Skip to contentSkip to conversation

Make NFSD use the vfs_iocb_iter APIs

RFC PATCH In review Chuck Lever 4 participants 7 messages Jun 13, 2025 – Jun 16, 2025

Subsystems (2)
lore
Share & tools
lore

Chuck Lever

author maintainer
#0 lore raw

v2 cover Make NFSD use the vfs_iocb_iter APIs

From: Chuck Lever <[email protected]>

Mike has expressed interest in making NFSD perform direct, uncached, or asynchronous I/O, independent of how the target file might have been opened by the file cache. To do that, the idea is to pass in RWF_ flags during each VFS read and write.

However, Christoph suggested APIs that already exist which streamline the I/O operation a bit and expose the per-I/O flag setting directly. The suggestion looks to me like a straightforward and sensible general clean up of these code paths.

This series refactors nfsd_iter_read() and nfsd_vfs_write() to use those APIs instead of vfs_iter_read() and vfs_iter_write(), respectively, as a first baby step down this path. No behavior change is expected.

Chuck Lever (2):
NFSD: Use vfs_iocb_iter_read()
NFSD: Use vfs_iocb_iter_write()

 fs/nfsd/vfs.c | 19 ++++++++++++-------
 1 file changed, 12 insertions(+), 7 deletions(-)
Signature · 1 line
2.49.0
author maintainer
#1 lore raw

RFC PATCH v2 1/2 NFSD: Use vfs_iocb_iter_read()

From: Chuck Lever <[email protected]>

Refactor: Enable the use of IOCB flags to control NFSD's individual read operations (when not using splice). This allows the eventual use of atomic, uncached, direct, or asynchronous reads.

---
 fs/nfsd/vfs.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)
fs/nfsd/vfs.c +6 -2
index cd689df2ca5d..7b3bd141d54f 100644
@@ -1086,10 +1086,14 @@ __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);
+	kiocb.ki_pos = offset;
+	kiocb.ki_flags = 0;
+
 	v = 0;
 	total = *count;
 	while (total) {
@@ -1104,7 +1108,7 @@ __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);
 	return nfsd_finish_read(rqstp, fhp, file, offset, count, eof, host_err);
 }
 
Signature · 1 line
2.49.0
author maintainer
#3 lore raw

RFC PATCH v2 2/2 NFSD: Use vfs_iocb_iter_write()

In reply to Chuck Lever

From: Chuck Lever <[email protected]>

Refactor: Enable the use of IOCB flags to control NFSD's individual write operations. This allows the eventual use of atomic, uncached, direct, or asynchronous writes.

---
 fs/nfsd/vfs.c | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)
fs/nfsd/vfs.c +6 -5
index 7b3bd141d54f..66adeb6ffff5 100644
@@ -1174,15 +1174,14 @@ nfsd_vfs_write(struct svc_rqst *rqstp, struct svc_fh *fhp,
 	struct nfsd_net		*nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
 	struct file		*file = nf->nf_file;
 	struct super_block	*sb = file_inode(file)->i_sb;
+	struct kiocb		kiocb;
 	struct svc_export	*exp;
 	struct iov_iter		iter;
 	errseq_t		since;
 	__be32			nfserr;
 	int			host_err;
-	loff_t			pos = offset;
 	unsigned long		exp_op_flags = 0;
 	unsigned int		pflags = current->flags;
-	rwf_t			flags = 0;
 	bool			restore_flags = false;
 	unsigned int		nvecs;
 
@@ -1208,16 +1207,18 @@ nfsd_vfs_write(struct svc_rqst *rqstp, struct svc_fh *fhp,
 
 	if (!EX_ISSYNC(exp))
 		stable = NFS_UNSTABLE;
-
+	init_sync_kiocb(&kiocb, file);
+	kiocb.ki_pos = offset;
+	kiocb.ki_flags = 0;
 	if (stable && !fhp->fh_use_wgather)
-		flags |= RWF_SYNC;
+		kiocb.ki_flags |= IOCB_DSYNC;
 
 	nvecs = xdr_buf_to_bvec(rqstp->rq_bvec, rqstp->rq_maxpages, payload);
 	iov_iter_bvec(&iter, ITER_SOURCE, rqstp->rq_bvec, nvecs, *cnt);
 	since = READ_ONCE(file->f_wb_err);
 	if (verf)
 		nfsd_copy_write_verifier(verf, nn);
-	host_err = vfs_iter_write(file, &iter, &pos, flags);
+	host_err = vfs_iocb_iter_write(file, &kiocb, &iter);
 	if (host_err < 0) {
 		commit_reset_write_verifier(nn, rqstp, host_err);
 		goto out_nfserr;
Signature · 1 line
2.49.0
maintainer Reviewed-by on the v2 series
#5 lore raw
In reply to Chuck Lever
Quoted 17 lines · Chuck Lever

From: Chuck Lever <[email protected]>

Mike has expressed interest in making NFSD perform direct, uncached, or asynchronous I/O, independent of how the target file might have been opened by the file cache. To do that, the idea is to pass in RWF_ flags during each VFS read and write.

However, Christoph suggested APIs that already exist which streamline the I/O operation a bit and expose the per-I/O flag setting directly. The suggestion looks to me like a straightforward and sensible general clean up of these code paths.

This series refactors nfsd_iter_read() and nfsd_vfs_write() to use those APIs instead of vfs_iter_read() and vfs_iter_write(), respectively, as a first baby step down this path. No behavior change is expected.

Quoted 6 lines

Chuck Lever (2):
NFSD: Use vfs_iocb_iter_read()
NFSD: Use vfs_iocb_iter_write()

 fs/nfsd/vfs.c | 19 ++++++++++++-------
 1 file changed, 12 insertions(+), 7 deletions(-)

Looks like a straightforward change to me. Why the RFC?

reviewer Reviewed-by on the v2 series
#6 lore raw
In reply to Chuck Lever
Quoted 12 lines · Chuck Lever

From: Chuck Lever <[email protected]>

Mike has expressed interest in making NFSD perform direct, uncached, or asynchronous I/O, independent of how the target file might have been opened by the file cache. To do that, the idea is to pass in RWF_ flags during each VFS read and write.

However, Christoph suggested APIs that already exist which streamline the I/O operation a bit and expose the per-I/O flag setting directly. The suggestion looks to me like a straightforward and sensible general clean up of these code paths.

Quoted 8 lines

This series refactors nfsd_iter_read() and nfsd_vfs_write() to use those APIs instead of vfs_iter_read() and vfs_iter_write(), respectively, as a first baby step down this path. No behavior change is expected.

Chuck Lever (2):
NFSD: Use vfs_iocb_iter_read()
NFSD: Use vfs_iocb_iter_write()

It important part of this change (as I see it) is that we no longer have to go through kiocb_set_rw_flags() (which vfs_iter_read/vfs_iter_write do) but can set the IOCB_ flags directly. Eminently sensible.

Thanks,
NeilBrown

Quoted 8 lines
 fs/nfsd/vfs.c | 19 ++++++++++++-------
 1 file changed, 12 insertions(+), 7 deletions(-)
Signature · 1 line
2.49.0
Reviewed-by on v2 1/2
#2 lore raw

Re: [RFC PATCH v2 1/2] NFSD: Use vfs_iocb_iter_read()

In reply to Chuck Lever

fs/nfsd/vfs.c __be32 nfsd_iter_read(struct svc_rqst *rqstp, struct svc_fh *fhp, · line 1095

Quoted 3 lines · Chuck Lever
fs/nfsd/vfs.c +3 -0
+	init_sync_kiocb(&kiocb, file);
+	kiocb.ki_pos = offset;
+	kiocb.ki_flags = 0;

ki_flags is already cleared by init_sync_kiocb, so the line doing that can be dropped.

Otherwise looks good: