[PATCH v2] riscv: mm: add paging_check() before paging_init()

Alexandre Ghiti alexghiti at rivosinc.com
Thu Aug 1 01:01:36 PDT 2024


Hi Yunhui,

On Thu, Jul 25, 2024 at 9:42 AM Yunhui Cui <cuiyunhui at bytedance.com> wrote:
>
> When establishing a linear mapping, the virtual address is obtained
> through __va(). If the physical address is too large, such as 1TB, then
> the virtual address will overflow in the address space of sv39.
> The log is as follows:
> [    0.000000] Unable to handle kernel paging request at virtual address 000000d97fdf7ad8
> [    0.000000] [000000d97fdf7ad8] pgd=000000407ff7e801, p4d=000000407ff7e801, pud=000000407ff7e801
> [    0.000000] Unable to handle kernel paging request at virtual address 000000d97fdfaff0
> [    0.000000] [000000d97fdfaff0] pgd=000000407ff7e801, p4d=000000407ff7e801, pud=000000407ff7e801
> ...
> [    0.000000] Insufficient stack space to handle exception!
> [    0.000000] Task stack:     [0xffffffff81400000..0xffffffff81404000]
> [    0.000000] Overflow stack: [0xffffffff80c67370..0xffffffff80c68370]
> [    0.000000] CPU: 0 PID: 0 Comm: swapper Tainted: G        W          6.6.3-00133-g60497fad461d-dirty #71
> [    0.000000] epc : die_kernel_fault+0x158/0x1c8
> [    0.000000]  ra : die_kernel_fault+0x12a/0x1c8
> [    0.000000] epc : ffffffff808cde36 ra : ffffffff808cde08 sp : ffffffff813fff80
> [    0.000000]  gp : ffffffff815a1678 tp : 0000000000000000 t0 : 0000003130386537
> [    0.000000]  t1 : 0000000000000031 t2 : 6537666637303430 s0 : ffffffff813fffc0
> [    0.000000]  s1 : ffffffff815b0b28 a0 : 0000000000000016 a1 : ffffffff81495298
> [    0.000000]  a2 : 0000000000000010 a3 : ffffffff81495298 a4 : 00000000000001fe
> [    0.000000]  a5 : 000000d97fdfa000 a6 : ffffffff814250d0 a7 : 0000000000000030
> [    0.000000]  s2 : 000000d97fdfaff0 s3 : ffffffff81400040 s4 : 000000d97fdfaff0
> [    0.000000]  s5 : ffffffff815a0ed0 s6 : 0000000000000000 s7 : 000000008f604390
> [    0.000000]  s8 : 0000000000000000 s9 : ffffffffffffffff s10: 0000000000000000
> [    0.000000]  s11: 0000000000000000 t3 : ffffffff815baa9b t4 : ffffffff815baa9b
> [    0.000000]  t5 : ffffffff815baa88 t6 : ffffffff813ffda8
> [    0.000000] status: 0000000200000100 badaddr: 000000d97fdfaff0 cause: 000000000000000d
> [    0.000000] Kernel panic - not syncing: Kernel stack overflow
> [    0.000000] CPU: 0 PID: 0 Comm: swapper Tainted: G        W          6.6.3-00133-g60497fad461d-dirty #71
> [    0.000000] Call Trace:
> [    0.000000] [<ffffffff800066bc>] dump_backtrace+0x28/0x30
> [    0.000000] [<ffffffff808cdac8>] show_stack+0x38/0x44
> [    0.000000] [<ffffffff808d9d40>] dump_stack_lvl+0x44/0x5c
> [    0.000000] [<ffffffff808d9d70>] dump_stack+0x18/0x20
> [    0.000000] [<ffffffff808cdfb6>] panic+0x110/0x2f2
> [    0.000000] [<ffffffff80006532>] walk_stackframe+0x0/0x120
> [    0.000000] [<ffffffff808cde08>] die_kernel_fault+0x12a/0x1c8
> [    0.000000] ---[ end Kernel panic - not syncing: Kernel stack overflow ]---
>
> In other words, the maximum value of the physical address needs to meet
> Documentation/riscv/vm-layout.rst to ensure that there is no overflow.
> For sv48/57, the actual virtual address space is huge, so this problem
> is generally not triggered, but it is also checked in the code.
>
> We give a warning for the overflowed physical address region and reverve it
> so that the kernel can bringup successfully.
>
> Signed-off-by: Yunhui Cui <cuiyunhui at bytedance.com>
> ---
>  arch/riscv/include/asm/page.h    |  9 +++++++++
>  arch/riscv/include/asm/pgtable.h |  1 +
>  arch/riscv/kernel/setup.c        |  1 +
>  arch/riscv/mm/init.c             | 30 ++++++++++++++++++++++++++++++
>  4 files changed, 41 insertions(+)
>
> diff --git a/arch/riscv/include/asm/page.h b/arch/riscv/include/asm/page.h
> index 57e887bfa34c..f6c0f6e14ecb 100644
> --- a/arch/riscv/include/asm/page.h
> +++ b/arch/riscv/include/asm/page.h
> @@ -38,6 +38,15 @@
>   */
>  #define PAGE_OFFSET_L4         _AC(0xffffaf8000000000, UL)
>  #define PAGE_OFFSET_L3         _AC(0xffffffd800000000, UL)
> +
> +/*
> + * See vm-layout.rst, the size of L3 direct mapping of all physical
> + * memory 124GB, L4 is 64TB, L5 is 32PB.
> + */
> +#define MAX_PFN_MEM_ADDR_L5    (0x80000000000000ULL)
> +#define MAX_PFN_MEM_ADDR_L4    (0x400000000000ULL)
> +#define MAX_PFN_MEM_ADDR_L3    (0x1F00000000ULL)
> +
>  #else
>  #define PAGE_OFFSET            _AC(CONFIG_PAGE_OFFSET, UL)
>  #endif /* CONFIG_64BIT */
> diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h
> index c8e8867c42f6..e4835de5a743 100644
> --- a/arch/riscv/include/asm/pgtable.h
> +++ b/arch/riscv/include/asm/pgtable.h
> @@ -915,6 +915,7 @@ extern uintptr_t _dtb_early_pa;
>  #endif /* CONFIG_XIP_KERNEL */
>  extern u64 satp_mode;
>
> +void paging_check(void);
>  void paging_init(void);
>  void misc_mem_init(void);
>
> diff --git a/arch/riscv/kernel/setup.c b/arch/riscv/kernel/setup.c
> index 2a79d4ed2660..366918578544 100644
> --- a/arch/riscv/kernel/setup.c
> +++ b/arch/riscv/kernel/setup.c
> @@ -273,6 +273,7 @@ void __init setup_arch(char **cmdline_p)
>         parse_early_param();
>
>         efi_init();
> +       paging_check();
>         paging_init();
>
>         /* Parse the ACPI tables for possible boot-time configuration */
> diff --git a/arch/riscv/mm/init.c b/arch/riscv/mm/init.c
> index b0cc28f7595f..aa25dcf8a0ff 100644
> --- a/arch/riscv/mm/init.c
> +++ b/arch/riscv/mm/init.c
> @@ -1482,6 +1482,36 @@ static void __init reserve_crashkernel(void)
>         crashk_res.end = crash_base + crash_size - 1;
>  }
>
> +static void __init phymem_addr_overflow(void)
> +{
> +       phys_addr_t end = memblock_end_of_DRAM();
> +
> +       if (pgtable_l5_enabled) {
> +               if (end > MAX_PFN_MEM_ADDR_L5) {
> +                       memblock_reserve(MAX_PFN_MEM_ADDR_L5, end - MAX_PFN_MEM_ADDR_L5);
> +                       WARN(true, "Phymem addr 0x%llx overflowed, reserve [0x%llx-0x%llx] directly.",
> +                            end, MAX_PFN_MEM_ADDR_L5, end);
> +               }
> +       }
> +       if (pgtable_l4_enabled) {
> +               if (end > MAX_PFN_MEM_ADDR_L4) {
> +                       memblock_reserve(MAX_PFN_MEM_ADDR_L4, end - MAX_PFN_MEM_ADDR_L4);
> +                       WARN(true, "Phymem addr 0x%llx overflowed, reserve [0x%llx-0x%llx] directly.",
> +                            end, MAX_PFN_MEM_ADDR_L4, end);
> +               }
> +       }
> +       if (end > MAX_PFN_MEM_ADDR_L3) {
> +               memblock_reserve(MAX_PFN_MEM_ADDR_L3, end - MAX_PFN_MEM_ADDR_L3);
> +               WARN(true, "Phymem addr 0x%llx overflowed, reserve [0x%llx-0x%llx] directly.",
> +                    end, MAX_PFN_MEM_ADDR_L3, end);
> +       }
> +}
> +
> +void __init paging_check(void)
> +{
> +       phymem_addr_overflow();
> +}
> +
>  void __init paging_init(void)
>  {
>         setup_bootmem();
> --
> 2.39.2
>

So the following patch should fix your issue and was posted some time
ago https://lore.kernel.org/linux-riscv/bdfbed9b-ea04-4415-8416-d6e9d0a643a3@ghiti.fr/T/#me26a82cf32f46cae12e2ea8892a3bc16d4fc37e3.
I prefer this solution as it does not introduce a bunch of hardcoded
defines and relies on the already existing memblock API.

Let me know if that's not the case!

Thanks,

Alex



More information about the linux-riscv mailing list