[12/20] abg-elf-helpers: migrate ppc64 specific helpers

Message ID 20210127125853.886677-13-maennich@google.com
State Committed
Headers
Series Refactor (k)symtab reader |

Commit Message

Matthias Männich Jan. 27, 2021, 12:58 p.m. UTC
  This migrates more helpers to abg-elf-helpers:
  lookup_ppc64_elf_fn_entry_point_address  with dependencies
    read_uint64_from_array_of_bytes
    read_int_from_array_of_bytes

  address_is_in_opd_section  with dependency
    address_is_in_section

read_context::find_opd_section and read_context::opd_section_ are obsolete.

	* src/abg-dwarf-reader.cc (read_context::opd_section_): Delete.
	(read_context::find_opd_section): Delete.
	(read_context::read_uint64_from_array_of_bytes): Delete.
	(read_context::read_int_from_array_of_bytes): Delete.
	(read_context::lookup_ppc64_elf_fn_entry_point_address): Delete.
	(read_context::address_is_in_opd_section): Delete.
	(read_context::address_is_in_section): Delete.
	(read_context::load_symbol_maps_from_symtab_section): Adjust.
	* src/abg-elf-helpers.cc (read_int_from_array_of_bytes): New.
	(read_uint64_from_array_of_bytes): New.
	(lookup_ppc64_elf_fn_entry_point_address): New.
	(address_is_in_section): New.
	(address_is_in_opd_section): New.
	* src/abg-elf-helpers.h
	(lookup_ppc64_elf_fn_entry_point_address): New declaration.
	(address_is_in_opd_section): New declaration.

Reviewed-by: Giuliano Procida <gprocida@google.com>
Signed-off-by: Matthias Maennich <maennich@google.com>
---
 src/abg-dwarf-reader.cc | 208 +---------------------------------------
 src/abg-elf-helpers.cc  | 186 +++++++++++++++++++++++++++++++++++
 src/abg-elf-helpers.h   |   8 ++
 3 files changed, 198 insertions(+), 204 deletions(-)
  

Comments

Dodji Seketeli March 16, 2021, 10:59 a.m. UTC | #1
Hello,

Matthias Maennich <maennich@google.com> a écrit:

[...]

> diff --git a/src/abg-dwarf-reader.cc b/src/abg-dwarf-reader.cc
> index 693fae82040e..b2a7144afe9d 100644
> --- a/src/abg-dwarf-reader.cc
> +++ b/src/abg-dwarf-reader.cc

[...]

>    /// Test if a given function symbol has been exported.
>    ///
>    /// @param symbol_address the address of the symbol we are looking
> @@ -5890,13 +5726,15 @@ public:
>  		      // symbol that are in the .opd section.
>  		      GElf_Addr fn_desc_addr = sym->st_value;
>  		      GElf_Addr fn_entry_point_addr =
> -			lookup_ppc64_elf_fn_entry_point_address(fn_desc_addr);
> +			  lookup_ppc64_elf_fn_entry_point_address(
> +			      elf_handle(), fn_desc_addr);

Please put the opening parenthesis of the function call on the new line.

[...]

> 	* src/abg-dwarf-reader.cc (read_context::opd_section_): Delete.
> 	(read_context::find_opd_section): Delete.
> 	(read_context::read_uint64_from_array_of_bytes): Delete.
> 	(read_context::read_int_from_array_of_bytes): Delete.
> 	(read_context::lookup_ppc64_elf_fn_entry_point_address): Delete.
> 	(read_context::address_is_in_opd_section): Delete.
> 	(read_context::address_is_in_section): Delete.
> 	(read_context::load_symbol_maps_from_symtab_section): Adjust.
> 	* src/abg-elf-helpers.cc (read_int_from_array_of_bytes): New.
> 	(read_uint64_from_array_of_bytes): New.
> 	(lookup_ppc64_elf_fn_entry_point_address): New.
> 	(address_is_in_section): New.
> 	(address_is_in_opd_section): New.
> 	* src/abg-elf-helpers.h
> 	(lookup_ppc64_elf_fn_entry_point_address): New declaration.
> 	(address_is_in_opd_section): New declaration.
>
> Reviewed-by: Giuliano Procida <gprocida@google.com>
> Signed-off-by: Matthias Maennich <maennich@google.com>

OK to apply to master with the above change once the prerequisite
patches are applied.

Thanks!

[...]

Cheers,
  

Patch

diff --git a/src/abg-dwarf-reader.cc b/src/abg-dwarf-reader.cc
index 693fae82040e..b2a7144afe9d 100644
--- a/src/abg-dwarf-reader.cc
+++ b/src/abg-dwarf-reader.cc
@@ -2141,10 +2141,6 @@  public:
   mutable Elf*			elf_handle_;
   string			elf_path_;
   mutable Elf_Scn*		symtab_section_;
-  // The "Official procedure descriptor section, aka .opd", used in
-  // ppc64 elf v1 binaries.  This section contains the procedure
-  // descriptors on that platform.
-  mutable Elf_Scn*		opd_section_;
   Dwarf_Die*			cur_tu_die_;
   mutable dwarf_expr_eval_context	dwarf_expr_eval_context_;
   // A set of maps (one per kind of die source) that associates a decl
@@ -2325,7 +2321,6 @@  public:
     elf_handle_ = 0;
     elf_path_ = elf_path;
     symtab_section_ = 0;
-    opd_section_ = 0;
     cur_tu_die_ =  0;
     exported_decls_builder_ = 0;
 
@@ -5312,19 +5307,6 @@  public:
     return symtab_section_;
   }
 
-  /// Return the "Official Procedure descriptors section."  This
-  /// section is named .opd, and is usually present only on PPC64
-  /// ELFv1 binaries.
-  ///
-  /// @return the .opd section, if found.  Return nil otherwise.
-  Elf_Scn*
-  find_opd_section() const
-  {
-    if (!opd_section_)
-      opd_section_ = elf_helpers::find_opd_section(elf_handle());
-    return opd_section_;
-  }
-
   /// Lookup an elf symbol, referred to by its index, from the .symtab
   /// section.
   ///
@@ -5436,152 +5418,6 @@  public:
     return sym;
   }
 
-  /// Read 8 bytes and convert their value into an uint64_t.
-  ///
-  /// @param bytes the array of bytes to read the next 8 bytes from.
-  /// Note that this array must be at least 8 bytes long.
-  ///
-  /// @param result where to store the resuting uint64_t that was read.
-  ///
-  /// @param is_big_endian if true, read the 8 bytes in Big Endian
-  /// mode, otherwise, read them in Little Endian.
-  ///
-  /// @param true if the 8 bytes could be read, false otherwise.
-  bool
-  read_uint64_from_array_of_bytes(const uint8_t	*bytes,
-				  bool			is_big_endian,
-				  uint64_t		&result) const
-  {
-    return read_int_from_array_of_bytes(bytes, 8, is_big_endian, result);
-  }
-
-  /// Read N bytes and convert their value into an integer type T.
-  ///
-  /// Note that N cannot be bigger than 8 for now. The type passed needs to be
-  /// at least of the size of number_of_bytes.
-  ///
-  /// @param bytes the array of bytes to read the next 8 bytes from.
-  /// Note that this array must be at least 8 bytes long.
-  ///
-  /// @param number_of_bytes the number of bytes to read.  This number
-  /// cannot be bigger than 8.
-  ///
-  /// @param is_big_endian if true, read the 8 bytes in Big Endian
-  /// mode, otherwise, read them in Little Endian.
-  ///
-  /// @param result where to store the resuting integer that was read.
-  ///
-  ///
-  /// @param true if the 8 bytes could be read, false otherwise.
-  template<typename T>
-  bool
-  read_int_from_array_of_bytes(const uint8_t	*bytes,
-			       unsigned char	number_of_bytes,
-			       bool		is_big_endian,
-			       T		&result) const
-  {
-    if (!bytes)
-      return false;
-
-    ABG_ASSERT(number_of_bytes <= 8);
-    ABG_ASSERT(number_of_bytes <= sizeof(T));
-
-    T res = 0;
-
-    const uint8_t *cur = bytes;
-    if (is_big_endian)
-      {
-	// In Big Endian, the most significant byte is at the lowest
-	// address.
-	const uint8_t* msb = cur;
-	res = *msb;
-
-	// Now read the remaining least significant bytes.
-	for (uint i = 1; i < number_of_bytes; ++i)
-	  res = (res << 8) | ((T)msb[i]);
-      }
-    else
-      {
-	// In Little Endian, the least significant byte is at the
-	// lowest address.
-	const uint8_t* lsb = cur;
-	res = *lsb;
-	// Now read the remaining most significant bytes.
-	for (uint i = 1; i < number_of_bytes; ++i)
-	  res = res | (((T)lsb[i]) << i * 8);
-      }
-
-    result = res;
-    return true;
-  }
-
-  /// Lookup the address of the function entry point that corresponds
-  /// to the address of a given function descriptor.
-  ///
-  /// On PPC64, a function pointer is the address of a function
-  /// descriptor.  Function descriptors are located in the .opd
-  /// section.  Each function descriptor is a triplet of three
-  /// addresses, each one on 64 bits.  Among those three address only
-  /// the first one is of any interest to us: the address of the entry
-  /// point of the function.
-  ///
-  /// This function returns the address of the entry point of the
-  /// function whose descriptor's address is given.
-  ///
-  /// http://refspecs.linuxfoundation.org/ELF/ppc64/PPC-elf64abi.html#FUNC-DES
-  ///
-  /// https://www.ibm.com/developerworks/community/blogs/5894415f-be62-4bc0-81c5-3956e82276f3/entry/deeply_understand_64_bit_powerpc_elf_abi_function_descriptors?lang=en
-  ///
-  /// @param fn_desc_address the address of the function descriptor to
-  /// consider.
-  ///
-  /// @return the address of the entry point of the function whose
-  /// descriptor has the address @p fn_desc_address.  If there is no
-  /// .opd section (e.g because we are not on ppc64) or more generally
-  /// if the function descriptor could not be found then this function
-  /// just returns the address of the fuction descriptor.
-  GElf_Addr
-  lookup_ppc64_elf_fn_entry_point_address(GElf_Addr fn_desc_address) const
-  {
-    if (!elf_handle())
-      return fn_desc_address;
-
-    if (!architecture_is_ppc64(elf_handle()))
-      return fn_desc_address;
-
-    bool is_big_endian = architecture_is_big_endian(elf_handle());
-
-    Elf_Scn *opd_section = find_opd_section();
-    if (!opd_section)
-      return fn_desc_address;
-
-    GElf_Shdr header_mem;
-    // The section header of the .opd section.
-    GElf_Shdr *opd_sheader = gelf_getshdr(opd_section, &header_mem);
-
-    // The offset of the function descriptor entry, in the .opd
-    // section.
-    size_t fn_desc_offset = fn_desc_address - opd_sheader->sh_addr;
-    Elf_Data *elf_data = elf_rawdata(opd_section, 0);
-
-    // Ensure that the opd_section has at least 8 bytes, starting from
-    // the offset we want read the data from.
-    if (elf_data->d_size <= fn_desc_offset + 8)
-      return fn_desc_address;
-
-    // A pointer to the data of the .opd section, that we can actually
-    // do something with.
-    uint8_t * bytes = (uint8_t*) elf_data->d_buf;
-
-    // The resulting address we are looking for is going to be formed
-    // in this variable.
-    GElf_Addr result = 0;
-    ABG_ASSERT(read_uint64_from_array_of_bytes(bytes + fn_desc_offset,
-					   is_big_endian, result));
-
-    return result;
-  }
-
   /// Test if a given function symbol has been exported.
   ///
   /// @param symbol_address the address of the symbol we are looking
@@ -5890,13 +5726,15 @@  public:
 		      // symbol that are in the .opd section.
 		      GElf_Addr fn_desc_addr = sym->st_value;
 		      GElf_Addr fn_entry_point_addr =
-			lookup_ppc64_elf_fn_entry_point_address(fn_desc_addr);
+			  lookup_ppc64_elf_fn_entry_point_address(
+			      elf_handle(), fn_desc_addr);
 		      addr_elf_symbol_sptr_map_type::const_iterator it2 =
 			fun_entry_addr_sym_map().find(fn_entry_point_addr);
 
 		      if (it2 == fun_entry_addr_sym_map().end())
 			fun_entry_addr_sym_map()[fn_entry_point_addr] = symbol;
-		      else if (address_is_in_opd_section(fn_desc_addr))
+		      else if (address_is_in_opd_section(elf_handle(),
+							 fn_desc_addr))
 			{
 			  // Either
 			  //
@@ -6059,24 +5897,6 @@  public:
     return true;
   }
 
-  /// Return true if an address is in the ".opd" section that is
-  /// present on the ppc64 platform.
-  ///
-  /// @param addr the address to consider.
-  ///
-  /// @return true iff @p addr is designates a word that is in the
-  /// ".opd" section.
-  bool
-  address_is_in_opd_section(Dwarf_Addr addr)
-  {
-    Elf_Scn * opd_section = find_opd_section();
-    if (!opd_section)
-      return false;
-    if (address_is_in_section(addr, opd_section))
-      return true;
-    return false;
-  }
-
   /// Load the symbol maps if necessary.
   ///
   /// @return true iff the symbol maps has been loaded by this
@@ -6227,26 +6047,6 @@  public:
     return addr;
   }
 
-  /// Test if a given address is in a given section.
-  ///
-  /// @param addr the address to consider.
-  ///
-  /// @param section the section to consider.
-  bool
-  address_is_in_section(Dwarf_Addr addr, Elf_Scn* section) const
-  {
-    if (!section)
-      return false;
-
-    GElf_Shdr sheader_mem;
-    GElf_Shdr* sheader = gelf_getshdr(section, &sheader_mem);
-
-    if (sheader->sh_addr <= addr && addr <= sheader->sh_addr + sheader->sh_size)
-      return true;
-
-    return false;
-  }
-
   /// For a relocatable (*.o) elf file, this function expects an
   /// absolute address, representing a global variable symbol.  It
   /// then extracts the address of the {.data,.data1,.rodata,.bss}
diff --git a/src/abg-elf-helpers.cc b/src/abg-elf-helpers.cc
index 5130e5b64f7a..fb9f3af67bfc 100644
--- a/src/abg-elf-helpers.cc
+++ b/src/abg-elf-helpers.cc
@@ -848,6 +848,153 @@  architecture_is_big_endian(Elf* elf_handle)
   return is_big_endian;
 }
 
+/// Read N bytes and convert their value into an integer type T.
+///
+/// Note that N cannot be bigger than 8 for now. The type passed needs to be at
+/// least of the size of number_of_bytes.
+///
+/// @param bytes the array of bytes to read the next 8 bytes from.
+/// Note that this array must be at least 8 bytes long.
+///
+/// @param number_of_bytes the number of bytes to read.  This number
+/// cannot be bigger than 8.
+///
+/// @param is_big_endian if true, read the 8 bytes in Big Endian
+/// mode, otherwise, read them in Little Endian.
+///
+/// @param result where to store the resuting integer that was read.
+///
+///
+/// @param true if the 8 bytes could be read, false otherwise.
+template <typename T>
+bool
+read_int_from_array_of_bytes(const uint8_t* bytes,
+			     unsigned char  number_of_bytes,
+			     bool	    is_big_endian,
+			     T&		    result)
+{
+  if (!bytes)
+    return false;
+
+  ABG_ASSERT(number_of_bytes <= 8);
+  ABG_ASSERT(number_of_bytes <= sizeof(T));
+
+  T res = 0;
+
+  const uint8_t* cur = bytes;
+  if (is_big_endian)
+    {
+      // In Big Endian, the most significant byte is at the lowest
+      // address.
+      const uint8_t* msb = cur;
+      res = *msb;
+
+      // Now read the remaining least significant bytes.
+      for (uint i = 1; i < number_of_bytes; ++i)
+	res = (res << 8) | ((T)msb[i]);
+    }
+  else
+    {
+      // In Little Endian, the least significant byte is at the
+      // lowest address.
+      const uint8_t* lsb = cur;
+      res = *lsb;
+      // Now read the remaining most significant bytes.
+      for (uint i = 1; i < number_of_bytes; ++i)
+	res = res | (((T)lsb[i]) << i * 8);
+    }
+
+  result = res;
+  return true;
+}
+
+/// Read 8 bytes and convert their value into an uint64_t.
+///
+/// @param bytes the array of bytes to read the next 8 bytes from.
+/// Note that this array must be at least 8 bytes long.
+///
+/// @param result where to store the resuting uint64_t that was read.
+///
+/// @param is_big_endian if true, read the 8 bytes in Big Endian
+/// mode, otherwise, read them in Little Endian.
+///
+/// @param true if the 8 bytes could be read, false otherwise.
+bool
+read_uint64_from_array_of_bytes(const uint8_t* bytes,
+				bool	       is_big_endian,
+				uint64_t&      result)
+{
+  return read_int_from_array_of_bytes(bytes, 8, is_big_endian, result);
+}
+
+
+/// Lookup the address of the function entry point that corresponds
+/// to the address of a given function descriptor.
+///
+/// On PPC64, a function pointer is the address of a function
+/// descriptor.  Function descriptors are located in the .opd
+/// section.  Each function descriptor is a triplet of three
+/// addresses, each one on 64 bits.  Among those three address only
+/// the first one is of any interest to us: the address of the entry
+/// point of the function.
+///
+/// This function returns the address of the entry point of the
+/// function whose descriptor's address is given.
+///
+/// http://refspecs.linuxfoundation.org/ELF/ppc64/PPC-elf64abi.html#FUNC-DES
+///
+/// https://www.ibm.com/developerworks/community/blogs/5894415f-be62-4bc0-81c5-3956e82276f3/entry/deeply_understand_64_bit_powerpc_elf_abi_function_descriptors?lang=en
+///
+/// @param fn_desc_address the address of the function descriptor to
+/// consider.
+///
+/// @return the address of the entry point of the function whose
+/// descriptor has the address @p fn_desc_address.  If there is no
+/// .opd section (e.g because we are not on ppc64) or more generally
+/// if the function descriptor could not be found then this function
+/// just returns the address of the fuction descriptor.
+GElf_Addr
+lookup_ppc64_elf_fn_entry_point_address(Elf* elf_handle, GElf_Addr fn_desc_address)
+{
+  if (!elf_handle)
+    return fn_desc_address;
+
+  if (!architecture_is_ppc64(elf_handle))
+    return fn_desc_address;
+
+  bool is_big_endian = architecture_is_big_endian(elf_handle);
+
+  Elf_Scn* opd_section = find_opd_section(elf_handle);
+  if (!opd_section)
+    return fn_desc_address;
+
+  GElf_Shdr header_mem;
+  // The section header of the .opd section.
+  GElf_Shdr* opd_sheader = gelf_getshdr(opd_section, &header_mem);
+
+  // The offset of the function descriptor entry, in the .opd
+  // section.
+  size_t    fn_desc_offset = fn_desc_address - opd_sheader->sh_addr;
+  Elf_Data* elf_data = elf_rawdata(opd_section, 0);
+
+  // Ensure that the opd_section has at least 8 bytes, starting from
+  // the offset we want read the data from.
+  if (elf_data->d_size <= fn_desc_offset + 8)
+    return fn_desc_address;
+
+  // A pointer to the data of the .opd section, that we can actually
+  // do something with.
+  uint8_t* bytes = (uint8_t*)elf_data->d_buf;
+
+  // The resulting address we are looking for is going to be formed
+  // in this variable.
+  GElf_Addr result = 0;
+  ABG_ASSERT(read_uint64_from_array_of_bytes(bytes + fn_desc_offset,
+					     is_big_endian, result));
+
+  return result;
+}
+
 /// Test if the ELF binary denoted by a given ELF handle is a Linux
 /// Kernel Module.
 ///
@@ -1012,5 +1159,44 @@  maybe_adjust_et_rel_sym_addr_to_abs_addr(Elf* elf_handle, GElf_Sym* sym)
   return addr + section_header.sh_addr;
 }
 
+/// Test if a given address is in a given section.
+///
+/// @param addr the address to consider.
+///
+/// @param section the section to consider.
+bool
+address_is_in_section(Dwarf_Addr addr, Elf_Scn* section)
+{
+  if (!section)
+    return false;
+
+  GElf_Shdr  sheader_mem;
+  GElf_Shdr* sheader = gelf_getshdr(section, &sheader_mem);
+
+  if (sheader->sh_addr <= addr && addr <= sheader->sh_addr + sheader->sh_size)
+    return true;
+
+  return false;
+}
+
+/// Return true if an address is in the ".opd" section that is
+/// present on the ppc64 platform.
+///
+/// @param addr the address to consider.
+///
+/// @return true iff @p addr is designates a word that is in the
+/// ".opd" section.
+bool
+address_is_in_opd_section(Elf* elf_handle, Dwarf_Addr addr)
+{
+  Elf_Scn * opd_section = find_opd_section(elf_handle);
+  if (!opd_section)
+    return false;
+  if (address_is_in_section(addr, opd_section))
+    return true;
+  return false;
+}
+
+
 } // end namespace elf_helpers
 } // end namespace abigail
diff --git a/src/abg-elf-helpers.h b/src/abg-elf-helpers.h
index b4f3293d2bd6..a06232738052 100644
--- a/src/abg-elf-helpers.h
+++ b/src/abg-elf-helpers.h
@@ -12,6 +12,7 @@ 
 
 #include "config.h"
 
+#include <elfutils/libdwfl.h>
 #include <gelf.h>
 #include <string>
 
@@ -133,6 +134,10 @@  architecture_is_ppc64(Elf* elf_handle);
 bool
 architecture_is_big_endian(Elf* elf_handle);
 
+GElf_Addr
+lookup_ppc64_elf_fn_entry_point_address(Elf*	  elf_handle,
+					GElf_Addr fn_desc_address);
+
 //
 // Helpers for Linux Kernel Binaries
 //
@@ -162,6 +167,9 @@  is_dso(Elf* elf_handle);
 GElf_Addr
 maybe_adjust_et_rel_sym_addr_to_abs_addr(Elf* elf_handle, GElf_Sym* sym);
 
+bool
+address_is_in_opd_section(Elf* elf_handle, Dwarf_Addr addr);
+
 } // end namespace elf_helpers
 } // end namespace abigail