[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <170906349657.3809281.8603439312604083486.b4-ty@google.com>
Date: Tue, 27 Feb 2024 13:28:18 -0800
From: Sean Christopherson <seanjc@...gle.com>
To: Sean Christopherson <seanjc@...gle.com>, Paolo Bonzini <pbonzini@...hat.com>,
Andrei Vagin <avagin@...gle.com>
Cc: linux-kernel@...r.kernel.org, x86@...nel.org, kvm@...r.kernel.org,
Zhenyu Wang <zhenyuw@...ux.intel.com>, Zhi Wang <zhi.a.wang@...el.com>,
Yan Zhao <yan.y.zhao@...el.com>, Yongwei Ma <yongwei.ma@...el.com>
Subject: Re: [PATCH v3] kvm/x86: allocate the write-tracking metadata on-demand
Added a few more KVM-GT folks. FYI, this changes KVM to allocate metadata for
write-tracking on-demand, i.e. when KVM-GT first registers with KVM. I tested
by hacking in usage of the external APIs, to register/unregister a node on every
vCPU before/after vcpu_run(), so I am fairly confident that it's functionally
correct. But just in case you see issues in linux-next... :-)
On Tue, 13 Feb 2024 11:23:40 -0800, Andrei Vagin wrote:
> The write-track is used externally only by the gpu/drm/i915 driver.
> Currently, it is always enabled, if a kernel has been compiled with this
> driver.
>
> Enabling the write-track mechanism adds a two-byte overhead per page across
> all memory slots. It isn't significant for regular VMs. However in gVisor,
> where the entire process virtual address space is mapped into the VM, even
> with a 39-bit address space, the overhead amounts to 256MB.
>
> [...]
Applied to kvm-x86 mmu, thanks!
[1/1] kvm/x86: allocate the write-tracking metadata on-demand
https://github.com/kvm-x86/linux/commit/a364c014a2c1
--
https://github.com/kvm-x86/linux/tree/next
Powered by blists - more mailing lists