KVM: arm64: Remove size-order align in the nVHE hyp private VA range

commit f922c13e778d ("KVM: arm64: Introduce
pkvm_alloc_private_va_range()") and commit 92abe0f81e13 ("KVM: arm64:
Introduce hyp_alloc_private_va_range()") added an alignment for the
start address of any allocation into the nVHE hypervisor private VA
range.

This alignment (order of the size of the allocation) intends to enable
efficient stack verification (if the PAGE_SHIFT bit is zero, the stack
pointer is on the guard page and a stack overflow occurred).

But this is only necessary for stack allocation and can waste a lot of
VA space. So instead make stack-specific functions, handling the guard
page requirements, while other users (e.g.  fixmap) will only get page
alignment.

Reviewed-by: Kalesh Singh <kaleshsingh@google.com>
Signed-off-by: Vincent Donnefort <vdonnefort@google.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20230811112037.1147863-1-vdonnefort@google.com

1 files changed, 66 insertions, 17 deletions

diff --git a/arch/arm64/kvm/hyp/nvhe/mm.c b/arch/arm64/kvm/hyp/nvhe/mm.c
index 318298eb3d6b..65a7a186d7b2 100644
--- a/arch/arm64/kvm/hyp/nvhe/mm.c
+++ b/arch/arm64/kvm/hyp/nvhe/mm.c
@@ -44,6 +44,27 @@ static int __pkvm_create_mappings(unsigned long start, unsigned long size,
 	return err;
 }
 
+static int __pkvm_alloc_private_va_range(unsigned long start, size_t size)
+{
+	unsigned long cur;
+
+	hyp_assert_lock_held(&pkvm_pgd_lock);
+
+	if (!start || start < __io_map_base)
+		return -EINVAL;
+
+	/* The allocated size is always a multiple of PAGE_SIZE */
+	cur = start + PAGE_ALIGN(size);
+
+	/* Are we overflowing on the vmemmap ? */
+	if (cur > __hyp_vmemmap)
+		return -ENOMEM;
+
+	__io_map_base = cur;
+
+	return 0;
+}
+
 /**
  * pkvm_alloc_private_va_range - Allocates a private VA range.
  * @size:	The size of the VA range to reserve.
@@ -56,27 +77,16 @@ static int __pkvm_create_mappings(unsigned long start, unsigned long size,
  */
 int pkvm_alloc_private_va_range(size_t size, unsigned long *haddr)
 {
-	unsigned long base, addr;
-	int ret = 0;
+	unsigned long addr;
+	int ret;
 
 	hyp_spin_lock(&pkvm_pgd_lock);
-
-	/* Align the allocation based on the order of its size */
-	addr = ALIGN(__io_map_base, PAGE_SIZE << get_order(size));
-
-	/* The allocated size is always a multiple of PAGE_SIZE */
-	base = addr + PAGE_ALIGN(size);
-
-	/* Are we overflowing on the vmemmap ? */
-	if (!addr || base > __hyp_vmemmap)
-		ret = -ENOMEM;
-	else {
-		__io_map_base = base;
-		*haddr = addr;
-	}
-
+	addr = __io_map_base;
+	ret = __pkvm_alloc_private_va_range(addr, size);
 	hyp_spin_unlock(&pkvm_pgd_lock);
 
+	*haddr = addr;
+
 	return ret;
 }
 
@@ -340,6 +350,45 @@ int hyp_create_idmap(u32 hyp_va_bits)
 	return __pkvm_create_mappings(start, end - start, start, PAGE_HYP_EXEC);
 }
 
+int pkvm_create_stack(phys_addr_t phys, unsigned long *haddr)
+{
+	unsigned long addr, prev_base;
+	size_t size;
+	int ret;
+
+	hyp_spin_lock(&pkvm_pgd_lock);
+
+	prev_base = __io_map_base;
+	/*
+	 * Efficient stack verification using the PAGE_SHIFT bit implies
+	 * an alignment of our allocation on the order of the size.
+	 */
+	size = PAGE_SIZE * 2;
+	addr = ALIGN(__io_map_base, size);
+
+	ret = __pkvm_alloc_private_va_range(addr, size);
+	if (!ret) {
+		/*
+		 * Since the stack grows downwards, map the stack to the page
+		 * at the higher address and leave the lower guard page
+		 * unbacked.
+		 *
+		 * Any valid stack address now has the PAGE_SHIFT bit as 1
+		 * and addresses corresponding to the guard page have the
+		 * PAGE_SHIFT bit as 0 - this is used for overflow detection.
+		 */
+		ret = kvm_pgtable_hyp_map(&pkvm_pgtable, addr + PAGE_SIZE,
+					  PAGE_SIZE, phys, PAGE_HYP);
+		if (ret)
+			__io_map_base = prev_base;
+	}
+	hyp_spin_unlock(&pkvm_pgd_lock);
+
+	*haddr = addr + size;
+
+	return ret;
+}
+
 static void *admit_host_page(void *arg)
 {
 	struct kvm_hyp_memcache *host_mc = arg;