Skip to contentSkip to conversation

cgroup/cpuset: Prevent UAF in proc_cpuset_show()

PATCH Applied Chen Ridong 4 participants 6 messages Jun 26, 2024 – Jun 27, 2024

Subsystems (2)

v3 of 4 Applied Latest is v4 →

lore
Share & tools
lore
#0 lore raw

PATCH v3 1/1 cgroup/cpuset: Prevent UAF in proc_cpuset_show()

An UAF can happen when /proc/cpuset is read as reported in [1].

This can be reproduced by the following methods:
1.add an mdelay(1000) before acquiring the cgroup_lock In the
cgroup_path_ns function.
2.$cat /proc/<pid>/cpuset repeatly.
3.$mount -t cgroup -o cpuset cpuset /sys/fs/cgroup/cpuset/
$umount /sys/fs/cgroup/cpuset/ repeatly.

The race that cause this bug can be shown as below:

(umount) | (cat /proc/<pid>/cpuset)
css_release | proc_cpuset_show
css_release_work_fn | css = task_get_css(tsk, cpuset_cgrp_id);
css_free_rwork_fn | cgroup_path_ns(css->cgroup, ...);
cgroup_destroy_root | mutex_lock(&cgroup_mutex);
rebind_subsystems |
cgroup_free_root |

			|	// cgrp was freed, UAF
			|	cgroup_path_ns_locked(cgrp,..);

When the cpuset is initialized, the root node top_cpuset.css.cgrp will point to &cgrp_dfl_root.cgrp. In cgroup v1, the mount operation will allocate cgroup_root, and top_cpuset.css.cgrp will point to the allocated &cgroup_root.cgrp. When the umount operation is executed, top_cpuset.css.cgrp will be rebound to &cgrp_dfl_root.cgrp.

The problem is that when rebinding to cgrp_dfl_root, there are cases where the cgroup_root allocated by setting up the root for cgroup v1 is cached. This could lead to a Use-After-Free (UAF) if it is subsequently freed. The descendant cgroups of cgroup v1 can only be freed after the css is released. However, the css of the root will never be released, yet the cgroup_root should be freed when it is unmounted. This means that obtaining a reference to the css of the root does not guarantee that css.cgrp->root will not be freed.

Fix this problem by using rcu_read_lock in proc_cpuset_show(). As cgroup root_list is already RCU-safe, css->cgroup is safe. This is similar to commit 9067d90006df ("cgroup: Eliminate the need for cgroup_mutex in proc_cgroup_show()")

[1] https://syzkaller.appspot.com/bug?extid=9b1ff7be974a403aa4cd

Fixes a79a908fd2b0 ("cgroup: introduce cgroup namespaces")
---
 kernel/cgroup/cpuset.c | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)
kernel/cgroup/cpuset.c +10 -2
index c12b9fdb22a4..7f4536c9ccce 100644
@@ -21,6 +21,7 @@
  *  License.  See the file COPYING in the main directory of the Linux
  *  distribution for more details.
  */
+#include "cgroup-internal.h"
 
 #include <linux/cpu.h>
 #include <linux/cpumask.h>
@@ -5052,8 +5053,15 @@ int proc_cpuset_show(struct seq_file *m, struct pid_namespace *ns,
 		goto out;
 
 	css = task_get_css(tsk, cpuset_cgrp_id);
-	retval = cgroup_path_ns(css->cgroup, buf, PATH_MAX,
-				current->nsproxy->cgroup_ns);
+	rcu_read_lock();
+	spin_lock_irq(&css_set_lock);
+	/* In case the root has already been unmounted */
+	if (css->cgroup)
+		retval = cgroup_path_ns_locked(css->cgroup, buf, PATH_MAX,
+					       current->nsproxy->cgroup_ns);
+
+	spin_unlock_irq(&css_set_lock);
+	rcu_read_unlock();
 	css_put(css);
 	if (retval == -E2BIG)
 		retval = -ENAMETOOLONG;
Signature · 1 line
2.34.1
maintainer Reviewed-by on v3 1/1
#1 lore raw
Quoted 69 lines · Chen Ridong

An UAF can happen when /proc/cpuset is read as reported in [1].

This can be reproduced by the following methods:
1.add an mdelay(1000) before acquiring the cgroup_lock In the
cgroup_path_ns function.
2.$cat /proc/<pid>/cpuset repeatly.
3.$mount -t cgroup -o cpuset cpuset /sys/fs/cgroup/cpuset/
$umount /sys/fs/cgroup/cpuset/ repeatly.

The race that cause this bug can be shown as below:

(umount) | (cat /proc/<pid>/cpuset)
css_release | proc_cpuset_show
css_release_work_fn | css = task_get_css(tsk, cpuset_cgrp_id);
css_free_rwork_fn | cgroup_path_ns(css->cgroup, ...);
cgroup_destroy_root | mutex_lock(&cgroup_mutex);
rebind_subsystems |
cgroup_free_root |

			|	// cgrp was freed, UAF
			|	cgroup_path_ns_locked(cgrp,..);

When the cpuset is initialized, the root node top_cpuset.css.cgrp will point to &cgrp_dfl_root.cgrp. In cgroup v1, the mount operation will allocate cgroup_root, and top_cpuset.css.cgrp will point to the allocated &cgroup_root.cgrp. When the umount operation is executed, top_cpuset.css.cgrp will be rebound to &cgrp_dfl_root.cgrp.

The problem is that when rebinding to cgrp_dfl_root, there are cases where the cgroup_root allocated by setting up the root for cgroup v1 is cached. This could lead to a Use-After-Free (UAF) if it is subsequently freed. The descendant cgroups of cgroup v1 can only be freed after the css is released. However, the css of the root will never be released, yet the cgroup_root should be freed when it is unmounted. This means that obtaining a reference to the css of the root does not guarantee that css.cgrp->root will not be freed.

Fix this problem by using rcu_read_lock in proc_cpuset_show(). As cgroup root_list is already RCU-safe, css->cgroup is safe. This is similar to commit 9067d90006df ("cgroup: Eliminate the need for cgroup_mutex in proc_cgroup_show()")

[1] https://syzkaller.appspot.com/bug?extid=9b1ff7be974a403aa4cd

Fixes a79a908fd2b0 ("cgroup: introduce cgroup namespaces")
---
  kernel/cgroup/cpuset.c | 12 ++++++++++--
  1 file changed, 10 insertions(+), 2 deletions(-)
kernel/cgroup/cpuset.c +3 -2
index c12b9fdb22a4..7f4536c9ccce 100644
@@ -21,6 +21,7 @@
   *  License.  See the file COPYING in the main directory of the Linux
   *  distribution for more details.
   */
+#include "cgroup-internal.h"
 
  #include <linux/cpu.h>
  #include <linux/cpumask.h>
@@ -5052,8 +5053,15 @@ int proc_cpuset_show(struct seq_file *m, struct pid_namespace *ns,
  		goto out;
 
  	css = task_get_css(tsk, cpuset_cgrp_id);
-	retval = cgroup_path_ns(css->cgroup, buf, PATH_MAX,
-				current->nsproxy->cgroup_ns);
+	rcu_read_lock();
+	spin_lock_irq(&css_set_lock);

kernel/cgroup/cpuset.c int proc_cpuset_show(struct seq_file *m, struct pid_namespace *ns, · line 5067

Quoted 10 lines
kernel/cgroup/cpuset.c +7 -0
+	/* In case the root has already been unmounted */
+	if (css->cgroup)
+		retval = cgroup_path_ns_locked(css->cgroup, buf, PATH_MAX,
+					       current->nsproxy->cgroup_ns);
+
+	spin_unlock_irq(&css_set_lock);
+	rcu_read_unlock();
  	css_put(css);
  	if (retval == -E2BIG)
  		retval = -ENAMETOOLONG;
maintainer Applied
#2 lore raw
In reply to Chen Ridong
Quoted 41 lines · Chen Ridong

An UAF can happen when /proc/cpuset is read as reported in [1].

This can be reproduced by the following methods:
1.add an mdelay(1000) before acquiring the cgroup_lock In the
cgroup_path_ns function.
2.$cat /proc/<pid>/cpuset repeatly.
3.$mount -t cgroup -o cpuset cpuset /sys/fs/cgroup/cpuset/
$umount /sys/fs/cgroup/cpuset/ repeatly.

The race that cause this bug can be shown as below:

(umount) | (cat /proc/<pid>/cpuset)
css_release | proc_cpuset_show
css_release_work_fn | css = task_get_css(tsk, cpuset_cgrp_id);
css_free_rwork_fn | cgroup_path_ns(css->cgroup, ...);
cgroup_destroy_root | mutex_lock(&cgroup_mutex);
rebind_subsystems |
cgroup_free_root |

			|	// cgrp was freed, UAF
			|	cgroup_path_ns_locked(cgrp,..);

When the cpuset is initialized, the root node top_cpuset.css.cgrp will point to &cgrp_dfl_root.cgrp. In cgroup v1, the mount operation will allocate cgroup_root, and top_cpuset.css.cgrp will point to the allocated &cgroup_root.cgrp. When the umount operation is executed, top_cpuset.css.cgrp will be rebound to &cgrp_dfl_root.cgrp.

The problem is that when rebinding to cgrp_dfl_root, there are cases where the cgroup_root allocated by setting up the root for cgroup v1 is cached. This could lead to a Use-After-Free (UAF) if it is subsequently freed. The descendant cgroups of cgroup v1 can only be freed after the css is released. However, the css of the root will never be released, yet the cgroup_root should be freed when it is unmounted. This means that obtaining a reference to the css of the root does not guarantee that css.cgrp->root will not be freed.

Fix this problem by using rcu_read_lock in proc_cpuset_show(). As cgroup root_list is already RCU-safe, css->cgroup is safe. This is similar to commit 9067d90006df ("cgroup: Eliminate the need for cgroup_mutex in proc_cgroup_show()")

Quoted 4 lines

[1] https://syzkaller.appspot.com/bug?extid=9b1ff7be974a403aa4cd

Fixes a79a908fd2b0 ("cgroup: introduce cgroup namespaces")

Applied to cgroup/for-6.10-fixes w/ stable cc added.

Thanks.

Signature · 1 line
tejun
maintainer
#3 lore raw
In reply to Chen Ridong
Quoted 18 lines · Chen Ridong

An UAF can happen when /proc/cpuset is read as reported in [1].

This can be reproduced by the following methods:
1.add an mdelay(1000) before acquiring the cgroup_lock In the
cgroup_path_ns function.
2.$cat /proc/<pid>/cpuset repeatly.
3.$mount -t cgroup -o cpuset cpuset /sys/fs/cgroup/cpuset/
$umount /sys/fs/cgroup/cpuset/ repeatly.

The race that cause this bug can be shown as below:

(umount) | (cat /proc/<pid>/cpuset)
css_release | proc_cpuset_show
css_release_work_fn | css = task_get_css(tsk, cpuset_cgrp_id);
css_free_rwork_fn | cgroup_path_ns(css->cgroup, ...);
cgroup_destroy_root | mutex_lock(&cgroup_mutex);
rebind_subsystems |
cgroup_free_root |

Quoted 2 lines
			|	// cgrp was freed, UAF
			|	cgroup_path_ns_locked(cgrp,..);

Thanks for this breakdown.

Quoted 5 lines

...
Fix this problem by using rcu_read_lock in proc_cpuset_show(). As cgroup root_list is already RCU-safe, css->cgroup is safe. This is similar to commit 9067d90006df ("cgroup: Eliminate the need for cgroup_mutex in proc_cgroup_show()")

Apologies for misleading you in my previous message about root_list. As I look better at proc_cpuset_show vs proc_cgroup_show, there's a difference and task_get_css() doesn't rely on root_list synchronization.

I think it could go like this (with my extra comments)

	rcu_read_lock();
	spin_lock_irq(&css_set_lock);
	css = task_css(tsk, cpuset_cgrp_id); // css is stable wrt task's migration thanks to css_set_lock
	cgrp = css->cgroup; // whatever we see here, won't be free'd thanks to RCU lock and cgroup_free_root/kfree_rcu
	retval = cgroup_path_ns_locked(cgrp, buf, PATH_MAX,
				       current->nsproxy->cgroup_ns);
	...

Your patch should work thanks to the rcu_read_lock and cgroup_free_root/kfree_rcu and the if (css->cgroup) guard is unnecessary.

So the patch is a functional fix, the reasoning in commit message is little off. Not sure if Tejun rebases his for-6.10-fixes (with a possible v4), full fixup commit for this may not be worthy.

Michal

maintainer
#4 lore raw

Hello,

Quoted 7 lines · Michal Koutný

Your patch should work thanks to the rcu_read_lock and cgroup_free_root/kfree_rcu and the if (css->cgroup) guard is unnecessary.

So the patch is a functional fix, the reasoning in commit message is little off. Not sure if Tejun rebases his for-6.10-fixes (with a possible v4), full fixup commit for this may not be worthy.

This one's on the top. If Chen can send me a patch with updated description, I will replace the patch.

Thanks.

Signature · 1 line
tejun
maintainer
#5 lore raw
Quoted 6 lines · Tejun Heo

Hello,

Quoted 3 lines · Michal Koutný

Your patch should work thanks to the rcu_read_lock and cgroup_free_root/kfree_rcu and the if (css->cgroup) guard is unnecessary.

I also notice that the if (css->cgroup) guard is not needed. I didn't reject it because it doesn't hurt.

Cheers,
Longman

Quoted 9 lines
Quoted 4 lines

So the patch is a functional fix, the reasoning in commit message is little off. Not sure if Tejun rebases his for-6.10-fixes (with a possible v4), full fixup commit for this may not be worthy.

This one's on the top. If Chen can send me a patch with updated description, I will replace the patch.

Thanks.