[RFC PATCH v1 10/28] riscv/mm : Introducing new protection flag "PROT_SHADOWSTACK"

From: debug
Date: Thu Jan 25 2024 - 01:32:05 EST


From: Deepak Gupta <debug@xxxxxxxxxxxx>

x86 and arm64 are using VM_SHADOW_STACK (which actually is VM_HIGH_ARCH_5)
vma flag and thus restrict it to 64bit implementation only. RISC-V is choosing
to encode presence of only VM_WRITE in vma flags as shadow stack vma. This allows
32bit RISC-V ecosystem leverage shadow stack as well.
This means that existing users of `do_mmap` who had been using `VM_WRITE` and
expecting read and write permissions will break.
Thus introducing `PROT_SHADOWSTACK` to allow `do_mmap` disambiguate between
read write v/s shadow stack mappings. Thus any kernel driver/module using `do_mmap`
and only passing `VM_WRITE` would still get read-write mappings. Although any user
of `do_mmap` intending to map a shaodw stack should pass `PROT_SHADOWSTACK` to get
a shadow stack mapping.

Although for userspace still want to rely on `map_shadow_stack` and not expose
`PROT_SHADOWSTACK` to userspace and that's why this prot flag is not exposed in uapi
headers.

Signed-off-by: Deepak Gupta <debug@xxxxxxxxxxxx>
---
arch/riscv/include/asm/mman.h | 25 +++++++++++++++++++++++++
mm/mmap.c | 1 +
2 files changed, 26 insertions(+)
create mode 100644 arch/riscv/include/asm/mman.h

diff --git a/arch/riscv/include/asm/mman.h b/arch/riscv/include/asm/mman.h
new file mode 100644
index 000000000000..4902d837e93c
--- /dev/null
+++ b/arch/riscv/include/asm/mman.h
@@ -0,0 +1,25 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef __ASM_MMAN_H__
+#define __ASM_MMAN_H__
+
+#include <linux/compiler.h>
+#include <linux/types.h>
+#include <uapi/asm/mman.h>
+
+/*
+ * Major architectures (x86, aarch64, riscv) have shadow stack now. x86 and
+ * arm64 choose to use VM_SHADOW_STACK (which actually is VM_HIGH_ARCH_5) vma
+ * flag, however that restrict it to 64bit implementation only. risc-v shadow
+ * stack encodings in page tables is PTE.R=0, PTE.W=1, PTE.D=1 which used to be
+ * reserved until now. risc-v is choosing to encode presence of only VM_WRITE in
+ * vma flags as shadow stack vma. However this means that existing users of mmap
+ * (and do_mmap) who were relying on passing only PROT_WRITE (or VM_WRITE from
+ * kernel driver) but still getting read and write mappings, should still work.
+ * x86 and arm64 followed the direction of a new system call `map_shadow_stack`.
+ * risc-v would like to converge on that so that shadow stacks flows are as much
+ * arch agnostic. Thus a conscious decision to define PROT_XXX definition for
+ * shadow stack here (and not exposed to uapi)
+ */
+#define PROT_SHADOWSTACK 0x40
+
+#endif /* ! __ASM_MMAN_H__ */
diff --git a/mm/mmap.c b/mm/mmap.c
index 1971bfffcc03..fab2acf21ce9 100644
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -47,6 +47,7 @@
#include <linux/oom.h>
#include <linux/sched/mm.h>
#include <linux/ksm.h>
+#include <linux/processor.h>

#include <linux/uaccess.h>
#include <asm/cacheflush.h>
--
2.43.0