When attempt is made to suspend a filesystem that is already syspended,

wait until the current suspension is lifted instead of silently returning
success immediately. The consequences of calling vfs_write() resume when
not owning the suspension are not well-defined at best.

Add the vfs_susp_clean() mount method to be called from
vfs_write_resume(). Set it to process_deferred_inactive() for ffs, and
stop calling it manually.

Add the thread flag TDP_IGNSUSP that allows to bypass the suspension
point in the vn_start_write. It is intended for use by VFS in the
situations where the suspender want to do some i/o requiring calls to
vn_start_write(), and this i/o cannot be done later.

Reviewed by:	tegge
In collaboration with:	pho
MFC after:	 1 month

6 files changed, 37 insertions, 14 deletions

diff --git a/sys/kern/vfs_vnops.c b/sys/kern/vfs_vnops.c
index 58863f4..c66e49c 100644
--- a/sys/kern/vfs_vnops.c
+++ b/sys/kern/vfs_vnops.c
@@ -947,15 +947,18 @@ vn_start_write(vp, mpp, flags)
 	/*
 	 * Check on status of suspension.
 	 */
-	while ((mp->mnt_kern_flag & MNTK_SUSPEND) != 0) {
-		if (flags & V_NOWAIT) {
-			error = EWOULDBLOCK;
-			goto unlock;
+	if ((curthread->td_pflags & TDP_IGNSUSP) == 0 ||
+	    mp->mnt_susp_owner != curthread) {
+		while ((mp->mnt_kern_flag & MNTK_SUSPEND) != 0) {
+			if (flags & V_NOWAIT) {
+				error = EWOULDBLOCK;
+				goto unlock;
+			}
+			error = msleep(&mp->mnt_flag, MNT_MTX(mp),
+			    (PUSER - 1) | (flags & PCATCH), "suspfs", 0);
+			if (error)
+				goto unlock;
 		}
-		error = msleep(&mp->mnt_flag, MNT_MTX(mp), 
-		    (PUSER - 1) | (flags & PCATCH), "suspfs", 0);
-		if (error)
-			goto unlock;
 	}
 	if (flags & V_XSLEEP)
 		goto unlock;
@@ -1079,11 +1082,14 @@ vfs_write_suspend(mp)
 	int error;
 
 	MNT_ILOCK(mp);
-	if (mp->mnt_kern_flag & MNTK_SUSPEND) {
+	if (mp->mnt_susp_owner == curthread) {
 		MNT_IUNLOCK(mp);
-		return (0);
+		return (EALREADY);
 	}
+	while (mp->mnt_kern_flag & MNTK_SUSPEND)
+		msleep(&mp->mnt_flag, MNT_MTX(mp), PUSER - 1, "wsuspfs", 0);
 	mp->mnt_kern_flag |= MNTK_SUSPEND;
+	mp->mnt_susp_owner = curthread;
 	if (mp->mnt_writeopcount > 0)
 		(void) msleep(&mp->mnt_writeopcount, 
 		    MNT_MTX(mp), (PUSER - 1)|PDROP, "suspwt", 0);
@@ -1104,12 +1110,17 @@ vfs_write_resume(mp)
 
 	MNT_ILOCK(mp);
 	if ((mp->mnt_kern_flag & MNTK_SUSPEND) != 0) {
+		KASSERT(mp->mnt_susp_owner == curthread, ("mnt_susp_owner"));
 		mp->mnt_kern_flag &= ~(MNTK_SUSPEND | MNTK_SUSPEND2 |
 				       MNTK_SUSPENDED);
+		mp->mnt_susp_owner = NULL;
 		wakeup(&mp->mnt_writeopcount);
 		wakeup(&mp->mnt_flag);
-	}
-	MNT_IUNLOCK(mp);
+		curthread->td_pflags &= ~TDP_IGNSUSP;
+		MNT_IUNLOCK(mp);
+		VFS_SUSP_CLEAN(mp);
+	} else
+		MNT_IUNLOCK(mp);
 }
 
 /*
diff --git a/sys/sys/mount.h b/sys/sys/mount.h
index 7e9534e..27a4a60 100644
--- a/sys/sys/mount.h
+++ b/sys/sys/mount.h
@@ -180,6 +180,7 @@ struct mount {
 	int		mnt_holdcntwaiters;	/* waits on hold count */
 	int		mnt_secondary_writes;   /* (i) # of secondary writes */
 	int		mnt_secondary_accwrites;/* (i) secondary wr. starts */
+	struct thread	*mnt_susp_owner;	/* (i) thread owning suspension */
 #define	mnt_endzero	mnt_gjprovider
 	char		*mnt_gjprovider;	/* gjournal provider name */
 	struct lock	mnt_explock;		/* vfs_export walkers lock */
@@ -544,6 +545,7 @@ typedef	int vfs_extattrctl_t(struct mount *mp, int cmd,
 typedef	int vfs_mount_t(struct mount *mp, struct thread *td);
 typedef int vfs_sysctl_t(struct mount *mp, fsctlop_t op,
 		    struct sysctl_req *req);
+typedef void vfs_susp_clean_t(struct mount *mp);
 
 struct vfsops {
 	vfs_mount_t		*vfs_mount;
@@ -560,6 +562,7 @@ struct vfsops {
 	vfs_uninit_t		*vfs_uninit;
 	vfs_extattrctl_t	*vfs_extattrctl;
 	vfs_sysctl_t		*vfs_sysctl;
+	vfs_susp_clean_t	*vfs_susp_clean;
 };
 
 vfs_statfs_t	__vfs_statfs;
@@ -581,6 +584,9 @@ vfs_statfs_t	__vfs_statfs;
 	(*(MP)->mnt_op->vfs_extattrctl)(MP, C, FN, NS, N, P)
 #define VFS_SYSCTL(MP, OP, REQ) \
 	(*(MP)->mnt_op->vfs_sysctl)(MP, OP, REQ)
+#define	VFS_SUSP_CLEAN(MP) \
+	({if (*(MP)->mnt_op->vfs_susp_clean != NULL)		\
+	       (*(MP)->mnt_op->vfs_susp_clean)(MP); })
 
 extern int mpsafe_vfs;
 
diff --git a/sys/sys/proc.h b/sys/sys/proc.h
index 2ec5d20..66e95ac 100644
--- a/sys/sys/proc.h
+++ b/sys/sys/proc.h
@@ -359,6 +359,7 @@ do {									\
 #define	TDP_INBDFLUSH	0x00100000 /* Already in BO_BDFLUSH, do not recurse */
 #define	TDP_KTHREAD	0x00200000 /* This is an official kernel thread */
 #define	TDP_CALLCHAIN	0x00400000 /* Capture thread's callchain */
+#define	TDP_IGNSUSP	0x00800000 /* Permission to ignore the MNTK_SUSPEND* */
 
 /*
  * Reasons that the current thread can not be run yet.
diff --git a/sys/ufs/ffs/ffs_extern.h b/sys/ufs/ffs/ffs_extern.h
index d5dfa12..4a2f3a2 100644
--- a/sys/ufs/ffs/ffs_extern.h
+++ b/sys/ufs/ffs/ffs_extern.h
@@ -80,6 +80,7 @@ void	ffs_snapremove(struct vnode *vp);
 int	ffs_snapshot(struct mount *mp, char *snapfile);
 void	ffs_snapshot_mount(struct mount *mp);
 void	ffs_snapshot_unmount(struct mount *mp);
+void	process_deferred_inactive(struct mount *mp);
 int	ffs_syncvnode(struct vnode *vp, int waitfor);
 int	ffs_truncate(struct vnode *, off_t, int, struct ucred *, struct thread *);
 int	ffs_update(struct vnode *, int);
diff --git a/sys/ufs/ffs/ffs_snapshot.c b/sys/ufs/ffs/ffs_snapshot.c
index c367d66..fab6798 100644
--- a/sys/ufs/ffs/ffs_snapshot.c
+++ b/sys/ufs/ffs/ffs_snapshot.c
@@ -167,7 +167,6 @@ static int snapacct_ufs2(struct vnode *, ufs2_daddr_t *, ufs2_daddr_t *,
 static int mapacct_ufs2(struct vnode *, ufs2_daddr_t *, ufs2_daddr_t *,
     struct fs *, ufs_lbn_t, int);
 static int readblock(struct vnode *vp, struct buf *, ufs2_daddr_t);
-static void process_deferred_inactive(struct mount *);
 static void try_free_snapdata(struct vnode *devvp);
 static struct snapdata *ffs_snapdata_acquire(struct vnode *devvp);
 static int ffs_bp_snapblk(struct vnode *, struct buf *);
@@ -2375,12 +2374,14 @@ readblock(vp, bp, lbn)
 	return (bp->b_error);
 }
 
+#endif
+
 /*
  * Process file deletes that were deferred by ufs_inactive() due to
  * the file system being suspended. Transfer IN_LAZYACCESS into
  * IN_MODIFIED for vnodes that were accessed during suspension.
  */
-static void
+void
 process_deferred_inactive(struct mount *mp)
 {
 	struct vnode *vp, *mvp;
@@ -2453,6 +2454,8 @@ process_deferred_inactive(struct mount *mp)
 	vn_finished_secondary_write(mp);
 }
 
+#ifndef NO_FFS_SNAPSHOT
+
 static struct snapdata *
 ffs_snapdata_alloc(void)
 {
diff --git a/sys/ufs/ffs/ffs_vfsops.c b/sys/ufs/ffs/ffs_vfsops.c
index 93e0b48..86b3f60 100644
--- a/sys/ufs/ffs/ffs_vfsops.c
+++ b/sys/ufs/ffs/ffs_vfsops.c
@@ -105,6 +105,7 @@ static struct vfsops ufs_vfsops = {
 	.vfs_uninit =		ffs_uninit,
 	.vfs_unmount =		ffs_unmount,
 	.vfs_vget =		ffs_vget,
+	.vfs_susp_clean =	process_deferred_inactive,
 };
 
 VFS_SET(ufs_vfsops, ufs, 0);