[PATCH/RFC V2] iommu/ipmmu-vmsa: Disable IPMMU when address expansion is not needed

[PATCH/RFC V2] iommu/ipmmu-vmsa: Disable IPMMU when address expansion is not needed

[Magnus Damm]

From: Magnus Damm <damm+renesas@opensource.se>

Add a memory bank location check to the whitelist handling.

Signed-off-by: Magnus Damm <damm+renesas@opensource.se>
---

 Changes since V1:
 - None, simply broke out this patch from the rest of the series

 drivers/iommu/ipmmu-vmsa.c |    7 +++++++
 1 file changed, 7 insertions(+)

--- 0001/drivers/iommu/ipmmu-vmsa.c
+++ work/drivers/iommu/ipmmu-vmsa.c	2019-03-19 21:04:43.000000000 +0900
@@ -17,6 +17,7 @@
 #include <linux/io.h>
 #include <linux/io-pgtable.h>
 #include <linux/iommu.h>
+#include <linux/memblock.h>
 #include <linux/of.h>
 #include <linux/of_device.h>
 #include <linux/of_iommu.h>
@@ -797,6 +798,12 @@ static bool ipmmu_slave_whitelist(struct
 	if (!soc_device_match(soc_rcar_gen3_whitelist))
 		return false;
 
+	/* In case all system memory fits within 32 bits of physical space
+	 * then assume the IPMMU will not be needed for address expansion.
+	 */
+	if (memblock_end_of_DRAM() <= SZ_4G)
+		return false;
+	
 	/* Check whether this slave device can work with the IPMMU */
 	for (i = 0; i < ARRAY_SIZE(rcar_gen3_slave_whitelist); i++) {
 		if (!strcmp(dev_name(dev), rcar_gen3_slave_whitelist[i]))

Re: [PATCH/RFC V2] iommu/ipmmu-vmsa: Disable IPMMU when address expansion is not needed

[Geert Uytterhoeven]

Hi Magnus,

On Tue, Mar 19, 2019 at 2:24 PM Magnus Damm <magnus.damm@gmail.com> wrote:
> From: Magnus Damm <damm+renesas@opensource.se>
>
> Add a memory bank location check to the whitelist handling.
>
> Signed-off-by: Magnus Damm <damm+renesas@opensource.se>

Thanks for your patch!

> --- 0001/drivers/iommu/ipmmu-vmsa.c
> +++ work/drivers/iommu/ipmmu-vmsa.c     2019-03-19 21:04:43.000000000 +0900
> @@ -17,6 +17,7 @@
>  #include <linux/io.h>
>  #include <linux/io-pgtable.h>
>  #include <linux/iommu.h>
> +#include <linux/memblock.h>
>  #include <linux/of.h>
>  #include <linux/of_device.h>
>  #include <linux/of_iommu.h>
> @@ -797,6 +798,12 @@ static bool ipmmu_slave_whitelist(struct
>         if (!soc_device_match(soc_rcar_gen3_whitelist))
>                 return false;
>
> +       /* In case all system memory fits within 32 bits of physical space
> +        * then assume the IPMMU will not be needed for address expansion.
> +        */
> +       if (memblock_end_of_DRAM() <= SZ_4G)
> +               return false;

The IPMMU may still be useful for IO virtualization, with KVM+VFIO.

Perhaps the check can be enhanced to keep it enabled when the CPU cores
are running in HYP mode?

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [PATCH/RFC V2] iommu/ipmmu-vmsa: Disable IPMMU when address expansion is not needed

[Magnus Damm]

Hi Geert,

On Tue, Mar 19, 2019 at 10:34 PM Geert Uytterhoeven
<geert@linux-m68k.org> wrote:
>
> Hi Magnus,
>
> On Tue, Mar 19, 2019 at 2:24 PM Magnus Damm <magnus.damm@gmail.com> wrote:
> > From: Magnus Damm <damm+renesas@opensource.se>
> >
> > Add a memory bank location check to the whitelist handling.
> >
> > Signed-off-by: Magnus Damm <damm+renesas@opensource.se>
>
> Thanks for your patch!
>
> > --- 0001/drivers/iommu/ipmmu-vmsa.c
> > +++ work/drivers/iommu/ipmmu-vmsa.c     2019-03-19 21:04:43.000000000 +0900
> > @@ -17,6 +17,7 @@
> >  #include <linux/io.h>
> >  #include <linux/io-pgtable.h>
> >  #include <linux/iommu.h>
> > +#include <linux/memblock.h>
> >  #include <linux/of.h>
> >  #include <linux/of_device.h>
> >  #include <linux/of_iommu.h>
> > @@ -797,6 +798,12 @@ static bool ipmmu_slave_whitelist(struct
> >         if (!soc_device_match(soc_rcar_gen3_whitelist))
> >                 return false;
> >
> > +       /* In case all system memory fits within 32 bits of physical space
> > +        * then assume the IPMMU will not be needed for address expansion.
> > +        */
> > +       if (memblock_end_of_DRAM() <= SZ_4G)
> > +               return false;
>
> The IPMMU may still be useful for IO virtualization, with KVM+VFIO.

You're right.

> Perhaps the check can be enhanced to keep it enabled when the CPU cores
> are running in HYP mode?

That might indeed be useful.

Next quarter perhaps you and I can discuss a bit about IO
virtualization with KVM?

Cheers,

/ magnus