[PATCH] ARM: davinci_all_defconfig: enable CMA options

[PATCH] ARM: davinci_all_defconfig: enable CMA options

[Suman Anna]

Enable the CMA and DMA_CMA Kconfig options by default for
Davinci platforms. Davinci remoteproc driver is one of the
modules that depends on these options, and this allows the
driver to be made visible for selection with menuconfig.

Signed-off-by: Suman Anna <s-anna@ti.com>
---
Hi Sekhar,

Here's the patch that enables CMA as per your request. Patch based
on your v4.15/defconfig branch.

regards
Suman

 arch/arm/configs/davinci_all_defconfig | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/arch/arm/configs/davinci_all_defconfig b/arch/arm/configs/davinci_all_defconfig
index 4d432cf1ee79..bd0cf22f9ceb 100644
--- a/arch/arm/configs/davinci_all_defconfig
+++ b/arch/arm/configs/davinci_all_defconfig
@@ -34,6 +34,7 @@ CONFIG_DAVINCI_MUX_WARNINGS=y
 CONFIG_DAVINCI_RESET_CLOCKS=y
 CONFIG_PREEMPT=y
 CONFIG_AEABI=y
+CONFIG_CMA=y
 CONFIG_SECCOMP=y
 CONFIG_ZBOOT_ROM_TEXT=0x0
 CONFIG_ZBOOT_ROM_BSS=0x0
@@ -56,6 +57,7 @@ CONFIG_NETFILTER=y
 CONFIG_DEVTMPFS=y
 CONFIG_DEVTMPFS_MOUNT=y
 # CONFIG_FW_LOADER is not set
+CONFIG_DMA_CMA=y
 CONFIG_DA8XX_MSTPRI=y
 CONFIG_MTD=m
 CONFIG_MTD_TESTS=m
-- 
2.13.1

Re: [PATCH] ARM: davinci_all_defconfig: enable CMA options

[Sekhar Nori]

On Wednesday 20 September 2017 11:31 PM, Suman Anna wrote:
> Enable the CMA and DMA_CMA Kconfig options by default for
> Davinci platforms. Davinci remoteproc driver is one of the
> modules that depends on these options, and this allows the
> driver to be made visible for selection with menuconfig.
> 
> Signed-off-by: Suman Anna <s-anna@ti.com>
> ---
> Hi Sekhar,
> 
> Here's the patch that enables CMA as per your request. Patch based
> on your v4.15/defconfig branch.

Applied! I guess you will be sending another patch enabling the
remoteproc driver itself?

Thanks,
Sekhar

Re: [PATCH] ARM: davinci_all_defconfig: enable CMA options

[Suman Anna]

On 09/21/2017 09:43 AM, Sekhar Nori wrote:
> On Wednesday 20 September 2017 11:31 PM, Suman Anna wrote:
>> Enable the CMA and DMA_CMA Kconfig options by default for
>> Davinci platforms. Davinci remoteproc driver is one of the
>> modules that depends on these options, and this allows the
>> driver to be made visible for selection with menuconfig.
>>
>> Signed-off-by: Suman Anna <s-anna@ti.com>
>> ---
>> Hi Sekhar,
>>
>> Here's the patch that enables CMA as per your request. Patch based
>> on your v4.15/defconfig branch.
> 
> Applied! I guess you will be sending another patch enabling the
> remoteproc driver itself?

Thanks. Well, one always needs a firmware image in their filesystem
without which enabling the driver is meaningless (it would fail). So was
not planning to enable it by default.

regards
Suman

Re: [PATCH] ARM: davinci_all_defconfig: enable CMA options

[Sekhar Nori]

Hi Suman,

On Thursday 21 September 2017 08:41 PM, Suman Anna wrote:
> On 09/21/2017 09:43 AM, Sekhar Nori wrote:
>> On Wednesday 20 September 2017 11:31 PM, Suman Anna wrote:
>>> Enable the CMA and DMA_CMA Kconfig options by default for
>>> Davinci platforms. Davinci remoteproc driver is one of the
>>> modules that depends on these options, and this allows the
>>> driver to be made visible for selection with menuconfig.
>>>
>>> Signed-off-by: Suman Anna <s-anna@ti.com>
>>> ---
>>> Hi Sekhar,
>>>
>>> Here's the patch that enables CMA as per your request. Patch based
>>> on your v4.15/defconfig branch.
>>
>> Applied! I guess you will be sending another patch enabling the
>> remoteproc driver itself?
> 
> Thanks. Well, one always needs a firmware image in their filesystem
> without which enabling the driver is meaningless (it would fail). So was
> not planning to enable it by default.

But the errors will show up only on DA850 LCDK where we have enabled the
DSP by default in DT because its an important usecase there. I think it
is important for users to see those errors.

(besides the usual benefits of having the module enabled to catch build
and runtime issues easily).

Thanks,
Sekhar

Re: [PATCH] ARM: davinci_all_defconfig: enable CMA options

[Suman Anna]

Hi Sekhar,

> 
> On Thursday 21 September 2017 08:41 PM, Suman Anna wrote:
>> On 09/21/2017 09:43 AM, Sekhar Nori wrote:
>>> On Wednesday 20 September 2017 11:31 PM, Suman Anna wrote:
>>>> Enable the CMA and DMA_CMA Kconfig options by default for
>>>> Davinci platforms. Davinci remoteproc driver is one of the
>>>> modules that depends on these options, and this allows the
>>>> driver to be made visible for selection with menuconfig.
>>>>
>>>> Signed-off-by: Suman Anna <s-anna@ti.com>
>>>> ---
>>>> Hi Sekhar,
>>>>
>>>> Here's the patch that enables CMA as per your request. Patch based
>>>> on your v4.15/defconfig branch.
>>>
>>> Applied! I guess you will be sending another patch enabling the
>>> remoteproc driver itself?
>>
>> Thanks. Well, one always needs a firmware image in their filesystem
>> without which enabling the driver is meaningless (it would fail). So was
>> not planning to enable it by default.
> 
> But the errors will show up only on DA850 LCDK where we have enabled the
> DSP by default in DT because its an important usecase there. I think it
> is important for users to see those errors.
> (besides the usual benefits of having the module enabled to catch build
> and runtime issues easily).

Yeah, DSP node is enabled only for DA850 LCDK among the DT-supported
boards, but I believe there are boards still using legacy boot, and it
shows up on those platforms as well. FYI, we do not have any rpmsg
drivers other than the kernel sample to make real usage of a booted
remoteproc. Anycase, if you still prefer this to be added, I don't mind.

regards
Suman

Re: [PATCH] ARM: davinci_all_defconfig: enable CMA options

[David Lechner]

On 09/22/2017 11:24 AM, Suman Anna wrote:
> Hi Sekhar,
> 
>>
>> On Thursday 21 September 2017 08:41 PM, Suman Anna wrote:
>>> On 09/21/2017 09:43 AM, Sekhar Nori wrote:
>>>> On Wednesday 20 September 2017 11:31 PM, Suman Anna wrote:
>>>>> Enable the CMA and DMA_CMA Kconfig options by default for
>>>>> Davinci platforms. Davinci remoteproc driver is one of the
>>>>> modules that depends on these options, and this allows the
>>>>> driver to be made visible for selection with menuconfig.
>>>>>
>>>>> Signed-off-by: Suman Anna <s-anna@ti.com>
>>>>> ---
>>>>> Hi Sekhar,
>>>>>
>>>>> Here's the patch that enables CMA as per your request. Patch based
>>>>> on your v4.15/defconfig branch.
>>>>
>>>> Applied! I guess you will be sending another patch enabling the
>>>> remoteproc driver itself?
>>>
>>> Thanks. Well, one always needs a firmware image in their filesystem
>>> without which enabling the driver is meaningless (it would fail). So was
>>> not planning to enable it by default.
>>
>> But the errors will show up only on DA850 LCDK where we have enabled the
>> DSP by default in DT because its an important usecase there. I think it
>> is important for users to see those errors.
>> (besides the usual benefits of having the module enabled to catch build
>> and runtime issues easily).
> 
> Yeah, DSP node is enabled only for DA850 LCDK among the DT-supported
> boards, but I believe there are boards still using legacy boot, and it
> shows up on those platforms as well. FYI, we do not have any rpmsg
> drivers other than the kernel sample to make real usage of a booted
> remoteproc. Anycase, if you still prefer this to be added, I don't mind.
> 
> regards
> Suman
> 

I am interested in using remoteproc on LEGO MINDSTORMS EV3 to implement 
I2C and UART in the PRU.

Re: [PATCH] ARM: davinci_all_defconfig: enable CMA options

[Sekhar Nori]

Hi David,

On Friday 22 September 2017 10:13 PM, David Lechner wrote:
> On 09/22/2017 11:24 AM, Suman Anna wrote:
>> Hi Sekhar,
>>
>>>
>>> On Thursday 21 September 2017 08:41 PM, Suman Anna wrote:
>>>> On 09/21/2017 09:43 AM, Sekhar Nori wrote:
>>>>> On Wednesday 20 September 2017 11:31 PM, Suman Anna wrote:
>>>>>> Enable the CMA and DMA_CMA Kconfig options by default for
>>>>>> Davinci platforms. Davinci remoteproc driver is one of the
>>>>>> modules that depends on these options, and this allows the
>>>>>> driver to be made visible for selection with menuconfig.
>>>>>>
>>>>>> Signed-off-by: Suman Anna <s-anna@ti.com>
>>>>>> ---
>>>>>> Hi Sekhar,
>>>>>>
>>>>>> Here's the patch that enables CMA as per your request. Patch based
>>>>>> on your v4.15/defconfig branch.
>>>>>
>>>>> Applied! I guess you will be sending another patch enabling the
>>>>> remoteproc driver itself?
>>>>
>>>> Thanks. Well, one always needs a firmware image in their filesystem
>>>> without which enabling the driver is meaningless (it would fail). So
>>>> was
>>>> not planning to enable it by default.
>>>
>>> But the errors will show up only on DA850 LCDK where we have enabled the
>>> DSP by default in DT because its an important usecase there. I think it
>>> is important for users to see those errors.
>>> (besides the usual benefits of having the module enabled to catch build
>>> and runtime issues easily).
>>
>> Yeah, DSP node is enabled only for DA850 LCDK among the DT-supported
>> boards, but I believe there are boards still using legacy boot, and it
>> shows up on those platforms as well. FYI, we do not have any rpmsg
>> drivers other than the kernel sample to make real usage of a booted
>> remoteproc. Anycase, if you still prefer this to be added, I don't mind.
>>
>> regards
>> Suman
>>
> 
> I am interested in using remoteproc on LEGO MINDSTORMS EV3 to implement
> I2C and UART in the PRU.

Do note that the remoteproc driver being discussed here is the DSP
remoteproc driver. TI releases have a remoteproc driver that talks to
PRUSS[1], but that has not been used to talk to PRUSS on OMAP-L138. And
AFAIK, there is no one working towards that in TI as well.

Thanks,
Sekhar

[1]
http://git.ti.com/cgit/cgit.cgi/ti-linux-kernel/ti-linux-kernel.git/tree/drivers/remoteproc/pru_rproc.c?h=ti-linux-4.9.y

Re: [PATCH] ARM: davinci_all_defconfig: enable CMA options

[David Lechner]

On 09/26/2017 04:56 AM, Sekhar Nori wrote:
> Hi David,
> 

>>
>> I am interested in using remoteproc on LEGO MINDSTORMS EV3 to implement
>> I2C and UART in the PRU.
> 
> Do note that the remoteproc driver being discussed here is the DSP
> remoteproc driver. TI releases have a remoteproc driver that talks to
> PRUSS[1], but that has not been used to talk to PRUSS on OMAP-L138. And
> AFAIK, there is no one working towards that in TI as well.
> 
> Thanks,
> Sekhar
> 
> [1]
> http://git.ti.com/cgit/cgit.cgi/ti-linux-kernel/ti-linux-kernel.git/tree/drivers/remoteproc/pru_rproc.c?h=ti-linux-4.9.y
> 

Darn. I've used pru_rproc on the BeagleBone. One more thing for the EV3 
list ;-).

Re: [PATCH] ARM: davinci_all_defconfig: enable CMA options

[Sekhar Nori]

On Friday 22 September 2017 09:54 PM, Suman Anna wrote:
> Hi Sekhar,
> 
>>
>> On Thursday 21 September 2017 08:41 PM, Suman Anna wrote:
>>> On 09/21/2017 09:43 AM, Sekhar Nori wrote:
>>>> On Wednesday 20 September 2017 11:31 PM, Suman Anna wrote:
>>>>> Enable the CMA and DMA_CMA Kconfig options by default for
>>>>> Davinci platforms. Davinci remoteproc driver is one of the
>>>>> modules that depends on these options, and this allows the
>>>>> driver to be made visible for selection with menuconfig.
>>>>>
>>>>> Signed-off-by: Suman Anna <s-anna@ti.com>
>>>>> ---
>>>>> Hi Sekhar,
>>>>>
>>>>> Here's the patch that enables CMA as per your request. Patch based
>>>>> on your v4.15/defconfig branch.
>>>>
>>>> Applied! I guess you will be sending another patch enabling the
>>>> remoteproc driver itself?
>>>
>>> Thanks. Well, one always needs a firmware image in their filesystem
>>> without which enabling the driver is meaningless (it would fail). So was
>>> not planning to enable it by default.
>>
>> But the errors will show up only on DA850 LCDK where we have enabled the
>> DSP by default in DT because its an important usecase there. I think it
>> is important for users to see those errors.
>> (besides the usual benefits of having the module enabled to catch build
>> and runtime issues easily).
> 
> Yeah, DSP node is enabled only for DA850 LCDK among the DT-supported
> boards, but I believe there are boards still using legacy boot, and it
> shows up on those platforms as well. FYI, we do not have any rpmsg

Thats right. Even there, my thinking is if a board is registering a
platform device, its because it wants to use it. And so the driver
should be enabled.

> drivers other than the kernel sample to make real usage of a booted
> remoteproc. Anycase, if you still prefer this to be added, I don't mind.

I think its better to enable the driver too. Please submit a patch. Thanks!

Regards,
Sekhar