linux-kernel - Re: [path][rfc] add PR

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Tue, 05 Apr 2011 00:05:52 +0400
From:	Stas Sergeev <stsp@...et.ru>
To:	Oleg Nesterov <oleg@...hat.com>
CC:	Linux kernel <linux-kernel@...r.kernel.org>
Subject: Re: [path][rfc] add PR_DETACH prctl command

04.04.2011 20:03, Oleg Nesterov wrote:
>> I still haven't solved the problems with checking
>> parent and checking ptrace, so ignore them for
>> now (or give me the hints:)
> Not sure I understand your question...
I wonder how to check whether the child was
reparented to init, and not is a child on init's
subthread. You suggested to check
same_thread_group(real_parent, ns->child_reaper),
but I don't understand that suggestion because in
our case real_parent==ns->child_reaper, yet it is not
enough, as it could be of CLONE_PARENT (I suppose).

>> +			while_each_thread(me, p) {
>> +				if (!ptrace_reparented(p))
>> +					p->parent = pid_ns->child_reaper;
>> +				p->real_parent = pid_ns->child_reaper;
> Eek. Even ignoring ptrace, this is weird. We change parent/real_parent,
> but we do not do list_move_tail(sibling) until wait_task_detached() !
> No, I think we should not do this even if this was correct. I'll try
> to nack this in any case, even if there were no immediate problems ;)
> IMHO, this is insane.
>
> But this is wrong. Well. Suppose that the caller of PR_DETACH exits
> before the old parent does do_wait(). What /sbin/init (who is the new
> parent) can do after it gets SIGCHLD? If can't see this zombie. Nor
> the old parent can release this task due to ->detaching. Eventually
> /sbin/init can reap it if it does, say, waitpid(-1), but still this
> is wrong.
No, the idea was like that: the old parent either wait()s or
exits, then init became a "real" parent of that process, and
reaps it immediately. I think that's natural: the very same will happen
if the child didn't do pr_detach. If it exits, its current parent
have to either wait(), or exit. If it doesn't do so - zombie.
I decided that if I still allow wait() for the old parent, I'd better
leave that logic untouched. So: until the old parent wait()s
or exits, we have zombie, and that's what I supposed to implement.

> Or. Suppose that the old parent exits after its child does PR_DETACH.
> You changed forget_original_parent() but this is not enough, note that
> find_new_reaper() can pick the live sub-thread. In this case the child
> will be moved to init's ->children list, and after that we are changing
> ->real_parent back.
How? I think I prevented that with this:
---

+			p->detaching = 0;
+			continue;
---
So it should not go down to get reparented back.
Or am I missing something?

Thanks for your time!

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/