[PPC64] Fix toc restore nops offset for V2 ABI

[syzaara]

Summary

The PPC64 V2 ABI restores the toc base by loading from an offset of 24 from r1. This patch fixes the offset and updates the testcases from V1 to V2. It also issues an error when a nop is missing after a call to an external function.

[ruiu]

LGTM with these fixes.

lld/ELF/InputSection.cpp

746 ↗

(On Diff #143350)

error() automatically adds "error: " at beginning of a message, so please remove it from your message. Error messages shouldn't end with a full stop.

If you flip the condition, you can reduce the nesting level like this.

if (BufLoc + 8 > BufEnd || read32(BufLoc + 4) != 0x60000000) {
  error(...);
  return;
}
write32(...)

[sfertile]

It would be nice to have a test similar to ppc64-error-toc-restore.s that makes sure we emit the error for a missing nop on a tail call as well.

ELF/InputSection.cpp

for (const Relocation &Rel : Relocations) {
     case R_RELAX_TLS_GD_TO_IE:
     case R_RELAX_TLS_GD_TO_IE_ABS:
     case R_RELAX_TLS_GD_TO_IE_PAGE_PC:
     case R_RELAX_TLS_GD_TO_IE_END:
       Target->relaxTlsGdToIe(BufLoc, Type, TargetVA);
       break;
     case R_PPC_PLT_OPD:
       // Patch a nop (0x60000000) to a ld.
-      if (BufLoc + 8 <= BufEnd && read32be(BufLoc + 4) == 0x60000000)
-        write32be(BufLoc + 4, 0xe8410028); // ld %r2, 40(%r1)
+      if (BufLoc + 8 <= BufEnd && read32(BufLoc + 4) == 0x60000000) {
+        write32(BufLoc + 4, 0xe8410018); // ld %r2, 24(%r1)
+      } else {
+        error(getErrorLocation(BufLoc) + "error: call lacks nop, can't restore toc.");
+        return;
+      }
       LLVM_FALLTHROUGH;
     default:
       Target->relocateOne(BufLoc, Type, TargetVA);
       break;
     }
   }
 }
 

test/ELF/Inputs/ppc64-func.s

+    .text
+    .abiversion 2
+    .globl  foo_not_shared
+    .p2align        4
+    .type   foo_not_shared,@function
+
+foo_not_shared:
+.Lfunc_begin0:
+  li 3, 55
+  blr
+  .long   0
+  .quad   0
+.Lfunc_end0:
+  .size foo_not_shared, .Lfunc_end0-.Lfunc_begin0

test/ELF/ppc64-error-toc-restore.s

+// RUN: llvm-mc -filetype=obj -triple=powerpc64le-unknown-linux %s -o %t.o
+// RUN: llvm-mc -filetype=obj -triple=powerpc64le-unknown-linux %p/Inputs/shared-ppc64le.s -o %t2.o
+// RUN: ld.lld -shared %t2.o -o %t2.so
+// RUN: not ld.lld %t.o %t2.so -o %t 2>&1 | FileCheck %s
+// REQUIRES: ppc
+
+# Calling external function bar needs a nop
+// CHECK: error: call lacks nop, can't restore toc
+    .text
+    .abiversion 2
+
+.global _start
+_start:
+  bl foo

test/ELF/ppc64-error-toc-tail-call.s

+// RUN: llvm-mc -filetype=obj -triple=powerpc64le-unknown-linux %s -o %t.o
+// RUN: llvm-mc -filetype=obj -triple=powerpc64le-unknown-linux %p/Inputs/shared-ppc64le.s -o %t2.o
+// RUN: ld.lld -shared %t2.o -o %t2.so
+// RUN: not ld.lld %t.o %t2.so -o %t 2>&1 | FileCheck %s
+// REQUIRES: ppc
+
+# A tail call to an external function without a nop should issue an error.
+// CHECK: error: call lacks nop, can't restore toc
+    .text
+    .abiversion 2
+
+.global _start
+_start:
+  b foo

test/ELF/ppc64-ifunc.s

 # REQUIRES: ppc
-# RUN: llvm-mc -filetype=obj -triple=powerpc64-unknown-linux %s -o %t.o
-# RUN: llvm-mc -filetype=obj -triple=powerpc64-unknown-linux %p/Inputs/shared-ppc64.s -o %t2.o
+# RUN: llvm-mc -filetype=obj -triple=powerpc64le-unknown-linux %s -o %t.o
+# RUN: llvm-mc -filetype=obj -triple=powerpc64le-unknown-linux %p/Inputs/shared-ppc64le.s -o %t2.o
 # RUN: ld.lld -shared %t2.o -o %t2.so
 # RUN: ld.lld %t.o %t2.so -o %t
 # RUN: llvm-objdump -d %t | FileCheck %s
 
 # CHECK:      _start:
-# CHECK-NEXT:  10010004: {{.*}} bl .+12
-# CHECK-NEXT:  10010008: {{.*}} bl .+40
+# CHECK-NEXT: 10010004:       1d 00 00 48     bl .+28
+# CHECK-NEXT: 10010008:       18 00 41 e8     ld 2, 24(1)
+# CHECK-NEXT: 1001000c:       35 00 00 48     bl .+52
+# CHECK-NEXT: 10010010:       18 00 41 e8     ld 2, 24(1)
 
-# 0x10010004 + 12 = 0x10010010 (PLT entry 0)
-# 0x10010008 + 40 = 0x10010030 (PLT entry 1)
+# 0x10010004 + 28 = 0x10010020 (PLT entry 0)
+# 0x1001000c + 52 = 0x10010040 (PLT entry 1)
 
 # CHECK:     Disassembly of section .plt:
-# CHECK:       10010010: {{.*}} std 2, 40(1)
-# CHECK-NEXT:  10010014: {{.*}} addis 11, 2, 4098
-# CHECK-NEXT:  10010018: {{.*}} ld 12, -32744(11)
-# CHECK-NEXT:  1001001c: {{.*}} ld 11, 0(12)
-# CHECK-NEXT:  10010020: {{.*}} mtctr 11
-# CHECK-NEXT:  10010024: {{.*}} ld 2, 8(12)
-# CHECK-NEXT:  10010028: {{.*}} ld 11, 16(12)
-# CHECK-NEXT:  1001002c: {{.*}} bctr
-# CHECK-NEXT:  10010030: {{.*}} std 2, 40(1)
-# CHECK-NEXT:  10010034: {{.*}} addis 11, 2, 4098
-# CHECK-NEXT:  10010038: {{.*}} ld 12, -32736(11)
-# CHECK-NEXT:  1001003c: {{.*}} ld 11, 0(12)
-# CHECK-NEXT:  10010040: {{.*}} mtctr 11
-# CHECK-NEXT:  10010044: {{.*}} ld 2, 8(12)
-# CHECK-NEXT:  10010048: {{.*}} ld 11, 16(12)
-# CHECK-NEXT:  1001004c: {{.*}} bctr
+# CHECK-NEXT: .plt:
+# CHECK-NEXT: 10010020:       18 00 41 f8     std 2, 24(1)
+# CHECK-NEXT: 10010024:       02 10 82 3d     addis 12, 2, 4098
+# CHECK-NEXT: 10010028:       10 80 8c e9     ld 12, -32752(12)
+# CHECK-NEXT: 1001002c:       a6 03 89 7d     mtctr 12
+# CHECK-NEXT: 10010030:       20 04 80 4e     bctr
+# CHECK-NEXT: 10010034:       08 00 e0 7f     trap
+# CHECK-NEXT: 10010038:       08 00 e0 7f     trap
+# CHECK-NEXT: 1001003c:       08 00 e0 7f     trap
+# CHECK-NEXT: 10010040:       18 00 41 f8     std 2, 24(1)
+# CHECK-NEXT: 10010044:       02 10 82 3d     addis 12, 2, 4098
+# CHECK-NEXT: 10010048:       18 80 8c e9     ld 12, -32744(12)
+# CHECK-NEXT: 1001004c:       a6 03 89 7d     mtctr 12
+    .text
+    .abiversion 2
 
 .type ifunc STT_GNU_IFUNC
 .globl ifunc
 ifunc:
  nop
 
 .global _start
 _start:
-  bl bar
+  bl foo
+  nop
   bl ifunc
+  nop

test/ELF/ppc64-toc-restore.s

-// RUN: llvm-mc -filetype=obj -triple=powerpc64-unknown-linux %s -o %t.o
-// RUN: llvm-mc -filetype=obj -triple=powerpc64-unknown-linux %p/Inputs/shared-ppc64.s -o %t2.o
+// RUN: llvm-mc -filetype=obj -triple=powerpc64le-unknown-linux %s -o %t.o
+// RUN: llvm-mc -filetype=obj -triple=powerpc64le-unknown-linux %p/Inputs/shared-ppc64le.s -o %t2.o
+// RUN: llvm-mc -filetype=obj -triple=powerpc64le-unknown-linux %p/Inputs/ppc64-func.s -o %t3.o
 // RUN: ld.lld -shared %t2.o -o %t2.so
-// RUN: ld.lld %t.o %t2.so -o %t
+// RUN: ld.lld %t.o %t2.so %t3.o -o %t
 // RUN: llvm-objdump -d %t | FileCheck %s
 // REQUIRES: ppc
+    .text
+    .abiversion 2
+.global bar_local
+bar_local:
+  li 3, 2
+  blr
 
+# Calling external function foo in a shared object needs a nop.
+# Calling local function bar_local doe snot need a nop.
 // CHECK: Disassembly of section .text:
-
 .global _start
 _start:
-  bl bar
+  bl foo
   nop
+  bl bar_local
 
 // CHECK: _start:
-// CHECK: 10010000:       48 00 00 21     bl .+32
-// CHECK-NOT: 10010004:       60 00 00 00     nop
-// CHECK: 10010004:       e8 41 00 28     ld 2, 40(1)
-
-.global noret
-noret:
-  bl bar
-  li 5, 7
-
-// CHECK: noret:
-// CHECK: 10010008: 48 00 00 19 bl .+24
-// CHECK: 1001000c: 38 a0 00 07 li 5, 7
-
-.global noretend
-noretend:
-  bl bar
-
-// CHECK: noretend:
-// CHECK: 10010010: 48 00 00 11 bl .+16
-
-.global noretb
-noretb:
-  b bar
+// CHECK: 10010008:       49 00 00 48     bl .+72
+// CHECK-NOT: 1001000c:       00 00 00 60     nop
+// CHECK: 1001000c:       18 00 41 e8     ld 2, 24(1)
+// CHECK: 10010010:       f1 ff ff 4b     bl .+67108848
+// CHECK-NOT: 10010014:       00 00 00 60     nop
+// CHECK-NOT: 10010014:       18 00 41 e8     ld 2, 24(1)
+
+# Calling a function in another object file which will have same
+# TOC base does not need a nop. If nop present, do not rewrite to
+# a toc restore
+.global diff_object
+_diff_object:
+  bl foo_not_shared
+  bl foo_not_shared
+  nop
 
-// CHECK: noretb:
-// CHECK: 10010014: 48 00 00 0c b .+12
+// CHECK: _diff_object:
+// CHECK-NEXT: 10010014:       1d 00 00 48     bl .+28
+// CHECK-NEXT: 10010018:       19 00 00 48     bl .+24
+// CHECK-NEXT: 1001001c:       00 00 00 60     nop
+
+# Branching to a local function does not need a nop
+.global noretbranch
+noretbranch:
+  b bar_local
+// CHECK: noretbranch:
+// CHECK: 10010020:       e0 ff ff 4b     b .+67108832
+// CHECK-NOT: 10010024:       00 00 00 60     nop
+// CHECK-NOT: 10010024:       18 00 41 e8     ld 2, 24(1)
 
 // This should come last to check the end-of-buffer condition.
 .global last
 last:
-  bl bar
+  bl foo
   nop
-
 // CHECK: last:
-// CHECK: 10010018: 48 00 00 09 bl .+8
-// CHECK: 1001001c: e8 41 00 28 ld 2, 40(1)
+// CHECK: 10010024:       2d 00 00 48     bl .+44
+// CHECK-NEXT: 10010028:       18 00 41 e8     ld 2, 24(1)
 
 // CHECK: Disassembly of section .plt:
 // CHECK: .plt:
-// CHECK: 10010020:       f8 41 00 28     std 2, 40(1)
-// CHECK: 10010024:       3d 62 10 02     addis 11, 2, 4098
-// CHECK: 10010028:       e9 8b 80 18     ld 12, -32744(11)
-// CHECK: 1001002c:       e9 6c 00 00     ld 11, 0(12)
-// CHECK: 10010030:       7d 69 03 a6     mtctr 11
-// CHECK: 10010034:       e8 4c 00 08     ld 2, 8(12)
-// CHECK: 10010038:       e9 6c 00 10     ld 11, 16(12)
-// CHECK: 1001003c:       4e 80 04 20     bctr
+// CHECK-NEXT: 10010050:       18 00 41 f8     std 2, 24(1)
+// CHECK-NEXT: 10010054:       02 10 82 3d     addis 12, 2, 4098
+// CHECK-NEXT: 10010058:       10 80 8c e9     ld 12, -32752(12)
+// CHECK-NEXT: 1001005c:       a6 03 89 7d     mtctr 12

test/ELF/ppc64-weak-undef-call-shared.s

 # RUN: llvm-mc -filetype=obj -triple=powerpc64-unknown-linux %s -o %t.o
 # RUN: ld.lld -shared %t.o -o %t.so
 # RUN: llvm-readobj -t -r -dyn-symbols %t.so | FileCheck %s
 # REQUIRES: ppc
 
 .section        ".toc","aw"
 .quad weakfunc
 // CHECK-NOT: R_PPC64_RELATIVE
 
 .text
 .Lfoo:
   bl weakfunc
+  nop
 // CHECK-NOT: R_PPC64_REL24
 
 .weak weakfunc
-