mbox series

[v2,0/4] riscv: Separate vendor extensions from standard extensions

Message ID 20240609-support_vendor_extensions-v2-0-9a43f1fdcbb9@rivosinc.com (mailing list archive)
Headers show
Series riscv: Separate vendor extensions from standard extensions | expand

Message

Charlie Jenkins June 10, 2024, 4:34 a.m. UTC
All extensions, both standard and vendor, live in one struct
"riscv_isa_ext". There is currently one vendor extension, xandespmu, but
it is likely that more vendor extensions will be added to the kernel in
the future. As more vendor extensions (and standard extensions) are
added, riscv_isa_ext will become more bloated with a mix of vendor and
standard extensions.

This also allows each vendor to be conditionally enabled through
Kconfig.

---
This has been split out from the previous series that contained the
addition of xtheadvector due to lack of reviews. The xtheadvector
support will be posted again separately from this.

The reviewed-bys on "riscv: Extend cpufeature.c to detect vendor extensions"
and "riscv: Introduce vendor variants of extension helpers" have been
dropped in this series. The majority of the code is the same in these
patches, but thead-specific code is swapped out with andes-specific
code. The changes are minimal, but I decided to drop the reviews in case
I inadvertently introduced issues.

The alternative patching code from "riscv: Introduce vendor variants of
extension helpers" has been migrated to "riscv: Extend cpufeature.c to
detect vendor extensions" such that the andes patching still works in
that patch.

I also fix a bug in this patch from the previous series that the Andes
extensions were not being properly enabled due to the manual alternative
patching not incrementing the id to be greater than
RISCV_VENDOR_EXT_ALTERNATIVES_BASE. 

To: Paul Walmsley <paul.walmsley@sifive.com>
To: Palmer Dabbelt <palmer@dabbelt.com>
To: Albert Ou <aou@eecs.berkeley.edu>
To: Conor Dooley <conor.dooley@microchip.com>
To: Evan Green <evan@rivosinc.com>
To: Andy Chiu <andy.chiu@sifive.com>
Cc: linux-riscv@lists.infradead.org
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Charlie Jenkins <charlie@rivosinc.com>

---
Changes in v2:
- Fixed issue in riscv_fill_vendor_ext_list() that initalizion was only
  happening properly for the first vendor. Add is_initialized field to
  riscv_isa_vendor_ext_data_list to allow intialization to be tracked on
  a per-vendor basis.
- Link to v1: https://lore.kernel.org/r/20240515-support_vendor_extensions-v1-0-b05dd5ea7d8d@rivosinc.com

---
Charlie Jenkins (4):
      riscv: Extend cpufeature.c to detect vendor extensions
      riscv: Add vendor extensions to /proc/cpuinfo
      riscv: Introduce vendor variants of extension helpers
      riscv: cpufeature: Extract common elements from extension checking

 arch/riscv/Kconfig                               |   2 +
 arch/riscv/Kconfig.vendor                        |  19 ++++
 arch/riscv/errata/andes/errata.c                 |   3 +
 arch/riscv/errata/sifive/errata.c                |   3 +
 arch/riscv/errata/thead/errata.c                 |   3 +
 arch/riscv/include/asm/cpufeature.h              |  96 ++++++++++------
 arch/riscv/include/asm/hwcap.h                   |   1 -
 arch/riscv/include/asm/vendor_extensions.h       | 104 +++++++++++++++++
 arch/riscv/include/asm/vendor_extensions/andes.h |  19 ++++
 arch/riscv/kernel/Makefile                       |   2 +
 arch/riscv/kernel/cpu.c                          |  35 +++++-
 arch/riscv/kernel/cpufeature.c                   | 137 +++++++++++++++++------
 arch/riscv/kernel/vendor_extensions.c            |  56 +++++++++
 arch/riscv/kernel/vendor_extensions/Makefile     |   3 +
 arch/riscv/kernel/vendor_extensions/andes.c      |  18 +++
 drivers/perf/riscv_pmu_sbi.c                     |  11 +-
 16 files changed, 438 insertions(+), 74 deletions(-)
---
base-commit: c3f38fa61af77b49866b006939479069cd451173
change-id: 20240515-support_vendor_extensions-aa80120e4230

Comments

Conor Dooley June 13, 2024, 2:45 p.m. UTC | #1
On Sun, Jun 09, 2024 at 09:34:13PM -0700, Charlie Jenkins wrote:
> All extensions, both standard and vendor, live in one struct
> "riscv_isa_ext". There is currently one vendor extension, xandespmu, but
> it is likely that more vendor extensions will be added to the kernel in
> the future. As more vendor extensions (and standard extensions) are
> added, riscv_isa_ext will become more bloated with a mix of vendor and
> standard extensions.
> 
> This also allows each vendor to be conditionally enabled through
> Kconfig.
> 
> ---
> This has been split out from the previous series that contained the
> addition of xtheadvector due to lack of reviews. The xtheadvector
> support will be posted again separately from this.

I think that's a good call.

> The reviewed-bys on "riscv: Extend cpufeature.c to detect vendor extensions"
> and "riscv: Introduce vendor variants of extension helpers" have been
> dropped in this series. The majority of the code is the same in these
> patches, but thead-specific code is swapped out with andes-specific
> code. The changes are minimal, but I decided to drop the reviews in case
> I inadvertently introduced issues.

Actually, you only completely did that on the first of the two patches
you mention, but I don't mind.
Charlie Jenkins June 13, 2024, 5:37 p.m. UTC | #2
On Thu, Jun 13, 2024 at 03:45:33PM +0100, Conor Dooley wrote:
> On Sun, Jun 09, 2024 at 09:34:13PM -0700, Charlie Jenkins wrote:
> > All extensions, both standard and vendor, live in one struct
> > "riscv_isa_ext". There is currently one vendor extension, xandespmu, but
> > it is likely that more vendor extensions will be added to the kernel in
> > the future. As more vendor extensions (and standard extensions) are
> > added, riscv_isa_ext will become more bloated with a mix of vendor and
> > standard extensions.
> > 
> > This also allows each vendor to be conditionally enabled through
> > Kconfig.
> > 
> > ---
> > This has been split out from the previous series that contained the
> > addition of xtheadvector due to lack of reviews. The xtheadvector
> > support will be posted again separately from this.
> 
> I think that's a good call.
> 
> > The reviewed-bys on "riscv: Extend cpufeature.c to detect vendor extensions"
> > and "riscv: Introduce vendor variants of extension helpers" have been
> > dropped in this series. The majority of the code is the same in these
> > patches, but thead-specific code is swapped out with andes-specific
> > code. The changes are minimal, but I decided to drop the reviews in case
> > I inadvertently introduced issues.
> 
> Actually, you only completely did that on the first of the two patches
> you mention, but I don't mind.

You reviewed the third patch of this series in the first revision of
this series and I should have updated this comment.

- Charlie