macho: extract parallel hasher into a generic helper struct

3 files changed, 67 insertions, 59 deletions

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 16c7dd3d0e..7a5726bdc5 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -594,6 +594,7 @@ set(ZIG_STAGE2_SOURCES
     "${CMAKE_SOURCE_DIR}/src/link/MachO/dead_strip.zig"
     "${CMAKE_SOURCE_DIR}/src/link/MachO/eh_frame.zig"
     "${CMAKE_SOURCE_DIR}/src/link/MachO/fat.zig"
+    "${CMAKE_SOURCE_DIR}/src/link/MachO/hasher.zig"
     "${CMAKE_SOURCE_DIR}/src/link/MachO/load_commands.zig"
     "${CMAKE_SOURCE_DIR}/src/link/MachO/thunks.zig"
     "${CMAKE_SOURCE_DIR}/src/link/MachO/zld.zig"
diff --git a/src/link/MachO/CodeSignature.zig b/src/link/MachO/CodeSignature.zig
index 02511dbe29..84c5b49362 100644
--- a/src/link/MachO/CodeSignature.zig
+++ b/src/link/MachO/CodeSignature.zig
@@ -7,11 +7,10 @@ const log = std.log.scoped(.link);
 const macho = std.macho;
 const mem = std.mem;
 const testing = std.testing;
-const ThreadPool = std.Thread.Pool;
-const WaitGroup = std.Thread.WaitGroup;
 
 const Allocator = mem.Allocator;
 const Compilation = @import("../../Compilation.zig");
+const Hasher = @import("hasher.zig").ParallelHasher;
 const Sha256 = std.crypto.hash.sha2.Sha256;
 
 const hash_size = Sha256.digest_length;
@@ -289,7 +288,11 @@ pub fn writeAdhocSignature(
     self.code_directory.inner.nCodeSlots = total_pages;
 
     // Calculate hash for each page (in file) and write it to the buffer
-    try self.parallelHash(gpa, comp.thread_pool, opts.file, opts.file_size);
+    var hasher = Hasher(Sha256){};
+    try hasher.hash(gpa, comp.thread_pool, opts.file, self.code_directory.code_slots.items, .{
+        .chunk_size = self.page_size,
+        .max_file_size = opts.file_size,
+    });
 
     try blobs.append(.{ .code_directory = &self.code_directory });
     header.length += @sizeOf(macho.BlobIndex);
@@ -348,62 +351,6 @@ pub fn writeAdhocSignature(
     }
 }
 
-fn parallelHash(
-    self: *CodeSignature,
-    gpa: Allocator,
-    pool: *ThreadPool,
-    file: fs.File,
-    file_size: u32,
-) !void {
-    var wg: WaitGroup = .{};
-
-    const total_num_chunks = mem.alignForward(usize, file_size, self.page_size) / self.page_size;
-    assert(self.code_directory.code_slots.items.len >= total_num_chunks);
-
-    const buffer = try gpa.alloc(u8, self.page_size * total_num_chunks);
-    defer gpa.free(buffer);
-
-    const results = try gpa.alloc(fs.File.PReadError!usize, total_num_chunks);
-    defer gpa.free(results);
-
-    {
-        wg.reset();
-        defer wg.wait();
-
-        var i: usize = 0;
-        while (i < total_num_chunks) : (i += 1) {
-            const fstart = i * self.page_size;
-            const fsize = if (fstart + self.page_size > file_size)
-                file_size - fstart
-            else
-                self.page_size;
-            wg.start();
-            try pool.spawn(worker, .{
-                file,
-                fstart,
-                buffer[fstart..][0..fsize],
-                &self.code_directory.code_slots.items[i],
-                &results[i],
-                &wg,
-            });
-        }
-    }
-    for (results) |result| _ = try result;
-}
-
-fn worker(
-    file: fs.File,
-    fstart: usize,
-    buffer: []u8,
-    out: *[hash_size]u8,
-    err: *fs.File.PReadError!usize,
-    wg: *WaitGroup,
-) void {
-    defer wg.finish();
-    err.* = file.preadAll(buffer, fstart);
-    Sha256.hash(buffer, out, .{});
-}
-
 pub fn size(self: CodeSignature) u32 {
     var ssize: u32 = @sizeOf(macho.SuperBlob) + @sizeOf(macho.BlobIndex) + self.code_directory.size();
     if (self.requirements) |req| {
diff --git a/src/link/MachO/hasher.zig b/src/link/MachO/hasher.zig
new file mode 100644
index 0000000000..d7bf6888b5
--- /dev/null
+++ b/src/link/MachO/hasher.zig
@@ -0,0 +1,60 @@
+const std = @import("std");
+const assert = std.debug.assert;
+const fs = std.fs;
+const mem = std.mem;
+
+const Allocator = mem.Allocator;
+const ThreadPool = std.Thread.Pool;
+const WaitGroup = std.Thread.WaitGroup;
+
+pub fn ParallelHasher(comptime Hasher: type) type {
+    const hash_size = Hasher.digest_length;
+
+    return struct {
+        pub fn hash(self: @This(), gpa: Allocator, pool: *ThreadPool, file: fs.File, out: [][hash_size]u8, opts: struct {
+            chunk_size: u16 = 0x4000,
+            max_file_size: ?u64 = null,
+        }) !void {
+            _ = self;
+
+            var wg: WaitGroup = .{};
+
+            const file_size = opts.max_file_size orelse try file.getEndPos();
+            const total_num_chunks = mem.alignForward(u64, file_size, opts.chunk_size) / opts.chunk_size;
+            assert(out.len >= total_num_chunks);
+
+            const buffer = try gpa.alloc(u8, opts.chunk_size * total_num_chunks);
+            defer gpa.free(buffer);
+
+            const results = try gpa.alloc(fs.File.PReadError!usize, total_num_chunks);
+            defer gpa.free(results);
+
+            {
+                wg.reset();
+                defer wg.wait();
+
+                var i: usize = 0;
+                while (i < total_num_chunks) : (i += 1) {
+                    const fstart = i * opts.chunk_size;
+                    const fsize = if (fstart + opts.chunk_size > file_size) file_size - fstart else opts.chunk_size;
+                    wg.start();
+                    try pool.spawn(worker, .{ file, fstart, buffer[fstart..][0..fsize], &out[i], &results[i], &wg });
+                }
+            }
+            for (results) |result| _ = try result;
+        }
+
+        fn worker(
+            file: fs.File,
+            fstart: usize,
+            buffer: []u8,
+            out: *[hash_size]u8,
+            err: *fs.File.PReadError!usize,
+            wg: *WaitGroup,
+        ) void {
+            defer wg.finish();
+            err.* = file.preadAll(buffer, fstart);
+            Hasher.hash(buffer, out, .{});
+        }
+    };
+}