[PATCH 0/2] fs: fix capable() call in simple_xattr_list()

[PATCH 0/2] fs: fix capable() call in simple_xattr_list()

[Ondrej Mosnacek]

The goal of these patches is to avoid calling capable() unconditionally
in simple_xattr_list(), which causes issues under SELinux (see
explanation in the second patch).

The first patch tries to make this change safer by converting
simple_xattrs to use the RCU mechanism, so that capable() is not called
while the xattrs->lock is held. I didn't find evidence that this is an
issue in the current code, but it can't hurt to make that change
either way (and it was quite straightforward).

Ondrej Mosnacek (2):
  fs: convert simple_xattrs to RCU list
  fs: don't call capable() prematurely in simple_xattr_list()

 fs/xattr.c            | 39 +++++++++++++++++++++++----------------
 include/linux/xattr.h |  1 +
 2 files changed, 24 insertions(+), 16 deletions(-)

-- 
2.37.2

[PATCH 1/2] fs: convert simple_xattrs to RCU list

[Ondrej Mosnacek]

Use the RCU list mechanism instead of a simple lock to access/modify
simple_xattrs. The performance benefit is probably negligible, but it
will help avoid lock nesting concerns for an upcoming patch.

Signed-off-by: Ondrej Mosnacek <omosnace@redhat.com>
---
 fs/xattr.c            | 36 ++++++++++++++++++++++--------------
 include/linux/xattr.h |  1 +
 2 files changed, 23 insertions(+), 14 deletions(-)

diff --git a/fs/xattr.c b/fs/xattr.c
index a1f4998bc6be..fad2344f1168 100644
--- a/fs/xattr.c
+++ b/fs/xattr.c
@@ -22,6 +22,8 @@
 #include <linux/audit.h>
 #include <linux/vmalloc.h>
 #include <linux/posix_acl_xattr.h>
+#include <linux/rculist.h>
+#include <linux/rcupdate.h>
 
 #include <linux/uaccess.h>
 
@@ -1030,8 +1032,8 @@ int simple_xattr_get(struct simple_xattrs *xattrs, const char *name,
 	struct simple_xattr *xattr;
 	int ret = -ENODATA;
 
-	spin_lock(&xattrs->lock);
-	list_for_each_entry(xattr, &xattrs->head, list) {
+	rcu_read_lock();
+	list_for_each_entry_rcu(xattr, &xattrs->head, list) {
 		if (strcmp(name, xattr->name))
 			continue;
 
@@ -1044,10 +1046,18 @@ int simple_xattr_get(struct simple_xattrs *xattrs, const char *name,
 		}
 		break;
 	}
-	spin_unlock(&xattrs->lock);
+	rcu_read_unlock();
 	return ret;
 }
 
+static void simple_xattr_free_rcu(struct rcu_head *rcu)
+{
+	struct simple_xattr *xattr = container_of(rcu, struct simple_xattr, rcu);
+
+	kfree(xattr->name);
+	kvfree(xattr);
+}
+
 /**
  * simple_xattr_set - xattr SET operation for in-memory/pseudo filesystems
  * @xattrs: target simple_xattr list
@@ -1094,11 +1104,11 @@ int simple_xattr_set(struct simple_xattrs *xattrs, const char *name,
 				xattr = new_xattr;
 				err = -EEXIST;
 			} else if (new_xattr) {
-				list_replace(&xattr->list, &new_xattr->list);
+				list_replace_rcu(&xattr->list, &new_xattr->list);
 				if (removed_size)
 					*removed_size = xattr->size;
 			} else {
-				list_del(&xattr->list);
+				list_del_rcu(&xattr->list);
 				if (removed_size)
 					*removed_size = xattr->size;
 			}
@@ -1109,15 +1119,13 @@ int simple_xattr_set(struct simple_xattrs *xattrs, const char *name,
 		xattr = new_xattr;
 		err = -ENODATA;
 	} else {
-		list_add(&new_xattr->list, &xattrs->head);
+		list_add_rcu(&new_xattr->list, &xattrs->head);
 		xattr = NULL;
 	}
 out:
 	spin_unlock(&xattrs->lock);
-	if (xattr) {
-		kfree(xattr->name);
-		kvfree(xattr);
-	}
+	if (xattr)
+		call_rcu(&xattr->rcu, simple_xattr_free_rcu);
 	return err;
 
 }
@@ -1169,8 +1177,8 @@ ssize_t simple_xattr_list(struct inode *inode, struct simple_xattrs *xattrs,
 	}
 #endif
 
-	spin_lock(&xattrs->lock);
-	list_for_each_entry(xattr, &xattrs->head, list) {
+	rcu_read_lock();
+	list_for_each_entry_rcu(xattr, &xattrs->head, list) {
 		/* skip "trusted." attributes for unprivileged callers */
 		if (!trusted && xattr_is_trusted(xattr->name))
 			continue;
@@ -1179,7 +1187,7 @@ ssize_t simple_xattr_list(struct inode *inode, struct simple_xattrs *xattrs,
 		if (err)
 			break;
 	}
-	spin_unlock(&xattrs->lock);
+	rcu_read_unlock();
 
 	return err ? err : size - remaining_size;
 }
@@ -1191,6 +1199,6 @@ void simple_xattr_list_add(struct simple_xattrs *xattrs,
 			   struct simple_xattr *new_xattr)
 {
 	spin_lock(&xattrs->lock);
-	list_add(&new_xattr->list, &xattrs->head);
+	list_add_rcu(&new_xattr->list, &xattrs->head);
 	spin_unlock(&xattrs->lock);
 }
diff --git a/include/linux/xattr.h b/include/linux/xattr.h
index 979a9d3e5bfb..3236c469aaac 100644
--- a/include/linux/xattr.h
+++ b/include/linux/xattr.h
@@ -86,6 +86,7 @@ struct simple_xattrs {
 
 struct simple_xattr {
 	struct list_head list;
+	struct rcu_head rcu;
 	char *name;
 	size_t size;
 	char value[];
-- 
2.37.2

[PATCH 2/2] fs: don't call capable() prematurely in simple_xattr_list()

[Ondrej Mosnacek]

Calling capable() pre-emptively causes a problem for SELinux, which will
normally log a denial whenever capable() is called and the task's
SELinux context doesn't have the corresponding capability permission
allowed.

With the current implementation of simple_xattr_list(), any time a
process without CAP_SYS_ADMIN calls listxattr(2) or similar on a
filesystem that uses this function, a denial is logged even if there are
no trusted.* xattrs on the inode in question. In such situation, policy
writers are forced to chose one of the following options:

1. Grant CAP_SYS_ADMIN to the given SELinux domain even though it
   doesn't really need it. (Not good for security.)
2. Add a rule to the policy that will silence CAP_SYS_ADMIN denials for
   the given domain without actually granting it. (Not good, because now
   denials that make actual difference may be hidden, making
   troubleshooting harder.)
3. Do nothing and let the denials appear. (Not good, because the audit
   spam could obscure actual important denials.)

To avoid this misery, only call capable() when an actual trusted.* xattr
is encountered. This is somewhat less optimal, since capable() will now
be called once per each trusted.* xattr, but that's pretty unlikely to
matter in practice.

Even after this fix any process listing xattrs on an inode that has one
or more trusted.* ones may trigger an "irrelevant" denial if it doesn't
actually care about the trusted.* xattrs, but such cases should be rare
and thus silencing the denial in such cases would not be as big of a
deal.

Fixes: b09e0fa4b4ea ("tmpfs: implement generic xattr support")
Reported-by: Martin Pitt <mpitt@redhat.com>
Signed-off-by: Ondrej Mosnacek <omosnace@redhat.com>
---
 fs/xattr.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/fs/xattr.c b/fs/xattr.c
index fad2344f1168..84a459ac779a 100644
--- a/fs/xattr.c
+++ b/fs/xattr.c
@@ -1155,7 +1155,6 @@ static int xattr_list_one(char **buffer, ssize_t *remaining_size,
 ssize_t simple_xattr_list(struct inode *inode, struct simple_xattrs *xattrs,
 			  char *buffer, size_t size)
 {
-	bool trusted = capable(CAP_SYS_ADMIN);
 	struct simple_xattr *xattr;
 	ssize_t remaining_size = size;
 	int err = 0;
@@ -1180,7 +1179,7 @@ ssize_t simple_xattr_list(struct inode *inode, struct simple_xattrs *xattrs,
 	rcu_read_lock();
 	list_for_each_entry_rcu(xattr, &xattrs->head, list) {
 		/* skip "trusted." attributes for unprivileged callers */
-		if (!trusted && xattr_is_trusted(xattr->name))
+		if (xattr_is_trusted(xattr->name) && !capable(CAP_SYS_ADMIN))
 			continue;
 
 		err = xattr_list_one(&buffer, &remaining_size, xattr->name);
-- 
2.37.2

Re: [PATCH 0/2] fs: fix capable() call in simple_xattr_list()

[Christian Brauner]

On Thu, Sep 01, 2022 at 05:26:30PM +0200, Ondrej Mosnacek wrote:
> The goal of these patches is to avoid calling capable() unconditionally
> in simple_xattr_list(), which causes issues under SELinux (see
> explanation in the second patch).
> 
> The first patch tries to make this change safer by converting
> simple_xattrs to use the RCU mechanism, so that capable() is not called
> while the xattrs->lock is held. I didn't find evidence that this is an
> issue in the current code, but it can't hurt to make that change
> either way (and it was quite straightforward).

Hey Ondrey,

There's another patchset I'd like to see first which switches from a
linked list to an rbtree to get rid of performance issues in this code
that can be used to dos tmpfs in containers:

https://lore.kernel.org/lkml/d73bd478-e373-f759-2acb-2777f6bba06f@openvz.org

I don't think Vasily has time to continue with this so I'll just pick it
up hopefully this or the week after LPC.

Christian

Re: [PATCH 0/2] fs: fix capable() call in simple_xattr_list()

[Ondrej Mosnacek]

On Mon, Sep 5, 2022 at 11:08 AM Christian Brauner <brauner@kernel.org> wrote:
> On Thu, Sep 01, 2022 at 05:26:30PM +0200, Ondrej Mosnacek wrote:
> > The goal of these patches is to avoid calling capable() unconditionally
> > in simple_xattr_list(), which causes issues under SELinux (see
> > explanation in the second patch).
> >
> > The first patch tries to make this change safer by converting
> > simple_xattrs to use the RCU mechanism, so that capable() is not called
> > while the xattrs->lock is held. I didn't find evidence that this is an
> > issue in the current code, but it can't hurt to make that change
> > either way (and it was quite straightforward).
>
> Hey Ondrey,
>
> There's another patchset I'd like to see first which switches from a
> linked list to an rbtree to get rid of performance issues in this code
> that can be used to dos tmpfs in containers:
>
> https://lore.kernel.org/lkml/d73bd478-e373-f759-2acb-2777f6bba06f@openvz.org
>
> I don't think Vasily has time to continue with this so I'll just pick it
> up hopefully this or the week after LPC.

Hm... does rbtree support lockless traversal? Because if not, that
would make it impossible to fix the issue without calling capable()
inside the critical section (or doing something complicated), AFAICT.
Would rhashtable be a workable alternative to rbtree for this use
case? Skimming <linux/rhashtable.h> it seems to support both lockless
lookup and traversal using RCU. And according to its manpage,
*listxattr(2) doesn't guarantee that the returned names are sorted.

--
Ondrej Mosnacek
Senior Software Engineer, Linux Security - SELinux kernel
Red Hat, Inc.

Re: [PATCH 0/2] fs: fix capable() call in simple_xattr_list()

[Christian Brauner]

On Mon, Sep 05, 2022 at 12:15:01PM +0200, Ondrej Mosnacek wrote:
> On Mon, Sep 5, 2022 at 11:08 AM Christian Brauner <brauner@kernel.org> wrote:
> > On Thu, Sep 01, 2022 at 05:26:30PM +0200, Ondrej Mosnacek wrote:
> > > The goal of these patches is to avoid calling capable() unconditionally
> > > in simple_xattr_list(), which causes issues under SELinux (see
> > > explanation in the second patch).
> > >
> > > The first patch tries to make this change safer by converting
> > > simple_xattrs to use the RCU mechanism, so that capable() is not called
> > > while the xattrs->lock is held. I didn't find evidence that this is an
> > > issue in the current code, but it can't hurt to make that change
> > > either way (and it was quite straightforward).
> >
> > Hey Ondrey,
> >
> > There's another patchset I'd like to see first which switches from a
> > linked list to an rbtree to get rid of performance issues in this code
> > that can be used to dos tmpfs in containers:
> >
> > https://lore.kernel.org/lkml/d73bd478-e373-f759-2acb-2777f6bba06f@openvz.org
> >
> > I don't think Vasily has time to continue with this so I'll just pick it
> > up hopefully this or the week after LPC.
> 
> Hm... does rbtree support lockless traversal? Because if not, that

The rfc that Vasily sent didn't allow for that at least.

> would make it impossible to fix the issue without calling capable()
> inside the critical section (or doing something complicated), AFAICT.
> Would rhashtable be a workable alternative to rbtree for this use
> case? Skimming <linux/rhashtable.h> it seems to support both lockless
> lookup and traversal using RCU. And according to its manpage,
> *listxattr(2) doesn't guarantee that the returned names are sorted.

I've never used the rhashtable infrastructure in any meaningful way. All
I can say from looking at current users that it looks like it could work
well for us here:

struct simple_xattr {
	struct rhlist_head rhlist_head;
	char *name;
	size_t size;
	char value[];
};

static const struct rhashtable_params simple_xattr_rhashtable = {
	.head_offset = offsetof(struct simple_xattr, rhlist_head),
	.key_offset = offsetof(struct simple_xattr, name),

or sm like this.