[dragonfly.git] / sys / vfs / hammer2 / TODO


* Kernel-side needs to clean up transaction queues and make appropriate
  callbacks.

* Userland side needs to do the same for any initiated transactions.

* Nesting problems in the flusher.

* Inefficient vfsync due to thousands of file buffers, one per-vnode.
  (need to aggregate using a device buffer?)

* Adjust the flusher to unlock the parent after the child is locked,
  then restart if the parent changed out from under us.  This will
  greatly reduce namecache contention.

* Use bp->b_dep to interlock the buffer with the chain structure so the
  strategy code can calculate the crc and assert that the chain is marked
  modified (not yet flushed).

* Deleted inode not reachable via tree for volume flush but still reachable
  via fsync/inactive/reclaim.  Its tree can be destroyed at that point.

* The direct write code needs to invalidate any underlying physical buffers.
  Direct write needs to be implemented.

* Make sure a resized block (hammer2_chain_resize()) calculates a new
  hash code in the parent bref

* The freemap allocator needs to getblk/clrbuf/bdwrite any partial
  block allocations (less than 64KB) that allocate out of a new 64K
  block, to avoid causing a read-before-write I/O.

* Check flush race upward recursion setting SUBMODIFIED vs downward
  recursion checking SUBMODIFIED then locking (must clear before the
  recursion and might need additional synchronization)

* There is definitely a flush race in the hardlink implementation between
  the forwarding entries and the actual (hidden) hardlink inode.

  This will require us to associate a small hard-link-adjust structure
  with the chain whenever we create or delete hardlinks, on top of
  adjusting the hardlink inode itself.  Any actual flush to the media
  has to synchronize the correct nlinks value based on whether related
  created or deleted hardlinks were also flushed.

* When a directory entry is created and also if an indirect block is
  created and entries moved into it, the directory seek position can
  potentially become incorrect during a scan.

* When a directory entry is deleted a directory seek position depending
  on that key can cause readdir to skip entries.

* TWO PHASE COMMIT - store two data offsets in the chain, and
  hammer2_chain_delete() needs to leave the chain intact if MODIFIED2 is
  set on its buffer until the flusher gets to it?


				OPTIMIZATIONS

* If a file is unlinked buts its descriptors is left open and used, we
  should allow data blocks on-media to be reused since there is no
  topology left to point at them.
Commit	Line	Data
8c280d5d MD	1
	2	* Kernel-side needs to clean up transaction queues and make appropriate
	3	callbacks.
	4
	5	* Userland side needs to do the same for any initiated transactions.
	6
222d9e22 MD	7	* Nesting problems in the flusher.
222d9e22 MD	8
01eabad4 MD	9	* Inefficient vfsync due to thousands of file buffers, one per-vnode.
	10	(need to aggregate using a device buffer?)
	11
222d9e22 MD	12	* Adjust the flusher to unlock the parent after the child is locked,
	13	then restart if the parent changed out from under us. This will
	14	greatly reduce namecache contention.
995e78dc	15
8cce658d MD	16	* Use bp->b_dep to interlock the buffer with the chain structure so the
	17	strategy code can calculate the crc and assert that the chain is marked
	18	modified (not yet flushed).
	19
	20	* Deleted inode not reachable via tree for volume flush but still reachable
	21	via fsync/inactive/reclaim. Its tree can be destroyed at that point.
	22
866d5273 MD	23	* The direct write code needs to invalidate any underlying physical buffers.
	24	Direct write needs to be implemented.
	25
	26	* Make sure a resized block (hammer2_chain_resize()) calculates a new
222d9e22	27	hash code in the parent bref
866d5273	28
995e78dc MD	29	* The freemap allocator needs to getblk/clrbuf/bdwrite any partial
	30	block allocations (less than 64KB) that allocate out of a new 64K
	31	block, to avoid causing a read-before-write I/O.
	32
	33	* Check flush race upward recursion setting SUBMODIFIED vs downward
	34	recursion checking SUBMODIFIED then locking (must clear before the
	35	recursion and might need additional synchronization)
	36
db0c2eb3 MD	37	* There is definitely a flush race in the hardlink implementation between
	38	the forwarding entries and the actual (hidden) hardlink inode.
	39
	40	This will require us to associate a small hard-link-adjust structure
	41	with the chain whenever we create or delete hardlinks, on top of
	42	adjusting the hardlink inode itself. Any actual flush to the media
	43	has to synchronize the correct nlinks value based on whether related
	44	created or deleted hardlinks were also flushed.
	45
995e78dc MD	46	* When a directory entry is created and also if an indirect block is
	47	created and entries moved into it, the directory seek position can
	48	potentially become incorrect during a scan.
	49
	50	* When a directory entry is deleted a directory seek position depending
	51	on that key can cause readdir to skip entries.
db0c2eb3	52
73e441b9 MD	53	* TWO PHASE COMMIT - store two data offsets in the chain, and
	54	hammer2_chain_delete() needs to leave the chain intact if MODIFIED2 is
	55	set on its buffer until the flusher gets to it?
	56
db0c2eb3 MD	57
	58	OPTIMIZATIONS
	59
	60	* If a file is unlinked buts its descriptors is left open and used, we
	61	should allow data blocks on-media to be reused since there is no
	62	topology left to point at them.