Re: [PATCH] nfsd: fix kernel crash when load nfsd in docker

["J. Bruce Fields" <bfields@fieldses.org>]

On Mon, Jun 15, 2020 at 03:12:11PM +0800, Luo Xiaogang wrote:
> We load nfsd module in the docker container, kernel crash as following.
> 
> The 'current->nsproxy->net_ns->gen->ptr[nfsd_net_id]' is overflow in the
> nfsd_init_net.
> 
> We should use the net_ns which is being init in the nfsd_init_net,
> not the 'current->nsproxy->net_ns'.

Thanks!  Actually, I think my problem was that net init and exit are
just the wrong place to be doing this--I moved them to nfsd start/stop
instead.

And then that exposed the fact that I had an inode leak.

Do the following two patches help?

--b.

From 16f954bd5c481596a63271a91963bf260e2f3f46 Mon Sep 17 00:00:00 2001
From: "J. Bruce Fields" <bfields@redhat.com>
Date: Tue, 23 Jun 2020 16:00:33 -0400
Subject: [PATCH 1/2] nfsd4: fix nfsdfs reference count loop

We don't drop the reference on the nfsdfs filesystem with
mntput(nn->nfsd_mnt) until nfsd_exit_net(), but that won't be called
until the nfsd module's unloaded, and we can't unload the module as long
as there's a reference on nfsdfs.  So this prevents module unloading.

Signed-off-by: J. Bruce Fields <bfields@redhat.com>
---
 fs/nfsd/nfs4state.c |  8 +++++++-
 fs/nfsd/nfsctl.c    | 22 ++++++++++++----------
 fs/nfsd/nfsd.h      |  3 +++
 3 files changed, 22 insertions(+), 11 deletions(-)

diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
index bb3d2c32664a..cce2510b2cca 100644
--- a/fs/nfsd/nfs4state.c
+++ b/fs/nfsd/nfs4state.c
@@ -7912,9 +7912,14 @@ nfs4_state_start_net(struct net *net)
 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 	int ret;
 
-	ret = nfs4_state_create_net(net);
+	ret = get_nfsdfs(net);
 	if (ret)
 		return ret;
+	ret = nfs4_state_create_net(net);
+	if (ret) {
+		mntput(nn->nfsd_mnt);
+		return ret;
+	}
 	locks_start_grace(net, &nn->nfsd4_manager);
 	nfsd4_client_tracking_init(net);
 	if (nn->track_reclaim_completes && nn->reclaim_str_hashtbl_size == 0)
@@ -7984,6 +7989,7 @@ nfs4_state_shutdown_net(struct net *net)
 
 	nfsd4_client_tracking_exit(net);
 	nfs4_state_destroy_net(net);
+	mntput(nn->nfsd_mnt);
 }
 
 void
diff --git a/fs/nfsd/nfsctl.c b/fs/nfsd/nfsctl.c
index b68e96681522..cf98a81ca1ea 100644
--- a/fs/nfsd/nfsctl.c
+++ b/fs/nfsd/nfsctl.c
@@ -1424,6 +1424,18 @@ static struct file_system_type nfsd_fs_type = {
 };
 MODULE_ALIAS_FS("nfsd");
 
+int get_nfsdfs(struct net *net)
+{
+	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
+	struct vfsmount *mnt;
+
+	mnt =  vfs_kern_mount(&nfsd_fs_type, SB_KERNMOUNT, "nfsd", NULL);
+	if (IS_ERR(mnt))
+		return PTR_ERR(mnt);
+	nn->nfsd_mnt = mnt;
+	return 0;
+}
+
 #ifdef CONFIG_PROC_FS
 static int create_proc_exports_entry(void)
 {
@@ -1451,7 +1463,6 @@ unsigned int nfsd_net_id;
 static __net_init int nfsd_init_net(struct net *net)
 {
 	int retval;
-	struct vfsmount *mnt;
 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 
 	retval = nfsd_export_init(net);
@@ -1478,16 +1489,8 @@ static __net_init int nfsd_init_net(struct net *net)
 	init_waitqueue_head(&nn->ntf_wq);
 	seqlock_init(&nn->boot_lock);
 
-	mnt =  vfs_kern_mount(&nfsd_fs_type, SB_KERNMOUNT, "nfsd", NULL);
-	if (IS_ERR(mnt)) {
-		retval = PTR_ERR(mnt);
-		goto out_mount_err;
-	}
-	nn->nfsd_mnt = mnt;
 	return 0;
 
-out_mount_err:
-	nfsd_reply_cache_shutdown(nn);
 out_drc_error:
 	nfsd_idmap_shutdown(net);
 out_idmap_error:
@@ -1500,7 +1503,6 @@ static __net_exit void nfsd_exit_net(struct net *net)
 {
 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 
-	mntput(nn->nfsd_mnt);
 	nfsd_reply_cache_shutdown(nn);
 	nfsd_idmap_shutdown(net);
 	nfsd_export_shutdown(net);
diff --git a/fs/nfsd/nfsd.h b/fs/nfsd/nfsd.h
index 36cdd81b6688..57c832d1b30f 100644
--- a/fs/nfsd/nfsd.h
+++ b/fs/nfsd/nfsd.h
@@ -90,6 +90,8 @@ void		nfsd_destroy(struct net *net);
 
 bool		i_am_nfsd(void);
 
+int get_nfsdfs(struct net *);
+
 struct nfsdfs_client {
 	struct kref cl_ref;
 	void (*cl_release)(struct kref *kref);
@@ -100,6 +102,7 @@ struct dentry *nfsd_client_mkdir(struct nfsd_net *nn,
 		struct nfsdfs_client *ncl, u32 id, const struct tree_descr *);
 void nfsd_client_rmdir(struct dentry *dentry);
 
+
 #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL)
 #ifdef CONFIG_NFSD_V2_ACL
 extern const struct svc_version nfsd_acl_version2;
-- 
2.26.2


From 51de3b460b39e862f7dcfd4d600e8de0afe73e29 Mon Sep 17 00:00:00 2001
From: "J. Bruce Fields" <bfields@redhat.com>
Date: Tue, 23 Jun 2020 21:01:19 -0400
Subject: [PATCH 2/2] nfsd: fix nfsdfs inode reference count leak

I don't understand this code well, but  I'm seeing a warning about a
still-referenced inode on unmount, and every other similar filesystem
does a dput() here.

Signed-off-by: J. Bruce Fields <bfields@redhat.com>
---
 fs/nfsd/nfsctl.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/fs/nfsd/nfsctl.c b/fs/nfsd/nfsctl.c
index cf98a81ca1ea..cd05732f8eaa 100644
--- a/fs/nfsd/nfsctl.c
+++ b/fs/nfsd/nfsctl.c
@@ -1335,6 +1335,7 @@ void nfsd_client_rmdir(struct dentry *dentry)
 	WARN_ON_ONCE(ret);
 	fsnotify_rmdir(dir, dentry);
 	d_delete(dentry);
+	dput(dentry);
 	inode_unlock(dir);
 }
 
-- 
2.26.2