[RFC PATCH] NFSD: Use vfs_iocb_iter_read()

[RFC PATCH] NFSD: Use vfs_iocb_iter_read()

[Chuck Lever]

From: Chuck Lever <chuck.lever@oracle.com>

Refactor: Enable the use of RWF_ flags to control individual I/O
operations.

Suggested-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 fs/nfsd/vfs.c | 13 +++++++++++--
 1 file changed, 11 insertions(+), 2 deletions(-)

diff --git a/fs/nfsd/vfs.c b/fs/nfsd/vfs.c
index cd689df2ca5d..f20bacd9b224 100644
--- a/fs/nfsd/vfs.c
+++ b/fs/nfsd/vfs.c
@@ -1086,10 +1086,18 @@ __be32 nfsd_iter_read(struct svc_rqst *rqstp, struct svc_fh *fhp,
 {
 	unsigned long v, total;
 	struct iov_iter iter;
-	loff_t ppos = offset;
+	struct kiocb kiocb;
 	ssize_t host_err;
 	size_t len;
 
+	init_sync_kiocb(&kiocb, file);
+	host_err = kiocb_set_rw_flags(&kiocb, 0, READ);
+	if (host_err) {
+		*count = 0;
+		goto out;
+	}
+	kiocb.ki_pos = offset;
+
 	v = 0;
 	total = *count;
 	while (total) {
@@ -1104,7 +1112,8 @@ __be32 nfsd_iter_read(struct svc_rqst *rqstp, struct svc_fh *fhp,
 
 	trace_nfsd_read_vector(rqstp, fhp, offset, *count);
 	iov_iter_bvec(&iter, ITER_DEST, rqstp->rq_bvec, v, *count);
-	host_err = vfs_iter_read(file, &iter, &ppos, 0);
+	host_err = vfs_iocb_iter_read(file, &kiocb, &iter);
+out:
 	return nfsd_finish_read(rqstp, fhp, file, offset, count, eof, host_err);
 }
 
-- 
2.49.0

Re: [RFC PATCH] NFSD: Use vfs_iocb_iter_read()

[NeilBrown]

On Fri, 13 Jun 2025, Chuck Lever wrote:
> From: Chuck Lever <chuck.lever@oracle.com>
> 
> Refactor: Enable the use of RWF_ flags to control individual I/O
> operations.

RWF_ flags such as .... ???

I'm guessing RWF_DIRECT and maybe RWF_NOWAIT based on the earlier
conversation.  Might be useful to provide this as explicit motivation. 

But seeing they will be passed to kiocb_set_rw_flags(), shouldn't we be
talking about the IOCB_ versions of the flags?  The only caller of
kiocb_set_rw_flags() I can find which doesn't pass zero is aio_prep_rw()
and it uses the IOCB_ versions.

And should we be changing the vfs_iter_write() to vfs_iocb_iter_write()
too, just for consistency?

> 
> Suggested-by: Christoph Hellwig <hch@infradead.org>
> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

Reviewed-by: NeilBrown <neil@brown.name>

3 more callers to go and vfs_iter_read() could be discarded.

NeilBrown


> ---
>  fs/nfsd/vfs.c | 13 +++++++++++--
>  1 file changed, 11 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/nfsd/vfs.c b/fs/nfsd/vfs.c
> index cd689df2ca5d..f20bacd9b224 100644
> --- a/fs/nfsd/vfs.c
> +++ b/fs/nfsd/vfs.c
> @@ -1086,10 +1086,18 @@ __be32 nfsd_iter_read(struct svc_rqst *rqstp, struct svc_fh *fhp,
>  {
>  	unsigned long v, total;
>  	struct iov_iter iter;
> -	loff_t ppos = offset;
> +	struct kiocb kiocb;
>  	ssize_t host_err;
>  	size_t len;
>  
> +	init_sync_kiocb(&kiocb, file);
> +	host_err = kiocb_set_rw_flags(&kiocb, 0, READ);
> +	if (host_err) {
> +		*count = 0;
> +		goto out;
> +	}
> +	kiocb.ki_pos = offset;
> +
>  	v = 0;
>  	total = *count;
>  	while (total) {
> @@ -1104,7 +1112,8 @@ __be32 nfsd_iter_read(struct svc_rqst *rqstp, struct svc_fh *fhp,
>  
>  	trace_nfsd_read_vector(rqstp, fhp, offset, *count);
>  	iov_iter_bvec(&iter, ITER_DEST, rqstp->rq_bvec, v, *count);
> -	host_err = vfs_iter_read(file, &iter, &ppos, 0);
> +	host_err = vfs_iocb_iter_read(file, &kiocb, &iter);
> +out:
>  	return nfsd_finish_read(rqstp, fhp, file, offset, count, eof, host_err);
>  }
>  
> -- 
> 2.49.0
> 
> 

Re: [RFC PATCH] NFSD: Use vfs_iocb_iter_read()

[Christoph Hellwig]

On Thu, Jun 12, 2025 at 08:36:53PM -0400, Chuck Lever wrote:
> From: Chuck Lever <chuck.lever@oracle.com>
> 
> Refactor: Enable the use of RWF_ flags to control individual I/O
> operations.

The IOCB_ ones, really.

> +	host_err = kiocb_set_rw_flags(&kiocb, 0, READ);
> +	if (host_err) {
> +		*count = 0;
> +		goto out;
> +	}

And if you don't want to explicitly convert from RWF_* flags, this
is pointless.  Just drop it for now and explicitly set the IOCB_
flags we want once we add some.