关于UNIX域套接字SOCK_STREAM模式下connect()报错106的技术咨询

Hey there, let's break down this error—it's error code 106, which maps to EISCONN on most Unix-like systems. This pops up when you try to call connect() on a socket that’s already successfully linked to a peer. Here are the most common causes and straightforward fixes:

Common Causes & Solutions #

1. Accidental repeated connect() calls in the client #

This is the #1 culprit. If your client code calls connect() more than once on the same socket file descriptor after a successful connection, you’ll hit this error every single time.

For example, a loop that checks connection status but fails to track whether you’re already connected:

// Bad example: unguarded repeated connect calls
int sockfd = socket(AF_UNIX, SOCK_STREAM, 0);
struct sockaddr_un server_addr;
// ... initialize server_addr details ...

// This loop will keep calling connect even after success
while (1) {
    if (connect(sockfd, (struct sockaddr*)&server_addr, sizeof(server_addr)) == -1) {
        perror("connect failed");
    }
    // Client work here
}

Fix: Add a simple flag to track connection state, so connect() only runs when the socket is unconnected:

// Better example: track connection status
int sockfd = socket(AF_UNIX, SOCK_STREAM, 0);
struct sockaddr_un server_addr;
// ... initialize server_addr details ...
int is_connected = 0;

while (1) {
    if (!is_connected) {
        if (connect(sockfd, (struct sockaddr*)&server_addr, sizeof(server_addr)) == 0) {
            is_connected = 1;
            printf("Connected to server!\n");
        } else {
            perror("Connect failed, retrying in 1s...");
            sleep(1);
        }
    }
    // Do client tasks (send/receive data)
}

2. Reusing an already-connected socket file descriptor #

If you close a socket but don’t properly reset the file descriptor (or accidentally reuse it before creating a new socket), you might end up calling connect() on a descriptor still tied to an active connection.

Fix: Always close the socket with close(sockfd) when done, mark the descriptor as invalid (set to -1), and create a new socket before attempting to reconnect:

// Proper reconnection flow
int sockfd = -1;
struct sockaddr_un server_addr;
// ... initialize server_addr details ...

// When you need to connect/reconnect:
if (sockfd != -1) {
    close(sockfd);
    sockfd = -1;
}
sockfd = socket(AF_UNIX, SOCK_STREAM, 0);
if (sockfd == -1) {
    perror("Failed to create socket");
    exit(EXIT_FAILURE);
}
if (connect(sockfd, (struct sockaddr*)&server_addr, sizeof(server_addr)) == -1) {
    perror("Connect failed");
    close(sockfd);
    sockfd = -1;
}

3. Edge case: Server-side connection mishandling #

Rarely, this can happen if the server accepts a connection but immediately closes it, and the client tries to reconnect without resetting its socket. Double-check that your server’s accept() logic isn’t closing accepted sockets unintentionally (unless that’s part of your design).

Quick Debugging Tips #

• Add debug prints right before every connect() call to log the socket descriptor value and your expected connection state.
• Use ss -x (or netstat -x on older systems) to list active Unix domain socket connections—this lets you confirm if the socket is already connected before your connect() attempt.

内容的提问来源于stack exchange，提问作者user9365205