advice needed for gentoo bug involving lvm2/udev

advice needed for gentoo bug involving lvm2/udev

[William Hubbs]

[-- Attachment #1: Type: text/plain, Size: 488 bytes --]

All,

we have the following bug posted in gentoo's bugzilla:

http://bugs.gentoo.org/show_bug.cgi?id=365227.

The reporter is telling me that we should use --action=change instead of
--action=add in the cold boot sequence when dev is devtmpfs. However,
this doesn't seem to be the correct fix based on earlier discussions on
this list.

Does anyone else have any suggestions for fixing this? My thought is
that the rules for lvm2 should be fixed. What does everyone else think?

William


[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]

Re: advice needed for gentoo bug involving lvm2/udev

[Kay Sievers]

On Mon, Oct 3, 2011 at 18:41, William Hubbs <w.d.hubbs@gmail.com> wrote:
> we have the following bug posted in gentoo's bugzilla:
>
> http://bugs.gentoo.org/show_bug.cgi?id65227.
>
> The reporter is telling me that we should use --action=change instead of
> --action­d in the cold boot sequence when dev is devtmpfs. However,
> this doesn't seem to be the correct fix based on earlier discussions on
> this list.
>
> Does anyone else have any suggestions for fixing this? My thought is
> that the rules for lvm2 should be fixed. What does everyone else think?

--action­d is still the recommended and default way of doing coldplug.

It should only be done once after udevd is started though, and never
again. All later triggers should be change only.

Devtmpfs does not really matter here, it's the replay of the missed
events, which should be add and not change.

Kay

Re: advice needed for gentoo bug involving lvm2/udev

[William Hubbs]

[-- Attachment #1: Type: text/plain, Size: 1345 bytes --]

On Tue, Oct 04, 2011 at 03:09:34AM +0200, Kay Sievers wrote:
> On Mon, Oct 3, 2011 at 18:41, William Hubbs <w.d.hubbs@gmail.com> wrote:
> > we have the following bug posted in gentoo's bugzilla:
> >
> > http://bugs.gentoo.org/show_bug.cgi?id=365227.
> >
> > The reporter is telling me that we should use --action=change instead of
> > --action=add in the cold boot sequence when dev is devtmpfs. However,
> > this doesn't seem to be the correct fix based on earlier discussions on
> > this list.
> >
> > Does anyone else have any suggestions for fixing this? My thought is
> > that the rules for lvm2 should be fixed. What does everyone else think?
> 
> --action=add is still the recommended and default way of doing coldplug.
> 
> It should only be done once after udevd is started though, and never
> again. All later triggers should be change only.

The reporter is now saying that --action=add does not touch nodes that
are already in the file system, so, for example, if you mount devtmpfs
on /dev then call udevadm trigger --action=add, the permissions,
ownership, etc, of nodes that already exist are not touched. So, he is
suggesting that we add another udevadm trigger call with --action=change
to the cold boot sequence.

Is this a bug in udev, or should we add this extra udevadm trigger call?

William


[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]

Re: advice needed for gentoo bug involving lvm2/udev

[William Hubbs]

[-- Attachment #1: Type: text/plain, Size: 1763 bytes --]

On Tue, Dec 06, 2011 at 04:10:13PM -0600, William Hubbs wrote:
> On Tue, Oct 04, 2011 at 03:09:34AM +0200, Kay Sievers wrote:
> > On Mon, Oct 3, 2011 at 18:41, William Hubbs <w.d.hubbs@gmail.com> wrote:
> > > we have the following bug posted in gentoo's bugzilla:
> > >
> > > http://bugs.gentoo.org/show_bug.cgi?id=365227.
> > >
> > > The reporter is telling me that we should use --action=change instead of
> > > --action=add in the cold boot sequence when dev is devtmpfs. However,
> > > this doesn't seem to be the correct fix based on earlier discussions on
> > > this list.
> > >
> > > Does anyone else have any suggestions for fixing this? My thought is
> > > that the rules for lvm2 should be fixed. What does everyone else think?
> > 
> > --action=add is still the recommended and default way of doing coldplug.
> > 
> > It should only be done once after udevd is started though, and never
> > again. All later triggers should be change only.
> 
> The reporter is now saying that --action=add does not touch nodes that
> are already in the file system, so, for example, if you mount devtmpfs
> on /dev then call udevadm trigger --action=add, the permissions,
> ownership, etc, of nodes that already exist are not touched. So, he is
> suggesting that we add another udevadm trigger call with --action=change
> to the cold boot sequence.
> 
> Is this a bug in udev, or should we add this extra udevadm trigger call?

Hi Kay,

I am re-sending this in case you didn't get it before.

Basically, the reporter is now saying that the udevadm trigger
--action=add is correct, but that the coldplug sequence should include
udevadm trigger --action=change after the --action=add call.

Is this correct?

Thanks,

William


[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]

Re: advice needed for gentoo bug involving lvm2/udev

[Kay Sievers]

On Mon, Dec 12, 2011 at 22:45, William Hubbs <w.d.hubbs@gmail.com> wrote:
> On Tue, Dec 06, 2011 at 04:10:13PM -0600, William Hubbs wrote:
>> On Tue, Oct 04, 2011 at 03:09:34AM +0200, Kay Sievers wrote:
>> > On Mon, Oct 3, 2011 at 18:41, William Hubbs <w.d.hubbs@gmail.com> wrote:
>> > > we have the following bug posted in gentoo's bugzilla:
>> > >
>> > > http://bugs.gentoo.org/show_bug.cgi?id65227.
>> > >
>> > > The reporter is telling me that we should use --action=change instead of
>> > > --action­d in the cold boot sequence when dev is devtmpfs. However,
>> > > this doesn't seem to be the correct fix based on earlier discussions on
>> > > this list.
>> > >
>> > > Does anyone else have any suggestions for fixing this? My thought is
>> > > that the rules for lvm2 should be fixed. What does everyone else think?
>> >
>> > --action­d is still the recommended and default way of doing coldplug.
>> >
>> > It should only be done once after udevd is started though, and never
>> > again. All later triggers should be change only.
>>
>> The reporter is now saying that --action­d does not touch nodes that
>> are already in the file system, so, for example, if you mount devtmpfs
>> on /dev then call udevadm trigger --action­d, the permissions,
>> ownership, etc, of nodes that already exist are not touched. So, he is
>> suggesting that we add another udevadm trigger call with --action=change
>> to the cold boot sequence.
>>
>> Is this a bug in udev, or should we add this extra udevadm trigger call?
>
> Hi Kay,
>
> I am re-sending this in case you didn't get it before.
>
> Basically, the reporter is now saying that the udevadm trigger
> --action­d is correct, but that the coldplug sequence should include
> udevadm trigger --action=change after the --action­d call.
>
> Is this correct?

No, udev should not trigger any 'change' event during bootup, it
should only trigger 'add' events once and very early during
bootup.

Any other issues need to be fixed by/in the tools, not hacked around in udev.

Kay