Re: [PATCH 0/3] hw/misc: QUP UART device

[Brian Cain <brian.cain@oss.qualcomm.com>]

On 10/27/2026 3:12 PM, Pierrick Bouvier wrote:
> On 10/27/2026 12:02 PM, Brian Cain wrote:
>> On 10/27/2026 1:43 PM, Pierrick Bouvier wrote:
>>> On 10/27/2026 10:38 AM, Brian Cain wrote:
>>>> Add the QUP + GENI uart device models.  The QUP is a "Universal
>>>> Peripheral"
>>>> present on Qualcomm SoCs, with connections for SPI, UART, I2C, I3C.  For
>>>> now we're only interested in UART.
>>>>
>>>> Will follow up with a v2 including wiring into machines + qtest after
>>>> https://lore.kernel.org/qemu-devel/20261027143721.4125584-1-
>>>> brian.cain@oss.qualcomm.com/
>>>> lands.
>>>>
>>>> Brian Cain (3):
>>>>     hw/misc: add the QUP wrapper device
>>>>     rust/chardev: fix write errors and add write watches
>>>>     rust/hw/char: add QUP GENI UART device model
>>>>
>>>>    MAINTAINERS                                 |    3 +
>>>>    docs/system/device-emulation.rst            |    1 +
>>>>    docs/system/devices/qup.rst                 |   10 +
>>>>    include/hw/misc/qup-wrapper.h               |   27 +
>>>>    hw/misc/qup-wrapper.c                       |   90 ++
>>>>    Cargo.lock                                  |   15 +
>>>>    Cargo.toml                                  |    1 +
>>>>    hw/misc/Kconfig                             |    3 +
>>>>    hw/misc/meson.build                         |    3 +
>>>>    rust/chardev/src/chardev.rs                 |   82 +-
>>>>    rust/hw/char/Kconfig                        |    4 +
>>>>    rust/hw/char/meson.build                    |    1 +
>>>>    rust/hw/char/qup_geni_uart/Cargo.toml       |   27 +
>>>>    rust/hw/char/qup_geni_uart/meson.build      |   28 +
>>>>    rust/hw/char/qup_geni_uart/src/device.rs    | 1293 +++++++++++++++++++
>>>>    rust/hw/char/qup_geni_uart/src/lib.rs       |    9 +
>>>>    rust/hw/char/qup_geni_uart/src/registers.rs |  160 +++
>>>>    17 files changed, 1741 insertions(+), 16 deletions(-)
>>>>    create mode 100644 docs/system/devices/qup.rst
>>>>    create mode 100644 include/hw/misc/qup-wrapper.h
>>>>    create mode 100644 hw/misc/qup-wrapper.c
>>>>    create mode 100644 rust/hw/char/qup_geni_uart/Cargo.toml
>>>>    create mode 100644 rust/hw/char/qup_geni_uart/meson.build
>>>>    create mode 100644 rust/hw/char/qup_geni_uart/src/device.rs
>>>>    create mode 100644 rust/hw/char/qup_geni_uart/src/lib.rs
>>>>    create mode 100644 rust/hw/char/qup_geni_uart/src/registers.rs
>>>>
>>> Any specific reason to have this implemented in Rust?
>>>
>>> Not a bad thing in itself, but it means it's only available with rust
>>> builds, which can be a problem if a machine defined in C wants to use it
>>> unconditionally. Rust is not (yet) a mandatory dep for QEMU.
>> A new device model -> a brand new day -> a new frontier?
>>
>> It was an interesting challenge and seemed appropriate given the nice
>> reference pl011 device.  But I will concede it can be impractical.
>> Making this device conditionally enabled in the machine that uses it
>> felt kinda "lame" and made me long for a day when it *was* a mandatory
>> dependency.
>>
>> Seeing as how we have a pl011 device in C, maybe it's better to start
>> with a C version of the QUP?  If so, that's okay with me.
>>
> If it's a mandatory device as part of a machine definition, it would
> probably be better to stick to C for now.
> I hope we can get Rust as a main dep at some point. I'm not sure what is
> the criteria to do this change, out of deciding "let's do this".

Hmm - so ... can I put in a vote for this, then?  Can "at some point" be 
now?  What are the criteria?


> Regards,
> Pierrick