Help debugging CF support under 3.0.18

Help debugging CF support under 3.0.18

[Chris Tapp]

I've been trying to get a BSP running under 6.0.1 using kernel 3.0.18 for an ALIX 3D3 board. I've done this before under 4.0.1 using 2.6.34.

It's nearly working, but I can't get partitions on the CF card (which works under 2.6.34) to show.

- The controller (CS5536) is showing as loaded in dmesg.
- hda is reported as using UDMA/33, but hda1 is not mentioned.
- /proc/partions does not show hda1.
- cat'ing /proc/ide/ide0/hda/model reports 'CF 1GB', which is what the CF is.
- The following kernel options are on : CONFIG_PCI, CONFIG_PCI_QUIRKS, CONFIG_BLK_IDEPCI, CONFIG_IDEPCI_PCIBUS_ORDER, CONFIG_BLK_DEV_IDEDMA.

I suspect I've turned 'something' off in the kernel configuration that I need, but it's not obvious (to me) what it could be. What's the best way of tracking this down? I'll start again using a base .config file, but it takes a long time to rebuild the kernel (I need a faster machine!) and it'll take a while to work out what I've touched in error.

On a related note, I use SRCREV_machine_pn-linux-yocto_mybsp ?= "d7e81e7f975c57c581ce13446adf023f95d9fd9f" in my kernel bbappend (base 3.0.18 kernel). If I do a 'cleanall' and then build I have to wait for the kernel to be downloaded, even though I have a DL_DIR specified. Is it possible to stop this happening? If not, is there a better way to rebuild the kernel and get a new image in deploy/install?

Chris Tapp

opensource@keylevel.com
www.keylevel.com

Re: Help debugging CF support under 3.0.18

[Gary Thomas]

On 2012-04-10 11:03, Chris Tapp wrote:
> I've been trying to get a BSP running under 6.0.1 using kernel 3.0.18 for an ALIX 3D3 board. I've done this before under 4.0.1 using 2.6.34.
>
> It's nearly working, but I can't get partitions on the CF card (which works under 2.6.34) to show.
>
> - The controller (CS5536) is showing as loaded in dmesg.
> - hda is reported as using UDMA/33, but hda1 is not mentioned.
> - /proc/partions does not show hda1.
> - cat'ing /proc/ide/ide0/hda/model reports 'CF 1GB', which is what the CF is.
> - The following kernel options are on : CONFIG_PCI, CONFIG_PCI_QUIRKS, CONFIG_BLK_IDEPCI, CONFIG_IDEPCI_PCIBUS_ORDER, CONFIG_BLK_DEV_IDEDMA.
>
> I suspect I've turned 'something' off in the kernel configuration that I need, but it's not obvious (to me) what it could be. What's the best way of tracking this down? I'll start again using a base .config file, but it takes a long time to rebuild the kernel (I need a faster machine!) and it'll take a while to work out what I've touched in error.
>
> On a related note, I use SRCREV_machine_pn-linux-yocto_mybsp ?= "d7e81e7f975c57c581ce13446adf023f95d9fd9f" in my kernel bbappend (base 3.0.18 kernel). If I do a 'cleanall' and then build I have to wait for the kernel to be downloaded, even though I have a DL_DIR specified. Is it possible to stop this happening? If not, is there a better way to rebuild the kernel and get a new image in deploy/install?

Do you have CONFIG_HOTPLUG=y?

-- 
------------------------------------------------------------
Gary Thomas                 |  Consulting for the
MLB Associates              |    Embedded world
------------------------------------------------------------

Re: Help debugging CF support under 3.0.18

[Chris Tapp]

On 10 Apr 2012, at 18:11, Gary Thomas wrote:

> On 2012-04-10 11:03, Chris Tapp wrote:
>> I've been trying to get a BSP running under 6.0.1 using kernel 3.0.18 for an ALIX 3D3 board. I've done this before under 4.0.1 using 2.6.34.
>> 
>> It's nearly working, but I can't get partitions on the CF card (which works under 2.6.34) to show.
>> 
>> - The controller (CS5536) is showing as loaded in dmesg.
>> - hda is reported as using UDMA/33, but hda1 is not mentioned.
>> - /proc/partions does not show hda1.
>> - cat'ing /proc/ide/ide0/hda/model reports 'CF 1GB', which is what the CF is.
>> - The following kernel options are on : CONFIG_PCI, CONFIG_PCI_QUIRKS, CONFIG_BLK_IDEPCI, CONFIG_IDEPCI_PCIBUS_ORDER, CONFIG_BLK_DEV_IDEDMA.
>> 
>> I suspect I've turned 'something' off in the kernel configuration that I need, but it's not obvious (to me) what it could be. What's the best way of tracking this down? I'll start again using a base .config file, but it takes a long time to rebuild the kernel (I need a faster machine!) and it'll take a while to work out what I've touched in error.
>> 
>> On a related note, I use SRCREV_machine_pn-linux-yocto_mybsp ?= "d7e81e7f975c57c581ce13446adf023f95d9fd9f" in my kernel bbappend (base 3.0.18 kernel). If I do a 'cleanall' and then build I have to wait for the kernel to be downloaded, even though I have a DL_DIR specified. Is it possible to stop this happening? If not, is there a better way to rebuild the kernel and get a new image in deploy/install?
> 
> Do you have CONFIG_HOTPLUG=y?

Yes, that's set. However, CONFIG_HOTPLUG_PCI isn't.

Do I need hotplug? The CF card is the boot media (currently booting from USB while getting this fixed) and is present when the system starts.

Chris

Re: Help debugging CF support under 3.0.18

[Gary Thomas]

On 2012-04-10 13:19, Chris Tapp wrote:
> On 10 Apr 2012, at 18:11, Gary Thomas wrote:
>
>> On 2012-04-10 11:03, Chris Tapp wrote:
>>> I've been trying to get a BSP running under 6.0.1 using kernel 3.0.18 for an ALIX 3D3 board. I've done this before under 4.0.1 using 2.6.34.
>>>
>>> It's nearly working, but I can't get partitions on the CF card (which works under 2.6.34) to show.
>>>
>>> - The controller (CS5536) is showing as loaded in dmesg.
>>> - hda is reported as using UDMA/33, but hda1 is not mentioned.
>>> - /proc/partions does not show hda1.
>>> - cat'ing /proc/ide/ide0/hda/model reports 'CF 1GB', which is what the CF is.
>>> - The following kernel options are on : CONFIG_PCI, CONFIG_PCI_QUIRKS, CONFIG_BLK_IDEPCI, CONFIG_IDEPCI_PCIBUS_ORDER, CONFIG_BLK_DEV_IDEDMA.
>>>
>>> I suspect I've turned 'something' off in the kernel configuration that I need, but it's not obvious (to me) what it could be. What's the best way of tracking this down? I'll start again using a base .config file, but it takes a long time to rebuild the kernel (I need a faster machine!) and it'll take a while to work out what I've touched in error.
>>>
>>> On a related note, I use SRCREV_machine_pn-linux-yocto_mybsp ?= "d7e81e7f975c57c581ce13446adf023f95d9fd9f" in my kernel bbappend (base 3.0.18 kernel). If I do a 'cleanall' and then build I have to wait for the kernel to be downloaded, even though I have a DL_DIR specified. Is it possible to stop this happening? If not, is there a better way to rebuild the kernel and get a new image in deploy/install?
>>
>> Do you have CONFIG_HOTPLUG=y?
>
> Yes, that's set. However, CONFIG_HOTPLUG_PCI isn't.
>
> Do I need hotplug? The CF card is the boot media (currently booting from USB while getting this fixed) and is present when the system starts.

I think so, yes, that's [part of] the machinery the kernel uses to find
and create devices, etc.

-- 
------------------------------------------------------------
Gary Thomas                 |  Consulting for the
MLB Associates              |    Embedded world
------------------------------------------------------------