[profile] Add __attribute__((used)) to zero size dummy sections

[MaskRay]

Summary

D14468 added these dummy sections. This patch adds __attribute__((used)) so
that when compiled by GCC>=11 or (expected, D96838) Clang>=13 on some ELF platforms,
these sections will get SHF_GNU_RETAIN to make sure they will not be discarded
by ld --gc-sections.

We are trying to get rid of LLD's "C identifier name sections are GC roots" semantics.
If LLD drops the rule in the future (we will retain compatibility for __llvm_prf_* for a while),
__llvm_prf_* will need to have the SHF_GNU_RETAIN flag, otherwise:

// __llvm_prf_cnts/__llvm_prf_data usually exist, but {names,vnds} may not exist.
// Such diagnostics will happen with {cnts,data} as well if no input object file is instrumented.
% clang++ -fprofile-generate a.cc -fuse-ld=lld -Wl,--gc-sections
ld.lld: error: undefined hidden symbol: __start___llvm_prf_names
>>> referenced by InstrProfilingPlatformLinux.c
>>>               InstrProfilingPlatformLinux.c.o:(__llvm_profile_begin_names) in archive /tmp/RelA/lib/clang/13.0.0/lib/linux/libclang_rt.profile-x86_64.a
...

[davidxl]

lgtm

[MaskRay]

Apologies that I might have committed this too fast (I wanted to unblock others quickly) but I just noticed there has been a discussion on whether __attribute__((used)) should have SHF_GNU_RETAIN behavior in GCC community.
I'll watch it closely: https://gcc.gnu.org/pipermail/gcc-patches/2021-February/565457.html
(I think even if they did not use used, they would pick a different attribute.)

compiler-rt/lib/profile/InstrProfilingPlatformLinux.c

 extern uint64_t PROF_CNTS_START COMPILER_RT_VISIBILITY;
 extern uint64_t PROF_CNTS_STOP COMPILER_RT_VISIBILITY;
 extern uint32_t PROF_ORDERFILE_START COMPILER_RT_VISIBILITY;
 extern char PROF_NAME_START COMPILER_RT_VISIBILITY;
 extern char PROF_NAME_STOP COMPILER_RT_VISIBILITY;
 extern ValueProfNode PROF_VNODES_START COMPILER_RT_VISIBILITY;
 extern ValueProfNode PROF_VNODES_STOP COMPILER_RT_VISIBILITY;
 
-/* Add dummy data to ensure the section is always created. */
-__llvm_profile_data
-    __prof_data_sect_data[0] COMPILER_RT_SECTION(INSTR_PROF_DATA_SECT_NAME);
-uint64_t
-    __prof_cnts_sect_data[0] COMPILER_RT_SECTION(INSTR_PROF_CNTS_SECT_NAME);
-uint32_t
-    __prof_orderfile_sect_data[0] COMPILER_RT_SECTION(INSTR_PROF_ORDERFILE_SECT_NAME);
-const char __prof_nms_sect_data[0] COMPILER_RT_SECTION(INSTR_PROF_NAME_SECT_NAME);
-ValueProfNode __prof_vnodes_sect_data[0] COMPILER_RT_SECTION(INSTR_PROF_VNODES_SECT_NAME);
+/* Add dummy data to ensure the section is always created. Add used attribute so
+ * that they are linker GC roots on supported ELF platforms.
+ */
+__llvm_profile_data __prof_data_sect_data[0] COMPILER_RT_SECTION(
+    INSTR_PROF_DATA_SECT_NAME) COMPILER_RT_USED;
+uint64_t __prof_cnts_sect_data[0] COMPILER_RT_SECTION(INSTR_PROF_CNTS_SECT_NAME)
+    COMPILER_RT_USED;
+uint32_t __prof_orderfile_sect_data[0] COMPILER_RT_SECTION(
+    INSTR_PROF_ORDERFILE_SECT_NAME) COMPILER_RT_USED;
+const char __prof_nms_sect_data[0] COMPILER_RT_SECTION(
+    INSTR_PROF_NAME_SECT_NAME) COMPILER_RT_USED;
+ValueProfNode __prof_vnodes_sect_data[0] COMPILER_RT_SECTION(
+    INSTR_PROF_VNODES_SECT_NAME) COMPILER_RT_USED;
 
 COMPILER_RT_VISIBILITY const __llvm_profile_data *
 __llvm_profile_begin_data(void) {
   return &PROF_DATA_START;
 }
 COMPILER_RT_VISIBILITY const __llvm_profile_data *
 __llvm_profile_end_data(void) {
   return &PROF_DATA_STOP;

compiler-rt/lib/profile/InstrProfilingPort.h

 /* FIXME: selectany does not have the same semantics as weak. */
 #define COMPILER_RT_WEAK __declspec(selectany)
 /* Need to include <windows.h> */
 #define COMPILER_RT_ALLOCA _alloca
 /* Need to include <stdio.h> and <io.h> */
 #define COMPILER_RT_FTRUNCATE(f,l) _chsize(_fileno(f),l)
 #define COMPILER_RT_ALWAYS_INLINE __forceinline
 #define COMPILER_RT_CLEANUP(x)
+#define COMPILER_RT_USED
 #elif __GNUC__
 #ifdef _WIN32
 #define COMPILER_RT_FTRUNCATE(f, l) _chsize(fileno(f), l)
 #define COMPILER_RT_VISIBILITY
 #define COMPILER_RT_WEAK __attribute__((selectany))
 #else
 #define COMPILER_RT_FTRUNCATE(f, l) ftruncate(fileno(f), l)
 #define COMPILER_RT_VISIBILITY __attribute__((visibility("hidden")))
 #define COMPILER_RT_WEAK __attribute__((weak))
 #endif
 #define COMPILER_RT_ALIGNAS(x) __attribute__((aligned(x)))
 #define COMPILER_RT_ALLOCA __builtin_alloca
 #define COMPILER_RT_ALWAYS_INLINE inline __attribute((always_inline))
 #define COMPILER_RT_CLEANUP(x) __attribute__((cleanup(x)))
+#define COMPILER_RT_USED __attribute__((used))
 #endif
 
 #if defined(__APPLE__)
 #define COMPILER_RT_SEG "__DATA,"
 #else
 #define COMPILER_RT_SEG ""
 #endif