[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-ppc] [RFC PATCH] spapr_vio/spapr_iommu: Move VIO bypass where
From: |
David Gibson |
Subject: |
Re: [Qemu-ppc] [RFC PATCH] spapr_vio/spapr_iommu: Move VIO bypass where it belongs |
Date: |
Wed, 14 Jan 2015 13:42:58 +1100 |
User-agent: |
Mutt/1.5.23 (2014-03-12) |
On Fri, Jan 09, 2015 at 12:02:51PM +1100, Alexey Kardashevskiy wrote:
> Instead of tweaking a TCE table device by adding there a bypass flag,
> let's add an alias to RAM and IOMMU memory region, and enable/disable
> those according to the selected bypass mode.
> This way IOMMU memory region can have size of the actual window rather
> than ram_size which is essential for upcoming DDW support.
>
> Signed-off-by: Alexey Kardashevskiy <address@hidden>
> ---
>
> The main reason for this patch is DDW and the fact that sPAPRTCETable
> used for DMA windows for VFIO. My latest approach removes all DMA windows
> on the guest reset (and creates a new 32bit one) which means than VFIO
> unmaps everything and this fails as normally sPAPRTCETable MemoryRegion is
> ram_size big (to support bypass) while it should be 1-2GB.
Paolo already mentioned the only significant problem I see with this,
which is migration backwards compat. Otherwise it looks good, sorry I
messed it up in the first place :).
[snip]
> @@ -456,14 +469,25 @@ static int spapr_vio_busdev_init(DeviceState *qdev)
>
> if (pc->rtce_window_size) {
> uint32_t liobn = SPAPR_VIO_BASE_LIOBN | dev->reg;
> +
> + memory_region_init(&dev->mrroot, OBJECT(dev), "iommu-spapr-root",
> + ram_size);
> + memory_region_init_alias(&dev->mrbypass, OBJECT(dev),
> + "iommu-spapr-bypass", get_system_memory(),
> + 0, ram_size);
> + memory_region_add_subregion_overlap(&dev->mrroot, 0, &dev->mrbypass,
> 1);
> + address_space_init(&dev->as, &dev->mrroot, qdev->id);
> +
> dev->tcet = spapr_tce_new_table(qdev, liobn,
> 0,
> SPAPR_TCE_PAGE_SHIFT,
> pc->rtce_window_size >>
> SPAPR_TCE_PAGE_SHIFT, false);
> - address_space_init(&dev->as, spapr_tce_get_iommu(dev->tcet),
> qdev->id);
> + memory_region_add_subregion_overlap(&dev->mrroot, 0,
> + spapr_tce_get_iommu(dev->tcet),
> 2);
> }
>
> +
^^ Tiny nit, extraneous whitespace change.
--
David Gibson | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_
| _way_ _around_!
http://www.ozlabs.org/~dgibson
pgpCa69uJhU2a.pgp
Description: PGP signature