breakage due to udev and hotplug both running hotplug scripts

breakage due to udev and hotplug both running hotplug scripts

[Duncan Sands]

I'm seeing the following: kobject_hotplug results in both
udev and /sbin/hotplug running hotplug scripts.  udevd gets
the hotplug event via netlink thanks to kobject_hotplug
calling send_uevent; /sbin/hotplug gets the hotplug event
thanks to kobject_hotplug calling call_usermodehelper.
Both calls eventually result in an agent being run.  Am
I right in thinking that my system is wrongly configured,
and /sbin/hotplug should be disabled?

The specific problem I'm seeing is that firmware loading
from user-space doesn't work reliably.  That is because
firmware.agent is being run simultaneously by two processes,
leading to uncoordinated, conflicting writes to the loading
and data files.  This causes all kinds of problems, in
particular partial or empty firmware uploads.

Thanks for any help.

All the best,

Duncan.


-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
_______________________________________________
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: breakage due to udev and hotplug both running hotplug scripts

[Marco d'Itri]

On Sep 06, Duncan Sands <baldrick@free.fr> wrote:

> I'm seeing the following: kobject_hotplug results in both
> udev and /sbin/hotplug running hotplug scripts.  udevd gets
> the hotplug event via netlink thanks to kobject_hotplug
> calling send_uevent; /sbin/hotplug gets the hotplug event
> thanks to kobject_hotplug calling call_usermodehelper.
> Both calls eventually result in an agent being run.  Am
> I right in thinking that my system is wrongly configured,
> and /sbin/hotplug should be disabled?
No, but there is still something wrong because udevd should ignore the
second event. I think this should fix it:

echo /sbin/udevsend > /proc/sys/kernel/hotplug

(or at least, I cannot reproduce this with debian/unstable.)

-- 
ciao,
Marco


-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
_______________________________________________
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: breakage due to udev and hotplug both running hotplug scripts

[Duncan Sands]

Hi Marco,

On Tuesday 06 September 2005 20:24, Marco d'Itri wrote:
> On Sep 06, Duncan Sands <baldrick@free.fr> wrote:
> > I'm seeing the following: kobject_hotplug results in both
> > udev and /sbin/hotplug running hotplug scripts.  udevd gets
> > the hotplug event via netlink thanks to kobject_hotplug
> > calling send_uevent; /sbin/hotplug gets the hotplug event
> > thanks to kobject_hotplug calling call_usermodehelper.
> > Both calls eventually result in an agent being run.  Am
> > I right in thinking that my system is wrongly configured,
> > and /sbin/hotplug should be disabled?
> No, but there is still something wrong because udevd should ignore the
> second event. I think this should fix it:
> 
> echo /sbin/udevsend > /proc/sys/kernel/hotplug

yes that works, and is clearly the right solution - thanks a lot.
 
> (or at least, I cannot reproduce this with debian/unstable.)

In debian unstable, who takes care of writing /sbin/udevsend
to /proc/sys/kernel/hotplug?

I'm running "ubuntu breezy badger" up-to-date as of today;

hotplug is 0.0.20040329-16ubuntu2-4.10ubp2
udev is 0.060-1ubuntu10

All the best,

Duncan.


-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
_______________________________________________
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: breakage due to udev and hotplug both running hotplug scripts

[Marco d'Itri]

On Sep 06, Duncan Sands <baldrick@free.fr> wrote:

> In debian unstable, who takes care of writing /sbin/udevsend
> to /proc/sys/kernel/hotplug?
/etc/init.d/udev, and Ubuntu is supposed to work the same way.

-- 
ciao,
Marco


-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
_______________________________________________
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: breakage due to udev and hotplug both running hotplug scripts

[Duncan Sands]

> > In debian unstable, who takes care of writing /sbin/udevsend
> > to /proc/sys/kernel/hotplug?
> /etc/init.d/udev, and Ubuntu is supposed to work the same way.

The problem on ubuntu seems to be that /etc/init.d/hotplug is
writing /sbin/hotplug back into /proc/sys/kernel/hotplug.  In
your release notes for udev I see:

  * Use udevsend as the hotplug multiplexer and depend on hotplug
    >= 0.0.20040329-17 to be sure that it will not be changed back by
    its init script.

ubuntu is shipping hotplug 20040329-16...  I will file a bug report
with ubuntu.

Thanks for sorting it all out.

Best wishes,

Duncan.


-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
_______________________________________________
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: breakage due to udev and hotplug both running hotplug scripts

[Scott James Remnant]

[-- Attachment #1: Type: text/plain, Size: 338 bytes --]

On Tue, 2005-09-06 at 20:35 +0200, Duncan Sands wrote:

> I'm running "ubuntu breezy badger" up-to-date as of today;
> 
> hotplug is 0.0.20040329-16ubuntu2-4.10ubp2
> udev is 0.060-1ubuntu10
> 
As discussed in the Ubuntu bug you filed, these weren't official ubuntu
packages.

Scott
-- 
Scott James Remnant
scott@ubuntu.com

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: breakage due to udev and hotplug both running hotplug scripts

[Duncan Sands]

> > I'm running "ubuntu breezy badger" up-to-date as of today;
> > 
> > hotplug is 0.0.20040329-16ubuntu2-4.10ubp2
> > udev is 0.060-1ubuntu10
> > 
> As discussed in the Ubuntu bug you filed, these weren't official ubuntu
> packages.

Just to clarify: udev was the ubuntu package, hotplug was not:
ubuntu is shipping hotplug 0.0.20040329-22ubuntu11.

Ciao,

Duncan.


-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
_______________________________________________
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