Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Fri, 28 Jan 2022 12:44:46 +0100
In-Reply-To: <20220128114446.740575-1-elver@google.com>
Message-Id: <20220128114446.740575-2-elver@google.com>
Mime-Version: 1.0
References: <20220128114446.740575-1-elver@google.com>
Subject: [PATCH 2/2] stack: Constrain stack offset randomization with Clang builds
From:   Marco Elver <elver@google.com>
To:     elver@google.com, Thomas Gleixner <tglx@linutronix.de>,
        Kees Cook <keescook@chromium.org>
Cc:     Peter Zijlstra <peterz@infradead.org>,
        Ingo Molnar <mingo@kernel.org>,
        Nathan Chancellor <nathan@kernel.org>,
        Nick Desaulniers <ndesaulniers@google.com>,
        Elena Reshetova <elena.reshetova@intel.com>,
        Alexander Potapenko <glider@google.com>, llvm@lists.linux.dev,
        kasan-dev@googlegroups.com, linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

All supported versions of Clang perform auto-init of __builtin_alloca()
when stack auto-init is on (CONFIG_INIT_STACK_ALL_{ZERO,PATTERN}).

add_random_kstack_offset() uses __builtin_alloca() to add a stack
offset. This means, when CONFIG_INIT_STACK_ALL_{ZERO,PATTERN} is
enabled, add_random_kstack_offset() will auto-init that unused portion
of the stack used to add an offset.

There are several problems with this:

	1. These offsets can be as large as 1023 bytes. Performing
	   memset() on them isn't exactly cheap, and this is done on
	   every syscall entry.

	2. Architectures adding add_random_kstack_offset() to syscall
	   entry implemented in C require them to be 'noinstr' (e.g. see
	   x86 and s390). The potential problem here is that a call to
	   memset may occur, which is not noinstr.

A x86_64 defconfig kernel with Clang 11 and CONFIG_VMLINUX_VALIDATION shows:

 | vmlinux.o: warning: objtool: do_syscall_64()+0x9d: call to memset() leaves .noinstr.text section
 | vmlinux.o: warning: objtool: do_int80_syscall_32()+0xab: call to memset() leaves .noinstr.text section
 | vmlinux.o: warning: objtool: __do_fast_syscall_32()+0xe2: call to memset() leaves .noinstr.text section
 | vmlinux.o: warning: objtool: fixup_bad_iret()+0x2f: call to memset() leaves .noinstr.text section

Clang 14 (unreleased) will introduce a way to skip alloca initialization
via __builtin_alloca_uninitialized() (https://reviews.llvm.org/D115440).

Constrain RANDOMIZE_KSTACK_OFFSET to only be enabled if no stack
auto-init is enabled, the compiler is GCC, or Clang is version 14+. Use
__builtin_alloca_uninitialized() if the compiler provides it, as is done
by Clang 14.

Link: https://lkml.kernel.org/r/YbHTKUjEejZCLyhX@elver.google.com
Fixes: 39218ff4c625 ("stack: Optionally randomize kernel stack offset each syscall")
Signed-off-by: Marco Elver <elver@google.com>
---
 arch/Kconfig                     |  1 +
 include/linux/randomize_kstack.h | 14 ++++++++++++--
 2 files changed, 13 insertions(+), 2 deletions(-)

diff --git a/arch/Kconfig b/arch/Kconfig
index 2cde48d9b77c..c5b50bfe31c1 100644
--- a/arch/Kconfig
+++ b/arch/Kconfig
@@ -1163,6 +1163,7 @@ config RANDOMIZE_KSTACK_OFFSET
 	bool "Support for randomizing kernel stack offset on syscall entry" if EXPERT
 	default y
 	depends on HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
+	depends on INIT_STACK_NONE || !CC_IS_CLANG || CLANG_VERSION >= 140000
 	help
 	  The kernel stack offset can be randomized (after pt_regs) by
 	  roughly 5 bits of entropy, frustrating memory corruption
diff --git a/include/linux/randomize_kstack.h b/include/linux/randomize_kstack.h
index 91f1b990a3c3..5c711d73ed10 100644
--- a/include/linux/randomize_kstack.h
+++ b/include/linux/randomize_kstack.h
@@ -17,8 +17,18 @@ DECLARE_PER_CPU(u32, kstack_offset);
  * alignment. Also, since this use is being explicitly masked to a max of
  * 10 bits, stack-clash style attacks are unlikely. For more details see
  * "VLAs" in Documentation/process/deprecated.rst
+ *
+ * The normal alloca() can be initialized with INIT_STACK_ALL. Initializing the
+ * unused area on each syscall entry is expensive, and generating an implicit
+ * call to memset() may also be problematic (such as in noinstr functions).
+ * Therefore, if the compiler provides it, use the "uninitialized" variant.
  */
-void *__builtin_alloca(size_t size);
+#if __has_builtin(__builtin_alloca_uninitialized)
+#define __kstack_alloca __builtin_alloca_uninitialized
+#else
+#define __kstack_alloca __builtin_alloca
+#endif
+
 /*
  * Use, at most, 10 bits of entropy. We explicitly cap this to keep the
  * "VLA" from being unbounded (see above). 10 bits leaves enough room for
@@ -37,7 +47,7 @@ void *__builtin_alloca(size_t size);
 	if (static_branch_maybe(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,	\
 				&randomize_kstack_offset)) {		\
 		u32 offset = raw_cpu_read(kstack_offset);		\
-		u8 *ptr = __builtin_alloca(KSTACK_OFFSET_MAX(offset));	\
+		u8 *ptr = __kstack_alloca(KSTACK_OFFSET_MAX(offset));	\
 		/* Keep allocation even after "ptr" loses scope. */	\
 		asm volatile("" :: "r"(ptr) : "memory");		\
 	}								\
-- 
2.35.0.rc0.227.g00780c9af4-goog