diff --git a/macho/dumper.cc b/macho/dumper.cc
index 613f1767..edc6c10a 100644
--- a/macho/dumper.cc
+++ b/macho/dumper.cc
@@ -476,25 +476,25 @@ void dump_file(std::string path) {
       LinkEditDataCommand &cmd = *(LinkEditDataCommand *)&lc;
 
       CodeSignatureHeader &sig = *(CodeSignatureHeader *)(buf + cmd.dataoff);
-      std::cout << " magic: " << read32be((u8 *)&sig.magic)
-                << "\n length: " << read32be((u8 *)&sig.length)
-                << "\n count: " << read32be((u8 *)&sig.count)
+      std::cout << " magic: " << read32be(&sig.magic)
+                << "\n length: " << read32be(&sig.length)
+                << "\n count: " << read32be(&sig.count)
                 << "\n";
 
-      for (i64 i = 0; i < read32be((u8 *)&sig.count); i++) {
-        CodeSignatureBlobIndex &idx = *(CodeSignatureBlobIndex *)(&sig + i + 1);
-        std::cout << " idx type: " << read32be((u8 *)&idx.type)
-                  << "\n idx offset: " << read32be((u8 *)&idx.offset)
+      for (i64 i = 0; i < read32be(&sig.count); i++) {
+        CodeSignatureBlobIndex &idx = ((CodeSignatureBlobIndex *)(&sig + 1))[i];
+        std::cout << " idx type: " << read32be(&idx.type)
+                  << "\n idx offset: " << read32be(&idx.offset)
                   << "\n";
 
         CodeSignatureDirectory &dir =
-          *(CodeSignatureDirectory *)((u8 *)&sig + read32be((u8 *)&idx.offset));
+          *(CodeSignatureDirectory *)((u8 *)&sig + read32be(&idx.offset));
         std::cout << std::hex
-                  << " magic: 0x" << read32be((u8 *)&dir.magic)
-                  << "\n version: 0x" << read32be((u8 *)&dir.version)
-                  << "\n flags: 0x" << read32be((u8 *)&dir.flags)
-                  << "\n hash_offset: 0x" << read32be((u8 *)&dir.hash_offset)
-                  << "\n n_code_slots: 0x" << read32be((u8 *)&dir.n_code_slots)
+                  << " magic: 0x" << read32be(&dir.magic)
+                  << "\n version: 0x" << read32be(&dir.version)
+                  << "\n flags: 0x" << read32be(&dir.flags)
+                  << "\n hash_offset: 0x" << read32be(&dir.hash_offset)
+                  << "\n n_code_slots: 0x" << read32be(&dir.n_code_slots)
                   << "\n hash_size: 0x" << (u32)dir.hash_size
                   << "\n hash_type: 0x" << (u32)dir.hash_type
                   << "\n page_size: 0x" << (1 << dir.page_size)
diff --git a/macho/main.cc b/macho/main.cc
index 7d36a062..b305f69b 100644
--- a/macho/main.cc
+++ b/macho/main.cc
@@ -288,6 +288,7 @@ int main(int argc, char **argv) {
 
   for (std::unique_ptr<OutputSegment> &seg : ctx.segments)
     seg->copy_buf(ctx);
+  ctx.code_sig.write_signature(ctx);
 
   ctx.output_file->close(ctx);
   ctx.checkpoint();
diff --git a/macho/mold.h b/macho/mold.h
index c4a48d65..e35c216a 100644
--- a/macho/mold.h
+++ b/macho/mold.h
@@ -14,6 +14,7 @@ namespace mold::macho {
 
 static constexpr i64 PAGE_SIZE = 0x4000;
 static constexpr i64 PAGE_ZERO_SIZE = 0x100000000;
+static constexpr i64 SHA256_SIZE = 32;
 
 class Chunk;
 class OutputSection;
@@ -458,6 +459,8 @@ public:
 
   void compute_size(Context &ctx) override;
   void write_signature(Context &ctx);
+
+  static constexpr i64 BLOCK_SIZE = 4096;
 };
 
 class StubsSection : public Chunk {
diff --git a/macho/output-chunks.cc b/macho/output-chunks.cc
index b63f5ac2..552c95e9 100644
--- a/macho/output-chunks.cc
+++ b/macho/output-chunks.cc
@@ -803,6 +803,16 @@ void OutputIndirectSymtabSection::copy_buf(Context &ctx) {
 }
 
 void CodeSignatureSection::compute_size(Context &ctx) {
+  i64 num_blocks = align_to(hdr.offset, BLOCK_SIZE) / BLOCK_SIZE;
+  hdr.size = sizeof(CodeSignatureHeader) + sizeof(CodeSignatureBlobIndex) +
+             num_blocks * SHA256_SIZE;
+}
+
+void CodeSignatureSection::write_signature(Context &ctx) {
+  u8 *buf = ctx.buf + hdr.offset;
+
+  CodeSignatureHeader &sighdr = *(CodeSignatureHeader *)buf;
+  write32be(&sighdr.magic, CSMAGIC_EMBEDDED_SIGNATURE);
 }
 
 void StubsSection::add(Context &ctx, Symbol *sym) {
diff --git a/mold.h b/mold.h
index 69c2f130..d8ef0ea2 100644
--- a/mold.h
+++ b/mold.h
@@ -178,45 +178,55 @@ inline void sort(T &vec, U less) {
   std::stable_sort(vec.begin(), vec.end(), less);
 }
 
-inline u64 read64be(u8 *buf) {
-  return ((u64)buf[0] << 56) | ((u64)buf[1] << 48) |
-         ((u64)buf[2] << 40) | ((u64)buf[3] << 32) |
-         ((u64)buf[4] << 24) | ((u64)buf[5] << 16) |
-         ((u64)buf[6] << 8)  | (u64)buf[7];
-}
-
-inline u64 read32be(u8 *buf) {
+template <typename T>
+inline u64 read64be(T *buf) {
   u8 *p = (u8 *)buf;
-  return ((u64)buf[0] << 24) | ((u64)buf[1] << 16) |
-         ((u64)buf[2] << 8)  | (u64)buf[3];
+  return ((u64)p[0] << 56) | ((u64)p[1] << 48) |
+         ((u64)p[2] << 40) | ((u64)p[3] << 32) |
+         ((u64)p[4] << 24) | ((u64)p[5] << 16) |
+         ((u64)p[6] << 8)  | (u64)p[7];
 }
 
-inline u64 read16be(u8 *buf) {
+template <typename T>
+inline u64 read32be(T *buf) {
   u8 *p = (u8 *)buf;
-  return ((u64)buf[0] << 8)  | (u64)buf[1];
+  return ((u64)p[0] << 24) | ((u64)p[1] << 16) |
+         ((u64)p[2] << 8)  | (u64)p[3];
 }
 
-inline void write64be(u8 *buf, u64 val) {
-  buf[0] = val >> 56;
-  buf[1] = val >> 48;
-  buf[2] = val >> 40;
-  buf[3] = val >> 32;
-  buf[4] = val >> 24;
-  buf[5] = val >> 16;
-  buf[6] = val >> 8;
-  buf[7] = val;
+template <typename T>
+inline u64 read16be(T *buf) {
+  u8 *p = (u8 *)buf;
+  return ((u64)p[0] << 8) | (u64)p[1];
 }
 
-inline void write32be(u8 *buf, u64 val) {
-  buf[0] = val >> 24;
-  buf[1] = val >> 16;
-  buf[2] = val >> 8;
-  buf[3] = val;
+template <typename T>
+inline void write64be(T *buf, u64 val) {
+  u8 *p = (u8 *)buf;
+  p[0] = val >> 56;
+  p[1] = val >> 48;
+  p[2] = val >> 40;
+  p[3] = val >> 32;
+  p[4] = val >> 24;
+  p[5] = val >> 16;
+  p[6] = val >> 8;
+  p[7] = val;
 }
 
-inline void write16be(u8 *buf, u64 val) {
-  buf[0] = val >> 8;
-  buf[1] = val;
+template <typename T>
+inline void write32be(T *buf, u32 val) {
+  u8 *p = (u8 *)buf;
+  p[0] = val >> 24;
+  p[1] = val >> 16;
+  p[2] = val >> 8;
+  p[3] = val;
+}
+
+template <typename T>
+inline void write16be(T *buf, u32 val) {
+  u8 *p = (u8 *)buf;
+  p[0] = val >> 8;
+  p[1] = val;
 }
 
 inline i64 write_string(u8 *buf, std::string_view str) {