Re: [PATCH v4 1/5] NFSD: filecache: add STATX_DIOALIGN and STATX_DIO_READ_ALIGN support

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

On 7/23/25 3:56 PM, Jeff Layton wrote:
> On Wed, 2025-07-23 at 15:43 -0400, Mike Snitzer wrote:
>> On Wed, Jul 23, 2025 at 03:06:00PM -0400, Mike Snitzer wrote:
>>> On Wed, Jul 23, 2025 at 02:58:19PM -0400, Jeff Layton wrote:
>>>> On Wed, 2025-07-23 at 11:43 -0400, Mike Snitzer wrote:
>>>>> Use STATX_DIOALIGN and STATX_DIO_READ_ALIGN to get and store DIO
>>>>> alignment attributes from underlying filesystem in associated
>>>>> nfsd_file.  This is done when the nfsd_file is first opened for
>>>>> a regular file.
>>>>>
>>>>> Signed-off-by: Mike Snitzer <snitzer@kernel.org>
>>>>> ---
>>>>>  fs/nfsd/filecache.c | 32 ++++++++++++++++++++++++++++++++
>>>>>  fs/nfsd/filecache.h |  4 ++++
>>>>>  fs/nfsd/nfsfh.c     |  4 ++++
>>>>>  3 files changed, 40 insertions(+)
>>>>>
>>>>> diff --git a/fs/nfsd/filecache.c b/fs/nfsd/filecache.c
>>>>> index 8581c131338b..5447dba6c5da 100644
>>>>> --- a/fs/nfsd/filecache.c
>>>>> +++ b/fs/nfsd/filecache.c
>>>>> @@ -231,6 +231,9 @@ nfsd_file_alloc(struct net *net, struct inode *inode, unsigned char need,
>>>>>  	refcount_set(&nf->nf_ref, 1);
>>>>>  	nf->nf_may = need;
>>>>>  	nf->nf_mark = NULL;
>>>>> +	nf->nf_dio_mem_align = 0;
>>>>> +	nf->nf_dio_offset_align = 0;
>>>>> +	nf->nf_dio_read_offset_align = 0;
>>>>>  	return nf;
>>>>>  }
>>>>>  
>>>>> @@ -1048,6 +1051,33 @@ nfsd_file_is_cached(struct inode *inode)
>>>>>  	return ret;
>>>>>  }
>>>>>  
>>>>> +static __be32
>>>>> +nfsd_file_getattr(const struct svc_fh *fhp, struct nfsd_file *nf)
>>>>> +{
>>>>> +	struct inode *inode = file_inode(nf->nf_file);
>>>>> +	struct kstat stat;
>>>>> +	__be32 status;
>>>>> +
>>>>> +	/* Currently only need to get DIO alignment info for regular files */
>>>>> +	if (!S_ISREG(inode->i_mode))
>>>>> +		return nfs_ok;
>>>>> +
>>>>> +	status = fh_getattr(fhp, &stat);
>>>>> +	if (status != nfs_ok)
>>>>> +		return status;
>>>>> +
>>>>> +	if (stat.result_mask & STATX_DIOALIGN) {
>>>>> +		nf->nf_dio_mem_align = stat.dio_mem_align;
>>>>> +		nf->nf_dio_offset_align = stat.dio_offset_align;
>>>>> +	}
>>>>> +	if (stat.result_mask & STATX_DIO_READ_ALIGN)
>>>>> +		nf->nf_dio_read_offset_align = stat.dio_read_offset_align;
>>>>> +	else
>>>>> +		nf->nf_dio_read_offset_align = nf->nf_dio_offset_align;
>>>>> +
>>>>> +	return status;
>>>>> +}
>>>>> +
>>>>>  static __be32
>>>>>  nfsd_file_do_acquire(struct svc_rqst *rqstp, struct net *net,
>>>>>  		     struct svc_cred *cred,
>>>>> @@ -1166,6 +1196,8 @@ nfsd_file_do_acquire(struct svc_rqst *rqstp, struct net *net,
>>>>>  			}
>>>>>  			status = nfserrno(ret);
>>>>>  			trace_nfsd_file_open(nf, status);
>>>>> +			if (status == nfs_ok)
>>>>> +				status = nfsd_file_getattr(fhp, nf);
>>>>
>>>>
>>>> Doing a getattr alongside every open could be expensive in some
>>>> configurations (like reexported NFS). We may want to skip doing this
>>>> getattr this if O_DIRECT isn't is use. Is that possible?
>>>
>>> Good point, yes, should be easy enough.  Will depend on the debugfs
>>> knobs, so will tack a patch on at the end.
>>
>> What is the best way to check for NFSD reexporting NFS?
> 
>> I've done stuff like that as a side-effect of setting/checking a
>> particular flag, e.g. commit 5cca2483b9fd ("nfsd: disallow file
>> locking and delegations for NFSv4 reexport")
>>
>> But not immediately seeing a more generic way to do the check...
>>
> 
> I don't think there is a reliable method, and we probably wouldn't want
> to just limit this to NFS. Other filesystems might have similar
> limitations (e.g. Ceph or Lustre).
> 
> I'd probably just base this on whether support is enabled in debugfs.
> If it is, then do the getattr. If that slows down reexports then we can
> figure out what to do then. If and when we get to converting this to an
> export option, we may need to do something more elaborate, but I
> wouldn't bother for now.

Fwiw, I would expect that re-exports would prefer to utilize the NFS
server's page cache.


-- 
Chuck Lever