Finally multiple-files-per-tape is working on this tape drive, so I started putting 4 dumps per tape. When I went to recover something from the last dump (restore ibsf 20 4 /dev/rst0), the "forward space file" operation timed out, along with driver complaints about "mailbox not free". Things went downhill rapidly as sd0 contains important parts of the system, and it was getting timeout errors also. Eventually it rebooted itself. 10 minutes is not enough time for a forward-space-file, or especially multiple forward-space-files. Writing or reading a full tape (Doesn't seem to vary a lot whether I use 525M, 250M, or 120M tapes, either) takes somewhere in the vicinity of 45 minutes to just under an hour. I suspect forward-space-file is done at the same speed as reading the tape. Separate timeouts for "forward space file" and "forward space record" might be appropriate. Fix: Apply the following patch, and rebuild the kernel. How-To-Repeat: Create a 6525 tape with multiple dumps on it. Fill up most of the tape, so the beginning of the last dump is well into the tape. restore ibsf 20 4 /dev/rst0 (where 4 represents the number of the last dump). With 6525 tapes and my tape drive, a blocking factor of 20 is required, according to complaints from the kernel if I forget to use it.
Responsible Changed From-To: freebsd-bugs->pst
State Changed From-To: open->closed Patch applied to -current.