| Message ID | 20240103084327.19955-1-yan.y.zhao@intel.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | KVM: allow mapping of compound tail pages for IO or PFNMAP mapping | expand |
On Wed, Jan 03, 2024, Yan Zhao wrote: > This is a v2 for previous series [1] to allow mapping for compound tail > pages for IO or PFNMAP mapping. > > Compared to v1, this version provides selftest to check functionality in > KVM to map memslots for MMIO BARs (VMAs with flag VM_IO | VM_PFNMAP), as > requested by Sean in [1]. Doh. So I didn't intend for you to have to create a mock device just to be able to run a selftest. I assumed it would be easy-ish to utilize an existing generic device. I take it that's not the case?
On Mon, Feb 12, 2024 at 07:07:25PM -0800, Sean Christopherson wrote: > On Wed, Jan 03, 2024, Yan Zhao wrote: > > This is a v2 for previous series [1] to allow mapping for compound tail > > pages for IO or PFNMAP mapping. > > > > Compared to v1, this version provides selftest to check functionality in > > KVM to map memslots for MMIO BARs (VMAs with flag VM_IO | VM_PFNMAP), as > > requested by Sean in [1]. > > Doh. So I didn't intend for you to have to create a mock device just to be able > to run a selftest. I assumed it would be easy-ish to utilize an existing generic > device. I take it that's not the case? The selftest requires a vma with flag (VM_IO | VM_PFNMAP) with non-reserved pages as backend. Without a mock device, I don't find a easy way to let the selftest take effect. So, I borrowed the way in "tools/testing/selftests/mm/hmm-tests.c" which uses a mock driver in "lib/test_hmm.c".
This is a v2 for previous series [1] to allow mapping for compound tail pages for IO or PFNMAP mapping. Compared to v1, this version provides selftest to check functionality in KVM to map memslots for MMIO BARs (VMAs with flag VM_IO | VM_PFNMAP), as requested by Sean in [1]. The selftest can also be used to test series "allow mapping non-refcounted pages" [2]. Tag RFC is added because a test driver is introduced in patch 2, which is new to KVM selftest, and test "set_memory_region_io" in patch 3 depends on that the test driver is compiled and loaded in kernel. Besides, patch 3 calls vm_set_user_memory_region() directly without modifying vm_mem_add(). So, this series is sent to ensure the main direction is right. Thanks Yan [1] https://lore.kernel.org/all/20230719083332.4584-1-yan.y.zhao@intel.com/ [2] https://lore.kernel.org/all/20230911021637.1941096-1-stevensd@google.com/ v2: added patch 2 and 3 to do selftest for patch 1 (Sean). Yan Zhao (3): KVM: allow mapping of compound tail pages for IO or PFNMAP mapping KVM: selftests: add selftest driver for KVM to test memory slots for MMIO BARs KVM: selftests: Add set_memory_region_io to test memslots for MMIO BARs lib/Kconfig.debug | 14 + lib/Makefile | 1 + lib/test_kvm_mock_device.c | 281 ++++++++++++++++++ lib/test_kvm_mock_device_uapi.h | 16 + tools/testing/selftests/kvm/Makefile | 1 + .../selftests/kvm/set_memory_region_io.c | 188 ++++++++++++ virt/kvm/kvm_main.c | 2 +- 7 files changed, 502 insertions(+), 1 deletion(-) create mode 100644 lib/test_kvm_mock_device.c create mode 100644 lib/test_kvm_mock_device_uapi.h create mode 100644 tools/testing/selftests/kvm/set_memory_region_io.c base-commit: 8ed26ab8d59111c2f7b86d200d1eb97d2a458fd1