Re: [PATCH v6 5/6] KVM: arm64: Initialize the ptdump parser with stage-2 attributes

[Sebastian Ene <sebastianene@google.com>]

On Tue, Feb 20, 2024 at 03:10:34PM +0000, Sebastian Ene wrote:
> Define a set of attributes used by the ptdump parser to display the
> properties of a guest memory region covered by a pagetable descriptor.
> Build a description of the pagetable levels and initialize the parser
> with this configuration.
> 
> Signed-off-by: Sebastian Ene <sebastianene@google.com>
> ---
>  arch/arm64/kvm/ptdump.c | 146 ++++++++++++++++++++++++++++++++++++++--
>  1 file changed, 139 insertions(+), 7 deletions(-)
> 
> diff --git a/arch/arm64/kvm/ptdump.c b/arch/arm64/kvm/ptdump.c
> index 9b04c24bb9be..2c4e0c122d23 100644
> --- a/arch/arm64/kvm/ptdump.c
> +++ b/arch/arm64/kvm/ptdump.c
> @@ -14,6 +14,61 @@
>  #include <kvm_ptdump.h>
>  
>  
> +#define MARKERS_LEN		(2)
> +#define KVM_PGTABLE_LEVELS	(KVM_PGTABLE_LAST_LEVEL + 1)
> +
> +struct kvm_ptdump_guest_state {
> +	struct kvm		*kvm;
> +	struct pg_state		parser_state;
> +	struct addr_marker	ipa_marker[MARKERS_LEN];
> +	struct pg_level		level[KVM_PGTABLE_LEVELS];
> +	struct ptdump_range	range[MARKERS_LEN];
> +};
> +
> +static const struct prot_bits stage2_pte_bits[] = {
> +	{
> +		.mask	= PTE_VALID,
> +		.val	= PTE_VALID,
> +		.set	= " ",
> +		.clear	= "F",
> +	}, {
> +		.mask	= KVM_PTE_LEAF_ATTR_HI_S2_XN | PTE_VALID,
> +		.val	= KVM_PTE_LEAF_ATTR_HI_S2_XN | PTE_VALID,
> +		.set	= "XN",
> +		.clear	= "  ",
> +	}, {
> +		.mask	= KVM_PTE_LEAF_ATTR_LO_S2_S2AP_R | PTE_VALID,
> +		.val	= KVM_PTE_LEAF_ATTR_LO_S2_S2AP_R | PTE_VALID,
> +		.set	= "R",
> +		.clear	= " ",
> +	}, {
> +		.mask	= KVM_PTE_LEAF_ATTR_LO_S2_S2AP_W | PTE_VALID,
> +		.val	= KVM_PTE_LEAF_ATTR_LO_S2_S2AP_W | PTE_VALID,
> +		.set	= "W",
> +		.clear	= " ",
> +	}, {
> +		.mask	= KVM_PTE_LEAF_ATTR_LO_S2_AF | PTE_VALID,
> +		.val	= KVM_PTE_LEAF_ATTR_LO_S2_AF | PTE_VALID,
> +		.set	= "AF",
> +		.clear	= "  ",
> +	}, {
> +		.mask	= PTE_NG,
> +		.val	= PTE_NG,
> +		.set	= "FnXS",
> +		.clear	= "  ",
> +	}, {
> +		.mask	= PTE_CONT | PTE_VALID,
> +		.val	= PTE_CONT | PTE_VALID,
> +		.set	= "CON",
> +		.clear	= "   ",
> +	}, {
> +		.mask	= PTE_TABLE_BIT,
> +		.val	= PTE_TABLE_BIT,
> +		.set	= "   ",
> +		.clear	= "BLK",
> +	},
> +};
> +
>  static int kvm_ptdump_visitor(const struct kvm_pgtable_visit_ctx *ctx,
>  			      enum kvm_pgtable_walk_flags visit)
>  {
> @@ -37,15 +92,78 @@ static int kvm_ptdump_show_common(struct seq_file *m,
>  	return kvm_pgtable_walk(pgtable, 0, BIT(pgtable->ia_bits), &walker);
>  }
>  
> +static int kvm_ptdump_build_levels(struct pg_level *level, u32 start_lvl)
> +{
> +	static const char * const level_names[] = {"PGD", "PUD", "PMD", "PTE"};
> +	u32 i = 0;
> +	u64 mask = 0;
> +
> +	if (start_lvl > 2) {
> +		pr_err("invalid start_lvl %u\n", start_lvl);
> +		return -EINVAL;
> +	}
> +
> +	for (i = 0; i < ARRAY_SIZE(stage2_pte_bits); i++)
> +		mask |= stage2_pte_bits[i].mask;
> +
> +	for (i = start_lvl; i <= KVM_PGTABLE_LAST_LEVEL; i++) {
> +		level[i].name	= level_names[i];
> +		level[i].num	= ARRAY_SIZE(stage2_pte_bits);
> +		level[i].bits	= stage2_pte_bits;
> +		level[i].mask	= mask;
> +	}
> +
> +	if (start_lvl > 0)
> +		level[start_lvl].name = level_names[0];
> +
> +	return 0;
> +}
> +
> +static struct kvm_ptdump_guest_state
> +*kvm_ptdump_parser_init(struct kvm *kvm)
> +{
> +	struct kvm_ptdump_guest_state *st;
> +	struct kvm_s2_mmu *mmu = &kvm->arch.mmu;
> +	struct kvm_pgtable *pgtable = mmu->pgt;
> +	int ret;
> +
> +	st = kzalloc(sizeof(struct kvm_ptdump_guest_state), GFP_KERNEL_ACCOUNT);
> +	if (!st)
> +		return NULL;
> +
> +	ret = kvm_ptdump_build_levels(&st->level[0], pgtable->start_level);
> +	if (ret)
> +		goto free_with_state;
> +
> +	st->ipa_marker[0].name		= "Guest IPA";
> +	st->ipa_marker[1].start_address = BIT(pgtable->ia_bits);
> +	st->range[0].end		= BIT(pgtable->ia_bits);
> +
> +	st->kvm				= kvm;
> +	st->parser_state = (struct pg_state) {
> +		.marker		= &st->ipa_marker[0],
> +		.level		= -1,
> +		.pg_level	= &st->level[0],
> +		.ptdump.range	= &st->range[0],
> +	};
> +
> +	return st;
> +free_with_state:
> +	kfree(st);
> +	return NULL;
> +}
> +
>  static int kvm_ptdump_guest_show(struct seq_file *m, void *)
>  {
> -	struct kvm *kvm = m->private;
> +	struct kvm_ptdump_guest_state *st = m->private;
> +	struct kvm *kvm = st->kvm;
>  	struct kvm_s2_mmu *mmu = &kvm->arch.mmu;
> -	struct pg_state parser_state = {0};
>  	int ret;
>  
> +	st->parser_state.seq = m;
> +
>  	write_lock(&kvm->mmu_lock);
> -	ret = kvm_ptdump_show_common(m, mmu->pgt, &parser_state);
> +	ret = kvm_ptdump_show_common(m, mmu->pgt, &st->parser_state);
>  	write_unlock(&kvm->mmu_lock);
>  
>  	return ret;
> @@ -54,22 +172,36 @@ static int kvm_ptdump_guest_show(struct seq_file *m, void *)
>  static int kvm_ptdump_guest_open(struct inode *m, struct file *file)
>  {
>  	struct kvm *kvm = m->i_private;
> +	struct kvm_ptdump_guest_state *st;
>  	int ret;
>  
> -	if (!kvm_get_kvm_safe(kvm))
> -		return -ENOENT;

FIXME: Keep kvm_get_kvm_safe before invoking kvm_ptdump_parser_init to
avoid UAF if kvm gets destroyed.

> +	st = kvm_ptdump_parser_init(kvm);
> +	if (!st)
> +		return -ENOMEM;
>  
> -	ret = single_open(file, kvm_ptdump_guest_show, m->i_private);
> +	if (!kvm_get_kvm_safe(kvm)) {
> +		ret = -ENOENT;
> +		goto free_with_state;
> +	}
> +
> +	ret = single_open(file, kvm_ptdump_guest_show, st);
>  	if (ret < 0)
> -		kvm_put_kvm(kvm);
> +		goto free_with_kvm_ref;
>  
>  	return ret;
> +free_with_kvm_ref:
> +	kvm_put_kvm(kvm);
> +free_with_state:
> +	kfree(st);
> +	return ret;
>  }
>  
>  static int kvm_ptdump_guest_close(struct inode *m, struct file *file)
>  {
>  	struct kvm *kvm = m->i_private;
> +	void *st = ((struct seq_file *)file->private_data)->private;
>  
> +	kfree(st);
>  	kvm_put_kvm(kvm);
>  	return single_release(m, file);
>  }
> -- 
> 2.44.0.rc0.258.g7320e95886-goog
>