yaffs New background garbage collector and related tweaks This commit - introduces the yaffs background garbage collection feature - tweaks the foregound garbage collection to do less work - changes the way auto-checkpointing works - tweaks the block refreshing logic. The aim of the background garbage collector is to do at least some of the garbage collection in the background so that writing operations will not have to do as much garbage collection which should make writes faster. The amount of background garbage collection is controlled by the ratio between the amount of erased space vs the amount of free space: * If less than quarter of free space is erased, then background gc is frequent. * Else if less than a half of the free space is erased, then background gc is still done reasonably frequently. * else (at least half of the free space is erased) the background gc is done infrequently. Background gc is not attempted if the partition is checkpointed since that would invalidate the checkpoint. The auto-checkpointing feature has changed slightly. If the yaffs_auto_checkpoint value is set to 1 or 2 then the auto checkpointing will be blocked if the erased space is less than half the free space (ie. the auto-checkpointing is blocked to allow background gc to progress). Oring in 4 into the yaffs_auto_checkpoint will do a one-shot override, forcing a checkpoint and suspending background gc until the partition is dirtied by a write, erase etc. The block refreshing control has been changed. The dev->param.refreshPeriod now controls how many blocks are garbage collected before another refresh is performed. Values around 1000 probably make the best sense. Signed-off-by: Charles Manning <cdhmanning@gmail.com>