(ugh, following up to my own mail a second time ;)
On Mon, 08 May 2000 00:53:46 Marcus Brinkmann wrote:
> The other direction seems to work, though. Currently, in the node cache, I
> hold the dirent (directory entry) address in disk_image.
I think I can move the struct dirrect * to the disknode struct and
have a reader/writer lock for it, like allocsize.
> I doubt that this will suffice, though. In isofs (and thus current fatfs),
> the cache consists of all nodes read so far. However, it makes sense only to
> cache a couple (ext2fs caches 512).
What a stupid mistake by me. 512 is the hash tables size of course.
All nodes are cached, so using the cache would work, but it is less
efficient.
Thanks,
Marcus
