Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Tue, 31 Mar 2020 12:53:04 -0700 (PDT)
Subject:     Re: [RFC PATCH 0/7] Introduce sv48 support 
In-Reply-To: <20200322110028.18279-1-alex@ghiti.fr>
CC:     Paul Walmsley <paul.walmsley@sifive.com>, zong.li@sifive.com,
        anup@brainfault.org, Christoph Hellwig <hch@lst.de>,
        linux-riscv@lists.infradead.org, linux-kernel@vger.kernel.org,
        alex@ghiti.fr
From:   Palmer Dabbelt <palmer@dabbelt.com>
To:     alex@ghiti.fr
Message-ID: <mhng-8dbb5aa6-857f-4321-a966-a370c0786a1c@palmerdabbelt-glaptop1>
Mime-Version: 1.0 (MHng)
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: 8bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Sun, 22 Mar 2020 04:00:21 PDT (-0700), alex@ghiti.fr wrote:
> This patchset implements sv48 support at runtime. The kernel will try to
> boot with 4-level page table and will fallback to 3-level if the HW does not
> support it.
>
> The biggest advantage is that we only have one kernel for 64bit, which
> is way easier to maintain.

Thanks, that's great!  This is something we've been missing for a long time
now.

> Folding the 4th level into a 3-level page table has almost no cost at
> runtime.
>
> At the moment, there is no way to enforce 3-level if the HW supports
> 4-level page table: early parameters are parsed after the choice must be
> made.

This is different than how I'd been considering doing it -- specifically, my
worry was that 4-level paging would have a meaningful performance hit and
therefor we'd want to allow users to select 3-level paging somehow.  I'd been
thinking of having a Kconfig option, so the options would be "3-level only" or
"3 or 4 level".  That came with a bunch of drawbacks, so I'd be much happier to
have a single kernel.

Where did you get your performance numbers from?  Appologies in advance if
there's more info in the patches, I'll look at those now...

> It is based on my relocatable patchset v3 that I have not posted yet,
> you can try the sv48 support by using the branch
> int/alex/riscv_sv48_runtime_v1 here:
>
> https://github.com/AlexGhiti/riscv-linux
>
> Any feedback appreciated,
>
> Thanks,
>
> Alexandre Ghiti (7):
>   riscv: Get rid of compile time logic with MAX_EARLY_MAPPING_SIZE
>   riscv: Allow to dynamically define VA_BITS
>   riscv: Simplify MAXPHYSMEM config
>   riscv: Implement sv48 support
>   riscv: Use pgtable_l4_enabled to output mmu type in cpuinfo
>   dt-bindings: riscv: Remove "riscv,svXX" property from device-tree
>   riscv: Explicit comment about user virtual address space size
>
>  .../devicetree/bindings/riscv/cpus.yaml       |  13 --
>  arch/riscv/Kconfig                            |  34 ++---
>  arch/riscv/boot/dts/sifive/fu540-c000.dtsi    |   4 -
>  arch/riscv/include/asm/csr.h                  |   3 +-
>  arch/riscv/include/asm/fixmap.h               |   1 +
>  arch/riscv/include/asm/page.h                 |  15 +-
>  arch/riscv/include/asm/pgalloc.h              |  36 +++++
>  arch/riscv/include/asm/pgtable-64.h           |  98 +++++++++++-
>  arch/riscv/include/asm/pgtable.h              |  24 ++-
>  arch/riscv/include/asm/sparsemem.h            |   2 +-
>  arch/riscv/kernel/cpu.c                       |  24 +--
>  arch/riscv/kernel/head.S                      |  37 ++++-
>  arch/riscv/mm/context.c                       |   4 +-
>  arch/riscv/mm/init.c                          | 142 +++++++++++++++---
>  14 files changed, 341 insertions(+), 96 deletions(-)