lld/MachO/EhFrame.h

   1 //===- EhFrame.h ------------------------------------------------*- C++ -*-===//
   2 //
   3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
   4 // See https://llvm.org/LICENSE.txt for license information.
   5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
   6 //
   7 //===----------------------------------------------------------------------===//
   8
   9 #ifndef LLD_MACHO_EH_FRAME_H
  10 #define LLD_MACHO_EH_FRAME_H
  11
  12 #include "InputSection.h"
  13 #include "Relocations.h"
  14
  15 #include "lld/Common/LLVM.h"
  16 #include "llvm/ADT/ArrayRef.h"
  17 #include "llvm/ADT/PointerUnion.h"
  18 #include "llvm/ADT/SmallVector.h"
  19
  20 /*
  21  * NOTE: The main bulk of the EH frame parsing logic is in InputFiles.cpp as it
  22  * is closely coupled with other file parsing logic; EhFrame.h just contains a
  23  * few helpers.
  24  */
  25
  26 /*
  27  * === The EH frame format ===
  28  *
  29  * EH frames can either be Common Information Entries (CIEs) or Frame
  30  * Description Entries (FDEs). CIEs contain information that is common amongst
  31  * several FDEs. Each FDE contains a pointer to its CIE. Thus all the EH frame
  32  * entries together form a forest of two-level trees, with CIEs as the roots
  33  * and FDEs as the leaves. Note that a CIE must precede the FDEs which point
  34  * to it.
  35  *
  36  * A CIE comprises the following fields in order:
  37  * 1.   Length of the entry (4 or 12 bytes)
  38  * 2.   CIE offset (4 bytes; always 0 for CIEs)
  39  * 3.   CIE version (byte)
  40  * 4.   Null-terminated augmentation string
  41  * 5-8. LEB128 values that we don't care about
  42  * 9.   Augmentation data, to be interpreted using the aug string
  43  * 10.  DWARF instructions (ignored by LLD)
  44  *
  45  * An FDE comprises of the following:
  46  * 1. Length of the entry (4 or 12 bytes)
  47  * 2. CIE offset (4 bytes pcrel offset that points backwards to this FDE's CIE)
  48  * 3. Function address (pointer-sized pcrel offset)
  49  * 4. (Optional) Augmentation data length
  50  * 5. (Optional) LSDA address (pointer-sized pcrel offset)
  51  * 6. DWARF instructions (ignored by LLD)
  52  */
  53 namespace lld::macho {
  54
  55 class EhReader {
  56 public:
  57   EhReader(const ObjFile *file, ArrayRef<uint8_t> data, size_t dataOff)
  58       : file(file), data(data), dataOff(dataOff) {}
  59   size_t size() const { return data.size(); }
  60   // Read and validate the length field.
  61   uint64_t readLength(size_t *off) const;
  62   // Skip the length field without doing validation.
  63   void skipValidLength(size_t *off) const;
  64   uint8_t readByte(size_t *off) const;
  65   uint32_t readU32(size_t *off) const;
  66   uint64_t readPointer(size_t *off, uint8_t size) const;
  67   StringRef readString(size_t *off) const;
  68   void skipLeb128(size_t *off) const;
  69   void failOn(size_t errOff, const Twine &msg) const;
  70
  71 private:
  72   const ObjFile *file;
  73   ArrayRef<uint8_t> data;
  74   // The offset of the data array within its section. Used only for error
  75   // reporting.
  76   const size_t dataOff;
  77 };
  78
  79 // The EH frame format, when emitted by llvm-mc, consists of a number of
  80 // "abs-ified" relocations, i.e. relocations that are implicitly encoded as
  81 // pcrel offsets in the section data. The offsets refer to the locations of
  82 // symbols in the input object file. When we ingest these EH frames, we convert
  83 // these implicit relocations into explicit Relocs.
  84 //
  85 // These pcrel relocations are semantically similar to X86_64_RELOC_SIGNED_4.
  86 // However, we need this operation to be cross-platform, and ARM does not have a
  87 // similar relocation that is applicable. We therefore use the more verbose (but
  88 // more generic) subtractor relocation to encode these pcrel values. ld64
  89 // appears to do something similar -- its `-r` output contains these explicit
  90 // subtractor relocations.
  91 class EhRelocator {
  92 public:
  93   EhRelocator(InputSection *isec) : isec(isec) {}
  94
  95   // For the next two methods, let `PC` denote `isec address + off`.
  96   // Create relocs writing the value of target - PC to PC.
  97   void makePcRel(uint64_t off,
  98                  llvm::PointerUnion<Symbol *, InputSection *> target,
  99                  uint8_t length);
 100   // Create relocs writing the value of PC - target to PC.
 101   void makeNegativePcRel(uint64_t off,
 102                          llvm::PointerUnion<Symbol *, InputSection *> target,
 103                          uint8_t length);
 104   // Insert the new relocations into isec->relocs.
 105   void commit();
 106
 107 private:
 108   InputSection *isec;
 109   // Insert new relocs here so that we don't invalidate iterators into the
 110   // existing relocs vector.
 111   SmallVector<Reloc, 6> newRelocs;
 112 };
 113
 114 } // namespace lld::macho
 115
 116 #endif