NFSD: Clean up legacy NFS WRITE argument XDR decoders

Move common code in NFSD's legacy NFS WRITE decoders into a helper.
The immediate benefit is reduction of code duplication and some nice
micro-optimizations (see below).

In the long term, this helper can perform a per-transport call-out
to fill the rq_vec (say, using RDMA Reads).

The legacy WRITE decoders and procs are changed to work like NFSv4,
which constructs the rq_vec just before it is about to call
vfs_writev.

Why? Calling a transport call-out from the proc instead of the XDR
decoder means that the incoming FH can be resolved to a particular
filesystem and file. This would allow pages from the backing file to
be presented to the transport to be filled, rather than presenting
anonymous pages and copying or flipping them into the file's page
cache later.

I also prefer using the pages in rq_arg.pages, instead of pulling
the data pages directly out of the rqstp::rq_pages array. This is
currently the way the NFSv3 write decoder works, but the other two
do not seem to take this approach. Fixing this removes the only
reference to rq_pages found in NFSD, eliminating an NFSD assumption
about how transports use the pages in rq_pages.

Lastly, avoid setting up the first element of rq_vec as a zero-
length buffer. This happens with an RDMA transport when a normal
Read chunk is present because the data payload is in rq_arg's
page list (none of it is in the head buffer).

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

8 files changed, 65 insertions, 30 deletions

diff --git a/fs/nfsd/nfs3proc.c b/fs/nfsd/nfs3proc.c
index 1d0ce3c57d93..2dd95ebf4935 100644
--- a/fs/nfsd/nfs3proc.c
+++ b/fs/nfsd/nfs3proc.c
@@ -192,6 +192,7 @@ nfsd3_proc_write(struct svc_rqst *rqstp)
 	struct nfsd3_writeres *resp = rqstp->rq_resp;
 	__be32	nfserr;
 	unsigned long cnt = argp->len;
+	unsigned int nvecs;
 
 	dprintk("nfsd: WRITE(3)    %s %d bytes at %Lu%s\n",
 				SVCFH_fmt(&argp->fh),
@@ -201,9 +202,12 @@ nfsd3_proc_write(struct svc_rqst *rqstp)
 
 	fh_copy(&resp->fh, &argp->fh);
 	resp->committed = argp->stable;
+	nvecs = svc_fill_write_vector(rqstp, &argp->first, cnt);
+	if (!nvecs)
+		RETURN_STATUS(nfserr_io);
 	nfserr = nfsd_write(rqstp, &resp->fh, argp->offset,
-				rqstp->rq_vec, argp->vlen,
-				&cnt, resp->committed);
+			    rqstp->rq_vec, nvecs, &cnt,
+			    resp->committed);
 	resp->count = cnt;
 	RETURN_STATUS(nfserr);
 }
diff --git a/fs/nfsd/nfs3xdr.c b/fs/nfsd/nfs3xdr.c
index 1a70581e1cb2..e19fc5d8bcb5 100644
--- a/fs/nfsd/nfs3xdr.c
+++ b/fs/nfsd/nfs3xdr.c
@@ -391,7 +391,7 @@ int
 nfs3svc_decode_writeargs(struct svc_rqst *rqstp, __be32 *p)
 {
 	struct nfsd3_writeargs *args = rqstp->rq_argp;
-	unsigned int len, v, hdr, dlen;
+	unsigned int len, hdr, dlen;
 	u32 max_blocksize = svc_max_payload(rqstp);
 	struct kvec *head = rqstp->rq_arg.head;
 	struct kvec *tail = rqstp->rq_arg.tail;
@@ -433,17 +433,9 @@ nfs3svc_decode_writeargs(struct svc_rqst *rqstp, __be32 *p)
 		args->count = max_blocksize;
 		len = args->len = max_blocksize;
 	}
-	rqstp->rq_vec[0].iov_base = (void*)p;
-	rqstp->rq_vec[0].iov_len = head->iov_len - hdr;
-	v = 0;
-	while (len > rqstp->rq_vec[v].iov_len) {
-		len -= rqstp->rq_vec[v].iov_len;
-		v++;
-		rqstp->rq_vec[v].iov_base = page_address(rqstp->rq_pages[v]);
-		rqstp->rq_vec[v].iov_len = PAGE_SIZE;
-	}
-	rqstp->rq_vec[v].iov_len = len;
-	args->vlen = v + 1;
+
+	args->first.iov_base = (void *)p;
+	args->first.iov_len = head->iov_len - hdr;
 	return 1;
 }
 
diff --git a/fs/nfsd/nfsproc.c b/fs/nfsd/nfsproc.c
index 43c0419b8ddb..1995ea6bfd2b 100644
--- a/fs/nfsd/nfsproc.c
+++ b/fs/nfsd/nfsproc.c
@@ -212,13 +212,18 @@ nfsd_proc_write(struct svc_rqst *rqstp)
 	struct nfsd_attrstat *resp = rqstp->rq_resp;
 	__be32	nfserr;
 	unsigned long cnt = argp->len;
+	unsigned int nvecs;
 
 	dprintk("nfsd: WRITE    %s %d bytes at %d\n",
 		SVCFH_fmt(&argp->fh),
 		argp->len, argp->offset);
 
-	nfserr = nfsd_write(rqstp, fh_copy(&resp->fh, &argp->fh), argp->offset,
-				rqstp->rq_vec, argp->vlen, &cnt, NFS_DATA_SYNC);
+	nvecs = svc_fill_write_vector(rqstp, &argp->first, cnt);
+	if (!nvecs)
+		return nfserr_io;
+	nfserr = nfsd_write(rqstp, fh_copy(&resp->fh, &argp->fh),
+			    argp->offset, rqstp->rq_vec, nvecs,
+			    &cnt, NFS_DATA_SYNC);
 	return nfsd_return_attrs(nfserr, resp);
 }
 
diff --git a/fs/nfsd/nfsxdr.c b/fs/nfsd/nfsxdr.c
index 79b6064f8977..db24ae8b67e0 100644
--- a/fs/nfsd/nfsxdr.c
+++ b/fs/nfsd/nfsxdr.c
@@ -287,7 +287,6 @@ nfssvc_decode_writeargs(struct svc_rqst *rqstp, __be32 *p)
 	struct nfsd_writeargs *args = rqstp->rq_argp;
 	unsigned int len, hdr, dlen;
 	struct kvec *head = rqstp->rq_arg.head;
-	int v;
 
 	p = decode_fh(p, &args->fh);
 	if (!p)
@@ -323,17 +322,8 @@ nfssvc_decode_writeargs(struct svc_rqst *rqstp, __be32 *p)
 	if (dlen < XDR_QUADLEN(len)*4)
 		return 0;
 
-	rqstp->rq_vec[0].iov_base = (void*)p;
-	rqstp->rq_vec[0].iov_len = head->iov_len - hdr;
-	v = 0;
-	while (len > rqstp->rq_vec[v].iov_len) {
-		len -= rqstp->rq_vec[v].iov_len;
-		v++;
-		rqstp->rq_vec[v].iov_base = page_address(rqstp->rq_pages[v]);
-		rqstp->rq_vec[v].iov_len = PAGE_SIZE;
-	}
-	rqstp->rq_vec[v].iov_len = len;
-	args->vlen = v + 1;
+	args->first.iov_base = (void *)p;
+	args->first.iov_len = head->iov_len - hdr;
 	return 1;
 }
 
diff --git a/fs/nfsd/xdr.h b/fs/nfsd/xdr.h
index 2f4f22e6b8cb..a765c414015e 100644
--- a/fs/nfsd/xdr.h
+++ b/fs/nfsd/xdr.h
@@ -34,7 +34,7 @@ struct nfsd_writeargs {
 	svc_fh			fh;
 	__u32			offset;
 	int			len;
-	int			vlen;
+	struct kvec		first;
 };
 
 struct nfsd_createargs {
diff --git a/fs/nfsd/xdr3.h b/fs/nfsd/xdr3.h
index 056bf8a7364e..deccf7f691e9 100644
--- a/fs/nfsd/xdr3.h
+++ b/fs/nfsd/xdr3.h
@@ -41,7 +41,7 @@ struct nfsd3_writeargs {
 	__u32			count;
 	int			stable;
 	__u32			len;
-	int			vlen;
+	struct kvec		first;
 };
 
 struct nfsd3_createargs {
diff --git a/include/linux/sunrpc/svc.h b/include/linux/sunrpc/svc.h
index dc4c009deec1..fb3fcacc1e98 100644
--- a/include/linux/sunrpc/svc.h
+++ b/include/linux/sunrpc/svc.h
@@ -495,6 +495,8 @@ void		   svc_wake_up(struct svc_serv *);
 void		   svc_reserve(struct svc_rqst *rqstp, int space);
 struct svc_pool *  svc_pool_for_cpu(struct svc_serv *serv, int cpu);
 char *		   svc_print_addr(struct svc_rqst *, char *, size_t);
+unsigned int	   svc_fill_write_vector(struct svc_rqst *rqstp,
+					 struct kvec *first, size_t total);
 
 #define	RPC_MAX_ADDRBUFLEN	(63U)
 
diff --git a/net/sunrpc/svc.c b/net/sunrpc/svc.c
index f19987f5d3b5..a155e2de19aa 100644
--- a/net/sunrpc/svc.c
+++ b/net/sunrpc/svc.c
@@ -1533,3 +1533,45 @@ u32 svc_max_payload(const struct svc_rqst *rqstp)
 	return max;
 }
 EXPORT_SYMBOL_GPL(svc_max_payload);
+
+/**
+ * svc_fill_write_vector - Construct data argument for VFS write call
+ * @rqstp: svc_rqst to operate on
+ * @first: buffer containing first section of write payload
+ * @total: total number of bytes of write payload
+ *
+ * Returns the number of elements populated in the data argument array.
+ */
+unsigned int svc_fill_write_vector(struct svc_rqst *rqstp, struct kvec *first,
+				   size_t total)
+{
+	struct kvec *vec = rqstp->rq_vec;
+	struct page **pages;
+	unsigned int i;
+
+	/* Some types of transport can present the write payload
+	 * entirely in rq_arg.pages. In this case, @first is empty.
+	 */
+	i = 0;
+	if (first->iov_len) {
+		vec[i].iov_base = first->iov_base;
+		vec[i].iov_len = min_t(size_t, total, first->iov_len);
+		total -= vec[i].iov_len;
+		++i;
+	}
+
+	WARN_ON_ONCE(rqstp->rq_arg.page_base != 0);
+	pages = rqstp->rq_arg.pages;
+	while (total) {
+		vec[i].iov_base = page_address(*pages);
+		vec[i].iov_len = min_t(size_t, total, PAGE_SIZE);
+		total -= vec[i].iov_len;
+		++i;
+
+		++pages;
+	}
+
+	WARN_ON_ONCE(i > ARRAY_SIZE(rqstp->rq_vec));
+	return i;
+}
+EXPORT_SYMBOL_GPL(svc_fill_write_vector);