JFFS3 & deletion dirents RFC
Artem B. Bityuckiy
dedekind at infradead.org
Thu Mar 24 04:18:22 EST 2005
Jared Hulbert wrote:
> Forgive me if my ignorance is obvious...
> Are there any implications to non-dirent nodes?
No, in data nodes versions play another and very important role.
Basically, they help to distinguish between valid and obsolete data when
the data nodes' data ranges overlap.
> Shouldn't this speed
> up the GC, mount time for any node? If not then the version field for
> non-dirent nodes is utterly wasted, right? If the version field is
> wasted for all non-dirent nodes (in other words, most nodes) then it
> seems like a bad idea to add a WORD or DWORD to each node. If there is
> a way to make version field only apply to dirent nodes then it would
> be much better, of course that sounds ugly without even thinking it
> through too far.
We can't get rid of version in data nodes.
data node A has data that covers file's range 0-100
data node B has data that covers file's range 10-110
How can we detect from which node should we fetch data of the 10-100
range? JFFS2 uses version for this reason.
Dirents' versions play another role.
Artem B. Bityuckiy,
More information about the linux-mtd