Re: [PATCH v6 25/43] KVM: selftests: Add support for mmap() on guest_memfd in core library
From: Ackerley Tng <hidden>
Date: 2026-05-22 23:02:38
Also in:
kvm, linux-coco, linux-doc, linux-kselftest, linux-mm, lkml
Ackerley Tng via B4 Relay [off-list ref] writes:
quoted hunk ↗ jump to hunk
[...snip...]@@ -1078,13 +1077,17 @@ void vm_mem_add(struct kvm_vm *vm, enum vm_mem_backing_src_type src_type, } region->fd = -1; - if (backing_src_is_shared(src_type)) + if (flags & KVM_MEM_GUEST_MEMFD && gmem_flags & GUEST_MEMFD_FLAG_MMAP) { + region->fd = kvm_dup(gmem_fd); + mmap_offset = gmem_offset; + } else if (backing_src_is_shared(src_type)) { region->fd = kvm_memfd_alloc(region->mmap_size, src_type == VM_MEM_SRC_SHARED_HUGETLB); + } - region->mmap_start = kvm_mmap(region->mmap_size, PROT_READ | PROT_WRITE, - vm_mem_backing_src_alias(src_type)->flag, - region->fd); + region->mmap_start = __kvm_mmap(region->mmap_size, PROT_READ | PROT_WRITE, + vm_mem_backing_src_alias(src_type)->flag, + region->fd, mmap_offset);
Sashiko pointed out these: 1. When mmap() is done for region->mmap_alias, it doesn't use mmap_offset. I'll fix that in the next revision. 2. mmap() may map past the end of the guest_memfd if, due to alignment, the mmap_size is increased. That is true, but I feel that that fix should go with a bigger clean up for vm_mem_add(). 3. vm_mem_backing_src_alias(src_type)->flag may contain incompatible mmap flags. This is true. For now, when guest_memfd is used with vm_mem_add, the src_type passed has to be VM_MEM_SRC_SHMEM. I think this also falls in the category of doing a bigger clean up for vm_mem_add().
[...snip...]