[lld][WebAssembly] Report undefined symbols in -shared/-pie builds #75242

sbc100 · 2023-12-12T20:30:12Z

Previously we would ignore all undefined symbols when using
-shared or -pie. All undefined symbols would be treated as imports
regardless of whether those symbols we defined in any shared library.

With this change we now track symbol in shared libraries and report
undefined symbols in the main program by default.

The old behavior is still available via the
--unresolved-symbols=import-dynamic command line flag.

This rationale for allowing this type of breaking change is that -pie
and -shared are both still experimental will warn as such, unless
--experimental-pic is passed.

As part of this change the linker now models shared library symbols
via new SharedFunctionSymbol and SharedDataSymbol types.

I've also added a new --no-shlib-sigcheck option that bypassed the
checking of functions signature in shared libraries. This is
specifically required by emscripten the case where the imports/exports
of shared libraries have been modified by via JS type legalization (this
is only needed when targeting old JS engines where bigint is not yet
available

See emscripten-core/emscripten#18198

github-actions · 2023-12-12T20:32:57Z

✅ With the latest revision this PR passed the C/C++ code formatter.

llvmbot · 2023-12-20T20:19:43Z

@llvm/pr-subscribers-llvm-binary-utilities

@llvm/pr-subscribers-lld-wasm

Author: Sam Clegg (sbc100)

Changes

Previously we would ignore/import all undefined symbols when using
-shared or -pie. With this change we now track symbol in shared
libraries and report undefined symbols by default.

The old behaviour is still possible via the
`--unresolved-symbols=import-dynamic flag.

The rationale here is that changing the semantics like this is
reasonable since -pie and -shared are both still experimental will
warn as such unless --experimental-pic is passed.

Depends on (and currently includes)
#75238

Patch is 34.18 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/75242.diff

23 Files Affected:

(modified) lld/test/wasm/Inputs/ret32.s (-1)
(added) lld/test/wasm/dylink.s (+30)
(modified) lld/test/wasm/emit-relocs.s (+1-1)
(modified) lld/test/wasm/libsearch.s (+31-31)
(modified) lld/test/wasm/pie.s (+3-3)
(modified) lld/test/wasm/shared-needed.s (+20-10)
(modified) lld/test/wasm/shared.s (+3-3)
(modified) lld/test/wasm/shared64.s (+1-1)
(modified) lld/test/wasm/signature-mismatch.s (+1-1)
(modified) lld/test/wasm/tag-section.ll (+1-1)
(added) lld/test/wasm/undef-shared.s (+12)
(modified) lld/test/wasm/undefined-data.s (+1-1)
(modified) lld/test/wasm/unresolved-symbols.s (+1-1)
(modified) lld/wasm/InputFiles.cpp (+39-4)
(modified) lld/wasm/InputFiles.h (+20-9)
(modified) lld/wasm/MarkLive.cpp (+3-2)
(modified) lld/wasm/Relocations.cpp (+9-4)
(modified) lld/wasm/SymbolTable.cpp (+76)
(modified) lld/wasm/SymbolTable.h (+3)
(modified) lld/wasm/Symbols.cpp (+6-1)
(modified) lld/wasm/Symbols.h (+25-1)
(modified) lld/wasm/SyntheticSections.cpp (+2-2)
(modified) lld/wasm/Writer.cpp (+6-2)

diff --git a/lld/test/wasm/Inputs/ret32.s b/lld/test/wasm/Inputs/ret32.s
index 5233455917e670..009f28c8cc9b88 100644
--- a/lld/test/wasm/Inputs/ret32.s
+++ b/lld/test/wasm/Inputs/ret32.s
@@ -1,4 +1,3 @@
-  .hidden ret32
   .globl  ret32
 ret32:
   .functype ret32 (f32) -> (i32)
diff --git a/lld/test/wasm/dylink.s b/lld/test/wasm/dylink.s
new file mode 100644
index 00000000000000..5326025edee018
--- /dev/null
+++ b/lld/test/wasm/dylink.s
@@ -0,0 +1,30 @@
+# RUN: llvm-mc -filetype=obj -triple=wasm32-unknown-emscripten -o %t.o %s
+# RUN: llvm-mc -filetype=obj -triple=wasm32-unknown-emscripten %p/Inputs/ret32.s -o %t.ret32.o
+# RUN: llvm-mc -filetype=obj -triple=wasm32-unknown-emscripten %p/Inputs/libsearch-dyn.s -o %t.dyn.o
+# RUN: wasm-ld --experimental-pic -shared %t.ret32.o %t.dyn.o -o %t.lib.so
+# RUN: not wasm-ld --experimental-pic -pie -o %t.wasm %t.o 2>&1 | FileCheck --check-prefix=ERROR %s
+# RUN: wasm-ld --experimental-pic -pie -o %t.wasm %t.o %t.lib.so
+# RUN: obj2yaml %t.wasm | FileCheck %s
+
+# ERROR: error: {{.*}}: undefined symbol: ret32
+.functype ret32 (f32) -> (i32)
+
+.globl _start
+_start:
+  .functype _start () -> ()
+  f32.const 0.0
+  call ret32
+  drop
+  i32.const _dynamic@GOT
+  drop
+  end_function
+
+# CHECK:      Sections:
+# CHECK-NEXT:   - Type:            CUSTOM
+# CHECK-NEXT:     Name:            dylink.0
+# CHECK-NEXT:     MemorySize:      0
+# CHECK-NEXT:     MemoryAlignment: 0
+# CHECK-NEXT:     TableSize:       0
+# CHECK-NEXT:     TableAlignment:  0
+# CHECK-NEXT:     Needed:
+# CHECK-NEXT:       - {{.*}}.lib.so
diff --git a/lld/test/wasm/emit-relocs.s b/lld/test/wasm/emit-relocs.s
index 91de6116164f72..bd136ba810b5ef 100644
--- a/lld/test/wasm/emit-relocs.s
+++ b/lld/test/wasm/emit-relocs.s
@@ -54,7 +54,7 @@ foo:
 # CHECK-NEXT:       - Index:           1
 # CHECK-NEXT:         Kind:            FUNCTION
 # CHECK-NEXT:         Name:            ret32
-# CHECK-NEXT:         Flags:           [ VISIBILITY_HIDDEN ]
+# CHECK-NEXT:         Flags:           [  ]
 # CHECK-NEXT:         Function:        1
 # CHECK-NEXT:       - Index:           2
 # CHECK-NEXT:         Kind:            DATA
diff --git a/lld/test/wasm/libsearch.s b/lld/test/wasm/libsearch.s
index 23336510748ce8..f321b859a8a179 100644
--- a/lld/test/wasm/libsearch.s
+++ b/lld/test/wasm/libsearch.s
@@ -32,66 +32,66 @@
 
 // Should use explicitly specified static library
 // Also ensure that we accept -L <arg>
-// RUN: wasm-ld --emit-relocs --no-gc-sections -o %t3 %t.o -L %t.dir -l:libls.a
-// RUN: llvm-readobj --symbols %t3 | FileCheck --check-prefix=STATIC %s
+// RUN: wasm-ld --emit-relocs --no-gc-sections -o %t4 %t.o -L %t.dir -l:libls.a
+// RUN: llvm-readobj --symbols %t4 | FileCheck --check-prefix=STATIC %s
 // STATIC: Symbols [
 // STATIC: Name: _static
 
 // Should use explicitly specified dynamic library
-// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t3 %t.o -L%t.dir -l:libls.so
-// RUN: llvm-readobj --symbols %t3 | FileCheck --check-prefix=DYNAMIC %s
+// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t5 %t.o -L%t.dir -l:libls.so
+// RUN: llvm-readobj --symbols %t5 | FileCheck --check-prefix=DYNAMIC %s
 // DYNAMIC: Symbols [
 // DYNAMIC-NOT: Name: _static
 
 // Should prefer static to dynamic when linking regular executable.
-// RUN: wasm-ld --emit-relocs --no-gc-sections -o %t3 %t.o -L%t.dir -lls
-// RUN: llvm-readobj --symbols %t3 | FileCheck --check-prefix=STATIC %s
+// RUN: wasm-ld --emit-relocs --no-gc-sections -o %t6 %t.o -L%t.dir -lls
+// RUN: llvm-readobj --symbols %t6 | FileCheck --check-prefix=STATIC %s
 
 // Should prefer dynamic when linking PIE.
-// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t3 %t.o -L%t.dir -lls
-// RUN: llvm-readobj --symbols %t3 | FileCheck --check-prefix=DYNAMIC %s
+// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t7 %t.o -L%t.dir -lls
+// RUN: llvm-readobj --symbols %t7 | FileCheck --check-prefix=DYNAMIC %s
 
 // Check for library search order
 // RUN: mkdir -p %t.dir2
 // RUN: cp %t.dir/libls.a %t.dir2
-// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t3 %t.o -L%t.dir2 -L%t.dir -lls
-// RUN: llvm-readobj --symbols %t3 | FileCheck --check-prefix=STATIC %s
+// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t8 %t.o -L%t.dir2 -L%t.dir -lls
+// RUN: llvm-readobj --symbols %t8 | FileCheck --check-prefix=STATIC %s
 
 // -L can be placed after -l
-// RUN: wasm-ld -o %t3 %t.o -lls -L%t.dir
+// RUN: wasm-ld -o %t8 %t.o -lls -L%t.dir
 
 // Check long forms as well
-// RUN: wasm-ld --emit-relocs --no-gc-sections -o %t3 %t.o --library-path=%t.dir --library=ls
-// RUN: wasm-ld --emit-relocs --no-gc-sections -o %t3 %t.o --library-path %t.dir --library ls
+// RUN: wasm-ld --emit-relocs --no-gc-sections -o %t8 %t.o --library-path=%t.dir --library=ls
+// RUN: wasm-ld --emit-relocs --no-gc-sections -o %t8 %t.o --library-path %t.dir --library ls
 
 // Should not search for dynamic libraries if -Bstatic is specified
-// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t3 %t.o -L%t.dir -Bstatic -lls
-// RUN: llvm-readobj --symbols %t3 | FileCheck --check-prefix=STATIC %s
+// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t10 %t.o -L%t.dir -Bstatic -lls
+// RUN: llvm-readobj --symbols %t10 | FileCheck --check-prefix=STATIC %s
 // RUN: not wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o /dev/null %t.o -L%t.dir -Bstatic -lls2 2>&1 \
 // RUN:   | FileCheck --check-prefix=NOLIB2 %s
 // NOLIB2: unable to find library -lls2
 
 // -Bdynamic should restore default behaviour
-// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t3 %t.o -L%t.dir -Bstatic -Bdynamic -lls
-// RUN: llvm-readobj --symbols %t3 | FileCheck --check-prefix=DYNAMIC %s
+// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t11 %t.o -L%t.dir -Bstatic -Bdynamic -lls
+// RUN: llvm-readobj --symbols %t11 | FileCheck --check-prefix=DYNAMIC %s
 
 // -Bstatic and -Bdynamic should affect only libraries which follow them
-// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t3 %t.o -L%t.dir -lls -Bstatic -Bdynamic
-// RUN: llvm-readobj --symbols %t3 | FileCheck --check-prefix=DYNAMIC %s
-// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t3 %t.o -L%t.dir -Bstatic -lls -Bdynamic
-// RUN: llvm-readobj --symbols %t3 | FileCheck --check-prefix=STATIC %s
+// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t12 %t.o -L%t.dir -lls -Bstatic -Bdynamic
+// RUN: llvm-readobj --symbols %t12 | FileCheck --check-prefix=DYNAMIC %s
+// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t13 %t.o -L%t.dir -Bstatic -lls -Bdynamic
+// RUN: llvm-readobj --symbols %t13 | FileCheck --check-prefix=STATIC %s
 
 // Check aliases as well
-// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t3 %t.o -L%t.dir -dn -lls
-// RUN: llvm-readobj --symbols %t3 | FileCheck --check-prefix=STATIC %s
-// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t3 %t.o -L%t.dir -non_shared -lls
-// RUN: llvm-readobj --symbols %t3 | FileCheck --check-prefix=STATIC %s
-// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t3 %t.o -L%t.dir -static -lls
-// RUN: llvm-readobj --symbols %t3 | FileCheck --check-prefix=STATIC %s
-// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t3 %t.o -L%t.dir -Bstatic -dy -lls
-// RUN: llvm-readobj --symbols %t3 | FileCheck --check-prefix=DYNAMIC %s
-// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t3 %t.o -L%t.dir -Bstatic -call_shared -lls
-// RUN: llvm-readobj --symbols %t3 | FileCheck --check-prefix=DYNAMIC %s
+// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t13 %t.o -L%t.dir -dn -lls
+// RUN: llvm-readobj --symbols %t13 | FileCheck --check-prefix=STATIC %s
+// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t14 %t.o -L%t.dir -non_shared -lls
+// RUN: llvm-readobj --symbols %t14 | FileCheck --check-prefix=STATIC %s
+// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t15 %t.o -L%t.dir -static -lls
+// RUN: llvm-readobj --symbols %t15 | FileCheck --check-prefix=STATIC %s
+// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t16 %t.o -L%t.dir -Bstatic -dy -lls
+// RUN: llvm-readobj --symbols %t16 | FileCheck --check-prefix=DYNAMIC %s
+// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t17 %t.o -L%t.dir -Bstatic -call_shared -lls
+// RUN: llvm-readobj --symbols %t17 | FileCheck --check-prefix=DYNAMIC %s
 
 .globl _start, _bar
 _start:
diff --git a/lld/test/wasm/pie.s b/lld/test/wasm/pie.s
index 887377043e555c..21eac792073187 100644
--- a/lld/test/wasm/pie.s
+++ b/lld/test/wasm/pie.s
@@ -1,6 +1,6 @@
 # RUN: llvm-mc -filetype=obj -triple=wasm32-unknown-emscripten -o %t.o %s
 # RUN: llvm-mc -filetype=obj -triple=wasm32-unknown-emscripten %S/Inputs/internal_func.s -o %t.internal_func.o
-# RUN: wasm-ld --no-gc-sections --experimental-pic -pie -o %t.wasm %t.o %t.internal_func.o
+# RUN: wasm-ld --no-gc-sections --experimental-pic -pie --unresolved-symbols=import-dynamic -o %t.wasm %t.o %t.internal_func.o
 # RUN: obj2yaml %t.wasm | FileCheck %s
 # RUN: llvm-objdump --disassemble-symbols=__wasm_call_ctors,__wasm_apply_data_relocs --no-show-raw-insn --no-leading-addr %t.wasm | FileCheck %s --check-prefixes DISASSEM
 
@@ -150,7 +150,7 @@ _start:
 # instruction in the InitExpr.  We also, therefore, do not need these globals
 # to be mutable.
 
-# RUN: wasm-ld --no-gc-sections --experimental-pic -pie --extra-features=extended-const -o %t.extended.wasm %t.o %t.internal_func.o
+# RUN: wasm-ld --no-gc-sections --experimental-pic -pie --unresolved-symbols=import-dynamic --extra-features=extended-const -o %t.extended.wasm %t.o %t.internal_func.o
 # RUN: obj2yaml %t.extended.wasm | FileCheck %s --check-prefix=EXTENDED-CONST
 
 # EXTENDED-CONST-NOT: __wasm_apply_global_relocs
@@ -207,7 +207,7 @@ _start:
 # to be generated along with __wasm_start as the start
 # function.
 
-# RUN: wasm-ld --no-gc-sections --shared-memory --experimental-pic -pie -o %t.shmem.wasm %t.o %t.internal_func.o
+# RUN: wasm-ld --no-gc-sections --shared-memory --experimental-pic -pie --unresolved-symbols=import-dynamic -o %t.shmem.wasm %t.o %t.internal_func.o
 # RUN: obj2yaml %t.shmem.wasm | FileCheck %s --check-prefix=SHMEM
 # RUN: llvm-objdump --disassemble-symbols=__wasm_start --no-show-raw-insn --no-leading-addr %t.shmem.wasm | FileCheck %s --check-prefix DISASSEM-SHMEM
 
diff --git a/lld/test/wasm/shared-needed.s b/lld/test/wasm/shared-needed.s
index 12c4597190a3b5..0176da832d2b19 100644
--- a/lld/test/wasm/shared-needed.s
+++ b/lld/test/wasm/shared-needed.s
@@ -1,17 +1,27 @@
 # RUN: llvm-mc -filetype=obj -triple=wasm32-unknown-unknown -o %t.o %s
 # RUN: llvm-mc -filetype=obj -triple=wasm32-unknown-unknown %p/Inputs/ret32.s -o %t.ret32.o
 
-# RUN: wasm-ld -shared --experimental-pic -o %t1.so %t.o
-# RUN: obj2yaml %t1.so | FileCheck %s -check-prefix=SO1
+# RUN: wasm-ld -shared --experimental-pic -o %t.ret32.so %t.ret32.o
+# RUN: obj2yaml %t.ret32.so | FileCheck %s -check-prefix=SO1
+
+# With linking against the ret32.so shared object we expect and undefined
+# symbol error
+# RUN: not wasm-ld -shared --experimental-pic -o %t.so %t.o 2>&1 | FileCheck %s --check-prefix=ERROR
+# ERROR: undefined symbol: ret32
+
+# RUN: wasm-ld -shared --experimental-pic -o %t.so %t.o %t.ret32.so
+# RUN: obj2yaml %t.so | FileCheck %s -check-prefix=SO2
 
-# RUN: wasm-ld -shared --experimental-pic -o %t2.so %t1.so %t.ret32.o
-# RUN: obj2yaml %t2.so | FileCheck %s -check-prefix=SO2
 
 .globl foo
 .globl data
 
+.functype ret32 (f32) -> (i32)
+
 foo:
-  .functype foo () -> ()
+  .functype foo (f32) -> (i32)
+  local.get 0
+  call ret32
   end_function
 
 .section .data,"",@
@@ -24,8 +34,8 @@ data:
 # SO1:      Sections:
 # SO1-NEXT:   - Type:            CUSTOM
 # SO1-NEXT:     Name:            dylink.0
-# SO1-NEXT:     MemorySize:      4
-# SO1-NEXT:     MemoryAlignment: 2
+# SO1-NEXT:     MemorySize:      0
+# SO1-NEXT:     MemoryAlignment: 0
 # SO1-NEXT:     TableSize:       0
 # SO1-NEXT:     TableAlignment:  0
 # SO1-NEXT:     Needed:          []
@@ -34,10 +44,10 @@ data:
 # SO2:      Sections:
 # SO2-NEXT:   - Type:            CUSTOM
 # SO2-NEXT:     Name:            dylink.0
-# SO2-NEXT:     MemorySize:      0
-# SO2-NEXT:     MemoryAlignment: 0
+# SO2-NEXT:     MemorySize:      4
+# SO2-NEXT:     MemoryAlignment: 2
 # SO2-NEXT:     TableSize:       0
 # SO2-NEXT:     TableAlignment:  0
 # SO2-NEXT:     Needed:
-# SO2-NEXT:       - shared-needed.s.tmp1.so
+# SO2-NEXT:       - shared-needed.s.tmp.ret32.so
 # SO2-NEXT:   - Type:            TYPE
diff --git a/lld/test/wasm/shared.s b/lld/test/wasm/shared.s
index a26f00163fea70..5b40d4ebee7abc 100644
--- a/lld/test/wasm/shared.s
+++ b/lld/test/wasm/shared.s
@@ -1,5 +1,5 @@
 # RUN: llvm-mc -filetype=obj -triple=wasm32-unknown-unknown -o %t.o %s
-# RUN: wasm-ld --experimental-pic -shared -o %t.wasm %t.o
+# RUN: wasm-ld --experimental-pic --unresolved-symbols=import-dynamic -shared -o %t.wasm %t.o
 # RUN: obj2yaml %t.wasm | FileCheck %s
 # RUN: llvm-objdump --disassemble-symbols=__wasm_call_ctors,__wasm_apply_data_relocs --no-show-raw-insn --no-leading-addr %t.wasm | FileCheck %s --check-prefixes DIS
 
@@ -7,8 +7,8 @@
 
 # Linker-synthesized globals
 .globaltype __stack_pointer, i32
-.globaltype	__table_base, i32, immutable
-.globaltype	__memory_base, i32, immutable
+.globaltype __table_base, i32, immutable
+.globaltype __memory_base, i32, immutable
 
 .section .data.data,"",@
 data:
diff --git a/lld/test/wasm/shared64.s b/lld/test/wasm/shared64.s
index 3401faed8610c1..fba73f187679a6 100644
--- a/lld/test/wasm/shared64.s
+++ b/lld/test/wasm/shared64.s
@@ -1,5 +1,5 @@
 # RUN: llvm-mc -filetype=obj -triple=wasm64-unknown-unknown -o %t.o %s
-# RUN: wasm-ld -mwasm64 --experimental-pic -shared -o %t.wasm %t.o
+# RUN: wasm-ld -mwasm64 --experimental-pic --unresolved-symbols=import-dynamic -shared -o %t.wasm %t.o
 # RUN: obj2yaml %t.wasm | FileCheck %s
 # RUN: llvm-objdump --disassemble-symbols=__wasm_call_ctors,__wasm_apply_data_relocs --no-show-raw-insn --no-leading-addr %t.wasm | FileCheck %s --check-prefixes DIS
 
diff --git a/lld/test/wasm/signature-mismatch.s b/lld/test/wasm/signature-mismatch.s
index 5d305efca24649..e9da3073dde87c 100644
--- a/lld/test/wasm/signature-mismatch.s
+++ b/lld/test/wasm/signature-mismatch.s
@@ -84,7 +84,7 @@ ret32_address_main:
 # RELOC-NEXT:       - Index:           1
 # RELOC-NEXT:         Kind:            FUNCTION
 # RELOC-NEXT:         Name:            ret32
-# RELOC-NEXT:         Flags:           [ VISIBILITY_HIDDEN ]
+# RELOC-NEXT:         Flags:           [ ]
 # RELOC-NEXT:         Function:        2
 # RELOC-NEXT:       - Index:           2
 # RELOC-NEXT:         Kind:            DATA
diff --git a/lld/test/wasm/tag-section.ll b/lld/test/wasm/tag-section.ll
index 4decdb58f952a9..20823c72c65116 100644
--- a/lld/test/wasm/tag-section.ll
+++ b/lld/test/wasm/tag-section.ll
@@ -11,7 +11,7 @@
 ; RUN: llc -filetype=obj -wasm-enable-eh -exception-model=wasm -mattr=+exception-handling -relocation-model=pic %p/Inputs/tag-section1.ll -o %t1.o
 ; RUN: llc -filetype=obj -wasm-enable-eh -exception-model=wasm -mattr=+exception-handling -relocation-model=pic %p/Inputs/tag-section2.ll -o %t2.o
 ; RUN: llc -filetype=obj -wasm-enable-eh -exception-model=wasm -mattr=+exception-handling -relocation-model=pic %s -o %t.o
-; RUN: wasm-ld --import-undefined --experimental-pic -pie -o %t.wasm %t.o %t1.o %t2.o
+; RUN: wasm-ld --import-undefined --experimental-pic --unresolved-symbols=import-dynamic -pie -o %t.wasm %t.o %t1.o %t2.o
 ; RUN: obj2yaml %t.wasm | FileCheck %s --check-prefix=PIC
 
 target datalayout = "e-m:e-p:32:32-i64:64-n32:64-S128"
diff --git a/lld/test/wasm/undef-shared.s b/lld/test/wasm/undef-shared.s
new file mode 100644
index 00000000000000..4c270880ef531d
--- /dev/null
+++ b/lld/test/wasm/undef-shared.s
@@ -0,0 +1,12 @@
+# RUN: llvm-mc -filetype=obj -triple=wasm32-unknown-unknown %s -o %t.o
+# RUN: not wasm-ld --experimental-pic %t.o -o /dev/null -shared 2>&1 | FileCheck %s
+
+# CHECK: error: {{.*}}: undefined symbol: hidden
+.global hidden
+.hidden hidden
+
+.global foo
+.section .data,"",@
+foo:
+ .int32 hidden
+ .size foo,4
diff --git a/lld/test/wasm/undefined-data.s b/lld/test/wasm/undefined-data.s
index d63b667c4ea38b..5e2a41606612ab 100644
--- a/lld/test/wasm/undefined-data.s
+++ b/lld/test/wasm/undefined-data.s
@@ -1,7 +1,7 @@
 # RUN: llvm-mc -filetype=obj -triple=wasm32-unknown-unknown -o %t.o %s
 # RUN: not wasm-ld -o %t.wasm %t.o 2>&1 | FileCheck %s -check-prefix=UNDEF
 # RUN: wasm-ld --allow-undefined -o %t.wasm %t.o
-# RUN: not wasm-ld --shared -o %t.wasm %t.o 2>&1 | FileCheck %s -check-prefix=SHARED
+# RUN: not wasm-ld --experimental-pic -shared --unresolved-symbols=import-dynamic -o %t.wasm %t.o 2>&1 | FileCheck %s -check-prefix=SHARED
 
 .globl  _start
 _start:
diff --git a/lld/test/wasm/unresolved-symbols.s b/lld/test/wasm/unresolved-symbols.s
index 5de54d76c6de81..ebb5acc6876c9d 100644
--- a/lld/test/wasm/unresolved-symbols.s
+++ b/lld/test/wasm/unresolved-symbols.s
@@ -80,7 +80,7 @@
 .functype get_func_addr () -> (i32)
 
 ## import-dynamic should fail due to incompatible relocations.
-# RUN: not wasm-ld %t1.o -o %t5.wasm --unresolved-symbols=import-dynamic 2>&1 | FileCheck -check-prefix=ERRNOPIC %s
+# RUN: not wasm-ld %t1.o -o %t5.wasm --experimental-pic --unresolved-symbols=import-dynamic 2>&1 | FileCheck -check-prefix=ERRNOPIC %s
 # ERRNOPIC: relocation R_WASM_MEMORY_ADDR_SLEB cannot be used against symbol `undef_data`; recompile with -fPIC
 # ERRNOPIC: relocation R_WASM_TABLE_INDEX_SLEB cannot be used against symbol `undef_func`; recompile with -fPIC
 
diff --git a/lld/wasm/InputFiles.cpp b/lld/wasm/InputFiles.cpp
index 96ac1e1610dd3b..6ef4769beed695 100644
--- a/lld/wasm/InputFiles.cpp
+++ b/lld/wasm/InputFiles.cpp
@@ -178,7 +178,7 @@ uint64_t ObjFile::calcNewValue(const WasmRelocation &reloc, uint64_t tombstone,
   case R_WASM_MEMORY_ADDR_TLS_SLEB:
   case R_WASM_MEMORY_ADDR_TLS_SLEB64:
   case R_WASM_MEMORY_ADDR_LOCREL_I32: {
-    if (isa<UndefinedData>(sym) || sym->isUndefWeak())
+    if (isa<UndefinedData>(sym) || sym->isShared() || sym->isUndefWeak())
       return 0;
     auto D = cast<DefinedData>(sym);
     uint64_t value = D->getVA() + reloc.Addend;
@@ -391,7 +391,36 @@ static bool shouldMerge(const WasmSegment &seg) {
   return true;
 }
 
-void ObjFile::parse(bool ignoreComdats) {
+void SharedFile::parse() {
+  WasmFileBase::parse();
+  assert(wasmObj->isSharedObject());
+
+  for (const SymbolRef &sym : wasmObj->symbols()) {
+    const WasmSymbol &wasmSym = wasmObj->getWasmSymbol(sym.getRawDataRefImpl());
+    if (wasmSym.isDefined()) {
+      StringRef name = wasmSym.Info.Name;
+      uint32_t flags = wasmSym.Info.Flags;
+      Symbol *s;
+      LLVM_DEBUG(dbgs() << "shared symbol: " << name << "\n");
+      switch (wasmSym.Info.Kind) {
+      case WASM_SYMBOL_TYPE_FUNCTION:
+        if (name == "__wasm_apply_data_relocs" || name == "__wasm_call_ctors") {
+          continue;
+        }
+        s = symtab->addSharedFunction(name, flags, this, wasmSym.Signature);
+        break;
+      case WASM_SYMBOL_TYPE_DATA:
+        s = symtab->addSharedData(name, flags, this);
+        break;
+      default:
+        continue;
+      }
+      symbols.push_back(s);
+    }
+  }
+}
+
+void WasmFileBase::parse() {
   // Parse a memory buffer as a wasm file.
   LLVM_DEBUG(dbgs() << "Parsing object: " << toString(this) << "\n");
   std::unique_ptr<Binary> bin = CHECK(createBinary(mb), toString(this));
@@ -399,13 +428,18 @@ void ObjFile::parse(bool ignoreComdats) {
   auto *obj = dyn_cast<WasmObjectFile>(bin.get());
   if (!obj)
     fatal(toString(this) + ": not a wasm file");
-  if (!obj->isRelocatableObject())
-    fatal(toString(this) + ": not a relocatable wasm file");
 
   bin.release();
   wasmObj.reset(obj);
 
   checkArch(obj->getArch());
+}
+
+void ObjFile::parse(bool ignoreComdats) {
+  WasmFileBase::parse();
+
+  if (!wasmObj->isRelocatableObject())
+    fatal(toString(this) + ": not a relocatable wasm file");
 
   // Build...
[truncated]

llvmbot · 2023-12-20T20:19:43Z

@llvm/pr-subscribers-lld

Author: Sam Clegg (sbc100)

Changes

Previously we would ignore/import all undefined symbols when using
-shared or -pie. With this change we now track symbol in shared
libraries and report undefined symbols by default.

The old behaviour is still possible via the
`--unresolved-symbols=import-dynamic flag.

The rationale here is that changing the semantics like this is
reasonable since -pie and -shared are both still experimental will
warn as such unless --experimental-pic is passed.

Depends on (and currently includes)
#75238

Patch is 34.18 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/75242.diff

23 Files Affected:

(modified) lld/test/wasm/Inputs/ret32.s (-1)
(added) lld/test/wasm/dylink.s (+30)
(modified) lld/test/wasm/emit-relocs.s (+1-1)
(modified) lld/test/wasm/libsearch.s (+31-31)
(modified) lld/test/wasm/pie.s (+3-3)
(modified) lld/test/wasm/shared-needed.s (+20-10)
(modified) lld/test/wasm/shared.s (+3-3)
(modified) lld/test/wasm/shared64.s (+1-1)
(modified) lld/test/wasm/signature-mismatch.s (+1-1)
(modified) lld/test/wasm/tag-section.ll (+1-1)
(added) lld/test/wasm/undef-shared.s (+12)
(modified) lld/test/wasm/undefined-data.s (+1-1)
(modified) lld/test/wasm/unresolved-symbols.s (+1-1)
(modified) lld/wasm/InputFiles.cpp (+39-4)
(modified) lld/wasm/InputFiles.h (+20-9)
(modified) lld/wasm/MarkLive.cpp (+3-2)
(modified) lld/wasm/Relocations.cpp (+9-4)
(modified) lld/wasm/SymbolTable.cpp (+76)
(modified) lld/wasm/SymbolTable.h (+3)
(modified) lld/wasm/Symbols.cpp (+6-1)
(modified) lld/wasm/Symbols.h (+25-1)
(modified) lld/wasm/SyntheticSections.cpp (+2-2)
(modified) lld/wasm/Writer.cpp (+6-2)

diff --git a/lld/test/wasm/Inputs/ret32.s b/lld/test/wasm/Inputs/ret32.s
index 5233455917e670..009f28c8cc9b88 100644
--- a/lld/test/wasm/Inputs/ret32.s
+++ b/lld/test/wasm/Inputs/ret32.s
@@ -1,4 +1,3 @@
-  .hidden ret32
   .globl  ret32
 ret32:
   .functype ret32 (f32) -> (i32)
diff --git a/lld/test/wasm/dylink.s b/lld/test/wasm/dylink.s
new file mode 100644
index 00000000000000..5326025edee018
--- /dev/null
+++ b/lld/test/wasm/dylink.s
@@ -0,0 +1,30 @@
+# RUN: llvm-mc -filetype=obj -triple=wasm32-unknown-emscripten -o %t.o %s
+# RUN: llvm-mc -filetype=obj -triple=wasm32-unknown-emscripten %p/Inputs/ret32.s -o %t.ret32.o
+# RUN: llvm-mc -filetype=obj -triple=wasm32-unknown-emscripten %p/Inputs/libsearch-dyn.s -o %t.dyn.o
+# RUN: wasm-ld --experimental-pic -shared %t.ret32.o %t.dyn.o -o %t.lib.so
+# RUN: not wasm-ld --experimental-pic -pie -o %t.wasm %t.o 2>&1 | FileCheck --check-prefix=ERROR %s
+# RUN: wasm-ld --experimental-pic -pie -o %t.wasm %t.o %t.lib.so
+# RUN: obj2yaml %t.wasm | FileCheck %s
+
+# ERROR: error: {{.*}}: undefined symbol: ret32
+.functype ret32 (f32) -> (i32)
+
+.globl _start
+_start:
+  .functype _start () -> ()
+  f32.const 0.0
+  call ret32
+  drop
+  i32.const _dynamic@GOT
+  drop
+  end_function
+
+# CHECK:      Sections:
+# CHECK-NEXT:   - Type:            CUSTOM
+# CHECK-NEXT:     Name:            dylink.0
+# CHECK-NEXT:     MemorySize:      0
+# CHECK-NEXT:     MemoryAlignment: 0
+# CHECK-NEXT:     TableSize:       0
+# CHECK-NEXT:     TableAlignment:  0
+# CHECK-NEXT:     Needed:
+# CHECK-NEXT:       - {{.*}}.lib.so
diff --git a/lld/test/wasm/emit-relocs.s b/lld/test/wasm/emit-relocs.s
index 91de6116164f72..bd136ba810b5ef 100644
--- a/lld/test/wasm/emit-relocs.s
+++ b/lld/test/wasm/emit-relocs.s
@@ -54,7 +54,7 @@ foo:
 # CHECK-NEXT:       - Index:           1
 # CHECK-NEXT:         Kind:            FUNCTION
 # CHECK-NEXT:         Name:            ret32
-# CHECK-NEXT:         Flags:           [ VISIBILITY_HIDDEN ]
+# CHECK-NEXT:         Flags:           [  ]
 # CHECK-NEXT:         Function:        1
 # CHECK-NEXT:       - Index:           2
 # CHECK-NEXT:         Kind:            DATA
diff --git a/lld/test/wasm/libsearch.s b/lld/test/wasm/libsearch.s
index 23336510748ce8..f321b859a8a179 100644
--- a/lld/test/wasm/libsearch.s
+++ b/lld/test/wasm/libsearch.s
@@ -32,66 +32,66 @@
 
 // Should use explicitly specified static library
 // Also ensure that we accept -L <arg>
-// RUN: wasm-ld --emit-relocs --no-gc-sections -o %t3 %t.o -L %t.dir -l:libls.a
-// RUN: llvm-readobj --symbols %t3 | FileCheck --check-prefix=STATIC %s
+// RUN: wasm-ld --emit-relocs --no-gc-sections -o %t4 %t.o -L %t.dir -l:libls.a
+// RUN: llvm-readobj --symbols %t4 | FileCheck --check-prefix=STATIC %s
 // STATIC: Symbols [
 // STATIC: Name: _static
 
 // Should use explicitly specified dynamic library
-// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t3 %t.o -L%t.dir -l:libls.so
-// RUN: llvm-readobj --symbols %t3 | FileCheck --check-prefix=DYNAMIC %s
+// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t5 %t.o -L%t.dir -l:libls.so
+// RUN: llvm-readobj --symbols %t5 | FileCheck --check-prefix=DYNAMIC %s
 // DYNAMIC: Symbols [
 // DYNAMIC-NOT: Name: _static
 
 // Should prefer static to dynamic when linking regular executable.
-// RUN: wasm-ld --emit-relocs --no-gc-sections -o %t3 %t.o -L%t.dir -lls
-// RUN: llvm-readobj --symbols %t3 | FileCheck --check-prefix=STATIC %s
+// RUN: wasm-ld --emit-relocs --no-gc-sections -o %t6 %t.o -L%t.dir -lls
+// RUN: llvm-readobj --symbols %t6 | FileCheck --check-prefix=STATIC %s
 
 // Should prefer dynamic when linking PIE.
-// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t3 %t.o -L%t.dir -lls
-// RUN: llvm-readobj --symbols %t3 | FileCheck --check-prefix=DYNAMIC %s
+// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t7 %t.o -L%t.dir -lls
+// RUN: llvm-readobj --symbols %t7 | FileCheck --check-prefix=DYNAMIC %s
 
 // Check for library search order
 // RUN: mkdir -p %t.dir2
 // RUN: cp %t.dir/libls.a %t.dir2
-// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t3 %t.o -L%t.dir2 -L%t.dir -lls
-// RUN: llvm-readobj --symbols %t3 | FileCheck --check-prefix=STATIC %s
+// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t8 %t.o -L%t.dir2 -L%t.dir -lls
+// RUN: llvm-readobj --symbols %t8 | FileCheck --check-prefix=STATIC %s
 
 // -L can be placed after -l
-// RUN: wasm-ld -o %t3 %t.o -lls -L%t.dir
+// RUN: wasm-ld -o %t8 %t.o -lls -L%t.dir
 
 // Check long forms as well
-// RUN: wasm-ld --emit-relocs --no-gc-sections -o %t3 %t.o --library-path=%t.dir --library=ls
-// RUN: wasm-ld --emit-relocs --no-gc-sections -o %t3 %t.o --library-path %t.dir --library ls
+// RUN: wasm-ld --emit-relocs --no-gc-sections -o %t8 %t.o --library-path=%t.dir --library=ls
+// RUN: wasm-ld --emit-relocs --no-gc-sections -o %t8 %t.o --library-path %t.dir --library ls
 
 // Should not search for dynamic libraries if -Bstatic is specified
-// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t3 %t.o -L%t.dir -Bstatic -lls
-// RUN: llvm-readobj --symbols %t3 | FileCheck --check-prefix=STATIC %s
+// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t10 %t.o -L%t.dir -Bstatic -lls
+// RUN: llvm-readobj --symbols %t10 | FileCheck --check-prefix=STATIC %s
 // RUN: not wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o /dev/null %t.o -L%t.dir -Bstatic -lls2 2>&1 \
 // RUN:   | FileCheck --check-prefix=NOLIB2 %s
 // NOLIB2: unable to find library -lls2
 
 // -Bdynamic should restore default behaviour
-// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t3 %t.o -L%t.dir -Bstatic -Bdynamic -lls
-// RUN: llvm-readobj --symbols %t3 | FileCheck --check-prefix=DYNAMIC %s
+// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t11 %t.o -L%t.dir -Bstatic -Bdynamic -lls
+// RUN: llvm-readobj --symbols %t11 | FileCheck --check-prefix=DYNAMIC %s
 
 // -Bstatic and -Bdynamic should affect only libraries which follow them
-// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t3 %t.o -L%t.dir -lls -Bstatic -Bdynamic
-// RUN: llvm-readobj --symbols %t3 | FileCheck --check-prefix=DYNAMIC %s
-// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t3 %t.o -L%t.dir -Bstatic -lls -Bdynamic
-// RUN: llvm-readobj --symbols %t3 | FileCheck --check-prefix=STATIC %s
+// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t12 %t.o -L%t.dir -lls -Bstatic -Bdynamic
+// RUN: llvm-readobj --symbols %t12 | FileCheck --check-prefix=DYNAMIC %s
+// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t13 %t.o -L%t.dir -Bstatic -lls -Bdynamic
+// RUN: llvm-readobj --symbols %t13 | FileCheck --check-prefix=STATIC %s
 
 // Check aliases as well
-// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t3 %t.o -L%t.dir -dn -lls
-// RUN: llvm-readobj --symbols %t3 | FileCheck --check-prefix=STATIC %s
-// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t3 %t.o -L%t.dir -non_shared -lls
-// RUN: llvm-readobj --symbols %t3 | FileCheck --check-prefix=STATIC %s
-// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t3 %t.o -L%t.dir -static -lls
-// RUN: llvm-readobj --symbols %t3 | FileCheck --check-prefix=STATIC %s
-// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t3 %t.o -L%t.dir -Bstatic -dy -lls
-// RUN: llvm-readobj --symbols %t3 | FileCheck --check-prefix=DYNAMIC %s
-// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t3 %t.o -L%t.dir -Bstatic -call_shared -lls
-// RUN: llvm-readobj --symbols %t3 | FileCheck --check-prefix=DYNAMIC %s
+// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t13 %t.o -L%t.dir -dn -lls
+// RUN: llvm-readobj --symbols %t13 | FileCheck --check-prefix=STATIC %s
+// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t14 %t.o -L%t.dir -non_shared -lls
+// RUN: llvm-readobj --symbols %t14 | FileCheck --check-prefix=STATIC %s
+// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t15 %t.o -L%t.dir -static -lls
+// RUN: llvm-readobj --symbols %t15 | FileCheck --check-prefix=STATIC %s
+// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t16 %t.o -L%t.dir -Bstatic -dy -lls
+// RUN: llvm-readobj --symbols %t16 | FileCheck --check-prefix=DYNAMIC %s
+// RUN: wasm-ld -pie --experimental-pic --emit-relocs --no-gc-sections -o %t17 %t.o -L%t.dir -Bstatic -call_shared -lls
+// RUN: llvm-readobj --symbols %t17 | FileCheck --check-prefix=DYNAMIC %s
 
 .globl _start, _bar
 _start:
diff --git a/lld/test/wasm/pie.s b/lld/test/wasm/pie.s
index 887377043e555c..21eac792073187 100644
--- a/lld/test/wasm/pie.s
+++ b/lld/test/wasm/pie.s
@@ -1,6 +1,6 @@
 # RUN: llvm-mc -filetype=obj -triple=wasm32-unknown-emscripten -o %t.o %s
 # RUN: llvm-mc -filetype=obj -triple=wasm32-unknown-emscripten %S/Inputs/internal_func.s -o %t.internal_func.o
-# RUN: wasm-ld --no-gc-sections --experimental-pic -pie -o %t.wasm %t.o %t.internal_func.o
+# RUN: wasm-ld --no-gc-sections --experimental-pic -pie --unresolved-symbols=import-dynamic -o %t.wasm %t.o %t.internal_func.o
 # RUN: obj2yaml %t.wasm | FileCheck %s
 # RUN: llvm-objdump --disassemble-symbols=__wasm_call_ctors,__wasm_apply_data_relocs --no-show-raw-insn --no-leading-addr %t.wasm | FileCheck %s --check-prefixes DISASSEM
 
@@ -150,7 +150,7 @@ _start:
 # instruction in the InitExpr.  We also, therefore, do not need these globals
 # to be mutable.
 
-# RUN: wasm-ld --no-gc-sections --experimental-pic -pie --extra-features=extended-const -o %t.extended.wasm %t.o %t.internal_func.o
+# RUN: wasm-ld --no-gc-sections --experimental-pic -pie --unresolved-symbols=import-dynamic --extra-features=extended-const -o %t.extended.wasm %t.o %t.internal_func.o
 # RUN: obj2yaml %t.extended.wasm | FileCheck %s --check-prefix=EXTENDED-CONST
 
 # EXTENDED-CONST-NOT: __wasm_apply_global_relocs
@@ -207,7 +207,7 @@ _start:
 # to be generated along with __wasm_start as the start
 # function.
 
-# RUN: wasm-ld --no-gc-sections --shared-memory --experimental-pic -pie -o %t.shmem.wasm %t.o %t.internal_func.o
+# RUN: wasm-ld --no-gc-sections --shared-memory --experimental-pic -pie --unresolved-symbols=import-dynamic -o %t.shmem.wasm %t.o %t.internal_func.o
 # RUN: obj2yaml %t.shmem.wasm | FileCheck %s --check-prefix=SHMEM
 # RUN: llvm-objdump --disassemble-symbols=__wasm_start --no-show-raw-insn --no-leading-addr %t.shmem.wasm | FileCheck %s --check-prefix DISASSEM-SHMEM
 
diff --git a/lld/test/wasm/shared-needed.s b/lld/test/wasm/shared-needed.s
index 12c4597190a3b5..0176da832d2b19 100644
--- a/lld/test/wasm/shared-needed.s
+++ b/lld/test/wasm/shared-needed.s
@@ -1,17 +1,27 @@
 # RUN: llvm-mc -filetype=obj -triple=wasm32-unknown-unknown -o %t.o %s
 # RUN: llvm-mc -filetype=obj -triple=wasm32-unknown-unknown %p/Inputs/ret32.s -o %t.ret32.o
 
-# RUN: wasm-ld -shared --experimental-pic -o %t1.so %t.o
-# RUN: obj2yaml %t1.so | FileCheck %s -check-prefix=SO1
+# RUN: wasm-ld -shared --experimental-pic -o %t.ret32.so %t.ret32.o
+# RUN: obj2yaml %t.ret32.so | FileCheck %s -check-prefix=SO1
+
+# With linking against the ret32.so shared object we expect and undefined
+# symbol error
+# RUN: not wasm-ld -shared --experimental-pic -o %t.so %t.o 2>&1 | FileCheck %s --check-prefix=ERROR
+# ERROR: undefined symbol: ret32
+
+# RUN: wasm-ld -shared --experimental-pic -o %t.so %t.o %t.ret32.so
+# RUN: obj2yaml %t.so | FileCheck %s -check-prefix=SO2
 
-# RUN: wasm-ld -shared --experimental-pic -o %t2.so %t1.so %t.ret32.o
-# RUN: obj2yaml %t2.so | FileCheck %s -check-prefix=SO2
 
 .globl foo
 .globl data
 
+.functype ret32 (f32) -> (i32)
+
 foo:
-  .functype foo () -> ()
+  .functype foo (f32) -> (i32)
+  local.get 0
+  call ret32
   end_function
 
 .section .data,"",@
@@ -24,8 +34,8 @@ data:
 # SO1:      Sections:
 # SO1-NEXT:   - Type:            CUSTOM
 # SO1-NEXT:     Name:            dylink.0
-# SO1-NEXT:     MemorySize:      4
-# SO1-NEXT:     MemoryAlignment: 2
+# SO1-NEXT:     MemorySize:      0
+# SO1-NEXT:     MemoryAlignment: 0
 # SO1-NEXT:     TableSize:       0
 # SO1-NEXT:     TableAlignment:  0
 # SO1-NEXT:     Needed:          []
@@ -34,10 +44,10 @@ data:
 # SO2:      Sections:
 # SO2-NEXT:   - Type:            CUSTOM
 # SO2-NEXT:     Name:            dylink.0
-# SO2-NEXT:     MemorySize:      0
-# SO2-NEXT:     MemoryAlignment: 0
+# SO2-NEXT:     MemorySize:      4
+# SO2-NEXT:     MemoryAlignment: 2
 # SO2-NEXT:     TableSize:       0
 # SO2-NEXT:     TableAlignment:  0
 # SO2-NEXT:     Needed:
-# SO2-NEXT:       - shared-needed.s.tmp1.so
+# SO2-NEXT:       - shared-needed.s.tmp.ret32.so
 # SO2-NEXT:   - Type:            TYPE
diff --git a/lld/test/wasm/shared.s b/lld/test/wasm/shared.s
index a26f00163fea70..5b40d4ebee7abc 100644
--- a/lld/test/wasm/shared.s
+++ b/lld/test/wasm/shared.s
@@ -1,5 +1,5 @@
 # RUN: llvm-mc -filetype=obj -triple=wasm32-unknown-unknown -o %t.o %s
-# RUN: wasm-ld --experimental-pic -shared -o %t.wasm %t.o
+# RUN: wasm-ld --experimental-pic --unresolved-symbols=import-dynamic -shared -o %t.wasm %t.o
 # RUN: obj2yaml %t.wasm | FileCheck %s
 # RUN: llvm-objdump --disassemble-symbols=__wasm_call_ctors,__wasm_apply_data_relocs --no-show-raw-insn --no-leading-addr %t.wasm | FileCheck %s --check-prefixes DIS
 
@@ -7,8 +7,8 @@
 
 # Linker-synthesized globals
 .globaltype __stack_pointer, i32
-.globaltype	__table_base, i32, immutable
-.globaltype	__memory_base, i32, immutable
+.globaltype __table_base, i32, immutable
+.globaltype __memory_base, i32, immutable
 
 .section .data.data,"",@
 data:
diff --git a/lld/test/wasm/shared64.s b/lld/test/wasm/shared64.s
index 3401faed8610c1..fba73f187679a6 100644
--- a/lld/test/wasm/shared64.s
+++ b/lld/test/wasm/shared64.s
@@ -1,5 +1,5 @@
 # RUN: llvm-mc -filetype=obj -triple=wasm64-unknown-unknown -o %t.o %s
-# RUN: wasm-ld -mwasm64 --experimental-pic -shared -o %t.wasm %t.o
+# RUN: wasm-ld -mwasm64 --experimental-pic --unresolved-symbols=import-dynamic -shared -o %t.wasm %t.o
 # RUN: obj2yaml %t.wasm | FileCheck %s
 # RUN: llvm-objdump --disassemble-symbols=__wasm_call_ctors,__wasm_apply_data_relocs --no-show-raw-insn --no-leading-addr %t.wasm | FileCheck %s --check-prefixes DIS
 
diff --git a/lld/test/wasm/signature-mismatch.s b/lld/test/wasm/signature-mismatch.s
index 5d305efca24649..e9da3073dde87c 100644
--- a/lld/test/wasm/signature-mismatch.s
+++ b/lld/test/wasm/signature-mismatch.s
@@ -84,7 +84,7 @@ ret32_address_main:
 # RELOC-NEXT:       - Index:           1
 # RELOC-NEXT:         Kind:            FUNCTION
 # RELOC-NEXT:         Name:            ret32
-# RELOC-NEXT:         Flags:           [ VISIBILITY_HIDDEN ]
+# RELOC-NEXT:         Flags:           [ ]
 # RELOC-NEXT:         Function:        2
 # RELOC-NEXT:       - Index:           2
 # RELOC-NEXT:         Kind:            DATA
diff --git a/lld/test/wasm/tag-section.ll b/lld/test/wasm/tag-section.ll
index 4decdb58f952a9..20823c72c65116 100644
--- a/lld/test/wasm/tag-section.ll
+++ b/lld/test/wasm/tag-section.ll
@@ -11,7 +11,7 @@
 ; RUN: llc -filetype=obj -wasm-enable-eh -exception-model=wasm -mattr=+exception-handling -relocation-model=pic %p/Inputs/tag-section1.ll -o %t1.o
 ; RUN: llc -filetype=obj -wasm-enable-eh -exception-model=wasm -mattr=+exception-handling -relocation-model=pic %p/Inputs/tag-section2.ll -o %t2.o
 ; RUN: llc -filetype=obj -wasm-enable-eh -exception-model=wasm -mattr=+exception-handling -relocation-model=pic %s -o %t.o
-; RUN: wasm-ld --import-undefined --experimental-pic -pie -o %t.wasm %t.o %t1.o %t2.o
+; RUN: wasm-ld --import-undefined --experimental-pic --unresolved-symbols=import-dynamic -pie -o %t.wasm %t.o %t1.o %t2.o
 ; RUN: obj2yaml %t.wasm | FileCheck %s --check-prefix=PIC
 
 target datalayout = "e-m:e-p:32:32-i64:64-n32:64-S128"
diff --git a/lld/test/wasm/undef-shared.s b/lld/test/wasm/undef-shared.s
new file mode 100644
index 00000000000000..4c270880ef531d
--- /dev/null
+++ b/lld/test/wasm/undef-shared.s
@@ -0,0 +1,12 @@
+# RUN: llvm-mc -filetype=obj -triple=wasm32-unknown-unknown %s -o %t.o
+# RUN: not wasm-ld --experimental-pic %t.o -o /dev/null -shared 2>&1 | FileCheck %s
+
+# CHECK: error: {{.*}}: undefined symbol: hidden
+.global hidden
+.hidden hidden
+
+.global foo
+.section .data,"",@
+foo:
+ .int32 hidden
+ .size foo,4
diff --git a/lld/test/wasm/undefined-data.s b/lld/test/wasm/undefined-data.s
index d63b667c4ea38b..5e2a41606612ab 100644
--- a/lld/test/wasm/undefined-data.s
+++ b/lld/test/wasm/undefined-data.s
@@ -1,7 +1,7 @@
 # RUN: llvm-mc -filetype=obj -triple=wasm32-unknown-unknown -o %t.o %s
 # RUN: not wasm-ld -o %t.wasm %t.o 2>&1 | FileCheck %s -check-prefix=UNDEF
 # RUN: wasm-ld --allow-undefined -o %t.wasm %t.o
-# RUN: not wasm-ld --shared -o %t.wasm %t.o 2>&1 | FileCheck %s -check-prefix=SHARED
+# RUN: not wasm-ld --experimental-pic -shared --unresolved-symbols=import-dynamic -o %t.wasm %t.o 2>&1 | FileCheck %s -check-prefix=SHARED
 
 .globl  _start
 _start:
diff --git a/lld/test/wasm/unresolved-symbols.s b/lld/test/wasm/unresolved-symbols.s
index 5de54d76c6de81..ebb5acc6876c9d 100644
--- a/lld/test/wasm/unresolved-symbols.s
+++ b/lld/test/wasm/unresolved-symbols.s
@@ -80,7 +80,7 @@
 .functype get_func_addr () -> (i32)
 
 ## import-dynamic should fail due to incompatible relocations.
-# RUN: not wasm-ld %t1.o -o %t5.wasm --unresolved-symbols=import-dynamic 2>&1 | FileCheck -check-prefix=ERRNOPIC %s
+# RUN: not wasm-ld %t1.o -o %t5.wasm --experimental-pic --unresolved-symbols=import-dynamic 2>&1 | FileCheck -check-prefix=ERRNOPIC %s
 # ERRNOPIC: relocation R_WASM_MEMORY_ADDR_SLEB cannot be used against symbol `undef_data`; recompile with -fPIC
 # ERRNOPIC: relocation R_WASM_TABLE_INDEX_SLEB cannot be used against symbol `undef_func`; recompile with -fPIC
 
diff --git a/lld/wasm/InputFiles.cpp b/lld/wasm/InputFiles.cpp
index 96ac1e1610dd3b..6ef4769beed695 100644
--- a/lld/wasm/InputFiles.cpp
+++ b/lld/wasm/InputFiles.cpp
@@ -178,7 +178,7 @@ uint64_t ObjFile::calcNewValue(const WasmRelocation &reloc, uint64_t tombstone,
   case R_WASM_MEMORY_ADDR_TLS_SLEB:
   case R_WASM_MEMORY_ADDR_TLS_SLEB64:
   case R_WASM_MEMORY_ADDR_LOCREL_I32: {
-    if (isa<UndefinedData>(sym) || sym->isUndefWeak())
+    if (isa<UndefinedData>(sym) || sym->isShared() || sym->isUndefWeak())
       return 0;
     auto D = cast<DefinedData>(sym);
     uint64_t value = D->getVA() + reloc.Addend;
@@ -391,7 +391,36 @@ static bool shouldMerge(const WasmSegment &seg) {
   return true;
 }
 
-void ObjFile::parse(bool ignoreComdats) {
+void SharedFile::parse() {
+  WasmFileBase::parse();
+  assert(wasmObj->isSharedObject());
+
+  for (const SymbolRef &sym : wasmObj->symbols()) {
+    const WasmSymbol &wasmSym = wasmObj->getWasmSymbol(sym.getRawDataRefImpl());
+    if (wasmSym.isDefined()) {
+      StringRef name = wasmSym.Info.Name;
+      uint32_t flags = wasmSym.Info.Flags;
+      Symbol *s;
+      LLVM_DEBUG(dbgs() << "shared symbol: " << name << "\n");
+      switch (wasmSym.Info.Kind) {
+      case WASM_SYMBOL_TYPE_FUNCTION:
+        if (name == "__wasm_apply_data_relocs" || name == "__wasm_call_ctors") {
+          continue;
+        }
+        s = symtab->addSharedFunction(name, flags, this, wasmSym.Signature);
+        break;
+      case WASM_SYMBOL_TYPE_DATA:
+        s = symtab->addSharedData(name, flags, this);
+        break;
+      default:
+        continue;
+      }
+      symbols.push_back(s);
+    }
+  }
+}
+
+void WasmFileBase::parse() {
   // Parse a memory buffer as a wasm file.
   LLVM_DEBUG(dbgs() << "Parsing object: " << toString(this) << "\n");
   std::unique_ptr<Binary> bin = CHECK(createBinary(mb), toString(this));
@@ -399,13 +428,18 @@ void ObjFile::parse(bool ignoreComdats) {
   auto *obj = dyn_cast<WasmObjectFile>(bin.get());
   if (!obj)
     fatal(toString(this) + ": not a wasm file");
-  if (!obj->isRelocatableObject())
-    fatal(toString(this) + ": not a relocatable wasm file");
 
   bin.release();
   wasmObj.reset(obj);
 
   checkArch(obj->getArch());
+}
+
+void ObjFile::parse(bool ignoreComdats) {
+  WasmFileBase::parse();
+
+  if (!wasmObj->isRelocatableObject())
+    fatal(toString(this) + ": not a relocatable wasm file");
 
   // Build...
[truncated]

dschuff

The biggest part of this change is the addition of the new symbol kind. Can you add to the commit message describing that (maybe it even needs to be the commit title? not sure). And can you also add a tool-conventions change to document?

lld/wasm/SymbolTable.cpp

lld/wasm/InputFiles.cpp

lld/test/wasm/dylink.s

lld/test/wasm/shared-needed.s

lld/wasm/SymbolTable.cpp

sbc100 · 2024-07-10T23:19:41Z

I've updated the PR description to give more detail. This change doesn't actually add a new symbol kind anywhere except internally within the linker. All we are doing is making use of existing information which was previously being ignored by the linker.

dschuff

I don't see any tests for -no-shlib-sigcheck?

lld/test/wasm/shared-needed.s

…uilds Previously we would ignore all undefined symbols when using `-shared` or `-pie`. All undefined symbols would be treated as imports regardless of whether those symbols we defined in any shared library. With this change we now track symbol in shared libraries and report undefined symbols in the main program by default. The old behavior is still available via the `--unresolved-symbols=import-dynamic` command line flag. This rationale for allowing this type of breaking change is that `-pie` and `-shared` are both still experimental will warn as such, unless `--experimental-pic` is passed. As part of this change the linker now models shared library symbols via new SharedFunctionSymbol and SharedDataSymbol types. I've also added a new `--no-shlib-sigcheck` option that bypassed the checking of functions signature in shared libraries. This is specifically required by emscripten the case where the imports/exports of shared libraries have been modified by via JS type legalization (this is only needed when targeting old JS engines where bigint is not yet available See emscripten-core/emscripten#18198

sbc100 · 2024-07-11T23:31:24Z

I don't see any tests for -no-shlib-sigcheck?

Test added.

sbc100 · 2024-07-12T17:01:58Z

PTAL

dschuff · 2024-07-12T19:45:15Z

lld/wasm/Symbols.h

@@ -323,6 +330,12 @@ class DefinedData : public DataSymbol {
  uint64_t size = 0;
 };

+class SharedData : public DataSymbol {


Should this class be called SharedDataSymbol for consistency with functions and with the base class?

Actually it looks like the larger symbol class hierarchy is fairly inconsistent here. so changing just this class wouldn't really fix that.

This allows llvm/llvm-project#75242 to roll in.

This allows llvm/llvm-project#75242 to roll in. Also rename MIN_INT53 in library_int53.js to avoid conflicts with runtime_debug.js (this is needed because I enabled WASM_BIGINT in these tests which pulls in library_int53.js in more tests).

This allows llvm/llvm-project#75242 to roll in. Also, rename MIN_INT53 in library_int53.js to avoid conflicts with runtime_debug.js (this is needed because I enabled WASM_BIGINT in these tests which pulls in library_int53.js in more tests).

This allows llvm/llvm-project#75242 to roll in. Also rename MIN_INT53 in library_int53.js to avoid conflicts with runtime_debug.js (this is needed because I enabled WASM_BIGINT in these tests which pulls in library_int53.js in more tests).

Followup to #22231 to allow the llvm roller to succeed. See llvm/llvm-project#75242.

The llvm change that includes `--no-shlib-sigcheck` has now langed (llvm/llvm-project#75242) so we can use that instead.

See llvm/llvm-project#98778 This issue only started showing up after the recent llvm change where shared libraries are actually processed by the linker: llvm/llvm-project#75242

…2240) See llvm/llvm-project#98778 This issue only started showing up after the recent llvm change where shared libraries are actually processed by the linker: llvm/llvm-project#75242

The llvm change that includes `--no-shlib-sigcheck` has now langed (llvm/llvm-project#75242) so we can use that instead.

…lvm#75242) Previously we would ignore all undefined symbols when using `-shared` or `-pie`. All undefined symbols would be treated as imports regardless of whether those symbols we defined in any shared library. With this change we now track symbol in shared libraries and report undefined symbols in the main program by default. The old behavior is still available via the `--unresolved-symbols=import-dynamic` command line flag. This rationale for allowing this type of breaking change is that `-pie` and `-shared` are both still experimental will warn as such, unless `--experimental-pic` is passed. As part of this change the linker now models shared library symbols via new SharedFunctionSymbol and SharedDataSymbol types. I've also added a new `--no-shlib-sigcheck` option that bypassed the checking of functions signature in shared libraries. This is specifically required by emscripten the case where the imports/exports of shared libraries have been modified by via JS type legalization (this is only needed when targeting old JS engines where bigint is not yet available See emscripten-core/emscripten#18198

The llvm change that includes `--no-shlib-sigcheck` has now langed (llvm/llvm-project#75242) so we can use that instead.

This allows llvm/llvm-project#75242 to roll in. Also, rename MIN_INT53 in library_int53.js to avoid conflicts with runtime_debug.js (this is needed because I enabled WASM_BIGINT in these tests which pulls in library_int53.js in more tests).

Followup to emscripten-core#22231 to allow the llvm roller to succeed. See llvm/llvm-project#75242.

…scripten-core#22240) See llvm/llvm-project#98778 This issue only started showing up after the recent llvm change where shared libraries are actually processed by the linker: llvm/llvm-project#75242

…ten-core#22238) The llvm change that includes `--no-shlib-sigcheck` has now langed (llvm/llvm-project#75242) so we can use that instead.

sbc100 mentioned this pull request Dec 12, 2023

[WebAssembly] Add symbol information for shared libraries #75238

Merged

sbc100 changed the title ~~Dylink symbols lld~~ [lld][WebAssembly] Report undefined symbols by default -shared/-pie builds Dec 12, 2023

sbc100 force-pushed the dylink_symbols_lld branch from 77e038c to edca2f9 Compare December 20, 2023 20:18

sbc100 marked this pull request as ready for review December 20, 2023 20:19

llvmbot added lld lld:wasm labels Dec 20, 2023

sbc100 requested a review from dschuff December 20, 2023 20:20

sbc100 force-pushed the dylink_symbols_lld branch 3 times, most recently from 0a14bef to 9b5bbe6 Compare December 20, 2023 20:25

dschuff reviewed Dec 21, 2023

View reviewed changes

lld/wasm/SymbolTable.cpp Outdated Show resolved Hide resolved

lld/wasm/InputFiles.cpp Outdated Show resolved Hide resolved

lld/test/wasm/dylink.s Outdated Show resolved Hide resolved

sbc100 force-pushed the dylink_symbols_lld branch from 9b5bbe6 to d35a345 Compare July 3, 2024 22:19

llvmbot added backend:WebAssembly llvm:binary-utilities labels Jul 3, 2024

sbc100 force-pushed the dylink_symbols_lld branch from d35a345 to 68803c1 Compare July 3, 2024 22:23

sbc100 requested a review from dschuff July 3, 2024 22:23

dschuff reviewed Jul 8, 2024

View reviewed changes

lld/test/wasm/shared-needed.s Outdated Show resolved Hide resolved

lld/wasm/SymbolTable.cpp Outdated Show resolved Hide resolved

sbc100 force-pushed the dylink_symbols_lld branch from 68803c1 to 38ff861 Compare July 10, 2024 23:03

sbc100 force-pushed the dylink_symbols_lld branch from 38ff861 to 3e5b127 Compare July 10, 2024 23:20

dschuff reviewed Jul 10, 2024

View reviewed changes

lld/test/wasm/shared-needed.s Outdated Show resolved Hide resolved

sbc100 force-pushed the dylink_symbols_lld branch 2 times, most recently from 6cf72c9 to 8d62e57 Compare July 11, 2024 16:31

sbc100 force-pushed the dylink_symbols_lld branch from 8d62e57 to 000519d Compare July 11, 2024 23:31

sbc100 changed the title ~~[lld][WebAssembly] Report undefined symbols by default -shared/-pie builds~~ [lld][WebAssembly] Report undefined symbols in -shared/-pie builds Jul 12, 2024

dschuff approved these changes Jul 12, 2024

View reviewed changes

sbc100 merged commit 22b7b84 into llvm:main Jul 12, 2024
7 checks passed

sbc100 deleted the dylink_symbols_lld branch July 12, 2024 20:26

sbc100 added a commit to sbc100/emscripten that referenced this pull request Jul 12, 2024

Temporarily add WASM_BIGINT to dylink tests

768bc8b

This allows llvm/llvm-project#75242 to roll in.

sbc100 mentioned this pull request Jul 12, 2024

Temporarily add WASM_BIGINT to dylink tests emscripten-core/emscripten#22231

Merged

sbc100 added a commit to emscripten-core/emscripten that referenced this pull request Jul 13, 2024

Temporarily add MEMORY64=2 dylink tests

ec14ddf

Followup to #22231 to allow the llvm roller to succeed. See llvm/llvm-project#75242.

sbc100 added a commit to sbc100/emscripten that referenced this pull request Jul 13, 2024

Remove temporary setting of WASM_BIGINT in dylink tests. NFC

d6160c4

The llvm change that includes `--no-shlib-sigcheck` has now langed (llvm/llvm-project#75242) so we can use that instead.

sbc100 mentioned this pull request Jul 13, 2024

Remove temporary setting of WASM_BIGINT in dylink tests. NFC emscripten-core/emscripten#22238

Merged

sbc100 mentioned this pull request Jul 14, 2024

Temporarily disable test_fetch_polyfill_shared_lib under memory64 emscripten-core/emscripten#22240

Merged

sbc100 added a commit to sbc100/emscripten that referenced this pull request Jul 14, 2024

Remove temporary setting of WASM_BIGINT in dylink tests. NFC

8473e07

The llvm change that includes `--no-shlib-sigcheck` has now langed (llvm/llvm-project#75242) so we can use that instead.

sbc100 added a commit to sbc100/emscripten that referenced this pull request Jul 14, 2024

Remove temporary setting of WASM_BIGINT in dylink tests. NFC

81aa16b

The llvm change that includes `--no-shlib-sigcheck` has now langed (llvm/llvm-project#75242) so we can use that instead.

verhovsky pushed a commit to verhovsky/emscripten that referenced this pull request Jul 30, 2024

Temporarily add MEMORY64=2 dylink tests

a6aecf1

Followup to emscripten-core#22231 to allow the llvm roller to succeed. See llvm/llvm-project#75242.

ryanking13 mentioned this pull request Jan 27, 2025

wasm-ld regression in 3.1.64: symbol type mismatch emscripten-core/emscripten#23501

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[lld][WebAssembly] Report undefined symbols in -shared/-pie builds #75242

[lld][WebAssembly] Report undefined symbols in -shared/-pie builds #75242

sbc100 commented Dec 12, 2023 •

edited

Loading

github-actions bot commented Dec 12, 2023 •

edited

Loading

llvmbot commented Dec 20, 2023 •

edited

Loading

llvmbot commented Dec 20, 2023

dschuff left a comment

sbc100 commented Jul 10, 2024

dschuff left a comment

sbc100 commented Jul 11, 2024

sbc100 commented Jul 12, 2024

dschuff Jul 12, 2024

dschuff Jul 12, 2024

[lld][WebAssembly] Report undefined symbols in -shared/-pie builds #75242

[lld][WebAssembly] Report undefined symbols in -shared/-pie builds #75242

Conversation

sbc100 commented Dec 12, 2023 • edited Loading

github-actions bot commented Dec 12, 2023 • edited Loading

llvmbot commented Dec 20, 2023 • edited Loading

llvmbot commented Dec 20, 2023

dschuff left a comment

Choose a reason for hiding this comment

sbc100 commented Jul 10, 2024

dschuff left a comment

Choose a reason for hiding this comment

sbc100 commented Jul 11, 2024

sbc100 commented Jul 12, 2024

dschuff Jul 12, 2024

Choose a reason for hiding this comment

dschuff Jul 12, 2024

Choose a reason for hiding this comment

sbc100 commented Dec 12, 2023 •

edited

Loading

github-actions bot commented Dec 12, 2023 •

edited

Loading

llvmbot commented Dec 20, 2023 •

edited

Loading