[Buildroot] /run/udev missing ?

[Buildroot] /run/udev missing ?

[Sven Neumann]

Hi,

after updating to latest buildroot, udevd has problems starting:

udevd[1114]: error: runtime directory '/run/udev' not writable, for now
falling back to '/dev/.udev'
udevd[1114]: failed to create queue file: No such file or directory
udevd[1114]: error creating queue file

It looks like udevd is not running or not working properly then as it
fails to load the firmware for the Wifi chip.

This is a configuration with static device management.

Any ideas? Should /run/udev exist or is the fallback OK and supposed to
work?


Sven

[Buildroot] /run/udev missing ?

[Peter Korsgaard]

>>>>> "Sven" == Sven Neumann <s.neumann@raumfeld.com> writes:

 Sven> Hi,
 Sven> after updating to latest buildroot, udevd has problems starting:

 Sven> udevd[1114]: error: runtime directory '/run/udev' not writable, for now
 Sven> falling back to '/dev/.udev'
 Sven> udevd[1114]: failed to create queue file: No such file or directory
 Sven> udevd[1114]: error creating queue file

 Sven> It looks like udevd is not running or not working properly then as it
 Sven> fails to load the firmware for the Wifi chip.

 Sven> This is a configuration with static device management.

All other configurations than static dev (devtmpfs/mdev/udev) expects
CONFIG_DEVTMPFS_MOUNT to be enabled in the kernel, not a static /dev.

-- 
Bye, Peter Korsgaard

[Buildroot] /run/udev missing ?

[Sven Neumann]

On Sun, 2011-08-14 at 08:43 +0200, Peter Korsgaard wrote:
> >>>>> "Sven" == Sven Neumann <s.neumann@raumfeld.com> writes:
> 
>  Sven> Hi,
>  Sven> after updating to latest buildroot, udevd has problems starting:
> 
>  Sven> udevd[1114]: error: runtime directory '/run/udev' not writable, for now
>  Sven> falling back to '/dev/.udev'
>  Sven> udevd[1114]: failed to create queue file: No such file or directory
>  Sven> udevd[1114]: error creating queue file
> 
>  Sven> It looks like udevd is not running or not working properly then as it
>  Sven> fails to load the firmware for the Wifi chip.
> 
>  Sven> This is a configuration with static device management.
> 
> All other configurations than static dev (devtmpfs/mdev/udev) expects
> CONFIG_DEVTMPFS_MOUNT to be enabled in the kernel, not a static /dev.

Yes, and it looks like linux/linux.mk is responsible for making sure
that the kernel is configured as expected. As far as I can see this
works just fine.

Still it may be desirable to run udev and use static device management
at the same time. At least this used to work and I would like to make it
work again.


Sven

[Buildroot] /run/udev missing ?

[Peter Korsgaard]

>>>>> "Sven" == Sven Neumann <s.neumann@raumfeld.com> writes:

Hi,

 Sven> Yes, and it looks like linux/linux.mk is responsible for making sure
 Sven> that the kernel is configured as expected. As far as I can see this
 Sven> works just fine.

Indeed, if you build your kernel with buildroot.

 Sven> Still it may be desirable to run udev and use static device
 Sven> management at the same time. At least this used to work and I
 Sven> would like to make it work again.

What would the advantage of this be? This is not how udev is normally
used nowadays (devtmpfs was explicitly made to fix various issues with
the initial static /dev and handover to udev) on PCs.

Compared to udev, devtmpfs is very lightweight. I would recommend pure
devtmpfs (no mdev/udev) to people using static /dev today.

-- 
Bye, Peter Korsgaard

[Buildroot] /run/udev missing ?

[Sven Neumann]

On Mon, 2011-08-15 at 11:05 +0200, Peter Korsgaard wrote:
> >>>>> "Sven" == Sven Neumann <s.neumann@raumfeld.com> writes:
> 
> Hi,
> 
>  Sven> Yes, and it looks like linux/linux.mk is responsible for making sure
>  Sven> that the kernel is configured as expected. As far as I can see this
>  Sven> works just fine.
> 
> Indeed, if you build your kernel with buildroot.
> 
>  Sven> Still it may be desirable to run udev and use static device
>  Sven> management at the same time. At least this used to work and I
>  Sven> would like to make it work again.
> 
> What would the advantage of this be? This is not how udev is normally
> used nowadays (devtmpfs was explicitly made to fix various issues with
> the initial static /dev and handover to udev) on PCs.
> 
> Compared to udev, devtmpfs is very lightweight. I would recommend pure
> devtmpfs (no mdev/udev) to people using static /dev today.

Well, we need udev to load the firmware of our Wifi module. Of course I
can try to switch to dynamic device management, but I thought I'd rather
point out that a setup broke that the buildroot configuration allows and
that used to work.

If this is not supposed to work, then it should probably be disallowed
in the buildroot configuration. In other words, if udev is selected,
static device management should not be selectable.


Sven

[Buildroot] /run/udev missing ?

[Peter Korsgaard]

>>>>> "Sven" == Sven Neumann <s.neumann@raumfeld.com> writes:

Hi,

 >> Compared to udev, devtmpfs is very lightweight. I would recommend
 >> pure devtmpfs (no mdev/udev) to people using static /dev today.

 Sven> Well, we need udev to load the firmware of our Wifi module. Of
 Sven> course I can try to switch to dynamic device management, but I
 Sven> thought I'd rather point out that a setup broke that the
 Sven> buildroot configuration allows and that used to work.

 Sven> If this is not supposed to work, then it should probably be
 Sven> disallowed in the buildroot configuration. In other words, if
 Sven> udev is selected, static device management should not be
 Sven> selectable.

Ahh yes, good idea - Fixed in git.

-- 
Bye, Peter Korsgaard

[Buildroot] /run/udev missing ?

[Sven Neumann]

On Mon, 2011-08-15 at 14:10 +0200, Peter Korsgaard wrote:
> >>>>> "Sven" == Sven Neumann <s.neumann@raumfeld.com> writes:
> 
> Hi,
> 
>  >> Compared to udev, devtmpfs is very lightweight. I would recommend
>  >> pure devtmpfs (no mdev/udev) to people using static /dev today.
> 
>  Sven> Well, we need udev to load the firmware of our Wifi module. Of
>  Sven> course I can try to switch to dynamic device management, but I
>  Sven> thought I'd rather point out that a setup broke that the
>  Sven> buildroot configuration allows and that used to work.
> 
>  Sven> If this is not supposed to work, then it should probably be
>  Sven> disallowed in the buildroot configuration. In other words, if
>  Sven> udev is selected, static device management should not be
>  Sven> selectable.
> 
> Ahh yes, good idea - Fixed in git.

Hmm, not really was I was asking for in the first place. But I guess
I'll have to deal with your decision then to drop support for this
configuration...

Still I wonder why udevd is looking for "/run/udev". Shouldn't this
be /var/run/udev ? Looks to me like udevd is configured with a wrong
prefix for the run directory.


Sven

[Buildroot] /run/udev missing ?

[Daniel Mack]

On Mon, Aug 15, 2011 at 4:01 PM, Sven Neumann <s.neumann@raumfeld.com> wrote:
> Still I wonder why udevd is looking for "/run/udev". Shouldn't this
> be /var/run/udev ? Looks to me like udevd is configured with a wrong
> prefix for the run directory.

/run was introduced recently by systemd and udev, considering the fact
that /var might be mounted on partitions that are not yet available
when systemd starts. So this sounds like the new standard, and It's
likely that buildroot needs to be amended in this case.


Daniel

[Buildroot] /run/udev missing ?

[Baruch Siach]

Hi Sven,

On Mon, Aug 15, 2011 at 04:01:02PM +0200, Sven Neumann wrote:
> Still I wonder why udevd is looking for "/run/udev". Shouldn't this
> be /var/run/udev ? Looks to me like udevd is configured with a wrong
> prefix for the run directory.

/run is a newly introduced toplevel directory for early boot runtime data. See 
http://lwn.net/Articles/436012/.

baruch

-- 
                                                     ~. .~   Tk Open Systems
=}------------------------------------------------ooO--U--Ooo------------{=
   - baruch at tkos.co.il - tel: +972.2.679.5364, http://www.tkos.co.il -

[Buildroot] /run/udev missing ?

[Sven Neumann]

On Sun, 2011-08-14 at 08:43 +0200, Peter Korsgaard wrote:
> >>>>> "Sven" == Sven Neumann <s.neumann@raumfeld.com> writes:
> 
>  Sven> Hi,
>  Sven> after updating to latest buildroot, udevd has problems starting:
> 
>  Sven> udevd[1114]: error: runtime directory '/run/udev' not writable, for now
>  Sven> falling back to '/dev/.udev'
>  Sven> udevd[1114]: failed to create queue file: No such file or directory
>  Sven> udevd[1114]: error creating queue file
> 
>  Sven> It looks like udevd is not running or not working properly then as it
>  Sven> fails to load the firmware for the Wifi chip.
> 
>  Sven> This is a configuration with static device management.
> 
> All other configurations than static dev (devtmpfs/mdev/udev) expects
> CONFIG_DEVTMPFS_MOUNT to be enabled in the kernel, not a static /dev.

Do you have a suggestion on how to use udev with a kernel that does not
yet have this option? We are stuck with 2.6.31 on one of our platforms
and I haven't been able to get udev working with recent buildroot on
this platform.


Sven

[Buildroot] /run/udev missing ?

[Diego Iastrubni]

On Tue, Aug 16, 2011 at 11:40 AM, Sven Neumann <s.neumann@raumfeld.com>wrote:

>
> Do you have a suggestion on how to use udev with a kernel that does not
> yet have this option? We are stuck with 2.6.31 on one of our platforms
> and I haven't been able to get udev working with recent buildroot on
> this platform.
>
>
This init.d file woks for me on 2.6.32. I am using tmpfs for mounting udev.

Note that for udev 173 you will also need  a patch to compile with older
toolchains (look for in in the ML or the bugzilla).
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.busybox.net/pipermail/buildroot/attachments/20110816/371d7630/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: S10udev
Type: application/octet-stream
Size: 1788 bytes
Desc: not available
URL: <http://lists.busybox.net/pipermail/buildroot/attachments/20110816/371d7630/attachment.obj>

[Buildroot] /run/udev missing ?

[Michael S. Zick]

On Tue August 16 2011, Diego Iastrubni wrote:
> On Tue, Aug 16, 2011 at 11:40 AM, Sven Neumann <s.neumann@raumfeld.com>wrote:
> 
> >
> > Do you have a suggestion on how to use udev with a kernel that does not
> > yet have this option? We are stuck with 2.6.31 on one of our platforms
> > and I haven't been able to get udev working with recent buildroot on
> > this platform.
> >
> >
> This init.d file woks for me on 2.6.32. I am using tmpfs for mounting udev.
>

Isn't: mount -t ramfs
A ram disk on the older kernels rather than a tmpfs (-t tmpfs) ?

Mike
> 
> Note that for udev 173 you will also need  a patch to compile with older
> toolchains (look for in in the ML or the bugzilla).
>