[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAHVXubg+aXS7-smmWS9LwUbAkpfGvet9Zk0UJs_AZhxV2vx6UQ@mail.gmail.com>
Date: Mon, 1 Jul 2024 13:58:27 +0200
From: Alexandre Ghiti <alexghiti@...osinc.com>
To: Stuart Menefy <stuart.menefy@...asip.com>
Cc: Jonathan Corbet <corbet@....net>, Paul Walmsley <paul.walmsley@...ive.com>,
Palmer Dabbelt <palmer@...belt.com>, Albert Ou <aou@...s.berkeley.edu>, linux-doc@...r.kernel.org,
linux-riscv@...ts.infradead.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2] riscv: Extend sv39 linear mapping max size to 128G
Hi Stuart,
On Sun, Jun 30, 2024 at 1:06 PM Stuart Menefy <stuart.menefy@...asip.com> wrote:
>
> This harmonizes all virtual addressing modes which can now all map
> (PGDIR_SIZE * PTRS_PER_PGD) / 4 of physical memory.
>
> The RISCV implementation of KASAN requires that the boundary between
> shallow mappings are aligned on an 8G boundary. In this case we need
> VMALLOC_START to be 8G aligned. So although we only need to move the
> start of the linear mapping down by 4GiB to allow 128GiB to be mapped,
> we actually move it down by 8GiB (creating a 4GiB hole between the
> linear mapping and KASAN shadow space) to maintain the alignment
> requirement.
>
> Signed-off-by: Stuart Menefy <stuart.menefy@...asip.com>
> ---
>
> Changes since V1:
> This is a modified version of the patch posted by Alexandre Ghiti
> <alexghiti@...osinc.com> [1]. That version moved the start of the
> liner mapping by 4GiB, but was subsequently found to trigger a problem
> in the KASAN code [2]. By moving the start address by 8GiB alignment
> requiremenents are maintained.
>
> [1] https://lore.kernel.org/linux-riscv/20240514133614.87813-1-alexghiti@rivosinc.com/
> [2] https://lore.kernel.org/linux-riscv/4011b34a-1b86-42c7-aaf6-3bd297149f06@ghiti.fr/
>
> ---
> Documentation/arch/riscv/vm-layout.rst | 11 ++++++-----
> arch/riscv/include/asm/page.h | 2 +-
> 2 files changed, 7 insertions(+), 6 deletions(-)
>
> diff --git a/Documentation/arch/riscv/vm-layout.rst b/Documentation/arch/riscv/vm-layout.rst
> index e476b4386bd9..077b968dcc81 100644
> --- a/Documentation/arch/riscv/vm-layout.rst
> +++ b/Documentation/arch/riscv/vm-layout.rst
> @@ -47,11 +47,12 @@ RISC-V Linux Kernel SV39
> | Kernel-space virtual memory, shared between all processes:
> ____________________________________________________________|___________________________________________________________
> | | | |
> - ffffffc6fea00000 | -228 GB | ffffffc6feffffff | 6 MB | fixmap
> - ffffffc6ff000000 | -228 GB | ffffffc6ffffffff | 16 MB | PCI io
> - ffffffc700000000 | -228 GB | ffffffc7ffffffff | 4 GB | vmemmap
> - ffffffc800000000 | -224 GB | ffffffd7ffffffff | 64 GB | vmalloc/ioremap space
> - ffffffd800000000 | -160 GB | fffffff6ffffffff | 124 GB | direct mapping of all physical memory
> + ffffffc4fea00000 | -236 GB | ffffffc4feffffff | 6 MB | fixmap
> + ffffffc4ff000000 | -236 GB | ffffffc4ffffffff | 16 MB | PCI io
> + ffffffc500000000 | -236 GB | ffffffc5ffffffff | 4 GB | vmemmap
> + ffffffc600000000 | -232 GB | ffffffd5ffffffff | 64 GB | vmalloc/ioremap space
> + ffffffd600000000 | -168 GB | fffffff5ffffffff | 128 GB | direct mapping of all physical memory
> + | | | |
> fffffff700000000 | -36 GB | fffffffeffffffff | 32 GB | kasan
> __________________|____________|__________________|_________|____________________________________________________________
> |
> diff --git a/arch/riscv/include/asm/page.h b/arch/riscv/include/asm/page.h
> index 115ac98b8d72..81d47fbbd927 100644
> --- a/arch/riscv/include/asm/page.h
> +++ b/arch/riscv/include/asm/page.h
> @@ -37,7 +37,7 @@
> * define the PAGE_OFFSET value for SV48 and SV39.
> */
> #define PAGE_OFFSET_L4 _AC(0xffffaf8000000000, UL)
> -#define PAGE_OFFSET_L3 _AC(0xffffffd800000000, UL)
> +#define PAGE_OFFSET_L3 _AC(0xffffffd600000000, UL)
> #else
> #define PAGE_OFFSET _AC(CONFIG_PAGE_OFFSET, UL)
> #endif /* CONFIG_64BIT */
> --
> 2.43.0
>
So the issue was with my patch, sorry! Your fix makes sense so you can add:
Reviewed-by: Alexandre Ghiti <alexghiti@...osinc.com>
Thanks,
Alex
Powered by blists - more mailing lists