Re: [PATCH 2/2] dFrom: Max Kellermann <mk@cm4all.com>

From: Trond Myklebust <trondmy@primarydata.com>
To: "hch@infradead.org" <hch@infradead.org>,
	"mk@cm4all.com" <mk@cm4all.com>,
	"linux-nfs@vger.kernel.org" <linux-nfs@vger.kernel.org>,
	"gregkh@linuxfoundation.org" <gregkh@linuxfoundation.org>,
	"linux-fsdevel@vger.kernel.org" <linux-fsdevel@vger.kernel.org>
Cc: "max.kellermann@gmail.com" <max.kellermann@gmail.com>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH 2/2] dFrom: Max Kellermann <mk@cm4all.com>
Date: Tue, 16 Jan 2018 15:06:44 +0000	[thread overview]
Message-ID: <1516115201.5784.4.camel@primarydata.com> (raw)
In-Reply-To: <151603745169.29035.6866708909114015882.stgit@rabbit.intern.cm-ag>

On Mon, 2018-01-15 at 18:30 +0100, Max Kellermann wrote:
> nfs/super: set MS_POSIXACL only if ACL support is enabled
> 
> The code comment says "We will [apply the umask] ourselves", but that
> happens in posix_acl_create() only if the kernel has POSIX ACL
> support.  Without it, posix_acl_create() is a is an empty dummy
> function.
> 
> So let's not pretend we will apply the umask if we can already know
> that we will never.
> 
> This fixes a problem where the umask is always ignored in the NFS
> client when compiled without CONFIG_FS_POSIX_ACL.  This is a 4 year
> old regression caused by commit 013cdf1088d723 which itself was not
> completely wrong, but failed to consider all the side effects by
> misdesigned VFS code.
> 
> There are two compile-time checks and one runtime check:
> 
> - If CONFIG_FS_POSIX_ACL=n, then MS_POSIXACL is never set.
> 
> - If CONFIG_FS_POSIX_ACL=y and CONFIG_NFS_V3_ACL=n, then only NFSv4
>   has ACL support (and cannot be disabled), and we need to check for
>   "version==4".
> 
> - If CONFIG_FS_POSIX_ACL=y and CONFIG_NFS_V3_ACL=y, MS_POSIXACL is
>   always set, as before.
> 
> Signed-off-by: Max Kellermann <mk@cm4all.com>
> ---
>  fs/nfs/super.c |   15 +++++++++++----
>  1 file changed, 11 insertions(+), 4 deletions(-)
> 
> diff --git a/fs/nfs/super.c b/fs/nfs/super.c
> index 216f67d628b3..ec4e1f2775e0 100644
> --- a/fs/nfs/super.c
> +++ b/fs/nfs/super.c
> @@ -2338,10 +2338,17 @@ void nfs_fill_super(struct super_block *sb,
> struct nfs_mount_info *mount_info)
>  		sb->s_blocksize = nfs_block_size(data->bsize, &sb-
> >s_blocksize_bits);
>  
>  	if (server->nfs_client->rpc_ops->version != 2) {
> -		/* The VFS shouldn't apply the umask to mode bits.
> We will do
> -		 * so ourselves when necessary.
> -		 */
> -		sb->s_flags |= MS_POSIXACL;
> +#ifdef CONFIG_FS_POSIX_ACL
> +#ifndef CONFIG_NFS_V3_ACL
> +		if (nfss->nfs_client->rpc_ops->version == 4)
> +#endif
> +			/* The VFS shouldn't apply the umask to mode
> +			 * bits. We will do so ourselves when
> +			 * necessary.
> +			 */
> +			sb->s_flags |= MS_POSIXACL;
> +#endif
> +
>  		sb->s_time_gran = 1;
>  		sb->s_export_op = &nfs_export_ops;
>  	}

The above illustrates exactly why I've asked people _never_ to make
anything conditional on rpc_ops->version. Please use a NFS capability
(i.e. NFS_SB(sb)->caps) for this kind of thing. That expresses the
condition in terms of the functionality we want instead of a whimsical
protocol version number.

Thanks
  Trond
-- 
Trond Myklebust
Linux NFS client maintainer, PrimaryData
trond.myklebust@primarydata.com