buffer: support aligned allocations

doc/api/buffer.md

@@ -793,11 +793,14 @@ data that might not have been allocated for `Buffer`s.
 
 A `TypeError` will be thrown if `size` is not a number.
 
-### Static method: `Buffer.allocUnsafe(size)`
+### Static method: `Buffer.allocUnsafe(size[, alignment])`
 
 <!-- YAML
 added: v5.10.0
 changes:
+  - version: REPLACEME
+    pr-url: https://github.com/nodejs/node/pull/65003
+    description: Added the `alignment` argument.
   - version: v20.0.0
     pr-url: https://github.com/nodejs/node/pull/45796
     description: Throw ERR_INVALID_ARG_TYPE or ERR_OUT_OF_RANGE instead of
@@ -812,6 +815,9 @@ changes:
 -->
 
 * `size` {integer} The desired length of the new `Buffer`.
+* `alignment` {integer} If given, the memory backing the new `Buffer` will start
+  at an address that is a multiple of `alignment`. Must be a power of two no
+  larger than `2 ** 30`. See [Aligned allocations][].
 * Returns: {Buffer}
 
 Allocates a new `Buffer` of `size` bytes. If `size` is larger than
@@ -867,11 +873,14 @@ pool, while `Buffer.allocUnsafe(size).fill(fill)` _will_ use the internal
 difference is subtle but can be important when an application requires the
 additional performance that [`Buffer.allocUnsafe()`][] provides.
 
-### Static method: `Buffer.allocUnsafeSlow(size)`
+### Static method: `Buffer.allocUnsafeSlow(size[, alignment])`
 
 <!-- YAML
 added: v5.12.0
 changes:
+  - version: REPLACEME
+    pr-url: https://github.com/nodejs/node/pull/65003
+    description: Added the `alignment` argument.
   - version: v20.0.0
     pr-url: https://github.com/nodejs/node/pull/45796
     description: Throw ERR_INVALID_ARG_TYPE or ERR_OUT_OF_RANGE instead of
@@ -883,6 +892,9 @@ changes:
 -->
 
 * `size` {integer} The desired length of the new `Buffer`.
+* `alignment` {integer} If given, the memory backing the new `Buffer` will start
+  at an address that is a multiple of `alignment`. Must be a power of two no
+  larger than `2 ** 30`. See [Aligned allocations][].
 * Returns: {Buffer}
 
 Allocates a new `Buffer` of `size` bytes. If `size` is larger than
@@ -5602,16 +5614,85 @@ While there are clear performance advantages to using
 [`Buffer.allocUnsafe()`][], extra care _must_ be taken in order to avoid
 introducing security vulnerabilities into an application.
 
+### Aligned allocations
+
+Some operating system interfaces require the memory they operate on to be
+aligned, and on some hardware alignment is merely faster. The most common
+example of the former is unbuffered ("direct") file I/O, which on Linux requires
+the buffer address, the file offset and the transfer length to all be multiples
+of the logical block size of the underlying device:
+
+```mjs
+import { open } from 'node:fs/promises';
+import { constants } from 'node:fs';
+import { Buffer } from 'node:buffer';
+
+const blockSize = 4096;
+
+// The buffer address must be block-aligned for O_DIRECT to accept it.
+const buf = Buffer.allocUnsafeSlow(blockSize, blockSize);
+
+const file = await open('/dev/sda', constants.O_RDONLY | constants.O_DIRECT);
+try {
+  await file.read(buf, 0, blockSize, 0);
+} finally {
+  await file.close();
+}
+```
+
+```cjs
+const fs = require('node:fs');
+const { Buffer } = require('node:buffer');
+
+const blockSize = 4096;
+
+// The buffer address must be block-aligned for O_DIRECT to accept it.
+const buf = Buffer.allocUnsafeSlow(blockSize, blockSize);
+
+const flags = fs.constants.O_RDONLY | fs.constants.O_DIRECT;
+fs.open('/dev/sda', flags, (err, fd) => {
+  if (err) throw err;
+  fs.read(fd, buf, 0, blockSize, 0, (err) => {
+    fs.close(fd, () => {});
+    if (err) throw err;
+  });
+});
+```
+
+Alignment can also be worth requesting purely for performance, even when no
+interface demands it. Aligning a hot `Buffer` to the cache line size (64 bytes on
+most contemporary CPUs) keeps it from straddling one more cache line than it
+needs to, so that a small structure is fetched with one cache miss instead of
+two, and page-aligned (4096 bytes) allocations similarly help interfaces that map
+or pin memory. These are micro-optimizations: measure before reaching for them,
+since the extra bytes are not free.
+
+Because the address of a `Buffer`'s memory cannot be chosen directly, extra bytes
+have to be allocated or skipped to reach an aligned address.
+[`Buffer.allocUnsafeSlow()`][] over-allocates up to `alignment - 1` bytes and
+positions the returned `Buffer` at the first suitably aligned byte within them.
+[`Buffer.allocUnsafe()`][] instead pads its offset into the shared internal pool,
+whose start is always aligned to 64 bytes, and only falls back to an allocation
+of its own when `alignment` is larger than that. Either way,
+[`buf.byteOffset`][] is usually not 0 and [`buf.buffer`][] is larger than `size`,
+so code that reaches past the `Buffer` into its underlying `ArrayBuffer` must
+take the offset into account, as it must for pooled `Buffer`s.
+
+The alignment is a property of the returned `Buffer` and is preserved for its
+whole lifetime, but it is not inherited by other views: [`buf.subarray`][],
+[`buf.slice()`][] and `structuredClone()` may all produce unaligned `Buffer`s.
+
 [ASCII]: https://en.wikipedia.org/wiki/ASCII
+[Aligned allocations]: #aligned-allocations
 [Base64]: https://en.wikipedia.org/wiki/Base64
 [ISO-8859-1]: https://en.wikipedia.org/wiki/ISO-8859-1
 [RFC 4648, Section 5]: https://tools.ietf.org/html/rfc4648#section-5
 [UTF-16]: https://en.wikipedia.org/wiki/UTF-16
 [UTF-8]: https://en.wikipedia.org/wiki/UTF-8
 [WHATWG Encoding Standard]: https://encoding.spec.whatwg.org/
 [`Buffer.alloc()`]: #static-method-bufferallocsize-fill-encoding
-[`Buffer.allocUnsafe()`]: #static-method-bufferallocunsafesize
-[`Buffer.allocUnsafeSlow()`]: #static-method-bufferallocunsafeslowsize
+[`Buffer.allocUnsafe()`]: #static-method-bufferallocunsafesize-alignment
+[`Buffer.allocUnsafeSlow()`]: #static-method-bufferallocunsafeslowsize-alignment
 [`Buffer.concat()`]: #static-method-bufferconcatlist-totallength
 [`Buffer.copyBytesFrom()`]: #static-method-buffercopybytesfromview-offset-length
 [`Buffer.from(array)`]: #static-method-bufferfromarray
@@ -5633,6 +5714,7 @@ introducing security vulnerabilities into an application.
 [`TypedArray.prototype.subarray()`]: https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/TypedArray/subarray
 [`blob.stream()`]: #blobstream
 [`buf.buffer`]: #bufbuffer
+[`buf.byteOffset`]: #bufbyteoffset
 [`buf.compare()`]: #bufcomparetarget-targetstart-targetend-sourcestart-sourceend
 [`buf.entries()`]: #bufentries
 [`buf.fill()`]: #buffillvalue-offset-end-encoding

doc/api/deprecations.md

@@ -4727,7 +4727,7 @@ calling or overriding `_listen2`.
 [`--pending-deprecation`]: cli.md#--pending-deprecation
 [`--throw-deprecation`]: cli.md#--throw-deprecation
 [`--unhandled-rejections`]: cli.md#--unhandled-rejectionsmode
-[`Buffer.allocUnsafeSlow(size)`]: buffer.md#static-method-bufferallocunsafeslowsize
+[`Buffer.allocUnsafeSlow(size)`]: buffer.md#static-method-bufferallocunsafeslowsize-alignment
 [`Buffer.from(array)`]: buffer.md#static-method-bufferfromarray
 [`Buffer.from(buffer)`]: buffer.md#static-method-bufferfrombuffer
 [`Buffer.isBuffer()`]: buffer.md#static-method-bufferisbufferobj
@@ -4862,7 +4862,7 @@ calling or overriding `_listen2`.
 [`writable.writableLength`]: stream.md#writablewritablelength
 [`zlib.bytesWritten`]: zlib.md#zlibbyteswritten
 [alloc]: buffer.md#static-method-bufferallocsize-fill-encoding
-[alloc_unsafe_size]: buffer.md#static-method-bufferallocunsafesize
+[alloc_unsafe_size]: buffer.md#static-method-bufferallocunsafesize-alignment
 [caveats of asynchronous customization hooks]: module.md#caveats-of-asynchronous-customization-hooks
 [from_arraybuffer]: buffer.md#static-method-bufferfromarraybuffer-byteoffset-length
 [from_string_encoding]: buffer.md#static-method-bufferfromstring-encoding

doc/api/worker_threads.md

@@ -2234,7 +2234,7 @@ thread spawned will spawn another until the application crashes.
 [`--max-old-space-size`]: cli.md#--max-old-space-sizesize-in-mib
 [`--max-semi-space-size`]: cli.md#--max-semi-space-sizesize-in-mib
 [`AsyncResource`]: async_hooks.md#class-asyncresource
-[`Buffer.allocUnsafe()`]: buffer.md#static-method-bufferallocunsafesize
+[`Buffer.allocUnsafe()`]: buffer.md#static-method-bufferallocunsafesize-alignment
 [`ERR_MISSING_MESSAGE_PORT_IN_TRANSFER_LIST`]: errors.md#err_missing_message_port_in_transfer_list
 [`ERR_WORKER_MESSAGING_ERRORED`]: errors.md#err_worker_messaging_errored
 [`ERR_WORKER_MESSAGING_FAILED`]: errors.md#err_worker_messaging_failed

lib/buffer.js

@@ -140,6 +140,7 @@ const {
   markAsUntransferable,
   addBufferPrototypeMethods,
   createUnsafeBuffer,
+  createUnsafeAlignedBuffer,
   asciiWrite,
   latin1Write,
   utf8Write,
@@ -171,13 +172,27 @@ const constants = ObjectDefineProperties({}, {
   },
 });
 
+// The largest alignment accepted by `Buffer.allocUnsafeSlow()`. Any plausible
+// I/O alignment requirement (logical block size, memory page size, huge page
+// size) is well below this.
+const kMaxAlignment = 2 ** 30;
+
+// Slices handed out of the pool are 8 byte aligned relative to the start of the
+// pool, so aligning the pool itself to a cache line keeps them from straddling
+// one more cache line than their size requires.
+const kPoolAlignment = 64;
+
 Buffer.poolSize = 64 * 1024;
-let poolSize, poolOffset, allocPool, allocBuffer;
+// `poolOffset` is relative to `poolBase`, which is where the pool starts inside
+// `allocPool`. The pool is over-allocated to be able to align it, so `poolBase`
+// is not necessarily 0.
+let poolSize, poolOffset, poolBase, allocPool, allocBuffer;
 
 function createPool() {
   poolSize = Buffer.poolSize;
-  allocBuffer = createUnsafeBuffer(poolSize);
+  allocBuffer = createUnsafeAlignedBuffer(poolSize, kPoolAlignment);
   allocPool = allocBuffer.buffer;
+  poolBase = TypedArrayPrototypeGetByteOffset(allocBuffer);
   markAsUntransferable(allocPool);
   poolOffset = 0;
 }
@@ -444,40 +459,92 @@ Buffer.alloc = function alloc(size, fill, encoding) {
 /**
  * Equivalent to Buffer(num), by default creates a non-zero-filled Buffer
  * instance. If `--zero-fill-buffers` is set, will zero-fill the buffer.
+ *
+ * If `alignment` is given, the memory backing the returned buffer starts at an
+ * address that is a multiple of `alignment`. See `Buffer.allocUnsafeSlow()`.
+ * @param {number} size
+ * @param {number} [alignment] A power of two, at most 2 ** 30
  * @returns {FastBuffer}
  */
-Buffer.allocUnsafe = function allocUnsafe(size) {
+Buffer.allocUnsafe = function allocUnsafe(size, alignment) {
   validateNumber(size, 'size', 0, kMaxLength);
-  return allocate(size);
+  if (alignment === undefined) {
+    return allocate(size);
+  }
+  validateAlignment(size, alignment);
+  return allocateAligned(size, alignment);
 };
 
 /**
  * By default creates a non-zero-filled Buffer instance that is not allocated
  * off the pre-initialized pool. If `--zero-fill-buffers` is set, will zero-fill
  * the buffer.
+ *
+ * If `alignment` is given, the memory backing the returned buffer starts at an
+ * address that is a multiple of `alignment`, which is required by e.g. reads
+ * and writes on file descriptors opened with `O_DIRECT`. Note that up to
+ * `alignment - 1` extra bytes are allocated to satisfy the request, and that
+ * the returned buffer's `byteOffset` is therefore usually non-zero.
  * @param {number} size
- * @returns {FastBuffer|undefined}
+ * @param {number} [alignment] A power of two, at most 2 ** 30
+ * @returns {FastBuffer}
  */
-Buffer.allocUnsafeSlow = function allocUnsafeSlow(size) {
+Buffer.allocUnsafeSlow = function allocUnsafeSlow(size, alignment) {
   validateNumber(size, 'size', 0, kMaxLength);
-  return createUnsafeBuffer(size);
+  if (alignment === undefined) {
+    return createUnsafeBuffer(size);
+  }
+  validateAlignment(size, alignment);
+  return createUnsafeAlignedBuffer(size, alignment);
 };
 
+function validateAlignment(size, alignment) {
+  validateInteger(alignment, 'alignment', 1, kMaxAlignment);
+  if ((alignment & (alignment - 1)) !== 0) {
+    throw new ERR_INVALID_ARG_VALUE(
+      'alignment', alignment, 'must be a power of two');
+  }
+  // Satisfying the alignment costs up to `alignment - 1` extra bytes.
+  if (size > kMaxLength - (alignment - 1)) {
+    throw new ERR_OUT_OF_RANGE(
+      'size', `<= ${kMaxLength - (alignment - 1)}`, size);
+  }
+}
+
 function allocate(size) {
   if (size <= 0) {
     return new FastBuffer();
   }
   if (size < (Buffer.poolSize >>> 1)) {
     if (size > (poolSize - poolOffset))
       createPool();
-    const b = new FastBuffer(allocPool, poolOffset, size);
+    const b = new FastBuffer(allocPool, poolBase + poolOffset, size);
     poolOffset += size;
     alignPool();
     return b;
   }
   return createUnsafeBuffer(size);
 }
 
+function allocateAligned(size, alignment) {
+  if (size <= 0) {
+    return new FastBuffer();
+  }
+  // The pool starts at a `kPoolAlignment` aligned address, so any alignment up
+  // to that can be satisfied by padding the offset into the pool. Stricter
+  // alignments need an allocation of their own.
+  if (alignment > kPoolAlignment || size >= (Buffer.poolSize >>> 1)) {
+    return createUnsafeAlignedBuffer(size, alignment);
+  }
+  poolOffset = (poolOffset + alignment - 1) & ~(alignment - 1);
+  if (size > (poolSize - poolOffset))
+    createPool();
+  const b = new FastBuffer(allocPool, poolBase + poolOffset, size);
+  poolOffset += size;
+  alignPool();
+  return b;
+}
+
 function fromStringFast(string, ops) {
   const maxLength = Buffer.poolSize >>> 1;
 
@@ -498,7 +565,7 @@ function fromStringFast(string, ops) {
     createPool();
 
   const actual = ops.write(allocBuffer, string, poolOffset, length);
-  const b = new FastBuffer(allocPool, poolOffset, actual);
+  const b = new FastBuffer(allocPool, poolBase + poolOffset, actual);
 
   poolOffset += actual;
   alignPool();
@@ -560,7 +627,7 @@ function fromArrayLike(obj) {
   if (length < (Buffer.poolSize >>> 1)) {
     if (length > (poolSize - poolOffset))
       createPool();
-    const b = new FastBuffer(allocPool, poolOffset, length);
+    const b = new FastBuffer(allocPool, poolBase + poolOffset, length);
     TypedArrayPrototypeSet(b, obj, 0);
     poolOffset += length;
     alignPool();

lib/internal/buffer.js

@@ -33,6 +33,7 @@ const {
   hexWrite,
   ucs2Write,
   utf8WriteStatic,
+  arrayBufferAlignedOffset,
   createUnsafeArrayBuffer,
   setDetachKey,
 } = internalBinding('buffer');
@@ -1104,12 +1105,28 @@ function createUnsafeBuffer(size) {
   return new FastBuffer(createUnsafeArrayBuffer(size));
 }
 
+// Returns an uninitialized buffer of `size` bytes whose first byte is located at
+// a memory address that is a multiple of `alignment`. `alignment` must be a
+// power of two, and `size + alignment - 1` must not exceed the maximum buffer
+// length. Since the address of a backing store cannot be chosen, `alignment - 1`
+// extra bytes are allocated and skipped, which leaves the returned buffer with a
+// non-zero `byteOffset` into a larger ArrayBuffer.
+function createUnsafeAlignedBuffer(size, alignment) {
+  if (size === 0) {
+    return new FastBuffer();
+  }
+
+  const ab = createUnsafeArrayBuffer(size + alignment - 1);
+  return new FastBuffer(ab, arrayBufferAlignedOffset(ab, alignment), size);
+}
+
 module.exports = {
   FastBuffer,
   addBufferPrototypeMethods,
   markAsUntransferable,
   isMarkedAsUntransferable,
   createUnsafeBuffer,
+  createUnsafeAlignedBuffer,
   readUInt16BE,
   readUInt32BE,
   asciiWrite,