[openib-general] Lustre over OpenIB Gen2

Hal Rosenstock halr at voltaire.com
Thu Nov 17 12:37:08 PST 2005


On Thu, 2005-11-10 at 14:50, Roland Dreier wrote:
>     Eric>    Or is it better to use FMR pools and take the map/unmap
>     Eric> overhead?  If so, is there a way to know when the unmap
>     Eric> actually hits the hardware and my memory is safe?
> 
> FMRs are only supported on Mellanox HCAs at the moment.  But they do
> have some advantages, like allowing you to convert a bunch of pages
> into a single virtually contiguous region.  You can use the
> ib_flush_fmr_pool() function to make sure that all unmapped FMRs are
> really and truly flushed, but that is a slow operation (since it
> incurs the penalty of flushing all in-flight operations in the HCA).

Assuming one had a machine with a mix of vendor HCAs (Mellanox,
PathScale), how would one determine what hardware/driver was being used 
(assuming one wanted to take advantage of a proprietary feature which
was supported only on one or the other) ? If it that EINVAL or the like
is returned from the ones that don't support that feature ?

-- Hal




More information about the general mailing list