Message-ID: <d9b78eb6-048a-dd17-4815-8d9635687e5b@loongson.cn>
Date:   Tue, 12 Sep 2023 17:42:37 +0800
From:   zhaotianrui <zhaotianrui@...ngson.cn>
To:     WANG Xuerui <kernel@...0n.name>, linux-kernel@...r.kernel.org,
        kvm@...r.kernel.org
Cc:     Paolo Bonzini <pbonzini@...hat.com>,
        Huacai Chen <chenhuacai@...nel.org>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        loongarch@...ts.linux.dev, Jens Axboe <axboe@...nel.dk>,
        Mark Brown <broonie@...nel.org>,
        Alex Deucher <alexander.deucher@....com>,
        Oliver Upton <oliver.upton@...ux.dev>, maobibo@...ngson.cn,
        Xi Ruoyao <xry111@...111.site>
Subject: Re: [PATCH v20 19/30] LoongArch: KVM: Implement kvm mmu operations


在 2023/9/8 上午3:57, WANG Xuerui 写道:
> On 8/31/23 16:30, Tianrui Zhao wrote:
>> Implement LoongArch kvm mmu, it is used to switch gpa to hpa when
>> guest exit because of address translation exception. This patch
>> implement allocate gpa page table, search gpa from it and flush guest
>> gpa in the table.
>>
>> Reviewed-by: Bibo Mao <maobibo@...ngson.cn>
>> Signed-off-by: Tianrui Zhao <zhaotianrui@...ngson.cn>
>> ---
>>   arch/loongarch/kvm/mmu.c | 678 +++++++++++++++++++++++++++++++++++++++
>>   1 file changed, 678 insertions(+)
>>   create mode 100644 arch/loongarch/kvm/mmu.c
>>
>> diff --git a/arch/loongarch/kvm/mmu.c b/arch/loongarch/kvm/mmu.c
>> new file mode 100644
>> index 0000000000..4bb20393f4
>> --- /dev/null
>> +++ b/arch/loongarch/kvm/mmu.c
>> @@ -0,0 +1,678 @@
>> +// SPDX-License-Identifier: GPL-2.0
>> +/*
>> + * Copyright (C) 2020-2023 Loongson Technology Corporation Limited
>> + */
>> +
>> +#include <linux/highmem.h>
>> +#include <linux/page-flags.h>
>> +#include <linux/kvm_host.h>
>> +#include <linux/uaccess.h>
>> +#include <asm/mmu_context.h>
>> +#include <asm/pgalloc.h>
>> +#include <asm/tlb.h>
>> +
>> +/*
>> + * KVM_MMU_CACHE_MIN_PAGES is the number of GPA page table 
>> translation levels
>> + * for which pages need to be cached.
>> + */
>> +#define KVM_MMU_CACHE_MIN_PAGES (CONFIG_PGTABLE_LEVELS - 1)
>> +
>> +static inline void kvm_set_pte(pte_t *ptep, pte_t pteval)
>> +{
>> +    *ptep = pteval;
>> +}
>> +
>> +/**
>> + * kvm_pgd_alloc() - Allocate and initialise a KVM GPA page directory.
>> + *
>> + * Allocate a blank KVM GPA page directory (PGD) for representing 
>> guest physical
>> + * to host physical page mappings.
>> + *
>> + * Returns:    Pointer to new KVM GPA page directory.
>> + *        NULL on allocation failure.
>> + */
>> +pgd_t *kvm_pgd_alloc(void)
>> +{
>> +    pgd_t *pgd;
>> +
>> +    pgd = (pgd_t *)__get_free_pages(GFP_KERNEL, 0);
>> +    if (pgd)
>> +        pgd_init((void *)pgd);
>> +
>> +    return pgd;
>> +}
>> +
>> +/*
>> + * Caller must hold kvm->mm_lock
>> + *
>> + * Walk the page tables of kvm to find the PTE corresponding to the
>> + * address @addr. If page tables don't exist for @addr, they will be 
>> created
>> + * from the MMU cache if @cache is not NULL.
>> + */
>> +static pte_t *kvm_populate_gpa(struct kvm *kvm,
>> +                struct kvm_mmu_memory_cache *cache,
>> +                unsigned long addr)
>> +{
>> +    pgd_t *pgd;
>> +    p4d_t *p4d;
>> +    pud_t *pud;
>> +    pmd_t *pmd;
>> +
>> +    pgd = kvm->arch.pgd + pgd_index(addr);
>> +    p4d = p4d_offset(pgd, addr);
>> +    if (p4d_none(*p4d)) {
>> +        if (!cache)
>> +            return NULL;
>> +
>> +        pud = kvm_mmu_memory_cache_alloc(cache);
>> +        pud_init(pud);
>> +        p4d_populate(NULL, p4d, pud);
>> +    }
>> +
>> +    pud = pud_offset(p4d, addr);
>> +    if (pud_none(*pud)) {
>> +        if (!cache)
>> +            return NULL;
>> +        pmd = kvm_mmu_memory_cache_alloc(cache);
>> +        pmd_init(pmd);
>> +        pud_populate(NULL, pud, pmd);
>> +    }
>> +
>> +    pmd = pmd_offset(pud, addr);
>> +    if (pmd_none(*pmd)) {
>> +        pte_t *pte;
>> +
>> +        if (!cache)
>> +            return NULL;
>> +        pte = kvm_mmu_memory_cache_alloc(cache);
>> +        clear_page(pte);
>> +        pmd_populate_kernel(NULL, pmd, pte);
>> +    }
>> +
>> +    return pte_offset_kernel(pmd, addr);
>> +}
>> +
>> +typedef int (*kvm_pte_ops)(pte_t *pte);
>> +
>> +struct kvm_ptw_ctx {
>> +    kvm_pte_ops    ops;
>> +    int        need_flush;
>> +};
>> +
>> +static int kvm_ptw_pte(pmd_t *pmd, unsigned long addr, unsigned long 
>> end,
>> +            struct kvm_ptw_ctx *context)
>> +{
>> +    pte_t *pte;
>> +    unsigned long next, start;
>> +    int ret;
>> +
>> +    ret = 0;
>> +    start = addr;
>> +    pte = pte_offset_kernel(pmd, addr);
>> +    do {
>> +        next = addr + PAGE_SIZE;
>> +        if (!pte_present(*pte))
>> +            continue;
>> +
>> +        ret |= context->ops(pte);
>> +    } while (pte++, addr = next, addr != end);
>> +
>> +    if (context->need_flush && (start + PMD_SIZE == end)) {
>> +        pte = pte_offset_kernel(pmd, 0);
>> +        pmd_clear(pmd);
>> +        free_page((unsigned long)pte);
>> +    }
>> +
>> +    return ret;
>> +}
>> +
>> +static int kvm_ptw_pmd(pud_t *pud, unsigned long addr, unsigned long 
>> end,
>> +            struct kvm_ptw_ctx *context)
>> +{
>> +    pmd_t *pmd;
>> +    unsigned long next, start;
>> +    int ret;
>> +
>> +    ret = 0;
>> +    start = addr;
>> +    pmd = pmd_offset(pud, addr);
>> +    do {
>> +        next = pmd_addr_end(addr, end);
>> +        if (!pmd_present(*pmd))
>> +            continue;
>> +
>> +        ret |= kvm_ptw_pte(pmd, addr, next, context);
>> +    } while (pmd++, addr = next, addr != end);
>> +
>> +#ifndef __PAGETABLE_PMD_FOLDED
>> +    if (context->need_flush && (start + PUD_SIZE == end)) {
>> +        pmd = pmd_offset(pud, 0);
>> +        pud_clear(pud);
>> +        free_page((unsigned long)pmd);
>> +    }
>> +#endif
>> +
>> +    return ret;
>> +}
>> +
>> +static int kvm_ptw_pud(pgd_t *pgd, unsigned long addr, unsigned long 
>> end,
>> +            struct kvm_ptw_ctx *context)
>> +{
>> +    p4d_t *p4d;
>> +    pud_t *pud;
>> +    int ret = 0;
>> +    unsigned long next;
>> +#ifndef __PAGETABLE_PUD_FOLDED
>> +    unsigned long start = addr;
>> +#endif
>> +
>> +    p4d = p4d_offset(pgd, addr);
>> +    pud = pud_offset(p4d, addr);
>> +    do {
>> +        next = pud_addr_end(addr, end);
>> +        if (!pud_present(*pud))
>> +            continue;
>> +
>> +        ret |= kvm_ptw_pmd(pud, addr, next, context);
>> +    } while (pud++, addr = next, addr != end);
>> +
>> +#ifndef __PAGETABLE_PUD_FOLDED
>> +    if (context->need_flush && (start + PGDIR_SIZE == end)) {
>> +        pud = pud_offset(p4d, 0);
>> +        p4d_clear(p4d);
>> +        free_page((unsigned long)pud);
>> +    }
>> +#endif
>> +
>> +    return ret;
>> +}
>> +
>> +static int kvm_ptw_pgd(pgd_t *pgd, unsigned long addr, unsigned long 
>> end,
>> +            struct kvm_ptw_ctx *context)
>> +{
>> +    unsigned long next;
>> +    int ret;
>> +
>> +    ret = 0;
>> +    if (addr > end - 1)
>> +        return ret;
>> +    pgd = pgd + pgd_index(addr);
>> +    do {
>> +        next = pgd_addr_end(addr, end);
>> +        if (!pgd_present(*pgd))
>> +            continue;
>> +
>> +        ret |= kvm_ptw_pud(pgd, addr, next, context);
>> +    }  while (pgd++, addr = next, addr != end);
>> +
>> +    return ret;
>> +}
>> +
>> +/*
>> + * clear pte entry
>> + */
>> +static int kvm_flush_pte(pte_t *pte)
>> +{
>> +    kvm_set_pte(pte, __pte(0));
>> +    return 1;
>> +}
>> +
>> +/**
>> + * kvm_flush_range() - Flush a range of guest physical addresses.
>> + * @kvm:    KVM pointer.
>> + * @start_gfn:    Guest frame number of first page in GPA range to 
>> flush.
>> + * @end_gfn:    Guest frame number of last page in GPA range to flush.
>> + *
>> + * Flushes a range of GPA mappings from the GPA page tables.
>> + *
>> + * The caller must hold the @kvm->mmu_lock spinlock.
>> + *
>> + * Returns:    Whether its safe to remove the top level page 
>> directory because
>> + *        all lower levels have been removed.
>> + */
>> +static bool kvm_flush_range(struct kvm *kvm, gfn_t start_gfn, gfn_t 
>> end_gfn)
>> +{
>> +    struct kvm_ptw_ctx ctx;
>> +
>> +    ctx.ops = kvm_flush_pte;
>> +    ctx.need_flush = 1;
>> +
>> +    return kvm_ptw_pgd(kvm->arch.pgd, start_gfn << PAGE_SHIFT,
>> +                end_gfn << PAGE_SHIFT, &ctx);
>> +}
>> +
>> +/*
>> + * kvm_mkclean_pte
>> + * Mark a range of guest physical address space clean (writes fault) 
>> in the VM's
>> + * GPA page table to allow dirty page tracking.
>> + */
>> +static int kvm_mkclean_pte(pte_t *pte)
>> +{
>> +    pte_t val;
>> +
>> +    val = *pte;
>> +    if (pte_dirty(val)) {
>> +        *pte = pte_mkclean(val);
>> +        return 1;
>> +    }
>> +    return 0;
>> +}
>> +
>> +/*
>> + * kvm_mkclean_gpa_pt() - Make a range of guest physical addresses 
>> clean.
>> + * @kvm:    KVM pointer.
>> + * @start_gfn:    Guest frame number of first page in GPA range to 
>> flush.
>> + * @end_gfn:    Guest frame number of last page in GPA range to flush.
>> + *
>> + * Make a range of GPA mappings clean so that guest writes will 
>> fault and
>> + * trigger dirty page logging.
>> + *
>> + * The caller must hold the @kvm->mmu_lock spinlock.
>> + *
>> + * Returns:    Whether any GPA mappings were modified, which would 
>> require
>> + *        derived mappings (GVA page tables & TLB enties) to be
>> + *        invalidated.
>> + */
>> +static int kvm_mkclean_gpa_pt(struct kvm *kvm, gfn_t start_gfn, 
>> gfn_t end_gfn)
>> +{
>> +    struct kvm_ptw_ctx ctx;
>> +
>> +    ctx.ops = kvm_mkclean_pte;
>> +    ctx.need_flush = 0;
>> +    return kvm_ptw_pgd(kvm->arch.pgd, start_gfn << PAGE_SHIFT,
>> +                end_gfn << PAGE_SHIFT, &ctx);
>> +}
>> +
>> +/*
>> + * kvm_arch_mmu_enable_log_dirty_pt_masked() - write protect dirty 
>> pages
>> + * @kvm:    The KVM pointer
>> + * @slot:    The memory slot associated with mask
>> + * @gfn_offset:    The gfn offset in memory slot
>> + * @mask:    The mask of dirty pages at offset 'gfn_offset' in this 
>> memory
>> + *        slot to be write protected
>> + *
>> + * Walks bits set in mask write protects the associated pte's. 
>> Caller must
>> + * acquire @kvm->mmu_lock.
>> + */
>> +void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
>> +        struct kvm_memory_slot *slot,
>> +        gfn_t gfn_offset, unsigned long mask)
>> +{
>> +    gfn_t base_gfn = slot->base_gfn + gfn_offset;
>> +    gfn_t start = base_gfn +  __ffs(mask);
> One extra space after the plus sign?
Thanks, I will remove the extra space.
>> +    gfn_t end = base_gfn + __fls(mask) + 1;
>> +
>> +    kvm_mkclean_gpa_pt(kvm, start, end);
>> +}
>> +
>> +void kvm_arch_commit_memory_region(struct kvm *kvm,
>> +                   struct kvm_memory_slot *old,
>> +                   const struct kvm_memory_slot *new,
>> +                   enum kvm_mr_change change)
>> +{
>> +    int needs_flush;
>> +
>> +    /*
>> +     * If dirty page logging is enabled, write protect all pages in 
>> the slot
>> +     * ready for dirty logging.
>> +     *
>> +     * There is no need to do this in any of the following cases:
>> +     * CREATE:    No dirty mappings will already exist.
>> +     * MOVE/DELETE:    The old mappings will already have been 
>> cleaned up by
>> +     *        kvm_arch_flush_shadow_memslot()
>> +     */
>> +    if (change == KVM_MR_FLAGS_ONLY &&
>> +        (!(old->flags & KVM_MEM_LOG_DIRTY_PAGES) &&
>> +         new->flags & KVM_MEM_LOG_DIRTY_PAGES)) {
>> +        spin_lock(&kvm->mmu_lock);
>> +        /* Write protect GPA page table entries */
>> +        needs_flush = kvm_mkclean_gpa_pt(kvm, new->base_gfn,
>> +                    new->base_gfn + new->npages);
>> +        if (needs_flush)
>> +            kvm_flush_remote_tlbs(kvm);
>> +        spin_unlock(&kvm->mmu_lock);
>> +    }
>> +}
>> +
>> +void kvm_arch_flush_shadow_all(struct kvm *kvm)
>> +{
>> +    /* Flush whole GPA */
>> +    kvm_flush_range(kvm, 0, kvm->arch.gpa_size >> PAGE_SHIFT);
>> +    /* Flush vpid for each vCPU individually */
>> +    kvm_flush_remote_tlbs(kvm);
>> +}
>> +
>> +void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
>> +        struct kvm_memory_slot *slot)
>> +{
>> +    int ret;
>> +
>> +    /*
>> +     * The slot has been made invalid (ready for moving or 
>> deletion), so we
>> +     * need to ensure that it can no longer be accessed by any guest 
>> vCPUs.
>> +     */
>> +    spin_lock(&kvm->mmu_lock);
>> +    /* Flush slot from GPA */
>> +    ret = kvm_flush_range(kvm, slot->base_gfn,
>> +            slot->base_gfn + slot->npages);
>> +    /* Let implementation do the rest */
>> +    if (ret)
>> +        kvm_flush_remote_tlbs(kvm);
>> +    spin_unlock(&kvm->mmu_lock);
>> +}
>> +
>> +void _kvm_destroy_mm(struct kvm *kvm)
>> +{
>> +    /* It should always be safe to remove after flushing the whole 
>> range */
>> +    kvm_flush_range(kvm, 0, kvm->arch.gpa_size >> PAGE_SHIFT);
>> +    pgd_free(NULL, kvm->arch.pgd);
>> +    kvm->arch.pgd = NULL;
>> +}
>> +
>> +/*
>> + * Mark a range of guest physical address space old (all accesses 
>> fault) in the
>> + * VM's GPA page table to allow detection of commonly used pages.
>> + */
>> +static int kvm_mkold_pte(pte_t *pte)
>> +{
>> +    pte_t val;
>> +
>> +    val = *pte;
> "pte_t val = *pte" would be enough... You may want to check the entire 
> patch series for simplifications like this.
Thanks, I will fix this.
>> +    if (pte_young(val)) {
>> +        *pte = pte_mkold(val);
>> +        return 1;
>> +    }
>> +    return 0;
>> +}
>> +
>> +bool kvm_unmap_gfn_range(struct kvm *kvm, struct kvm_gfn_range *range)
>> +{
>> +    return kvm_flush_range(kvm, range->start, range->end);
>> +}
>> +
>> +bool kvm_set_spte_gfn(struct kvm *kvm, struct kvm_gfn_range *range)
>> +{
>> +    gpa_t gpa = range->start << PAGE_SHIFT;
>> +    pte_t hva_pte = range->pte;
> This has become "range->arg.pte" since commit 3e1efe2b67d3 ("KVM: Wrap 
> kvm_{gfn,hva}_range.pte in a per-action union") which is already 
> inside linux-next.
Thanks, I will update it.
>> +    pte_t *ptep = kvm_populate_gpa(kvm, NULL, gpa);
>> +    pte_t old_pte;
>> +
>> +    if (!ptep)
>> +        return false;
>> +
>> +    /* Mapping may need adjusting depending on memslot flags */
>> +    old_pte = *ptep;
>> +    if (range->slot->flags & KVM_MEM_LOG_DIRTY_PAGES && 
>> !pte_dirty(old_pte))
>> +        hva_pte = pte_mkclean(hva_pte);
>> +    else if (range->slot->flags & KVM_MEM_READONLY)
>> +        hva_pte = pte_wrprotect(hva_pte);
>> +
>> +    kvm_set_pte(ptep, hva_pte);
>> +
>> +    /* Replacing an absent or old page doesn't need flushes */
>> +    if (!pte_present(old_pte) || !pte_young(old_pte))
>> +        return false;
>> +
>> +    /* Pages swapped, aged, moved, or cleaned require flushes */
>> +    return !pte_present(hva_pte) ||
>> +           !pte_young(hva_pte) ||
>> +           pte_pfn(old_pte) != pte_pfn(hva_pte) ||
>> +           (pte_dirty(old_pte) && !pte_dirty(hva_pte));
>> +}
>> +
>> +bool kvm_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range)
>> +{
>> +    struct kvm_ptw_ctx ctx;
>> +
>> +    ctx.ops = kvm_mkold_pte;
>> +    ctx.need_flush = 0;
>> +    return kvm_ptw_pgd(kvm->arch.pgd, range->start << PAGE_SHIFT,
>> +                range->end << PAGE_SHIFT, &ctx);
>> +}
>> +
>> +bool kvm_test_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range)
>> +{
>> +    gpa_t gpa = range->start << PAGE_SHIFT;
>> +    pte_t *ptep = kvm_populate_gpa(kvm, NULL, gpa);
>> +
>> +    if (ptep && pte_present(*ptep) && pte_young(*ptep))
>> +        return true;
>> +
>> +    return false;
>> +}
>> +
>> +/**
>> + * kvm_map_page_fast() - Fast path GPA fault handler.
>> + * @vcpu:        vCPU pointer.
>> + * @gpa:        Guest physical address of fault.
>> + * @write:    Whether the fault was due to a write.
>> + *
>> + * Perform fast path GPA fault handling, doing all that can be done 
>> without
>> + * calling into KVM. This handles marking old pages young (for idle 
>> page
>> + * tracking), and dirtying of clean pages (for dirty page logging).
>> + *
>> + * Returns:    0 on success, in which case we can update derived 
>> mappings and
>> + *        resume guest execution.
>> + *        -EFAULT on failure due to absent GPA mapping or write to
>> + *        read-only page, in which case KVM must be consulted.
>> + */
>> +static int kvm_map_page_fast(struct kvm_vcpu *vcpu, unsigned long gpa,
>> +                   bool write)
>> +{
>> +    struct kvm *kvm = vcpu->kvm;
>> +    gfn_t gfn = gpa >> PAGE_SHIFT;
>> +    pte_t *ptep;
>> +    kvm_pfn_t pfn = 0;
>> +    bool pfn_valid = false, pfn_dirty = false;
>> +    int ret = 0;
>> +
>> +    spin_lock(&kvm->mmu_lock);
>> +
>> +    /* Fast path - just check GPA page table for an existing entry */
>> +    ptep = kvm_populate_gpa(kvm, NULL, gpa);
>> +    if (!ptep || !pte_present(*ptep)) {
>> +        ret = -EFAULT;
>> +        goto out;
>> +    }
>> +
>> +    /* Track access to pages marked old */
>> +    if (!pte_young(*ptep)) {
>> +        kvm_set_pte(ptep, pte_mkyoung(*ptep));
>> +        pfn = pte_pfn(*ptep);
>> +        pfn_valid = true;
>> +        /* call kvm_set_pfn_accessed() after unlock */
>> +    }
>> +    if (write && !pte_dirty(*ptep)) {
>> +        if (!pte_write(*ptep)) {
>> +            ret = -EFAULT;
>> +            goto out;
>> +        }
>> +
>> +        /* Track dirtying of writeable pages */
>> +        kvm_set_pte(ptep, pte_mkdirty(*ptep));
>> +        pfn = pte_pfn(*ptep);
>> +        pfn_dirty = true;
>> +    }
>> +
>> +out:
>> +    spin_unlock(&kvm->mmu_lock);
>> +    if (pfn_valid)
>> +        kvm_set_pfn_accessed(pfn);
>> +    if (pfn_dirty) {
>> +        mark_page_dirty(kvm, gfn);
>> +        kvm_set_pfn_dirty(pfn);
>> +    }
>> +    return ret;
>> +}
>> +
>> +/**
>> + * kvm_map_page() - Map a guest physical page.
>> + * @vcpu:        vCPU pointer.
>> + * @gpa:        Guest physical address of fault.
>> + * @write:    Whether the fault was due to a write.
>> + *
>> + * Handle GPA faults by creating a new GPA mapping (or updating an 
>> existing
>> + * one).
>> + *
>> + * This takes care of marking pages young or dirty (idle/dirty page 
>> tracking),
>> + * asking KVM for the corresponding PFN, and creating a mapping in 
>> the GPA page
>> + * tables. Derived mappings (GVA page tables and TLBs) must be 
>> handled by the
>> + * caller.
>> + *
>> + * Returns:    0 on success
>> + *        -EFAULT if there is no memory region at @gpa or a write was
>> + *        attempted to a read-only memory region. This is usually 
>> handled
>> + *        as an MMIO access.
>> + */
>> +static int kvm_map_page(struct kvm_vcpu *vcpu, unsigned long gpa, 
>> bool write)
>> +{
>> +    bool writeable;
>> +    int srcu_idx, err = 0, retry_no = 0;
>> +    unsigned long hva;
>> +    unsigned long mmu_seq;
>> +    unsigned long prot_bits;
>> +    pte_t *ptep, new_pte;
>> +    kvm_pfn_t pfn;
>> +    gfn_t gfn = gpa >> PAGE_SHIFT;
>> +    struct vm_area_struct *vma;
>> +    struct kvm *kvm = vcpu->kvm;
>> +    struct kvm_memory_slot *memslot;
>> +    struct kvm_mmu_memory_cache *memcache = &vcpu->arch.mmu_page_cache;
>> +
>> +    /* Try the fast path to handle old / clean pages */
>> +    srcu_idx = srcu_read_lock(&kvm->srcu);
>> +    err = kvm_map_page_fast(vcpu, gpa, write);
>> +    if (!err)
>> +        goto out;
>> +
>> +    memslot = gfn_to_memslot(kvm, gfn);
>> +    hva = gfn_to_hva_memslot_prot(memslot, gfn, &writeable);
>> +    if (kvm_is_error_hva(hva) || (write && !writeable))
>> +        goto out;
>> +
>> +    mmap_read_lock(current->mm);
>> +    vma = find_vma_intersection(current->mm, hva, hva + 1);
>> +    if (unlikely(!vma)) {
>> +        kvm_err("Failed to find VMA for hva 0x%lx\n", hva);
>> +        mmap_read_unlock(current->mm);
>> +        err = -EFAULT;
>> +        goto out;
>> +    }
>> +    mmap_read_unlock(current->mm);
>> +
>> +    /* We need a minimum of cached pages ready for page table 
>> creation */
>> +    err = kvm_mmu_topup_memory_cache(memcache, 
>> KVM_MMU_CACHE_MIN_PAGES);
>> +    if (err)
>> +        goto out;
>> +
>> +retry:
>> +    /*
>> +     * Used to check for invalidations in progress, of the pfn that is
>> +     * returned by pfn_to_pfn_prot below.
>> +     */
>> +    mmu_seq = kvm->mmu_invalidate_seq;
>> +    /*
>> +     * Ensure the read of mmu_invalidate_seq isn't reordered with 
>> PTE reads in
>> +     * gfn_to_pfn_prot() (which calls get_user_pages()), so that we 
>> don't
>> +     * risk the page we get a reference to getting unmapped before 
>> we have a
>> +     * chance to grab the mmu_lock without mmu_invalidate_retry() 
>> noticing.
>> +     *
>> +     * This smp_rmb() pairs with the effective smp_wmb() of the 
>> combination
>> +     * of the pte_unmap_unlock() after the PTE is zapped, and the
>> +     * spin_lock() in 
>> kvm_mmu_invalidate_invalidate_<page|range_end>() before
>> +     * mmu_invalidate_seq is incremented.
>> +     */
>> +    smp_rmb();
>> +
>> +    /* Slow path - ask KVM core whether we can access this GPA */
>> +    pfn = gfn_to_pfn_prot(kvm, gfn, write, &writeable);
>> +    if (is_error_noslot_pfn(pfn)) {
>> +        err = -EFAULT;
>> +        goto out;
>> +    }
>> +
>> +    /* Check if an invalidation has taken place since we got pfn */
>> +    if (mmu_invalidate_retry(kvm, mmu_seq)) {
>> +        /*
> Wrong indentation?
I will fix this indentation.

Thanks
Tianrui Zhao
>> +         * This can happen when mappings are changed asynchronously, 
>> but
>> +         * also synchronously if a COW is triggered by
>> +         * gfn_to_pfn_prot().
>> +         */
>> +        kvm_set_pfn_accessed(pfn);
>> +        kvm_release_pfn_clean(pfn);
>> +        if (retry_no > 100) {
>> +            retry_no = 0;
>> +            schedule();
>> +        }
>> +        retry_no++;
>> +        goto retry;
>> +    }
>> +
>> +    /*
>> +     * For emulated devices such virtio device, actual cache 
>> attribute is
>> +     * determined by physical machine.
>> +     * For pass through physical device, it should be uncachable
>> +     */
>> +    prot_bits = _PAGE_PRESENT | __READABLE;
>> +    if (vma->vm_flags & (VM_IO | VM_PFNMAP))
>> +        prot_bits |= _CACHE_SUC;
>> +    else
>> +        prot_bits |= _CACHE_CC;
>> +
>> +    if (writeable) {
>> +        prot_bits |= _PAGE_WRITE;
>> +        if (write)
>> +            prot_bits |= __WRITEABLE;
>> +    }
>> +
>> +    /* Ensure page tables are allocated */
>> +    spin_lock(&kvm->mmu_lock);
>> +    ptep = kvm_populate_gpa(kvm, memcache, gpa);
>> +    new_pte = pfn_pte(pfn, __pgprot(prot_bits));
>> +    kvm_set_pte(ptep, new_pte);
>> +
>> +    err = 0;
>> +    spin_unlock(&kvm->mmu_lock);
>> +
>> +    if (prot_bits & _PAGE_DIRTY) {
>> +        mark_page_dirty(kvm, gfn);
>> +        kvm_set_pfn_dirty(pfn);
>> +    }
>> +
>> +    kvm_set_pfn_accessed(pfn);
>> +    kvm_release_pfn_clean(pfn);
>> +out:
>> +    srcu_read_unlock(&kvm->srcu, srcu_idx);
>> +    return err;
>> +}
>> +
>> +int kvm_handle_mm_fault(struct kvm_vcpu *vcpu, unsigned long gpa, 
>> bool write)
>> +{
>> +    int ret;
>> +
>> +    ret = kvm_map_page(vcpu, gpa, write);
>> +    if (ret)
>> +        return ret;
>> +
>> +    /* Invalidate this entry in the TLB */
>> +    return kvm_flush_tlb_gpa(vcpu, gpa);
>> +}
>> +
>> +void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot 
>> *memslot)
>> +{
>> +
>> +}
>> +
>> +int kvm_arch_prepare_memory_region(struct kvm *kvm,
>> +                   const struct kvm_memory_slot *old,
>> +                   struct kvm_memory_slot *new,
>> +                   enum kvm_mr_change change)
>> +{
>> +    return 0;
>> +}
>> +
>> +void kvm_arch_flush_remote_tlbs_memslot(struct kvm *kvm,
>> +                    const struct kvm_memory_slot *memslot)
>> +{
>> +    kvm_flush_remote_tlbs(kvm);
>> +}
>

Powered by blists - more mailing lists