June 2019 - Wine-Devel - list.winehq.org

[PATCH 1/6] qcap/avimux: Use strmbase_sink_init() and heap_alloc_zero() in create_input_pin().
by Zebediah Figura June 13, 2019

June 13, 2019

Signed-off-by: Zebediah Figura <z.figura12(a)gmail.com> --- dlls/qcap/avimux.c | 51 +++++++++++++++++++++++----------------------- 1 file changed, 26 insertions(+), 25 deletions(-) diff --git a/dlls/qcap/avimux.c b/dlls/qcap/avimux.c index d289d2b9fed..9dcff5074e6 100644 --- a/dlls/qcap/avimux.c +++ b/dlls/qcap/avimux.c @@ -139,7 +139,8 @@ static void avi_mux_destroy(BaseFilter *iface) IPin_Disconnect(&filter->in[i]->pin.pin.IPin_iface); IMemAllocator_Release(filter->in[i]->samples_allocator); filter->in[i]->samples_allocator = NULL; - BaseInputPinImpl_Release(&filter->in[i]->pin.pin.IPin_iface); + strmbase_sink_cleanup(&filter->in[i]->pin); + heap_free(filter->in[i]); } heap_free(filter->idx1); @@ -2214,6 +2215,7 @@ static const IQualityControlVtbl AviMuxIn_QualityControlVtbl = { static HRESULT create_input_pin(AviMux *avimux) { static const WCHAR name[] = {'I','n','p','u','t',' ','0','0',0}; + AviMuxIn *object; PIN_INFO info; HRESULT hr; @@ -2226,40 +2228,39 @@ static HRESULT create_input_pin(AviMux *avimux) info.achName[7] = '0' + (avimux->input_pin_no+1) % 10; info.achName[6] = '0' + (avimux->input_pin_no+1) / 10; - hr = BaseInputPin_Construct(&AviMuxIn_PinVtbl, sizeof(AviMuxIn), &info, - &AviMuxIn_BaseInputFuncTable, &avimux->filter.csFilter, NULL, (IPin**)&avimux->in[avimux->input_pin_no]); - if(FAILED(hr)) - return hr; - avimux->in[avimux->input_pin_no]->pin.IMemInputPin_iface.lpVtbl = &AviMuxIn_MemInputPinVtbl; - avimux->in[avimux->input_pin_no]->IAMStreamControl_iface.lpVtbl = &AviMuxIn_AMStreamControlVtbl; - avimux->in[avimux->input_pin_no]->IPropertyBag_iface.lpVtbl = &AviMuxIn_PropertyBagVtbl; - avimux->in[avimux->input_pin_no]->IQualityControl_iface.lpVtbl = &AviMuxIn_QualityControlVtbl; + if (!(object = heap_alloc_zero(sizeof(*object)))) + return E_OUTOFMEMORY; + + strmbase_sink_init(&object->pin, &AviMuxIn_PinVtbl, &info, + &AviMuxIn_BaseInputFuncTable, &avimux->filter.csFilter, NULL); + object->pin.IMemInputPin_iface.lpVtbl = &AviMuxIn_MemInputPinVtbl; + object->IAMStreamControl_iface.lpVtbl = &AviMuxIn_AMStreamControlVtbl; + object->IPropertyBag_iface.lpVtbl = &AviMuxIn_PropertyBagVtbl; + object->IQualityControl_iface.lpVtbl = &AviMuxIn_QualityControlVtbl; - avimux->in[avimux->input_pin_no]->samples_head = NULL; hr = CoCreateInstance(&CLSID_MemoryAllocator, NULL, CLSCTX_INPROC_SERVER, - &IID_IMemAllocator, (void**)&avimux->in[avimux->input_pin_no]->samples_allocator); - if(FAILED(hr)) { - BaseInputPinImpl_Release(&avimux->in[avimux->input_pin_no]->pin.pin.IPin_iface); + &IID_IMemAllocator, (void **)&object->samples_allocator); + if (FAILED(hr)) + { + strmbase_sink_cleanup(&object->pin); + heap_free(object); return hr; } hr = CoCreateInstance(&CLSID_MemoryAllocator, NULL, CLSCTX_INPROC_SERVER, - &IID_IMemAllocator, (void**)&avimux->in[avimux->input_pin_no]->pin.pAllocator); - if(FAILED(hr)) { - IMemAllocator_Release(avimux->in[avimux->input_pin_no]->samples_allocator); - BaseInputPinImpl_Release(&avimux->in[avimux->input_pin_no]->pin.pin.IPin_iface); + &IID_IMemAllocator, (void **)&object->pin.pAllocator); + if (FAILED(hr)) + { + IMemAllocator_Release(object->samples_allocator); + strmbase_sink_cleanup(&object->pin); + heap_free(object); return hr; } - avimux->in[avimux->input_pin_no]->stream_time = 0; - memset(&avimux->in[avimux->input_pin_no]->strh, 0, sizeof(avimux->in[avimux->input_pin_no]->strh)); - avimux->in[avimux->input_pin_no]->strf = NULL; - memset(&avimux->in[avimux->input_pin_no]->indx_data, 0, sizeof(avimux->in[avimux->input_pin_no]->indx_data)); - memset(&avimux->in[avimux->input_pin_no]->ix_data, 0, sizeof(avimux->in[avimux->input_pin_no]->ix_data)); - avimux->in[avimux->input_pin_no]->indx = (AVISUPERINDEX*)&avimux->in[avimux->input_pin_no]->indx_data; - avimux->in[avimux->input_pin_no]->ix = (AVISTDINDEX*)avimux->in[avimux->input_pin_no]->ix_data; + object->indx = (AVISUPERINDEX *)&object->indx_data; + object->ix = (AVISTDINDEX *)object->ix_data; - avimux->input_pin_no++; + avimux->in[avimux->input_pin_no++] = object; return S_OK; } -- 2.20.1

1 5

[PATCH 1/2] widl: Mark unions, user types, and ranges as IsSimpleRef.
by Zebediah Figura June 13, 2019

June 13, 2019

This fixes a regression introduced by 02976b655a5d9ad8ee25dbfb2051acb5b7431381. That patch should only have affected arrays. Signed-off-by: Zebediah Figura <z.figura12(a)gmail.com> --- tools/widl/typegen.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/tools/widl/typegen.c b/tools/widl/typegen.c index f320f57789a..fbc9f3cc5bd 100644 --- a/tools/widl/typegen.c +++ b/tools/widl/typegen.c @@ -1064,7 +1064,8 @@ static unsigned char get_parameter_fc( const var_t *var, int is_return, unsigned case TGT_UNION: case TGT_USER_TYPE: case TGT_RANGE: - *flags |= MustFree; + *flags |= MustFree | IsSimpleRef; + *typestring_offset = ref->typestring_offset; if (!is_in && is_out) server_size = type_memsize( ref ); break; case TGT_ARRAY: -- 2.20.1

1 1

[PATCH v2 1/3] ntdll: Implement RtlCreateUserStack() and RtlFreeUserStack().
by Zebediah Figura June 13, 2019

June 13, 2019

Signed-off-by: Zebediah Figura <zfigura(a)codeweavers.com> --- dlls/ntdll/loader.c | 7 ++++++- dlls/ntdll/ntdll.spec | 3 ++- dlls/ntdll/ntdll_misc.h | 2 +- dlls/ntdll/thread.c | 7 ++++++- dlls/ntdll/virtual.c | 46 +++++++++++++++++++++++++++++++++++++---- include/winternl.h | 12 ++++++----- 6 files changed, 64 insertions(+), 13 deletions(-) diff --git a/dlls/ntdll/loader.c b/dlls/ntdll/loader.c index 3f422a9c0bee..3d747edaba3b 100644 --- a/dlls/ntdll/loader.c +++ b/dlls/ntdll/loader.c @@ -3787,6 +3787,7 @@ void __wine_process_init(void) ANSI_STRING func_name; UNICODE_STRING nt_name; void * (CDECL *init_func)(void); + INITIAL_TEB stack; thread_init(); @@ -3838,11 +3839,15 @@ void __wine_process_init(void) RemoveEntryList( &wm->ldr.InMemoryOrderModuleList ); InsertHeadList( &NtCurrentTeb()->Peb->LdrData->InMemoryOrderModuleList, &wm->ldr.InMemoryOrderModuleList ); - if ((status = virtual_alloc_thread_stack( NtCurrentTeb(), 0, 0, NULL )) != STATUS_SUCCESS) + if ((status = virtual_alloc_thread_stack( &stack, 0, 0, NULL )) != STATUS_SUCCESS) { ERR( "Main exe initialization for %s failed, status %x\n", debugstr_w(wm->ldr.FullDllName.Buffer), status ); NtTerminateProcess( GetCurrentProcess(), status ); } + NtCurrentTeb()->Tib.StackBase = stack.StackBase; + NtCurrentTeb()->Tib.StackLimit = stack.StackLimit; + NtCurrentTeb()->DeallocationStack = stack.DeallocationStack; + server_init_process_done(); } diff --git a/dlls/ntdll/ntdll.spec b/dlls/ntdll/ntdll.spec index aeb9735ba100..e0def228de07 100644 --- a/dlls/ntdll/ntdll.spec +++ b/dlls/ntdll/ntdll.spec @@ -526,6 +526,7 @@ @ stdcall RtlCreateUnicodeStringFromAsciiz(ptr str) @ stdcall RtlCreateUserProcess(ptr long ptr ptr ptr long long long long ptr) @ stub RtlCreateUserSecurityObject +@ stdcall RtlCreateUserStack(long long long long long ptr) @ stdcall RtlCreateUserThread(long ptr long ptr long long ptr ptr ptr ptr) @ stub RtlCustomCPToUnicodeN @ stub RtlCutoverTimeToSystemTime @@ -644,7 +645,7 @@ @ stdcall RtlFreeSid (ptr) @ stdcall RtlFreeThreadActivationContextStack() @ stdcall RtlFreeUnicodeString(ptr) -@ stub RtlFreeUserThreadStack +@ stdcall RtlFreeUserStack(ptr) @ stdcall RtlGUIDFromString(ptr ptr) @ stub RtlGenerate8dot3Name @ stdcall RtlGetAce(ptr long ptr) diff --git a/dlls/ntdll/ntdll_misc.h b/dlls/ntdll/ntdll_misc.h index 3463ebd38adc..d1b895cd364a 100644 --- a/dlls/ntdll/ntdll_misc.h +++ b/dlls/ntdll/ntdll_misc.h @@ -173,7 +173,7 @@ extern NTSTATUS virtual_map_section( HANDLE handle, PVOID *addr_ptr, ULONG zero_ pe_image_info_t *image_info ) DECLSPEC_HIDDEN; extern void virtual_get_system_info( SYSTEM_BASIC_INFORMATION *info ) DECLSPEC_HIDDEN; extern NTSTATUS virtual_create_builtin_view( void *base ) DECLSPEC_HIDDEN; -extern NTSTATUS virtual_alloc_thread_stack( TEB *teb, SIZE_T reserve_size, +extern NTSTATUS virtual_alloc_thread_stack( INITIAL_TEB *stack, SIZE_T reserve_size, SIZE_T commit_size, SIZE_T *pthread_size ) DECLSPEC_HIDDEN; extern void virtual_clear_thread_stack( void *stack_end ) DECLSPEC_HIDDEN; extern BOOL virtual_handle_stack_fault( void *addr ) DECLSPEC_HIDDEN; diff --git a/dlls/ntdll/thread.c b/dlls/ntdll/thread.c index 46de839400d2..f612d135273f 100644 --- a/dlls/ntdll/thread.c +++ b/dlls/ntdll/thread.c @@ -416,6 +416,7 @@ NTSTATUS WINAPI RtlCreateUserThread( HANDLE process, SECURITY_DESCRIPTOR *descr, SIZE_T extra_stack = PTHREAD_STACK_MIN; data_size_t len = 0; struct object_attributes *objattr = NULL; + INITIAL_TEB stack; if (process != NtCurrentProcess()) { @@ -507,9 +508,13 @@ NTSTATUS WINAPI RtlCreateUserThread( HANDLE process, SECURITY_DESCRIPTOR *descr, info->entry_point = start; info->entry_arg = param; - if ((status = virtual_alloc_thread_stack( teb, stack_reserve, stack_commit, &extra_stack ))) + if ((status = virtual_alloc_thread_stack( &stack, stack_reserve, stack_commit, &extra_stack ))) goto error; + teb->Tib.StackBase = stack.StackBase; + teb->Tib.StackLimit = stack.StackLimit; + teb->DeallocationStack = stack.DeallocationStack; + thread_data = (struct ntdll_thread_data *)&teb->GdiTebBatch; thread_data->request_fd = request_pipe[1]; thread_data->reply_fd = -1; diff --git a/dlls/ntdll/virtual.c b/dlls/ntdll/virtual.c index 78973a8cda43..a3ea195c917b 100644 --- a/dlls/ntdll/virtual.c +++ b/dlls/ntdll/virtual.c @@ -1924,7 +1924,7 @@ NTSTATUS virtual_create_builtin_view( void *module ) /*********************************************************************** * virtual_alloc_thread_stack */ -NTSTATUS virtual_alloc_thread_stack( TEB *teb, SIZE_T reserve_size, SIZE_T commit_size, SIZE_T *pthread_size ) +NTSTATUS virtual_alloc_thread_stack( INITIAL_TEB *stack, SIZE_T reserve_size, SIZE_T commit_size, SIZE_T *pthread_size ) { struct file_view *view; NTSTATUS status; @@ -1978,9 +1978,11 @@ NTSTATUS virtual_alloc_thread_stack( TEB *teb, SIZE_T reserve_size, SIZE_T commi } /* note: limit is lower than base since the stack grows down */ - teb->DeallocationStack = view->base; - teb->Tib.StackBase = (char *)view->base + view->size; - teb->Tib.StackLimit = (char *)view->base + 2 * page_size; + stack->OldStackBase = 0; + stack->OldStackLimit = 0; + stack->DeallocationStack = view->base; + stack->StackBase = (char *)view->base + view->size; + stack->StackLimit = (char *)view->base + 2 * page_size; done: server_leave_uninterrupted_section( &csVirtual, &sigset ); return status; @@ -2001,6 +2003,42 @@ void virtual_clear_thread_stack( void *stack_end ) if (force_exec_prot) mprotect( stack, size, PROT_READ | PROT_WRITE | PROT_EXEC ); } +/********************************************************************** + * RtlCreateUserStack (NTDLL.@) + */ +NTSTATUS WINAPI RtlCreateUserStack( SIZE_T commit, SIZE_T reserve, ULONG zero_bits, + SIZE_T commit_align, SIZE_T reserve_align, INITIAL_TEB *stack ) +{ + TRACE("commit %#lx, reserve %#lx, zero_bits %u, commit_align %#lx, reserve_align %#lx, stack %p\n", + commit, reserve, zero_bits, commit_align, reserve_align, stack); + + if (!commit_align || !reserve_align) + return STATUS_INVALID_PARAMETER; + + if (!commit || !reserve) + { + IMAGE_NT_HEADERS *nt = RtlImageNtHeader( NtCurrentTeb()->Peb->ImageBaseAddress ); + if (!reserve) reserve = nt->OptionalHeader.SizeOfStackReserve; + if (!commit) commit = nt->OptionalHeader.SizeOfStackCommit; + } + + reserve = (reserve + reserve_align - 1) & ~(reserve_align - 1); + commit = (commit + commit_align - 1) & ~(commit_align - 1); + + return virtual_alloc_thread_stack( stack, reserve, commit, NULL ); +} + +/********************************************************************** + * RtlFreeUserStack (NTDLL.@) + */ +void WINAPI RtlFreeUserStack( void *stack ) +{ + SIZE_T size = 0; + + TRACE("stack %p\n", stack); + + NtFreeVirtualMemory( NtCurrentProcess(), &stack, &size, MEM_RELEASE ); +} /*********************************************************************** * virtual_handle_fault diff --git a/include/winternl.h b/include/winternl.h index e7f89b005922..b87aa2298c6b 100644 --- a/include/winternl.h +++ b/include/winternl.h @@ -2104,11 +2104,11 @@ typedef struct _DIRECTORY_BASIC_INFORMATION { } DIRECTORY_BASIC_INFORMATION, *PDIRECTORY_BASIC_INFORMATION; typedef struct _INITIAL_TEB { - PVOID StackBase; - PVOID StackLimit; - PVOID StackCommit; - PVOID StackCommitMax; - PVOID StackReserved; + void *OldStackBase; + void *OldStackLimit; + void *StackBase; + void *StackLimit; + void *DeallocationStack; } INITIAL_TEB, *PINITIAL_TEB; typedef enum _PORT_INFORMATION_CLASS { @@ -2618,6 +2618,7 @@ NTSYSAPI BOOLEAN WINAPI RtlCreateUnicodeString(PUNICODE_STRING,LPCWSTR); NTSYSAPI BOOLEAN WINAPI RtlCreateUnicodeStringFromAsciiz(PUNICODE_STRING,LPCSTR); NTSYSAPI NTSTATUS WINAPI RtlCreateUserProcess(UNICODE_STRING*,ULONG,RTL_USER_PROCESS_PARAMETERS*,SECURITY_DESCRIPTOR*,SECURITY_DESCRIPTOR*,HANDLE,BOOLEAN,HANDLE,HANDLE,RTL_USER_PROCESS_INFORMATION*); NTSYSAPI NTSTATUS WINAPI RtlCreateUserThread(HANDLE,SECURITY_DESCRIPTOR*,BOOLEAN,PVOID,SIZE_T,SIZE_T,PRTL_THREAD_START_ROUTINE,void*,HANDLE*,CLIENT_ID*); +NTSYSAPI NTSTATUS WINAPI RtlCreateUserStack(SIZE_T,SIZE_T,ULONG,SIZE_T,SIZE_T,INITIAL_TEB*); NTSYSAPI void WINAPI RtlDeactivateActivationContext(DWORD,ULONG_PTR); NTSYSAPI PVOID WINAPI RtlDecodePointer(PVOID); NTSYSAPI NTSTATUS WINAPI RtlDecompressBuffer(USHORT,PUCHAR,ULONG,PUCHAR,ULONG,PULONG); @@ -2691,6 +2692,7 @@ NTSYSAPI void WINAPI RtlFreeOemString(POEM_STRING); NTSYSAPI DWORD WINAPI RtlFreeSid(PSID); NTSYSAPI void WINAPI RtlFreeThreadActivationContextStack(void); NTSYSAPI void WINAPI RtlFreeUnicodeString(PUNICODE_STRING); +NTSYSAPI void WINAPI RtlFreeUserStack(void*); NTSYSAPI NTSTATUS WINAPI RtlGetAce(PACL,DWORD,LPVOID *); NTSYSAPI NTSTATUS WINAPI RtlGetActiveActivationContext(HANDLE*); NTSYSAPI NTSTATUS WINAPI RtlGetCompressionWorkSpaceSize(USHORT,PULONG,PULONG); -- 2.20.1

2 3

[PATCH v5] ntdll/directory: Add support for EXT4 case folding per directory
by Gabriel Ivăncescu June 13, 2019

June 13, 2019

This adds support for checking the case-insensitive attribute on ext4 with newer kernels so that Wine can rely on it for performance. Checking for the EXT4_CASEFOLD_FL attribute involves an ioctl, which operates on file descriptors, while all the former checks operated directly on the dir pathname (e.g. statfs). Obviously, it's best to avoid looking up the directory multiple times (also for correctness, so it refers to the same dir). So in the case that we *do* have a file descriptor, then use it everywhere, with e.g. fstatfs. Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=47099 Signed-off-by: Gabriel Ivăncescu <gabrielopcode(a)gmail.com> --- v5: Only use the file descriptor on Linux. The second patch is not resent since it's in staging for users to test out first before going upstream. dlls/ntdll/directory.c | 44 +++++++++++++++++++++++++++--------------- 1 file changed, 28 insertions(+), 16 deletions(-) diff --git a/dlls/ntdll/directory.c b/dlls/ntdll/directory.c index bbdbbe9..9b4e61b 100644 --- a/dlls/ntdll/directory.c +++ b/dlls/ntdll/directory.c @@ -115,6 +115,8 @@ WINE_DEFAULT_DEBUG_CHANNEL(file); /* just in case... */ #undef VFAT_IOCTL_READDIR_BOTH +#undef EXT2_IOC_GETFLAGS +#undef EXT4_CASEFOLD_FL #ifdef linux @@ -130,6 +132,12 @@ typedef struct /* Define the VFAT ioctl to get both short and long file names */ #define VFAT_IOCTL_READDIR_BOTH _IOR('r', 1, KERNEL_DIRENT [2] ) +/* Define the ext2 ioctl for handling extra attributes */ +#define EXT2_IOC_GETFLAGS _IOR('f', 1, long) + +/* Case-insensitivity attribute */ +#define EXT4_CASEFOLD_FL 0x40000000 + #ifndef O_DIRECTORY # define O_DIRECTORY 0200000 /* must be directory */ #endif @@ -1113,7 +1121,7 @@ static int get_dir_case_sensitivity_attr( const char *dir ) * get_dir_case_sensitivity_stat * * Checks if the volume containing the specified directory is case - * sensitive or not. Uses statfs(2) or statvfs(2). + * sensitive or not. Uses (f)statfs(2), statvfs(2), fstatat(2), or ioctl(2). */ static BOOLEAN get_dir_case_sensitivity_stat( const char *dir ) { @@ -1165,14 +1173,23 @@ static BOOLEAN get_dir_case_sensitivity_stat( const char *dir ) return FALSE; #elif defined(__linux__) + BOOLEAN sens = FALSE; struct statfs stfs; struct stat st; - char *cifile; + int fd, flags; + + if ((fd = open( dir, O_RDONLY | O_NONBLOCK | O_LARGEFILE )) == -1) + return FALSE; + + if (ioctl( fd, EXT2_IOC_GETFLAGS, &flags ) != -1 && (flags & EXT4_CASEFOLD_FL)) + goto end; /* Only assume CIOPFS is case insensitive. */ - if (statfs( dir, &stfs ) == -1) return FALSE; + if (fstatfs( fd, &stfs ) == -1) goto end; + + sens = TRUE; if (stfs.f_type != 0x65735546 /* FUSE_SUPER_MAGIC */) - return TRUE; + goto end; /* Normally, we'd have to parse the mtab to find out exactly what * kind of FUSE FS this is. But, someone on wine-devel suggested * a shortcut. We'll stat a special file in the directory. If it's @@ -1180,17 +1197,12 @@ static BOOLEAN get_dir_case_sensitivity_stat( const char *dir ) * This will break if somebody puts a file named ".ciopfs" in a non- * CIOPFS directory. */ - cifile = RtlAllocateHeap( GetProcessHeap(), 0, strlen( dir )+sizeof("/.ciopfs") ); - if (!cifile) return TRUE; - strcpy( cifile, dir ); - strcat( cifile, "/.ciopfs" ); - if (stat( cifile, &st ) == 0) - { - RtlFreeHeap( GetProcessHeap(), 0, cifile ); - return FALSE; - } - RtlFreeHeap( GetProcessHeap(), 0, cifile ); - return TRUE; + if (fstatat( fd, ".ciopfs", &st, AT_NO_AUTOMOUNT ) == 0) + sens = FALSE; + +end: + close( fd ); + return sens; #else return TRUE; #endif @@ -1201,7 +1213,7 @@ static BOOLEAN get_dir_case_sensitivity_stat( const char *dir ) * get_dir_case_sensitivity * * Checks if the volume containing the specified directory is case - * sensitive or not. Uses statfs(2) or statvfs(2). + * sensitive or not. Uses multiple methods, depending on platform. */ static BOOLEAN get_dir_case_sensitivity( const char *dir ) { -- 2.21.0

2 2

[PATCH v4 1/3] wined3d: Use 3 component norm for 'nrm' opcode in GLSL backend.
by Paul Gofman June 13, 2019

June 13, 2019

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=8848 Signed-off-by: Paul Gofman <gofmanp(a)gmail.com> --- v2: no changes v3: - Split off w component handling for zero norm in a separate patch. v4: no changes dlls/wined3d/glsl_shader.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) diff --git a/dlls/wined3d/glsl_shader.c b/dlls/wined3d/glsl_shader.c index c934528981..f2fa18dba8 100644 --- a/dlls/wined3d/glsl_shader.c +++ b/dlls/wined3d/glsl_shader.c @@ -4260,8 +4260,12 @@ static void shader_glsl_nrm(const struct wined3d_shader_instruction *ins) mask_size = shader_glsl_get_write_mask_size(write_mask); shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src_param); - shader_addline(buffer, "tmp0.x = dot(%s, %s);\n", - src_param.param_str, src_param.param_str); + if (mask_size > 3) + shader_addline(buffer, "tmp0.x = dot(vec3(%s), vec3(%s));\n", + src_param.param_str, src_param.param_str); + else + shader_addline(buffer, "tmp0.x = dot(%s, %s);\n", + src_param.param_str, src_param.param_str); shader_glsl_append_dst(buffer, ins); if (mask_size > 1) -- 2.21.0

3 6

[PATCH vkd3d 4/4] vkd3d: Use FIXME_ONCE() for AntialiasedLineEnable.
by Józef Kucia June 13, 2019

June 13, 2019

From: Józef Kucia <jkucia(a)codeweavers.com> Signed-off-by: Józef Kucia <jkucia(a)codeweavers.com> --- libs/vkd3d/state.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/libs/vkd3d/state.c b/libs/vkd3d/state.c index 64be72d5af9c..794668041fa2 100644 --- a/libs/vkd3d/state.c +++ b/libs/vkd3d/state.c @@ -1722,7 +1722,7 @@ static void rs_desc_from_d3d12(VkPipelineRasterizationStateCreateInfo *vk_desc, if (d3d12_desc->MultisampleEnable) FIXME_ONCE("Ignoring MultisampleEnable %#x.\n", d3d12_desc->MultisampleEnable); if (d3d12_desc->AntialiasedLineEnable) - FIXME("Ignoring AntialiasedLineEnable %#x.\n", d3d12_desc->AntialiasedLineEnable); + FIXME_ONCE("Ignoring AntialiasedLineEnable %#x.\n", d3d12_desc->AntialiasedLineEnable); if (d3d12_desc->ForcedSampleCount) FIXME("Ignoring ForcedSampleCount %#x.\n", d3d12_desc->ForcedSampleCount); if (d3d12_desc->ConservativeRaster) -- 2.21.0

2 1

[PATCH vkd3d 3/4] vkd3d: Avoid busy-waiting in fence worker thread.
by Józef Kucia June 13, 2019

June 13, 2019

From: Józef Kucia <jkucia(a)codeweavers.com> This might delay updating a d3d12_fence when a fence enqueued later than other fences is signaled before them. On the other hand, it significantly reduces CPU usage. I haven't found a program negatively impacted by this change so far. Signed-off-by: Józef Kucia <jkucia(a)codeweavers.com> --- libs/vkd3d/command.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/libs/vkd3d/command.c b/libs/vkd3d/command.c index 7634e71b32ee..75556f5335e3 100644 --- a/libs/vkd3d/command.c +++ b/libs/vkd3d/command.c @@ -362,7 +362,7 @@ static void vkd3d_wait_for_gpu_fences(struct vkd3d_fence_worker *worker) return; vr = VK_CALL(vkWaitForFences(device->vk_device, - worker->fence_count, worker->vk_fences, VK_FALSE, 0)); + worker->fence_count, worker->vk_fences, VK_FALSE, ~(uint64_t)0)); if (vr == VK_TIMEOUT) return; if (vr != VK_SUCCESS) -- 2.21.0

2 1

[PATCH vkd3d 2/4] vkd3d: Avoid holding mutex while waiting for fences.
by Józef Kucia June 13, 2019

June 13, 2019

From: Józef Kucia <jkucia(a)codeweavers.com> We maintain separate arrays for enqueued fences and fences owned by the fence worker thread. Signed-off-by: Józef Kucia <jkucia(a)codeweavers.com> --- include/private/vkd3d_common.h | 10 +- libs/vkd3d/command.c | 164 +++++++++++++++++++++++---------- libs/vkd3d/vkd3d_private.h | 15 +++ 3 files changed, 137 insertions(+), 52 deletions(-) diff --git a/include/private/vkd3d_common.h b/include/private/vkd3d_common.h index 244209243594..16c31e633ab1 100644 --- a/include/private/vkd3d_common.h +++ b/include/private/vkd3d_common.h @@ -118,7 +118,7 @@ static inline LONG InterlockedIncrement(LONG volatile *x) return __sync_add_and_fetch(x, 1); } # else -# error "InterlockedIncrement not implemented for this platform" +# error "InterlockedIncrement() not implemented for this platform" # endif /* HAVE_SYNC_ADD_AND_FETCH */ # if HAVE_SYNC_SUB_AND_FETCH @@ -127,8 +127,14 @@ static inline LONG InterlockedDecrement(LONG volatile *x) return __sync_sub_and_fetch(x, 1); } # else -# error "InterlockedDecrement not implemented for this platform" +# error "InterlockedDecrement() not implemented for this platform" # endif #endif /* _WIN32 */ +#if HAVE_SYNC_ADD_AND_FETCH +# define atomic_add_fetch(ptr, val) __sync_add_and_fetch(ptr, val) +#else +# error "atomic_add_fetch() not implemented for this platform" +#endif /* HAVE_SYNC_ADD_AND_FETCH */ + #endif /* __VKD3D_COMMON_H */ diff --git a/libs/vkd3d/command.c b/libs/vkd3d/command.c index 25c8005dc445..7634e71b32ee 100644 --- a/libs/vkd3d/command.c +++ b/libs/vkd3d/command.c @@ -257,27 +257,22 @@ static HRESULT vkd3d_enqueue_gpu_fence(struct vkd3d_fence_worker *worker, return hresult_from_errno(rc); } - if (!vkd3d_array_reserve((void **)&worker->vk_fences, &worker->vk_fences_size, - worker->fence_count + 1, sizeof(*worker->vk_fences))) - { - ERR("Failed to add GPU fence.\n"); - pthread_mutex_unlock(&worker->mutex); - return E_OUTOFMEMORY; - } - if (!vkd3d_array_reserve((void **)&worker->fences, &worker->fences_size, - worker->fence_count + 1, sizeof(*worker->fences))) + 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 fence.\n"); pthread_mutex_unlock(&worker->mutex); return E_OUTOFMEMORY; } - worker->vk_fences[worker->fence_count] = vk_fence; - worker->fences[worker->fence_count].fence = fence; - worker->fences[worker->fence_count].value = value; - worker->fences[worker->fence_count].queue = queue; - worker->fences[worker->fence_count].queue_sequence_number = queue_sequence_number; - ++worker->fence_count; + worker->enqueued_fences[worker->enqueued_fence_count].vk_fence = vk_fence; + worker->enqueued_fences[worker->enqueued_fence_count].fence = fence; + worker->enqueued_fences[worker->enqueued_fence_count].value = value; + worker->enqueued_fences[worker->enqueued_fence_count].queue = queue; + worker->enqueued_fences[worker->enqueued_fence_count].queue_sequence_number = queue_sequence_number; + ++worker->enqueued_fence_count; + + InterlockedIncrement(&fence->pending_worker_operation_count); pthread_cond_signal(&worker->cond); pthread_mutex_unlock(&worker->mutex); @@ -287,37 +282,73 @@ static HRESULT vkd3d_enqueue_gpu_fence(struct vkd3d_fence_worker *worker, static void vkd3d_fence_worker_remove_fence(struct vkd3d_fence_worker *worker, struct d3d12_fence *fence) { - struct d3d12_device *device = worker->device; - const struct vkd3d_vk_device_procs *vk_procs = &device->vk_procs; - unsigned int i, j; + LONG count; int rc; + if (!(count = atomic_add_fetch(&fence->pending_worker_operation_count, 0))) + return; + + WARN("Waiting for %u pending fence operations (fence %p).\n", count, fence); + if ((rc = pthread_mutex_lock(&worker->mutex))) { ERR("Failed to lock mutex, error %d.\n", rc); return; } - for (i = 0, j = 0; i < worker->fence_count; ++i) + while ((count = atomic_add_fetch(&fence->pending_worker_operation_count, 0))) { - if (worker->fences[i].fence == fence) - { - VK_CALL(vkDestroyFence(device->vk_device, worker->vk_fences[i], NULL)); - continue; - } + TRACE("Still waiting for %u pending fence operations (fence %p).\n", count, fence); - if (i != j) - { - worker->vk_fences[j] = worker->vk_fences[i]; - worker->fences[j] = worker->fences[i]; - } - ++j; + worker->pending_fence_destruction = true; + pthread_cond_signal(&worker->cond); + + pthread_cond_wait(&worker->fence_destruction_cond, &worker->mutex); } - worker->fence_count = j; + + TRACE("Removed fence %p.\n", fence); pthread_mutex_unlock(&worker->mutex); } +static void vkd3d_fence_worker_move_enqueued_fences_locked(struct vkd3d_fence_worker *worker) +{ + unsigned int i; + size_t count; + bool ret; + + if (!worker->enqueued_fence_count) + return; + + 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)); + ret &= vkd3d_array_reserve((void **)&worker->fences, &worker->fences_size, + count, sizeof(*worker->fences)); + if (!ret) + { + ERR("Failed to reserve memory.\n"); + return; + } + + for (i = 0; i < worker->enqueued_fence_count; ++i) + { + struct vkd3d_enqueued_fence *current = &worker->enqueued_fences[i]; + + worker->vk_fences[worker->fence_count] = current->vk_fence; + + worker->fences[worker->fence_count].fence = current->fence; + worker->fences[worker->fence_count].value = current->value; + worker->fences[worker->fence_count].queue = current->queue; + worker->fences[worker->fence_count].queue_sequence_number = current->queue_sequence_number; + + ++worker->fence_count; + } + assert(worker->fence_count == count); + worker->enqueued_fence_count = 0; +} + static void vkd3d_wait_for_gpu_fences(struct vkd3d_fence_worker *worker) { struct d3d12_device *device = worker->device; @@ -351,6 +382,8 @@ static void vkd3d_wait_for_gpu_fences(struct vkd3d_fence_worker *worker) if (FAILED(hr = d3d12_fence_signal(current->fence, current->value, vk_fence))) ERR("Failed to signal D3D12 fence, hr %#x.\n", hr); + InterlockedDecrement(&current->fence->pending_worker_operation_count); + vkd3d_queue_update_sequence_number(current->queue, current->queue_sequence_number, device); continue; } @@ -373,35 +406,48 @@ static void *vkd3d_fence_worker_main(void *arg) struct vkd3d_fence_worker *worker = arg; int rc; - vkd3d_set_thread_name("vkd3d_worker"); + vkd3d_set_thread_name("vkd3d_fence"); for (;;) { - if ((rc = pthread_mutex_lock(&worker->mutex))) - { - ERR("Failed to lock mutex, error %d.\n", rc); - return NULL; - } + vkd3d_wait_for_gpu_fences(worker); - if (worker->should_exit && !worker->fence_count) + if (!worker->fence_count || atomic_add_fetch(&worker->enqueued_fence_count, 0)) { - pthread_mutex_unlock(&worker->mutex); - break; - } + if ((rc = pthread_mutex_lock(&worker->mutex))) + { + ERR("Failed to lock mutex, error %d.\n", rc); + break; + } - if (!worker->fence_count) - { - if ((rc = pthread_cond_wait(&worker->cond, &worker->mutex))) + if (worker->pending_fence_destruction) { - ERR("Failed to wait on condition variable, error %d.\n", rc); - pthread_mutex_unlock(&worker->mutex); - return NULL; + pthread_cond_broadcast(&worker->fence_destruction_cond); + worker->pending_fence_destruction = false; } - } - vkd3d_wait_for_gpu_fences(worker); + if (worker->enqueued_fence_count) + { + vkd3d_fence_worker_move_enqueued_fences_locked(worker); + } + else + { + if (worker->should_exit) + { + pthread_mutex_unlock(&worker->mutex); + break; + } - pthread_mutex_unlock(&worker->mutex); + if ((rc = pthread_cond_wait(&worker->cond, &worker->mutex))) + { + ERR("Failed to wait on condition variable, error %d.\n", rc); + pthread_mutex_unlock(&worker->mutex); + break; + } + } + + pthread_mutex_unlock(&worker->mutex); + } } return NULL; @@ -416,8 +462,13 @@ HRESULT vkd3d_fence_worker_start(struct vkd3d_fence_worker *worker, TRACE("worker %p.\n", worker); worker->should_exit = false; + worker->pending_fence_destruction = false; worker->device = device; + worker->enqueued_fence_count = 0; + worker->enqueued_fences = NULL; + worker->enqueued_fences_size = 0; + worker->fence_count = 0; worker->vk_fences = NULL; @@ -438,11 +489,20 @@ HRESULT vkd3d_fence_worker_start(struct vkd3d_fence_worker *worker, return hresult_from_errno(rc); } + if ((rc = pthread_cond_init(&worker->fence_destruction_cond, NULL))) + { + ERR("Failed to initialize condition variable, error %d.\n", rc); + pthread_mutex_destroy(&worker->mutex); + pthread_cond_destroy(&worker->cond); + return hresult_from_errno(rc); + } + if (FAILED(hr = vkd3d_create_thread(device->vkd3d_instance, vkd3d_fence_worker_main, worker, &worker->thread))) { pthread_mutex_destroy(&worker->mutex); pthread_cond_destroy(&worker->cond); + pthread_cond_destroy(&worker->fence_destruction_cond); } return hr; @@ -472,7 +532,9 @@ HRESULT vkd3d_fence_worker_stop(struct vkd3d_fence_worker *worker, pthread_mutex_destroy(&worker->mutex); pthread_cond_destroy(&worker->cond); + pthread_cond_destroy(&worker->fence_destruction_cond); + vkd3d_free(worker->enqueued_fences); vkd3d_free(worker->vk_fences); vkd3d_free(worker->fences); @@ -1047,6 +1109,8 @@ static HRESULT d3d12_fence_init(struct d3d12_fence *fence, struct d3d12_device * memset(fence->old_vk_fences, 0, sizeof(fence->old_vk_fences)); + fence->pending_worker_operation_count = 0; + if (FAILED(hr = vkd3d_private_store_init(&fence->private_store))) { pthread_mutex_destroy(&fence->mutex); diff --git a/libs/vkd3d/vkd3d_private.h b/libs/vkd3d/vkd3d_private.h index 9f00d327b73b..3af8a9594a5e 100644 --- a/libs/vkd3d/vkd3d_private.h +++ b/libs/vkd3d/vkd3d_private.h @@ -156,7 +156,20 @@ struct vkd3d_fence_worker union vkd3d_thread_handle thread; pthread_mutex_t mutex; pthread_cond_t cond; + pthread_cond_t fence_destruction_cond; bool should_exit; + bool pending_fence_destruction; + + size_t enqueued_fence_count; + struct vkd3d_enqueued_fence + { + VkFence vk_fence; + struct d3d12_fence *fence; + uint64_t value; + struct vkd3d_queue *queue; + uint64_t queue_sequence_number; + } *enqueued_fences; + size_t enqueued_fences_size; size_t fence_count; VkFence *vk_fences; @@ -317,6 +330,8 @@ struct d3d12_fence struct list semaphores; unsigned int semaphore_count; + LONG pending_worker_operation_count; + VkFence old_vk_fences[VKD3D_MAX_VK_SYNC_OBJECTS]; struct d3d12_device *device; -- 2.21.0

2 1

[PATCH vkd3d 1/4] tests: Add bug_if() for Nvidia tessellation shader bugs.
by Józef Kucia June 13, 2019

June 13, 2019

From: Józef Kucia <jkucia(a)codeweavers.com> Signed-off-by: Józef Kucia <jkucia(a)codeweavers.com> --- Those are fixed but the fixes are not released yet. --- tests/d3d12.c | 1 + 1 file changed, 1 insertion(+) diff --git a/tests/d3d12.c b/tests/d3d12.c index a57a0453123f..81fb26ad7d60 100644 --- a/tests/d3d12.c +++ b/tests/d3d12.c @@ -27042,6 +27042,7 @@ static void test_line_tessellation(void) data->x, data->y, data->z, data->w, expected->x, expected->y, expected->z, expected->w, i); ++data; expected = &expected_data[3 * i + 1]; + bug_if(is_nvidia_device(context.device)) ok(compare_vec4(data, expected, 1), "Got color {%.8e, %.8e, %.8e, %.8e}, expected {%.8e, %.8e, %.8e, %.8e} at %u.\n", data->x, data->y, data->z, data->w, expected->x, expected->y, expected->z, expected->w, i); -- 2.21.0

2 1

[PATCH v3 1/3] wined3d: Use 3 component norm for 'nrm' opcode in GLSL backend.
by Paul Gofman June 13, 2019

June 13, 2019

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=8848 Signed-off-by: Paul Gofman <gofmanp(a)gmail.com> --- v2: no changes v3: - Split off w component handling for zero norm in a separate patch. dlls/wined3d/glsl_shader.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) diff --git a/dlls/wined3d/glsl_shader.c b/dlls/wined3d/glsl_shader.c index c934528981..f2fa18dba8 100644 --- a/dlls/wined3d/glsl_shader.c +++ b/dlls/wined3d/glsl_shader.c @@ -4260,8 +4260,12 @@ static void shader_glsl_nrm(const struct wined3d_shader_instruction *ins) mask_size = shader_glsl_get_write_mask_size(write_mask); shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src_param); - shader_addline(buffer, "tmp0.x = dot(%s, %s);\n", - src_param.param_str, src_param.param_str); + if (mask_size > 3) + shader_addline(buffer, "tmp0.x = dot(vec3(%s), vec3(%s));\n", + src_param.param_str, src_param.param_str); + else + shader_addline(buffer, "tmp0.x = dot(%s, %s);\n", + src_param.param_str, src_param.param_str); shader_glsl_append_dst(buffer, ins); if (mask_size > 1) -- 2.21.0

2 3