SGI: Hardware

canavan wrote: Looks like I should have tried something smaller than 4GB cards. 2GB cards actually do work

If I could ask, with which of the card readers you mentioned? It would be an excellent opportunity to expand the list of "known working" firewire card readers beyond the current list of one.

The problem with 4GB cards is really just that IRIX only supports - at best - FAT16, apparently limited to 32k sectors and no support for either the non-standard 64k sectors required for a 4GB filesystem, or FAT32.

I've tried all three of my firewire card readers, and they are all working perfectly with 2GB media. Note that the label on the RW-19 calls it a RW-19 Rev. B.

recondas wrote: mediad ignores mine too, but manual mounting works ok. Since my main goal is to access photos from a digital camera, I leave the the CF card with the camera formatted FAT file system, and tell mount which type of file system to expect <and partition to examine>. Something like:

Code: Select all

mount -t dos -o partition=1 /hw/rdisk/1394/30e00100001419/lun0vol/c4p0 /usr/people/Fuel/FW/

The specific device ID <"30e00100001419" in my example> used in the mount command will be different for each device connected to the FW bus <but can easily be obtained from the output of fwprobe>.

Thanks to this lead, I was able to fx, mkfs_xfs and then mount a 1TB LACIE d2 Quadra drive this evening:
(where X for me is 6 - the FW card shows up as SCSI controller 6), then partition as an option drive
Note that the mount command uses /hw/disk rather than /hw/rdisk so as to access the block device file rather than the character file.

I do get a "Disk(6,2)" icon on the desktop, but Disk Panel tells me, "No volume header is seen on this disk" (though fx did not complain!) and suggests I initial the disk. But that fails with the error, "error code 256 on ppclose of fx".

At least the disk seems to be accessible, even if mounting it is not as easy as for a SCSI disk. I have updated the wiki for the drive and type of FW card.

Afterthought: if we can get the major/minor block device numbers from /hw/disk/1394/[device ID]/lun0s7/cXp0, would it be out of order to "mknod" the corresponding block device in /dev/dsk ? (as I understand it, IRIX populates /dev automagically from /hw so manually adding devices might foul things up; mknod takes me back to moving RISC iX from an ST506 drive to SCSI, where nothing was automagic!)

Andrew

The real problem is that "standard" FAT16 supports only 2GB. For anything larger, you'd have to use FAT32, which IRIX doesn't support. The 64k clusters I mentioned are non-standard and don't work for either IRIX or my Canon 40D.

I'm trying to set up an XFS-format drive. I think the work-around for failure to access the volume header is to run fx, sync the label to update /hw with the existing partition info, which generates block and character files to which a mount command can refer. Before fx "sync", I only see lun0vol in the FW device's folder, but afterwards, lun0vh and lun0sX where X is the partition number (i.e. 0, 1, 6 or 7 for root, swap, usr and option disk partitions).

16 Oct update - in the most recent session, the /dev/dsk/1394/[long device number]/lun0vol was joined by s0, s1 and vh so I can mount directly. Time to set up a small mounting script to do this from desktop.

But more testing to do before I start relying on it for any significant data ....