[Nouveau] [PATCH v2 6/6] instmem/gk20a: add IOMMU support

[Alexandre Courbot]

On Wed, Feb 11, 2015 at 5:03 PM, Ilia Mirkin <imirkin at alum.mit.edu> wrote:
> On Wed, Feb 11, 2015 at 2:21 AM, Alexandre Courbot <gnurou at gmail.com> wrote:
>> Let GK20A's instmem take advantage of the IOMMU if it is present. Having
>> an IOMMU means that instmem is no longer allocated using the DMA API,
>> but instead obtained through page_alloc and made contiguous to the GPU
>> by IOMMU mappings.
>>
>> Signed-off-by: Alexandre Courbot <acourbot at nvidia.com>
>> ---
>>  drm/nouveau/nvkm/subdev/instmem/gk20a.c | 272 ++++++++++++++++++++++++++++----
>>  1 file changed, 241 insertions(+), 31 deletions(-)
>>
>> diff --git a/drm/nouveau/nvkm/subdev/instmem/gk20a.c b/drm/nouveau/nvkm/subdev/instmem/gk20a.c
>> index 4c8af6e..a20b93c 100644
>> --- a/drm/nouveau/nvkm/subdev/instmem/gk20a.c
>> +++ b/drm/nouveau/nvkm/subdev/instmem/gk20a.c
>> @@ -20,12 +20,32 @@
>>   * DEALINGS IN THE SOFTWARE.
>>   */
>>
>> +/*
>> + * GK20A does not have dedicated video memory, and to accurately represent this
>> + * fact Nouveau will not create a RAM device for it. Therefore its instmem
>> + * implementation must be done directly on top of system memory, while providing
>> + * coherent read and write operations.
>> + *
>> + * Instmem can be allocated through two means:
>> + * 1) If a IOMMU mapping has been probed, the IOMMU API is used to make memory
>
> an IOMMU mapping
>
>> + *    pages contiguous to the GPU. This is the preferred way.
>> + * 2) If no IOMMU mapping is probed, the DMA API is used to allocated physically
>
> to allocate physically

Both fixed, thanks!

>
>> + *    contiguous memory.
>> + *
>> + * In both cases CPU read and writes are performed using PRAMIN (i.e. using the
>> + * GPU path) to ensure these operations are coherent for the GPU. This allows us
>> + * to use more "relaxed" allocation parameters when using the DMA API, since we
>> + * never need a kernel mapping.
>> + */
>
> Er yeah... obviously ignore my comments about sharing nv50_instmem_priv stuff.

Yeah, actually these functions are not even supposed to stay in the
long term. I am using them because it solves the coherency issues, but
later I'd like to replace them with a temporary coherent CPU mapping
if I can figure out a way to do it...

Thanks for the review!