[Xenomai] patch for 3.14 kernel ?

[Xenomai] patch for 3.14 kernel ?

[Huub Van Niekerk]

Hi,

After struggling with the 3.10.18 kernel on armhf (both with and without the patch), I tried the 3.14 kernel and to my delight some issues are solved: the clock runs fine again, roscore and top work great, while they kept failing on 3.10.18. So my question is if (or when) there will be a patch for the 3.14 kernel line ?

Thanks,

hvn

Re: [Xenomai] patch for 3.14 kernel ?

[Gilles Chanteperdrix]

On 04/07/2014 02:21 PM, Huub Van Niekerk wrote:
> Hi,
>
> After struggling with the 3.10.18 kernel on armhf (both with and
> without the patch), I tried the 3.14 kernel and to my delight some
> issues are solved: the clock runs fine again, roscore and top work
> great, while they kept failing on 3.10.18.

I would say you have a timer issue, or perhaps the same issue with 
SD-card as Andreas. Have you tried turning off the options I said Adreas 
to disable yesterday? (dynamic ticks and high res timers)?

When the system is stuck, do you still see the timer interrupt count 
incrementing in /proc/interrupts? Same question for the MMC interrupts.

Since you are the one having the issue, you will have to do the 
investigation. If you do not do that, do not expect the issue to 
magically go away.

> So my question is if (or
> when) there will be a patch for the 3.14 kernel line ?

I work for the I-pipe patch for the ARM architecture on my free time. 
So, I do things when I want to, without any deadline, and certainly not 
under pressure.

-- 
					    Gilles.

Re: [Xenomai] patch for 3.14 kernel ?

[Huub Van Niekerk]

On Monday, April 7, 2014 2:51 PM, Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org> wrote:
 
On 04/07/2014 02:21 PM, Huub Van Niekerk wrote:
> Hi,
>
> After struggling with the 3.10.18 kernel on armhf (both with and
> without the patch), I tried the 3.14 kernel and to my delight some
> issues are solved: the clock runs fine again, roscore and top work
> great, while they kept failing on 3.10.18.

>I would say you have a timer issue, or perhaps the same issue with 
>SD-card as Andreas. Have you tried turning off the options I said Adreas 
>to disable yesterday? (dynamic ticks and high res timers)?

>When the system is stuck, do you still see the timer interrupt count 
>incrementing in /proc/interrupts? Same question for the MMC interrupts.

>Since you are the one having the issue, you will have to do the 
>investigation. If you do not do that, do not expect the issue to 
>magically go away.


> So my question is if (or
> when) there will be a patch for the 3.14 kernel line ?

>I work for the I-pipe patch for the ARM architecture on my free time. 
>So, I do things when I want to, without any deadline, and certainly not 
>under pressure.

Right now I can't test anything under the 3.10.18 kernel since for some weird reason (which has nothing to do with Xenomai) my network and USB (combined on my board) is off. With the same configuration file, the 3.14 kernel works fine. 

Re: [Xenomai] patch for 3.14 kernel ?

[Gilles Chanteperdrix]

On 04/08/2014 02:47 PM, Huub Van Niekerk wrote:
> On Monday, April 7, 2014 2:51 PM, Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org> wrote:
>
> On 04/07/2014 02:21 PM, Huub Van Niekerk wrote:
>> Hi,
>>
>> After struggling with the 3.10.18 kernel on armhf (both with and
>> without the patch), I tried the 3.14 kernel and to my delight some
>> issues are solved: the clock runs fine again, roscore and top work
>> great, while they kept failing on 3.10.18.
>
>> I would say you have a timer issue, or perhaps the same issue with
>> SD-card as Andreas. Have you tried turning off the options I said Adreas
>> to disable yesterday? (dynamic ticks and high res timers)?
>
>> When the system is stuck, do you still see the timer interrupt count
>> incrementing in /proc/interrupts? Same question for the MMC interrupts.
>
>> Since you are the one having the issue, you will have to do the
>> investigation. If you do not do that, do not expect the issue to
>> magically go away.
>
>
>> So my question is if (or
>> when) there will be a patch for the 3.14 kernel line ?
>
>> I work for the I-pipe patch for the ARM architecture on my free time.
>> So, I do things when I want to, without any deadline, and certainly not
>> under pressure.
>
> Right now I can't test anything under the 3.10.18 kernel since for some weird reason (which has nothing to do with Xenomai) my network and USB (combined on my board) is off. With the same configuration file, the 3.14 kernel works fine.
>

In 3.10 kernel, a new configuration disabled by default is the USB PHY. 
You need to select the USB PHY corresponding to your board. If you want 
to boot using NFS, you have to enable the PHY as built-in in addition to 
the built-in network controller. I did that for the pandaboard (which 
uses the same network adapter as Beagle-XM I believe), and it boots just 
fine using NFS.


-- 
					    Gilles.

Re: [Xenomai] patch for 3.14 kernel ?

[Gilles Chanteperdrix]

On 04/08/2014 02:55 PM, Gilles Chanteperdrix wrote:
> On 04/08/2014 02:47 PM, Huub Van Niekerk wrote:
>> On Monday, April 7, 2014 2:51 PM, Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org> wrote:
>>
>> On 04/07/2014 02:21 PM, Huub Van Niekerk wrote:
>>> Hi,
>>>
>>> After struggling with the 3.10.18 kernel on armhf (both with and
>>> without the patch), I tried the 3.14 kernel and to my delight some
>>> issues are solved: the clock runs fine again, roscore and top work
>>> great, while they kept failing on 3.10.18.
>>
>>> I would say you have a timer issue, or perhaps the same issue with
>>> SD-card as Andreas. Have you tried turning off the options I said Adreas
>>> to disable yesterday? (dynamic ticks and high res timers)?
>>
>>> When the system is stuck, do you still see the timer interrupt count
>>> incrementing in /proc/interrupts? Same question for the MMC interrupts.
>>
>>> Since you are the one having the issue, you will have to do the
>>> investigation. If you do not do that, do not expect the issue to
>>> magically go away.
>>
>>
>>> So my question is if (or
>>> when) there will be a patch for the 3.14 kernel line ?
>>
>>> I work for the I-pipe patch for the ARM architecture on my free time.
>>> So, I do things when I want to, without any deadline, and certainly not
>>> under pressure.
>>
>> Right now I can't test anything under the 3.10.18 kernel since for some weird reason (which has nothing to do with Xenomai) my network and USB (combined on my board) is off. With the same configuration file, the 3.14 kernel works fine.
>>
>
> In 3.10 kernel, a new configuration disabled by default is the USB PHY.
> You need to select the USB PHY corresponding to your board. If you want
> to boot using NFS, you have to enable the PHY as built-in in addition to
> the built-in network controller. I did that for the pandaboard (which
> uses the same network adapter as Beagle-XM I believe), and it boots just
> fine using NFS.
>
>
The required configuration is CONFIG_NOP_USB_XCEIV, and you do not have 
the problem with 3.14 because CONFIG_NOP_USB_XCEIV is automatically 
selected by CONFIG_USB_EHCI_HCD_OMAP, I guess too many people had the 
issue.

-- 
					    Gilles.

Re: [Xenomai] patch for 3.14 kernel ?

[Huub Van Niekerk]

On Tuesday, April 8, 2014 2:55 PM, Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org> wrote:
 
On 04/08/2014 02:47 PM, Huub Van Niekerk wrote:
> On Monday, April 7, 2014 2:51 PM, Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org> wrote:
>
> On 04/07/2014 02:21 PM, Huub Van Niekerk wrote:
>> Hi,
>>
>> After struggling with the 3.10.18 kernel on armhf (both with and
>> without the patch), I tried the 3.14 kernel and to my delight some
>> issues are solved: the clock runs fine again, roscore and top work
>> great, while they kept failing on 3.10.18.
>
>> I would say you have a timer issue, or perhaps the same issue with
>> SD-card as Andreas. Have you tried turning off the options I said Adreas
>> to disable yesterday? (dynamic ticks and high res timers)?
>
>> When the system is stuck, do you still see the timer interrupt count
>> incrementing in /proc/interrupts? Same question for the MMC interrupts.
>
>> Since you are the one having the issue, you will have to do the
>> investigation. If you do not do that, do not expect the issue to
>> magically go away.
>
>
>> So my question is if (or
>> when) there will be a patch for the 3.14 kernel line ?
>
>> I work for the I-pipe patch for the ARM architecture on my free time.
>> So, I do things when I want to, without any deadline, and certainly not
>> under pressure.
>
> Right now I can't test anything under the 3.10.18 kernel since for some weird reason (which has nothing to do with Xenomai) my network and USB (combined on my board) is off. With the same configuration file, the 3.14 kernel works fine.
>

>In 3.10 kernel, a new configuration disabled by default is the USB PHY. 
>You need to select the USB PHY corresponding to your board. If you want 
>to boot using NFS, you have to enable the PHY as built-in in addition to 
>the built-in network controller. I did that for the pandaboard (which 
>uses the same network adapter as Beagle-XM I believe), and it boots just 
>fine using NFS.

I had the 3.10.18 kernel working (with xenomai), including the smsc95xx, usb phy, usb hid, etc. Just the clock didnt work. For some reason now all of USB and networking is off. I already triple-checked all options, and so far no solution. 

Re: [Xenomai] patch for 3.14 kernel ?

[Gilles Chanteperdrix]

On 04/08/2014 03:59 PM, Huub Van Niekerk wrote:
> On Tuesday, April 8, 2014 2:55 PM, Gilles Chanteperdrix
> <gilles.chanteperdrix@xenomai.org> wrote:
>
> On 04/08/2014 02:47 PM, Huub Van Niekerk wrote:
>> Right now I can't test anything under the 3.10.18 kernel since for
>> some weird reason (which has nothing to do with Xenomai) my network
>> and USB (combined on my board) is off. With the same configuration
>> file, the 3.14 kernel works fine.
>>
>
>> In 3.10 kernel, a new configuration disabled by default is the USB
>> PHY. You need to select the USB PHY corresponding to your board. If
>> you want to boot using NFS, you have to enable the PHY as built-in
>> in addition to the built-in network controller. I did that for the
>> pandaboard (which uses the same network adapter as Beagle-XM I
>> believe), and it boots just fine using NFS.
>
> I had the 3.10.18 kernel working (with xenomai), including the
> smsc95xx, usb phy, usb hid, etc. Just the clock didnt work. For some
> reason now all of USB and networking is off. I already triple-checked
> all options, and so far no solution.

If everything runs as module, have you installed the module and run 
depmod -a ?


-- 
					    Gilles.

Re: [Xenomai] patch for 3.14 kernel ?

[Huub Van Niekerk]

On Tuesday, April 8, 2014 5:01 PM, Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org> wrote:
 
On 04/08/2014 03:59 PM, Huub Van Niekerk wrote:
> On Tuesday, April 8, 2014 2:55 PM, Gilles Chanteperdrix
> <gilles.chanteperdrix@xenomai.org> wrote:
>
> On 04/08/2014 02:47 PM, Huub Van Niekerk wrote:
>> Right now I can't test anything under the 3.10.18 kernel since for
>> some weird reason (which has nothing to do with Xenomai) my network
>> and USB (combined on my board) is off. With the same configuration
>> file, the 3.14 kernel works fine.
>>
>
>> In 3.10 kernel, a new configuration disabled by default is the USB
>> PHY. You need to select the USB PHY corresponding to your board. If
>> you want to boot using NFS, you have to enable the PHY as built-in
>> in addition to the built-in network controller. I did that for the
>> pandaboard (which uses the same network adapter as Beagle-XM I
>> believe), and it boots just fine using NFS.
>
> I had the 3.10.18 kernel working (with xenomai), including the
> smsc95xx, usb phy, usb hid, etc. Just the clock didnt work. For some
> reason now all of USB and networking is off. I already triple-checked
> all options, and so far no solution.

>If everything runs as module, have you installed the module and run 
>depmod -a ?

What I can put as built-in, I do. So that's what I did for all of them.

Re: [Xenomai] patch for 3.14 kernel ?

[Gilles Chanteperdrix]

On 04/08/2014 03:59 PM, Huub Van Niekerk wrote:
> I had the 3.10.18 kernel working (with xenomai), including the
> smsc95xx, usb phy, usb hid, etc. Just the clock didnt work.

What does the "clock didn't work" mean, clock_gettime does not return 
monotonically increasing values?

-- 
					    Gilles.

Re: [Xenomai] patch for 3.14 kernel ?

[Huub Van Niekerk]

On Tuesday, April 8, 2014 5:02 PM, Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org> wrote:
 
On 04/08/2014 03:59 PM, Huub Van Niekerk wrote:
> I had the 3.10.18 kernel working (with xenomai), including the
> smsc95xx, usb phy, usb hid, etc. Just the clock didnt work.

>What does the "clock didn't work" mean, clock_gettime does not return 
>monotonically increasing values?

Correct.

Re: [Xenomai] patch for 3.14 kernel ?

[Gilles Chanteperdrix]

On 04/08/2014 03:59 PM, Huub Van Niekerk wrote:
> On Tuesday, April 8, 2014 2:55 PM, Gilles Chanteperdrix
> <gilles.chanteperdrix@xenomai.org> wrote:
> 
> On 04/08/2014 02:47 PM, Huub Van Niekerk wrote:
>> On Monday, April 7, 2014 2:51 PM, Gilles Chanteperdrix
>> <gilles.chanteperdrix@xenomai.org> wrote:
>> 
>> On 04/07/2014 02:21 PM, Huub Van Niekerk wrote:
>>> Hi,
>>> 
>>> After struggling with the 3.10.18 kernel on armhf (both with and 
>>> without the patch), I tried the 3.14 kernel and to my delight
>>> some issues are solved: the clock runs fine again, roscore and
>>> top work great, while they kept failing on 3.10.18.
>> 
>>> I would say you have a timer issue, or perhaps the same issue
>>> with SD-card as Andreas. Have you tried turning off the options I
>>> said Adreas to disable yesterday? (dynamic ticks and high res
>>> timers)?
>> 
>>> When the system is stuck, do you still see the timer interrupt
>>> count incrementing in /proc/interrupts? Same question for the MMC
>>> interrupts.
>> 
>>> Since you are the one having the issue, you will have to do the 
>>> investigation. If you do not do that, do not expect the issue to 
>>> magically go away.
>> 
>> 
>>> So my question is if (or when) there will be a patch for the 3.14
>>> kernel line ?
>> 
>>> I work for the I-pipe patch for the ARM architecture on my free
>>> time. So, I do things when I want to, without any deadline, and
>>> certainly not under pressure.
>> 
>> Right now I can't test anything under the 3.10.18 kernel since for
>> some weird reason (which has nothing to do with Xenomai) my network
>> and USB (combined on my board) is off. With the same configuration
>> file, the 3.14 kernel works fine.
>> 
> 
>> In 3.10 kernel, a new configuration disabled by default is the USB
>> PHY. You need to select the USB PHY corresponding to your board. If
>> you want to boot using NFS, you have to enable the PHY as built-in
>> in addition to the built-in network controller. I did that for the
>> pandaboard (which uses the same network adapter as Beagle-XM I
>> believe), and it boots just fine using NFS.
> 
> I had the 3.10.18 kernel working (with xenomai), including the
> smsc95xx, usb phy, usb hid, etc. Just the clock didnt work. For some
> reason now all of USB and networking is off. I already triple-checked
> all options, and so far no solution.

Just had an issue ressembling this on omap4: booting the panda board
with 3.10 using DT fails;  USB is never started, and network controller
never discovered, which since I am booting with NFS means kernel panic.
Booting with the old style board file works, on the other hand. All
other configuration options being the same. I do not know if this is
specific to Xenomai or not.

-- 
                                                                Gilles.

Re: [Xenomai] patch for 3.14 kernel ?

[Philippe Gerum]

On 04/07/2014 02:21 PM, Huub Van Niekerk wrote:
> Hi,
>
> After struggling with the 3.10.18 kernel on armhf (both with and without the patch), I tried the 3.14 kernel and to my delight some issues are solved: the clock runs fine again, roscore and top work great, while they kept failing on 3.10.18. So my question is if (or when) there will be a patch for the 3.14 kernel line ?
>

raw/ipipe-3.14.0 is currently work in progress. ppc6xx and x86 32/64 
boot and run the basic tests here, but mainline is so much in a state of 
flux for ARM with the dt introduction, that Gilles needs time to fix up 
the pipeline for several SoCs. He's enduring most of the pain these days 
when it comes to upgrading to recent kernel releases, even worse than I 
used to suffer for x86 a few years ago, so please be patient.

We will keep the list posted.

-- 
Philippe.