Patchwork [v2,3/3] ARM: imx_v6_v7_defconfig: Select MXS_DMA

login
register
mail settings
Submitter Fabio Estevam
Date July 25, 2012, 11:18 a.m.
Message ID <1343215100-24005-3-git-send-email-festevam@gmail.com>
Download mbox | patch
Permalink /patch/173137/
State New
Headers show

Comments

Fabio Estevam - July 25, 2012, 11:18 a.m.
From: Fabio Estevam <fabio.estevam@freescale.com>

Let MXS_DMA be selected by default.

Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
---
Changes since v1:
- No changes. Newly introduced in this series
 arch/arm/configs/imx_v6_v7_defconfig |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)
Huang Shijie - July 25, 2012, 3:05 p.m.
Hi Fabio:

On Wed, Jul 25, 2012 at 7:18 AM, Fabio Estevam <festevam@gmail.com> wrote:
> From: Fabio Estevam <fabio.estevam@freescale.com>
>
> Let MXS_DMA be selected by default.
>
> Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
> ---
> Changes since v1:
> - No changes. Newly introduced in this series
>  arch/arm/configs/imx_v6_v7_defconfig |    1 +
>  1 files changed, 1 insertions(+), 0 deletions(-)
>
> diff --git a/arch/arm/configs/imx_v6_v7_defconfig b/arch/arm/configs/imx_v6_v7_defconfig
> index f725b96..3c9f32f 100644
> --- a/arch/arm/configs/imx_v6_v7_defconfig
> +++ b/arch/arm/configs/imx_v6_v7_defconfig
> @@ -192,6 +192,7 @@ CONFIG_RTC_DRV_MC13XXX=y
>  CONFIG_RTC_DRV_MXC=y
>  CONFIG_DMADEVICES=y
>  CONFIG_IMX_SDMA=y
> +CONFIG_MXS_DMA=y
Is this patch needed? The mxs-dma only has one user, the gpmi-nand in mx6q.
this patch makes the kernel built-in the mxs-dma, even the gpmi-nand
is not enabled.

Huang Shijie

>  CONFIG_COMMON_CLK_DEBUG=y
>  # CONFIG_IOMMU_SUPPORT is not set
>  CONFIG_EXT2_FS=y
> --
> 1.7.1
>
>
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Fabio Estevam - July 25, 2012, 3:23 p.m.
On Wed, Jul 25, 2012 at 12:05 PM, Huang Shijie <shijie8@gmail.com> wrote:

> Is this patch needed? The mxs-dma only has one user, the gpmi-nand in mx6q.
> this patch makes the kernel built-in the mxs-dma, even the gpmi-nand
> is not enabled.

Well, as the driver currently stands: yes, it is needed.

Without CONFIG_MXS_DMA, the GPMI driver cannot even be built.

This is why I thought that 'select MXS_DMA' was the way to go, but
according to the thread you pointed me at, it was Nacked by Shawn.

Regards,

Fabio Estevam
Huang Shijie - July 26, 2012, 2:24 a.m.
On Wed, Jul 25, 2012 at 11:23 PM, Fabio Estevam <festevam@gmail.com> wrote:
> On Wed, Jul 25, 2012 at 12:05 PM, Huang Shijie <shijie8@gmail.com> wrote:
>
>> Is this patch needed? The mxs-dma only has one user, the gpmi-nand in mx6q.
>> this patch makes the kernel built-in the mxs-dma, even the gpmi-nand
>> is not enabled.
>
> Well, as the driver currently stands: yes, it is needed.
>
> Without CONFIG_MXS_DMA, the GPMI driver cannot even be built.
But you have added "depend on MXS_DMA". So If you do not select the
mxs-dma, you can not see the
gpmi-nand.
that's why i think this patch is not needed.

Huang Shijie

>
> This is why I thought that 'select MXS_DMA' was the way to go, but
> according to the thread you pointed me at, it was Nacked by Shawn.
>
> Regards,
>
> Fabio Estevam
David Woodhouse - July 26, 2012, 10:45 a.m.
On Thu, 2012-07-26 at 10:24 +0800, Huang Shijie wrote:
> But you have added "depend on MXS_DMA". So If you do not select the
> mxs-dma, you can not see the gpmi-nand. 

That's an issue with the 'make *config' tools. I recommend fixing those
instead; it's horrid to use 'select' on an option which *is*
user-visible, such as CONFIG_MXS_DMA.

In the mid-1990s, there was a variant of the tcl 'make xconfig' which
would *show* you the dependencies for things, and would allow you to
turn them on with a single click. Is that really so hard for us to do?
Fabio Estevam - July 26, 2012, 10:53 a.m.
On Wed, Jul 25, 2012 at 11:24 PM, Huang Shijie <shijie8@gmail.com> wrote:

> But you have added "depend on MXS_DMA". So If you do not select the
> mxs-dma, you can not see the
> gpmi-nand.
> that's why i think this patch is not needed.

So how CONFIG_MXS_DMA would be selected if not via defconfig?
Huang Shijie - July 26, 2012, 10:56 a.m.
On Thu, Jul 26, 2012 at 6:53 AM, Fabio Estevam <festevam@gmail.com> wrote:
> On Wed, Jul 25, 2012 at 11:24 PM, Huang Shijie <shijie8@gmail.com> wrote:
>
>> But you have added "depend on MXS_DMA". So If you do not select the
>> mxs-dma, you can not see the
>> gpmi-nand.
>> that's why i think this patch is not needed.
>
> So how CONFIG_MXS_DMA would be selected if not via defconfig?
Use the make menuconfig ARCH=arm:

"device drivers"-->"DMA engine support" -->"mxs dma support"

Huang Shijie
Fabio Estevam - July 26, 2012, 11:08 a.m.
On Thu, Jul 26, 2012 at 7:56 AM, Huang Shijie <shijie8@gmail.com> wrote:
> On Thu, Jul 26, 2012 at 6:53 AM, Fabio Estevam <festevam@gmail.com> wrote:
>> On Wed, Jul 25, 2012 at 11:24 PM, Huang Shijie <shijie8@gmail.com> wrote:
>>
>>> But you have added "depend on MXS_DMA". So If you do not select the
>>> mxs-dma, you can not see the
>>> gpmi-nand.
>>> that's why i think this patch is not needed.
>>
>> So how CONFIG_MXS_DMA would be selected if not via defconfig?
> Use the make menuconfig ARCH=arm:
>
> "device drivers"-->"DMA engine support" -->"mxs dma support"

I prefer to have selected in the defconfig by default instead.

It is not obvious for the end user that they need to manually select
this option.

mxs_defconfig also selects CONFIG_MXS_DMA.

Can't we do the same here and let imx_v6_v7_defconfig be built again?
Huang Shijie - July 26, 2012, 11:32 a.m.
On Thu, Jul 26, 2012 at 7:08 AM, Fabio Estevam <festevam@gmail.com> wrote:
> On Thu, Jul 26, 2012 at 7:56 AM, Huang Shijie <shijie8@gmail.com> wrote:
>> On Thu, Jul 26, 2012 at 6:53 AM, Fabio Estevam <festevam@gmail.com> wrote:
>>> On Wed, Jul 25, 2012 at 11:24 PM, Huang Shijie <shijie8@gmail.com> wrote:
>>>
>>>> But you have added "depend on MXS_DMA". So If you do not select the
>>>> mxs-dma, you can not see the
>>>> gpmi-nand.
>>>> that's why i think this patch is not needed.
>>>
>>> So how CONFIG_MXS_DMA would be selected if not via defconfig?
>> Use the make menuconfig ARCH=arm:
>>
>> "device drivers"-->"DMA engine support" -->"mxs dma support"
>
> I prefer to have selected in the defconfig by default instead.
>
> It is not obvious for the end user that they need to manually select
> this option.
>
> mxs_defconfig also selects CONFIG_MXS_DMA.
In mx23/mx28, the mxs-dma has many users, so it's proper to make it default.

>
> Can't we do the same here and let imx_v6_v7_defconfig be built again?
Shawn, do you have any opinion about this?

Huang Shijie
David Woodhouse - July 26, 2012, 11:54 a.m.
On Thu, 2012-07-26 at 08:08 -0300, Fabio Estevam wrote:
> It is not obvious for the end user that they need to manually select
> this option.

I repeat:

THIS IS A TOOLS ISSUE. FIX THE TOOLS.
Richard Zhao - July 27, 2012, 9:26 a.m.
On Thu, Jul 26, 2012 at 07:32:54AM -0400, Huang Shijie wrote:
> On Thu, Jul 26, 2012 at 7:08 AM, Fabio Estevam <festevam@gmail.com> wrote:
> > On Thu, Jul 26, 2012 at 7:56 AM, Huang Shijie <shijie8@gmail.com> wrote:
> >> On Thu, Jul 26, 2012 at 6:53 AM, Fabio Estevam <festevam@gmail.com> wrote:
> >>> On Wed, Jul 25, 2012 at 11:24 PM, Huang Shijie <shijie8@gmail.com> wrote:
> >>>
> >>>> But you have added "depend on MXS_DMA". So If you do not select the
> >>>> mxs-dma, you can not see the
> >>>> gpmi-nand.
> >>>> that's why i think this patch is not needed.
> >>>
> >>> So how CONFIG_MXS_DMA would be selected if not via defconfig?
> >> Use the make menuconfig ARCH=arm:
> >>
> >> "device drivers"-->"DMA engine support" -->"mxs dma support"
> >
> > I prefer to have selected in the defconfig by default instead.
> >
> > It is not obvious for the end user that they need to manually select
> > this option.
> >
> > mxs_defconfig also selects CONFIG_MXS_DMA.
> In mx23/mx28, the mxs-dma has many users, so it's proper to make it default.
> 
> >
> > Can't we do the same here and let imx_v6_v7_defconfig be built again?
> Shawn, do you have any opinion about this?
Most drivers should be kernel modules. mxs_dma needs changes to support
building as module.

Thanks
Richard
> 
> Huang Shijie
> 
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
>
Dirk Behme - July 27, 2012, 9:30 a.m.
On 26.07.2012 13:32, Huang Shijie wrote:
> On Thu, Jul 26, 2012 at 7:08 AM, Fabio Estevam <festevam@gmail.com> wrote:
>> On Thu, Jul 26, 2012 at 7:56 AM, Huang Shijie <shijie8@gmail.com> wrote:
>>> On Thu, Jul 26, 2012 at 6:53 AM, Fabio Estevam <festevam@gmail.com> wrote:
>>>> On Wed, Jul 25, 2012 at 11:24 PM, Huang Shijie <shijie8@gmail.com> wrote:
>>>>
>>>>> But you have added "depend on MXS_DMA". So If you do not select the
>>>>> mxs-dma, you can not see the
>>>>> gpmi-nand.
>>>>> that's why i think this patch is not needed.
>>>> So how CONFIG_MXS_DMA would be selected if not via defconfig?
>>> Use the make menuconfig ARCH=arm:
>>>
>>> "device drivers"-->"DMA engine support" -->"mxs dma support"
>> I prefer to have selected in the defconfig by default instead.
>>
>> It is not obvious for the end user that they need to manually select
>> this option.
>>
>> mxs_defconfig also selects CONFIG_MXS_DMA.
> In mx23/mx28, the mxs-dma has many users, so it's proper to make it default.
> 
>> Can't we do the same here and let imx_v6_v7_defconfig be built again?
> Shawn, do you have any opinion about this?

First, I think Shawn is on holiday at the moment.

Second, recent kernel mainline from today building imx_v6_v7_defconfig 
is broken due to this.

So to make it build again, we have to enable CONFIG_MXS_DMA anyhow. As 
it seems that GPMI_NAND needs it to build, I would vote for the

depends on MXS_DMA

fix proposed by Richard

http://www.spinics.net/lists/arm-kernel/msg186111.html

Anyhow, any chance to get the kernel mainline fixed asap?

Many thanks and best regards

Dirk
Huang Shijie - July 27, 2012, 10:21 a.m.
于 2012年07月27日 17:30, Dirk Behme 写道:
> On 26.07.2012 13:32, Huang Shijie wrote:
>> On Thu, Jul 26, 2012 at 7:08 AM, Fabio Estevam <festevam@gmail.com> 
>> wrote:
>>> On Thu, Jul 26, 2012 at 7:56 AM, Huang Shijie <shijie8@gmail.com> 
>>> wrote:
>>>> On Thu, Jul 26, 2012 at 6:53 AM, Fabio Estevam <festevam@gmail.com> 
>>>> wrote:
>>>>> On Wed, Jul 25, 2012 at 11:24 PM, Huang Shijie <shijie8@gmail.com> 
>>>>> wrote:
>>>>>
>>>>>> But you have added "depend on MXS_DMA". So If you do not select the
>>>>>> mxs-dma, you can not see the
>>>>>> gpmi-nand.
>>>>>> that's why i think this patch is not needed.
>>>>> So how CONFIG_MXS_DMA would be selected if not via defconfig?
>>>> Use the make menuconfig ARCH=arm:
>>>>
>>>> "device drivers"-->"DMA engine support" -->"mxs dma support"
>>> I prefer to have selected in the defconfig by default instead.
>>>
>>> It is not obvious for the end user that they need to manually select
>>> this option.
>>>
>>> mxs_defconfig also selects CONFIG_MXS_DMA.
>> In mx23/mx28, the mxs-dma has many users, so it's proper to make it 
>> default.
>>
>>> Can't we do the same here and let imx_v6_v7_defconfig be built again?
>> Shawn, do you have any opinion about this?
>
> First, I think Shawn is on holiday at the moment.
>
> Second, recent kernel mainline from today building imx_v6_v7_defconfig 
> is broken due to this.
Sascha did not CC to me when he submitted the patch 3def84ed9 "ARM: 
imx_v6_v7_defconfig: ..".

I would check this patch if he CCed to me.

BR
Huang Shijie


>
> So to make it build again, we have to enable CONFIG_MXS_DMA anyhow. As 
> it seems that GPMI_NAND needs it to build, I would vote for the
>
> depends on MXS_DMA
>
> fix proposed by Richard
>
> http://www.spinics.net/lists/arm-kernel/msg186111.html
>
> Anyhow, any chance to get the kernel mainline fixed asap?
>
> Many thanks and best regards
>
> Dirk
>
>
Dirk Behme - Aug. 3, 2012, 8:40 a.m.
On 25.07.2012 13:18, Fabio Estevam wrote:
> From: Fabio Estevam <fabio.estevam@freescale.com>
> 
> Let MXS_DMA be selected by default.
> 
> Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
> ---
> Changes since v1:
> - No changes. Newly introduced in this series
>  arch/arm/configs/imx_v6_v7_defconfig |    1 +
>  1 files changed, 1 insertions(+), 0 deletions(-)
> 
> diff --git a/arch/arm/configs/imx_v6_v7_defconfig b/arch/arm/configs/imx_v6_v7_defconfig
> index f725b96..3c9f32f 100644
> --- a/arch/arm/configs/imx_v6_v7_defconfig
> +++ b/arch/arm/configs/imx_v6_v7_defconfig
> @@ -192,6 +192,7 @@ CONFIG_RTC_DRV_MC13XXX=y
>  CONFIG_RTC_DRV_MXC=y
>  CONFIG_DMADEVICES=y
>  CONFIG_IMX_SDMA=y
> +CONFIG_MXS_DMA=y
>  CONFIG_COMMON_CLK_DEBUG=y
>  # CONFIG_IOMMU_SUPPORT is not set
>  CONFIG_EXT2_FS=y

Building mainline 3.6-rc1 with imx_v6_v7_defconfig this is still broken 
due to

drivers/built-in.o: In function `gpmi_dma_filter':
clk-fixed-factor.c:(.text+0xab170): undefined reference to `mxs_dma_is_apbh'

Is any fix for this on its way to mainline? Either the above or any 
other one?

Best regards

Dirk
Huang Shijie - Aug. 3, 2012, 8:59 a.m.
于 2012年08月03日 16:40, Dirk Behme 写道:
> On 25.07.2012 13:18, Fabio Estevam wrote:
>> From: Fabio Estevam <fabio.estevam@freescale.com>
>>
>> Let MXS_DMA be selected by default.
>>
>> Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
>> ---
>> Changes since v1:
>> - No changes. Newly introduced in this series
>> arch/arm/configs/imx_v6_v7_defconfig | 1 +
>> 1 files changed, 1 insertions(+), 0 deletions(-)
>>
>> diff --git a/arch/arm/configs/imx_v6_v7_defconfig 
>> b/arch/arm/configs/imx_v6_v7_defconfig
>> index f725b96..3c9f32f 100644
>> --- a/arch/arm/configs/imx_v6_v7_defconfig
>> +++ b/arch/arm/configs/imx_v6_v7_defconfig
>> @@ -192,6 +192,7 @@ CONFIG_RTC_DRV_MC13XXX=y
>> CONFIG_RTC_DRV_MXC=y
>> CONFIG_DMADEVICES=y
>> CONFIG_IMX_SDMA=y
>> +CONFIG_MXS_DMA=y
>> CONFIG_COMMON_CLK_DEBUG=y
>> # CONFIG_IOMMU_SUPPORT is not set
>> CONFIG_EXT2_FS=y
>
> Building mainline 3.6-rc1 with imx_v6_v7_defconfig this is still 
> broken due to
>
> drivers/built-in.o: In function `gpmi_dma_filter':
> clk-fixed-factor.c:(.text+0xab170): undefined reference to 
> `mxs_dma_is_apbh'
>
> Is any fix for this on its way to mainline? Either the above or any 
> other one?
>
> Best regards
>
> Dirk
>
Since the gpmi has been enabled in the config file, I think it's better 
to select the mxs-dma in the config file too.

Acked-by: Huang Shijie <b32955@freescale.com>

Huang Shijie

Patch

diff --git a/arch/arm/configs/imx_v6_v7_defconfig b/arch/arm/configs/imx_v6_v7_defconfig
index f725b96..3c9f32f 100644
--- a/arch/arm/configs/imx_v6_v7_defconfig
+++ b/arch/arm/configs/imx_v6_v7_defconfig
@@ -192,6 +192,7 @@  CONFIG_RTC_DRV_MC13XXX=y
 CONFIG_RTC_DRV_MXC=y
 CONFIG_DMADEVICES=y
 CONFIG_IMX_SDMA=y
+CONFIG_MXS_DMA=y
 CONFIG_COMMON_CLK_DEBUG=y
 # CONFIG_IOMMU_SUPPORT is not set
 CONFIG_EXT2_FS=y