Messages in this thread | | | Subject | Re: [PATCH] arm64: remove page granularity limitation from KFENCE | From | Liu Shixin <> | Date | Tue, 28 Sep 2021 15:03:21 +0800 |
| |
On 2021/9/18 19:50, Alexander Potapenko wrote: > On Sat, Sep 18, 2021 at 10:10 AM Liu Shixin <liushixin2@huawei.com> wrote: >> Currently if KFENCE is enabled in arm64, the entire linear map will be >> mapped at page granularity which seems overkilled. Actually only the >> kfence pool requires to be mapped at page granularity. We can remove the >> restriction from KFENCE and force the linear mapping of the kfence pool >> at page granularity later in arch_kfence_init_pool(). > There was a previous patch by Jisheng Zhang intended to remove this > requirement: https://lore.kernel.org/linux-arm-kernel/20210524180656.395e45f6@xhacker.debian/ > Which of the two is more preferable? The previous patch by Jisheng Zhang guarantees kfence pool to be mapped at page granularity by allocating kfence pool before paging_init(), and then map it at page granularity during map_mem().
The previous patch has a problem: Even if kfence is disabled in cmdline, kfence pool is still allocated, which is a waste of memory.
I'm sorry for sending it repeatly, and I have no idea how to limit the email format to TEXT/PLAIN.
thanks.
>> Signed-off-by: Liu Shixin <liushixin2@huawei.com> >> --- >> arch/arm64/include/asm/kfence.h | 69 ++++++++++++++++++++++++++++++++- >> arch/arm64/mm/mmu.c | 4 +- >> 2 files changed, 70 insertions(+), 3 deletions(-) >> >> diff --git a/arch/arm64/include/asm/kfence.h b/arch/arm64/include/asm/kfence.h >> index aa855c6a0ae6..bee101eced0b 100644 >> --- a/arch/arm64/include/asm/kfence.h >> +++ b/arch/arm64/include/asm/kfence.h >> @@ -8,9 +8,76 @@ >> #ifndef __ASM_KFENCE_H >> #define __ASM_KFENCE_H >> >> +#include <linux/kfence.h> >> #include <asm/set_memory.h> >> +#include <asm/pgalloc.h> >> >> -static inline bool arch_kfence_init_pool(void) { return true; } >> +static inline int split_pud_page(pud_t *pud, unsigned long addr) >> +{ >> + int i; >> + pmd_t *pmd = pmd_alloc_one(&init_mm, addr); >> + unsigned long pfn = PFN_DOWN(__pa(addr)); >> + >> + if (!pmd) >> + return -ENOMEM; >> + >> + for (i = 0; i < PTRS_PER_PMD; i++) >> + set_pmd(pmd + i, pmd_mkhuge(pfn_pmd(pfn + i * PTRS_PER_PTE, PAGE_KERNEL))); >> + >> + smp_wmb(); /* See comment in __pte_alloc */ >> + pud_populate(&init_mm, pud, pmd); >> + flush_tlb_kernel_range(addr, addr + PUD_SIZE); >> + return 0; >> +} >> + >> +static inline int split_pmd_page(pmd_t *pmd, unsigned long addr) >> +{ >> + int i; >> + pte_t *pte = pte_alloc_one_kernel(&init_mm); >> + unsigned long pfn = PFN_DOWN(__pa(addr)); >> + >> + if (!pte) >> + return -ENOMEM; >> + >> + for (i = 0; i < PTRS_PER_PTE; i++) >> + set_pte(pte + i, pfn_pte(pfn + i, PAGE_KERNEL)); >> + >> + smp_wmb(); /* See comment in __pte_alloc */ >> + pmd_populate_kernel(&init_mm, pmd, pte); >> + >> + flush_tlb_kernel_range(addr, addr + PMD_SIZE); >> + return 0; >> +} >> + >> +static inline bool arch_kfence_init_pool(void) >> +{ >> + unsigned long addr; >> + pgd_t *pgd; >> + p4d_t *p4d; >> + pud_t *pud; >> + pmd_t *pmd; >> + >> + for (addr = (unsigned long)__kfence_pool; is_kfence_address((void *)addr); >> + addr += PAGE_SIZE) { >> + pgd = pgd_offset(&init_mm, addr); >> + if (pgd_leaf(*pgd)) >> + return false; >> + p4d = p4d_offset(pgd, addr); >> + if (p4d_leaf(*p4d)) >> + return false; >> + pud = pud_offset(p4d, addr); >> + if (pud_leaf(*pud)) { >> + if (split_pud_page(pud, addr & PUD_MASK)) >> + return false; >> + } >> + pmd = pmd_offset(pud, addr); >> + if (pmd_leaf(*pmd)) { >> + if (split_pmd_page(pmd, addr & PMD_MASK)) >> + return false; >> + } >> + } >> + return true; >> +} >> >> static inline bool kfence_protect_page(unsigned long addr, bool protect) >> { >> diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c >> index cfd9deb347c3..b2c79ccfb1c5 100644 >> --- a/arch/arm64/mm/mmu.c >> +++ b/arch/arm64/mm/mmu.c >> @@ -516,7 +516,7 @@ static void __init map_mem(pgd_t *pgdp) >> */ >> BUILD_BUG_ON(pgd_index(direct_map_end - 1) == pgd_index(direct_map_end)); >> >> - if (can_set_direct_map() || crash_mem_map || IS_ENABLED(CONFIG_KFENCE)) >> + if (can_set_direct_map() || crash_mem_map) >> flags |= NO_BLOCK_MAPPINGS | NO_CONT_MAPPINGS; >> >> /* >> @@ -1485,7 +1485,7 @@ int arch_add_memory(int nid, u64 start, u64 size, >> * KFENCE requires linear map to be mapped at page granularity, so that >> * it is possible to protect/unprotect single pages in the KFENCE pool. >> */ >> - if (can_set_direct_map() || IS_ENABLED(CONFIG_KFENCE)) >> + if (can_set_direct_map()) >> flags |= NO_BLOCK_MAPPINGS | NO_CONT_MAPPINGS; >> >> __create_pgd_mapping(swapper_pg_dir, start, __phys_to_virt(start), >> -- >> 2.18.0.huawei.25 >> >> -- >> You received this message because you are subscribed to the Google Groups "kasan-dev" group. >> To unsubscribe from this group and stop receiving emails from it, send an email to kasan-dev+unsubscribe@googlegroups.com. >> To view this discussion on the web visit https://groups.google.com/d/msgid/kasan-dev/20210918083849.2696287-1-liushixin2%40huawei.com. > >
| |