[binutils-gdb] bfd: aarch64: Avoid BTI stub for a PLT that has BTI

[binutils-gdb] bfd: aarch64: Avoid BTI stub for a PLT that has BTI

[Szabolcs Nagy]

https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;h=fc48504c7abe8eb9d9723632b2d53504927f46ff

commit fc48504c7abe8eb9d9723632b2d53504927f46ff
Author: Szabolcs Nagy <szabolcs.nagy@arm.com>
Date:   Wed Oct 18 16:12:56 2023 +0100

    bfd: aarch64: Avoid BTI stub for a PLT that has BTI
    
    We decide to emit BTI stubs based on the instruction at the target
    location. But PLT code is generated later than the stubs so we always
    read 0 which is not a valid BTI.
    
    Fix the logic to special case the PLT section: this is code the linker
    generates so we know when it will have BTI.
    
    This avoids BTI stubs in large executables where the PLTs have them
    already. An alternative is to never emit BTI stubs for PLTs, instead
    use BTI in the PLT if a library gets too big, however that may be
    more tricky given the ordering of PLT sizing and stub insertion.
    
    Related to bug 30957.

Diff:
---
 bfd/elfnn-aarch64.c | 17 ++++++++++++++---
 1 file changed, 14 insertions(+), 3 deletions(-)

diff --git a/bfd/elfnn-aarch64.c b/bfd/elfnn-aarch64.c
index 3adece30250..4faf642b422 100644
--- a/bfd/elfnn-aarch64.c
+++ b/bfd/elfnn-aarch64.c
@@ -3675,7 +3675,8 @@ group_sections (struct elf_aarch64_link_hash_table *htab,
 /* True if the inserted stub does not break BTI compatibility.  */
 
 static bool
-aarch64_bti_stub_p (struct elf_aarch64_stub_hash_entry *stub_entry)
+aarch64_bti_stub_p (struct bfd_link_info *info,
+		    struct elf_aarch64_stub_hash_entry *stub_entry)
 {
   /* Stubs without indirect branch are BTI compatible.  */
   if (stub_entry->stub_type != aarch64_stub_adrp_branch
@@ -3684,12 +3685,22 @@ aarch64_bti_stub_p (struct elf_aarch64_stub_hash_entry *stub_entry)
 
   /* Return true if the target instruction is compatible with BR x16.  */
 
+  struct elf_aarch64_link_hash_table *globals = elf_aarch64_hash_table (info);
   asection *section = stub_entry->target_section;
   bfd_byte loc[4];
   file_ptr off = stub_entry->target_value;
   bfd_size_type count = sizeof (loc);
 
-  if (!bfd_get_section_contents (section->owner, section, loc, off, count))
+  /* PLT code is not generated yet, so treat it specially.
+     Note: Checking elf_aarch64_obj_tdata.plt_type & PLT_BTI is not
+     enough because it only implies BTI in the PLT0 and tlsdesc PLT
+     entries. Normal PLT entries don't have BTI in a shared library
+     (because such PLT is normally not called indirectly and adding
+     the BTI when a stub targets a PLT would change the PLT layout
+     and it's too late for that here).  */
+  if (section == globals->root.splt)
+    memcpy (loc, globals->plt_entry, count);
+  else if (!bfd_get_section_contents (section->owner, section, loc, off, count))
     return false;
 
   uint32_t insn = bfd_getl32 (loc);
@@ -4636,7 +4647,7 @@ _bfd_aarch64_add_call_stub_entries (bool *stub_changed, bfd *output_bfd,
 
 	      /* A stub with indirect jump may break BTI compatibility, so
 		 insert another stub with direct jump near the target then.  */
-	      if (need_bti && !aarch64_bti_stub_p (stub_entry))
+	      if (need_bti && !aarch64_bti_stub_p (info, stub_entry))
 		{
 		  id_sec_bti = htab->stub_group[sym_sec->id].link_sec;