From: Jack Thomson <jackabt@amazon.com>

Adding __gmem_abort and __user_mem_abort that preserve -EAGAIN results.
These will be used by the pre-fault implementation which needs to retry
on -EAGAIN.

Also add an optional page_size output parameter to __user_mem_abort to
return the VMA page size, which will be needed for pre-faulting.

No functional changes are intended

Signed-off-by: Jack Thomson <jackabt@amazon.com>
---
 arch/arm64/kvm/mmu.c | 36 +++++++++++++++++++++++++++++-------
 1 file changed, 29 insertions(+), 7 deletions(-)

diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
index a36426ccd9b5..082e7d8ae655 100644
--- a/arch/arm64/kvm/mmu.c
+++ b/arch/arm64/kvm/mmu.c
@@ -1521,9 +1521,9 @@ static void adjust_nested_fault_perms(struct kvm_s2_trans *nested,
 
 #define KVM_PGTABLE_WALK_MEMABORT_FLAGS (KVM_PGTABLE_WALK_HANDLE_FAULT | KVM_PGTABLE_WALK_SHARED)
 
-static int gmem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
-		      struct kvm_s2_trans *nested,
-		      struct kvm_memory_slot *memslot, bool is_perm)
+static int __gmem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
+			struct kvm_s2_trans *nested,
+			struct kvm_memory_slot *memslot, bool is_perm)
 {
 	bool write_fault, exec_fault, writable;
 	enum kvm_pgtable_walk_flags flags = KVM_PGTABLE_WALK_MEMABORT_FLAGS;
@@ -1592,13 +1592,22 @@ static int gmem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
 	if (writable && !ret)
 		mark_page_dirty_in_slot(kvm, memslot, gfn);
 
+	return ret;
+}
+
+static int gmem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
+		      struct kvm_s2_trans *nested,
+		      struct kvm_memory_slot *memslot, bool is_perm)
+{
+	int ret = __gmem_abort(vcpu, fault_ipa, nested, memslot, is_perm);
 	return ret != -EAGAIN ? ret : 0;
 }
 
-static int user_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
-			  struct kvm_s2_trans *nested,
-			  struct kvm_memory_slot *memslot, unsigned long hva,
-			  bool fault_is_perm)
+static int __user_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
+			    struct kvm_s2_trans *nested,
+			    struct kvm_memory_slot *memslot,
+			    long *page_size, unsigned long hva,
+			    bool fault_is_perm)
 {
 	int ret = 0;
 	bool topup_memcache;
@@ -1871,10 +1880,23 @@ static int user_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
 	kvm_release_faultin_page(kvm, page, !!ret, writable);
 	kvm_fault_unlock(kvm);
 
+	if (page_size)
+		*page_size = vma_pagesize;
+
 	/* Mark the page dirty only if the fault is handled successfully */
 	if (writable && !ret)
 		mark_page_dirty_in_slot(kvm, memslot, gfn);
 
+	return ret;
+}
+
+static int user_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
+			  struct kvm_s2_trans *nested,
+			  struct kvm_memory_slot *memslot, unsigned long hva,
+			  bool fault_is_perm)
+{
+	int ret = __user_mem_abort(vcpu, fault_ipa, nested, memslot, NULL,
+			    hva, fault_is_perm);
 	return ret != -EAGAIN ? ret : 0;
 }
 
-- 
2.43.0

Hi Jack,

On Thu, Sep 11, 2025 at 02:46:43PM +0100, Jack Thomson wrote:
> From: Jack Thomson <jackabt@amazon.com>
> 
> Adding __gmem_abort and __user_mem_abort that preserve -EAGAIN results.
> These will be used by the pre-fault implementation which needs to retry
> on -EAGAIN.

-EAGAIN is a pretty clear signal that another vCPU has faulted on this
memory and is in the middle of installing a mapping. Why bother with
retrying?

If we conceptually treat this thing as a synthetic stage-2 abort then it
should use the same EAGAIN handling as a literal stage-2 abort.

Thanks,
Oliver