Re: [PATCH v7 2/7] kvmppc: Shared pages support for secure guests

[Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>]

> A secure guest will share some of its pages with hypervisor (Eg. virtio
> bounce buffers etc). Support sharing of pages between hypervisor and
> ultravisor.
> 
> Once a secure page is converted to shared page, the device page is
> unmapped from the HV side page tables.
> 
> Signed-off-by: Bharata B Rao <bharata@linux.ibm.com>
> ---
>  arch/powerpc/include/asm/hvcall.h |  3 ++
>  arch/powerpc/kvm/book3s_hv_devm.c | 70 +++++++++++++++++++++++++++++--
>  2 files changed, 69 insertions(+), 4 deletions(-)
> 
> diff --git a/arch/powerpc/include/asm/hvcall.h b/arch/powerpc/include/asm/hvcall.h
> index 2f6b952deb0f..05b8536f6653 100644
> --- a/arch/powerpc/include/asm/hvcall.h
> +++ b/arch/powerpc/include/asm/hvcall.h
> @@ -337,6 +337,9 @@
>  #define H_TLB_INVALIDATE	0xF808
>  #define H_COPY_TOFROM_GUEST	0xF80C
> 
> +/* Flags for H_SVM_PAGE_IN */
> +#define H_PAGE_IN_SHARED        0x1
> +
>  /* Platform-specific hcalls used by the Ultravisor */
>  #define H_SVM_PAGE_IN		0xEF00
>  #define H_SVM_PAGE_OUT		0xEF04
> diff --git a/arch/powerpc/kvm/book3s_hv_devm.c b/arch/powerpc/kvm/book3s_hv_devm.c
> index 13722f27fa7d..6a3229b78fed 100644
> --- a/arch/powerpc/kvm/book3s_hv_devm.c
> +++ b/arch/powerpc/kvm/book3s_hv_devm.c
> @@ -46,6 +46,7 @@ struct kvmppc_devm_page_pvt {
>  	unsigned long *rmap;
>  	unsigned int lpid;
>  	unsigned long gpa;
> +	bool skip_page_out;
>  };
> 
>  /*
> @@ -139,6 +140,54 @@ kvmppc_devm_migrate_alloc_and_copy(struct migrate_vma *mig,
>  	return 0;
>  }
> 
> +/*
> + * Shares the page with HV, thus making it a normal page.
> + *
> + * - If the page is already secure, then provision a new page and share
> + * - If the page is a normal page, share the existing page
> + *
> + * In the former case, uses the dev_pagemap_ops migrate_to_ram handler
> + * to unmap the device page from QEMU's page tables.
> + */
> +static unsigned long
> +kvmppc_share_page(struct kvm *kvm, unsigned long gpa, unsigned long page_shift)
> +{
> +
> +	int ret = H_PARAMETER;
> +	struct page *devm_page;
> +	struct kvmppc_devm_page_pvt *pvt;
> +	unsigned long pfn;
> +	unsigned long *rmap;
> +	struct kvm_memory_slot *slot;
> +	unsigned long gfn = gpa >> page_shift;
> +	int srcu_idx;
> +
> +	srcu_idx = srcu_read_lock(&kvm->srcu);
> +	slot = gfn_to_memslot(kvm, gfn);
> +	if (!slot)
> +		goto out;
> +
> +	rmap = &slot->arch.rmap[gfn - slot->base_gfn];
> +	if (kvmppc_rmap_is_devm_pfn(*rmap)) {
> +		devm_page = pfn_to_page(*rmap & ~KVMPPC_RMAP_DEVM_PFN);
> +		pvt = (struct kvmppc_devm_page_pvt *)
> +			devm_page->zone_device_data;
> +		pvt->skip_page_out = true;
> +	}
> +
> +	pfn = gfn_to_pfn(kvm, gpa >> page_shift);

Use 'gfn'?

> +	if (is_error_noslot_pfn(pfn))
> +		goto out;
> +
> +	ret = uv_page_in(kvm->arch.lpid, pfn << page_shift, gpa, 0, page_shift);
> +	if (ret == U_SUCCESS)
> +		ret = H_SUCCESS;
> +	kvm_release_pfn_clean(pfn);

Nit: Blank line?
> +out:
> +	srcu_read_unlock(&kvm->srcu, srcu_idx);
> +	return ret;
> +}
> +
>  /*
>   * Move page from normal memory to secure memory.
>   */
> @@ -159,9 +208,12 @@ kvmppc_h_svm_page_in(struct kvm *kvm, unsigned long gpa,
>  	if (page_shift != PAGE_SHIFT)
>  		return H_P3;
> 
> -	if (flags)
> +	if (flags & ~H_PAGE_IN_SHARED)
>  		return H_P2;
> 
> +	if (flags & H_PAGE_IN_SHARED)
> +		return kvmppc_share_page(kvm, gpa, page_shift);
> +
>  	ret = H_PARAMETER;
>  	down_read(&kvm->mm->mmap_sem);
>  	srcu_idx = srcu_read_lock(&kvm->srcu);
> @@ -211,7 +263,7 @@ kvmppc_devm_fault_migrate_alloc_and_copy(struct migrate_vma *mig,
>  	struct page *dpage, *spage;
>  	struct kvmppc_devm_page_pvt *pvt;
>  	unsigned long pfn;
> -	int ret;
> +	int ret = U_SUCCESS;
> 
>  	spage = migrate_pfn_to_page(*mig->src);
>  	if (!spage || !(*mig->src & MIGRATE_PFN_MIGRATE))
> @@ -226,8 +278,18 @@ kvmppc_devm_fault_migrate_alloc_and_copy(struct migrate_vma *mig,
>  	pvt = spage->zone_device_data;
> 
>  	pfn = page_to_pfn(dpage);
> -	ret = uv_page_out(pvt->lpid, pfn << page_shift, pvt->gpa, 0,
> -			  page_shift);
> +
> +	/*
> +	 * This same function is used in two cases:

Nit: s/same//

> +	 * - When HV touches a secure page, for which we do page-out

Better to qualify page-out with "uv page-out"? its kind of counterintuitive
to do a page-out on a fault!

> +	 * - When a secure page is converted to shared page, we touch
> +	 *   the page to essentially unmap the device page. In this
> +	 *   case we skip page-out.
> +	 */
> +	if (!pvt->skip_page_out)
> +		ret = uv_page_out(pvt->lpid, pfn << page_shift, pvt->gpa, 0,
> +				  page_shift);
> +
>  	if (ret == U_SUCCESS)
>  		*mig->dst = migrate_pfn(pfn) | MIGRATE_PFN_LOCKED;
>  	else {
> -- 
> 2.21.0