xref: /dflybsd-src/contrib/binutils-2.34/gold/dwarf_reader.h (revision b52ef7118d1621abed722c5bbbd542210290ecef)
1*fae548d3Szrj // dwarf_reader.h -- parse dwarf2/3 debug information for gold  -*- C++ -*-
2*fae548d3Szrj 
3*fae548d3Szrj // Copyright (C) 2007-2020 Free Software Foundation, Inc.
4*fae548d3Szrj // Written by Ian Lance Taylor <iant@google.com>.
5*fae548d3Szrj 
6*fae548d3Szrj // This file is part of gold.
7*fae548d3Szrj 
8*fae548d3Szrj // This program is free software; you can redistribute it and/or modify
9*fae548d3Szrj // it under the terms of the GNU General Public License as published by
10*fae548d3Szrj // the Free Software Foundation; either version 3 of the License, or
11*fae548d3Szrj // (at your option) any later version.
12*fae548d3Szrj 
13*fae548d3Szrj // This program is distributed in the hope that it will be useful,
14*fae548d3Szrj // but WITHOUT ANY WARRANTY; without even the implied warranty of
15*fae548d3Szrj // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
16*fae548d3Szrj // GNU General Public License for more details.
17*fae548d3Szrj 
18*fae548d3Szrj // You should have received a copy of the GNU General Public License
19*fae548d3Szrj // along with this program; if not, write to the Free Software
20*fae548d3Szrj // Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston,
21*fae548d3Szrj // MA 02110-1301, USA.
22*fae548d3Szrj 
23*fae548d3Szrj #ifndef GOLD_DWARF_READER_H
24*fae548d3Szrj #define GOLD_DWARF_READER_H
25*fae548d3Szrj 
26*fae548d3Szrj #include <vector>
27*fae548d3Szrj #include <map>
28*fae548d3Szrj #include <limits.h>
29*fae548d3Szrj #include <sys/types.h>
30*fae548d3Szrj 
31*fae548d3Szrj #include "elfcpp.h"
32*fae548d3Szrj #include "elfcpp_swap.h"
33*fae548d3Szrj #include "dwarf.h"
34*fae548d3Szrj #include "reloc.h"
35*fae548d3Szrj 
36*fae548d3Szrj namespace gold
37*fae548d3Szrj {
38*fae548d3Szrj 
39*fae548d3Szrj class Dwarf_info_reader;
40*fae548d3Szrj struct LineStateMachine;
41*fae548d3Szrj 
42*fae548d3Szrj // This class is used to extract the section index and offset of
43*fae548d3Szrj // the target of a relocation for a given offset within the section.
44*fae548d3Szrj 
45*fae548d3Szrj class Elf_reloc_mapper
46*fae548d3Szrj {
47*fae548d3Szrj  public:
Elf_reloc_mapper()48*fae548d3Szrj   Elf_reloc_mapper()
49*fae548d3Szrj   { }
50*fae548d3Szrj 
51*fae548d3Szrj   virtual
~Elf_reloc_mapper()52*fae548d3Szrj   ~Elf_reloc_mapper()
53*fae548d3Szrj   { }
54*fae548d3Szrj 
55*fae548d3Szrj   // Initialize the relocation tracker for section RELOC_SHNDX.
56*fae548d3Szrj   bool
initialize(unsigned int reloc_shndx,unsigned int reloc_type)57*fae548d3Szrj   initialize(unsigned int reloc_shndx, unsigned int reloc_type)
58*fae548d3Szrj   { return this->do_initialize(reloc_shndx, reloc_type); }
59*fae548d3Szrj 
60*fae548d3Szrj   // Return the next reloc_offset.
61*fae548d3Szrj   off_t
next_offset()62*fae548d3Szrj   next_offset()
63*fae548d3Szrj   { return this->do_next_offset(); }
64*fae548d3Szrj 
65*fae548d3Szrj   // Advance to the next relocation past OFFSET.
66*fae548d3Szrj   void
advance(off_t offset)67*fae548d3Szrj   advance(off_t offset)
68*fae548d3Szrj   { this->do_advance(offset); }
69*fae548d3Szrj 
70*fae548d3Szrj   // Return the section index and offset within the section of the target
71*fae548d3Szrj   // of the relocation for RELOC_OFFSET in the referring section.
72*fae548d3Szrj   unsigned int
get_reloc_target(off_t reloc_offset,off_t * target_offset)73*fae548d3Szrj   get_reloc_target(off_t reloc_offset, off_t* target_offset)
74*fae548d3Szrj   { return this->do_get_reloc_target(reloc_offset, target_offset); }
75*fae548d3Szrj 
76*fae548d3Szrj   // Checkpoint the current position in the reloc section.
77*fae548d3Szrj   uint64_t
checkpoint()78*fae548d3Szrj   checkpoint() const
79*fae548d3Szrj   { return this->do_checkpoint(); }
80*fae548d3Szrj 
81*fae548d3Szrj   // Reset the current position to the CHECKPOINT.
82*fae548d3Szrj   void
reset(uint64_t checkpoint)83*fae548d3Szrj   reset(uint64_t checkpoint)
84*fae548d3Szrj   { this->do_reset(checkpoint); }
85*fae548d3Szrj 
86*fae548d3Szrj  protected:
87*fae548d3Szrj   virtual bool
88*fae548d3Szrj   do_initialize(unsigned int, unsigned int) = 0;
89*fae548d3Szrj 
90*fae548d3Szrj   // Return the next reloc_offset.
91*fae548d3Szrj   virtual off_t
92*fae548d3Szrj   do_next_offset() = 0;
93*fae548d3Szrj 
94*fae548d3Szrj   // Advance to the next relocation past OFFSET.
95*fae548d3Szrj   virtual void
96*fae548d3Szrj   do_advance(off_t offset) = 0;
97*fae548d3Szrj 
98*fae548d3Szrj   virtual unsigned int
99*fae548d3Szrj   do_get_reloc_target(off_t reloc_offset, off_t* target_offset) = 0;
100*fae548d3Szrj 
101*fae548d3Szrj   // Checkpoint the current position in the reloc section.
102*fae548d3Szrj   virtual uint64_t
103*fae548d3Szrj   do_checkpoint() const = 0;
104*fae548d3Szrj 
105*fae548d3Szrj   // Reset the current position to the CHECKPOINT.
106*fae548d3Szrj   virtual void
107*fae548d3Szrj   do_reset(uint64_t checkpoint) = 0;
108*fae548d3Szrj };
109*fae548d3Szrj 
110*fae548d3Szrj template<int size, bool big_endian>
111*fae548d3Szrj class Sized_elf_reloc_mapper : public Elf_reloc_mapper
112*fae548d3Szrj {
113*fae548d3Szrj  public:
Sized_elf_reloc_mapper(Object * object,const unsigned char * symtab,off_t symtab_size)114*fae548d3Szrj   Sized_elf_reloc_mapper(Object* object, const unsigned char* symtab,
115*fae548d3Szrj 			 off_t symtab_size)
116*fae548d3Szrj     : object_(object), symtab_(symtab), symtab_size_(symtab_size),
117*fae548d3Szrj       reloc_type_(0), track_relocs_()
118*fae548d3Szrj   { }
119*fae548d3Szrj 
120*fae548d3Szrj  protected:
121*fae548d3Szrj   bool
122*fae548d3Szrj   do_initialize(unsigned int reloc_shndx, unsigned int reloc_type);
123*fae548d3Szrj 
124*fae548d3Szrj   // Return the next reloc_offset.
125*fae548d3Szrj   virtual off_t
do_next_offset()126*fae548d3Szrj   do_next_offset()
127*fae548d3Szrj   { return this->track_relocs_.next_offset(); }
128*fae548d3Szrj 
129*fae548d3Szrj   // Advance to the next relocation past OFFSET.
130*fae548d3Szrj   virtual void
do_advance(off_t offset)131*fae548d3Szrj   do_advance(off_t offset)
132*fae548d3Szrj   { this->track_relocs_.advance(offset); }
133*fae548d3Szrj 
134*fae548d3Szrj   unsigned int
135*fae548d3Szrj   do_get_reloc_target(off_t reloc_offset, off_t* target_offset);
136*fae548d3Szrj 
137*fae548d3Szrj   // Checkpoint the current position in the reloc section.
138*fae548d3Szrj   uint64_t
do_checkpoint()139*fae548d3Szrj   do_checkpoint() const
140*fae548d3Szrj   { return this->track_relocs_.checkpoint(); }
141*fae548d3Szrj 
142*fae548d3Szrj   // Reset the current position to the CHECKPOINT.
143*fae548d3Szrj   void
do_reset(uint64_t checkpoint)144*fae548d3Szrj   do_reset(uint64_t checkpoint)
145*fae548d3Szrj   { this->track_relocs_.reset(checkpoint); }
146*fae548d3Szrj 
147*fae548d3Szrj  private:
148*fae548d3Szrj   typedef typename elfcpp::Elf_types<size>::Elf_Addr Address;
149*fae548d3Szrj 
150*fae548d3Szrj   // Return the section index of symbol SYMNDX, and copy its value to *VALUE.
151*fae548d3Szrj   // Set *IS_ORDINARY true if the section index is an ordinary section index.
152*fae548d3Szrj   unsigned int
153*fae548d3Szrj   symbol_section(unsigned int symndx, Address* value, bool* is_ordinary);
154*fae548d3Szrj 
155*fae548d3Szrj   // The object file.
156*fae548d3Szrj   Object* object_;
157*fae548d3Szrj   // The ELF symbol table.
158*fae548d3Szrj   const unsigned char* symtab_;
159*fae548d3Szrj   // The size of the ELF symbol table.
160*fae548d3Szrj   off_t symtab_size_;
161*fae548d3Szrj   // Type of the relocation section (SHT_REL or SHT_RELA).
162*fae548d3Szrj   unsigned int reloc_type_;
163*fae548d3Szrj   // Relocations for the referring section.
164*fae548d3Szrj   Track_relocs<size, big_endian> track_relocs_;
165*fae548d3Szrj };
166*fae548d3Szrj 
167*fae548d3Szrj // This class is used to read the abbreviations table from the
168*fae548d3Szrj // .debug_abbrev section of the object file.
169*fae548d3Szrj 
170*fae548d3Szrj class Dwarf_abbrev_table
171*fae548d3Szrj {
172*fae548d3Szrj  public:
173*fae548d3Szrj   // An attribute list entry.
174*fae548d3Szrj   struct Attribute
175*fae548d3Szrj   {
AttributeAttribute176*fae548d3Szrj     Attribute(unsigned int a, unsigned int f)
177*fae548d3Szrj       : attr(a), form(f)
178*fae548d3Szrj     { }
179*fae548d3Szrj     unsigned int attr;
180*fae548d3Szrj     unsigned int form;
181*fae548d3Szrj   };
182*fae548d3Szrj 
183*fae548d3Szrj   // An abbrev code entry.
184*fae548d3Szrj   struct Abbrev_code
185*fae548d3Szrj   {
Abbrev_codeAbbrev_code186*fae548d3Szrj     Abbrev_code(unsigned int t, bool hc)
187*fae548d3Szrj       : tag(t), has_children(hc), has_sibling_attribute(false), attributes()
188*fae548d3Szrj     {
189*fae548d3Szrj       this->attributes.reserve(10);
190*fae548d3Szrj     }
191*fae548d3Szrj 
192*fae548d3Szrj     void
add_attributeAbbrev_code193*fae548d3Szrj     add_attribute(unsigned int attr, unsigned int form)
194*fae548d3Szrj     {
195*fae548d3Szrj       this->attributes.push_back(Attribute(attr, form));
196*fae548d3Szrj     }
197*fae548d3Szrj 
198*fae548d3Szrj     // The DWARF tag.
199*fae548d3Szrj     unsigned int tag;
200*fae548d3Szrj     // True if the DIE has children.
201*fae548d3Szrj     bool has_children : 1;
202*fae548d3Szrj     // True if the DIE has a sibling attribute.
203*fae548d3Szrj     bool has_sibling_attribute : 1;
204*fae548d3Szrj     // The list of attributes and forms.
205*fae548d3Szrj     std::vector<Attribute> attributes;
206*fae548d3Szrj   };
207*fae548d3Szrj 
Dwarf_abbrev_table()208*fae548d3Szrj   Dwarf_abbrev_table()
209*fae548d3Szrj     : abbrev_shndx_(0), abbrev_offset_(0), buffer_(NULL), buffer_end_(NULL),
210*fae548d3Szrj       owns_buffer_(false), buffer_pos_(NULL), high_abbrev_codes_()
211*fae548d3Szrj   {
212*fae548d3Szrj     memset(this->low_abbrev_codes_, 0, sizeof(this->low_abbrev_codes_));
213*fae548d3Szrj   }
214*fae548d3Szrj 
~Dwarf_abbrev_table()215*fae548d3Szrj   ~Dwarf_abbrev_table()
216*fae548d3Szrj   {
217*fae548d3Szrj     if (this->owns_buffer_ && this->buffer_ != NULL)
218*fae548d3Szrj       delete[] this->buffer_;
219*fae548d3Szrj     this->clear_abbrev_codes();
220*fae548d3Szrj   }
221*fae548d3Szrj 
222*fae548d3Szrj   // Read the abbrev table from an object file.
223*fae548d3Szrj   bool
read_abbrevs(Relobj * object,unsigned int abbrev_shndx,off_t abbrev_offset)224*fae548d3Szrj   read_abbrevs(Relobj* object,
225*fae548d3Szrj 	       unsigned int abbrev_shndx,
226*fae548d3Szrj 	       off_t abbrev_offset)
227*fae548d3Szrj   {
228*fae548d3Szrj     // If we've already read this abbrev table, return immediately.
229*fae548d3Szrj     if (this->abbrev_shndx_ > 0
230*fae548d3Szrj 	&& this->abbrev_shndx_ == abbrev_shndx
231*fae548d3Szrj 	&& this->abbrev_offset_ == abbrev_offset)
232*fae548d3Szrj       return true;
233*fae548d3Szrj     return this->do_read_abbrevs(object, abbrev_shndx, abbrev_offset);
234*fae548d3Szrj   }
235*fae548d3Szrj 
236*fae548d3Szrj   // Return the abbrev code entry for CODE.  This is a fast path for
237*fae548d3Szrj   // abbrev codes that are in the direct lookup table.  If not found
238*fae548d3Szrj   // there, we call do_get_abbrev() to do the hard work.
239*fae548d3Szrj   const Abbrev_code*
get_abbrev(unsigned int code)240*fae548d3Szrj   get_abbrev(unsigned int code)
241*fae548d3Szrj   {
242*fae548d3Szrj     if (code < this->low_abbrev_code_max_
243*fae548d3Szrj 	&& this->low_abbrev_codes_[code] != NULL)
244*fae548d3Szrj       return this->low_abbrev_codes_[code];
245*fae548d3Szrj     return this->do_get_abbrev(code);
246*fae548d3Szrj   }
247*fae548d3Szrj 
248*fae548d3Szrj  private:
249*fae548d3Szrj   // Read the abbrev table from an object file.
250*fae548d3Szrj   bool
251*fae548d3Szrj   do_read_abbrevs(Relobj* object,
252*fae548d3Szrj 		  unsigned int abbrev_shndx,
253*fae548d3Szrj 		  off_t abbrev_offset);
254*fae548d3Szrj 
255*fae548d3Szrj   // Lookup the abbrev code entry for CODE.
256*fae548d3Szrj   const Abbrev_code*
257*fae548d3Szrj   do_get_abbrev(unsigned int code);
258*fae548d3Szrj 
259*fae548d3Szrj   // Store an abbrev code entry for CODE.
260*fae548d3Szrj   void
store_abbrev(unsigned int code,const Abbrev_code * entry)261*fae548d3Szrj   store_abbrev(unsigned int code, const Abbrev_code* entry)
262*fae548d3Szrj   {
263*fae548d3Szrj     if (code < this->low_abbrev_code_max_)
264*fae548d3Szrj       this->low_abbrev_codes_[code] = entry;
265*fae548d3Szrj     else
266*fae548d3Szrj       this->high_abbrev_codes_[code] = entry;
267*fae548d3Szrj   }
268*fae548d3Szrj 
269*fae548d3Szrj   // Clear the abbrev code table and release the memory it uses.
270*fae548d3Szrj   void
271*fae548d3Szrj   clear_abbrev_codes();
272*fae548d3Szrj 
273*fae548d3Szrj   typedef Unordered_map<unsigned int, const Abbrev_code*> Abbrev_code_table;
274*fae548d3Szrj 
275*fae548d3Szrj   // The section index of the current abbrev table.
276*fae548d3Szrj   unsigned int abbrev_shndx_;
277*fae548d3Szrj   // The offset within the section of the current abbrev table.
278*fae548d3Szrj   off_t abbrev_offset_;
279*fae548d3Szrj   // The buffer containing the .debug_abbrev section.
280*fae548d3Szrj   const unsigned char* buffer_;
281*fae548d3Szrj   const unsigned char* buffer_end_;
282*fae548d3Szrj   // True if this object owns the buffer and needs to delete it.
283*fae548d3Szrj   bool owns_buffer_;
284*fae548d3Szrj   // Pointer to the current position in the buffer.
285*fae548d3Szrj   const unsigned char* buffer_pos_;
286*fae548d3Szrj   // The table of abbrev codes.
287*fae548d3Szrj   // We use a direct-lookup array for low abbrev codes,
288*fae548d3Szrj   // and store the rest in a hash table.
289*fae548d3Szrj   static const unsigned int low_abbrev_code_max_ = 256;
290*fae548d3Szrj   const Abbrev_code* low_abbrev_codes_[low_abbrev_code_max_];
291*fae548d3Szrj   Abbrev_code_table high_abbrev_codes_;
292*fae548d3Szrj };
293*fae548d3Szrj 
294*fae548d3Szrj // A DWARF range list.  The start and end offsets are relative
295*fae548d3Szrj // to the input section SHNDX.  Each range must lie entirely
296*fae548d3Szrj // within a single section.
297*fae548d3Szrj 
298*fae548d3Szrj class Dwarf_range_list
299*fae548d3Szrj {
300*fae548d3Szrj  public:
301*fae548d3Szrj   struct Range
302*fae548d3Szrj   {
RangeRange303*fae548d3Szrj     Range(unsigned int a_shndx, off_t a_start, off_t a_end)
304*fae548d3Szrj       : shndx(a_shndx), start(a_start), end(a_end)
305*fae548d3Szrj     { }
306*fae548d3Szrj 
307*fae548d3Szrj     unsigned int shndx;
308*fae548d3Szrj     off_t start;
309*fae548d3Szrj     off_t end;
310*fae548d3Szrj   };
311*fae548d3Szrj 
Dwarf_range_list()312*fae548d3Szrj   Dwarf_range_list()
313*fae548d3Szrj     : range_list_()
314*fae548d3Szrj   { }
315*fae548d3Szrj 
316*fae548d3Szrj   void
add(unsigned int shndx,off_t start,off_t end)317*fae548d3Szrj   add(unsigned int shndx, off_t start, off_t end)
318*fae548d3Szrj   { this->range_list_.push_back(Range(shndx, start, end)); }
319*fae548d3Szrj 
320*fae548d3Szrj   size_t
size()321*fae548d3Szrj   size() const
322*fae548d3Szrj   { return this->range_list_.size(); }
323*fae548d3Szrj 
324*fae548d3Szrj   const Range&
325*fae548d3Szrj   operator[](off_t i) const
326*fae548d3Szrj   { return this->range_list_[i]; }
327*fae548d3Szrj 
328*fae548d3Szrj  private:
329*fae548d3Szrj   std::vector<Range> range_list_;
330*fae548d3Szrj };
331*fae548d3Szrj 
332*fae548d3Szrj // This class is used to read the ranges table from the
333*fae548d3Szrj // .debug_ranges section of the object file.
334*fae548d3Szrj 
335*fae548d3Szrj class Dwarf_ranges_table
336*fae548d3Szrj {
337*fae548d3Szrj  public:
Dwarf_ranges_table(Dwarf_info_reader * dwinfo)338*fae548d3Szrj   Dwarf_ranges_table(Dwarf_info_reader* dwinfo)
339*fae548d3Szrj     : dwinfo_(dwinfo), ranges_shndx_(0), ranges_buffer_(NULL),
340*fae548d3Szrj       ranges_buffer_end_(NULL), owns_ranges_buffer_(false),
341*fae548d3Szrj       ranges_reloc_mapper_(NULL), reloc_type_(0), output_section_offset_(0)
342*fae548d3Szrj   { }
343*fae548d3Szrj 
~Dwarf_ranges_table()344*fae548d3Szrj   ~Dwarf_ranges_table()
345*fae548d3Szrj   {
346*fae548d3Szrj     if (this->owns_ranges_buffer_ && this->ranges_buffer_ != NULL)
347*fae548d3Szrj       delete[] this->ranges_buffer_;
348*fae548d3Szrj     if (this->ranges_reloc_mapper_ != NULL)
349*fae548d3Szrj       delete this->ranges_reloc_mapper_;
350*fae548d3Szrj   }
351*fae548d3Szrj 
352*fae548d3Szrj   // Read the ranges table from an object file.
353*fae548d3Szrj   bool
354*fae548d3Szrj   read_ranges_table(Relobj* object,
355*fae548d3Szrj 		    const unsigned char* symtab,
356*fae548d3Szrj 		    off_t symtab_size,
357*fae548d3Szrj 		    unsigned int ranges_shndx);
358*fae548d3Szrj 
359*fae548d3Szrj   // Read the range table from an object file.
360*fae548d3Szrj   Dwarf_range_list*
361*fae548d3Szrj   read_range_list(Relobj* object,
362*fae548d3Szrj 		  const unsigned char* symtab,
363*fae548d3Szrj 		  off_t symtab_size,
364*fae548d3Szrj 		  unsigned int address_size,
365*fae548d3Szrj 		  unsigned int ranges_shndx,
366*fae548d3Szrj 		  off_t ranges_offset);
367*fae548d3Szrj 
368*fae548d3Szrj   // Look for a relocation at offset OFF in the range table,
369*fae548d3Szrj   // and return the section index and offset of the target.
370*fae548d3Szrj   unsigned int
371*fae548d3Szrj   lookup_reloc(off_t off, off_t* target_off);
372*fae548d3Szrj 
373*fae548d3Szrj  private:
374*fae548d3Szrj   // The Dwarf_info_reader, for reading data.
375*fae548d3Szrj   Dwarf_info_reader* dwinfo_;
376*fae548d3Szrj   // The section index of the ranges table.
377*fae548d3Szrj   unsigned int ranges_shndx_;
378*fae548d3Szrj   // The buffer containing the .debug_ranges section.
379*fae548d3Szrj   const unsigned char* ranges_buffer_;
380*fae548d3Szrj   const unsigned char* ranges_buffer_end_;
381*fae548d3Szrj   // True if this object owns the buffer and needs to delete it.
382*fae548d3Szrj   bool owns_ranges_buffer_;
383*fae548d3Szrj   // Relocation mapper for the .debug_ranges section.
384*fae548d3Szrj   Elf_reloc_mapper* ranges_reloc_mapper_;
385*fae548d3Szrj   // Type of the relocation section (SHT_REL or SHT_RELA).
386*fae548d3Szrj   unsigned int reloc_type_;
387*fae548d3Szrj   // For incremental update links, this will hold the offset of the
388*fae548d3Szrj   // input section within the output section.  Offsets read from
389*fae548d3Szrj   // relocated data will be relative to the output section, and need
390*fae548d3Szrj   // to be corrected before reading data from the input section.
391*fae548d3Szrj   uint64_t output_section_offset_;
392*fae548d3Szrj };
393*fae548d3Szrj 
394*fae548d3Szrj // This class is used to read the pubnames and pubtypes tables from the
395*fae548d3Szrj // .debug_pubnames and .debug_pubtypes sections of the object file.
396*fae548d3Szrj 
397*fae548d3Szrj class Dwarf_pubnames_table
398*fae548d3Szrj {
399*fae548d3Szrj  public:
Dwarf_pubnames_table(Dwarf_info_reader * dwinfo,bool is_pubtypes)400*fae548d3Szrj   Dwarf_pubnames_table(Dwarf_info_reader* dwinfo, bool is_pubtypes)
401*fae548d3Szrj     : dwinfo_(dwinfo), buffer_(NULL), buffer_end_(NULL), owns_buffer_(false),
402*fae548d3Szrj       offset_size_(0), pinfo_(NULL), end_of_table_(NULL),
403*fae548d3Szrj       is_pubtypes_(is_pubtypes), is_gnu_style_(false),
404*fae548d3Szrj       unit_length_(0), cu_offset_(0)
405*fae548d3Szrj   { }
406*fae548d3Szrj 
~Dwarf_pubnames_table()407*fae548d3Szrj   ~Dwarf_pubnames_table()
408*fae548d3Szrj   {
409*fae548d3Szrj     if (this->owns_buffer_ && this->buffer_ != NULL)
410*fae548d3Szrj       delete[] this->buffer_;
411*fae548d3Szrj   }
412*fae548d3Szrj 
413*fae548d3Szrj   // Read the pubnames section from the object file, using the symbol
414*fae548d3Szrj   // table for relocating it.
415*fae548d3Szrj   bool
416*fae548d3Szrj   read_section(Relobj* object, const unsigned char* symbol_table,
417*fae548d3Szrj                off_t symtab_size);
418*fae548d3Szrj 
419*fae548d3Szrj   // Read the header for the set at OFFSET.
420*fae548d3Szrj   bool
421*fae548d3Szrj   read_header(off_t offset);
422*fae548d3Szrj 
423*fae548d3Szrj   // Return the offset to the cu within the info or types section.
424*fae548d3Szrj   off_t
cu_offset()425*fae548d3Szrj   cu_offset()
426*fae548d3Szrj   { return this->cu_offset_; }
427*fae548d3Szrj 
428*fae548d3Szrj   // Return the size of this subsection of the table.  The unit length
429*fae548d3Szrj   // doesn't include the size of its own field.
430*fae548d3Szrj   off_t
subsection_size()431*fae548d3Szrj   subsection_size()
432*fae548d3Szrj   { return this->unit_length_; }
433*fae548d3Szrj 
434*fae548d3Szrj   // Read the next name from the set.  If the pubname table is gnu-style,
435*fae548d3Szrj   // FLAG_BYTE is set to the high-byte of a gdb_index version 7 cu_index.
436*fae548d3Szrj   const char*
437*fae548d3Szrj   next_name(uint8_t* flag_byte);
438*fae548d3Szrj 
439*fae548d3Szrj  private:
440*fae548d3Szrj   // The Dwarf_info_reader, for reading data.
441*fae548d3Szrj   Dwarf_info_reader* dwinfo_;
442*fae548d3Szrj   // The buffer containing the .debug_ranges section.
443*fae548d3Szrj   const unsigned char* buffer_;
444*fae548d3Szrj   const unsigned char* buffer_end_;
445*fae548d3Szrj   // True if this object owns the buffer and needs to delete it.
446*fae548d3Szrj   bool owns_buffer_;
447*fae548d3Szrj   // The size of a DWARF offset for the current set.
448*fae548d3Szrj   unsigned int offset_size_;
449*fae548d3Szrj   // The current position within the buffer.
450*fae548d3Szrj   const unsigned char* pinfo_;
451*fae548d3Szrj   // The end of the current pubnames table.
452*fae548d3Szrj   const unsigned char* end_of_table_;
453*fae548d3Szrj   // TRUE if this is a .debug_pubtypes section.
454*fae548d3Szrj   bool is_pubtypes_;
455*fae548d3Szrj   // Gnu-style pubnames table. This style has an extra flag byte between the
456*fae548d3Szrj   // offset and the name, and is used for generating version 7 of gdb-index.
457*fae548d3Szrj   bool is_gnu_style_;
458*fae548d3Szrj   // Fields read from the header.
459*fae548d3Szrj   uint64_t unit_length_;
460*fae548d3Szrj   off_t cu_offset_;
461*fae548d3Szrj 
462*fae548d3Szrj   // Track relocations for this table so we can find the CUs that
463*fae548d3Szrj   // correspond to the subsections.
464*fae548d3Szrj   Elf_reloc_mapper* reloc_mapper_;
465*fae548d3Szrj   // Type of the relocation section (SHT_REL or SHT_RELA).
466*fae548d3Szrj   unsigned int reloc_type_;
467*fae548d3Szrj };
468*fae548d3Szrj 
469*fae548d3Szrj // This class represents a DWARF Debug Info Entry (DIE).
470*fae548d3Szrj 
471*fae548d3Szrj class Dwarf_die
472*fae548d3Szrj {
473*fae548d3Szrj  public:
474*fae548d3Szrj   // An attribute value.
475*fae548d3Szrj   struct Attribute_value
476*fae548d3Szrj   {
477*fae548d3Szrj     unsigned int attr;
478*fae548d3Szrj     unsigned int form;
479*fae548d3Szrj     union
480*fae548d3Szrj     {
481*fae548d3Szrj       int64_t intval;
482*fae548d3Szrj       uint64_t uintval;
483*fae548d3Szrj       const char* stringval;
484*fae548d3Szrj       const unsigned char* blockval;
485*fae548d3Szrj       off_t refval;
486*fae548d3Szrj     } val;
487*fae548d3Szrj     union
488*fae548d3Szrj     {
489*fae548d3Szrj       // Section index for reference forms.
490*fae548d3Szrj       unsigned int shndx;
491*fae548d3Szrj       // Block length for block forms.
492*fae548d3Szrj       unsigned int blocklen;
493*fae548d3Szrj       // Attribute offset for DW_FORM_strp.
494*fae548d3Szrj       unsigned int attr_off;
495*fae548d3Szrj     } aux;
496*fae548d3Szrj   };
497*fae548d3Szrj 
498*fae548d3Szrj   // A list of attribute values.
499*fae548d3Szrj   typedef std::vector<Attribute_value> Attributes;
500*fae548d3Szrj 
501*fae548d3Szrj   Dwarf_die(Dwarf_info_reader* dwinfo,
502*fae548d3Szrj 	    off_t die_offset,
503*fae548d3Szrj 	    Dwarf_die* parent);
504*fae548d3Szrj 
505*fae548d3Szrj   // Return the DWARF tag for this DIE.
506*fae548d3Szrj   unsigned int
tag()507*fae548d3Szrj   tag() const
508*fae548d3Szrj   {
509*fae548d3Szrj     if (this->abbrev_code_ == NULL)
510*fae548d3Szrj       return 0;
511*fae548d3Szrj     return this->abbrev_code_->tag;
512*fae548d3Szrj   }
513*fae548d3Szrj 
514*fae548d3Szrj   // Return true if this DIE has children.
515*fae548d3Szrj   bool
has_children()516*fae548d3Szrj   has_children() const
517*fae548d3Szrj   {
518*fae548d3Szrj     gold_assert(this->abbrev_code_ != NULL);
519*fae548d3Szrj     return this->abbrev_code_->has_children;
520*fae548d3Szrj   }
521*fae548d3Szrj 
522*fae548d3Szrj   // Return true if this DIE has a sibling attribute.
523*fae548d3Szrj   bool
has_sibling_attribute()524*fae548d3Szrj   has_sibling_attribute() const
525*fae548d3Szrj   {
526*fae548d3Szrj     gold_assert(this->abbrev_code_ != NULL);
527*fae548d3Szrj     return this->abbrev_code_->has_sibling_attribute;
528*fae548d3Szrj   }
529*fae548d3Szrj 
530*fae548d3Szrj   // Return the value of attribute ATTR.
531*fae548d3Szrj   const Attribute_value*
532*fae548d3Szrj   attribute(unsigned int attr);
533*fae548d3Szrj 
534*fae548d3Szrj   // Return the value of the DW_AT_name attribute.
535*fae548d3Szrj   const char*
name()536*fae548d3Szrj   name()
537*fae548d3Szrj   {
538*fae548d3Szrj     if (this->name_ == NULL)
539*fae548d3Szrj       this->set_name();
540*fae548d3Szrj     return this->name_;
541*fae548d3Szrj   }
542*fae548d3Szrj 
543*fae548d3Szrj   // Return the value of the DW_AT_linkage_name
544*fae548d3Szrj   // or DW_AT_MIPS_linkage_name attribute.
545*fae548d3Szrj   const char*
linkage_name()546*fae548d3Szrj   linkage_name()
547*fae548d3Szrj   {
548*fae548d3Szrj     if (this->linkage_name_ == NULL)
549*fae548d3Szrj       this->set_linkage_name();
550*fae548d3Szrj     return this->linkage_name_;
551*fae548d3Szrj   }
552*fae548d3Szrj 
553*fae548d3Szrj   // Return the value of the DW_AT_specification attribute.
554*fae548d3Szrj   off_t
specification()555*fae548d3Szrj   specification()
556*fae548d3Szrj   {
557*fae548d3Szrj     if (!this->attributes_read_)
558*fae548d3Szrj       this->read_attributes();
559*fae548d3Szrj     return this->specification_;
560*fae548d3Szrj   }
561*fae548d3Szrj 
562*fae548d3Szrj   // Return the value of the DW_AT_abstract_origin attribute.
563*fae548d3Szrj   off_t
abstract_origin()564*fae548d3Szrj   abstract_origin()
565*fae548d3Szrj   {
566*fae548d3Szrj     if (!this->attributes_read_)
567*fae548d3Szrj       this->read_attributes();
568*fae548d3Szrj     return this->abstract_origin_;
569*fae548d3Szrj   }
570*fae548d3Szrj 
571*fae548d3Szrj   // Return the value of attribute ATTR as a string.
572*fae548d3Szrj   const char*
573*fae548d3Szrj   string_attribute(unsigned int attr);
574*fae548d3Szrj 
575*fae548d3Szrj   // Return the value of attribute ATTR as an integer.
576*fae548d3Szrj   int64_t
577*fae548d3Szrj   int_attribute(unsigned int attr);
578*fae548d3Szrj 
579*fae548d3Szrj   // Return the value of attribute ATTR as an unsigned integer.
580*fae548d3Szrj   uint64_t
581*fae548d3Szrj   uint_attribute(unsigned int attr);
582*fae548d3Szrj 
583*fae548d3Szrj   // Return the value of attribute ATTR as a reference.
584*fae548d3Szrj   off_t
585*fae548d3Szrj   ref_attribute(unsigned int attr, unsigned int* shndx);
586*fae548d3Szrj 
587*fae548d3Szrj   // Return the value of attribute ATTR as a address.
588*fae548d3Szrj   off_t
589*fae548d3Szrj   address_attribute(unsigned int attr, unsigned int* shndx);
590*fae548d3Szrj 
591*fae548d3Szrj   // Return the value of attribute ATTR as a flag.
592*fae548d3Szrj   bool
flag_attribute(unsigned int attr)593*fae548d3Szrj   flag_attribute(unsigned int attr)
594*fae548d3Szrj   { return this->int_attribute(attr) != 0; }
595*fae548d3Szrj 
596*fae548d3Szrj   // Return true if this DIE is a declaration.
597*fae548d3Szrj   bool
is_declaration()598*fae548d3Szrj   is_declaration()
599*fae548d3Szrj   { return this->flag_attribute(elfcpp::DW_AT_declaration); }
600*fae548d3Szrj 
601*fae548d3Szrj   // Return the parent of this DIE.
602*fae548d3Szrj   Dwarf_die*
parent()603*fae548d3Szrj   parent() const
604*fae548d3Szrj   { return this->parent_; }
605*fae548d3Szrj 
606*fae548d3Szrj   // Return the offset of this DIE.
607*fae548d3Szrj   off_t
offset()608*fae548d3Szrj   offset() const
609*fae548d3Szrj   { return this->die_offset_; }
610*fae548d3Szrj 
611*fae548d3Szrj   // Return the offset of this DIE's first child.
612*fae548d3Szrj   off_t
613*fae548d3Szrj   child_offset();
614*fae548d3Szrj 
615*fae548d3Szrj   // Set the offset of this DIE's next sibling.
616*fae548d3Szrj   void
set_sibling_offset(off_t sibling_offset)617*fae548d3Szrj   set_sibling_offset(off_t sibling_offset)
618*fae548d3Szrj   { this->sibling_offset_ = sibling_offset; }
619*fae548d3Szrj 
620*fae548d3Szrj   // Return the offset of this DIE's next sibling.
621*fae548d3Szrj   off_t
622*fae548d3Szrj   sibling_offset();
623*fae548d3Szrj 
624*fae548d3Szrj  private:
625*fae548d3Szrj   typedef Dwarf_abbrev_table::Abbrev_code Abbrev_code;
626*fae548d3Szrj 
627*fae548d3Szrj   // Read all the attributes of the DIE.
628*fae548d3Szrj   bool
629*fae548d3Szrj   read_attributes();
630*fae548d3Szrj 
631*fae548d3Szrj   // Set the name of the DIE if present.
632*fae548d3Szrj   void
633*fae548d3Szrj   set_name();
634*fae548d3Szrj 
635*fae548d3Szrj   // Set the linkage name if present.
636*fae548d3Szrj   void
637*fae548d3Szrj   set_linkage_name();
638*fae548d3Szrj 
639*fae548d3Szrj   // Skip all the attributes of the DIE and return the offset
640*fae548d3Szrj   // of the next DIE.
641*fae548d3Szrj   off_t
642*fae548d3Szrj   skip_attributes();
643*fae548d3Szrj 
644*fae548d3Szrj   // The Dwarf_info_reader, for reading attributes.
645*fae548d3Szrj   Dwarf_info_reader* dwinfo_;
646*fae548d3Szrj   // The parent of this DIE.
647*fae548d3Szrj   Dwarf_die* parent_;
648*fae548d3Szrj   // Offset of this DIE within its compilation unit.
649*fae548d3Szrj   off_t die_offset_;
650*fae548d3Szrj   // Offset of the first attribute, relative to the beginning of the DIE.
651*fae548d3Szrj   off_t attr_offset_;
652*fae548d3Szrj   // Offset of the first child, relative to the compilation unit.
653*fae548d3Szrj   off_t child_offset_;
654*fae548d3Szrj   // Offset of the next sibling, relative to the compilation unit.
655*fae548d3Szrj   off_t sibling_offset_;
656*fae548d3Szrj   // The abbreviation table entry.
657*fae548d3Szrj   const Abbrev_code* abbrev_code_;
658*fae548d3Szrj   // The list of attributes.
659*fae548d3Szrj   Attributes attributes_;
660*fae548d3Szrj   // True if the attributes have been read.
661*fae548d3Szrj   bool attributes_read_;
662*fae548d3Szrj   // The following fields hold common attributes to avoid a linear
663*fae548d3Szrj   // search through the attribute list.
664*fae548d3Szrj   // The DIE name (DW_AT_name).
665*fae548d3Szrj   const char* name_;
666*fae548d3Szrj   // Offset of the name in the string table (for DW_FORM_strp).
667*fae548d3Szrj   off_t name_off_;
668*fae548d3Szrj   // The linkage name (DW_AT_linkage_name or DW_AT_MIPS_linkage_name).
669*fae548d3Szrj   const char* linkage_name_;
670*fae548d3Szrj   // Offset of the linkage name in the string table (for DW_FORM_strp).
671*fae548d3Szrj   off_t linkage_name_off_;
672*fae548d3Szrj   // Section index of the string table (for DW_FORM_strp).
673*fae548d3Szrj   unsigned int string_shndx_;
674*fae548d3Szrj   // The value of a DW_AT_specification attribute.
675*fae548d3Szrj   off_t specification_;
676*fae548d3Szrj   // The value of a DW_AT_abstract_origin attribute.
677*fae548d3Szrj   off_t abstract_origin_;
678*fae548d3Szrj };
679*fae548d3Szrj 
680*fae548d3Szrj // This class is used to read the debug info from the .debug_info
681*fae548d3Szrj // or .debug_types sections.  This is a base class that implements
682*fae548d3Szrj // the generic parsing of the compilation unit header and DIE
683*fae548d3Szrj // structure.  The parse() method parses the entire section, and
684*fae548d3Szrj // calls the various visit_xxx() methods for each header.  Clients
685*fae548d3Szrj // should derive a new class from this one and implement the
686*fae548d3Szrj // visit_compilation_unit() and visit_type_unit() functions.
687*fae548d3Szrj 
688*fae548d3Szrj class Dwarf_info_reader
689*fae548d3Szrj {
690*fae548d3Szrj  public:
Dwarf_info_reader(bool is_type_unit,Relobj * object,const unsigned char * symtab,off_t symtab_size,unsigned int shndx,unsigned int reloc_shndx,unsigned int reloc_type)691*fae548d3Szrj   Dwarf_info_reader(bool is_type_unit,
692*fae548d3Szrj 		    Relobj* object,
693*fae548d3Szrj 		    const unsigned char* symtab,
694*fae548d3Szrj 		    off_t symtab_size,
695*fae548d3Szrj 		    unsigned int shndx,
696*fae548d3Szrj 		    unsigned int reloc_shndx,
697*fae548d3Szrj 		    unsigned int reloc_type)
698*fae548d3Szrj     : is_type_unit_(is_type_unit), object_(object), symtab_(symtab),
699*fae548d3Szrj       symtab_size_(symtab_size), shndx_(shndx), reloc_shndx_(reloc_shndx),
700*fae548d3Szrj       reloc_type_(reloc_type), abbrev_shndx_(0), string_shndx_(0),
701*fae548d3Szrj       buffer_(NULL), buffer_end_(NULL), cu_offset_(0), cu_length_(0),
702*fae548d3Szrj       offset_size_(0), address_size_(0), cu_version_(0),
703*fae548d3Szrj       abbrev_table_(), ranges_table_(this),
704*fae548d3Szrj       reloc_mapper_(NULL), string_buffer_(NULL), string_buffer_end_(NULL),
705*fae548d3Szrj       owns_string_buffer_(false), string_output_section_offset_(0)
706*fae548d3Szrj   { }
707*fae548d3Szrj 
708*fae548d3Szrj   virtual
~Dwarf_info_reader()709*fae548d3Szrj   ~Dwarf_info_reader()
710*fae548d3Szrj   {
711*fae548d3Szrj     if (this->reloc_mapper_ != NULL)
712*fae548d3Szrj       delete this->reloc_mapper_;
713*fae548d3Szrj     if (this->owns_string_buffer_ && this->string_buffer_ != NULL)
714*fae548d3Szrj       delete[] this->string_buffer_;
715*fae548d3Szrj   }
716*fae548d3Szrj 
717*fae548d3Szrj   // Begin parsing the debug info.  This calls visit_compilation_unit()
718*fae548d3Szrj   // or visit_type_unit() for each compilation or type unit found in the
719*fae548d3Szrj   // section, and visit_die() for each top-level DIE.
720*fae548d3Szrj   void
721*fae548d3Szrj   parse();
722*fae548d3Szrj 
723*fae548d3Szrj   // Return the abbrev code entry for a CODE.
724*fae548d3Szrj   const Dwarf_abbrev_table::Abbrev_code*
get_abbrev(unsigned int code)725*fae548d3Szrj   get_abbrev(unsigned int code)
726*fae548d3Szrj   { return this->abbrev_table_.get_abbrev(code); }
727*fae548d3Szrj 
728*fae548d3Szrj   // Return a pointer to the DWARF info buffer at OFFSET.
729*fae548d3Szrj   const unsigned char*
buffer_at_offset(off_t offset)730*fae548d3Szrj   buffer_at_offset(off_t offset) const
731*fae548d3Szrj   {
732*fae548d3Szrj     const unsigned char* p = this->buffer_ + this->cu_offset_ + offset;
733*fae548d3Szrj     if (this->check_buffer(p + 1))
734*fae548d3Szrj       return p;
735*fae548d3Szrj     return NULL;
736*fae548d3Szrj   }
737*fae548d3Szrj 
738*fae548d3Szrj   // Read a possibly unaligned integer of SIZE.
739*fae548d3Szrj   template <int valsize>
740*fae548d3Szrj   inline typename elfcpp::Valtype_base<valsize>::Valtype
741*fae548d3Szrj   read_from_pointer(const unsigned char* source);
742*fae548d3Szrj 
743*fae548d3Szrj   // Read a possibly unaligned integer of SIZE.  Update SOURCE after read.
744*fae548d3Szrj   template <int valsize>
745*fae548d3Szrj   inline typename elfcpp::Valtype_base<valsize>::Valtype
746*fae548d3Szrj   read_from_pointer(const unsigned char** source);
747*fae548d3Szrj 
748*fae548d3Szrj   // Look for a relocation at offset ATTR_OFF in the dwarf info,
749*fae548d3Szrj   // and return the section index and offset of the target.
750*fae548d3Szrj   unsigned int
751*fae548d3Szrj   lookup_reloc(off_t attr_off, off_t* target_off);
752*fae548d3Szrj 
753*fae548d3Szrj   // Return a string from the DWARF string table.
754*fae548d3Szrj   const char*
755*fae548d3Szrj   get_string(off_t str_off, unsigned int string_shndx);
756*fae548d3Szrj 
757*fae548d3Szrj   // Return the size of a DWARF offset.
758*fae548d3Szrj   unsigned int
offset_size()759*fae548d3Szrj   offset_size() const
760*fae548d3Szrj   { return this->offset_size_; }
761*fae548d3Szrj 
762*fae548d3Szrj   // Return the size of an address.
763*fae548d3Szrj   unsigned int
address_size()764*fae548d3Szrj   address_size() const
765*fae548d3Szrj   { return this->address_size_; }
766*fae548d3Szrj 
767*fae548d3Szrj   // Return the size of a DW_FORM_ref_addr.
768*fae548d3Szrj   // In DWARF v2, this was the size of an address; in DWARF v3 and later,
769*fae548d3Szrj   // it is the size of an DWARF offset.
770*fae548d3Szrj   unsigned int
ref_addr_size()771*fae548d3Szrj   ref_addr_size() const
772*fae548d3Szrj   { return this->cu_version_ > 2 ? this->offset_size_ : this->address_size_; }
773*fae548d3Szrj 
774*fae548d3Szrj   // Set the section index of the .debug_abbrev section.
775*fae548d3Szrj   // We use this if there are no relocations for the .debug_info section.
776*fae548d3Szrj   // If not set, the code parse() routine will search for the section by name.
777*fae548d3Szrj   void
set_abbrev_shndx(unsigned int abbrev_shndx)778*fae548d3Szrj   set_abbrev_shndx(unsigned int abbrev_shndx)
779*fae548d3Szrj   { this->abbrev_shndx_ = abbrev_shndx; }
780*fae548d3Szrj 
781*fae548d3Szrj   // Return a pointer to the object file's ELF symbol table.
782*fae548d3Szrj   const unsigned char*
symtab()783*fae548d3Szrj   symtab() const
784*fae548d3Szrj   { return this->symtab_; }
785*fae548d3Szrj 
786*fae548d3Szrj   // Return the size of the object file's ELF symbol table.
787*fae548d3Szrj   off_t
symtab_size()788*fae548d3Szrj   symtab_size() const
789*fae548d3Szrj   { return this->symtab_size_; }
790*fae548d3Szrj 
791*fae548d3Szrj   // Return the offset of the current compilation unit.
792*fae548d3Szrj   off_t
cu_offset()793*fae548d3Szrj   cu_offset() const
794*fae548d3Szrj   { return this->cu_offset_; }
795*fae548d3Szrj 
796*fae548d3Szrj  protected:
797*fae548d3Szrj   // Begin parsing the debug info.  This calls visit_compilation_unit()
798*fae548d3Szrj   // or visit_type_unit() for each compilation or type unit found in the
799*fae548d3Szrj   // section, and visit_die() for each top-level DIE.
800*fae548d3Szrj   template<bool big_endian>
801*fae548d3Szrj   void
802*fae548d3Szrj   do_parse();
803*fae548d3Szrj 
804*fae548d3Szrj   // The following methods are hooks that are meant to be implemented
805*fae548d3Szrj   // by a derived class.  A default, do-nothing, implementation of
806*fae548d3Szrj   // each is provided for this base class.
807*fae548d3Szrj 
808*fae548d3Szrj   // Visit a compilation unit.
809*fae548d3Szrj   virtual void
810*fae548d3Szrj   visit_compilation_unit(off_t cu_offset, off_t cu_length, Dwarf_die* root_die);
811*fae548d3Szrj 
812*fae548d3Szrj   // Visit a type unit.
813*fae548d3Szrj   virtual void
814*fae548d3Szrj   visit_type_unit(off_t tu_offset, off_t tu_length, off_t type_offset,
815*fae548d3Szrj 		  uint64_t signature, Dwarf_die* root_die);
816*fae548d3Szrj 
817*fae548d3Szrj   // Read the range table.
818*fae548d3Szrj   Dwarf_range_list*
read_range_list(unsigned int ranges_shndx,off_t ranges_offset)819*fae548d3Szrj   read_range_list(unsigned int ranges_shndx, off_t ranges_offset)
820*fae548d3Szrj   {
821*fae548d3Szrj     return this->ranges_table_.read_range_list(this->object_,
822*fae548d3Szrj 					       this->symtab_,
823*fae548d3Szrj 					       this->symtab_size_,
824*fae548d3Szrj 					       this->address_size_,
825*fae548d3Szrj 					       ranges_shndx,
826*fae548d3Szrj 					       ranges_offset);
827*fae548d3Szrj   }
828*fae548d3Szrj 
829*fae548d3Szrj   // Return the object.
830*fae548d3Szrj   Relobj*
object()831*fae548d3Szrj   object() const
832*fae548d3Szrj   { return this->object_; }
833*fae548d3Szrj 
834*fae548d3Szrj   // Checkpoint the relocation tracker.
835*fae548d3Szrj   uint64_t
get_reloc_checkpoint()836*fae548d3Szrj   get_reloc_checkpoint() const
837*fae548d3Szrj   { return this->reloc_mapper_->checkpoint(); }
838*fae548d3Szrj 
839*fae548d3Szrj   // Reset the relocation tracker to the CHECKPOINT.
840*fae548d3Szrj   void
reset_relocs(uint64_t checkpoint)841*fae548d3Szrj   reset_relocs(uint64_t checkpoint)
842*fae548d3Szrj   { this->reloc_mapper_->reset(checkpoint); }
843*fae548d3Szrj 
844*fae548d3Szrj  private:
845*fae548d3Szrj   // Print a warning about a corrupt debug section.
846*fae548d3Szrj   void
847*fae548d3Szrj   warn_corrupt_debug_section() const;
848*fae548d3Szrj 
849*fae548d3Szrj   // Check that P is within the bounds of the current section.
850*fae548d3Szrj   bool
check_buffer(const unsigned char * p)851*fae548d3Szrj   check_buffer(const unsigned char* p) const
852*fae548d3Szrj   {
853*fae548d3Szrj     if (p > this->buffer_ + this->cu_offset_ + this->cu_length_)
854*fae548d3Szrj       {
855*fae548d3Szrj 	this->warn_corrupt_debug_section();
856*fae548d3Szrj 	return false;
857*fae548d3Szrj       }
858*fae548d3Szrj     return true;
859*fae548d3Szrj   }
860*fae548d3Szrj 
861*fae548d3Szrj   // Read the DWARF string table.
862*fae548d3Szrj   bool
read_string_table(unsigned int string_shndx)863*fae548d3Szrj   read_string_table(unsigned int string_shndx)
864*fae548d3Szrj   {
865*fae548d3Szrj     // If we've already read this string table, return immediately.
866*fae548d3Szrj     if (this->string_shndx_ > 0 && this->string_shndx_ == string_shndx)
867*fae548d3Szrj       return true;
868*fae548d3Szrj     if (string_shndx == 0 && this->string_shndx_ > 0)
869*fae548d3Szrj       return true;
870*fae548d3Szrj     return this->do_read_string_table(string_shndx);
871*fae548d3Szrj   }
872*fae548d3Szrj 
873*fae548d3Szrj   bool
874*fae548d3Szrj   do_read_string_table(unsigned int string_shndx);
875*fae548d3Szrj 
876*fae548d3Szrj   // True if this is a type unit; false for a compilation unit.
877*fae548d3Szrj   bool is_type_unit_;
878*fae548d3Szrj   // The object containing the .debug_info or .debug_types input section.
879*fae548d3Szrj   Relobj* object_;
880*fae548d3Szrj   // The ELF symbol table.
881*fae548d3Szrj   const unsigned char* symtab_;
882*fae548d3Szrj   // The size of the ELF symbol table.
883*fae548d3Szrj   off_t symtab_size_;
884*fae548d3Szrj   // Index of the .debug_info or .debug_types section.
885*fae548d3Szrj   unsigned int shndx_;
886*fae548d3Szrj   // Index of the relocation section.
887*fae548d3Szrj   unsigned int reloc_shndx_;
888*fae548d3Szrj   // Type of the relocation section (SHT_REL or SHT_RELA).
889*fae548d3Szrj   unsigned int reloc_type_;
890*fae548d3Szrj   // Index of the .debug_abbrev section (0 if not known).
891*fae548d3Szrj   unsigned int abbrev_shndx_;
892*fae548d3Szrj   // Index of the .debug_str section.
893*fae548d3Szrj   unsigned int string_shndx_;
894*fae548d3Szrj   // The buffer for the debug info.
895*fae548d3Szrj   const unsigned char* buffer_;
896*fae548d3Szrj   const unsigned char* buffer_end_;
897*fae548d3Szrj   // Offset of the current compilation unit.
898*fae548d3Szrj   off_t cu_offset_;
899*fae548d3Szrj   // Length of the current compilation unit.
900*fae548d3Szrj   off_t cu_length_;
901*fae548d3Szrj   // Size of a DWARF offset for the current compilation unit.
902*fae548d3Szrj   unsigned int offset_size_;
903*fae548d3Szrj   // Size of an address for the target architecture.
904*fae548d3Szrj   unsigned int address_size_;
905*fae548d3Szrj   // Compilation unit version number.
906*fae548d3Szrj   unsigned int cu_version_;
907*fae548d3Szrj   // Abbreviations table for current compilation unit.
908*fae548d3Szrj   Dwarf_abbrev_table abbrev_table_;
909*fae548d3Szrj   // Ranges table for the current compilation unit.
910*fae548d3Szrj   Dwarf_ranges_table ranges_table_;
911*fae548d3Szrj   // Relocation mapper for the section.
912*fae548d3Szrj   Elf_reloc_mapper* reloc_mapper_;
913*fae548d3Szrj   // The buffer for the debug string table.
914*fae548d3Szrj   const char* string_buffer_;
915*fae548d3Szrj   const char* string_buffer_end_;
916*fae548d3Szrj   // True if this object owns the buffer and needs to delete it.
917*fae548d3Szrj   bool owns_string_buffer_;
918*fae548d3Szrj   // For incremental update links, this will hold the offset of the
919*fae548d3Szrj   // input .debug_str section within the output section.  Offsets read
920*fae548d3Szrj   // from relocated data will be relative to the output section, and need
921*fae548d3Szrj   // to be corrected before reading data from the input section.
922*fae548d3Szrj   uint64_t string_output_section_offset_;
923*fae548d3Szrj };
924*fae548d3Szrj 
925*fae548d3Szrj // We can't do better than to keep the offsets in a sorted vector.
926*fae548d3Szrj // Here, offset is the key, and file_num/line_num is the value.
927*fae548d3Szrj struct Offset_to_lineno_entry
928*fae548d3Szrj {
929*fae548d3Szrj   off_t offset;
930*fae548d3Szrj   int header_num;  // which file-list to use (i.e. which .o file are we in)
931*fae548d3Szrj   // A pointer into files_.
932*fae548d3Szrj   unsigned int file_num : sizeof(int) * CHAR_BIT - 1;
933*fae548d3Szrj   // True if this was the last entry for the current offset, meaning
934*fae548d3Szrj   // it's the line that actually applies.
935*fae548d3Szrj   unsigned int last_line_for_offset : 1;
936*fae548d3Szrj   // The line number in the source file.  -1 to indicate end-of-function.
937*fae548d3Szrj   int line_num;
938*fae548d3Szrj 
939*fae548d3Szrj   // This sorts by offsets first, and then puts the correct line to
940*fae548d3Szrj   // report for a given offset at the beginning of the run of equal
941*fae548d3Szrj   // offsets (so that asking for 1 line gives the best answer).  This
942*fae548d3Szrj   // is not a total ordering.
943*fae548d3Szrj   bool operator<(const Offset_to_lineno_entry& that) const
944*fae548d3Szrj   {
945*fae548d3Szrj     if (this->offset != that.offset)
946*fae548d3Szrj       return this->offset < that.offset;
947*fae548d3Szrj     // Note the '>' which makes this sort 'true' first.
948*fae548d3Szrj     return this->last_line_for_offset > that.last_line_for_offset;
949*fae548d3Szrj   }
950*fae548d3Szrj };
951*fae548d3Szrj 
952*fae548d3Szrj // This class is used to read the line information from the debugging
953*fae548d3Szrj // section of an object file.
954*fae548d3Szrj 
955*fae548d3Szrj class Dwarf_line_info
956*fae548d3Szrj {
957*fae548d3Szrj  public:
Dwarf_line_info()958*fae548d3Szrj   Dwarf_line_info()
959*fae548d3Szrj   { }
960*fae548d3Szrj 
961*fae548d3Szrj   virtual
~Dwarf_line_info()962*fae548d3Szrj   ~Dwarf_line_info()
963*fae548d3Szrj   { }
964*fae548d3Szrj 
965*fae548d3Szrj   // Given a section number and an offset, returns the associated
966*fae548d3Szrj   // file and line-number, as a string: "file:lineno".  If unable
967*fae548d3Szrj   // to do the mapping, returns the empty string.  You must call
968*fae548d3Szrj   // read_line_mappings() before calling this function.  If
969*fae548d3Szrj   // 'other_lines' is non-NULL, fills that in with other line
970*fae548d3Szrj   // numbers assigned to the same offset.
971*fae548d3Szrj   std::string
addr2line(unsigned int shndx,off_t offset,std::vector<std::string> * other_lines)972*fae548d3Szrj   addr2line(unsigned int shndx, off_t offset,
973*fae548d3Szrj             std::vector<std::string>* other_lines)
974*fae548d3Szrj   { return this->do_addr2line(shndx, offset, other_lines); }
975*fae548d3Szrj 
976*fae548d3Szrj   // A helper function for a single addr2line lookup.  It also keeps a
977*fae548d3Szrj   // cache of the last CACHE_SIZE Dwarf_line_info objects it created;
978*fae548d3Szrj   // set to 0 not to cache at all.  The larger CACHE_SIZE is, the more
979*fae548d3Szrj   // chance this routine won't have to re-create a Dwarf_line_info
980*fae548d3Szrj   // object for its addr2line computation; such creations are slow.
981*fae548d3Szrj   // NOTE: Not thread-safe, so only call from one thread at a time.
982*fae548d3Szrj   static std::string
983*fae548d3Szrj   one_addr2line(Object* object, unsigned int shndx, off_t offset,
984*fae548d3Szrj                 size_t cache_size, std::vector<std::string>* other_lines);
985*fae548d3Szrj 
986*fae548d3Szrj   // This reclaims all the memory that one_addr2line may have cached.
987*fae548d3Szrj   // Use this when you know you will not be calling one_addr2line again.
988*fae548d3Szrj   static void
989*fae548d3Szrj   clear_addr2line_cache();
990*fae548d3Szrj 
991*fae548d3Szrj  private:
992*fae548d3Szrj   virtual std::string
993*fae548d3Szrj   do_addr2line(unsigned int shndx, off_t offset,
994*fae548d3Szrj                std::vector<std::string>* other_lines) = 0;
995*fae548d3Szrj };
996*fae548d3Szrj 
997*fae548d3Szrj template<int size, bool big_endian>
998*fae548d3Szrj class Sized_dwarf_line_info : public Dwarf_line_info
999*fae548d3Szrj {
1000*fae548d3Szrj  public:
1001*fae548d3Szrj   // Initializes a .debug_line reader for a given object file.
1002*fae548d3Szrj   // If SHNDX is specified and non-negative, only read the debug
1003*fae548d3Szrj   // information that pertains to the specified section.
1004*fae548d3Szrj   Sized_dwarf_line_info(Object* object, unsigned int read_shndx = -1U);
1005*fae548d3Szrj 
1006*fae548d3Szrj   virtual
~Sized_dwarf_line_info()1007*fae548d3Szrj   ~Sized_dwarf_line_info()
1008*fae548d3Szrj   {
1009*fae548d3Szrj     if (this->buffer_start_ != NULL)
1010*fae548d3Szrj       delete[] this->buffer_start_;
1011*fae548d3Szrj   }
1012*fae548d3Szrj 
1013*fae548d3Szrj  private:
1014*fae548d3Szrj   std::string
1015*fae548d3Szrj   do_addr2line(unsigned int shndx, off_t offset,
1016*fae548d3Szrj                std::vector<std::string>* other_lines);
1017*fae548d3Szrj 
1018*fae548d3Szrj   // Formats a file and line number to a string like "dirname/filename:lineno".
1019*fae548d3Szrj   std::string
1020*fae548d3Szrj   format_file_lineno(const Offset_to_lineno_entry& lineno) const;
1021*fae548d3Szrj 
1022*fae548d3Szrj   // Start processing line info, and populates the offset_map_.
1023*fae548d3Szrj   // If SHNDX is non-negative, only store debug information that
1024*fae548d3Szrj   // pertains to the specified section.
1025*fae548d3Szrj   void
1026*fae548d3Szrj   read_line_mappings(unsigned int shndx);
1027*fae548d3Szrj 
1028*fae548d3Szrj   // Reads the relocation section associated with .debug_line and
1029*fae548d3Szrj   // stores relocation information in reloc_map_.
1030*fae548d3Szrj   void
1031*fae548d3Szrj   read_relocs();
1032*fae548d3Szrj 
1033*fae548d3Szrj   // Reads the DWARF2/3 header for this line info.  Each takes as input
1034*fae548d3Szrj   // a starting buffer position, and returns the ending position.
1035*fae548d3Szrj   const unsigned char*
1036*fae548d3Szrj   read_header_prolog(const unsigned char* lineptr);
1037*fae548d3Szrj 
1038*fae548d3Szrj   const unsigned char*
1039*fae548d3Szrj   read_header_tables(const unsigned char* lineptr);
1040*fae548d3Szrj 
1041*fae548d3Szrj   // Reads the DWARF2/3 line information.  If shndx is non-negative,
1042*fae548d3Szrj   // discard all line information that doesn't pertain to the given
1043*fae548d3Szrj   // section.
1044*fae548d3Szrj   const unsigned char*
1045*fae548d3Szrj   read_lines(const unsigned char* lineptr, unsigned int shndx);
1046*fae548d3Szrj 
1047*fae548d3Szrj   // Process a single line info opcode at START using the state
1048*fae548d3Szrj   // machine at LSM.  Return true if we should define a line using the
1049*fae548d3Szrj   // current state of the line state machine.  Place the length of the
1050*fae548d3Szrj   // opcode in LEN.
1051*fae548d3Szrj   bool
1052*fae548d3Szrj   process_one_opcode(const unsigned char* start,
1053*fae548d3Szrj                      struct LineStateMachine* lsm, size_t* len);
1054*fae548d3Szrj 
1055*fae548d3Szrj   // Some parts of processing differ depending on whether the input
1056*fae548d3Szrj   // was a .o file or not.
1057*fae548d3Szrj   bool input_is_relobj();
1058*fae548d3Szrj 
1059*fae548d3Szrj   // If we saw anything amiss while parsing, we set this to false.
1060*fae548d3Szrj   // Then addr2line will always fail (rather than return possibly-
1061*fae548d3Szrj   // corrupt data).
1062*fae548d3Szrj   bool data_valid_;
1063*fae548d3Szrj 
1064*fae548d3Szrj   // A DWARF2/3 line info header.  This is not the same size as in the
1065*fae548d3Szrj   // actual file, as the one in the file may have a 32 bit or 64 bit
1066*fae548d3Szrj   // lengths.
1067*fae548d3Szrj 
1068*fae548d3Szrj   struct Dwarf_line_infoHeader
1069*fae548d3Szrj   {
1070*fae548d3Szrj     off_t total_length;
1071*fae548d3Szrj     int version;
1072*fae548d3Szrj     off_t prologue_length;
1073*fae548d3Szrj     int min_insn_length; // insn stands for instruction
1074*fae548d3Szrj     int max_ops_per_insn; // Added in DWARF-4.
1075*fae548d3Szrj     bool default_is_stmt; // stmt stands for statement
1076*fae548d3Szrj     signed char line_base;
1077*fae548d3Szrj     int line_range;
1078*fae548d3Szrj     unsigned char opcode_base;
1079*fae548d3Szrj     std::vector<unsigned char> std_opcode_lengths;
1080*fae548d3Szrj     int offset_size;
1081*fae548d3Szrj   } header_;
1082*fae548d3Szrj 
1083*fae548d3Szrj   // buffer is the buffer for our line info, starting at exactly where
1084*fae548d3Szrj   // the line info to read is.
1085*fae548d3Szrj   const unsigned char* buffer_;
1086*fae548d3Szrj   const unsigned char* buffer_end_;
1087*fae548d3Szrj   // If the buffer was allocated temporarily, and therefore must be
1088*fae548d3Szrj   // deallocated in the dtor, this contains a pointer to the start
1089*fae548d3Szrj   // of the buffer.
1090*fae548d3Szrj   const unsigned char* buffer_start_;
1091*fae548d3Szrj 
1092*fae548d3Szrj   // This has relocations that point into buffer.
1093*fae548d3Szrj   Sized_elf_reloc_mapper<size, big_endian>* reloc_mapper_;
1094*fae548d3Szrj   // The type of the reloc section in track_relocs_--SHT_REL or SHT_RELA.
1095*fae548d3Szrj   unsigned int track_relocs_type_;
1096*fae548d3Szrj 
1097*fae548d3Szrj   // This is used to figure out what section to apply a relocation to.
1098*fae548d3Szrj   const unsigned char* symtab_buffer_;
1099*fae548d3Szrj   section_size_type symtab_buffer_size_;
1100*fae548d3Szrj 
1101*fae548d3Szrj   // Holds the directories and files as we see them.  We have an array
1102*fae548d3Szrj   // of directory-lists, one for each .o file we're reading (usually
1103*fae548d3Szrj   // there will just be one, but there may be more if input is a .so).
1104*fae548d3Szrj   std::vector<std::vector<std::string> > directories_;
1105*fae548d3Szrj   // The first part is an index into directories_, the second the filename.
1106*fae548d3Szrj   std::vector<std::vector< std::pair<int, std::string> > > files_;
1107*fae548d3Szrj 
1108*fae548d3Szrj   // An index into the current directories_ and files_ vectors.
1109*fae548d3Szrj   int current_header_index_;
1110*fae548d3Szrj 
1111*fae548d3Szrj   // A sorted map from offset of the relocation target to the shndx
1112*fae548d3Szrj   // and addend for the relocation.
1113*fae548d3Szrj   typedef std::map<off_t, std::pair<unsigned int, off_t> >
1114*fae548d3Szrj   Reloc_map;
1115*fae548d3Szrj   Reloc_map reloc_map_;
1116*fae548d3Szrj 
1117*fae548d3Szrj   // We have a vector of offset->lineno entries for every input section.
1118*fae548d3Szrj   typedef Unordered_map<unsigned int, std::vector<Offset_to_lineno_entry> >
1119*fae548d3Szrj   Lineno_map;
1120*fae548d3Szrj 
1121*fae548d3Szrj   Lineno_map line_number_map_;
1122*fae548d3Szrj };
1123*fae548d3Szrj 
1124*fae548d3Szrj } // End namespace gold.
1125*fae548d3Szrj 
1126*fae548d3Szrj #endif // !defined(GOLD_DWARF_READER_H)
1127