@samitouri / QOSAMI-WSL-2 / commits / 17852001

Serialize move/resize/set-sparse against in-progress conversions (#41170)

* Serialize move/resize/set-sparse against in-progress conversions MoveDistribution, SetSparse, and ResizeDistribution mutate a distribution's VHD but never checked m_lockedDistributions. A conversion (--set-version) or export releases m_instanceLock while running yet keeps the distribution in m_lockedDistributions, so these operations could start mid-conversion and race on the same VHD (e.g. a corrupt export, or the VHD moved/resized out from under a conversion). Add _EnsureNotLocked under m_instanceLock to each so they fail with E_ILLEGAL_STATE_CHANGE while a conversion/export/compaction is in progress, matching how the conversion paths already guard each other via _ConversionBegin. Add a regression test that holds a distribution in the Exporting state (via a blocked export pipe) and verifies --resize, --set-sparse, and --move are all rejected. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: 37ad1fd7-a452-4056-a8cc-e91cf7870a60 * Generalize lock-guard comments to conversion/export Compaction isn't part of this branch, so drop it from the comments. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: 37ad1fd7-a452-4056-a8cc-e91cf7870a60 * Address review: deterministic pipe block and accurate cleanup comment - Use an explicit 1-byte pipe buffer so the export blocks immediately regardless of test distro size, instead of relying on the default buffer filling. - Fix the cleanup comment: closing the read end unblocks the export via a broken pipe, it does not drain it to completion. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: 37ad1fd7-a452-4056-a8cc-e91cf7870a60 --------- Co-authored-by: Ben Hillis <benhill@ntdev.microsoft.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: 37ad1fd7-a452-4056-a8cc-e91cf7870a60

Ben Hillis committed Jul 27, 2026 at 13:24 UTC 178520010b3e88a89fe526727cea96599f78112c

2 files changed +68

src/windows/service/exe/LxssUserSession.cpp

+14

@@ -914,6 +914,11 @@ HRESULT LxssUserSessionImpl::MoveDistribution(_In_ LPCGUID DistroGuid, _In_ LPCW

914 // Fail if the distribution is running.

915 RETURN_HR_IF(WSL_E_DISTRO_NOT_STOPPED, m_runningInstances.contains(*DistroGuid));

916

917 + // Fail if a conversion or export is in progress for this distribution. Those operations release

918 + // m_instanceLock while running but keep the distribution in m_lockedDistributions, so mutating

919 + // the VHD here would race with them.

920 + _EnsureNotLocked(DistroGuid);

921 +

922 // Lookup the distribution configuration

923 const auto lxssKey = s_OpenLxssUserKey();

924 _ValidateDistributionNameAndPathNotInUse(lxssKey.get(), Location, nullptr);

@@ -1774,6 +1779,10 @@ try

1779 // Don't attempt if running

1780 RETURN_HR_IF(WSL_E_DISTRO_NOT_STOPPED, m_runningInstances.contains(*DistroGuid));

1781

1782 + // Don't attempt while a conversion or export holds this distribution; those operations release

1783 + // m_instanceLock while running but keep the entry in m_lockedDistributions.

1784 + _EnsureNotLocked(DistroGuid);

1785 +

1786 const wil::unique_hfile vhd{::CreateFileW(configuration.VhdFilePath.c_str(), GENERIC_WRITE, 0, nullptr, OPEN_EXISTING, 0, nullptr)};

1787 if (!vhd)

1788 {

@@ -1803,6 +1812,11 @@ try

1812 const auto configuration = s_GetDistributionConfiguration(registration);

1813 RETURN_HR_IF(WSL_E_WSL2_NEEDED, WI_IsFlagClear(configuration.Flags, LXSS_DISTRO_FLAGS_VM_MODE));

1814

1815 + // Fail if a conversion or export is in progress; those operations release m_instanceLock while

1816 + // running but keep this distribution in m_lockedDistributions, so resizing its VHD now would

1817 + // race with them.

1818 + _EnsureNotLocked(DistroGuid);

1819 +

1820 const auto& vhdPath = configuration.VhdFilePath;

1821 if (m_utilityVm && m_utilityVm->IsVhdAttached(vhdPath.c_str()))

1822 {

test/windows/UnitTests.cpp

+54

@@ -3269,6 +3269,60 @@ Error code: Wsl/InstallDistro/WSL_E_DISTRO_NOT_FOUND

3269 }

3270 }

3271

3272 + // Verifies that VHD-mutating manage operations (--resize, --set-sparse, --move) are rejected while a

3273 + // long-running conversion/export holds the distribution lock, rather than racing with it on the VHD.

3274 + WSL2_TEST_METHOD(ManageRejectedWhileLocked)

3275 + {

3276 + constexpr auto name = L"manage-locked-test-distro";

3277 +

3278 + VERIFY_ARE_EQUAL(LxsstuLaunchWsl(std::format(L"--import {} . \"{}\" --version 2", name, g_testDistroPath)), 0L);

3279 + auto cleanupName =

3280 + wil::scope_exit_log(WI_DIAGNOSTICS_INFO, [name]() { LxsstuLaunchWsl(std::format(L"--unregister {}", name)); });

3281 + WslShutdown();

3282 +

3283 + // Start an export to a pipe we deliberately don't drain. Use a tiny buffer so the export blocks

3284 + // as soon as it writes any data, regardless of the test distro's size, deterministically holding

3285 + // the distribution in the "Exporting" locked state.

3286 + auto [readPipe, writePipe] = CreateSubprocessPipe(false, true, 1);

3287 +

3288 + std::thread exportThread([&]() { LxsstuLaunchWsl(std::format(L"--export {} -", name), nullptr, writePipe.get()); });

3289 +

3290 + auto joinExport = wil::scope_exit_log(WI_DIAGNOSTICS_INFO, [&]() {

3291 + // Close the read end so the blocked export fails with a broken pipe and returns, then join.

3292 + readPipe.reset();

3293 + if (exportThread.joinable())

3294 + {

3295 + exportThread.join();

3296 + }

3297 + });

3298 +

3299 + // Wait until the service reports the distribution as Exporting (i.e. the lock is held).

3300 + bool locked = false;

3301 + for (int i = 0; i < 100 && !locked; ++i)

3302 + {

3303 + auto [out, _] = LxsstuLaunchWslAndCaptureOutput(L"--list --verbose");

3304 + locked = (out.find(name) != std::wstring::npos) && (out.find(L"Exporting") != std::wstring::npos);

3305 + if (!locked)

3306 + {

3307 + std::this_thread::sleep_for(std::chrono::milliseconds(100));

3308 + }

3309 + }

3310 +

3311 + VERIFY_IS_TRUE(locked);

3312 +

3313 + // Each VHD-mutating manage operation must be rejected with E_ILLEGAL_STATE_CHANGE while the lock is held.

3314 + auto verifyRejected = [&](const std::wstring& command) {

3315 + auto [out, _] = LxsstuLaunchWslAndCaptureOutput(command, -1);

3316 + VERIFY_IS_TRUE(out.find(L"E_ILLEGAL_STATE_CHANGE") != std::wstring::npos);

3317 + };

3318 +

3319 + verifyRejected(std::format(L"--manage {} --resize 2GB", name));

3320 + verifyRejected(std::format(L"--manage {} --set-sparse false", name));

3321 +

3322 + const auto moveTarget = std::filesystem::absolute(L"manage-locked-move-target").wstring();

3323 + verifyRejected(std::format(L"--manage {} --move \"{}\"", name, moveTarget));

3324 + }

3325 +

3326 WSL2_TEST_METHOD(FileOffsets)

3327 {

3328 auto cleanup = wil::scope_exit_log(WI_DIAGNOSTICS_INFO, []() { DeleteFile(L"output.txt"); });