lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20230325081117.93245-4-ashimida.1990@gmail.com>
Date:   Sat, 25 Mar 2023 01:11:17 -0700
From:   Dan Li <ashimida.1990@...il.com>
To:     gcc-patches@....gnu.org,
        Richard Sandiford <richard.sandiford@....com>,
        Masahiro Yamada <masahiroy@...nel.org>,
        Michal Marek <michal.lkml@...kovi.net>,
        Nick Desaulniers <ndesaulniers@...gle.com>,
        Catalin Marinas <catalin.marinas@....com>,
        Will Deacon <will@...nel.org>,
        Sami Tolvanen <samitolvanen@...gle.com>,
        Kees Cook <keescook@...omium.org>,
        Nathan Chancellor <nathan@...nel.org>,
        Tom Rix <trix@...hat.com>,
        Peter Zijlstra <peterz@...radead.org>,
        "Paul E. McKenney" <paulmck@...nel.org>,
        Mark Rutland <mark.rutland@....com>,
        Josh Poimboeuf <jpoimboe@...nel.org>,
        Frederic Weisbecker <frederic@...nel.org>,
        "Eric W. Biederman" <ebiederm@...ssion.com>,
        Dan Li <ashimida.1990@...il.com>,
        Marco Elver <elver@...gle.com>,
        Christophe Leroy <christophe.leroy@...roup.eu>,
        Song Liu <song@...nel.org>,
        Andrew Morton <akpm@...ux-foundation.org>,
        Uros Bizjak <ubizjak@...il.com>,
        Kumar Kartikeya Dwivedi <memxor@...il.com>,
        Juergen Gross <jgross@...e.com>,
        Luis Chamberlain <mcgrof@...nel.org>,
        Borislav Petkov <bp@...e.de>,
        Masami Hiramatsu <mhiramat@...nel.org>,
        Dmitry Torokhov <dmitry.torokhov@...il.com>,
        Aaron Tomlin <atomlin@...hat.com>,
        Kalesh Singh <kaleshsingh@...gle.com>,
        Yuntao Wang <ytcoode@...il.com>,
        Changbin Du <changbin.du@...el.com>
Cc:     linux-kbuild@...r.kernel.org, linux-kernel@...r.kernel.org,
        linux-arm-kernel@...ts.infradead.org, llvm@...ts.linux.dev,
        linux-hardening@...r.kernel.org
Subject: [RFC/RFT,V2 3/3] [PR102768] aarch64: Add support for Kernel Control Flow Integrity

In the AArch64 platform, typeid can be directly inserted in front
of the function header (offset is patch_area_entry + 4), it should
be assumed that patch_area_entry is the same for all functions.

For all functions that will not be called indirectly, insert the
reserved RESERVED_CFI_TYPEID (0x0) as typeid in front of them. If
not, the attacker may use the instruction/data before the function
as typeid to bypass CFI.

All typeids ignore some bits (& AARCH64_UNALLOCATED_INSN_MASK) to
avoid conflicts with the AArch64 instruction set (see AAPCS64 for
details).

Signed-off-by: Dan Li <ashimida.1990@...il.com>

gcc/ChangeLog:

	* config/aarch64/aarch64.cc (RESERVED_CFI_TYPEID): Macro definition.
	(DEFAULT_CFI_TYPEID): Likewise.
	(AARCH64_UNALLOCATED_INSN_MASK): Likewise.
	(aarch64_calc_func_cfi_typeid): Platform-dependent CFI function.
	(cgraph_indirectly_callable): Determine whether a funtion may
	be called indirectly.
	(aarch64_output_func_kcfi_typeid): Platform-dependent CFI function.
	(aarch64_output_icall_kcfi_check): Likewise.
	(TARGET_HAVE_KCFI): New hook.
	(TARGET_CALC_FUNC_CFI_TYPEID): Likewise.
	(TARGET_ASM_OUTPUT_FUNC_KCFI_TYPEID): Likewise.
	(TARGET_ASM_OUTPUT_ICALL_KCFI_CHECK): Likewise.
	* doc/invoke.texi: Document -fsanitize=kcfi.
---
 gcc/config/aarch64/aarch64.cc | 166 ++++++++++++++++++++++++++++++++++
 gcc/doc/invoke.texi           |  36 ++++++++
 2 files changed, 202 insertions(+)

diff --git a/gcc/config/aarch64/aarch64.cc b/gcc/config/aarch64/aarch64.cc
index 5c9e7791a12..5b55541d437 100644
--- a/gcc/config/aarch64/aarch64.cc
+++ b/gcc/config/aarch64/aarch64.cc
@@ -5450,6 +5450,160 @@ aarch64_output_sve_addvl_addpl (rtx offset)
   return buffer;
 }
 
+/* Reserved for all functions that cannot be called indirectly.  */
+#define RESERVED_CFI_TYPEID 0x0U
+
+/* If the typeid of a function that can be called indirectly is equal to
+   RESERVED_CFI_TYPEID, change it to DEFAULT_CFI_TYPEID.  */
+#define DEFAULT_CFI_TYPEID 0x00000ADAU
+
+/* Mask of reserved and unallocated instructions in AArch64 platform.  */
+#define AARCH64_UNALLOCATED_INSN_MASK 0xE7FFFFFFU
+
+static unsigned int
+aarch64_calc_func_cfi_typeid (const_tree fntype)
+{
+  unsigned int hash;
+
+  /* The value of typeid has a probability of being the same as the encoding
+     of an instruction.  If the attacker can find the same encoding as the
+     typeid in the assembly code, then he has found a usable jump location.
+     So here, a platform-related mask is used when generating a typeid to
+     avoid such conflicts as much as possible.  */
+  hash = unified_type_hash (fntype) & AARCH64_UNALLOCATED_INSN_MASK;
+
+  /* RESERVED_CFI_TYPEID is reserved for functions that cannot
+     be called indirectly.  */
+  if (hash == RESERVED_CFI_TYPEID)
+    hash = DEFAULT_CFI_TYPEID;
+
+  return hash;
+}
+
+static bool
+cgraph_indirectly_callable (struct cgraph_node *node,
+			    void *data ATTRIBUTE_UNUSED)
+{
+  if (node->externally_visible || node->address_taken)
+    return true;
+
+  return false;
+}
+
+static void
+aarch64_output_func_kcfi_typeid (FILE * stream, tree decl)
+{
+  struct cgraph_node *node;
+  unsigned int cur_func_typeid;
+
+  node = cgraph_node::get (decl);
+
+  if (!node->call_for_symbol_thunks_and_aliases (cgraph_indirectly_callable,
+						 NULL, true))
+    /* CFI's typeid check always considers that there is a typeid before the
+       target function, so it is also necessary to output typeid for functions
+       that cannot be called indirectly to prevent attackers from bypassing
+       CFI by using instructions/data before those functions.
+       The typeid inserted before such a function is RESERVED_CFI_TYPEID,
+       and the calculation of the typeid must ensure that this value is always
+       reserved.  */
+    cur_func_typeid = RESERVED_CFI_TYPEID;
+  else
+    cur_func_typeid = aarch64_calc_func_cfi_typeid (TREE_TYPE (decl));
+
+  fprintf (stream, "__kcfi_%s:\n", get_name (decl));
+  fprintf (stream, "\t.4byte %#010x\n", cur_func_typeid);
+}
+
+/* This function outputs assembly instructions to check cfi typeid before
+   indirect call (blr Xn), which may destroy x16, x17, x9 registers (according
+   to the AAPCS64 specification, these registers do not need to be restored
+   after the function call).
+   The assembly code output by this function is as follows:
+	ldur    w16, [x1, #-4]
+	movk    w17, #13570
+	movk    w17, #17309, lsl #16
+	cmp     w16, w17
+	b.eq	.Lkcfi8
+	brk     #0x8221
+.Lkcfi8:
+	blr     x1
+ */
+
+static void
+aarch64_output_icall_kcfi_check (rtx reg, unsigned int value)
+{
+  unsigned int addr_reg, scratch_reg1, scratch_reg2;
+  unsigned int esr, addr_index, type_index;
+  char label_buf[256];
+  const char *label_ptr;
+  unsigned HOST_WIDE_INT patch_area_entry = crtl->patch_area_entry;
+  rtx_code_label * tmp_label = gen_label_rtx ();
+
+  gcc_assert (GET_CODE (reg) == REG);
+
+  addr_reg = REGNO (reg);
+
+  /* The typeid read from the front of the callee is saved in the
+     register specified by scratch_reg1, the default is R16_REGNUM.  */
+  scratch_reg1 = R16_REGNUM;
+
+  /* The expected typeid of the caller is saved in the register
+     specified by scratch_reg2, which defaults to R17_REGNUM.  */
+  scratch_reg2 = R17_REGNUM;
+
+  gcc_assert (GP_REGNUM_P (addr_reg));
+
+  /* If one of the scratch registers is used for the call target,
+     we can clobber another caller-saved temporary register instead
+     (in this case, R9_REGNUM) as the check is immediately followed
+     by the call instruction.  */
+  if (addr_reg == R16_REGNUM)
+    {
+      scratch_reg1 = R9_REGNUM;
+    }
+  else if (addr_reg == R17_REGNUM)
+    {
+      scratch_reg2 = R9_REGNUM;
+    }
+
+  gcc_assert ((scratch_reg1 != addr_reg) && (scratch_reg2 != addr_reg));
+
+  ASM_GENERATE_INTERNAL_LABEL (label_buf, "Lkcfi",
+			       CODE_LABEL_NUMBER (tmp_label));
+  label_ptr = targetm.strip_name_encoding (label_buf);
+
+  /* The offset of callee's typeid needs to be adjusted according to
+     patch_area_entry.  This assumes that patch_area_entry is the
+     same for all functions.  */
+  fprintf (asm_out_file, "\tldur\tw%d, [x%d, #-%ld]\n",
+	   scratch_reg1, addr_reg, patch_area_entry * 4 + 4);
+
+  fprintf (asm_out_file, "\tmovk\tw%d, #%d\n", scratch_reg2, value & 0xFFFF);
+
+  fprintf (asm_out_file, "\tmovk\tw%d, #%d, lsl #16\n",
+	   scratch_reg2, (value >> 16) & 0xFFFF);
+
+  fprintf (asm_out_file, "\tcmp\tw%d, w%d\n", scratch_reg1, scratch_reg2);
+
+  fprintf (asm_out_file, "\tb.eq\t%s\n", label_ptr);
+
+  /* The base ESR for brk is 0x8000 and the register information is
+     encoded in bits 0-9 as follows:
+     - 0-4: n, where the register Xn contains the callee address
+     - 5-9: m, where the register Wm contains the expected typeid
+     Where n, m are in[0,30].
+  */
+  addr_index = addr_reg - R0_REGNUM;
+  type_index = scratch_reg2 - R0_REGNUM;
+  esr = 0x8000 | ((type_index & 31) << 5) | (addr_index & 31);
+  fprintf (asm_out_file, "\tbrk\t#0x%x\n", esr);
+
+  fprintf (asm_out_file, "%s:\n", label_ptr);
+
+  return;
+}
+
 /* Return true if X is a valid immediate for an SVE vector INC or DEC
    instruction.  If it is, store the number of elements in each vector
    quadword in *NELTS_PER_VQ_OUT (if nonnull) and store the multiplication
@@ -27823,6 +27977,18 @@ aarch64_libgcc_floating_mode_supported_p
 #undef TARGET_HAVE_SHADOW_CALL_STACK
 #define TARGET_HAVE_SHADOW_CALL_STACK true
 
+#undef TARGET_HAVE_KCFI
+#define TARGET_HAVE_KCFI true
+
+#undef TARGET_CALC_FUNC_CFI_TYPEID
+#define TARGET_CALC_FUNC_CFI_TYPEID aarch64_calc_func_cfi_typeid
+
+#undef TARGET_ASM_OUTPUT_FUNC_KCFI_TYPEID
+#define TARGET_ASM_OUTPUT_FUNC_KCFI_TYPEID aarch64_output_func_kcfi_typeid
+
+#undef TARGET_ASM_OUTPUT_ICALL_KCFI_CHECK
+#define TARGET_ASM_OUTPUT_ICALL_KCFI_CHECK aarch64_output_icall_kcfi_check
+
 struct gcc_target targetm = TARGET_INITIALIZER;
 
 #include "gt-aarch64.h"
diff --git a/gcc/doc/invoke.texi b/gcc/doc/invoke.texi
index ff6c338bedb..1b2ba7a0f29 100644
--- a/gcc/doc/invoke.texi
+++ b/gcc/doc/invoke.texi
@@ -15736,6 +15736,42 @@ to turn off exceptions.
 See @uref{https://clang.llvm.org/docs/ShadowCallStack.html} for more
 details.
 
+@...m -fsanitize=kcfi
+@...ndex fsanitize=kcfi
+The KCFI sanitizer, enabled with @option{-fsanitize=kcfi}, implements a
+forward-edge control flow integrity scheme for indirect calls.  It
+attaches a type identifier (@code{typeid}) for each function and injects
+verification code before indirect calls.
+
+A @code{typeid} is a 32-bit constant, its value is mainly related to the
+return value type and all parameter types of the function, and is invariant
+for each compilation.  Since the value of @code{typeid} may conflict with
+the instruction set encoding of the current platform, some bits may be
+ignored on different platforms.
+
+At compile time, the compiler inserts checking code on all indirect calls,
+and at run time, before any indirect calls occur, the code checks that
+the @code{typeid} before the callee function matches the @code{typeid}
+requested by the caller.  If the match fails, an exception instruction
+will be triggered, such as a @code{brk} in aarch64.  This mechanism is
+mainly designed for low-level codes, such as operating systems, and the
+system needs to handle those exceptions by itself.
+
+If a program contains indirect calls to assembly functions, they must be
+manually annotated with the expected type identifiers to prevent errors.
+To make this easier, CFI generates a weak SHN_ABS
+@...e{__kcfi_typeid_<function>} symbol for each address-taken function
+declaration, which can be used to annotate functions in assembly as long
+as at least one C translation unit linked into the program takes the
+function address.
+
+Currently this feature only supports the aarch64 platform, mainly for
+the linux kernel.  Users who want to use this feature in other system
+need to provide their own support for the exception handling.
+
+See @uref{https://clang.llvm.org/docs/ControlFlowIntegrity.html} for
+more details.
+
 @item -fsanitize=thread
 @opindex fsanitize=thread
 Enable ThreadSanitizer, a fast data race detector.
-- 
2.17.1

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ