Re: [PATCH rdma-core 1/2] mlx5: Use non-cached mapping when Blueflame is not used

[Yishai Hadas <yishaih-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org>]

On 10/30/2017 4:57 PM, Jason Gunthorpe wrote:
> On Sun, Oct 29, 2017 at 11:26:04PM +0200, Yishai Hadas wrote:
>>>> @@ -632,6 +632,11 @@ enum mlx5dv_set_ctx_attr_type {
>>>>   	MLX5DV_CTX_ATTR_BUF_ALLOCATORS = 1,
>>>>   };
>>>> +enum {
>>>> +	MLX5_MMAP_GET_REGULAR_PAGES_CMD	= 0,
>>>> +	MLX5_MMAP_GET_NC_PAGES_CMD	= 3,
>>>> +};
>>>> +
>>>
>>> These are added to a public header, but I couldn't see where they
>>> would get used by a DV user??
>>
>> A DV user already had access to the mlx5 mmap command before this series by
>> using the mlx5dv_get_qp() API:
>> qp_out->uar_mmap_offset = mqp->bf->uar_mmap_offset;
> 
> Okay.
> 
> As long as existing users can reasonably be expected to handle this
> new value in the uar_mmap_offset output field...
> 

Yes, we expect existing users to work properly.

This output field was exposed by the DV API to enable re-mmap of the UAR 
address by a secondary process by using the same 'uar_mmap_offset' 
regardless its internal command value.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html