[PATCH v2] KVM: arm64: Don't WARN on SMCCC filter reserved range allocation failure

[Fuad Tabba]

The first time userspace sets an SMCCC filter, kvm_smccc_set_filter()
inserts the reserved architecture ranges into the maple tree and WARNs
if that fails. The insertion allocates with GFP_KERNEL_ACCOUNT, so a
tight memory cgroup limit is enough for userspace to hit the WARN, even
though the -ENOMEM is also returned to it. The WARN dates back to when
the ranges were inserted at VM creation.

Leave -ENOMEM out of the WARN and keep it for every other error. The
two reserved ranges are fixed and disjoint and go into an empty tree
under config_lock, so any other error is a KVM bug.

Fixes: fb88707dd39bd ("KVM: arm64: Use a maple tree to represent the SMCCC filter")
Reported-by: Sashiko <sashiko-bot@xxxxxxxxxx>
Closes: https://lore.kernel.org/all/20260829055809.29AD31F000E9@xxxxxxxxxxxxxxx/
Suggested-by: Oliver Upton <oupton@xxxxxxxxxx>
Signed-off-by: Fuad Tabba <fuad.tabba@xxxxxxxxx>
---
Changes since v1 [1]:
  - Keep the WARN for any error other than -ENOMEM. (Oliver)
  - Split out of v1 as its own patch. (Marc)

Based on Linux 7.3-rc3 (fd73f4a665989).

[1] https://lore.kernel.org/all/20260928152507.2116110-2-fuad.tabba@xxxxxxxxx/

 arch/arm64/kvm/hypercalls.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/arch/arm64/kvm/hypercalls.c b/arch/arm64/kvm/hypercalls.c
index b11b8821c9fbc..c9d488c586e5f 100644
--- a/arch/arm64/kvm/hypercalls.c
+++ b/arch/arm64/kvm/hypercalls.c
@@ -197,8 +197,10 @@ static int kvm_smccc_set_filter(struct kvm *kvm, struct kvm_smccc_filter __user
 
 	if (!kvm_smccc_filter_configured(kvm)) {
 		r = kvm_smccc_filter_insert_reserved(kvm);
-		if (WARN_ON_ONCE(r))
+		if (r) {
+			WARN_ON_ONCE(r != -ENOMEM);
 			goto out_unlock;
+		}
 	}
 
 	r = mtree_insert_range(&kvm->arch.smccc_filter, start, end,

base-commit: fd73f4a6659897191fa0d40695fe370925dd3780
-- 
2.39.5