suddenly no /dev/hda3 after update?

Message

alastairII · Post by **alastairII** » Sun Jan 22, 2006 2:36 am

Since this seems to be a problem in Portage, perhaps one of the mods should move it to the Portage forum.

glober · Post by **glober** » Sun Jan 22, 2006 3:35 am

mwojc wrote:I think, you should recompile your kernel with the options:
Code: Select all
File systems --->
  Pseudo Filesystems --->
    [*] /dev file system support (OBSOLETE)
      [ ]   Automatically mount at boot
where 'Automatically mount at boot' HAVE TO BE DISABLED. See http://www.gentoo.org/doc/en/udev-guide.xml

This change in my kernel configuration helped, and the usual names of the devices are still in my /etc/fstab...

I tried this and it doesn't work for me. When I uncheck "Automaticaly mount at boot" X doesn't start because there is a problem loading the mouse, saying that "core device is missing" or some such message.

Genone · Post by **Genone** » Sun Jan 22, 2006 7:15 am

Did anyone file a bugreport about this yet?

glober · Post by **glober** » Sun Jan 22, 2006 9:40 am

created bugreport #119894 just now

Paapaa · Post by **Paapaa** » Sun Jan 22, 2006 1:15 pm

glober wrote:created bugreport #119894 just now

I added my case (x86/stable) there:

- I'm using baselayout-1.11.14-r2
- I'm using udev-079-r1
- Everything works fine with gentoo-sources-2.6.14-r5

I get the error only with gentoo-sources-2.6.15-r1. So this might be a kernel issue also?

[vector] · Post by **[vector]** » Sun Jan 22, 2006 10:04 pm

Exaclty same problem, all works fine with baselayout-1.11.14-r2, udev-079-r1 AND kernel-2.6.14-r5 but boot fails with

Code: Select all

/dev/hda6 no such file or directory

if i try kernel-2.6.15-r1.

glober · Post by **glober** » Mon Jan 23, 2006 3:38 am

glober wrote:
mwojc wrote:I think, you should recompile your kernel with the options:
Code: Select all
File systems --->
  Pseudo Filesystems --->
    [*] /dev file system support (OBSOLETE)
      [ ]   Automatically mount at boot
where 'Automatically mount at boot' HAVE TO BE DISABLED. See http://www.gentoo.org/doc/en/udev-guide.xml

This change in my kernel configuration helped, and the usual names of the devices are still in my /etc/fstab...
I tried this and it doesn't work for me. When I uncheck "Automaticaly mount at boot" X doesn't start because there is a problem loading the mouse, saying that "core device is missing" or some such message.

It turns out that disabling 'Automatically mount at boot' does fix the problem with using a newer baselayout. I had to change my xorg.conf file form /dev/mouse to /dev/input/mice to take care of the X problem, but after that I emerged baselayout 1.11.14-r2 , and all is working now. Thankyou mwojc

Paapaa · Post by **Paapaa** » Mon Jan 23, 2006 12:21 pm

glober wrote:It turns out that disabling 'Automatically mount at boot' does fix the problem with using a newer baselayout. I had to change my xorg.conf file form /dev/mouse to /dev/input/mice to take care of the X problem, but after that I emerged baselayout 1.11.14-r2 , and all is working now. Thankyou mwojc

Yes, but you are using a very old kernel without udev. That option is no longer present in newer kernels.

There is, however, still the issue with 2.6.15-r1 and latest udev. I'll report a bug if noone has done that already.

Basically:

baselayout-1.11.14-r2 + udev-079-r1 + gentoo-sources-2.6.15-r1 == won't work.
baselayout-1.11.14-r2 + udev-070-r1 + gentoo-sources-2.6.15-r1 == works.

baselayout-1.11.14-r2 + udev-079-r1 + gentoo-sources-2.6.14-r5 == works.

I don't know if baselayout version has anything to do with this.

minaguib · Post by **minaguib** » Mon Jan 23, 2006 6:10 pm

I've just updateded bug #119894 with some more details regarding this.

chance2105 · Post by **chance2105** » Mon Jan 23, 2006 7:19 pm

Yeah .. the bug on b.g.o was closed:

------- Comment #10 From Greg Kroah-Hartman 2006-01-22 21:03 PST [reply] -------

That means that you are using devfs (or were using devfs), so why
were you trying to use udev?

Anyway, am closing this.

This is totally a bug somebody needs to look at. It is not a "mystery" of "misconfiguration" bug. DEVFS isn't even an *option* in newer kernels.

minaguib · Post by **minaguib** » Mon Jan 23, 2006 8:09 pm

chance2105 wrote:Yeah .. the bug on b.g.o was closed:
------- Comment #10 From Greg Kroah-Hartman 2006-01-22 21:03 PST [reply] -------

That means that you are using devfs (or were using devfs), so why
were you trying to use udev?

Anyway, am closing this.
This is totally a bug somebody needs to look at. It is not a "mystery" of "misconfiguration" bug. DEVFS isn't even an *option* in newer kernels.

Re-check the bug report.

Enabling HOTPLUG support in the kernel fixed it for me and at least one other person.

Paapaa · Post by **Paapaa** » Mon Jan 23, 2006 9:24 pm

minaguib wrote:Enabling HOTPLUG support in the kernel fixed it for me and at least one other person.

Is that a proper fix or a workaround? Why did it work without HOTPLUG in 2.6.14 and why is it now required? Will an updated udev remove the need for HOTPLUG?

Paapaa · Post by **Paapaa** » Mon Jan 23, 2006 10:26 pm

Paapaa wrote:Is that a proper fix or a workaround? Why did it work without HOTPLUG in 2.6.14 and why is it now required? Will an updated udev remove the need for HOTPLUG?

This is from http://www.kernel.org/pub/linux/utils/k ... /udev.html:

It (udev) requires a 2.6 Linux kernel with CONFIG_HOTPLUG enabled to run. It is recommended that you also have the Linux Hotplug scripts installed, but it is not necessary for it to work properly.

So yes, Hotplug support is required to be able to use udev. I wonder why udev worked without Hotplug in 2.6.14?

chance2105 · Post by **chance2105** » Mon Jan 23, 2006 11:46 pm

Code: Select all

chance@hoss / $ cat /usr/src/linux-2.6.15.1/.config |grep HOTPLUG
CONFIG_HOTPLUG=y
chance@hoss / $ uname -a
Linux hoss 2.6.15.1 #3 SMP PREEMPT Tue Jan 17 11:24:22 CST 2006 sparc64 sun4u TI UltraSparc II  (BlackBird) GNU/Linux

Well, it's enabled in my config. Although it seems unlikely, maybe this is an exception for my architecture (new udev+baselayout not working with 2.6.15 kernels).

chance2105 · Post by **chance2105** » Tue Jan 24, 2006 12:13 am

Ok, so I reviewed everything in my setup, looking at http://www.gentoo.org/doc/en/udev-guide.xml

The following are enabled in my kernel config:

Code: Select all

General setup --->
  [*] Support for hot-pluggable devices

File systems --->
  Pseudo filesystems --->
    [*] /proc file system support
    [*] Virtual memory file system support (former shm fs)

.. exactly as listed. I can't enable DEVFS because it's non-existant in this kernel.

The guide then goes on to list "Known Issues: Missing device node files at boot", and gives as an exaple /dev/null and /dev/console missing giving trouble on boot. So I followed the next little snippet to check what /dev/ nodes are available at boot:

Code: Select all

# mkdir test
# mount --bind / test
# cd test/dev
# ls

.. and all /dev/sdaX nodes are there. I noticed when I had trouble booting, that in /dev/ I had sda10-sda15, but not sda1-sda9.

The other section deals with coldplug and hotplug. I have neither installed on my system, but if I'm correct, neither of them have any bearing on the root filesystem mounting (is this true?).

Other than that, I'm not sure what to check out. Am I missing something obvious?

lost-distance · Post by **lost-distance** » Tue Jan 24, 2006 12:01 pm

This morning I also found my system to be broken when booting up. I quickly determined that the /dev/hda and /dev/hda[1-4] nodes were missing. These are precisely the nodes present under /sys/block/hda. Strangely, all other /dev/hd* nodes were present, even /dev/hda5 onwards. I use udev with Linux 2.6.15.

I fixed my system by setting RC_DEVICES="static" in /etc/conf.d/rc and thus falling back to static /dev nodes.

What had changed? Yesterday I had updated udev to udev-079-r1. So my feeling is that udev-079-r1 is the culprit.

Paapaa · Post by **Paapaa** » Tue Jan 24, 2006 12:41 pm

lost-distance wrote:This morning I also found my system to be broken when booting up. I quickly determined that the /dev/hda and /dev/hda[1-4] nodes were missing. These are precisely the nodes present under /sys/block/hda. Strangely, all other /dev/hd* nodes were present, even /dev/hda5 onwards. I use udev with Linux 2.6.15.

Just to make sure: do you have Hotplug support enabled in kernel?

chance2105 · Post by **chance2105** » Tue Jan 24, 2006 1:19 pm

lost-distance wrote:This morning I also found my system to be broken when booting up. I quickly determined that the /dev/hda and /dev/hda[1-4] nodes were missing. These are precisely the nodes present under /sys/block/hda. Strangely, all other /dev/hd* nodes were present, even /dev/hda5 onwards. I use udev with Linux 2.6.15..

That sounds strikingly familiar to my post immediately prior to yours -- the /dev/ nodes that are needed are missing, but the *others* are present. Do check to make sure CONFIG_HOTPLUG is enabled in your kernel, and follow the advice in the Gentoo UDEV guide.

I haven't tried it yet, but this evening when I get home (gah, the day job) I'll give my system a shot with a 2.6.14 kernel, to see if there's any difference for myself -- others have reported it works, but the fix others are using on their systems (enable CONFIG_HOTPLUG) with 2.6.15 doesn't help me since it was already enabled ..

lost-distance · Post by **lost-distance** » Tue Jan 24, 2006 3:08 pm

My 2.6.15 kernel didn't have CONFIG_HOTPLUG set. So I set it, and udev-079-r1 now works properly. Thanks guys.