]> git.ipfire.org Git - thirdparty/kernel/stable.git/commit
KVM: SVM: Allow guest writes to set MSR_AMD64_DE_CFG bits
authorSean Christopherson <seanjc@google.com>
Wed, 11 Dec 2024 17:29:52 +0000 (09:29 -0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Fri, 27 Dec 2024 13:02:17 +0000 (14:02 +0100)
commit3931567bf1757a2ce9c03bb3c935a7c036636c4c
tree7eef04fcbe2be17dcdec3f8d809380eb4305d0a4
parent16846e8e0ace406f7a8887c8c1eb060136291229
KVM: SVM: Allow guest writes to set MSR_AMD64_DE_CFG bits

commit 4d5163cba43fe96902165606fa54e1aecbbb32de upstream.

Drop KVM's arbitrary behavior of making DE_CFG.LFENCE_SERIALIZE read-only
for the guest, as rejecting writes can lead to guest crashes, e.g. Windows
in particular doesn't gracefully handle unexpected #GPs on the WRMSR, and
nothing in the AMD manuals suggests that LFENCE_SERIALIZE is read-only _if
it exists_.

KVM only allows LFENCE_SERIALIZE to be set, by the guest or host, if the
underlying CPU has X86_FEATURE_LFENCE_RDTSC, i.e. if LFENCE is guaranteed
to be serializing.  So if the guest sets LFENCE_SERIALIZE, KVM will provide
the desired/correct behavior without any additional action (the guest's
value is never stuffed into hardware).  And having LFENCE be serializing
even when it's not _required_ to be is a-ok from a functional perspective.

Fixes: 74a0e79df68a ("KVM: SVM: Disallow guest from changing userspace's MSR_AMD64_DE_CFG value")
Fixes: d1d93fa90f1a ("KVM: SVM: Add MSR-based feature support for serializing LFENCE")
Reported-by: Simon Pilkington <simonp.git@mailbox.org>
Closes: https://lore.kernel.org/all/52914da7-a97b-45ad-86a0-affdf8266c61@mailbox.org
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: stable@vger.kernel.org
Reviewed-by: Tom Lendacky <thomas.lendacky@amd.com>
Link: https://lore.kernel.org/r/20241211172952.1477605-1-seanjc@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
arch/x86/kvm/svm/svm.c