Cylinder group bitmaps and blocks containing inode for a snapshot

file are after snaplock, while other ffs device buffers are before
snaplock in global lock order. By itself, this could cause deadlock
when bdwrite() tries to flush dirty buffers on snapshotted ffs. If,
during the flush, COW activity for snapshot needs to allocate block
and ffs_alloccg() selects the cylinder group that is being written
by bdwrite(), then kernel would panic due to recursive buffer lock
acquision.

Avoid dealing with buffers in bdwrite() that are from other side of
snaplock divisor in the lock order then the buffer being written. Add
new BOP, bop_bdwrite(), to do dirty buffer flushing for same vnode in
the bdwrite(). Default implementation, bufbdflush(), refactors the code
from bdwrite(). For ffs device buffers, specialized implementation is
used.

Reviewed by:	tegge, jeff, Russell Cattelan (cattelan xfs org, xfs changes)
Tested by:	Peter Holm
X-MFC after:	3 weeks (if ever: it changes ABI)

1 files changed, 4 insertions, 0 deletions

diff --git a/sys/sys/bufobj.h b/sys/sys/bufobj.h
index ea568e4..6650702 100644
--- a/sys/sys/bufobj.h
+++ b/sys/sys/bufobj.h
@@ -70,17 +70,20 @@ struct bufv {
 typedef void b_strategy_t(struct bufobj *, struct buf *);
 typedef int b_write_t(struct buf *);
 typedef int b_sync_t(struct bufobj *, int waitfor, struct thread *td);
+typedef void b_bdflush_t(struct bufobj *, struct buf *);
 
 struct buf_ops {
 	char		*bop_name;
 	b_write_t	*bop_write;
 	b_strategy_t	*bop_strategy;
 	b_sync_t	*bop_sync;
+	b_bdflush_t	*bop_bdflush;
 };
 
 #define BO_STRATEGY(bo, bp)	((bo)->bo_ops->bop_strategy((bo), (bp)))
 #define BO_SYNC(bo, w, td)	((bo)->bo_ops->bop_sync((bo), (w), (td)))
 #define BO_WRITE(bo, bp)	((bo)->bo_ops->bop_write((bp)))
+#define BO_BDFLUSH(bo, bp)	((bo)->bo_ops->bop_bdflush((bo), (bp)))
 
 struct bufobj {
 	struct mtx	*bo_mtx;	/* Mutex which protects "i" things */
@@ -130,6 +133,7 @@ void bufobj_wrefl(struct bufobj *bo);
 int bufobj_invalbuf(struct bufobj *bo, int flags, struct thread *td, int slpflag, int slptimeo);
 int bufobj_wwait(struct bufobj *bo, int slpflag, int timeo);
 int bufsync(struct bufobj *bo, int waitfor, struct thread *td);
+void bufbdflush(struct bufobj *bo, struct buf *bp);
 
 #endif /* defined(_KERNEL) || defined(_KVM_VNODE) */
 #endif /* _SYS_BUFOBJ_H_ */