Changes to make udelay a bit easier to compile in. There are now three version.

src/cpu/p5/Config

@@ -1,4 +1,5 @@
+option CONFIG_TSC_X86RDTSC_CALIBRATE_WITH_TIMER2=0
 option i586=1
 object cpuid.o
-object delay_tsc.o
+object delay_tsc.o  CONFIG_UDELAY_TSC
 #object tsc.o

src/cpu/p5/delay_tsc.c

@@ -4,7 +4,7 @@
 
 static unsigned long clocks_per_usec;
 
-
+#if (CONFIG_TSC_X86RDTSC_CALIBRATE_WITH_TIMER2 == 1)
 #define CLOCK_TICK_RATE	1193180U /* Underlying HZ */
 
 /* ------ Calibrate the TSC ------- 
@@ -79,6 +79,73 @@ bad_ctc:
 	return 0;
 }
 
+#else /*  CONFIG_TSC_X86RDTSC_CALIBRATE_WITH_TIMER2 */
+
+/*
+ * this is the "no timer2" version.
+ * to calibrate tsc, we get a TSC reading, then do 1,000,000 outbs to port 0x80
+ * then we read TSC again, and divide the difference by 1,000,000
+ * we have found on a wide range of machines that this gives us a a 
+ * good microsecond value
+ * to +- 10%. On a dual AMD 1.6 Ghz box, it gives us .97 microseconds, and on a
+ * 267 Mhz. p5, it gives us 1.1 microseconds.
+ * also, since gcc now supports long long, we use that.
+ * also no unsigned long long / operator, so we play games.
+ * about the only thing you can do with long longs, it seems, 
+ *is return them and assign them.
+ * (and do asm on them, yuck)
+ * so avoid all ops on long longs.
+ */
+static unsigned long long calibrate_tsc(void)
+{
+    unsigned long long retval, start, end, delta;
+    unsigned long allones = (unsigned long) -1, result;
+    unsigned long startlow, starthigh;
+    unsigned long endlow, endhigh;
+    unsigned long count;
+
+    rdtsc(startlow,starthigh);
+    // no udivdi3, dammit.
+    // so we count to 1<< 20 and then right shift 20
+    for(count = 0; count < (1<<20); count ++)
+	outb(0x80, 0x80);
+    rdtsc(endlow,endhigh);
+
+    // make delta be (endhigh - starthigh) + (endlow - startlow)
+    // but >> 20
+    // do it this way to avoid gcc warnings.
+    start = starthigh;
+    start <<= 32;
+    start |= startlow;
+    end = endhigh;
+    end <<= 32;
+    end |= endlow;
+    delta = end - start;
+    // at this point we have a delta for 1,000,000 outbs. Now rescale for one microsecond.
+    delta >>= 20;
+    // save this for microsecond timing.
+    clocks_per_usec = delta;
+#ifdef DEBUG
+    printf("end %x:%x, start %x:%x\n",
+	endhigh, endlow, starthigh, startlow);
+    printf("32-bit delta %d\n", (unsigned long) delta);
+#endif
+
+    // ok now we don't have divide for unsigned long long.
+    // so what we do is set an unsigned long to all 1's, which is just 1 less than
+    // 1 << 32. then we divide by an unsigned long version of delta (we now that delta fits
+    // easily in 32 bits). that way we avoid assembly code.
+    // avoid gcc complaints.
+    result = allones / (unsigned long) delta;
+    retval = result;
+#ifdef DEBUG
+    printf(__FUNCTION__ " 32-bit result is %d\n", result);
+#endif
+    return retval;
+}
+
+
+#endif /* CONFIG_TSC_X86RDTSC_CALIBRATE_WITH_TIMER2*/
 
 void udelay(unsigned long us)
 {

src/lib/delay.c

@@ -1,59 +1,4 @@
 #include <delay.h>
-#include <arch/io.h>
-
-/* Ports for the 8254 timer chip */
-#define	TIMER2_PORT	0x42
-#define	TIMER_MODE_PORT	0x43
-
-/* Meaning of the mode bits */
-#define	TIMER0_SEL	0x00
-#define	TIMER1_SEL	0x40
-#define	TIMER2_SEL	0x80
-#define	READBACK_SEL	0xC0
-
-#define	LATCH_COUNT	0x00
-#define	LOBYTE_ACCESS	0x10
-#define	HIBYTE_ACCESS	0x20
-#define	WORD_ACCESS	0x30
-
-#define	MODE0		0x00
-#define	MODE1		0x02
-#define	MODE2		0x04
-#define	MODE3		0x06
-#define	MODE4		0x08
-#define	MODE5		0x0A
-
-#define	BINARY_COUNT	0x00
-#define	BCD_COUNT	0x01
-
-/* Timers tick over at this rate */
-#define	TICKS_PER_MS	1193
-
-/* Parallel Peripheral Controller Port B */
-#define	PPC_PORTB	0x61
-
-/* Meaning of the port bits */
-#define	PPCB_T2OUT	0x20	/* Bit 5 */
-#define	PPCB_SPKR	0x02	/* Bit 1 */
-#define	PPCB_T2GATE	0x01	/* Bit 0 */
-
-
-static void load_timer2(unsigned int ticks)
-{
-	/* Set up the timer gate, turn off the speaker */
-	outb((inb(PPC_PORTB) & ~PPCB_SPKR) | PPCB_T2GATE, PPC_PORTB);
-	outb(TIMER2_SEL|WORD_ACCESS|MODE0|BINARY_COUNT, TIMER_MODE_PORT);
-	outb(ticks & 0xFF, TIMER2_PORT);
-	outb(ticks >> 8, TIMER2_PORT);
-}
-
-
-void udelay(int usecs)
-{
-	load_timer2((usecs*TICKS_PER_MS)/1000);
-	while ((inb(PPC_PORTB) & PPCB_T2OUT) == 0)
-		;
-}
 void mdelay(int msecs)
 {
 	int i;

src/pc80/Config

@@ -2,5 +2,6 @@ object keyboard.o
 object mc146818rtc.o
 object isa-dma.o
 object i8259.o
+object udelay_timer2.o CONFIG_UDELAY_TIMER2
 dir ide
 

src/pc80/udelay_timer2.c

@@ -0,0 +1,56 @@
+#include <delay.h>
+#include <arch/io.h>
+
+/* Ports for the 8254 timer chip */
+#define	TIMER2_PORT	0x42
+#define	TIMER_MODE_PORT	0x43
+
+/* Meaning of the mode bits */
+#define	TIMER0_SEL	0x00
+#define	TIMER1_SEL	0x40
+#define	TIMER2_SEL	0x80
+#define	READBACK_SEL	0xC0
+
+#define	LATCH_COUNT	0x00
+#define	LOBYTE_ACCESS	0x10
+#define	HIBYTE_ACCESS	0x20
+#define	WORD_ACCESS	0x30
+
+#define	MODE0		0x00
+#define	MODE1		0x02
+#define	MODE2		0x04
+#define	MODE3		0x06
+#define	MODE4		0x08
+#define	MODE5		0x0A
+
+#define	BINARY_COUNT	0x00
+#define	BCD_COUNT	0x01
+
+/* Timers tick over at this rate */
+#define	TICKS_PER_MS	1193
+
+/* Parallel Peripheral Controller Port B */
+#define	PPC_PORTB	0x61
+
+/* Meaning of the port bits */
+#define	PPCB_T2OUT	0x20	/* Bit 5 */
+#define	PPCB_SPKR	0x02	/* Bit 1 */
+#define	PPCB_T2GATE	0x01	/* Bit 0 */
+
+
+static void load_timer2(unsigned int ticks)
+{
+	/* Set up the timer gate, turn off the speaker */
+	outb((inb(PPC_PORTB) & ~PPCB_SPKR) | PPCB_T2GATE, PPC_PORTB);
+	outb(TIMER2_SEL|WORD_ACCESS|MODE0|BINARY_COUNT, TIMER_MODE_PORT);
+	outb(ticks & 0xFF, TIMER2_PORT);
+	outb(ticks >> 8, TIMER2_PORT);
+}
+
+
+void udelay(int usecs)
+{
+	load_timer2((usecs*TICKS_PER_MS)/1000);
+	while ((inb(PPC_PORTB) & PPCB_T2OUT) == 0)
+		;
+}

src/southbridge/amd/amd766/Config

@@ -1,3 +1,4 @@
+option CONFIG_UDELAY_TSC
 object southbridge.o
 object nvram.o
 object amd766_ioapic.o
@@ -11,6 +12,3 @@ object amd766_power.o
 object amd766_reboot.o
 object amd766_ide.o
 
-
-
-