January 2022 - Wine-Devel - list.winehq.org

[tools] testbot/SetWinLocale: Explicitly set ResetAllSettings=true for <gs:Locale>.
by Francois Gouget Jan. 25, 2022

Jan. 25, 2022

The ResetAllSettings default value is probably true but that is not documented. Signed-off-by: Francois Gouget <fgouget(a)codeweavers.com> --- testbot/bin/SetWinLocale | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/testbot/bin/SetWinLocale b/testbot/bin/SetWinLocale index 83cb581883..911aa452d9 100755 --- a/testbot/bin/SetWinLocale +++ b/testbot/bin/SetWinLocale @@ -768,7 +768,8 @@ if ($KeyboardIds) if ($OptLocale) { push @Config, " <gs:UserLocale>", - " <gs:Locale Name='$OptLocale' SetAsCurrent='true'/>", + " <gs:Locale Name='$OptLocale' SetAsCurrent='true' ResetAllSettings='true'>", + " </gs:Locale>", " </gs:UserLocale>"; } push @Config, "</gs:GlobalizationServices>"; -- 2.30.2

1 0

[PATCH] ntdll: use the correct amd64 trap codes on *BSD
by Damjan Jovanovic Jan. 25, 2022

Jan. 25, 2022

Fix the *BSD trap codes on amd64 like 148120f21ecf7ad30aaa2136f8d9f00f2f241231 did on i386. Signed-off-by: Damjan Jovanovic <damjan.jov(a)gmail.com> --- dlls/ntdll/unix/signal_x86_64.c | 24 ++++++++++++++++++++++++ 1 file changed, 24 insertions(+)

1 0

[PATCH] WBEMProxy: Implemented Complete SoftwareLicensingProduct row and data
by Ben Shefte Jan. 25, 2022

Jan. 25, 2022

Signed-off-by: Ben Shefte<shefben(a)gmail.com> --- dlls/wbemprox/builtin.c | 98 ++++++++++++++++++++++++++++++++++++++--- 1 file changed, 92 insertions(+), 6 deletions(-) diff --git a/dlls/wbemprox/builtin.c b/dlls/wbemprox/builtin.c index 5a09d64b8fb..4122afee59d 100644 --- a/dlls/wbemprox/builtin.c +++ b/dlls/wbemprox/builtin.c @@ -388,9 +388,51 @@ static const struct column col_sid[] = { L"SidLength", CIM_UINT32 }, }; static const struct column col_softwarelicensingproduct[] = -{ - { L"LicenseIsAddon", CIM_BOOLEAN }, - { L"LicenseStatus", CIM_UINT32 }, + { + { L"ID", CIM_STRING | COL_FLAG_KEY }, + { L"Name", CIM_STRING }, + { L"Description", CIM_STRING }, + { L"ApplicationID", CIM_STRING }, + { L"ProcessorURL", CIM_STRING }, + { L"MachineURL", CIM_STRING }, + { L"ProductKeyURL", CIM_STRING }, + { L"UseLicenseURL", CIM_STRING }, + { L"LicenseStatus", CIM_UINT32 }, + { L"LicenseStatusReason", CIM_UINT32 }, + { L"GracePeriodRemaining", CIM_UINT32 }, + { L"EvaluationEndDate", CIM_DATETIME }, + { L"OfflineInstallationId", CIM_STRING }, + { L"PartialProductKey", CIM_STRING }, + { L"ProductKeyID", CIM_STRING }, + { L"LicenseFamily", CIM_STRING }, + { L"LicenseDependsOn", CIM_STRING }, + { L"LicenseIsAddon", CIM_BOOLEAN}, + { L"VLActivationInterval", CIM_UINT32 }, + { L"VLRenewalInterval", CIM_UINT32 }, + { L"KeyManagementServiceProductKeyID", CIM_STRING }, + { L"KeyManagementServiceMachine", CIM_STRING }, + { L"KeyManagementServicePort", CIM_UINT32 }, + { L"DiscoveredKeyManagementServiceMachineName", CIM_STRING }, + { L"DiscoveredKeyManagementServiceMachinePort", CIM_UINT32 }, + { L"IsKeyManagementServiceMachine", CIM_UINT32 }, + { L"KeyManagementServiceCurrentCount", CIM_UINT32 }, + { L"RequiredClientCount", CIM_UINT32 }, + { L"KeyManagementServiceUnlicensedRequests", CIM_UINT32 }, + { L"KeyManagementServiceLicensedRequests", CIM_UINT32 }, + { L"KeyManagementServiceOOBGraceRequests", CIM_UINT32 }, + { L"KeyManagementServiceOOTGraceRequests", CIM_UINT32 }, + { L"KeyManagementServiceNonGenuineGraceRequests", CIM_UINT32 }, + { L"KeyManagementServiceTotalRequests", CIM_UINT32 }, + { L"KeyManagementServiceFailedRequests", CIM_UINT32 }, + { L"KeyManagementServiceNotificationRequests", CIM_UINT32 }, + { L"GenuineStatus", CIM_UINT32 }, + { L"ExtendedGrace", CIM_UINT32 }, + { L"TokenActivationILID", CIM_STRING }, + { L"TokenActivationILVID", CIM_UINT32 }, + { L"TokenActivationGrantNumber", CIM_UINT32 }, + { L"TokenActivationCertificateThumbprint", CIM_STRING }, + { L"TokenActivationAdditionalInfo", CIM_STRING }, + { L"TrustedTime", CIM_DATETIME }, }; static const struct column col_sounddevice[] = { @@ -816,8 +858,50 @@ struct record_sid }; struct record_softwarelicensingproduct { - int license_is_addon; - UINT32 license_status; + const WCHAR *id; + const WCHAR *name; + const WCHAR *description; + const WCHAR *appid; + const WCHAR *processorurl; + const WCHAR *machineurl; + const WCHAR *productkeyurl; + const WCHAR *useLicenseurl; + UINT32 licensestatus; + UINT32 licensestatusreason; + UINT32 graceperiodremaining; + const WCHAR *evaluationenddate; + const WCHAR *offlineinstallationid; + const WCHAR *partialproductkey; + const WCHAR *productkeyid; + const WCHAR *licensefamily; + const WCHAR *licensedependson; + BOOLEAN licenseisaddon; + UINT32 vlactivationinterval; + UINT32 vlrenewalinterval; + const WCHAR *keymanagementserviceproductkeyid; + const WCHAR *keymanagementservicemachine; + UINT32 keymanagementserviceport; + const WCHAR *discoveredkeymanagementservicemachinename; + UINT32 discoveredkeymanagementservicemachineport; + UINT32 isKeymanagementservicemachine; + UINT32 keymanagementservicecurrentcount; + UINT32 requiredclientcount; + UINT32 keymanagementserviceunlicensedrequests; + UINT32 keymanagementservicelicensedrequests; + UINT32 keymanagementserviceoobgracerequests; + UINT32 keymanagementserviceootgracerequests; + UINT32 keymanagementservicenongenuinegracerequests; + UINT32 keymanagementservicetotalrequests; + UINT32 keymanagementservicefailedrequests; + UINT32 keymanagementservicenotificationrequests; + UINT32 genuinestatus; + UINT32 extendedgrace; + const WCHAR *tokenactivationilid; + UINT32 tokenactivationilvid; + UINT32 tokenactivationgrantnumber; + const WCHAR *tokenactivationcertificatethumbprint; + const WCHAR *tokenactivationadditionalinfo; + const WCHAR *trustedtime; }; struct record_sounddevice { @@ -990,7 +1074,9 @@ static const struct record_quickfixengineering data_quickfixengineering[] = static const struct record_softwarelicensingproduct data_softwarelicensingproduct[] = { - { 0, 1 }, + { L"0", L"Nonexistent", L"first entry into table", L"0", L"localhost", L"localhost", L"localhost", + L"localhost", 1, 1, 0, L"1-1-1991", L"0", L"inval", L"0", L"0", NULL, FALSE, 0, 0, L"0", L"localhost", + 80, L"localhost", 80, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, L"noid", 0, 0, L"none", L"none", L"1-1-1991"}, }; static const struct record_stdregprov data_stdregprov[] = -- 2.27.0

2 1

[PATCH v3 resend 1/2] ntdll/tests: Add more tests for \Device\NamedPipe\ root directory.
by Jinoh Kang Jan. 25, 2022

Jan. 25, 2022

Signed-off-by: Jinoh Kang <jinoh.kang.kr(a)gmail.com> --- Notes: v2 -> v3: move some tests from om.c to pipe.c dlls/ntdll/tests/om.c | 9 ++++ dlls/ntdll/tests/pipe.c | 93 ++++++++++++++++++++++++++++++++++++++++- 2 files changed, 101 insertions(+), 1 deletion(-) diff --git a/dlls/ntdll/tests/om.c b/dlls/ntdll/tests/om.c index f872d6709a1..04acf5b2241 100644 --- a/dlls/ntdll/tests/om.c +++ b/dlls/ntdll/tests/om.c @@ -1871,6 +1871,15 @@ static void test_query_object(void) pNtClose( handle ); + handle = CreateFileA( "\\\\.\\pipe\\", 0, 0, NULL, OPEN_EXISTING, 0, 0 ); + ok( handle != INVALID_HANDLE_VALUE, "CreateFile failed (%d)\n", GetLastError() ); + + test_object_name( handle, L"\\Device\\NamedPipe\\", TRUE ); + test_object_type( handle, L"File" ); + test_file_info( handle ); + + pNtClose( handle ); + RtlInitUnicodeString( &path, L"\\REGISTRY\\Machine" ); status = pNtCreateKey( &handle, KEY_READ, &attr, 0, 0, 0, 0 ); ok( status == STATUS_SUCCESS, "NtCreateKey failed status %x\n", status ); diff --git a/dlls/ntdll/tests/pipe.c b/dlls/ntdll/tests/pipe.c index 6940e0020b7..f5ca6077605 100644 --- a/dlls/ntdll/tests/pipe.c +++ b/dlls/ntdll/tests/pipe.c @@ -51,6 +51,13 @@ typedef struct { ULONG NamedPipeEnd; } FILE_PIPE_LOCAL_INFORMATION; +typedef struct _FILE_PIPE_WAIT_FOR_BUFFER { + LARGE_INTEGER Timeout; + ULONG NameLength; + BOOLEAN TimeoutSpecified; + WCHAR Name[1]; +} FILE_PIPE_WAIT_FOR_BUFFER, *PFILE_PIPE_WAIT_FOR_BUFFER; + #ifndef FILE_SYNCHRONOUS_IO_ALERT #define FILE_SYNCHRONOUS_IO_ALERT 0x10 #endif @@ -62,6 +69,10 @@ typedef struct { #ifndef FSCTL_PIPE_LISTEN #define FSCTL_PIPE_LISTEN CTL_CODE(FILE_DEVICE_NAMED_PIPE, 2, METHOD_BUFFERED, FILE_ANY_ACCESS) #endif + +#ifndef FSCTL_PIPE_WAIT +#define FSCTL_PIPE_WAIT CTL_CODE(FILE_DEVICE_NAMED_PIPE, 6, METHOD_BUFFERED, FILE_ANY_ACCESS) +#endif #endif static NTSTATUS (WINAPI *pNtFsControlFile) (HANDLE handle, HANDLE event, PIO_APC_ROUTINE apc, PVOID apc_context, PIO_STATUS_BLOCK io, ULONG code, PVOID in_buffer, ULONG in_size, PVOID out_buffer, ULONG out_size); @@ -177,6 +188,51 @@ static NTSTATUS listen_pipe(HANDLE hPipe, HANDLE hEvent, PIO_STATUS_BLOCK iosb, return pNtFsControlFile(hPipe, hEvent, use_apc ? &ioapc: NULL, use_apc ? &dummy: NULL, iosb, FSCTL_PIPE_LISTEN, 0, 0, 0, 0); } +static NTSTATUS wait_pipe(HANDLE handle, PUNICODE_STRING name, const LARGE_INTEGER* timeout) +{ + HANDLE event; + NTSTATUS status; + OBJECT_ATTRIBUTES attr; + IO_STATUS_BLOCK iosb; + FILE_PIPE_WAIT_FOR_BUFFER *pipe_wait; + ULONG pipe_wait_size; + + pipe_wait_size = offsetof(FILE_PIPE_WAIT_FOR_BUFFER, Name[0]) + name->Length; + pipe_wait = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, pipe_wait_size); + if (!pipe_wait) return STATUS_NO_MEMORY; + + pipe_wait->TimeoutSpecified = !!timeout; + pipe_wait->NameLength = name->Length; + if (timeout) pipe_wait->Timeout = *timeout; + memcpy(pipe_wait->Name, name->Buffer, name->Length); + + InitializeObjectAttributes(&attr, NULL, 0, 0, NULL); + status = NtCreateEvent(&event, GENERIC_ALL, &attr, NotificationEvent, FALSE); + if (status != STATUS_SUCCESS) + { + ok(0, "NtCreateEvent failure: %08x\n", status); + HeapFree(GetProcessHeap(), 0, pipe_wait); + return status; + } + + memset(&iosb, 0, sizeof(iosb)); + iosb.Status = STATUS_PENDING; + status = pNtFsControlFile(handle, event, NULL, NULL, &iosb, FSCTL_PIPE_WAIT, + pipe_wait, pipe_wait_size, NULL, 0); + if (status == STATUS_PENDING && iosb.Status == STATUS_PENDING) + { + WaitForSingleObject(event, INFINITE); + } + if (SUCCEEDED(status) && iosb.Status != STATUS_PENDING) + { + status = iosb.Status; + } + + NtClose(event); + HeapFree(GetProcessHeap(), 0, pipe_wait); + return status; +} + static void test_create_invalid(void) { IO_STATUS_BLOCK iosb; @@ -2420,6 +2476,7 @@ static void test_security_info(void) static void test_empty_name(void) { + static const LARGE_INTEGER zero_timeout = {{ 0 }}; HANDLE hdirectory, hpipe, hpipe2, hwrite, hwrite2, handle; OBJECT_TYPE_INFORMATION *type_info; OBJECT_NAME_INFORMATION *name_info; @@ -2450,6 +2507,10 @@ static void test_empty_name(void) FILE_SHARE_READ|FILE_SHARE_WRITE, FILE_OPEN, 0, NULL, 0 ); ok(!status, "Got unexpected status %#x.\n", status); + pRtlInitUnicodeString(&name, L"nonexistent_pipe"); + status = wait_pipe(hdirectory, &name, &zero_timeout); + todo_wine ok(status == STATUS_ILLEGAL_FUNCTION, "unexpected status for FSCTL_PIPE_WAIT on \\Device\\NamedPipe: %08x\n", status); + name.Buffer = NULL; name.Length = 0; name.MaximumLength = 0; @@ -2462,6 +2523,17 @@ static void test_empty_name(void) todo_wine ok(status == STATUS_OBJECT_NAME_INVALID, "Got unexpected status %#x.\n", status); if (!status) CloseHandle(hpipe); + + pRtlInitUnicodeString(&name, L"test3\\pipe"); + attr.RootDirectory = hdirectory; + attr.ObjectName = &name; + timeout.QuadPart = -(LONG64)10000000; + status = pNtCreateNamedPipeFile(&hpipe, GENERIC_READ|GENERIC_WRITE, &attr, &io, FILE_SHARE_READ|FILE_SHARE_WRITE, + FILE_CREATE, FILE_PIPE_FULL_DUPLEX, 0, 0, 0, 1, 256, 256, &timeout); + ok(status == STATUS_OBJECT_NAME_INVALID, "unexpected status from NtCreateNamedPipeFile: %08x\n", status); + if (!status) + CloseHandle(hpipe); + CloseHandle(hdirectory); pRtlInitUnicodeString(&name, L"\\Device\\NamedPipe\\"); @@ -2475,6 +2547,10 @@ static void test_empty_name(void) FILE_SHARE_READ|FILE_SHARE_WRITE, FILE_OPEN, 0, NULL, 0 ); ok(!status, "Got unexpected status %#x.\n", status); + pRtlInitUnicodeString(&name, L"nonexistent_pipe"); + status = wait_pipe(hdirectory, &name, &zero_timeout); + ok(status == STATUS_OBJECT_NAME_NOT_FOUND, "unexpected status for FSCTL_PIPE_WAIT on \\Device\\NamedPipe\\: %08x\n", status); + name.Buffer = NULL; name.Length = 0; name.MaximumLength = 0; @@ -2565,9 +2641,24 @@ static void test_empty_name(void) CloseHandle(hwrite); CloseHandle(hpipe); - CloseHandle(hdirectory); CloseHandle(hpipe2); CloseHandle(hwrite2); + + pRtlInitUnicodeString(&name, L"test3\\pipe"); + attr.RootDirectory = hdirectory; + attr.ObjectName = &name; + timeout.QuadPart = -(LONG64)10000000; + status = pNtCreateNamedPipeFile(&hpipe, GENERIC_READ|GENERIC_WRITE, &attr, &io, FILE_SHARE_READ|FILE_SHARE_WRITE, + FILE_CREATE, FILE_PIPE_FULL_DUPLEX, 0, 0, 0, 1, 256, 256, &timeout); + todo_wine ok(!status, "unexpected failure from NtCreateNamedPipeFile: %08x\n", status); + + handle = CreateFileA("\\\\.\\pipe\\test3\\pipe", GENERIC_READ, FILE_SHARE_READ|FILE_SHARE_WRITE, NULL, + OPEN_EXISTING, 0, 0 ); + todo_wine ok(handle != INVALID_HANDLE_VALUE, "Failed to open NamedPipe (%u)\n", GetLastError()); + + CloseHandle(handle); + CloseHandle(hpipe); + CloseHandle(hdirectory); } START_TEST(pipe) -- 2.31.1

2 4

[PATCH vkd3d] vkd3d: Use Vulkan timeline semaphores for D3D12 fences.
by Conor McCarthy Jan. 25, 2022

Jan. 25, 2022

If KHR_timeline_semaphore is not available, the old implementation will be used. Based on a vkd3d-proton patch by Hans-Kristian Arntzen. Signed-off-by: Conor McCarthy <cmccarthy(a)codeweavers.com> --- libs/vkd3d/command.c | 479 +++++++++++++++++++++++++++++++++---- libs/vkd3d/device.c | 14 ++ libs/vkd3d/vkd3d_private.h | 11 + libs/vkd3d/vulkan_procs.h | 5 + tests/d3d12.c | 11 +- 5 files changed, 463 insertions(+), 57 deletions(-) diff --git a/libs/vkd3d/command.c b/libs/vkd3d/command.c index 2cf1eba2..3203cb51 100644 --- a/libs/vkd3d/command.c +++ b/libs/vkd3d/command.c @@ -268,6 +268,7 @@ static HRESULT vkd3d_enqueue_gpu_fence(struct vkd3d_fence_worker *worker, } worker->enqueued_fences[worker->enqueued_fence_count].vk_fence = vk_fence; + worker->enqueued_fences[worker->enqueued_fence_count].vk_semaphore = VK_NULL_HANDLE; waiting_fence = &worker->enqueued_fences[worker->enqueued_fence_count].waiting_fence; waiting_fence->fence = fence; waiting_fence->value = value; @@ -317,6 +318,7 @@ static void vkd3d_fence_worker_remove_fence(struct vkd3d_fence_worker *worker, s static void vkd3d_fence_worker_move_enqueued_fences_locked(struct vkd3d_fence_worker *worker) { unsigned int i; + bool timeline; size_t count; bool ret; @@ -325,8 +327,18 @@ static void vkd3d_fence_worker_move_enqueued_fences_locked(struct vkd3d_fence_wo count = worker->fence_count + worker->enqueued_fence_count; - ret = vkd3d_array_reserve((void **)&worker->vk_fences, &worker->vk_fences_size, - count, sizeof(*worker->vk_fences)); + if ((timeline = worker->device->vk_info.KHR_timeline_semaphore)) + { + ret = vkd3d_array_reserve((void **) &worker->vk_semaphores, &worker->vk_semaphores_size, + count, sizeof(*worker->vk_semaphores)); + ret &= vkd3d_array_reserve((void **) &worker->semaphore_wait_values, &worker->semaphore_wait_values_size, + count, sizeof(*worker->semaphore_wait_values)); + } + else + { + ret = vkd3d_array_reserve((void **)&worker->vk_fences, &worker->vk_fences_size, + count, sizeof(*worker->vk_fences)); + } ret &= vkd3d_array_reserve((void **)&worker->fences, &worker->fences_size, count, sizeof(*worker->fences)); if (!ret) @@ -339,7 +351,16 @@ static void vkd3d_fence_worker_move_enqueued_fences_locked(struct vkd3d_fence_wo { struct vkd3d_enqueued_fence *current = &worker->enqueued_fences[i]; - worker->vk_fences[worker->fence_count] = current->vk_fence; + if (timeline) + { + worker->vk_semaphores[worker->fence_count] = current->vk_semaphore; + worker->semaphore_wait_values[worker->fence_count] = current->waiting_fence.value; + } + else + { + worker->vk_fences[worker->fence_count] = current->vk_fence; + } + worker->fences[worker->fence_count] = current->waiting_fence; ++worker->fence_count; } @@ -347,6 +368,66 @@ static void vkd3d_fence_worker_move_enqueued_fences_locked(struct vkd3d_fence_wo worker->enqueued_fence_count = 0; } +static void vkd3d_wait_for_gpu_timeline_semaphores(struct vkd3d_fence_worker *worker) +{ + struct d3d12_device *device = worker->device; + const struct vkd3d_vk_device_procs *vk_procs = &device->vk_procs; + VkSemaphoreWaitInfoKHR wait_info; + VkSemaphore vk_semaphore; + uint64_t counter_value; + unsigned int i, j; + HRESULT hr; + int vr; + + if (!worker->fence_count) + return; + + wait_info.sType = VK_STRUCTURE_TYPE_SEMAPHORE_WAIT_INFO_KHR; + wait_info.pNext = NULL; + wait_info.flags = VK_SEMAPHORE_WAIT_ANY_BIT_KHR; + wait_info.pSemaphores = worker->vk_semaphores; + wait_info.semaphoreCount = worker->fence_count; + wait_info.pValues = worker->semaphore_wait_values; + + vr = VK_CALL(vkWaitSemaphoresKHR(device->vk_device, &wait_info, ~(uint64_t)0)); + if (vr == VK_TIMEOUT) + return; + if (vr != VK_SUCCESS) + { + ERR("Failed to wait for Vulkan timeline semaphores, vr %d.\n", vr); + return; + } + + for (i = 0, j = 0; i < worker->fence_count; ++i) + { + struct vkd3d_waiting_fence *current = &worker->fences[i]; + + vk_semaphore = worker->vk_semaphores[i]; + if ((vr = VK_CALL(vkGetSemaphoreCounterValueKHR(device->vk_device, vk_semaphore, &counter_value))) < 0) + { + ERR("Failed to get Vulkan semaphore status, vr %d.\n", vr); + } + else if (counter_value >= current->value) + { + TRACE("Signaling fence %p value %#"PRIx64".\n", current->fence, current->value); + if (FAILED(hr = d3d12_fence_signal(current->fence, counter_value, VK_NULL_HANDLE))) + ERR("Failed to signal D3D12 fence, hr %#x.\n", hr); + + InterlockedDecrement(&current->fence->pending_worker_operation_count); + continue; + } + + if (i != j) + { + worker->vk_semaphores[j] = worker->vk_semaphores[i]; + worker->semaphore_wait_values[j] = worker->semaphore_wait_values[i]; + worker->fences[j] = worker->fences[i]; + } + ++j; + } + worker->fence_count = j; +} + static void vkd3d_wait_for_gpu_fences(struct vkd3d_fence_worker *worker) { struct d3d12_device *device = worker->device; @@ -402,13 +483,19 @@ static void vkd3d_wait_for_gpu_fences(struct vkd3d_fence_worker *worker) static void *vkd3d_fence_worker_main(void *arg) { struct vkd3d_fence_worker *worker = arg; + bool timeline; int rc; vkd3d_set_thread_name("vkd3d_fence"); + timeline = worker->device->vk_info.KHR_timeline_semaphore; + for (;;) { - vkd3d_wait_for_gpu_fences(worker); + if (timeline) + vkd3d_wait_for_gpu_timeline_semaphores(worker); + else + vkd3d_wait_for_gpu_fences(worker); if (!worker->fence_count || atomic_add_fetch(&worker->enqueued_fence_count, 0)) { @@ -473,6 +560,10 @@ HRESULT vkd3d_fence_worker_start(struct vkd3d_fence_worker *worker, worker->vk_fences_size = 0; worker->fences = NULL; worker->fences_size = 0; + worker->vk_semaphores = NULL; + worker->vk_semaphores_size = 0; + worker->semaphore_wait_values = NULL; + worker->semaphore_wait_values_size = 0; if ((rc = pthread_mutex_init(&worker->mutex, NULL))) { @@ -535,6 +626,8 @@ HRESULT vkd3d_fence_worker_stop(struct vkd3d_fence_worker *worker, vkd3d_free(worker->enqueued_fences); vkd3d_free(worker->vk_fences); vkd3d_free(worker->fences); + vkd3d_free(worker->vk_semaphores); + vkd3d_free(worker->semaphore_wait_values); return S_OK; } @@ -684,6 +777,7 @@ static void d3d12_fence_destroy_vk_objects(struct d3d12_fence *fence) } d3d12_fence_garbage_collect_vk_semaphores_locked(fence, true); + VK_CALL(vkDestroySemaphore(device->vk_device, fence->timeline_semaphore, NULL)); pthread_mutex_unlock(&fence->mutex); } @@ -802,31 +896,21 @@ static HRESULT d3d12_fence_add_vk_semaphore(struct d3d12_fence *fence, return hr; } -static HRESULT d3d12_fence_signal(struct d3d12_fence *fence, uint64_t value, VkFence vk_fence) +static bool d3d12_fence_signal_external_events_locked(struct d3d12_fence *fence) { struct d3d12_device *device = fence->device; - struct vkd3d_signaled_semaphore *current; bool signal_null_event_cond = false; unsigned int i, j; - int rc; - - if ((rc = pthread_mutex_lock(&fence->mutex))) - { - ERR("Failed to lock mutex, error %d.\n", rc); - return hresult_from_errno(rc); - } - - fence->value = value; for (i = 0, j = 0; i < fence->event_count; ++i) { struct vkd3d_waiting_event *current = &fence->events[i]; - if (current->value <= value) + if (current->value <= fence->value) { if (current->event) { - fence->device->signal_event(current->event); + device->signal_event(current->event); } else { @@ -841,9 +925,36 @@ static HRESULT d3d12_fence_signal(struct d3d12_fence *fence, uint64_t value, VkF ++j; } } + fence->event_count = j; - if (signal_null_event_cond) + return signal_null_event_cond; +} + +static HRESULT d3d12_fence_signal(struct d3d12_fence *fence, uint64_t value, VkFence vk_fence) +{ + struct d3d12_device *device = fence->device; + struct vkd3d_signaled_semaphore *current; + unsigned int i; + int rc; + + if ((rc = pthread_mutex_lock(&fence->mutex))) + { + ERR("Failed to lock mutex, error %d.\n", rc); + return hresult_from_errno(rc); + } + + if (value < fence->value && device->vk_info.KHR_timeline_semaphore) + { + FIXME("Fence values must be monotonically increasing. Fence %p, was %"PRIx64", now %"PRIx64".\n", + fence, fence->value, value); + } + else + { + fence->value = value; + } + + if (d3d12_fence_signal_external_events_locked(fence)) pthread_cond_broadcast(&fence->null_event_cond); if (vk_fence) @@ -1069,12 +1180,71 @@ static HRESULT STDMETHODCALLTYPE d3d12_fence_SetEventOnCompletion(ID3D12Fence *i return S_OK; } +static HRESULT d3d12_fence_signal_cpu_timeline_semaphore(struct d3d12_fence *fence, uint64_t value) +{ + struct d3d12_device *device = fence->device; + VkSemaphoreSignalInfoKHR info; + VkResult vr = VK_SUCCESS; + int rc; + + if ((rc = pthread_mutex_lock(&fence->mutex))) + { + ERR("Failed to lock mutex, error %d.\n", rc); + return hresult_from_errno(rc); + } + + /* We must only signal a value which is greater than the current value. + * That value can be in the range of current known value (fence->value), or as large as pending_timeline_value. + * Pending timeline value signal might be blocked by another synchronization primitive, and thus statically + * cannot be that value, so the safest thing to do is to check the current value which is updated by the fence + * wait thread continuously. This check is technically racy since the value might be immediately out of date, + * but there is no way to avoid this. */ + if (value > fence->value) + { + struct vkd3d_vk_device_procs *vk_procs = &device->vk_procs; + + /* Sanity check against the delta limit. */ + if (value - fence->value > device->vk_info.timeline_semaphore_properties.maxTimelineSemaphoreValueDifference) + { + FIXME("Timeline semaphore delta is %"PRIu64", but implementation only supports a delta of %"PRIu64".\n", + value - fence->value, device->vk_info.timeline_semaphore_properties.maxTimelineSemaphoreValueDifference); + } + + info.sType = VK_STRUCTURE_TYPE_SEMAPHORE_SIGNAL_INFO_KHR; + info.pNext = NULL; + info.semaphore = fence->timeline_semaphore; + info.value = value; + if ((vr = VK_CALL(vkSignalSemaphoreKHR(device->vk_device, &info))) >= 0) + { + fence->value = value; + if (value > fence->pending_timeline_value) + fence->pending_timeline_value = value; + } + else + { + ERR("Failed to signal timeline semaphore, vr %d.\n", vr); + } + } + else if (value != fence->value) + { + FIXME("Attempting to signal fence %p with %"PRIu64", but value is currently %"PRIu64".\n", + fence, value, fence->value); + } + + d3d12_fence_signal_external_events_locked(fence); + + pthread_mutex_unlock(&fence->mutex); + return hresult_from_vk_result(vr); +} + static HRESULT STDMETHODCALLTYPE d3d12_fence_Signal(ID3D12Fence *iface, UINT64 value) { struct d3d12_fence *fence = impl_from_ID3D12Fence(iface); TRACE("iface %p, value %#"PRIx64".\n", iface, value); + if (fence->timeline_semaphore) + return d3d12_fence_signal_cpu_timeline_semaphore(fence, value); return d3d12_fence_signal(fence, value, VK_NULL_HANDLE); } @@ -1105,6 +1275,34 @@ static struct d3d12_fence *unsafe_impl_from_ID3D12Fence(ID3D12Fence *iface) return impl_from_ID3D12Fence(iface); } +static HRESULT d3d12_fence_init_timeline_semaphore(struct d3d12_fence *fence, struct d3d12_device *device, + uint64_t initial_value) +{ + const struct vkd3d_vk_device_procs *vk_procs = &device->vk_procs; + VkSemaphoreTypeCreateInfoKHR type_info; + VkSemaphoreCreateInfo info; + VkResult vr; + + info.sType = VK_STRUCTURE_TYPE_SEMAPHORE_CREATE_INFO; + info.pNext = &type_info; + info.flags = 0; + + type_info.sType = VK_STRUCTURE_TYPE_SEMAPHORE_TYPE_CREATE_INFO_KHR; + type_info.pNext = NULL; + type_info.semaphoreType = VK_SEMAPHORE_TYPE_TIMELINE_KHR; + type_info.initialValue = initial_value; + + if ((vr = VK_CALL(vkCreateSemaphore(device->vk_device, &info, NULL, &fence->timeline_semaphore))) < 0) + { + WARN("Failed to create timeline semaphore, vr %d.\n", vr); + return hresult_from_vk_result(vr); + } + + fence->pending_timeline_value = initial_value; + + return S_OK; +} + static HRESULT d3d12_fence_init(struct d3d12_fence *fence, struct d3d12_device *device, UINT64 initial_value, D3D12_FENCE_FLAGS flags) { @@ -1136,6 +1334,11 @@ static HRESULT d3d12_fence_init(struct d3d12_fence *fence, struct d3d12_device * fence->events_size = 0; fence->event_count = 0; + fence->timeline_semaphore = NULL; + if (device->vk_info.KHR_timeline_semaphore && FAILED(hr = d3d12_fence_init_timeline_semaphore(fence, + device, initial_value))) + return hr; + list_init(&fence->semaphores); fence->semaphore_count = 0; @@ -6000,18 +6203,97 @@ static void STDMETHODCALLTYPE d3d12_command_queue_EndEvent(ID3D12CommandQueue *i FIXME("iface %p stub!\n", iface); } +static bool d3d12_fence_update_gpu_signal_timeline_semaphore(struct d3d12_fence *fence, uint64_t value) +{ + struct d3d12_device *device = fence->device; + bool need_signal = false; + int rc; + + if ((rc = pthread_mutex_lock(&fence->mutex))) + { + ERR("Failed to lock mutex, error %d.\n", rc); + return false; + } + + /* If we're attempting to async signal a fence with a value which is not monotonically increasing the payload value, + * warn about this case. Do not treat this as an error since it might work. */ + if (value > fence->pending_timeline_value) + { + /* Sanity check against the delta limit. Use the current fence value. */ + if (value - fence->value > device->vk_info.timeline_semaphore_properties.maxTimelineSemaphoreValueDifference) + { + FIXME("Timeline semaphore delta is %"PRIu64", but implementation only supports a delta of %"PRIu64".\n", + value - fence->value, device->vk_info.timeline_semaphore_properties.maxTimelineSemaphoreValueDifference); + } + + fence->pending_timeline_value = value; + need_signal = true; + } + else + { + FIXME("Fence %p is being signalled non-monotonically. Old pending value %"PRIu64", new pending value %"PRIu64".\n", + fence, fence->pending_timeline_value, value); + + /* Mostly to be safe against weird, unknown use cases. The pending signal might be blocked + * by another fence, so we'll base this on the actual, currently visible count value. */ + need_signal = value > fence->value; + } + + pthread_mutex_unlock(&fence->mutex); + + return need_signal; +} + +static HRESULT vkd3d_enqueue_timeline_semaphore(struct vkd3d_fence_worker *worker, VkSemaphore vk_semaphore, + struct d3d12_fence *fence, uint64_t value, struct vkd3d_queue *queue) +{ + struct vkd3d_waiting_fence *waiting_fence; + int rc; + + TRACE("worker %p, fence %p, value %#"PRIx64".\n", worker, fence, value); + + if ((rc = pthread_mutex_lock(&worker->mutex))) + { + ERR("Failed to lock mutex, error %d.\n", rc); + return hresult_from_errno(rc); + } + + if (!vkd3d_array_reserve((void **)&worker->enqueued_fences, &worker->enqueued_fences_size, + worker->enqueued_fence_count + 1, sizeof(*worker->enqueued_fences))) + { + ERR("Failed to add GPU timeline semaphore.\n"); + pthread_mutex_unlock(&worker->mutex); + return E_OUTOFMEMORY; + } + + worker->enqueued_fences[worker->enqueued_fence_count].vk_semaphore = vk_semaphore; + waiting_fence = &worker->enqueued_fences[worker->enqueued_fence_count].waiting_fence; + waiting_fence->fence = fence; + waiting_fence->value = value; + waiting_fence->queue = queue; + ++worker->enqueued_fence_count; + + InterlockedIncrement(&fence->pending_worker_operation_count); + + pthread_cond_signal(&worker->cond); + pthread_mutex_unlock(&worker->mutex); + + return S_OK; +} + static HRESULT STDMETHODCALLTYPE d3d12_command_queue_Signal(ID3D12CommandQueue *iface, ID3D12Fence *fence_iface, UINT64 value) { struct d3d12_command_queue *command_queue = impl_from_ID3D12CommandQueue(iface); + VkTimelineSemaphoreSubmitInfoKHR timeline_submit_info; const struct vkd3d_vk_device_procs *vk_procs; VkSemaphore vk_semaphore = VK_NULL_HANDLE; VkFence vk_fence = VK_NULL_HANDLE; struct vkd3d_queue *vkd3d_queue; + uint64_t sequence_number = 0; struct d3d12_device *device; struct d3d12_fence *fence; VkSubmitInfo submit_info; - uint64_t sequence_number; VkQueue vk_queue; VkResult vr; HRESULT hr; @@ -6024,10 +6306,27 @@ static HRESULT STDMETHODCALLTYPE d3d12_command_queue_Signal(ID3D12CommandQueue * fence = unsafe_impl_from_ID3D12Fence(fence_iface); - if ((vr = d3d12_fence_create_vk_fence(fence, &vk_fence)) < 0) + if (device->vk_info.KHR_timeline_semaphore) { - WARN("Failed to create Vulkan fence, vr %d.\n", vr); - goto fail_vkresult; + if (d3d12_fence_update_gpu_signal_timeline_semaphore(fence, value)) + { + vk_semaphore = fence->timeline_semaphore; + assert(vk_semaphore); + } + else + { + /* If we are not incrementing the counter, this is a noop since we cannot + * signal a timeline semaphore non-monotonically in Vulkan. */ + return S_OK; + } + } + else + { + if ((vr = d3d12_fence_create_vk_fence(fence, &vk_fence)) < 0) + { + WARN("Failed to create Vulkan fence, vr %d.\n", vr); + goto fail_vkresult; + } } if (!(vk_queue = vkd3d_queue_acquire(vkd3d_queue))) @@ -6037,7 +6336,8 @@ static HRESULT STDMETHODCALLTYPE d3d12_command_queue_Signal(ID3D12CommandQueue * goto fail; } - if ((vr = vkd3d_queue_create_vk_semaphore_locked(vkd3d_queue, device, &vk_semaphore)) < 0) + if (!device->vk_info.KHR_timeline_semaphore && (vr = vkd3d_queue_create_vk_semaphore_locked(vkd3d_queue, + device, &vk_semaphore)) < 0) { ERR("Failed to create Vulkan semaphore, vr %d.\n", vr); vk_semaphore = VK_NULL_HANDLE; @@ -6053,7 +6353,19 @@ static HRESULT STDMETHODCALLTYPE d3d12_command_queue_Signal(ID3D12CommandQueue * submit_info.signalSemaphoreCount = vk_semaphore ? 1 : 0; submit_info.pSignalSemaphores = &vk_semaphore; - if ((vr = VK_CALL(vkQueueSubmit(vk_queue, 1, &submit_info, vk_fence))) >= 0) + if (device->vk_info.KHR_timeline_semaphore) + { + timeline_submit_info.sType = VK_STRUCTURE_TYPE_TIMELINE_SEMAPHORE_SUBMIT_INFO_KHR; + timeline_submit_info.pNext = NULL; + timeline_submit_info.pSignalSemaphoreValues = &value; + timeline_submit_info.signalSemaphoreValueCount = 1; + timeline_submit_info.waitSemaphoreValueCount = 0; + timeline_submit_info.pWaitSemaphoreValues = NULL; + submit_info.pNext = &timeline_submit_info; + } + + vr = VK_CALL(vkQueueSubmit(vk_queue, 1, &submit_info, vk_fence)); + if (!device->vk_info.KHR_timeline_semaphore && vr >= 0) { sequence_number = ++vkd3d_queue->submitted_sequence_number; @@ -6070,26 +6382,34 @@ static HRESULT STDMETHODCALLTYPE d3d12_command_queue_Signal(ID3D12CommandQueue * goto fail_vkresult; } - if (vk_semaphore && SUCCEEDED(hr = d3d12_fence_add_vk_semaphore(fence, vk_semaphore, vk_fence, value))) - vk_semaphore = VK_NULL_HANDLE; - - vr = VK_CALL(vkGetFenceStatus(device->vk_device, vk_fence)); - if (vr == VK_NOT_READY) - { - if (SUCCEEDED(hr = vkd3d_enqueue_gpu_fence(&device->fence_worker, vk_fence, fence, value, vkd3d_queue, sequence_number))) - vk_fence = VK_NULL_HANDLE; - } - else if (vr == VK_SUCCESS) + if (device->vk_info.KHR_timeline_semaphore) { - TRACE("Already signaled %p, value %#"PRIx64".\n", fence, value); - hr = d3d12_fence_signal(fence, value, vk_fence); - vk_fence = VK_NULL_HANDLE; - vkd3d_queue_update_sequence_number(vkd3d_queue, sequence_number, device); + if (SUCCEEDED(hr = vkd3d_enqueue_timeline_semaphore(&device->fence_worker, vk_semaphore, fence, value, vkd3d_queue))) + vk_semaphore = VK_NULL_HANDLE; } else { - FIXME("Failed to get fence status, vr %d.\n", vr); - hr = hresult_from_vk_result(vr); + if (vk_semaphore && SUCCEEDED(hr = d3d12_fence_add_vk_semaphore(fence, vk_semaphore, vk_fence, value))) + vk_semaphore = VK_NULL_HANDLE; + + vr = VK_CALL(vkGetFenceStatus(device->vk_device, vk_fence)); + if (vr == VK_NOT_READY) + { + if (SUCCEEDED(hr = vkd3d_enqueue_gpu_fence(&device->fence_worker, vk_fence, fence, value, vkd3d_queue, sequence_number))) + vk_fence = VK_NULL_HANDLE; + } + else if (vr == VK_SUCCESS) + { + TRACE("Already signaled %p, value %#"PRIx64".\n", fence, value); + hr = d3d12_fence_signal(fence, value, vk_fence); + vk_fence = VK_NULL_HANDLE; + vkd3d_queue_update_sequence_number(vkd3d_queue, sequence_number, device); + } + else + { + FIXME("Failed to get fence status, vr %d.\n", vr); + hr = hresult_from_vk_result(vr); + } } if (vk_fence || vk_semaphore) @@ -6105,32 +6425,27 @@ fail_vkresult: hr = hresult_from_vk_result(vr); fail: VK_CALL(vkDestroyFence(device->vk_device, vk_fence, NULL)); - VK_CALL(vkDestroySemaphore(device->vk_device, vk_semaphore, NULL)); + if (!device->vk_info.KHR_timeline_semaphore) + VK_CALL(vkDestroySemaphore(device->vk_device, vk_semaphore, NULL)); return hr; } -static HRESULT STDMETHODCALLTYPE d3d12_command_queue_Wait(ID3D12CommandQueue *iface, - ID3D12Fence *fence_iface, UINT64 value) +static HRESULT d3d12_command_queue_wait_binary_semaphore(struct d3d12_command_queue *command_queue, + struct d3d12_fence *fence, uint64_t value) { static const VkPipelineStageFlagBits wait_stage_mask = VK_PIPELINE_STAGE_ALL_COMMANDS_BIT; - struct d3d12_command_queue *command_queue = impl_from_ID3D12CommandQueue(iface); const struct vkd3d_vk_device_procs *vk_procs; struct vkd3d_signaled_semaphore *semaphore; uint64_t completed_value = 0; struct vkd3d_queue *queue; - struct d3d12_fence *fence; VkSubmitInfo submit_info; VkQueue vk_queue; VkResult vr; HRESULT hr; - TRACE("iface %p, fence %p, value %#"PRIx64".\n", iface, fence_iface, value); - vk_procs = &command_queue->device->vk_procs; queue = command_queue->vkd3d_queue; - fence = unsafe_impl_from_ID3D12Fence(fence_iface); - semaphore = d3d12_fence_acquire_vk_semaphore(fence, value, &completed_value); if (!semaphore && completed_value >= value) { @@ -6208,6 +6523,72 @@ fail: return hr; } +static HRESULT d3d12_command_queue_wait_timeline_semaphore(struct d3d12_command_queue *command_queue, + struct d3d12_fence *fence, uint64_t value) +{ + static const VkPipelineStageFlagBits wait_stage_mask = VK_PIPELINE_STAGE_ALL_COMMANDS_BIT; + VkTimelineSemaphoreSubmitInfoKHR timeline_submit_info; + const struct vkd3d_vk_device_procs *vk_procs; + struct vkd3d_queue *queue; + VkSubmitInfo submit_info; + VkQueue vk_queue; + VkResult vr; + + vk_procs = &command_queue->device->vk_procs; + queue = command_queue->vkd3d_queue; + + assert(fence->timeline_semaphore); + timeline_submit_info.sType = VK_STRUCTURE_TYPE_TIMELINE_SEMAPHORE_SUBMIT_INFO_KHR; + timeline_submit_info.pNext = NULL; + timeline_submit_info.signalSemaphoreValueCount = 0; + timeline_submit_info.pSignalSemaphoreValues = NULL; + timeline_submit_info.waitSemaphoreValueCount = 1; + timeline_submit_info.pWaitSemaphoreValues = &value; + + if (!(vk_queue = vkd3d_queue_acquire(queue))) + { + ERR("Failed to acquire queue %p.\n", queue); + return E_FAIL; + } + + submit_info.sType = VK_STRUCTURE_TYPE_SUBMIT_INFO; + submit_info.pNext = &timeline_submit_info; + submit_info.waitSemaphoreCount = 1; + submit_info.pWaitSemaphores = &fence->timeline_semaphore; + submit_info.pWaitDstStageMask = &wait_stage_mask; + submit_info.commandBufferCount = 0; + submit_info.pCommandBuffers = NULL; + submit_info.signalSemaphoreCount = 0; + submit_info.pSignalSemaphores = NULL; + + vr = VK_CALL(vkQueueSubmit(vk_queue, 1, &submit_info, VK_NULL_HANDLE)); + + vkd3d_queue_release(queue); + + if (vr < 0) + { + WARN("Failed to submit wait operation, vr %d.\n", vr); + return hresult_from_vk_result(vr); + } + + return S_OK; +} + +static HRESULT STDMETHODCALLTYPE d3d12_command_queue_Wait(ID3D12CommandQueue *iface, + ID3D12Fence *fence_iface, UINT64 value) +{ + struct d3d12_command_queue *command_queue = impl_from_ID3D12CommandQueue(iface); + struct d3d12_fence *fence = unsafe_impl_from_ID3D12Fence(fence_iface); + + TRACE("iface %p, fence %p, value %#"PRIx64".\n", iface, fence_iface, value); + + if (command_queue->device->vk_info.KHR_timeline_semaphore) + return d3d12_command_queue_wait_timeline_semaphore(command_queue, fence, value); + + FIXME_ONCE("KHR_timeline_semaphore is not available. Some wait commands may be unsupported.\n"); + return d3d12_command_queue_wait_binary_semaphore(command_queue, fence, value); +} + static HRESULT STDMETHODCALLTYPE d3d12_command_queue_GetTimestampFrequency(ID3D12CommandQueue *iface, UINT64 *frequency) { diff --git a/libs/vkd3d/device.c b/libs/vkd3d/device.c index 4bcb5efc..bc5cad76 100644 --- a/libs/vkd3d/device.c +++ b/libs/vkd3d/device.c @@ -129,6 +129,7 @@ static const struct vkd3d_optional_extension_info optional_device_extensions[] = VK_EXTENSION(KHR_MAINTENANCE3, KHR_maintenance3), VK_EXTENSION(KHR_PUSH_DESCRIPTOR, KHR_push_descriptor), VK_EXTENSION(KHR_SAMPLER_MIRROR_CLAMP_TO_EDGE, KHR_sampler_mirror_clamp_to_edge), + VK_EXTENSION(KHR_TIMELINE_SEMAPHORE, KHR_timeline_semaphore), /* EXT extensions */ VK_EXTENSION(EXT_CALIBRATED_TIMESTAMPS, EXT_calibrated_timestamps), VK_EXTENSION(EXT_CONDITIONAL_RENDERING, EXT_conditional_rendering), @@ -683,6 +684,7 @@ struct vkd3d_physical_device_info VkPhysicalDeviceTexelBufferAlignmentPropertiesEXT texel_buffer_alignment_properties; VkPhysicalDeviceTransformFeedbackPropertiesEXT xfb_properties; VkPhysicalDeviceVertexAttributeDivisorPropertiesEXT vertex_divisor_properties; + VkPhysicalDeviceTimelineSemaphorePropertiesKHR timeline_semaphore_properties; VkPhysicalDeviceProperties2KHR properties2; @@ -694,6 +696,7 @@ struct vkd3d_physical_device_info VkPhysicalDeviceTexelBufferAlignmentFeaturesEXT texel_buffer_alignment_features; VkPhysicalDeviceTransformFeedbackFeaturesEXT xfb_features; VkPhysicalDeviceVertexAttributeDivisorFeaturesEXT vertex_divisor_features; + VkPhysicalDeviceTimelineSemaphoreFeaturesKHR timeline_semaphore_features; VkPhysicalDeviceFeatures2 features2; }; @@ -705,10 +708,12 @@ static void vkd3d_physical_device_info_init(struct vkd3d_physical_device_info *i VkPhysicalDeviceDescriptorIndexingPropertiesEXT *descriptor_indexing_properties; VkPhysicalDeviceVertexAttributeDivisorPropertiesEXT *vertex_divisor_properties; VkPhysicalDeviceTexelBufferAlignmentPropertiesEXT *buffer_alignment_properties; + VkPhysicalDeviceTimelineSemaphorePropertiesKHR *timeline_semaphore_properties; VkPhysicalDeviceDescriptorIndexingFeaturesEXT *descriptor_indexing_features; VkPhysicalDeviceVertexAttributeDivisorFeaturesEXT *vertex_divisor_features; VkPhysicalDeviceTexelBufferAlignmentFeaturesEXT *buffer_alignment_features; VkPhysicalDeviceShaderDemoteToHelperInvocationFeaturesEXT *demote_features; + VkPhysicalDeviceTimelineSemaphoreFeaturesKHR *timeline_semaphore_features; VkPhysicalDeviceDepthClipEnableFeaturesEXT *depth_clip_features; VkPhysicalDeviceMaintenance3Properties *maintenance3_properties; VkPhysicalDeviceTransformFeedbackPropertiesEXT *xfb_properties; @@ -727,6 +732,8 @@ static void vkd3d_physical_device_info_init(struct vkd3d_physical_device_info *i buffer_alignment_properties = &info->texel_buffer_alignment_properties; vertex_divisor_features = &info->vertex_divisor_features; vertex_divisor_properties = &info->vertex_divisor_properties; + timeline_semaphore_features = &info->timeline_semaphore_features; + timeline_semaphore_properties = &info->timeline_semaphore_properties; xfb_features = &info->xfb_features; xfb_properties = &info->xfb_properties; @@ -746,6 +753,8 @@ static void vkd3d_physical_device_info_init(struct vkd3d_physical_device_info *i vk_prepend_struct(&info->features2, xfb_features); vertex_divisor_features->sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VERTEX_ATTRIBUTE_DIVISOR_FEATURES_EXT; vk_prepend_struct(&info->features2, vertex_divisor_features); + timeline_semaphore_features->sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_TIMELINE_SEMAPHORE_FEATURES_KHR; + vk_prepend_struct(&info->features2, timeline_semaphore_features); if (vulkan_info->KHR_get_physical_device_properties2) VK_CALL(vkGetPhysicalDeviceFeatures2KHR(physical_device, &info->features2)); @@ -764,6 +773,8 @@ static void vkd3d_physical_device_info_init(struct vkd3d_physical_device_info *i vk_prepend_struct(&info->properties2, xfb_properties); vertex_divisor_properties->sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VERTEX_ATTRIBUTE_DIVISOR_PROPERTIES_EXT; vk_prepend_struct(&info->properties2, vertex_divisor_properties); + timeline_semaphore_properties->sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_TIMELINE_SEMAPHORE_PROPERTIES_KHR; + vk_prepend_struct(&info->properties2, timeline_semaphore_properties); if (vulkan_info->KHR_get_physical_device_properties2) VK_CALL(vkGetPhysicalDeviceProperties2KHR(physical_device, &info->properties2)); @@ -1312,6 +1323,7 @@ static HRESULT vkd3d_init_device_caps(struct d3d12_device *device, vulkan_info->rasterization_stream = physical_device_info->xfb_properties.transformFeedbackRasterizationStreamSelect; vulkan_info->transform_feedback_queries = physical_device_info->xfb_properties.transformFeedbackQueries; vulkan_info->max_vertex_attrib_divisor = max(physical_device_info->vertex_divisor_properties.maxVertexAttribDivisor, 1); + vulkan_info->timeline_semaphore_properties = physical_device_info->timeline_semaphore_properties; device->feature_options.DoublePrecisionFloatShaderOps = features->shaderFloat64; device->feature_options.OutputMergerLogicOp = features->logicOp; @@ -1434,6 +1446,8 @@ static HRESULT vkd3d_init_device_caps(struct d3d12_device *device, vulkan_info->EXT_shader_demote_to_helper_invocation = false; if (!physical_device_info->texel_buffer_alignment_features.texelBufferAlignment) vulkan_info->EXT_texel_buffer_alignment = false; + if (!physical_device_info->timeline_semaphore_features.timelineSemaphore) + vulkan_info->KHR_timeline_semaphore = false; vulkan_info->texel_buffer_alignment_properties = physical_device_info->texel_buffer_alignment_properties; diff --git a/libs/vkd3d/vkd3d_private.h b/libs/vkd3d/vkd3d_private.h index 136b0203..088d4312 100644 --- a/libs/vkd3d/vkd3d_private.h +++ b/libs/vkd3d/vkd3d_private.h @@ -107,6 +107,7 @@ struct vkd3d_vulkan_info bool KHR_maintenance3; bool KHR_push_descriptor; bool KHR_sampler_mirror_clamp_to_edge; + bool KHR_timeline_semaphore; /* EXT device extensions */ bool EXT_calibrated_timestamps; bool EXT_conditional_rendering; @@ -130,6 +131,8 @@ struct vkd3d_vulkan_info VkPhysicalDeviceTexelBufferAlignmentPropertiesEXT texel_buffer_alignment_properties; + VkPhysicalDeviceTimelineSemaphorePropertiesKHR timeline_semaphore_properties; + unsigned int shader_extension_count; enum vkd3d_shader_spirv_extension shader_extensions[VKD3D_MAX_SHADER_EXTENSIONS]; @@ -196,6 +199,7 @@ struct vkd3d_fence_worker struct vkd3d_enqueued_fence { VkFence vk_fence; + VkSemaphore vk_semaphore; struct vkd3d_waiting_fence waiting_fence; } *enqueued_fences; size_t enqueued_fences_size; @@ -205,6 +209,10 @@ struct vkd3d_fence_worker size_t vk_fences_size; struct vkd3d_waiting_fence *fences; size_t fences_size; + VkSemaphore *vk_semaphores; + size_t vk_semaphores_size; + uint64_t *semaphore_wait_values; + size_t semaphore_wait_values_size; struct d3d12_device *device; }; @@ -378,6 +386,9 @@ struct d3d12_fence size_t events_size; size_t event_count; + VkSemaphore timeline_semaphore; + uint64_t pending_timeline_value; + struct list semaphores; unsigned int semaphore_count; diff --git a/libs/vkd3d/vulkan_procs.h b/libs/vkd3d/vulkan_procs.h index 60556735..34e0ab4b 100644 --- a/libs/vkd3d/vulkan_procs.h +++ b/libs/vkd3d/vulkan_procs.h @@ -195,6 +195,11 @@ VK_DEVICE_EXT_PFN(vkGetDescriptorSetLayoutSupportKHR) /* VK_KHR_push_descriptor */ VK_DEVICE_EXT_PFN(vkCmdPushDescriptorSetKHR) +/* VK_KHR_timeline_semaphore */ +VK_DEVICE_EXT_PFN(vkGetSemaphoreCounterValueKHR) +VK_DEVICE_EXT_PFN(vkWaitSemaphoresKHR) +VK_DEVICE_EXT_PFN(vkSignalSemaphoreKHR) + /* VK_EXT_calibrated_timestamps */ VK_DEVICE_EXT_PFN(vkGetCalibratedTimestampsEXT) diff --git a/tests/d3d12.c b/tests/d3d12.c index 82f353a7..b5a9b89f 100644 --- a/tests/d3d12.c +++ b/tests/d3d12.c @@ -33238,7 +33238,9 @@ static void test_queue_wait(void) command_list = context.list; queue = context.queue; - queue2 = create_command_queue(device, D3D12_COMMAND_LIST_TYPE_DIRECT, D3D12_COMMAND_QUEUE_PRIORITY_NORMAL); + /* 'queue2' must not map to the same command queue as 'queue', or Wait() before GPU signal will fail. + * Using a compute queue fixes this on most hardware, but it may still fail on low spec hardware. */ + queue2 = create_command_queue(device, D3D12_COMMAND_LIST_TYPE_COMPUTE, D3D12_COMMAND_QUEUE_PRIORITY_NORMAL); event = create_event(); ok(event, "Failed to create event.\n"); @@ -33303,12 +33305,6 @@ static void test_queue_wait(void) check_readback_data_uint(&rb, NULL, 0xff00ff00, 0); release_resource_readback(&rb); - if (!vkd3d_test_platform_is_windows()) - { - skip("Wait() is not implemented yet.\n"); /* FIXME */ - goto skip_tests; - } - /* Wait() before CPU signal */ update_buffer_data(cb, 0, sizeof(blue), &blue); queue_wait(queue, fence, 2); @@ -33384,7 +33380,6 @@ static void test_queue_wait(void) check_readback_data_uint(&rb, NULL, 0xff00ff00, 0); release_resource_readback(&rb); -skip_tests: /* Signal() and Wait() in the same command queue */ update_buffer_data(cb, 0, sizeof(blue), &blue); queue_signal(queue, fence, 7); -- 2.34.1

1 0

[PATCH 2/2] msvcp140_atomic_wait: Implement __std_parallel_algorithms_hw_threads.
by Daniel Lehman Jan. 25, 2022

Jan. 25, 2022

2 1

[PATCH 1/2] winevulkan: Update to VK spec version 1.3.204.
by Georg Lehmann Jan. 25, 2022

Jan. 25, 2022

Signed-off-by: Georg Lehmann <dadschoorse(a)gmail.com> --- dlls/winevulkan/make_vulkan | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/dlls/winevulkan/make_vulkan b/dlls/winevulkan/make_vulkan index e548025a6ad..33c4495d8aa 100755 --- a/dlls/winevulkan/make_vulkan +++ b/dlls/winevulkan/make_vulkan @@ -64,7 +64,7 @@ from enum import Enum LOGGER = logging.Logger("vulkan") LOGGER.addHandler(logging.StreamHandler()) -VK_XML_VERSION = "1.2.203" +VK_XML_VERSION = "1.3.204" WINE_VK_VERSION = (1, 2) # Filenames to create. -- 2.35.0

1 1

[PATCH 1/9] winhttp: Return ERROR_INVALID_OPERATION when websocket is in the wrong state.
by Paul Gofman Jan. 25, 2022

Jan. 25, 2022

Signed-off-by: Paul Gofman <pgofman(a)codeweavers.com> --- The main idea behind this series is to address a few things regarind socket shutdown and close: - avoid directly sending shutdown frame from WinHttpWebSocketClose() which doesn't cope nicely with sync / async send duality; - address the case of closing websocket which has receive waiting. - changing the socket status right away before the close or shutdown operation is complete through the queue (which is needed to straigten up the above and the following later bit which didn't fit in this series). In theory we could maybe cancel socket I/O on the socket to abort receive instead of teaching socket receive to read the close status but that seems harder to synchronize and I suppose that would complicate the things considerably without a clear advantage. The remaining part of WinHttpWebSocketClose() changes which didn't fit in in this patch series concerns the way the callback is sent when receive operation is pending during WinHttpWebSocketClose() call. Windows always calls the callback synchronously during WinHttpWebSocketClose() with dwResult 0 and dwError ERROR_WINHTTP_OPERATION_CANCELLED, the async receive thread does not call any callback in this case. dlls/winhttp/request.c | 10 +++++----- dlls/winhttp/tests/notification.c | 15 ++++++++++++++- 2 files changed, 19 insertions(+), 6 deletions(-) diff --git a/dlls/winhttp/request.c b/dlls/winhttp/request.c index 31e65a68359..fc2c06bd11f 100644 --- a/dlls/winhttp/request.c +++ b/dlls/winhttp/request.c @@ -3324,7 +3324,7 @@ DWORD WINAPI WinHttpWebSocketSend( HINTERNET hsocket, WINHTTP_WEB_SOCKET_BUFFER_ if (socket->state != SOCKET_STATE_OPEN) { release_object( &socket->hdr ); - return ERROR_WINHTTP_INCORRECT_HANDLE_STATE; + return ERROR_INVALID_OPERATION; } if (socket->request->connect->hdr.flags & WINHTTP_FLAG_ASYNC) @@ -3625,7 +3625,7 @@ DWORD WINAPI WinHttpWebSocketReceive( HINTERNET hsocket, void *buf, DWORD len, D if (socket->state > SOCKET_STATE_SHUTDOWN) { release_object( &socket->hdr ); - return ERROR_WINHTTP_INCORRECT_HANDLE_STATE; + return ERROR_INVALID_OPERATION; } if (socket->request->connect->hdr.flags & WINHTTP_FLAG_ASYNC) @@ -3704,7 +3704,7 @@ DWORD WINAPI WinHttpWebSocketShutdown( HINTERNET hsocket, USHORT status, void *r if (socket->state >= SOCKET_STATE_SHUTDOWN) { release_object( &socket->hdr ); - return ERROR_WINHTTP_INCORRECT_HANDLE_STATE; + return ERROR_INVALID_OPERATION; } if (socket->request->connect->hdr.flags & WINHTTP_FLAG_ASYNC) @@ -3814,7 +3814,7 @@ DWORD WINAPI WinHttpWebSocketClose( HINTERNET hsocket, USHORT status, void *reas if (socket->state >= SOCKET_STATE_CLOSED) { release_object( &socket->hdr ); - return ERROR_WINHTTP_INCORRECT_HANDLE_STATE; + return ERROR_INVALID_OPERATION; } if (socket->request->connect->hdr.flags & WINHTTP_FLAG_ASYNC) @@ -3859,7 +3859,7 @@ DWORD WINAPI WinHttpWebSocketQueryCloseStatus( HINTERNET hsocket, USHORT *status if (socket->state < SOCKET_STATE_CLOSED) { release_object( &socket->hdr ); - return ERROR_WINHTTP_INCORRECT_HANDLE_STATE; + return ERROR_INVALID_OPERATION; } *status = socket->status; diff --git a/dlls/winhttp/tests/notification.c b/dlls/winhttp/tests/notification.c index e919159ba98..0c3ff98c342 100644 --- a/dlls/winhttp/tests/notification.c +++ b/dlls/winhttp/tests/notification.c @@ -816,6 +816,12 @@ static void test_websocket(BOOL secure) ok( err == ERROR_SUCCESS, "got %u\n", err ); WaitForSingleObject( info.wait, INFINITE ); + err = pWinHttpWebSocketShutdown( socket, 1000, (void *)"success", sizeof("success") ); + ok( err == ERROR_INVALID_OPERATION, "got %u\n", err ); + err = pWinHttpWebSocketSend( socket, WINHTTP_WEB_SOCKET_BINARY_MESSAGE_BUFFER_TYPE, + (void*)"hello", sizeof("hello") ); + ok( err == ERROR_INVALID_OPERATION, "got %u\n", err ); + setup_test( &info, winhttp_websocket_receive, __LINE__ ); buffer[0] = 0; size = 0xdeadbeef; @@ -838,8 +844,15 @@ static void test_websocket(BOOL secure) ok( type == 0xdeadbeef, "got %u\n", type ); ok( buffer[0] == 'h', "unexpected data\n" ); + close_status = 0xdead; + size = sizeof(buffer) + 1; + err = pWinHttpWebSocketQueryCloseStatus( socket, &close_status, buffer, sizeof(buffer), &size ); + ok( err == ERROR_INVALID_OPERATION, "got %u\n", err ); + ok( close_status == 0xdead, "got %u\n", close_status ); + ok( size == sizeof(buffer) + 1, "got %u\n", size ); + setup_test( &info, winhttp_websocket_close, __LINE__ ); - ret = pWinHttpWebSocketClose( socket, 1000, (void *)"success", sizeof("success") ); + err = pWinHttpWebSocketClose( socket, 1000, (void *)"success", sizeof("success") ); ok( err == ERROR_SUCCESS, "got %u\n", err ); WaitForSingleObject( info.wait, INFINITE ); -- 2.34.1

3 21

[PATCH v6] ntdll: Use case insensitive comparison in is_import_dll_system().
by Paul Gofman Jan. 25, 2022

Jan. 25, 2022

From: Ben Shefte <shefben(a)gmail.com> Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=52446 Fixes a regression introduced by commit 250c113169217933b1fffa8024fa958f71b8aee7. Signed-off-by: Paul Gofman <pgofman(a)codeweavers.com> --- Supersedes 224209. v6: - make patch compile; - change subject; - use _stricmp() instead of stricmp(). changelog: v1-v4 -> 5 removed unicode code referemces Fixes Regression from: Wine 6.21 dlls/ntdll/loader.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/dlls/ntdll/loader.c b/dlls/ntdll/loader.c index 362e1c66be5..b50b9e61bbc 100644 --- a/dlls/ntdll/loader.c +++ b/dlls/ntdll/loader.c @@ -628,7 +628,7 @@ static BOOL is_import_dll_system( LDR_DATA_TABLE_ENTRY *mod, const IMAGE_IMPORT_ { const char *name = get_rva( mod->DllBase, import->Name ); - return !strcmp( name, "ntdll.dll" ) || !strcmp( name, "kernel32.dll" ); + return !_stricmp( name, "ntdll.dll" ) || !_stricmp( name, "kernel32.dll" ); } /********************************************************************** -- 2.34.1

1 0

[PATCH 2/2] server: Strip terminating '\0' from the registry key name. (Resend)
by Dmitry Timoshkov Jan. 25, 2022

Jan. 25, 2022

Signed-off-by: Dmitry Timoshkov <dmitry(a)baikal.ru> --- dlls/ntdll/tests/reg.c | 18 +++++++++++++++++- server/registry.c | 2 ++ 2 files changed, 19 insertions(+), 1 deletion(-) diff --git a/dlls/ntdll/tests/reg.c b/dlls/ntdll/tests/reg.c index 445c05a02e1..956df02fae4 100644 --- a/dlls/ntdll/tests/reg.c +++ b/dlls/ntdll/tests/reg.c @@ -456,6 +456,14 @@ todo_wine ok( status == STATUS_OBJECT_TYPE_MISMATCH, "NtOpenKey failed: 0x%08x\n", status ); pRtlFreeUnicodeString( &str ); + pRtlCreateUnicodeStringFromAsciiz( &str, "\\REGISTRY\\Machine\\Software\\Classes" ); + str.Length += sizeof(WCHAR); /* include terminating \0 in the length */ + status = pNtOpenKey( &key, KEY_READ, &attr ); + ok( status == STATUS_SUCCESS || broken(status == STATUS_OBJECT_NAME_NOT_FOUND) /* win8 */, + "NtOpenKey failed: 0x%08x\n", status ); + if (!status) pNtClose( key ); + pRtlFreeUnicodeString( &str ); + if (!pNtOpenKeyEx) { win_skip("NtOpenKeyEx not available\n"); @@ -630,13 +638,21 @@ static void test_NtCreateKey(void) status = pNtCreateKey( &subkey, am, &attr, 0, 0, 0, 0 ); ok( status == STATUS_SUCCESS, "NtCreateKey failed: 0x%08x\n", status ); - if (!status) pNtClose( subkey ); + pNtClose( subkey ); pRtlFreeUnicodeString( &str ); pRtlCreateUnicodeStringFromAsciiz( &str, "\\REGISTRY\\MACHINE\\SOFTWARE\\CLASSES" ); status = pNtCreateKey( &subkey, am, &attr, 0, 0, 0, 0 ); ok( status == STATUS_SUCCESS, "NtCreateKey failed: 0x%08x\n", status ); + pNtClose( subkey ); + pRtlFreeUnicodeString( &str ); + + pRtlCreateUnicodeStringFromAsciiz( &str, "\\REGISTRY\\Machine\\Software\\Classes" ); + str.Length += sizeof(WCHAR); /* include terminating \0 in the length */ + status = pNtCreateKey( &subkey, am, &attr, 0, 0, 0, 0 ); + ok( status == STATUS_SUCCESS || broken(status == STATUS_OBJECT_PATH_NOT_FOUND) /* win8 */, + "NtCreateKey failed: 0x%08x\n", status ); if (!status) pNtClose( subkey ); pRtlFreeUnicodeString( &str ); diff --git a/server/registry.c b/server/registry.c index 2a2bc9a4e68..6a03a41a9de 100644 --- a/server/registry.c +++ b/server/registry.c @@ -477,6 +477,8 @@ static inline void get_req_path( struct unicode_str *str ) { str->str = get_req_data(); str->len = (get_req_data_size() / sizeof(WCHAR)) * sizeof(WCHAR); + + while (str->len && !str->str[str->len / sizeof(WCHAR) - 1]) str->len -= sizeof(WCHAR); } /* return the next token in a given path */ -- 2.34.1

2 3