Re: [PATCH] selftests: kvm: make allocation of extra memory take effect

Re: [PATCH] selftests: kvm: make allocation of extra memory take effect

[Andrew Jones]

On Wed, May 12, 2021 at 12:31:06PM +0800, Zhenzhong Duan wrote:
> The extra memory pages is missed to be allocated during VM creating.
> perf_test_util and kvm_page_table_test use it to alloc extra memory
> currently.
> 
> Fix it by adding extra_mem_pages to the total memory calculation before
> allocate.
> 
> Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
> ---
>  tools/testing/selftests/kvm/lib/kvm_util.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/tools/testing/selftests/kvm/lib/kvm_util.c b/tools/testing/selftests/kvm/lib/kvm_util.c
> index fc83f6c5902d..159f4d62241d 100644
> --- a/tools/testing/selftests/kvm/lib/kvm_util.c
> +++ b/tools/testing/selftests/kvm/lib/kvm_util.c
> @@ -295,7 +295,7 @@ struct kvm_vm *vm_create_with_vcpus(enum vm_guest_mode mode, uint32_t nr_vcpus,
>  	 */
>  	uint64_t vcpu_pages = (DEFAULT_STACK_PGS + num_percpu_pages) * nr_vcpus;
>  	uint64_t extra_pg_pages = (extra_mem_pages + vcpu_pages) / PTES_PER_MIN_PAGE * 2;
> -	uint64_t pages = DEFAULT_GUEST_PHY_PAGES + vcpu_pages + extra_pg_pages;
> +	uint64_t pages = DEFAULT_GUEST_PHY_PAGES + extra_mem_pages + vcpu_pages + extra_pg_pages;
>  	struct kvm_vm *vm;
>  	int i;
>  
> -- 
> 2.25.1
>

Reviewed-by: Andrew Jones <drjones@redhat.com>

Thanks,
drew

[PATCH] selftests: kvm: make allocation of extra memory take effect

[Zhenzhong Duan]

The extra memory pages is missed to be allocated during VM creating.
perf_test_util and kvm_page_table_test use it to alloc extra memory
currently.

Fix it by adding extra_mem_pages to the total memory calculation before
allocate.

Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
---
 tools/testing/selftests/kvm/lib/kvm_util.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tools/testing/selftests/kvm/lib/kvm_util.c b/tools/testing/selftests/kvm/lib/kvm_util.c
index fc83f6c5902d..159f4d62241d 100644
--- a/tools/testing/selftests/kvm/lib/kvm_util.c
+++ b/tools/testing/selftests/kvm/lib/kvm_util.c
@@ -295,7 +295,7 @@ struct kvm_vm *vm_create_with_vcpus(enum vm_guest_mode mode, uint32_t nr_vcpus,
 	 */
 	uint64_t vcpu_pages = (DEFAULT_STACK_PGS + num_percpu_pages) * nr_vcpus;
 	uint64_t extra_pg_pages = (extra_mem_pages + vcpu_pages) / PTES_PER_MIN_PAGE * 2;
-	uint64_t pages = DEFAULT_GUEST_PHY_PAGES + vcpu_pages + extra_pg_pages;
+	uint64_t pages = DEFAULT_GUEST_PHY_PAGES + extra_mem_pages + vcpu_pages + extra_pg_pages;
 	struct kvm_vm *vm;
 	int i;
 
-- 
2.25.1

[PATCH] selftests: kvm: Fix a potential elf loading issue

[Zhenzhong Duan]

vm_vaddr_alloc() setup GVA to GPA mapping page by page, then GPA may not be
continuous if same memslot is used for data and page table allocation.

kvm_vm_elf_load() expects a continuous memory of GPA or else it need to
read file data page by page. Fix it by adding a check in vm_vaddr_alloc()
to ensure memory is allocated in a whole if same memslot is used for data
and page table.

Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
---
 tools/testing/selftests/kvm/lib/kvm_util.c | 13 +++++++++++--
 1 file changed, 11 insertions(+), 2 deletions(-)

diff --git a/tools/testing/selftests/kvm/lib/kvm_util.c b/tools/testing/selftests/kvm/lib/kvm_util.c
index 159f4d62241d..6118a2f8906e 100644
--- a/tools/testing/selftests/kvm/lib/kvm_util.c
+++ b/tools/testing/selftests/kvm/lib/kvm_util.c
@@ -1097,9 +1097,14 @@ vm_vaddr_t vm_vaddr_alloc(struct kvm_vm *vm, size_t sz, vm_vaddr_t vaddr_min,
 			  uint32_t data_memslot, uint32_t pgd_memslot)
 {
 	uint64_t pages = (sz >> vm->page_shift) + ((sz % vm->page_size) != 0);
+	vm_paddr_t paddr_start = 0;
 
 	virt_pgd_alloc(vm, pgd_memslot);
 
+	if (data_memslot == pgd_memslot)
+		paddr_start = vm_phy_pages_alloc(vm, pages,
+				KVM_UTIL_MIN_PFN * vm->page_size, data_memslot);
+
 	/*
 	 * Find an unused range of virtual page addresses of at least
 	 * pages in length.
@@ -1111,8 +1116,12 @@ vm_vaddr_t vm_vaddr_alloc(struct kvm_vm *vm, size_t sz, vm_vaddr_t vaddr_min,
 		pages--, vaddr += vm->page_size) {
 		vm_paddr_t paddr;
 
-		paddr = vm_phy_page_alloc(vm,
-				KVM_UTIL_MIN_PFN * vm->page_size, data_memslot);
+		if (paddr_start) {
+			paddr = paddr_start;
+			paddr_start += vm->page_size;
+		} else
+			paddr = vm_phy_page_alloc(vm,
+					KVM_UTIL_MIN_PFN * vm->page_size, data_memslot);
 
 		virt_pg_map(vm, vaddr, paddr, pgd_memslot);
 
-- 
2.25.1

Re: [PATCH] selftests: kvm: Fix a potential elf loading issue

[Paolo Bonzini]

On 12/05/21 06:31, Zhenzhong Duan wrote:
> vm_vaddr_alloc() setup GVA to GPA mapping page by page, then GPA may not be
> continuous if same memslot is used for data and page table allocation.
> 
> kvm_vm_elf_load() expects a continuous memory of GPA or else it need to
> read file data page by page. Fix it by adding a check in vm_vaddr_alloc()
> to ensure memory is allocated in a whole if same memslot is used for data
> and page table.
> 
> Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
> ---
Why not do

diff --git a/tools/testing/selftests/kvm/lib/kvm_util.c b/tools/testing/selftests/kvm/lib/kvm_util.c
index 7426163d448a..f362a066f37a 100644
--- a/tools/testing/selftests/kvm/lib/kvm_util.c
+++ b/tools/testing/selftests/kvm/lib/kvm_util.c
@@ -1170,6 +1170,9 @@ vm_vaddr_t vm_vaddr_alloc(struct kvm_vm *vm, size_t sz, vm_vaddr_t vaddr_min,
  	uint64_t pages = (sz >> vm->page_shift) + ((sz % vm->page_size) != 0);
  
  	virt_pgd_alloc(vm, pgd_memslot);
+	vm_paddr_t paddr = vm_phy_pages_alloc(vm, pages,
+					      KVM_UTIL_MIN_PFN * vm->page_size,
+					      data_memslot);
  
  	/*
  	 * Find an unused range of virtual page addresses of at least
@@ -1179,11 +1182,7 @@ vm_vaddr_t vm_vaddr_alloc(struct kvm_vm *vm, size_t sz, vm_vaddr_t vaddr_min,
  
  	/* Map the virtual pages. */
  	for (vm_vaddr_t vaddr = vaddr_start; pages > 0;
-		pages--, vaddr += vm->page_size) {
-		vm_paddr_t paddr;
-
-		paddr = vm_phy_page_alloc(vm,
-				KVM_UTIL_MIN_PFN * vm->page_size, data_memslot);
+		pages--, vaddr += vm->page_size, paddr += vm->page_size) {
  
  		virt_pg_map(vm, vaddr, paddr, pgd_memslot);
  

instead?

Paolo

RE: [PATCH] selftests: kvm: Fix a potential elf loading issue

[Duan, Zhenzhong]

> -----Original Message-----
> From: Paolo Bonzini <pbonzini@redhat.com>
> Sent: Monday, May 24, 2021 8:28 PM
> To: Duan, Zhenzhong <zhenzhong.duan@intel.com>; linux-
> kernel@vger.kernel.org
> Cc: linux-kselftest@vger.kernel.org; kvm@vger.kernel.org; shuah@kernel.org
> Subject: Re: [PATCH] selftests: kvm: Fix a potential elf loading issue
> 
> On 12/05/21 06:31, Zhenzhong Duan wrote:
> > vm_vaddr_alloc() setup GVA to GPA mapping page by page, then GPA may
> > not be continuous if same memslot is used for data and page table
> allocation.
> >
> > kvm_vm_elf_load() expects a continuous memory of GPA or else it need
> > to read file data page by page. Fix it by adding a check in
> > vm_vaddr_alloc() to ensure memory is allocated in a whole if same
> > memslot is used for data and page table.
> >
> > Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
> > ---
> Why not do
> 
> diff --git a/tools/testing/selftests/kvm/lib/kvm_util.c
> b/tools/testing/selftests/kvm/lib/kvm_util.c
> index 7426163d448a..f362a066f37a 100644
> --- a/tools/testing/selftests/kvm/lib/kvm_util.c
> +++ b/tools/testing/selftests/kvm/lib/kvm_util.c
> @@ -1170,6 +1170,9 @@ vm_vaddr_t vm_vaddr_alloc(struct kvm_vm *vm,
> size_t sz, vm_vaddr_t vaddr_min,
>   	uint64_t pages = (sz >> vm->page_shift) + ((sz % vm->page_size) != 0);
> 
>   	virt_pgd_alloc(vm, pgd_memslot);
> +	vm_paddr_t paddr = vm_phy_pages_alloc(vm, pages,
> +					      KVM_UTIL_MIN_PFN * vm-
> >page_size,
> +					      data_memslot);
> 
>   	/*
>   	 * Find an unused range of virtual page addresses of at least @@ -
> 1179,11 +1182,7 @@ vm_vaddr_t vm_vaddr_alloc(struct kvm_vm *vm,
> size_t sz, vm_vaddr_t vaddr_min,
> 
>   	/* Map the virtual pages. */
>   	for (vm_vaddr_t vaddr = vaddr_start; pages > 0;
> -		pages--, vaddr += vm->page_size) {
> -		vm_paddr_t paddr;
> -
> -		paddr = vm_phy_page_alloc(vm,
> -				KVM_UTIL_MIN_PFN * vm->page_size,
> data_memslot);
> +		pages--, vaddr += vm->page_size, paddr += vm->page_size) {
> 
>   		virt_pg_map(vm, vaddr, paddr, pgd_memslot);
> 
> 
> instead?

Yes, this is better, thanks for fixing.

Regards
Zhenzhong