diff --git a/benchmark/fs/readfile-partitioned.js b/benchmark/fs/readfile-partitioned.js
new file mode 100644
index 00000000000000..be3b7fd057bbe0
--- /dev/null
+++ b/benchmark/fs/readfile-partitioned.js
@@ -0,0 +1,86 @@
+// Submit a mix of short and long jobs to the threadpool.
+// Report total job throughput.
+// If we partition the long job, overall job throughput goes up significantly.
+// However, this comes at the cost of the long job throughput.
+//
+// Short jobs: small zip jobs.
+// Long jobs: fs.readFile on a large file.
+
+'use strict';
+
+const path = require('path');
+const common = require('../common.js');
+const filename = path.resolve(__dirname,
+                              `.removeme-benchmark-garbage-${process.pid}`);
+const fs = require('fs');
+const zlib = require('zlib');
+const assert = require('assert');
+
+const bench = common.createBenchmark(main, {
+  dur: [5],
+  len: [1024, 16 * 1024 * 1024],
+  concurrent: [1, 10]
+});
+
+function main(conf) {
+  const len = +conf.len;
+  try { fs.unlinkSync(filename); } catch (e) {}
+  var data = Buffer.alloc(len, 'x');
+  fs.writeFileSync(filename, data);
+  data = null;
+
+  var zipData = Buffer.alloc(1024, 'a');
+
+  var reads = 0;
+  var zips = 0;
+  var benchEnded = false;
+  bench.start();
+  setTimeout(function() {
+    const totalOps = reads + zips;
+    benchEnded = true;
+    bench.end(totalOps);
+    try { fs.unlinkSync(filename); } catch (e) {}
+  }, +conf.dur * 1000);
+
+  function read() {
+    fs.readFile(filename, afterRead);
+  }
+
+  function afterRead(er, data) {
+    if (er) {
+      if (er.code === 'ENOENT') {
+        // Only OK if unlinked by the timer from main.
+        assert.ok(benchEnded);
+        return;
+      }
+      throw er;
+    }
+
+    if (data.length !== len)
+      throw new Error('wrong number of bytes returned');
+
+    reads++;
+    if (!benchEnded)
+      read();
+  }
+
+  function zip() {
+    zlib.deflate(zipData, afterZip);
+  }
+
+  function afterZip(er, data) {
+    if (er)
+      throw er;
+
+    zips++;
+    if (!benchEnded)
+      zip();
+  }
+
+  // Start reads
+  var cur = +conf.concurrent;
+  while (cur--) read();
+
+  // Start a competing zip
+  zip();
+}
diff --git a/benchmark/fs/readfile.js b/benchmark/fs/readfile.js
index 7da7758ed06638..282b4ac7621340 100644
--- a/benchmark/fs/readfile.js
+++ b/benchmark/fs/readfile.js
@@ -8,6 +8,7 @@ const common = require('../common.js');
 const filename = path.resolve(process.env.NODE_TMPDIR || __dirname,
                               `.removeme-benchmark-garbage-${process.pid}`);
 const fs = require('fs');
+const assert = require('assert');
 
 const bench = common.createBenchmark(main, {
   dur: [5],
@@ -22,10 +23,10 @@ function main({ len, dur, concurrent }) {
   data = null;
 
   var reads = 0;
-  var bench_ended = false;
+  var benchEnded = false;
   bench.start();
   setTimeout(function() {
-    bench_ended = true;
+    benchEnded = true;
     bench.end(reads);
     try { fs.unlinkSync(filename); } catch (e) {}
     process.exit(0);
@@ -36,14 +37,20 @@ function main({ len, dur, concurrent }) {
   }
 
   function afterRead(er, data) {
-    if (er)
+    if (er) {
+      if (er.code === 'ENOENT') {
+        // Only OK if unlinked by the timer from main.
+        assert.ok(benchEnded);
+        return;
+      }
       throw er;
+    }
 
     if (data.length !== len)
       throw new Error('wrong number of bytes returned');
 
     reads++;
-    if (!bench_ended)
+    if (!benchEnded)
       read();
   }
 
diff --git a/doc/api/fs.md b/doc/api/fs.md
index d5e7f75ebe8bc1..00c6d946a0b355 100644
--- a/doc/api/fs.md
+++ b/doc/api/fs.md
@@ -2250,10 +2250,9 @@ Any specified file descriptor has to support reading.
 *Note*: If a file descriptor is specified as the `path`, it will not be closed
 automatically.
 
-*Note*: `fs.readFile()` reads the entire file in a single threadpool request.
-To minimize threadpool task length variation, prefer the partitioned APIs
-`fs.read()` and `fs.createReadStream()` when reading files as part of
-fulfilling a client request.
+*Note*: `fs.readFile()` buffers the entire file.
+To minimize memory costs, when possible prefer streaming via
+`fs.createReadStream()`.
 
 ## fs.readFileSync(path[, options])
 <!-- YAML
diff --git a/lib/fs.js b/lib/fs.js
index d4f5ee0909bcef..0b9cf0cc9b8190 100644
--- a/lib/fs.js
+++ b/lib/fs.js
@@ -508,7 +508,7 @@ ReadFileContext.prototype.read = function() {
   } else {
     buffer = this.buffer;
     offset = this.pos;
-    length = this.size - this.pos;
+    length = Math.min(kReadFileBufferLength, this.size - this.pos);
   }
 
   var req = new FSReqWrap();
diff --git a/test/parallel/test-fs-readfile.js b/test/parallel/test-fs-readfile.js
new file mode 100644
index 00000000000000..689f98ff849ded
--- /dev/null
+++ b/test/parallel/test-fs-readfile.js
@@ -0,0 +1,58 @@
+'use strict';
+const common = require('../common');
+
+// This test ensures that fs.readFile correctly returns the
+// contents of varying-sized files.
+
+const assert = require('assert');
+const fs = require('fs');
+const path = require('path');
+
+const prefix = `.removeme-fs-readfile-${process.pid}`;
+
+common.refreshTmpDir();
+
+const fileInfo = [
+  { name: path.join(common.tmpDir, `${prefix}-1K.txt`),
+    len: 1024,
+  },
+  { name: path.join(common.tmpDir, `${prefix}-64K.txt`),
+    len: 64 * 1024,
+  },
+  { name: path.join(common.tmpDir, `${prefix}-64KLessOne.txt`),
+    len: (64 * 1024) - 1,
+  },
+  { name: path.join(common.tmpDir, `${prefix}-1M.txt`),
+    len: 1 * 1024 * 1024,
+  },
+  { name: path.join(common.tmpDir, `${prefix}-1MPlusOne.txt`),
+    len: (1 * 1024 * 1024) + 1,
+  },
+];
+
+// Populate each fileInfo (and file) with unique fill.
+const sectorSize = 512;
+for (const e of fileInfo) {
+  e.contents = Buffer.allocUnsafe(e.len);
+
+  // This accounts for anything unusual in Node's implementation of readFile.
+  // Using e.g. 'aa...aa' would miss bugs like Node re-reading
+  // the same section twice instead of two separate sections.
+  for (let offset = 0; offset < e.len; offset += sectorSize) {
+    const fillByte = 256 * Math.random();
+    const nBytesToFill = Math.min(sectorSize, e.len - offset);
+    e.contents.fill(fillByte, offset, offset + nBytesToFill);
+  }
+
+  fs.writeFileSync(e.name, e.contents);
+}
+// All files are now populated.
+
+// Test readFile on each size.
+for (const e of fileInfo) {
+  fs.readFile(e.name, common.mustCall((err, buf) => {
+    console.log(`Validating readFile on file ${e.name} of length ${e.len}`);
+    assert.ifError(err, 'An error occurred');
+    assert.deepStrictEqual(buf, e.contents, 'Incorrect file contents');
+  }));
+}