[PATCH 1/3] ceph: report the requested length in the netfs read metrics

[Xiubo Li via B4 Relay]

From: Xiubo Li <xiubo.li@xxxxxxxxx>

finish_netfs_read() passes osd_data->length to
ceph_update_read_metrics(), but a buffered read of an unencrypted
file attaches its data as CEPH_OSD_DATA_TYPE_ITER, and
ceph_osd_iter_init() only sets the type and the iov_iter.  length
shares a union with the iov_iter, so the read size metrics record
some unrelated iov_iter field, which is normally 0, for every page
cache read, and metrics/size is useless for buffered reads.

Report op->extent.length, the length the OSD read asked for, which is
set for every data type and equals osd_data->length for the page
vector that encrypted files still use.

Fixes: 4de77f25fd85 ("ceph: use osd_req_op_extent_osd_iter for netfs reads")
Reported-by: Dan van der Ster <dan.vanderster@xxxxxxxxx>
Signed-off-by: Xiubo Li <xiubo.li@xxxxxxxxx>
---
 fs/ceph/addr.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/ceph/addr.c b/fs/ceph/addr.c
index dd2bd8ed2424..8500d72702d6 100644
--- a/fs/ceph/addr.c
+++ b/fs/ceph/addr.c
@@ -236,7 +236,7 @@ static void finish_netfs_read(struct ceph_osd_request *req)
 	ceph_blog_enter_gfp(fsc, &__ji, GFP_ATOMIC);
 
 	ceph_update_read_metrics(&fsc->mdsc->metric, req->r_start_latency,
-				 req->r_end_latency, osd_data->length, err);
+				 req->r_end_latency, op->extent.length, err);
 
 	boutc(cl, "result %d subreq->len=%zu i_size=%lld\n", req->r_result,
 	      subreq->len, i_size_read(req->r_inode));

-- 
2.53.0