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 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 has been modified to work with this, fixing a
really old/rare swapoff bug.
A patch to allow asynchronous I/O to swapfiles.
This is an beta release, aimed at producing a stable version shortly.
My code is at:
http://www.npwt.net/~ebiederm/files/
files:
shmfs-0.1.009.tar.gz shmfs-0.3.001.tar.gz
shmfs-0.3.001 is a developmental version written primarily by
Gabor Kuti <seasons@falcon.sch.bme.hu> where new features are being
worked upon. Since I just started integrating his patches I may have
done something stupid...
shmfs-0.1.009 has hit the code freeze before becoming shmfs-0.2.000
a stable and usefull version. Strange interactions with kernel memory
management are the only things left before 0.2
This works with both kernels 2.0.32 and 2.1.101.
2.1.102-2.1.105 should also work but I haven't tested those.
2.0.32 support is almost necessary so I can tell my bugs from, those
of others :)
Changes since 0.1.008:
Fixed minor bugs and ported to kernel 101. Now I should have a
correct kernel patch, on a kernel that handles swapping correctly.
Use on development kernels is no longer discouraged, only minor
performance glitches appear to remain as problems there.
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