QEMU 1.3 supports new device pass-through technology
After six years of work, the official QEMU now includes all the changes that have been in QEMU-KVM, a fork modified to support KVM virtualisation. The news comes as part of the announcement of the release of QEMU 1.3. This should allow the KVM developers to drop their QEMU fork. The QEMU fork has been used by default together with the KVM hypervisor in the Linux kernel to virtualise PC hardware for operating systems; some distributions already use QEMU and KVM already, rather than QEMU-KVM.
QEMU 1.3 also offers support for VFIO (Virtual Function I/O) as introduced in the Linux kernel 3.6. VFIO is a technique for passing through of PCI and PCIe devices by the host system to guest systems to give them low latency and high throughput access without the risk of them being accessed by the host. Several bugs in QEMU's AHCI controller have also been fixed so that it works better with recent Windows versions. New support has been added for live disk mirroring, also known as "storage migration", which allows blocks to be copied from one running QEMU image to another, ready to switch to that other image. Another block enhancement is the addition of live block commits ("snapshot deletion") which can move data from a snaphot of an image in the image chain to another. A QEMU server for Network Block Devices (NBDs) has also been added and the developers have extended support for the Gluster cluster file system.
According to the developers, QEMU 1.3 also offers much better support for USB3 and the new Paravirtualised Random Number Generator allows guest systems to obtain random data from a hosts hardware random number generator. When working with Spice clients, QEMU will now only send changed areas of the screen to the client when working in legacy VGA mode. The TCG (Tiny Code Generator) now supports the SMEP (Supervisor Mode Execution Protection and SMAP (Supervisor Mode Access Prevention) features of the kernel. Other changes are detailed in the changelog for QEMU 1.3.