linux-kernel - Re: [PATCH v8 3/4] cgroups: allow a cgroup subsystem to reject a fork

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, 7 Apr 2015 23:47:02 +1000
From:	Aleksa Sarai <cyphar@...har.com>
To:	Tejun Heo <tj@...nel.org>
Cc:	lizefan@...wei.com, mingo@...hat.com, peterz@...radead.org,
	richard@....at,
	Frédéric Weisbecker <fweisbec@...il.com>,
	linux-kernel@...r.kernel.org, cgroups@...r.kernel.org
Subject: Re: [PATCH v8 3/4] cgroups: allow a cgroup subsystem to reject a fork

Hi Tejun,

>> I tried doing this and the kernel would refuse to boot. I believe it has
>> something to do with the ordering of early_init subsystems, but I'm not
>
> Hmmm... yeah, failure in early_init can be tricky to debug.

Okay, I took another look and it isn't that there are hard ordering
dependencies, it's because of how I wrote the tagging macro for cgroup_subsys.h
that caused a NULL deref during the early_init. Whoops.

>> entirely sure (this optimisation can be dealt with later [it's non-critical],
>> so IMO this should be done in a separate patchset [if at all]). Also, your
>> later comments would fix the subsys bitmask problem (we can just pass the
>> default %NULL), we don't even need to test the index.
>
> [...] I don't
> think it's a good idea to send the patches as-are because we can't
> debug and fix them properly, right? [...]

In what way are they hard to debug?

> [...] If there are hard ordering
> dependencies, the range of subsystems which require fork/exit doesn't
> have to be at the beginning.

I just spent a few hours trying to write it and it's just too dodgy to live.
Either you have a solution that won't work with nested tags, and is generally
just kind of bad or you have a solution that requires you to keep two separate
enumerations in sync. And of course you need to offset the index you're looking
stuff up in ss_state so that also looks bad.

But that's all besides the point because, *even if* it I had a clean solution,
it still wouldn't solve the fact that we are
> [...] adding quite a few loops in relatively hot paths.
You kind of _need_ to loop over all the subsystems in either case, the only
difference between having an array of CGROUP_SUBSYS_COUNT pointers or
CGROUP_PREFORK_COUNT is the few bytes of memory you've "saved" (at the expense
of making the callback code essentially unreadable).

--
Aleksa Sarai (cyphar)
www.cyphar.com
--
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/