Hi Matt,
Here is my janitoring item for tmpfs.
I would really appreciate your effort to polish on tmpfs,
Also I would do small pieces of janitors in parallel.
The attachment is a patch to enable option -u, -g and -m as NetBSD
does. We can mount with non root uid, gid or even with sticky bit by
applying this.
See mount_tmpfs(8) for reference.
My next plan is to write a patch allowing users full notation (i.e. -o
uid=hoge,mode=1777 instead of -o -u=hoge -o -m=1777) as we see by
tmpfs(5)
thank you,
- Naoya
On 2/16/10, Matthew Dillon <[email protected]> wrote:
> We now have a very nice port by Naoya Sugioka of TMPFS from NetBSD
> in our main development branch. It is brand new and still considered
> experimental but should be reasonably stable.
>
> TMPFS is a better alternative to MFS and MD for temporary filesystems.
> It doesn't have the data duplication issue that MFS has and its data
> is swap-backed using a very nice algorithm that only pages data to swap
> when the system is under real memory pressure (and not due to buffer
> cache flushes, fsyncs, or anything else).
>
> -Matt
> Matthew Dillon
> <[email protected]>
>
diff --git a/sys/vfs/tmpfs/tmpfs_vfsops.c b/sys/vfs/tmpfs/tmpfs_vfsops.c
index 5fd5d38..d183889 100644
--- a/sys/vfs/tmpfs/tmpfs_vfsops.c
+++ b/sys/vfs/tmpfs/tmpfs_vfsops.c
@@ -171,6 +171,9 @@ tmpfs_mount(struct mount *mp, char *path, caddr_t data,
struct ucred *cred)
}
size_max = args.ta_size_max;
nodes_max = args.ta_nodes_max;
+ root_uid = args.ta_root_uid;
+ root_gid = args.ta_root_gid;
+ root_mode = args.ta_root_mode;
}
/*
