diff options
Diffstat (limited to 'arch/powerpc')
24 files changed, 844 insertions, 201 deletions
diff --git a/arch/powerpc/boot/dts/gef_sbc610.dts b/arch/powerpc/boot/dts/gef_sbc610.dts index 9708b3423bb..e78c355c7ba 100644 --- a/arch/powerpc/boot/dts/gef_sbc610.dts +++ b/arch/powerpc/boot/dts/gef_sbc610.dts @@ -88,6 +88,21 @@ compatible = "gef,fpga-regs"; reg = <0x4 0x0 0x40>; }; + + wdt@4,2000 { + compatible = "gef,fpga-wdt"; + reg = <0x4 0x2000 0x8>; + interrupts = <0x1a 0x4>; + interrupt-parent = <&gef_pic>; + }; + /* Second watchdog available, driver currently supports one. + wdt@4,2010 { + compatible = "gef,fpga-wdt"; + reg = <0x4 0x2010 0x8>; + interrupts = <0x1b 0x4>; + interrupt-parent = <&gef_pic>; + }; + */ gef_pic: pic@4,4000 { #interrupt-cells = <1>; interrupt-controller; diff --git a/arch/powerpc/boot/dts/mpc8544ds.dts b/arch/powerpc/boot/dts/mpc8544ds.dts index b9da4210506..0668d104877 100644 --- a/arch/powerpc/boot/dts/mpc8544ds.dts +++ b/arch/powerpc/boot/dts/mpc8544ds.dts @@ -313,7 +313,7 @@ 0x1000000 0x0 0x0 0xe1010000 0x0 0x10000>; clock-frequency = <33333333>; interrupt-parent = <&mpic>; - interrupts = <26 2>; + interrupts = <25 2>; interrupt-map-mask = <0xf800 0x0 0x0 0x7>; interrupt-map = < /* IDSEL 0x0 */ @@ -350,7 +350,7 @@ 0x1000000 0x0 0x0 0xe1020000 0x0 0x10000>; clock-frequency = <33333333>; interrupt-parent = <&mpic>; - interrupts = <25 2>; + interrupts = <26 2>; interrupt-map-mask = <0xf800 0x0 0x0 0x7>; interrupt-map = < /* IDSEL 0x0 */ diff --git a/arch/powerpc/boot/dts/mpc8572ds.dts b/arch/powerpc/boot/dts/mpc8572ds.dts index 21459e161d0..3dcc001b8ed 100644 --- a/arch/powerpc/boot/dts/mpc8572ds.dts +++ b/arch/powerpc/boot/dts/mpc8572ds.dts @@ -724,7 +724,7 @@ 0x1000000 0x0 0x00000000 0 0xffc10000 0x0 0x00010000>; clock-frequency = <33333333>; interrupt-parent = <&mpic>; - interrupts = <26 2>; + interrupts = <25 2>; interrupt-map-mask = <0xf800 0x0 0x0 0x7>; interrupt-map = < /* IDSEL 0x0 */ @@ -761,7 +761,7 @@ 0x1000000 0x0 0x00000000 0 0xffc20000 0x0 0x00010000>; clock-frequency = <33333333>; interrupt-parent = <&mpic>; - interrupts = <27 2>; + interrupts = <26 2>; interrupt-map-mask = <0xf800 0x0 0x0 0x7>; interrupt-map = < /* IDSEL 0x0 */ diff --git a/arch/powerpc/boot/dts/mpc8572ds_camp_core0.dts b/arch/powerpc/boot/dts/mpc8572ds_camp_core0.dts index c114c4ee993..fd462efa9e6 100644 --- a/arch/powerpc/boot/dts/mpc8572ds_camp_core0.dts +++ b/arch/powerpc/boot/dts/mpc8572ds_camp_core0.dts @@ -457,7 +457,7 @@ 0x1000000 0x0 0x0 0xffc10000 0x0 0x10000>; clock-frequency = <33333333>; interrupt-parent = <&mpic>; - interrupts = <26 2>; + interrupts = <25 2>; interrupt-map-mask = <0xf800 0x0 0x0 0x7>; interrupt-map = < /* IDSEL 0x0 */ diff --git a/arch/powerpc/boot/dts/mpc8572ds_camp_core1.dts b/arch/powerpc/boot/dts/mpc8572ds_camp_core1.dts index 04ecda18d20..e35230f2ac9 100644 --- a/arch/powerpc/boot/dts/mpc8572ds_camp_core1.dts +++ b/arch/powerpc/boot/dts/mpc8572ds_camp_core1.dts @@ -208,7 +208,7 @@ 0x1000000 0x0 0x0 0xffc20000 0x0 0x10000>; clock-frequency = <33333333>; interrupt-parent = <&mpic>; - interrupts = <27 2>; + interrupts = <26 2>; interrupt-map-mask = <0xf800 0x0 0x0 0x7>; interrupt-map = < /* IDSEL 0x0 */ diff --git a/arch/powerpc/configs/86xx/gef_sbc610_defconfig b/arch/powerpc/configs/86xx/gef_sbc610_defconfig index cd1ffa44932..391874c7b43 100644 --- a/arch/powerpc/configs/86xx/gef_sbc610_defconfig +++ b/arch/powerpc/configs/86xx/gef_sbc610_defconfig @@ -1164,6 +1164,7 @@ CONFIG_WATCHDOG=y # CONFIG_SOFT_WATCHDOG is not set # CONFIG_ALIM7101_WDT is not set # CONFIG_8xxx_WDT is not set +CONFIG_GEF_WDT=y # # PCI-based Watchdog Cards diff --git a/arch/powerpc/include/asm/cell-pmu.h b/arch/powerpc/include/asm/cell-pmu.h index 8066eede3a0..b4b7338ad79 100644 --- a/arch/powerpc/include/asm/cell-pmu.h +++ b/arch/powerpc/include/asm/cell-pmu.h @@ -37,9 +37,11 @@ #define CBE_PM_STOP_AT_MAX 0x40000000 #define CBE_PM_TRACE_MODE_GET(pm_control) (((pm_control) >> 28) & 0x3) #define CBE_PM_TRACE_MODE_SET(mode) (((mode) & 0x3) << 28) +#define CBE_PM_TRACE_BUF_OVFLW(bit) (((bit) & 0x1) << 17) #define CBE_PM_COUNT_MODE_SET(count) (((count) & 0x3) << 18) #define CBE_PM_FREEZE_ALL_CTRS 0x00100000 #define CBE_PM_ENABLE_EXT_TRACE 0x00008000 +#define CBE_PM_SPU_ADDR_TRACE_SET(msk) (((msk) & 0x3) << 9) /* Macros for the trace_address register. */ #define CBE_PM_TRACE_BUF_FULL 0x00000800 diff --git a/arch/powerpc/include/asm/oprofile_impl.h b/arch/powerpc/include/asm/oprofile_impl.h index 95035c602ba..639dc96077a 100644 --- a/arch/powerpc/include/asm/oprofile_impl.h +++ b/arch/powerpc/include/asm/oprofile_impl.h @@ -32,6 +32,12 @@ struct op_system_config { unsigned long mmcr0; unsigned long mmcr1; unsigned long mmcra; +#ifdef CONFIG_OPROFILE_CELL + /* Register for oprofile user tool to check cell kernel profiling + * suport. + */ + unsigned long cell_support; +#endif #endif unsigned long enable_kernel; unsigned long enable_user; diff --git a/arch/powerpc/include/asm/qe.h b/arch/powerpc/include/asm/qe.h index a0a15311d0d..2701753d993 100644 --- a/arch/powerpc/include/asm/qe.h +++ b/arch/powerpc/include/asm/qe.h @@ -624,7 +624,7 @@ struct ucc_slow_pram { #define UCC_GETH_UCCE_RXF1 0x00000002 #define UCC_GETH_UCCE_RXF0 0x00000001 -/* UPSMR, when used as a UART */ +/* UCC Protocol Specific Mode Register (UPSMR), when used for UART */ #define UCC_UART_UPSMR_FLC 0x8000 #define UCC_UART_UPSMR_SL 0x4000 #define UCC_UART_UPSMR_CL_MASK 0x3000 @@ -652,6 +652,23 @@ struct ucc_slow_pram { #define UCC_UART_UPSMR_TPM_EVEN 0x0002 #define UCC_UART_UPSMR_TPM_HIGH 0x0003 +/* UCC Protocol Specific Mode Register (UPSMR), when used for Ethernet */ +#define UCC_GETH_UPSMR_FTFE 0x80000000 +#define UCC_GETH_UPSMR_PTPE 0x40000000 +#define UCC_GETH_UPSMR_ECM 0x04000000 +#define UCC_GETH_UPSMR_HSE 0x02000000 +#define UCC_GETH_UPSMR_PRO 0x00400000 +#define UCC_GETH_UPSMR_CAP 0x00200000 +#define UCC_GETH_UPSMR_RSH 0x00100000 +#define UCC_GETH_UPSMR_RPM 0x00080000 +#define UCC_GETH_UPSMR_R10M 0x00040000 +#define UCC_GETH_UPSMR_RLPB 0x00020000 +#define UCC_GETH_UPSMR_TBIM 0x00010000 +#define UCC_GETH_UPSMR_RES1 0x00002000 +#define UCC_GETH_UPSMR_RMM 0x00001000 +#define UCC_GETH_UPSMR_CAM 0x00000400 +#define UCC_GETH_UPSMR_BRO 0x00000200 + /* UCC Transmit On Demand Register (UTODR) */ #define UCC_SLOW_TOD 0x8000 #define UCC_FAST_TOD 0x8000 diff --git a/arch/powerpc/kernel/asm-offsets.c b/arch/powerpc/kernel/asm-offsets.c index 9937fe44555..19ee491e9e2 100644 --- a/arch/powerpc/kernel/asm-offsets.c +++ b/arch/powerpc/kernel/asm-offsets.c @@ -56,6 +56,10 @@ #include "head_booke.h" #endif +#if defined(CONFIG_FSL_BOOKE) +#include "../mm/mmu_decl.h" +#endif + int main(void) { DEFINE(THREAD, offsetof(struct task_struct, thread)); @@ -382,6 +386,9 @@ int main(void) DEFINE(PGD_T_LOG2, PGD_T_LOG2); DEFINE(PTE_T_LOG2, PTE_T_LOG2); #endif +#ifdef CONFIG_FSL_BOOKE + DEFINE(TLBCAM_SIZE, sizeof(struct tlbcam)); +#endif #ifdef CONFIG_KVM_EXIT_TIMING DEFINE(VCPU_TIMING_EXIT_TBU, offsetof(struct kvm_vcpu, diff --git a/arch/powerpc/kernel/head_fsl_booke.S b/arch/powerpc/kernel/head_fsl_booke.S index 11b549acc03..2f32720a44a 100644 --- a/arch/powerpc/kernel/head_fsl_booke.S +++ b/arch/powerpc/kernel/head_fsl_booke.S @@ -389,10 +389,6 @@ skpinv: addi r6,r6,1 /* Increment */ #endif #endif - mfspr r3,SPRN_TLB1CFG - andi. r3,r3,0xfff - lis r4,num_tlbcam_entries@ha - stw r3,num_tlbcam_entries@l(r4) /* * Decide what sort of machine this is and initialize the MMU. */ @@ -909,7 +905,7 @@ KernelSPE: _GLOBAL(loadcam_entry) lis r4,TLBCAM@ha addi r4,r4,TLBCAM@l - mulli r5,r3,20 + mulli r5,r3,TLBCAM_SIZE add r3,r5,r4 lwz r4,0(r3) mtspr SPRN_MAS0,r4 diff --git a/arch/powerpc/mm/fsl_booke_mmu.c b/arch/powerpc/mm/fsl_booke_mmu.c index 23cee39534f..1971e4ee3d6 100644 --- a/arch/powerpc/mm/fsl_booke_mmu.c +++ b/arch/powerpc/mm/fsl_booke_mmu.c @@ -56,18 +56,11 @@ extern void loadcam_entry(unsigned int index); unsigned int tlbcam_index; -unsigned int num_tlbcam_entries; static unsigned long __cam0, __cam1, __cam2; #define NUM_TLBCAMS (16) -struct tlbcam { - u32 MAS0; - u32 MAS1; - u32 MAS2; - u32 MAS3; - u32 MAS7; -} TLBCAM[NUM_TLBCAMS]; +struct tlbcam TLBCAM[NUM_TLBCAMS]; struct tlbcamrange { unsigned long start; diff --git a/arch/powerpc/mm/mmu_decl.h b/arch/powerpc/mm/mmu_decl.h index ad123bced40..d1f9c62dc17 100644 --- a/arch/powerpc/mm/mmu_decl.h +++ b/arch/powerpc/mm/mmu_decl.h @@ -75,6 +75,15 @@ extern void _tlbia(void); #endif /* CONFIG_PPC_MMU_NOHASH */ #ifdef CONFIG_PPC32 + +struct tlbcam { + u32 MAS0; + u32 MAS1; + u32 MAS2; + u32 MAS3; + u32 MAS7; +}; + extern void mapin_ram(void); extern int map_page(unsigned long va, phys_addr_t pa, int flags); extern void setbat(int index, unsigned long virt, phys_addr_t phys, @@ -90,8 +99,6 @@ extern unsigned int rtas_data, rtas_size; struct hash_pte; extern struct hash_pte *Hash, *Hash_end; extern unsigned long Hash_size, Hash_mask; - -extern unsigned int num_tlbcam_entries; #endif extern unsigned long ioremap_bot; diff --git a/arch/powerpc/oprofile/cell/pr_util.h b/arch/powerpc/oprofile/cell/pr_util.h index dfdbffa0681..964b93974d8 100644 --- a/arch/powerpc/oprofile/cell/pr_util.h +++ b/arch/powerpc/oprofile/cell/pr_util.h @@ -30,6 +30,10 @@ extern struct delayed_work spu_work; extern int spu_prof_running; +#define TRACE_ARRAY_SIZE 1024 + +extern spinlock_t oprof_spu_smpl_arry_lck; + struct spu_overlay_info { /* map of sections within an SPU overlay */ unsigned int vma; /* SPU virtual memory address from elf */ unsigned int size; /* size of section from elf */ @@ -89,10 +93,11 @@ void vma_map_free(struct vma_to_fileoffset_map *map); * Entry point for SPU profiling. * cycles_reset is the SPU_CYCLES count value specified by the user. */ -int start_spu_profiling(unsigned int cycles_reset); - -void stop_spu_profiling(void); +int start_spu_profiling_cycles(unsigned int cycles_reset); +void start_spu_profiling_events(void); +void stop_spu_profiling_cycles(void); +void stop_spu_profiling_events(void); /* add the necessary profiling hooks */ int spu_sync_start(void); diff --git a/arch/powerpc/oprofile/cell/spu_profiler.c b/arch/powerpc/oprofile/cell/spu_profiler.c index 83faa958b9d..9305ddaac51 100644 --- a/arch/powerpc/oprofile/cell/spu_profiler.c +++ b/arch/powerpc/oprofile/cell/spu_profiler.c @@ -18,11 +18,21 @@ #include <asm/cell-pmu.h> #include "pr_util.h" -#define TRACE_ARRAY_SIZE 1024 #define SCALE_SHIFT 14 static u32 *samples; +/* spu_prof_running is a flag used to indicate if spu profiling is enabled + * or not. It is set by the routines start_spu_profiling_cycles() and + * start_spu_profiling_events(). The flag is cleared by the routines + * stop_spu_profiling_cycles() and stop_spu_profiling_events(). These + * routines are called via global_start() and global_stop() which are called in + * op_powerpc_start() and op_powerpc_stop(). These routines are called once + * per system as a result of the user starting/stopping oprofile. Hence, only + * one CPU per user at a time will be changing the value of spu_prof_running. + * In general, OProfile does not protect against multiple users trying to run + * OProfile at a time. + */ int spu_prof_running; static unsigned int profiling_interval; @@ -31,8 +41,8 @@ static unsigned int profiling_interval; #define SPU_PC_MASK 0xFFFF -static DEFINE_SPINLOCK(sample_array_lock); -unsigned long sample_array_lock_flags; +DEFINE_SPINLOCK(oprof_spu_smpl_arry_lck); +unsigned long oprof_spu_smpl_arry_lck_flags; void set_spu_profiling_frequency(unsigned int freq_khz, unsigned int cycles_reset) { @@ -145,13 +155,13 @@ static enum hrtimer_restart profile_spus(struct hrtimer *timer) * sample array must be loaded and then processed for a given * cpu. The sample array is not per cpu. */ - spin_lock_irqsave(&sample_array_lock, - sample_array_lock_flags); + spin_lock_irqsave(&oprof_spu_smpl_arry_lck, + oprof_spu_smpl_arry_lck_flags); num_samples = cell_spu_pc_collection(cpu); if (num_samples == 0) { - spin_unlock_irqrestore(&sample_array_lock, - sample_array_lock_flags); + spin_unlock_irqrestore(&oprof_spu_smpl_arry_lck, + oprof_spu_smpl_arry_lck_flags); continue; } @@ -162,8 +172,8 @@ static enum hrtimer_restart profile_spus(struct hrtimer *timer) num_samples); } - spin_unlock_irqrestore(&sample_array_lock, - sample_array_lock_flags); + spin_unlock_irqrestore(&oprof_spu_smpl_arry_lck, + oprof_spu_smpl_arry_lck_flags); } smp_wmb(); /* insure spu event buffer updates are written */ @@ -182,13 +192,13 @@ static enum hrtimer_restart profile_spus(struct hrtimer *timer) static struct hrtimer timer; /* - * Entry point for SPU profiling. + * Entry point for SPU cycle profiling. * NOTE: SPU profiling is done system-wide, not per-CPU. * * cycles_reset is the count value specified by the user when * setting up OProfile to count SPU_CYCLES. */ -int start_spu_profiling(unsigned int cycles_reset) +int start_spu_profiling_cycles(unsigned int cycles_reset) { ktime_t kt; @@ -212,10 +222,30 @@ int start_spu_profiling(unsigned int cycles_reset) return 0; } -void stop_spu_profiling(void) +/* + * Entry point for SPU event profiling. + * NOTE: SPU profiling is done system-wide, not per-CPU. + * + * cycles_reset is the count value specified by the user when + * setting up OProfile to count SPU_CYCLES. + */ +void start_spu_profiling_events(void) +{ + spu_prof_running = 1; + schedule_delayed_work(&spu_work, DEFAULT_TIMER_EXPIRE); + + return; +} + +void stop_spu_profiling_cycles(void) { spu_prof_running = 0; hrtimer_cancel(&timer); kfree(samples); - pr_debug("SPU_PROF: stop_spu_profiling issued\n"); + pr_debug("SPU_PROF: stop_spu_profiling_cycles issued\n"); +} + +void stop_spu_profiling_events(void) +{ + spu_prof_running = 0; } diff --git a/arch/powerpc/oprofile/common.c b/arch/powerpc/oprofile/common.c index 17807acb05d..21f16edf6c8 100644 --- a/arch/powerpc/oprofile/common.c +++ b/arch/powerpc/oprofile/common.c @@ -132,6 +132,28 @@ static int op_powerpc_create_files(struct super_block *sb, struct dentry *root) oprofilefs_create_ulong(sb, root, "mmcr0", &sys.mmcr0); oprofilefs_create_ulong(sb, root, "mmcr1", &sys.mmcr1); oprofilefs_create_ulong(sb, root, "mmcra", &sys.mmcra); +#ifdef CONFIG_OPROFILE_CELL + /* create a file the user tool can check to see what level of profiling + * support exits with this kernel. Initialize bit mask to indicate + * what support the kernel has: + * bit 0 - Supports SPU event profiling in addition to PPU + * event and cycles; and SPU cycle profiling + * bits 1-31 - Currently unused. + * + * If the file does not exist, then the kernel only supports SPU + * cycle profiling, PPU event and cycle profiling. + */ + oprofilefs_create_ulong(sb, root, "cell_support", &sys.cell_support); + sys.cell_support = 0x1; /* Note, the user OProfile tool must check + * that this bit is set before attempting to + * user SPU event profiling. Older kernels + * will not have this file, hence the user + * tool is not allowed to do SPU event + * profiling on older kernels. Older kernels + * will accept SPU events but collected data + * is garbage. + */ +#endif #endif for (i = 0; i < model->num_counters; ++i) { diff --git a/arch/powerpc/oprofile/op_model_cell.c b/arch/powerpc/oprofile/op_model_cell.c index 25a4ec2514a..ae06c6236d9 100644 --- a/arch/powerpc/oprofile/op_model_cell.c +++ b/arch/powerpc/oprofile/op_model_cell.c @@ -40,14 +40,15 @@ #include "../platforms/cell/interrupt.h" #include "cell/pr_util.h" -static void cell_global_stop_spu(void); +#define PPU_PROFILING 0 +#define SPU_PROFILING_CYCLES 1 +#define SPU_PROFILING_EVENTS 2 -/* - * spu_cycle_reset is the number of cycles between samples. - * This variable is used for SPU profiling and should ONLY be set - * at the beginning of cell_reg_setup; otherwise, it's read-only. - */ -static unsigned int spu_cycle_reset; +#define SPU_EVENT_NUM_START 4100 +#define SPU_EVENT_NUM_STOP 4399 +#define SPU_PROFILE_EVENT_ADDR 4363 /* spu, address trace, decimal */ +#define SPU_PROFILE_EVENT_ADDR_MASK_A 0x146 /* sub unit set to zero */ +#define SPU_PROFILE_EVENT_ADDR_MASK_B 0x186 /* sub unit set to zero */ #define NUM_SPUS_PER_NODE 8 #define SPU_CYCLES_EVENT_NUM 2 /* event number for SPU_CYCLES */ @@ -66,6 +67,21 @@ static unsigned int spu_cycle_reset; #define MAX_SPU_COUNT 0xFFFFFF /* maximum 24 bit LFSR value */ +/* Minumum HW interval timer setting to send value to trace buffer is 10 cycle. + * To configure counter to send value every N cycles set counter to + * 2^32 - 1 - N. + */ +#define NUM_INTERVAL_CYC 0xFFFFFFFF - 10 + +/* + * spu_cycle_reset is the number of cycles between samples. + * This variable is used for SPU profiling and should ONLY be set + * at the beginning of cell_reg_setup; otherwise, it's read-only. + */ +static unsigned int spu_cycle_reset; +static unsigned int profiling_mode; +static int spu_evnt_phys_spu_indx; + struct pmc_cntrl_data { unsigned long vcntr; unsigned long evnts; @@ -105,6 +121,8 @@ struct pm_cntrl { u16 trace_mode; u16 freeze; u16 count_mode; + u16 spu_addr_trace; + u8 trace_buf_ovflw; }; static struct { @@ -122,7 +140,7 @@ static struct { #define GET_INPUT_CONTROL(x) ((x & 0x00000004) >> 2) static DEFINE_PER_CPU(unsigned long[NR_PHYS_CTRS], pmc_values); - +static unsigned long spu_pm_cnt[MAX_NUMNODES * NUM_SPUS_PER_NODE]; static struct pmc_cntrl_data pmc_cntrl[NUM_THREADS][NR_PHYS_CTRS]; /* @@ -152,6 +170,7 @@ static u32 hdw_thread; static u32 virt_cntr_inter_mask; static struct timer_list timer_virt_cntr; +static struct timer_list timer_spu_event_swap; /* * pm_signal needs to be global since it is initialized in @@ -165,7 +184,7 @@ static int spu_rtas_token; /* token for SPU cycle profiling */ static u32 reset_value[NR_PHYS_CTRS]; static int num_counters; static int oprofile_running; -static DEFINE_SPINLOCK(virt_cntr_lock); +static DEFINE_SPINLOCK(cntr_lock); static u32 ctr_enabled; @@ -336,13 +355,13 @@ static void set_pm_event(u32 ctr, int event, u32 unit_mask) for (i = 0; i < NUM_DEBUG_BUS_WORDS; i++) { if (bus_word & (1 << i)) { pm_regs.debug_bus_control |= - (bus_type << (30 - (2 * i))); + (bus_type << (30 - (2 * i))); for (j = 0; j < NUM_INPUT_BUS_WORDS; j++) { if (input_bus[j] == 0xff) { input_bus[j] = i; pm_regs.group_control |= - (i << (30 - (2 * j))); + (i << (30 - (2 * j))); break; } @@ -367,12 +386,16 @@ static void write_pm_cntrl(int cpu) if (pm_regs.pm_cntrl.stop_at_max == 1) val |= CBE_PM_STOP_AT_MAX; - if (pm_regs.pm_cntrl.trace_mode == 1) + if (pm_regs.pm_cntrl.trace_mode != 0) val |= CBE_PM_TRACE_MODE_SET(pm_regs.pm_cntrl.trace_mode); + if (pm_regs.pm_cntrl.trace_buf_ovflw == 1) + val |= CBE_PM_TRACE_BUF_OVFLW(pm_regs.pm_cntrl.trace_buf_ovflw); if (pm_regs.pm_cntrl.freeze == 1) val |= CBE_PM_FREEZE_ALL_CTRS; + val |= CBE_PM_SPU_ADDR_TRACE_SET(pm_regs.pm_cntrl.spu_addr_trace); + /* * Routine set_count_mode must be called previously to set * the count mode based on the user selection of user and kernel. @@ -441,7 +464,7 @@ static void cell_virtual_cntr(unsigned long data) * not both playing with the counters on the same node. */ - spin_lock_irqsave(&virt_cntr_lock, flags); + spin_lock_irqsave(&cntr_lock, flags); prev_hdw_thread = hdw_thread; @@ -480,7 +503,7 @@ static void cell_virtual_cntr(unsigned long data) cbe_disable_pm_interrupts(cpu); for (i = 0; i < num_counters; i++) { per_cpu(pmc_values, cpu + prev_hdw_thread)[i] - = cbe_read_ctr(cpu, i); + = cbe_read_ctr(cpu, i); if (per_cpu(pmc_values, cpu + next_hdw_thread)[i] == 0xFFFFFFFF) @@ -527,7 +550,7 @@ static void cell_virtual_cntr(unsigned long data) cbe_enable_pm(cpu); } - spin_unlock_irqrestore(&virt_cntr_lock, flags); + spin_unlock_irqrestore(&cntr_lock, flags); mod_timer(&timer_virt_cntr, jiffies + HZ / 10); } @@ -541,38 +564,146 @@ static void start_virt_cntrs(void) add_timer(&timer_virt_cntr); } -/* This function is called once for all cpus combined */ -static int cell_reg_setup(struct op_counter_config *ctr, +static int cell_reg_setup_spu_cycles(struct op_counter_config *ctr, struct op_system_config *sys, int num_ctrs) { - int i, j, cpu; - spu_cycle_reset = 0; + spu_cycle_reset = ctr[0].count; - if (ctr[0].event == SPU_CYCLES_EVENT_NUM) { - spu_cycle_reset = ctr[0].count; + /* + * Each node will need to make the rtas call to start + * and stop SPU profiling. Get the token once and store it. + */ + spu_rtas_token = rtas_token("ibm,cbe-spu-perftools"); + + if (unlikely(spu_rtas_token == RTAS_UNKNOWN_SERVICE)) { + printk(KERN_ERR + "%s: rtas token ibm,cbe-spu-perftools unknown\n", + __func__); + return -EIO; + } + return 0; +} + +/* Unfortunately, the hardware will only support event profiling + * on one SPU per node at a time. Therefore, we must time slice + * the profiling across all SPUs in the node. Note, we do this + * in parallel for each node. The following routine is called + * periodically based on kernel timer to switch which SPU is + * being monitored in a round robbin fashion. + */ +static void spu_evnt_swap(unsigned long data) +{ + int node; + int cur_phys_spu, nxt_phys_spu, cur_spu_evnt_phys_spu_indx; + unsigned long flags; + int cpu; + int ret; + u32 interrupt_mask; + + + /* enable interrupts on cntr 0 */ + interrupt_mask = CBE_PM_CTR_OVERFLOW_INTR(0); + + hdw_thread = 0; + + /* Make sure spu event interrupt handler and spu event swap + * don't access the counters simultaneously. + */ + spin_lock_irqsave(&cntr_lock, flags); + + cur_spu_evnt_phys_spu_indx = spu_evnt_phys_spu_indx; + + if (++(spu_evnt_phys_spu_indx) == NUM_SPUS_PER_NODE) + spu_evnt_phys_spu_indx = 0; + + pm_signal[0].sub_unit = spu_evnt_phys_spu_indx; + pm_signal[1].sub_unit = spu_evnt_phys_spu_indx; + pm_signal[2].sub_unit = spu_evnt_phys_spu_indx; + + /* switch the SPU being profiled on each node */ + for_each_online_cpu(cpu) { + if (cbe_get_hw_thread_id(cpu)) + continue; + + node = cbe_cpu_to_node(cpu); + cur_phys_spu = (node * NUM_SPUS_PER_NODE) + + cur_spu_evnt_phys_spu_indx; + nxt_phys_spu = (node * NUM_SPUS_PER_NODE) + + spu_evnt_phys_spu_indx; /* - * Each node will need to make the rtas call to start - * and stop SPU profiling. Get the token once and store it. + * stop counters, save counter values, restore counts + * for previous physical SPU */ - spu_rtas_token = rtas_token("ibm,cbe-spu-perftools"); + cbe_disable_pm(cpu); + cbe_disable_pm_interrupts(cpu); - if (unlikely(spu_rtas_token == RTAS_UNKNOWN_SERVICE)) { - printk(KERN_ERR - "%s: rtas token ibm,cbe-spu-perftools unknown\n", - __func__); - return -EIO; - } + spu_pm_cnt[cur_phys_spu] + = cbe_read_ctr(cpu, 0); + + /* restore previous count for the next spu to sample */ + /* NOTE, hardware issue, counter will not start if the + * counter value is at max (0xFFFFFFFF). + */ + if (spu_pm_cnt[nxt_phys_spu] >= 0xFFFFFFFF) + cbe_write_ctr(cpu, 0, 0xFFFFFFF0); + else + cbe_write_ctr(cpu, 0, spu_pm_cnt[nxt_phys_spu]); + + pm_rtas_reset_signals(cbe_cpu_to_node(cpu)); + + /* setup the debug bus measure the one event and + * the two events to route the next SPU's PC on + * the debug bus + */ + ret = pm_rtas_activate_signals(cbe_cpu_to_node(cpu), 3); + if (ret) + printk(KERN_ERR "%s: pm_rtas_activate_signals failed, " + "SPU event swap\n", __func__); + + /* clear the trace buffer, don't want to take PC for + * previous SPU*/ + cbe_write_pm(cpu, trace_address, 0); + + enable_ctr(cpu, 0, pm_regs.pm07_cntrl); + + /* Enable interrupts on the CPU thread that is starting */ + cbe_enable_pm_interrupts(cpu, hdw_thread, + interrupt_mask); + cbe_enable_pm(cpu); } - pm_rtas_token = rtas_token("ibm,cbe-perftools"); + spin_unlock_irqrestore(&cntr_lock, flags); + /* swap approximately every 0.1 seconds */ + mod_timer(&timer_spu_event_swap, jiffies + HZ / 25); +} + +static void start_spu_event_swap(void) +{ + init_timer(&timer_spu_event_swap); + timer_spu_event_swap.function = spu_evnt_swap; + timer_spu_event_swap.data = 0UL; + timer_spu_event_swap.expires = jiffies + HZ / 25; + add_timer(&timer_spu_event_swap); +} + +static int cell_reg_setup_spu_events(struct op_counter_config *ctr, + struct op_system_config *sys, int num_ctrs) +{ + int i; + + /* routine is called once for all nodes */ + + spu_evnt_phys_spu_indx = 0; /* - * For all events excetp PPU CYCLEs, each node will need to make + * For all events except PPU CYCLEs, each node will need to make * the rtas cbe-perftools call to setup and reset the debug bus. * Make the token lookup call once and store it in the global * variable pm_rtas_token. */ + pm_rtas_token = rtas_token("ibm,cbe-perftools"); + if (unlikely(pm_rtas_token == RTAS_UNKNOWN_SERVICE)) { printk(KERN_ERR "%s: rtas token ibm,cbe-perftools unknown\n", @@ -580,6 +711,58 @@ static int cell_reg_setup(struct op_counter_config *ctr, return -EIO; } + /* setup the pm_control register settings, + * settings will be written per node by the + * cell_cpu_setup() function. + */ + pm_regs.pm_cntrl.trace_buf_ovflw = 1; + + /* Use the occurrence trace mode to have SPU PC saved + * to the trace buffer. Occurrence data in trace buffer + * is not used. Bit 2 must be set to store SPU addresses. + */ + pm_regs.pm_cntrl.trace_mode = 2; + + pm_regs.pm_cntrl.spu_addr_trace = 0x1; /* using debug bus + event 2 & 3 */ + + /* setup the debug bus event array with the SPU PC routing events. + * Note, pm_signal[0] will be filled in by set_pm_event() call below. + */ + pm_signal[1].signal_group = SPU_PROFILE_EVENT_ADDR / 100; + pm_signal[1].bus_word = GET_BUS_WORD(SPU_PROFILE_EVENT_ADDR_MASK_A); + pm_signal[1].bit = SPU_PROFILE_EVENT_ADDR % 100; + pm_signal[1].sub_unit = spu_evnt_phys_spu_indx; + + pm_signal[2].signal_group = SPU_PROFILE_EVENT_ADDR / 100; + pm_signal[2].bus_word = GET_BUS_WORD(SPU_PROFILE_EVENT_ADDR_MASK_B); + pm_signal[2].bit = SPU_PROFILE_EVENT_ADDR % 100; + pm_signal[2].sub_unit = spu_evnt_phys_spu_indx; + + /* Set the user selected spu event to profile on, + * note, only one SPU profiling event is supported + */ + num_counters = 1; /* Only support one SPU event at a time */ + set_pm_event(0, ctr[0].event, ctr[0].unit_mask); + + reset_value[0] = 0xFFFFFFFF - ctr[0].count; + + /* global, used by cell_cpu_setup */ + ctr_enabled |= 1; + + /* Initialize the count for each SPU to the reset value */ + for (i=0; i < MAX_NUMNODES * NUM_SPUS_PER_NODE; i++) + spu_pm_cnt[i] = reset_value[0]; + + return 0; +} + +static int cell_reg_setup_ppu(struct op_counter_config *ctr, + struct op_system_config *sys, int num_ctrs) +{ + /* routine is called once for all nodes */ + int i, j, cpu; + num_counters = num_ctrs; if (unlikely(num_ctrs > NR_PHYS_CTRS)) { @@ -589,14 +772,6 @@ static int cell_reg_setup(struct op_counter_config *ctr, __func__); return -EIO; } - pm_regs.group_control = 0; - pm_regs.debug_bus_control = 0; - - /* setup the pm_control register */ - memset(&pm_regs.pm_cntrl, 0, sizeof(struct pm_cntrl)); - pm_regs.pm_cntrl.stop_at_max = 1; - pm_regs.pm_cntrl.trace_mode = 0; - pm_regs.pm_cntrl.freeze = 1; set_count_mode(sys->enable_kernel, sys->enable_user); @@ -665,6 +840,63 @@ static int cell_reg_setup(struct op_counter_config *ctr, } +/* This function is called once for all cpus combined */ +static int cell_reg_setup(struct op_counter_config *ctr, + struct op_system_config *sys, int num_ctrs) +{ + int ret=0; + spu_cycle_reset = 0; + + /* initialize the spu_arr_trace value, will be reset if + * doing spu event profiling. + */ + pm_regs.group_control = 0; + pm_regs.debug_bus_control = 0; + pm_regs.pm_cntrl.stop_at_max = 1; + pm_regs.pm_cntrl.trace_mode = 0; + pm_regs.pm_cntrl.freeze = 1; + pm_regs.pm_cntrl.trace_buf_ovflw = 0; + pm_regs.pm_cntrl.spu_addr_trace = 0; + + /* + * For all events except PPU CYCLEs, each node will need to make + * the rtas cbe-perftools call to setup and reset the debug bus. + * Make the token lookup call once and store it in the global + * variable pm_rtas_token. + */ + pm_rtas_token = rtas_token("ibm,cbe-perftools"); + + if (unlikely(pm_rtas_token == RTAS_UNKNOWN_SERVICE)) { + printk(KERN_ERR + "%s: rtas token ibm,cbe-perftools unknown\n", + __func__); + return -EIO; + } + + if (ctr[0].event == SPU_CYCLES_EVENT_NUM) { + profiling_mode = SPU_PROFILING_CYCLES; + ret = cell_reg_setup_spu_cycles(ctr, sys, num_ctrs); + } else if ((ctr[0].event >= SPU_EVENT_NUM_START) && + (ctr[0].event <= SPU_EVENT_NUM_STOP)) { + profiling_mode = SPU_PROFILING_EVENTS; + spu_cycle_reset = ctr[0].count; + + /* for SPU event profiling, need to setup the + * pm_signal array with the events to route the + * SPU PC before making the FW call. Note, only + * one SPU event for profiling can be specified + * at a time. + */ + cell_reg_setup_spu_events(ctr, sys, num_ctrs); + } else { + profiling_mode = PPU_PROFILING; + ret = cell_reg_setup_ppu(ctr, sys, num_ctrs); + } + + return ret; +} + + /* This function is called once for each cpu */ static int cell_cpu_setup(struct op_counter_config *cntr) @@ -672,8 +904,13 @@ static int cell_cpu_setup(struct op_counter_config *cntr) u32 cpu = smp_processor_id(); u32 num_enabled = 0; int i; + int ret; - if (spu_cycle_reset) + /* Cycle based SPU profiling does not use the performance + * counters. The trace array is configured to collect + * the data. + */ + if (profiling_mode == SPU_PROFILING_CYCLES) return 0; /* There is one performance monitor per processor chip (i.e. node), @@ -686,7 +923,6 @@ static int cell_cpu_setup(struct op_counter_config *cntr) cbe_disable_pm(cpu); cbe_disable_pm_interrupts(cpu); - cbe_write_pm(cpu, pm_interval, 0); cbe_write_pm(cpu, pm_start_stop, 0); cbe_write_pm(cpu, group_control, pm_regs.group_control); cbe_write_pm(cpu, debug_bus_control, pm_regs.debug_bus_control); @@ -703,7 +939,20 @@ static int cell_cpu_setup(struct op_counter_config *cntr) * The pm_rtas_activate_signals will return -EIO if the FW * call failed. */ - return pm_rtas_activate_signals(cbe_cpu_to_node(cpu), num_enabled); + if (profiling_mode == SPU_PROFILING_EVENTS) { + /* For SPU event profiling also need to setup the + * pm interval timer + */ + ret = pm_rtas_activate_signals(cbe_cpu_to_node(cpu), + num_enabled+2); + /* store PC from debug bus to Trace buffer as often + * as possible (every 10 cycles) + */ + cbe_write_pm(cpu, pm_interval, NUM_INTERVAL_CYC); + return ret; + } else + return pm_rtas_activate_signals(cbe_cpu_to_node(cpu), + num_enabled); } #define ENTRIES 303 @@ -885,7 +1134,122 @@ static struct notifier_block cpu_freq_notifier_block = { }; #endif -static int cell_global_start_spu(struct op_counter_config *ctr) +/* + * Note the generic OProfile stop calls do not support returning + * an error on stop. Hence, will not return an error if the FW + * calls fail on stop. Failure to reset the debug bus is not an issue. + * Failure to disable the SPU profiling is not an issue. The FW calls + * to enable the performance counters and debug bus will work even if + * the hardware was not cleanly reset. + */ +static void cell_global_stop_spu_cycles(void) +{ + int subfunc, rtn_value; + unsigned int lfsr_value; + int cpu; + + oprofile_running = 0; + smp_wmb(); + +#ifdef CONFIG_CPU_FREQ + cpufreq_unregister_notifier(&cpu_freq_notifier_block, + CPUFREQ_TRANSITION_NOTIFIER); +#endif + + for_each_online_cpu(cpu) { + if (cbe_get_hw_thread_id(cpu)) + continue; + + subfunc = 3; /* + * 2 - activate SPU tracing, + * 3 - deactivate + */ + lfsr_value = 0x8f100000; + + rtn_value = rtas_call(spu_rtas_token, 3, 1, NULL, + subfunc, cbe_cpu_to_node(cpu), + lfsr_value); + + if (unlikely(rtn_value != 0)) { + printk(KERN_ERR + "%s: rtas call ibm,cbe-spu-perftools " \ + "failed, return = %d\n", + __func__, rtn_value); + } + + /* Deactivate the signals */ + pm_rtas_reset_signals(cbe_cpu_to_node(cpu)); + } + + stop_spu_profiling_cycles(); +} + +static void cell_global_stop_spu_events(void) +{ + int cpu; + oprofile_running = 0; + + stop_spu_profiling_events(); + smp_wmb(); + + for_each_online_cpu(cpu) { + if (cbe_get_hw_thread_id(cpu)) + continue; + + cbe_sync_irq(cbe_cpu_to_node(cpu)); + /* Stop the counters */ + cbe_disable_pm(cpu); + cbe_write_pm07_control(cpu, 0, 0); + + /* Deactivate the signals */ + pm_rtas_reset_signals(cbe_cpu_to_node(cpu)); + + /* Deactivate interrupts */ + cbe_disable_pm_interrupts(cpu); + } + del_timer_sync(&timer_spu_event_swap); +} + +static void cell_global_stop_ppu(void) +{ + int cpu; + + /* + * This routine will be called once for the system. + * There is one performance monitor per node, so we + * only need to perform this function once per node. + */ + del_timer_sync(&timer_virt_cntr); + oprofile_running = 0; + smp_wmb(); + + for_each_online_cpu(cpu) { + if (cbe_get_hw_thread_id(cpu)) + continue; + + cbe_sync_irq(cbe_cpu_to_node(cpu)); + /* Stop the counters */ + cbe_disable_pm(cpu); + + /* Deactivate the signals */ + pm_rtas_reset_signals(cbe_cpu_to_node(cpu)); + + /* Deactivate interrupts */ + cbe_disable_pm_interrupts(cpu); + } +} + +static void cell_global_stop(void) +{ + if (profiling_mode == PPU_PROFILING) + cell_global_stop_ppu(); + else if (profiling_mode == SPU_PROFILING_EVENTS) + cell_global_stop_spu_events(); + else + cell_global_stop_spu_cycles(); +} + +static int cell_global_start_spu_cycles(struct op_counter_config *ctr) { int subfunc; unsigned int lfsr_value; @@ -951,18 +1315,18 @@ static int cell_global_start_spu(struct op_counter_config *ctr) /* start profiling */ ret = rtas_call(spu_rtas_token, 3, 1, NULL, subfunc, - cbe_cpu_to_node(cpu), lfsr_value); + cbe_cpu_to_node(cpu), lfsr_value); if (unlikely(ret != 0)) { printk(KERN_ERR - "%s: rtas call ibm,cbe-spu-perftools failed, return = %d\n", - __func__, ret); + "%s: rtas call ibm,cbe-spu-perftools failed, " \ + "return = %d\n", __func__, ret); rtas_error = -EIO; goto out; } } - rtas_error = start_spu_profiling(spu_cycle_reset); + rtas_error = start_spu_profiling_cycles(spu_cycle_reset); if (rtas_error) goto out_stop; @@ -970,11 +1334,74 @@ static int cell_global_start_spu(struct op_counter_config *ctr) return 0; out_stop: - cell_global_stop_spu(); /* clean up the PMU/debug bus */ + cell_global_stop_spu_cycles(); /* clean up the PMU/debug bus */ out: return rtas_error; } +static int cell_global_start_spu_events(struct op_counter_config *ctr) +{ + int cpu; + u32 interrupt_mask = 0; + int rtn = 0; + + hdw_thread = 0; + + /* spu event profiling, uses the performance counters to generate + * an interrupt. The hardware is setup to store the SPU program + * counter into the trace array. The occurrence mode is used to + * enable storing data to the trace buffer. The bits are set + * to send/store the SPU address in the trace buffer. The debug + * bus must be setup to route the SPU program counter onto the + * debug bus. The occurrence data in the trace buffer is not used. + */ + + /* This routine gets called once for the system. + * There is one performance monitor per node, so we + * only need to perform this function once per node. + */ + + for_each_online_cpu(cpu) { + if (cbe_get_hw_thread_id(cpu)) + continue; + + /* + * Setup SPU event-based profiling. + * Set perf_mon_control bit 0 to a zero before + * enabling spu collection hardware. + * + * Only support one SPU event on one SPU per node. + */ + if (ctr_enabled & 1) { + cbe_write_ctr(cpu, 0, reset_value[0]); + enable_ctr(cpu, 0, pm_regs.pm07_cntrl); + interrupt_mask |= + CBE_PM_CTR_OVERFLOW_INTR(0); + } else { + /* Disable counter */ + cbe_write_pm07_control(cpu, 0, 0); + } + + cbe_get_and_clear_pm_interrupts(cpu); + cbe_enable_pm_interrupts(cpu, hdw_thread, interrupt_mask); + cbe_enable_pm(cpu); + + /* clear the trace buffer */ + cbe_write_pm(cpu, trace_address, 0); + } + + /* Start the timer to time slice collecting the event profile + * on each of the SPUs. Note, can collect profile on one SPU + * per node at a time. + */ + start_spu_event_swap(); + start_spu_profiling_events(); + oprofile_running = 1; + smp_wmb(); + + return rtn; +} + static int cell_global_start_ppu(struct op_counter_config *ctr) { u32 cpu, i; @@ -994,8 +1421,7 @@ static int cell_global_start_ppu(struct op_counter_config *ctr) if (ctr_enabled & (1 << i)) { cbe_write_ctr(cpu, i, reset_value[i]); enable_ctr(cpu, i, pm_regs.pm07_cntrl); - interrupt_mask |= - CBE_PM_CTR_OVERFLOW_INTR(i); + interrupt_mask |= CBE_PM_CTR_OVERFLOW_INTR(i); } else { /* Disable counter */ cbe_write_pm07_control(cpu, i, 0); @@ -1024,99 +1450,162 @@ static int cell_global_start_ppu(struct op_counter_config *ctr) static int cell_global_start(struct op_counter_config *ctr) { - if (spu_cycle_reset) - return cell_global_start_spu(ctr); + if (profiling_mode == SPU_PROFILING_CYCLES) + return cell_global_start_spu_cycles(ctr); + else if (profiling_mode == SPU_PROFILING_EVENTS) + return cell_global_start_spu_events(ctr); else return cell_global_start_ppu(ctr); } -/* - * Note the generic OProfile stop calls do not support returning - * an error on stop. Hence, will not return an error if the FW - * calls fail on stop. Failure to reset the debug bus is not an issue. - * Failure to disable the SPU profiling is not an issue. The FW calls - * to enable the performance counters and debug bus will work even if - * the hardware was not cleanly reset. + +/* The SPU interrupt handler + * + * SPU event profiling works as follows: + * The pm_signal[0] holds the one SPU event to be measured. It is routed on + * the debug bus using word 0 or 1. The value of pm_signal[1] and + * pm_signal[2] contain the necessary events to route the SPU program + * counter for the selected SPU onto the debug bus using words 2 and 3. + * The pm_interval register is setup to write the SPU PC value into the + * trace buffer at the maximum rate possible. The trace buffer is configured + * to store the PCs, wrapping when it is full. The performance counter is + * intialized to the max hardware count minus the number of events, N, between + * samples. Once the N events have occured, a HW counter overflow occurs + * causing the generation of a HW counter interrupt which also stops the + * writing of the SPU PC values to the trace buffer. Hence the last PC + * written to the trace buffer is the SPU PC that we want. Unfortunately, + * we have to read from the beginning of the trace buffer to get to the + * last value written. We just hope the PPU has nothing better to do then + * service this interrupt. The PC for the specific SPU being profiled is + * extracted from the trace buffer processed and stored. The trace buffer + * is cleared, interrupts are cleared, the counter is reset to max - N. + * A kernel timer is used to periodically call the routine spu_evnt_swap() + * to switch to the next physical SPU in the node to profile in round robbin + * order. This way data is collected for all SPUs on the node. It does mean + * that we need to use a relatively small value of N to ensure enough samples + * on each SPU are collected each SPU is being profiled 1/8 of the time. + * It may also be necessary to use a longer sample collection period. */ -static void cell_global_stop_spu(void) +static void cell_handle_interrupt_spu(struct pt_regs *regs, + struct op_counter_config *ctr) { - int subfunc, rtn_value; - unsigned int lfsr_value; - int cpu; + u32 cpu, cpu_tmp; + u64 trace_entry; + u32 interrupt_mask; + u64 trace_buffer[2]; + u64 last_trace_buffer; + u32 sample; + u32 trace_addr; + unsigned long sample_array_lock_flags; + int spu_num; + unsigned long flags; - oprofile_running = 0; + /* Make sure spu event interrupt handler and spu event swap + * don't access the counters simultaneously. + */ + cpu = smp_processor_id(); + spin_lock_irqsave(&cntr_lock, flags); -#ifdef CONFIG_CPU_FREQ - cpufreq_unregister_notifier(&cpu_freq_notifier_block, - CPUFREQ_TRANSITION_NOTIFIER); -#endif + cpu_tmp = cpu; + cbe_disable_pm(cpu); - for_each_online_cpu(cpu) { - if (cbe_get_hw_thread_id(cpu)) - continue; + interrupt_mask = cbe_get_and_clear_pm_interrupts(cpu); - subfunc = 3; /* - * 2 - activate SPU tracing, - * 3 - deactivate - */ - lfsr_value = 0x8f100000; + sample = 0xABCDEF; + trace_entry = 0xfedcba; + last_trace_buffer = 0xdeadbeaf; - rtn_value = rtas_call(spu_rtas_token, 3, 1, NULL, - subfunc, cbe_cpu_to_node(cpu), - lfsr_value); + if ((oprofile_running == 1) && (interrupt_mask != 0)) { + /* disable writes to trace buff */ + cbe_write_pm(cpu, pm_interval, 0); - if (unlikely(rtn_value != 0)) { - printk(KERN_ERR - "%s: rtas call ibm,cbe-spu-perftools failed, return = %d\n", - __func__, rtn_value); + /* only have one perf cntr being used, cntr 0 */ + if ((interrupt_mask & CBE_PM_CTR_OVERFLOW_INTR(0)) + && ctr[0].enabled) + /* The SPU PC values will be read + * from the trace buffer, reset counter + */ + + cbe_write_ctr(cpu, 0, reset_value[0]); + + trace_addr = cbe_read_pm(cpu, trace_address); + + while (!(trace_addr & CBE_PM_TRACE_BUF_EMPTY)) { + /* There is data in the trace buffer to process + * Read the buffer until you get to the last + * entry. This is the value we want. + */ + + cbe_read_trace_buffer(cpu, trace_buffer); + trace_addr = cbe_read_pm(cpu, trace_address); } - /* Deactivate the signals */ - pm_rtas_reset_signals(cbe_cpu_to_node(cpu)); - } + /* SPU Address 16 bit count format for 128 bit + * HW trace buffer is used for the SPU PC storage + * HDR bits 0:15 + * SPU Addr 0 bits 16:31 + * SPU Addr 1 bits 32:47 + * unused bits 48:127 + * + * HDR: bit4 = 1 SPU Address 0 valid + * HDR: bit5 = 1 SPU Address 1 valid + * - unfortunately, the valid bits don't seem to work + * + * Note trace_buffer[0] holds bits 0:63 of the HW + * trace buffer, trace_buffer[1] holds bits 64:127 + */ - stop_spu_profiling(); -} + trace_entry = trace_buffer[0] + & 0x00000000FFFF0000; -static void cell_global_stop_ppu(void) -{ - int cpu; + /* only top 16 of the 18 bit SPU PC address + * is stored in trace buffer, hence shift right + * by 16 -2 bits */ + sample = trace_entry >> 14; + last_trace_buffer = trace_buffer[0]; - /* - * This routine will be called once for the system. - * There is one performance monitor per node, so we - * only need to perform this function once per node. - */ - del_timer_sync(&timer_virt_cntr); - oprofile_running = 0; - smp_wmb(); + spu_num = spu_evnt_phys_spu_indx + + (cbe_cpu_to_node(cpu) * NUM_SPUS_PER_NODE); - for_each_online_cpu(cpu) { - if (cbe_get_hw_thread_id(cpu)) - continue; + /* make sure only one process at a time is calling + * spu_sync_buffer() + */ + spin_lock_irqsave(&oprof_spu_smpl_arry_lck, + sample_array_lock_flags); + spu_sync_buffer(spu_num, &sample, 1); + spin_unlock_irqrestore(&oprof_spu_smpl_arry_lck, + sample_array_lock_flags); - cbe_sync_irq(cbe_cpu_to_node(cpu)); - /* Stop the counters */ - cbe_disable_pm(cpu); + smp_wmb(); /* insure spu event buffer updates are written + * don't want events intermingled... */ - /* Deactivate the signals */ - pm_rtas_reset_signals(cbe_cpu_to_node(cpu)); + /* The counters were frozen by the interrupt. + * Reenable the interrupt and restart the counters. + */ + cbe_write_pm(cpu, pm_interval, NUM_INTERVAL_CYC); + cbe_enable_pm_interrupts(cpu, hdw_thread, + virt_cntr_inter_mask); - /* Deactivate interrupts */ - cbe_disable_pm_interrupts(cpu); - } -} + /* clear the trace buffer, re-enable writes to trace buff */ + cbe_write_pm(cpu, trace_address, 0); + cbe_write_pm(cpu, pm_interval, NUM_INTERVAL_CYC); -static void cell_global_stop(void) -{ - if (spu_cycle_reset) - cell_global_stop_spu(); - else - cell_global_stop_ppu(); + /* The writes to the various performance counters only writes + * to a latch. The new values (interrupt setting bits, reset + * counter value etc.) are not copied to the actual registers + * until the performance monitor is enabled. In order to get + * this to work as desired, the permormance monitor needs to + * be disabled while writing to the latches. This is a + * HW design issue. + */ + write_pm_cntrl(cpu); + cbe_enable_pm(cpu); + } + spin_unlock_irqrestore(&cntr_lock, flags); } -static void cell_handle_interrupt(struct pt_regs *regs, - struct op_counter_config *ctr) +static void cell_handle_interrupt_ppu(struct pt_regs *regs, + struct op_counter_config *ctr) { u32 cpu; u64 pc; @@ -1132,7 +1621,7 @@ static void cell_handle_interrupt(struct pt_regs *regs, * routine are not running at the same time. See the * cell_virtual_cntr() routine for additional comments. */ - spin_lock_irqsave(&virt_cntr_lock, flags); + spin_lock_irqsave(&cntr_lock, flags); /* * Need to disable and reenable the performance counters @@ -1185,7 +1674,16 @@ static void cell_handle_interrupt(struct pt_regs *regs, */ cbe_enable_pm(cpu); } - spin_unlock_irqrestore(&virt_cntr_lock, flags); + spin_unlock_irqrestore(&cntr_lock, flags); +} + +static void cell_handle_interrupt(struct pt_regs *regs, + struct op_counter_config *ctr) +{ + if (profiling_mode == PPU_PROFILING) + cell_handle_interrupt_ppu(regs, ctr); + else + cell_handle_interrupt_spu(regs, ctr); } /* @@ -1195,7 +1693,8 @@ static void cell_handle_interrupt(struct pt_regs *regs, */ static int cell_sync_start(void) { - if (spu_cycle_reset) + if ((profiling_mode == SPU_PROFILING_CYCLES) || + (profiling_mode == SPU_PROFILING_EVENTS)) return spu_sync_start(); else return DO_GENERIC_SYNC; @@ -1203,7 +1702,8 @@ static int cell_sync_start(void) static int cell_sync_stop(void) { - if (spu_cycle_reset) + if ((profiling_mode == SPU_PROFILING_CYCLES) || + (profiling_mode == SPU_PROFILING_EVENTS)) return spu_sync_stop(); else return 1; diff --git a/arch/powerpc/platforms/cell/cbe_cpufreq.c b/arch/powerpc/platforms/cell/cbe_cpufreq.c index ec7c8f45a21..e6506cd0ff9 100644 --- a/arch/powerpc/platforms/cell/cbe_cpufreq.c +++ b/arch/powerpc/platforms/cell/cbe_cpufreq.c @@ -118,7 +118,7 @@ static int cbe_cpufreq_cpu_init(struct cpufreq_policy *policy) policy->cur = cbe_freqs[cur_pmode].frequency; #ifdef CONFIG_SMP - policy->cpus = per_cpu(cpu_sibling_map, policy->cpu); + cpumask_copy(policy->cpus, &per_cpu(cpu_sibling_map, policy->cpu)); #endif cpufreq_frequency_table_get_attr(cbe_freqs, policy->cpu); diff --git a/arch/powerpc/platforms/cell/cpufreq_spudemand.c b/arch/powerpc/platforms/cell/cpufreq_spudemand.c index a3c6c01bd6d..968c1c0b4d5 100644 --- a/arch/powerpc/platforms/cell/cpufreq_spudemand.c +++ b/arch/powerpc/platforms/cell/cpufreq_spudemand.c @@ -110,7 +110,7 @@ static int spu_gov_govern(struct cpufreq_policy *policy, unsigned int event) } /* initialize spu_gov_info for all affected cpus */ - for_each_cpu_mask(i, policy->cpus) { + for_each_cpu(i, policy->cpus) { affected_info = &per_cpu(spu_gov_info, i); affected_info->policy = policy; } @@ -127,7 +127,7 @@ static int spu_gov_govern(struct cpufreq_policy *policy, unsigned int event) spu_gov_cancel_work(info); /* clean spu_gov_info for all affected cpus */ - for_each_cpu_mask (i, policy->cpus) { + for_each_cpu (i, policy->cpus) { info = &per_cpu(spu_gov_info, i); info->policy = NULL; } diff --git a/arch/powerpc/platforms/cell/spufs/spufs.h b/arch/powerpc/platforms/cell/spufs/spufs.h index 15c62d3ca12..3bf908e2873 100644 --- a/arch/powerpc/platforms/cell/spufs/spufs.h +++ b/arch/powerpc/platforms/cell/spufs/spufs.h @@ -314,7 +314,7 @@ extern char *isolated_loader; * we need to call spu_release(ctx) before sleeping, and * then spu_acquire(ctx) when awoken. * - * Returns with state_mutex re-acquired when successfull or + * Returns with state_mutex re-acquired when successful or * with -ERESTARTSYS and the state_mutex dropped when interrupted. */ diff --git a/arch/powerpc/platforms/fsl_uli1575.c b/arch/powerpc/platforms/fsl_uli1575.c index 8c619963bec..1db6b9e037f 100644 --- a/arch/powerpc/platforms/fsl_uli1575.c +++ b/arch/powerpc/platforms/fsl_uli1575.c @@ -249,6 +249,7 @@ DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_AL, 0x5288, quirk_uli5288); DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_AL, 0x5229, quirk_uli5229); DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AL, 0x5249, quirk_final_uli5249); DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AL, 0x1575, quirk_final_uli1575); +DECLARE_PCI_FIXUP_RESUME(PCI_VENDOR_ID_AL, 0x5229, quirk_uli5229); static void __devinit hpcd_quirk_uli1575(struct pci_dev *dev) { diff --git a/arch/powerpc/platforms/pasemi/cpufreq.c b/arch/powerpc/platforms/pasemi/cpufreq.c index 86db47c1b66..be2527a516e 100644 --- a/arch/powerpc/platforms/pasemi/cpufreq.c +++ b/arch/powerpc/platforms/pasemi/cpufreq.c @@ -213,7 +213,7 @@ static int pas_cpufreq_cpu_init(struct cpufreq_policy *policy) pr_debug("current astate is at %d\n",cur_astate); policy->cur = pas_freqs[cur_astate].frequency; - policy->cpus = cpu_online_map; + cpumask_copy(policy->cpus, &cpu_online_map); ppc_proc_freq = policy->cur * 1000ul; diff --git a/arch/powerpc/platforms/powermac/cpufreq_64.c b/arch/powerpc/platforms/powermac/cpufreq_64.c index 4dfb4bc242b..beb38333b6d 100644 --- a/arch/powerpc/platforms/powermac/cpufreq_64.c +++ b/arch/powerpc/platforms/powermac/cpufreq_64.c @@ -362,7 +362,7 @@ static int g5_cpufreq_cpu_init(struct cpufreq_policy *policy) /* secondary CPUs are tied to the primary one by the * cpufreq core if in the secondary policy we tell it that * it actually must be one policy together with all others. */ - policy->cpus = cpu_online_map; + cpumask_copy(policy->cpus, &cpu_online_map); cpufreq_frequency_table_get_attr(g5_cpu_freqs, policy->cpu); return cpufreq_frequency_table_cpuinfo(policy, diff --git a/arch/powerpc/sysdev/fsl_pci.c b/arch/powerpc/sysdev/fsl_pci.c index f611d0369cc..9817f63723d 100644 --- a/arch/powerpc/sysdev/fsl_pci.c +++ b/arch/powerpc/sysdev/fsl_pci.c @@ -28,63 +28,104 @@ #include <sysdev/fsl_pci.h> #if defined(CONFIG_PPC_85xx) || defined(CONFIG_PPC_86xx) +static int __init setup_one_atmu(struct ccsr_pci __iomem *pci, + unsigned int index, const struct resource *res, + resource_size_t offset) +{ + resource_size_t pci_addr = res->start - offset; + resource_size_t phys_addr = res->start; + resource_size_t size = res->end - res->start + 1; + u32 flags = 0x80044000; /* enable & mem R/W */ + unsigned int i; + + pr_debug("PCI MEM resource start 0x%016llx, size 0x%016llx.\n", + (u64)res->start, (u64)size); + + if (res->flags & IORESOURCE_PREFETCH) + flags |= 0x10000000; /* enable relaxed ordering */ + + for (i = 0; size > 0; i++) { + unsigned int bits = min(__ilog2(size), + __ffs(pci_addr | phys_addr)); + + if (index + i >= 5) + return -1; + + out_be32(&pci->pow[index + i].potar, pci_addr >> 12); + out_be32(&pci->pow[index + i].potear, (u64)pci_addr >> 44); + out_be32(&pci->pow[index + i].powbar, phys_addr >> 12); + out_be32(&pci->pow[index + i].powar, flags | (bits - 1)); + + pci_addr += (resource_size_t)1U << bits; + phys_addr += (resource_size_t)1U << bits; + size -= (resource_size_t)1U << bits; + } + + return i; +} + /* atmu setup for fsl pci/pcie controller */ static void __init setup_pci_atmu(struct pci_controller *hose, struct resource *rsrc) { struct ccsr_pci __iomem *pci; - int i; + int i, j, n; pr_debug("PCI memory map start 0x%016llx, size 0x%016llx\n", (u64)rsrc->start, (u64)rsrc->end - (u64)rsrc->start + 1); pci = ioremap(rsrc->start, rsrc->end - rsrc->start + 1); + if (!pci) { + dev_err(hose->parent, "Unable to map ATMU registers\n"); + return; + } - /* Disable all windows (except powar0 since its ignored) */ + /* Disable all windows (except powar0 since it's ignored) */ for(i = 1; i < 5; i++) out_be32(&pci->pow[i].powar, 0); for(i = 0; i < 3; i++) out_be32(&pci->piw[i].piwar, 0); /* Setup outbound MEM window */ - for(i = 0; i < 3; i++) - if (hose->mem_resources[i].flags & IORESOURCE_MEM){ - resource_size_t pci_addr_start = - hose->mem_resources[i].start - - hose->pci_mem_offset; - pr_debug("PCI MEM resource start 0x%016llx, size 0x%016llx.\n", - (u64)hose->mem_resources[i].start, - (u64)hose->mem_resources[i].end - - (u64)hose->mem_resources[i].start + 1); - out_be32(&pci->pow[i+1].potar, (pci_addr_start >> 12)); - out_be32(&pci->pow[i+1].potear, 0); - out_be32(&pci->pow[i+1].powbar, - (hose->mem_resources[i].start >> 12)); - /* Enable, Mem R/W */ - out_be32(&pci->pow[i+1].powar, 0x80044000 - | (__ilog2(hose->mem_resources[i].end - - hose->mem_resources[i].start + 1) - 1)); - } + for(i = 0, j = 1; i < 3; i++) { + if (!(hose->mem_resources[i].flags & IORESOURCE_MEM)) + continue; + + n = setup_one_atmu(pci, j, &hose->mem_resources[i], + hose->pci_mem_offset); + + if (n < 0 || j >= 5) { + pr_err("Ran out of outbound PCI ATMUs for resource %d!\n", i); + hose->mem_resources[i].flags |= IORESOURCE_DISABLED; + } else + j += n; + } /* Setup outbound IO window */ - if (hose->io_resource.flags & IORESOURCE_IO){ - pr_debug("PCI IO resource start 0x%016llx, size 0x%016llx, " - "phy base 0x%016llx.\n", - (u64)hose->io_resource.start, - (u64)hose->io_resource.end - (u64)hose->io_resource.start + 1, - (u64)hose->io_base_phys); - out_be32(&pci->pow[i+1].potar, (hose->io_resource.start >> 12)); - out_be32(&pci->pow[i+1].potear, 0); - out_be32(&pci->pow[i+1].powbar, (hose->io_base_phys >> 12)); - /* Enable, IO R/W */ - out_be32(&pci->pow[i+1].powar, 0x80088000 - | (__ilog2(hose->io_resource.end - - hose->io_resource.start + 1) - 1)); + if (hose->io_resource.flags & IORESOURCE_IO) { + if (j >= 5) { + pr_err("Ran out of outbound PCI ATMUs for IO resource\n"); + } else { + pr_debug("PCI IO resource start 0x%016llx, size 0x%016llx, " + "phy base 0x%016llx.\n", + (u64)hose->io_resource.start, + (u64)hose->io_resource.end - (u64)hose->io_resource.start + 1, + (u64)hose->io_base_phys); + out_be32(&pci->pow[j].potar, (hose->io_resource.start >> 12)); + out_be32(&pci->pow[j].potear, 0); + out_be32(&pci->pow[j].powbar, (hose->io_base_phys >> 12)); + /* Enable, IO R/W */ + out_be32(&pci->pow[j].powar, 0x80088000 + | (__ilog2(hose->io_resource.end + - hose->io_resource.start + 1) - 1)); + } } /* Setup 2G inbound Memory Window @ 1 */ out_be32(&pci->piw[2].pitar, 0x00000000); out_be32(&pci->piw[2].piwbar,0x00000000); out_be32(&pci->piw[2].piwar, PIWAR_2G); + + iounmap(pci); } static void __init setup_pci_cmd(struct pci_controller *hose) |