KVM: arm64: Fix race when enabling KVM_ARM_CAP_MTE
authorSteven Price <steven.price@arm.com>
Thu, 29 Jul 2021 16:00:36 +0000 (17:00 +0100)
committerMarc Zyngier <maz@kernel.org>
Thu, 29 Jul 2021 16:34:01 +0000 (17:34 +0100)
commitc4d7c51845af9542d42cd18a25c570583abf2768
treea192daad41b4e977b196a9ee9cf06cb8b357f067
parentfacee1be7689f8cf573b9ffee6a5c28ee193615e
KVM: arm64: Fix race when enabling KVM_ARM_CAP_MTE

When enabling KVM_CAP_ARM_MTE the ioctl checks that there are no VCPUs
created to ensure that the capability is enabled before the VM is
running. However no locks are held at that point so it is
(theoretically) possible for another thread in the VMM to create VCPUs
between the check and actually setting mte_enabled. Close the race by
taking kvm->lock.

Reported-by: Alexandru Elisei <alexandru.elisei@arm.com>
Fixes: 673638f434ee ("KVM: arm64: Expose KVM_ARM_CAP_MTE")
Signed-off-by: Steven Price <steven.price@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20210729160036.20433-1-steven.price@arm.com
arch/arm64/kvm/arm.c