[IA64] VIRT_CPU_ACCOUNTING (accurate cpu time accounting)

This patch implements VIRT_CPU_ACCOUNTING for ia64,
which enable us to use more accurate cpu time accounting.

The VIRT_CPU_ACCOUNTING is an item of kernel config, which s390
and powerpc arch have.  By turning this config on, these archs
change the mechanism of cpu time accounting from tick-sampling
based one to state-transition based one.

The state-transition based accounting is done by checking time
(cycle counter in processor) at every state-transition point,
such as entrance/exit of kernel, interrupt, softirq etc.
The difference between point to point is the actual time consumed
during in the state. There is no doubt about that this value is
more accurate than that of tick-sampling based accounting.

Signed-off-by: Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>

7 files changed, 278 insertions, 0 deletions

diff --git a/arch/ia64/kernel/asm-offsets.c b/arch/ia64/kernel/asm-offsets.c
index 0aebc6f..5865130 100644
--- a/arch/ia64/kernel/asm-offsets.c
+++ b/arch/ia64/kernel/asm-offsets.c
@@ -39,6 +39,12 @@ void foo(void)
 	DEFINE(TI_FLAGS, offsetof(struct thread_info, flags));
 	DEFINE(TI_CPU, offsetof(struct thread_info, cpu));
 	DEFINE(TI_PRE_COUNT, offsetof(struct thread_info, preempt_count));
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	DEFINE(TI_AC_STAMP, offsetof(struct thread_info, ac_stamp));
+	DEFINE(TI_AC_LEAVE, offsetof(struct thread_info, ac_leave));
+	DEFINE(TI_AC_STIME, offsetof(struct thread_info, ac_stime));
+	DEFINE(TI_AC_UTIME, offsetof(struct thread_info, ac_utime));
+#endif
 
 	BLANK();
 
diff --git a/arch/ia64/kernel/entry.S b/arch/ia64/kernel/entry.S
index 3c331c4..b0be4a2 100644
--- a/arch/ia64/kernel/entry.S
+++ b/arch/ia64/kernel/entry.S
@@ -710,6 +710,16 @@ ENTRY(ia64_leave_syscall)
 (pUStk)	cmp.eq.unc p6,p0=r0,r0		// p6 <- pUStk
 #endif
 .work_processed_syscall:
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	adds r2=PT(LOADRS)+16,r12
+(pUStk)	mov.m r22=ar.itc			// fetch time at leave
+	adds r18=TI_FLAGS+IA64_TASK_SIZE,r13
+	;;
+(p6)	ld4 r31=[r18]				// load current_thread_info()->flags
+	ld8 r19=[r2],PT(B6)-PT(LOADRS)		// load ar.rsc value for "loadrs"
+	adds r3=PT(AR_BSPSTORE)+16,r12		// deferred
+	;;
+#else
 	adds r2=PT(LOADRS)+16,r12
 	adds r3=PT(AR_BSPSTORE)+16,r12
 	adds r18=TI_FLAGS+IA64_TASK_SIZE,r13
@@ -718,6 +728,7 @@ ENTRY(ia64_leave_syscall)
 	ld8 r19=[r2],PT(B6)-PT(LOADRS)		// load ar.rsc value for "loadrs"
 	nop.i 0
 	;;
+#endif
 	mov r16=ar.bsp				// M2  get existing backing store pointer
 	ld8 r18=[r2],PT(R9)-PT(B6)		// load b6
 (p6)	and r15=TIF_WORK_MASK,r31		// any work other than TIF_SYSCALL_TRACE?
@@ -737,12 +748,21 @@ ENTRY(ia64_leave_syscall)
 
 	ld8 r29=[r2],16		// M0|1 load cr.ipsr
 	ld8 r28=[r3],16		// M0|1 load cr.iip
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+(pUStk) add r14=TI_AC_LEAVE+IA64_TASK_SIZE,r13
+	;;
+	ld8 r30=[r2],16		// M0|1 load cr.ifs
+	ld8 r25=[r3],16		// M0|1 load ar.unat
+(pUStk) add r15=IA64_TASK_THREAD_ON_USTACK_OFFSET,r13
+	;;
+#else
 	mov r22=r0		// A    clear r22
 	;;
 	ld8 r30=[r2],16		// M0|1 load cr.ifs
 	ld8 r25=[r3],16		// M0|1 load ar.unat
 (pUStk) add r14=IA64_TASK_THREAD_ON_USTACK_OFFSET,r13
 	;;
+#endif
 	ld8 r26=[r2],PT(B0)-PT(AR_PFS)	// M0|1 load ar.pfs
 (pKStk)	mov r22=psr			// M2   read PSR now that interrupts are disabled
 	nop 0
@@ -759,7 +779,11 @@ ENTRY(ia64_leave_syscall)
 	ld8.fill r1=[r3],16			// M0|1 load r1
 (pUStk) mov r17=1				// A
 	;;
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+(pUStk) st1 [r15]=r17				// M2|3
+#else
 (pUStk) st1 [r14]=r17				// M2|3
+#endif
 	ld8.fill r13=[r3],16			// M0|1
 	mov f8=f0				// F    clear f8
 	;;
@@ -775,12 +799,22 @@ ENTRY(ia64_leave_syscall)
 	shr.u r18=r19,16		// I0|1 get byte size of existing "dirty" partition
 	cover				// B    add current frame into dirty partition & set cr.ifs
 	;;
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	mov r19=ar.bsp			// M2   get new backing store pointer
+	st8 [r14]=r22			// M	save time at leave
+	mov f10=f0			// F    clear f10
+
+	mov r22=r0			// A	clear r22
+	movl r14=__kernel_syscall_via_epc // X
+	;;
+#else
 	mov r19=ar.bsp			// M2   get new backing store pointer
 	mov f10=f0			// F    clear f10
 
 	nop.m 0
 	movl r14=__kernel_syscall_via_epc // X
 	;;
+#endif
 	mov.m ar.csd=r0			// M2   clear ar.csd
 	mov.m ar.ccv=r0			// M2   clear ar.ccv
 	mov b7=r14			// I0   clear b7 (hint with __kernel_syscall_via_epc)
@@ -913,10 +947,18 @@ GLOBAL_ENTRY(ia64_leave_kernel)
 	adds r16=PT(CR_IPSR)+16,r12
 	adds r17=PT(CR_IIP)+16,r12
 
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	.pred.rel.mutex pUStk,pKStk
+(pKStk)	mov r22=psr		// M2 read PSR now that interrupts are disabled
+(pUStk)	mov.m r22=ar.itc	// M  fetch time at leave
+	nop.i 0
+	;;
+#else
 (pKStk)	mov r22=psr		// M2 read PSR now that interrupts are disabled
 	nop.i 0
 	nop.i 0
 	;;
+#endif
 	ld8 r29=[r16],16	// load cr.ipsr
 	ld8 r28=[r17],16	// load cr.iip
 	;;
@@ -938,15 +980,37 @@ GLOBAL_ENTRY(ia64_leave_kernel)
 	;;
 	ld8.fill r12=[r16],16
 	ld8.fill r13=[r17],16
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+(pUStk)	adds r3=TI_AC_LEAVE+IA64_TASK_SIZE,r18
+#else
 (pUStk)	adds r18=IA64_TASK_THREAD_ON_USTACK_OFFSET,r18
+#endif
 	;;
 	ld8 r20=[r16],16	// ar.fpsr
 	ld8.fill r15=[r17],16
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+(pUStk)	adds r18=IA64_TASK_THREAD_ON_USTACK_OFFSET,r18	// deferred
+#endif
 	;;
 	ld8.fill r14=[r16],16
 	ld8.fill r2=[r17]
 (pUStk)	mov r17=1
 	;;
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	//  mmi_ :  ld8 st1 shr;;         mmi_ : st8 st1 shr;;
+	//  mib  :  mov add br        ->  mib  : ld8 add br
+	//  bbb_ :  br  nop cover;;       mbb_ : mov br  cover;;
+	//
+	//  no one require bsp in r16 if (pKStk) branch is selected.
+(pUStk)	st8 [r3]=r22		// save time at leave
+(pUStk)	st1 [r18]=r17		// restore current->thread.on_ustack
+	shr.u r18=r19,16	// get byte size of existing "dirty" partition
+	;;
+	ld8.fill r3=[r16]	// deferred
+	LOAD_PHYS_STACK_REG_SIZE(r17)
+(pKStk)	br.cond.dpnt skip_rbs_switch
+	mov r16=ar.bsp		// get existing backing store pointer
+#else
 	ld8.fill r3=[r16]
 (pUStk)	st1 [r18]=r17		// restore current->thread.on_ustack
 	shr.u r18=r19,16	// get byte size of existing "dirty" partition
@@ -954,6 +1018,7 @@ GLOBAL_ENTRY(ia64_leave_kernel)
 	mov r16=ar.bsp		// get existing backing store pointer
 	LOAD_PHYS_STACK_REG_SIZE(r17)
 (pKStk)	br.cond.dpnt skip_rbs_switch
+#endif
 
 	/*
 	 * Restore user backing store.
diff --git a/arch/ia64/kernel/fsys.S b/arch/ia64/kernel/fsys.S
index 4484197..c932d86 100644
--- a/arch/ia64/kernel/fsys.S
+++ b/arch/ia64/kernel/fsys.S
@@ -660,7 +660,11 @@ GLOBAL_ENTRY(fsys_bubble_down)
 	nop.i 0
 	;;
 	mov ar.rsc=0				// M2   set enforced lazy mode, pl 0, LE, loadrs=0
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	mov.m r30=ar.itc			// M    get cycle for accounting
+#else
 	nop.m 0
+#endif
 	nop.i 0
 	;;
 	mov r23=ar.bspstore			// M2 (12 cyc) save ar.bspstore
@@ -682,6 +686,28 @@ GLOBAL_ENTRY(fsys_bubble_down)
 	cmp.ne pKStk,pUStk=r0,r0		// A    set pKStk <- 0, pUStk <- 1
 	br.call.sptk.many b7=ia64_syscall_setup	// B
 	;;
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	// mov.m r30=ar.itc is called in advance
+	add r16=TI_AC_STAMP+IA64_TASK_SIZE,r2
+	add r17=TI_AC_LEAVE+IA64_TASK_SIZE,r2
+	;;
+	ld8 r18=[r16],TI_AC_STIME-TI_AC_STAMP	// time at last check in kernel
+	ld8 r19=[r17],TI_AC_UTIME-TI_AC_LEAVE	// time at leave kernel
+	;;
+	ld8 r20=[r16],TI_AC_STAMP-TI_AC_STIME	// cumulated stime
+	ld8 r21=[r17]				// cumulated utime
+	sub r22=r19,r18				// stime before leave kernel
+	;;
+	st8 [r16]=r30,TI_AC_STIME-TI_AC_STAMP	// update stamp
+	sub r18=r30,r19				// elapsed time in user mode
+	;;
+	add r20=r20,r22				// sum stime
+	add r21=r21,r18				// sum utime
+	;;
+	st8 [r16]=r20				// update stime
+	st8 [r17]=r21				// update utime
+	;;
+#endif
 	mov ar.rsc=0x3				// M2   set eager mode, pl 0, LE, loadrs=0
 	mov rp=r14				// I0   set the real return addr
 	and r3=_TIF_SYSCALL_TRACEAUDIT,r3	// A
diff --git a/arch/ia64/kernel/head.S b/arch/ia64/kernel/head.S
index d3a41d5..ddeab4e 100644
--- a/arch/ia64/kernel/head.S
+++ b/arch/ia64/kernel/head.S
@@ -1002,6 +1002,26 @@ GLOBAL_ENTRY(sched_clock)
 	br.ret.sptk.many rp
 END(sched_clock)
 
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+GLOBAL_ENTRY(cycle_to_cputime)
+	alloc r16=ar.pfs,1,0,0,0
+	addl r8=THIS_CPU(cpu_info) + IA64_CPUINFO_NSEC_PER_CYC_OFFSET,r0
+	;;
+	ldf8 f8=[r8]
+	;;
+	setf.sig f9=r32
+	;;
+	xmpy.lu f10=f9,f8	// calculate low 64 bits of 128-bit product	(4 cyc)
+	xmpy.hu f11=f9,f8	// calculate high 64 bits of 128-bit product
+	;;
+	getf.sig r8=f10		//						(5 cyc)
+	getf.sig r9=f11
+	;;
+	shrp r8=r9,r8,IA64_NSEC_PER_CYC_SHIFT
+	br.ret.sptk.many rp
+END(cycle_to_cputime)
+#endif /* CONFIG_VIRT_CPU_ACCOUNTING */
+
 GLOBAL_ENTRY(start_kernel_thread)
 	.prologue
 	.save rp, r0				// this is the end of the call-chain
diff --git a/arch/ia64/kernel/ivt.S b/arch/ia64/kernel/ivt.S
index 34f44d8..6678c49 100644
--- a/arch/ia64/kernel/ivt.S
+++ b/arch/ia64/kernel/ivt.S
@@ -805,8 +805,13 @@ ENTRY(break_fault)
 
 (p8)	adds r28=16,r28				// A    switch cr.iip to next bundle
 (p9)	adds r8=1,r8				// A    increment ei to next slot
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	;;
+	mov b6=r30				// I0   setup syscall handler branch reg early
+#else
 	nop.i 0
 	;;
+#endif
 
 	mov.m r25=ar.unat			// M2 (5 cyc)
 	dep r29=r8,r29,41,2			// I0   insert new ei into cr.ipsr
@@ -817,7 +822,11 @@ ENTRY(break_fault)
 	//
 ///////////////////////////////////////////////////////////////////////
 	st1 [r16]=r0				// M2|3 clear current->thread.on_ustack flag
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	mov.m r30=ar.itc			// M    get cycle for accounting
+#else
 	mov b6=r30				// I0   setup syscall handler branch reg early
+#endif
 	cmp.eq pKStk,pUStk=r0,r17		// A    were we on kernel stacks already?
 
 	and r9=_TIF_SYSCALL_TRACEAUDIT,r9	// A    mask trace or audit
@@ -829,6 +838,30 @@ ENTRY(break_fault)
 	cmp.eq p14,p0=r9,r0			// A    are syscalls being traced/audited?
 	br.call.sptk.many b7=ia64_syscall_setup	// B
 1:
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	// mov.m r30=ar.itc is called in advance, and r13 is current
+	add r16=TI_AC_STAMP+IA64_TASK_SIZE,r13	// A
+	add r17=TI_AC_LEAVE+IA64_TASK_SIZE,r13	// A
+(pKStk)	br.cond.spnt .skip_accounting		// B	unlikely skip
+	;;
+	ld8 r18=[r16],TI_AC_STIME-TI_AC_STAMP	// M  get last stamp
+	ld8 r19=[r17],TI_AC_UTIME-TI_AC_LEAVE	// M  time at leave
+	;;
+	ld8 r20=[r16],TI_AC_STAMP-TI_AC_STIME	// M  cumulated stime
+	ld8 r21=[r17]				// M  cumulated utime
+	sub r22=r19,r18				// A  stime before leave
+	;;
+	st8 [r16]=r30,TI_AC_STIME-TI_AC_STAMP	// M  update stamp
+	sub r18=r30,r19				// A  elapsed time in user
+	;;
+	add r20=r20,r22				// A  sum stime
+	add r21=r21,r18				// A  sum utime
+	;;
+	st8 [r16]=r20				// M  update stime
+	st8 [r17]=r21				// M  update utime
+	;;
+.skip_accounting:
+#endif
 	mov ar.rsc=0x3				// M2   set eager mode, pl 0, LE, loadrs=0
 	nop 0
 	bsw.1					// B (6 cyc) regs are saved, switch to bank 1
@@ -928,6 +961,7 @@ END(interrupt)
 	 *	- r27: saved ar.rsc
 	 *	- r28: saved cr.iip
 	 *	- r29: saved cr.ipsr
+	 *	- r30: ar.itc for accounting (don't touch)
 	 *	- r31: saved pr
 	 *	-  b0: original contents (to be saved)
 	 * On exit:
@@ -1090,6 +1124,41 @@ END(dispatch_illegal_op_fault)
 	DBG_FAULT(16)
 	FAULT(16)
 
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	/*
+	 * There is no particular reason for this code to be here, other than
+	 * that there happens to be space here that would go unused otherwise.
+	 * If this fault ever gets "unreserved", simply moved the following
+	 * code to a more suitable spot...
+	 *
+	 * account_sys_enter is called from SAVE_MIN* macros if accounting is
+	 * enabled and if the macro is entered from user mode.
+	 */
+ENTRY(account_sys_enter)
+	// mov.m r20=ar.itc is called in advance, and r13 is current
+	add r16=TI_AC_STAMP+IA64_TASK_SIZE,r13
+	add r17=TI_AC_LEAVE+IA64_TASK_SIZE,r13
+	;;
+	ld8 r18=[r16],TI_AC_STIME-TI_AC_STAMP	// time at last check in kernel
+	ld8 r19=[r17],TI_AC_UTIME-TI_AC_LEAVE	// time at left from kernel
+        ;;
+	ld8 r23=[r16],TI_AC_STAMP-TI_AC_STIME	// cumulated stime
+	ld8 r21=[r17]				// cumulated utime
+	sub r22=r19,r18				// stime before leave kernel
+	;;
+	st8 [r16]=r20,TI_AC_STIME-TI_AC_STAMP	// update stamp
+	sub r18=r20,r19				// elapsed time in user mode
+	;;
+	add r23=r23,r22				// sum stime
+	add r21=r21,r18				// sum utime
+	;;
+	st8 [r16]=r23				// update stime
+	st8 [r17]=r21				// update utime
+	;;
+	br.ret.sptk.many rp
+END(account_sys_enter)
+#endif
+
 	.org ia64_ivt+0x4400
 /////////////////////////////////////////////////////////////////////////////////////////
 // 0x4400 Entry 17 (size 64 bundles) Reserved
diff --git a/arch/ia64/kernel/minstate.h b/arch/ia64/kernel/minstate.h
index c9ac8ba..7c548ac 100644
--- a/arch/ia64/kernel/minstate.h
+++ b/arch/ia64/kernel/minstate.h
@@ -3,6 +3,18 @@
 
 #include "entry.h"
 
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+/* read ar.itc in advance, and use it before leaving bank 0 */
+#define ACCOUNT_GET_STAMP				\
+(pUStk) mov.m r20=ar.itc;
+#define ACCOUNT_SYS_ENTER				\
+(pUStk) br.call.spnt rp=account_sys_enter		\
+	;;
+#else
+#define ACCOUNT_GET_STAMP
+#define ACCOUNT_SYS_ENTER
+#endif
+
 /*
  * DO_SAVE_MIN switches to the kernel stacks (if necessary) and saves
  * the minimum state necessary that allows us to turn psr.ic back
@@ -122,11 +134,13 @@
 	;;											\
 .mem.offset 0,0; st8.spill [r16]=r2,16;								\
 .mem.offset 8,0; st8.spill [r17]=r3,16;								\
+	ACCOUNT_GET_STAMP									\
 	adds r2=IA64_PT_REGS_R16_OFFSET,r1;							\
 	;;											\
 	EXTRA;											\
 	movl r1=__gp;		/* establish kernel global pointer */				\
 	;;											\
+	ACCOUNT_SYS_ENTER									\
 	bsw.1;			/* switch back to bank 1 (must be last in insn group) */	\
 	;;
 
diff --git a/arch/ia64/kernel/time.c b/arch/ia64/kernel/time.c
index 17fda52..48e15a5 100644
--- a/arch/ia64/kernel/time.c
+++ b/arch/ia64/kernel/time.c
@@ -59,6 +59,84 @@ static struct clocksource clocksource_itc = {
 };
 static struct clocksource *itc_clocksource;
 
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+
+#include <linux/kernel_stat.h>
+
+extern cputime_t cycle_to_cputime(u64 cyc);
+
+/*
+ * Called from the context switch with interrupts disabled, to charge all
+ * accumulated times to the current process, and to prepare accounting on
+ * the next process.
+ */
+void ia64_account_on_switch(struct task_struct *prev, struct task_struct *next)
+{
+	struct thread_info *pi = task_thread_info(prev);
+	struct thread_info *ni = task_thread_info(next);
+	cputime_t delta_stime, delta_utime;
+	__u64 now;
+
+	now = ia64_get_itc();
+
+	delta_stime = cycle_to_cputime(pi->ac_stime + (now - pi->ac_stamp));
+	account_system_time(prev, 0, delta_stime);
+	account_system_time_scaled(prev, delta_stime);
+
+	if (pi->ac_utime) {
+		delta_utime = cycle_to_cputime(pi->ac_utime);
+		account_user_time(prev, delta_utime);
+		account_user_time_scaled(prev, delta_utime);
+	}
+
+	pi->ac_stamp = ni->ac_stamp = now;
+	ni->ac_stime = ni->ac_utime = 0;
+}
+
+/*
+ * Account time for a transition between system, hard irq or soft irq state.
+ * Note that this function is called with interrupts enabled.
+ */
+void account_system_vtime(struct task_struct *tsk)
+{
+	struct thread_info *ti = task_thread_info(tsk);
+	unsigned long flags;
+	cputime_t delta_stime;
+	__u64 now;
+
+	local_irq_save(flags);
+
+	now = ia64_get_itc();
+
+	delta_stime = cycle_to_cputime(ti->ac_stime + (now - ti->ac_stamp));
+	account_system_time(tsk, 0, delta_stime);
+	account_system_time_scaled(tsk, delta_stime);
+	ti->ac_stime = 0;
+
+	ti->ac_stamp = now;
+
+	local_irq_restore(flags);
+}
+
+/*
+ * Called from the timer interrupt handler to charge accumulated user time
+ * to the current process.  Must be called with interrupts disabled.
+ */
+void account_process_tick(struct task_struct *p, int user_tick)
+{
+	struct thread_info *ti = task_thread_info(p);
+	cputime_t delta_utime;
+
+	if (ti->ac_utime) {
+		delta_utime = cycle_to_cputime(ti->ac_utime);
+		account_user_time(p, delta_utime);
+		account_user_time_scaled(p, delta_utime);
+		ti->ac_utime = 0;
+	}
+}
+
+#endif /* CONFIG_VIRT_CPU_ACCOUNTING */
+
 static irqreturn_t
 timer_interrupt (int irq, void *dev_id)
 {