Message-ID: <5d7466ad-7790-4a08-86ba-82b0059562c1@schaufler-ca.com>
Date: Tue, 31 Dec 2024 08:43:09 -0800
From: Casey Schaufler <casey@...aufler-ca.com>
To: Shervin Oloumi <enlightened@...omium.org>, mic@...ikod.net,
 viro@...iv.linux.org.uk
Cc: brauner@...nel.org, jack@...e.cz, paul@...l-moore.com, jmorris@...ei.org,
 serge@...lyn.com, linux-kernel@...r.kernel.org,
 linux-security-module@...r.kernel.org, gnoack@...gle.com, shuah@...nel.org,
 jorgelo@...omium.org, allenwebb@...omium.org,
 Casey Schaufler <casey@...aufler-ca.com>
Subject: Re: [PATCH 1/2] fs: add loopback/bind mount specific security hook

On 12/30/2024 5:46 PM, Shervin Oloumi wrote:
> The main mount security hook (security_sb_mount) is called early in the
> process before the mount type is determined and the arguments are
> validated and converted to the appropriate format. Specifically, the
> source path is surfaced as a string, which is not appropriate for
> checking bind mount requests. For bind mounts the source should be
> validated and passed as a path struct (same as destination), after the
> mount type is determined. This allows the hook users to evaluate the
> mount attributes without the need to perform any validations or
> conversions out of band, which can introduce a TOCTOU race condition.
>
> The newly introduced hook is invoked only if the security_sb_mount hook
> passes, and only if the MS_BIND flag is detected. At this point the
> source of the mount has been successfully converted to a path struct
> using the kernel's kern_path API. This allows LSMs to target bind mount
> requests at the right stage, and evaluate the attributes in the right
> format, based on the type of mount.

I am not very happy about an LSM hook that is this specific.
Why restrict it to bind mounts? It seems that there might be other
valuable restrictions on mounts that are based on the path.

>
> This does not affect the functionality of the existing mount security
> hooks, including security_sb_mount. The new hook, can be utilized as a
> supplement to the main hook for further analyzing bind mount requests.
> This means that there is still the option of only using the main hook
> function, if all one wants to do is indiscriminately reject all bind
> mount requests, regardless of the source and destination arguments.
> However, if one needs to evaluate the source and destination of a bind
> mount request before making a decision, this hook function should be
> preferred. Of course, if a bind mount request does not make it past the
> security_sb_mount check, the bind mount hook function is never invoked.
>
> Signed-off-by: Shervin Oloumi <enlightened@...omium.org>
> ---
>  fs/namespace.c                |  4 ++++
>  include/linux/lsm_hook_defs.h |  1 +
>  include/linux/security.h      |  1 +
>  security/security.c           | 16 ++++++++++++++++
>  4 files changed, 22 insertions(+)
>
> diff --git a/fs/namespace.c b/fs/namespace.c
> index 23e81c2a1e3f..c902608c9759 100644
> --- a/fs/namespace.c
> +++ b/fs/namespace.c
> @@ -2765,6 +2765,10 @@ static int do_loopback(struct path *path, const char *old_name,
>  	if (err)
>  		return err;
>  
> +	err = security_sb_bindmount(&old_path, path);

I can easily envision uses for this other than bind mounts.
Perhaps security_sb_mount_path()?

> +	if (err)
> +		goto out;
> +
>  	err = -EINVAL;
>  	if (mnt_ns_loop(old_path.dentry))
>  		goto out;
> diff --git a/include/linux/lsm_hook_defs.h b/include/linux/lsm_hook_defs.h
> index eb2937599cb0..3d1940239556 100644
> --- a/include/linux/lsm_hook_defs.h
> +++ b/include/linux/lsm_hook_defs.h
> @@ -71,6 +71,7 @@ LSM_HOOK(int, 0, sb_show_options, struct seq_file *m, struct super_block *sb)
>  LSM_HOOK(int, 0, sb_statfs, struct dentry *dentry)
>  LSM_HOOK(int, 0, sb_mount, const char *dev_name, const struct path *path,
>  	 const char *type, unsigned long flags, void *data)
> +LSM_HOOK(int, 0, sb_bindmount, const struct path *old_path, const struct path *path)
>  LSM_HOOK(int, 0, sb_umount, struct vfsmount *mnt, int flags)
>  LSM_HOOK(int, 0, sb_pivotroot, const struct path *old_path,
>  	 const struct path *new_path)
> diff --git a/include/linux/security.h b/include/linux/security.h
> index cbdba435b798..512ac656500e 100644
> --- a/include/linux/security.h
> +++ b/include/linux/security.h
> @@ -365,6 +365,7 @@ int security_sb_show_options(struct seq_file *m, struct super_block *sb);
>  int security_sb_statfs(struct dentry *dentry);
>  int security_sb_mount(const char *dev_name, const struct path *path,
>  		      const char *type, unsigned long flags, void *data);
> +int security_sb_bindmount(const struct path *old_path, const struct path *path);
>  int security_sb_umount(struct vfsmount *mnt, int flags);
>  int security_sb_pivotroot(const struct path *old_path, const struct path *new_path);
>  int security_sb_set_mnt_opts(struct super_block *sb,
> diff --git a/security/security.c b/security/security.c
> index 09664e09fec9..bd7cb3df16f4 100644
> --- a/security/security.c
> +++ b/security/security.c
> @@ -1564,6 +1564,22 @@ int security_sb_mount(const char *dev_name, const struct path *path,
>  	return call_int_hook(sb_mount, dev_name, path, type, flags, data);
>  }
>  
> +/**
> + * security_sb_bindmount() - Loopback/bind mount specific permission check
> + * @old_path: source of loopback/bind mount
> + * @path: mount point
> + *
> + * This check is performed in addition to security_sb_mount and only if the
> + * mount type is determined to be loopback/bind mount (flags & MS_BIND).  It
> + * surfaces the mount source as a path struct.
> + *
> + * Return: Returns 0 if permission is granted.
> + */
> +int security_sb_bindmount(const struct path *old_path, const struct path *path)
> +{
> +	return call_int_hook(sb_bindmount, old_path, path);
> +}
> +
>  /**
>   * security_sb_umount() - Check permission for unmounting a filesystem
>   * @mnt: mounted filesystem
>
> base-commit: fc033cf25e612e840e545f8d5ad2edd6ba613ed5

Powered by blists - more mailing lists