[LU-20037] (osc_page.c:183:osc_page_delete()) LBUG: do not return error from mdc_io_fsync_start in case of CL_FSYNC_DISCARD

• Type: Bug
• Resolution: Unresolved
• Priority: Medium
• Severity: 3
• Assignee: WC Triage
• Reporter: Vladimir Saveliev

Description

"LU-16043 osc: allow error for write on CL_FSYNC_DISCARD" makes osc_io_fsync_start() to not error in case of CL_FSYNC_DISCARD:

@@ -967,6 +967,13 @@ int osc_io_fsync_start(const struct lu_env *env,

        result = osc_cache_writeback_range(env, osc, start, end, 0,
                                           fio->fi_mode == CL_FSYNC_DISCARD);
+       if (result < 0 && fio->fi_mode == CL_FSYNC_DISCARD) {
+               CDEBUG(D_CACHE,
+                      "%s: ignore error %d on discarding "DFID":[%lu-%lu]\n",
+                      cli_name(osc_cli(osc)), result, PFID(fio->fi_fid),
+                      start, end);
+               result = 0;
+       }
        if (result > 0) {
                fio->fi_nr_written += result;
                result = 0;

The same should be done in mdc_io_fsync_start():

static int mdc_io_fsync_start(const struct lu_env *env,
                              const struct cl_io_slice *slice)
..
        result = osc_cache_writeback_range(env, osc, 0, CL_PAGE_EOF, 0,
                                           fio->fi_mode == CL_FSYNC_DISCARD);
        if (result > 0) {
                fio->fi_nr_written += result;
                result = 0;
        }

If mdc_io_fsync_start() returned an error running for ll_delete_inode()->cl_sync_file_range(), that would cause the LBUG in osc_page_delete():

[ 1647.787680] LustreError: lustre-MDT0000-mdc-ffff97980c315800: operation ost_write to node 192.168.176.171@tcp failed: rc = -3
..
[ 1647.831003] LustreError: 14269:0:(osc_page.c:144:osc_page_delete()) Trying to teardown failed: -16
[ 1647.834241] LustreError: 14269:0:(osc_page.c:146:osc_page_delete()) ASSERTION( 0 ) failed: