The memory core now has double indirection during dma or kvm I/O
dispatch: first we lookup a MemoryRegionSection from the memory page
table, then we look up the ram_addr to dispatch the I/O operation.
This patchset eliminates the second lookup. To do that, we now store a
MemoryRegionSection index in the iotlb instead of a ram_addr.
The patches were posted on the list a few days ago and received no
negative (or positive) review, and testing with the qemu images seems to
work fine.
Please pull from:
git://git.kernel.org/pub/scm/virt/kvm/qemu-kvm.git memory/core
----------------------------------------------------------------
Avi Kivity (5):
memory: make phys_page_find() return an unadjusted section
memory: store section indices in iotlb instead of io indices
exec: fix code tlb entry misused as iotlb in get_page_addr_code()
memory: dispatch directly via MemoryRegion
memory: get rid of cpu_register_io_memory()
cpu-all.h | 8 -
exec-all.h | 9 +-
exec-obsolete.h | 3 -
exec.c | 362
++++++++++++++++++++--------------------------------
memory.c | 13 +-
softmmu_template.h | 40 +++---
6 files changed, 168 insertions(+), 267 deletions(-)
--
error compiling committee.c: too many arguments to function
