2 * The freemap allocator needs to getblk/clrbuf/bdwrite any partial
3 block allocations (less than 64KB) that allocate out of a new 64K
4 block, to avoid causing a read-before-write I/O.
6 * Check flush race upward recursion setting SUBMODIFIED vs downward
7 recursion checking SUBMODIFIED then locking (must clear before the
8 recursion and might need additional synchronization)
10 * There is definitely a flush race in the hardlink implementation between
11 the forwarding entries and the actual (hidden) hardlink inode.
13 This will require us to associate a small hard-link-adjust structure
14 with the chain whenever we create or delete hardlinks, on top of
15 adjusting the hardlink inode itself. Any actual flush to the media
16 has to synchronize the correct nlinks value based on whether related
17 created or deleted hardlinks were also flushed.
19 * When a directory entry is created and also if an indirect block is
20 created and entries moved into it, the directory seek position can
21 potentially become incorrect during a scan.
23 * When a directory entry is deleted a directory seek position depending
24 on that key can cause readdir to skip entries.
26 * TWO PHASE COMMIT - store two data offsets in the chain, and
27 hammer2_chain_delete() needs to leave the chain intact if MODIFIED2 is
28 set on its buffer until the flusher gets to it?
33 * If a file is unlinked buts its descriptors is left open and used, we
34 should allow data blocks on-media to be reused since there is no
35 topology left to point at them.