staging/lustre/llite: deadlock for page write

Writing thread already locked page #1, and then wait for the
Writeback bit of page #2;

Ptlrpc thread is composing a write RPC, so it sets Writeback on
page #2 and tries to lock page #1 to make it ready.

Deadlocked.

Signed-off-by: Jinshan Xiong <jinshan.xiong@intel.com>
Reviewed-on: http://review.whamcloud.com/9036
Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-4540
Reviewed-by: wangdi <di.wang@intel.com>
Signed-off-by: Oleg Drokin <green@linuxhacker.ru>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

diff --git a/drivers/staging/lustre/lustre/llite/rw26.c b/drivers/staging/lustre/lustre/llite/rw26.c
index 50d8289afe06b5f86b49788a31eaa6fcf046f95a..f87238b3373cad82aacfc3ea5023c821bb8113d8 100644 (file)
--- a/drivers/staging/lustre/lustre/llite/rw26.c
+++ b/drivers/staging/lustre/lustre/llite/rw26.c
@@ -512,7 +512,7 @@ static int ll_write_begin(struct file *file, struct address_space *mapping,
 
        /* To avoid deadlock, try to lock page first. */
        vmpage = grab_cache_page_nowait(mapping, index);
-       if (unlikely(!vmpage || PageDirty(vmpage))) {
+       if (unlikely(!vmpage || PageDirty(vmpage) || PageWriteback(vmpage))) {
                struct ccc_io *cio = ccc_env_io(env);
                struct cl_page_list *plist = &cio->u.write.cui_queue;
 
@@ -522,7 +522,7 @@ static int ll_write_begin(struct file *file, struct address_space *mapping,
                 * more grants. It's okay for the dirty page to be the first
                 * one in commit page list, though.
                 */
-               if (vmpage && PageDirty(vmpage) && plist->pl_nr > 0) {
+               if (vmpage && plist->pl_nr > 0) {
                        unlock_page(vmpage);
                        page_cache_release(vmpage);
                        vmpage = NULL;