Re: [PATCHv2 03/16] Documentation: dt: add OMAP iommu bindings

Hi Suman,

On Wednesday 26 February 2014 16:28:08 Suman Anna wrote:

On 02/26/2014 04:18 PM, Suman Anna wrote:

quoted

On 02/26/2014 02:36 PM, Laurent Pinchart wrote:

quoted

On Wednesday 26 February 2014 14:23:03 Suman Anna wrote:

quoted

quoted

On Wednesday 26 February 2014 11:02:24 Suman Anna wrote:

quoted

On 02/25/2014 08:13 PM, Laurent Pinchart wrote:

quoted

On Tuesday 25 February 2014 17:02:35 Suman Anna wrote:

quoted

On 02/25/2014 03:26 PM, Laurent Pinchart wrote:

quoted

On Thursday 13 February 2014 12:15:34 Suman Anna wrote:

quoted

From: Florian Vaussard <redacted>

This patch adds the iommu bindings for all OMAP2+ SoCs. Apart from
the standard bindings used by OMAP peripherals, this patch uses a
'dma-window' (already used by Tegra SMMU) and adds two OMAP custom
bindings - 'ti,#tlb-entries' and 'ti,iommu-bus-err-back'.

Signed-off-by: Florian Vaussard <redacted>
[s-anna@ti.com: split bindings document, add dra7 and bus error
back]
Signed-off-by: Suman Anna <redacted>
---

 .../devicetree/bindings/iommu/ti,omap-iommu.txt    | 28 +++++++++
 1 file changed, 28 insertions(+)
 create mode 100644
 Documentation/devicetree/bindings/iommu/ti,omap-iommu.txt

diff --git
a/Documentation/devicetree/bindings/iommu/ti,omap-iommu.txt
b/Documentation/devicetree/bindings/iommu/ti,omap-iommu.txt new
file
mode 100644
index 0000000..116492d

--- /dev/null
+++ b/Documentation/devicetree/bindings/iommu/ti,omap-iommu.txt

@@ -0,0 +1,28 @@
+OMAP2+ IOMMU
+
+Required properties:
+- compatible : Should be one of,
+        "ti,omap2-iommu" for OMAP2/OMAP3 IOMMU instances
+        "ti,omap4-iommu" for OMAP4/OMAP5 IOMMU instances
+        "ti,dra7-iommu" for DRA7xx IOMMU instances
+- ti,hwmods  : Name of the hwmod associated with the IOMMU

instance
+- reg        : Address space for the configuration registers
+- interrupts : Interrupt specifier for the IOMMU instance
+- dma-window : IOVA start address and length

Isn't the dma window more of a system configuration property than a
hardware property ? How do you expect it to be set?

We are setting it based on the addressable range for the MMU.

A quick look at the ISP and IVA IOMMUs in the OMAP3 shows that both
support the full 4GB VA space. Why do you need to restrict it ?

I should have rephrased it better when I said addressable range. While
the MMUs are capable of programming the full 4GB space, there are some
address ranges that are private from the processor view. This
window is currently used to set the range for the omap-iovmm driver
(which only OMAP3 ISP is using atm), and there is no point in allowing
the omap-iovmm driver the full range when the processor could never
reach/access those addresses.

But the IOMMU VA space is from a device point of view, not from a CPU
point of view. Could you point me to where those private ranges are
documented, in order to understand the problem correctly ?

Yes, they are indeed from the device perspective. I meant DSP and/or IPU
by processor.

For example on OMAP3, you can refer to Table 2-9 in section 2.4.5 "DSP
Subsystem Memory Space Mapping" of the OMAP36xx TRM, and the external
addressable range starts from 0x11000000.

OK, so it looks more like a property of the IOMMU master than a
property of the IOMMU itself. It would be better to express it as such,
but I wonder how that could be done, and if it would be worth it in this
case.

This property is currently solely used to configure the range for the
omap-iovmm module, which were supplied through platform data in the
non-DT case.

If I'm not mistaken omap-iovmm is something we want to get rid of. I know that 
the OMAP3 ISP driver is the only user of that module, and I've started working 
on fixing that, but I'm currently lacking time to complete the work.

Now, if we get rid of omap-iovmm, does that mean that the dma-window property 
won't need to be specified anymore ? If so, given that the only omap-iovmm 
user is the OMAP3 ISP driver, I would propose to drop the property and just 
hardcode the value.

Please let me know if there's something I've missed.

quoted

I am wondering if the way to go here is to use iommu_domain_set_attr() and
use the domain geometry values.

The other option is to supply these as driver match data, and switching
the compatible strings to identify the MMU instance precisely.

regards
Suman

quoted

quoted

As not all masters (the OMAP3 ISP doesn't for instance) have restrictions
regarding the VA range they can address, should this property be at
least made
optional ?

quoted

quoted

quoted

quoted

quoted

We are reusing the existing defined property and it allows us to get
rid of the IOVA start and end addresses defined in the pre-DT OMAP
iommu platform data.

[snip]

-- 
Regards,

Laurent Pinchart