3 * Copyright (C) 2000 David Welch <welch@cwcom.net>
4 * Copyright (C) 1999 Gareth Owen <gaz@athene.co.uk>, Ramon von Handel
5 * Copyright (C) 1991, 1992 Linus Torvalds
7 * This software is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU General Public License as
9 * published by the Free Software Foundation; either version 2 of the
10 * License, or (at your option) any later version.
12 * This software is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License for more details.
17 * You should have received a copy of the GNU General Public License
18 * along with this software; see the file COPYING. If not, write
19 * to the Free Software Foundation, Inc., 675 Mass Ave, Cambridge,
25 * PROJECT: ReactOS kernel
26 * FILE: ntoskrnl/hal/x86/udelay.c
27 * PURPOSE: Busy waiting
28 * PROGRAMMER: David Welch (david.welch@seh.ox.ac.uk)
33 /* INCLUDES ***************************************************************/
35 #include <ddk/ntddk.h>
36 #include <internal/ps.h>
40 #include <internal/debug.h>
42 /* GLOBALS ******************************************************************/
44 #define TMR_CTRL 0x43 /* I/O for control */
45 #define TMR_CNT0 0x40 /* I/O for counter 0 */
46 #define TMR_CNT1 0x41 /* I/O for counter 1 */
47 #define TMR_CNT2 0x42 /* I/O for counter 2 */
49 #define TMR_SC0 0 /* Select channel 0 */
50 #define TMR_SC1 0x40 /* Select channel 1 */
51 #define TMR_SC2 0x80 /* Select channel 2 */
53 #define TMR_LOW 0x10 /* RW low byte only */
54 #define TMR_HIGH 0x20 /* RW high byte only */
55 #define TMR_BOTH 0x30 /* RW both bytes */
57 #define TMR_MD0 0 /* Mode 0 */
58 #define TMR_MD1 0x2 /* Mode 1 */
59 #define TMR_MD2 0x4 /* Mode 2 */
60 #define TMR_MD3 0x6 /* Mode 3 */
61 #define TMR_MD4 0x8 /* Mode 4 */
62 #define TMR_MD5 0xA /* Mode 5 */
64 #define TMR_BCD 1 /* BCD mode */
66 #define TMR_LATCH 0 /* Latch command */
68 #define TMR_READ 0xF0 /* Read command */
69 #define TMR_CNT 0x20 /* CNT bit (Active low, subtract it) */
70 #define TMR_STAT 0x10 /* Status bit (Active low, subtract it) */
71 #define TMR_CH2 0x8 /* Channel 2 bit */
72 #define TMR_CH1 0x4 /* Channel 1 bit */
73 #define TMR_CH0 0x2 /* Channel 0 bit */
75 #define MILLISEC 10 /* Number of millisec between interrupts */
76 #define HZ (1000 / MILLISEC) /* Number of interrupts per second */
77 #define CLOCK_TICK_RATE 1193182 /* Clock frequency of the timer chip */
78 #define LATCH (CLOCK_TICK_RATE / HZ) /* Count to program into the timer chip */
79 #define PRECISION 8 /* Number of bits to calibrate for delay loop */
81 static BOOLEAN UdelayCalibrated
= FALSE
;
83 /* FUNCTIONS **************************************************************/
86 * NOTE: This function MUST NOT be optimized by the compiler!
87 * If it is, it obviously will not delay AT ALL, and the system
88 * will appear completely frozen at boot since
89 * HalpCalibrateStallExecution will never return.
90 * There are three options to stop optimization:
91 * 1. Use a volatile automatic variable. Making it delay quite a bit
92 * due to memory accesses, and keeping the code portable. However,
93 * as this involves memory access it depends on both the CPU cache,
94 * e.g. if the stack used is already in a cache line or not, and
95 * whether or not we're MP. If MP, another CPU could (probably would)
96 * also access RAM at the same time - making the delay imprecise.
97 * 2. Use compiler-specific #pragma's to disable optimization.
98 * 3. Use inline assembly, making it equally unportable as #2.
99 * For supported compilers we use inline assembler. For the others,
102 VOID STDCALL
__attribute__((noinline
))
103 __KeStallExecutionProcessor(ULONG Loops
)
109 #if defined(__GNUC__)
110 __asm__
__volatile__ (
113 "jnz ROSL1" : : "d" (Loops
));
115 #elif defined(_MSC_VER)
121 volatile unsigned int target
= Loops
;
123 for (i
=0; i
<target
;i
++);
127 VOID STDCALL
KeStallExecutionProcessor(ULONG Microseconds
)
129 PKPCR Pcr
= KeGetCurrentKPCR();
131 if (Pcr
->PrcbData
.FeatureBits
& X86_FEATURE_TSC
)
133 LARGE_INTEGER EndCount
, CurrentCount
;
134 Ki386RdTSC(EndCount
);
135 EndCount
.QuadPart
+= Microseconds
* (ULONGLONG
)Pcr
->PrcbData
.MHz
;
138 Ki386RdTSC(CurrentCount
);
140 while (CurrentCount
.QuadPart
< EndCount
.QuadPart
);
144 __KeStallExecutionProcessor((Pcr
->StallScaleFactor
*Microseconds
)/1000);
148 static ULONG
Read8254Timer(VOID
)
153 /* save flags and disable interrupts */
154 Ki386SaveFlags(flags
);
155 Ki386DisableInterrupts();
157 WRITE_PORT_UCHAR((PUCHAR
) TMR_CTRL
, TMR_SC0
| TMR_LATCH
);
158 Count
= READ_PORT_UCHAR((PUCHAR
) TMR_CNT0
);
159 Count
|= READ_PORT_UCHAR((PUCHAR
) TMR_CNT0
) << 8;
162 Ki386RestoreFlags(flags
);
168 static VOID
WaitFor8254Wraparound(VOID
)
170 ULONG CurCount
, PrevCount
= ~0;
173 CurCount
= Read8254Timer();
177 PrevCount
= CurCount
;
178 CurCount
= Read8254Timer();
179 Delta
= CurCount
- PrevCount
;
182 * This limit for delta seems arbitrary, but it isn't, it's
183 * slightly above the level of error a buggy Mercury/Neptune
184 * chipset timer can cause.
191 VOID
HalpCalibrateStallExecution(VOID
)
197 LARGE_INTEGER StartCount
, EndCount
;
199 if (UdelayCalibrated
)
204 UdelayCalibrated
= TRUE
;
205 Pcr
= KeGetCurrentKPCR();
207 /* Initialise timer interrupt with MILLISEC ms interval */
208 WRITE_PORT_UCHAR((PUCHAR
) TMR_CTRL
, TMR_SC0
| TMR_BOTH
| TMR_MD2
); /* binary, mode 2, LSB/MSB, ch 0 */
209 WRITE_PORT_UCHAR((PUCHAR
) TMR_CNT0
, LATCH
& 0xff); /* LSB */
210 WRITE_PORT_UCHAR((PUCHAR
) TMR_CNT0
, LATCH
>> 8); /* MSB */
212 if (Pcr
->PrcbData
.FeatureBits
& X86_FEATURE_TSC
)
215 WaitFor8254Wraparound();
216 Ki386RdTSC(StartCount
);
218 WaitFor8254Wraparound();
219 Ki386RdTSC(EndCount
);
221 Pcr
->PrcbData
.MHz
= (ULONG
)(EndCount
.QuadPart
- StartCount
.QuadPart
) / 10000;
222 DPRINT("%dMHz\n", Pcr
->PrcbData
.MHz
);
227 DbgPrint("Calibrating delay loop... [");
229 /* Stage 1: Coarse calibration */
231 WaitFor8254Wraparound();
233 Pcr
->StallScaleFactor
= 1;
237 Pcr
->StallScaleFactor
<<= 1; /* Next delay count to try */
239 WaitFor8254Wraparound();
241 __KeStallExecutionProcessor(Pcr
->StallScaleFactor
); /* Do the delay */
243 CurCount
= Read8254Timer();
245 while (CurCount
> LATCH
/ 2);
247 Pcr
->StallScaleFactor
>>= 1; /* Get bottom value for delay */
249 /* Stage 2: Fine calibration */
250 DbgPrint("delay_count: %d", Pcr
->StallScaleFactor
);
252 calib_bit
= Pcr
->StallScaleFactor
; /* Which bit are we going to test */
254 for (i
= 0; i
< PRECISION
; i
++)
256 calib_bit
>>= 1; /* Next bit to calibrate */
259 break; /* If we have done all bits, stop */
262 Pcr
->StallScaleFactor
|= calib_bit
; /* Set the bit in delay_count */
264 WaitFor8254Wraparound();
266 __KeStallExecutionProcessor(Pcr
->StallScaleFactor
); /* Do the delay */
268 CurCount
= Read8254Timer();
269 if (CurCount
<= LATCH
/ 2) /* If a tick has passed, turn the */
270 { /* calibrated bit back off */
271 Pcr
->StallScaleFactor
&= ~calib_bit
;
275 /* We're finished: Do the finishing touches */
277 Pcr
->StallScaleFactor
/= (MILLISEC
/ 2); /* Calculate delay_count for 1ms */
280 DbgPrint("delay_count: %d\n", Pcr
->StallScaleFactor
);
281 DbgPrint("CPU speed: %d\n", Pcr
->StallScaleFactor
/ 250);
283 DbgPrint("About to start delay loop test\n");
284 DbgPrint("Waiting for five minutes...");
285 for (i
= 0; i
< (5*60*1000*20); i
++)
287 KeStallExecutionProcessor(50);
289 DbgPrint("finished\n");
296 HalCalibratePerformanceCounter(ULONG Count
)
300 /* save flags and disable interrupts */
301 Ki386SaveFlags(flags
);
302 Ki386DisableInterrupts();
304 __KeStallExecutionProcessor(Count
);
307 Ki386RestoreFlags(flags
);
311 LARGE_INTEGER STDCALL
312 KeQueryPerformanceCounter(PLARGE_INTEGER PerformanceFreq
)
314 * FUNCTION: Queries the finest grained running count available in the system
316 * PerformanceFreq (OUT) = The routine stores the number of
317 * performance counter ticks per second here
318 * RETURNS: The number of performance counter ticks since boot
325 Ki386SaveFlags(Flags
);
326 Ki386DisableInterrupts();
328 Pcr
= KeGetCurrentKPCR();
330 if (Pcr
->PrcbData
.FeatureBits
& X86_FEATURE_TSC
)
332 Ki386RestoreFlags(Flags
);
333 if (NULL
!= PerformanceFreq
)
335 PerformanceFreq
->QuadPart
= Pcr
->PrcbData
.MHz
* (ULONGLONG
)1000000;
341 LARGE_INTEGER TicksOld
;
342 LARGE_INTEGER TicksNew
;
345 Ki386RestoreFlags(Flags
);
347 if (NULL
!= PerformanceFreq
)
349 PerformanceFreq
->QuadPart
= CLOCK_TICK_RATE
;
354 KeQueryTickCount(&TicksOld
);
355 CountsLeft
= Read8254Timer();
356 Value
.QuadPart
= TicksOld
.QuadPart
* LATCH
+ (LATCH
- CountsLeft
);
357 KeQueryTickCount(&TicksNew
);
359 while (TicksOld
.QuadPart
!= TicksNew
.QuadPart
);