On Sat, Jun 29, 2024 at 06:18:42PM -0400, Chuck Lever wrote:

quoted

+
+	/* nfs_fh -> svc_fh */
+	if (nfs_fh->size > NFS4_FHSIZE) {
+		status = -EINVAL;
+		goto out;
+	}
+	fh_init(&fh, NFS4_FHSIZE);
+	fh.fh_handle.fh_size = nfs_fh->size;
+	memcpy(fh.fh_handle.fh_raw, nfs_fh->data, nfs_fh->size);
+
+	if (fmode & FMODE_READ)
+		mayflags |= NFSD_MAY_READ;
+	if (fmode & FMODE_WRITE)
+		mayflags |= NFSD_MAY_WRITE;
+
+	beres = nfsd_file_acquire(rqstp, &fh, mayflags, &nf);
+	if (beres) {
+		status = nfs_stat_to_errno(be32_to_cpu(beres));
+		goto out_fh_put;
+	}

So I'm wondering whether just calling fh_verify() and then
nfsd_open_verified() would be simpler and/or good enough here. Is
there a strong reason to use the file cache for locally opened
files? Jeff, any thoughts?

Will there be writeback ramifications for
doing this? Maybe we need a comment here explaining how these files
are garbage collected (just an fput by the local I/O client, I would
guess).

OK, going back to this: Since right here we immediately call 

	nfsd_file_put(nf);

There are no writeback ramifications nor any need to comment about
garbage collection. But this still seems like a lot of (possibly
unnecessary) overhead for simply obtaining a struct file.

quoted

+
+	*pfilp = get_file(nf->nf_file);
+
+	nfsd_file_put(nf);
+out_fh_put:
+	fh_put(&fh);
+
+out:
+	nfsd_local_fakerqst_destroy(rqstp);
+out_revertcred:
+	revert_creds(save_cred);
+	return status;
+}
+EXPORT_SYMBOL_GPL(nfsd_open_local_fh);

-- 
Chuck Lever