From: Sean Christopherson <seanjc@google.com>
Date: Tue, 10 Jun 2025 22:57:07 +0000 (-0700)
Subject: KVM: SVM: Allocate IOPM pages after initial setup in svm_hardware_setup()
X-Git-Url: http://git.ipfire.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=fb96d5cf0fdaacd2165797982502d5b157f8bcf5;p=thirdparty%2Flinux.git

KVM: SVM: Allocate IOPM pages after initial setup in svm_hardware_setup()

Allocate pages for the IOPM after initial setup has been completed in
svm_hardware_setup(), so that sanity checks can be added in the setup flow
without needing to free the IOPM pages.  The IOPM is only referenced (via
iopm_base) in init_vmcb() and svm_hardware_unsetup(), so there's no need
to allocate it early on.

No functional change intended (beyond the obvious ordering differences,
e.g. if the allocation fails).

Link: https://lore.kernel.org/r/20250610225737.156318-3-seanjc@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
---

diff --git a/arch/x86/kvm/svm/svm.c b/arch/x86/kvm/svm/svm.c
index 9c6ec5e881411..6ce35f01d410e 100644
--- a/arch/x86/kvm/svm/svm.c
+++ b/arch/x86/kvm/svm/svm.c
@@ -5494,15 +5494,6 @@ static __init int svm_hardware_setup(void)
 	}
 	kvm_enable_efer_bits(EFER_NX);
 
-	iopm_pages = alloc_pages(GFP_KERNEL, order);
-
-	if (!iopm_pages)
-		return -ENOMEM;
-
-	iopm_va = page_address(iopm_pages);
-	memset(iopm_va, 0xff, PAGE_SIZE * (1 << order));
-	iopm_base = __sme_page_pa(iopm_pages);
-
 	init_msrpm_offsets();
 
 	kvm_caps.supported_xcr0 &= ~(XFEATURE_MASK_BNDREGS |
@@ -5569,6 +5560,15 @@ static __init int svm_hardware_setup(void)
 		else
 			pr_info("LBR virtualization supported\n");
 	}
+
+	iopm_pages = alloc_pages(GFP_KERNEL, order);
+	if (!iopm_pages)
+		return -ENOMEM;
+
+	iopm_va = page_address(iopm_pages);
+	memset(iopm_va, 0xff, PAGE_SIZE * (1 << order));
+	iopm_base = __sme_page_pa(iopm_pages);
+
 	/*
 	 * Note, SEV setup consumes npt_enabled and enable_mmio_caching (which
 	 * may be modified by svm_adjust_mmio_mask()), as well as nrips.