It doesn't make sense to pollute 'mdconfig' with functions that we already
have an API available to perform. An API called 'mount'.
Why not write a 'mount_md' program to do all the magic based on fstab
options, similar to what mount_mfs used to do for MFS? A 'mount_md'
would give us instant integration into existing kernel mechanisms,
including startup (/etc/rc) mechanisms.
-Matt
:[cvs-(all|commiters) -> -hackers]
:
:> >> If anybody writes a patch to mdconfig to DTRT based on some
:> >> less bogus /etc/fstab entries, I'll happily review and commit it.
:> >
:> >Does this sort of functionality really belong in mdconfig?
:>
:> Not by definition, and I'm not religious about it.
:
:Okay, any objections then to jhb's configuration file idea? The idea
:is that mdconfig parses a file (md.conf or something) and sets up some
:memory disks accordingly, including optional disklabel/newfs stuff.
:Later, a simple 'ffs'-type fstab entry will mount it wherever the user
:wants. This may be a little complicated for the simple /tmp case, but
:I can think of some environments where this would be useful. vnconfig
:also supports a similar config file option, so I think it's fitting
:that mdconfig should, too.
:
:I've come up with a preliminary patch to teach mdconfig(8) to parse a
:config file. It's a little early to be of practical use (and it lacks
:the disklabel/newfs option), but it's a start. It's a little long
:because I made some of the code in the switch(ch) block into functions
:to avoid having two blocks doing the same thing, so I've decided not
:to attach it not to bother the uninterested parties. If you could
:take a minute or two and comment on it I'd appreciate it. You can
:find it at http://www.unixfreak.org/~dima/mdconfig1.diff
:
:The current format of the config file is:
:
: <unit> <type> <argument=value>,<arg=value>,...
:
:for example:
:
: md0 swap size=64m
:
:Comments? Suggestions?
:
:Thanks in advance
:
: Dima Dorfman
: [EMAIL PROTECTED]
:
:P.S. Is there any reason the -b (baseaddr) option for preload disks
:is currently unimplemented? The patch implements it (it was one line,
:so I thought 'what the heck'); the ioctl works, but I was unable to
:test it.
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-hackers" in the body of the message
