[BOLT][test] Fix cross-compilation tests after D151920

[Elvina]

Summary

Fix tests that are failing in cross-compilation after D151920
(https://lab.llvm.org/buildbot/#/builders/221/builds/17715):

• instrumentation-ind-call, basic-instrumentation: add -mno-outline-atomics flag to runtime lib
• bolt-address-translation-internal-call, internal-call-instrument: add %cflags
• meta-merge-fdata: restrict to x86_64

[Elvina]

Trying another approach, with UNSUPPORTED target

[maksfb]

@Elvina, could you please try %cflags change for the two test cases instead of disabling them?

bolt/test/X86/bolt-address-translation-internal-call.test

12

bolt/test/X86/internal-call-instrument.s

8

[Elvina]

added %cflags to bolt-address-translation-internal-call.test

[Elvina]

In D159094#4626791, @maksfb wrote:

@Elvina, could you please try %cflags change for the two test cases instead of disabling them?

Thanks for the suggestion! I added it to the bolt-address-translation-internal-call.test, but for internal-call-instrument.s I need to modify test further, because instrumentation asks for DT_FINI entry. Will try to do it asap :)

[Elvina]

added fix for instrumentation-ind-call.c and basic-instrumentation.test

[Elvina]

rebase

[Amir]

Please update the title and summary.

bolt/test/X86/instrumentation-eh_frame_hdr.cpp

4–5

bolt/test/X86/internal-call-instrument.s

3–4

[Elvina]

Fixed internal-call-instrument.s test

[Elvina]

@Amir I disabled meta-merge-fdata.test for aarch64, as we need to correctly handle .plt.got for this case. I'll add separate patch for this soon

[Elvina]

@Amir gentle ping

[Amir]

@Elvina – please update the title and summary. Looks good otherwise.

[Elvina]

In D159094#4640029, @Amir wrote:

@Elvina – please update the title and summary. Looks good otherwise.

Sorry, I forgot about adding [BOLT] to the title, updated now.

[Amir]

Can you please also tidy up the summary using 50-72 rule?

[Amir]

(Please make sure to amend your local commit before pushing - use arc land)

[Elvina]

@Amir Thanks for the review and for your help!

bolt/runtime/CMakeLists.txt

set(BOLT_RT_FLAGS
   -fno-exceptions
   -fno-rtti
   -fno-stack-protector
   -fPIC
   -mgeneral-regs-only)
 if (CMAKE_SYSTEM_PROCESSOR MATCHES "x86_64")
   set(BOLT_RT_FLAGS ${BOLT_RT_FLAGS} "-mno-sse")
 endif()
+if (CMAKE_SYSTEM_PROCESSOR MATCHES "aarch64")
+  set(BOLT_RT_FLAGS ${BOLT_RT_FLAGS} "-mno-outline-atomics")
+endif()
 
 # Don't let the compiler think it can create calls to standard libs
 target_compile_options(bolt_rt_instr PRIVATE ${BOLT_RT_FLAGS})
 target_include_directories(bolt_rt_instr PRIVATE ${CMAKE_CURRENT_BINARY_DIR})
 target_compile_options(bolt_rt_hugify PRIVATE ${BOLT_RT_FLAGS})
 target_include_directories(bolt_rt_hugify PRIVATE ${CMAKE_CURRENT_BINARY_DIR})
 
 install(TARGETS bolt_rt_instr DESTINATION "lib${LLVM_LIBDIR_SUFFIX}")

bolt/test/X86/bolt-address-translation-internal-call.test

 # This checks for an issue with internal calls and BAT (BOLT address
 # translation). BAT needs to map every output block back to an input
 # block, but passes that introduce new blocks (such as validate
 # internal calls) might create new blocks without a mapping to an
 # input block.
 
 # REQUIRES: x86_64-linux,bolt-runtime
 
 # RUN: llvm-mc -filetype=obj -triple x86_64-unknown-unknown %s -o %t.o
 # Delete our BB symbols so BOLT doesn't mark them as entry points
 # RUN: llvm-strip --strip-unneeded %t.o
-# RUN: %clang %t.o -o %t.exe -Wl,-q
+# RUN: %clang %cflags %t.o -o %t.exe -Wl,-q
 
 # RUN: llvm-bolt --enable-bat %t.exe --relocs -o %t.out | FileCheck %s
 # CHECK: BOLT-INFO: Wrote {{.*}} BAT maps
 
 # RUN: llvm-bat-dump %t.out --dump-all | \
 # RUN:   FileCheck %s --check-prefix=CHECK-BAT-DUMP
 # CHECK-BAT-DUMP:  BAT tables for {{.*}} functions
 
   .text
   .globl  main
   .type main, %function
   .p2align  4
 main:
   push   %rbp
   mov    %rsp,%rbp
   push   %r12
   push   %rbx
   sub    $0x120,%rsp
   mov    $0x3,%rbx
+  movq   rel(%rip), %rdi
 .J1:
   cmp    $0x0,%rbx
   je     .J2
   callq  .J3
   nopl   (%rax,%rax,1)
   movabs $0xdeadbeef,%rax
   retq
 .J2:
   add    $0x120,%rsp
   pop    %rbx
   pop    %r12
   jmp    .J4
 .J3:
   pop    %rax
   add    $0x4,%rax
   dec    %rbx
   jmp    .J1
 .J4:
   pop    %rbp
   retq
+end:
   .size main, .-main
+
+  .data
+rel: .quad end

bolt/test/X86/instrumentation-eh_frame_hdr.cpp

 // This test checks that .eh_frame_hdr address is in bounds of the last LOAD
 // end address i.e. the section address is smaller then the LOAD end address.
 
-// REQUIRES: system-linux,bolt-runtime
+// REQUIRES: system-linux,bolt-runtime,target=x86_64{{.*}}
 
 // RUN: %clangxx %cxxflags -static -Wl,-q %s -o %t.exe -Wl,--entry=_start
 // RUN: llvm-bolt %t.exe -o %t.instr -instrument \
 // RUN:   --instrumentation-file=%t.fdata -instrumentation-sleep-time=1
 // RUN: (llvm-readelf -SW %t.instr | grep -v bolt; llvm-readelf -lW %t.instr | \
 // RUN:   grep LOAD | tail -n 1) | FileCheck %s
 
 // CHECK: {{.*}} .eh_frame_hdr PROGBITS [[#%x, EH_ADDR:]]
 // CHECK: LOAD 0x[[#%x, LD_OFFSET:]] 0x[[#%x, LD_VADDR:]] 0x[[#%x, LD_FSIZE:]]

bolt/test/X86/internal-call-instrument.s

 # This reproduces a bug with instrumentation crashes on internal call
 
-# REQUIRES: x86_64-linux,bolt-runtime
+# REQUIRES: x86_64-linux,bolt-runtime,target=x86_64{{.*}}
 
 # RUN: llvm-mc -filetype=obj -triple x86_64-unknown-unknown %s -o %t.o
 # Delete our BB symbols so BOLT doesn't mark them as entry points
 # RUN: llvm-strip --strip-unneeded %t.o
-# RUN: %clang %t.o -o %t.exe -Wl,-q
+# RUN: %clang %cflags %t.o -o %t.exe -Wl,-q
 
 # RUN: llvm-bolt --instrument %t.exe --relocs -o %t.out
 
   .text
+  .globl _start
+  .type _start, %function
+  .p2align  4
+_start:
+  call main
+  ret
+  .size _start, .-_start
+
   .globl  main
   .type main, %function
   .p2align  4
 main:
   push   %rbp
   mov    %rsp,%rbp
   push   %r12
   push   %rbx
   sub    $0x120,%rsp
   mov    $0x3,%rbx
+  movq   rel(%rip), %rdi
 .J1:
   cmp    $0x0,%rbx
   je     .J2
   callq  .J3
   nopl   (%rax,%rax,1)
   movabs $0xdeadbeef,%rax
   retq
 .J2:
   add    $0x120,%rsp
   pop    %rbx
   pop    %r12
   jmp    .J4
 .J3:
   pop    %rax
   add    $0x4,%rax
   dec    %rbx
   jmp    .J1
 .J4:
   pop    %rbp
   retq
+end:
   .size main, .-main
+
+  .globl  _fini
+  .type _fini, %function
+  .p2align  4
+_fini:
+  hlt
+  .size _fini, .-_fini
+
+  .data
+rel: .quad end

bolt/test/runtime/meta-merge-fdata.test

 # Meta test using merge-fdata binary
 UNSUPPORTED: asan
 # Instrumentation currently only works on X86
-REQUIRES: bolt-runtime
+REQUIRES: x86_64-linux,bolt-runtime
 
 # Instrumentation, should test:
 # - Direct branches
 # - Unmapped calls coming from uninstrumented code (libc)
 # - Direct calls
 RUN: llvm-bolt merge-fdata -o %t.inst --instrument \
 RUN:   --instrumentation-file=%t.fdata