Fix serious bug in ZIL that can lead to pool corruption

in the case of a held dataset during remount.

Detailed description is available at:
https://www.illumos.org/issues/883

illumos-gate revision:	13380:161b964a0e10

Reviewed by:	pjd
Approved by:	re (kib)
Obtained from:	Illumos (Bug #883)
MFC after:	3 days

1 files changed, 35 insertions, 0 deletions

diff --git a/cddl/contrib/opensolaris/cmd/ztest/ztest.c b/cddl/contrib/opensolaris/cmd/ztest/ztest.c
index 62c98f4..d90961e 100644
--- a/cddl/contrib/opensolaris/cmd/ztest/ztest.c
+++ b/cddl/contrib/opensolaris/cmd/ztest/ztest.c
@@ -205,6 +205,7 @@ typedef struct ztest_od {
  */
 typedef struct ztest_ds {
 	objset_t	*zd_os;
+	rwlock_t	zd_zilog_lock;
 	zilog_t		*zd_zilog;
 	uint64_t	zd_seq;
 	ztest_od_t	*zd_od;		/* debugging aid */
@@ -238,6 +239,7 @@ ztest_func_t ztest_dmu_commit_callbacks;
 ztest_func_t ztest_zap;
 ztest_func_t ztest_zap_parallel;
 ztest_func_t ztest_zil_commit;
+ztest_func_t ztest_zil_remount;
 ztest_func_t ztest_dmu_read_write_zcopy;
 ztest_func_t ztest_dmu_objset_create_destroy;
 ztest_func_t ztest_dmu_prealloc;
@@ -273,6 +275,7 @@ ztest_info_t ztest_info[] = {
 	{ ztest_zap_parallel,			100,	&zopt_always	},
 	{ ztest_split_pool,			1,	&zopt_always	},
 	{ ztest_zil_commit,			1,	&zopt_incessant	},
+	{ ztest_zil_remount,			1,	&zopt_sometimes	},
 	{ ztest_dmu_read_write_zcopy,		1,	&zopt_often	},
 	{ ztest_dmu_objset_create_destroy,	1,	&zopt_often	},
 	{ ztest_dsl_prop_get_set,		1,	&zopt_often	},
@@ -986,6 +989,7 @@ ztest_zd_init(ztest_ds_t *zd, objset_t *os)
 	zd->zd_seq = 0;
 	dmu_objset_name(os, zd->zd_name);
 
+	VERIFY(rwlock_init(&zd->zd_zilog_lock, USYNC_THREAD, NULL) == 0);
 	VERIFY(_mutex_init(&zd->zd_dirobj_lock, USYNC_THREAD, NULL) == 0);
 
 	for (int l = 0; l < ZTEST_OBJECT_LOCKS; l++)
@@ -1965,6 +1969,8 @@ ztest_io(ztest_ds_t *zd, uint64_t object, uint64_t offset)
 	if (ztest_random(2) == 0)
 		io_type = ZTEST_IO_WRITE_TAG;
 
+	(void) rw_rdlock(&zd->zd_zilog_lock);
+
 	switch (io_type) {
 
 	case ZTEST_IO_WRITE_TAG:
@@ -2000,6 +2006,8 @@ ztest_io(ztest_ds_t *zd, uint64_t object, uint64_t offset)
 		break;
 	}
 
+	(void) rw_unlock(&zd->zd_zilog_lock);
+
 	umem_free(data, blocksize);
 }
 
@@ -2054,6 +2062,8 @@ ztest_zil_commit(ztest_ds_t *zd, uint64_t id)
 {
 	zilog_t *zilog = zd->zd_zilog;
 
+	(void) rw_rdlock(&zd->zd_zilog_lock);
+
 	zil_commit(zilog, ztest_random(ZTEST_OBJECTS));
 
 	/*
@@ -2065,6 +2075,31 @@ ztest_zil_commit(ztest_ds_t *zd, uint64_t id)
 	ASSERT(zd->zd_seq <= zilog->zl_commit_lr_seq);
 	zd->zd_seq = zilog->zl_commit_lr_seq;
 	mutex_exit(&zilog->zl_lock);
+
+	(void) rw_unlock(&zd->zd_zilog_lock);
+}
+
+/*
+ * This function is designed to simulate the operations that occur during a
+ * mount/unmount operation.  We hold the dataset across these operations in an
+ * attempt to expose any implicit assumptions about ZIL management.
+ */
+/* ARGSUSED */
+void
+ztest_zil_remount(ztest_ds_t *zd, uint64_t id)
+{
+	objset_t *os = zd->zd_os;
+
+	(void) rw_wrlock(&zd->zd_zilog_lock);
+
+	/* zfsvfs_teardown() */
+	zil_close(zd->zd_zilog);
+
+	/* zfsvfs_setup() */
+	VERIFY(zil_open(os, ztest_get_data) == zd->zd_zilog);
+	zil_replay(os, zd, ztest_replay_vector);
+
+	(void) rw_unlock(&zd->zd_zilog_lock);
 }
 
 /*