Received-SPF: pass (google.com: domain of linux-kernel+bounces-72757-linux.lists.archive=gmail.com@vger.kernel.org designates 139.178.88.99 as permitted sender) client-ip=139.178.88.99;
Date: Tue, 20 Feb 2024 10:56:53 +0100
From: Christian Brauner <brauner@kernel.org>
To: Chuck Lever <cel@kernel.org>
Cc: viro@zeniv.linux.org.uk, jack@suse.cz, hughd@google.com, 
	akpm@linux-foundation.org, Liam.Howlett@oracle.com, oliver.sang@intel.com, 
	feng.tang@intel.com, linux-kernel@vger.kernel.org, linux-fsdevel@vger.kernel.org, 
	maple-tree@lists.infradead.org, linux-mm@kvack.org, lkp@intel.com
Subject: Re: [PATCH v2 1/6] libfs: Re-arrange locking in offset_iterate_dir()
Message-ID: <20240220-ortsrand-initialen-43550ee746ed@brauner>
References: <170820083431.6328.16233178852085891453.stgit@91.116.238.104.host.secureserver.net>
 <170820142021.6328.15047865406275957018.stgit@91.116.238.104.host.secureserver.net>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
In-Reply-To: <170820142021.6328.15047865406275957018.stgit@91.116.238.104.host.secureserver.net>

On Sat, Feb 17, 2024 at 03:23:40PM -0500, Chuck Lever wrote:
> From: Chuck Lever <chuck.lever@oracle.com>
> 
> Liam and Matthew say that once the RCU read lock is released,
> xa_state is not safe to re-use for the next xas_find() call. But the
> RCU read lock must be released on each loop iteration so that
> dput(), which might_sleep(), can be called safely.

Fwiw, functions like this:

static struct dentry *offset_find_next(struct xa_state *xas)
{
        struct dentry *child, *found = NULL;

        rcu_read_lock();
        child = xas_next_entry(xas, U32_MAX);
        if (!child)
                goto out;
        spin_lock(&child->d_lock);
        if (simple_positive(child))
                found = dget_dlock(child);
        spin_unlock(&child->d_lock);
out:
        rcu_read_unlock();
        return found;
}

should use the new guard feature going forward imho. IOW, in the future such
helpers should be written as:

static struct dentry *offset_find_next(struct xa_state *xas)
{
        struct dentry *child, *found = NULL;

	guard(rcu)();
        child = xas_next_entry(xas, U32_MAX);
        if (!child)
		return NULL;
        spin_lock(&child->d_lock);
        if (simple_positive(child))
                found = dget_dlock(child);
        spin_unlock(&child->d_lock);
        return found;
}

which allows you to eliminate the goto and to have the guarantee that the rcu
lock is released when you return. This also works for other locks btw.