Redo cpuid feature detection, part 2

Turns out using GCC cpuid() clobbers registers with first symptoms dying in tunables. MS has intrinsic cpuid() so we might as well use it. Also make sure the tnable get functions use static memory, as they are used after function returns. Signed-off-by: Jorgen Lundman <[email protected]>
openzfsonwindows · Mar 18, 2024 · ad3d9f7 · ad3d9f7
1 parent fe2ce35
commit ad3d9f7
Show file tree

Hide file tree

Showing 6 changed files with 22 additions and 36 deletions.
diff --git a/include/os/windows/spl/sys/processor.h b/include/os/windows/spl/sys/processor.h
@@ -3,6 +3,7 @@
 #define	_SPL_PROCESSOR_H
 
 #include <sys/types.h>
+#include <intrin.h>
 
 extern uint32_t getcpuid();
 
@@ -14,41 +15,22 @@ extern int __cpuid_count(unsigned int __level, unsigned int __sublevel,
 	unsigned int __eax, unsigned int __ebx,
 	unsigned int __ecx, unsigned int __edx);
 
-#define	__cpuid_count(level, count, a, b, c, d) \
-	__asm__("xchg{l}\t{%%}ebx, %1\n\t" \
-		"cpuid\n\t" \
-		"xchg{l}\t{%%}ebx, %1\n\t" \
-		: "=a" (a), "=r" (b), "=c" (c), "=d" (d) \
-		: "0" (level), "2" (count))
-
-#define	__cpuid(level, a, b, c, d) \
-	__asm__("xchg{l}\t{%%}ebx, %1\n\t" \
-		"cpuid\n\t" \
-		"xchg{l}\t{%%}ebx, %1\n\t" \
-		: "=a" (a), "=r" (b), "=c" (c), "=d" (d) \
-		: "0" (level))
-
 static inline unsigned int
 __get_cpuid_max(unsigned int __ext, unsigned int *__sig)
 {
+#ifdef _WIN32
+	int32_t r[4];
+	__cpuid(r, __ext);
+	if (__sig)
+		*__sig = (uint32_t)r[1];
+	return ((uint32_t)r[0]);
+#else
 	unsigned int __eax, __ebx, __ecx, __edx;
 	__cpuid(__ext, __eax, __ebx, __ecx, __edx);
 	if (__sig)
 		*__sig = __ebx;
 	return (__eax);
-}
-
-/* macOS does have do_cpuid() macro */
-static inline int
-__get_cpuid(unsigned int __level,
-    unsigned int *__eax, unsigned int *__ebx,
-    unsigned int *__ecx, unsigned int *__edx)
-{
-	unsigned int __ext = __level & 0x80000000;
-	if (__get_cpuid_max(__ext, 0) < __level)
-		return (0);
-	__cpuid(__level, *__eax, *__ebx, *__ecx, *__edx);
-	return (1);
+#endif
 }
 
 #endif // x86

diff --git a/include/os/windows/spl/sys/simd.h b/include/os/windows/spl/sys/simd.h
@@ -262,8 +262,12 @@ __cpuid_check_feature(const cpuid_feature_desc_t *desc)
 		 * for AVX2. It is a macro, so return parameters
 		 * are passed by value.
 		 */
+#ifdef _WIN32
+		__cpuidex((int32_t *)r, desc->leaf, desc->subleaf);
+#else
 		__cpuid_count(desc->leaf, desc->subleaf,
 		    r[EAX], r[EBX], r[ECX], r[EDX]);
+#endif
 		return ((r[desc->reg] & desc->flag) == desc->flag);
 	}
 	return (B_FALSE);

diff --git a/module/icp/algs/blake3/blake3_impl.c b/module/icp/algs/blake3/blake3_impl.c
@@ -355,10 +355,10 @@ blake3_param_set(const char *val, zfs_kernel_param_t *unused)
 int
 win32_blake3_param_set(ZFS_MODULE_PARAM_ARGS)
 {
+	static char buffer[PAGE_SIZE]; /* Looks like they use page size */
 	*type = ZT_TYPE_STRING;
 
 	if (set == B_FALSE) {
-		char buffer[PAGE_SIZE]; /* Looks like they use page size */
 		blake3_param_get(buffer, NULL);
 		*ptr = buffer;
 		*len = strlen(buffer);

diff --git a/module/icp/algs/modes/gcm.c b/module/icp/algs/modes/gcm.c
@@ -1031,15 +1031,15 @@ icp_gcm_impl_get(char *buffer, zfs_kernel_param_t *kp)
 int
 win32_icp_gcm_impl_set(ZFS_MODULE_PARAM_ARGS)
 {
-	static char str[1024] = "";
+	static char buffer[PAGE_SIZE] = "";
 
 	*type = ZT_TYPE_STRING;
 
 	if (set == B_FALSE) {
 		if (gcm_impl_initialized)
-			icp_gcm_impl_get(str, NULL);
-		*ptr = str;
-		*len = strlen(str);
+			icp_gcm_impl_get(buffer, NULL);
+		*ptr = buffer;
+		*len = strlen(buffer);
 		return (0);
 	}
 

diff --git a/module/os/windows/zfs/sysctl_os.c b/module/os/windows/zfs/sysctl_os.c
@@ -567,7 +567,7 @@ param_set_arc_int(ZFS_MODULE_PARAM_ARGS)
 int
 param_set_active_allocator(ZFS_MODULE_PARAM_ARGS)
 {
-	char buf[16];
+	static char buf[16];
 
 	*type = ZT_TYPE_STRING;
 
@@ -668,7 +668,7 @@ param_set_deadman_ziotime(ZFS_MODULE_PARAM_ARGS)
 int
 param_set_deadman_failmode(ZFS_MODULE_PARAM_ARGS)
 {
-	char buf[16];
+	static char buf[16];
 
 	*type = ZT_TYPE_STRING;
 

diff --git a/module/zfs/spa.c b/module/zfs/spa.c
@@ -1485,7 +1485,7 @@ spa_taskq_write_param(ZFS_MODULE_PARAM_ARGS)
 static int
 win32_spa_taskq_read_param_set(ZFS_MODULE_PARAM_ARGS)
 {
-	char str[1024] = "";
+	static char str[1024] = "";
 
 	*type = ZT_TYPE_STRING;
 
@@ -1506,7 +1506,7 @@ win32_spa_taskq_read_param_set(ZFS_MODULE_PARAM_ARGS)
 static int
 win32_spa_taskq_write_param_set(ZFS_MODULE_PARAM_ARGS)
 {
-	char str[1024] = "";
+	static char str[1024] = "";
 
 	*type = ZT_TYPE_STRING;