Re: [RESEND PATCH v13 8/8] Documentation/cxl: Document DPA partition layout and ordering rules

[Anisa Su <anisa.su887@gmail.com>]

On Tue, Sep 08, 2026 at 11:17:00PM +0100, Jonathan Cameron wrote:
> On Tue,  8 Sep 2026 03:15:12 -0700
> Anisa Su <anisa.su887@gmail.com> wrote:
> 
> > DC Partitions complicate DPA ordering. Add a DPA Partitions section to
> > cxl-driver.rst describing spec-mandated and Linux requirements for the
> > layout.
> > 
> > Suggested-by: Gregory Price <gourry@gourry.net>
> > Signed-off-by: Anisa Su <anisa.su@samsung.com>
> A couple of minor potential tweaks.  Otherwise LGTM
> Reviewed-by: Jonathan Cameron <jonathan.cameron@oss.qualcomm.com>
> 
> > 
> > ---
> > New patch in v13.
> > ---
> >  .../driver-api/cxl/linux/cxl-driver.rst       | 38 +++++++++++++++++++
> >  1 file changed, 38 insertions(+)
> > 
> > diff --git a/Documentation/driver-api/cxl/linux/cxl-driver.rst b/Documentation/driver-api/cxl/linux/cxl-driver.rst
> > index dd6dd17dc536..f0742d6c86c3 100644
> > --- a/Documentation/driver-api/cxl/linux/cxl-driver.rst
> > +++ b/Documentation/driver-api/cxl/linux/cxl-driver.rst
> > @@ -181,6 +181,44 @@ A Memory Device is a discrete base object that is not a port.  While the
> >  physical device it belongs to may also host an `endpoint`, the relationship
> >  between an `endpoint` and a `memdev` is not captured in sysfs.
> >  
> > +DPA Partitions
> > +~~~~~~~~~~~~~~
> > +A memory device presents its capacity as one flat `Device Physical Address`
> > +(DPA) space divided into `partitions`, which Linux lays out in a fixed
> > +order::
> > +
> > +  DPA 0                                                     end
> > +  +---------------+---------------+---------------------------+
> > +  |      ram      |     pmem      |       dynamic_ram_1       |
> > +  +---------------+---------------+---------------------------+
> > +  part[0]         part[1]         part[2]
> > +
> > +Part of that order is required by the CXL specification and part of it is a
> > +Linux choice.
> > +
> > +The `ram` and `pmem` order is mandated.  CXL r4.0 section 8.2.10.9.2.1 "Get
> > +Partition Info" (4100h), Table 8-310, mandates that volatile capacity starts
> > +at DPA 0 and pmem starts at the DPA immediately following it.
> > +
> > +Dynamic Capacity partitions only need to be 256MB aligned according to
> > +CXL r4.0 section 8.2.10.9.9.1 "Get Dynamic Capacity Configuration"
> > +(opcode 4800h), Table 8-347. So a device could leave a gap between ram/pmem
> > +(static) capacity and its first DC partition, or between one DC partition
> > +and the next.
> > +
> > +Linux follows the static precedent anyway for the partition it maps: the
> 
> Maybe "Linux chooses to only support...
> 
I went with "Linux chooses not to support gaps between static and
dynamic capacity: the first DC partition must begin..."

> > +first DC partition must begin at the DPA immediately following static
> > +capacity -- after pmem, after ram on a device with no pmem, or at DPA 0 on
> > +a device with no static capacity at all.
> > +
> > +Currently, only one dynamic partition is supported. A device may report up
> > +to eight (CXL r4.0 Table 8-346); Linux configures the first and exposes it as
> > +`dynamic_ram_1`.
> > +
> > +Support for additional dynamic partitions may be added if devices appear
> > +that need it, which is what the `dynamic_ram_1` name leaves room for.  Until
> > +then a device offering more than one is still usable, just not in full.
> I'd drop this last paragraph. Predicting the future is tricky, even with a may.
> Hopefully anyone realises that if something is needed Linux doesn't support
> they should propose patches to add it!
> 
Agreed, dropped this paragraph.

> Jonathan

Thanks,
Anisa
> 
> > +
> >  Port Relationships
> >  ~~~~~~~~~~~~~~~~~~
> >  In our example described above, there are four host bridges attached to the
>