Keep symbols passed by -init and -fini

[ruiu]

Summary

Previously, symbols passed by -init and -fini look as if they are
not referenced by anyone, and the LTO might eliminate them.
This patch fixes the issue.

Fixes a bug reported in https://bugs.llvm.org/show_bug.cgi?id=43927

[grimar]

lld/ELF/Driver.cpp

1790

Perhaps a stupid question, but still: can these symbols be lazy? i.e. should handleUndefined be used here, like for config->entry?

[MaskRay]

--gc-sections retains --init= and --fini= symbols. Adding the rule to LTO seems to improve consistency.

lld/test/ELF/lto/init-fini.ll

8

llvm-objdump -t -> llvm-nm

13

Enhance the test to check llvm-readelf -d output for DT_INIT and DT_FINI as well? It will need -pie (or -shared or --export-dynamic).

[MaskRay]

lld/ELF/Driver.cpp

1790

Calling handleUndefined will fetch the lazy symbol. However, ld.bfd and gold do not fetch lazy --init or --fini. I think this is a very minor edge case, being incompatible with GNU linkers should be fine.

Calling handleUndefined LGTM because it makes code prettier.

[ruiu]

• address review comments

[ruiu]

lld/ELF/Driver.cpp

1790

I don't worry too much about compatibility in this edge case, but I don't think there's a reason to intentionally diverge from GNU too in this case, so I prefer keeping this as is.

lld/test/ELF/lto/init-fini.ll

13

Good suggestion. Done.

[grimar]

LGTM

[grimar]

lld/test/ELF/lto/init-fini.ll

27

nit: you can use --implicit-check-not=DT_INIT --implicit-check-not=DT_FINI

[ruiu]

• simplify test

lld/ELF/Driver.cpp

template <class ELFT> void LinkerDriver::link(opt::InputArgList &args) {
   // If an entry symbol is in a static archive, pull out that file now.
   if (Symbol *sym = symtab->find(config->entry))
     handleUndefined(sym);
 
   // Handle the `--undefined-glob <pattern>` options.
   for (StringRef pat : args::getStrings(args, OPT_undefined_glob))
     handleUndefinedGlob(pat);
 
+  // Mark -init and -fini symbols so that the LTO doesn't eliminate them.
+  if (Symbol *sym = symtab->find(config->init))
+    sym->isUsedInRegularObj = true;
+  if (Symbol *sym = symtab->find(config->fini))
+    sym->isUsedInRegularObj = true;
+
   // If any of our inputs are bitcode files, the LTO code generator may create
   // references to certain library functions that might not be explicit in the
   // bitcode file's symbol table. If any of those library functions are defined
   // in a bitcode file in an archive member, we need to arrange to use LTO to
   // compile those archive members by adding them to the link beforehand.
   //
   // However, adding all libcall symbols to the link can have undesired
   // consequences. For example, the libgcc implementation of

lld/test/ELF/lto/init-fini.ll

+; REQUIRES: x86
+; RUN: llvm-as %s -o %t.o
+
+;;
+;; Verify that symbols given by -init and -fini are preserved and
+;; DT_INIT/DT_FINI are created.
+;;
+
+; RUN: ld.lld -o %t.exe -pie %t.o
+; RUN: llvm-nm %t.exe | FileCheck -check-prefix=TEST1 --allow-empty %s
+; RUN: llvm-readelf -d %t.exe | FileCheck -check-prefix=TEST2 %s
+
+; TEST1-NOT: foo
+; TEST1-NOT: bar
+
+; TEST2-NOT: INIT
+; TEST2-NOT: FINI
+
+; RUN: ld.lld -o %t.exe -pie -init=foo -fini=bar %t.o
+; RUN: llvm-nm %t.exe | FileCheck -check-prefix=TEST3 %s
+; RUN: llvm-readelf -d %t.exe | FileCheck -check-prefix=TEST4 %s
+
+; TEST3: bar
+; TEST3: foo
+
+; TEST4: INIT
+; TEST4: FINI
+
+target triple = "x86_64-unknown-linux-gnu"
+target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
+
+define void @foo() {
+  ret void
+}
+
+define void @bar() {
+  ret void
+}