Re: [PATCH] fix SEM_UNDO with namespaces, take 2

[Manfred Spraul <manfred@colorfullife.com>]

[-- Attachment #1: Type: text/plain, Size: 817 bytes --]

Pavel Emelyanov wrote:
> No, switch_task_namespaces is the wrong place to do this. It is to be 
> done in copy_ipc_ns. If you need a task for which a new namespace is 
> being prepared, then pass one into it.
>
>   
copy_ipc_ns() is also the wrong place:
There are calls after copy_ipc_ns() that can fail (e.g. copy_pid_ns()), 
and if they fail, then the whole syscall is aborted.
But undoing outstanding semaphore operations cannot be undone.
Or simpler: the copy_whatever() functions do not modify current.

Another option would be within sys_unshare():
sys_unshare() first creates all new pointers, and then actual unsharing 
is performed.

What do you think? I that the right place? The implementation could be 
moved into a seperate function, perhaps some of the NULL tests are 
superflous, too.

--
    Manfred

[-- Attachment #2: patch-detach-in-sys_unshare --]
[-- Type: text/plain, Size: 1274 bytes --]

diff --git a/ipc/sem.c b/ipc/sem.c
index 0b45a4d..35841bd 100644
--- a/ipc/sem.c
+++ b/ipc/sem.c
@@ -1298,6 +1298,7 @@ void exit_sem(struct task_struct *tsk)
 	undo_list = tsk->sysvsem.undo_list;
 	if (!undo_list)
 		return;
+	tsk->sysvsem.undo_list = NULL;
 
 	if (!atomic_dec_and_test(&undo_list->refcnt))
 		return;
diff --git a/kernel/fork.c b/kernel/fork.c
index 9c042f9..a3f3abb 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -1733,6 +1733,18 @@ asmlinkage long sys_unshare(unsigned long unshare_flags)
 	if (new_fs ||  new_mm || new_fd || new_ulist || new_nsproxy) {
 
 		if (new_nsproxy) {
+			if ((current->nsproxy == NULL && new_nsproxy != NULL) ||
+				(current->nsproxy != NULL && new_nsproxy == NULL) ||
+				(current->nsproxy != NULL && new_nsproxy != NULL && current->nsproxy->ipc_ns != new_nsproxy->ipc_ns)) {
+				/* switching the IPC namespace is considered equivalent to sys_exit() wrt.
+				 * to outstanding SEM_UNDO undos: After switching to the new IPC namespace,
+				 * the semaphore arrays from the old namespace are not accessible anymore.
+				 * 
+				 * Additionally, an implicit sys_unshare(CLONE_SYSVSEM) is performed.
+				 */
+				exit_sem(current);
+			}
+
 			switch_task_namespaces(current, new_nsproxy);
 			new_nsproxy = NULL;
 		}