Bug 32831

Simple use of the HP Colorado tape drive frequently results in the problems that can only be resolved by rebooting.  The process doing i/o is blocked on physst at an uninteruptable priority.  For example, I use the Amanda amlabel command to label a new tape and got the following:

ast0: WEOF command timeout - resetting
ata1: resetting devices ..  device dissapeared! 3 done
ata1-slave: timeout waiting to give command=a0 s=d0 e=00
ast0: failure to send ATAPI packet command
ast0: WEOF command timeout - resetting
ata1-slave: timeout waiting to give command=a0 s=d0 e=00
ast0: failure to send ATAPI packet command
ast0: WEOF command timeout - resetting
ata1-slave: timeout waiting to give command=a0 s=d0 e=00
ast0: failure to send ATAPI packet command
ast0: WEOF command timeout - resetting
ata1-slave: timeout waiting to give command=a0 s=d0 e=00
ast0: failure to send ATAPI packet command
ast0: TEST_UNIT_READY command timeout - resetting
ata1-slave: timeout waiting to give command=a0 s=d0 e=00
ast0: failure to send ATAPI packet command

ebb# ps alx|grep amlabel
 1000  4037  4036   0  -6  0  1524 1044 physst DL+   v0    0:10.08 amlabel Dail

This may be crappy hardware but the system shouldn't get wedged like this.

Fix: 

Haven't had a chance but submitting this in case there's a known fix or workaround (haven't looked in the code to see if perhaps there's a quirk table that needs tweaking).
How-To-Repeat: One case was when I ran the Amanda tapetype program that deduces the amount of data that'll fit on a tape by writing 32k records until it gets an error.  On hitting the end of tape the system wedged with the same kernel printfs as shown above.

The second case was running amlabel as described above.