NFS: More O_DIRECT accounting fixes for error paths
authorTrond Myklebust <trond.myklebust@hammerspace.com>
Mon, 4 Sep 2023 16:34:39 +0000 (12:34 -0400)
committerAnna Schumaker <Anna.Schumaker@Netapp.com>
Wed, 13 Sep 2023 15:51:10 +0000 (11:51 -0400)
If we hit a fatal error when retransmitting, we do need to record the
removal of the request from the count of written bytes.

Fixes: 031d73ed768a ("NFS: Fix O_DIRECT accounting of number of bytes read/written")
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
fs/nfs/direct.c

index e8a1645857dd6dee6ccfa7dc8908f1c3dd6b00b6..a53e501234993afd7e43505d9e1f8810e4bcfbf0 100644 (file)
@@ -93,12 +93,10 @@ nfs_direct_handle_truncated(struct nfs_direct_req *dreq,
                dreq->max_count = dreq_len;
                if (dreq->count > dreq_len)
                        dreq->count = dreq_len;
-
-               if (test_bit(NFS_IOHDR_ERROR, &hdr->flags))
-                       dreq->error = hdr->error;
-               else /* Clear outstanding error if this is EOF */
-                       dreq->error = 0;
        }
+
+       if (test_bit(NFS_IOHDR_ERROR, &hdr->flags) && !dreq->error)
+               dreq->error = hdr->error;
 }
 
 static void
@@ -120,6 +118,18 @@ nfs_direct_count_bytes(struct nfs_direct_req *dreq,
                dreq->count = dreq_len;
 }
 
+static void nfs_direct_truncate_request(struct nfs_direct_req *dreq,
+                                       struct nfs_page *req)
+{
+       loff_t offs = req_offset(req);
+       size_t req_start = (size_t)(offs - dreq->io_start);
+
+       if (req_start < dreq->max_count)
+               dreq->max_count = req_start;
+       if (req_start < dreq->count)
+               dreq->count = req_start;
+}
+
 /**
  * nfs_swap_rw - NFS address space operation for swap I/O
  * @iocb: target I/O control block
@@ -537,10 +547,6 @@ static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq)
 
        nfs_direct_join_group(&reqs, dreq->inode);
 
-       dreq->count = 0;
-       dreq->max_count = 0;
-       list_for_each_entry(req, &reqs, wb_list)
-               dreq->max_count += req->wb_bytes;
        nfs_clear_pnfs_ds_commit_verifiers(&dreq->ds_cinfo);
        get_dreq(dreq);
 
@@ -574,10 +580,14 @@ static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq)
                req = nfs_list_entry(reqs.next);
                nfs_list_remove_request(req);
                nfs_unlock_and_release_request(req);
-               if (desc.pg_error == -EAGAIN)
+               if (desc.pg_error == -EAGAIN) {
                        nfs_mark_request_commit(req, NULL, &cinfo, 0);
-               else
+               } else {
+                       spin_lock(&dreq->lock);
+                       nfs_direct_truncate_request(dreq, req);
+                       spin_unlock(&dreq->lock);
                        nfs_release_request(req);
+               }
        }
 
        if (put_dreq(dreq))
@@ -597,8 +607,6 @@ static void nfs_direct_commit_complete(struct nfs_commit_data *data)
        if (status < 0) {
                /* Errors in commit are fatal */
                dreq->error = status;
-               dreq->max_count = 0;
-               dreq->count = 0;
                dreq->flags = NFS_ODIRECT_DONE;
        } else {
                status = dreq->error;
@@ -609,7 +617,12 @@ static void nfs_direct_commit_complete(struct nfs_commit_data *data)
        while (!list_empty(&data->pages)) {
                req = nfs_list_entry(data->pages.next);
                nfs_list_remove_request(req);
-               if (status >= 0 && !nfs_write_match_verf(verf, req)) {
+               if (status < 0) {
+                       spin_lock(&dreq->lock);
+                       nfs_direct_truncate_request(dreq, req);
+                       spin_unlock(&dreq->lock);
+                       nfs_release_request(req);
+               } else if (!nfs_write_match_verf(verf, req)) {
                        dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
                        /*
                         * Despite the reboot, the write was successful,
@@ -617,7 +630,7 @@ static void nfs_direct_commit_complete(struct nfs_commit_data *data)
                         */
                        req->wb_nio = 0;
                        nfs_mark_request_commit(req, NULL, &cinfo, 0);
-               } else /* Error or match */
+               } else
                        nfs_release_request(req);
                nfs_unlock_and_release_request(req);
        }
@@ -670,6 +683,7 @@ static void nfs_direct_write_clear_reqs(struct nfs_direct_req *dreq)
        while (!list_empty(&reqs)) {
                req = nfs_list_entry(reqs.next);
                nfs_list_remove_request(req);
+               nfs_direct_truncate_request(dreq, req);
                nfs_release_request(req);
                nfs_unlock_and_release_request(req);
        }
@@ -719,7 +733,8 @@ static void nfs_direct_write_completion(struct nfs_pgio_header *hdr)
        }
 
        nfs_direct_count_bytes(dreq, hdr);
-       if (test_bit(NFS_IOHDR_UNSTABLE_WRITES, &hdr->flags)) {
+       if (test_bit(NFS_IOHDR_UNSTABLE_WRITES, &hdr->flags) &&
+           !test_bit(NFS_IOHDR_ERROR, &hdr->flags)) {
                if (!dreq->flags)
                        dreq->flags = NFS_ODIRECT_DO_COMMIT;
                flags = dreq->flags;