Skip to content

Fix 64-bit file offsets on Windows #348

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
YayoRazo wants to merge 1 commit into
base: main
Choose a base branch
from
+77 −19
Open

Fix 64-bit file offsets on Windows #348

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 9 additions & 0 deletions Sources/System/Internals/Exports.swift
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -32,7 +32,16 @@ import Android
#error("Unsupported Platform")
#endif

#if os(Windows)
// On Windows, C `off_t` is `long`, i.e. 32-bit, so it cannot represent file
// offsets or sizes at or beyond 2 GiB. Use a 64-bit offset type so that
// seeking, positioned I/O, and resizing work on large files, matching the
// 64-bit `off_t` on Linux and Darwin. The Windows syscall adapters route
// this through the 64-bit `_lseeki64` / `OVERLAPPED` / `LARGE_INTEGER` APIs.

@glessard glessard Jul 17, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Could you make this comment less wordy? It's probably enough to say that we use a 64-bit value because off_t is a 32-bit type on Windows.

internal typealias _COffT = Int64

@compnerd compnerd Jul 17, 2026

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think I would slightly prefer _off64_t as the spelling. This is still Int64 under the hood, but more explicit and uses the libc spelling.

#else
internal typealias _COffT = off_t
#endif

// MARK: syscalls and variables

Expand Down
10 changes: 5 additions & 5 deletions Sources/System/Internals/Syscalls.swift
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -72,7 +72,7 @@ internal func system_read(

// pread
internal func system_pread(
_ fd: Int32, _ buf: UnsafeMutableRawPointer?, _ nbyte: Int, _ offset: off_t
_ fd: Int32, _ buf: UnsafeMutableRawPointer?, _ nbyte: Int, _ offset: _COffT
) -> Int {
#if ENABLE_MOCKING
if mockingEnabled { return _mockInt(fd, buf, nbyte, offset) }
Expand All @@ -90,8 +90,8 @@ internal func system_pread(

// lseek
internal func system_lseek(
_ fd: Int32, _ off: off_t, _ whence: Int32
) -> off_t {
_ fd: Int32, _ off: _COffT, _ whence: Int32
) -> _COffT {
#if ENABLE_MOCKING
if mockingEnabled { return _mockOffT(fd, off, whence) }
#endif
Expand All @@ -110,7 +110,7 @@ internal func system_write(

// pwrite
internal func system_pwrite(
_ fd: Int32, _ buf: UnsafeRawPointer?, _ nbyte: Int, _ offset: off_t
_ fd: Int32, _ buf: UnsafeRawPointer?, _ nbyte: Int, _ offset: _COffT
) -> Int {
#if ENABLE_MOCKING
if mockingEnabled { return _mockInt(fd, buf, nbyte, offset) }
Expand Down Expand Up @@ -165,7 +165,7 @@ internal func system_pipe2(_ fds: UnsafeMutablePointer<Int32>, _ oflag: Int32) -
}
#endif

internal func system_ftruncate(_ fd: Int32, _ length: off_t) -> Int32 {
internal func system_ftruncate(_ fd: Int32, _ length: _COffT) -> Int32 {
#if ENABLE_MOCKING
if mockingEnabled { return _mock(fd, length) }
#endif
Expand Down
25 changes: 11 additions & 14 deletions Sources/System/Internals/WindowsSyscallAdapters.swift
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -121,13 +121,6 @@ internal func write(
Int(_write(fd, buf, numericCast(nbyte)))
}

@inline(__always)
internal func lseek(
_ fd: Int32, _ off: off_t, _ whence: Int32
) -> off_t {
_lseek(fd, off, whence)
}

@inline(__always)
internal func dup(_ fd: Int32) -> Int32 {
_dup(fd)
Expand All @@ -144,7 +137,7 @@ internal func dup2(_ fd: Int32, _ fd2: Int32) -> Int32 {

@inline(__always)
internal func pread(
_ fd: Int32, _ buf: UnsafeMutableRawPointer!, _ nbyte: Int, _ offset: off_t
_ fd: Int32, _ buf: UnsafeMutableRawPointer!, _ nbyte: Int, _ offset: Int64

@compnerd compnerd Jul 17, 2026

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

These can then use the same spelling

) -> Int {
let handle: intptr_t = _get_osfhandle(fd)
if handle == /* INVALID_HANDLE_VALUE */ -1 { ucrt._set_errno(EBADF); return -1 }
Expand All @@ -159,8 +152,10 @@ internal func pread(
let hFile: HANDLE = HANDLE(bitPattern: handle)!

var ovlOverlapped: OVERLAPPED = OVERLAPPED()
ovlOverlapped.OffsetHigh = DWORD(UInt32(offset >> 32) & 0xffffffff)
ovlOverlapped.Offset = DWORD(UInt32(offset >> 0) & 0xffffffff)
// Split the 64-bit offset into high/low DWORDs. Use truncating conversions:
// `DWORD(UInt32(offset))` would trap for any offset >= 4 GiB.
Comment on lines +155 to +156

@glessard glessard Jul 17, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This comment isn't necessary. The truncating API is wordy for a reason!

ovlOverlapped.OffsetHigh = DWORD(truncatingIfNeeded: offset >> 32)
ovlOverlapped.Offset = DWORD(truncatingIfNeeded: offset)

var nNumberOfBytesRead: DWORD = 0
if !ReadFile(hFile, buf, DWORD(nbyte), &nNumberOfBytesRead, &ovlOverlapped) {
Expand All @@ -172,7 +167,7 @@ internal func pread(

@inline(__always)
internal func pwrite(
_ fd: Int32, _ buf: UnsafeRawPointer!, _ nbyte: Int, _ offset: off_t
_ fd: Int32, _ buf: UnsafeRawPointer!, _ nbyte: Int, _ offset: Int64
) -> Int {
let handle: intptr_t = _get_osfhandle(fd)
if handle == /* INVALID_HANDLE_VALUE */ -1 { ucrt._set_errno(EBADF); return -1 }
Expand All @@ -187,8 +182,10 @@ internal func pwrite(
let hFile: HANDLE = HANDLE(bitPattern: handle)!

var ovlOverlapped: OVERLAPPED = OVERLAPPED()
ovlOverlapped.OffsetHigh = DWORD(UInt32(offset >> 32) & 0xffffffff)
ovlOverlapped.Offset = DWORD(UInt32(offset >> 0) & 0xffffffff)
// Split the 64-bit offset into high/low DWORDs. Use truncating conversions:
// `DWORD(UInt32(offset))` would trap for any offset >= 4 GiB.
Comment on lines +185 to +186

@glessard glessard Jul 17, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Same as above

ovlOverlapped.OffsetHigh = DWORD(truncatingIfNeeded: offset >> 32)
ovlOverlapped.Offset = DWORD(truncatingIfNeeded: offset)

var nNumberOfBytesWritten: DWORD = 0
if !WriteFile(hFile, buf, DWORD(nbyte), &nNumberOfBytesWritten,
Expand All @@ -214,7 +211,7 @@ internal func csystem_posix_pipe2(
}

@inline(__always)
internal func ftruncate(_ fd: Int32, _ length: off_t) -> Int32 {
internal func ftruncate(_ fd: Int32, _ length: Int64) -> Int32 {
let handle: intptr_t = _get_osfhandle(fd)
if handle == /* INVALID_HANDLE_VALUE */ -1 { ucrt._set_errno(EBADF); return -1 }

Expand Down
52 changes: 52 additions & 0 deletions Tests/SystemTests/FileOperationsTestWindows.swift
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -328,6 +328,58 @@ final class FileOperationsTestWindows: XCTestCase {
}
}
}

/// Regression test for 64-bit file offsets on Windows.
///
/// C `off_t` is 32-bit on Windows, so before the offset type was widened,
/// seeking or performing positioned I/O at or beyond 2 GiB trapped at
/// runtime, and the `OVERLAPPED` high dword was always zero so offsets past
/// 4 GiB were unreachable.
func testLargeFileOffsets() throws {
try withTemporaryFilePath(basename: "testLargeFileOffsets") { path in
let fd = try FileDescriptor.open(
path.appending("large.bin"), .readWrite,
options: [.create, .truncate],
permissions: .ownerReadWrite
)
defer { try? fd.close() }

// 2 GiB is > Int32.max; 5 GiB is > UInt32.max. Both trapped before the fix.
let twoGiB: Int64 = 1 << 31
let fiveGiB: Int64 = 5 << 30

// Seeking allocates no storage; it exercises `_lseeki64` and must
// round-trip the full 64-bit position rather than trapping or truncating.
XCTAssertEqual(try fd.seek(offset: twoGiB, from: .start), twoGiB)
XCTAssertEqual(try fd.seek(offset: fiveGiB, from: .start), fiveGiB)

// Positioned read/write beyond 4 GiB exercises both dwords of the
// `OVERLAPPED` offset. Mark the file sparse first so the test does not
// allocate several gigabytes of real storage.
let handle = try XCTUnwrap(HANDLE(bitPattern: _get_osfhandle(fd.rawValue)))
var bytesReturned: DWORD = 0
let FSCTL_SET_SPARSE: DWORD = 0x000900C4
try XCTSkipUnless(
DeviceIoControl(handle, FSCTL_SET_SPARSE, nil, 0, nil, 0,
&bytesReturned, nil),
Comment on lines +363 to +364

@glessard glessard Jul 17, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Our coding style is to have all the parameters on one line, or to have one line per parameter.

Suggested change
DeviceIoControl(handle, FSCTL_SET_SPARSE, nil, 0, nil, 0,
&bytesReturned, nil),
DeviceIoControl(
handle, FSCTL_SET_SPARSE, nil, 0, nil, 0, &bytesReturned, nil
),

"filesystem does not support sparse files"
)

let marker = Array("swift-system".utf8)
let offset = fiveGiB + 123
let written = try marker.withUnsafeBytes {
try fd.write(toAbsoluteOffset: offset, $0)
}
XCTAssertEqual(written, marker.count)

var readBack = [UInt8](repeating: 0, count: marker.count)
let read = try readBack.withUnsafeMutableBytes {
try fd.read(fromAbsoluteOffset: offset, into: $0)
}
XCTAssertEqual(read, marker.count)
XCTAssertEqual(readBack, marker)
}
}
}

#endif // os(Windows)
Loading