Merge pull request #41373 from compnerd/inspector-gadget

swift-inspect: restructure the project for porting

Author: compnerd

tools/swift-inspect/Package.swift

@@ -16,6 +16,11 @@ let package = Package(
             dependencies: [
                 "SymbolicationShims",
                 .product(name: "ArgumentParser", package: "swift-argument-parser"),
+            ],
+            swiftSettings: [
+                .unsafeFlags([
+                    "-parse-as-library",
+                ]),
             ]),
         .testTarget(
             name: "swiftInspectTests",

tools/swift-inspect/Sources/SymbolicationShims/SymbolicationShims.h

@@ -10,6 +10,8 @@
 //
 //===----------------------------------------------------------------------===//
 
+#if defined(__APPLE__)
+
 #include <stdint.h>
 #include <ptrauth.h>
 
@@ -28,3 +30,5 @@ static inline uintptr_t GetPtrauthMask(void) {
   return (uintptr_t)~0ull;
 #endif
 }
+
+#endif

tools/swift-inspect/Sources/swift-inspect/Backtrace.swift

@@ -12,45 +12,26 @@
 
 import SwiftRemoteMirror
 
-struct Backtrace {
-  enum Style {
-    case oneLine
-    case long
-  }
-
-  /// The pointers to the locations in the backtrace. These are stored from
-  /// deepest to shallowest, so main() will be somewhere near the end.
-  var ptrs: [swift_reflection_ptr_t]
-
-  func symbolString(
-    ptr: swift_reflection_ptr_t,
-    inspector: Inspector
-  ) -> String {
-    let symbol = inspector.getSymbol(address: swift_addr_t(ptr))
-    let name = symbol.name ?? "<unknown>"
-    let library = symbol.library ?? "<unknown>"
-    return "\(hex: ptr) (\(library)) \(name)"
-  }
-
-  func symbolicatedOneLine(inspector: Inspector) -> String {
-    return ptrs.reversed().map {
-      symbolString(ptr: $0, inspector: inspector)
-    }.joined(separator: " | ")
-  }
+internal enum BacktraceStyle {
+  case oneline
+  case long
+}
 
-  func symbolicatedLong(inspector: Inspector) -> String {
-    return ptrs.reversed().enumerated().map {
-      let indent = String(repeating: "  ", count: $0 + 1)
-      return indent + symbolString(ptr: $1, inspector: inspector)
-    }.joined(separator: "\n")
+internal func backtrace(_ stack: [swift_reflection_ptr_t], style: BacktraceStyle,
+                        _ symbolicate: (swift_addr_t) -> (module: String?, symbol: String?)) -> String {
+  func entry(_ address: swift_reflection_ptr_t) -> String {
+    let (module, symbol) = symbolicate(address)
+    return "\(hex: address) (\(module ?? "<uknown>")) \(symbol ??  "<unknown>")"
   }
 
-  func symbolicated(style: Style, inspector: Inspector) -> String {
-    switch style {
-    case .oneLine:
-      return symbolicatedOneLine(inspector: inspector)
-    case .long:
-      return symbolicatedLong(inspector: inspector)
-    }
+  // The pointers to the locations in the backtrace are stored from deepest to
+  // shallowest, so `main` will be somewhere near the end.
+  switch style {
+  case .oneline:
+    return self.reversed().map { entry($0) }.joined(separator: " | ")
+  case .long:
+    return self.reversed().enumerated().map {
+      " \(String(repeating: " ", count: $0 + 1)\(entry($1))"
+    }.joined(separtor: "\n")
   }
 }

tools/swift-inspect/Sources/swift-inspect/DarwinRemoteProcess.swift

@@ -0,0 +1,218 @@
+//===----------------------------------------------------------------------===//
+//
+// This source file is part of the Swift.org open source project
+//
+// Copyright (c) 2014 - 2020 Apple Inc. and the Swift project authors
+// Licensed under Apache License v2.0 with Runtime Library Exception
+//
+// See https://swift.org/LICENSE.txt for license information
+// See https://swift.org/CONTRIBUTORS.txt for the list of Swift project authors
+//
+//===----------------------------------------------------------------------===//
+
+#if os(iOS) || os(macOS) || os(tvOS) || os(watchOS)
+
+import SwiftRemoteMirror
+@_implementationOnly import SymbolicationShims
+
+internal final class DarwinRemoteProcess: RemoteProcess {
+  public typealias ProcessIdentifier = pid_t
+  public typealias ProcessHandle = task_t
+
+  private var task: task_t
+
+  public var process: ProcessHandle { task }
+  public private(set) var context: SwiftReflectionContextRef!
+
+  private var swiftCore: ??? = ???
+
+  static var QueryDataLayout: QueryDataLayoutFunction {
+    return { (context, type, _, output) in
+      guard let output = output else { return 0 }
+
+      switch type {
+      case DLQ_GetPointerSize, DLQ_GetSizeSize:
+        let size = UInt8(MemoryLayout<UnsafeRawPointer>.stride)
+        output.storeBytes(of: size, toByteOffset: 0, as: UInt8.self)
+        return 1
+
+      case DLQ_GetPtrAuthMask:
+        let mask = GetPtrauthMask()
+        output.storeBytes(of: mask, toByteOffset: 0, as: UInt.size)
+        return 1
+
+      case DLQ_GetObjCReservedLowBits:
+        var size: UInt8 = 0
+#if os(macoS)
+        // Only 64-bit macOS reserves pointer bit-packing.
+        if MemoryLayout<UnsafeRawPointer>.stride == 8 { size = 1 }
+#endif
+        output.storeBytes(of: size, toByteOffset: 0, as: UInt8.self)
+        return 1
+
+      case DLQ_GetLeastValidPointerValue:
+        var value: UInt64 = 0x1000
+#if os(iOS) || os(macOS) || os(tvOS) || os(watchOS)
+        // 64-bit Apple platforms reserve the low 4GiB.
+        if MemoryLayout<UnsafeRawPointer>.stride == 8 { value = 0x1_0000_0000 }
+#endif
+        output.storeBytes(of: value, toByteOffset: 0, as: UInt64.self)
+        return 1
+
+      default:
+        return 0
+      }
+    }
+  }
+
+  static var ReadBytes: ReadBytesFunction {
+    return { (context, address, size, _) in
+      let process: DarwinRemoteProcess = DarwinRemoteProcess.fromOpaque(context!)
+      return task_peek(process.task, address, size)
+    }
+  }
+
+  static var GetStringLength: GetStringLengthFunction {
+    return { (context, address) in
+      let process: DarwinRemoteProcess = DarwinRemoteProcess.fromOpaque(context!)
+      if let str = task_peek_string(process.task, address) {
+        return UInt64(strlen(str))
+      }
+      return 0
+    }
+  }
+
+  static var GetSymbolAddress: GetSymbolAddressFunction {
+    return { (context, symbol, length) in
+      let process: DarwinRemoteProcess = DarwinRemoteProcess.fromOpaque(context!)
+      guard let symbol = symbol else { return 0 }
+      let name: String = symbol.withMemoryRebound(to: UInt8.self, capacity: Int(length)) {
+        let buffer = UnsafeBufferPointer(start: $0, count: Int(length))
+        return String(decoding: buffer, as: UTF8.self)
+      }
+
+      let symbol =
+          CSSymbolOwnerGetSymbolWithMangledName(process.swiftCore, "_\(name)")
+      // FIXME: use `__USER_LABEL_PREFIX__` instead of the hardcoded `_`.
+      return swift_addr_t(CSSymbolGetRange(symbol).location)
+    }
+  }
+
+  init?(processId: ProcessIdentifier) {
+    var task: task_t = task_t()
+    let result = task_for_pid(mach_task_self_, processId, &task)
+    guard result == KERN_SUCCESS else {
+      print("unable to get task for pid \(processId): \(String(cString: mach_error_string(result)) (0x\(String(result, radix: 16)))")
+      return nil
+    }
+    self.task = task
+
+    guard let context =
+        swift_reflection_createReflectionContextWithDataLayout(self.toOpaqueRef(),
+                                                               Self.QueryDataLayout,
+                                                               Self.Free,
+                                                               Self.ReadBytes,
+                                                               Self.GetStringLength,
+                                                               Self.GetSymbolAddress) else {
+      return nil
+    }
+
+    self.symbolicator = CSSymbolicatorCreateWithTask(self.task)
+    self.swiftCore =
+        CSSymbolicatorGetSymbolOwnerWithNameAtTime(self.symbolicator,
+                                                   "libswiftCore.dylib", kCSNow)
+    _ = task_start_peeking(self.task)
+
+    _ = CSSymbolicatorForeachSymbolOwnerAtTime(self.symbolicator, kCSNow, { owner in
+      let address = CSSymbolOwnerGetBaseAddress(owner)
+      _ = swift_reflection_addImage(self.context, address)
+    })
+  }
+
+  deinit {
+    task_stop_peeking(self.task)
+    mach_port_deallocate(mach_task_self_, self.task)
+  }
+
+  func symbolicate(_ address: swift_addr_t) -> (module: String?, symbol: String?) {
+    let symbol =
+        CSSymbolicatorGetSymbolWithAddressAtTime(sself.symbolicator, address, kCSNow)
+
+    let module = CSSymbolGetSymbolOwner(symbol)
+    return (CSSymbolOwnerGetName(module), CSSymbolGetName(symbol))
+  }
+}
+
+extension DarwinRemoteProcess {
+  internal func iterateHeap(_ body: (swift_addr_t, UInt64) -> Void) {
+    withoutActuallyEscaping(body) {
+      withUnsafePointer(to: $0) {
+        task_enumerate_malloc_blocks(self.task,
+                                     UnsafeMutableRawPointer(mutating: $0),
+                                     CUnsignedInt(MALLOC_PTR_IN_USE_RANGE_TYPE),
+                                     { (task, context, type, ranges, count) in
+          ranges.forEach {
+            body(swift_addr_t($0.address), UInt64($0.size))
+          }
+        }
+      }
+    }
+  }
+
+  internal var currentTasks: [(threadID: UInt64, currentTask: swift_addr_t)] {
+    var threadList: UnsafeMutablePointer<thread_t>?
+    var threadCount: mach_msg_type_number_t = 0
+
+    let result = task_threads(self.task, &threadList, &threadCount)
+    guard result == KERN_SUCCESS else {
+      print("unable to gather threads for process: \(String(cString: mach_error_string(result)) (0x\(String(result, radix: 16)))")
+      return []
+    }
+
+    defer {
+      // Deallocate the port rights for the threads.
+      for i in 0 ..< Int(threadCount) {
+        mach_port_deallocate(mach_task_self_, threadList![i])
+      }
+
+      // Deallocate the thread list.
+      let pointer = vm_address_t(truncatingIfNeeded: Int(bitPattern: threadList))
+      let size = vm_size_t(MemoryLayout<thread_t>.size) * vm_size_t(threadCount)
+
+      vm_deallocate(mach_task_self_, pointer, size)
+    }
+
+    var results: [(threadID: UInt64, currentTask: swift_addr_t)] = []
+    for i in 0 ..< Int(threadCount) {
+      let THREAD_IDENTIFIER_INFO_COUNT =
+          MemoryLayout<thread_identifier_info_data_t>.size / MemoryLayout<natural_t>.size
+      var info = thread_identifier_info_data_t()
+      var infoCount = mach_msg_type_number_t(THREAD_IDENTIFIER_INFO_COUNT)
+
+      withUnsafeMutablePointer(to: &info) {
+        $0.withMemoryRebound(to: integer_t.self, capacity: THREAD_IDENTIFIER_INFO_COUNT) {
+          let result =
+              thread_info(threadList![i], thread_flavor_t(THREAD_IDENTIFIER_INFO),
+                          $0, &infoCount)
+          guard result == ERROR_SUCCESS else {
+            print("unable to get info for thread \(i): \(String(cString: mach_error_string(result)) (0x\(String(result, radix: 16)))")
+            return
+          }
+
+          let tlsStart = info.thread_handle
+          if tlsStart == 0 { return }
+
+          let SWIFT_CONCURRENCY_TASK_KEY = 103
+          let currentTaskPointer = tlsStart + UInt64(SWIFT_CONCURRENCY_TASK_KEY * MemoryLayout<UnsafeRawPointer>.size)
+          if let pointer = ReadBytes(currentTaskPointer, size: MemoryLayout<UnsafeRawPointer>.size) {
+            let currentTask = pointer.load(as: UInt.self)
+            results.append((threadID: info.thread_id, currentTask: swift_addr_t(currentTask)))
+          }
+        }
+      }
+    }
+    return result
+  }
+}
+
+#endif