It also occured to me this would be a good opportunity to work on
adding dirty page support to the page cache code and I have done that
as well.
The quick synopsis of what I have written is as follows:
A complete filesystem that resides in swap, and for kicks and reduced
space consumption uses a btree for all of it's directories.
A patch to allow sync to sync mmaped files, as well as everything else.
A patch to allow the page cache to handle dirty files
A patch to allow the swapoff to work with swap pages that do not
reside in page tables.
SYSV shared memory probably should be modified to work with this.
This is an beta release. The filesystem is fully functional, but I am the
only one who has tested it so far. If I could find bugs it would be
alpha but since I can't, it's beta :) Extra eyes, and testers welcome.
Well that is the case with 2.0.32 port 2.1.91 has a major glitch.
I suspect it is a bug in another part of the kernel, randomly
trashing, parts of inodes, but I haven't been able to track it down to
know :(
My code is at:
http://www.npwt.net/~ebiederm/files/shmfs-0.1.007.tgz
This works with both kernels 2.0.32 and 2.1.91
2.0.32 support is almost necessary so I can tell my bugs from, those
of others :)
Changes since 0.1.004:
I have synchronized my two versions, and rewritten a bunch of namei
code in the search for races. And I have found and fixed my
mysterious mutating symlink bug.
Eric
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu