Re: [RFC, PATCH] fix SEM_UNDO with namespaces

[Pavel Emelyanov <xemul@openvz.org>]

Manfred Spraul wrote:
> Pavel Emelyanov wrote:
>> Manfred Spraul wrote:
>>   
>>> Hi,
>>>
>>> the attached patch should fix the combination of CLONE_NEWIPC with 
>>> shared sysv undo structures (the common case, just 
>>> sys_unshare(CLONE_NEWIPC)):
>>> lookup_undo() now locates the undo array based on both semid and the 
>>> namespace pointer.
>>>     
>> If you start using any IPC object and then call unshare with CLONE_NEWIPC,
>> then it's your problem, but not the kernel.
>>   
> The result is a kernel memory corruption, and kernel memory corruptions 
> are always the kernel's problem.

Agree. Must be fixed, but I'm not sure we should try handling this
case by trying to de-op semaphores for former task namespace. I think
that destroying this list or returning -EBUSY for this case is OK.

> The code assumed that a semaphore id is globally unique. With 
> namespaces,  this is not true anymore.
> If two semaphore arrays exist with the same id, but different sizes, 
> then semops will cause memory corruptions: The undo structure contains 
> one element for each semaphore, thus the semop will write behind the end 
> of the memory allocation.
> 
>> I agree, that we should probably destroy this one when the task calls 
>> unshare, but trying to keep this list relevant is useless.
>>   
> A very tricky question: Let's assume we have a process with two threads.
> The undo structure is shared, as per opengroup standard.
> Now one thread calls unshare(CLONE_NEWIPC). What should happen? We 
> cannot destroy the undo structure, the other thread might be still 
> interested in it.
> If we allow sys_unshare() for multithreaded processes with CLONE_NEWIPC 
> and without CLONE_SYSVSEM, then we must handle this case.

Hm... I'd simply disable creating any new namespaces for threads.
I think other namespaces developers agree with me. Serge, Suka, Eric
what do you think?

> --
>     Manfred
>