From owner-freebsd-current Tue Oct 29 00:21:44 1996 Return-Path: owner-current Received: (from root@localhost) by freefall.freebsd.org (8.7.5/8.7.3) id AAA05615 for current-outgoing; Tue, 29 Oct 1996 00:21:44 -0800 (PST) Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.19]) by freefall.freebsd.org (8.7.5/8.7.3) with ESMTP id AAA05605 for ; Tue, 29 Oct 1996 00:21:35 -0800 (PST) Received: (from bde@localhost) by godzilla.zeta.org.au (8.7.6/8.6.9) id TAA00564; Tue, 29 Oct 1996 19:17:53 +1100 Date: Tue, 29 Oct 1996 19:17:53 +1100 From: Bruce Evans Message-Id: <199610290817.TAA00564@godzilla.zeta.org.au> To: jehamby@lightside.com, phk@critter.tfs.com Subject: Re: test DEVFS! Cc: current@freebsd.org Sender: owner-current@freebsd.org X-Loop: FreeBSD.org Precedence: bulk >Okay, on casual inspection, I've found three problems with DEVFS, and one >observation: > >1) Audio devices are not being created correctly. I have a Soundblaster >16, and the following lines in my kernel: I fixed almost everything except the audio devices a long time ago, so it's not surprising that the problems are mostly for audio devices. >2) There is no entry for /dev/sysmouse. This is a device, managed by >syscons and moused, which allows both XFree86 and syscons to use the >mouse. It should be: > >crw------- 1 root wheel 12, 128 Oct 28 21:29 /dev/sysmouse > >and /dev/mouse should be a symlink to /dev/sysmouse. Driver authors are responsible for new devices. >3) The disk devices are not being created, as you mentioned. However, This will be fixed soon. >I've noticed an interesting explanation for this. The disk device is only >created AFTER you mount that device node! For example (/dev is a regular >filesystem, DEVFS is mounted on /mnt/dev): Opening a slice creates all the partitions on the slice. I forget if opening the whole disk creates all the slices. Perhaps not. The root device is opened without going through the file system, so all the partitions on the same slice as it are there to begin with. >Notice how the entries for /dev/sd2s4 showed up after I mounted it? >Because of removable media (in this case, a Zip drive), it seems we need a >"disks" utility (indeed, this is the name of the Solaris equivalent) to >probe the SCSI chain and recreate the nodes if a disk has been added or >removed. I hope you have already thought of this problem! Opening something on it will soon work right (all the old devices will go away and all the new ones will appear), but this should not be required. Just changing the disk should do this, except if one of the old devices is open you should expect problems if you changing. >4) Finally, an observation. There are a few devices which are created by >LKM's, for example joy_mod.o. These LKM's would need to be recompiled >with -DDEVFS or else they won't work for a DEVFS system. I didn't test >this theory extensively, however I tried loading joy_mod.o and the joy0 >device of course did not show up in DEVFS. When I unloaded joy_mod.o, the >kernel panicked! I hate ifdefs. Bruce