[lld][WebAssembly] Initialize bss segments using memory.fill

Previously we were relying on the dynamic loader to take care of this
but it simple and correct for us to do it here instead.

Now we initialize bss segments as part of `__wasm_init_memory` at the
same time we initialize passive segments.

In addition we extent the us of `__wasm_init_memory` outside of shared
memory situations.  Specifically it is now used to initialize bss
segments when the memory is imported.

Differential Revision: https://reviews.llvm.org/D112667

lld/test/wasm/data-segments.ll

@@ -187,6 +187,13 @@
 ; DIS-NEXT:            i32.const       0
 ; DIS-NEXT:            i32.const       20
 ; DIS-NEXT:            memory.init     1, 0
+; NOPIC-DIS-NEXT:      [[PTR]].const   1060
+; PIC-DIS-NEXT:        [[PTR]].const   36
+; PIC-DIS-NEXT:        global.get      1
+; PIC-DIS-NEXT:        [[PTR]].add
+; DIS-NEXT:            i32.const       0
+; DIS-NEXT:            i32.const       10000
+; DIS-NEXT:            memory.fill     0
 
 ; NOPIC-DIS-NEXT:      [[PTR]].const   11060
 ; PIC-DIS-NEXT:        local.get       0

lld/test/wasm/shared-memory-bss.s

@@ -0,0 +1,72 @@
+# RUN: llvm-mc -filetype=obj -triple=wasm32-unknown-unknown -o %t.o %s
+# RUN: wasm-ld --experimental-pic -shared --shared-memory -o %t.so %t.o
+# RUN: llvm-objdump -d --no-show-raw-insn --no-leading-addr %t.so | FileCheck %s
+# RUN: obj2yaml %t.so | FileCheck %s --check-prefix=YAML
+
+.section  .bss.foo,"",@
+.globl  foo
+.p2align  2
+foo:
+  .int32  0
+  .size foo, 4
+
+.section  .data.bar,"",@
+.globl  bar
+.p2align  2
+bar:
+  .int32  42
+  .size bar, 4
+
+.section  .custom_section.target_features,"",@
+  .int8 2
+  .int8 43
+  .int8 7
+  .ascii  "atomics"
+  .int8 43
+  .int8 11
+  .ascii  "bulk-memory"
+
+# Verify that there is only a single data segment and no bss
+# in the binary:
+
+#      YAML:  - Type:            DATA{{$}}
+# YAML-NEXT:    Segments:
+# YAML-NEXT:      - SectionOffset:   3
+# YAML-NEXT:        InitFlags:       1
+# YAML-NEXT:        Content:         2A000000
+# YAML-NEXT:  - Type:            CUSTOM
+
+# CHECK:      <__wasm_init_memory>:
+# CHECK-NEXT:    .local i32
+# CHECK-NEXT:            global.get      0
+# CHECK-NEXT:            i32.const       8
+# CHECK-NEXT:            i32.add
+# CHECK-NEXT:            local.set       0
+# CHECK-NEXT:            block
+# CHECK-NEXT:            block
+# CHECK-NEXT:            block
+# CHECK-NEXT:            local.get       0
+# CHECK-NEXT:            i32.const       0
+# CHECK-NEXT:            i32.const       1
+# CHECK-NEXT:            i32.atomic.rmw.cmpxchg  0
+# CHECK-NEXT:            br_table        {0, 1, 2}       # 1: down to label1
+# CHECK-NEXT:                                            # 2: down to label0
+# CHECK-NEXT:            end
+
+# Regular data gets initialized with memory.init
+
+# CHECK-NEXT:            i32.const       0
+# CHECK-NEXT:            global.get      0
+# CHECK-NEXT:            i32.add
+# CHECK-NEXT:            i32.const       0
+# CHECK-NEXT:            i32.const       4
+# CHECK-NEXT:            memory.init     0, 0
+
+# BSS gets initialized with memory.fill
+
+# CHECK-NEXT:            i32.const       4
+# CHECK-NEXT:            global.get      0
+# CHECK-NEXT:            i32.add
+# CHECK-NEXT:            i32.const       0
+# CHECK-NEXT:            i32.const       4
+# CHECK-NEXT:            memory.fill     0

lld/test/wasm/tls.s

@@ -87,7 +87,7 @@ tls3:
 # CHECK-NEXT:       Mutable:         true
 # CHECK-NEXT:       InitExpr:
 # CHECK-NEXT:         Opcode:          I32_CONST
-# CHECK-NEXT:         Value:           66576
+# CHECK-NEXT:         Value:           66592
 
 # __tls_base
 # CHECK-NEXT:     - Index:           1

lld/wasm/Config.h

@@ -91,6 +91,10 @@ struct Configuration {
   // for shared libraries (since they always added to a dynamic offset at
   // runtime).
   uint32_t tableBase = 0;
+
+  // Will be set to true if bss data segments should be emitted. In most cases
+  // this is not necessary.
+  bool emitBssSegments = false;
 };
 
 // The only instance of Configuration struct.

lld/wasm/OutputSections.cpp

@@ -133,10 +133,9 @@ void CodeSection::writeRelocations(raw_ostream &os) const {
 
 void DataSection::finalizeContents() {
   raw_string_ostream os(dataSectionHeader);
-  unsigned segmentCount =
-      std::count_if(segments.begin(), segments.end(),
-                    [](OutputSegment *segment) { return !segment->isBss; });
-
+  unsigned segmentCount = std::count_if(
+      segments.begin(), segments.end(),
+      [](OutputSegment *segment) { return segment->requiredInBinary(); });
 #ifndef NDEBUG
   unsigned activeCount = std::count_if(
       segments.begin(), segments.end(), [](OutputSegment *segment) {
@@ -152,7 +151,7 @@ void DataSection::finalizeContents() {
   bodySize = dataSectionHeader.size();
 
   for (OutputSegment *segment : segments) {
-    if (segment->isBss)
+    if (!segment->requiredInBinary())
       continue;
     raw_string_ostream os(segment->header);
     writeUleb128(os, segment->initFlags, "init flags");
@@ -199,7 +198,7 @@ void DataSection::writeTo(uint8_t *buf) {
   memcpy(buf, dataSectionHeader.data(), dataSectionHeader.size());
 
   for (const OutputSegment *segment : segments) {
-    if (segment->isBss)
+    if (!segment->requiredInBinary())
       continue;
     // Write data segment header
     uint8_t *segStart = buf + segment->sectionOffset;
@@ -227,7 +226,7 @@ void DataSection::writeRelocations(raw_ostream &os) const {
 
 bool DataSection::isNeeded() const {
   for (const OutputSegment *seg : segments)
-    if (!seg->isBss)
+    if (seg->requiredInBinary())
       return true;
   return false;
 }

lld/wasm/OutputSegment.h

@@ -24,6 +24,11 @@ class OutputSegment {
 
   void addInputSegment(InputChunk *inSeg);
   void finalizeInputSegments();
+  // In most circumstances BSS segments don't need to be written
+  // to the output binary.  However if the memory is imported, and
+  // we can't use memory.fill during startup (due to lack of bulk
+  // memory feature) then we include BSS segments verbatim.
+  bool requiredInBinary() const { return !isBss || config->emitBssSegments; }
 
   bool isTLS() const { return name == ".tdata"; }
 

lld/wasm/SyntheticSections.cpp

@@ -558,9 +558,10 @@ void ElemSection::writeBody() {
 
 DataCountSection::DataCountSection(ArrayRef<OutputSegment *> segments)
     : SyntheticSection(llvm::wasm::WASM_SEC_DATACOUNT),
-      numSegments(std::count_if(
-          segments.begin(), segments.end(),
-          [](OutputSegment *const segment) { return !segment->isBss; })) {}
+      numSegments(std::count_if(segments.begin(), segments.end(),
+                                [](OutputSegment *const segment) {
+                                  return segment->requiredInBinary();
+                                })) {}
 
 void DataCountSection::writeBody() {
   writeUleb128(bodyOutputStream, numSegments, "data count");
@@ -716,7 +717,7 @@ unsigned NameSection::numNamedDataSegments() const {
   unsigned numNames = 0;
 
   for (const OutputSegment *s : segments)
-    if (!s->name.empty() && !s->isBss)
+    if (!s->name.empty() && s->requiredInBinary())
       ++numNames;
 
   return numNames;
@@ -789,7 +790,7 @@ void NameSection::writeBody() {
     writeUleb128(sub.os, count, "name count");
 
     for (OutputSegment *s : segments) {
-      if (!s->name.empty() && !s->isBss) {
+      if (!s->name.empty() && s->requiredInBinary()) {
         writeUleb128(sub.os, s->index, "global index");
         writeStr(sub.os, s->name, "segment name");
       }