RISC-V: Support PCREL_* relocations agaist weak undefined symbols

I recently modified our Linux port's base address such the absolute
address 0 is no longer addressable as a 32-bit PC-relative offset.
Since Linux links a weak undefined symbol in an intermediate binary, it
needs to be able to reference absolute address 0.

This patch changes R_RISCV_PCREL_* relocations to absolute relocations
while resolving them in order to allow these symbols to be referenced in
PC-relative programs linked at high addresses.  Note that this doesn't
apply to PIC, which also uses PC-relative relocations, just to
position-dependent objects, which we use to allow programs to be linked
at high addresses.

In case some of our embedded users are using R_RISCV_PCREL_* as a hacked
up method of getting position-independent binaries (which can work if
you have very simple programs), we only convert the relocations when the
PC-relative version would overflow.

bfd/ChangeLog:

2017-09-07  Palmer Dabbelt  <palmer@dabbelt.com>

        * elfnn-riscv.c (riscv_zero_pcrel_hi_reloc): New function.
        (riscv_record_pcrel_hi_reloc): Add absolute argument.
        (riscv_elf_relocate_section): Call riscv_zero_pcrel_hi_reloc for
        R_RISCV_PCREL_HI20 relocs, and pass the result to
        riscv_record_pcrel_hi_reloc.

diff --git a/bfd/ChangeLog b/bfd/ChangeLog
index a8db91a3fa43c81d2dafbfe6e76eb9ddb1eb2f51..8096040a48309dc73eb2b66224d0988572698517 100644 (file)
--- a/bfd/ChangeLog
+++ b/bfd/ChangeLog
@@ -1,3 +1,11 @@
+2017-09-07  Palmer Dabbelt  <palmer@dabbelt.com>
+
+       * elfnn-riscv.c (riscv_zero_pcrel_hi_reloc): New function.
+       (riscv_record_pcrel_hi_reloc): Add absolute argument.
+       (riscv_elf_relocate_section): Call riscv_zero_pcrel_hi_reloc for
+       R_RISCV_PCREL_HI20 relocs, and pass the result to
+       riscv_record_pcrel_hi_reloc.
+
 2017-09-07  H.J. Lu  <hongjiu.lu@intel.com>
 
        * elf32-i386.c (elf_i386_convert_load_reloc): Add an argument,

diff --git a/bfd/elfnn-riscv.c b/bfd/elfnn-riscv.c
index 973f471e05293ceffb0e7bf11d5439c9e8aa5d1f..fdb151be827f248d30142fdc5ebe0e7d2a4bc784 100644 (file)
--- a/bfd/elfnn-riscv.c
+++ b/bfd/elfnn-riscv.c
@@ -1651,9 +1651,50 @@ riscv_free_pcrel_relocs (riscv_pcrel_relocs *p)
 }
 
 static bfd_boolean
-riscv_record_pcrel_hi_reloc (riscv_pcrel_relocs *p, bfd_vma addr, bfd_vma value)
+riscv_zero_pcrel_hi_reloc (Elf_Internal_Rela *rel,
+                          struct bfd_link_info *info,
+                          bfd_vma pc,
+                          bfd_vma addr,
+                          bfd_byte *contents,
+                          const reloc_howto_type *howto,
+                          bfd *input_bfd)
 {
-  riscv_pcrel_hi_reloc entry = {addr, value - addr};
+  /* We may need to reference low addreses in PC-relative modes even when the
+   * PC is far away from these addresses.  For example, undefweak references
+   * need to produce the address 0 when linked.  As 0 is far from the arbitrary
+   * addresses that we can link PC-relative programs at, the linker can't
+   * actually relocate references to those symbols.  In order to allow these
+   * programs to work we simply convert the PC-relative auipc sequences to
+   * 0-relative lui sequences.  */
+  if (bfd_link_pic (info))
+    return FALSE;
+
+  /* If it's possible to reference the symbol using auipc we do so, as that's
+   * more in the spirit of the PC-relative relocations we're processing.  */
+  bfd_vma offset = addr - pc;
+  if (ARCH_SIZE == 32 || VALID_UTYPE_IMM (RISCV_CONST_HIGH_PART (offset)))
+    return FALSE;
+
+  /* If it's impossible to reference this with a LUI-based offset then don't
+   * bother to convert it at all so users still see the PC-relative relocation
+   * in the truncation message.  */
+  if (ARCH_SIZE > 32 && !VALID_UTYPE_IMM (RISCV_CONST_HIGH_PART (addr)))
+    return FALSE;
+
+  rel->r_info = ELFNN_R_INFO(addr, R_RISCV_HI20);
+
+  bfd_vma insn = bfd_get(howto->bitsize, input_bfd, contents + rel->r_offset);
+  insn = (insn & ~MASK_AUIPC) | MATCH_LUI;
+  bfd_put(howto->bitsize, input_bfd, insn, contents + rel->r_offset);
+  return TRUE;
+}
+
+static bfd_boolean
+riscv_record_pcrel_hi_reloc (riscv_pcrel_relocs *p, bfd_vma addr,
+                            bfd_vma value, bfd_boolean absolute)
+{
+  bfd_vma offset = absolute ? value : value - addr;
+  riscv_pcrel_hi_reloc entry = {addr, offset};
   riscv_pcrel_hi_reloc **slot =
     (riscv_pcrel_hi_reloc **) htab_find_slot (p->hi_relocs, &entry, INSERT);
 
@@ -1758,6 +1799,7 @@ riscv_elf_relocate_section (bfd *output_bfd,
   Elf_Internal_Shdr *symtab_hdr = &elf_symtab_hdr (input_bfd);
   struct elf_link_hash_entry **sym_hashes = elf_sym_hashes (input_bfd);
   bfd_vma *local_got_offsets = elf_local_got_offsets (input_bfd);
+  bfd_boolean absolute;
 
   if (!riscv_init_pcrel_relocs (&pcrel_relocs))
     return FALSE;
@@ -1931,7 +1973,17 @@ riscv_elf_relocate_section (bfd *output_bfd,
                }
            }
          relocation = sec_addr (htab->elf.sgot) + off;
-         if (!riscv_record_pcrel_hi_reloc (&pcrel_relocs, pc, relocation))
+         absolute = riscv_zero_pcrel_hi_reloc (rel,
+                                               info,
+                                               pc,
+                                               relocation,
+                                               contents,
+                                               howto,
+                                               input_bfd);
+         r_type = ELFNN_R_TYPE (rel->r_info);
+         howto = riscv_elf_rtype_to_howto (r_type);
+         if (!riscv_record_pcrel_hi_reloc (&pcrel_relocs, pc,
+                                           relocation, absolute))
            r = bfd_reloc_overflow;
          break;
 
@@ -2017,8 +2069,18 @@ riscv_elf_relocate_section (bfd *output_bfd,
          }
 
        case R_RISCV_PCREL_HI20:
+         absolute = riscv_zero_pcrel_hi_reloc (rel,
+                                               info,
+                                               pc,
+                                               relocation,
+                                               contents,
+                                               howto,
+                                               input_bfd);
+         r_type = ELFNN_R_TYPE (rel->r_info);
+         howto = riscv_elf_rtype_to_howto (r_type);
          if (!riscv_record_pcrel_hi_reloc (&pcrel_relocs, pc,
-                                           relocation + rel->r_addend))
+                                           relocation + rel->r_addend,
+                                           absolute))
            r = bfd_reloc_overflow;
          break;
 
@@ -2214,7 +2276,8 @@ riscv_elf_relocate_section (bfd *output_bfd,
 
          BFD_ASSERT (off < (bfd_vma) -2);
          relocation = sec_addr (htab->elf.sgot) + off + (is_ie ? ie_off : 0);
-         if (!riscv_record_pcrel_hi_reloc (&pcrel_relocs, pc, relocation))
+         if (!riscv_record_pcrel_hi_reloc (&pcrel_relocs, pc,
+                                           relocation, FALSE))
            r = bfd_reloc_overflow;
          unresolved_reloc = FALSE;
          break;