Re: wait_for_syfs errors

[Kay Sievers <kay.sievers@vrfy.org>]

[-- Attachment #1: Type: text/plain, Size: 1012 bytes --]

On Fri, Oct 22, 2004 at 04:23:52PM -0700, Greg KH wrote:
> On Fri, Oct 22, 2004 at 10:39:07PM +0200, Kay Sievers wrote:
> > On Fri, Oct 22, 2004 at 10:22:29PM +0200, Marcel Paulekuhn wrote:
> > > Hm yeah, I am, 2.6.9-mm1.
> > > 
> > > So there's no other way to fix this than changing kernel? Because I
> > > want to keep reiser4-support.
> > > 
> > > 
> > > On Fri, 22 Oct 2004 22:17:46 +0200, Kay Sievers <kay.sievers@vrfy.org> wrote:
> > > > 
> > > > You are running a -mm kernel, right? There is a bug in there, that
> > > > causes this.
> > 
> > The message is not critical. It just means, that you get (useless) hotplug
> > events for the class itself which should not happen. This will hopefully
> > fixed soon in the -mm tree and will go away.
> 
> Actually I just commited that change to the main kernel tree, so we need
> to fix this up in wait_for_sysfs before we start to get a lot of people
> complaining :)

Ok, here is the fix. Bad events with no device in the class are now ignored.

Thanks,
Kay

[-- Attachment #2: wait-bad_event-01.patch --]
[-- Type: text/plain, Size: 626 bytes --]

===== wait_for_sysfs.c 1.27 vs edited =====
--- 1.27/wait_for_sysfs.c	2004-10-19 03:50:52 +02:00
+++ edited/wait_for_sysfs.c	2004-10-23 03:57:33 +02:00
@@ -99,6 +99,12 @@
 		snprintf(filename, SYSFS_PATH_MAX-1, "%s%s", sysfs_mnt_path, devpath);
 		filename[SYSFS_PATH_MAX-1] = '\0';
 
+		/* skip bad events where we get no device for the class */
+		if (strncmp(devpath, "/class/", 7) == 0 && strchr(&devpath[7], '/') == NULL) {
+			dbg("no device name for '%s', bad event", devpath);
+			goto exit;
+		}
+
 		/* open the class device we are called for */
 		class_dev = open_class_device_wait(filename);
 		if (!class_dev) {