[Info-vax] SUCCESS Re: Error using fibre channel target on FreeBSD from OpenVMS

[Matthew R. Wilson]

SUCCESS!

It turns out there was another confounding factor in my experiments:
dodgy hardware.

Not sure if it's a dirty optic, failing SFP port, or what. But I
switched to doing some heavy I/O testing by installing Linux on the
Itanium system and using the same FC targets... and Linux started
logging all sort of dropped frame messages on the console. This after I
started realizing I was encountering some very odd seemingly "stateful"
behavior in OpenVMS when trying to use the FC target: sometimes commands
would work, sometimes they wouldn't, sometimes they'd hang for a while,
etc., all seemingly without reason. Should have suspected a lower-level
hardware problem earlier...

I switched to a second FC HBA I have for the Itanium, switched SFPs in
the SAN switch, and while I still can't get the errors entirely
eliminated (I don't have anything to actually clean the optics here at
home with me at the moment), I can get through a couple GBs of data
transfer in Linux at a time without errors. So I figured I'd try from
VMS again... and it works perfectly. INIT runs instantly, I can mount
the volume, I can write and read data, etc.

<sigh>

Always something dumb/simple.

In the end, the *only* change I needed to make was to have the FreeBSD
target emulation report a lower SCSI version. I'm having luck with
SCSI-3.

For those playing along at home, around line 10211 of sys/cam/ctl/ctl.c,
you can change:

	inq_ptr->version = SCSI_REV_SPC5;
to
	inq_ptr->version = SCSI_REV_SPC;

Then rebuild the FreeBSD kernel and OpenVMS appears to be happy.

When defining your LUNs in FreeBSD, you must also set option ident_info
to a four-byte number to control the UDID; for example 0x00000007 will
make the LUN appear as $1$DGA7 in VMS.

I still don't trust the FC connectivity until I can repeat the testing
from Linux without any interface errors, but with things mostly
reliable, I'm happy to report OpenVMS is working well with the FreeBSD
SAN volumes so far.

Thank you for all of your input and help throughout this process! I
ended up going on a wild goose chase with some of the SCSI behavior that
was easily explained away by hardware problems, but at least I learned a
lot.

Thanks,
Matthew