OSNews

And what ?
If they are modules, should they not also compile under the newer kernel ?

have you ever built custom kernel?
Cisco VPN client, ipw3945 in past required constant patching otherwise module would not load under new kernel. nVidia occasionally would not even build.

Nothing new.
It does not mean that specific module will not built though.

I meant the modules for his Freeview card, they are pretty much a standard affair covered extensively on the linuxtv pages.

http://www.linuxtv.org/wiki/index.php/How_to_install_DVB_device_dri...

I have got them to work on almost all kernels, stock or custom.

Doubt it. Most mainframe shops probably run more than linux. All IBM has to do is say zOS is only supported under VM, and that's that.

I suppose there are some linux-only mainframe shops, but I need to see a couple of cases where $100,000 per processor hardware makes sense in a linux-only shop before I believe it's more than just a good way for IBM salespeople to make Porsche payments.

Huh? Do you realize that IBM has contributed the patches? Read the actual commit messages? IBM has been contributing in the past to Xen too. When virtualization becomes more of a commodity, they still get to make money from services from a expanded market.

What I don't get is what it is actually for. I mean, if you paid for the mainframe for the usual reasons, it seems natural you wouldn't mind paying for VM, which has been doing the job for 30 years. Heck, if IBM thought it would matter, sales wise, they could give VM away.

Not that I care that they did it, I just wonder who is going to use it. But then I feel that people read too much into the whole 390 Linux thing: IBM had no viable UNIX for the mainframe, some customers have the mainframe as their strategic platform and wanted to consolidate a little UNIX here and there, and Linux was convenient. Some see it a crown jewel in OSS acceptance (because mainframe is teh awesome!), but I don't really get how enabling one of the world's most closed computing systems is consistent with OSS goals. But then again, I don't buy mainframes

One possible reason: consistency. When you have x86 in more numbers, the management frameworks using it such as libvirt or virt-manager can be reused for other architectures if the underlying technology is same or similar. While libvirt can abstract away these details, it is useful when it is ported and available everywhere.

S/390 can support nested virtualisation in hardware, I think. You might well be running z/VM on the bare hardware and Linux in guests (I understand this is normal). But this should make it easy to run guest operating systems inside those guests. Since I think the hardware is designed with this in mind, it might not be as prohibitive, performance-wise, as you might think.

One advantage of this would be that you can carve up the machine coarsely using z/VM and give the resulting VMs to various business groups or individuals and then say "If you need more partitioning, use KVM and do it yourself".

I don't know if anybody ever runs Linux on the bare metal hardware. I guess it's not inconceivable that IBM are looking at deprecating z/VM in favour of leveraging Linux+KVM but I'm not sure getting rid of z/VM would be The IBM Way. Just having KVM available for nested virtualisation seems like a useful result in itself, so I assume that's what they're thinking of.

I'll buy that explanation

Look, if you want Windows only news why don't you bugger off to a Windows-only site then This is OSnews after all. Linux news might not be riveting, but it is on-topic.

And yes, Linux has only around 1% percent of the desktop market. So what. Only 2% of people have IQs over 130 (or below 70, if you think that's the Linux user-base, ). So fr!kken what. Linux has far greater penetration in the server market, which some of OSnews' readers might actually be interested in as they do real work instead of playing games.

I ran redhat 5.2 on an old ultra 5. Everything worked, audio was a bit of a chore, if I remember correctly.

Heh, yeah, I actually tried Ubuntu on the Sun T1000, but really I am interested in hearing about commercial deployments.

One interesting thing that I did not have time to explore was that the T1 uses 8k pages, but 8k-block ext3 would not mount. I think it was defined out at kernel compile time, or maybe not in newer kernels at all, but regardless I guess the mythical 16tb ext3 filesystem still is only a reality on Alpha