Unable to read tape after writing with 128k blocksize

[gschadow]

(sa1:isp0:0:5:1): 131072-byte tape record bigger than supplied buffer
(sa1:isp0:0:5:1): tape is now frozen- use an OFFLINE, REWIND or MTEOM command to clear this state.

of course I am reading with

dd if=/dev/nsa1 bs=$((128*1024))

so the buffer is exactly 131072. And of course I have tried to set mt blocksize 131072 or larger to no avail.

I am getting these errors when trying to read from any of my backup tapes written with 128k blocksize. Writing was no problem, but I can't figure out how to read. I have that on FreeBSD 6.2 and 6.0 with HP LTO-2 drives through QLogic fibrechannel HBA.

The strange thing is that on a different machine of the same hardware (may be slightly different QLogic HBA) with HP LTO-1 drive I can read the tapes that I write (with exactly the same method.)

I can read from the exact same tape from a Solaris 10 system.

I am afraid that something is limiting my buffer size to 64k even inside the system. I would upgrade to 7.2 if this was a known bug, but don't have time to upgrade just for trying it blindly.

thanks,
-Gunther

 

[mav@]

Before 8.x, FreeBSD didn't bother much with maximum supported transfer size. I don't know about sa, but da and cd drivers and most of HBA drivers were limited by 64K transactions. Now that area was improved for da, cd and some HBAs, but I have no idea about sa and isp, at least isp doesn't report this capability.

 

[gschadow]

Thanks. Funny, this stuff today seems to be from stone age. I have long done away with my cage in the server room, no more tapes.