Eric Lee Green wrote:
> ftp://ftp.estinc.com/pub/unsupported/mtxl-1.3.tar.gz
And no sooner do I announce it, than I find some problems with it on Linux.
Digging through Leonard's code I see he's using the SCSI_IOCTL_SEND_COMMAND
ioctl() rather than the SCSI generic interface. The problem is that, as I dig
through the kernel code for SCSI_IOCTL_SEND_COMMAND in the Linux kernel, I
find that it is limited to requests of PAGE_SIZE or less (hmm, 8K?). The
problem is that a fully loaded tape library could need me to request as much
as 11K of data for READ_ELEMENT_STATUS, thus causing SCSI_IOCTL_SEND_COMMAND
to burp :-(. Well, at least that explains why requesting the full 11K failed
every time (thus causing me to do the funky "request 8 bytes" trick to try to
get the # of bytes that would be reported, then allocating only that # of
bytes before issuing the 'real' SCSI READ_ELEMENT_STATUS command, but
apparently I don't really need to do that).
I'm looking at the SCSI generic interface and not liking what I see, but it
looks like I'm going to have to migrate the code to that on Linux if I want to
be able to work with BIG tape libraries... and besides, the Seagate DDS-4
autoloader appears to be burping on the little tricky thingy, though it works
fine with an Exabyte EZ-17.
Sigh. Back to coding. I'll let you know when I have something that does what
it's supposed to do :-}.
--
Eric Lee Green [EMAIL PROTECTED]
Software Engineer Visit our Web page:
Enhanced Software Technologies, Inc. http://www.estinc.com/
(602) 470-1115 voice (602) 470-1116 fax
