Skip to contentSkip to conversation

: i386: make x86_64 tsc header require i386 rather than vice-versa

PATCH In review Andres Salomon 2 participants 2 messages Mar 3, 2007 – Mar 6, 2007

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 : i386: make x86_64 tsc header require i386 rather than vice-versa

Prior to commit 95492e4646e5de8b43d9a7908d6177fb737b61f0 ([PATCH] x86:
rewrite SMP TSC sync code), the headers in asm-i386 did not really require anything in include/asm-x86_64. This means that distributions such as fedora did not include asm-x86_64 in kernel-devel headers for i386. Ingo's commit changed that, and broke things. This is easy enough to hack around in package builds by just including asm-x86_64 on i386, but that's kind of annoying. If anything, x86_64 should depend upon i386, not the other way around.

This patch changes it so that asm-x86_64/tsc.h includes asm-i386/tsc.h, rather than vice-versa.

include/asm-i386/tsc.h +66 -1
index e997891..6207fc0 100644
@@ -1 +1,66 @@
-#include <asm-x86_64/tsc.h>
+/*
+ * linux/include/asm-i386/tsc.h
+ *
+ * x86 TSC related functions
+ */
+#ifndef _ASM_i386_TSC_H
+#define _ASM_i386_TSC_H
+
+#include <asm/processor.h>
+
+/*
+ * Standard way to access the cycle counter.
+ */
+typedef unsigned long long cycles_t;
+
+extern unsigned int cpu_khz;
+extern unsigned int tsc_khz;
+
+static inline cycles_t get_cycles(void)
+{
+	unsigned long long ret = 0;
+
+#ifndef CONFIG_X86_TSC
+	if (!cpu_has_tsc)
+		return 0;
+#endif
+
+#if defined(CONFIG_X86_GENERIC) || defined(CONFIG_X86_TSC)
+	rdtscll(ret);
+#endif
+	return ret;
+}
+
+/* Like get_cycles, but make sure the CPU is synchronized. */
+static __always_inline cycles_t get_cycles_sync(void)
+{
+	unsigned long long ret;
+#ifdef X86_FEATURE_SYNC_RDTSC
+	unsigned eax;
+
+	/*
+	 * Don't do an additional sync on CPUs where we know
+	 * RDTSC is already synchronous:
+	 */
+	alternative_io("cpuid", ASM_NOP2, X86_FEATURE_SYNC_RDTSC,
+			  "=a" (eax), "0" (1) : "ebx","ecx","edx","memory");
+#else
+	sync_core();
+#endif
+	rdtscll(ret);
+
+	return ret;
+}
+
+extern void tsc_init(void);
+extern void mark_tsc_unstable(void);
+extern int unsynchronized_tsc(void);
+
+/*
+ * Boot-time check whether the TSCs are synchronized across
+ * all CPUs/cores:
+ */
+extern void check_tsc_sync_source(int cpu);
+extern void check_tsc_sync_target(void);
+
+#endif
include/asm-x86_64/tsc.h +1 -66

67 lines not on this page · raw message

#1 lore raw
Quoted 11 lines · On Fri, 02 Mar 2007 22:10:33 -0500 Andres Salomon <[email protected]> wrote:

Prior to commit 95492e4646e5de8b43d9a7908d6177fb737b61f0 ([PATCH] x86:
rewrite SMP TSC sync code), the headers in asm-i386 did not really require anything in include/asm-x86_64. This means that distributions such as fedora did not include asm-x86_64 in kernel-devel headers for i386. Ingo's commit changed that, and broke things. This is easy enough to hack around in package builds by just including asm-x86_64 on i386, but that's kind of annoying. If anything, x86_64 should depend upon i386, not the other way around.

This patch changes it so that asm-x86_64/tsc.h includes asm-i386/tsc.h, rather than vice-versa.

Agree - I don't see any gain in making the references be bidirectional, and it just makes life more complicated. Knowing "x86_64 always references i386" is a useful mental mark.

That, plus the kernel-headers point which you describe..

<looks at early_printk.c>