Bug 186164

I'm seeing the exact same problem on a Toshiba Tecra S2.  FC5t3 did not have 
this problem for me.

I used the rescue CD to mount my new system and changed /etc/rc.d/rc.sysinit to
"sh -x /sbin/start_udev", in an effort to see what's [not] happening.

Here is the last lines on the console before freeze:
---8<----
loop=176
test -d /dev/.udev/queue
usleep 100000
--->8----
As you can imagine, there were lots of "loop=177 ..." friends above it.

I thought I'd report my findings in case that helps anyone, or if those CC'd on
this bug want to try on their system, too.

Added acpi=off to the kernel cmdline and was able to get past the lockup.

you could also boot from the rescue CD and set in /etc/udev/udev.conf
udev_log="debug"

I suspect, that a kernel module is causing the hangup, so look out for the last
"modprobe" to be executed.

Same problem here on a NEC P550 laptop.
udev hangs at first boot after install.
And I confirm that added "acpi=off" option solved the udev hang.
Unfortunately the kernel crashes a bit later (general protection fault).
But it seems to be another problem.

I edited /etc/udev/udev.conf and set: udev_log="debug"
The last log message I see before the hang is:

udevd-event[1104]: wait_for_sysfs: wait for
'/sys/devices/pci0000:00/0000:00:1d.2/usb4/4-0:1.0/bus' for 20 mseconds

any "modprobe" message before that?

(In reply to comment #7)
> any "modprobe" message before that?

I'm not trying to be naive about this, but what is the recommended way to
capture the console messages since once it locks up, the scrollback is gone?

I do not have a serial port on this machine, otherwise I'd ask about booting
over serial line (I've heard that's possible, anyway).

I tried the acpi=off as suggested, and I, too, receive a kernel panic but I
cannot confirm its impact on udev as my system never makes it that far (with
acpi=off).

I also edited the udev.conf and there were no modprobe messages on the screen at
hang-time. If there were modprobe messages at all, they were shoved off-screen
by the plethora of udevd-event messages (similar to comment #6).

I have the same problem on the identical hardware with FC5 as described in

http://www.redhat.com/archives/fedora-devel-list/2006-March/msg00456.html

ok, next step... comment out all modprobe lines from:
/etc/udev/rules.d/*
try to boot

I played a bit with udev rules this morning:

1. I commented all lines containing modprobes in *.rules files:
   --> it doesn't solve the problem. udev is still hanging.

2. I removed some udev rules file. I remembered pcmcia daemon made 
   this same laptop crash at boot with FC4, I add some problem with 
   the sound card too and I don't own a Wacom tablet: pcmcia.rules,
   wacom.rules and alsa.rules removed.
   --> problem seems to be solved! It seems PCMCIA on this laptop is
       supported.

3. I added back wacom.rules and alsa.rules to confirm the above
   statement:
   --> udev is ok, but HAL daemon crashes later in boot process

4. I removed alsa.rules again.
   --> udev still ok and HAL daemon starts without problem.

Conclusion:

PCMCIA seems to be the cause for udev hang.
There is also a problem with alsa but it may be unrelated.

this is NOT an UDEV hang... this is the KERNEL which hangs ...

please try to isolate the line in 60-pcmcia.rules, which crashes the kernel.

*** Bug 186386 has been marked as a duplicate of this bug. ***

(In reply to comment #12)
> this is NOT an UDEV hang... this is the KERNEL which hangs ...

When I say 'udev hangs', I only mean "the boot process halts after the 'Starting
 udev' message is displayed on screen and the machine seems to hang". 
No need to yell, I'm no specialist. :)
I'm just trying to help and make this kernel boot cleanly on my laptop.

As asked in comment #13, I re-installed 60-pcmcia.rules and tried to find the
line which causes of the hang:
  - comments every ACTION lines
  - loop:
      - uncomment one more ACTION line
      - reboot

ACTIONs 1 + 2 : OK
ACTIONs 1 + 2 + 3 : udev OK, but hald crashes 
                    (may be hald needs action 4 if 3 is executed)
                    (I can provide the dmesg output if needed)
ACTIONs 1 + 2 + 3 + 4 : udev OK, hald OK
ACTIONS 1 + 2 + 3 + 4 + 5 : machine hangs

Hope this helps.

To avoid any confusion, action 5 is the last one in 60-pcmcia.rules. It is:

# if this is a PCMCIA socket which needs a resource database,
# pcmcia-socket-startup sets it up
ACTION=="add", SUBSYSTEM=="pcmcia_socket", \
                RUN+="/sbin/pcmcia-socket-startup"

I realized it wasn't very clear in my previous comment.

ok...

could you try to run as root:
# /sbin/pcmcia-socket-startup

if the laptop hangs, could you try to update pcmciautils to
ftp://people.redhat.com/harald/pcmciautils-012-1.i386.rpm

and rerun:
# /sbin/pcmcia-socket-startup

I've tried it. Laptop hangs.
Downloaded pcmciautils. Now it doesn't hang.

As expected, running /sbin/pcmcia-socket-startup as root hangs the computer.

I installed your updated package and ran succesfully the new
/sbin/pcmcia-socket-startup: no crash.

I checked that every lines in 60-pcmcia.rules that no lines are commented.
It rebooted like a charm: no hang.

It seems you found the cause.

(BTW, I can't tell you if pcmcia actually works as I have no pcmcia card at
disposable here.)

Cool! Good News! Thanks for all the help!

ok, please also test:
ftp://people.redhat.com/harald/pcmciautils/012-0.FC5.1

for those using pcmciautils-012-1.i386.rpm:
# rpm -Uvh --oldpackage

No problem here with pcmciautils-012-0.FC5.1.i386.rpm.
This rpm seems as good as the previous one.

*** Bug 185989 has been marked as a duplicate of this bug. ***

Please test:
http://www.redhat.com/archives/fedora-test-list/2006-March/msg01542.html 

I've installed pcmciautils-012-0.FC5.2.i386.rpm from comment #25.
It works well on my laptop.

Same story (problem and workaround) with Acer Travelmate 4100.

This thing must be a total showstopper for an average user.

After installing the RPM from comment #18, my Acer Travelmate 4600 boots up
fine, too. Thanks!

udev freezes here as well, and this is not on a laptop with PCMCIA.

If I supply the "acpi=off" switch through GRUB, then the freeze won't happen,
but udev will complain about "No devices found" and there's a message saying
"ERROR: No arguments allowed with -i". There are no apparent ill effects from
these messages, as far as I can tell as a mere user. Not other than losing ACPI
functionality, that is.

I don't know if this could be relevant, but when I first tried to do an FC4 -> 5
upgrade, and later when I out of frustration wiped everything and made a fresh
install, I got strange messages seemingly pertaining to some RAID issue. That's
weird because I don't have any RAID controller. The message, right before the
mid-installation lockups, was "ERROR: pdc: wrong # of devices in RAID set"
followed by something about "pdc_<some seemingly random combination of letters>".

My affected hardware:
Asus A7M-266D mobo (AMD-760MPX) w. 2x Athlon MP, 1 GB RAM
Fedora on sda, connected to a Sym 53C896-based SCSI controller
hda (Windows) + hdb (storage) connected to on-board IDE (AMD-768)
CMI8738 on-board audio
VT6105 [Rhine-III] eth0
3c905C eth1
NEC USB2 PCI
Radeon 9600XT (mis-identified as RV350)
USB mouse, trackball, joystick, scanner, webcam
PS/2 keyboard with built-in serial touchpad

first, i got no problem with udev or pcmcia and maybe it is not the right place
here, but i can confirm, that i see the line
"ERROR: No arguments allowed with -i" in fc5-devel since 2-3 month and in FC5.
but only on my amd 64 with i386 and x64_86 packages, not on an pentium 3 or my
notebook.
i search the bugzilla for a while, daily reading the mailing lists, no results,
only a discussion/bug about env in fc4-test?.

it seemed that i am quite alone or most people boot with rhgb.
so, since this line doesn't affect my work and remembering the discussion about
the "reorganisation" of bugzilla and the slip of the fc5 release day,i decided
to wait for fc5.
i still see that line. 
i search my /etc, no results.
i played with the thought to fill a bug, but on what component ???

nevertheless, FC5 is quit good !!!
thx to the developers, (maybe) the whole day hearing/reading, what is NOT
functioning. i know !

i think, the last 2 sentences have to be broadcasted or duplicated in, maybe,
the bugzilla, but on what component ...

on my thoshiba M30x : default install hang on udev.
after apply the new pcmcia rpm from comment #18 : now it boot 
but hald coredump and now computer is unable to start X11 !!
I remove the hald start from init ... and it boot's OK (without hald ;-)

on my thoshiba M30x : default install hang on udev.
after apply the new pcmcia rpm from comment #18 : now it boot 
but hald coredump and now computer is unable to start X11 !!
I remove the hald start from init ... and it boot's OK (without hald ;-)

Confirmed for my Acer Travlemate 8202 WLMi. 
The rpm in testing solves the problem.

Thanks a lot

Let me add just a twist to this problem. I installed FC5 earlier this week. Was 
running FC4 just fine. FC5 installed with no problem and I was able to run it 
for several days. Then, one morning, I got the exact same 'starting udev:' 
problem......after being successful for days that is. I re-installed FC4 and 
was not able to boot. Then, I reinstalled FC5 and witnessed the same udev 
situation. However, I left the machine for at least 3 hours, and when I 
returned, the machine had booted and was at a user prompt. I logged on fine, 
but the machine is now running slower than my old 8086 DOS machine.

Dell Inspiron 4000.....Any thoughts or ideas?

Doug

I see the same problem on a new hp dv800t CTO laptop.
At the starting udev prompt, the system hangs (completely unresponsive...can
only power down by holding power button)

Seems to happen when udevstart is called from the /sbin/start_udev script. 
Commenting this out allows the system to boot (obviously not very well without
the /dev/ entries)

Will run strace or enable udev logging at a later time.  

Also seeing a bunch of PCI allocation errors just before the "...press "I" for
interactive startup" is displayed, along with a temperature error (0 degrees).

Maybe there is a problem with the drivers for intel 945pm chipset on this machine?

Was able to use FC4 (final) with a different problem (pcmcia service would hand
in a similar manner), but with that service disabled, was able to get everything
but the accelerated graphics to work (although the generic 'nv' driver looked
good enough for now...).

This is still an issue for myself. I have removed the pcmciautils package 
completely. I am not using a laptop. 

Removing the udev package allowed my machine to boot to a bash login 
(obviously with errors). I will try to get a sysreport of the machine

Created attachment 127189 [details]
sysreport output

I have installed udev from FC4 and now my system boots to the desktop, my sound
card is nowhere to be seen, but I will take what I can get for now. 

Let me know if any further info is required. 

I also find it hangs on boot.
- upgrading pcmciautils did not help
- kernel-2.6.16-1.2080_FC5 has the bug
- kernel-2.6.15-1.2054_FC5 is fine


Hardware: IBM T42p

More information on my t42p report above:

In my laptop there is a COnexant HSF modem. Downloading the stacksize=16k kernel 
from Linuxant fixed it.

It looks to me that there is a whole slew of potential causes for this bug.

Created attachment 127223 [details]
Sysreport of my PC

I hope this can help

I have same problem here.
I use a normal desktop (not notebook) with Asus K8V-X mainboard, Geforce Fx5200
with 256MB memory, 512 MB DDR RAM and 80 Gb PATA hard disk.

The systemm boots when i use the 

pci=off 
acpi-off

both options in kernel boot line.
Obviously, boot without AGP and PCI devices.

At this, when i clear /etc/udev/rules.d/ files, the system boots in text mode.
No graphic support available.

Look at sysreport file attached.

I confirm upgrading pcmciautils fixes the udev problem for me:

Problem
============
Sony Vaio VGN FS115S
Fresh Fedora Core 5 install
Kernel 2.6.15-1.2054_FC5
On boot it would not get past "Starting udev"
It would boot if I set the kernel flag acpi=off 
But no access to battery status, a real pain on a laptop!

Solution
=============
rpm -Uvh pcmciautils-012-0.FC5.1.i386.rpm
ftp://people.redhat.com/harald/pcmciautils/012-0.FC5.1
Now boots fine and everything is working

Since we have multiple indications of success, do we want to mark this bug as FIXED?

No, we have people that do not have a fix. Like myself for example

Me too.

By the way, I configured my X Server successfully, but my other devices doesn´t
work.

i hope someone help us.

Regards.

I agree with Michael Knox.....we need a definitive answer before this should be 
closed. Unfortunately, some of us, (like me), don't have the Linux savy to 
employ these 'workarounds'. I have tried some of the remedies posted, but still 
have problems. I am re-installing FC4 on my laptop as we speak.

Doug

FWIW, the pcmciautils-012-0.FC5.1.i386.rpm package solved my problem with Acer
Aspire 1692. The report is at
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=187798 and is clearly a
duplicate of this one.

*** Bug 187798 has been marked as a duplicate of this bug. ***

For those who have still problems, _please_ open a seperate bug report along
with your possible workarounds.

*** Bug 187956 has been marked as a duplicate of this bug. ***

I found this problem and after a lot of reinstall's I found that it didnt hang 
when the monitor was plugged into the onboard video port, and the video card 
removed. Im not sure if the card wasn't supported but it worked fine with FC4

Robin Taylor: which video card?

try disabling selinux by pressing a on grub and appending selinux=disabled to
the kernel parameters. For some reason selinux came with enforced policy and the
labels on the files were wrong and boot stopped. 

The card is a "Sparkle 8MB PCI SIS 6326", it wouldnt install in graphical form
either, the X server wouldnt load

I've had the udev bug since I did the yum upgrade thing, so I blamed myself for
awhile. Mine complains about missing /etc/fstab.sys. I don't beleieve I recall
having that file in the 'old days', so it's a new one on me. <g> 

I'm trying the pcmcia rpm thingie, but it doesn't seem likely that it would be
the problem, for me. Ric

Still no relief... how do I generate /etc/fstab.sys? The pcmcia update did not
help. What do you need of me to get some sort of resolution? I want to help... Ric

Rickey Moore, I don't know of any /etc/fstab.sys... What is this? Where does
this come from?

Having similar problems with Lenovo N100. Haven't tried the RPMs yet - not quite
sure how to add rpms when the computer doesn't boot.

(In reply to comment #59)
> Having similar problems with Lenovo N100. Haven't tried the RPMs yet - not quite
> sure how to add rpms when the computer doesn't boot.

I booted with the installer disc and activated rescue mode (linux rescue).
If your network card works in rescue mode, just do a "chroot /mnt/sysimage" at
the prompt and then a "yum update pcmciautils"

After this, the boot should work

Yup - that worked. Now the Lenovo N100 no longer hangs on udev. Thanks.

I tried this, yum reports that I already have this installed... so no relief. :(

This bug report has practically developed its own forum here. Can we assign this
is someone else or _close_it_ so I'm not getting CC'ed on all this tech support?

  Thanks,
  -- /v\atthew

(In reply to comment #54)
> try disabling selinux by pressing a on grub and appending selinux=disabled to
> the kernel parameters. For some reason selinux came with enforced policy and 
the
> labels on the files were wrong and boot stopped. 

This solved my long delay at starting udev...

More info about that bug here: 
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=189856

Thanks

   Starting udev on my system results in a shutdown, as if I switched off the 
power, immediate with no warning. It is a pentium 4, 3.2 GHz, 1 Gig ram, Soyo 
motherboard, two large hard drives, USB optical wheel mouse, LCD screen.
   I shall try the above fixes, will keep in touch as to progress.
TTYL,
Rob Linklater.

I think the initial bug report was fixed with the pcmcia utils update, so the
bug IMHO should be closed as resolved.
Most of the other comments should belong to other (new?) bugs

RE: #58
Sorry, missed that question. When I boot the screen says /etc/fstab.sys missing.
Dmesg nor /var/log/messages say nothing about this.  Strange. I do have /etc/fstab. 

The "fstab.sys" problem is not a udev problem... you may google a bit and find,
e.g.:
http://www.linuxquestions.org/questions/showthread.php?p=2191333
Rickey Moore, your problem is in the initrd.

I erased 60-pcmcia.rules and alsa.rules in my rules.d subdirectory, but my 
system still shuts down when it reachea the " starting udev " line. Will 
downloading the fix file work if this test didn't?

I am going to try restoring FC1 for now, I miss having my Linux up and having 
to use WinXP to do everything. I'll watch for a future fix of this bug, so that 
I can try upgrading again in the future.
TTYL.

Rob, please have a look at:
http://www.redhat.com/archives/fedora-test-list/2006-March/msg01542.html

Thanks Harald...I've re-installed Fedora Core 1 and it's working well, although
all my email addresses, etc are gone. I looked at the above message, the fix it
refers to is no longer available. The oldest file is April...
Anyway, I have a desktop unit with no PCMCIA stuff on it. Thanks for trying to help.
TTYL,
Rob.

oops, should be:
https://www.redhat.com/archives/fedora-announce-list/2006-April/msg00007.html

(In reply to comment #73)
> oops, should be:
> https://www.redhat.com/archives/fedora-announce-list/2006-April/msg00007.html

Still doesnt fix it for me. Running on a T-series Lifebook from Fujitsu-Siemens.
Followed the instructions for moving aside the start up script, that worked, got
a full system up, patched pcmciautils as described, rebooted and still get:

Starting udev: Wait timeout. Will continue in the background [FAILED]

Then a hang. Holding in the power button is the only way out.

Rozak, you have to play around with the ioports in /etc/pcmcia/config.opts and
exclude the port which makes problems.

AFAICT, this is a kernel-level issue or a bug in udev. I've got a pci wireless
NIC (00:08.0 Ethernet controller: Marvell Technology Group Ltd. 88w8335
[Libertas] 802.11b/g Wireless (rev 03) 00:08.0 0200: 11ab:1faa (rev 03)) in a
desktop workstation that <b><i>intermittently</i><b> (even better) hangs when
trying to boot kernel-2.6.17-1.2157_FC5. It gets to the point where it's
starting udev then this after about 20 seconds:
Starting udev: Wait timeout. Will continue in the backgroun[FAILED]
and appears to completely hang. acpi=off didn't help.

Same exact hardware, and i can boot just fine via kernel-2.6.15-1.2054_FC5
(haven't tried any other kernels yet). Then when i shut down and try to boot
into the most recent kernel, it happens again. When i pull the card out i can
boot fine via most recent kernel. This kinda bites. I'm hard-pressed to envision
a simple workaround at this point. I think i'll try the previous kernel release
for now and see what kind of results i get.

I am able to boot with kernel-2.6.16-1.2133_FC5.i686.rpm, but every kernel
revision package after this one seems to manifest the udev hang problem.
Hopefully this helps narrow down the source (pun intended:) of the problem.

Gabriel, kernel problem with the module.

Created attachment 133805 [details]
udev hang - sysreport

Hi
I get same problem - Starting udev .... [FAILED] then hang - on desktop
workstation. kernel-2.6.16-1.2133_FC5.x86_64 is fine, later kernels
(e.g kernel-2.6.17-1.2157_FC5.x86_64) give problem. Attaching sysreport.
Thanks.

I ran into this udev/kernel hang problem.  My machine:

Acer Aspire 3102WLMi
AMD Sempron 3200+
ATI Radeon 1100 video
512 MB RAM
Internal 802.11 b/g wireless, internal card reader, etc.

I tried the solution stated here (updating pcmciautils) and that did not help.  
However, I figured that as long as I could chroot via the rescue CD, I might as 
well update everything while I was at it.  I simply did a 'yum update' to the 
stock FC5 install.  *THIS* worked; I am now able to boot.  I realise that with 
such a sledgehammer solution it isn't possible to determine which package 
update did the trick, but in my case it wasn't pcmciautils.

I ran also into the udev/kernel problem

My system is: 
Fedora Core 5 x86_64 on an Acer Aspire 5102WLMi
AMD Turion 64 x2
ATI Radeon Xpress 1100 (fc recognize it as a Radeon Xpress M200)

I tried to install the recommended pcmciautils package above without success.
Then I downloaded the kernel from updates on fedora.redhat.com and install it
under rescue mode (chroot /mnt/sysimage) with the command rpm -i
kernel-2.6.x-x.x86_64.FC5.rpm (or something like that). I reboot with success.

Patrik Andersson

(Just another data point if someone is trying to track down this problem.)

I upgraded to FC5, from FC4, a week ago, and it has been running fine, yet in
this morning's boot it hung just after the udev message.

Yesterday's only change was re-detecting the sound card (I have two sound cards,
an onboard via82xx which has never worked (hardware problem I believe) but seems
to keep becoming the default, and a cmipci). However removing alsa.rules from
/etc/udev/rules.d made no difference (I also removed the pcmciautils rpm - this
is a desktop machine - and that also made no difference).

The fix was to add acpi=off to the grub boot settings.
I'm using 2.6.17-1.2174_FC5 kernel.
HTH - if you want more information let me know.

Today I experienced udev hang after system crash.
I added "udevtimeout=3" to kernel commandline.
The udev timeouted, and system started.
Once the system was started, the system was started well without "udevtimeout=3".

Additional info (following from comment #83): acpi=off only worked for a couple
of boots. Then I had to use "acpi=off apm=off" to get it to boot.
Then after I installed the kernel update (2.6.17-1.2187_FC5) even that didn't work.
The solution I found was to remove 51-hotplug.rules from /etc/udev/rules.d/.
After that FC5 boots, even without the "acpi=off apm=off" parameters; I've had 6
perfect reboots in a row - a personal best since installing FC5 ;-).
Incidentally my USB scanner still gets auto-detected when I plug it in, so it
doesn't seem hotplug.rules was required? Or it is for some other functionality I
have no need for?
If you need more info let me know the commandline to run.

I have the same problem - system hangs after " Starting udev: ". My PC has the
foll. config -
AMD Athlon64 2800+ processor
Seagate SATA hard disk 
ATI Radeon X200 (onboard)

(In reply to comment #81)
Hi 
my system would also not boot after installing FC5 (at the hald message).
Thanks for your tips in comment #60 and #81. I hadn't realised that the 
downloads of the ISO images are so outdated!
I also managed to do a 
yum update 
from the rescue CD as described. 1 Gb download!
Still didn't work.
Tried again
got an error during update that named.conf bind... conflicts with caching-
nameserver...
So I yum remove bind and then again a yum update
another 1Gb download!
If my machine wouldn't be so heavy I probably would have thrown it all the way 
to Redhat.
At last another yum update told me there is nothing else to update.
Reboot and IT WORKED!
Thanks guys, I couldn't have done it without you!

my system is:
Dell Poweredge 2900 with one dual core Xeon, 2 HDD( SAS/SATA), 2GB RAM

After installing FC6 on my system udev would hang and then the kernel would
abend. One work a round I found that works on a Dell DX370 and my homegrown
Intel-based system (VIA Technologies chipset) is to comment out the line:

ACTION=="add", SUBSYSTEM=="?*", MODALIAS=="?*", RUN+="/sbin/modprobe $modalias"

in /etc/udev/rules.d/50-udev.rules. This has also worked for co-workers
Intel-based workstation. Not sure of all the ramifications of soing this but,
the system boots up normally and all sub-systems appear to be working correctly.

Please debug which module is causing the system hang:
Create /sbin/modprobe.sh:
#!/bin/sh
/sbin/modprobe -v -s $@ &>/dev/console

# chmod 0644 /sbin/modprobe.sh
and change the rule to:
ACTION=="add", SUBSYSTEM=="?*", MODALIAS=="?*", RUN+="/sbin/modprobe.sh $modalias"

It seems that the problem is with my Colorgraphic Predator LT 4 Video adapter. I
removed the card from my system and it booted up. Although, I couldn't see the
message for it scrolled off the screen. I'll see if I can it slow down...

The problem appears to be with the i2c_savage4.ko module. The Colorgraphic LT 4
Video adapter is a 4 port vga adapter that has worked under previous kernels.
Any ideas?

Created attachment 144343 [details]
Abend messages while attempting to load 12c-savage4 module

add to /etc/modprobe.conf:
blacklist i2c_savage4

and open a new bug against component kernel with the backtrace.

Hi, I have tried to install FC6 on my new computer and it hangs on "starting
udev". I have packard bell computer with IntelCore 2 Duo E6300 processor and
first I tried with a Nvidia Geforce 7300SE graphics card and also with a ATI
Radeon X1300 Pro, but the same problem with both. My computer hangs and the
screen is turned off.