linux-kernel - Re: [PATCH 2/2] drivers: core: Remove glue dirs from sysfs earlier

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]

Message-ID: <20180703154610.GR533219@devbig577.frc2.facebook.com>
Date:   Tue, 3 Jul 2018 08:46:10 -0700
From:   Tejun Heo <tj@...nel.org>
To:     Benjamin Herrenschmidt <benh@...nel.crashing.org>
Cc:     Linus Torvalds <torvalds@...ux-foundation.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        "Eric W. Biederman" <ebiederm@...ssion.com>,
        Joel Stanley <joel@....id.au>
Subject: Re: [PATCH 2/2] drivers: core: Remove glue dirs from sysfs earlier

Hello,

On Tue, Jul 03, 2018 at 03:22:47PM +1000, Benjamin Herrenschmidt wrote:
> +bool kernfs_has_children(struct kernfs_node *kn)
> +{
> +	bool has_children = false;
> +	struct kernfs_node *pos;
> +
> +	/* Lockless shortcut */
> +	if (RB_EMPTY_NODE(&kn->rb))
> +		return false;

Hmm... shouldn't it be testing !rb_first(kn->dir.children)?  The above
would test whether @kn itself is unlinked.

> +
> +	/* Now check for active children */
> +	mutex_lock(&kernfs_mutex);
> +	pos = NULL;
> +	while ((pos = kernfs_next_descendant_post(pos, kn)) && pos != kn) {
> +		if (kernfs_active(pos)) {
> +			has_children = true;
> +			break;
> +		}
> +	}
> +	mutex_unlock(&kernfs_mutex);
> +
> +	return has_children;

The active ref is there to synchronize removal against in-flight reads
so that kernfs_remove() can wait for them to drain.  On return from
kernfs_remove(), the node is guaranteed to be off the rbtree and the
above test isn't necessary.  !rb_first() test should be enough
(provided that there's external synchronization, of course).

Thanks.

-- 
tejun