On Fri, Apr 22, 2022 at 10:08 AM Linus Torvalds
<torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:
>
> Just opt-in with the mappings that matter.
Actually, we could automatically opt-in a few common cases that we
know are fundamentally ok, because they already can't play protection
games.
In particular, kvmalloc().
So I think something like this patch - along with Song's patch to
enable it for alloc_large_system_hash() - would be fairly safe, and
avoid any nasty cases.
And probably catch quite a lot of the cases that matter that can grow large.
Linus
mm/util.c | 11 +++++++++--
1 file changed, 9 insertions(+), 2 deletions(-)
diff --git a/mm/util.c b/mm/util.c
index 54e5e761a9a9..3492a9e81aa3 100644
--- a/mm/util.c
+++ b/mm/util.c
@@ -592,8 +592,15 @@ void *kvmalloc_node(size_t size, gfp_t flags, int node)
return NULL;
}
- return __vmalloc_node(size, 1, flags, node,
- __builtin_return_address(0));
+ /*
+ * kvmalloc() can always use VM_ALLOW_HUGE_VMAP,
+ * since the callers already cannot assume anything
+ * about the resulting pointer, and cannot play
+ * protection games.
+ */
+ return __vmalloc_node_range(size, 1, VMALLOC_START, VMALLOC_END,
+ flags, PAGE_KERNEL, VM_ALLOW_HUGE_VMAP,
+ node, __builtin_return_address(0));
}
EXPORT_SYMBOL(kvmalloc_node);
