guest_memfd就地转换 · v10

guest_memfd populate 路径改用 kvm_mem_is_private() 替代手写等价判断，简化当前逻辑并避免日后引入 range-based gmem 查找 API；无功能变化。

为 guest_memfd 引入 per-inode 的 maple tree 属性跟踪，支持 private/shared 原位转换；用 filemap invalidation lock 保护，避免新增锁带来的死锁复杂度。

将 KVM_GENERIC_MEMORY_ATTRIBUTES 改名为 KVM_VM_MEMORY_ATTRIBUTES，明确其表示 per-VM 属性，为后续 guest_memfd 支持内存属性做准备；纯改名无功能变化。

仅当架构定义 kvm_arch_has_private_mem() 时才上报 KVM_MEMORY_ATTRIBUTE_PRIVATE，避免 guest_memfd 属性通用化后 arm64 等系统级误报；与 CONFIG_KVM_VM_MEMORY_ATTRIBUTES 解耦。

重命名内存属性 API，加入 vm_ 前缀以便与后续 guest_memfd 属性区分；kvm_vm_set_mem_attributes() 改为 kvm_set_vm_mem_attributes()，trace 函数同步更新；无功能变化。

新增 kvm_mem_is_private() 静态调用接口，运行时选择 guest_memfd 或 VM 属性实现；原基于 VM 属性的检查改名 kvm_vm_mem_is_private()，为后续实现选择做准备。

加入 gmem_in_place_conversion 全局开关（后续暴露为模块参数），启用时把 __kvm_mem_is_private 静态调用接到 guest_memfd 版本，使 gmem 成为 private/shared 唯一判定源。

整理 kvm_host.h 中 kvm_arch_has_private_mem 桩和 guest_memfd 声明，复用现有 #ifdef 使相关代码聚集；无功能变化。

guest_memfd 失效时同时设置 shared/private 过滤位，因为转换后同一区间可能混合两种页，只按初始状态过滤会漏；让 zapping 逻辑自行遍历属性，简化处理。

新增 KVM_SET_MEMORY_ATTRIBUTES2 ioctl，KVM 对 payload 改为可写以支持部分/增量转换，出错时记录失败偏移；新结构带 padding，并引入 KVM_CAP_GUEST_MEMFD_MEMORY_ATTRIBUTES 通告。

On Fri, Aug 07, 2026 at 02:52:50PM -0700, Ackerley Tng via B4 Relay wrote: > From: Ackerley Tng ackerleytng@google.com > > When converting memory to private in guest_memfd, it is necessary to ensur

private 转 shared 时遍历受影响 folio，调用 kvm_arch_gmem_make_shared() 清理平台私有状态/加密元数据；在 MMU 失效开始后调用以阻止并发页故障；不实现 make-private，因 SNP 偏好 fault 时再 private。

新增区间属性检查函数，若目标区间所有页已具有请求属性则提前返回，避免多余转换和 unmap 操作；优化 KVM_SET_MEMORY_ATTRIBUTES2 路径。

将 LRU cache 冲刷逻辑抽取为 lru_cache_drain_for_folio()，供 guest_memfd 转换前移除 per-CPU LRU cache 中的额外 folio 引用；保留两次预期 refcount 检查以降低并发漏检风险。

转换安全检查中处理 lru_add fbatch 导致的额外 refcount：先排空 per-CPU fbatch 再判断是否仅剩 guest_memfd 引用；若已 DMA pinned 或仍映射则提前失败；引入 drain 状态避免重复排空。

将 folio 初始化/清零逻辑移入 __kvm_gmem_get_pfn()，使 gmem_populate 获取页面时也清零，避免未初始化主机内存泄漏给 guest；用 uptodate 保证页生命周期内最多清零一次。

KVM_SEV_SNP_LAUNCH_UPDATE 的 uaddr 在启用 in-place 转换时可为 NULL，表示直接从已写好的 guest_memfd 页原位填充；per-VM 属性模式仍强制非 NULL。

TDX 的 init_mem_region 允许无源页：启用 gmem_in_place_conversion 时，tdx_gmem_post_populate 直接对目标 PFN 执行 TDH.MEM.PAGE.ADD 原位填充；否则拒绝，避免支持已废弃路径或隐式零页优化。

将 KVM_VM_MEMORY_ATTRIBUTES Kconfig 移到 x86，明确 per-VM 属性仅限 x86/CoCo，阻止其他架构使用已弃用的 per-VM 追踪；公共代码保留便于回退。

将 gmem_in_place_conversion 暴露为模块参数（仅 CONFIG_KVM_VM_MEMORY_ATTRIBUTES=y 时），并让该 Kconfig 可被选择；vm_memory_attributes 参数设为只读防不一致，kvm_arch_has_private_mem 改由平台配置决定。

From: Sean Christopherson seanjc@google.com Now that guest_memfd supports tracking private vs. shared within gmem itself, allow userspace to specify INIT_SHARED on a guest_memfd instance for x86 Co

selftests 添加 memslot 时先创建/dup guest_memfd，再创建普通 backing file，为将来 guest_memfd 支持 mmap 后让同一 fd 同时作为普通 fd 做准备；调整顺序无功能影响。

From: Sean Christopherson seanjc@google.com Rename local variables and function parameters for the guest memory file descriptor and its offset to use a “gmem_” prefix instead of “guest_memfd_”. No

From: Sean Christopherson seanjc@google.com Accept gmem_flags in vm_mem_add() to be able to create a guest_memfd within vm_mem_add(). When vm_mem_add() is used to set up a guest_memfd for a memslo

From: Sean Christopherson seanjc@google.com Add a global variable, kvm_has_gmem_attributes, to make the result of checking for KVM_CAP_GUEST_MEMFD_MEMORY_ATTRIBUTES available to all tests. kvm_has

From: Sean Christopherson seanjc@google.com Add helper functions to kvm_util.h to support calling ioctls, specifically KVM_SET_MEMORY_ATTRIBUTES2, on a guest_memfd file descriptor. Introduce gmem_

新增 x86 selftest 验证 guest_memfd 单页 private→shared→private 基础转换流程，并搭建可扩展的转换测试基础设施；使用 KVM_X86_SW_PROTECTED_VM 作为测试载体。

From: Ackerley Tng ackerleytng@google.com Add a test case to verify that conversions between private and shared memory work correctly when the memory is initially created as shared. Co-developed-b

From: Ackerley Tng ackerleytng@google.com The existing guest_memfd conversion tests only use single-page memory regions. This provides no coverage for multi-page guest_memfd objects, specifically w

From: Ackerley Tng ackerleytng@google.com Add two test cases to the guest_memfd conversions selftest to cover the scenario where a conversion is requested before any memory has been allocated in th

From: Ackerley Tng ackerleytng@google.com Add a guest_memfd selftest to verify that memory conversions work correctly with allocated folios in different layouts. By iterating through which pages a

From: Ackerley Tng ackerleytng@google.com Add a test to verify that deallocating a page in a guest memfd region via fallocate() with FALLOC_FL_PUNCH_HOLE does not alter the shared or private status

新增跨进程一致性测试：fork 子进程后两进程交替访问共享区间，验证 guest_memfd 的 shared/private 状态在所有进程一致，且某进程映射的 shared 页在转换后从所有进程 unmap。

From: Ackerley Tng ackerleytng@google.com Add helper functions to allow KVM selftests to pin memory using CONFIG_GUP_TEST. This is useful for testing scenarios where some page has an increased refc

From: Ackerley Tng ackerleytng@google.com Add a selftest to verify that converting a shared guest_memfd page to a private page fails if the page has an elevated reference count. When KVM converts

From: Ackerley Tng ackerleytng@google.com private_mem_conversions_test used to reset the shared memory that was used for the test to an initial pattern at the end of each test iteration. Then, it w

From: Ackerley Tng ackerleytng@google.com Introduce a new helper, kvm_gpa_to_guest_memfd(), to find the guest_memfd-related details of a memory region that contains a given guest physical address (

From: Sean Christopherson seanjc@google.com Introduce vm_mem_set_memory_attributes(), which handles setting of memory attributes for a range of guest physical addresses, regardless of whether the a

From: Ackerley Tng ackerleytng@google.com The TEST_EXPECT_SIGBUS macro is not thread-safe as it uses a global sigjmp_buf and installs a global SIGBUS signal handler. If multiple threads execute the

From: Ackerley Tng ackerleytng@google.com Update the private memory conversions selftest to also test conversions that are done “in-place” via per-guest_memfd memory attributes. In-place conversion

From: Sean Christopherson seanjc@google.com Skip setting memory to private in the private memory exits test when using per-gmem memory attributes, as memory is initialized to private by default for