[openib-general] mthca and DDR not hidden

[Michael S. Tsirkin]

Hello!
Quoting r. Roland Dreier (roland at topspin.com) "Re: [openib-general] mthca and DDR not hidden":
>     Sean> Having an allocator routine might force users to perform
>     Sean> data copies when sending data.
> 
>     Sean> Do all of the existing MAD implementations have routines to
>     Sean> allocate MADs when sending data, and require those routines
>     Sean> to be used?
> 
> Not Topspin's.
> 
> I think moving allocation into the MAD layer is probably a bad idea,
> especially if the only motivation is to handle a DMA address of
> 0xffffffffffffffff for Tavor (I don't think that will ever happen in
> practice).

Actually saw some mail on lkml the other day about a 64 bit system
where memory is fragmented - PCI at 0 and actual memory near -1.
Anyway, Tavor is likely not only device with restrictions on
maximum region size?

> The argument about IOMMU resources seems to be an argument
> in favor of letting the consumer handle allocation.
> 
>  - R.

Are you saying consumer shall play with IOMMU? But then
PCI addresses of memory will change each time consumer
does this and you wont be able to use your default region?


MSt