diff -rupN linux-2.6.18/CREDITS linux-2.6.18.perfctr27/CREDITS --- linux-2.6.18/CREDITS 2007-09-27 20:28:30.000000000 +0200 +++ linux-2.6.18.perfctr27/CREDITS 2007-09-27 21:04:27.000000000 +0200 @@ -2629,9 +2629,10 @@ S: Ottawa, Ontario S: Canada K2P 0X8 N: Mikael Pettersson -E: mikpe@csd.uu.se -W: http://www.csd.uu.se/~mikpe/ +E: mikpe@it.uu.se +W: http://user.it.uu.se/~mikpe/ D: Miscellaneous fixes +D: Performance-monitoring counters driver N: Reed H. Petty E: rhp@draper.net diff -rupN linux-2.6.18/MAINTAINERS linux-2.6.18.perfctr27/MAINTAINERS --- linux-2.6.18/MAINTAINERS 2007-09-27 20:28:31.000000000 +0200 +++ linux-2.6.18.perfctr27/MAINTAINERS 2007-09-27 20:35:38.000000000 +0200 @@ -2272,6 +2272,12 @@ M: nagar@watson.ibm.com L: linux-kernel@vger.kernel.org S: Maintained +PERFORMANCE-MONITORING COUNTERS DRIVER +P: Mikael Pettersson +M: mikpe@it.uu.se +W: http://user.it.uu.se/~mikpe/linux/perfctr/ +S: Maintained + PERSONALITY HANDLING P: Christoph Hellwig M: hch@infradead.org diff -rupN linux-2.6.18/arch/i386/Kconfig linux-2.6.18.perfctr27/arch/i386/Kconfig --- linux-2.6.18/arch/i386/Kconfig 2007-09-27 20:28:32.000000000 +0200 +++ linux-2.6.18.perfctr27/arch/i386/Kconfig 2007-09-27 20:34:06.000000000 +0200 @@ -737,6 +737,8 @@ config SECCOMP If unsure, say Y. Only embedded should say N here. +source "drivers/perfctr/Kconfig" + source kernel/Kconfig.hz config KEXEC diff -rupN linux-2.6.18/arch/i386/kernel/entry.S linux-2.6.18.perfctr27/arch/i386/kernel/entry.S --- linux-2.6.18/arch/i386/kernel/entry.S 2007-09-27 20:28:32.000000000 +0200 +++ linux-2.6.18.perfctr27/arch/i386/kernel/entry.S 2007-09-27 20:34:06.000000000 +0200 @@ -591,6 +591,22 @@ ENTRY(name) \ /* The include is where all of the SMP etc. interrupts come from */ #include "entry_arch.h" +#if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_PERFCTR) +ENTRY(perfctr_interrupt) + RING0_INT_FRAME + pushl $~(LOCAL_PERFCTR_VECTOR) + CFI_ADJUST_CFA_OFFSET 4 + SAVE_ALL + TRACE_IRQS_OFF + pushl %esp + CFI_ADJUST_CFA_OFFSET 4 + call smp_perfctr_interrupt + addl $4, %esp + CFI_ADJUST_CFA_OFFSET -4 + jmp ret_from_intr + CFI_ENDPROC +#endif + ENTRY(divide_error) RING0_INT_FRAME pushl $0 # no error code diff -rupN linux-2.6.18/arch/i386/kernel/i8259.c linux-2.6.18.perfctr27/arch/i386/kernel/i8259.c --- linux-2.6.18/arch/i386/kernel/i8259.c 2007-09-27 20:28:32.000000000 +0200 +++ linux-2.6.18.perfctr27/arch/i386/kernel/i8259.c 2007-09-27 20:34:06.000000000 +0200 @@ -22,6 +22,7 @@ #include #include #include +#include #include @@ -420,6 +421,8 @@ void __init init_IRQ(void) */ intr_init_hook(); + perfctr_vector_init(); + /* * Set the clock to HZ Hz, we already have a valid * vector now: diff -rupN linux-2.6.18/arch/i386/kernel/process.c linux-2.6.18.perfctr27/arch/i386/kernel/process.c --- linux-2.6.18/arch/i386/kernel/process.c 2007-09-27 20:28:32.000000000 +0200 +++ linux-2.6.18.perfctr27/arch/i386/kernel/process.c 2007-09-27 20:34:06.000000000 +0200 @@ -32,6 +32,7 @@ #include #include #include +#include #include #include #include @@ -379,6 +380,7 @@ void exit_thread(void) tss->io_bitmap_base = INVALID_IO_BITMAP_OFFSET; put_cpu(); } + perfctr_exit_thread(¤t->thread); } void flush_thread(void) @@ -431,6 +433,8 @@ int copy_thread(int nr, unsigned long cl savesegment(fs,p->thread.fs); savesegment(gs,p->thread.gs); + perfctr_copy_task(p, regs); + tsk = current; if (unlikely(test_tsk_thread_flag(tsk, TIF_IO_BITMAP))) { p->thread.io_bitmap_ptr = kmalloc(IO_BITMAP_BYTES, GFP_KERNEL); @@ -696,6 +700,8 @@ struct task_struct fastcall * __switch_t disable_tsc(prev_p, next_p); + perfctr_resume_thread(next); + return prev_p; } diff -rupN linux-2.6.18/arch/i386/kernel/syscall_table.S linux-2.6.18.perfctr27/arch/i386/kernel/syscall_table.S --- linux-2.6.18/arch/i386/kernel/syscall_table.S 2007-09-27 20:28:32.000000000 +0200 +++ linux-2.6.18.perfctr27/arch/i386/kernel/syscall_table.S 2007-09-27 20:49:23.000000000 +0200 @@ -317,3 +317,14 @@ ENTRY(sys_call_table) .long sys_tee /* 315 */ .long sys_vmsplice .long sys_move_pages + .long sys_ni_syscall + .long sys_ni_syscall + .long sys_ni_syscall /* 320 */ + .long sys_ni_syscall + .long sys_ni_syscall + .long sys_ni_syscall + .long sys_ni_syscall + .long sys_vperfctr_open /* 325 */ + .long sys_vperfctr_control + .long sys_vperfctr_write + .long sys_vperfctr_read diff -rupN linux-2.6.18/arch/powerpc/Kconfig linux-2.6.18.perfctr27/arch/powerpc/Kconfig --- linux-2.6.18/arch/powerpc/Kconfig 2007-09-27 20:28:33.000000000 +0200 +++ linux-2.6.18.perfctr27/arch/powerpc/Kconfig 2007-09-27 20:34:06.000000000 +0200 @@ -320,6 +320,9 @@ config NOT_COHERENT_CACHE bool depends on 4xx || 8xx || E200 default y + +source "drivers/perfctr/Kconfig" + endmenu source "init/Kconfig" diff -rupN linux-2.6.18/arch/powerpc/kernel/process.c linux-2.6.18.perfctr27/arch/powerpc/kernel/process.c --- linux-2.6.18/arch/powerpc/kernel/process.c 2007-09-27 20:28:33.000000000 +0200 +++ linux-2.6.18.perfctr27/arch/powerpc/kernel/process.c 2007-09-27 20:34:06.000000000 +0200 @@ -31,6 +31,7 @@ #include #include #include +#include #include #include #include @@ -332,7 +333,9 @@ struct task_struct *__switch_to(struct t account_process_vtime(current); calculate_steal_time(); + perfctr_suspend_thread(&prev->thread); last = _switch(old_thread, new_thread); + perfctr_resume_thread(¤t->thread); local_irq_restore(flags); @@ -464,6 +467,7 @@ void show_regs(struct pt_regs * regs) void exit_thread(void) { discard_lazy_cpu_state(); + perfctr_exit_thread(¤t->thread); } void flush_thread(void) @@ -576,6 +580,8 @@ int copy_thread(int nr, unsigned long cl p->thread.last_syscall = -1; #endif + perfctr_copy_task(p, regs); + return 0; } diff -rupN linux-2.6.18/arch/x86_64/Kconfig linux-2.6.18.perfctr27/arch/x86_64/Kconfig --- linux-2.6.18/arch/x86_64/Kconfig 2007-09-27 20:28:35.000000000 +0200 +++ linux-2.6.18.perfctr27/arch/x86_64/Kconfig 2007-09-27 20:34:06.000000000 +0200 @@ -526,6 +526,8 @@ config SECCOMP If unsure, say Y. Only embedded should say N here. +source "drivers/perfctr/Kconfig" + source kernel/Kconfig.hz config REORDER diff -rupN linux-2.6.18/arch/x86_64/ia32/ia32entry.S linux-2.6.18.perfctr27/arch/x86_64/ia32/ia32entry.S --- linux-2.6.18/arch/x86_64/ia32/ia32entry.S 2007-09-27 20:28:35.000000000 +0200 +++ linux-2.6.18.perfctr27/arch/x86_64/ia32/ia32entry.S 2007-09-27 20:50:45.000000000 +0200 @@ -713,4 +713,15 @@ ia32_sys_call_table: .quad sys_tee .quad compat_sys_vmsplice .quad compat_sys_move_pages + .quad quiet_ni_syscall + .quad quiet_ni_syscall + .quad quiet_ni_syscall /* 320 */ + .quad quiet_ni_syscall + .quad quiet_ni_syscall + .quad quiet_ni_syscall + .quad quiet_ni_syscall + .quad sys_vperfctr_open /* 325 */ + .quad sys_vperfctr_control + .quad sys_vperfctr_write + .quad sys_vperfctr_read ia32_syscall_end: diff -rupN linux-2.6.18/arch/x86_64/kernel/entry.S linux-2.6.18.perfctr27/arch/x86_64/kernel/entry.S --- linux-2.6.18/arch/x86_64/kernel/entry.S 2007-09-27 20:28:35.000000000 +0200 +++ linux-2.6.18.perfctr27/arch/x86_64/kernel/entry.S 2007-09-27 20:34:06.000000000 +0200 @@ -693,6 +693,12 @@ ENTRY(spurious_interrupt) END(spurious_interrupt) #endif +#if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_PERFCTR) +ENTRY(perfctr_interrupt) + apicinterrupt LOCAL_PERFCTR_VECTOR,smp_perfctr_interrupt +END(perfctr_interrupt) +#endif + /* * Exception entry points. */ diff -rupN linux-2.6.18/arch/x86_64/kernel/i8259.c linux-2.6.18.perfctr27/arch/x86_64/kernel/i8259.c --- linux-2.6.18/arch/x86_64/kernel/i8259.c 2007-09-27 20:28:35.000000000 +0200 +++ linux-2.6.18.perfctr27/arch/x86_64/kernel/i8259.c 2007-09-27 20:34:06.000000000 +0200 @@ -22,6 +22,7 @@ #include #include #include +#include /* * Common place to define all x86 IRQ vectors @@ -590,6 +591,8 @@ void __init init_IRQ(void) set_intr_gate(ERROR_APIC_VECTOR, error_interrupt); #endif + perfctr_vector_init(); + /* * Set the clock to HZ Hz, we already have a valid * vector now: diff -rupN linux-2.6.18/arch/x86_64/kernel/process.c linux-2.6.18.perfctr27/arch/x86_64/kernel/process.c --- linux-2.6.18/arch/x86_64/kernel/process.c 2007-09-27 20:28:35.000000000 +0200 +++ linux-2.6.18.perfctr27/arch/x86_64/kernel/process.c 2007-09-27 20:34:06.000000000 +0200 @@ -32,6 +32,7 @@ #include #include #include +#include #include #include #include @@ -357,6 +358,7 @@ void exit_thread(void) t->io_bitmap_max = 0; put_cpu(); } + perfctr_exit_thread(&me->thread); } void flush_thread(void) @@ -461,6 +463,8 @@ int copy_thread(int nr, unsigned long cl asm("mov %%es,%0" : "=m" (p->thread.es)); asm("mov %%ds,%0" : "=m" (p->thread.ds)); + perfctr_copy_task(p, regs); + if (unlikely(me->thread.io_bitmap_ptr != NULL)) { p->thread.io_bitmap_ptr = kmalloc(IO_BITMAP_BYTES, GFP_KERNEL); if (!p->thread.io_bitmap_ptr) { @@ -618,6 +622,8 @@ __switch_to(struct task_struct *prev_p, } } + perfctr_resume_thread(next); + return prev_p; } diff -rupN linux-2.6.18/drivers/Makefile linux-2.6.18.perfctr27/drivers/Makefile --- linux-2.6.18/drivers/Makefile 2007-09-27 20:28:35.000000000 +0200 +++ linux-2.6.18.perfctr27/drivers/Makefile 2007-09-27 20:34:06.000000000 +0200 @@ -71,6 +71,7 @@ obj-$(CONFIG_NEW_LEDS) += leds/ obj-$(CONFIG_INFINIBAND) += infiniband/ obj-$(CONFIG_IPATH_CORE) += infiniband/ obj-$(CONFIG_SGI_SN) += sn/ +obj-$(CONFIG_PERFCTR) += perfctr/ obj-y += firmware/ obj-$(CONFIG_CRYPTO) += crypto/ obj-$(CONFIG_SUPERH) += sh/ diff -rupN linux-2.6.18/include/asm-i386/mach-default/irq_vectors.h linux-2.6.18.perfctr27/include/asm-i386/mach-default/irq_vectors.h --- linux-2.6.18/include/asm-i386/mach-default/irq_vectors.h 2007-02-04 19:44:54.000000000 +0100 +++ linux-2.6.18.perfctr27/include/asm-i386/mach-default/irq_vectors.h 2007-09-27 20:34:06.000000000 +0200 @@ -56,14 +56,15 @@ * sources per level' errata. */ #define LOCAL_TIMER_VECTOR 0xef +#define LOCAL_PERFCTR_VECTOR 0xee /* - * First APIC vector available to drivers: (vectors 0x30-0xee) + * First APIC vector available to drivers: (vectors 0x30-0xed) * we start at 0x31 to spread out vectors evenly between priority * levels. (0x80 is the syscall vector) */ #define FIRST_DEVICE_VECTOR 0x31 -#define FIRST_SYSTEM_VECTOR 0xef +#define FIRST_SYSTEM_VECTOR 0xee #define TIMER_IRQ 0 diff -rupN linux-2.6.18/include/asm-i386/mach-visws/irq_vectors.h linux-2.6.18.perfctr27/include/asm-i386/mach-visws/irq_vectors.h --- linux-2.6.18/include/asm-i386/mach-visws/irq_vectors.h 2007-02-04 19:44:54.000000000 +0100 +++ linux-2.6.18.perfctr27/include/asm-i386/mach-visws/irq_vectors.h 2007-09-27 20:34:06.000000000 +0200 @@ -35,14 +35,15 @@ * sources per level' errata. */ #define LOCAL_TIMER_VECTOR 0xef +#define LOCAL_PERFCTR_VECTOR 0xee /* - * First APIC vector available to drivers: (vectors 0x30-0xee) + * First APIC vector available to drivers: (vectors 0x30-0xed) * we start at 0x31 to spread out vectors evenly between priority * levels. (0x80 is the syscall vector) */ #define FIRST_DEVICE_VECTOR 0x31 -#define FIRST_SYSTEM_VECTOR 0xef +#define FIRST_SYSTEM_VECTOR 0xee #define TIMER_IRQ 0 diff -rupN linux-2.6.18/include/asm-i386/processor.h linux-2.6.18.perfctr27/include/asm-i386/processor.h --- linux-2.6.18/include/asm-i386/processor.h 2007-09-27 20:28:47.000000000 +0200 +++ linux-2.6.18.perfctr27/include/asm-i386/processor.h 2007-09-27 20:34:06.000000000 +0200 @@ -471,6 +471,8 @@ struct thread_struct { unsigned long iopl; /* max allowed port in the bitmap, in bytes: */ unsigned long io_bitmap_max; +/* performance counters */ + struct vperfctr *perfctr; }; #define INIT_THREAD { \ diff -rupN linux-2.6.18/include/asm-i386/system.h linux-2.6.18.perfctr27/include/asm-i386/system.h --- linux-2.6.18/include/asm-i386/system.h 2007-09-27 20:28:47.000000000 +0200 +++ linux-2.6.18.perfctr27/include/asm-i386/system.h 2007-09-27 20:36:58.000000000 +0200 @@ -17,6 +17,7 @@ extern struct task_struct * FASTCALL(__s */ #define switch_to(prev,next,last) do { \ unsigned long esi,edi; \ + perfctr_suspend_thread(&(prev)->thread); \ asm volatile("pushfl\n\t" /* Save flags */ \ "pushl %%ebp\n\t" \ "movl %%esp,%0\n\t" /* save ESP */ \ diff -rupN linux-2.6.18/include/asm-i386/unistd.h linux-2.6.18.perfctr27/include/asm-i386/unistd.h --- linux-2.6.18/include/asm-i386/unistd.h 2007-09-27 20:28:47.000000000 +0200 +++ linux-2.6.18.perfctr27/include/asm-i386/unistd.h 2007-09-27 20:47:19.000000000 +0200 @@ -323,10 +323,14 @@ #define __NR_tee 315 #define __NR_vmsplice 316 #define __NR_move_pages 317 +#define __NR_vperfctr_open 325 +#define __NR_vperfctr_control (__NR_vperfctr_open+1) +#define __NR_vperfctr_write (__NR_vperfctr_open+2) +#define __NR_vperfctr_read (__NR_vperfctr_open+3) #ifdef __KERNEL__ -#define NR_syscalls 318 +#define NR_syscalls 329 /* * user-visible error numbers are in the range -1 - -128: see diff -rupN linux-2.6.18/include/asm-powerpc/processor.h linux-2.6.18.perfctr27/include/asm-powerpc/processor.h --- linux-2.6.18/include/asm-powerpc/processor.h 2007-09-27 20:28:48.000000000 +0200 +++ linux-2.6.18.perfctr27/include/asm-powerpc/processor.h 2007-09-27 20:34:06.000000000 +0200 @@ -169,6 +169,9 @@ struct thread_struct { unsigned long spefscr; /* SPE & eFP status */ int used_spe; /* set if process has used spe */ #endif /* CONFIG_SPE */ +#ifdef CONFIG_PERFCTR_VIRTUAL + struct vperfctr *perfctr; /* performance counters */ +#endif }; #define ARCH_MIN_TASKALIGN 16 diff -rupN linux-2.6.18/include/asm-powerpc/reg.h linux-2.6.18.perfctr27/include/asm-powerpc/reg.h --- linux-2.6.18/include/asm-powerpc/reg.h 2007-09-27 20:28:48.000000000 +0200 +++ linux-2.6.18.perfctr27/include/asm-powerpc/reg.h 2007-09-27 20:34:06.000000000 +0200 @@ -365,10 +365,8 @@ #define SPRN_PURR 0x135 /* Processor Utilization of Resources Reg */ #define SPRN_PVR 0x11F /* Processor Version Register */ #define SPRN_RPA 0x3D6 /* Required Physical Address Register */ -#define SPRN_SDA 0x3BF /* Sampled Data Address Register */ #define SPRN_SDR1 0x019 /* MMU Hash Base Register */ #define SPRN_ASR 0x118 /* Address Space Register */ -#define SPRN_SIA 0x3BB /* Sampled Instruction Address Register */ #define SPRN_SPRG0 0x110 /* Special Purpose Register General 0 */ #define SPRN_SPRG1 0x111 /* Special Purpose Register General 1 */ #define SPRN_SPRG2 0x112 /* Special Purpose Register General 2 */ @@ -405,13 +403,6 @@ #define SPRN_THRM3 0x3FE /* Thermal Management Register 3 */ #define THRM3_E (1<<0) #define SPRN_TLBMISS 0x3D4 /* 980 7450 TLB Miss Register */ -#define SPRN_UMMCR0 0x3A8 /* User Monitor Mode Control Register 0 */ -#define SPRN_UMMCR1 0x3AC /* User Monitor Mode Control Register 0 */ -#define SPRN_UPMC1 0x3A9 /* User Performance Counter Register 1 */ -#define SPRN_UPMC2 0x3AA /* User Performance Counter Register 2 */ -#define SPRN_UPMC3 0x3AD /* User Performance Counter Register 3 */ -#define SPRN_UPMC4 0x3AE /* User Performance Counter Register 4 */ -#define SPRN_USIA 0x3AB /* User Sampled Instruction Address Register */ #define SPRN_VRSAVE 0x100 /* Vector Register Save Register */ #define SPRN_XER 0x001 /* Fixed Point Exception Register */ @@ -461,33 +452,7 @@ #define SPRN_SDAR 781 #else /* 32-bit */ -#define SPRN_MMCR0 952 /* Monitor Mode Control Register 0 */ -#define MMCR0_FC 0x80000000UL /* freeze counters */ -#define MMCR0_FCS 0x40000000UL /* freeze in supervisor state */ -#define MMCR0_FCP 0x20000000UL /* freeze in problem state */ -#define MMCR0_FCM1 0x10000000UL /* freeze counters while MSR mark = 1 */ -#define MMCR0_FCM0 0x08000000UL /* freeze counters while MSR mark = 0 */ -#define MMCR0_PMXE 0x04000000UL /* performance monitor exception enable */ -#define MMCR0_FCECE 0x02000000UL /* freeze ctrs on enabled cond or event */ -#define MMCR0_TBEE 0x00400000UL /* time base exception enable */ -#define MMCR0_PMC1CE 0x00008000UL /* PMC1 count enable*/ #define MMCR0_PMCnCE 0x00004000UL /* count enable for all but PMC 1*/ -#define MMCR0_TRIGGER 0x00002000UL /* TRIGGER enable */ -#define MMCR0_PMC1SEL 0x00001fc0UL /* PMC 1 Event */ -#define MMCR0_PMC2SEL 0x0000003fUL /* PMC 2 Event */ - -#define SPRN_MMCR1 956 -#define MMCR1_PMC3SEL 0xf8000000UL /* PMC 3 Event */ -#define MMCR1_PMC4SEL 0x07c00000UL /* PMC 4 Event */ -#define MMCR1_PMC5SEL 0x003e0000UL /* PMC 5 Event */ -#define MMCR1_PMC6SEL 0x0001f800UL /* PMC 6 Event */ -#define SPRN_MMCR2 944 -#define SPRN_PMC1 953 /* Performance Counter Register 1 */ -#define SPRN_PMC2 954 /* Performance Counter Register 2 */ -#define SPRN_PMC3 957 /* Performance Counter Register 3 */ -#define SPRN_PMC4 958 /* Performance Counter Register 4 */ -#define SPRN_PMC5 945 /* Performance Counter Register 5 */ -#define SPRN_PMC6 946 /* Performance Counter Register 6 */ #define SPRN_SIAR 955 /* Sampled Instruction Address Register */ @@ -499,6 +464,77 @@ #define MMCR0_PMC2_CYCLES 0x1 #define MMCR0_PMC2_ITLB 0x7 #define MMCR0_PMC2_LOADMISSTIME 0x5 + +/* Performance-monitoring control and counter registers */ +#define SPRN_MMCR0 0x3B8 /* Monitor Mode Control Register 0 (604 and up) */ +#define SPRN_MMCR1 0x3BC /* Monitor Mode Control Register 1 (604e and up) */ +#define SPRN_MMCR2 0x3B0 /* Monitor Mode Control Register 2 (7400 and up) */ +#define SPRN_PMC1 0x3B9 /* Performance Counter Register 1 (604 and up) */ +#define SPRN_PMC2 0x3BA /* Performance Counter Register 2 (604 and up) */ +#define SPRN_PMC3 0x3BD /* Performance Counter Register 3 (604e and up) */ +#define SPRN_PMC4 0x3BE /* Performance Counter Register 4 (604e and up) */ +#define SPRN_PMC5 0x3B1 /* Performance Counter Register 5 (7450 and up) */ +#define SPRN_PMC6 0x3B2 /* Performance Counter Register 6 (7450 and up) */ +#define SPRN_SIA 0x3BB /* Sampled Instruction Address Register (604 and up) */ +#define SPRN_SDA 0x3BF /* Sampled Data Address Register (604/604e only) */ +#define SPRN_BAMR 0x3B7 /* Breakpoint Address Mask Register (7400 and up) */ + +#define SPRN_UMMCR0 0x3A8 /* User Monitor Mode Control Register 0 (750 and up) */ +#define SPRN_UMMCR1 0x3AC /* User Monitor Mode Control Register 0 (750 and up) */ +#define SPRN_UMMCR2 0x3A0 /* User Monitor Mode Control Register 0 (7400 and up) */ +#define SPRN_UPMC1 0x3A9 /* User Performance Counter Register 1 (750 and up) */ +#define SPRN_UPMC2 0x3AA /* User Performance Counter Register 2 (750 and up) */ +#define SPRN_UPMC3 0x3AD /* User Performance Counter Register 3 (750 and up) */ +#define SPRN_UPMC4 0x3AE /* User Performance Counter Register 4 (750 and up) */ +#define SPRN_UPMC5 0x3A1 /* User Performance Counter Register 5 (7450 and up) */ +#define SPRN_UPMC6 0x3A2 /* User Performance Counter Register 5 (7450 and up) */ +#define SPRN_USIA 0x3AB /* User Sampled Instruction Address Register (750 and up) */ +#define SPRN_UBAMR 0x3A7 /* User Breakpoint Address Mask Register (7400 and up) */ + +/* MMCR0 layout (74xx terminology) */ +#define MMCR0_FC 0x80000000 /* Freeze counters unconditionally. */ +#define MMCR0_FCS 0x40000000 /* Freeze counters while MSR[PR]=0 (supervisor mode). */ +#define MMCR0_FCP 0x20000000 /* Freeze counters while MSR[PR]=1 (user mode). */ +#define MMCR0_FCM1 0x10000000 /* Freeze counters while MSR[PM]=1. */ +#define MMCR0_FCM0 0x08000000 /* Freeze counters while MSR[PM]=0. */ +#define MMCR0_PMXE 0x04000000 /* Enable performance monitor exceptions. + * Cleared by hardware when a PM exception occurs. + * 604: PMXE is not cleared by hardware. + */ +#define MMCR0_FCECE 0x02000000 /* Freeze counters on enabled condition or event. + * FCECE is treated as 0 if TRIGGER is 1. + * 74xx: FC is set when the event occurs. + * 604/750: ineffective when PMXE=0. + */ +#define MMCR0_TBSEL 0x01800000 /* Time base lower (TBL) bit selector. + * 00: bit 31, 01: bit 23, 10: bit 19, 11: bit 15. + */ +#define MMCR0_TBEE 0x00400000 /* Enable event on TBL bit transition from 0 to 1. */ +#define MMCR0_THRESHOLD 0x003F0000 /* Threshold value for certain events. */ +#define MMCR0_PMC1CE 0x00008000 /* Enable event on PMC1 overflow. */ +#define MMCR0_PMCjCE 0x00004000 /* Enable event on PMC2-PMC6 overflow. + * 604/750: Overrides FCECE (DISCOUNT). + */ +#define MMCR0_TRIGGER 0x00002000 /* Disable PMC2-PMC6 until PMC1 overflow or other event. + * 74xx: cleared by hardware when the event occurs. + */ +#define MMCR0_PMC1SEL 0x00001FC0 /* PMC1 event selector, 7 bits. */ +#define MMCR0_PMC2SEL 0x0000003F /* PMC2 event selector, 6 bits. */ + +/* MMCR1 layout (604e-7457) */ +#define MMCR1_PMC3SEL 0xF8000000 /* PMC3 event selector, 5 bits. */ +#define MMCR1_PMC4SEL 0x07C00000 /* PMC4 event selector, 5 bits. */ +#define MMCR1_PMC5SEL 0x003E0000 /* PMC5 event selector, 5 bits. (745x only) */ +#define MMCR1_PMC6SEL 0x0001F800 /* PMC6 event selector, 6 bits. (745x only) */ +#define MMCR1__RESERVED 0x000007FF /* should be zero */ + +/* MMCR2 layout (7400-7457) */ +#define MMCR2_THRESHMULT 0x80000000 /* MMCR0[THRESHOLD] multiplier. */ +#define MMCR2_SMCNTEN 0x40000000 /* 7400/7410 only, should be zero. */ +#define MMCR2_SMINTEN 0x20000000 /* 7400/7410 only, should be zero. */ +#define MMCR2__RESERVED 0x1FFFFFFF /* should be zero */ +#define MMCR2_RESERVED (MMCR2_SMCNTEN | MMCR2_SMINTEN | MMCR2__RESERVED) + #endif /* diff -rupN linux-2.6.18/include/asm-powerpc/systbl.h linux-2.6.18.perfctr27/include/asm-powerpc/systbl.h --- linux-2.6.18/include/asm-powerpc/systbl.h 2007-09-27 20:28:48.000000000 +0200 +++ linux-2.6.18.perfctr27/include/asm-powerpc/systbl.h 2007-09-27 20:45:09.000000000 +0200 @@ -303,4 +303,17 @@ SYSCALL_SPU(readlinkat) SYSCALL_SPU(fchmodat) SYSCALL_SPU(faccessat) COMPAT_SYS_SPU(get_robust_list) -COMPAT_SYS_SPU(set_robust_list) +COMPAT_SYS_SPU(set_robust_list) /* 300 */ +SYSCALL(ni_syscall) /* 301 */ +SYSCALL(ni_syscall) /* 302 */ +SYSCALL(ni_syscall) /* 303 */ +SYSCALL(ni_syscall) /* 304 */ +SYSCALL(ni_syscall) /* 305 */ +SYSCALL(ni_syscall) /* 306 */ +SYSCALL(ni_syscall) /* 307 */ +SYSCALL(ni_syscall) /* 308 */ +SYSCALL(ni_syscall) /* 309 */ +SYSCALL(vperfctr_open) /* 310 */ +SYSCALL(vperfctr_control) +SYSCALL(vperfctr_write) +SYSCALL(vperfctr_read) diff -rupN linux-2.6.18/include/asm-powerpc/unistd.h linux-2.6.18.perfctr27/include/asm-powerpc/unistd.h --- linux-2.6.18/include/asm-powerpc/unistd.h 2007-09-27 20:28:48.000000000 +0200 +++ linux-2.6.18.perfctr27/include/asm-powerpc/unistd.h 2007-09-27 20:42:27.000000000 +0200 @@ -323,10 +323,14 @@ #define __NR_faccessat 298 #define __NR_get_robust_list 299 #define __NR_set_robust_list 300 +#define __NR_vperfctr_open 310 +#define __NR_vperfctr_control (__NR_vperfctr_open+1) +#define __NR_vperfctr_write (__NR_vperfctr_open+2) +#define __NR_vperfctr_read (__NR_vperfctr_open+3) #ifdef __KERNEL__ -#define __NR_syscalls 301 +#define __NR_syscalls 314 #define __NR__exit __NR_exit #define NR_syscalls __NR_syscalls diff -rupN linux-2.6.18/include/asm-x86_64/hw_irq.h linux-2.6.18.perfctr27/include/asm-x86_64/hw_irq.h --- linux-2.6.18/include/asm-x86_64/hw_irq.h 2007-09-27 20:28:48.000000000 +0200 +++ linux-2.6.18.perfctr27/include/asm-x86_64/hw_irq.h 2007-09-27 20:34:06.000000000 +0200 @@ -64,14 +64,15 @@ struct hw_interrupt_type; * sources per level' errata. */ #define LOCAL_TIMER_VECTOR 0xef +#define LOCAL_PERFCTR_VECTOR 0xee /* - * First APIC vector available to drivers: (vectors 0x30-0xee) + * First APIC vector available to drivers: (vectors 0x30-0xed) * we start at 0x31 to spread out vectors evenly between priority * levels. (0x80 is the syscall vector) */ #define FIRST_DEVICE_VECTOR 0x31 -#define FIRST_SYSTEM_VECTOR 0xef /* duplicated in irq.h */ +#define FIRST_SYSTEM_VECTOR 0xee /* duplicated in irq.h */ #ifndef __ASSEMBLY__ diff -rupN linux-2.6.18/include/asm-x86_64/irq.h linux-2.6.18.perfctr27/include/asm-x86_64/irq.h --- linux-2.6.18/include/asm-x86_64/irq.h 2007-09-27 18:15:36.000000000 +0200 +++ linux-2.6.18.perfctr27/include/asm-x86_64/irq.h 2007-09-27 20:34:06.000000000 +0200 @@ -29,7 +29,7 @@ */ #define NR_VECTORS 256 -#define FIRST_SYSTEM_VECTOR 0xef /* duplicated in hw_irq.h */ +#define FIRST_SYSTEM_VECTOR 0xee /* duplicated in hw_irq.h */ #ifdef CONFIG_PCI_MSI #define NR_IRQS FIRST_SYSTEM_VECTOR diff -rupN linux-2.6.18/include/asm-x86_64/processor.h linux-2.6.18.perfctr27/include/asm-x86_64/processor.h --- linux-2.6.18/include/asm-x86_64/processor.h 2007-09-27 20:28:48.000000000 +0200 +++ linux-2.6.18.perfctr27/include/asm-x86_64/processor.h 2007-09-27 20:34:06.000000000 +0200 @@ -274,6 +274,8 @@ struct thread_struct { unsigned io_bitmap_max; /* cached TLS descriptors. */ u64 tls_array[GDT_ENTRY_TLS_ENTRIES]; +/* performance counters */ + struct vperfctr *perfctr; } __attribute__((aligned(16))); #define INIT_THREAD { \ diff -rupN linux-2.6.18/include/asm-x86_64/system.h linux-2.6.18.perfctr27/include/asm-x86_64/system.h --- linux-2.6.18/include/asm-x86_64/system.h 2007-09-27 20:28:48.000000000 +0200 +++ linux-2.6.18.perfctr27/include/asm-x86_64/system.h 2007-09-27 20:34:06.000000000 +0200 @@ -20,7 +20,8 @@ #define __EXTRA_CLOBBER \ ,"rcx","rbx","rdx","r8","r9","r10","r11","r12","r13","r14","r15" -#define switch_to(prev,next,last) \ +#define switch_to(prev,next,last) do { \ + perfctr_suspend_thread(&(prev)->thread); \ asm volatile(SAVE_CONTEXT \ "movq %%rsp,%P[threadrsp](%[prev])\n\t" /* save RSP */ \ "movq %P[threadrsp](%[next]),%%rsp\n\t" /* restore RSP */ \ @@ -40,7 +41,8 @@ [tif_fork] "i" (TIF_FORK), \ [thread_info] "i" (offsetof(struct task_struct, thread_info)), \ [pda_pcurrent] "i" (offsetof(struct x8664_pda, pcurrent)) \ - : "memory", "cc" __EXTRA_CLOBBER) + : "memory", "cc" __EXTRA_CLOBBER); \ +} while (0) extern void load_gs_index(unsigned); diff -rupN linux-2.6.18/include/asm-x86_64/unistd.h linux-2.6.18.perfctr27/include/asm-x86_64/unistd.h --- linux-2.6.18/include/asm-x86_64/unistd.h 2007-09-27 20:28:48.000000000 +0200 +++ linux-2.6.18.perfctr27/include/asm-x86_64/unistd.h 2007-09-27 20:56:09.000000000 +0200 @@ -619,10 +619,24 @@ __SYSCALL(__NR_sync_file_range, sys_sync __SYSCALL(__NR_vmsplice, sys_vmsplice) #define __NR_move_pages 279 __SYSCALL(__NR_move_pages, sys_move_pages) +__SYSCALL(280, sys_ni_syscall) /* utimensat */ +__SYSCALL(281, sys_ni_syscall) /* epoll_wait */ +__SYSCALL(282, sys_ni_syscall) /* signalfd */ +__SYSCALL(283, sys_ni_syscall) /* timerfd */ +__SYSCALL(284, sys_ni_syscall) /* eventfd */ +__SYSCALL(285, sys_ni_syscall) /* fallocate */ +#define __NR_vperfctr_open 286 +__SYSCALL(__NR_vperfctr_open, sys_vperfctr_open) +#define __NR_vperfctr_control (__NR_vperfctr_open+1) +__SYSCALL(__NR_vperfctr_control, sys_vperfctr_control) +#define __NR_vperfctr_write (__NR_vperfctr_open+2) +__SYSCALL(__NR_vperfctr_write, sys_vperfctr_write) +#define __NR_vperfctr_read (__NR_vperfctr_open+3) +__SYSCALL(__NR_vperfctr_read, sys_vperfctr_read) #ifdef __KERNEL__ -#define __NR_syscall_max __NR_move_pages +#define __NR_syscall_max __NR_vperfctr_read #ifndef __NO_STUBS diff -rupN linux-2.6.18/include/linux/sched.h linux-2.6.18.perfctr27/include/linux/sched.h --- linux-2.6.18/include/linux/sched.h 2007-09-27 20:28:49.000000000 +0200 +++ linux-2.6.18.perfctr27/include/linux/sched.h 2007-09-27 20:34:06.000000000 +0200 @@ -1351,6 +1351,9 @@ static inline int thread_group_empty(str * subscriptions and synchronises with wait4(). Also used in procfs. Also * pins the final release of task.io_context. Also protects ->cpuset. * + * Synchronises set_cpus_allowed(), unlink, and creat of ->thread.perfctr. + * [if CONFIG_PERFCTR_VIRTUAL] + * * Nests both inside and outside of read_lock(&tasklist_lock). * It must not be nested with write_lock_irq(&tasklist_lock), * neither inside nor outside. diff -rupN linux-2.6.18/kernel/exit.c linux-2.6.18.perfctr27/kernel/exit.c --- linux-2.6.18/kernel/exit.c 2007-09-27 20:28:49.000000000 +0200 +++ linux-2.6.18.perfctr27/kernel/exit.c 2007-09-27 20:34:06.000000000 +0200 @@ -25,6 +25,7 @@ #include #include #include +#include #include #include #include @@ -166,6 +167,7 @@ repeat: zap_leader = (leader->exit_signal == -1); } + perfctr_release_task(p); sched_exit(p); write_unlock_irq(&tasklist_lock); proc_flush_task(p); diff -rupN linux-2.6.18/kernel/sched.c linux-2.6.18.perfctr27/kernel/sched.c --- linux-2.6.18/kernel/sched.c 2007-09-27 20:28:49.000000000 +0200 +++ linux-2.6.18.perfctr27/kernel/sched.c 2007-09-27 20:34:06.000000000 +0200 @@ -45,6 +45,7 @@ #include #include #include +#include #include #include #include @@ -4845,6 +4846,8 @@ int set_cpus_allowed(struct task_struct struct rq *rq; int ret = 0; + perfctr_set_cpus_allowed(p, new_mask); + rq = task_rq_lock(p, &flags); if (!cpus_intersects(new_mask, cpu_online_map)) { ret = -EINVAL; diff -rupN linux-2.6.18/kernel/sys_ni.c linux-2.6.18.perfctr27/kernel/sys_ni.c --- linux-2.6.18/kernel/sys_ni.c 2007-09-27 20:28:49.000000000 +0200 +++ linux-2.6.18.perfctr27/kernel/sys_ni.c 2007-09-27 20:34:06.000000000 +0200 @@ -72,6 +72,10 @@ cond_syscall(compat_sys_mq_timedsend); cond_syscall(compat_sys_mq_timedreceive); cond_syscall(compat_sys_mq_notify); cond_syscall(compat_sys_mq_getsetattr); +cond_syscall(sys_vperfctr_open); +cond_syscall(sys_vperfctr_control); +cond_syscall(sys_vperfctr_write); +cond_syscall(sys_vperfctr_read); cond_syscall(sys_mbind); cond_syscall(sys_get_mempolicy); cond_syscall(sys_set_mempolicy); diff -rupN linux-2.6.18/kernel/timer.c linux-2.6.18.perfctr27/kernel/timer.c --- linux-2.6.18/kernel/timer.c 2007-09-27 20:28:49.000000000 +0200 +++ linux-2.6.18.perfctr27/kernel/timer.c 2007-09-27 20:34:06.000000000 +0200 @@ -32,6 +32,7 @@ #include #include #include +#include #include #include @@ -1181,6 +1182,7 @@ void update_process_times(int user_tick) account_user_time(p, jiffies_to_cputime(1)); else account_system_time(p, HARDIRQ_OFFSET, jiffies_to_cputime(1)); + perfctr_sample_thread(&p->thread); run_local_timers(); if (rcu_pending(cpu)) rcu_check_callbacks(cpu, user_tick);