MIRRORS/linux-imx
1
0
Fork 0
mirror of https://github.com/nxp-imx/linux-imx.git synced 2025-07-10 19:35:21 +02:00
Code Issues Actions Packages Projects Releases Wiki Activity

ANDROID: KVM: Fix fast-forward size in pkvm_mem_abort_range()

Browse Source 
When a block mapping could be installed, pkvm_mem_abort_range() was not
skipping the entire block range from the iteration leading to up to 511
useless calls to pkvm_mem_abort().

Also, it was possible to cover less than the requested range if the
original fault IPA wasn't PMD-aligned but a PMD mapping has been
installed.

The system can recover from both problems... however fix the iteration
to avoid useless hyp/host back and forths.

Bug: 278749606
Bug: 243642516
Fixes: 5d9808b907 ("ANDROID: KVM: arm64: THP support for pKVM guests")
Change-Id: I99e5225a6aea3869aa78ff52ada6949748b78c6f
Signed-off-by: Vincent Donnefort <vdonnefort@google.com>
This commit is contained in:
Vincent Donnefort 2024-08-09 10:26:31 +01:00
parent 92db83f6bf
commit aeaa3ba902
1 changed files with 3 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
arch/arm64/kvm/mmu.c
View File

@ -1804,7 +1804,7 @@ int pkvm_mem_abort_range(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa, size_t si
read_lock(&vcpu->kvm->mmu_lock);
ppage = find_ppage_or_above(vcpu->kvm, fault_ipa);
while (size) {
while (fault_ipa < ipa_end) {
if (ppage && ppage->ipa == fault_ipa) {
page_size = PAGE_SIZE << ppage->order;
ppage = mt_next(&vcpu->kvm->arch.pkvm.pinned_pages,
@ -1832,11 +1832,10 @@ int pkvm_mem_abort_range(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa, size_t si
* We had to release the mmu_lock so let's update the
* reference.
*/
ppage = find_ppage_or_above(vcpu->kvm, fault_ipa + PAGE_SIZE);
ppage = find_ppage_or_above(vcpu->kvm, fault_ipa + page_size);
}
size = size_sub(size, PAGE_SIZE);
fault_ipa += PAGE_SIZE;
fault_ipa += page_size;
}
end:
read_unlock(&vcpu->kvm->mmu_lock);