Re: [PATCH v3 00/14] arm: rework id register storage

[Sebastian Ott <sebott@redhat.com>]

On Tue, 11 Mar 2025, Cornelia Huck wrote:
> Yet another update of the id register series, less changes this time
> around.
>
> Changed from v2:
> - changed generation of the various register defines via the "DEF"
>  magic suggested by Richard
> - some kvm-only code moved to kvm.c; some code potentially useful to
>  non-kvm code stayed out of there (the cpu model code will make use
>  of it, and that one should be extendable outside of kvm -- a
>  revised version of those patches is still in the works, but I'll be
>  off for a few days and rather wanted to get this one out first)
>
> Also available at
> https://gitlab.com/cohuck/qemu/-/commits/arm-rework-idreg-storage-v3
>
> <v2 cover letter>
>
> Changed from v1:
> - Noticed that we missed the hvf code. Converted, compiled, but not tested
>  as I'm lacking an environment for testing.
> - Hopefully incorporated most of the suggested changes -- if I missed
>  something, it was unintentional unless mentioned below.
>  - fixed repeated inclusion of definitions
>  - hopefully made macros more robust
>  - removed distinction between reading 32/64 values, which was mostly
>    adding churn for little value
>  - postponed generating property definitions to the cpu model patches,
>    where they are actually used
>  - juggled hunks and moved them to the right patches
>  - fixed some typos
> - rebased to a more recent code base
>
> NOT changed from v1:
> - definitions are still generated from the Linux sysregs file
>  - I still think updating the generated files on demand (so that we can
>    double check the result) is the right thing to do
>  - I'm open to changing the source of the definitions from the sysregs
>    file to the JSON definitions published by Arm; however, I first wanted
>    to get the code using it right -- we can switch out the code generating
>    the file to use a different source easily later on, and I'd also like
>    to steal parts of the script from Linux once integrated (which I think
>    hasn't happened yet?)
>
> <v1 cover letter>
>
> [Note: I've kept the cc list from the last round of cpu model patches;
> so if you're confused as to why you're cc:ed here, take it as a
> heads-up that a new cpu model series will come along soon]
>
> This patch series contains patches extracted from the larger cpu model
> series (RFC v2 last posted at
> https://lore.kernel.org/qemu-devel/20241206112213.88394-1-cohuck@redhat.com/)
> and aims at providing a base upon which we can continue with building
> support for cpu models, but which is hopefully already an improvement
> on its own.
>
> Main changes from the patches in that series include:
> - post-pone the changes to handle KVM writable ID registers for cpu models
>  (I have a series including that on top of this one)
> - change how we store the list of ID registers, and access them
>  basically, use an enum for indexing, and an enum doing encodings in a
>  pattern similar to cpregs
> - move some hunks to different patches
> - update the scripts to generate the register descriptions, and run
>  them against a recent Linux sysregs file
>
> What I've kept:
> - generating the register descriptions from the Linux sysregs file
>  I think that file is still our best bet to generate the descriptions
>  easily, and updating the definitions is a manual step that can be checked
>  for unintended changes
> - most of the hard work that Eric had been doing; all new bugs in there
>  are my own :)
>
> </v1 cover letter>
> </v2 cover letter>
>
> Cornelia Huck (2):
>  arm/kvm: add accessors for storing host features into idregs
>  arm/cpu: switch to a generated cpu-sysregs.h.inc
>
> Eric Auger (12):
>  arm/cpu: Add sysreg definitions in cpu-sysregs.h
>  arm/cpu: Store aa64isar0/aa64zfr0 into the idregs arrays
>  arm/cpu: Store aa64isar1/2 into the idregs array
>  arm/cpu: Store aa64pfr0/1 into the idregs array
>  arm/cpu: Store aa64mmfr0-3 into the idregs array
>  arm/cpu: Store aa64dfr0/1 into the idregs array
>  arm/cpu: Store aa64smfr0 into the idregs array
>  arm/cpu: Store id_isar0-7 into the idregs array
>  arm/cpu: Store id_pfr0/1/2 into the idregs array
>  arm/cpu: Store id_dfr0/1 into the idregs array
>  arm/cpu: Store id_mmfr0-5 into the idregs array
>  arm/cpu: Add sysreg generation scripts
>
> hw/intc/armv7m_nvic.c                 |  27 +-
> scripts/gen-cpu-sysregs-header.awk    |  39 +++
> scripts/update-aarch64-sysreg-code.sh |  25 ++
> target/arm/cpu-features.h             | 317 +++++++++---------
> target/arm/cpu-sysregs.h              |  41 +++
> target/arm/cpu-sysregs.h.inc          | 170 ++++++++++
> target/arm/cpu.c                      | 111 +++----
> target/arm/cpu.h                      |  80 +++--
> target/arm/cpu64.c                    | 128 +++----
> target/arm/helper.c                   |  68 ++--
> target/arm/hvf/hvf.c                  |  36 +-
> target/arm/internals.h                |   6 +-
> target/arm/kvm.c                      | 129 ++++----
> target/arm/ptw.c                      |   6 +-
> target/arm/tcg/cpu-v7m.c              | 174 +++++-----
> target/arm/tcg/cpu32.c                | 320 +++++++++---------
> target/arm/tcg/cpu64.c                | 460 +++++++++++++-------------
> 17 files changed, 1218 insertions(+), 919 deletions(-)
> create mode 100755 scripts/gen-cpu-sysregs-header.awk
> create mode 100755 scripts/update-aarch64-sysreg-code.sh
> create mode 100644 target/arm/cpu-sysregs.h
> create mode 100644 target/arm/cpu-sysregs.h.inc
>

Reviewed-by: Sebastian Ott <sebott@redhat.com>