The release of project QEMU 7.0 has been announced. As an emulator, QEMU allows you to run programs built for one hardware platform on a system with a completely different architecture, for example, running ARM applications on an x86-compatible PC. In virtualization mode, QEMU's performance of executing code in an isolated environment is close to that of a physical system due to direct execution of instructions on the CPU and the utilization of the Xen hypervisor or KVM module.
Initially, the project was created by Fabrice Bellard to provide the ability to run Linux executable files compiled for the x86 platform on architectures other than x86. Over the years of development, full emulation support for 14 hardware architectures has been added, with the number of emulated hardware devices exceeding 400. More than 2500 changes from 225 developers have been made in preparation for version 7.0.
Key improvements added in QEMU 7.0:
- Support for the Intel AMX (Advanced Matrix Extensions) instruction set has been added in the x86 architecture emulator, implemented in Intel Xeon Scalable server processors. AMX introduces new configurable TMM 'TILE' registers and instructions for manipulating data in these registers, such as TMUL (Tile matrix MULtiply) for matrix multiplication.
- The ability to log ACPI events from the guest system through the ACPI ERST interface has been provided.
- In the virtiofs module, used for passing part of the host environment's file system to the guest system, improved support for security labels has been implemented. The vulnerability CVE-2022-0358, which allowed privilege escalation in the system by creating executable files in directories passed through virtiofs that belong to another group and are marked with the SGID flag, has been fixed.
- Increased flexibility for backing up active system images in operation (a snapshot is created, after which a copy-before-write (CBW) filter is applied to update the snapshot state, copying data from areas where the guest system writes). Support for images in formats other than qcow2 has been added. Access to the snapshot with a backup is now possible not directly, but through the snapshot-access block device driver. The capabilities to manage the CBW filter have been expanded, allowing for the exclusion of specific bitmaps from processing.
- The ARM emulator for 'virt' machines has added support for virtio-mem-pci, CPU topology definition for the guest system, and PAuth activation when using the hypervisor. KVM with the hvf accelerator. The 'xlnx-versal-virt' board emulator has added support for PMC SLCR and emulation of the Flash memory controller OSPI. New models for controlling CRF and APU have been added for the 'xlnx-zynqmp' emulated machines. Emulation of the extensions FEAT_LVA2, FEAT_LVA (Large Virtual Address space), and FEAT_LPA (Large Physical Address space) has been implemented.
- In the classic TCG (Tiny Code Generator), support for hosts with ARMv4 and ARMv5 CPUs, which do not support unaligned memory access and lack sufficient RAM to run QEMU, has been discontinued.
- In the RISC-V architecture emulator, support for the KVM hypervisor has been added, along with the implementation of vector extensions Vector 1.0 and instructions Zve64f, Zve32f, Zfhmin, Zfh, zfinx, zdinx, and zhinx{min}. Support for loading OpenSBI binary files (RISC-V Supervisor Binary Interface) has been added for 'spike' emulated machines. For 'virt' emulated machines, the capability to use up to 32 processor cores has been implemented alongside AIA support.
- In the HPPA architecture emulator, the ability to use up to 16 vCPUs has been provided, and the graphical driver for user environments HP-UX VDE/CDE has been improved. The ability to change the boot order for SCSI devices has been added.
- In the OpenRISC architecture emulator for 'sim' boards, support for using up to 4 CPU cores, loading an external initrd image, and automatic device tree generation for the booted kernel has been added.
- The PowerPC architecture emulator for 'pseries' machines has implemented the ability to run guest systems under the nested hypervisor KVM. Support for the spapr-nvdimm device has been added. For 'powernv' machines, support for the XIVE2 interrupt controller and PHB5 controllers has been added, along with enhanced support for XIVE and PHB 3/4.
- The s390x architecture emulator has added support for z15 extensions (Miscellaneous-Instruction-Extensions Facility 3).
Source: opennet.ru
