From: donsheng dongsheng.x.zhang@intel.com
If the host was booted with the "default_hugepagesz=1G" kernel command-line parameter, running the NX hugepage test will fail with error "Invalid argument" at the TEST_ASSERT line in kvm_util.c's __vm_mem_region_delete() function: static void __vm_mem_region_delete(struct kvm_vm *vm, struct userspace_mem_region *region, bool unlink) { int ret; ... ret = munmap(region->mmap_start, region->mmap_size); TEST_ASSERT(!ret, __KVM_SYSCALL_ERROR("munmap()", ret)); ... }
NX hugepage test creates a VM with a data slot of 6M size backed with huge pages. If the default hugetlb page size is set to 1G, calling mmap() with MAP_HUGETLB and a length of 6M will succeed but calling its matching munmap() will fail. Documentation/admin-guide/mm/hugetlbpage.rst specifies this behavior:
"Syscalls that operate on memory backed by hugetlb pages only have their lengths aligned to the native page size of the processor; they will normally fail with errno set to EINVAL or exclude hugetlb pages that extend beyond the length if not hugepage aligned. For example, munmap(2) will fail if memory is backed by a hugetlb page and the length is smaller than the hugepage size."
Explicitly use MAP_HUGE_2MB in conjunction with MAP_HUGETLB to fix the issue.
Signed-off-by: donsheng dongsheng.x.zhang@intel.com Suggested-by: Zide Chen zide.chen@intel.com --- tools/testing/selftests/kvm/x86_64/nx_huge_pages_test.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/tools/testing/selftests/kvm/x86_64/nx_huge_pages_test.c b/tools/testing/selftests/kvm/x86_64/nx_huge_pages_test.c index 17bbb96fc4df..146e9033e206 100644 --- a/tools/testing/selftests/kvm/x86_64/nx_huge_pages_test.c +++ b/tools/testing/selftests/kvm/x86_64/nx_huge_pages_test.c @@ -129,7 +129,7 @@ void run_test(int reclaim_period_ms, bool disable_nx_huge_pages,
vcpu = vm_vcpu_add(vm, 0, guest_code);
- vm_userspace_mem_region_add(vm, VM_MEM_SRC_ANONYMOUS_HUGETLB, + vm_userspace_mem_region_add(vm, VM_MEM_SRC_ANONYMOUS_HUGETLB_2MB, HPAGE_GPA, HPAGE_SLOT, HPAGE_SLOT_NPAGES, 0);
From: donsheng dongsheng.x.zhang@intel.com
If mmap() is called with MAP_HUGETLB and the requested mapping size is not a multiple of the underlying hugetlb page size, kernel will automatically align the mmap size, but munmap() will fail if you specify the size you asked for rather than the size that was assigned by Linux kernel mmap().
To avoid munmap() failure, add sanity check for mem_size in vm_mem_add() to ensure that mem_size is aligned to page size of the underlying hugetlb backing src.
This assertion helps to ensure that vm_mem_add() is called with aligned mapping size, especially when vm_mem_add() is called to add guest memory backed by VM_MEM_SRC_ANONYMOUS_HUGETLB which relies on the default hugetlb page size.
Signed-off-by: donsheng dongsheng.x.zhang@intel.com --- tools/testing/selftests/kvm/lib/kvm_util.c | 5 +++++ 1 file changed, 5 insertions(+)
diff --git a/tools/testing/selftests/kvm/lib/kvm_util.c b/tools/testing/selftests/kvm/lib/kvm_util.c index b2262b5fad9e..827bb3d57815 100644 --- a/tools/testing/selftests/kvm/lib/kvm_util.c +++ b/tools/testing/selftests/kvm/lib/kvm_util.c @@ -985,6 +985,11 @@ void vm_mem_add(struct kvm_vm *vm, enum vm_mem_backing_src_type src_type, " vm->max_gfn: 0x%lx vm->page_size: 0x%x", guest_paddr, npages, vm->max_gfn, vm->page_size);
+ TEST_ASSERT(!(is_backing_src_hugetlb(src_type) && + (mem_size & (backing_src_pagesz - 1))), + "mem_size 0x%lx is not aligned to backing src %s's page size 0x%lx", + mem_size, vm_mem_backing_src_alias(src_type)->name, backing_src_pagesz); + /* * Confirm a mem region with an overlapping address doesn't * already exist.
Hi, Shuah, Any comment on this series?
Thanks, don
On 4/24/2024 3:44 PM, Dongsheng Zhang wrote:
From: donsheng dongsheng.x.zhang@intel.com
If the host was booted with the "default_hugepagesz=1G" kernel command-line parameter, running the NX hugepage test will fail with error "Invalid argument" at the TEST_ASSERT line in kvm_util.c's __vm_mem_region_delete() function: static void __vm_mem_region_delete(struct kvm_vm *vm, struct userspace_mem_region *region, bool unlink) { int ret; ... ret = munmap(region->mmap_start, region->mmap_size); TEST_ASSERT(!ret, __KVM_SYSCALL_ERROR("munmap()", ret)); ... }
NX hugepage test creates a VM with a data slot of 6M size backed with huge pages. If the default hugetlb page size is set to 1G, calling mmap() with MAP_HUGETLB and a length of 6M will succeed but calling its matching munmap() will fail. Documentation/admin-guide/mm/hugetlbpage.rst specifies this behavior:
"Syscalls that operate on memory backed by hugetlb pages only have their lengths aligned to the native page size of the processor; they will normally fail with errno set to EINVAL or exclude hugetlb pages that extend beyond the length if not hugepage aligned. For example, munmap(2) will fail if memory is backed by a hugetlb page and the length is smaller than the hugepage size."
Explicitly use MAP_HUGE_2MB in conjunction with MAP_HUGETLB to fix the issue.
Signed-off-by: donsheng dongsheng.x.zhang@intel.com Suggested-by: Zide Chen zide.chen@intel.com
tools/testing/selftests/kvm/x86_64/nx_huge_pages_test.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/tools/testing/selftests/kvm/x86_64/nx_huge_pages_test.c b/tools/testing/selftests/kvm/x86_64/nx_huge_pages_test.c index 17bbb96fc4df..146e9033e206 100644 --- a/tools/testing/selftests/kvm/x86_64/nx_huge_pages_test.c +++ b/tools/testing/selftests/kvm/x86_64/nx_huge_pages_test.c @@ -129,7 +129,7 @@ void run_test(int reclaim_period_ms, bool disable_nx_huge_pages, vcpu = vm_vcpu_add(vm, 0, guest_code);
- vm_userspace_mem_region_add(vm, VM_MEM_SRC_ANONYMOUS_HUGETLB,
- vm_userspace_mem_region_add(vm, VM_MEM_SRC_ANONYMOUS_HUGETLB_2MB, HPAGE_GPA, HPAGE_SLOT, HPAGE_SLOT_NPAGES, 0);
linux-kselftest-mirror@lists.linaro.org