SuSE 10.0 udev stuck with bogus name / rule entry

SuSE 10.0 udev stuck with bogus name / rule entry

[Ken Ryan]

Hello.

I'm trying to recover from a failed attempt at creating a udev rule.

I'm running SuSE 10.0, pretty well up to date via YOU.

At one point I tried to make world-access permissions for /dev/ttyS0 
stick by writing a udev rule.  I clearly did not do it correctly. 
Worse, I did not keep the rule as I wrote it.

The problem now, though, is udev is incapable of sorting out non-static
devices.  This includes /dev/hda*, so the boot process crashes.

I do have a laptop configured with 10.0, and I did some comparing of
file contents.  I have verified that the boot scripts, udev rules,
and udev binaries are, at this instant in time, correct (they match
the working system).  My bogus changes are fully reverted.

Now, when I run udevinfo -q all -p /sys/block/hda/hda1 (for example)
all of the lines returned look fine EXCEPT the N: line.  It says

	N: ttyS0

for all my hd* devices, loop devices, ram devices, etc.  These
also appear in the files in /dev/.udevdb - ALL of the files in
there contain the above N: line instead of the proper line.

The system fails some fscks, so drops me to single mode.  If I then
run "/etc/init.d/boot.udev force-reload" I get the proper entries
created.  When I shut down and restart, however, I'm back to square 1.

So it seems that somewhere it is remembering something about my bogus 
ttyS0 rule and screqing up the detection process.

I searched the /etc and /var trees for any mention of ttyS0; everywhere
that comes up makes perfect sense for serial-port things.

I would appreciate any suggestions on what to try or where else to look.

BTW, none of my log files are getting updated.  I presume it's because 
of those missing /dev entries.  So I can't see any output from udevd 
there.  I haven't yet gotten a serial console kicking, though since I
don't see any abnormal messages on the boot screens I'm not sure it'll
help (note I keep the boot splash turned off so I can see the boot 
messages; I've also booted into "failsafe" and normal modes with 
equivalent results).

Thanks in advance!

		ken

p.s. I'm not subscribed to the list but I'll check back on marc.  I'd 
appreciate if responses are CCed to me though.



_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: SuSE 10.0 udev stuck with bogus name / rule entry

[Greg KH]

On Thu, Jun 08, 2006 at 02:37:44PM -0400, Ken Ryan wrote:
> Hello.
> 
> I'm trying to recover from a failed attempt at creating a udev rule.
> 
> I'm running SuSE 10.0, pretty well up to date via YOU.

Any chance of trying out 10.1 now?  This list isn't for distro specific
support, sorry, try one of the opensuse mailing lists (but they will
probably just recommend upgrading to 10.1...)

good luck,

greg k-h


_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: SuSE 10.0 udev stuck with bogus name / rule entry

[Ken Ryan]

Greg KH wrote:

>On Thu, Jun 08, 2006 at 02:37:44PM -0400, Ken Ryan wrote:
>  
>
>>Hello.
>>
>>I'm trying to recover from a failed attempt at creating a udev rule.
>>
>>I'm running SuSE 10.0, pretty well up to date via YOU.
>>    
>>
>
>Any chance of trying out 10.1 now?  This list isn't for distro specific
>support, sorry, try one of the opensuse mailing lists (but they will
>probably just recommend upgrading to 10.1...)
>
>good luck,
>
>greg k-h
>  
>
I am planning on upgrading to 10.1 whenever I get my shipment from 
Novell (I took
Novell up on their free shipping offer, now the box has been on 
backorder for a couple
weeks.  My net connection is only a measly 128kb so the idea of 
downloading iso
images doesn't appeal to me).  I have to admit I've become a bit leery 
about upgrading
because of all the folks griping on the suse lists about how buggy 10.1 is.

I know the list isn't for distro-specific stuff; I was hoping a 
udev/hotplug guru would be
able to help me understand what the heck happened, and where ttyS0 is 
getting remembered.
As I continue to dig into this I'm getting the impression that udev and 
hotplug are necessarily
very distro-specific so I can understand if you can't really help.  I 
did post to a couple suse
lists, I just thought I might get lucky.

I did find that if I put lines at the end of the start case in boot.udev 
which mimic the
force-reload activity plus a small delay the system then comes up fine.  
I'm now trying
to figure out how to get the corrected entries remembered by suse's boot 
process.  With
the hack in place I boot almost normally, I get xdm and everything since 
/dev is properly
populated.  If I take the hack back out it reverts to the broken state.  
Since I can't figure out
what got ttyS0 stuck in there I don't know how to overwrite it.

Now that I am able to boot after a fashion, though, I'm going to pull 
out my last backup and
see if I can find the bogus rule...perhaps you or someone else on the 
list can give me a clue
as to what my erroneous rule might have done.

I miss my Slackware installation.  All this idiot-resistant gui stuff is 
great until something
breaks, then it seems to be impossible to figure out.

       ken




_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: SuSE 10.0 udev stuck with bogus name / rule entry

[Kay Sievers]

On Thu, 2006-06-08 at 14:37 -0400, Ken Ryan wrote:

> Now, when I run udevinfo -q all -p /sys/block/hda/hda1 (for example)
> all of the lines returned look fine EXCEPT the N: line.  It says
> 
> 	N: ttyS0

Try to run:
  udevtest /block/hda/hda1

What does it print?

Kay



_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: SuSE 10.0 udev stuck with bogus name / rule entry

[Ken Ryan]

On Fri, 9 Jun 2006, Kay Sievers wrote:

> On Thu, 2006-06-08 at 14:37 -0400, Ken Ryan wrote:
>
>> Now, when I run udevinfo -q all -p /sys/block/hda/hda1 (for example)
>> all of the lines returned look fine EXCEPT the N: line.  It says
>>
>> 	N: ttyS0
>
> Try to run:
>  udevtest /block/hda/hda1
>
> What does it print?
>
> Kay
>

(please forgive any typos, copying by hand))

# udevinfo -q all -p /sys/block/hda/hda1
F: /block/hda/hda1
N: ttyS0		<------- Huh?
S: disk/by-uuid/33c4c080-aa21-4209-bc03-880042e9af82
E: ID_FS_USAGE=filesystem
E: ID_FS_TYPE=ext3
E: ID_FS_VERSIONE: ID_FS_UUID3c4c080-aa21-4209-bc03-880042e9af82
E: ID_FS_LABELE: ID_FS_LABEL_SAFE#
# udevtest /block/hda/hda1 block
main: looking at device '/block/hda/hda1' from subsystem 'block'
main: opened class_dev->name='hda1'
udev_rules_get_name: add symlink 'disk/by-id/ata-WDC_WD1200BB-00CAA0_WD-WMA8C1127397-part1'
udev_rules_get_name: add symlink 'disk/by-path/pci-0000:00:1f.1-ide-0:0-part1'
run_program: '/sbin/vol_id --export /dev/.tmp-3-1'
run_program: '/sbin/vol_id' (stdout) 'ID_FS_USAGE=filesystem'
run_program: '/sbin/vol_id' (stdout) 'ID_FS_TYPE=ext3'
run_program: '/sbin/vol_id' (stdout) 'ID_FS_VERSION='
run_program: '/sbin/vol_id' (stdout) 'ID_FS_UUID3c4c080-aa21-4209-bc03-880042e9af82'
run_program: '/sbin/vol_id' (stdout) 'ID_FS_LABEL='
run_program: '/sbin/vol_id' (stdout) 'ID_FS_LABEL_SAFE='
run_program: '/sbin/vol_id' returned with status 0
udev_rules_get_name: add symlink 'disk/by-uuid/33c4c080-aa21-4209-bc03-880042e9af82'
udev_rules_get_name: no node name set, will use kernel name 'hda1'
create_node: creating device node '/dev/hda1', major = '3', minor = '1', mode='0640', uid = '0', gid = '6'
create_node: creating symlink 'disk/by-id/ata-WDC_WD1200BB-00CAA0_WD-WMA8C1127397-part1' to '../../hda1'
create_node: creating symlink 'disk/by-path/pci-0000:00:1f.1-ide-0:0-part1' to '../../hda1'
create_node: creating symlink 'disk/by-uuid/33c4c080-aa21-4209-bc03-880042e9af82' to '../../hda1'


A few observations:

- At the time the above were run the system had failed boot by failing to 
open /dev/hda1 for fsck; it dropped to the single-user mode.  The 
currently executing script is boot.localfs.  udevd is running with the
"--daemon" and "-stop-exec-queue" options (according to ps).

- The results of udevinfo reflect the values of the file under 
/dev/.udevdb

- At the time I ran the above commands, there are no /dev/hd* files. 
This applies to the following /dev/files:
 	fd0
 	hd[acg]
 	hd[ag][1234]
 	loop[0-7]
 	ram[0-9]
 	ram1[0-5]
 	input/
 	kmsg
 	psaux
 	rtc
 	hpet
 	fb0

- /dev/md* are present (/dev/md0 through /dev/md31).  Of these md0 and md1 
are actual devices (software RAID1 on hda3+hdg3, hda4+hdg4).

- The entries in /dev/.udevdb corresponding to each of the above all 
include the "N: ttyS0" line instead of their proper names

- If I run "/etc/init.d/boot.udev force-reload" in this state all the 
proper device nodes get created.  Stopping and restarting udev does
not help.

- If I modify /etc/init.d/boot.udev (the SuSE script that starts udev) and 
add to the end of the 'start' case all of the same commands that are in 
the 'force-reload' case plus a sleep of a few seconds /dev gets populated 
properly.  SuSE is able to bring xdm up and run the failsafe root xterm 
(no window manager), but bringing up gnome or kde fails immediately (I 
haven't pursued that bit at all).  If I leave out the sleep after 
force-reload, boot fails with all the characteristics above.

- The logging daemons aren't working but dmesg lists no warnings or 
errors.  The hda/hdg and partitions are getting detected properly by the 
drivers (as far as the boot messages say).  dmesg only shows lines as far 
back as "io scheduler deadline registered", before floppy drive detect.


Thanks, that's all I've come up with so far...

 	ken



_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: SuSE 10.0 udev stuck with bogus name / rule entry

[Kay Sievers]

On Fri, 2006-06-09 at 21:40 -0400, Ken Ryan wrote:
> On Fri, 9 Jun 2006, Kay Sievers wrote:
> > Try to run:
> >  udevtest /block/hda/hda1
> >
> > What does it print?

> # udevinfo -q all -p /sys/block/hda/hda1
> F: /block/hda/hda1
> N: ttyS0		<------- Huh?

That must be a broken rule.

> # udevtest /block/hda/hda1 block
> udev_rules_get_name: no node name set, will use kernel name 'hda1'

The rules in the root filesystem seems to be ok.

Try to run mkinitrd. You may still have the broken rules that create ttyS0
for all devices in the initramfs image.

Kay



_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel