[PATCH v3 00/17] Add support for Clang CFI

[PATCH v3 00/17] Add support for Clang CFI

[Sami Tolvanen]

This series adds support for Clang's Control-Flow Integrity (CFI)
checking. With CFI, the compiler injects a runtime check before each
indirect function call to ensure the target is a valid function with
the correct static type. This restricts possible call targets and
makes it more difficult for an attacker to exploit bugs that allow the
modification of stored function pointers. For more details, see:

  https://clang.llvm.org/docs/ControlFlowIntegrity.html

The first patch contains build system changes and error handling,
and implements support for cross-module indirect call checking. The
remaining patches address issues caused by the compiler
instrumentation. These include fixing known type mismatches, as well
as issues with address space confusion and cross-module function
address equality.

These patches add support only for arm64, but I'll post patches also
for x86_64 after we address the remaining issues there, including
objtool support.

You can also pull this series from

  https://github.com/samitolvanen/linux.git cfi-v3

---
Changes in v3:
 - Added a patch to change list_sort() callers treewide to use
   const pointers instead of simply removing the internal casts.
 - Changed cleanup_symbol_name() to return bool.
 - Changed module.lds.S to drop the .eh_frame section only with
   CONFIG_CFI_CLANG.
 - Switched to synchronize_rcu() in update_shadow().

Changes in v2:
 - Fixed .text merging in module.lds.S.
 - Added WARN_ON_FUNCTION_MISMATCH() and changed kernel/thread.c
   and kernel/workqueue.c to use the macro instead.


Sami Tolvanen (17):
  add support for Clang CFI
  cfi: add __cficanonical
  mm: add generic __va_function and __pa_function macros
  module: ensure __cfi_check alignment
  workqueue: use WARN_ON_FUNCTION_MISMATCH
  kthread: use WARN_ON_FUNCTION_MISMATCH
  kallsyms: strip ThinLTO hashes from static functions
  bpf: disable CFI in dispatcher functions
  treewide: Change list_sort to use const pointers
  lkdtm: use __va_function
  psci: use __pa_function for cpu_resume
  arm64: implement __va_function
  arm64: use __pa_function
  arm64: add __nocfi to functions that jump to a physical address
  arm64: add __nocfi to __apply_alternatives
  KVM: arm64: Disable CFI for nVHE
  arm64: allow CONFIG_CFI_CLANG to be selected

 Makefile                                      |  17 +
 arch/Kconfig                                  |  45 +++
 arch/arm64/Kconfig                            |   1 +
 arch/arm64/include/asm/memory.h               |  15 +
 arch/arm64/include/asm/mmu_context.h          |   4 +-
 arch/arm64/kernel/acpi_parking_protocol.c     |   2 +-
 arch/arm64/kernel/alternative.c               |   4 +-
 arch/arm64/kernel/cpu-reset.h                 |  10 +-
 arch/arm64/kernel/cpufeature.c                |   4 +-
 arch/arm64/kernel/psci.c                      |   3 +-
 arch/arm64/kernel/smp_spin_table.c            |   2 +-
 arch/arm64/kvm/hyp/nvhe/Makefile              |   6 +-
 arch/arm64/kvm/vgic/vgic-its.c                |   8 +-
 arch/arm64/kvm/vgic/vgic.c                    |   3 +-
 block/blk-mq-sched.c                          |   3 +-
 block/blk-mq.c                                |   3 +-
 drivers/acpi/nfit/core.c                      |   3 +-
 drivers/acpi/numa/hmat.c                      |   3 +-
 drivers/clk/keystone/sci-clk.c                |   4 +-
 drivers/firmware/psci/psci.c                  |   4 +-
 drivers/gpu/drm/drm_modes.c                   |   3 +-
 drivers/gpu/drm/i915/gt/intel_engine_user.c   |   3 +-
 drivers/gpu/drm/i915/gvt/debugfs.c            |   2 +-
 drivers/gpu/drm/i915/selftests/i915_gem_gtt.c |   3 +-
 drivers/gpu/drm/radeon/radeon_cs.c            |   4 +-
 .../hw/usnic/usnic_uiom_interval_tree.c       |   3 +-
 drivers/interconnect/qcom/bcm-voter.c         |   2 +-
 drivers/md/raid5.c                            |   3 +-
 drivers/misc/lkdtm/usercopy.c                 |   2 +-
 drivers/misc/sram.c                           |   4 +-
 drivers/nvme/host/core.c                      |   3 +-
 .../controller/cadence/pcie-cadence-host.c    |   3 +-
 drivers/spi/spi-loopback-test.c               |   3 +-
 fs/btrfs/raid56.c                             |   3 +-
 fs/btrfs/tree-log.c                           |   3 +-
 fs/btrfs/volumes.c                            |   3 +-
 fs/ext4/fsmap.c                               |   4 +-
 fs/gfs2/glock.c                               |   3 +-
 fs/gfs2/log.c                                 |   2 +-
 fs/gfs2/lops.c                                |   3 +-
 fs/iomap/buffered-io.c                        |   3 +-
 fs/ubifs/gc.c                                 |   7 +-
 fs/ubifs/replay.c                             |   4 +-
 fs/xfs/scrub/bitmap.c                         |   4 +-
 fs/xfs/xfs_bmap_item.c                        |   4 +-
 fs/xfs/xfs_buf.c                              |   6 +-
 fs/xfs/xfs_extent_busy.c                      |   4 +-
 fs/xfs/xfs_extent_busy.h                      |   3 +-
 fs/xfs/xfs_extfree_item.c                     |   4 +-
 fs/xfs/xfs_refcount_item.c                    |   4 +-
 fs/xfs/xfs_rmap_item.c                        |   4 +-
 include/asm-generic/bug.h                     |  16 +
 include/asm-generic/vmlinux.lds.h             |  20 +-
 include/linux/bpf.h                           |   4 +-
 include/linux/cfi.h                           |  41 +++
 include/linux/compiler-clang.h                |   3 +
 include/linux/compiler_types.h                |   8 +
 include/linux/init.h                          |   6 +-
 include/linux/list_sort.h                     |   7 +-
 include/linux/mm.h                            |   8 +
 include/linux/module.h                        |  13 +-
 include/linux/pci.h                           |   4 +-
 init/Kconfig                                  |   2 +-
 kernel/Makefile                               |   4 +
 kernel/cfi.c                                  | 329 ++++++++++++++++++
 kernel/kallsyms.c                             |  55 ++-
 kernel/kthread.c                              |   3 +-
 kernel/module.c                               |  43 +++
 kernel/workqueue.c                            |   2 +-
 lib/list_sort.c                               |  17 +-
 lib/test_list_sort.c                          |   3 +-
 net/tipc/name_table.c                         |   4 +-
 scripts/Makefile.modfinal                     |   2 +-
 scripts/module.lds.S                          |  20 +-
 74 files changed, 752 insertions(+), 112 deletions(-)
 create mode 100644 include/linux/cfi.h
 create mode 100644 kernel/cfi.c


base-commit: 7acac4b3196caee5e21fb5ea53f8bc124e6a16fc
-- 
2.31.0.291.g576ba9dcdaf-goog

[PATCH v3 01/17] add support for Clang CFI

[Sami Tolvanen]

This change adds support for Clang’s forward-edge Control Flow
Integrity (CFI) checking. With CONFIG_CFI_CLANG, the compiler
injects a runtime check before each indirect function call to ensure
the target is a valid function with the correct static type. This
restricts possible call targets and makes it more difficult for
an attacker to exploit bugs that allow the modification of stored
function pointers. For more details, see:

  https://clang.llvm.org/docs/ControlFlowIntegrity.html

Clang requires CONFIG_LTO_CLANG to be enabled with CFI to gain
visibility to possible call targets. Kernel modules are supported
with Clang’s cross-DSO CFI mode, which allows checking between
independently compiled components.

With CFI enabled, the compiler injects a __cfi_check() function into
the kernel and each module for validating local call targets. For
cross-module calls that cannot be validated locally, the compiler
calls the global __cfi_slowpath_diag() function, which determines
the target module and calls the correct __cfi_check() function. This
patch includes a slowpath implementation that uses __module_address()
to resolve call targets, and with CONFIG_CFI_CLANG_SHADOW enabled, a
shadow map that speeds up module look-ups by ~3x.

Clang implements indirect call checking using jump tables and
offers two methods of generating them. With canonical jump tables,
the compiler renames each address-taken function to <function>.cfi
and points the original symbol to a jump table entry, which passes
__cfi_check() validation. This isn’t compatible with stand-alone
assembly code, which the compiler doesn’t instrument, and would
result in indirect calls to assembly code to fail. Therefore, we
default to using non-canonical jump tables instead, where the compiler
generates a local jump table entry <function>.cfi_jt for each
address-taken function, and replaces all references to the function
with the address of the jump table entry.

Note that because non-canonical jump table addresses are local
to each component, they break cross-module function address
equality. Specifically, the address of a global function will be
different in each module, as it's replaced with the address of a local
jump table entry. If this address is passed to a different module,
it won’t match the address of the same function taken there. This
may break code that relies on comparing addresses passed from other
components.

CFI checking can be disabled in a function with the __nocfi attribute.
Additionally, CFI can be disabled for an entire compilation unit by
filtering out CC_FLAGS_CFI.

By default, CFI failures result in a kernel panic to stop a potential
exploit. CONFIG_CFI_PERMISSIVE enables a permissive mode, where the
kernel prints out a rate-limited warning instead, and allows execution
to continue. This option is helpful for locating type mismatches, but
should only be enabled during development.

Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
---
 Makefile                          |  17 ++
 arch/Kconfig                      |  45 ++++
 include/asm-generic/bug.h         |  16 ++
 include/asm-generic/vmlinux.lds.h |  20 +-
 include/linux/cfi.h               |  41 ++++
 include/linux/compiler-clang.h    |   2 +
 include/linux/compiler_types.h    |   4 +
 include/linux/init.h              |   2 +-
 include/linux/module.h            |  13 +-
 init/Kconfig                      |   2 +-
 kernel/Makefile                   |   4 +
 kernel/cfi.c                      | 329 ++++++++++++++++++++++++++++++
 kernel/module.c                   |  43 ++++
 scripts/Makefile.modfinal         |   2 +-
 14 files changed, 534 insertions(+), 6 deletions(-)
 create mode 100644 include/linux/cfi.h
 create mode 100644 kernel/cfi.c

diff --git a/Makefile b/Makefile
index d4784d181123..2b90a5cc08c8 100644
--- a/Makefile
+++ b/Makefile
@@ -920,6 +920,23 @@ KBUILD_AFLAGS	+= -fno-lto
 export CC_FLAGS_LTO
 endif
 
+ifdef CONFIG_CFI_CLANG
+CC_FLAGS_CFI	:= -fsanitize=cfi \
+		   -fsanitize-cfi-cross-dso \
+		   -fno-sanitize-cfi-canonical-jump-tables \
+		   -fno-sanitize-trap=cfi \
+		   -fno-sanitize-blacklist
+
+ifdef CONFIG_CFI_PERMISSIVE
+CC_FLAGS_CFI	+= -fsanitize-recover=cfi
+endif
+
+# If LTO flags are filtered out, we must also filter out CFI.
+CC_FLAGS_LTO	+= $(CC_FLAGS_CFI)
+KBUILD_CFLAGS	+= $(CC_FLAGS_CFI)
+export CC_FLAGS_CFI
+endif
+
 ifdef CONFIG_DEBUG_FORCE_FUNCTION_ALIGN_32B
 KBUILD_CFLAGS += -falign-functions=32
 endif
diff --git a/arch/Kconfig b/arch/Kconfig
index ecfd3520b676..f6a85ba6cba2 100644
--- a/arch/Kconfig
+++ b/arch/Kconfig
@@ -692,6 +692,51 @@ config LTO_CLANG_THIN
 	  If unsure, say Y.
 endchoice
 
+config ARCH_SUPPORTS_CFI_CLANG
+	bool
+	help
+	  An architecture should select this option if it can support Clang's
+	  Control-Flow Integrity (CFI) checking.
+
+config CFI_CLANG
+	bool "Use Clang's Control Flow Integrity (CFI)"
+	depends on LTO_CLANG && ARCH_SUPPORTS_CFI_CLANG
+	# Clang >= 12:
+	# - https://bugs.llvm.org/show_bug.cgi?id=46258
+	# - https://bugs.llvm.org/show_bug.cgi?id=47479
+	depends on CLANG_VERSION >= 120000
+	select KALLSYMS
+	help
+	  This option enables Clang’s forward-edge Control Flow Integrity
+	  (CFI) checking, where the compiler injects a runtime check to each
+	  indirect function call to ensure the target is a valid function with
+	  the correct static type. This restricts possible call targets and
+	  makes it more difficult for an attacker to exploit bugs that allow
+	  the modification of stored function pointers. More information can be
+	  found from Clang's documentation:
+
+	    https://clang.llvm.org/docs/ControlFlowIntegrity.html
+
+config CFI_CLANG_SHADOW
+	bool "Use CFI shadow to speed up cross-module checks"
+	default y
+	depends on CFI_CLANG && MODULES
+	help
+	  If you select this option, the kernel builds a fast look-up table of
+	  CFI check functions in loaded modules to reduce performance overhead.
+
+	  If unsure, say Y.
+
+config CFI_PERMISSIVE
+	bool "Use CFI in permissive mode"
+	depends on CFI_CLANG
+	help
+	  When selected, Control Flow Integrity (CFI) violations result in a
+	  warning instead of a kernel panic. This option should only be used
+	  for finding indirect call type mismatches during development.
+
+	  If unsure, say N.
+
 config HAVE_ARCH_WITHIN_STACK_FRAMES
 	bool
 	help
diff --git a/include/asm-generic/bug.h b/include/asm-generic/bug.h
index 76a10e0dca9f..b402494883b6 100644
--- a/include/asm-generic/bug.h
+++ b/include/asm-generic/bug.h
@@ -241,6 +241,22 @@ void __warn(const char *file, int line, void *caller, unsigned taint,
 # define WARN_ON_SMP(x)			({0;})
 #endif
 
+/*
+ * WARN_ON_FUNCTION_MISMATCH() warns if a value doesn't match a
+ * function address, and can be useful for catching issues with
+ * callback functions, for example.
+ *
+ * With CONFIG_CFI_CLANG, the warning is disabled because the
+ * compiler replaces function addresses taken in C code with
+ * local jump table addresses, which breaks cross-module function
+ * address equality.
+ */
+#if defined(CONFIG_CFI_CLANG) && defined(CONFIG_MODULES)
+# define WARN_ON_FUNCTION_MISMATCH(x, fn) ({ 0; })
+#else
+# define WARN_ON_FUNCTION_MISMATCH(x, fn) WARN_ON_ONCE((x) != (fn))
+#endif
+
 #endif /* __ASSEMBLY__ */
 
 #endif
diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h
index 0331d5d49551..40a9c101565e 100644
--- a/include/asm-generic/vmlinux.lds.h
+++ b/include/asm-generic/vmlinux.lds.h
@@ -544,6 +544,22 @@
 	. = ALIGN((align));						\
 	__end_rodata = .;
 
+
+/*
+ * .text..L.cfi.jumptable.* contain Control-Flow Integrity (CFI)
+ * jump table entries.
+ */
+#ifdef CONFIG_CFI_CLANG
+#define TEXT_CFI_JT							\
+		. = ALIGN(PMD_SIZE);					\
+		__cfi_jt_start = .;					\
+		*(.text..L.cfi.jumptable .text..L.cfi.jumptable.*)	\
+		. = ALIGN(PMD_SIZE);					\
+		__cfi_jt_end = .;
+#else
+#define TEXT_CFI_JT
+#endif
+
 /*
  * Non-instrumentable text section
  */
@@ -570,6 +586,7 @@
 		NOINSTR_TEXT						\
 		*(.text..refcount)					\
 		*(.ref.text)						\
+		TEXT_CFI_JT						\
 	MEM_KEEP(init.text*)						\
 	MEM_KEEP(exit.text*)						\
 
@@ -974,7 +991,8 @@
  * keep any .init_array.* sections.
  * https://bugs.llvm.org/show_bug.cgi?id=46478
  */
-#if defined(CONFIG_GCOV_KERNEL) || defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KCSAN)
+#if defined(CONFIG_GCOV_KERNEL) || defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KCSAN) || \
+	defined(CONFIG_CFI_CLANG)
 # ifdef CONFIG_CONSTRUCTORS
 #  define SANITIZER_DISCARDS						\
 	*(.eh_frame)
diff --git a/include/linux/cfi.h b/include/linux/cfi.h
new file mode 100644
index 000000000000..879744aaa6e0
--- /dev/null
+++ b/include/linux/cfi.h
@@ -0,0 +1,41 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Clang Control Flow Integrity (CFI) support.
+ *
+ * Copyright (C) 2021 Google LLC
+ */
+#ifndef _LINUX_CFI_H
+#define _LINUX_CFI_H
+
+#ifdef CONFIG_CFI_CLANG
+typedef void (*cfi_check_fn)(uint64_t id, void *ptr, void *diag);
+
+/* Compiler-generated function in each module, and the kernel */
+extern void __cfi_check(uint64_t id, void *ptr, void *diag);
+
+/*
+ * Force the compiler to generate a CFI jump table entry for a function
+ * and store the jump table address to __cfi_jt_<function>.
+ */
+#define __CFI_ADDRESSABLE(fn, __attr) \
+	const void *__cfi_jt_ ## fn __visible __attr = (void *)&fn
+
+#ifdef CONFIG_CFI_CLANG_SHADOW
+
+extern void cfi_module_add(struct module *mod, unsigned long base_addr);
+extern void cfi_module_remove(struct module *mod, unsigned long base_addr);
+
+#else
+
+static inline void cfi_module_add(struct module *mod, unsigned long base_addr) {}
+static inline void cfi_module_remove(struct module *mod, unsigned long base_addr) {}
+
+#endif /* CONFIG_CFI_CLANG_SHADOW */
+
+#else /* !CONFIG_CFI_CLANG */
+
+#define __CFI_ADDRESSABLE(fn, __attr)
+
+#endif /* CONFIG_CFI_CLANG */
+
+#endif /* _LINUX_CFI_H */
diff --git a/include/linux/compiler-clang.h b/include/linux/compiler-clang.h
index d217c382b02d..6de9d0c9377e 100644
--- a/include/linux/compiler-clang.h
+++ b/include/linux/compiler-clang.h
@@ -61,3 +61,5 @@
 #if __has_feature(shadow_call_stack)
 # define __noscs	__attribute__((__no_sanitize__("shadow-call-stack")))
 #endif
+
+#define __nocfi		__attribute__((__no_sanitize__("cfi")))
diff --git a/include/linux/compiler_types.h b/include/linux/compiler_types.h
index e5dd5a4ae946..796935a37e37 100644
--- a/include/linux/compiler_types.h
+++ b/include/linux/compiler_types.h
@@ -242,6 +242,10 @@ struct ftrace_likely_data {
 # define __noscs
 #endif
 
+#ifndef __nocfi
+# define __nocfi
+#endif
+
 #ifndef asm_volatile_goto
 #define asm_volatile_goto(x...) asm goto(x)
 #endif
diff --git a/include/linux/init.h b/include/linux/init.h
index 31f54de58429..b3ea15348fbd 100644
--- a/include/linux/init.h
+++ b/include/linux/init.h
@@ -47,7 +47,7 @@
 
 /* These are for everybody (although not all archs will actually
    discard it in modules) */
-#define __init		__section(".init.text") __cold  __latent_entropy __noinitretpoline
+#define __init		__section(".init.text") __cold  __latent_entropy __noinitretpoline __nocfi
 #define __initdata	__section(".init.data")
 #define __initconst	__section(".init.rodata")
 #define __exitdata	__section(".exit.data")
diff --git a/include/linux/module.h b/include/linux/module.h
index da4b6fbe8ebe..8100bb477d86 100644
--- a/include/linux/module.h
+++ b/include/linux/module.h
@@ -26,6 +26,7 @@
 #include <linux/tracepoint-defs.h>
 #include <linux/srcu.h>
 #include <linux/static_call_types.h>
+#include <linux/cfi.h>
 
 #include <linux/percpu.h>
 #include <asm/module.h>
@@ -128,13 +129,17 @@ extern void cleanup_module(void);
 #define module_init(initfn)					\
 	static inline initcall_t __maybe_unused __inittest(void)		\
 	{ return initfn; }					\
-	int init_module(void) __copy(initfn) __attribute__((alias(#initfn)));
+	int init_module(void) __copy(initfn)			\
+		__attribute__((alias(#initfn)));		\
+	__CFI_ADDRESSABLE(init_module, __initdata);
 
 /* This is only required if you want to be unloadable. */
 #define module_exit(exitfn)					\
 	static inline exitcall_t __maybe_unused __exittest(void)		\
 	{ return exitfn; }					\
-	void cleanup_module(void) __copy(exitfn) __attribute__((alias(#exitfn)));
+	void cleanup_module(void) __copy(exitfn)		\
+		__attribute__((alias(#exitfn)));		\
+	__CFI_ADDRESSABLE(cleanup_module, __exitdata);
 
 #endif
 
@@ -376,6 +381,10 @@ struct module {
 	const s32 *crcs;
 	unsigned int num_syms;
 
+#ifdef CONFIG_CFI_CLANG
+	cfi_check_fn cfi_check;
+#endif
+
 	/* Kernel parameters. */
 #ifdef CONFIG_SYSFS
 	struct mutex param_lock;
diff --git a/init/Kconfig b/init/Kconfig
index 5f5c776ef192..5babea38e346 100644
--- a/init/Kconfig
+++ b/init/Kconfig
@@ -2296,7 +2296,7 @@ endif # MODULES
 
 config MODULES_TREE_LOOKUP
 	def_bool y
-	depends on PERF_EVENTS || TRACING
+	depends on PERF_EVENTS || TRACING || CFI_CLANG
 
 config INIT_ALL_POSSIBLE
 	bool
diff --git a/kernel/Makefile b/kernel/Makefile
index 320f1f3941b7..e8a6715f38dc 100644
--- a/kernel/Makefile
+++ b/kernel/Makefile
@@ -41,6 +41,9 @@ KCSAN_SANITIZE_kcov.o := n
 UBSAN_SANITIZE_kcov.o := n
 CFLAGS_kcov.o := $(call cc-option, -fno-conserve-stack) -fno-stack-protector
 
+# Don't instrument error handlers
+CFLAGS_REMOVE_cfi.o := $(CC_FLAGS_CFI)
+
 obj-y += sched/
 obj-y += locking/
 obj-y += power/
@@ -111,6 +114,7 @@ obj-$(CONFIG_BPF) += bpf/
 obj-$(CONFIG_KCSAN) += kcsan/
 obj-$(CONFIG_SHADOW_CALL_STACK) += scs.o
 obj-$(CONFIG_HAVE_STATIC_CALL_INLINE) += static_call.o
+obj-$(CONFIG_CFI_CLANG) += cfi.o
 
 obj-$(CONFIG_PERF_EVENTS) += events/
 
diff --git a/kernel/cfi.c b/kernel/cfi.c
new file mode 100644
index 000000000000..e17a56639766
--- /dev/null
+++ b/kernel/cfi.c
@@ -0,0 +1,329 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Clang Control Flow Integrity (CFI) error and slowpath handling.
+ *
+ * Copyright (C) 2021 Google LLC
+ */
+
+#include <linux/hardirq.h>
+#include <linux/kallsyms.h>
+#include <linux/module.h>
+#include <linux/mutex.h>
+#include <linux/printk.h>
+#include <linux/ratelimit.h>
+#include <linux/rcupdate.h>
+#include <linux/vmalloc.h>
+#include <asm/cacheflush.h>
+#include <asm/set_memory.h>
+
+/* Compiler-defined handler names */
+#ifdef CONFIG_CFI_PERMISSIVE
+#define cfi_failure_handler	__ubsan_handle_cfi_check_fail
+#else
+#define cfi_failure_handler	__ubsan_handle_cfi_check_fail_abort
+#endif
+
+static inline void handle_cfi_failure(void *ptr)
+{
+	if (IS_ENABLED(CONFIG_CFI_PERMISSIVE))
+		WARN_RATELIMIT(1, "CFI failure (target: %pS):\n", ptr);
+	else
+		panic("CFI failure (target: %pS)\n", ptr);
+}
+
+#ifdef CONFIG_MODULES
+#ifdef CONFIG_CFI_CLANG_SHADOW
+/*
+ * Index type. A 16-bit index can address at most (2^16)-2 pages (taking
+ * into account SHADOW_INVALID), i.e. ~256M with 4k pages.
+ */
+typedef u16 shadow_t;
+#define SHADOW_INVALID		((shadow_t)~0UL)
+
+struct cfi_shadow {
+	/* Page index for the beginning of the shadow */
+	unsigned long base;
+	/* An array of __cfi_check locations (as indices to the shadow) */
+	shadow_t shadow[1];
+} __packed;
+
+/*
+ * The shadow covers ~128M from the beginning of the module region. If
+ * the region is larger, we fall back to __module_address for the rest.
+ */
+#define __SHADOW_RANGE		(_UL(SZ_128M) >> PAGE_SHIFT)
+
+/* The in-memory size of struct cfi_shadow, always at least one page */
+#define __SHADOW_PAGES		((__SHADOW_RANGE * sizeof(shadow_t)) >> PAGE_SHIFT)
+#define SHADOW_PAGES		max(1UL, __SHADOW_PAGES)
+#define SHADOW_SIZE		(SHADOW_PAGES << PAGE_SHIFT)
+
+/* The actual size of the shadow array, minus metadata */
+#define SHADOW_ARR_SIZE		(SHADOW_SIZE - offsetof(struct cfi_shadow, shadow))
+#define SHADOW_ARR_SLOTS	(SHADOW_ARR_SIZE / sizeof(shadow_t))
+
+static DEFINE_MUTEX(shadow_update_lock);
+static struct cfi_shadow __rcu *cfi_shadow __read_mostly;
+
+/* Returns the index in the shadow for the given address */
+static inline int ptr_to_shadow(const struct cfi_shadow *s, unsigned long ptr)
+{
+	unsigned long index;
+	unsigned long page = ptr >> PAGE_SHIFT;
+
+	if (unlikely(page < s->base))
+		return -1; /* Outside of module area */
+
+	index = page - s->base;
+
+	if (index >= SHADOW_ARR_SLOTS)
+		return -1; /* Cannot be addressed with shadow */
+
+	return (int)index;
+}
+
+/* Returns the page address for an index in the shadow */
+static inline unsigned long shadow_to_ptr(const struct cfi_shadow *s,
+	int index)
+{
+	if (unlikely(index < 0 || index >= SHADOW_ARR_SLOTS))
+		return 0;
+
+	return (s->base + index) << PAGE_SHIFT;
+}
+
+/* Returns the __cfi_check function address for the given shadow location */
+static inline unsigned long shadow_to_check_fn(const struct cfi_shadow *s,
+	int index)
+{
+	if (unlikely(index < 0 || index >= SHADOW_ARR_SLOTS))
+		return 0;
+
+	if (unlikely(s->shadow[index] == SHADOW_INVALID))
+		return 0;
+
+	/* __cfi_check is always page aligned */
+	return (s->base + s->shadow[index]) << PAGE_SHIFT;
+}
+
+static void prepare_next_shadow(const struct cfi_shadow __rcu *prev,
+		struct cfi_shadow *next)
+{
+	int i, index, check;
+
+	/* Mark everything invalid */
+	memset(next->shadow, 0xFF, SHADOW_ARR_SIZE);
+
+	if (!prev)
+		return; /* No previous shadow */
+
+	/* If the base address didn't change, an update is not needed */
+	if (prev->base == next->base) {
+		memcpy(next->shadow, prev->shadow, SHADOW_ARR_SIZE);
+		return;
+	}
+
+	/* Convert the previous shadow to the new address range */
+	for (i = 0; i < SHADOW_ARR_SLOTS; ++i) {
+		if (prev->shadow[i] == SHADOW_INVALID)
+			continue;
+
+		index = ptr_to_shadow(next, shadow_to_ptr(prev, i));
+		if (index < 0)
+			continue;
+
+		check = ptr_to_shadow(next,
+				shadow_to_check_fn(prev, prev->shadow[i]));
+		if (check < 0)
+			continue;
+
+		next->shadow[index] = (shadow_t)check;
+	}
+}
+
+static void add_module_to_shadow(struct cfi_shadow *s, struct module *mod,
+			unsigned long min_addr, unsigned long max_addr)
+{
+	int check_index;
+	unsigned long check = (unsigned long)mod->cfi_check;
+	unsigned long ptr;
+
+	if (unlikely(!PAGE_ALIGNED(check))) {
+		pr_warn("cfi: not using shadow for module %s\n", mod->name);
+		return;
+	}
+
+	check_index = ptr_to_shadow(s, check);
+	if (check_index < 0)
+		return; /* Module not addressable with shadow */
+
+	/* For each page, store the check function index in the shadow */
+	for (ptr = min_addr; ptr <= max_addr; ptr += PAGE_SIZE) {
+		int index = ptr_to_shadow(s, ptr);
+
+		if (index >= 0) {
+			/* Each page must only contain one module */
+			WARN_ON_ONCE(s->shadow[index] != SHADOW_INVALID);
+			s->shadow[index] = (shadow_t)check_index;
+		}
+	}
+}
+
+static void remove_module_from_shadow(struct cfi_shadow *s, struct module *mod,
+		unsigned long min_addr, unsigned long max_addr)
+{
+	unsigned long ptr;
+
+	for (ptr = min_addr; ptr <= max_addr; ptr += PAGE_SIZE) {
+		int index = ptr_to_shadow(s, ptr);
+
+		if (index >= 0)
+			s->shadow[index] = SHADOW_INVALID;
+	}
+}
+
+typedef void (*update_shadow_fn)(struct cfi_shadow *, struct module *,
+			unsigned long min_addr, unsigned long max_addr);
+
+static void update_shadow(struct module *mod, unsigned long base_addr,
+		update_shadow_fn fn)
+{
+	struct cfi_shadow *prev;
+	struct cfi_shadow *next;
+	unsigned long min_addr, max_addr;
+
+	next = vmalloc(SHADOW_SIZE);
+
+	mutex_lock(&shadow_update_lock);
+	prev = rcu_dereference_protected(cfi_shadow,
+					 mutex_is_locked(&shadow_update_lock));
+
+	if (next) {
+		next->base = base_addr >> PAGE_SHIFT;
+		prepare_next_shadow(prev, next);
+
+		min_addr = (unsigned long)mod->core_layout.base;
+		max_addr = min_addr + mod->core_layout.text_size;
+		fn(next, mod, min_addr & PAGE_MASK, max_addr & PAGE_MASK);
+
+		set_memory_ro((unsigned long)next, SHADOW_PAGES);
+	}
+
+	rcu_assign_pointer(cfi_shadow, next);
+	mutex_unlock(&shadow_update_lock);
+	synchronize_rcu();
+
+	if (prev) {
+		set_memory_rw((unsigned long)prev, SHADOW_PAGES);
+		vfree(prev);
+	}
+}
+
+void cfi_module_add(struct module *mod, unsigned long base_addr)
+{
+	update_shadow(mod, base_addr, add_module_to_shadow);
+}
+
+void cfi_module_remove(struct module *mod, unsigned long base_addr)
+{
+	update_shadow(mod, base_addr, remove_module_from_shadow);
+}
+
+static inline cfi_check_fn ptr_to_check_fn(const struct cfi_shadow __rcu *s,
+	unsigned long ptr)
+{
+	int index;
+
+	if (unlikely(!s))
+		return NULL; /* No shadow available */
+
+	index = ptr_to_shadow(s, ptr);
+	if (index < 0)
+		return NULL; /* Cannot be addressed with shadow */
+
+	return (cfi_check_fn)shadow_to_check_fn(s, index);
+}
+
+static inline cfi_check_fn find_shadow_check_fn(unsigned long ptr)
+{
+	cfi_check_fn fn;
+
+	rcu_read_lock_sched();
+	fn = ptr_to_check_fn(rcu_dereference_sched(cfi_shadow), ptr);
+	rcu_read_unlock_sched();
+
+	return fn;
+}
+
+#else /* !CONFIG_CFI_CLANG_SHADOW */
+
+static inline cfi_check_fn find_shadow_check_fn(unsigned long ptr)
+{
+	return NULL;
+}
+
+#endif /* CONFIG_CFI_CLANG_SHADOW */
+
+static inline cfi_check_fn find_module_check_fn(unsigned long ptr)
+{
+	cfi_check_fn fn = NULL;
+	struct module *mod;
+
+	rcu_read_lock_sched();
+	mod = __module_address(ptr);
+	if (mod)
+		fn = mod->cfi_check;
+	rcu_read_unlock_sched();
+
+	return fn;
+}
+
+static inline cfi_check_fn find_check_fn(unsigned long ptr)
+{
+	cfi_check_fn fn = NULL;
+
+	if (is_kernel_text(ptr))
+		return __cfi_check;
+
+	/*
+	 * Indirect call checks can happen when RCU is not watching. Both
+	 * the shadow and __module_address use RCU, so we need to wake it
+	 * up if necessary.
+	 */
+	RCU_NONIDLE({
+		if (IS_ENABLED(CONFIG_CFI_CLANG_SHADOW))
+			fn = find_shadow_check_fn(ptr);
+
+		if (!fn)
+			fn = find_module_check_fn(ptr);
+	});
+
+	return fn;
+}
+
+void __cfi_slowpath_diag(uint64_t id, void *ptr, void *diag)
+{
+	cfi_check_fn fn = find_check_fn((unsigned long)ptr);
+
+	if (likely(fn))
+		fn(id, ptr, diag);
+	else /* Don't allow unchecked modules */
+		handle_cfi_failure(ptr);
+}
+EXPORT_SYMBOL(__cfi_slowpath_diag);
+
+#else /* !CONFIG_MODULES */
+
+void __cfi_slowpath_diag(uint64_t id, void *ptr, void *diag)
+{
+	handle_cfi_failure(ptr); /* No modules */
+}
+EXPORT_SYMBOL(__cfi_slowpath_diag);
+
+#endif /* CONFIG_MODULES */
+
+void cfi_failure_handler(void *data, void *ptr, void *vtable)
+{
+	handle_cfi_failure(ptr);
+}
+EXPORT_SYMBOL(cfi_failure_handler);
diff --git a/kernel/module.c b/kernel/module.c
index 30479355ab85..20fb004e7d8d 100644
--- a/kernel/module.c
+++ b/kernel/module.c
@@ -2146,6 +2146,8 @@ void __weak module_arch_freeing_init(struct module *mod)
 {
 }
 
+static void cfi_cleanup(struct module *mod);
+
 /* Free a module, remove from lists, etc. */
 static void free_module(struct module *mod)
 {
@@ -2187,6 +2189,9 @@ static void free_module(struct module *mod)
 	synchronize_rcu();
 	mutex_unlock(&module_mutex);
 
+	/* Clean up CFI for the module. */
+	cfi_cleanup(mod);
+
 	/* This may be empty, but that's OK */
 	module_arch_freeing_init(mod);
 	module_memfree(mod->init_layout.base);
@@ -3866,6 +3871,8 @@ static int unknown_module_param_cb(char *param, char *val, const char *modname,
 	return 0;
 }
 
+static void cfi_init(struct module *mod);
+
 /*
  * Allocate and load the module: note that size of section 0 is always
  * zero, and we rely on this for optional sections.
@@ -3997,6 +4004,9 @@ static int load_module(struct load_info *info, const char __user *uargs,
 
 	flush_module_icache(mod);
 
+	/* Setup CFI for the module. */
+	cfi_init(mod);
+
 	/* Now copy in args */
 	mod->args = strndup_user(uargs, ~0UL >> 1);
 	if (IS_ERR(mod->args)) {
@@ -4070,6 +4080,7 @@ static int load_module(struct load_info *info, const char __user *uargs,
 	synchronize_rcu();
 	kfree(mod->args);
  free_arch_cleanup:
+	cfi_cleanup(mod);
 	module_arch_cleanup(mod);
  free_modinfo:
 	free_modinfo(mod);
@@ -4415,6 +4426,38 @@ int module_kallsyms_on_each_symbol(int (*fn)(void *, const char *,
 #endif /* CONFIG_LIVEPATCH */
 #endif /* CONFIG_KALLSYMS */
 
+static void cfi_init(struct module *mod)
+{
+#ifdef CONFIG_CFI_CLANG
+	initcall_t *init;
+	exitcall_t *exit;
+
+	rcu_read_lock_sched();
+	mod->cfi_check = (cfi_check_fn)
+		find_kallsyms_symbol_value(mod, "__cfi_check");
+	init = (initcall_t *)
+		find_kallsyms_symbol_value(mod, "__cfi_jt_init_module");
+	exit = (exitcall_t *)
+		find_kallsyms_symbol_value(mod, "__cfi_jt_cleanup_module");
+	rcu_read_unlock_sched();
+
+	/* Fix init/exit functions to point to the CFI jump table */
+	if (init)
+		mod->init = *init;
+	if (exit)
+		mod->exit = *exit;
+
+	cfi_module_add(mod, module_addr_min);
+#endif
+}
+
+static void cfi_cleanup(struct module *mod)
+{
+#ifdef CONFIG_CFI_CLANG
+	cfi_module_remove(mod, module_addr_min);
+#endif
+}
+
 /* Maximum number of characters written by module_flags() */
 #define MODULE_FLAGS_BUF_SIZE (TAINT_FLAGS_COUNT + 4)
 
diff --git a/scripts/Makefile.modfinal b/scripts/Makefile.modfinal
index 735e11e9041b..dd87cea9fba7 100644
--- a/scripts/Makefile.modfinal
+++ b/scripts/Makefile.modfinal
@@ -23,7 +23,7 @@ modname = $(notdir $(@:.mod.o=))
 part-of-module = y
 
 quiet_cmd_cc_o_c = CC [M]  $@
-      cmd_cc_o_c = $(CC) $(c_flags) -c -o $@ $<
+      cmd_cc_o_c = $(CC) $(filter-out $(CC_FLAGS_CFI), $(c_flags)) -c -o $@ $<
 
 %.mod.o: %.mod.c FORCE
 	$(call if_changed_dep,cc_o_c)
-- 
2.31.0.291.g576ba9dcdaf-goog

[PATCH v3 02/17] cfi: add __cficanonical

[Sami Tolvanen]

With CONFIG_CFI_CLANG, the compiler replaces a function address taken
in C code with the address of a local jump table entry, which passes
runtime indirect call checks. However, the compiler won't replace
addresses taken in assembly code, which will result in a CFI failure
if we later jump to such an address in instrumented C code. The code
generated for the non-canonical jump table looks this:

  <noncanonical.cfi_jt>: /* In C, &noncanonical points here */
	jmp noncanonical
  ...
  <noncanonical>:        /* function body */
	...

This change adds the __cficanonical attribute, which tells the
compiler to use a canonical jump table for the function instead. This
means the compiler will rename the actual function to <function>.cfi
and points the original symbol to the jump table entry instead:

  <canonical>:           /* jump table entry */
	jmp canonical.cfi
  ...
  <canonical.cfi>:       /* function body */
	...

As a result, the address taken in assembly, or other non-instrumented
code always points to the jump table and therefore, can be used for
indirect calls in instrumented code without tripping CFI checks.

Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>   # pci.h
---
 include/linux/compiler-clang.h | 1 +
 include/linux/compiler_types.h | 4 ++++
 include/linux/init.h           | 4 ++--
 include/linux/pci.h            | 4 ++--
 4 files changed, 9 insertions(+), 4 deletions(-)

diff --git a/include/linux/compiler-clang.h b/include/linux/compiler-clang.h
index 6de9d0c9377e..adbe76b203e2 100644
--- a/include/linux/compiler-clang.h
+++ b/include/linux/compiler-clang.h
@@ -63,3 +63,4 @@
 #endif
 
 #define __nocfi		__attribute__((__no_sanitize__("cfi")))
+#define __cficanonical	__attribute__((__cfi_canonical_jump_table__))
diff --git a/include/linux/compiler_types.h b/include/linux/compiler_types.h
index 796935a37e37..d29bda7f6ebd 100644
--- a/include/linux/compiler_types.h
+++ b/include/linux/compiler_types.h
@@ -246,6 +246,10 @@ struct ftrace_likely_data {
 # define __nocfi
 #endif
 
+#ifndef __cficanonical
+# define __cficanonical
+#endif
+
 #ifndef asm_volatile_goto
 #define asm_volatile_goto(x...) asm goto(x)
 #endif
diff --git a/include/linux/init.h b/include/linux/init.h
index b3ea15348fbd..045ad1650ed1 100644
--- a/include/linux/init.h
+++ b/include/linux/init.h
@@ -220,8 +220,8 @@ extern bool initcall_debug;
 	__initcall_name(initstub, __iid, id)
 
 #define __define_initcall_stub(__stub, fn)			\
-	int __init __stub(void);				\
-	int __init __stub(void)					\
+	int __init __cficanonical __stub(void);			\
+	int __init __cficanonical __stub(void)			\
 	{ 							\
 		return fn();					\
 	}							\
diff --git a/include/linux/pci.h b/include/linux/pci.h
index 86c799c97b77..39684b72db91 100644
--- a/include/linux/pci.h
+++ b/include/linux/pci.h
@@ -1944,8 +1944,8 @@ enum pci_fixup_pass {
 #ifdef CONFIG_LTO_CLANG
 #define __DECLARE_PCI_FIXUP_SECTION(sec, name, vendor, device, class,	\
 				  class_shift, hook, stub)		\
-	void stub(struct pci_dev *dev);					\
-	void stub(struct pci_dev *dev)					\
+	void __cficanonical stub(struct pci_dev *dev);			\
+	void __cficanonical stub(struct pci_dev *dev)			\
 	{ 								\
 		hook(dev); 						\
 	}								\
-- 
2.31.0.291.g576ba9dcdaf-goog

Re: [PATCH v3 02/17] cfi: add __cficanonical

[Rasmus Villemoes]

On 23/03/2021 21.39, Sami Tolvanen wrote:
> With CONFIG_CFI_CLANG, the compiler replaces a function address taken
> in C code with the address of a local jump table entry, which passes
> runtime indirect call checks. However, the compiler won't replace
> addresses taken in assembly code, which will result in a CFI failure
> if we later jump to such an address in instrumented C code. The code
> generated for the non-canonical jump table looks this:
> 
>   <noncanonical.cfi_jt>: /* In C, &noncanonical points here */
> 	jmp noncanonical
>   ...
>   <noncanonical>:        /* function body */
> 	...
> 
> This change adds the __cficanonical attribute, which tells the
> compiler to use a canonical jump table for the function instead. This
> means the compiler will rename the actual function to <function>.cfi
> and points the original symbol to the jump table entry instead:
> 
>   <canonical>:           /* jump table entry */
> 	jmp canonical.cfi
>   ...
>   <canonical.cfi>:       /* function body */
> 	...
> 
> As a result, the address taken in assembly, or other non-instrumented
> code always points to the jump table and therefore, can be used for
> indirect calls in instrumented code without tripping CFI checks.

Random ramblings, I'm trying to understand how this CFI stuff works.

First, patch 1 and 2 explain the pros and cons of canonical vs
non-canonical jump tables, in either case, there's problems with stuff
implemented in assembly. But I don't understand why those pros and cons
then end up with using the non-canonical jump tables by default. IIUC,
with canonical jump tables, function pointer equality would keep working
for functions implemented in C, because &func would always refer to the
same stub "function" that lives in the same object file as func.cfi,
whereas with the non-canonical version, each TU (or maybe DSO) that
takes the address of func ends up with its own func.cfi_jt.

There are of course lots of direct calls of assembly functions, but
I don't think we take the address of such functions very often. So why
can't we instead equip the declarations of those with a
__cfi_noncanonical attribute?

And now, more directed at the clang folks on cc:

As to how CFI works, I've tried to make sense of the clang docs. So at
place where some int (*)(long, int) function pointer is called, the
compiler computes (roughly) md5sum("int (*)(long, int)") and uses the
first 8 bytes as a cookie representing that type. It then goes to some
global table of jump table ranges indexed by that cookie and checks that
the address it is about to call is within that range. All jump table
entries for one type of function are consecutive in memory (with
complications arising from cross-DSO calls).

What I don't understand about all this is why that indirection through
some hidden global table and magic jump table (whether canonical or not)
is even needed in the simple common case of ordinary C functions. Why
can't the compiler just emit the cookie corresponding to a given
function's prototype immediately prior to the function? Then the inline
check would just be "if (*(u64*)((void*)func - 8) == cookie)" and
function pointer comparison would just work because there's no magic
involved when doing &func. Cross-DSO calls of C function have no extra
cost to look up a __cfi_check function in the target DSO. An indirect
call doesn't touch at least two extra cache lines (the range table and
the jump table entry). It seems to rely on LTO anyway, so it's not even
that the compiler would have to emit that cookie for every single
function, it knows at link time which functions have their address
taken. Calling functions implemented in assembly through a function
pointer will have the same problem as with the "canonical" jump table
approach, but with a suitable attribute on those surely the compiler
could emit a func.cfi_hoop

  .quad 0x1122334455667788 // cookie
  <func.cfi_hoop>:
	jmp func

and perhaps no such attribute would even be needed (with LTO, the
compiler should be able to see "hey, I don't know that function, it's
probably implemented in assembly, so lemme emit that trampoline with a
cookie in front and redirect address-of to that").

Rasmus

Re: [PATCH v3 02/17] cfi: add __cficanonical

[Sami Tolvanen]

On Wed, Mar 24, 2021 at 8:31 AM Rasmus Villemoes
<linux@rasmusvillemoes.dk> wrote:
>
> On 23/03/2021 21.39, Sami Tolvanen wrote:
> > With CONFIG_CFI_CLANG, the compiler replaces a function address taken
> > in C code with the address of a local jump table entry, which passes
> > runtime indirect call checks. However, the compiler won't replace
> > addresses taken in assembly code, which will result in a CFI failure
> > if we later jump to such an address in instrumented C code. The code
> > generated for the non-canonical jump table looks this:
> >
> >   <noncanonical.cfi_jt>: /* In C, &noncanonical points here */
> >       jmp noncanonical
> >   ...
> >   <noncanonical>:        /* function body */
> >       ...
> >
> > This change adds the __cficanonical attribute, which tells the
> > compiler to use a canonical jump table for the function instead. This
> > means the compiler will rename the actual function to <function>.cfi
> > and points the original symbol to the jump table entry instead:
> >
> >   <canonical>:           /* jump table entry */
> >       jmp canonical.cfi
> >   ...
> >   <canonical.cfi>:       /* function body */
> >       ...
> >
> > As a result, the address taken in assembly, or other non-instrumented
> > code always points to the jump table and therefore, can be used for
> > indirect calls in instrumented code without tripping CFI checks.
>
> Random ramblings, I'm trying to understand how this CFI stuff works.
>
> First, patch 1 and 2 explain the pros and cons of canonical vs
> non-canonical jump tables, in either case, there's problems with stuff
> implemented in assembly. But I don't understand why those pros and cons
> then end up with using the non-canonical jump tables by default. IIUC,
> with canonical jump tables, function pointer equality would keep working
> for functions implemented in C, because &func would always refer to the
> same stub "function" that lives in the same object file as func.cfi,
> whereas with the non-canonical version, each TU (or maybe DSO) that
> takes the address of func ends up with its own func.cfi_jt.

Correct.

> There are of course lots of direct calls of assembly functions, but
> I don't think we take the address of such functions very often. So why
> can't we instead equip the declarations of those with a
> __cfi_noncanonical attribute?

Clang doesn't support these attributes in function declarations,
unfortunately. If it did, that would certainly help, until someone
wants to compare addresses of assembly functions, in which case we
would again have a problem.

Another way to work around the issue with canonical CFI would be to
add C wrappers for all address-taken assembly functions, but that's
not quite ideal either. I think most indirect calls to assembly
functions happen in the crypto code, which would have required so many
changes that we decided to default to non-canonical CFI instead. This
resulted in far fewer kernel changes despite the cross-module function
address equality issue.

Sami

[PATCH v3 03/17] mm: add generic __va_function and __pa_function macros

[Sami Tolvanen]

With CONFIG_CFI_CLANG, the compiler replaces function addresses
in instrumented C code with jump table addresses. This means that
__pa_symbol(function) returns the physical address of the jump table
entry instead of the actual function, which may not work as the jump
table code will immediately jump to a virtual address that may not be
mapped.

To avoid this address space confusion, this change adds generic
definitions for __va_function and __pa_function, which architectures
that support CFI can override. The typical implementation of the
__va_function macro would use inline assembly to take the function
address, which avoids compiler instrumentation.

Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
---
 include/linux/mm.h | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/include/linux/mm.h b/include/linux/mm.h
index 64a71bf20536..a0d285cd59ce 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -116,6 +116,14 @@ extern int mmap_rnd_compat_bits __read_mostly;
 #define __pa_symbol(x)  __pa(RELOC_HIDE((unsigned long)(x), 0))
 #endif
 
+#ifndef __va_function
+#define __va_function(x) (x)
+#endif
+
+#ifndef __pa_function
+#define __pa_function(x) __pa_symbol(__va_function(x))
+#endif
+
 #ifndef page_to_virt
 #define page_to_virt(x)	__va(PFN_PHYS(page_to_pfn(x)))
 #endif
-- 
2.31.0.291.g576ba9dcdaf-goog

Re: [PATCH v3 03/17] mm: add generic __va_function and __pa_function macros

[Christoph Hellwig]

On Tue, Mar 23, 2021 at 01:39:32PM -0700, Sami Tolvanen wrote:
> With CONFIG_CFI_CLANG, the compiler replaces function addresses
> in instrumented C code with jump table addresses. This means that
> __pa_symbol(function) returns the physical address of the jump table
> entry instead of the actual function, which may not work as the jump
> table code will immediately jump to a virtual address that may not be
> mapped.
> 
> To avoid this address space confusion, this change adds generic
> definitions for __va_function and __pa_function, which architectures
> that support CFI can override. The typical implementation of the
> __va_function macro would use inline assembly to take the function
> address, which avoids compiler instrumentation.

I think these helper are sensible, but shouldn't they have somewhat
less arcane names and proper documentation?

Re: [PATCH v3 03/17] mm: add generic __va_function and __pa_function macros

[Sami Tolvanen]

On Wed, Mar 24, 2021 at 12:14 AM Christoph Hellwig <hch@infradead.org> wrote:
>
> On Tue, Mar 23, 2021 at 01:39:32PM -0700, Sami Tolvanen wrote:
> > With CONFIG_CFI_CLANG, the compiler replaces function addresses
> > in instrumented C code with jump table addresses. This means that
> > __pa_symbol(function) returns the physical address of the jump table
> > entry instead of the actual function, which may not work as the jump
> > table code will immediately jump to a virtual address that may not be
> > mapped.
> >
> > To avoid this address space confusion, this change adds generic
> > definitions for __va_function and __pa_function, which architectures
> > that support CFI can override. The typical implementation of the
> > __va_function macro would use inline assembly to take the function
> > address, which avoids compiler instrumentation.
>
> I think these helper are sensible, but shouldn't they have somewhat
> less arcane names and proper documentation?

Good point, I'll add comments in the next version. I thought
__pa_function would be a fairly straightforward replacement for
__pa_symbol, but I'm fine with renaming these. Any suggestions for
less arcane names?

Sami

Re: [PATCH v3 03/17] mm: add generic __va_function and __pa_function macros

[Mark Rutland]

On Wed, Mar 24, 2021 at 08:54:18AM -0700, Sami Tolvanen wrote:
> On Wed, Mar 24, 2021 at 12:14 AM Christoph Hellwig <hch@infradead.org> wrote:
> >
> > On Tue, Mar 23, 2021 at 01:39:32PM -0700, Sami Tolvanen wrote:
> > > With CONFIG_CFI_CLANG, the compiler replaces function addresses
> > > in instrumented C code with jump table addresses. This means that
> > > __pa_symbol(function) returns the physical address of the jump table
> > > entry instead of the actual function, which may not work as the jump
> > > table code will immediately jump to a virtual address that may not be
> > > mapped.
> > >
> > > To avoid this address space confusion, this change adds generic
> > > definitions for __va_function and __pa_function, which architectures
> > > that support CFI can override. The typical implementation of the
> > > __va_function macro would use inline assembly to take the function
> > > address, which avoids compiler instrumentation.
> >
> > I think these helper are sensible, but shouldn't they have somewhat
> > less arcane names and proper documentation?
> 
> Good point, I'll add comments in the next version. I thought
> __pa_function would be a fairly straightforward replacement for
> __pa_symbol, but I'm fine with renaming these. Any suggestions for
> less arcane names?

I think dropping 'nocfi' into the name would be clear enough. I think
that given the usual fun with {symbol,module,virt}->phys conversions
it's not worth having the __pa_* form, and we can leave the phys
conversion to the caller that knows where the function lives.

How about we just add `function_nocfi()` ?

Callers can then do `__pa_symbol(function_nocfi(foo))` and similar.

Thanks,
Mark.

Re: [PATCH v3 03/17] mm: add generic __va_function and __pa_function macros

[Sami Tolvanen]

On Thu, Mar 25, 2021 at 3:17 AM Mark Rutland <mark.rutland@arm.com> wrote:
>
> On Wed, Mar 24, 2021 at 08:54:18AM -0700, Sami Tolvanen wrote:
> > On Wed, Mar 24, 2021 at 12:14 AM Christoph Hellwig <hch@infradead.org> wrote:
> > >
> > > On Tue, Mar 23, 2021 at 01:39:32PM -0700, Sami Tolvanen wrote:
> > > > With CONFIG_CFI_CLANG, the compiler replaces function addresses
> > > > in instrumented C code with jump table addresses. This means that
> > > > __pa_symbol(function) returns the physical address of the jump table
> > > > entry instead of the actual function, which may not work as the jump
> > > > table code will immediately jump to a virtual address that may not be
> > > > mapped.
> > > >
> > > > To avoid this address space confusion, this change adds generic
> > > > definitions for __va_function and __pa_function, which architectures
> > > > that support CFI can override. The typical implementation of the
> > > > __va_function macro would use inline assembly to take the function
> > > > address, which avoids compiler instrumentation.
> > >
> > > I think these helper are sensible, but shouldn't they have somewhat
> > > less arcane names and proper documentation?
> >
> > Good point, I'll add comments in the next version. I thought
> > __pa_function would be a fairly straightforward replacement for
> > __pa_symbol, but I'm fine with renaming these. Any suggestions for
> > less arcane names?
>
> I think dropping 'nocfi' into the name would be clear enough. I think
> that given the usual fun with {symbol,module,virt}->phys conversions
> it's not worth having the __pa_* form, and we can leave the phys
> conversion to the caller that knows where the function lives.
>
> How about we just add `function_nocfi()` ?
>
> Callers can then do `__pa_symbol(function_nocfi(foo))` and similar.

Sounds reasonable. I'll drop __pa_function() and rename
__va_function() to function_nocfi() in the next version.

Sami

[PATCH v3 04/17] module: ensure __cfi_check alignment

[Sami Tolvanen]

CONFIG_CFI_CLANG_SHADOW assumes the __cfi_check() function is page
aligned and at the beginning of the .text section. While Clang would
normally align the function correctly, it fails to do so for modules
with no executable code.

This change ensures the correct __cfi_check() location and
alignment. It also discards the .eh_frame section, which Clang can
generate with certain sanitizers, such as CFI.

Link: https://bugs.llvm.org/show_bug.cgi?id=46293
Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
---
 scripts/module.lds.S | 20 +++++++++++++++++++-
 1 file changed, 19 insertions(+), 1 deletion(-)

diff --git a/scripts/module.lds.S b/scripts/module.lds.S
index 168cd27e6122..2ba9e5ce71df 100644
--- a/scripts/module.lds.S
+++ b/scripts/module.lds.S
@@ -3,10 +3,21 @@
  * Archs are free to supply their own linker scripts.  ld will
  * combine them automatically.
  */
+#include <asm/page.h>
+
+#ifdef CONFIG_CFI_CLANG
+# define ALIGN_CFI 		ALIGN(PAGE_SIZE)
+# define SANITIZER_DISCARDS	*(.eh_frame)
+#else
+# define ALIGN_CFI
+# define SANITIZER_DISCARDS
+#endif
+
 SECTIONS {
 	/DISCARD/ : {
 		*(.discard)
 		*(.discard.*)
+		SANITIZER_DISCARDS
 	}
 
 	__ksymtab		0 : { *(SORT(___ksymtab+*)) }
@@ -40,7 +51,14 @@ SECTIONS {
 		*(.rodata..L*)
 	}
 
-	.text : { *(.text .text.[0-9a-zA-Z_]*) }
+	/*
+	 * With CONFIG_CFI_CLANG, we assume __cfi_check is at the beginning
+	 * of the .text section, and is aligned to PAGE_SIZE.
+	 */
+	.text : ALIGN_CFI {
+		*(.text.__cfi_check)
+		*(.text .text.[0-9a-zA-Z_]* .text..L.cfi*)
+	}
 }
 
 /* bring in arch-specific sections */
-- 
2.31.0.291.g576ba9dcdaf-goog

Re: [PATCH v3 04/17] module: ensure __cfi_check alignment

[Kees Cook]

On Tue, Mar 23, 2021 at 01:39:33PM -0700, Sami Tolvanen wrote:
> CONFIG_CFI_CLANG_SHADOW assumes the __cfi_check() function is page
> aligned and at the beginning of the .text section. While Clang would
> normally align the function correctly, it fails to do so for modules
> with no executable code.
> 
> This change ensures the correct __cfi_check() location and
> alignment. It also discards the .eh_frame section, which Clang can
> generate with certain sanitizers, such as CFI.
> 
> Link: https://bugs.llvm.org/show_bug.cgi?id=46293
> Signed-off-by: Sami Tolvanen <samitolvanen@google.com>

Yay macros! :)

Reviewed-by: Kees Cook <keescook@chromium.org>

-- 
Kees Cook

Re: [PATCH v3 04/17] module: ensure __cfi_check alignment

[Jessica Yu]

+++ Sami Tolvanen [23/03/21 13:39 -0700]:
>CONFIG_CFI_CLANG_SHADOW assumes the __cfi_check() function is page
>aligned and at the beginning of the .text section. While Clang would
>normally align the function correctly, it fails to do so for modules
>with no executable code.
>
>This change ensures the correct __cfi_check() location and
>alignment. It also discards the .eh_frame section, which Clang can
>generate with certain sanitizers, such as CFI.
>
>Link: https://bugs.llvm.org/show_bug.cgi?id=46293
>Signed-off-by: Sami Tolvanen <samitolvanen@google.com>

Acked-by: Jessica Yu <jeyu@kernel.org>

Thanks!

[PATCH v3 05/17] workqueue: use WARN_ON_FUNCTION_MISMATCH

[Sami Tolvanen]

With CONFIG_CFI_CLANG, a callback function passed to
__queue_delayed_work from a module points to a jump table entry
defined in the module instead of the one used in the core kernel,
which breaks function address equality in this check:

  WARN_ON_ONCE(timer->function != delayed_work_timer_fn);

Use WARN_ON_FUNCTION_MISMATCH() instead to disable the warning
when CFI and modules are both enabled.

Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
---
 kernel/workqueue.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/workqueue.c b/kernel/workqueue.c
index 0d150da252e8..03fe07d2f39f 100644
--- a/kernel/workqueue.c
+++ b/kernel/workqueue.c
@@ -1630,7 +1630,7 @@ static void __queue_delayed_work(int cpu, struct workqueue_struct *wq,
 	struct work_struct *work = &dwork->work;
 
 	WARN_ON_ONCE(!wq);
-	WARN_ON_ONCE(timer->function != delayed_work_timer_fn);
+	WARN_ON_FUNCTION_MISMATCH(timer->function, delayed_work_timer_fn);
 	WARN_ON_ONCE(timer_pending(timer));
 	WARN_ON_ONCE(!list_empty(&work->entry));
 
-- 
2.31.0.291.g576ba9dcdaf-goog

Re: [PATCH v3 05/17] workqueue: use WARN_ON_FUNCTION_MISMATCH

[Kees Cook]

On Tue, Mar 23, 2021 at 01:39:34PM -0700, Sami Tolvanen wrote:
> With CONFIG_CFI_CLANG, a callback function passed to
> __queue_delayed_work from a module points to a jump table entry
> defined in the module instead of the one used in the core kernel,
> which breaks function address equality in this check:
> 
>   WARN_ON_ONCE(timer->function != delayed_work_timer_fn);
> 
> Use WARN_ON_FUNCTION_MISMATCH() instead to disable the warning
> when CFI and modules are both enabled.
> 
> Signed-off-by: Sami Tolvanen <samitolvanen@google.com>

Reviewed-by: Kees Cook <keescook@chromium.org>

-- 
Kees Cook

[PATCH v3 06/17] kthread: use WARN_ON_FUNCTION_MISMATCH

[Sami Tolvanen]

With CONFIG_CFI_CLANG, a callback function passed to
__kthread_queue_delayed_work from a module points to a jump table
entry defined in the module instead of the one used in the core
kernel, which breaks function address equality in this check:

  WARN_ON_ONCE(timer->function != ktead_delayed_work_timer_fn);

Use WARN_ON_FUNCTION_MISMATCH() instead to disable the warning
when CFI and modules are both enabled.

Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
---
 kernel/kthread.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/kernel/kthread.c b/kernel/kthread.c
index 1578973c5740..a1972eba2917 100644
--- a/kernel/kthread.c
+++ b/kernel/kthread.c
@@ -963,7 +963,8 @@ static void __kthread_queue_delayed_work(struct kthread_worker *worker,
 	struct timer_list *timer = &dwork->timer;
 	struct kthread_work *work = &dwork->work;
 
-	WARN_ON_ONCE(timer->function != kthread_delayed_work_timer_fn);
+	WARN_ON_FUNCTION_MISMATCH(timer->function,
+				  kthread_delayed_work_timer_fn);
 
 	/*
 	 * If @delay is 0, queue @dwork->work immediately.  This is for
-- 
2.31.0.291.g576ba9dcdaf-goog

Re: [PATCH v3 06/17] kthread: use WARN_ON_FUNCTION_MISMATCH

[Kees Cook]

On Tue, Mar 23, 2021 at 01:39:35PM -0700, Sami Tolvanen wrote:
> With CONFIG_CFI_CLANG, a callback function passed to
> __kthread_queue_delayed_work from a module points to a jump table
> entry defined in the module instead of the one used in the core
> kernel, which breaks function address equality in this check:
> 
>   WARN_ON_ONCE(timer->function != ktead_delayed_work_timer_fn);
> 
> Use WARN_ON_FUNCTION_MISMATCH() instead to disable the warning
> when CFI and modules are both enabled.
> 
> Signed-off-by: Sami Tolvanen <samitolvanen@google.com>

Reviewed-by: Kees Cook <keescook@chromium.org>

-- 
Kees Cook

[PATCH v3 07/17] kallsyms: strip ThinLTO hashes from static functions

[Sami Tolvanen]

With CONFIG_CFI_CLANG and ThinLTO, Clang appends a hash to the names
of all static functions not marked __used. This can break userspace
tools that don't expect the function name to change, so strip out the
hash from the output.

Suggested-by: Jack Pham <jackp@codeaurora.org>
Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
---
 kernel/kallsyms.c | 55 ++++++++++++++++++++++++++++++++++++++++++-----
 1 file changed, 50 insertions(+), 5 deletions(-)

diff --git a/kernel/kallsyms.c b/kernel/kallsyms.c
index 8043a90aa50e..c851ca0ed357 100644
--- a/kernel/kallsyms.c
+++ b/kernel/kallsyms.c
@@ -161,6 +161,27 @@ static unsigned long kallsyms_sym_address(int idx)
 	return kallsyms_relative_base - 1 - kallsyms_offsets[idx];
 }
 
+#if defined(CONFIG_CFI_CLANG) && defined(CONFIG_LTO_CLANG_THIN)
+/*
+ * LLVM appends a hash to static function names when ThinLTO and CFI are
+ * both enabled, i.e. foo() becomes foo$707af9a22804d33c81801f27dcfe489b.
+ * This causes confusion and potentially breaks user space tools, so we
+ * strip the suffix from expanded symbol names.
+ */
+static inline bool cleanup_symbol_name(char *s)
+{
+	char *res;
+
+	res = strrchr(s, '$');
+	if (res)
+		*res = '\0';
+
+	return res != NULL;
+}
+#else
+static inline bool cleanup_symbol_name(char *s) { return false; }
+#endif
+
 /* Lookup the address for this symbol. Returns 0 if not found. */
 unsigned long kallsyms_lookup_name(const char *name)
 {
@@ -173,6 +194,9 @@ unsigned long kallsyms_lookup_name(const char *name)
 
 		if (strcmp(namebuf, name) == 0)
 			return kallsyms_sym_address(i);
+
+		if (cleanup_symbol_name(namebuf) && strcmp(namebuf, name) == 0)
+			return kallsyms_sym_address(i);
 	}
 	return module_kallsyms_lookup_name(name);
 }
@@ -303,7 +327,9 @@ const char *kallsyms_lookup(unsigned long addr,
 				       namebuf, KSYM_NAME_LEN);
 		if (modname)
 			*modname = NULL;
-		return namebuf;
+
+		ret = namebuf;
+		goto found;
 	}
 
 	/* See if it's in a module or a BPF JITed image. */
@@ -316,11 +342,16 @@ const char *kallsyms_lookup(unsigned long addr,
 	if (!ret)
 		ret = ftrace_mod_address_lookup(addr, symbolsize,
 						offset, modname, namebuf);
+
+found:
+	cleanup_symbol_name(namebuf);
 	return ret;
 }
 
 int lookup_symbol_name(unsigned long addr, char *symname)
 {
+	int res;
+
 	symname[0] = '\0';
 	symname[KSYM_NAME_LEN - 1] = '\0';
 
@@ -331,15 +362,23 @@ int lookup_symbol_name(unsigned long addr, char *symname)
 		/* Grab name */
 		kallsyms_expand_symbol(get_symbol_offset(pos),
 				       symname, KSYM_NAME_LEN);
-		return 0;
+		goto found;
 	}
 	/* See if it's in a module. */
-	return lookup_module_symbol_name(addr, symname);
+	res = lookup_module_symbol_name(addr, symname);
+	if (res)
+		return res;
+
+found:
+	cleanup_symbol_name(symname);
+	return 0;
 }
 
 int lookup_symbol_attrs(unsigned long addr, unsigned long *size,
 			unsigned long *offset, char *modname, char *name)
 {
+	int res;
+
 	name[0] = '\0';
 	name[KSYM_NAME_LEN - 1] = '\0';
 
@@ -351,10 +390,16 @@ int lookup_symbol_attrs(unsigned long addr, unsigned long *size,
 		kallsyms_expand_symbol(get_symbol_offset(pos),
 				       name, KSYM_NAME_LEN);
 		modname[0] = '\0';
-		return 0;
+		goto found;
 	}
 	/* See if it's in a module. */
-	return lookup_module_symbol_attrs(addr, size, offset, modname, name);
+	res = lookup_module_symbol_attrs(addr, size, offset, modname, name);
+	if (res)
+		return res;
+
+found:
+	cleanup_symbol_name(name);
+	return 0;
 }
 
 /* Look up a kernel symbol and return it in a text buffer. */
-- 
2.31.0.291.g576ba9dcdaf-goog

[PATCH v3 08/17] bpf: disable CFI in dispatcher functions

[Sami Tolvanen]

BPF dispatcher functions are patched at runtime to perform direct
instead of indirect calls. Disable CFI for the dispatcher functions to
avoid conflicts.

Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
---
 include/linux/bpf.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/include/linux/bpf.h b/include/linux/bpf.h
index cccaef1088ea..9acdca574527 100644
--- a/include/linux/bpf.h
+++ b/include/linux/bpf.h
@@ -635,7 +635,7 @@ struct bpf_dispatcher {
 	struct bpf_ksym ksym;
 };
 
-static __always_inline unsigned int bpf_dispatcher_nop_func(
+static __always_inline __nocfi unsigned int bpf_dispatcher_nop_func(
 	const void *ctx,
 	const struct bpf_insn *insnsi,
 	unsigned int (*bpf_func)(const void *,
@@ -663,7 +663,7 @@ void bpf_trampoline_put(struct bpf_trampoline *tr);
 }
 
 #define DEFINE_BPF_DISPATCHER(name)					\
-	noinline unsigned int bpf_dispatcher_##name##_func(		\
+	noinline __nocfi unsigned int bpf_dispatcher_##name##_func(	\
 		const void *ctx,					\
 		const struct bpf_insn *insnsi,				\
 		unsigned int (*bpf_func)(const void *,			\
-- 
2.31.0.291.g576ba9dcdaf-goog

[PATCH v3 09/17] treewide: Change list_sort to use const pointers

[Sami Tolvanen]

list_sort() internally casts the comparison function passed to it
to a different type with constant struct list_head pointers, and
uses this pointer to call the functions, which trips indirect call
Control-Flow Integrity (CFI) checking.

Instead of removing the consts, this change defines the
list_cmp_func_t type and changes the comparison function types of
all list_sort() callers to use const pointers, thus avoiding type
mismatches.

Suggested-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
---
 arch/arm64/kvm/vgic/vgic-its.c                  |  8 ++++----
 arch/arm64/kvm/vgic/vgic.c                      |  3 ++-
 block/blk-mq-sched.c                            |  3 ++-
 block/blk-mq.c                                  |  3 ++-
 drivers/acpi/nfit/core.c                        |  3 ++-
 drivers/acpi/numa/hmat.c                        |  3 ++-
 drivers/clk/keystone/sci-clk.c                  |  4 ++--
 drivers/gpu/drm/drm_modes.c                     |  3 ++-
 drivers/gpu/drm/i915/gt/intel_engine_user.c     |  3 ++-
 drivers/gpu/drm/i915/gvt/debugfs.c              |  2 +-
 drivers/gpu/drm/i915/selftests/i915_gem_gtt.c   |  3 ++-
 drivers/gpu/drm/radeon/radeon_cs.c              |  4 ++--
 .../hw/usnic/usnic_uiom_interval_tree.c         |  3 ++-
 drivers/interconnect/qcom/bcm-voter.c           |  2 +-
 drivers/md/raid5.c                              |  3 ++-
 drivers/misc/sram.c                             |  4 ++--
 drivers/nvme/host/core.c                        |  3 ++-
 .../pci/controller/cadence/pcie-cadence-host.c  |  3 ++-
 drivers/spi/spi-loopback-test.c                 |  3 ++-
 fs/btrfs/raid56.c                               |  3 ++-
 fs/btrfs/tree-log.c                             |  3 ++-
 fs/btrfs/volumes.c                              |  3 ++-
 fs/ext4/fsmap.c                                 |  4 ++--
 fs/gfs2/glock.c                                 |  3 ++-
 fs/gfs2/log.c                                   |  2 +-
 fs/gfs2/lops.c                                  |  3 ++-
 fs/iomap/buffered-io.c                          |  3 ++-
 fs/ubifs/gc.c                                   |  7 ++++---
 fs/ubifs/replay.c                               |  4 ++--
 fs/xfs/scrub/bitmap.c                           |  4 ++--
 fs/xfs/xfs_bmap_item.c                          |  4 ++--
 fs/xfs/xfs_buf.c                                |  6 +++---
 fs/xfs/xfs_extent_busy.c                        |  4 ++--
 fs/xfs/xfs_extent_busy.h                        |  3 ++-
 fs/xfs/xfs_extfree_item.c                       |  4 ++--
 fs/xfs/xfs_refcount_item.c                      |  4 ++--
 fs/xfs/xfs_rmap_item.c                          |  4 ++--
 include/linux/list_sort.h                       |  7 ++++---
 lib/list_sort.c                                 | 17 ++++++-----------
 lib/test_list_sort.c                            |  3 ++-
 net/tipc/name_table.c                           |  4 ++--
 41 files changed, 90 insertions(+), 72 deletions(-)

diff --git a/arch/arm64/kvm/vgic/vgic-its.c b/arch/arm64/kvm/vgic/vgic-its.c
index 40cbaca81333..b9518f94bd43 100644
--- a/arch/arm64/kvm/vgic/vgic-its.c
+++ b/arch/arm64/kvm/vgic/vgic-its.c
@@ -2190,8 +2190,8 @@ static int vgic_its_restore_ite(struct vgic_its *its, u32 event_id,
 	return offset;
 }
 
-static int vgic_its_ite_cmp(void *priv, struct list_head *a,
-			    struct list_head *b)
+static int vgic_its_ite_cmp(void *priv, const struct list_head *a,
+			    const struct list_head *b)
 {
 	struct its_ite *itea = container_of(a, struct its_ite, ite_list);
 	struct its_ite *iteb = container_of(b, struct its_ite, ite_list);
@@ -2329,8 +2329,8 @@ static int vgic_its_restore_dte(struct vgic_its *its, u32 id,
 	return offset;
 }
 
-static int vgic_its_device_cmp(void *priv, struct list_head *a,
-			       struct list_head *b)
+static int vgic_its_device_cmp(void *priv, const struct list_head *a,
+			       const struct list_head *b)
 {
 	struct its_device *deva = container_of(a, struct its_device, dev_list);
 	struct its_device *devb = container_of(b, struct its_device, dev_list);
diff --git a/arch/arm64/kvm/vgic/vgic.c b/arch/arm64/kvm/vgic/vgic.c
index 1c597c9885fa..15b666200f0b 100644
--- a/arch/arm64/kvm/vgic/vgic.c
+++ b/arch/arm64/kvm/vgic/vgic.c
@@ -255,7 +255,8 @@ static struct kvm_vcpu *vgic_target_oracle(struct vgic_irq *irq)
  * Return negative if "a" sorts before "b", 0 to preserve order, and positive
  * to sort "b" before "a".
  */
-static int vgic_irq_cmp(void *priv, struct list_head *a, struct list_head *b)
+static int vgic_irq_cmp(void *priv, const struct list_head *a,
+			const struct list_head *b)
 {
 	struct vgic_irq *irqa = container_of(a, struct vgic_irq, ap_list);
 	struct vgic_irq *irqb = container_of(b, struct vgic_irq, ap_list);
diff --git a/block/blk-mq-sched.c b/block/blk-mq-sched.c
index e1e997af89a0..3ebd6f10f728 100644
--- a/block/blk-mq-sched.c
+++ b/block/blk-mq-sched.c
@@ -75,7 +75,8 @@ void blk_mq_sched_restart(struct blk_mq_hw_ctx *hctx)
 	blk_mq_run_hw_queue(hctx, true);
 }
 
-static int sched_rq_cmp(void *priv, struct list_head *a, struct list_head *b)
+static int sched_rq_cmp(void *priv, const struct list_head *a,
+			const struct list_head *b)
 {
 	struct request *rqa = container_of(a, struct request, queuelist);
 	struct request *rqb = container_of(b, struct request, queuelist);
diff --git a/block/blk-mq.c b/block/blk-mq.c
index d4d7c1caa439..4e3a70ab5be1 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -1895,7 +1895,8 @@ void blk_mq_insert_requests(struct blk_mq_hw_ctx *hctx, struct blk_mq_ctx *ctx,
 	spin_unlock(&ctx->lock);
 }
 
-static int plug_rq_cmp(void *priv, struct list_head *a, struct list_head *b)
+static int plug_rq_cmp(void *priv, const struct list_head *a,
+		       const struct list_head *b)
 {
 	struct request *rqa = container_of(a, struct request, queuelist);
 	struct request *rqb = container_of(b, struct request, queuelist);
diff --git a/drivers/acpi/nfit/core.c b/drivers/acpi/nfit/core.c
index 8c5dde628405..d15e3ee93b5b 100644
--- a/drivers/acpi/nfit/core.c
+++ b/drivers/acpi/nfit/core.c
@@ -1195,7 +1195,8 @@ static int __nfit_mem_init(struct acpi_nfit_desc *acpi_desc,
 	return 0;
 }
 
-static int nfit_mem_cmp(void *priv, struct list_head *_a, struct list_head *_b)
+static int nfit_mem_cmp(void *priv, const struct list_head *_a,
+		const struct list_head *_b)
 {
 	struct nfit_mem *a = container_of(_a, typeof(*a), list);
 	struct nfit_mem *b = container_of(_b, typeof(*b), list);
diff --git a/drivers/acpi/numa/hmat.c b/drivers/acpi/numa/hmat.c
index cb73a5d6ea76..137a5dd880c2 100644
--- a/drivers/acpi/numa/hmat.c
+++ b/drivers/acpi/numa/hmat.c
@@ -558,7 +558,8 @@ static bool hmat_update_best(u8 type, u32 value, u32 *best)
 	return updated;
 }
 
-static int initiator_cmp(void *priv, struct list_head *a, struct list_head *b)
+static int initiator_cmp(void *priv, const struct list_head *a,
+			 const struct list_head *b)
 {
 	struct memory_initiator *ia;
 	struct memory_initiator *ib;
diff --git a/drivers/clk/keystone/sci-clk.c b/drivers/clk/keystone/sci-clk.c
index aaf31abe1c8f..7e1b136e71ae 100644
--- a/drivers/clk/keystone/sci-clk.c
+++ b/drivers/clk/keystone/sci-clk.c
@@ -503,8 +503,8 @@ static int ti_sci_scan_clocks_from_fw(struct sci_clk_provider *provider)
 
 #else
 
-static int _cmp_sci_clk_list(void *priv, struct list_head *a,
-			     struct list_head *b)
+static int _cmp_sci_clk_list(void *priv, const struct list_head *a,
+			     const struct list_head *b)
 {
 	struct sci_clk *ca = container_of(a, struct sci_clk, node);
 	struct sci_clk *cb = container_of(b, struct sci_clk, node);
diff --git a/drivers/gpu/drm/drm_modes.c b/drivers/gpu/drm/drm_modes.c
index 1ac67d4505e0..6662d0457ad6 100644
--- a/drivers/gpu/drm/drm_modes.c
+++ b/drivers/gpu/drm/drm_modes.c
@@ -1290,7 +1290,8 @@ EXPORT_SYMBOL(drm_mode_prune_invalid);
  * Negative if @lh_a is better than @lh_b, zero if they're equivalent, or
  * positive if @lh_b is better than @lh_a.
  */
-static int drm_mode_compare(void *priv, struct list_head *lh_a, struct list_head *lh_b)
+static int drm_mode_compare(void *priv, const struct list_head *lh_a,
+			    const struct list_head *lh_b)
 {
 	struct drm_display_mode *a = list_entry(lh_a, struct drm_display_mode, head);
 	struct drm_display_mode *b = list_entry(lh_b, struct drm_display_mode, head);
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_user.c b/drivers/gpu/drm/i915/gt/intel_engine_user.c
index 34e6096f196e..da21d2a10cc9 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_user.c
+++ b/drivers/gpu/drm/i915/gt/intel_engine_user.c
@@ -49,7 +49,8 @@ static const u8 uabi_classes[] = {
 	[VIDEO_ENHANCEMENT_CLASS] = I915_ENGINE_CLASS_VIDEO_ENHANCE,
 };
 
-static int engine_cmp(void *priv, struct list_head *A, struct list_head *B)
+static int engine_cmp(void *priv, const struct list_head *A,
+		      const struct list_head *B)
 {
 	const struct intel_engine_cs *a =
 		container_of((struct rb_node *)A, typeof(*a), uabi_node);
diff --git a/drivers/gpu/drm/i915/gvt/debugfs.c b/drivers/gpu/drm/i915/gvt/debugfs.c
index 62e6a14ad58e..9f1c209d9251 100644
--- a/drivers/gpu/drm/i915/gvt/debugfs.c
+++ b/drivers/gpu/drm/i915/gvt/debugfs.c
@@ -41,7 +41,7 @@ struct diff_mmio {
 
 /* Compare two diff_mmio items. */
 static int mmio_offset_compare(void *priv,
-	struct list_head *a, struct list_head *b)
+	const struct list_head *a, const struct list_head *b)
 {
 	struct diff_mmio *ma;
 	struct diff_mmio *mb;
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
index c1adea8765a9..52b9c39e0155 100644
--- a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
@@ -1076,7 +1076,8 @@ static int igt_ppgtt_shrink_boom(void *arg)
 	return exercise_ppgtt(arg, shrink_boom);
 }
 
-static int sort_holes(void *priv, struct list_head *A, struct list_head *B)
+static int sort_holes(void *priv, const struct list_head *A,
+		      const struct list_head *B)
 {
 	struct drm_mm_node *a = list_entry(A, typeof(*a), hole_stack);
 	struct drm_mm_node *b = list_entry(B, typeof(*b), hole_stack);
diff --git a/drivers/gpu/drm/radeon/radeon_cs.c b/drivers/gpu/drm/radeon/radeon_cs.c
index 35e937d39b51..1a5c3db1d53b 100644
--- a/drivers/gpu/drm/radeon/radeon_cs.c
+++ b/drivers/gpu/drm/radeon/radeon_cs.c
@@ -393,8 +393,8 @@ int radeon_cs_parser_init(struct radeon_cs_parser *p, void *data)
 	return 0;
 }
 
-static int cmp_size_smaller_first(void *priv, struct list_head *a,
-				  struct list_head *b)
+static int cmp_size_smaller_first(void *priv, const struct list_head *a,
+				  const struct list_head *b)
 {
 	struct radeon_bo_list *la = list_entry(a, struct radeon_bo_list, tv.head);
 	struct radeon_bo_list *lb = list_entry(b, struct radeon_bo_list, tv.head);
diff --git a/drivers/infiniband/hw/usnic/usnic_uiom_interval_tree.c b/drivers/infiniband/hw/usnic/usnic_uiom_interval_tree.c
index d399523206c7..29d71267af78 100644
--- a/drivers/infiniband/hw/usnic/usnic_uiom_interval_tree.c
+++ b/drivers/infiniband/hw/usnic/usnic_uiom_interval_tree.c
@@ -83,7 +83,8 @@ usnic_uiom_interval_node_alloc(long int start, long int last, int ref_cnt,
 	return interval;
 }
 
-static int interval_cmp(void *priv, struct list_head *a, struct list_head *b)
+static int interval_cmp(void *priv, const struct list_head *a,
+			const struct list_head *b)
 {
 	struct usnic_uiom_interval_node *node_a, *node_b;
 
diff --git a/drivers/interconnect/qcom/bcm-voter.c b/drivers/interconnect/qcom/bcm-voter.c
index 1cc565bce2f4..d1591a28b743 100644
--- a/drivers/interconnect/qcom/bcm-voter.c
+++ b/drivers/interconnect/qcom/bcm-voter.c
@@ -39,7 +39,7 @@ struct bcm_voter {
 	u32 tcs_wait;
 };
 
-static int cmp_vcd(void *priv, struct list_head *a, struct list_head *b)
+static int cmp_vcd(void *priv, const struct list_head *a, const struct list_head *b)
 {
 	const struct qcom_icc_bcm *bcm_a = list_entry(a, struct qcom_icc_bcm, list);
 	const struct qcom_icc_bcm *bcm_b = list_entry(b, struct qcom_icc_bcm, list);
diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index 5d57a5bd171f..841e1c1aa5e6 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -953,7 +953,8 @@ static void dispatch_bio_list(struct bio_list *tmp)
 		submit_bio_noacct(bio);
 }
 
-static int cmp_stripe(void *priv, struct list_head *a, struct list_head *b)
+static int cmp_stripe(void *priv, const struct list_head *a,
+		      const struct list_head *b)
 {
 	const struct r5pending_data *da = list_entry(a,
 				struct r5pending_data, sibling);
diff --git a/drivers/misc/sram.c b/drivers/misc/sram.c
index 6c1a23cb3e8c..202bf951e909 100644
--- a/drivers/misc/sram.c
+++ b/drivers/misc/sram.c
@@ -144,8 +144,8 @@ static void sram_free_partitions(struct sram_dev *sram)
 	}
 }
 
-static int sram_reserve_cmp(void *priv, struct list_head *a,
-					struct list_head *b)
+static int sram_reserve_cmp(void *priv, const struct list_head *a,
+					const struct list_head *b)
 {
 	struct sram_reserve *ra = list_entry(a, struct sram_reserve, list);
 	struct sram_reserve *rb = list_entry(b, struct sram_reserve, list);
diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
index 0896e21642be..5eaaa51a5e30 100644
--- a/drivers/nvme/host/core.c
+++ b/drivers/nvme/host/core.c
@@ -3855,7 +3855,8 @@ static int nvme_init_ns_head(struct nvme_ns *ns, unsigned nsid,
 	return ret;
 }
 
-static int ns_cmp(void *priv, struct list_head *a, struct list_head *b)
+static int ns_cmp(void *priv, const struct list_head *a,
+		const struct list_head *b)
 {
 	struct nvme_ns *nsa = container_of(a, struct nvme_ns, list);
 	struct nvme_ns *nsb = container_of(b, struct nvme_ns, list);
diff --git a/drivers/pci/controller/cadence/pcie-cadence-host.c b/drivers/pci/controller/cadence/pcie-cadence-host.c
index 73dcf8cf98fb..ae1c55503513 100644
--- a/drivers/pci/controller/cadence/pcie-cadence-host.c
+++ b/drivers/pci/controller/cadence/pcie-cadence-host.c
@@ -345,7 +345,8 @@ static int cdns_pcie_host_bar_config(struct cdns_pcie_rc *rc,
 	return 0;
 }
 
-static int cdns_pcie_host_dma_ranges_cmp(void *priv, struct list_head *a, struct list_head *b)
+static int cdns_pcie_host_dma_ranges_cmp(void *priv, const struct list_head *a,
+					 const struct list_head *b)
 {
 	struct resource_entry *entry1, *entry2;
 
diff --git a/drivers/spi/spi-loopback-test.c b/drivers/spi/spi-loopback-test.c
index df981e55c24c..f1cf2232f0b5 100644
--- a/drivers/spi/spi-loopback-test.c
+++ b/drivers/spi/spi-loopback-test.c
@@ -454,7 +454,8 @@ struct rx_ranges {
 	u8 *end;
 };
 
-static int rx_ranges_cmp(void *priv, struct list_head *a, struct list_head *b)
+static int rx_ranges_cmp(void *priv, const struct list_head *a,
+			 const struct list_head *b)
 {
 	struct rx_ranges *rx_a = list_entry(a, struct rx_ranges, list);
 	struct rx_ranges *rx_b = list_entry(b, struct rx_ranges, list);
diff --git a/fs/btrfs/raid56.c b/fs/btrfs/raid56.c
index 8c31357f08ed..f4139de63b2e 100644
--- a/fs/btrfs/raid56.c
+++ b/fs/btrfs/raid56.c
@@ -1634,7 +1634,8 @@ struct btrfs_plug_cb {
 /*
  * rbios on the plug list are sorted for easier merging.
  */
-static int plug_cmp(void *priv, struct list_head *a, struct list_head *b)
+static int plug_cmp(void *priv, const struct list_head *a,
+		    const struct list_head *b)
 {
 	struct btrfs_raid_bio *ra = container_of(a, struct btrfs_raid_bio,
 						 plug_list);
diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
index 92a368627791..00a88bd8105e 100644
--- a/fs/btrfs/tree-log.c
+++ b/fs/btrfs/tree-log.c
@@ -4136,7 +4136,8 @@ static noinline int copy_items(struct btrfs_trans_handle *trans,
 	return ret;
 }
 
-static int extent_cmp(void *priv, struct list_head *a, struct list_head *b)
+static int extent_cmp(void *priv, const struct list_head *a,
+		      const struct list_head *b)
 {
 	struct extent_map *em1, *em2;
 
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index bc3b33efddc5..70dc25c8198e 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -1224,7 +1224,8 @@ static int open_fs_devices(struct btrfs_fs_devices *fs_devices,
 	return 0;
 }
 
-static int devid_cmp(void *priv, struct list_head *a, struct list_head *b)
+static int devid_cmp(void *priv, const struct list_head *a,
+		     const struct list_head *b)
 {
 	struct btrfs_device *dev1, *dev2;
 
diff --git a/fs/ext4/fsmap.c b/fs/ext4/fsmap.c
index 4c2a9fe30067..4493ef0c715e 100644
--- a/fs/ext4/fsmap.c
+++ b/fs/ext4/fsmap.c
@@ -354,8 +354,8 @@ static unsigned int ext4_getfsmap_find_sb(struct super_block *sb,
 
 /* Compare two fsmap items. */
 static int ext4_getfsmap_compare(void *priv,
-				 struct list_head *a,
-				 struct list_head *b)
+				 const struct list_head *a,
+				 const struct list_head *b)
 {
 	struct ext4_fsmap *fa;
 	struct ext4_fsmap *fb;
diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index 9567520d79f7..c06a6cdf05de 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -1732,7 +1732,8 @@ void gfs2_glock_complete(struct gfs2_glock *gl, int ret)
 	spin_unlock(&gl->gl_lockref.lock);
 }
 
-static int glock_cmp(void *priv, struct list_head *a, struct list_head *b)
+static int glock_cmp(void *priv, const struct list_head *a,
+		     const struct list_head *b)
 {
 	struct gfs2_glock *gla, *glb;
 
diff --git a/fs/gfs2/log.c b/fs/gfs2/log.c
index 6410281546f9..88649b43fcff 100644
--- a/fs/gfs2/log.c
+++ b/fs/gfs2/log.c
@@ -695,7 +695,7 @@ void log_flush_wait(struct gfs2_sbd *sdp)
 	}
 }
 
-static int ip_cmp(void *priv, struct list_head *a, struct list_head *b)
+static int ip_cmp(void *priv, const struct list_head *a, const struct list_head *b)
 {
 	struct gfs2_inode *ipa, *ipb;
 
diff --git a/fs/gfs2/lops.c b/fs/gfs2/lops.c
index a82f4747aa8d..b4809967efc6 100644
--- a/fs/gfs2/lops.c
+++ b/fs/gfs2/lops.c
@@ -634,7 +634,8 @@ static void gfs2_check_magic(struct buffer_head *bh)
 	kunmap_atomic(kaddr);
 }
 
-static int blocknr_cmp(void *priv, struct list_head *a, struct list_head *b)
+static int blocknr_cmp(void *priv, const struct list_head *a,
+		       const struct list_head *b)
 {
 	struct gfs2_bufdata *bda, *bdb;
 
diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c
index 414769a6ad11..0129e6bab985 100644
--- a/fs/iomap/buffered-io.c
+++ b/fs/iomap/buffered-io.c
@@ -1155,7 +1155,8 @@ iomap_ioend_try_merge(struct iomap_ioend *ioend, struct list_head *more_ioends,
 EXPORT_SYMBOL_GPL(iomap_ioend_try_merge);
 
 static int
-iomap_ioend_compare(void *priv, struct list_head *a, struct list_head *b)
+iomap_ioend_compare(void *priv, const struct list_head *a,
+		const struct list_head *b)
 {
 	struct iomap_ioend *ia = container_of(a, struct iomap_ioend, io_list);
 	struct iomap_ioend *ib = container_of(b, struct iomap_ioend, io_list);
diff --git a/fs/ubifs/gc.c b/fs/ubifs/gc.c
index a4aaeea63893..dc3e26e9ed7b 100644
--- a/fs/ubifs/gc.c
+++ b/fs/ubifs/gc.c
@@ -102,7 +102,8 @@ static int switch_gc_head(struct ubifs_info *c)
  * This function compares data nodes @a and @b. Returns %1 if @a has greater
  * inode or block number, and %-1 otherwise.
  */
-static int data_nodes_cmp(void *priv, struct list_head *a, struct list_head *b)
+static int data_nodes_cmp(void *priv, const struct list_head *a,
+			  const struct list_head *b)
 {
 	ino_t inuma, inumb;
 	struct ubifs_info *c = priv;
@@ -145,8 +146,8 @@ static int data_nodes_cmp(void *priv, struct list_head *a, struct list_head *b)
  * first and sorted by length in descending order. Directory entry nodes go
  * after inode nodes and are sorted in ascending hash valuer order.
  */
-static int nondata_nodes_cmp(void *priv, struct list_head *a,
-			     struct list_head *b)
+static int nondata_nodes_cmp(void *priv, const struct list_head *a,
+			     const struct list_head *b)
 {
 	ino_t inuma, inumb;
 	struct ubifs_info *c = priv;
diff --git a/fs/ubifs/replay.c b/fs/ubifs/replay.c
index 0f8a6a16421b..4d17e5382b74 100644
--- a/fs/ubifs/replay.c
+++ b/fs/ubifs/replay.c
@@ -298,8 +298,8 @@ static int apply_replay_entry(struct ubifs_info *c, struct replay_entry *r)
  * entries @a and @b by comparing their sequence numer.  Returns %1 if @a has
  * greater sequence number and %-1 otherwise.
  */
-static int replay_entries_cmp(void *priv, struct list_head *a,
-			      struct list_head *b)
+static int replay_entries_cmp(void *priv, const struct list_head *a,
+			      const struct list_head *b)
 {
 	struct ubifs_info *c = priv;
 	struct replay_entry *ra, *rb;
diff --git a/fs/xfs/scrub/bitmap.c b/fs/xfs/scrub/bitmap.c
index f88694f22d05..813b5f219113 100644
--- a/fs/xfs/scrub/bitmap.c
+++ b/fs/xfs/scrub/bitmap.c
@@ -63,8 +63,8 @@ xbitmap_init(
 static int
 xbitmap_range_cmp(
 	void			*priv,
-	struct list_head	*a,
-	struct list_head	*b)
+	const struct list_head	*a,
+	const struct list_head	*b)
 {
 	struct xbitmap_range	*ap;
 	struct xbitmap_range	*bp;
diff --git a/fs/xfs/xfs_bmap_item.c b/fs/xfs/xfs_bmap_item.c
index 2344757ede63..e3a691937e92 100644
--- a/fs/xfs/xfs_bmap_item.c
+++ b/fs/xfs/xfs_bmap_item.c
@@ -265,8 +265,8 @@ xfs_trans_log_finish_bmap_update(
 static int
 xfs_bmap_update_diff_items(
 	void				*priv,
-	struct list_head		*a,
-	struct list_head		*b)
+	const struct list_head		*a,
+	const struct list_head		*b)
 {
 	struct xfs_bmap_intent		*ba;
 	struct xfs_bmap_intent		*bb;
diff --git a/fs/xfs/xfs_buf.c b/fs/xfs/xfs_buf.c
index 37a1d12762d8..592800c8852f 100644
--- a/fs/xfs/xfs_buf.c
+++ b/fs/xfs/xfs_buf.c
@@ -2124,9 +2124,9 @@ xfs_buf_delwri_queue(
  */
 static int
 xfs_buf_cmp(
-	void		*priv,
-	struct list_head *a,
-	struct list_head *b)
+	void			*priv,
+	const struct list_head	*a,
+	const struct list_head	*b)
 {
 	struct xfs_buf	*ap = container_of(a, struct xfs_buf, b_list);
 	struct xfs_buf	*bp = container_of(b, struct xfs_buf, b_list);
diff --git a/fs/xfs/xfs_extent_busy.c b/fs/xfs/xfs_extent_busy.c
index ef17c1f6db32..a4075685d9eb 100644
--- a/fs/xfs/xfs_extent_busy.c
+++ b/fs/xfs/xfs_extent_busy.c
@@ -629,8 +629,8 @@ xfs_extent_busy_wait_all(
 int
 xfs_extent_busy_ag_cmp(
 	void			*priv,
-	struct list_head	*l1,
-	struct list_head	*l2)
+	const struct list_head	*l1,
+	const struct list_head	*l2)
 {
 	struct xfs_extent_busy	*b1 =
 		container_of(l1, struct xfs_extent_busy, list);
diff --git a/fs/xfs/xfs_extent_busy.h b/fs/xfs/xfs_extent_busy.h
index 990ab3891971..8aea07100092 100644
--- a/fs/xfs/xfs_extent_busy.h
+++ b/fs/xfs/xfs_extent_busy.h
@@ -58,7 +58,8 @@ void
 xfs_extent_busy_wait_all(struct xfs_mount *mp);
 
 int
-xfs_extent_busy_ag_cmp(void *priv, struct list_head *a, struct list_head *b);
+xfs_extent_busy_ag_cmp(void *priv, const struct list_head *a,
+	const struct list_head *b);
 
 static inline void xfs_extent_busy_sort(struct list_head *list)
 {
diff --git a/fs/xfs/xfs_extfree_item.c b/fs/xfs/xfs_extfree_item.c
index 93223ebb3372..2424230ca2c3 100644
--- a/fs/xfs/xfs_extfree_item.c
+++ b/fs/xfs/xfs_extfree_item.c
@@ -397,8 +397,8 @@ xfs_trans_free_extent(
 static int
 xfs_extent_free_diff_items(
 	void				*priv,
-	struct list_head		*a,
-	struct list_head		*b)
+	const struct list_head		*a,
+	const struct list_head		*b)
 {
 	struct xfs_mount		*mp = priv;
 	struct xfs_extent_free_item	*ra;
diff --git a/fs/xfs/xfs_refcount_item.c b/fs/xfs/xfs_refcount_item.c
index 07ebccbbf4df..746f4eda724c 100644
--- a/fs/xfs/xfs_refcount_item.c
+++ b/fs/xfs/xfs_refcount_item.c
@@ -269,8 +269,8 @@ xfs_trans_log_finish_refcount_update(
 static int
 xfs_refcount_update_diff_items(
 	void				*priv,
-	struct list_head		*a,
-	struct list_head		*b)
+	const struct list_head		*a,
+	const struct list_head		*b)
 {
 	struct xfs_mount		*mp = priv;
 	struct xfs_refcount_intent	*ra;
diff --git a/fs/xfs/xfs_rmap_item.c b/fs/xfs/xfs_rmap_item.c
index 49cebd68b672..dc4f0c9f0897 100644
--- a/fs/xfs/xfs_rmap_item.c
+++ b/fs/xfs/xfs_rmap_item.c
@@ -337,8 +337,8 @@ xfs_trans_log_finish_rmap_update(
 static int
 xfs_rmap_update_diff_items(
 	void				*priv,
-	struct list_head		*a,
-	struct list_head		*b)
+	const struct list_head		*a,
+	const struct list_head		*b)
 {
 	struct xfs_mount		*mp = priv;
 	struct xfs_rmap_intent		*ra;
diff --git a/include/linux/list_sort.h b/include/linux/list_sort.h
index 20f178c24e9d..453105f74e05 100644
--- a/include/linux/list_sort.h
+++ b/include/linux/list_sort.h
@@ -6,8 +6,9 @@
 
 struct list_head;
 
+typedef int __attribute__((nonnull(2,3))) (*list_cmp_func_t)(void *,
+		const struct list_head *, const struct list_head *);
+
 __attribute__((nonnull(2,3)))
-void list_sort(void *priv, struct list_head *head,
-	       int (*cmp)(void *priv, struct list_head *a,
-			  struct list_head *b));
+void list_sort(void *priv, struct list_head *head, list_cmp_func_t cmp);
 #endif
diff --git a/lib/list_sort.c b/lib/list_sort.c
index 52f0c258c895..a926d96ffd44 100644
--- a/lib/list_sort.c
+++ b/lib/list_sort.c
@@ -7,16 +7,13 @@
 #include <linux/list_sort.h>
 #include <linux/list.h>
 
-typedef int __attribute__((nonnull(2,3))) (*cmp_func)(void *,
-		struct list_head const *, struct list_head const *);
-
 /*
  * Returns a list organized in an intermediate format suited
  * to chaining of merge() calls: null-terminated, no reserved or
  * sentinel head node, "prev" links not maintained.
  */
 __attribute__((nonnull(2,3,4)))
-static struct list_head *merge(void *priv, cmp_func cmp,
+static struct list_head *merge(void *priv, list_cmp_func_t cmp,
 				struct list_head *a, struct list_head *b)
 {
 	struct list_head *head, **tail = &head;
@@ -52,7 +49,7 @@ static struct list_head *merge(void *priv, cmp_func cmp,
  * throughout.
  */
 __attribute__((nonnull(2,3,4,5)))
-static void merge_final(void *priv, cmp_func cmp, struct list_head *head,
+static void merge_final(void *priv, list_cmp_func_t cmp, struct list_head *head,
 			struct list_head *a, struct list_head *b)
 {
 	struct list_head *tail = head;
@@ -185,9 +182,7 @@ static void merge_final(void *priv, cmp_func cmp, struct list_head *head,
  * 2^(k+1) - 1 (second merge of case 5 when x == 2^(k-1) - 1).
  */
 __attribute__((nonnull(2,3)))
-void list_sort(void *priv, struct list_head *head,
-		int (*cmp)(void *priv, struct list_head *a,
-			struct list_head *b))
+void list_sort(void *priv, struct list_head *head, list_cmp_func_t cmp)
 {
 	struct list_head *list = head->next, *pending = NULL;
 	size_t count = 0;	/* Count of pending */
@@ -227,7 +222,7 @@ void list_sort(void *priv, struct list_head *head,
 		if (likely(bits)) {
 			struct list_head *a = *tail, *b = a->prev;
 
-			a = merge(priv, (cmp_func)cmp, b, a);
+			a = merge(priv, cmp, b, a);
 			/* Install the merged result in place of the inputs */
 			a->prev = b->prev;
 			*tail = a;
@@ -249,10 +244,10 @@ void list_sort(void *priv, struct list_head *head,
 
 		if (!next)
 			break;
-		list = merge(priv, (cmp_func)cmp, pending, list);
+		list = merge(priv, cmp, pending, list);
 		pending = next;
 	}
 	/* The final merge, rebuilding prev links */
-	merge_final(priv, (cmp_func)cmp, head, pending, list);
+	merge_final(priv, cmp, head, pending, list);
 }
 EXPORT_SYMBOL(list_sort);
diff --git a/lib/test_list_sort.c b/lib/test_list_sort.c
index 1f017d3b610e..00daaf23316f 100644
--- a/lib/test_list_sort.c
+++ b/lib/test_list_sort.c
@@ -56,7 +56,8 @@ static int __init check(struct debug_el *ela, struct debug_el *elb)
 	return 0;
 }
 
-static int __init cmp(void *priv, struct list_head *a, struct list_head *b)
+static int __init cmp(void *priv, const struct list_head *a,
+		      const struct list_head *b)
 {
 	struct debug_el *ela, *elb;
 
diff --git a/net/tipc/name_table.c b/net/tipc/name_table.c
index ee5ac40ea2b6..f8141443f2e2 100644
--- a/net/tipc/name_table.c
+++ b/net/tipc/name_table.c
@@ -397,8 +397,8 @@ static struct publication *tipc_service_remove_publ(struct service_range *sr,
  * Code reused: time_after32() for the same purpose
  */
 #define publication_after(pa, pb) time_after32((pa)->id, (pb)->id)
-static int tipc_publ_sort(void *priv, struct list_head *a,
-			  struct list_head *b)
+static int tipc_publ_sort(void *priv, const struct list_head *a,
+			  const struct list_head *b)
 {
 	struct publication *pa, *pb;
 
-- 
2.31.0.291.g576ba9dcdaf-goog

Re: [PATCH v3 09/17] treewide: Change list_sort to use const pointers

[Nick Desaulniers]

On Tue, Mar 23, 2021 at 1:40 PM Sami Tolvanen <samitolvanen@google.com> wrote:
>
> list_sort() internally casts the comparison function passed to it
> to a different type with constant struct list_head pointers, and
> uses this pointer to call the functions, which trips indirect call
> Control-Flow Integrity (CFI) checking.
>
> Instead of removing the consts, this change defines the
> list_cmp_func_t type and changes the comparison function types of
> all list_sort() callers to use const pointers, thus avoiding type
> mismatches.
>
> Suggested-by: Nick Desaulniers <ndesaulniers@google.com>
> Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
> ---
>  arch/arm64/kvm/vgic/vgic-its.c                  |  8 ++++----
>  arch/arm64/kvm/vgic/vgic.c                      |  3 ++-
>  block/blk-mq-sched.c                            |  3 ++-
>  block/blk-mq.c                                  |  3 ++-
>  drivers/acpi/nfit/core.c                        |  3 ++-
>  drivers/acpi/numa/hmat.c                        |  3 ++-
>  drivers/clk/keystone/sci-clk.c                  |  4 ++--
>  drivers/gpu/drm/drm_modes.c                     |  3 ++-
>  drivers/gpu/drm/i915/gt/intel_engine_user.c     |  3 ++-
>  drivers/gpu/drm/i915/gvt/debugfs.c              |  2 +-
>  drivers/gpu/drm/i915/selftests/i915_gem_gtt.c   |  3 ++-
>  drivers/gpu/drm/radeon/radeon_cs.c              |  4 ++--
>  .../hw/usnic/usnic_uiom_interval_tree.c         |  3 ++-
>  drivers/interconnect/qcom/bcm-voter.c           |  2 +-
>  drivers/md/raid5.c                              |  3 ++-
>  drivers/misc/sram.c                             |  4 ++--
>  drivers/nvme/host/core.c                        |  3 ++-
>  .../pci/controller/cadence/pcie-cadence-host.c  |  3 ++-
>  drivers/spi/spi-loopback-test.c                 |  3 ++-
>  fs/btrfs/raid56.c                               |  3 ++-
>  fs/btrfs/tree-log.c                             |  3 ++-
>  fs/btrfs/volumes.c                              |  3 ++-
>  fs/ext4/fsmap.c                                 |  4 ++--
>  fs/gfs2/glock.c                                 |  3 ++-
>  fs/gfs2/log.c                                   |  2 +-
>  fs/gfs2/lops.c                                  |  3 ++-
>  fs/iomap/buffered-io.c                          |  3 ++-
>  fs/ubifs/gc.c                                   |  7 ++++---
>  fs/ubifs/replay.c                               |  4 ++--
>  fs/xfs/scrub/bitmap.c                           |  4 ++--
>  fs/xfs/xfs_bmap_item.c                          |  4 ++--
>  fs/xfs/xfs_buf.c                                |  6 +++---
>  fs/xfs/xfs_extent_busy.c                        |  4 ++--
>  fs/xfs/xfs_extent_busy.h                        |  3 ++-
>  fs/xfs/xfs_extfree_item.c                       |  4 ++--
>  fs/xfs/xfs_refcount_item.c                      |  4 ++--
>  fs/xfs/xfs_rmap_item.c                          |  4 ++--
>  include/linux/list_sort.h                       |  7 ++++---
>  lib/list_sort.c                                 | 17 ++++++-----------
>  lib/test_list_sort.c                            |  3 ++-
>  net/tipc/name_table.c                           |  4 ++--
>  41 files changed, 90 insertions(+), 72 deletions(-)

This looks like all of the call sites I could find.  Yes, this looks
much better, thank you for taking the time to fix all of these. I ran
this through some build tests of ARCH=arm64 defconfig, allyesconfig,
and same for my host (x86_64). All LGTM.

Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Tested-by: Nick Desaulniers <ndesaulniers@google.com>
-- 
Thanks,
~Nick Desaulniers

Re: [PATCH v3 09/17] treewide: Change list_sort to use const pointers

[Christoph Hellwig]

On Tue, Mar 23, 2021 at 01:39:38PM -0700, Sami Tolvanen wrote:
> list_sort() internally casts the comparison function passed to it
> to a different type with constant struct list_head pointers, and
> uses this pointer to call the functions, which trips indirect call
> Control-Flow Integrity (CFI) checking.
> 
> Instead of removing the consts, this change defines the
> list_cmp_func_t type and changes the comparison function types of
> all list_sort() callers to use const pointers, thus avoiding type
> mismatches.
> 
> Suggested-by: Nick Desaulniers <ndesaulniers@google.com>
> Signed-off-by: Sami Tolvanen <samitolvanen@google.com>

Looks good:
Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [PATCH v3 09/17] treewide: Change list_sort to use const pointers

[Kees Cook]

On Tue, Mar 23, 2021 at 01:39:38PM -0700, Sami Tolvanen wrote:
> list_sort() internally casts the comparison function passed to it
> to a different type with constant struct list_head pointers, and
> uses this pointer to call the functions, which trips indirect call
> Control-Flow Integrity (CFI) checking.
> 
> Instead of removing the consts, this change defines the
> list_cmp_func_t type and changes the comparison function types of
> all list_sort() callers to use const pointers, thus avoiding type
> mismatches.
> 
> Suggested-by: Nick Desaulniers <ndesaulniers@google.com>
> Signed-off-by: Sami Tolvanen <samitolvanen@google.com>

Reviewed-by: Kees Cook <keescook@chromium.org>

-- 
Kees Cook

[PATCH v3 10/17] lkdtm: use __va_function

[Sami Tolvanen]

To ensure we take the actual address of a function in kernel text, use
__va_function. Otherwise, with CONFIG_CFI_CLANG, the compiler replaces
the address with a pointer to the CFI jump table, which is actually in
the module when compiled with CONFIG_LKDTM=m.

Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
Acked-by: Kees Cook <keescook@chromium.org>
---
 drivers/misc/lkdtm/usercopy.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/misc/lkdtm/usercopy.c b/drivers/misc/lkdtm/usercopy.c
index 109e8d4302c1..d173d6175c87 100644
--- a/drivers/misc/lkdtm/usercopy.c
+++ b/drivers/misc/lkdtm/usercopy.c
@@ -314,7 +314,7 @@ void lkdtm_USERCOPY_KERNEL(void)
 
 	pr_info("attempting bad copy_to_user from kernel text: %px\n",
 		vm_mmap);
-	if (copy_to_user((void __user *)user_addr, vm_mmap,
+	if (copy_to_user((void __user *)user_addr, __va_function(vm_mmap),
 			 unconst + PAGE_SIZE)) {
 		pr_warn("copy_to_user failed, but lacked Oops\n");
 		goto free_user;
-- 
2.31.0.291.g576ba9dcdaf-goog

[PATCH v3 11/17] psci: use __pa_function for cpu_resume

[Sami Tolvanen]

With CONFIG_CFI_CLANG, the compiler replaces function pointers with
jump table addresses, which results in __pa_symbol returning the
physical address of the jump table entry. As the jump table contains
an immediate jump to an EL1 virtual address, this typically won't
work as intended. Use __pa_function instead to get the address to
cpu_resume.

Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
---
 drivers/firmware/psci/psci.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/firmware/psci/psci.c b/drivers/firmware/psci/psci.c
index f5fc429cae3f..facd3cce3244 100644
--- a/drivers/firmware/psci/psci.c
+++ b/drivers/firmware/psci/psci.c
@@ -326,7 +326,7 @@ static int psci_suspend_finisher(unsigned long state)
 {
 	u32 power_state = state;
 
-	return psci_ops.cpu_suspend(power_state, __pa_symbol(cpu_resume));
+	return psci_ops.cpu_suspend(power_state, __pa_function(cpu_resume));
 }
 
 int psci_cpu_suspend_enter(u32 state)
@@ -345,7 +345,7 @@ int psci_cpu_suspend_enter(u32 state)
 static int psci_system_suspend(unsigned long unused)
 {
 	return invoke_psci_fn(PSCI_FN_NATIVE(1_0, SYSTEM_SUSPEND),
-			      __pa_symbol(cpu_resume), 0, 0);
+			      __pa_function(cpu_resume), 0, 0);
 }
 
 static int psci_system_suspend_enter(suspend_state_t state)
-- 
2.31.0.291.g576ba9dcdaf-goog

Re: [PATCH v3 11/17] psci: use __pa_function for cpu_resume

[Mark Rutland]

On Tue, Mar 23, 2021 at 01:39:40PM -0700, Sami Tolvanen wrote:
> With CONFIG_CFI_CLANG, the compiler replaces function pointers with
> jump table addresses, which results in __pa_symbol returning the
> physical address of the jump table entry. As the jump table contains
> an immediate jump to an EL1 virtual address, this typically won't
> work as intended. Use __pa_function instead to get the address to
> cpu_resume.
> 
> Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
> Reviewed-by: Kees Cook <keescook@chromium.org>
> ---
>  drivers/firmware/psci/psci.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/firmware/psci/psci.c b/drivers/firmware/psci/psci.c
> index f5fc429cae3f..facd3cce3244 100644
> --- a/drivers/firmware/psci/psci.c
> +++ b/drivers/firmware/psci/psci.c
> @@ -326,7 +326,7 @@ static int psci_suspend_finisher(unsigned long state)
>  {
>  	u32 power_state = state;
>  
> -	return psci_ops.cpu_suspend(power_state, __pa_symbol(cpu_resume));
> +	return psci_ops.cpu_suspend(power_state, __pa_function(cpu_resume));

As mentioned on the patch adding __pa_function(), I'd prefer to keep the
whatever->phys conversion separate from the CFI removal, since we have a
number of distinct virtual address ranges with differing conversions to
phys, and I don't think it's clear that __pa_function() only works for
kernel symbols (rather than module addresses, linear map addresses,
etc).

Other than that, I'm happy in principle with wrapping this. I suspect
that for clarity we should add an intermediate variable, e.g.

| phys_addr_t pa_cpu_resume = pa_symbol(function_nocfi(cpu_resume));
| return psci_ops.cpu_suspend(power_state, pa_cpu_resume)

Thanks,
Mark.

>  }
>  
>  int psci_cpu_suspend_enter(u32 state)
> @@ -345,7 +345,7 @@ int psci_cpu_suspend_enter(u32 state)
>  static int psci_system_suspend(unsigned long unused)
>  {
>  	return invoke_psci_fn(PSCI_FN_NATIVE(1_0, SYSTEM_SUSPEND),
> -			      __pa_symbol(cpu_resume), 0, 0);
> +			      __pa_function(cpu_resume), 0, 0);
>  }
>  
>  static int psci_system_suspend_enter(suspend_state_t state)
> -- 
> 2.31.0.291.g576ba9dcdaf-goog
> 

[PATCH v3 12/17] arm64: implement __va_function

[Sami Tolvanen]

With CONFIG_CFI_CLANG, the compiler replaces function addresses in
instrumented C code with jump table addresses. This change implements
the __va_function() macro, which returns the actual function address
instead.

Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
---
 arch/arm64/include/asm/memory.h | 15 +++++++++++++++
 1 file changed, 15 insertions(+)

diff --git a/arch/arm64/include/asm/memory.h b/arch/arm64/include/asm/memory.h
index 0aabc3be9a75..9a4887808681 100644
--- a/arch/arm64/include/asm/memory.h
+++ b/arch/arm64/include/asm/memory.h
@@ -321,6 +321,21 @@ static inline void *phys_to_virt(phys_addr_t x)
 #define virt_to_pfn(x)		__phys_to_pfn(__virt_to_phys((unsigned long)(x)))
 #define sym_to_pfn(x)		__phys_to_pfn(__pa_symbol(x))
 
+#ifdef CONFIG_CFI_CLANG
+/*
+ * With CONFIG_CFI_CLANG, the compiler replaces function address
+ * references with the address of the function's CFI jump table
+ * entry. The __va_function macro always returns the address of the
+ * actual function instead.
+ */
+#define __va_function(x) ({						\
+	void *addr;							\
+	asm("adrp %0, " __stringify(x) "\n\t"				\
+	    "add  %0, %0, :lo12:" __stringify(x) : "=r" (addr));	\
+	addr;								\
+})
+#endif
+
 /*
  *  virt_to_page(x)	convert a _valid_ virtual address to struct page *
  *  virt_addr_valid(x)	indicates whether a virtual address is valid
-- 
2.31.0.291.g576ba9dcdaf-goog

Re: [PATCH v3 12/17] arm64: implement __va_function

[Mark Rutland]

On Tue, Mar 23, 2021 at 01:39:41PM -0700, Sami Tolvanen wrote:
> With CONFIG_CFI_CLANG, the compiler replaces function addresses in
> instrumented C code with jump table addresses. This change implements
> the __va_function() macro, which returns the actual function address
> instead.
> 
> Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
> Reviewed-by: Kees Cook <keescook@chromium.org>

Is there really no attribute or builtin that can be used to do this
without assembly?

IIUC from other patches the symbol tables will contain the "real"
non-cfi entry points (unless we explciitly asked to make the jump table
address canonical), so AFAICT here the compiler should have all the
necessary information to generate either the CFI or non-CFI entry point
addresses, even if it doesn't expose an interface for that today.

It'd be a lot nicer if we could get the compiler to do this for us.

Thanks,
Mark.

> ---
>  arch/arm64/include/asm/memory.h | 15 +++++++++++++++
>  1 file changed, 15 insertions(+)
> 
> diff --git a/arch/arm64/include/asm/memory.h b/arch/arm64/include/asm/memory.h
> index 0aabc3be9a75..9a4887808681 100644
> --- a/arch/arm64/include/asm/memory.h
> +++ b/arch/arm64/include/asm/memory.h
> @@ -321,6 +321,21 @@ static inline void *phys_to_virt(phys_addr_t x)
>  #define virt_to_pfn(x)		__phys_to_pfn(__virt_to_phys((unsigned long)(x)))
>  #define sym_to_pfn(x)		__phys_to_pfn(__pa_symbol(x))
>  
> +#ifdef CONFIG_CFI_CLANG
> +/*
> + * With CONFIG_CFI_CLANG, the compiler replaces function address
> + * references with the address of the function's CFI jump table
> + * entry. The __va_function macro always returns the address of the
> + * actual function instead.
> + */
> +#define __va_function(x) ({						\
> +	void *addr;							\
> +	asm("adrp %0, " __stringify(x) "\n\t"				\
> +	    "add  %0, %0, :lo12:" __stringify(x) : "=r" (addr));	\
> +	addr;								\
> +})
> +#endif
> +
>  /*
>   *  virt_to_page(x)	convert a _valid_ virtual address to struct page *
>   *  virt_addr_valid(x)	indicates whether a virtual address is valid
> -- 
> 2.31.0.291.g576ba9dcdaf-goog
> 

Re: [PATCH v3 12/17] arm64: implement __va_function

[Sami Tolvanen]

On Thu, Mar 25, 2021 at 3:38 AM Mark Rutland <mark.rutland@arm.com> wrote:
>
> On Tue, Mar 23, 2021 at 01:39:41PM -0700, Sami Tolvanen wrote:
> > With CONFIG_CFI_CLANG, the compiler replaces function addresses in
> > instrumented C code with jump table addresses. This change implements
> > the __va_function() macro, which returns the actual function address
> > instead.
> >
> > Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
> > Reviewed-by: Kees Cook <keescook@chromium.org>
>
> Is there really no attribute or builtin that can be used to do this
> without assembly?

I don't think the compiler currently offers anything that could help
us here. Peter, can you think of another way to avoid the function
address to jump table address conversion with
-fno-sanitize-cfi-canonical-jump-tables?

> IIUC from other patches the symbol tables will contain the "real"
> non-cfi entry points (unless we explciitly asked to make the jump table
> address canonical), so AFAICT here the compiler should have all the
> necessary information to generate either the CFI or non-CFI entry point
> addresses, even if it doesn't expose an interface for that today.
>
> It'd be a lot nicer if we could get the compiler to do this for us.

I agree, that would be quite useful in the kernel.

Sami

Re: [PATCH v3 12/17] arm64: implement __va_function

[Peter Collingbourne]

On Thu, Mar 25, 2021 at 4:28 PM Sami Tolvanen <samitolvanen@google.com> wrote:
>
> On Thu, Mar 25, 2021 at 3:38 AM Mark Rutland <mark.rutland@arm.com> wrote:
> >
> > On Tue, Mar 23, 2021 at 01:39:41PM -0700, Sami Tolvanen wrote:
> > > With CONFIG_CFI_CLANG, the compiler replaces function addresses in
> > > instrumented C code with jump table addresses. This change implements
> > > the __va_function() macro, which returns the actual function address
> > > instead.
> > >
> > > Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
> > > Reviewed-by: Kees Cook <keescook@chromium.org>
> >
> > Is there really no attribute or builtin that can be used to do this
> > without assembly?
>
> I don't think the compiler currently offers anything that could help
> us here. Peter, can you think of another way to avoid the function
> address to jump table address conversion with
> -fno-sanitize-cfi-canonical-jump-tables?

No, the assembly seems like the best way at the moment.

> > IIUC from other patches the symbol tables will contain the "real"
> > non-cfi entry points (unless we explciitly asked to make the jump table
> > address canonical), so AFAICT here the compiler should have all the
> > necessary information to generate either the CFI or non-CFI entry point
> > addresses, even if it doesn't expose an interface for that today.
> >
> > It'd be a lot nicer if we could get the compiler to do this for us.
>
> I agree, that would be quite useful in the kernel.

Maybe. The kernel's requirements seem quite specialized here though. A
normal C or C++ program has little need for the actual entry point, so
if you need it you are probably doing something low level enough to
require assembly anyway.

Peter

[PATCH v3 13/17] arm64: use __pa_function

[Sami Tolvanen]

With CONFIG_CFI_CLANG, the compiler replaces function address
references with the address of the function's CFI jump table
entry. This means that __pa_symbol(function) returns the physical
address of the jump table entry, which can lead to address space
confusion as the jump table points to the function's virtual
address. Therefore, use the __pa_function() macro to ensure we are
always taking the address of the actual function instead.

Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
---
 arch/arm64/include/asm/mmu_context.h      | 2 +-
 arch/arm64/kernel/acpi_parking_protocol.c | 2 +-
 arch/arm64/kernel/cpu-reset.h             | 2 +-
 arch/arm64/kernel/cpufeature.c            | 2 +-
 arch/arm64/kernel/psci.c                  | 3 ++-
 arch/arm64/kernel/smp_spin_table.c        | 2 +-
 6 files changed, 7 insertions(+), 6 deletions(-)

diff --git a/arch/arm64/include/asm/mmu_context.h b/arch/arm64/include/asm/mmu_context.h
index bd02e99b1a4c..16cc9a694bb2 100644
--- a/arch/arm64/include/asm/mmu_context.h
+++ b/arch/arm64/include/asm/mmu_context.h
@@ -140,7 +140,7 @@ static inline void cpu_replace_ttbr1(pgd_t *pgdp)
 		ttbr1 |= TTBR_CNP_BIT;
 	}
 
-	replace_phys = (void *)__pa_symbol(idmap_cpu_replace_ttbr1);
+	replace_phys = (void *)__pa_function(idmap_cpu_replace_ttbr1);
 
 	cpu_install_idmap();
 	replace_phys(ttbr1);
diff --git a/arch/arm64/kernel/acpi_parking_protocol.c b/arch/arm64/kernel/acpi_parking_protocol.c
index e7c941d8340d..e7f3af6043c5 100644
--- a/arch/arm64/kernel/acpi_parking_protocol.c
+++ b/arch/arm64/kernel/acpi_parking_protocol.c
@@ -99,7 +99,7 @@ static int acpi_parking_protocol_cpu_boot(unsigned int cpu)
 	 * that read this address need to convert this address to the
 	 * Boot-Loader's endianness before jumping.
 	 */
-	writeq_relaxed(__pa_symbol(secondary_entry), &mailbox->entry_point);
+	writeq_relaxed(__pa_function(secondary_entry), &mailbox->entry_point);
 	writel_relaxed(cpu_entry->gic_cpu_id, &mailbox->cpu_id);
 
 	arch_send_wakeup_ipi_mask(cpumask_of(cpu));
diff --git a/arch/arm64/kernel/cpu-reset.h b/arch/arm64/kernel/cpu-reset.h
index ed50e9587ad8..dfba8cf921e5 100644
--- a/arch/arm64/kernel/cpu-reset.h
+++ b/arch/arm64/kernel/cpu-reset.h
@@ -22,7 +22,7 @@ static inline void __noreturn cpu_soft_restart(unsigned long entry,
 
 	unsigned long el2_switch = !is_kernel_in_hyp_mode() &&
 		is_hyp_mode_available();
-	restart = (void *)__pa_symbol(__cpu_soft_restart);
+	restart = (void *)__pa_function(__cpu_soft_restart);
 
 	cpu_install_idmap();
 	restart(el2_switch, entry, arg0, arg1, arg2);
diff --git a/arch/arm64/kernel/cpufeature.c b/arch/arm64/kernel/cpufeature.c
index 066030717a4c..7ec1c2ccdc0b 100644
--- a/arch/arm64/kernel/cpufeature.c
+++ b/arch/arm64/kernel/cpufeature.c
@@ -1460,7 +1460,7 @@ kpti_install_ng_mappings(const struct arm64_cpu_capabilities *__unused)
 	if (arm64_use_ng_mappings)
 		return;
 
-	remap_fn = (void *)__pa_symbol(idmap_kpti_install_ng_mappings);
+	remap_fn = (void *)__pa_function(idmap_kpti_install_ng_mappings);
 
 	cpu_install_idmap();
 	remap_fn(cpu, num_online_cpus(), __pa_symbol(swapper_pg_dir));
diff --git a/arch/arm64/kernel/psci.c b/arch/arm64/kernel/psci.c
index 62d2bda7adb8..bfb1a6f8282d 100644
--- a/arch/arm64/kernel/psci.c
+++ b/arch/arm64/kernel/psci.c
@@ -38,7 +38,8 @@ static int __init cpu_psci_cpu_prepare(unsigned int cpu)
 
 static int cpu_psci_cpu_boot(unsigned int cpu)
 {
-	int err = psci_ops.cpu_on(cpu_logical_map(cpu), __pa_symbol(secondary_entry));
+	int err = psci_ops.cpu_on(cpu_logical_map(cpu),
+				  __pa_function(secondary_entry));
 	if (err)
 		pr_err("failed to boot CPU%d (%d)\n", cpu, err);
 
diff --git a/arch/arm64/kernel/smp_spin_table.c b/arch/arm64/kernel/smp_spin_table.c
index 056772c26098..a80ff9092e86 100644
--- a/arch/arm64/kernel/smp_spin_table.c
+++ b/arch/arm64/kernel/smp_spin_table.c
@@ -88,7 +88,7 @@ static int smp_spin_table_cpu_prepare(unsigned int cpu)
 	 * boot-loader's endianness before jumping. This is mandated by
 	 * the boot protocol.
 	 */
-	writeq_relaxed(__pa_symbol(secondary_holding_pen), release_addr);
+	writeq_relaxed(__pa_function(secondary_holding_pen), release_addr);
 	__flush_dcache_area((__force void *)release_addr,
 			    sizeof(*release_addr));
 
-- 
2.31.0.291.g576ba9dcdaf-goog

[PATCH v3 14/17] arm64: add __nocfi to functions that jump to a physical address

[Sami Tolvanen]

Disable CFI checking for functions that switch to linear mapping and
make an indirect call to a physical address, since the compiler only
understands virtual addresses and the CFI check for such indirect calls
would always fail.

Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
---
 arch/arm64/include/asm/mmu_context.h | 2 +-
 arch/arm64/kernel/cpu-reset.h        | 8 ++++----
 arch/arm64/kernel/cpufeature.c       | 2 +-
 3 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/arch/arm64/include/asm/mmu_context.h b/arch/arm64/include/asm/mmu_context.h
index 16cc9a694bb2..270ba8761a23 100644
--- a/arch/arm64/include/asm/mmu_context.h
+++ b/arch/arm64/include/asm/mmu_context.h
@@ -119,7 +119,7 @@ static inline void cpu_install_idmap(void)
  * Atomically replaces the active TTBR1_EL1 PGD with a new VA-compatible PGD,
  * avoiding the possibility of conflicting TLB entries being allocated.
  */
-static inline void cpu_replace_ttbr1(pgd_t *pgdp)
+static inline void __nocfi cpu_replace_ttbr1(pgd_t *pgdp)
 {
 	typedef void (ttbr_replace_func)(phys_addr_t);
 	extern ttbr_replace_func idmap_cpu_replace_ttbr1;
diff --git a/arch/arm64/kernel/cpu-reset.h b/arch/arm64/kernel/cpu-reset.h
index dfba8cf921e5..a05bda363272 100644
--- a/arch/arm64/kernel/cpu-reset.h
+++ b/arch/arm64/kernel/cpu-reset.h
@@ -13,10 +13,10 @@
 void __cpu_soft_restart(unsigned long el2_switch, unsigned long entry,
 	unsigned long arg0, unsigned long arg1, unsigned long arg2);
 
-static inline void __noreturn cpu_soft_restart(unsigned long entry,
-					       unsigned long arg0,
-					       unsigned long arg1,
-					       unsigned long arg2)
+static inline void __noreturn __nocfi cpu_soft_restart(unsigned long entry,
+						       unsigned long arg0,
+						       unsigned long arg1,
+						       unsigned long arg2)
 {
 	typeof(__cpu_soft_restart) *restart;
 
diff --git a/arch/arm64/kernel/cpufeature.c b/arch/arm64/kernel/cpufeature.c
index 7ec1c2ccdc0b..473212ff4d70 100644
--- a/arch/arm64/kernel/cpufeature.c
+++ b/arch/arm64/kernel/cpufeature.c
@@ -1443,7 +1443,7 @@ static bool unmap_kernel_at_el0(const struct arm64_cpu_capabilities *entry,
 }
 
 #ifdef CONFIG_UNMAP_KERNEL_AT_EL0
-static void
+static void __nocfi
 kpti_install_ng_mappings(const struct arm64_cpu_capabilities *__unused)
 {
 	typedef void (kpti_remap_fn)(int, int, phys_addr_t);
-- 
2.31.0.291.g576ba9dcdaf-goog

[PATCH v3 15/17] arm64: add __nocfi to __apply_alternatives

[Sami Tolvanen]

__apply_alternatives makes indirect calls to functions whose address
is taken in assembly code using the alternative_cb macro. With
non-canonical CFI, the compiler won't replace these function
references with the jump table addresses, which trips CFI. Disable CFI
checking in the function to work around the issue.

Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
---
 arch/arm64/kernel/alternative.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/arch/arm64/kernel/alternative.c b/arch/arm64/kernel/alternative.c
index 1184c44ea2c7..abc84636af07 100644
--- a/arch/arm64/kernel/alternative.c
+++ b/arch/arm64/kernel/alternative.c
@@ -133,8 +133,8 @@ static void clean_dcache_range_nopatch(u64 start, u64 end)
 	} while (cur += d_size, cur < end);
 }
 
-static void __apply_alternatives(void *alt_region,  bool is_module,
-				 unsigned long *feature_mask)
+static void __nocfi __apply_alternatives(void *alt_region,  bool is_module,
+					 unsigned long *feature_mask)
 {
 	struct alt_instr *alt;
 	struct alt_region *region = alt_region;
-- 
2.31.0.291.g576ba9dcdaf-goog

[PATCH v3 16/17] KVM: arm64: Disable CFI for nVHE

[Sami Tolvanen]

Disable CFI for the nVHE code to avoid address space confusion.

Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
---
 arch/arm64/kvm/hyp/nvhe/Makefile | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/arch/arm64/kvm/hyp/nvhe/Makefile b/arch/arm64/kvm/hyp/nvhe/Makefile
index a6707df4f6c0..fb24a0f022ad 100644
--- a/arch/arm64/kvm/hyp/nvhe/Makefile
+++ b/arch/arm64/kvm/hyp/nvhe/Makefile
@@ -75,9 +75,9 @@ quiet_cmd_hyprel = HYPREL  $@
 quiet_cmd_hypcopy = HYPCOPY $@
       cmd_hypcopy = $(OBJCOPY) --prefix-symbols=__kvm_nvhe_ $< $@
 
-# Remove ftrace and Shadow Call Stack CFLAGS.
-# This is equivalent to the 'notrace' and '__noscs' annotations.
-KBUILD_CFLAGS := $(filter-out $(CC_FLAGS_FTRACE) $(CC_FLAGS_SCS), $(KBUILD_CFLAGS))
+# Remove ftrace, Shadow Call Stack, and CFI CFLAGS.
+# This is equivalent to the 'notrace', '__noscs', and '__nocfi' annotations.
+KBUILD_CFLAGS := $(filter-out $(CC_FLAGS_FTRACE) $(CC_FLAGS_SCS) $(CC_FLAGS_CFI), $(KBUILD_CFLAGS))
 
 # KVM nVHE code is run at a different exception code with a different map, so
 # compiler instrumentation that inserts callbacks or checks into the code may
-- 
2.31.0.291.g576ba9dcdaf-goog

[PATCH v3 17/17] arm64: allow CONFIG_CFI_CLANG to be selected

[Sami Tolvanen]

Select ARCH_SUPPORTS_CFI_CLANG to allow CFI to be enabled.

Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
---
 arch/arm64/Kconfig | 1 +
 1 file changed, 1 insertion(+)

diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
index 5656e7aacd69..2eefdbc3e3c9 100644
--- a/arch/arm64/Kconfig
+++ b/arch/arm64/Kconfig
@@ -75,6 +75,7 @@ config ARM64
 	select ARCH_SUPPORTS_SHADOW_CALL_STACK if CC_HAVE_SHADOW_CALL_STACK
 	select ARCH_SUPPORTS_LTO_CLANG if CPU_LITTLE_ENDIAN
 	select ARCH_SUPPORTS_LTO_CLANG_THIN
+	select ARCH_SUPPORTS_CFI_CLANG
 	select ARCH_SUPPORTS_ATOMIC_RMW
 	select ARCH_SUPPORTS_INT128 if CC_HAS_INT128 && (GCC_VERSION >= 50000 || CC_IS_CLANG)
 	select ARCH_SUPPORTS_NUMA_BALANCING
-- 
2.31.0.291.g576ba9dcdaf-goog