So my son decided that he liked the graphics on my XFX GeForce 7950GT. So I decided to see if it would work in his system. It currently has a PCI card - Pny GeoForce 5500 or 5200. When I first got the MSI K8N MASTER2-FAR motherboard in his system, it would not boot a PCIX card.

Well, I got it to boot this one. I ended up flipping their SLI/non-SLI card thing and it was a pain. It was hard to insert correctly. And form web searches I'm not the only one who had problems with both it and getting PCI Express video cards working.

Anyway, I couldn't get the drivers to install on the card. It would get almost to the end and lock up. I'm pretty sure it was changing video modes to one higher than my monitor supported. And unlike X, I couldn't reset the mode via a keyboard. With the aid of XFX Customer Support, I did everything - all to no avail. I was happy with their support.

I'm pretty sure it is the MSI motherboard. After I got the card booting into low res, it started to complain about "Other PCI bridge device". And I think that was something to do with their little SLI/non-SLI card. Their manual sucks about explaining it - the pictures show it oriented in the same way for different textual positions.

And their driver disks still look exactly the same. There are two of them and they are not labeled as to what they contain. I could go on - I don't plan to buy from MSI again.

But I would buy XFX again.

So, in In Kernel Sharetab: Have a single file psuedo-fs working!, I wrote about how I was close to having that work done. I got a request from Peter Tribble (comments) to make sure that the ignore option was set such that normal df output would not show the sharetab? I replied yes, since it sounded reasonable.

Well, I was wrong. :-) It turns out that the pseudo-filesystems like mntfs and objfs are loaded directly in vfs.c. And as such, they have no knowledge of /etc/vfstab. So they gleefully ignore 'ignore'.

I looked at changing the interface into vfs_mountfs(), but I hadn't run that past a design check and when I did it informally, the decision was that all of the pseudo-filesystems had to work the same way. And to change the others would require a PSARC review. I.e., we would be changing the expectation that these filesystems would appear in a published public interface.

So I decided to ship the code without setting 'ignore' and to work in the background on getting a new review approved. It turns out we have a collection of bugs requesting this change for the other filesystems.