devicetree.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* Using iommu-addresses beyond reserved-memory
@ 2025-04-30 20:35 Konrad Dybcio
  2025-05-10  8:57 ` Vikash Garodia
  0 siblings, 1 reply; 2+ messages in thread
From: Konrad Dybcio @ 2025-04-30 20:35 UTC (permalink / raw)
  To: dt, Rob Herring; +Cc: Vikash Garodia, Bjorn Andersson

Hi,

We're exploring the use of iommu-addresses due to hw specifics where memory
regions for some devices must be mapped in a given range.

iommu-addresses currently allows us to achieve that, but it's incredibly janky..
To achieve what we need, one has to:

1. define a negative of the desired region (i.e. reserve all memory except where
we want the allocations to happen)
2. pass that to a dt (sub)node as a memory region
(3. pray that you reserved enough memory after the region and there's no board
    with a larger amount of RAM down the line)

Now, the obvious idea is to redefine this property to consume the **allowed**
ranges on ""normal"" devices' nodes (i.e. not under /reserved-memory) - would
that be an acceptable approach? We're totally open to suggestions

Konrad

^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: Using iommu-addresses beyond reserved-memory
  2025-04-30 20:35 Using iommu-addresses beyond reserved-memory Konrad Dybcio
@ 2025-05-10  8:57 ` Vikash Garodia
  0 siblings, 0 replies; 2+ messages in thread
From: Vikash Garodia @ 2025-05-10  8:57 UTC (permalink / raw)
  To: Konrad Dybcio, dt, Rob Herring; +Cc: Bjorn Andersson


On 5/1/2025 2:05 AM, Konrad Dybcio wrote:
> Hi,
> 
> We're exploring the use of iommu-addresses due to hw specifics where memory
> regions for some devices must be mapped in a given range.
> 
> iommu-addresses currently allows us to achieve that, but it's incredibly janky..
> To achieve what we need, one has to:
> 
> 1. define a negative of the desired region (i.e. reserve all memory except where
> we want the allocations to happen)
> 2. pass that to a dt (sub)node as a memory region
> (3. pray that you reserved enough memory after the region and there's no board
>     with a larger amount of RAM down the line)
> 
> Now, the obvious idea is to redefine this property to consume the **allowed**
> ranges on ""normal"" devices' nodes (i.e. not under /reserved-memory) - would
> that be an acceptable approach? We're totally open to suggestions

Any suggestions ?

Regards,
Vikash

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2025-05-10  8:57 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2025-04-30 20:35 Using iommu-addresses beyond reserved-memory Konrad Dybcio
2025-05-10  8:57 ` Vikash Garodia

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).