[S390] virtualization aware cpu measurement

Use the SPP instruction to set a tag on entry to / exit of the virtual
machine context. This allows the cpu measurement facility to distinguish
the samples from the host and the different guests.
Signed-off-by: Carsten Otte <cotte@de.ibm.com>

arch/s390/include/asm/lowcore.h

@@ -267,7 +267,8 @@ struct _lowcore {
 	__u64	vdso_per_cpu_data;		/* 0x0358 */
 	__u64	machine_flags;			/* 0x0360 */
 	__u64	ftrace_func;			/* 0x0368 */
-	__u8	pad_0x0370[0x0380-0x0370];	/* 0x0370 */
+	__u64	sie_hook;			/* 0x0370 */
+	__u64	cmf_hpp;			/* 0x0378 */
 
 	/* Interrupt response block. */
 	__u8	irb[64];			/* 0x0380 */

arch/s390/include/asm/setup.h

@@ -2,7 +2,7 @@
  *  include/asm-s390/setup.h
  *
  *  S390 version
- *    Copyright IBM Corp. 1999,2006
+ *    Copyright IBM Corp. 1999,2010
  */
 
 #ifndef _ASM_S390_SETUP_H
@@ -72,6 +72,7 @@ extern unsigned int user_mode;
 #define MACHINE_FLAG_HPAGE	(1UL << 10)
 #define MACHINE_FLAG_PFMF	(1UL << 11)
 #define MACHINE_FLAG_LPAR	(1UL << 12)
+#define MACHINE_FLAG_SPP	(1UL << 13)
 
 #define MACHINE_IS_VM		(S390_lowcore.machine_flags & MACHINE_FLAG_VM)
 #define MACHINE_IS_KVM		(S390_lowcore.machine_flags & MACHINE_FLAG_KVM)
@@ -88,6 +89,7 @@ extern unsigned int user_mode;
 #define MACHINE_HAS_MVCOS	(0)
 #define MACHINE_HAS_HPAGE	(0)
 #define MACHINE_HAS_PFMF	(0)
+#define MACHINE_HAS_SPP		(0)
 #else /* __s390x__ */
 #define MACHINE_HAS_IEEE	(1)
 #define MACHINE_HAS_CSP		(1)
@@ -97,6 +99,7 @@ extern unsigned int user_mode;
 #define MACHINE_HAS_MVCOS	(S390_lowcore.machine_flags & MACHINE_FLAG_MVCOS)
 #define MACHINE_HAS_HPAGE	(S390_lowcore.machine_flags & MACHINE_FLAG_HPAGE)
 #define MACHINE_HAS_PFMF	(S390_lowcore.machine_flags & MACHINE_FLAG_PFMF)
+#define MACHINE_HAS_SPP		(S390_lowcore.machine_flags & MACHINE_FLAG_SPP)
 #endif /* __s390x__ */
 
 #define ZFCPDUMP_HSA_SIZE	(32UL<<20)

arch/s390/kernel/asm-offsets.c

@@ -131,6 +131,8 @@ int main(void)
 	DEFINE(__LC_MCCK_CLOCK, offsetof(struct _lowcore, mcck_clock));
 	DEFINE(__LC_MACHINE_FLAGS, offsetof(struct _lowcore, machine_flags));
 	DEFINE(__LC_FTRACE_FUNC, offsetof(struct _lowcore, ftrace_func));
+	DEFINE(__LC_SIE_HOOK, offsetof(struct _lowcore, sie_hook));
+	DEFINE(__LC_CMF_HPP, offsetof(struct _lowcore, cmf_hpp));
 	DEFINE(__LC_IRB, offsetof(struct _lowcore, irb));
 	DEFINE(__LC_CPU_TIMER_SAVE_AREA, offsetof(struct _lowcore, cpu_timer_save_area));
 	DEFINE(__LC_CLOCK_COMP_SAVE_AREA, offsetof(struct _lowcore, clock_comp_save_area));

arch/s390/kernel/early.c

@@ -356,6 +356,7 @@ static __init void detect_machine_facilities(void)
 {
 #ifdef CONFIG_64BIT
 	unsigned int facilities;
+	unsigned long long facility_bits;
 
 	facilities = stfl();
 	if (facilities & (1 << 28))
@@ -364,6 +365,9 @@ static __init void detect_machine_facilities(void)
 		S390_lowcore.machine_flags |= MACHINE_FLAG_PFMF;
 	if (facilities & (1 << 4))
 		S390_lowcore.machine_flags |= MACHINE_FLAG_MVCOS;
+	if ((stfle(&facility_bits, 1) > 0) &&
+	    (facility_bits & (1ULL << (63 - 40))))
+		S390_lowcore.machine_flags |= MACHINE_FLAG_SPP;
 #endif
 }
 

arch/s390/kernel/entry64.S

@@ -2,7 +2,7 @@
  *  arch/s390/kernel/entry64.S
  *    S390 low-level entry points.
  *
- *    Copyright (C) IBM Corp. 1999,2006
+ *    Copyright (C) IBM Corp. 1999,2010
  *    Author(s): Martin Schwidefsky (schwidefsky@de.ibm.com),
  *		 Hartmut Penner (hp@de.ibm.com),
  *		 Denis Joseph Barrow (djbarrow@de.ibm.com,barrow_dj@yahoo.com),
@@ -59,6 +59,16 @@ _TIF_SYSCALL = (_TIF_SYSCALL_TRACE>>8 | _TIF_SYSCALL_AUDIT>>8 | \
 
 #define BASED(name) name-system_call(%r13)
 
+	.macro	HANDLE_SIE_INTERCEPT
+#if defined(CONFIG_KVM) || defined(CONFIG_KVM_MODULE)
+	lg	%r3,__LC_SIE_HOOK
+	ltgr	%r3,%r3
+	jz	0f
+	basr	%r14,%r3
+	0:
+#endif
+	.endm
+
 #ifdef CONFIG_TRACE_IRQFLAGS
 	.macro	TRACE_IRQS_ON
 	basr	%r2,%r0
@@ -466,6 +476,7 @@ pgm_check_handler:
 	UPDATE_VTIME __LC_LAST_UPDATE_TIMER,__LC_EXIT_TIMER,__LC_SYSTEM_TIMER
 	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_SYNC_ENTER_TIMER
 pgm_no_vtime:
+	HANDLE_SIE_INTERCEPT
 	TRACE_IRQS_CHECK_OFF
 	lg	%r9,__LC_THREAD_INFO	# load pointer to thread_info struct
 	mvc	SP_ARGS(8,%r15),__LC_LAST_BREAK
@@ -507,6 +518,7 @@ pgm_per_std:
 	UPDATE_VTIME __LC_LAST_UPDATE_TIMER,__LC_EXIT_TIMER,__LC_SYSTEM_TIMER
 	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_SYNC_ENTER_TIMER
 pgm_no_vtime2:
+	HANDLE_SIE_INTERCEPT
 	TRACE_IRQS_CHECK_OFF
 	lg	%r9,__LC_THREAD_INFO	# load pointer to thread_info struct
 	lg	%r1,__TI_task(%r9)
@@ -570,6 +582,7 @@ io_int_handler:
 	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_ASYNC_ENTER_TIMER
 io_no_vtime:
 	lg	%r9,__LC_THREAD_INFO	# load pointer to thread_info struct
+	HANDLE_SIE_INTERCEPT
 	TRACE_IRQS_OFF
 	la	%r2,SP_PTREGS(%r15)	# address of register-save area
 	brasl	%r14,do_IRQ		# call standard irq handler
@@ -595,15 +608,6 @@ io_done:
 io_work:
 	tm	SP_PSW+1(%r15),0x01	# returning to user ?
 	jo	io_work_user		# yes -> do resched & signal
-#if defined(CONFIG_KVM) || defined(CONFIG_KVM_MODULE)
-	lg	%r2,SP_PSW+8(%r15)	# check if current instruction is SIE
-	lh	%r1,0(%r2)
-	chi	%r1,-19948		# signed 16 bit compare with 0xb214
-	jne	0f			# no -> leave PSW alone
-	aghi	%r2,4			# yes-> add 4 bytes to leave SIE
-	stg	%r2,SP_PSW+8(%r15)
-0:
-#endif
 #ifdef CONFIG_PREEMPT
 	# check for preemptive scheduling
 	icm	%r0,15,__TI_precount(%r9)
@@ -712,6 +716,7 @@ ext_int_handler:
 	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_ASYNC_ENTER_TIMER
 ext_no_vtime:
 	lg	%r9,__LC_THREAD_INFO	# load pointer to thread_info struct
+	HANDLE_SIE_INTERCEPT
 	TRACE_IRQS_OFF
 	la	%r2,SP_PTREGS(%r15)	# address of register-save area
 	llgh	%r3,__LC_EXT_INT_CODE	# get interruption code
@@ -786,6 +791,7 @@ mcck_no_vtime:
 	stosm	__SF_EMPTY(%r15),0x04	# turn dat on
 	tm	__TI_flags+7(%r9),_TIF_MCCK_PENDING
 	jno	mcck_return
+	HANDLE_SIE_INTERCEPT
 	TRACE_IRQS_OFF
 	brasl	%r14,s390_handle_mcck
 	TRACE_IRQS_ON

arch/s390/kernel/setup.c

@@ -2,7 +2,7 @@
  *  arch/s390/kernel/setup.c
  *
  *  S390 version
- *    Copyright (C) 1999,2000 IBM Deutschland Entwicklung GmbH, IBM Corporation
+ *    Copyright (C) IBM Corp. 1999,2010
  *    Author(s): Hartmut Penner (hp@de.ibm.com),
  *               Martin Schwidefsky (schwidefsky@de.ibm.com)
  *
@@ -401,6 +401,7 @@ setup_lowcore(void)
 	lc->io_new_psw.mask = psw_kernel_bits;
 	lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler;
 	lc->clock_comparator = -1ULL;
+	lc->cmf_hpp = -1ULL;
 	lc->kernel_stack = ((unsigned long) &init_thread_union) + THREAD_SIZE;
 	lc->async_stack = (unsigned long)
 		__alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0) + ASYNC_SIZE;

arch/s390/kvm/Kconfig

@@ -33,6 +33,17 @@ config KVM
 
 	  If unsure, say N.
 
+config KVM_AWARE_CMF
+	depends on KVM
+	bool "KVM aware sampling"
+	---help---
+	  This option enhances the sampling data from the CPU Measurement
+	  Facility with additional information, that allows to distinguish
+	  guest(s) and host when using the kernel based virtual machine
+	  functionality.
+
+	  If unsure, say N.
+
 # OK, it's a little counter-intuitive to do this, but it puts it neatly under
 # the virtualization menu.
 source drivers/vhost/Kconfig

arch/s390/kvm/sie64a.S

 /*
  * sie64a.S - low level sie call
  *
- * Copyright IBM Corp. 2008
+ * Copyright IBM Corp. 2008,2010
  *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of the GNU General Public License (version 2 only)
  * as published by the Free Software Foundation.
  *
  *    Author(s): Heiko Carstens <heiko.carstens@de.ibm.com>
+ *		 Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
  */
 
 #include <linux/errno.h>
 #include <asm/asm-offsets.h>
+#include <asm/setup.h>
+#include <asm/asm-offsets.h>
+#include <asm/ptrace.h>
+#include <asm/thread_info.h>
+
+_TIF_EXIT_SIE = (_TIF_SIGPENDING | _TIF_NEED_RESCHED | _TIF_MCCK_PENDING)
+
+/*
+ * offsets into stackframe
+ * SP_	= offsets into stack sie64 is called with
+ * SPI_ = offsets into irq stack
+ */
+SP_GREGS = __SF_EMPTY
+SP_HOOK  = __SF_EMPTY+8
+SP_GPP	 = __SF_EMPTY+16
+SPI_PSW  = STACK_FRAME_OVERHEAD + __PT_PSW
+
 
-SP_R5 =	5 * 8	# offset into stackframe
-SP_R6 =	6 * 8
+	.macro SPP newpp
+#ifdef CONFIG_KVM_AWARE_CMF
+	tm	__LC_MACHINE_FLAGS+6,0x20	# MACHINE_FLAG_SPP
+	jz	0f
+	.insn	s,0xb2800000,\newpp
+	0:
+#endif
+	.endm
+
+sie_irq_handler:
+	SPP	__LC_CMF_HPP			# set host id
+	larl	%r2,sie_inst
+	clg	%r2,SPI_PSW+8(0,%r15)		# intercepted sie
+	jne	1f
+	xc	__LC_SIE_HOOK(8),__LC_SIE_HOOK
+	lg	%r2,__LC_THREAD_INFO		# pointer thread_info struct
+	tm	__TI_flags+7(%r2),_TIF_EXIT_SIE
+	jz	0f
+	larl	%r2,sie_exit			# work pending, leave sie
+	stg	%r2,__LC_RETURN_PSW+8
+	br	%r14
+0:	larl	%r2,sie_reenter			# re-enter with guest id
+	stg	%r2,__LC_RETURN_PSW+8
+1:	br	%r14
 
 /*
  * sie64a calling convention:
@@ -23,23 +63,34 @@ SP_R6 =	6 * 8
  */
 	.globl	sie64a
 sie64a:
-	lgr	%r5,%r3
-	stmg	%r5,%r14,SP_R5(%r15)	# save register on entry
-	lgr	%r14,%r2		# pointer to sie control block
-	lmg	%r0,%r13,0(%r3)		# load guest gprs 0-13
+	stg	%r3,SP_GREGS(%r15)		# save guest register save area
+	stmg	%r6,%r14,__SF_GPRS(%r15)	# save registers on entry
+	lgr	%r14,%r2			# pointer to sie control block
+	larl	%r5,sie_irq_handler
+	stg	%r2,SP_GPP(%r15)
+	stg	%r5,SP_HOOK(%r15)		# save hook target
+	lmg	%r0,%r13,0(%r3)			# load guest gprs 0-13
+sie_reenter:
+	mvc	__LC_SIE_HOOK(8),SP_HOOK(%r15)
+	SPP	SP_GPP(%r15)			# set guest id
 sie_inst:
 	sie	0(%r14)
-	lg	%r14,SP_R5(%r15)
-	stmg	%r0,%r13,0(%r14)	# save guest gprs 0-13
+	xc	__LC_SIE_HOOK(8),__LC_SIE_HOOK
+	SPP	__LC_CMF_HPP			# set host id
+sie_exit:
+	lg	%r14,SP_GREGS(%r15)
+	stmg	%r0,%r13,0(%r14)		# save guest gprs 0-13
 	lghi	%r2,0
-	lmg	%r6,%r14,SP_R6(%r15)
+	lmg	%r6,%r14,__SF_GPRS(%r15)
 	br	%r14
 
 sie_err:
-	lg	%r14,SP_R5(%r15)
-	stmg	%r0,%r13,0(%r14)	# save guest gprs 0-13
+	xc	__LC_SIE_HOOK(8),__LC_SIE_HOOK
+	SPP	__LC_CMF_HPP			# set host id
+	lg	%r14,SP_GREGS(%r15)
+	stmg	%r0,%r13,0(%r14)		# save guest gprs 0-13
 	lghi	%r2,-EFAULT
-	lmg	%r6,%r14,SP_R6(%r15)
+	lmg	%r6,%r14,__SF_GPRS(%r15)
 	br	%r14
 
 	.section __ex_table,"a"