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

[yunhui cui <cuiyunhui@bytedance.com>]

Add mailing list:
linux-riscv@lists.infradead.org,linux-kernel@vger.kernel.org,linux-mm@kvack.org

On Wed, Jul 24, 2024 at 10:22 PM Yunhui Cui <cuiyunhui@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@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             | 44 ++++++++++++++++++++++++++++++++
>  include/linux/memblock.h         |  5 ++++
>  mm/memblock.c                    |  5 ----
>  6 files changed, 60 insertions(+), 5 deletions(-)
>
> diff --git a/arch/riscv/include/asm/page.h b/arch/riscv/include/asm/page.h
> index 235fd45d998d..60c3db47c6d8 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 089f3c9f56a3..b4ff4791e1dc 100644
> --- a/arch/riscv/include/asm/pgtable.h
> +++ b/arch/riscv/include/asm/pgtable.h
> @@ -947,6 +947,7 @@ 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 4f73c0ae44b2..24fedaa7ae93 100644
> --- a/arch/riscv/kernel/setup.c
> +++ b/arch/riscv/kernel/setup.c
> @@ -259,6 +259,7 @@ 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 bfa2dea95354..1c475c1d4c1b 100644
> --- a/arch/riscv/mm/init.c
> +++ b/arch/riscv/mm/init.c
> @@ -1385,6 +1385,50 @@ static void __init arch_reserve_crashkernel(void)
>                                     low_size, high);
>  }
>
> +static inline bool phymem_addr_overflow(phys_addr_t start, phys_addr_t end)
> +{
> +       if (pgtable_l5_enabled) {
> +               if (start > MAX_PFN_MEM_ADDR_L5 || end > MAX_PFN_MEM_ADDR_L5)
> +                       goto out;
> +       }
> +       if (pgtable_l4_enabled) {
> +               if (start > MAX_PFN_MEM_ADDR_L4 || end > MAX_PFN_MEM_ADDR_L4)
> +                       goto out;
> +       }
> +       if (start > MAX_PFN_MEM_ADDR_L3 || end > MAX_PFN_MEM_ADDR_L3)
> +               goto out;
> +
> +       return false;
> +
> +out:
> +       WARN(true, "Physical memory address overflowed!");
> +       return true;
> +}
> +
> +static void __init phymem_check(struct memblock_type *type)
> +{
> +       phys_addr_t base, end, size;
> +       int idx;
> +       struct memblock_region *rgn;
> +
> +       for_each_memblock_type(idx, type, rgn) {
> +               base = rgn->base;
> +               size = rgn->size;
> +               end = base + size - 1;
> +
> +               if (phymem_addr_overflow(base, end)) {
> +                       pr_warn("Region: [0x%llx-0x%llx] reserved.", base, end);
> +                       memblock_reserve(base, size);
> +               }
> +       }
> +}
> +
> +void __init paging_check(void)
> +{
> +       phymem_check(&memblock.reserved);
> +       phymem_check(&memblock.memory);
> +}
> +
>  void __init paging_init(void)
>  {
>         setup_bootmem();
> diff --git a/include/linux/memblock.h b/include/linux/memblock.h
> index fc4d75c6cec3..36a38c326b7a 100644
> --- a/include/linux/memblock.h
> +++ b/include/linux/memblock.h
> @@ -251,6 +251,11 @@ static inline void __next_physmem_range(u64 *idx, struct memblock_type *type,
>         __for_each_mem_range(i, &memblock.reserved, NULL, NUMA_NO_NODE, \
>                              MEMBLOCK_NONE, p_start, p_end, NULL)
>
> +#define for_each_memblock_type(i, memblock_type, rgn)                  \
> +       for (i = 0, rgn = &memblock_type->regions[0];                   \
> +            i < memblock_type->cnt;                                    \
> +            i++, rgn = &memblock_type->regions[i])
> +
>  static inline bool memblock_is_hotpluggable(struct memblock_region *m)
>  {
>         return m->flags & MEMBLOCK_HOTPLUG;
> diff --git a/mm/memblock.c b/mm/memblock.c
> index 3b9dc2d89b8a..f992050093f1 100644
> --- a/mm/memblock.c
> +++ b/mm/memblock.c
> @@ -141,11 +141,6 @@ struct memblock_type physmem = {
>   */
>  static __refdata struct memblock_type *memblock_memory = &memblock.memory;
>
> -#define for_each_memblock_type(i, memblock_type, rgn)                  \
> -       for (i = 0, rgn = &memblock_type->regions[0];                   \
> -            i < memblock_type->cnt;                                    \
> -            i++, rgn = &memblock_type->regions[i])
> -
>  #define memblock_dbg(fmt, ...)                                         \
>         do {                                                            \
>                 if (memblock_debug)                                     \
> --
> 2.39.2
>

Thanks,
Yunhui