Re: fsnotify_mark_srcu wtf?

From: Amir Goldstein <amir73il@gmail.com>
To: Jan Kara <jack@suse.cz>
Cc: Miklos Szeredi <miklos@szeredi.hu>,
	Eric Paris <eparis@redhat.com>,
	linux-fsdevel <linux-fsdevel@vger.kernel.org>,
	linux-kernel <linux-kernel@vger.kernel.org>
Subject: Re: fsnotify_mark_srcu wtf?
Date: Sun, 13 Nov 2016 20:43:12 +0200	[thread overview]
Message-ID: <CAOQ4uxjaPj6XfU_r1CLXaMsPupjDvjXh8ae00DjrwgvJMvOzLA@mail.gmail.com> (raw)
In-Reply-To: <20161110194625.GG31098@quack2.suse.cz>

On Thu, Nov 10, 2016 at 9:46 PM, Jan Kara <jack@suse.cz> wrote:

...

>
> Well but how would you like to protect the mark list hanging off the inode
> / mountpoint with two SRCUs? You'd need two lists hanging off the inode &
> mountpoint (for different priorities) and that's too big cost to pay to
> accomodate broken userspace...
>

This is the plan.
I have a rough implementation, which still requires some testing:

/*
 * fsnotify_perm_mark[1] protects reads of the first half of inode/vfsmount
 * mark lists, which consist of the high priority (>0) marks, whose masks
 * may contain permission events.
 *
 * fsnotify_perm_mark[0] protects reads of the second half of inode/vfsmount
 * mark lists, which consist of the low priority (0) marks, whose masks
 * do not contain permission events.
 *
 * High priority marks and low priority marks are added to different
 * destroy lists and freed by different reapers, who synchronize only
 * the relevant srcu.
 *
 * Before traversing to first mark on the list we must hold both srcu.
 * While traversing first half, we may temporary drop fsnotify_perm_mark[0]
 * to handle a blocking permission event.
 * When we reach the first low level mark, we may drop fsnotify_perm_mark[1]
 * for good.
 */