Booting with graphical renderer leaves the system at textual "login:" screen

[Scott James Remnant (Canonical)]

Binary package hint: plymouth

This bug occurs with the Framebuffer and DRM renderer backends (KMS and non-KMS framebuffers), I have not yet been able to replicate it with the text plugin ("Ubuntu 10.04" across the middle of the screen) -- if you can replicate it with that, it is unlikely to be a different bug, so please add details to this one.

You have this bug if after booting, you see a text "login:" screen rather than the X server you were expecting - AND - if pressing Alt-F7 takes you to the X server with no further problems. Your boot has worked, you've just ended up on the wrong VT. If you have further problems, or no X server whatsoever, you have a different but so please file a new one.

Given we end up on VT1, it's strongly likely that plymouth has simply changed the VT on quitting like it does for those who don't have an X server. This is quite odd, because gdm should call plymouth with the --retain-splash argument that prevents this.

[Sindhudweep Sarkar (sindhudweep-sarkar)]

@scott: I see this behavior in lucid post alpha 3 with a radeon x1400 lenovo laptop.

The only difference is when i switch from vt1 -> vt7 and login (in gdm) x/gdm will restart after a few seconds. x/gdm lives after the second login attempt.

[Scott James Remnant (Canonical) (canonical-scott)]

Sindhudweep Sarkar: which version of Plymouth do you have?

[vmc (vmclark)]

I now get the exact error. What information do you need?

$ aptitude show mountall|grep Ver
Version: 2.8
$ aptitude show plymouth|grep Ver
Version: 0.8.0~-13

Updated as of Mar12th.

[Scott James Remnant (Canonical) (canonical-scott)]

vmc: just whether your symptoms match this bug

[vmc (vmclark)]

Scott, Thanks. This bug matches 'to a tee'.

[IKT (ikt)]

Indeed, I have the exact same thing, this has only happened after updating to the latest round of updates.

[Hernando Torque (htorque)]

I'm seeing this too. Noticed something strange:

[14:44] ~ $ cat /proc/fb
0 nouveaufb
1 VGA16 VGA

IIRC they never were there at the same time. Blacklisting 'vga16fb' in '/etc/modprobe.d/blacklist-framebuffer.conf' followed by an 'update-initramfs -c -k all' brought me back to a clean boot to desktop.

[teh603 (darth-giles)]

I had this last night as well, on an Arrandale-based system. Didn't know there was a keystroke to get to the desktop, and it also explains why startx failed with a message saying that X was already running.

[William Shand (williamshand14)]

Yes got put into text login, Alt-F7 got me to the desktop with no further problems

[Doug McMahon (mc3man)]

Had the same behavior this morning using plymouth (Version: 0.8.0~-14, ubuntu-logo), though it wouldn't occur if a disk check took place
removing plymouth, plymouth-x11, rebooting and re-installing both has solved here atm ( thru 4 restarts

[Doug McMahon (mc3man)]

Returned on the 5th restart, I seem to get a few 'good' starts after a re-install, then behavior returns (w/ the fsck exception

[Andreas Modinos (blueturtl)]

Me too.

~ $ cat /proc/fb

0 radeondrmfb
1 VGA16 VGA

If that means anything.

[Wolfgang Kufner (wolfgangkufner)]

Booting the 20100313 64bit lucid iso from hd via grub2 loopback the picture is mixed depending on boot parameters:
"quiet splash": 5 good; 1 buggy
"": 2 good; 4 buggy
"quiet": 3 good; 1 buggy

With the installed lucid (with the newest plymouth):
"quiet splash": 1 good; 1 buggy
"quiet": 1 good; 1 buggy
"": 0 good; 5 buggy

with acer extensa 5630z (intel core2duo generation with intel graphics)

I get for cat /proc/fb:
0 inteldrmfb
1 VGA16 VGA
4 times out of 4, even without the bug

[Noel J. Bergman (noeljb)]

I had this happen today for the first time. A disk check took place, and then I was left at tty1 afterwards.

[teh603 (darth-giles)]

Why was this listed as milestoned for Beta 2 instead of Beta 1? This bug makes Ubuntu completely unusable for people who don't know or don't remember the keystroke that switches them back to the desktop.

[pablomme (pablomme)]

@teh603: probably because beta-1 is due this Thursday and devs can't guarantee to get it fixed in 4 days. People running beta releases should know their way out of trouble, or at least should read the "Known Problems" section of the release notes before jumping in.

[Scott James Remnant (Canonical) (canonical-scott)]

On Sat, 2010-03-13 at 13:49 +0000, Hernando Torque wrote:

> I'm seeing this too. Noticed something strange:
>
> [14:44] ~ $ cat /proc/fb
> 0 nouveaufb
> 1 VGA16 VGA
>
> IIRC they never were there at the same time.
>
Actually it's normal for them to be; as long as fb0 is the nouveau one -
things work out. (vga16fb is later in modules.order for this reason)

Scott
--
Scott James Remnant
<email address hidden>

[Scott James Remnant (Canonical) (canonical-scott)]

On Sun, 2010-03-14 at 09:36 +0000, Andreas Modinos wrote:

> Me too.
>
> ~ $ cat /proc/fb
>
> 0 radeondrmfb
> 1 VGA16 VGA
>
> If that means anything.
>
Useful, thanks. How many monitors do you have attached?

Scott
--
Scott James Remnant
<email address hidden>

[Scott James Remnant (Canonical) (canonical-scott)]

On Sun, 2010-03-14 at 13:16 +0000, Wolfgang Kufner wrote:
> with acer extensa 5630z (intel core2duo generation with intel graphics)
>
> I get for cat /proc/fb:
> 0 inteldrmfb
> 1 VGA16 VGA
> 4 times out of 4, even without the bug
>
You see the "login:" issue on this too?

Scott
--
Scott James Remnant
<email address hidden>

[Scott James Remnant (Canonical) (canonical-scott)]

On Sun, 2010-03-14 at 19:09 +0000, teh603 wrote:

> Why was this listed as milestoned for Beta 2 instead of Beta 1? This bug
> makes Ubuntu completely unusable for people who don't know or don't
> remember the keystroke that switches them back to the desktop.
>
We can release note things like that ;-)

The reason I put Beta 2 is because I don't know what causes this bug
yet, so I can't guarantee fixing it by tomorrow (which is the absolute
last chance cut off point for Beta 1)

Scott
--
Scott James Remnant
<email address hidden>

[IKT (ikt)]

Anything useful we can do?

[Scott James Remnant (Canonical) (canonical-scott)]

On Mon, 2010-03-15 at 14:55 +0000, IKT wrote:

> Anything useful we can do?
>
Not unless you can step through the plymouth code line-by-line while the
X server starts to see why it's invoking a VT switch back to the
original VT when it shouldn't <g>

Scott
--
Scott James Remnant
<email address hidden>

[Scott James Remnant (Canonical) (canonical-scott)]

Actually, there is something you can all do to help!

If you could boot with "--verbose" on the kernel command-line and attach your /var/log/syslog files - that would be very useful!

[Scott James Remnant (Canonical) (canonical-scott)]

(obviously of a boot where you end up at the VT rather than in X)

[vmc (vmclark)]

LOL, so adding "--verbose", is the fix :)

I did that and thought to myself, watch it NOT fail.

Sure enough adding "--verbose", and rebooting 6 times and I cannot get to just VT1 text screen.

Will keep at it until it fails or I do...

[vmc (vmclark)]

Finally, after 8 restarts and one cold boot, I get to text screen VT1
Here's the attached syslog

[Hernando Torque (htorque)]

@#7: For the record, blacklisting 'vga16fb' didn't fix anything - I was just "lucky" to get ten consecutive boots straight to desktop...

Adding '--verbose' results in a black screen (LCD backlight turned off, no blind/remote login possible) - at least the last 20 boots (some restarts, some cold boots).

[Lion1969 (lion-box)]

and here is my syslog with system booting to vt1 after plymouth appears for no more than a couple seconds

[Scott James Remnant (Canonical) (canonical-scott)]

On Mon, 2010-03-15 at 16:48 +0000, vmc wrote:

> Finally, after 8 restarts and one cold boot, I get to text screen VT1
> Here's the attached syslog
>
>
> ** Attachment added: "/var/log/syslog"
> http://launchpadlibrarian.net/40961349/syslog
>
This logs confirms my suspicions.

Scott
--
Scott James Remnant
<email address hidden>

[Scott James Remnant (Canonical) (canonical-scott)]

On Mon, 2010-03-15 at 17:09 +0000, Lion1969 wrote:

> and here is my syslog with system booting to vt1 after plymouth appears
> for no more than a couple seconds
>
>
> ** Attachment added: "syslog"
> http://launchpadlibrarian.net/40962034/syslog
>
Same as vmc.

Scott
--
Scott James Remnant
<email address hidden>

[Wolfgang Kufner (wolfgangkufner)]

>You see the "login:" issue on this too?
All my comments so far about these plymouth bugs are based exclusively on testing with this acer extensa 5630Z notebook (intel core2duo generation with intel graphics) without an external monitor.

Just booted again into the installed system with "quiet splash --verbose" and got the "login:" bug and I get for cat /proc/fb:
0 inteldrmfb
1 VGA16 VGA

[Hernando Torque (htorque)]

FWIW, booting with '--verbose' leaves me with:

[18:59] ~ $ cat /proc/fb
0 VGA16 VGA

[Рустам Second_Fry Губайдуллин (secondfry)]

Before and after "--verbose" option:
krypt@krypt-laptop:~$ cat /proc/fb
0 radeondrmfb
1 VGA16 VGA

[Рустам Second_Fry Губайдуллин (secondfry)]

Just downloaded source, compiled with error on "test" hook, installed and... nothing changed.

[Scott James Remnant (Canonical) (canonical-scott)]

On Mon, 2010-03-15 at 18:00 +0000, Wolfgang Kufner wrote:

> >You see the "login:" issue on this too?
> All my comments so far about these plymouth bugs are based exclusively
> on testing with this acer extensa 5630Z notebook (intel core2duo
> generation with intel graphics) without an external monitor.
>
Graphics probably doesn't make a difference, just how fast your PC can
boot. To confirm, your syslog shows the same result

Scott
--
Scott James Remnant
<email address hidden>

[Jim Rorie (jfrorie)]

Seeing the same thing. One thing I've noticed is that if I Ctrl-Alt-SysReq-B from the console (I know, bad form) the system pretty consistently comes up correctly the next time.

[Steve Langasek (vorlon)]

Based on discussion on IRC today, I understand it's possible to get us booting to the DM consistently, possibly at the expense of also leaving server installs at the splash screen instead of the login prompt. I think this is the right tradeoff for us to make for beta1 (reopening bug #506297), so resetting the milestone; if we end up deferring to beta-2, this should certainly be documented in the errata.

[Peter B P (peterbp)]

Got this bug today after applying recent Updates. Lucid ALpha 3. P5QPL-AM motherboard, Intel chipset.

Tends to NOT happen when cold booting, but consistently happens on reboots. Alt+F7 consistently a useful workaround. (Thanks!)

[Scott James Remnant (Canonical) (canonical-scott)]

There's a new plymouth package in my PPA - https://launchpad.net/~scott/+archive/ppa

Please try it and see whether it fixes this bug for you

[Colin (colinnc)]

The new package fixed the bug for me.

Thanks!

[Kiwinote (kiwinote)]

Same here. Thanks.

[Lion1969 (lion-box)]

fixed for me too
checked on 3 cold boots and 3 reboots with no problem at all
great work!
hope it goes in beta 1

[Launchpad Janitor (janitor)]

This bug was fixed in the package plymouth - 0.8.0~-15

---------------
plymouth (0.8.0~-15) lucid; urgency=low

  * debian/plymouth.upstart, debian/plymouth.plymouth-stop.upstart:
    - Replace the "stop on" clause in plymouth's own job with a job that
      handles stopping plymouth for the normal cases.

      For a display manager that handles the plymouth transition properly
      (gdm & ubiquity only at this point), this simply appears to be running
      until plymouth is stopped by that display manager. This blocks plymouth
      from being stopped for other reasons, e.g. rc2 exiting. LP: #538214.

      For other display managers, and when we reach the end of rc2 or
      other reasons, this simply asks plymouth to quit. We never use
      --retain-splash, if the display manager can't do plymouth transitions,
      we never want to leave the VT in KD_GRAPHICS in case the display
      manager gets things wrong (hi, kdm!). LP: #538524.
  * debian/plymouth.plymouth-splash.upstart:
    - Add LXDM. LP: #526748

  * debian/plymouth.plymouth-splash.upstart:
    - Remove the "starting gdm" bit from the job, since that won't have any
      effect (job is already started by the time gdm is); and as a result of
      the above changes could end up with plymouth --show-splash and quit
      being called together if someone runs "start gdm" themselves.
 -- Scott James Remnant <email address hidden> Wed, 17 Mar 2010 01:19:37 +0000

[vmc (vmclark)]

Strange, but I keep getting 404:
$ sudo add-apt-repository ppa:1024R/4929B943
Error reading https://launchpad.net/api/beta/~1024R/+archive/4929B943: HTTP Error 404: Not Found

W: GPG error: http://ppa.launchpad.net lucid Release: The following signatures couldn't be verified because the public key is not available: NO_PUBKEY 7D0AC3554929B943

[Scott James Remnant (Canonical) (canonical-scott)]

On Wed, 2010-03-17 at 01:48 +0000, vmc wrote:

> Strange, but I keep getting 404:
> $ sudo add-apt-repository ppa:1024R/4929B943
> Error reading https://launchpad.net/api/beta/~1024R/+archive/4929B943: HTTP Error 404: Not Found
>
Err, that's a GPG key, not a PPA :)

 YM add-apt-repository ppa:scott/ppa

but NM, it's in lucid now

Scott
--
Scott James Remnant
<email address hidden>

[vmc (vmclark)]

Thanks, I got it keyed this way:

For future reference...
I added your ppa this way:
$ sudo add-apt-repository ppa:scott/ppa

and sources.list, added:

deb http://ppa.launchpad.net/scott/ppa/ubuntu lucid main
deb-src http://ppa.launchpad.net/scott/ppa/ubuntu lucid main

but still get 404 on that ppa:

Err http://ppa.launchpad.net lucid/main Packages
  404 Not Found
Err http://ppa.launchpad.net lucid/main Packages
  404 Not Found

Hopefully an update will come through.
Thanks for all your hard work. Just in time to squeeze in before beta.

[teh603 (darth-giles)]

Doesn't seem to be in the suppository, or at least not as of now. Should I give it a few minutes more and try later?

Also: Did this make it in time for the Beta 1 freeze?

[teh603 (darth-giles)]

Ok, got it and installed.

On initial shutdown, I saw a "broken pipe" for a few moments.

On restart, it flickered thru several text screens and showed an fcsk check (sp?) before finally dumping me into the bootsplash. Boot went normally from there.

Next shutdown showed some text before the closing splash. No "broken pipe."

Restart didn't show the bootsplash at all (!), but went ok.

Shutdown showed normal splash again, no "broken pipe."

Cold startup (not restart) again showed no bootsplash, but had no overwhelming issues. Lots of text, though.

Following shutdown showed "broken pipe" for a few moments before turning off.

Massive improvement over before, but its still a little rough. The bootsplash still doesn't show as often as its supposed to for some reason.

[vmc (vmclark)]

I finally got the PPA in and it now works without going to textual VT login.

I also got a ubuntu logo on first couple of boots and then nothing on the next 10. I always get a ubuntu logo on reboot and shutdown though, if any of that matters.

[Scott James Remnant (Canonical) (canonical-scott)]

On Wed, 2010-03-17 at 03:19 +0000, teh603 wrote:

> Massive improvement over before, but its still a little rough. The
> bootsplash still doesn't show as often as its supposed to for some
> reason.
>
We have a switch to smooth off the messages appearing over top; will be
doing that for Beta 2.

Scott
--
Scott James Remnant
<email address hidden>

[Bernhard (b.a.koenig)]

An update from 2 weeks ago fixed this for me, but today I had another interesting variation: after checking the disk during boot, it reverted back to tty1.

[Peter B P (peterbp)]

Seems to be fixed after applying todays Updatemanager patches. Tested with bot cold boot and reboot.

Thanks for the quick fix!