Am 26.07.2011 16:46, schrieb Corey Bryant:
On 07/26/2011 10:05 AM, Kevin Wolf wrote:
Am 26.07.2011 15:02, schrieb Christoph Hellwig:
I have to say I really hate it. We've been working hard on getting rid of special cases in the qemu block layer, and this sprinkles them all over. I'd recommend to fix your security model instead. I think the problem here is more with the implementation that with the intention.
I agree that you just can't do this. A patch adding support for a fd: protocol should touch block/fd.c and nothing else. You can add some supporting patches that extend the generic block layer to support e.g. formats that can't reopen. However, if you touch the code of other block drivers, you're doing it wrong.
I'll look into this approach, but on the surface it seems like this could prevent a lot of code reuse in block/raw-posix.c.
What I meant here is more about modifying qcow2, VMDK etc. I still need to look into the details of what you share with raw-posix.c. Kevin