install: Re-implement hostns to work around supermin

I'm experimenting with doing `bootc install` in a supermin VM
like we do in coreos-assembler, but hit upon the fact that supermin's
`/sbin/init` implementation incorrectly does `chroot()` instead of
a `pivot_root`:

https://github.com/libguestfs/supermin/blob/5230e2c3cd07e82bd6431e871e239f7056bf25ad/init/init.c#L288

The problem here then is doing `nsenter` just takes us into the
original rootfs, and all the real content is in `/root`.

Re-implement this by re-executing our own binary and doing the `setns()`
and `chroot()` dance internally, because we can't really do it
with external binaries.

Signed-off-by: Colin Walters <[email protected]>

Author: cgwalters

lib/Cargo.toml

@@ -29,7 +29,7 @@ liboverdrop = "0.1.0"
 once_cell = "1.9"
 openssl = "^0.10"
 # TODO drop this in favor of rustix
-nix = { version = "0.27", features = ["ioctl"] }
+nix = { version = "0.27", features = ["ioctl", "sched"] }
 regex = "1.7.1"
 rustix = { "version" = "0.38", features = ["thread", "fs", "system", "process"] }
 schemars = { version = "0.8.6", features = ["chrono"] }

lib/src/blockdev.rs

@@ -1,5 +1,5 @@
+use crate::install::run_in_host_mountns;
 use crate::task::Task;
-use crate::utils::run_in_host_mountns;
 use anyhow::{anyhow, Context, Result};
 use camino::Utf8Path;
 use fn_error_context::context;

lib/src/cli.rs

@@ -155,6 +155,11 @@ pub(crate) enum Opt {
     /// Install to the target block device
     #[cfg(feature = "install")]
     Install(crate::install::InstallOpts),
+    /// Execute the given command in the host mount namespace
+    #[cfg(feature = "install")]
+    #[clap(hide = true)]
+    #[command(external_subcommand)]
+    ExecInHostMountNamespace(Vec<OsString>),
     /// Install to the target filesystem.
     #[cfg(feature = "install")]
     InstallToFilesystem(crate::install::InstallToFilesystemOpts),
@@ -503,6 +508,10 @@ async fn run_from_opt(opt: Opt) -> Result<()> {
         Opt::Install(opts) => crate::install::install(opts).await,
         #[cfg(feature = "install")]
         Opt::InstallToFilesystem(opts) => crate::install::install_to_filesystem(opts).await,
+        #[cfg(feature = "install")]
+        Opt::ExecInHostMountNamespace(args) => {
+            crate::install::exec_in_host_mountns(args.as_slice())
+        }
         Opt::Status(opts) => super::status::status(opts).await,
         #[cfg(feature = "internal-testing-api")]
         Opt::InternalTests(opts) => crate::privtests::run(opts).await,

lib/src/install.rs

@@ -11,6 +11,8 @@ mod baseline;
 use std::io::BufWriter;
 use std::io::Write;
 use std::os::fd::AsFd;
+use std::os::unix::process::CommandExt;
+use std::process::Command;
 use std::str::FromStr;
 use std::sync::Arc;
 
@@ -37,7 +39,6 @@ use serde::{Deserialize, Serialize};
 use self::baseline::InstallBlockDeviceOpts;
 use crate::containerenv::ContainerExecutionInfo;
 use crate::task::Task;
-use crate::utils::run_in_host_mountns;
 
 /// The default "stateroot" or "osname"; see https://github.com/ostreedev/ostree/issues/2794
 const STATEROOT_DEFAULT: &str = "default";
@@ -608,6 +609,30 @@ fn copy_to_oci(
     Ok(dest_imageref)
 }
 
+/// Run a command in the host mount namespace
+pub(crate) fn run_in_host_mountns(cmd: &str) -> Command {
+    let mut c = Command::new("/proc/self/exe");
+    c.args(["exec-in-host-mount-namespace", cmd]);
+    c
+}
+
+#[context("Re-exec in host mountns")]
+pub(crate) fn exec_in_host_mountns(args: &[std::ffi::OsString]) -> Result<()> {
+    let (cmd, args) = args[1..]
+        .split_first()
+        .ok_or_else(|| anyhow::anyhow!("Missing command"))?;
+    let pid1mountns = std::fs::File::open("/proc/1/ns/mnt")?;
+    nix::sched::setns(pid1mountns.as_fd(), nix::sched::CloneFlags::CLONE_NEWNS).context("setns")?;
+    rustix::process::chdir("/")?;
+    // Work around supermin doing chroot() and not pivot_root
+    // https://github.com/libguestfs/supermin/blob/5230e2c3cd07e82bd6431e871e239f7056bf25ad/init/init.c#L288
+    if !Utf8Path::new("/usr").try_exists()? && Utf8Path::new("/root/usr").try_exists()? {
+        tracing::debug!("Using supermin workaround");
+        rustix::process::chroot("/root")?;
+    }
+    Err(Command::new(cmd).args(args).exec())?
+}
+
 #[context("Querying skopeo version")]
 fn skopeo_supports_containers_storage() -> Result<bool> {
     let o = run_in_host_mountns("skopeo").arg("--version").output()?;

lib/src/podman.rs

@@ -1,7 +1,7 @@
 use anyhow::{anyhow, Result};
 use serde::Deserialize;
 
-use crate::utils::run_in_host_mountns;
+use crate::install::run_in_host_mountns;
 
 #[derive(Deserialize)]
 #[serde(rename_all = "PascalCase")]

lib/src/utils.rs

@@ -32,14 +32,6 @@ pub(crate) fn find_mount_option<'a>(
         .next()
 }
 
-/// Run a command in the host mount namespace
-#[allow(dead_code)]
-pub(crate) fn run_in_host_mountns(cmd: &str) -> Command {
-    let mut c = Command::new("nsenter");
-    c.args(["-m", "-t", "1", "--", cmd]);
-    c
-}
-
 pub(crate) fn spawn_editor(tmpf: &tempfile::NamedTempFile) -> Result<()> {
     let v = "EDITOR";
     let editor = std::env::var_os(v)