staging/lustre/llite: don't add to page cache upon failure
authorLai Siyao <lai.siyao@intel.com>
Mon, 2 Feb 2015 02:52:17 +0000 (21:52 -0500)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sat, 7 Feb 2015 09:32:34 +0000 (17:32 +0800)
Reading directory pages may fail on MDS, in this case client should
not cache a non-up-to-date directory page, because it will cause
a later read on the same page fail.

Signed-off-by: Lai Siyao <lai.siyao@intel.com>
Signed-off-by: Oleg Drokin <oleg.drokin@intel.com>
Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-5461
Reviewed-on: http://review.whamcloud.com/11450
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/staging/lustre/lustre/llite/dir.c

index 69a4a63..a182019 100644 (file)
@@ -183,7 +183,10 @@ static int ll_dir_filler(void *_hash, struct page *page0)
        op_data->op_offset = hash;
        rc = md_readpage(exp, op_data, page_pool, &request);
        ll_finish_md_op_data(op_data);
-       if (rc == 0) {
+       if (rc < 0) {
+               /* page0 is special, which was added into page cache early */
+               delete_from_page_cache(page0);
+       } else if (rc == 0) {
                body = req_capsule_server_get(&request->rq_pill, &RMF_MDT_BODY);
                /* Checked by mdc_readpage() */
                LASSERT(body != NULL);