Re: [PATCH v11 0/2] arm/virt: CXL support via pxb_cxl

[Jonathan Cameron via <qemu-devel@nongnu.org>]

On Thu, 16 Jun 2022 15:19:48 +0100
Jonathan Cameron via <qemu-devel@nongnu.org> wrote:

> Previously patches 40 and 41 of
> [PATCH v10 00/45] CXl 2.0 emulation Support
> https://lore.kernel.org/qemu-devel/20220429144110.25167-45-Jonathan.Cameron@huawei.com/#r
> 
> Now the base CXL support including for x86/pc is upstream (patches 1-39)
> there are no dependencies between the next few CXL elements in my queue
> so they can be reviewed / merged in parallel.  Hence I'll be sending switch
> support (43-45) separately and hopefully DOE / CDAT support in a few days.
> I'm assuming this particular series should go through the arm tree if
> the maintainers are happy?

Hi All,

If Peter or anyone else has time to look at this with a view to getting
ARM support on par with x86 that would be great.  I 'think' this should
be uncontroversial but I'm far from an expert!  I'm particularly keen on
getting this upstream as most of my testing is on ARM/virt.

Thanks,

Jonathan

> 
> Changes since v10:
>  - CXL machine setup is now entirely from the supporting machines rather
>    than via code in machine.c and vl.c.  Change made for x86 in:
>    https://lore.kernel.org/qemu-devel/20220608145440.26106-1-Jonathan.Cameron@huawei.com/
>  - Dropped Ben's sign off from patch 1 which resulted from him carrying
>    these patches of mine for a while. It isn't a useful bit of history
>    to carry now they are back to me.
> 
> This short series adds support for CXL host bridges and CXL fixed memory
> windows on arm/virt.  Two types of memory region are needed:
> 1. Register space for CXL host bridges (static allowance for 16)
> 2. CXL fixed memory windows: Ranges of host PA space which
>    are statically mapped to an interleave across 1 or more CXL host
>    bridges.
> 
> Both of these types of region are described via appropriate ACPI tables.
> As the CEDT table is created with the same code as for x86 I don't think
> there is much value in duplicating the existing CXL bios-tables test.
> 
> The second patch adds a single complex test. We test a lot more configurations
> on x86 but it does not seem useful to duplicate them all on ARM and this single
> test should act as a smoke test for any problems that occur.
> 
> Run through CI at:
> https://gitlab.com/jic23/qemu/-/pipelines/564934276
> Intermittent (unrelated I assume) failure in msys64 aio-test resolved
> with a retry.
> 
> Thanks,
> 
> Jonathan
> 
> Jonathan Cameron (2):
>   hw/arm/virt: Basic CXL enablement on pci_expander_bridge instances
>     pxb-cxl
>   qtest/cxl: Add aarch64 virt test for CXL
> 
>  hw/arm/virt-acpi-build.c | 34 ++++++++++++++++++++++++++++
>  hw/arm/virt.c            | 44 ++++++++++++++++++++++++++++++++++++
>  include/hw/arm/virt.h    |  3 +++
>  tests/qtest/cxl-test.c   | 48 ++++++++++++++++++++++++++++++++--------
>  tests/qtest/meson.build  |  1 +
>  5 files changed, 121 insertions(+), 9 deletions(-)
>