Re: [PATCH 1/2] unshare: Unsharing a thread does not require unsharing a vm

[ebiederm@xmission.com (Eric W. Biederman)]

Oleg Nesterov <oleg@redhat.com> writes:

> Let me first say that CLONE_SIGHAND must die, I think ;) and perhaps
> even sighand_struct... I am wondering if we can add something like
>
> 	if ((clone_flags & (CLONE_THREAD | CLONE_SIGHAND)) == CLONE_SIGHAND)
> 		pr_info("You are crazy, please report this to lkml\n");
>
> into copy_process().

The only way killing CLONE_SIGHAND would be viable would be with a
config option.  There are entire generations of linux where libpthreads
used this before CLONE_THREAD was implemented.  Now perhaps no one cares
anymore, but there are a lot of historic binairies that used it, even to
the point where I know of at least one user outside of glibc's pthread
implementation.

> On 08/12, Eric W. Biederman wrote:
>>
>> Oleg Nesterov <oleg@redhat.com> writes:
>>
>> > On 08/11, Eric W. Biederman wrote:
>> >>  	if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) {
>> >> -		/* FIXME: get_task_mm() increments ->mm_users */
>> >> -		if (atomic_read(&current->mm->mm_users) > 1)
>> >> +		if (!thread_group_empty(current))
>> >> +			return -EINVAL;
>> >> +	}
>> >> +	if (unshare_flags & CLONE_VM) {
>> >> +		if (!current_is_single_threaded())
>> >>  			return -EINVAL;
>> >>  	}
>> >
>> > OK, but then you can remove "| CLONE_VM" from the previous check...
>>
>> As an optimization, but I don't think anything cares enough for the
>> optimization to be worth the confusion.
>
> current_is_single_threaded() checks task->signal->live at the start,
> so there is no optimization. But I won't argue, this doesn't hurt.

>> >>  	/*
>> >> +	 * If unsharing a signal handlers, must also unshare the signal queues.
>> >> +	 */
>> >> +	if (unshare_flags & CLONE_SIGHAND)
>> >> +		unshare_flags |= CLONE_THREAD;
>> >
>> > This looks unnecessary, check_unshare_flags() checks "THREAD | SIGHAND".
>> > And to me the comment looks misleading although I won't argue.
>>
>> I absolutely can not understand this code if we jump 5 steps ahead
>> and optimize out the individual dependencies, and try for a flattened
>> dependency tree instead.  I can validate the individual dependencies
>> from first principles.
>>
>> If we jump several steps ahead I can not validate the individual
>> dependencies.
>
> OK,
>
>> > And in fact this doesn't look exactly right, or I am totally confused.
>> > Shouldn't we do
>> >
>> > 	if (unshare_flags & CLONE_SIGHAND)
>> > 		unshare_flags |= CLONE_VM;
>>
>> Nope.  The backward definitions of the flags in unshare has gotten you.
>
> See below,
>
>> CLONE_SIGHAND means that you want a struct sighand_struct with a count
>> of 1.
>
> This is (almost) true,
>
>> Nothing about a sighand_struct with a count of 1 implies or
>> requires mm_users == 1.  clone can quite happily create those.
>
> See
>
> 	if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM))
>
> in copy_process(). So if you have a shared sighand_struct, your ->mm
> is also shared, current_is_single_threaded() will notice this.

Yes.  A shared sighand_struct will have a shared ->mm.  But a private
sighand_struct with count == 1 may also have a shared ->mm.

>> > Otherwise suppose that a single threaded process does clone(VM | SIGHAND)
>> > and (say) child does sys_unshare(SIGHAND). This will wrongly succeed
>> > afaics.
>>
>> Why would it be wrong to succeed in that case?  struct sighand_struct
>> has a count of 1.
>
> How that? clone(VM | SIGHAND) will share ->sighand and increment its
> count.
>
>> unshare(CLONE_SIGHAND) requests a sighand_struct with
>> a count of 1.
>
> Exactly, that is why it is wrong to succeed.

Now that I am clear about what you are talking about I agree with you.

My apologies I clearly misread what you were saying yesterday.

>> unshare(SIGHAND) needs to guarantee that when it returns sighand->count == 1.
>> So unshare(SIGHAND) needs to test for sighand->count == 1.
>
> Oh, I do not think we should check sighand->count. This can lead to
> the same problem we have with the current current->mm->mm_users check.
>
> Most probably today nobody increments sighand->count (I didn't even
> try to verify). But this is possible, and I saw the code which did
> this to pin ->sighand...

I have verified that copy_sighand is the only place in the kernel where
we increment sighand->count today.  de_thread in fs/exec.c even seems to
rely on that.

So while I agree with you that the sighand->count could suffer a similar
fate as mm_users it does not.

Eric