linux-kernel - [PATCH 0/27] Nested Paging Virtualization for KVM v3

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Date:	Mon, 6 Sep 2010 17:01:35 +0200
From:	Joerg Roedel <joerg.roedel@....com>
To:	Avi Kivity <avi@...hat.com>, Marcelo Tosatti <mtosatti@...hat.com>
CC:	"http://kvm.qumranet.com"@osrc.amd.com,
	linux-kernel@...r.kernel.org
Subject: [PATCH 0/27] Nested Paging Virtualization for KVM v3

Hi Avi, Marcelo,

here is finally the third round of my NPT virtualization patches for KVM. It
took a while to get everything running (including KVM itself) on 32 bit again
to actually test it. But testing on 32 bit host and with a 32 bit hypervisor
was a very good idea. I found some serious bugs and shortcomings in my code
that are fixed now in v3.

The patchset was tested in a number of combinations:

	host(64|32e)
		->kvm(shadow|npt)
			->guest(64|32e|32)
				->test(boot|kbuild)

	host(64|32e)
		->kvm(npt)
			->guest(64|32e|32)
				->kvm(shadow|kvm)
					->guest(64|32e|32)
						->test(boot|kbuild)

Only the valid combinations where tested of course, so no 64 bit on 32 bit
combinations were tested. Except for that I tested all of the above
combinations and all worked without any regressions.

Other changes since v2 are:

	* Addressed the review comments from v2:
		- Rebased everything to latest upstream code
		- renamed nested_mmu to walk_mmu to make its
		  meaning more clear
		- the gva_to_gpa functions are not longer swapped
		  between the two mmu states which makes it more
		  consistent
		- Moved struct vcpu page fault data into seperate
		  sub-struct for better readability
		- Other minor stuff (coding style, typos)
		- Renamed the kvm_*_page_x86 functions to kvm_*_page_mmu so
		  that they can be made more generic later.
	* Made everything work on 32 bit
		- Introduced mmu->lm_root pointer to let the softmmu shadow 32
		  bit page tables with a long-mode page table. The lm_root
		  page-table root always just points to the mmu.pae_root, so
		  this builds entirely on the pae-shadow code.
		- Split mmu_alloc_roots into a shadow and direct_map version to
		  simplify the code and to not break the direct_map paths anymore
		  when changing something in that function.
	* Probably other changes I forgot about

This patchset applies on todays avi/master + the three patches I sent end of
last week. These patches are necessary for some of the tests above to run.

For the curious and impatient user I put everything in a branch on kernel.org.
If you want to test it you can pull the tree from

	git://git.kernel.org/pub/scm/linux/kernel/git/joro/linux-2.6-kvm.git npt-virt-v3

Please review and/or apply these patches if considered good enough. Otherwise I
appreciate your feedback.

Thanks,

	Joerg

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/