Filesystem Operations in libuv

Simple filesystem read/write is achieved using the uv_fs_* functions and the uv_fs_t struct.

The libuv filesystem operations are different from socket operations (that we’ll cover in the next chapter.) Socket operations use the non-blocking operations provided by the operating system. Filesystem operations use blocking functions internally, but invoke these functions in a thread pool and notify watchers registered with the event loop when application interaction is required.

All filesystem functions have two forms - synchronous and asynchronous.

The synchronous forms automatically get called (and block) if the callback is null. The return value of functions is a libuv error code. This is usually only useful for synchronous calls. The asynchronous form is called when a callback is passed and the return value is 0.

Reading/Writing files using libuv

A file descriptor is obtained using

int uv_fs_open(uv_loop_t* loop, uv_fs_t* req, const char* path, int flags, int mode, uv_fs_cb cb)

flags and mode are standard Unix flags. libuv takes care of converting to the appropriate Windows flags.

File descriptors are closed using

int uv_fs_close(uv_loop_t* loop, uv_fs_t* req, uv_file file, uv_fs_cb cb)

Filesystem operation callbacks have the signature:

void callback(uv_fs_t* req);

Let’s see a simple implementation of cat using event loop and libuv. We start with registering a callback for when the file is opened:

#include <assert.h>
#include <stdio.h>
#include <fcntl.h>
#include <unistd.h>
#include <uv.h>

void on_read(uv_fs_t *req);

uv_fs_t open_req;
uv_fs_t read_req;
uv_fs_t write_req;

static char buffer[1024];

static uv_buf_t iov;

void on_write(uv_fs_t *req) {
    if (req->result < 0) {
        fprintf(stderr, "Write error: %s\n", uv_strerror((int)req->result));
    else {
        uv_fs_read(uv_default_loop(), &read_req, open_req.result, &iov, 1, -1, on_read);

void on_read(uv_fs_t *req) {
    if (req->result < 0) {
        fprintf(stderr, "Read error: %s\n", uv_strerror(req->result));
    else if (req->result == 0) {
        uv_fs_t close_req;
        // synchronous
        uv_fs_close(uv_default_loop(), &close_req, open_req.result, NULL);
    else if (req->result > 0) {
        iov.len = req->result;
        uv_fs_write(uv_default_loop(), &write_req, 1, &iov, 1, -1, on_write);

void on_open(uv_fs_t *req) {
    // The request passed to the callback is the same as the one the call setup
    // function was passed.
    assert(req == &open_req);
    if (req->result >= 0) {
        iov = uv_buf_init(buffer, sizeof(buffer));
        uv_fs_read(uv_default_loop(), &read_req, req->result,
                   &iov, 1, -1, on_read);
    else {
        fprintf(stderr, "error opening file: %s\n", uv_strerror((int)req->result));

int main(int argc, char **argv) {
    uv_fs_open(uv_default_loop(), &open_req, argv[1], O_RDONLY, 0, on_open);
    uv_run(uv_default_loop(), UV_RUN_DEFAULT);

    return 0;

The result field of a uv_fs_t is the file descriptor in case of the uv_fs_open callback. If the file is successfully opened, we start reading it.

In the case of a read call, you should pass an initialized buffer which will be filled with data before the read callback is triggered. The uv_fs_* operations map almost directly to certain POSIX functions, so EOF is indicated in this case by result being 0. In the case of streams or pipes, the UV_EOF constant would have been passed as a status instead.

Here you see a common pattern when writing asynchronous programs. The uv_fs_close() call is performed synchronously. Usually tasks which are one-off, or are done as part of the startup or shutdown stage are performed synchronously, since we are interested in fast I/O when the program is going about its primary task and dealing with multiple I/O sources. For solo tasks the performance difference usually is negligible and may lead to simpler code.

Filesystem writing is similarly simple using uv_fs_write(). Your callback will be triggered after the write is complete. In our case the callback simply drives the next read. Thus read and write proceed in lockstep via callbacks.


Due to the way filesystems and disk drives are configured for performance, a write that ‘succeeds’ may not be committed to disk yet.

We set the dominos rolling in main():


The uv_fs_req_cleanup() function must always be called on filesystem requests to free internal memory allocations in libuv.

Filesystem operations

All the standard filesystem operations like unlink, rmdir, stat are supported asynchronously and have intuitive argument order. They follow the same patterns as the read/write/open calls, returning the result in the uv_fs_t.result field. The full list:

Filesystem operations

Buffers and Streams

The basic I/O handle in libuv is the stream (uv_stream_t). TCP sockets, UDP sockets, and pipes for file I/O and IPC are all treated as stream subclasses.

Streams are initialized using custom functions for each subclass, then operated upon using

int uv_read_start(uv_stream_t*, uv_alloc_cb alloc_cb, uv_read_cb read_cb);
int uv_read_stop(uv_stream_t*);
int uv_write(uv_write_t* req, uv_stream_t* handle,
             const uv_buf_t bufs[], unsigned int nbufs, uv_write_cb cb);

The stream based functions are simpler to use than the filesystem ones and libuv will automatically keep reading from a stream when uv_read_start() is called once, until uv_read_stop() is called.

The discrete unit of data is the buffer – uv_buf_t. This is simply a collection of a pointer to bytes (uv_buf_t.base) and the length (uv_buf_t.len). The uv_buf_t is lightweight and passed around by value. What does require management is the actual bytes, which have to be allocated and freed by the application.



To demonstrate streams we will need to use uv_pipe_t. This allows streaming local files. Here is a simple tee utility using libuv. Doing all operations asynchronously shows the power of evented I/O. The two writes won’t block each other, but we have to be careful to copy over the buffer data to ensure we don’t free a buffer until it has been written.

The program is to be executed as:

./uvtee <output_file>

We start off opening pipes on the files we require. libuv pipes to a file are opened as bidirectional by default.

#include <stdio.h>
#include <fcntl.h>
#include <unistd.h>
#include <string.h>
#include <stdlib.h>

#include <uv.h>

typedef struct {
    uv_write_t req;
    uv_buf_t buf;
} write_req_t;

uv_loop_t *loop;
uv_pipe_t stdin_pipe;
uv_pipe_t stdout_pipe;
uv_pipe_t file_pipe;

void alloc_buffer(uv_handle_t *handle, size_t suggested_size, uv_buf_t *buf) {
    *buf = uv_buf_init((char*) malloc(suggested_size), suggested_size);

void free_write_req(uv_write_t *req) {
    write_req_t *wr = (write_req_t*) req;

void on_stdout_write(uv_write_t *req, int status) {

void on_file_write(uv_write_t *req, int status) {

void write_data(uv_stream_t *dest, size_t size, uv_buf_t buf, uv_write_cb cb) {
    write_req_t *req = (write_req_t*) malloc(sizeof(write_req_t));
    req->buf = uv_buf_init((char*) malloc(size), size);
    memcpy(req->buf.base, buf.base, size);
    uv_write((uv_write_t*) req, (uv_stream_t*)dest, &req->buf, 1, cb);

void read_stdin(uv_stream_t *stream, ssize_t nread, const uv_buf_t *buf) {
    if (nread < 0){
        if (nread == UV_EOF){
            // end of file
            uv_close((uv_handle_t *)&stdin_pipe, NULL);
            uv_close((uv_handle_t *)&stdout_pipe, NULL);
            uv_close((uv_handle_t *)&file_pipe, NULL);
    } else if (nread > 0) {
        write_data((uv_stream_t *)&stdout_pipe, nread, *buf, on_stdout_write);
        write_data((uv_stream_t *)&file_pipe, nread, *buf, on_file_write);

    // OK to free buffer as write_data copies it.
    if (buf->base)

int main(int argc, char **argv) {
    loop = uv_default_loop();

    uv_pipe_init(loop, &stdin_pipe, 0);
    uv_pipe_open(&stdin_pipe, 0);

    uv_pipe_init(loop, &stdout_pipe, 0);
    uv_pipe_open(&stdout_pipe, 1);
    uv_fs_t file_req;
    int fd = uv_fs_open(loop, &file_req, argv[1], O_CREAT | O_RDWR, 0644, NULL);
    uv_pipe_init(loop, &file_pipe, 0);
    uv_pipe_open(&file_pipe, fd);

    uv_read_start((uv_stream_t*)&stdin_pipe, alloc_buffer, read_stdin);

    uv_run(loop, UV_RUN_DEFAULT);
    return 0;

The third argument of uv_pipe_init() should be set to 1 for IPC using named pipes. This is covered in processes. The uv_pipe_open() call associates the pipe with the file descriptor, in this case 0 (standard input).

We start monitoring stdin. The alloc_buffer callback is invoked as new buffers are required to hold incoming data. read_stdin will be called with these buffers.

The standard malloc is sufficient here, but you can use any memory allocation scheme. For example, node.js uses its own slab allocator which associates buffers with V8 objects.

The read callback nread parameter is less than 0 on any error. This error might be EOF, in which case we close all the streams, using the generic close function uv_close() which deals with the handle based on its internal type. Otherwise nread is a non-negative number and we can attempt to write that many bytes to the output streams. Finally remember that buffer allocation and deallocation is application responsibility, so we free the data.

The allocation callback may return a buffer with length zero if it fails to allocate memory. In this case, the read callback is invoked with error UV_ENOBUFS. libuv will continue to attempt to read the stream though, so you must explicitly call uv_close() if you want to stop when allocation fails.

The read callback may be called with nread = 0, indicating that at this point there is nothing to be read. Most applications will just ignore this.

write_data() makes a copy of the buffer obtained from read. This buffer does not get passed through to the write callback trigged on write completion. To get around this we wrap a write request and a buffer in write_req_t and unwrap it in the callbacks. We make a copy so we can free the two buffers from the two calls to write_data independently of each other. While acceptable for a demo program like this, you’ll probably want smarter memory management, like reference counted buffers or a pool of buffers in any major application.


If your program is meant to be used with other programs it may knowingly or unknowingly be writing to a pipe. This makes it susceptible to aborting on receiving a SIGPIPE. It is a good idea to insert:


in the initialization stages of your application.

File change events

All modern operating systems provide APIs to put watches on individual files or directories and be informed when the files are modified. libuv wraps common file change notification libraries (inotify on Linux, FSEvents on Darwin, kqueue on BSDs, ReadDirectoryChangesW on Windows, event ports on Solaris, unsupported on Cygwin). This is one of the more inconsistent parts of libuv. File change notification systems are themselves extremely varied across platforms so getting everything working everywhere is difficult. To demonstrate, I’m going to build a simple utility which runs a command whenever any of the watched files change:

./onchange <command> <file1> [file2] ...

The file change notification is started using uv_fs_event_init():

#include <stdio.h>
#include <stdlib.h>

#include <uv.h>

uv_loop_t *loop;
const char *command;

void run_command(uv_fs_event_t *handle, const char *filename, int events, int status) {
    char path[1024];
    size_t size = 1023;
    // Does not handle error if path is longer than 1023.
    uv_fs_event_getpath(handle, path, &size);
    path[size] = '\0';

    fprintf(stderr, "Change detected in %s: ", path);
    if (events & UV_RENAME)
        fprintf(stderr, "renamed");
    if (events & UV_CHANGE)
        fprintf(stderr, "changed");

    fprintf(stderr, " %s\n", filename ? filename : "");

int main(int argc, char **argv) {
    if (argc <= 2) {
        fprintf(stderr, "Usage: %s <command> <file1> [file2 ...]\n", argv[0]);
        return 1;

    loop = uv_default_loop();
    command = argv[1];

    while (argc-- > 2) {
        fprintf(stderr, "Adding watch on %s\n", argv[argc]);
        uv_fs_event_t *fs_event_req = malloc(sizeof(uv_fs_event_t));
        uv_fs_event_init(loop, fs_event_req);
        // The recursive flag watches subdirectories too.
        uv_fs_event_start(fs_event_req, run_command, argv[argc], UV_FS_EVENT_RECURSIVE);

    return uv_run(loop, UV_RUN_DEFAULT);

The third argument is the actual file or directory to monitor. The last argument, flags, can be:

UV_FS_EVENT_WATCH_ENTRY and UV_FS_EVENT_STAT don’t do anything (yet). UV_FS_EVENT_RECURSIVE will start watching subdirectories as well on supported platforms.

The callback will receive the following arguments:

  1. uv_fs_event_t *handle - The handle. The path field of the handle is the file on which the watch was set.
  2. const char *filename - If a directory is being monitored, this is the file which was changed. Only non-null on Linux and Windows. May be null even on those platforms.
  3. int flags - one of UV_RENAME or UV_CHANGE, or a bitwise OR of


  4. int status - Currently 0.

In our example we simply print the arguments and run the command using system().

Licenses and Attributions

Speak Your Mind