netdev - Re: [RFC] breakage in sysfs_readdir() and s

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:	Sun, 12 Jun 2011 19:17:10 +0100
From:	Al Viro <viro@...IV.linux.org.uk>
To:	Linus Torvalds <torvalds@...ux-foundation.org>
Cc:	"Eric W. Biederman" <ebiederm@...ssion.com>,
	linux-fsdevel@...r.kernel.org, netdev@...r.kernel.org,
	Linux Containers <containers@...ts.osdl.org>
Subject: Re: [RFC] breakage in sysfs_readdir() and s_instances abuse in sysfs

On Sun, Jun 12, 2011 at 10:59:42AM -0700, Linus Torvalds wrote:
> On Sun, Jun 12, 2011 at 12:15 AM, Eric W. Biederman
> <ebiederm@...ssion.com> wrote:
> >
> > I honestly hate the pattern that is being used here. ?Holding a
> > reference count because we can't be bothered to free things reliably
> > when we actually stop using them.
> 
> WHAT?
> 
> That's what a reference count *is*. It's all about "free things
> reliably when we actually stop using them".
> 
> Your comment makes zero sense.
> 
> EVERY SINGLE kernel data structure should be reference counted. Read
> Documentation/CodingStyle, or look at any of the good code in the
> kernel (ie core process or VFS code). A non-refcounted data structure
> that is used by more than one entity IS A BUG!
> 
> Quite frankly, your objection sounds moronic. If there is more than
> one user, then a reference count is _always_ the right thing. Nothing
> else ever works, and trust me, people have tried. They've tried
> locking, they've tried luck, they've tried crazy things. Nothing but
> refcounts works.

No, what the current code is trying to do is to have two kinds of references -
contributing to refcount (they do have one, all right) and non-contributing.
*AND* it attempts to hunt non-contributing ones down and replace them with
NULL when refcount hits zero.  And fscks up in dealing with the results.

What this patch does is pretty much the same thing we do for mm_struct and
superblocks - two refcounts, one controlling the shutdown of object and another
controlling the actual freeing of memory.  The second kind of references
contributes to the "memory" refcount and so does having non-zero "active"
refcount.  No games with replacing references with NULL, no races around those,
etc.

Eric's objection is that sysfs superblock would pin the memory occupied by
struct net down until it's unmounted.  Frankly, I think it's a BS -
aforementioned 2.5K are trivial to pin down *anyway*.  Just chdir deep
enough into that instance of sysfs tree and inodes/dentries you've pinned
down by that will easily eat this much.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html