Add zfs_iput_async() interface

Handle all iputs in zfs_purgedir() and zfs_inode_destroy() asynchronously to prevent deadlocks. When the iputs are allowed to run synchronously in the destroy call path deadlocks between xattr directory inodes and their parent file inodes are possible. Signed-off-by: Brian Behlendorf <[email protected]> Issue openzfs#457 Conflicts: module/zfs/zfs_vnops.c
ryao · Aug 11, 2014 · 4858825 · 4858825
1 parent fbd975b
commit 4858825
Show file tree

Hide file tree

Showing 4 changed files with 17 additions and 11 deletions.
diff --git a/include/sys/zfs_vnops.h b/include/sys/zfs_vnops.h
@@ -79,6 +79,7 @@ extern int zfs_putpage(struct inode *ip, struct page *pp,
 extern int zfs_dirty_inode(struct inode *ip, int flags);
 extern int zfs_map(struct inode *ip, offset_t off, caddr_t *addrp,
     size_t len, unsigned long vm_flags);
+extern void zfs_iput_async(struct inode *ip);
 
 #ifdef	__cplusplus
 }

diff --git a/module/zfs/zfs_dir.c b/module/zfs/zfs_dir.c
@@ -46,6 +46,7 @@
 #include <sys/policy.h>
 #include <sys/zfs_dir.h>
 #include <sys/zfs_acl.h>
+#include <sys/zfs_vnops.h>
 #include <sys/fs/zfs.h>
 #include "fs/fs_subr.h"
 #include <sys/zap.h>
@@ -528,7 +529,7 @@ zfs_purgedir(znode_t *dzp)
 		error = dmu_tx_assign(tx, TXG_WAIT);
 		if (error) {
 			dmu_tx_abort(tx);
-			iput(ZTOI(xzp));
+			zfs_iput_async(ZTOI(xzp));
 			skipped += 1;
 			continue;
 		}
@@ -541,7 +542,7 @@ zfs_purgedir(znode_t *dzp)
 			skipped += 1;
 		dmu_tx_commit(tx);
 
-		iput(ZTOI(xzp));
+		zfs_iput_async(ZTOI(xzp));
 	}
 	zap_cursor_fini(&zc);
 	if (error != ENOENT)
@@ -729,7 +730,7 @@ zfs_rmnode(znode_t *zp)
 	dmu_tx_commit(tx);
 out:
 	if (xzp)
-		iput(ZTOI(xzp));
+		zfs_iput_async(ZTOI(xzp));
 }
 
 static uint64_t

diff --git a/module/zfs/zfs_vnops.c b/module/zfs/zfs_vnops.c
@@ -101,7 +101,7 @@
  *	pushing cached pages (which acquires range locks) and syncing out
  *	cached atime changes.  Third, zfs_zinactive() may require a new tx,
  *	which could deadlock the system if you were already holding one.
- *	If you must call iput() within a tx then use iput_ASYNC().
+ *	If you must call iput() within a tx then use zfs_iput_async().
  *
  *  (3)	All range locks must be grabbed before calling dmu_tx_assign(),
  *	as they can span dmu_tx_assign() calls.
@@ -927,12 +927,17 @@ zfs_write(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr)
 }
 EXPORT_SYMBOL(zfs_write);
 
-static void
-iput_async(struct inode *ip, taskq_t *taskq)
+void
+zfs_iput_async(struct inode *ip)
 {
+	objset_t *os = ITOZSB(ip)->z_os;
+
 	ASSERT(atomic_read(&ip->i_count) > 0);
+	ASSERT(os != NULL);
+
 	if (atomic_read(&ip->i_count) == 1)
-		taskq_dispatch(taskq, (task_func_t *)iput, ip, TQ_PUSHPAGE);
+		taskq_dispatch(dsl_pool_iput_taskq(dmu_objset_pool(os)),
+		    (task_func_t *)iput, ip, TQ_SLEEP);
 	else
 		iput(ip);
 }
@@ -941,7 +946,6 @@ void
 zfs_get_done(zgd_t *zgd, int error)
 {
 	znode_t *zp = zgd->zgd_private;
-	objset_t *os = ZTOZSB(zp)->z_os;
 
 	if (zgd->zgd_db)
 		dmu_buf_rele(zgd->zgd_db, zgd);
@@ -952,7 +956,7 @@ zfs_get_done(zgd_t *zgd, int error)
 	 * Release the vnode asynchronously as we currently have the
 	 * txg stopped from syncing.
 	 */
-	iput_async(ZTOI(zp), dsl_pool_iput_taskq(dmu_objset_pool(os)));
+	zfs_iput_async(ZTOI(zp));
 
 	if (error == 0 && zgd->zgd_bp)
 		zil_add_block(zgd->zgd_zilog, zgd->zgd_bp);
@@ -994,7 +998,7 @@ zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio)
 		 * Release the vnode asynchronously as we currently have the
 		 * txg stopped from syncing.
 		 */
-		iput_async(ZTOI(zp), dsl_pool_iput_taskq(dmu_objset_pool(os)));
+		zfs_iput_async(ZTOI(zp));
 		return (SET_ERROR(ENOENT));
 	}
 

diff --git a/module/zfs/zfs_znode.c b/module/zfs/zfs_znode.c
@@ -292,7 +292,7 @@ zfs_inode_destroy(struct inode *ip)
 	}
 
 	if (zp->z_xattr_parent) {
-		iput(ZTOI(zp->z_xattr_parent));
+		zfs_iput_async(ZTOI(zp->z_xattr_parent));
 		zp->z_xattr_parent = NULL;
 	}