kvm: Use the address finalization code in the TLB

Information
Submitter:	Andreas Sandberg
Repository:	gem5
Branch:	default
Bugs:
Depends On:
Reviewers
Groups:	Default
People:

Description

Changeset 9739:8afcb103a82e
---------------------------
kvm: Use the address finalization code in the TLB

Reuse the address finalization code in the TLB instead of replicating
it when handling MMIO. This patch also adds support for injecting
memory mapped IPR requests into the memory system.

Testing Done

Quick regressions pass (with the exception of 02.insttest and 40.m5threads-test-atomic where I can't find the binaries) for all supported architectures. Boots Linux in KVM mode on x86 when applied together with #1860 (x86 requires APIC remapping magic from the TLB). Shouldn't affect ARM since finalizePhysical() is a no-op.

Issue Summary

Description	From	Last Updated	Status
Why here and not when it is used?	Andreas Hansson	June 2, 2013, 10:05 p.m.	Resolved
What happens to the packet on the IprRead/Write? The reason I'm asking is that the packet in question will go ...	Andreas Hansson	June 2, 2013, 10:05 p.m.	Dropped
Are there any locations in the code base where we assume that a virtual address is present? If so, could ...	Andreas Hansson	June 2, 2013, 10:12 p.m.	Dropped

A few questions, but overall it looks good.

src/cpu/kvm/base.cc (Diff revision 1)

Why here and not when it is used?

Show all issues

src/cpu/kvm/base.cc (Diff revision 1)

What happens to the packet on the IprRead/Write? The reason I'm asking is that the packet in question will go out of scope here and I'm simply not sure if it is still in use or not.

Show all issues

Andreas Sandberg June 2, 2013, 10:12 p.m. (June 2, 2013, 10:12 p.m.)

The function isn't supposed to hold on to the pointer, a similar pattern is used in the AtomicSimpleCPU.

src/mem/request.hh (Diff revision 1)

Are there any locations in the code base where we assume that a virtual address is present? If so, could you point out where in the comment above?

Show all issues

Andreas Sandberg June 2, 2013, 10:12 p.m. (June 2, 2013, 10:12 p.m.)

I don't think so. AFAIK, the only reason we make this check is because it is called from TLBs and it doesn't make sense to assign a new physical address to a request that doesn't have a virtual address. Devices normally use setPhys() instead, which updates both the address and length of the request.

Description:

~		Changeset 9692:651d765c1975
	~	Changeset 9739:8afcb103a82e

		kvm: Use the address finalization code in the TLB

		Reuse the address finalization code in the TLB instead of replicating
		it when handling MMIO. This patch also adds support for injecting
		memory mapped IPR requests into the memory system.

Diff:

Revision 2 (+27 -6)

Show changes

	src/cpu/kvm/base.cc
	src/mem/request.hh

Ship It!

You have a pending review.

Review Board 2.0.15

This change has been marked as submitted.

Screenshots

Files

Issue Summary

Description:

Changeset 9692:651d765c1975

Changeset 9739:8afcb103a82e

Diff:

Status: Closed (submitted)