On 02/07/2012 02:59 PM, Christian Hoff wrote:
Instead the format has some disadvantages: - It uses up 8 bytes where 3 bytes would be sufficient in order to store both the target ID and LUN number information - The format limits us to 255 target IDs. I agree that the LUN limit is probably more a theoretical and not a practical one, but 255 target IDs could become a limitation in the future.
It also provides better upwards-compatibility in case the limitations are actually hit. If I had used "uint8_t target; uint16_t lun;" an extension would require a feature bit and a new struct. With 8-bytes, you can just expand the definition. That pretty much sums it up.
But again, I don't think the limitations are serious. A MegaSAS header has room for 256 targets too, VMWare has only 15, Hyper-V has 1 (and 2 channels, but I think that's an off-by-one), and you can always have multiple HBAs on the same guest.
Nonetheless I think that virtio-scsi is a useful project and addresses many of the limitations imposed by virtio-block. The fact that I am still persisting has more to do with interest in the project rather than wanting to keep the code from going upstream.
No problem. :) Paolo -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
