This message is from the T13 list server.
On Mon, Apr 29 2002, Eschmann, Michael K wrote:
> This message is from the T13 list server.
>
>
> Ok, I'll bite.
>
> Can you explain the use model for your suggestion. I may not be thinking of
> the same thing when you say "Ordered Tag Support". If you want to set the
> completion order, then switching to non-queued DMA commands would achieve
> the same effect.
Yes that would work [1], but then I might as well just queue a cache flush
command instead. The goal is to insert a barrier but still allow a full
queue of commands. The NOP allows that, as would an 'ordered' bit
associated with the dma_queued command.
[1] provided it's guarenteed that when the write finishes, the data is
on disk.
--
Jens Axboe
