I wrote:
>This might be the fix for PR #2727 submitted by me. My panic reason was
>done < 0, though. I haven't tested this patch yet, but will do so asap. 

I patched scsi/st.c and tried to do the long overdue backup with a new
kernel. dump ran for a while, but then the kernel panicked with a DMA
poll time out. I had si_options=1 in that kernel.

Okay, I patched arch/sun3/dev/si_obio.c to make si_options=7 and tried
again.

This time I got through dumping / , /var , /usr and /home with no
panics, but I still get the occasional
	st:si(0,4,0) soft error; data 00 00 00

I had about one per tape, but I wasn't worried, because a FreeBSD
machine with the same tape drive get those as well, without causing
any further problems.

However, I restored the /home tape to /scratch/tmp and ran a diff and
sure enough, I had two files corrupted, and I had had two soft errors
dumping /home. The files were the right size, but corrupted. Both just
happened to be gifs.

Also, using tar to write a tape, whenever a soft error occurs, tar
moans about a short write and quits.

How would one go about updating a PR, because I fear this info might
be lost being only posted to this list?

Jari Kokko