[PATCH] NFSv4.1: Kill nfs4_ds_disconnect()

[PATCH] NFSv4.1: Kill nfs4_ds_disconnect()

[Trond Myklebust]

There is nothing to prevent another thread from dereferencing ds->ds_clp
during or after the call to nfs4_ds_disconnect(), and Oopsing due to the
resulting NULL pointer.

Instead, we should just rely on filelayout_mark_devid_invalid() to keep
us out of trouble by avoiding that deviceid.

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
---
 fs/nfs/nfs4filelayout.c    |  1 -
 fs/nfs/nfs4filelayout.h    |  1 -
 fs/nfs/nfs4filelayoutdev.c | 22 ----------------------
 3 files changed, 24 deletions(-)

diff --git a/fs/nfs/nfs4filelayout.c b/fs/nfs/nfs4filelayout.c
index 52d8472..816c2d0 100644
--- a/fs/nfs/nfs4filelayout.c
+++ b/fs/nfs/nfs4filelayout.c
@@ -207,7 +207,6 @@ static int filelayout_async_handle_error(struct rpc_task *task,
 		clear_bit(NFS_INO_LAYOUTCOMMIT, &NFS_I(inode)->flags);
 		_pnfs_return_layout(inode);
 		rpc_wake_up(&tbl->slot_tbl_waitq);
-		nfs4_ds_disconnect(clp);
 		/* fall through */
 	default:
 reset:
diff --git a/fs/nfs/nfs4filelayout.h b/fs/nfs/nfs4filelayout.h
index dca47d78..8c07241 100644
--- a/fs/nfs/nfs4filelayout.h
+++ b/fs/nfs/nfs4filelayout.h
@@ -149,6 +149,5 @@ extern void nfs4_fl_put_deviceid(struct nfs4_file_layout_dsaddr *dsaddr);
 extern void nfs4_fl_free_deviceid(struct nfs4_file_layout_dsaddr *dsaddr);
 struct nfs4_file_layout_dsaddr *
 filelayout_get_device_info(struct inode *inode, struct nfs4_deviceid *dev_id, gfp_t gfp_flags);
-void nfs4_ds_disconnect(struct nfs_client *clp);
 
 #endif /* FS_NFS_NFS4FILELAYOUT_H */
diff --git a/fs/nfs/nfs4filelayoutdev.c b/fs/nfs/nfs4filelayoutdev.c
index 3336d5e..a8eaa9b 100644
--- a/fs/nfs/nfs4filelayoutdev.c
+++ b/fs/nfs/nfs4filelayoutdev.c
@@ -149,28 +149,6 @@ _data_server_lookup_locked(const struct list_head *dsaddrs)
 }
 
 /*
- * Lookup DS by nfs_client pointer. Zero data server client pointer
- */
-void nfs4_ds_disconnect(struct nfs_client *clp)
-{
-	struct nfs4_pnfs_ds *ds;
-	struct nfs_client *found = NULL;
-
-	dprintk("%s clp %p\n", __func__, clp);
-	spin_lock(&nfs4_ds_cache_lock);
-	list_for_each_entry(ds, &nfs4_data_server_cache, ds_node)
-		if (ds->ds_clp && ds->ds_clp == clp) {
-			found = ds->ds_clp;
-			ds->ds_clp = NULL;
-		}
-	spin_unlock(&nfs4_ds_cache_lock);
-	if (found) {
-		set_bit(NFS_CS_STOP_RENEW, &clp->cl_res_state);
-		nfs_put_client(clp);
-	}
-}
-
-/*
  * Create an rpc connection to the nfs4_pnfs_ds data server
  * Currently only supports IPv4 and IPv6 addresses
  */
-- 
1.7.11.7

Re: [PATCH] NFSv4.1: Kill nfs4_ds_disconnect()

[Andy Adamson]

On Thu, Oct 11, 2012 at 12:51 PM, Trond Myklebust
<Trond.Myklebust@netapp.com> wrote:
> There is nothing to prevent another thread from dereferencing ds->ds_clp
> during or after the call to nfs4_ds_disconnect(), and Oopsing due to the
> resulting NULL pointer.

Yes.

>
> Instead, we should just rely on filelayout_mark_devid_invalid() to keep
> us out of trouble by avoiding that deviceid.

Why do we want to keep the connection and the nfs_client? For connection retry?

-->Andy

>
> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
> ---
>  fs/nfs/nfs4filelayout.c    |  1 -
>  fs/nfs/nfs4filelayout.h    |  1 -
>  fs/nfs/nfs4filelayoutdev.c | 22 ----------------------
>  3 files changed, 24 deletions(-)
>
> diff --git a/fs/nfs/nfs4filelayout.c b/fs/nfs/nfs4filelayout.c
> index 52d8472..816c2d0 100644
> --- a/fs/nfs/nfs4filelayout.c
> +++ b/fs/nfs/nfs4filelayout.c
> @@ -207,7 +207,6 @@ static int filelayout_async_handle_error(struct rpc_task *task,
>                 clear_bit(NFS_INO_LAYOUTCOMMIT, &NFS_I(inode)->flags);
>                 _pnfs_return_layout(inode);
>                 rpc_wake_up(&tbl->slot_tbl_waitq);
> -               nfs4_ds_disconnect(clp);
>                 /* fall through */
>         default:
>  reset:
> diff --git a/fs/nfs/nfs4filelayout.h b/fs/nfs/nfs4filelayout.h
> index dca47d78..8c07241 100644
> --- a/fs/nfs/nfs4filelayout.h
> +++ b/fs/nfs/nfs4filelayout.h
> @@ -149,6 +149,5 @@ extern void nfs4_fl_put_deviceid(struct nfs4_file_layout_dsaddr *dsaddr);
>  extern void nfs4_fl_free_deviceid(struct nfs4_file_layout_dsaddr *dsaddr);
>  struct nfs4_file_layout_dsaddr *
>  filelayout_get_device_info(struct inode *inode, struct nfs4_deviceid *dev_id, gfp_t gfp_flags);
> -void nfs4_ds_disconnect(struct nfs_client *clp);
>
>  #endif /* FS_NFS_NFS4FILELAYOUT_H */
> diff --git a/fs/nfs/nfs4filelayoutdev.c b/fs/nfs/nfs4filelayoutdev.c
> index 3336d5e..a8eaa9b 100644
> --- a/fs/nfs/nfs4filelayoutdev.c
> +++ b/fs/nfs/nfs4filelayoutdev.c
> @@ -149,28 +149,6 @@ _data_server_lookup_locked(const struct list_head *dsaddrs)
>  }
>
>  /*
> - * Lookup DS by nfs_client pointer. Zero data server client pointer
> - */
> -void nfs4_ds_disconnect(struct nfs_client *clp)
> -{
> -       struct nfs4_pnfs_ds *ds;
> -       struct nfs_client *found = NULL;
> -
> -       dprintk("%s clp %p\n", __func__, clp);
> -       spin_lock(&nfs4_ds_cache_lock);
> -       list_for_each_entry(ds, &nfs4_data_server_cache, ds_node)
> -               if (ds->ds_clp && ds->ds_clp == clp) {
> -                       found = ds->ds_clp;
> -                       ds->ds_clp = NULL;
> -               }
> -       spin_unlock(&nfs4_ds_cache_lock);
> -       if (found) {
> -               set_bit(NFS_CS_STOP_RENEW, &clp->cl_res_state);
> -               nfs_put_client(clp);
> -       }
> -}
> -
> -/*
>   * Create an rpc connection to the nfs4_pnfs_ds data server
>   * Currently only supports IPv4 and IPv6 addresses
>   */
> --
> 1.7.11.7
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

RE: [PATCH] NFSv4.1: Kill nfs4_ds_disconnect()

[Myklebust, Trond]

> -----Original Message-----
> From: Andy Adamson [mailto:androsadamson@gmail.com]
> Sent: Thursday, October 11, 2012 1:51 PM
> To: Myklebust, Trond
> Cc: linux-nfs@vger.kernel.org
> Subject: Re: [PATCH] NFSv4.1: Kill nfs4_ds_disconnect()
> 
> On Thu, Oct 11, 2012 at 12:51 PM, Trond Myklebust
> <Trond.Myklebust@netapp.com> wrote:
> > There is nothing to prevent another thread from dereferencing
> > ds->ds_clp during or after the call to nfs4_ds_disconnect(), and
> > Oopsing due to the resulting NULL pointer.
> 
> Yes.
> 
> >
> > Instead, we should just rely on filelayout_mark_devid_invalid() to
> > keep us out of trouble by avoiding that deviceid.
> 
> Why do we want to keep the connection and the nfs_client? For connection
> retry?
> 

The deviceid doesn't really want to keep the nfs_client, but I don't see how you can remove it safely in the current code. There is too much that relies on being able to dereference ds->ds_clp without any locking.

As for the TCP connection: that is a property of the nfs_client and its RPC client, and as such should not be manipulated by pNFS. If there are other users, then they may be waiting for it to reconnect...