TODO: Difference between revisions

From KVM
(remove packaging todo; picked up by Fedora)
(add struct kvm_vcpu cleanup)
Line 15: Line 15:
* Consolidate the various functions that read and write guest memory.  There is some duplication there.
* Consolidate the various functions that read and write guest memory.  There is some duplication there.
* Consolidate the inb/outb emulation helpers into x86_emulate.c
* Consolidate the inb/outb emulation helpers into x86_emulate.c
* <code><nowiki>struct kvm_vcpu</nowiki></code> has many vmx specific fields.  Move them into a new member <code><nowiki>vmx</nowiki></code>.  Have each arch allocate the vcpu so its arch-specific members can be addressed without indirection.


MMU related:
MMU related:

Revision as of 11:37, 1 April 2007

TODO


The following items need some love. Please post to the list if you are interested in helping out:

  • Real mode support: VT support for real mode is terrible, so we need to do it in software. This means extending the x86 emulator (x86_emulate.c) to handle more instructions, and changing the execution loop to call the emulator for real mode.
  • Move the apic/ioapic/pic/pit emulation into the kernel. This will simplify in-kernel paravirtualized devices (as there is no need to exit to userspace to inject an interrupt) and guest SMP. Some bitrotted code already exists.
  • Drop the preemption counter while doing mmu stuff, and turn kvm->lock into a mutex.
  • Emulate the architectural performance monitor counters/msrs, for Linux nmi watchdog support.

The following smaller scale tasks can be a nice entry point to someone wishing to get involved:

  • Enforce the virtual time stamp counter monotonicity even when a vcpu is migrated to another physical cpu. (in progress, Leonard Norrgård)
  • Avoid saving/loading the floating point unit state on each vm entry/exit (Anthony Liguori has partially working code).
  • Trap #UD and emulate sysenter/syscall/sysret/sysexit. These instructions don't exist on all cpus in all modes, so they hinder cross-vendor migration
  • Consolidate the various functions that read and write guest memory. There is some duplication there.
  • Consolidate the inb/outb emulation helpers into x86_emulate.c
  • struct kvm_vcpu has many vmx specific fields. Move them into a new member vmx. Have each arch allocate the vcpu so its arch-specific members can be addressed without indirection.

MMU related:

  • Attach kvm memory to a Linux address_space so that guest memory can be paged out.
  • Support large pages (in conjunction with the item above) so that if the guest uses a large page mapping and the guest memory is backed by hugetlbfs, a large-page pte is created.
  • Improve mmu page eviction algorithm (currently FIFO, change to approximate LRU).
  • Implement kvm_mmu_post_write() so that page table updates don't take an immediate page fault after the write.
  • Add a read-only memory type. This can be used to speed up APIC reads, which is fairly important for Windows guests.

For the adventurous:

  • Emulate the VT and SVM instructions, so that kvm can run in a virtual machine. Test by running a VM in a VT guest in an SVM guest on VT hardware, as well as running a VM in an SVM guest in a VT guest on SVM hardware.