Messages in this thread | | | Date | Mon, 18 Oct 1999 04:14:13 +0400 | From | Hans Reiser <> | Subject | Re: fs specific union in struct buffer head? Optimal implementation of allocate on flush seems to require a b_parent field in struct buffer_head. |
| |
I want to first implement allocate, then implement re-allocate. Note the synergy between this, and online shrinking the fs, and journaling, and wandering logs, and implementing online reordering for higher performance during the night.
Hans
"Stephen C. Tweedie" wrote:
> Hi, > > On Fri, 15 Oct 1999 23:22:23 +0400, Hans Reiser <reiser@idiom.com> > said: > > > Vladimir and I are discussing how to implement allocate on flush in our next > > major version (not for 2.4) of reiserfs. > > >> If you are going to reallocate on flush unformatted nodes you must be able > >> to: > > Remember, there is a huge difference between allocate-on-flush and > reallocate-on-flush. In the former case, you simply don't have > buffer_heads present until the final flush: everything has to be > maintained in the filesystem itself or the VM, not in the buffer > cache, until the final flush (because the buffer cache is really only > for storing blocks whose on-disk location is known). > > In the reallocate case, a buffer_head b_fs_private pointer to > fs-specific metadata would allow the fs to deal with this > independently of the buffer cache layer itself. For > allocate-on-flush, though, you need to be able to look up the > information from the VM, not from the buffer cache. > > --Stephen
-- Get Linux (http://www.kernel.org) plus ReiserFS (http://devlinux.org/namesys). If you sell an OS or internet appliance, buy a port of ReiserFS! If you need customizations and industrial grade support, we sell them.
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |