LU-17115 quota: fix race of acquiring locks in qmt

In qmt_delete_qid and qmt_reset_qid, the order to require
the lock of lquota_entry and journal is different from that
in qmt_dqacq0, which could cause deadlock in some cases.

Signed-off-by: Hongchao Zhang <hongchao@whamcloud.com>
Change-Id: Ic439f2c5d6ca22429422b87f0dde65e0d2e6113d
Reviewed-on: https://review.whamcloud.com/c/fs/lustre-release/+/52371
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Sergey Cheremencev <scherementsev@ddn.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

diff --git a/lustre/quota/qmt_handler.c b/lustre/quota/qmt_handler.c
index b25ee34..f6f830c 100644 (file)
--- a/lustre/quota/qmt_handler.c
+++ b/lustre/quota/qmt_handler.c
@@ -384,8 +384,6 @@ static int qmt_delete_qid(const struct lu_env *env, struct qmt_device *qmt,
        if (IS_ERR(lqe))
                RETURN(PTR_ERR(lqe));
 
-       lqe_write_lock(lqe);
-
        qpi = qmt_pool_lookup_glb(env, qmt, restype);
        if (IS_ERR(qpi))
                GOTO(out, rc = -ENOMEM);
@@ -394,9 +392,11 @@ static int qmt_delete_qid(const struct lu_env *env, struct qmt_device *qmt,
        if (IS_ERR(th))
                GOTO(out, rc = PTR_ERR(th));
 
+       lqe_write_lock(lqe);
        rc = lquota_disk_delete(env, th,
                                qpi->qpi_glb_obj[qtype], qid, &ver);
 
+       lqe_write_unlock(lqe);
        dt_trans_stop(env, qmt->qmt_child, th);
 
        if (rc == 0) {
@@ -410,7 +410,6 @@ out:
        if (!IS_ERR_OR_NULL(qpi))
                qpi_putref(env, qpi);
 
-       lqe_write_unlock(lqe);
        lqe_putref(lqe);
 
        RETURN(rc);
@@ -491,9 +490,7 @@ static int qmt_reset_qid(const struct lu_env *env, struct qmt_device *qmt,
        if (IS_ERR(lqe))
                RETURN(PTR_ERR(lqe));
 
-       lqe_write_lock(lqe);
-
-       qpi = qmt_pool_lookup_glb(env, qmt, restype);
+       qpi = lqe2qpi(lqe);
        if (IS_ERR(qpi))
                GOTO(out, rc = -ENOMEM);
 
@@ -501,6 +498,8 @@ static int qmt_reset_qid(const struct lu_env *env, struct qmt_device *qmt,
        if (IS_ERR(th))
                GOTO(out, rc = PTR_ERR(th));
 
+       lqe_write_lock(lqe);
+
        softlimit = lqe->lqe_softlimit;
        hardlimit = lqe->lqe_hardlimit;
 
@@ -513,8 +512,17 @@ static int qmt_reset_qid(const struct lu_env *env, struct qmt_device *qmt,
        lqe->lqe_is_deleted = 0;
        lqe->lqe_is_reset = 1;
        rc = qmt_glb_write(env, th, lqe, LQUOTA_BUMP_VER, &ver);
-       if (rc)
+       if (rc) {
                LQUOTA_ERROR(lqe, "failed to write quota global rec\n");
+
+               if (softlimit != 0)
+                       lqe->lqe_softlimit = softlimit;
+               if (hardlimit != 0)
+                       lqe->lqe_hardlimit = hardlimit;
+               lqe->lqe_is_reset = 0;
+       }
+
+       lqe_write_unlock(lqe);
        dt_trans_stop(env, qmt->qmt_child, th);
        if (rc)
                GOTO(out, rc);
@@ -527,18 +535,6 @@ static int qmt_reset_qid(const struct lu_env *env, struct qmt_device *qmt,
        qmt_glb_lock_notify(env, lqe, ver);
 
 out:
-       if (rc) {
-               if (softlimit != 0)
-                       lqe->lqe_softlimit = softlimit;
-               if (hardlimit != 0)
-                       lqe->lqe_hardlimit = hardlimit;
-               lqe->lqe_is_reset = 0;
-       }
-
-       if (!IS_ERR_OR_NULL(qpi))
-               qpi_putref(env, qpi);
-
-       lqe_write_unlock(lqe);
        lqe_putref(lqe);
 
        RETURN(rc);