summaryrefslogtreecommitdiff
path: root/ipc/shm.c
diff options
context:
space:
mode:
authorAlexander Mikhalitsyn <alexander.mikhalitsyn@virtuozzo.com>2021-11-20 03:43:21 +0300
committerLinus Torvalds <torvalds@linux-foundation.org>2021-11-20 21:35:54 +0300
commit85b6d24646e4125c591639841169baa98a2da503 (patch)
treebe067ff74ef80dbe3fa459d81a88ac852c7e6684 /ipc/shm.c
parent126e8bee943e9926238c891e2df5b5573aee76bc (diff)
downloadlinux-85b6d24646e4125c591639841169baa98a2da503.tar.xz
shm: extend forced shm destroy to support objects from several IPC nses
Currently, the exit_shm() function not designed to work properly when task->sysvshm.shm_clist holds shm objects from different IPC namespaces. This is a real pain when sysctl kernel.shm_rmid_forced = 1, because it leads to use-after-free (reproducer exists). This is an attempt to fix the problem by extending exit_shm mechanism to handle shm's destroy from several IPC ns'es. To achieve that we do several things: 1. add a namespace (non-refcounted) pointer to the struct shmid_kernel 2. during new shm object creation (newseg()/shmget syscall) we initialize this pointer by current task IPC ns 3. exit_shm() fully reworked such that it traverses over all shp's in task->sysvshm.shm_clist and gets IPC namespace not from current task as it was before but from shp's object itself, then call shm_destroy(shp, ns). Note: We need to be really careful here, because as it was said before (1), our pointer to IPC ns non-refcnt'ed. To be on the safe side we using special helper get_ipc_ns_not_zero() which allows to get IPC ns refcounter only if IPC ns not in the "state of destruction". Q/A Q: Why can we access shp->ns memory using non-refcounted pointer? A: Because shp object lifetime is always shorther than IPC namespace lifetime, so, if we get shp object from the task->sysvshm.shm_clist while holding task_lock(task) nobody can steal our namespace. Q: Does this patch change semantics of unshare/setns/clone syscalls? A: No. It's just fixes non-covered case when process may leave IPC namespace without getting task->sysvshm.shm_clist list cleaned up. Link: https://lkml.kernel.org/r/67bb03e5-f79c-1815-e2bf-949c67047418@colorfullife.com Link: https://lkml.kernel.org/r/20211109151501.4921-1-manfred@colorfullife.com Fixes: ab602f79915 ("shm: make exit_shm work proportional to task activity") Co-developed-by: Manfred Spraul <manfred@colorfullife.com> Signed-off-by: Manfred Spraul <manfred@colorfullife.com> Signed-off-by: Alexander Mikhalitsyn <alexander.mikhalitsyn@virtuozzo.com> Cc: "Eric W. Biederman" <ebiederm@xmission.com> Cc: Davidlohr Bueso <dave@stgolabs.net> Cc: Greg KH <gregkh@linuxfoundation.org> Cc: Andrei Vagin <avagin@gmail.com> Cc: Pavel Tikhomirov <ptikhomirov@virtuozzo.com> Cc: Vasily Averin <vvs@virtuozzo.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Diffstat (limited to 'ipc/shm.c')
-rw-r--r--ipc/shm.c189
1 files changed, 143 insertions, 46 deletions
diff --git a/ipc/shm.c b/ipc/shm.c
index 4942bdd65748..b3048ebd5c31 100644
--- a/ipc/shm.c
+++ b/ipc/shm.c
@@ -62,9 +62,18 @@ struct shmid_kernel /* private to the kernel */
struct pid *shm_lprid;
struct ucounts *mlock_ucounts;
- /* The task created the shm object. NULL if the task is dead. */
+ /*
+ * The task created the shm object, for
+ * task_lock(shp->shm_creator)
+ */
struct task_struct *shm_creator;
- struct list_head shm_clist; /* list by creator */
+
+ /*
+ * List by creator. task_lock(->shm_creator) required for read/write.
+ * If list_empty(), then the creator is dead already.
+ */
+ struct list_head shm_clist;
+ struct ipc_namespace *ns;
} __randomize_layout;
/* shm_mode upper byte flags */
@@ -115,6 +124,7 @@ static void do_shm_rmid(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp)
struct shmid_kernel *shp;
shp = container_of(ipcp, struct shmid_kernel, shm_perm);
+ WARN_ON(ns != shp->ns);
if (shp->shm_nattch) {
shp->shm_perm.mode |= SHM_DEST;
@@ -225,10 +235,43 @@ static void shm_rcu_free(struct rcu_head *head)
kfree(shp);
}
-static inline void shm_rmid(struct ipc_namespace *ns, struct shmid_kernel *s)
+/*
+ * It has to be called with shp locked.
+ * It must be called before ipc_rmid()
+ */
+static inline void shm_clist_rm(struct shmid_kernel *shp)
{
- list_del(&s->shm_clist);
- ipc_rmid(&shm_ids(ns), &s->shm_perm);
+ struct task_struct *creator;
+
+ /* ensure that shm_creator does not disappear */
+ rcu_read_lock();
+
+ /*
+ * A concurrent exit_shm may do a list_del_init() as well.
+ * Just do nothing if exit_shm already did the work
+ */
+ if (!list_empty(&shp->shm_clist)) {
+ /*
+ * shp->shm_creator is guaranteed to be valid *only*
+ * if shp->shm_clist is not empty.
+ */
+ creator = shp->shm_creator;
+
+ task_lock(creator);
+ /*
+ * list_del_init() is a nop if the entry was already removed
+ * from the list.
+ */
+ list_del_init(&shp->shm_clist);
+ task_unlock(creator);
+ }
+ rcu_read_unlock();
+}
+
+static inline void shm_rmid(struct shmid_kernel *s)
+{
+ shm_clist_rm(s);
+ ipc_rmid(&shm_ids(s->ns), &s->shm_perm);
}
@@ -283,7 +326,7 @@ static void shm_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp)
shm_file = shp->shm_file;
shp->shm_file = NULL;
ns->shm_tot -= (shp->shm_segsz + PAGE_SIZE - 1) >> PAGE_SHIFT;
- shm_rmid(ns, shp);
+ shm_rmid(shp);
shm_unlock(shp);
if (!is_file_hugepages(shm_file))
shmem_lock(shm_file, 0, shp->mlock_ucounts);
@@ -303,10 +346,10 @@ static void shm_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp)
*
* 2) sysctl kernel.shm_rmid_forced is set to 1.
*/
-static bool shm_may_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp)
+static bool shm_may_destroy(struct shmid_kernel *shp)
{
return (shp->shm_nattch == 0) &&
- (ns->shm_rmid_forced ||
+ (shp->ns->shm_rmid_forced ||
(shp->shm_perm.mode & SHM_DEST));
}
@@ -337,7 +380,7 @@ static void shm_close(struct vm_area_struct *vma)
ipc_update_pid(&shp->shm_lprid, task_tgid(current));
shp->shm_dtim = ktime_get_real_seconds();
shp->shm_nattch--;
- if (shm_may_destroy(ns, shp))
+ if (shm_may_destroy(shp))
shm_destroy(ns, shp);
else
shm_unlock(shp);
@@ -358,10 +401,10 @@ static int shm_try_destroy_orphaned(int id, void *p, void *data)
*
* As shp->* are changed under rwsem, it's safe to skip shp locking.
*/
- if (shp->shm_creator != NULL)
+ if (!list_empty(&shp->shm_clist))
return 0;
- if (shm_may_destroy(ns, shp)) {
+ if (shm_may_destroy(shp)) {
shm_lock_by_ptr(shp);
shm_destroy(ns, shp);
}
@@ -379,48 +422,97 @@ void shm_destroy_orphaned(struct ipc_namespace *ns)
/* Locking assumes this will only be called with task == current */
void exit_shm(struct task_struct *task)
{
- struct ipc_namespace *ns = task->nsproxy->ipc_ns;
- struct shmid_kernel *shp, *n;
+ for (;;) {
+ struct shmid_kernel *shp;
+ struct ipc_namespace *ns;
- if (list_empty(&task->sysvshm.shm_clist))
- return;
+ task_lock(task);
+
+ if (list_empty(&task->sysvshm.shm_clist)) {
+ task_unlock(task);
+ break;
+ }
+
+ shp = list_first_entry(&task->sysvshm.shm_clist, struct shmid_kernel,
+ shm_clist);
- /*
- * If kernel.shm_rmid_forced is not set then only keep track of
- * which shmids are orphaned, so that a later set of the sysctl
- * can clean them up.
- */
- if (!ns->shm_rmid_forced) {
- down_read(&shm_ids(ns).rwsem);
- list_for_each_entry(shp, &task->sysvshm.shm_clist, shm_clist)
- shp->shm_creator = NULL;
/*
- * Only under read lock but we are only called on current
- * so no entry on the list will be shared.
+ * 1) Get pointer to the ipc namespace. It is worth to say
+ * that this pointer is guaranteed to be valid because
+ * shp lifetime is always shorter than namespace lifetime
+ * in which shp lives.
+ * We taken task_lock it means that shp won't be freed.
*/
- list_del(&task->sysvshm.shm_clist);
- up_read(&shm_ids(ns).rwsem);
- return;
- }
+ ns = shp->ns;
- /*
- * Destroy all already created segments, that were not yet mapped,
- * and mark any mapped as orphan to cover the sysctl toggling.
- * Destroy is skipped if shm_may_destroy() returns false.
- */
- down_write(&shm_ids(ns).rwsem);
- list_for_each_entry_safe(shp, n, &task->sysvshm.shm_clist, shm_clist) {
- shp->shm_creator = NULL;
+ /*
+ * 2) If kernel.shm_rmid_forced is not set then only keep track of
+ * which shmids are orphaned, so that a later set of the sysctl
+ * can clean them up.
+ */
+ if (!ns->shm_rmid_forced)
+ goto unlink_continue;
- if (shm_may_destroy(ns, shp)) {
- shm_lock_by_ptr(shp);
- shm_destroy(ns, shp);
+ /*
+ * 3) get a reference to the namespace.
+ * The refcount could be already 0. If it is 0, then
+ * the shm objects will be free by free_ipc_work().
+ */
+ ns = get_ipc_ns_not_zero(ns);
+ if (!ns) {
+unlink_continue:
+ list_del_init(&shp->shm_clist);
+ task_unlock(task);
+ continue;
}
- }
- /* Remove the list head from any segments still attached. */
- list_del(&task->sysvshm.shm_clist);
- up_write(&shm_ids(ns).rwsem);
+ /*
+ * 4) get a reference to shp.
+ * This cannot fail: shm_clist_rm() is called before
+ * ipc_rmid(), thus the refcount cannot be 0.
+ */
+ WARN_ON(!ipc_rcu_getref(&shp->shm_perm));
+
+ /*
+ * 5) unlink the shm segment from the list of segments
+ * created by current.
+ * This must be done last. After unlinking,
+ * only the refcounts obtained above prevent IPC_RMID
+ * from destroying the segment or the namespace.
+ */
+ list_del_init(&shp->shm_clist);
+
+ task_unlock(task);
+
+ /*
+ * 6) we have all references
+ * Thus lock & if needed destroy shp.
+ */
+ down_write(&shm_ids(ns).rwsem);
+ shm_lock_by_ptr(shp);
+ /*
+ * rcu_read_lock was implicitly taken in shm_lock_by_ptr, it's
+ * safe to call ipc_rcu_putref here
+ */
+ ipc_rcu_putref(&shp->shm_perm, shm_rcu_free);
+
+ if (ipc_valid_object(&shp->shm_perm)) {
+ if (shm_may_destroy(shp))
+ shm_destroy(ns, shp);
+ else
+ shm_unlock(shp);
+ } else {
+ /*
+ * Someone else deleted the shp from namespace
+ * idr/kht while we have waited.
+ * Just unlock and continue.
+ */
+ shm_unlock(shp);
+ }
+
+ up_write(&shm_ids(ns).rwsem);
+ put_ipc_ns(ns); /* paired with get_ipc_ns_not_zero */
+ }
}
static vm_fault_t shm_fault(struct vm_fault *vmf)
@@ -676,7 +768,11 @@ static int newseg(struct ipc_namespace *ns, struct ipc_params *params)
if (error < 0)
goto no_id;
+ shp->ns = ns;
+
+ task_lock(current);
list_add(&shp->shm_clist, &current->sysvshm.shm_clist);
+ task_unlock(current);
/*
* shmid gets reported as "inode#" in /proc/pid/maps.
@@ -1567,7 +1663,8 @@ out_nattch:
down_write(&shm_ids(ns).rwsem);
shp = shm_lock(ns, shmid);
shp->shm_nattch--;
- if (shm_may_destroy(ns, shp))
+
+ if (shm_may_destroy(shp))
shm_destroy(ns, shp);
else
shm_unlock(shp);