Re: [patch 21/26] mount options: partially fix nfs

[Chuck Lever <chuck.lever@oracle.com>]

On Jan 25, 2008, at 4:39 AM, Miklos Szeredi wrote:
>> Miklos Szeredi wrote:
>>> From: Miklos Szeredi <mszeredi@suse.cz>
>>>
>>> Add posix, bsize=, namelen= options to /proc/mounts for nfs
>>> filesystems.
>>>
>>> Document several other options that are still missing.
>>
>> NFS lists only some options in /proc/mounts on purpose: only the
>> essential options are mentioned there to keep clutter down.  The  
>> three
>> you've added here are for all intents and purposes deprecated,  
>> which is
>> why they are not supported.
>>
>> NFS lists a more complete set of mount options for a mount point in
>> /proc/self/mountstats.  See nfs_show_stats().
>>
>> Since your cover letter does not explain why you are changing this  
>> code,
>> can you refer me to a description of why you are doing this?
>
> Descritption is in the 01/26 patch.
>
>> More below.
>>
>>> Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
>>> ---
>>>
>>> Index: linux/fs/nfs/super.c
>>> ===================================================================
>>> --- linux.orig/fs/nfs/super.c	2008-01-19 11:56:34.000000000 +0100
>>> +++ linux/fs/nfs/super.c	2008-01-21 20:41:30.000000000 +0100
>>> @@ -449,6 +449,7 @@ static void nfs_show_mount_options(struc
>>>  	} nfs_info[] = {
>>>  		{ NFS_MOUNT_SOFT, ",soft", ",hard" },
>>>  		{ NFS_MOUNT_INTR, ",intr", ",nointr" },
>>> +		{ NFS_MOUNT_POSIX, ",posix", "" },
>>>  		{ NFS_MOUNT_NOCTO, ",nocto", "" },
>>>  		{ NFS_MOUNT_NOAC, ",noac", "" },
>>>  		{ NFS_MOUNT_NONLM, ",nolock", "" },
>>> @@ -459,10 +460,17 @@ static void nfs_show_mount_options(struc
>>>  	};
>>>  	const struct proc_nfs_info *nfs_infop;
>>>  	struct nfs_client *clp = nfss->nfs_client;
>>> +	unsigned int default_namelen =
>>> +		clp->rpc_ops->version == 4 ? NFS4_MAXNAMLEN :
>>> +		clp->rpc_ops->version == 3 ? NFS3_MAXNAMLEN : NFS2_MAXNAMLEN;
>>>
>>>  	seq_printf(m, ",vers=%d", clp->rpc_ops->version);
>>>  	seq_printf(m, ",rsize=%d", nfss->rsize);
>>>  	seq_printf(m, ",wsize=%d", nfss->wsize);
>>> +	if (nfss->bsize != 0)
>>> +		seq_printf(m, ",bsize=%d", nfss->bsize);
>>> +	if (nfss->namelen != default_namelen)
>>> +		seq_printf(m, ",namelen=%d", nfss->namelen);
>>>  	if (nfss->acregmin != 3*HZ || showdefaults)
>>>  		seq_printf(m, ",acregmin=%d", nfss->acregmin/HZ);
>>>  	if (nfss->acregmax != 60*HZ || showdefaults)
>>> @@ -482,6 +490,18 @@ static void nfs_show_mount_options(struc
>>>  	seq_printf(m, ",timeo=%lu", 10U * nfss->client->cl_timeout- 
>>> >to_initval / HZ);
>>>  	seq_printf(m, ",retrans=%u", nfss->client->cl_timeout- 
>>> >to_retries);
>>>  	seq_printf(m, ",sec=%s", nfs_pseudoflavour_to_name(nfss->client- 
>>> >cl_auth->au_flavor));
>>> +
>>> +	/*
>>> +	 * Missing options:
>>> +	 * port=
>>
>> Probably should be supported.
>>
>>> +	 * addr=
>>
>> This one is already supported; see nfs_show_options().
>
> Right, thanks.
>
>>
>>> +	 * clientaddr=
>>
>> This one isn't, and should be... would be useful for tracking down
>> certain NFSv4 problems.
>>
>>> +	 * mounthost=
>>> +	 * mountaddr=
>>> +	 * mountport=
>>> +	 * mountvers=
>>> +	 * mountproto=
>>
>> And these mount* options are for the kernel's new mount protocol  
>> client.
>>   They aren't really useful for understanding steady-state NFS client
>> behavior, they only effect mount-time behavior.
>
> All mount options should be shown, which are needed to reconstruct a
> previous mount.

Ah, OK.

I'm happy to implement logic to display the all missing options.  I  
should have updated nfs_show_mount_options() when I wrote the NFS  
mount option parser.

Let me know your preference.

> For example, if you copy options out from /proc/mount, umount the
> filesystem, and then create a new mount with the copied options, you
> should get the same mount.

For NFS, umount also needs to read some of the options in order to  
determine how mountd is to connect to the server for the unmount.   
(That's why we have addr= in the first place).

--
Chuck Lever
chuck[dot]lever[at]oracle[dot]com