Messages in this thread | | | Date | Mon, 11 Oct 1999 17:55:16 -0400 | From | Jeff Garzik <> | Subject | Re: My $0.02 on devd and devfs |
| |
Shawn Leas wrote: > > From: Jeff Garzik [mailto:jgarzik@pobox.com] > Subject: Re: My $0.02 on devd and devfs > > >Khimenko Victor wrote: > >> See above. Try to turn on printer, do `cat something > /dev/printer/0` > with > >> devfs and explain how it'll be handled with devd. > >Are you implying here that /dev/printer/0 is created at open() time? > >IMHO that is a bad idea. > > I think you're mis-understanding, and you didn't read the entire post. > If the device node isn't there, devfs can locate the module needed, > load it, and then the DRIVER, implicitely by having been loaded, > puts it's device node into the namespace. (So to speak)
You seem to have repeated what I just said.
Does "/dev/printer/0" exist without the lp module being loaded? If yes, that is wasteful like the current system. If no, you must go outside normal filesystem semantics and talk to a _non-existent_ inode as if it exists.
Drivers which are not explicitly hot-plug need some sort of "kick" before the kernel knows it needs to load a module/driver. Having that kick come in the form of open(), or worse stat(), is the part I don't like. Creating a file implicitly like that just seems alien to the traditional Unix fs.
Jeff
-- Custom driver development | Never worry about theory as long Open source programming | as the machinery does what it's | supposed to do. -- R. A. Heinlein
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |