[PATCH] connman: fixed init script, so it can run with nfs

[PATCH] connman: fixed init script, so it can run with nfs

[eu]

From: "Felipe F. Tonello" <ftonello@cercacor.com>

Adding -I to the arguments, connmand will ignore the eth interface, so if you are
using nfs it will not be disconnected.

FIXME: it might not work if using more than one eth interface. But it's better
than not using connman with nfs at all.

Signed-off-by: Felipe F. Tonello <ftonello@cercacor.com>
---
 meta/recipes-connectivity/connman/connman/connman |    7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/meta/recipes-connectivity/connman/connman/connman b/meta/recipes-connectivity/connman/connman/connman
index 4a0017f..e90925e 100644
--- a/meta/recipes-connectivity/connman/connman/connman
+++ b/meta/recipes-connectivity/connman/connman/connman
@@ -26,11 +26,14 @@ while read dev mtpt fstype rest; do
 	fi
 done
 
+ethn=`ifconfig | grep eth | sed -e "s/\(eth[0-9]\)\(.*\)/\1/"`
+
 do_start() {
 	EXTRA_PARAM=""
-	if test $nfsroot -eq 0 ; then
-		$DAEMON $EXTRA_PARAM
+	if test $nfsroot -eq 1 ; then
+		EXTRA_PARAM="-I $ethn"
 	fi
+	$DAEMON $EXTRA_PARAM
 }
 
 do_stop() {
-- 
1.7.9.5

Re: [PATCH] connman: fixed init script, so it can run with nfs

[Iorga, Cristian]

Hello Felipe,

I have some questions regarding this patch:

1. On which hardware/machine have you tested this patch? (real hardware, qemu machine, etc)
2. In what way is connman affecting nfs? In other words, what is the root cause?
3. Which version of connman are you using?

I am asking these questions because I am working on something somehow related, and I think a more general fix might be needed.

Regards,
Cristian Iorga


-----Original Message-----
From: poky-bounces@yoctoproject.org [mailto:poky-bounces@yoctoproject.org] On Behalf Of eu@felipetonello.com
Sent: Wednesday, January 09, 2013 2:37 AM
To: poky@yoctoproject.org
Cc: Felipe F. Tonello
Subject: [poky] [PATCH] connman: fixed init script, so it can run with nfs

From: "Felipe F. Tonello" <ftonello@cercacor.com>

Adding -I to the arguments, connmand will ignore the eth interface, so if you are using nfs it will not be disconnected.

FIXME: it might not work if using more than one eth interface. But it's better than not using connman with nfs at all.

Signed-off-by: Felipe F. Tonello <ftonello@cercacor.com>
---
 meta/recipes-connectivity/connman/connman/connman |    7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/meta/recipes-connectivity/connman/connman/connman b/meta/recipes-connectivity/connman/connman/connman
index 4a0017f..e90925e 100644
--- a/meta/recipes-connectivity/connman/connman/connman
+++ b/meta/recipes-connectivity/connman/connman/connman
@@ -26,11 +26,14 @@ while read dev mtpt fstype rest; do
 	fi
 done
 
+ethn=`ifconfig | grep eth | sed -e "s/\(eth[0-9]\)\(.*\)/\1/"`
+
 do_start() {
 	EXTRA_PARAM=""
-	if test $nfsroot -eq 0 ; then
-		$DAEMON $EXTRA_PARAM
+	if test $nfsroot -eq 1 ; then
+		EXTRA_PARAM="-I $ethn"
 	fi
+	$DAEMON $EXTRA_PARAM
 }
 
 do_stop() {
--
1.7.9.5

_______________________________________________
poky mailing list
poky@yoctoproject.org
https://lists.yoctoproject.org/listinfo/poky

Re: [PATCH] connman: fixed init script, so it can run with nfs

[Richard Purdie]

On Wed, 2013-01-09 at 10:12 +0000, Iorga, Cristian wrote:
> Hello Felipe,
> 
> I have some questions regarding this patch:
> 
> 1. On which hardware/machine have you tested this patch? (real hardware, qemu machine, etc)
> 2. In what way is connman affecting nfs? In other words, what is the root cause?
> 3. Which version of connman are you using?
> 
> I am asking these questions because I am working on something somehow related, and I think a more general fix might be needed.

I'd also add that:

a) Patches to this part of the tree need to go to the OE-Core mailing list
b) Unconditionally running the expression is inefficient, it should move
inside the if statement as it triggers a number of exec calls. This may
sound trivial but that all mount up.

Cheers,

Richard

Re: [PATCH] connman: fixed init script, so it can run with nfs

[Felipe Ferreri Tonello]

Hi Richard,

On 01/09/2013 05:03 AM, Richard Purdie wrote:
 > On Wed, 2013-01-09 at 10:12 +0000, Iorga, Cristian wrote:
 >> Hello Felipe,
 >>
 >> I have some questions regarding this patch:
 >>
 >> 1. On which hardware/machine have you tested this patch? (real 
hardware, qemu machine, etc)
 >> 2. In what way is connman affecting nfs? In other words, what is the 
root cause?
 >> 3. Which version of connman are you using?
 >>
 >> I am asking these questions because I am working on something 
somehow related, and I think a more general fix might be needed.
 >
 > I'd also add that:
 >
 > a) Patches to this part of the tree need to go to the OE-Core mailing 
list
Ok.

 > b) Unconditionally running the expression is inefficient, it should move
 > inside the if statement as it triggers a number of exec calls. This may
 > sound trivial but that all mount up.
I agree. I will fix that before sending to OE-Core.

Felipe

Re: [PATCH] connman: fixed init script, so it can run with nfs

[Felipe Ferreri Tonello]

Richard and Cristian,

On 01/09/2013 05:03 AM, Richard Purdie wrote:
> On Wed, 2013-01-09 at 10:12 +0000, Iorga, Cristian wrote:
>> Hello Felipe,
>>
>> I have some questions regarding this patch:
>>
>> 1. On which hardware/machine have you tested this patch? (real hardware, qemu machine, etc)
>> 2. In what way is connman affecting nfs? In other words, what is the root cause?
>> 3. Which version of connman are you using?
>>
>> I am asking these questions because I am working on something somehow related, and I think a more general fix might be needed.
>
> I'd also add that:
>
> a) Patches to this part of the tree need to go to the OE-Core mailing list

I saw on openembedded tree that connman recipes are old. They are not 
the same as in poky.

I'm always confused about this. Where is oe-core that poky and oe share?

Felipe

Re: [PATCH] connman: fixed init script, so it can run with nfs

[Richard Purdie]

On Thu, 2013-01-10 at 11:23 -0800, Felipe Ferreri Tonello wrote:
> Richard and Cristian,
> 
> On 01/09/2013 05:03 AM, Richard Purdie wrote:
> > On Wed, 2013-01-09 at 10:12 +0000, Iorga, Cristian wrote:
> >> Hello Felipe,
> >>
> >> I have some questions regarding this patch:
> >>
> >> 1. On which hardware/machine have you tested this patch? (real hardware, qemu machine, etc)
> >> 2. In what way is connman affecting nfs? In other words, what is the root cause?
> >> 3. Which version of connman are you using?
> >>
> >> I am asking these questions because I am working on something somehow related, and I think a more general fix might be needed.
> >
> > I'd also add that:
> >
> > a) Patches to this part of the tree need to go to the OE-Core mailing list
> 
> I saw on openembedded tree that connman recipes are old. They are not 
> the same as in poky.
> 
> I'm always confused about this. Where is oe-core that poky and oe share?

These two repositories should be in sync in terms of file level content
of the meta directory:

http://cgit.openembedded.org/openembedded-core/log/

http://git.yoctoproject.org/cgit/cgit.cgi/poky/log/

Were you looking at openembedded or openembedded-core?

Cheers,

Richard

Re: [PATCH] connman: fixed init script, so it can run with nfs

[Felipe Ferreri Tonello]

Hi Richard,

On 01/10/2013 03:57 PM, Richard Purdie wrote:
> On Thu, 2013-01-10 at 11:23 -0800, Felipe Ferreri Tonello wrote:
>> Richard and Cristian,
>>
>> On 01/09/2013 05:03 AM, Richard Purdie wrote:
>>> On Wed, 2013-01-09 at 10:12 +0000, Iorga, Cristian wrote:
>>>> Hello Felipe,
>>>>
>>>> I have some questions regarding this patch:
>>>>
>>>> 1. On which hardware/machine have you tested this patch? (real hardware, qemu machine, etc)
>>>> 2. In what way is connman affecting nfs? In other words, what is the root cause?
>>>> 3. Which version of connman are you using?
>>>>
>>>> I am asking these questions because I am working on something somehow related, and I think a more general fix might be needed.
>>>
>>> I'd also add that:
>>>
>>> a) Patches to this part of the tree need to go to the OE-Core mailing list
>>
>> I saw on openembedded tree that connman recipes are old. They are not
>> the same as in poky.
>>
>> I'm always confused about this. Where is oe-core that poky and oe share?
>
> These two repositories should be in sync in terms of file level content
> of the meta directory:
>
> http://cgit.openembedded.org/openembedded-core/log/
>
> http://git.yoctoproject.org/cgit/cgit.cgi/poky/log/
>
> Were you looking at openembedded or openembedded-core?

I was looking into openembedded, now I got it.

btw, What's the difference between them? openembedded tree is the 
meta-oe? and openembedded-core ie oe-core?

Thanks,

Felipe

Re: [PATCH] connman: fixed init script, so it can run with nfs

[Richard Purdie]

On Thu, 2013-01-10 at 17:17 -0800, Felipe Ferreri Tonello wrote:
> Hi Richard,
> 
> On 01/10/2013 03:57 PM, Richard Purdie wrote:
> > On Thu, 2013-01-10 at 11:23 -0800, Felipe Ferreri Tonello wrote:
> >> Richard and Cristian,
> >>
> >> On 01/09/2013 05:03 AM, Richard Purdie wrote:
> >>> On Wed, 2013-01-09 at 10:12 +0000, Iorga, Cristian wrote:
> >>>> Hello Felipe,
> >>>>
> >>>> I have some questions regarding this patch:
> >>>>
> >>>> 1. On which hardware/machine have you tested this patch? (real hardware, qemu machine, etc)
> >>>> 2. In what way is connman affecting nfs? In other words, what is the root cause?
> >>>> 3. Which version of connman are you using?
> >>>>
> >>>> I am asking these questions because I am working on something somehow related, and I think a more general fix might be needed.
> >>>
> >>> I'd also add that:
> >>>
> >>> a) Patches to this part of the tree need to go to the OE-Core mailing list
> >>
> >> I saw on openembedded tree that connman recipes are old. They are not
> >> the same as in poky.
> >>
> >> I'm always confused about this. Where is oe-core that poky and oe share?
> >
> > These two repositories should be in sync in terms of file level content
> > of the meta directory:
> >
> > http://cgit.openembedded.org/openembedded-core/log/
> >
> > http://git.yoctoproject.org/cgit/cgit.cgi/poky/log/
> >
> > Were you looking at openembedded or openembedded-core?
> 
> I was looking into openembedded, now I got it.
> 
> btw, What's the difference between them? openembedded tree is the 
> meta-oe? and openembedded-core ie oe-core?

openembedded is OE-Classic, i.e. from before it got split into various
pieces such as oe-core and meta-oe.

Cheers,

Richard

Re: [PATCH] connman: fixed init script, so it can run with nfs

[Felipe Ferreri Tonello]

Hi Iorga,

On 01/09/2013 02:12 AM, Iorga, Cristian wrote:
 > Hello Felipe,
 >
 > I have some questions regarding this patch:
 >
 > 1. On which hardware/machine have you tested this patch? (real 
hardware, qemu machine, etc)
Real hardware.

 > 2. In what way is connman affecting nfs? In other words, what is the 
root cause?
When connman tried to take ownership of eth, it crashes the nfs, that is 
hosting rootfs for instance, connection.

 > 3. Which version of connman are you using?
I'm using 1.9 but I tested this patch on 1.6 as well.

 >
 > I am asking these questions because I am working on something somehow 
related, and I think a more general fix might be needed.

I agree, this is not a generic patch. But it works on most of the times. 
Also I think it's better then just disabling ConnMan with nfs.

Felipe

Re: [PATCH] connman: fixed init script, so it can run with nfs

[Iorga, Cristian]

Hello Felipe,

My first name is Cristian.

I have reviewed your patch and it does look OK in regard to functionality.

Regards,
Cristian

-----Original Message-----
From: Felipe Ferreri Tonello [mailto:eu@felipetonello.com] 
Sent: Wednesday, January 09, 2013 8:15 PM
To: Iorga, Cristian
Cc: poky@yoctoproject.org; Felipe F. Tonello
Subject: Re: [poky] [PATCH] connman: fixed init script, so it can run with nfs


Hi Iorga,

On 01/09/2013 02:12 AM, Iorga, Cristian wrote:
 > Hello Felipe,
 >
 > I have some questions regarding this patch:
 >
 > 1. On which hardware/machine have you tested this patch? (real hardware, qemu machine, etc) Real hardware.

 > 2. In what way is connman affecting nfs? In other words, what is the root cause?
When connman tried to take ownership of eth, it crashes the nfs, that is hosting rootfs for instance, connection.

 > 3. Which version of connman are you using?
I'm using 1.9 but I tested this patch on 1.6 as well.

 >
 > I am asking these questions because I am working on something somehow related, and I think a more general fix might be needed.

I agree, this is not a generic patch. But it works on most of the times. 
Also I think it's better then just disabling ConnMan with nfs.

Felipe

[PATCH] connman: fixed init script, so it can run with nfs

[Felipe Ferreri Tonello]

From: "Felipe F. Tonello" <ftonello@cercacor.com>

Adding -I to the arguments, connmand will ignore the eth interface, so 
if you are
using nfs it will not be disconnected.

FIXME: it might not work if using more than one eth interface. But it's 
better
than not using connman with nfs at all.

Signed-off-by: Felipe F. Tonello <ftonello@cercacor.com>
---
  meta/recipes-connectivity/connman/connman/connman |    7 +++++--
  1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/meta/recipes-connectivity/connman/connman/connman 
b/meta/recipes-connectivity/connman/connman/connman
index 4a0017f..e90925e 100644
--- a/meta/recipes-connectivity/connman/connman/connman
+++ b/meta/recipes-connectivity/connman/connman/connman
@@ -26,11 +26,14 @@ while read dev mtpt fstype rest; do
  	fi
  done

+ethn=`ifconfig | grep eth | sed -e "s/\(eth[0-9]\)\(.*\)/\1/"`
+
  do_start() {
  	EXTRA_PARAM=""
-	if test $nfsroot -eq 0 ; then
-		$DAEMON $EXTRA_PARAM
+	if test $nfsroot -eq 1 ; then
+		EXTRA_PARAM="-I $ethn"
  	fi
+	$DAEMON $EXTRA_PARAM
  }

  do_stop() {
-- 
1.7.9.5