[llvm-exegesis] Use mmap2 when mmap is unavailable to fix riscv32 build

[raj.khem]

Summary

SYS_mmap is not available on RV32 and it fails to build

| /mnt/b/yoe/master/build/tmp/work-shared/llvm-project-source-17.0.0-r0/git/llvm/tools/llvm-exegesis/lib/X86/Target.cpp:1116:19: error: use of undeclared identifier 'SYS_mmap'
|  1116 |   generateSyscall(SYS_mmap, MmapCode);                                    
|       |                   ^                                                                                                                                                               
| /mnt/b/yoe/master/build/tmp/work-shared/llvm-project-source-17.0.0-r0/git/llvm/tools/llvm-exegesis/lib/X86/Target.cpp:1134:19: error: use of undeclared identifier 'SYS_mmap'
|  1134 |   generateSyscall(SYS_mmap, GeneratedCode);                                                                                                                                       |       |                   ^                                                                 
| 1 warning and 2 errors generated.

[raj.khem]

This is also needed on 17.0.0 branch too.

[MaskRay]

llvm/tools/llvm-exegesis/lib/X86/Target.cpp

1094

I suggest that we write #if defined(SYS_mmap2) && !defined(SYS_mmap), which will help future 32-bit architectures.

// Some 32-bit architectures don't have mmap and use mmap2 instead ...

[MaskRay]

@aidengrossman This patch fixes an issue in D151023 that is not completely fixed by 82909f4e0529a0ac26adbd05eda7244b11d99cba

[aidengrossman]

Once you adopt @MaskRay's comments regarding the comment wording and preprocessor directives, this should LGTM.

I'm surprised this wasn't caught earlier on (ideally soon after commit). It doesn't seem like there's any buildbot coverage for 32-bit RISCV? Not sure what would be involved in setting that up, but it would be nice to have the coverage to catch obvious regressions like this one more quickly rather than late in the release process.

[MaskRay]

.

[MaskRay]

Given how late we are in the release process, I'll go ahead, apply the changes, and land this patch.
I've landed other fixes for @raj.khem and hope that he doesn't mind:)

[MaskRay]

@raj.khem Thank you for update! Already committed... Created https://github.com/llvm/llvm-project/issues/64994 for release/17.x backport.

[raj.khem]

In D158375#4617637, @MaskRay wrote:

Given how late we are in the release process, I'll go ahead, apply the changes, and land this patch.
I've landed other fixes for @raj.khem and hope that he doesn't mind:)

Not at all :) infact thank you for doing it

llvm/tools/llvm-exegesis/lib/X86/Target.cpp

 
 // Before kernel 4.17, Linux did not support MAP_FIXED_NOREPLACE, so if it is
 // not available, simplfy define it as MAP_FIXED which performs the same
 // function but does not guarantee existing mappings won't get clobbered.
 #ifndef MAP_FIXED_NOREPLACE
 #define MAP_FIXED_NOREPLACE MAP_FIXED
 #endif
 
-// 32 bit ARM doesn't have mmap and uses mmap2 instead. The only difference
-// between the two syscalls is that mmap2's offset parameter is in terms 4096
-// byte offsets rather than individual bytes, so for our purposes they are
-// effectively the same as all ofsets here are set to 0.
-#ifdef __arm__
+// some 32-bit architectures e.g. ARM, RISCV32 don't have mmap and use mmap2
+// instead. The only difference between the two syscalls is that mmap2's
+// offset parameter is in terms of 4096 byte offsets rather than individual
+// bytes, so for our purposes they are effectively the same as all ofsets
+// here are set to 0.
+#if defined(SYS_mmap2) && !defined(SYS_mmap)
 #define SYS_mmap SYS_mmap2
 #endif
 
 std::vector<MCInst>
 ExegesisX86Target::generateMmap(intptr_t Address, size_t Length,
                                 intptr_t FileDescriptorAddress) const {
   std::vector<MCInst> MmapCode;
   MmapCode.push_back(loadImmediate(X86::RDI, 64, APInt(64, Address)));

llvm/unittests/tools/llvm-exegesis/X86/TargetTest.cpp

 
 // Before kernel 4.17, Linux did not support MAP_FIXED_NOREPLACE, so if it is
 // not available, simplfy define it as MAP_FIXED which performs the same
 // function but does not guarantee existing mappings won't get clobbered.
 #ifndef MAP_FIXED_NOREPLACE
 #define MAP_FIXED_NOREPLACE MAP_FIXED
 #endif
 
-// 32 bit ARM doesn't have mmap and uses mmap2 instead. The only difference
-// between the two syscalls is that mmap2's offset parameter is in terms 4096
-// byte offsets rather than individual bytes, so for our purposes they are
-// effectively the same as all ofsets here are set to 0.
-#ifdef __arm__
+// some 32-bit architectures e.g. ARM, RISCV32 don't have mmap and use mmap2
+// instead. The only difference between the two syscalls is that mmap2's
+// offset parameter is in terms of 4096 byte offsets rather than individual
+// bytes, so for our purposes they are effectively the same as all ofsets
+// here are set to 0.
+#if defined(SYS_mmap2) && !defined(SYS_mmap)
 #define SYS_mmap SYS_mmap2
 #endif
 
 TEST_F(X86Core2TargetTest, GenerateMmapTest) {
   EXPECT_THAT(State.getExegesisTarget().generateMmap(0x1000, 4096, 0x2000),
               ElementsAre(IsMovImmediate(X86::MOV64ri, X86::RDI, 0x1000),
                           IsMovImmediate(X86::MOV64ri, X86::RSI, 4096),
                           IsMovImmediate(X86::MOV64ri, X86::RDX,