Received-SPF: pass (google.com: domain of linux-kernel+bounces-147302-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:4601:e00::3 as permitted sender) client-ip=2604:1380:4601:e00::3;
Message-Id: <20240416122254.868007168-2-mbland@motorola.com>
In-Reply-To: <20240416122254.868007168-1-mbland@motorola.com>
References: <20240416122254.868007168-1-mbland@motorola.com>
To: linux-mm@kvack.org
Cc: Maxwell Bland <mbland@motorola.com>, linux-kernel@vger.kernel.org,
        Andrew Morton <akpm@linux-foundation.org>,
        Uladzislau Rezki <urezki@gmail.com>,
        Christoph Hellwig <hch@infradead.org>,
        Lorenzo Stoakes <lstoakes@gmail.com>
From: Maxwell Bland <mbland@motorola.com>
Date: Tue, 2 Apr 2024 15:15:01 -0500
Subject: [PATCH 1/5] mm: allow arch refinement/skip for vmap alloc
Precedence: bulk

Makes red black tree allocation more flexible on a per-architecture
basis by introducing an optional hooks to refine the red-black tree
structuring and exposing vmalloc functions for clipping vmap areas,
finding vmap areas, and inserting vmap areas.

With this patch, the red-black vmap tree can be refined to account for
architecture-specific memory management operations, most notably address
space layout randomization, as these features conflict with generic
management of a single vmalloc_start to vmalloc_end range as given by
mm/vmalloc.c.

For example, x86 is forced to restrict aslr to 1024 possible locations,
which is a very, very small number, and arm64 breaks standard code/data
partitioning altogether, which prevents the enforcement of performant
immmutability on kernel page tables.

Signed-off-by: Maxwell Bland <mbland@motorola.com>
---
 include/linux/vmalloc.h | 24 ++++++++++++++++++++++++
 mm/vmalloc.c            | 16 ++++++++++------
 2 files changed, 34 insertions(+), 6 deletions(-)

diff --git a/include/linux/vmalloc.h b/include/linux/vmalloc.h
index 98ea90e90439..3c5ce7ee0bea 100644
--- a/include/linux/vmalloc.h
+++ b/include/linux/vmalloc.h
@@ -12,6 +12,7 @@
 
 #include <asm/vmalloc.h>
 
+struct kmem_cache;
 struct vm_area_struct;		/* vma defining user mapping in mm_types.h */
 struct notifier_block;		/* in notifier.h */
 struct iov_iter;		/* in uio.h */
@@ -125,6 +126,21 @@ static inline pgprot_t arch_vmap_pgprot_tagged(pgprot_t prot)
 }
 #endif
 
+#ifndef arch_skip_va
+static inline bool arch_skip_va(struct vmap_area *va, unsigned long vstart)
+{
+	return false;
+}
+#endif
+
+#ifndef arch_refine_vmap_space
+static inline void arch_refine_vmap_space(struct rb_root *root,
+					  struct list_head *head,
+					  struct kmem_cache *cachep)
+{
+}
+#endif
+
 /*
  *	Highlevel APIs for driver use
  */
@@ -214,6 +230,14 @@ extern struct vm_struct *__get_vm_area_caller(unsigned long size,
 void free_vm_area(struct vm_struct *area);
 extern struct vm_struct *remove_vm_area(const void *addr);
 extern struct vm_struct *find_vm_area(const void *addr);
+extern void insert_vmap_area_augment(struct vmap_area *va, struct rb_node *from,
+				     struct rb_root *root,
+				     struct list_head *head);
+extern int va_clip(struct rb_root *root, struct list_head *head,
+		   struct vmap_area *va, unsigned long nva_start_addr,
+		   unsigned long size);
+extern struct vmap_area *__find_vmap_area(unsigned long addr,
+					  struct rb_root *root);
 struct vmap_area *find_vmap_area(unsigned long addr);
 
 static inline bool is_vm_area_hugepages(const void *addr)
diff --git a/mm/vmalloc.c b/mm/vmalloc.c
index 68fa001648cc..de4577a3708e 100644
--- a/mm/vmalloc.c
+++ b/mm/vmalloc.c
@@ -989,7 +989,7 @@ unsigned long vmalloc_nr_pages(void)
 	return atomic_long_read(&nr_vmalloc_pages);
 }
 
-static struct vmap_area *__find_vmap_area(unsigned long addr, struct rb_root *root)
+struct vmap_area *__find_vmap_area(unsigned long addr, struct rb_root *root)
 {
 	struct rb_node *n = root->rb_node;
 
@@ -1322,7 +1322,7 @@ insert_vmap_area(struct vmap_area *va,
 		link_va(va, root, parent, link, head);
 }
 
-static void
+void
 insert_vmap_area_augment(struct vmap_area *va,
 	struct rb_node *from, struct rb_root *root,
 	struct list_head *head)
@@ -1501,7 +1501,7 @@ find_vmap_lowest_match(struct rb_root *root, unsigned long size,
 				vstart < va->va_start) {
 			node = node->rb_left;
 		} else {
-			if (is_within_this_va(va, size, align, vstart))
+			if (!arch_skip_va(va, vstart) && is_within_this_va(va, size, align, vstart))
 				return va;
 
 			/*
@@ -1522,7 +1522,8 @@ find_vmap_lowest_match(struct rb_root *root, unsigned long size,
 			 */
 			while ((node = rb_parent(node))) {
 				va = rb_entry(node, struct vmap_area, rb_node);
-				if (is_within_this_va(va, size, align, vstart))
+				if (!arch_skip_va(va, vstart) &&
+				    is_within_this_va(va, size, align, vstart))
 					return va;
 
 				if (get_subtree_max_size(node->rb_right) >= length &&
@@ -1554,7 +1555,7 @@ find_vmap_lowest_linear_match(struct list_head *head, unsigned long size,
 	struct vmap_area *va;
 
 	list_for_each_entry(va, head, list) {
-		if (!is_within_this_va(va, size, align, vstart))
+		if (arch_skip_va(va, vstart) || !is_within_this_va(va, size, align, vstart))
 			continue;
 
 		return va;
@@ -1617,7 +1618,7 @@ classify_va_fit_type(struct vmap_area *va,
 	return type;
 }
 
-static __always_inline int
+__always_inline int
 va_clip(struct rb_root *root, struct list_head *head,
 		struct vmap_area *va, unsigned long nva_start_addr,
 		unsigned long size)
@@ -5129,4 +5130,7 @@ void __init vmalloc_init(void)
 	vmap_node_shrinker->count_objects = vmap_node_shrink_count;
 	vmap_node_shrinker->scan_objects = vmap_node_shrink_scan;
 	shrinker_register(vmap_node_shrinker);
+
+	arch_refine_vmap_space(&free_vmap_area_root, &free_vmap_area_list,
+			       vmap_area_cachep);
 }
-- 
2.39.2