Re: Fix nagging make_dev() warnings

["Frank W. Josellis" <frank@xxxxxxxxxxxxxxxxxxxxx>]

Well, I see there is a hidden bug, that's not just a feature. This is not 
a fix, but at least it gives a hint where the problem is located. Please 
have a look at the patches below. With patch #1 for hashdev() "consolectl" 
is correctly retrieved from the device hash table and, consequently, the 
warning doesn't show up. Patch #2 indicates why the original hashdev() 
failed. It provides some debugging output for spec_open(), and in case 
somebody is interested, I have posted a corresponding dmesg output at

http://www.senax.net/downloads/dmesg.debug

For "consolectl" it shows that dev->si_ops->head.name was changed from 
"sc" to "intercept", which should not happen. Apparently something is 
messed up in sys/kern/tty_cons.c, but I haven't yet checked the details.

Regards,
Frank Josellis


--- patch_1 begins here ---
--- src/sys/kern/kern_conf.c.orig	2006-09-10 03:26:39.000000000 +0200
+++ src/sys/kern/kern_conf.c	2006-09-14 21:16:53.000000000 +0200
@@ -124,8 +124,8 @@
 	udev = makeudev(x, y);
 	hash = udev % DEVT_HASH;
 	LIST_FOREACH(si, &dev_hash[hash], si_hash) {
-		if (si->si_ops == ops && si->si_udev == udev)
-			return (si);
+	  if (si->si_ops->head.maj == ops->head.maj && si->si_udev == udev)
+	    return (si);
 	}
 	if (stashed >= DEVT_STASH) {
 		MALLOC(si, struct cdev *, sizeof(*si), M_DEVT,
--- patch_1 ends here ---

--- patch_2 begins here ---
--- src/sys/vfs/specfs/spec_vnops.c.orig	2006-09-10 03:26:41.000000000 +0200
+++ src/sys/vfs/specfs/spec_vnops.c	2006-09-14 21:44:43.000000000 +0200
@@ -266,6 +266,10 @@
 			dev->si_bsize_phys = DEV_BSIZE;
 		vinitvmio(vp, IDX_TO_OFF(INT_MAX));
 	}
+	if ( dev->si_ops->head.maj != 2) {
+	  printf("spec_open: \"%s\" (%s, %p)\n",
+		 devtoname(dev), dev_dname(dev), dev);
+	}
 	if ((dev_dflags(dev) & D_DISK) == 0) {
 		cp = devtoname(dev);
 		if (*cp == '#') {
--- patch_2 ends here ---