Re: [PATCH] kernel32: implement GetLargePageMinimum
On 14.06.2016 09:09, Marcus Meissner wrote:
From: Austin English <austinenglish(a)gmail.com>
https://bugs.winehq.org/show_bug.cgi?id=18745
Report 2MB minimum page size on x86 32bit and 64bit on Linux.
This assumes that all x86 processors and the 64/32bit emulation mode have hugetable support and it is always enabled.
Signed-off-by: Marcus Meissner <marcus(a)jet.franken.de> Signed-off-by: Austin English <austinenglish(a)gmail.com>
As pointed out on IRC, the order of the sign-offs is wrong here. Your own signoff should always be the last because you have to take responsibility when you send a patch.
--- dlls/kernel32/cpu.c | 15 +++++++++++++++ dlls/kernel32/kernel32.spec | 2 +- dlls/kernel32/tests/process.c | 16 ++++++++++++++++ 3 files changed, 32 insertions(+), 1 deletion(-)
diff --git a/dlls/kernel32/cpu.c b/dlls/kernel32/cpu.c index f48fcf0..f8b2c3d 100644 --- a/dlls/kernel32/cpu.c +++ b/dlls/kernel32/cpu.c @@ -291,3 +291,18 @@ err: } return TRUE; } + +/*********************************************************************** + * GetLargePageMinimum (KERNEL32.@) + */ +SIZE_T WINAPI GetLargePageMinimum(void) +{ +#ifdef linux
I do not see any need for that #ifdef linux. Its more architecture specific than linux specific. Dmitry already mentioned the same concern on the mailing list.
+#if defined(__i386___) || defined(__x86_64__) + /* Linux x86 hugepages are minimum 2 MB */ + return 2*1024*1024; +#endif +#endif + FIXME("stub: not implemented on your platform/architecture.\n"); + return 0; +} diff --git a/dlls/kernel32/kernel32.spec b/dlls/kernel32/kernel32.spec index ec7155b6e..0fc4dcb 100644 --- a/dlls/kernel32/kernel32.spec +++ b/dlls/kernel32/kernel32.spec @@ -696,7 +696,7 @@ @ stdcall GetHandleInformation(long ptr) @ stub -i386 GetLSCallbackTarget @ stub -i386 GetLSCallbackTemplate -# @ stub GetLargePageMinimum +@ stdcall GetLargePageMinimum() @ stdcall GetLargestConsoleWindowSize(long) @ stdcall GetLastError() @ stub GetLinguistLangSize diff --git a/dlls/kernel32/tests/process.c b/dlls/kernel32/tests/process.c index 61dcdc2..e9acfd3 100644 --- a/dlls/kernel32/tests/process.c +++ b/dlls/kernel32/tests/process.c @@ -88,6 +88,7 @@ static BOOL (WINAPI *pProcess32Next)(HANDLE, PROCESSENTRY32*); static BOOL (WINAPI *pThread32First)(HANDLE, THREADENTRY32*); static BOOL (WINAPI *pThread32Next)(HANDLE, THREADENTRY32*); static BOOL (WINAPI *pGetLogicalProcessorInformationEx)(LOGICAL_PROCESSOR_RELATIONSHIP,SYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX*,DWORD*); +static SIZE_T (WINAPI *pGetLargePageMinimum)(void);
/* ############################### */ static char base[MAX_PATH]; @@ -252,6 +253,7 @@ static BOOL init(void) pThread32First = (void *)GetProcAddress(hkernel32, "Thread32First"); pThread32Next = (void *)GetProcAddress(hkernel32, "Thread32Next"); pGetLogicalProcessorInformationEx = (void *)GetProcAddress(hkernel32, "GetLogicalProcessorInformationEx"); + pGetLargePageMinimum = (void *)GetProcAddress(hkernel32, "GetLargePageMinimum");
return TRUE; } @@ -3138,6 +3140,19 @@ static void test_GetLogicalProcessorInformationEx(void) HeapFree(GetProcessHeap(), 0, info); }
+static void test_largepages(void) +{ + SIZE_T size; + + if (!pGetLargePageMinimum) { + skip("No GetLargePageMinimum support.\n"); + return; + } + size = pGetLargePageMinimum(); + + ok((size == 0) || (size == 2*1024*1024) || (size == 4*1024*1024), "GetLargePageMinimum reports %ld size\n", size); +} + START_TEST(process) { HANDLE job; @@ -3210,6 +3225,7 @@ START_TEST(process) test_GetNumaProcessorNode(); test_session_info(); test_GetLogicalProcessorInformationEx(); + test_largepages();
/* things that can be tested: * lookup: check the way program to be executed is searched
On Jun 14, 2016 9:33 AM, "Sebastian Lackner" <sebastian(a)fds-team.de> wrote:
On 14.06.2016 09:09, Marcus Meissner wrote:
From: Austin English <austinenglish(a)gmail.com>
https://bugs.winehq.org/show_bug.cgi?id=18745
Report 2MB minimum page size on x86 32bit and 64bit on Linux.
This assumes that all x86 processors and the 64/32bit emulation mode have hugetable support and it is always enabled.
Signed-off-by: Marcus Meissner <marcus(a)jet.franken.de> Signed-off-by: Austin English <austinenglish(a)gmail.com>
As pointed out on IRC, the order of the sign-offs is wrong here. Your own
signoff
should always be the last because you have to take responsibility when you send a patch.
--- dlls/kernel32/cpu.c | 15 +++++++++++++++ dlls/kernel32/kernel32.spec | 2 +- dlls/kernel32/tests/process.c | 16 ++++++++++++++++ 3 files changed, 32 insertions(+), 1 deletion(-)
diff --git a/dlls/kernel32/cpu.c b/dlls/kernel32/cpu.c index f48fcf0..f8b2c3d 100644 --- a/dlls/kernel32/cpu.c +++ b/dlls/kernel32/cpu.c @@ -291,3 +291,18 @@ err: } return TRUE; } +
+/***********************************************************************
+ * GetLargePageMinimum (KERNEL32.@) + */ +SIZE_T WINAPI GetLargePageMinimum(void) +{ +#ifdef linux
I do not see any need for that #ifdef linux. Its more architecture specific than linux specific. Dmitry already mentioned the same concern on the mailing list.
+#if defined(__i386___) || defined(__x86_64__) + /* Linux x86 hugepages are minimum 2 MB */ + return 2*1024*1024; +#endif +#endif + FIXME("stub: not implemented on your platform/architecture.\n"); + return 0; +} diff --git a/dlls/kernel32/kernel32.spec b/dlls/kernel32/kernel32.spec index ec7155b6e..0fc4dcb 100644 --- a/dlls/kernel32/kernel32.spec +++ b/dlls/kernel32/kernel32.spec @@ -696,7 +696,7 @@ @ stdcall GetHandleInformation(long ptr) @ stub -i386 GetLSCallbackTarget @ stub -i386 GetLSCallbackTemplate -# @ stub GetLargePageMinimum +@ stdcall GetLargePageMinimum() @ stdcall GetLargestConsoleWindowSize(long) @ stdcall GetLastError() @ stub GetLinguistLangSize diff --git a/dlls/kernel32/tests/process.c b/dlls/kernel32/tests/process.c index 61dcdc2..e9acfd3 100644 --- a/dlls/kernel32/tests/process.c +++ b/dlls/kernel32/tests/process.c @@ -88,6 +88,7 @@ static BOOL (WINAPI *pProcess32Next)(HANDLE, PROCESSENTRY32*); static BOOL (WINAPI *pThread32First)(HANDLE, THREADENTRY32*); static BOOL (WINAPI *pThread32Next)(HANDLE, THREADENTRY32*); static BOOL (WINAPI *pGetLogicalProcessorInformationEx)(LOGICAL_PROCESSOR_RELATIONSHIP,SYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX*,DWORD*); +static SIZE_T (WINAPI *pGetLargePageMinimum)(void);
/* ############################### */ static char base[MAX_PATH]; @@ -252,6 +253,7 @@ static BOOL init(void) pThread32First = (void *)GetProcAddress(hkernel32, "Thread32First"); pThread32Next = (void *)GetProcAddress(hkernel32, "Thread32Next"); pGetLogicalProcessorInformationEx = (void *)GetProcAddress(hkernel32, "GetLogicalProcessorInformationEx"); + pGetLargePageMinimum = (void *)GetProcAddress(hkernel32, "GetLargePageMinimum");
return TRUE; } @@ -3138,6 +3140,19 @@ static void test_GetLogicalProcessorInformationEx(void) HeapFree(GetProcessHeap(), 0, info); }
+static void test_largepages(void) +{ + SIZE_T size; + + if (!pGetLargePageMinimum) { + skip("No GetLargePageMinimum support.\n"); + return; + } + size = pGetLargePageMinimum(); + + ok((size == 0) || (size == 2*1024*1024) || (size == 4*1024*1024), "GetLargePageMinimum reports %ld size\n", size); +} + START_TEST(process) { HANDLE job; @@ -3210,6 +3225,7 @@ START_TEST(process) test_GetNumaProcessorNode(); test_session_info(); test_GetLogicalProcessorInformationEx(); + test_largepages();
/* things that can be tested: * lookup: check the way program to be executed is searched
Perhaps ARM should be listed as well.
Hi, I fixed that order, and also made a smaller patch that just adds the initial function without large page support. Ciao, Marcus On Tue, Jun 14, 2016 at 04:33:15PM +0200, Sebastian Lackner wrote:
On 14.06.2016 09:09, Marcus Meissner wrote:
From: Austin English <austinenglish(a)gmail.com>
https://bugs.winehq.org/show_bug.cgi?id=18745
Report 2MB minimum page size on x86 32bit and 64bit on Linux.
This assumes that all x86 processors and the 64/32bit emulation mode have hugetable support and it is always enabled.
Signed-off-by: Marcus Meissner <marcus(a)jet.franken.de> Signed-off-by: Austin English <austinenglish(a)gmail.com>
As pointed out on IRC, the order of the sign-offs is wrong here. Your own signoff should always be the last because you have to take responsibility when you send a patch.
--- dlls/kernel32/cpu.c | 15 +++++++++++++++ dlls/kernel32/kernel32.spec | 2 +- dlls/kernel32/tests/process.c | 16 ++++++++++++++++ 3 files changed, 32 insertions(+), 1 deletion(-)
diff --git a/dlls/kernel32/cpu.c b/dlls/kernel32/cpu.c index f48fcf0..f8b2c3d 100644 --- a/dlls/kernel32/cpu.c +++ b/dlls/kernel32/cpu.c @@ -291,3 +291,18 @@ err: } return TRUE; } + +/*********************************************************************** + * GetLargePageMinimum (KERNEL32.@) + */ +SIZE_T WINAPI GetLargePageMinimum(void) +{ +#ifdef linux
I do not see any need for that #ifdef linux. Its more architecture specific than linux specific. Dmitry already mentioned the same concern on the mailing list.
+#if defined(__i386___) || defined(__x86_64__) + /* Linux x86 hugepages are minimum 2 MB */ + return 2*1024*1024; +#endif +#endif + FIXME("stub: not implemented on your platform/architecture.\n"); + return 0; +} diff --git a/dlls/kernel32/kernel32.spec b/dlls/kernel32/kernel32.spec index ec7155b6e..0fc4dcb 100644 --- a/dlls/kernel32/kernel32.spec +++ b/dlls/kernel32/kernel32.spec @@ -696,7 +696,7 @@ @ stdcall GetHandleInformation(long ptr) @ stub -i386 GetLSCallbackTarget @ stub -i386 GetLSCallbackTemplate -# @ stub GetLargePageMinimum +@ stdcall GetLargePageMinimum() @ stdcall GetLargestConsoleWindowSize(long) @ stdcall GetLastError() @ stub GetLinguistLangSize diff --git a/dlls/kernel32/tests/process.c b/dlls/kernel32/tests/process.c index 61dcdc2..e9acfd3 100644 --- a/dlls/kernel32/tests/process.c +++ b/dlls/kernel32/tests/process.c @@ -88,6 +88,7 @@ static BOOL (WINAPI *pProcess32Next)(HANDLE, PROCESSENTRY32*); static BOOL (WINAPI *pThread32First)(HANDLE, THREADENTRY32*); static BOOL (WINAPI *pThread32Next)(HANDLE, THREADENTRY32*); static BOOL (WINAPI *pGetLogicalProcessorInformationEx)(LOGICAL_PROCESSOR_RELATIONSHIP,SYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX*,DWORD*); +static SIZE_T (WINAPI *pGetLargePageMinimum)(void);
/* ############################### */ static char base[MAX_PATH]; @@ -252,6 +253,7 @@ static BOOL init(void) pThread32First = (void *)GetProcAddress(hkernel32, "Thread32First"); pThread32Next = (void *)GetProcAddress(hkernel32, "Thread32Next"); pGetLogicalProcessorInformationEx = (void *)GetProcAddress(hkernel32, "GetLogicalProcessorInformationEx"); + pGetLargePageMinimum = (void *)GetProcAddress(hkernel32, "GetLargePageMinimum");
return TRUE; } @@ -3138,6 +3140,19 @@ static void test_GetLogicalProcessorInformationEx(void) HeapFree(GetProcessHeap(), 0, info); }
+static void test_largepages(void) +{ + SIZE_T size; + + if (!pGetLargePageMinimum) { + skip("No GetLargePageMinimum support.\n"); + return; + } + size = pGetLargePageMinimum(); + + ok((size == 0) || (size == 2*1024*1024) || (size == 4*1024*1024), "GetLargePageMinimum reports %ld size\n", size); +} + START_TEST(process) { HANDLE job; @@ -3210,6 +3225,7 @@ START_TEST(process) test_GetNumaProcessorNode(); test_session_info(); test_GetLogicalProcessorInformationEx(); + test_largepages();
/* things that can be tested: * lookup: check the way program to be executed is searched
On 15.06.2016 08:18, Marcus Meissner wrote:
Hi,
I fixed that order, and also made a smaller patch that just adds the initial function without large page support.
Ciao, Marcus
According to the people on the bug report, a return value of zero is not sufficient to fix this issue (see comment 8).
participants (3)
-
Austin English -
Marcus Meissner -
Sebastian Lackner