[PATCH] eCryptfs: Fix refs to pid and user_ns

Serge E. Hallyn serue at us.ibm.com
Thu Apr 17 10:41:48 PDT 2008


Quoting Michael Halcrow (mhalcrow at us.ibm.com):
> On Thu, Apr 17, 2008 at 10:34:06AM -0500, Serge E. Hallyn wrote:
> > Quoting Michael Halcrow (mhalcrow at us.ibm.com):
> > > @@ -206,6 +210,7 @@ ecryptfs_spawn_daemon(struct ecryptfs_daemon **daemon, uid_t euid, pid_t pid)
> > >  		goto out;
> > >  	}
> > >  	(*daemon)->euid = euid;
> > > +	(*daemon)->user_ns = user_ns;
> > >  	(*daemon)->pid = pid;
> > 
> > You'll want to do a get_pid() here, no?
> > 
> > And get_user_ns().
> > 
> 
> > It's not because you particulary need them to stick around, but just
> > to ensure no wraparound causes another daemon with the same struct
> > pid or user_namespace to be spawned.  Pretty gosh-darned unlikely,
> > but still...
> > ...
> > > @@ -372,12 +383,24 @@ int ecryptfs_process_response(struct ecryptfs_message *msg, uid_t euid,
> > >  	msg_ctx = &ecryptfs_msg_ctx_arr[msg->index];
> > >  	mutex_lock(&msg_ctx->mux);
> > >  	mutex_lock(&ecryptfs_daemon_hash_mux);
> > > -	rc = ecryptfs_find_daemon_by_euid(&daemon, msg_ctx->task->euid);
> > > +	rcu_read_lock();
> > > +	nsproxy = task_nsproxy(msg_ctx->task);
> > > +	if (nsproxy == NULL) {
> > > +		rc = -EBADMSG;
> > > +		printk(KERN_ERR "%s: Receiving process is a zombie. Dropping "
> > > +		       "message.\n", __func__);
> > > +		rcu_read_unlock();
> > > +		mutex_unlock(&ecryptfs_daemon_hash_mux);
> > > +		goto wake_up;
> > > +	}
> > > +	rc = ecryptfs_find_daemon_by_euid(&daemon, msg_ctx->task->euid,
> > > +					  nsproxy->user_ns);
> > > +	rcu_read_unlock();
> > >  	mutex_unlock(&ecryptfs_daemon_hash_mux);
> > >  	if (rc) {
> > >  		rc = -EBADMSG;
> > >  		printk(KERN_WARNING "%s: User [%d] received a "
> > > -		       "message response from process [%d] but does "
> > > +		       "message response from process [0x%p] but does "
> > >  		       "not have a registered daemon\n", __func__,
> > >  		       msg_ctx->task->euid, pid);
> > >  		goto wake_up;
> > > @@ -389,10 +412,17 @@ int ecryptfs_process_response(struct ecryptfs_message *msg, uid_t euid,
> > >  		       euid, msg_ctx->task->euid);
> > >  		goto unlock;
> > >  	}
> > > +	if (nsproxy->user_ns != user_ns) {
> > 
> > Since you didn't grab a ref to the nsproxy, it's possible that it
> > will have been freed before this, right?  So you probably just want
> > to grab a copy of nsproxy->user_ns while under the rcu_read_lock,
> > where you can be sure it's still around.
> 
> Have eCryptfs properly reference the pid and user_ns objects. Copy
> user_ns out of nsproxy in case nsproxy goes away after we drop the
> lock.
> 
> Signed-off-by: Michael Halcrow <mhalcrow at us.ibm.com>

Looks good.  Thanks, Mike.

Acked-by: Serge Hallyn <serue at us.ibm.com>

for both.

-serge

> ---
>  fs/ecryptfs/messaging.c |   16 ++++++++++++----
>  1 files changed, 12 insertions(+), 4 deletions(-)
> 
> diff --git a/fs/ecryptfs/messaging.c b/fs/ecryptfs/messaging.c
> index f0d74b8..61506e5 100644
> --- a/fs/ecryptfs/messaging.c
> +++ b/fs/ecryptfs/messaging.c
> @@ -20,6 +20,8 @@
>   * 02111-1307, USA.
>   */
>  #include <linux/sched.h>
> +#include <linux/user_namespace.h>
> +#include <linux/nsproxy.h>
>  #include "ecryptfs_kernel.h"
> 
>  static LIST_HEAD(ecryptfs_msg_ctx_free_list);
> @@ -208,8 +210,8 @@ ecryptfs_spawn_daemon(struct ecryptfs_daemon **daemon, uid_t euid,
>  		goto out;
>  	}
>  	(*daemon)->euid = euid;
> -	(*daemon)->user_ns = user_ns;
> -	(*daemon)->pid = pid;
> +	(*daemon)->user_ns = get_user_ns(user_ns);
> +	(*daemon)->pid = get_pid(pid);
>  	(*daemon)->task = current;
>  	mutex_init(&(*daemon)->mux);
>  	INIT_LIST_HEAD(&(*daemon)->msg_ctx_out_queue);
> @@ -298,6 +300,10 @@ int ecryptfs_exorcise_daemon(struct ecryptfs_daemon *daemon)
>  	hlist_del(&daemon->euid_chain);
>  	if (daemon->task)
>  		wake_up_process(daemon->task);
> +	if (daemon->pid)
> +		put_pid(daemon->pid);
> +	if (daemon->user_ns)
> +		put_user_ns(daemon->user_ns);
>  	mutex_unlock(&daemon->mux);
>  	memset(daemon, 0, sizeof(*daemon));
>  	kfree(daemon);
> @@ -368,6 +374,7 @@ int ecryptfs_process_response(struct ecryptfs_message *msg, uid_t euid,
>  	struct ecryptfs_msg_ctx *msg_ctx;
>  	size_t msg_size;
>  	struct nsproxy *nsproxy;
> +	struct user_namespace *current_user_ns;
>  	int rc;
> 
>  	if (msg->index >= ecryptfs_message_buf_len) {
> @@ -391,8 +398,9 @@ int ecryptfs_process_response(struct ecryptfs_message *msg, uid_t euid,
>  		mutex_unlock(&ecryptfs_daemon_hash_mux);
>  		goto wake_up;
>  	}
> +	current_user_ns = nsproxy->user_ns;
>  	rc = ecryptfs_find_daemon_by_euid(&daemon, msg_ctx->task->euid,
> -					  nsproxy->user_ns);
> +					  current_user_ns);
>  	rcu_read_unlock();
>  	mutex_unlock(&ecryptfs_daemon_hash_mux);
>  	if (rc) {
> @@ -410,7 +418,7 @@ int ecryptfs_process_response(struct ecryptfs_message *msg, uid_t euid,
>  		       euid, msg_ctx->task->euid);
>  		goto unlock;
>  	}
> -	if (nsproxy->user_ns != user_ns) {
> +	if (current_user_ns != user_ns) {
>  		rc = -EBADMSG;
>  		printk(KERN_WARNING "%s: Received message from user_ns "
>  		       "[0x%p]; expected message from user_ns [0x%p]\n",
> -- 
> 1.5.1.6


More information about the Containers mailing list