tech-kern archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: mount(2) and properties



On Wed Jul 02 2008 at 15:21:10 +1000, Simon Burge wrote:
> Another option altogether is to add a new system call entirely.  Antti
> pointed out the FreeBSD nmount() system call, which is declared as:
> 
>       int
>       nmount(struct iovec *iov, u_int niov, int flags);
> 
> but rather than use iovecs I think we'd use a prop_dictionary instead.
> As to a name for this new system call - any suggestions?  :-)
> 
> Comments either way?

A mount syscall which uses proplists and not wildly opaque data would
be quite desireable.

What I want to do is to have the userland mount utility code usable in
the rump utilities.  It would also be nice to have the used data types
kernel-clean, so that the code could be compiled as host binaries on other
operating systems (can't #include from a NetBSD sys tree in that case).
The importance of this is more obvious once Arnaud finishes his SoC work
(*) and we want to access NetBSD file systems on, say, Linux.  There are
of course other ways around this, but I think proplistifying mount is
the best option.

And, as David pointed out, it would help with folding common stuff into
common routines.  Approximately 99.99% of our file systems pass only
the device path as the argument.

We'd obviously call ours fmount() ;)

  - antti

*) http://netbsd-soc.sourceforge.net/projects/fs-utils/


Home | Main Index | Thread Index | Old Index