/*

* Copyright (c) 2008-2009 Nokia Corporation and/or its subsidiary(-ies).

* All rights reserved.

* This component and the accompanying materials are made available

* under the terms of "Eclipse Public License v1.0"

* which accompanies this distribution, and is available

* at the URL "http://www.eclipse.org/legal/epl-v10.html".

*

* Initial Contributors:

* Nokia Corporation - initial contribution.

*

* Contributors:

*

* Description:  

* os\kernelhwsrv\kernel\eka\drivers\power\smpidlehelper.cpp

* Impelentation of helper classes required to implement CPU idle

* functionality in a SMP BSP.

*

*/

/**

 @file

 @prototype

*/

#ifdef __SMP__

#include <smppower/idlehelper.h>

#ifndef DISABLE_TRACE

extern "C" void btrace_printfn(const TAny *aPtr,const TInt regNum)

    {

    PMBTRACE4(KPrintReg,regNum,aPtr);

    }

extern "C" void btrace_printfnId(const TAny *aPtr,const TInt regNum, TUint aId)

    {

    PMBTRACE8(KPrintReg,0xffu,regNum,aPtr);

    }

#define PRINTREG(Rn)                 \

        asm("stmfd sp!,{r0-r12,lr}"); \

        asm("mov r0,r"#Rn); \

        asm("mov r1,#"#Rn); \

        asm("bl btrace_printfn"); \

        asm("ldmfd sp!,{r0-r12,lr}"); 

#define PRINTREGID(Rn,n)               \

        asm("stmfd sp!,{r0-r12,lr}"); \

        asm("mov r0,r"#Rn); \

        asm("mov r1,#"#Rn); \

        asm("mov r2,#"#n); \

        asm("bl btrace_printfnId"); \

        asm("ldmfd sp!,{r0-r12,lr}"); 

#else

#define PRINTREG(Rn)

#define PRINTREGID(Rn,n)

#endif

/**

   Atomically does the following:

     sets the current cpu idle mask bit to indicate current core wants to idle

     if all enaged cores have set their bit the flag KGlobalIdleFlag is also 

     orred into the idle mask to indicate all cores are going down. In this case

     the function returned true. False otherwise

   aCMask- Bit mask with only current CPU bit set

   Normal Usage:use in idle handler before waiting for all cores down IPI

   @pre 

 */		

__NAKED__ TBool TIdleSupport::SetLocalAndCheckSetGlobalIdle(TUint32 /*aCpuMask*/)

    {

	asm("ldr    r1,__iAllEngagedCpusMask");                           //r1 = address of iAllEngagedCpusMask

	asm("ldr	r2, [r1]");                                           //r2 = iAllEngagedCpusMask

    asm("add    r1,r1,#4");                                           //r1 = address of iIdlingCpus 

    __DATA_MEMORY_BARRIER_Z__(r12);   

    asm("1: ");

	LDREX(3,1);                                                       // r3 = iIdlingCpus

    asm("orr    r3,r0,r3");                                           // orr in mask for this CPU

    asm("cmp    r3,r2");                                              // compare to iAllEngagedCpusMask

    asm("orreq  r3,r3,#%a0" : : "i" (TIdleSupport::KGlobalIdleFlag)); // if equal orr in KGlobalIdleFlag

    STREX(12,3,1);

    asm("cmp    r12, #0 ");                                              // 

	asm("bne    1b ");                                                   // write didn't succeed try again

    __DATA_MEMORY_BARRIER__(r12);

    asm("and    r0,r3,#%a0" : : "i" (TIdleSupport::KGlobalIdleFlag));

	__JUMP(,lr);

    asm("__iAllEngagedCpusMask:");

    asm(".word %a0" : : "i" ((TInt)&TIdleSupport::iAllEngagedCpusMask));//

    }

/** 

Wait for all CPUs to reach the sync point. A CPU will only exit this function when all other CPUs

have reached it. 

Works like this:

cpuMask = 1 << NKern::CurrentCpu()

BEGIN_ATOMIC

    stage = iStageAndCPUWaitingMask >> 16

    waitingCpus = iStageAndCPUWaitingMask&iAllCpusMask

    oldstage = stage;

    if (waitingCpus == iAllCpusMask) // we synched already and this is new 

             waitingCpus = 0;

     waitingCpus |= cpuMask

     if (waitingCpus == iAllCpusMask) stage++

     iStageAndCPUWaitingMask = (stage << 16) | waitingCpus

END_ATOMIC

FOREVER 

   if (oldstage!=stage) return

   stage = iStageAndCPUWaitingMask >> 16 // reread stage

END_FOREVER   

*/

__NAKED__ void TSyncPoint::DoSW(TUint32 /*aCpuMask*/)

    {

    asm("stmfd sp!, {r4-r5,lr} ");	

    asm("add r0,r0,#%a0" : : "i"  _FOFF(TSyncPointBase, iStageAndCPUWaitingMask)); // skip vt

    asm("ldr r4,[r0,#4]"); 

    asm("ldr r4,[r4]")

   __DATA_MEMORY_BARRIER_Z__(r12);          // 

    asm("1: ");

	LDREX(2,0);                             // r2 =  iStageAndCPUWaitingMask, r4 = iAllEnagedCpusMask

    asm("mov r5,r2,lsr #16");               // r5 has old staging value

    asm("and r2,r2,r4");                    // r2 has currently waiting cpus

    asm("cmp r2,r4");                       // if r2 == r4 then we previously have had all cpus synched 

    asm("moveq r2,#0");                     // reset 

    asm("orr r2, r2, r1");                  // orr mask for this CPU 

    asm("mov r3,r5");                       // r3 will have new stage

    asm("cmp r2,r4");                       // if r2 == r4 then all cpus have set

    asm("addeq r3,r3,#1");                  // increment new stage count

    asm("orr r2,r2,r3, lsl #16");          

	STREX(12,2,0);                           // try to atomically iStageAndCPUWaitingMask

	asm("cmp r12, #0 ");

	asm("bne 1b ");                         // write didn't succeed try again

    __DATA_MEMORY_BARRIER__(r12);          // ensure that's written

#ifdef SYNCPOINT_WFE	

	asm("ands r2,r2,r4");

    asm("cmp r2,r4");                       // if r2 == r4 then all cpus have set      			

	ARM_SEVcc(CC_EQ);

#endif		

    asm("2: ");

    asm("cmp r3,r5");                       // all (old stage does not equal new stage)

    asm("ldmfdne sp!, {r4-r5,pc}");         // yup return

#ifdef SYNCPOINT_WFE		

	__DATA_MEMORY_BARRIER__(r12);        

	ARM_WFE;

#endif	

    asm("ldr r2,[r0]");                     // otherwise re read iWaitingCpusMask into r5

    __DATA_MEMORY_BARRIER__(r12);           // ensure read is observed

    asm("mov r3,r2,lsr #16");               // re-read new stage

    asm("b 2b");                            // loop back

    }

/** 

Wait for all CPUs to reach the sync point. A CPU will only exit this function when all other CPUs

have reached it or if another CPU has called the Break function. An attempt to wait on a broken

syncpoint will return immediately.

Works like this:

cpuMask = 1 << NKern::CurrentCpu()

BEGIN_ATOMIC

     waitingCpus = iStageAndCPUWaitingMask&iAllEnagedCpusMask

     if (iStageAndCPUWaitingMask & 0x80000000) // sync point is broken

        END_ATOMIC and return

     waitingCpus |= cpuMask

     if (waitingCpus == iAllEnagedCpusMask) waitingCpus |= 0x80000000

     iStageAndCPUWaitingMask = waitingCpus

END_ATOMIC

FOREVER 

     if (iStageAndCPUWaitingMask&0x80000000) break

END_FOREVER   

*/

__NAKED__ void TBreakableSyncPoint::DoSW(TUint32 /*aCpuMask*/)

    {

    asm("stmfd sp!, {r4,lr} ");	

    asm("add r0,r0,#%a0" : : "i"  _FOFF(TSyncPointBase, iStageAndCPUWaitingMask)); // skip vt

    asm("ldr r4,[r0,#4]");

    asm("ldr r4,[r4]")

    __DATA_MEMORY_BARRIER_Z__(r12);          // 

    asm("1: ");

	LDREX(2,0);                             // r2 =  iStageAndCPUWaitingMask, r4 = iAllEnagedCpusMask

    asm("ands r3,r2,#0x80000000");          // 

    asm("bne 3f");                          // sync point broken so return

    asm("and r2,r2,r4");                    // r2 has currently waiting cpus

    asm("orr r2, r2, r1");                  // orr mask for this CPU 

    asm("cmp r2,r4");                       // if r2 == r4 then all cpus have set

    asm("orreq r2,r2,#0x80000000");         // set MSB

	STREX(12,2,0);                          // try to atomically iStageAndCPUWaitingMask

	asm("cmp r12, #0 ");

	asm("bne 1b ");                         // write didn't succeed try again		

    __DATA_MEMORY_BARRIER__(r12);           // ensure that's written

#ifdef SYNCPOINT_WFE	

	asm("ands r3,r2,#0x80000000");          // MSB set?

	ARM_SEVcc(CC_NE);

#endif		

    asm("2: ");

    asm("ands r3,r2,#0x80000000");          // MSB set?	

    asm("ldmfdne sp!, {r4,pc}");            // yup return

#ifdef SYNCPOINT_WFE		

	__DATA_MEMORY_BARRIER__(r12);

	ARM_WFE;

#endif	

    asm("ldr r2,[r0]");                     // otherwise re read iWaitingCpusMask into r5

    //__DATA_MEMORY_BARRIER_Z__(r12);        // ensure read is observed

    asm("b 2b");                            // loop back

    asm("3:");

    CLREX;

#ifdef SYNCPOINT_WFE	 

    __DATA_MEMORY_BARRIER__(r12);           // ensure that's written

	ARM_SEV;

#endif	

    asm("ldmfd sp!, {r4,pc}");            // yup return

    }

#ifdef PROPER_WFI

__NAKED__ void TIdleSupport::DoWFI()

    {

    __DATA_SYNC_BARRIER_Z__(r12);  // generally good idea to a barrier before WFI            

    ARM_WFI;

	__JUMP(,lr);

    }

#else

void TIdleSupport::DoWFI()

    {

    TInt c=NKern::CurrentCpu();

    FOREVER

        {

        TInt isr = Pending();

        if (isr!=1023) 

            {

            BTRACE0(KIsrPendingCat,isr&0xff);

            break;

            }

        }

    }

#endif

__NAKED__ void TIdleSupport::DoIdleIPI(TUint32 /*aMask*/) 

    {

	//r0 = cpu mask

    asm("ldr    r2,__EnagedCpusMask");  // only IPI enaged cores r2 has enaged core mask addr

    asm("ldr    r2,[r2]");   

    asm("and    r0,r0,r2");            // and out retired cores

    asm("ldr	r1,__KGICAddr");//r1 = address off iGlobalIntDistAddress

    asm("ldr	r1, [r1]");//r1 = address of Hw GIC interrupt dispatcher base			

    __DATA_SYNC_BARRIER_Z__(r12);			// need DSB before sending any IPI

    asm("movs	r0, r0, lsl #16 ");		// CPU mask into bits 16-23 - any bits set in aMask?

    asm("orrne	r0, r0, #%a0" : : "i" ((TInt) IDLE_WAKEUP_IPI_VECTOR));

    asm("strne	r0, [r1, #%a0]" : : "i" _FOFF(GicDistributor, iSoftIrq));	// trigger IPIs if any

    __JUMP(,lr);

    asm("__KGICAddr:");

    asm(".word %a0" : : "i" ((TInt)&TIdleSupport::iGlobalIntDistAddress));

    asm("__EnagedCpusMask:");

    asm(".word %a0" : : "i" ((TInt)&TIdleSupport::iAllEngagedCpusMask));

    }

#ifdef _DEBUG

__NAKED__  TInt TIdleSupport::DoClearIdleIPI()

#else

__NAKED__  void TIdleSupport::ClearIdleIPI()

#endif

    {

    __DATA_SYNC_BARRIER_Z__(r12);			// DSB    

    asm("ldr    r1,__KCPUIFAddr");//r1 = address of iBaseIntIfAddress

    asm("ldr	r1, [r1]");//r1 = address of Hw GIC CPU interrupt interface base address

    asm("ldr    r0,[r1, #%a0]" : : "i" _FOFF(GicCpuIfc, iAck));

    // asm("mov    r0,#%a0" : : "i" ((TInt) IDLE_WAKEUP_IPI_VECTOR)); // has to be!

    asm("str    r0, [r1, #%a0]" : : "i" _FOFF(GicCpuIfc, iEoi));

	__JUMP(,lr);

    asm("__KCPUIFAddr:");

    asm(".word %a0" : : "i" ((TInt)&TIdleSupport::iBaseIntIfAddress));// CPU interrupt interface base address

    }

#ifndef _DEBUG

TInt TIdleSupport::DoClearIdleIPI()

    {

    return 0;

    }

#endif

__NAKED__  TInt TIdleSupport::IntPending()  

    {

	asm("ldr    r1,__KCPUIFAddr");//r1 = address of iBaseIntIfAddress

	asm("ldr	r1, [r1]");//r1 = address of Hw GIC CPU interrupt interface base address

	asm("ldr    r0, [r1, #%a0]" : : "i" _FOFF(GicCpuIfc, iHighestPending));    

	__JUMP(,lr);

    }	

#endif // ifdef __SMP__