Re: Intel IOMMU (and IOMMU for Virtualization) performances

[Grant Grundler]

On Fri, Jun 6, 2008 at 2:36 PM, Muli Ben-Yehuda <muli@xxxxxxxxxx> wrote:
...
> The logic seems pretty simple to me, all we need to do is keep track
> of how many entries are currently allocated in the bitmap vs. the size
> of the bitmap. Once we get to the half-way point, double the size of
> the bitmap.

"half-full" would be one hueristic but others might be more suitable
(e.g. wait until it's
nearly full or bitmap search fails).

> Having said that, I'm not sure it's worth it, and have no
> plans to implement it :-)

The historical DMA mapping "failure mode" is a kernel panic.  Resizing or
creating  a new IOMMU page table would avoid a panic, my guess is people
would prefer _anything_ that avoids a panic. :) Especially in the context of
Virtual Machines and multiply guests.

And I'm not exposed to this problem (yet)...I expect Mark Gross
(Intel) is on the
hook for this and has access to all the documentation.

I'll let him thank you for suggesting more work. :P

thanks,
grant
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/