lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Thu, 16 May 2024 22:03:33 +0300
From: Ariel Miculas <amiculas@...co.com>
To: rust-for-linux@...r.kernel.org
Cc: linux-kernel@...r.kernel.org, linux-fsdevel@...r.kernel.org,
        tycho@...ho.pizza, brauner@...nel.org, viro@...iv.linux.org.uk,
        ojeda@...nel.org, alex.gaynor@...il.com, wedsonaf@...il.com,
        shallyn@...co.com, Ariel Miculas <amiculas@...co.com>
Subject: [RFC PATCH v3 10/22] rust: kernel: add an abstraction over vfsmount to allow cloning a new private mount

Provide an abstraction over vfsmount so that we can create a new private
mount from Rust code.

PuzzleFS needs a private mount of the PuzzleFS image, so it can access
the chunks from the data store independent of the future changes to the
mount namespace.

Signed-off-by: Ariel Miculas <amiculas@...co.com>
---
 rust/bindings/bindings_helper.h |  1 +
 rust/kernel/lib.rs              |  1 +
 rust/kernel/mount.rs            | 73 +++++++++++++++++++++++++++++++++
 3 files changed, 75 insertions(+)
 create mode 100644 rust/kernel/mount.rs

diff --git a/rust/bindings/bindings_helper.h b/rust/bindings/bindings_helper.h
index 629fce394dbe..2d87bb9f87c9 100644
--- a/rust/bindings/bindings_helper.h
+++ b/rust/bindings/bindings_helper.h
@@ -16,6 +16,7 @@
 #include <linux/iomap.h>
 #include <linux/jiffies.h>
 #include <linux/mdio.h>
+#include <linux/namei.h>
 #include <linux/pagemap.h>
 #include <linux/phy.h>
 #include <linux/refcount.h>
diff --git a/rust/kernel/lib.rs b/rust/kernel/lib.rs
index 445599d4bff6..7f0d89b902ce 100644
--- a/rust/kernel/lib.rs
+++ b/rust/kernel/lib.rs
@@ -37,6 +37,7 @@
 #[cfg(CONFIG_KUNIT)]
 pub mod kunit;
 pub mod mem_cache;
+pub mod mount;
 #[cfg(CONFIG_NET)]
 pub mod net;
 pub mod prelude;
diff --git a/rust/kernel/mount.rs b/rust/kernel/mount.rs
new file mode 100644
index 000000000000..50cfe7f7437a
--- /dev/null
+++ b/rust/kernel/mount.rs
@@ -0,0 +1,73 @@
+//! Mount interface
+//!
+//! C headers: [`include/linux/mount.h`](../../../../include/linux/mount.h)
+
+use kernel::bindings;
+use kernel::error::from_err_ptr;
+use kernel::pr_err;
+use kernel::prelude::*;
+use kernel::str::CStr;
+use kernel::types::Opaque;
+
+/// Wraps the kernel's `struct path`.
+#[repr(transparent)]
+pub struct Path(pub(crate) Opaque<bindings::path>);
+
+/// Wraps the kernel's `struct vfsmount`.
+#[repr(transparent)]
+#[derive(Debug)]
+pub struct Vfsmount {
+    vfsmount: *mut bindings::vfsmount,
+}
+
+// SAFETY: No one besides us has the raw pointer, so we can safely transfer Vfsmount to another thread
+unsafe impl Send for Vfsmount {}
+// SAFETY: It's OK to access `Vfsmount` through references from other threads because we're not
+// accessing any properties from the underlying raw pointer
+unsafe impl Sync for Vfsmount {}
+
+impl Vfsmount {
+    /// Create a new private mount clone based on a path name
+    pub fn new_private_mount(path_name: &CStr) -> Result<Self> {
+        let path: Path = Path(Opaque::uninit());
+        // SAFETY: path_name is a &CStr, so it's a valid string pointer; path is an uninitialized
+        // struct stored on the stack and it's ok because kern_path expects an out parameter
+        let err = unsafe {
+            bindings::kern_path(
+                path_name.as_ptr().cast::<i8>(),
+                bindings::LOOKUP_FOLLOW | bindings::LOOKUP_DIRECTORY,
+                path.0.get(),
+            )
+        };
+        if err != 0 {
+            pr_err!("failed to resolve '{}': {}\n", path_name, err);
+            return Err(EINVAL);
+        }
+
+        // SAFETY: path is a struct stored on the stack and it is  initialized because the call to
+        // kern_path succeeded
+        let vfsmount = unsafe { from_err_ptr(bindings::clone_private_mount(path.0.get()))? };
+
+        // Don't inherit atime flags
+        // CAST: these flags fit into i32
+        let skip_flags =
+            !(bindings::MNT_NOATIME | bindings::MNT_NODIRATIME | bindings::MNT_RELATIME) as i32;
+        // SAFETY: we called from_err_ptr so it's safe to dereference this pointer
+        unsafe {
+            (*vfsmount).mnt_flags &= skip_flags;
+        }
+        Ok(Self { vfsmount })
+    }
+
+    /// Returns a raw pointer to vfsmount
+    pub fn get(&self) -> *mut bindings::vfsmount {
+        self.vfsmount
+    }
+}
+
+impl Drop for Vfsmount {
+    fn drop(&mut self) {
+        // SAFETY new_private_mount makes sure to return a valid pointer
+        unsafe { bindings::kern_unmount(self.vfsmount) };
+    }
+}
-- 
2.34.1


Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ