[libunwind] Support rtld-c18n as the runtime linker. (#731)

This commit adds support for backtrace and exception handling in
libunwind when the process is running with the compartmentalization
runtime linker. The unwinding process remains the same until a
trampoline is encountered as the return address. This means that we are
crossing compartment boundaries and we need to gather the unwind
information from the runtime linker. We do this by reading information
from the executive stack that the runtime linker populates for us in
unw_getcontext.

It also adds a new class, CompartmentInfo, which is responsible for
abstracting away the details of c18n compartments. Currently, it is only
used to define the constants relating to the trusted frame layout.

The otype allocated to libunwind is given to libunwind by the runtime
linker via the _rtld_unw_getsealer function, and as such this code is
guarded by a LIBUNWIND_CHERI_C18N_SUPPORT define. The sealer is
used to internally access the executive stack pointer in order to unwind
through compartment boundaries. This design may change in the future.

This functionality only works on Morello right now.

libunwind/CMakeLists.txt

@@ -50,6 +50,7 @@ option(LIBUNWIND_IS_BAREMETAL "Build libunwind for baremetal targets." OFF)
 option(LIBUNWIND_USE_FRAME_HEADER_CACHE "Cache frame headers for unwinding. Requires locking dl_iterate_phdr." OFF)
 option(LIBUNWIND_REMEMBER_HEAP_ALLOC "Use heap instead of the stack for .cfi_remember_state." OFF)
 option(LIBUNWIND_INSTALL_HEADERS "Install the libunwind headers." OFF)
+option(LIBUNWIND_CHERI_C18N_SUPPORT "Use a libunwind implementation that supports a CHERI c18n RTLD." OFF)
 
 set(LIBUNWIND_LIBDIR_SUFFIX "${LLVM_LIBDIR_SUFFIX}" CACHE STRING
     "Define suffix of library directory name (32/64)")
@@ -293,6 +294,14 @@ if (NOT LIBUNWIND_ENABLE_THREADS)
   add_compile_flags(-D_LIBUNWIND_HAS_NO_THREADS)
 endif()
 
+# Sandboxing and c18n support
+if (LIBUNWIND_CHERI_C18N_SUPPORT)
+  if (NOT CMAKE_SYSTEM_PROCESSOR MATCHES "ARM64" OR NOT CMAKE_SYSTEM_NAME MATCHES "FreeBSD")
+    message(FATAL_ERROR "LIBUNWIND_CHERI_C18N_SUPPORT is not supported for ${CMAKE_SYSTEM_NAME}-${CMAKE_SYSTEM_PROCESSOR}")
+  endif()
+  add_compile_flags(-D_LIBUNWIND_CHERI_C18N_SUPPORT)
+endif()
+
 # ARM WMMX register support
 if (LIBUNWIND_ENABLE_ARM_WMMX)
   # __ARM_WMMX is a compiler pre-define (as per the ACLE 2.0). Clang does not

libunwind/include/__libunwind_config.h

@@ -76,11 +76,11 @@
 # elif defined(__aarch64__)
 #  define _LIBUNWIND_TARGET_AARCH64 1
 #  if defined(__CHERI_PURE_CAPABILITY__)
-#    define _LIBUNWIND_CONTEXT_SIZE 100
+#    define _LIBUNWIND_CONTEXT_SIZE 102
 #    if defined(__SEH__)
 #      error "Pure-capability aarch64 SEH not supported"
 #    else
-#      define _LIBUNWIND_CURSOR_SIZE 124
+#      define _LIBUNWIND_CURSOR_SIZE 126
 #    endif
 #    define _LIBUNWIND_HIGHEST_DWARF_REGISTER _LIBUNWIND_HIGHEST_DWARF_REGISTER_MORELLO
 #  else
@@ -235,4 +235,9 @@
 # define _LIBUNWIND_HIGHEST_DWARF_REGISTER 287
 #endif // _LIBUNWIND_IS_NATIVE_ONLY
 
+#if defined(_LIBUNWIND_CHERI_C18N_SUPPORT) &&                                  \
+    !defined(_LIBUNWIND_TARGET_AARCH64)
+# error "LIBUNWIND_CHERI_C18N_SUPPORT is only supported on Morello"
+#endif
+
 #endif // ____LIBUNWIND_CONFIG_H__

libunwind/include/libunwind.h

@@ -678,7 +678,11 @@ enum {
   UNW_ARM64_C30 = 228,
   UNW_ARM64_CLR = 228,
   UNW_ARM64_C31 = 229,
-  UNW_ARM64_CSP = 229
+  UNW_ARM64_CSP = 229,
+  // Use 240 for ECSP (executive stack pointer). ECSP is not a real DWARF
+  // register, but we need it to implement c18n-aware unwinding. We pick 240
+  // because it is far enough away from the range of reserved registers on Arm.
+  UNW_ARM64_ECSP = 240,
 };
 
 // 32-bit ARM registers. Numbers match DWARF for ARM spec #3.1 Table 1.

libunwind/src/AddressSpace.hpp

@@ -321,6 +321,12 @@ class _LIBUNWIND_HIDDEN LocalAddressSpace {
     return get<v128>(addr);
   }
   capability_t     getCapability(pint_t addr) { return get<capability_t>(addr); }
+#if defined(__CHERI_PURE_CAPABILITY__) && defined(_LIBUNWIND_CHERI_C18N_SUPPORT)
+  static pint_t getUnwindSealer();
+  static bool isValidSealer(pint_t sealer) {
+    return __builtin_cheri_tag_get(sealer);
+  }
+#endif // __CHERI_PURE_CAPABILITY__ && _LIBUNWIND_CHERI_C18N_SUPPORT
   __attribute__((always_inline))
   uintptr_t       getP(pint_t addr);
   uint64_t        getRegister(pint_t addr);
@@ -409,6 +415,24 @@ inline uint64_t LocalAddressSpace::getRegister(pint_t addr) {
 #endif
 }
 
+#if defined(__CHERI_PURE_CAPABILITY__) && defined(_LIBUNWIND_CHERI_C18N_SUPPORT)
+extern "C" {
+/// Call into the RTLD to get a sealer capability. This sealer will be used to
+/// seal information in the unwinding context.
+uintptr_t _rtld_unw_getsealer();
+uintptr_t __rtld_unw_getsealer();
+_LIBUNWIND_HIDDEN uintptr_t __rtld_unw_getsealer() {
+  return (uintptr_t)0;
+}
+_LIBUNWIND_WEAK_ALIAS(__rtld_unw_getsealer, _rtld_unw_getsealer)
+}
+
+/// C++ wrapper for calling into RTLD.
+inline LocalAddressSpace::pint_t LocalAddressSpace::getUnwindSealer() {
+  return _rtld_unw_getsealer();
+}
+#endif // __CHERI_PURE_CAPABILITY__ && _LIBUNWIND_CHERI_C18N_SUPPORT
+
 /// Read a ULEB128 into a 64-bit word.
 inline uint64_t LocalAddressSpace::getULEB128(pint_t &addr, pint_t end) {
   const uint8_t *p = (uint8_t *)addr;
@@ -933,7 +957,8 @@ inline bool LocalAddressSpace::findUnwindSections(pc_t targetAddr,
     return true;
 #elif defined(_LIBUNWIND_USE_DL_ITERATE_PHDR)
   dl_iterate_cb_data cb_data = {this, &info, targetAddr};
-  CHERI_DBG("Calling dl_iterate_phdr()\n");
+  CHERI_DBG("Calling dl_iterate_phdr(0x%jx)\n",
+            (uintmax_t)targetAddr.address());
   int found = dl_iterate_phdr(findUnwindSectionsByPhdr, &cb_data);
   return static_cast<bool>(found);
 #endif

libunwind/src/CompartmentInfo.hpp

@@ -0,0 +1,33 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//
+// Abstracts unwind information when used with a compartmentalizing runtime
+// linker.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef __COMPARTMENT_INFO_HPP__
+#define __COMPARTMENT_INFO_HPP__
+
+namespace libunwind {
+class _LIBUNWIND_HIDDEN CompartmentInfo {
+public:
+#if defined(__CHERI_PURE_CAPABILITY__) && defined(_LIBUNWIND_CHERI_C18N_SUPPORT)
+  static CompartmentInfo sThisCompartmentInfo;
+  // Per-architecture trusted stack frame layout.
+#if defined(_LIBUNWIND_TARGET_AARCH64)
+  static const uint32_t kNewSPOffset = 12 * sizeof(void *);
+  static const uint32_t kNextOffset = 14 * sizeof(void *);
+  static const uint32_t kCalleeSavedOffset = 2 * sizeof(void *);
+  static const uint32_t kCalleeSavedCount = 10;
+  static const uint32_t kReturnAddressOffset = 15 * sizeof(void *) + 8;
+  static const uint32_t kPCOffset = sizeof(void *);
+#endif // _LIBUNWIND_TARGET_AARCH64
+#endif // __CHERI_PURE_CAPABILITY__ && _LIBUNWIND_CHERI_C18N_SUPPORT
+};
+} // namespace libunwind
+#endif // __COMPARTMENT_INFO_HPP__

libunwind/src/DwarfInstructions.hpp

@@ -20,6 +20,7 @@
 #include "Registers.hpp"
 #include "DwarfParser.hpp"
 #include "config.h"
+#include "CompartmentInfo.hpp"
 
 
 namespace libunwind {
@@ -54,6 +55,14 @@ class DwarfInstructions {
   typedef typename CFI_Parser<A>::FDE_Info          FDE_Info;
   typedef typename CFI_Parser<A>::CIE_Info          CIE_Info;
 
+#if defined(__CHERI_PURE_CAPABILITY__) && defined(_LIBUNWIND_CHERI_C18N_SUPPORT)
+  static pint_t restoreRegistersFromSandbox(pint_t csp, A &addressSpace,
+                                            R &newRegisters,
+                                            CompartmentInfo &CI, pint_t sealer);
+  static bool isCompartmentTransitionTrampoline(pint_t ecsp, A &addressSpace,
+                                                CompartmentInfo &CI,
+                                                pint_t returnAddress);
+#endif
   static pint_t evaluateExpression(pint_t expression, A &addressSpace,
                                    const R &registers,
                                    pint_t initialStackValue);
@@ -246,6 +255,75 @@ bool DwarfInstructions<A, R>::getRA_SIGN_STATE(A &addressSpace, R registers,
 }
 #endif
 
+#if defined(__CHERI_PURE_CAPABILITY__) && defined(_LIBUNWIND_CHERI_C18N_SUPPORT)
+#if defined(_LIBUNWIND_TARGET_AARCH64)
+template <typename A, typename R>
+typename A::pint_t DwarfInstructions<A, R>::restoreRegistersFromSandbox(
+    pint_t csp, A &addressSpace, R &newRegisters, CompartmentInfo &CI,
+    pint_t sealer) {
+  // Get the unsealed executive CSP
+  assert(__builtin_cheri_tag_get((void *)csp) &&
+         "Executive stack should be tagged!");
+  // Derive the new executive CSP
+  pint_t nextCSP = addressSpace.getP(csp + CI.kNextOffset);
+  // Seal ECSP
+  nextCSP = __builtin_cheri_seal(nextCSP, sealer);
+  assert(__builtin_cheri_tag_get((void *)nextCSP) &&
+         "Next executive stack should be tagged!");
+  CHERI_DBG("SANDBOX: SETTING EXECUTIVE CSP %#p\n", (void *)nextCSP);
+  newRegisters.setTrustedStack(nextCSP);
+  // Restore the next RCSP
+  pint_t nextRCSP = addressSpace.getP(csp + CI.kNewSPOffset);
+  newRegisters.setSP(nextRCSP);
+  CHERI_DBG("SANDBOX: SETTING RESTRICTED CSP: %#p\n",
+            (void *)newRegisters.getSP());
+  // Restore callee-saved registers
+  // Restore: c19-c28
+  for (size_t i = 0, offset = CI.kCalleeSavedOffset; i < CI.kCalleeSavedCount;
+       ++i, offset += sizeof(void *)) {
+    pint_t regValue = addressSpace.getP(csp + offset);
+    newRegisters.setCapabilityRegister(UNW_ARM64_C19 + i, regValue);
+    CHERI_DBG("SETTING CALLEE SAVED CAPABILITY REGISTER: %lu (%s): %#p "
+              "(offset=%zu)\n",
+              UNW_ARM64_C19 + i,
+              newRegisters.getRegisterName(UNW_ARM64_C19 + i), (void *)regValue,
+              offset);
+  }
+  // Restore the frame pointer
+  pint_t newFP = addressSpace.getP(csp);
+  CHERI_DBG("SANDBOX: SETTING CFP %#p\n", (void *)newFP);
+  newRegisters.setFP(newFP);
+  // Get the new return address.
+  return addressSpace.getP(csp + CI.kPCOffset);
+}
+
+template <typename A, typename R>
+bool DwarfInstructions<A, R>::isCompartmentTransitionTrampoline(
+    pint_t ecsp, A &addressSpace, CompartmentInfo &CI, pint_t returnAddress) {
+  ptraddr_t expectedReturnAddress =
+      addressSpace.template get<ptraddr_t>(ecsp + CI.kReturnAddressOffset);
+  CHERI_DBG(
+      "isCompartmentTransitionTrampoline(): expectedReturnAddress: 0x%lx\n",
+      expectedReturnAddress);
+  return expectedReturnAddress == returnAddress;
+}
+#else // _LIBUNWIND_TARGET_AARCH64
+template <typename A, typename R>
+typename A::pint_t DwarfInstructions<A, R>::restoreRegistersFromSandbox(
+    pint_t csp, A &addressSpace, R &newRegisters, CompartmentInfo &CI,
+    pint_t sealer) {
+  assert(0 && "not implemented on this architecture");
+  return (pint_t)0;
+}
+template <typename A, typename R>
+bool DwarfInstructions<A, R>::isCompartmentTransitionTrampoline(
+    pint_t ecsp, A &addressSpace, CompartmentInfo &CI, pint_t returnAddress) {
+  assert(0 && "not implemented on this architecture");
+  return false;
+}
+#endif // _LIBUNWIND_TARGET_AARCH64
+#endif // __CHERI_PURE_CAPABILITY__ && _LIBUNWIND_CHERI_C18N_SUPPORT
+
 template <typename A, typename R>
 int DwarfInstructions<A, R>::stepWithDwarf(A &addressSpace, pc_t pc,
                                            pint_t fdeStart, R &registers,
@@ -274,6 +352,7 @@ int DwarfInstructions<A, R>::stepWithDwarf(A &addressSpace, pc_t pc,
       //
       // We set the SP here to the CFA, allowing for it to be overridden
       // by a CFI directive later on.
+      CHERI_DBG("SETTING SP: %#p\n", (void *)cfa);
       newRegisters.setSP(cfa);
 
       pint_t returnAddress = 0;
@@ -297,25 +376,26 @@ int DwarfInstructions<A, R>::stepWithDwarf(A &addressSpace, pc_t pc,
           else if (i == (int)cieInfo.returnAddressRegister) {
             returnAddress = getSavedRegister(i, addressSpace, registers, cfa,
                                              prolog.savedRegisters[i]);
-            CHERI_DBG("SETTING RETURN REGISTER %d (%s): %#p \n",
-                      i, newRegisters.getRegisterName(i), (void*)returnAddress);
+            CHERI_DBG("GETTING RETURN ADDRESS (saved) %d (%s): %#p \n", i,
+                      newRegisters.getRegisterName(i), (void *)returnAddress);
           } else if (registers.validCapabilityRegister(i)) {
-            newRegisters.setCapabilityRegister(
-                i, getSavedCapabilityRegister(addressSpace, registers, cfa,
-                                              prolog.savedRegisters[i]));
-            CHERI_DBG("SETTING CAPABILITY REGISTER %d (%s): %#p \n",
-                      i, newRegisters.getRegisterName(i),
-                      (void*)A::to_pint_t(newRegisters.getCapabilityRegister(i)));
+            capability_t savedReg = getSavedCapabilityRegister(
+                addressSpace, registers, cfa, prolog.savedRegisters[i]);
+            newRegisters.setCapabilityRegister(i, savedReg);
+            CHERI_DBG("SETTING CAPABILITY REGISTER %d (%s): %#p \n", i,
+                      newRegisters.getRegisterName(i), (void *)savedReg);
           } else if (registers.validRegister(i))
             newRegisters.setRegister(
                 i, getSavedRegister(i, addressSpace, registers, cfa,
                                     prolog.savedRegisters[i]));
           else
             return UNW_EBADREG;
         } else if (i == (int)cieInfo.returnAddressRegister) {
-            // Leaf function keeps the return address in register and there is no
-            // explicit intructions how to restore it.
-            returnAddress = registers.getRegister(cieInfo.returnAddressRegister);
+          // Leaf function keeps the return address in register and there is no
+          // explicit intructions how to restore it.
+          returnAddress = registers.getRegister(cieInfo.returnAddressRegister);
+          CHERI_DBG("GETTING RETURN ADDRESS (leaf) %d (%s): %#p \n", i,
+                    registers.getRegisterName(i), (void *)returnAddress);
         }
       }
 
@@ -403,9 +483,32 @@ int DwarfInstructions<A, R>::stepWithDwarf(A &addressSpace, pc_t pc,
       }
 #endif
 
+#if defined(__CHERI_PURE_CAPABILITY__) && defined(_LIBUNWIND_CHERI_C18N_SUPPORT)
+      // If the sealer is not valid (only the case when we're running without
+      // c18n), check if the return address has the executive mode bit set.
+      // If so, we should be calling into the c18n RTLD as this is a
+      // compartment boundary. We need to restore registers from the executive
+      // stack and ask rtld for it.
+      uintptr_t sealer = addressSpace.getUnwindSealer();
+      if (addressSpace.isValidSealer(sealer)) {
+        pint_t csp = registers.getTrustedStack();
+        if (__builtin_cheri_sealed_get(csp))
+          csp = __builtin_cheri_unseal(csp, sealer);
+        CompartmentInfo &CI = CompartmentInfo::sThisCompartmentInfo;
+        if (csp != 0 && isCompartmentTransitionTrampoline(csp, addressSpace, CI,
+                                                          returnAddress)) {
+          CHERI_DBG("%#p: detected a trampoline, unwinding from sandbox\n",
+                    (void *)returnAddress);
+          returnAddress = restoreRegistersFromSandbox(
+              csp, addressSpace, newRegisters, CI, sealer);
+        }
+      }
+#endif
+
       // Return address is address after call site instruction, so setting IP to
       // that does simualates a return.
       newRegisters.setIP(returnAddress);
+      CHERI_DBG("SETTING RETURN ADDRESS %#p\n", (void *)returnAddress);
 
       // Simulate the step by replacing the register set with the new ones.
       registers = newRegisters;

libunwind/src/Registers.hpp

@@ -1866,14 +1866,18 @@ class _LIBUNWIND_HIDDEN Registers_arm64 {
   void        setCapabilityRegister(int num, uintcap_t value);
 #else
   CAPABILITIES_NOT_SUPPORTED
-#endif
+#endif // __CHERI_PURE_CAPABILITY__
 
   uintptr_t  getSP() const          { return _registers.__sp; }
   void       setSP(uintptr_t value) { _registers.__sp = value; }
   uintptr_t  getIP() const          { return _registers.__pc; }
   void       setIP(uintptr_t value) { _registers.__pc = value; }
   uintptr_t  getFP() const          { return _registers.__fp; }
   void       setFP(uintptr_t value) { _registers.__fp = value; }
+#if defined(__CHERI_PURE_CAPABILITY__) && defined(_LIBUNWIND_CHERI_C18N_SUPPORT)
+  uintptr_t  getTrustedStack() const { return _registers.__ecsp; }
+  void       setTrustedStack(uintptr_t value) { _registers.__ecsp = value; }
+#endif // __CHERI_PURE_CAPABILITY__ && _LIBUNWIND_CHERI_C18N_SUPPORT
 
 private:
   struct GPRs {
@@ -1882,6 +1886,9 @@ class _LIBUNWIND_HIDDEN Registers_arm64 {
     uintptr_t __lr;    // Link register r30
     uintptr_t __sp;    // Stack pointer r31
     uintptr_t __pc;    // Program counter
+#ifdef __CHERI_PURE_CAPABILITY__
+    uintptr_t __ecsp;  // Executive stack pointer.
+#endif
     uint64_t __ra_sign_state; // RA sign state register
   };
 
@@ -1898,8 +1905,8 @@ inline Registers_arm64::Registers_arm64(const void *registers) {
                 "arm64 registers do not fit into unw_context_t");
   memcpy(&_registers, registers, sizeof(_registers));
 #ifdef __CHERI_PURE_CAPABILITY__
-  static_assert(sizeof(GPRs) == 0x220,
-                "expected VFP registers to be at offset 544");
+  static_assert(sizeof(GPRs) == 0x230,
+                "expected VFP registers to be at offset 560");
 #else
   static_assert(sizeof(GPRs) == 0x110,
                 "expected VFP registers to be at offset 272");
@@ -2198,6 +2205,8 @@ inline const char *Registers_arm64::getRegisterName(int regNum) {
     return "clr";
   case UNW_ARM64_C31:
     return "csp";
+  case UNW_ARM64_ECSP:
+    return "ecsp";
   default:
     return "unknown register";
   }