* Re: [PATCH 00/02] ARM: shmobile: R-car Gen2 CMA reservation support
2014-06-09 12:38 [PATCH 00/02] ARM: shmobile: R-car Gen2 CMA reservation support Magnus Damm
@ 2014-06-10 1:37 ` Simon Horman
2014-06-10 4:10 ` Magnus Damm
2014-06-10 6:57 ` Simon Horman
2 siblings, 0 replies; 4+ messages in thread
From: Simon Horman @ 2014-06-10 1:37 UTC (permalink / raw)
To: linux-sh
On Mon, Jun 09, 2014 at 09:38:37PM +0900, Magnus Damm wrote:
> ARM: shmobile: R-car Gen2 CMA reservation support
>
> [PATCH 01/02] ARM: shmobile: Add shared R-Car Gen2 CMA reservation code
> [PATCH 02/02] ARM: shmobile: Enable R-Car Gen2 CMA code in board files
>
> These patches add CMA reservation code for R-Car Gen2 SoCs and boards.
>
> Instead of using custom memory reservation hacks for demos and other
> things, make use of standard CMA to reserve a pool of memory. This memory
> is suitable for devices that take physically contiguous memory such as
> DU and VSP1 multimedia devices. At this point no device assignment is
> included - only memory reservation code is added for now.
>
> All R-Car Gen2 SoCs so far have LPAE support and boards tend to use more
> than 32-bits of physical addresses to access the system memory. These
> patches add code that during run time parses the DT memory banks and nick
> 256MiB from the highest memory bank within the legacy 32-bit physical memory
> space and use that for CMA.
>
> On Lager and Koelsch boards these 256 MiB is at 0x7000000 -> 0x7fffffff.
>
> The code uses dynamic calculation of the CMA area base address over static
> DT reservation since this kind of configuration is based on software policy.
> DT is however still used to describe the system memory banks as expected.
> Software memory partitioning is excluded from DT by design.
>
> Signed-off-by: Magnus Damm <damm+renesas@opensource.se>
Hi Magnus,
this seems clean to me. Shall I queue it up?
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH 00/02] ARM: shmobile: R-car Gen2 CMA reservation support
2014-06-09 12:38 [PATCH 00/02] ARM: shmobile: R-car Gen2 CMA reservation support Magnus Damm
2014-06-10 1:37 ` Simon Horman
@ 2014-06-10 4:10 ` Magnus Damm
2014-06-10 6:57 ` Simon Horman
2 siblings, 0 replies; 4+ messages in thread
From: Magnus Damm @ 2014-06-10 4:10 UTC (permalink / raw)
To: linux-sh
Hi Simon,
On Tue, Jun 10, 2014 at 10:37 AM, Simon Horman <horms@verge.net.au> wrote:
> On Mon, Jun 09, 2014 at 09:38:37PM +0900, Magnus Damm wrote:
>> ARM: shmobile: R-car Gen2 CMA reservation support
>>
>> [PATCH 01/02] ARM: shmobile: Add shared R-Car Gen2 CMA reservation code
>> [PATCH 02/02] ARM: shmobile: Enable R-Car Gen2 CMA code in board files
>>
>> These patches add CMA reservation code for R-Car Gen2 SoCs and boards.
>>
>> Instead of using custom memory reservation hacks for demos and other
>> things, make use of standard CMA to reserve a pool of memory. This memory
>> is suitable for devices that take physically contiguous memory such as
>> DU and VSP1 multimedia devices. At this point no device assignment is
>> included - only memory reservation code is added for now.
>>
>> All R-Car Gen2 SoCs so far have LPAE support and boards tend to use more
>> than 32-bits of physical addresses to access the system memory. These
>> patches add code that during run time parses the DT memory banks and nick
>> 256MiB from the highest memory bank within the legacy 32-bit physical memory
>> space and use that for CMA.
>>
>> On Lager and Koelsch boards these 256 MiB is at 0x7000000 -> 0x7fffffff.
>>
>> The code uses dynamic calculation of the CMA area base address over static
>> DT reservation since this kind of configuration is based on software policy.
>> DT is however still used to describe the system memory banks as expected.
>> Software memory partitioning is excluded from DT by design.
>>
>> Signed-off-by: Magnus Damm <damm+renesas@opensource.se>
>
> Hi Magnus,
>
> this seems clean to me. Shall I queue it up?
Thanks, yes please queue up!
/ magnus
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH 00/02] ARM: shmobile: R-car Gen2 CMA reservation support
2014-06-09 12:38 [PATCH 00/02] ARM: shmobile: R-car Gen2 CMA reservation support Magnus Damm
2014-06-10 1:37 ` Simon Horman
2014-06-10 4:10 ` Magnus Damm
@ 2014-06-10 6:57 ` Simon Horman
2 siblings, 0 replies; 4+ messages in thread
From: Simon Horman @ 2014-06-10 6:57 UTC (permalink / raw)
To: linux-sh
On Tue, Jun 10, 2014 at 01:10:39PM +0900, Magnus Damm wrote:
> Hi Simon,
>
> On Tue, Jun 10, 2014 at 10:37 AM, Simon Horman <horms@verge.net.au> wrote:
> > On Mon, Jun 09, 2014 at 09:38:37PM +0900, Magnus Damm wrote:
> >> ARM: shmobile: R-car Gen2 CMA reservation support
> >>
> >> [PATCH 01/02] ARM: shmobile: Add shared R-Car Gen2 CMA reservation code
> >> [PATCH 02/02] ARM: shmobile: Enable R-Car Gen2 CMA code in board files
> >>
> >> These patches add CMA reservation code for R-Car Gen2 SoCs and boards.
> >>
> >> Instead of using custom memory reservation hacks for demos and other
> >> things, make use of standard CMA to reserve a pool of memory. This memory
> >> is suitable for devices that take physically contiguous memory such as
> >> DU and VSP1 multimedia devices. At this point no device assignment is
> >> included - only memory reservation code is added for now.
> >>
> >> All R-Car Gen2 SoCs so far have LPAE support and boards tend to use more
> >> than 32-bits of physical addresses to access the system memory. These
> >> patches add code that during run time parses the DT memory banks and nick
> >> 256MiB from the highest memory bank within the legacy 32-bit physical memory
> >> space and use that for CMA.
> >>
> >> On Lager and Koelsch boards these 256 MiB is at 0x7000000 -> 0x7fffffff.
> >>
> >> The code uses dynamic calculation of the CMA area base address over static
> >> DT reservation since this kind of configuration is based on software policy.
> >> DT is however still used to describe the system memory banks as expected.
> >> Software memory partitioning is excluded from DT by design.
> >>
> >> Signed-off-by: Magnus Damm <damm+renesas@opensource.se>
> >
> > Hi Magnus,
> >
> > this seems clean to me. Shall I queue it up?
>
> Thanks, yes please queue up!
Thanks, done.
^ permalink raw reply [flat|nested] 4+ messages in thread