source form. (I seem to recall a company named Contex in Boston that did sign such a contract so that they could support a 9 track SCSI tape drive of their own choice.) | Is there a new asynch version? Was this an oversight? The driver changed to support a few new features, and some bugs were fixed, but no significant changes in driver operation were made. | Any help on resolving this would be greatly appreciated!!! I would be happy to try and help you (unofficially!) if you would E-Mail me the specifics of your system configuration and what you are doing. -- Dave Olson Life would be so much easier if we could just look at the source code.