diff options
Diffstat (limited to 'arch/powerpc/perf')
32 files changed, 3579 insertions, 982 deletions
diff --git a/arch/powerpc/perf/8xx-pmu.c b/arch/powerpc/perf/8xx-pmu.c index 19124b0b171a..308a2e40d7be 100644 --- a/arch/powerpc/perf/8xx-pmu.c +++ b/arch/powerpc/perf/8xx-pmu.c @@ -15,6 +15,7 @@ #include <asm/firmware.h> #include <asm/ptrace.h> #include <asm/code-patching.h> +#include <asm/inst.h> #define PERF_8xx_ID_CPU_CYCLES 1 #define PERF_8xx_ID_HW_INSTRUCTIONS 2 @@ -99,9 +100,6 @@ static int mpc8xx_pmu_add(struct perf_event *event, int flags) unsigned long target = patch_site_addr(&patch__itlbmiss_perf); patch_branch_site(&patch__itlbmiss_exit_1, target, 0); -#ifndef CONFIG_PIN_TLB_TEXT - patch_branch_site(&patch__itlbmiss_exit_2, target, 0); -#endif } val = itlb_miss_counter; break; @@ -110,8 +108,6 @@ static int mpc8xx_pmu_add(struct perf_event *event, int flags) unsigned long target = patch_site_addr(&patch__dtlbmiss_perf); patch_branch_site(&patch__dtlbmiss_exit_1, target, 0); - patch_branch_site(&patch__dtlbmiss_exit_2, target, 0); - patch_branch_site(&patch__dtlbmiss_exit_3, target, 0); } val = dtlb_miss_counter; break; @@ -157,13 +153,11 @@ static void mpc8xx_pmu_read(struct perf_event *event) static void mpc8xx_pmu_del(struct perf_event *event, int flags) { - /* mfspr r10, SPRN_SPRG_SCRATCH0 */ - unsigned int insn = PPC_INST_MFSPR | __PPC_RS(R10) | - __PPC_SPR(SPRN_SPRG_SCRATCH0); + ppc_inst_t insn = ppc_inst(PPC_RAW_MFSPR(10, SPRN_SPRG_SCRATCH2)); mpc8xx_pmu_read(event); - /* If it was the last user, stop counting to avoid useles overhead */ + /* If it was the last user, stop counting to avoid useless overhead */ switch (event_type(event)) { case PERF_8xx_ID_CPU_CYCLES: break; @@ -172,19 +166,12 @@ static void mpc8xx_pmu_del(struct perf_event *event, int flags) mtspr(SPRN_ICTRL, 7); break; case PERF_8xx_ID_ITLB_LOAD_MISS: - if (atomic_dec_return(&itlb_miss_ref) == 0) { + if (atomic_dec_return(&itlb_miss_ref) == 0) patch_instruction_site(&patch__itlbmiss_exit_1, insn); -#ifndef CONFIG_PIN_TLB_TEXT - patch_instruction_site(&patch__itlbmiss_exit_2, insn); -#endif - } break; case PERF_8xx_ID_DTLB_LOAD_MISS: - if (atomic_dec_return(&dtlb_miss_ref) == 0) { + if (atomic_dec_return(&dtlb_miss_ref) == 0) patch_instruction_site(&patch__dtlbmiss_exit_1, insn); - patch_instruction_site(&patch__dtlbmiss_exit_2, insn); - patch_instruction_site(&patch__dtlbmiss_exit_3, insn); - } break; } } diff --git a/arch/powerpc/perf/Makefile b/arch/powerpc/perf/Makefile index c155dcbb8691..4f53d0b97539 100644 --- a/arch/powerpc/perf/Makefile +++ b/arch/powerpc/perf/Makefile @@ -1,12 +1,13 @@ # SPDX-License-Identifier: GPL-2.0 -obj-$(CONFIG_PERF_EVENTS) += callchain.o perf_regs.o +obj-y += callchain.o callchain_$(BITS).o perf_regs.o +obj-$(CONFIG_COMPAT) += callchain_32.o -obj-$(CONFIG_PPC_PERF_CTRS) += core-book3s.o bhrb.o +obj-$(CONFIG_PPC_PERF_CTRS) += core-book3s.o obj64-$(CONFIG_PPC_PERF_CTRS) += ppc970-pmu.o power5-pmu.o \ power5+-pmu.o power6-pmu.o power7-pmu.o \ isa207-common.o power8-pmu.o power9-pmu.o \ - generic-compat-pmu.o + generic-compat-pmu.o power10-pmu.o bhrb.o obj32-$(CONFIG_PPC_PERF_CTRS) += mpc7450-pmu.o obj-$(CONFIG_PPC_POWERNV) += imc-pmu.o diff --git a/arch/powerpc/perf/bhrb.S b/arch/powerpc/perf/bhrb.S index 1aa3259716b8..47ba05d5ae76 100644 --- a/arch/powerpc/perf/bhrb.S +++ b/arch/powerpc/perf/bhrb.S @@ -21,7 +21,7 @@ _GLOBAL(read_bhrb) cmpldi r3,31 bgt 1f - ld r4,bhrb_table@got(r2) + LOAD_REG_ADDR(r4, bhrb_table) sldi r3,r3,3 add r3,r4,r3 mtctr r3 diff --git a/arch/powerpc/perf/callchain.c b/arch/powerpc/perf/callchain.c index 35d542515faf..6b4434dd0ff3 100644 --- a/arch/powerpc/perf/callchain.c +++ b/arch/powerpc/perf/callchain.c @@ -11,15 +11,12 @@ #include <linux/uaccess.h> #include <linux/mm.h> #include <asm/ptrace.h> -#include <asm/pgtable.h> #include <asm/sigcontext.h> #include <asm/ucontext.h> #include <asm/vdso.h> -#ifdef CONFIG_PPC64 -#include "../kernel/ppc32.h" -#endif #include <asm/pte-walk.h> +#include "callchain.h" /* * Is sp valid as the address of the next kernel stack frame after prev_sp? @@ -30,7 +27,7 @@ static int valid_next_sp(unsigned long sp, unsigned long prev_sp) { if (sp & 0xf) return 0; /* must be 16-byte aligned */ - if (!validate_sp(sp, current, STACK_FRAME_OVERHEAD)) + if (!validate_sp(sp, current)) return 0; if (sp >= prev_sp + STACK_FRAME_MIN_SIZE) return 1; @@ -43,7 +40,7 @@ static int valid_next_sp(unsigned long sp, unsigned long prev_sp) return 0; } -void +void __no_sanitize_address perf_callchain_kernel(struct perf_callchain_entry_ctx *entry, struct pt_regs *regs) { unsigned long sp, next_sp; @@ -56,7 +53,7 @@ perf_callchain_kernel(struct perf_callchain_entry_ctx *entry, struct pt_regs *re sp = regs->gpr[1]; perf_callchain_store(entry, perf_instruction_pointer(regs)); - if (!validate_sp(sp, current, STACK_FRAME_OVERHEAD)) + if (!validate_sp(sp, current)) return; for (;;) { @@ -64,12 +61,13 @@ perf_callchain_kernel(struct perf_callchain_entry_ctx *entry, struct pt_regs *re next_sp = fp[0]; if (next_sp == sp + STACK_INT_FRAME_SIZE && - fp[STACK_FRAME_MARKER] == STACK_FRAME_REGS_MARKER) { + validate_sp_size(sp, current, STACK_INT_FRAME_SIZE) && + fp[STACK_INT_FRAME_MARKER_LONGS] == STACK_FRAME_REGS_MARKER) { /* * This looks like an interrupt frame for an * interrupt that occurred in the kernel */ - regs = (struct pt_regs *)(sp + STACK_FRAME_OVERHEAD); + regs = (struct pt_regs *)(sp + STACK_INT_FRAME_REGS); next_ip = regs->nip; lr = regs->link; level = 0; @@ -102,372 +100,6 @@ perf_callchain_kernel(struct perf_callchain_entry_ctx *entry, struct pt_regs *re } } -#ifdef CONFIG_PPC64 -/* - * On 64-bit we don't want to invoke hash_page on user addresses from - * interrupt context, so if the access faults, we read the page tables - * to find which page (if any) is mapped and access it directly. - */ -static int read_user_stack_slow(void __user *ptr, void *buf, int nb) -{ - int ret = -EFAULT; - pgd_t *pgdir; - pte_t *ptep, pte; - unsigned shift; - unsigned long addr = (unsigned long) ptr; - unsigned long offset; - unsigned long pfn, flags; - void *kaddr; - - pgdir = current->mm->pgd; - if (!pgdir) - return -EFAULT; - - local_irq_save(flags); - ptep = find_current_mm_pte(pgdir, addr, NULL, &shift); - if (!ptep) - goto err_out; - if (!shift) - shift = PAGE_SHIFT; - - /* align address to page boundary */ - offset = addr & ((1UL << shift) - 1); - - pte = READ_ONCE(*ptep); - if (!pte_present(pte) || !pte_user(pte)) - goto err_out; - pfn = pte_pfn(pte); - if (!page_is_ram(pfn)) - goto err_out; - - /* no highmem to worry about here */ - kaddr = pfn_to_kaddr(pfn); - memcpy(buf, kaddr + offset, nb); - ret = 0; -err_out: - local_irq_restore(flags); - return ret; -} - -static int read_user_stack_64(unsigned long __user *ptr, unsigned long *ret) -{ - if ((unsigned long)ptr > TASK_SIZE - sizeof(unsigned long) || - ((unsigned long)ptr & 7)) - return -EFAULT; - - pagefault_disable(); - if (!__get_user_inatomic(*ret, ptr)) { - pagefault_enable(); - return 0; - } - pagefault_enable(); - - return read_user_stack_slow(ptr, ret, 8); -} - -static int read_user_stack_32(unsigned int __user *ptr, unsigned int *ret) -{ - if ((unsigned long)ptr > TASK_SIZE - sizeof(unsigned int) || - ((unsigned long)ptr & 3)) - return -EFAULT; - - pagefault_disable(); - if (!__get_user_inatomic(*ret, ptr)) { - pagefault_enable(); - return 0; - } - pagefault_enable(); - - return read_user_stack_slow(ptr, ret, 4); -} - -static inline int valid_user_sp(unsigned long sp, int is_64) -{ - if (!sp || (sp & 7) || sp > (is_64 ? TASK_SIZE : 0x100000000UL) - 32) - return 0; - return 1; -} - -/* - * 64-bit user processes use the same stack frame for RT and non-RT signals. - */ -struct signal_frame_64 { - char dummy[__SIGNAL_FRAMESIZE]; - struct ucontext uc; - unsigned long unused[2]; - unsigned int tramp[6]; - struct siginfo *pinfo; - void *puc; - struct siginfo info; - char abigap[288]; -}; - -static int is_sigreturn_64_address(unsigned long nip, unsigned long fp) -{ - if (nip == fp + offsetof(struct signal_frame_64, tramp)) - return 1; - if (vdso64_rt_sigtramp && current->mm->context.vdso_base && - nip == current->mm->context.vdso_base + vdso64_rt_sigtramp) - return 1; - return 0; -} - -/* - * Do some sanity checking on the signal frame pointed to by sp. - * We check the pinfo and puc pointers in the frame. - */ -static int sane_signal_64_frame(unsigned long sp) -{ - struct signal_frame_64 __user *sf; - unsigned long pinfo, puc; - - sf = (struct signal_frame_64 __user *) sp; - if (read_user_stack_64((unsigned long __user *) &sf->pinfo, &pinfo) || - read_user_stack_64((unsigned long __user *) &sf->puc, &puc)) - return 0; - return pinfo == (unsigned long) &sf->info && - puc == (unsigned long) &sf->uc; -} - -static void perf_callchain_user_64(struct perf_callchain_entry_ctx *entry, - struct pt_regs *regs) -{ - unsigned long sp, next_sp; - unsigned long next_ip; - unsigned long lr; - long level = 0; - struct signal_frame_64 __user *sigframe; - unsigned long __user *fp, *uregs; - - next_ip = perf_instruction_pointer(regs); - lr = regs->link; - sp = regs->gpr[1]; - perf_callchain_store(entry, next_ip); - - while (entry->nr < entry->max_stack) { - fp = (unsigned long __user *) sp; - if (!valid_user_sp(sp, 1) || read_user_stack_64(fp, &next_sp)) - return; - if (level > 0 && read_user_stack_64(&fp[2], &next_ip)) - return; - - /* - * Note: the next_sp - sp >= signal frame size check - * is true when next_sp < sp, which can happen when - * transitioning from an alternate signal stack to the - * normal stack. - */ - if (next_sp - sp >= sizeof(struct signal_frame_64) && - (is_sigreturn_64_address(next_ip, sp) || - (level <= 1 && is_sigreturn_64_address(lr, sp))) && - sane_signal_64_frame(sp)) { - /* - * This looks like an signal frame - */ - sigframe = (struct signal_frame_64 __user *) sp; - uregs = sigframe->uc.uc_mcontext.gp_regs; - if (read_user_stack_64(&uregs[PT_NIP], &next_ip) || - read_user_stack_64(&uregs[PT_LNK], &lr) || - read_user_stack_64(&uregs[PT_R1], &sp)) - return; - level = 0; - perf_callchain_store_context(entry, PERF_CONTEXT_USER); - perf_callchain_store(entry, next_ip); - continue; - } - - if (level == 0) - next_ip = lr; - perf_callchain_store(entry, next_ip); - ++level; - sp = next_sp; - } -} - -#else /* CONFIG_PPC64 */ -/* - * On 32-bit we just access the address and let hash_page create a - * HPTE if necessary, so there is no need to fall back to reading - * the page tables. Since this is called at interrupt level, - * do_page_fault() won't treat a DSI as a page fault. - */ -static int read_user_stack_32(unsigned int __user *ptr, unsigned int *ret) -{ - int rc; - - if ((unsigned long)ptr > TASK_SIZE - sizeof(unsigned int) || - ((unsigned long)ptr & 3)) - return -EFAULT; - - pagefault_disable(); - rc = __get_user_inatomic(*ret, ptr); - pagefault_enable(); - - return rc; -} - -static inline void perf_callchain_user_64(struct perf_callchain_entry_ctx *entry, - struct pt_regs *regs) -{ -} - -static inline int valid_user_sp(unsigned long sp, int is_64) -{ - if (!sp || (sp & 7) || sp > TASK_SIZE - 32) - return 0; - return 1; -} - -#define __SIGNAL_FRAMESIZE32 __SIGNAL_FRAMESIZE -#define sigcontext32 sigcontext -#define mcontext32 mcontext -#define ucontext32 ucontext -#define compat_siginfo_t struct siginfo - -#endif /* CONFIG_PPC64 */ - -/* - * Layout for non-RT signal frames - */ -struct signal_frame_32 { - char dummy[__SIGNAL_FRAMESIZE32]; - struct sigcontext32 sctx; - struct mcontext32 mctx; - int abigap[56]; -}; - -/* - * Layout for RT signal frames - */ -struct rt_signal_frame_32 { - char dummy[__SIGNAL_FRAMESIZE32 + 16]; - compat_siginfo_t info; - struct ucontext32 uc; - int abigap[56]; -}; - -static int is_sigreturn_32_address(unsigned int nip, unsigned int fp) -{ - if (nip == fp + offsetof(struct signal_frame_32, mctx.mc_pad)) - return 1; - if (vdso32_sigtramp && current->mm->context.vdso_base && - nip == current->mm->context.vdso_base + vdso32_sigtramp) - return 1; - return 0; -} - -static int is_rt_sigreturn_32_address(unsigned int nip, unsigned int fp) -{ - if (nip == fp + offsetof(struct rt_signal_frame_32, - uc.uc_mcontext.mc_pad)) - return 1; - if (vdso32_rt_sigtramp && current->mm->context.vdso_base && - nip == current->mm->context.vdso_base + vdso32_rt_sigtramp) - return 1; - return 0; -} - -static int sane_signal_32_frame(unsigned int sp) -{ - struct signal_frame_32 __user *sf; - unsigned int regs; - - sf = (struct signal_frame_32 __user *) (unsigned long) sp; - if (read_user_stack_32((unsigned int __user *) &sf->sctx.regs, ®s)) - return 0; - return regs == (unsigned long) &sf->mctx; -} - -static int sane_rt_signal_32_frame(unsigned int sp) -{ - struct rt_signal_frame_32 __user *sf; - unsigned int regs; - - sf = (struct rt_signal_frame_32 __user *) (unsigned long) sp; - if (read_user_stack_32((unsigned int __user *) &sf->uc.uc_regs, ®s)) - return 0; - return regs == (unsigned long) &sf->uc.uc_mcontext; -} - -static unsigned int __user *signal_frame_32_regs(unsigned int sp, - unsigned int next_sp, unsigned int next_ip) -{ - struct mcontext32 __user *mctx = NULL; - struct signal_frame_32 __user *sf; - struct rt_signal_frame_32 __user *rt_sf; - - /* - * Note: the next_sp - sp >= signal frame size check - * is true when next_sp < sp, for example, when - * transitioning from an alternate signal stack to the - * normal stack. - */ - if (next_sp - sp >= sizeof(struct signal_frame_32) && - is_sigreturn_32_address(next_ip, sp) && - sane_signal_32_frame(sp)) { - sf = (struct signal_frame_32 __user *) (unsigned long) sp; - mctx = &sf->mctx; - } - - if (!mctx && next_sp - sp >= sizeof(struct rt_signal_frame_32) && - is_rt_sigreturn_32_address(next_ip, sp) && - sane_rt_signal_32_frame(sp)) { - rt_sf = (struct rt_signal_frame_32 __user *) (unsigned long) sp; - mctx = &rt_sf->uc.uc_mcontext; - } - - if (!mctx) - return NULL; - return mctx->mc_gregs; -} - -static void perf_callchain_user_32(struct perf_callchain_entry_ctx *entry, - struct pt_regs *regs) -{ - unsigned int sp, next_sp; - unsigned int next_ip; - unsigned int lr; - long level = 0; - unsigned int __user *fp, *uregs; - - next_ip = perf_instruction_pointer(regs); - lr = regs->link; - sp = regs->gpr[1]; - perf_callchain_store(entry, next_ip); - - while (entry->nr < entry->max_stack) { - fp = (unsigned int __user *) (unsigned long) sp; - if (!valid_user_sp(sp, 0) || read_user_stack_32(fp, &next_sp)) - return; - if (level > 0 && read_user_stack_32(&fp[1], &next_ip)) - return; - - uregs = signal_frame_32_regs(sp, next_sp, next_ip); - if (!uregs && level <= 1) - uregs = signal_frame_32_regs(sp, next_sp, lr); - if (uregs) { - /* - * This looks like an signal frame, so restart - * the stack trace with the values in it. - */ - if (read_user_stack_32(&uregs[PT_NIP], &next_ip) || - read_user_stack_32(&uregs[PT_LNK], &lr) || - read_user_stack_32(&uregs[PT_R1], &sp)) - return; - level = 0; - perf_callchain_store_context(entry, PERF_CONTEXT_USER); - perf_callchain_store(entry, next_ip); - continue; - } - - if (level == 0) - next_ip = lr; - perf_callchain_store(entry, next_ip); - ++level; - sp = next_sp; - } -} - void perf_callchain_user(struct perf_callchain_entry_ctx *entry, struct pt_regs *regs) { diff --git a/arch/powerpc/perf/callchain.h b/arch/powerpc/perf/callchain.h new file mode 100644 index 000000000000..19a8d051ddf1 --- /dev/null +++ b/arch/powerpc/perf/callchain.h @@ -0,0 +1,35 @@ +/* SPDX-License-Identifier: GPL-2.0-or-later */ +#ifndef _POWERPC_PERF_CALLCHAIN_H +#define _POWERPC_PERF_CALLCHAIN_H + +void perf_callchain_user_64(struct perf_callchain_entry_ctx *entry, + struct pt_regs *regs); +void perf_callchain_user_32(struct perf_callchain_entry_ctx *entry, + struct pt_regs *regs); + +static inline bool invalid_user_sp(unsigned long sp) +{ + unsigned long mask = is_32bit_task() ? 3 : 7; + unsigned long top = STACK_TOP - (is_32bit_task() ? 16 : 32); + + return (!sp || (sp & mask) || (sp > top)); +} + +/* + * On 32-bit we just access the address and let hash_page create a + * HPTE if necessary, so there is no need to fall back to reading + * the page tables. Since this is called at interrupt level, + * do_page_fault() won't treat a DSI as a page fault. + */ +static inline int __read_user_stack(const void __user *ptr, void *ret, + size_t size) +{ + unsigned long addr = (unsigned long)ptr; + + if (addr > TASK_SIZE - size || (addr & (size - 1))) + return -EFAULT; + + return copy_from_user_nofault(ret, ptr, size); +} + +#endif /* _POWERPC_PERF_CALLCHAIN_H */ diff --git a/arch/powerpc/perf/callchain_32.c b/arch/powerpc/perf/callchain_32.c new file mode 100644 index 000000000000..ea8cfe3806dc --- /dev/null +++ b/arch/powerpc/perf/callchain_32.c @@ -0,0 +1,178 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * Performance counter callchain support - powerpc architecture code + * + * Copyright © 2009 Paul Mackerras, IBM Corporation. + */ +#include <linux/kernel.h> +#include <linux/sched.h> +#include <linux/perf_event.h> +#include <linux/percpu.h> +#include <linux/uaccess.h> +#include <linux/mm.h> +#include <asm/ptrace.h> +#include <asm/sigcontext.h> +#include <asm/ucontext.h> +#include <asm/vdso.h> +#include <asm/pte-walk.h> + +#include "callchain.h" + +#ifdef CONFIG_PPC64 +#include <asm/syscalls_32.h> +#else /* CONFIG_PPC64 */ + +#define __SIGNAL_FRAMESIZE32 __SIGNAL_FRAMESIZE +#define sigcontext32 sigcontext +#define mcontext32 mcontext +#define ucontext32 ucontext +#define compat_siginfo_t struct siginfo + +#endif /* CONFIG_PPC64 */ + +static int read_user_stack_32(const unsigned int __user *ptr, unsigned int *ret) +{ + return __read_user_stack(ptr, ret, sizeof(*ret)); +} + +/* + * Layout for non-RT signal frames + */ +struct signal_frame_32 { + char dummy[__SIGNAL_FRAMESIZE32]; + struct sigcontext32 sctx; + struct mcontext32 mctx; + int abigap[56]; +}; + +/* + * Layout for RT signal frames + */ +struct rt_signal_frame_32 { + char dummy[__SIGNAL_FRAMESIZE32 + 16]; + compat_siginfo_t info; + struct ucontext32 uc; + int abigap[56]; +}; + +static int is_sigreturn_32_address(unsigned int nip, unsigned int fp) +{ + if (nip == fp + offsetof(struct signal_frame_32, mctx.mc_pad)) + return 1; + if (current->mm->context.vdso && + nip == VDSO32_SYMBOL(current->mm->context.vdso, sigtramp32)) + return 1; + return 0; +} + +static int is_rt_sigreturn_32_address(unsigned int nip, unsigned int fp) +{ + if (nip == fp + offsetof(struct rt_signal_frame_32, + uc.uc_mcontext.mc_pad)) + return 1; + if (current->mm->context.vdso && + nip == VDSO32_SYMBOL(current->mm->context.vdso, sigtramp_rt32)) + return 1; + return 0; +} + +static int sane_signal_32_frame(unsigned int sp) +{ + struct signal_frame_32 __user *sf; + unsigned int regs; + + sf = (struct signal_frame_32 __user *) (unsigned long) sp; + if (read_user_stack_32((unsigned int __user *) &sf->sctx.regs, ®s)) + return 0; + return regs == (unsigned long) &sf->mctx; +} + +static int sane_rt_signal_32_frame(unsigned int sp) +{ + struct rt_signal_frame_32 __user *sf; + unsigned int regs; + + sf = (struct rt_signal_frame_32 __user *) (unsigned long) sp; + if (read_user_stack_32((unsigned int __user *) &sf->uc.uc_regs, ®s)) + return 0; + return regs == (unsigned long) &sf->uc.uc_mcontext; +} + +static unsigned int __user *signal_frame_32_regs(unsigned int sp, + unsigned int next_sp, unsigned int next_ip) +{ + struct mcontext32 __user *mctx = NULL; + struct signal_frame_32 __user *sf; + struct rt_signal_frame_32 __user *rt_sf; + + /* + * Note: the next_sp - sp >= signal frame size check + * is true when next_sp < sp, for example, when + * transitioning from an alternate signal stack to the + * normal stack. + */ + if (next_sp - sp >= sizeof(struct signal_frame_32) && + is_sigreturn_32_address(next_ip, sp) && + sane_signal_32_frame(sp)) { + sf = (struct signal_frame_32 __user *) (unsigned long) sp; + mctx = &sf->mctx; + } + + if (!mctx && next_sp - sp >= sizeof(struct rt_signal_frame_32) && + is_rt_sigreturn_32_address(next_ip, sp) && + sane_rt_signal_32_frame(sp)) { + rt_sf = (struct rt_signal_frame_32 __user *) (unsigned long) sp; + mctx = &rt_sf->uc.uc_mcontext; + } + + if (!mctx) + return NULL; + return mctx->mc_gregs; +} + +void perf_callchain_user_32(struct perf_callchain_entry_ctx *entry, + struct pt_regs *regs) +{ + unsigned int sp, next_sp; + unsigned int next_ip; + unsigned int lr; + long level = 0; + unsigned int __user *fp, *uregs; + + next_ip = perf_instruction_pointer(regs); + lr = regs->link; + sp = regs->gpr[1]; + perf_callchain_store(entry, next_ip); + + while (entry->nr < entry->max_stack) { + fp = (unsigned int __user *) (unsigned long) sp; + if (invalid_user_sp(sp) || read_user_stack_32(fp, &next_sp)) + return; + if (level > 0 && read_user_stack_32(&fp[1], &next_ip)) + return; + + uregs = signal_frame_32_regs(sp, next_sp, next_ip); + if (!uregs && level <= 1) + uregs = signal_frame_32_regs(sp, next_sp, lr); + if (uregs) { + /* + * This looks like an signal frame, so restart + * the stack trace with the values in it. + */ + if (read_user_stack_32(&uregs[PT_NIP], &next_ip) || + read_user_stack_32(&uregs[PT_LNK], &lr) || + read_user_stack_32(&uregs[PT_R1], &sp)) + return; + level = 0; + perf_callchain_store_context(entry, PERF_CONTEXT_USER); + perf_callchain_store(entry, next_ip); + continue; + } + + if (level == 0) + next_ip = lr; + perf_callchain_store(entry, next_ip); + ++level; + sp = next_sp; + } +} diff --git a/arch/powerpc/perf/callchain_64.c b/arch/powerpc/perf/callchain_64.c new file mode 100644 index 000000000000..488e8a21a11e --- /dev/null +++ b/arch/powerpc/perf/callchain_64.c @@ -0,0 +1,120 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * Performance counter callchain support - powerpc architecture code + * + * Copyright © 2009 Paul Mackerras, IBM Corporation. + */ +#include <linux/kernel.h> +#include <linux/sched.h> +#include <linux/perf_event.h> +#include <linux/percpu.h> +#include <linux/uaccess.h> +#include <linux/mm.h> +#include <asm/ptrace.h> +#include <asm/sigcontext.h> +#include <asm/ucontext.h> +#include <asm/vdso.h> +#include <asm/pte-walk.h> + +#include "callchain.h" + +static int read_user_stack_64(const unsigned long __user *ptr, unsigned long *ret) +{ + return __read_user_stack(ptr, ret, sizeof(*ret)); +} + +/* + * 64-bit user processes use the same stack frame for RT and non-RT signals. + */ +struct signal_frame_64 { + char dummy[__SIGNAL_FRAMESIZE]; + struct ucontext uc; + unsigned long unused[2]; + unsigned int tramp[6]; + struct siginfo *pinfo; + void *puc; + struct siginfo info; + char abigap[288]; +}; + +static int is_sigreturn_64_address(unsigned long nip, unsigned long fp) +{ + if (nip == fp + offsetof(struct signal_frame_64, tramp)) + return 1; + if (current->mm->context.vdso && + nip == VDSO64_SYMBOL(current->mm->context.vdso, sigtramp_rt64)) + return 1; + return 0; +} + +/* + * Do some sanity checking on the signal frame pointed to by sp. + * We check the pinfo and puc pointers in the frame. + */ +static int sane_signal_64_frame(unsigned long sp) +{ + struct signal_frame_64 __user *sf; + unsigned long pinfo, puc; + + sf = (struct signal_frame_64 __user *) sp; + if (read_user_stack_64((unsigned long __user *) &sf->pinfo, &pinfo) || + read_user_stack_64((unsigned long __user *) &sf->puc, &puc)) + return 0; + return pinfo == (unsigned long) &sf->info && + puc == (unsigned long) &sf->uc; +} + +void perf_callchain_user_64(struct perf_callchain_entry_ctx *entry, + struct pt_regs *regs) +{ + unsigned long sp, next_sp; + unsigned long next_ip; + unsigned long lr; + long level = 0; + struct signal_frame_64 __user *sigframe; + unsigned long __user *fp, *uregs; + + next_ip = perf_instruction_pointer(regs); + lr = regs->link; + sp = regs->gpr[1]; + perf_callchain_store(entry, next_ip); + + while (entry->nr < entry->max_stack) { + fp = (unsigned long __user *) sp; + if (invalid_user_sp(sp) || read_user_stack_64(fp, &next_sp)) + return; + if (level > 0 && read_user_stack_64(&fp[2], &next_ip)) + return; + + /* + * Note: the next_sp - sp >= signal frame size check + * is true when next_sp < sp, which can happen when + * transitioning from an alternate signal stack to the + * normal stack. + */ + if (next_sp - sp >= sizeof(struct signal_frame_64) && + (is_sigreturn_64_address(next_ip, sp) || + (level <= 1 && is_sigreturn_64_address(lr, sp))) && + sane_signal_64_frame(sp)) { + /* + * This looks like an signal frame + */ + sigframe = (struct signal_frame_64 __user *) sp; + uregs = sigframe->uc.uc_mcontext.gp_regs; + if (read_user_stack_64(&uregs[PT_NIP], &next_ip) || + read_user_stack_64(&uregs[PT_LNK], &lr) || + read_user_stack_64(&uregs[PT_R1], &sp)) + return; + level = 0; + perf_callchain_store_context(entry, PERF_CONTEXT_USER); + perf_callchain_store(entry, next_ip); + continue; + } + + if (level == 0) + next_ip = lr; + perf_callchain_store(entry, next_ip); + ++level; + sp = next_sp; + } +} diff --git a/arch/powerpc/perf/core-book3s.c b/arch/powerpc/perf/core-book3s.c index 48604625ab31..6b5f8a94e7d8 100644 --- a/arch/powerpc/perf/core-book3s.c +++ b/arch/powerpc/perf/core-book3s.c @@ -17,6 +17,8 @@ #include <asm/firmware.h> #include <asm/ptrace.h> #include <asm/code-patching.h> +#include <asm/hw_irq.h> +#include <asm/interrupt.h> #ifdef CONFIG_PPC64 #include "internal.h" @@ -37,12 +39,7 @@ struct cpu_hw_events { struct perf_event *event[MAX_HWEVENTS]; u64 events[MAX_HWEVENTS]; unsigned int flags[MAX_HWEVENTS]; - /* - * The order of the MMCR array is: - * - 64-bit, MMCR0, MMCR1, MMCRA, MMCR2 - * - 32-bit, MMCR0, MMCR1, MMCR2 - */ - unsigned long mmcr[4]; + struct mmcr_regs mmcr; struct perf_event *limited_counter[MAX_LIMITED_HWCOUNTERS]; u8 limited_hwidx[MAX_LIMITED_HWCOUNTERS]; u64 alternatives[MAX_HWEVENTS][MAX_EVENT_ALTERNATIVES]; @@ -59,6 +56,9 @@ struct cpu_hw_events { struct perf_branch_stack bhrb_stack; struct perf_branch_entry bhrb_entries[BHRB_MAX_ENTRIES]; u64 ic_init; + + /* Store the PMC values */ + unsigned long pmcs[MAX_HWEVENTS]; }; static DEFINE_PER_CPU(struct cpu_hw_events, cpu_hw_events); @@ -77,6 +77,11 @@ static unsigned int freeze_events_kernel = MMCR0_FCS; /* * 32-bit doesn't have MMCRA but does have an MMCR2, * and a few other names are different. + * Also 32-bit doesn't have MMCR3, SIER2 and SIER3. + * Define them as zero knowing that any code path accessing + * these registers (via mtspr/mfspr) are done under ppmu flag + * check for PPMU_ARCH_31 and we will not enter that code path + * for 32-bit. */ #ifdef CONFIG_PPC32 @@ -90,7 +95,12 @@ static unsigned int freeze_events_kernel = MMCR0_FCS; #define MMCR0_PMCC_U6 0 #define SPRN_MMCRA SPRN_MMCR2 +#define SPRN_MMCR3 0 +#define SPRN_SIER2 0 +#define SPRN_SIER3 0 #define MMCRA_SAMPLE_ENABLE 0 +#define MMCRA_BHRB_DISABLE 0 +#define MMCR0_PMCCEXT 0 static inline unsigned long perf_ip_adjust(struct pt_regs *regs) { @@ -105,10 +115,6 @@ static inline void perf_read_regs(struct pt_regs *regs) { regs->result = 0; } -static inline int perf_intr_is_nmi(struct pt_regs *regs) -{ - return 0; -} static inline int siar_valid(struct pt_regs *regs) { @@ -121,24 +127,38 @@ static void ebb_event_add(struct perf_event *event) { } static void ebb_switch_out(unsigned long mmcr0) { } static unsigned long ebb_switch_in(bool ebb, struct cpu_hw_events *cpuhw) { - return cpuhw->mmcr[0]; + return cpuhw->mmcr.mmcr0; } static inline void power_pmu_bhrb_enable(struct perf_event *event) {} static inline void power_pmu_bhrb_disable(struct perf_event *event) {} -static void power_pmu_sched_task(struct perf_event_context *ctx, bool sched_in) {} +static void power_pmu_sched_task(struct perf_event_pmu_context *pmu_ctx, bool sched_in) {} static inline void power_pmu_bhrb_read(struct perf_event *event, struct cpu_hw_events *cpuhw) {} static void pmao_restore_workaround(bool ebb) { } #endif /* CONFIG_PPC32 */ bool is_sier_available(void) { + if (!ppmu) + return false; + if (ppmu->flags & PPMU_HAS_SIER) return true; return false; } +/* + * Return PMC value corresponding to the + * index passed. + */ +unsigned long get_pmcs_ext_regs(int idx) +{ + struct cpu_hw_events *cpuhw = this_cpu_ptr(&cpu_hw_events); + + return cpuhw->pmcs[idx]; +} + static bool regs_use_siar(struct pt_regs *regs) { /* @@ -150,7 +170,7 @@ static bool regs_use_siar(struct pt_regs *regs) * they have not been setup using perf_read_regs() and so regs->result * is something random. */ - return ((TRAP(regs) == 0xf00) && regs->result); + return ((TRAP(regs) == INTERRUPT_PERFMON) && regs->result); } /* @@ -204,7 +224,7 @@ static inline void perf_get_data_addr(struct perf_event *event, struct pt_regs * if (!(mmcra & MMCRA_SAMPLE_ENABLE) || sdar_valid) *addrp = mfspr(SPRN_SDAR); - if (is_kernel_addr(mfspr(SPRN_SDAR)) && perf_allow_kernel(&event->attr) != 0) + if (is_kernel_addr(mfspr(SPRN_SDAR)) && event->attr.exclude_kernel) *addrp = 0; } @@ -236,7 +256,7 @@ static bool regs_sipr(struct pt_regs *regs) static inline u32 perf_flags_from_msr(struct pt_regs *regs) { - if (regs->msr & MSR_PR) + if (user_mode(regs)) return PERF_RECORD_MISC_USER; if ((regs->msr & MSR_HV) && freeze_events_kernel != MMCR0_FCHV) return PERF_RECORD_MISC_HYPERVISOR; @@ -246,11 +266,32 @@ static inline u32 perf_flags_from_msr(struct pt_regs *regs) static inline u32 perf_get_misc_flags(struct pt_regs *regs) { bool use_siar = regs_use_siar(regs); + unsigned long mmcra = regs->dsisr; + int marked = mmcra & MMCRA_SAMPLE_ENABLE; if (!use_siar) return perf_flags_from_msr(regs); /* + * Check the address in SIAR to identify the + * privilege levels since the SIER[MSR_HV, MSR_PR] + * bits are not set for marked events in power10 + * DD1. + */ + if (marked && (ppmu->flags & PPMU_P10_DD1)) { + unsigned long siar = mfspr(SPRN_SIAR); + if (siar) { + if (is_kernel_addr(siar)) + return PERF_RECORD_MISC_KERNEL; + return PERF_RECORD_MISC_USER; + } else { + if (is_kernel_addr(regs->nip)) + return PERF_RECORD_MISC_KERNEL; + return PERF_RECORD_MISC_USER; + } + } + + /* * If we don't have flags in MMCRA, rather than using * the MSR, we intuit the flags from the address in * SIAR which should give slightly more reliable @@ -300,6 +341,13 @@ static inline void perf_read_regs(struct pt_regs *regs) * If the PMU doesn't update the SIAR for non marked events use * pt_regs. * + * If regs is a kernel interrupt, always use SIAR. Some PMUs have an + * issue with regs_sipr not being in synch with SIAR in interrupt entry + * and return sequences, which can result in regs_sipr being true for + * kernel interrupts and SIAR, which has the effect of causing samples + * to pile up at mtmsrd MSR[EE] 0->1 or pending irq replay around + * interrupt entry/exit. + * * If the PMU has HV/PR flags then check to see if they * place the exception in userspace. If so, use pt_regs. In * continuous sampling mode the SIAR and the PMU exception are @@ -308,7 +356,7 @@ static inline void perf_read_regs(struct pt_regs *regs) * hypervisor samples as well as samples in the kernel with * interrupts off hence the userspace check. */ - if (TRAP(regs) != 0xf00) + if (TRAP(regs) != INTERRUPT_PERFMON) use_siar = 0; else if ((ppmu->flags & PPMU_NO_SIAR)) use_siar = 0; @@ -316,6 +364,8 @@ static inline void perf_read_regs(struct pt_regs *regs) use_siar = 1; else if ((ppmu->flags & PPMU_NO_CONT_SAMPLING)) use_siar = 0; + else if (!user_mode(regs)) + use_siar = 1; else if (!(ppmu->flags & PPMU_NO_SIPR) && regs_sipr(regs)) use_siar = 0; else @@ -325,15 +375,6 @@ static inline void perf_read_regs(struct pt_regs *regs) } /* - * If interrupts were soft-disabled when a PMU interrupt occurs, treat - * it as an NMI. - */ -static inline int perf_intr_is_nmi(struct pt_regs *regs) -{ - return (regs->softe & IRQS_DISABLED); -} - -/* * On processors like P7+ that have the SIAR-Valid bit, marked instructions * must be sampled only if the SIAR-valid bit is set. * @@ -346,7 +387,14 @@ static inline int siar_valid(struct pt_regs *regs) int marked = mmcra & MMCRA_SAMPLE_ENABLE; if (marked) { - if (ppmu->flags & PPMU_HAS_SIER) + /* + * SIER[SIAR_VALID] is not set for some + * marked events on power10 DD1, so drop + * the check for SIER[SIAR_VALID] and return true. + */ + if (ppmu->flags & PPMU_P10_DD1) + return 0x1; + else if (ppmu->flags & PPMU_HAS_SIER) return regs->dar & SIER_SIAR_VALID; if (ppmu->flags & PPMU_SIAR_VALID) @@ -376,7 +424,7 @@ static void power_pmu_bhrb_enable(struct perf_event *event) cpuhw->bhrb_context = event->ctx; } cpuhw->bhrb_users++; - perf_sched_cb_inc(event->ctx->pmu); + perf_sched_cb_inc(event->pmu); } static void power_pmu_bhrb_disable(struct perf_event *event) @@ -388,7 +436,7 @@ static void power_pmu_bhrb_disable(struct perf_event *event) WARN_ON_ONCE(!cpuhw->bhrb_users); cpuhw->bhrb_users--; - perf_sched_cb_dec(event->ctx->pmu); + perf_sched_cb_dec(event->pmu); if (!cpuhw->disabled && !cpuhw->bhrb_users) { /* BHRB cannot be turned off when other @@ -403,7 +451,7 @@ static void power_pmu_bhrb_disable(struct perf_event *event) /* Called from ctxsw to prevent one process's branch entries to * mingle with the other process's entries during context switch. */ -static void power_pmu_sched_task(struct perf_event_context *ctx, bool sched_in) +static void power_pmu_sched_task(struct perf_event_pmu_context *pmu_ctx, bool sched_in) { if (!ppmu->bhrb_nr) return; @@ -415,24 +463,20 @@ static void power_pmu_sched_task(struct perf_event_context *ctx, bool sched_in) static __u64 power_pmu_bhrb_to(u64 addr) { unsigned int instr; - int ret; __u64 target; if (is_kernel_addr(addr)) { - if (probe_kernel_read(&instr, (void *)addr, sizeof(instr))) + if (copy_from_kernel_nofault(&instr, (void *)addr, + sizeof(instr))) return 0; return branch_target(&instr); } /* Userspace: need copy instruction here then translate it */ - pagefault_disable(); - ret = __get_user_inatomic(instr, (unsigned int __user *)addr); - if (ret) { - pagefault_enable(); + if (copy_from_user_nofault(&instr, (unsigned int __user *)addr, + sizeof(instr))) return 0; - } - pagefault_enable(); target = branch_target(&instr); if ((!target) || (instr & BRANCH_ABSOLUTE)) @@ -470,8 +514,11 @@ static void power_pmu_bhrb_read(struct perf_event *event, struct cpu_hw_events * * addresses at this point. Check the privileges before * exporting it to userspace (avoid exposure of regions * where we could have speculative execution) + * Incase of ISA v3.1, BHRB will capture only user-space + * addresses, hence include a check before filtering code */ - if (is_kernel_addr(addr) && perf_allow_kernel(&event->attr) != 0) + if (!(ppmu->flags & PPMU_ARCH_31) && + is_kernel_addr(addr) && event->attr.exclude_kernel) continue; /* Branches are read most recent first (ie. mfbhrb 0 is @@ -524,6 +571,7 @@ static void power_pmu_bhrb_read(struct perf_event *event, struct cpu_hw_events * } } cpuhw->bhrb_stack.nr = u_index; + cpuhw->bhrb_stack.hw_idx = -1ULL; return; } @@ -589,11 +637,16 @@ static void ebb_switch_out(unsigned long mmcr0) current->thread.sdar = mfspr(SPRN_SDAR); current->thread.mmcr0 = mmcr0 & MMCR0_USER_MASK; current->thread.mmcr2 = mfspr(SPRN_MMCR2) & MMCR2_USER_MASK; + if (ppmu->flags & PPMU_ARCH_31) { + current->thread.mmcr3 = mfspr(SPRN_MMCR3); + current->thread.sier2 = mfspr(SPRN_SIER2); + current->thread.sier3 = mfspr(SPRN_SIER3); + } } static unsigned long ebb_switch_in(bool ebb, struct cpu_hw_events *cpuhw) { - unsigned long mmcr0 = cpuhw->mmcr[0]; + unsigned long mmcr0 = cpuhw->mmcr.mmcr0; if (!ebb) goto out; @@ -627,7 +680,13 @@ static unsigned long ebb_switch_in(bool ebb, struct cpu_hw_events *cpuhw) * unfreeze counters, it should not set exclude_xxx in its events and * instead manage the MMCR2 entirely by itself. */ - mtspr(SPRN_MMCR2, cpuhw->mmcr[3] | current->thread.mmcr2); + mtspr(SPRN_MMCR2, cpuhw->mmcr.mmcr2 | current->thread.mmcr2); + + if (ppmu->flags & PPMU_ARCH_31) { + mtspr(SPRN_MMCR3, current->thread.mmcr3); + mtspr(SPRN_SIER2, current->thread.sier2); + mtspr(SPRN_SIER3, current->thread.sier3); + } out: return mmcr0; } @@ -717,6 +776,34 @@ static void pmao_restore_workaround(bool ebb) mtspr(SPRN_PMC6, pmcs[5]); } +/* + * If the perf subsystem wants performance monitor interrupts as soon as + * possible (e.g., to sample the instruction address and stack chain), + * this should return true. The IRQ masking code can then enable MSR[EE] + * in some places (e.g., interrupt handlers) that allows PMI interrupts + * through to improve accuracy of profiles, at the cost of some performance. + * + * The PMU counters can be enabled by other means (e.g., sysfs raw SPR + * access), but in that case there is no need for prompt PMI handling. + * + * This currently returns true if any perf counter is being used. It + * could possibly return false if only events are being counted rather than + * samples being taken, but for now this is good enough. + */ +bool power_pmu_wants_prompt_pmi(void) +{ + struct cpu_hw_events *cpuhw; + + /* + * This could simply test local_paca->pmcregs_in_use if that were not + * under ifdef KVM. + */ + if (!ppmu) + return false; + + cpuhw = this_cpu_ptr(&cpu_hw_events); + return cpuhw->n_events; +} #endif /* CONFIG_PPC64 */ static void perf_event_interrupt(struct pt_regs *regs); @@ -799,6 +886,19 @@ static void write_pmc(int idx, unsigned long val) } } +static int any_pmc_overflown(struct cpu_hw_events *cpuhw) +{ + int i, idx; + + for (i = 0; i < cpuhw->n_events; i++) { + idx = cpuhw->event[i]->hw.idx; + if ((idx) && ((int)read_pmc(idx) < 0)) + return idx; + } + + return 0; +} + /* Called from sysrq_handle_showregs() */ void perf_event_print_debug(void) { @@ -848,6 +948,11 @@ void perf_event_print_debug(void) pr_info("EBBRR: %016lx BESCR: %016lx\n", mfspr(SPRN_EBBRR), mfspr(SPRN_BESCR)); } + + if (ppmu->flags & PPMU_ARCH_31) { + pr_info("MMCR3: %016lx SIER2: %016lx SIER3: %016lx\n", + mfspr(SPRN_MMCR3), mfspr(SPRN_SIER2), mfspr(SPRN_SIER3)); + } #endif pr_info("SIAR: %016lx SDAR: %016lx SIER: %016lx\n", mfspr(SPRN_SIAR), sdar, sier); @@ -863,7 +968,7 @@ void perf_event_print_debug(void) */ static int power_check_constraints(struct cpu_hw_events *cpuhw, u64 event_id[], unsigned int cflags[], - int n_ev) + int n_ev, struct perf_event **event) { unsigned long mask, value, nv; unsigned long smasks[MAX_HWEVENTS], svalues[MAX_HWEVENTS]; @@ -886,7 +991,7 @@ static int power_check_constraints(struct cpu_hw_events *cpuhw, event_id[i] = cpuhw->alternatives[i][0]; } if (ppmu->get_constraint(event_id[i], &cpuhw->amasks[i][0], - &cpuhw->avalues[i][0])) + &cpuhw->avalues[i][0], event[i]->attr.config1)) return -1; } value = mask = 0; @@ -921,7 +1026,8 @@ static int power_check_constraints(struct cpu_hw_events *cpuhw, for (j = 1; j < n_alt[i]; ++j) ppmu->get_constraint(cpuhw->alternatives[i][j], &cpuhw->amasks[i][j], - &cpuhw->avalues[i][j]); + &cpuhw->avalues[i][j], + event[i]->attr.config1); } /* enumerate all possibilities and see if any will work */ @@ -1036,7 +1142,7 @@ static u64 check_and_compute_delta(u64 prev, u64 val) /* * POWER7 can roll back counter values, if the new value is smaller * than the previous value it will cause the delta and the counter to - * have bogus values unless we rolled a counter over. If a coutner is + * have bogus values unless we rolled a counter over. If a counter is * rolled back, it will be smaller, but within 256, which is the maximum * number of events to rollback at once. If we detect a rollback * return 0. This can lead to a small lack of precision in the @@ -1199,7 +1305,7 @@ static void write_mmcr0(struct cpu_hw_events *cpuhw, unsigned long mmcr0) static void power_pmu_disable(struct pmu *pmu) { struct cpu_hw_events *cpuhw; - unsigned long flags, mmcr0, val; + unsigned long flags, mmcr0, val, mmcra; if (!ppmu) return; @@ -1217,11 +1323,16 @@ static void power_pmu_disable(struct pmu *pmu) /* * Set the 'freeze counters' bit, clear EBE/BHRBA/PMCC/PMAO/FC56 + * Also clear PMXE to disable PMI's getting triggered in some + * corner cases during PMU disable. */ val = mmcr0 = mfspr(SPRN_MMCR0); val |= MMCR0_FC; val &= ~(MMCR0_EBE | MMCR0_BHRBA | MMCR0_PMCC | MMCR0_PMAO | - MMCR0_FC56); + MMCR0_PMXE | MMCR0_FC56); + /* Set mmcr0 PMCCEXT for p10 */ + if (ppmu->flags & PPMU_ARCH_31) + val |= MMCR0_PMCCEXT; /* * The barrier is to make sure the mtspr has been @@ -1233,11 +1344,45 @@ static void power_pmu_disable(struct pmu *pmu) isync(); /* + * Some corner cases could clear the PMU counter overflow + * while a masked PMI is pending. One such case is when + * a PMI happens during interrupt replay and perf counter + * values are cleared by PMU callbacks before replay. + * + * Disable the interrupt by clearing the paca bit for PMI + * since we are disabling the PMU now. Otherwise provide a + * warning if there is PMI pending, but no counter is found + * overflown. + * + * Since power_pmu_disable runs under local_irq_save, it + * could happen that code hits a PMC overflow without PMI + * pending in paca. Hence only clear PMI pending if it was + * set. + * + * If a PMI is pending, then MSR[EE] must be disabled (because + * the masked PMI handler disabling EE). So it is safe to + * call clear_pmi_irq_pending(). + */ + if (pmi_irq_pending()) + clear_pmi_irq_pending(); + + val = mmcra = cpuhw->mmcr.mmcra; + + /* * Disable instruction sampling if it was enabled */ - if (cpuhw->mmcr[2] & MMCRA_SAMPLE_ENABLE) { - mtspr(SPRN_MMCRA, - cpuhw->mmcr[2] & ~MMCRA_SAMPLE_ENABLE); + val &= ~MMCRA_SAMPLE_ENABLE; + + /* Disable BHRB via mmcra (BHRBRD) for p10 */ + if (ppmu->flags & PPMU_ARCH_31) + val |= MMCRA_BHRB_DISABLE; + + /* + * Write SPRN_MMCRA if mmcra has either disabled + * instruction sampling or BHRB. + */ + if (val != mmcra) { + mtspr(SPRN_MMCRA, val); mb(); isync(); } @@ -1311,18 +1456,29 @@ static void power_pmu_enable(struct pmu *pmu) * (possibly updated for removal of events). */ if (!cpuhw->n_added) { - mtspr(SPRN_MMCRA, cpuhw->mmcr[2] & ~MMCRA_SAMPLE_ENABLE); - mtspr(SPRN_MMCR1, cpuhw->mmcr[1]); + /* + * If there is any active event with an overflown PMC + * value, set back PACA_IRQ_PMI which would have been + * cleared in power_pmu_disable(). + */ + hard_irq_disable(); + if (any_pmc_overflown(cpuhw)) + set_pmi_irq_pending(); + + mtspr(SPRN_MMCRA, cpuhw->mmcr.mmcra & ~MMCRA_SAMPLE_ENABLE); + mtspr(SPRN_MMCR1, cpuhw->mmcr.mmcr1); + if (ppmu->flags & PPMU_ARCH_31) + mtspr(SPRN_MMCR3, cpuhw->mmcr.mmcr3); goto out_enable; } /* * Clear all MMCR settings and recompute them for the new set of events. */ - memset(cpuhw->mmcr, 0, sizeof(cpuhw->mmcr)); + memset(&cpuhw->mmcr, 0, sizeof(cpuhw->mmcr)); if (ppmu->compute_mmcr(cpuhw->events, cpuhw->n_events, hwc_index, - cpuhw->mmcr, cpuhw->event)) { + &cpuhw->mmcr, cpuhw->event, ppmu->flags)) { /* shouldn't ever get here */ printk(KERN_ERR "oops compute_mmcr failed\n"); goto out; @@ -1336,11 +1492,11 @@ static void power_pmu_enable(struct pmu *pmu) */ event = cpuhw->event[0]; if (event->attr.exclude_user) - cpuhw->mmcr[0] |= MMCR0_FCP; + cpuhw->mmcr.mmcr0 |= MMCR0_FCP; if (event->attr.exclude_kernel) - cpuhw->mmcr[0] |= freeze_events_kernel; + cpuhw->mmcr.mmcr0 |= freeze_events_kernel; if (event->attr.exclude_hv) - cpuhw->mmcr[0] |= MMCR0_FCHV; + cpuhw->mmcr.mmcr0 |= MMCR0_FCHV; } /* @@ -1349,12 +1505,15 @@ static void power_pmu_enable(struct pmu *pmu) * Then unfreeze the events. */ ppc_set_pmu_inuse(1); - mtspr(SPRN_MMCRA, cpuhw->mmcr[2] & ~MMCRA_SAMPLE_ENABLE); - mtspr(SPRN_MMCR1, cpuhw->mmcr[1]); - mtspr(SPRN_MMCR0, (cpuhw->mmcr[0] & ~(MMCR0_PMC1CE | MMCR0_PMCjCE)) + mtspr(SPRN_MMCRA, cpuhw->mmcr.mmcra & ~MMCRA_SAMPLE_ENABLE); + mtspr(SPRN_MMCR1, cpuhw->mmcr.mmcr1); + mtspr(SPRN_MMCR0, (cpuhw->mmcr.mmcr0 & ~(MMCR0_PMC1CE | MMCR0_PMCjCE)) | MMCR0_FC); if (ppmu->flags & PPMU_ARCH_207S) - mtspr(SPRN_MMCR2, cpuhw->mmcr[3]); + mtspr(SPRN_MMCR2, cpuhw->mmcr.mmcr2); + + if (ppmu->flags & PPMU_ARCH_31) + mtspr(SPRN_MMCR3, cpuhw->mmcr.mmcr3); /* * Read off any pre-existing events that need to move @@ -1405,7 +1564,7 @@ static void power_pmu_enable(struct pmu *pmu) perf_event_update_userpage(event); } cpuhw->n_limited = n_lim; - cpuhw->mmcr[0] |= MMCR0_PMXE | MMCR0_FCECE; + cpuhw->mmcr.mmcr0 |= MMCR0_PMXE | MMCR0_FCECE; out_enable: pmao_restore_workaround(ebb); @@ -1421,9 +1580,9 @@ static void power_pmu_enable(struct pmu *pmu) /* * Enable instruction sampling if necessary */ - if (cpuhw->mmcr[2] & MMCRA_SAMPLE_ENABLE) { + if (cpuhw->mmcr.mmcra & MMCRA_SAMPLE_ENABLE) { mb(); - mtspr(SPRN_MMCRA, cpuhw->mmcr[2]); + mtspr(SPRN_MMCRA, cpuhw->mmcr.mmcra); } out: @@ -1507,7 +1666,7 @@ static int power_pmu_add(struct perf_event *event, int ef_flags) if (check_excludes(cpuhw->event, cpuhw->flags, n0, 1)) goto out; - if (power_check_constraints(cpuhw, cpuhw->events, cpuhw->flags, n0 + 1)) + if (power_check_constraints(cpuhw, cpuhw->events, cpuhw->flags, n0 + 1, cpuhw->event)) goto out; event->hw.config = cpuhw->events[n0]; @@ -1520,9 +1679,16 @@ nocheck: ret = 0; out: if (has_branch_stack(event)) { - power_pmu_bhrb_enable(event); - cpuhw->bhrb_filter = ppmu->bhrb_filter_map( - event->attr.branch_sample_type); + u64 bhrb_filter = -1; + + if (ppmu->bhrb_filter_map) + bhrb_filter = ppmu->bhrb_filter_map( + event->attr.branch_sample_type); + + if (bhrb_filter != -1) { + cpuhw->bhrb_filter = bhrb_filter; + power_pmu_bhrb_enable(event); + } } perf_pmu_enable(event->pmu); @@ -1553,7 +1719,7 @@ static void power_pmu_del(struct perf_event *event, int ef_flags) cpuhw->flags[i-1] = cpuhw->flags[i]; } --cpuhw->n_events; - ppmu->disable_pmc(event->hw.idx - 1, cpuhw->mmcr); + ppmu->disable_pmc(event->hw.idx - 1, &cpuhw->mmcr); if (event->hw.idx) { write_pmc(event->hw.idx, 0); event->hw.idx = 0; @@ -1574,7 +1740,7 @@ static void power_pmu_del(struct perf_event *event, int ef_flags) } if (cpuhw->n_events == 0) { /* disable exceptions if no events are running */ - cpuhw->mmcr[0] &= ~(MMCR0_PMXE | MMCR0_FCECE); + cpuhw->mmcr.mmcr0 &= ~(MMCR0_PMXE | MMCR0_FCECE); } if (has_branch_stack(event)) @@ -1710,7 +1876,7 @@ static int power_pmu_commit_txn(struct pmu *pmu) n = cpuhw->n_events; if (check_excludes(cpuhw->event, cpuhw->flags, 0, n)) return -EAGAIN; - i = power_check_constraints(cpuhw, cpuhw->events, cpuhw->flags, n); + i = power_check_constraints(cpuhw, cpuhw->events, cpuhw->flags, n, cpuhw->event); if (i < 0) return -EAGAIN; @@ -1798,7 +1964,7 @@ static void hw_perf_event_destroy(struct perf_event *event) static int hw_perf_cache_event(u64 config, u64 *eventp) { unsigned long type, op, result; - int ev; + u64 ev; if (!ppmu->cache_events) return -EINVAL; @@ -1837,14 +2003,13 @@ static bool is_event_blacklisted(u64 ev) static int power_pmu_event_init(struct perf_event *event) { u64 ev; - unsigned long flags; + unsigned long flags, irq_flags; struct perf_event *ctrs[MAX_HWEVENTS]; u64 events[MAX_HWEVENTS]; unsigned int cflags[MAX_HWEVENTS]; int n; int err; struct cpu_hw_events *cpuhw; - u64 bhrb_filter; if (!ppmu) return -ENOENT; @@ -1883,6 +2048,17 @@ static int power_pmu_event_init(struct perf_event *event) return -ENOENT; } + /* + * PMU config registers have fields that are + * reserved and some specific values for bit fields are reserved. + * For ex., MMCRA[61:62] is Random Sampling Mode (SM) + * and value of 0b11 to this field is reserved. + * Check for invalid values in attr.config. + */ + if (ppmu->check_attr_config && + ppmu->check_attr_config(event)) + return -EINVAL; + event->hw.config_base = ev; event->hw.idx = 0; @@ -1946,21 +2122,43 @@ static int power_pmu_event_init(struct perf_event *event) if (check_excludes(ctrs, cflags, n, 1)) return -EINVAL; - cpuhw = &get_cpu_var(cpu_hw_events); - err = power_check_constraints(cpuhw, events, cflags, n + 1); + local_irq_save(irq_flags); + cpuhw = this_cpu_ptr(&cpu_hw_events); + + err = power_check_constraints(cpuhw, events, cflags, n + 1, ctrs); if (has_branch_stack(event)) { - bhrb_filter = ppmu->bhrb_filter_map( + u64 bhrb_filter = -1; + + /* + * Currently no PMU supports having multiple branch filters + * at the same time. Branch filters are set via MMCRA IFM[32:33] + * bits for Power8 and above. Return EOPNOTSUPP when multiple + * branch filters are requested in the event attr. + * + * When opening event via perf_event_open(), branch_sample_type + * gets adjusted in perf_copy_attr(). Kernel will automatically + * adjust the branch_sample_type based on the event modifier + * settings to include PERF_SAMPLE_BRANCH_PLM_ALL. Hence drop + * the check for PERF_SAMPLE_BRANCH_PLM_ALL. + */ + if (hweight64(event->attr.branch_sample_type & ~PERF_SAMPLE_BRANCH_PLM_ALL) > 1) { + local_irq_restore(irq_flags); + return -EOPNOTSUPP; + } + + if (ppmu->bhrb_filter_map) + bhrb_filter = ppmu->bhrb_filter_map( event->attr.branch_sample_type); if (bhrb_filter == -1) { - put_cpu_var(cpu_hw_events); + local_irq_restore(irq_flags); return -EOPNOTSUPP; } cpuhw->bhrb_filter = bhrb_filter; } - put_cpu_var(cpu_hw_events); + local_irq_restore(irq_flags); if (err) return -EINVAL; @@ -2028,6 +2226,9 @@ static struct pmu power_pmu = { .sched_task = power_pmu_sched_task, }; +#define PERF_SAMPLE_ADDR_TYPE (PERF_SAMPLE_ADDR | \ + PERF_SAMPLE_PHYS_ADDR | \ + PERF_SAMPLE_DATA_PAGE_SIZE) /* * A counter has overflowed; update its count and record * things if requested. Note that interrupts are hard-disabled @@ -2063,7 +2264,17 @@ static void record_and_restart(struct perf_event *event, unsigned long val, left += period; if (left <= 0) left = period; - record = siar_valid(regs); + + /* + * If address is not requested in the sample via + * PERF_SAMPLE_IP, just record that sample irrespective + * of SIAR valid check. + */ + if (event->attr.sample_type & PERF_SAMPLE_IP) + record = siar_valid(regs); + else + record = 1; + event->hw.last_period = event->hw.sample_period; } if (left < 0x80000000LL) @@ -2076,6 +2287,17 @@ static void record_and_restart(struct perf_event *event, unsigned long val, perf_event_update_userpage(event); /* + * Due to hardware limitation, sometimes SIAR could sample a kernel + * address even when freeze on supervisor state (kernel) is set in + * MMCR2. Check attr.exclude_kernel and address to drop the sample in + * these cases. + */ + if (event->attr.exclude_kernel && + (event->attr.sample_type & PERF_SAMPLE_IP) && + is_kernel_addr(mfspr(SPRN_SIAR))) + record = 0; + + /* * Finally record data if requested. */ if (record) { @@ -2083,27 +2305,33 @@ static void record_and_restart(struct perf_event *event, unsigned long val, perf_sample_data_init(&data, ~0ULL, event->hw.last_period); - if (event->attr.sample_type & - (PERF_SAMPLE_ADDR | PERF_SAMPLE_PHYS_ADDR)) + if (event->attr.sample_type & PERF_SAMPLE_ADDR_TYPE) perf_get_data_addr(event, regs, &data.addr); if (event->attr.sample_type & PERF_SAMPLE_BRANCH_STACK) { struct cpu_hw_events *cpuhw; cpuhw = this_cpu_ptr(&cpu_hw_events); power_pmu_bhrb_read(event, cpuhw); - data.br_stack = &cpuhw->bhrb_stack; + perf_sample_save_brstack(&data, event, &cpuhw->bhrb_stack, NULL); } if (event->attr.sample_type & PERF_SAMPLE_DATA_SRC && - ppmu->get_mem_data_src) + ppmu->get_mem_data_src) { ppmu->get_mem_data_src(&data.data_src, ppmu->flags, regs); + data.sample_flags |= PERF_SAMPLE_DATA_SRC; + } - if (event->attr.sample_type & PERF_SAMPLE_WEIGHT && - ppmu->get_mem_weight) - ppmu->get_mem_weight(&data.weight); - + if (event->attr.sample_type & PERF_SAMPLE_WEIGHT_TYPE && + ppmu->get_mem_weight) { + ppmu->get_mem_weight(&data.weight.full, event->attr.sample_type); + data.sample_flags |= PERF_SAMPLE_WEIGHT_TYPE; + } if (perf_event_overflow(event, &data, regs)) power_pmu_stop(event, 0); + } else if (period) { + /* Account for interrupt in case of invalid SIAR */ + if (perf_event_account_interrupt(event)) + power_pmu_stop(event, 0); } } @@ -2127,12 +2355,10 @@ unsigned long perf_misc_flags(struct pt_regs *regs) */ unsigned long perf_instruction_pointer(struct pt_regs *regs) { - bool use_siar = regs_use_siar(regs); + unsigned long siar = mfspr(SPRN_SIAR); - if (use_siar && siar_valid(regs)) - return mfspr(SPRN_SIAR) + perf_ip_adjust(regs); - else if (use_siar) - return 0; // no valid instruction pointer + if (regs_use_siar(regs) && siar_valid(regs) && siar) + return siar + perf_ip_adjust(regs); else return regs->nip; } @@ -2172,9 +2398,7 @@ static void __perf_event_interrupt(struct pt_regs *regs) int i, j; struct cpu_hw_events *cpuhw = this_cpu_ptr(&cpu_hw_events); struct perf_event *event; - unsigned long val[8]; int found, active; - int nmi; if (cpuhw->n_limited) freeze_limited_counters(cpuhw, mfspr(SPRN_PMC5), @@ -2182,20 +2406,14 @@ static void __perf_event_interrupt(struct pt_regs *regs) perf_read_regs(regs); - nmi = perf_intr_is_nmi(regs); - if (nmi) - nmi_enter(); - else - irq_enter(); - /* Read all the PMCs since we'll need them a bunch of times */ for (i = 0; i < ppmu->n_counter; ++i) - val[i] = read_pmc(i + 1); + cpuhw->pmcs[i] = read_pmc(i + 1); /* Try to find what caused the IRQ */ found = 0; for (i = 0; i < ppmu->n_counter; ++i) { - if (!pmc_overflow(val[i])) + if (!pmc_overflow(cpuhw->pmcs[i])) continue; if (is_limited_pmc(i + 1)) continue; /* these won't generate IRQs */ @@ -2210,10 +2428,18 @@ static void __perf_event_interrupt(struct pt_regs *regs) event = cpuhw->event[j]; if (event->hw.idx == (i + 1)) { active = 1; - record_and_restart(event, val[i], regs); + record_and_restart(event, cpuhw->pmcs[i], regs); break; } } + + /* + * Clear PACA_IRQ_PMI in case it was set by + * set_pmi_irq_pending() when PMU was enabled + * after accounting for interrupts. + */ + clear_pmi_irq_pending(); + if (!active) /* reset non active counters that have overflowed */ write_pmc(i + 1, 0); @@ -2224,17 +2450,24 @@ static void __perf_event_interrupt(struct pt_regs *regs) event = cpuhw->event[i]; if (!event->hw.idx || is_limited_pmc(event->hw.idx)) continue; - if (pmc_overflow_power7(val[event->hw.idx - 1])) { + if (pmc_overflow_power7(cpuhw->pmcs[event->hw.idx - 1])) { /* event has overflowed in a buggy way*/ found = 1; record_and_restart(event, - val[event->hw.idx - 1], + cpuhw->pmcs[event->hw.idx - 1], regs); } } } - if (!found && !nmi && printk_ratelimit()) - printk(KERN_WARNING "Can't find PMC that caused IRQ\n"); + + /* + * During system wide profiling or while specific CPU is monitored for an + * event, some corner cases could cause PMC to overflow in idle path. This + * will trigger a PMI after waking up from idle. Since counter values are _not_ + * saved/restored in idle path, can lead to below "Can't find PMC" message. + */ + if (unlikely(!found) && !arch_irq_disabled_regs(regs)) + printk_ratelimited(KERN_WARNING "Can't find PMC that caused IRQ\n"); /* * Reset MMCR0 to its normal value. This will set PMXE and @@ -2243,12 +2476,11 @@ static void __perf_event_interrupt(struct pt_regs *regs) * XXX might want to use MSR.PM to keep the events frozen until * we get back out of this interrupt. */ - write_mmcr0(cpuhw, cpuhw->mmcr[0]); + write_mmcr0(cpuhw, cpuhw->mmcr.mmcr0); + + /* Clear the cpuhw->pmcs */ + memset(&cpuhw->pmcs, 0, sizeof(cpuhw->pmcs)); - if (nmi) - nmi_exit(); - else - irq_exit(); } static void perf_event_interrupt(struct pt_regs *regs) @@ -2265,12 +2497,39 @@ static int power_pmu_prepare_cpu(unsigned int cpu) if (ppmu) { memset(cpuhw, 0, sizeof(*cpuhw)); - cpuhw->mmcr[0] = MMCR0_FC; + cpuhw->mmcr.mmcr0 = MMCR0_FC; } return 0; } -int register_power_pmu(struct power_pmu *pmu) +static ssize_t pmu_name_show(struct device *cdev, + struct device_attribute *attr, + char *buf) +{ + if (ppmu) + return sysfs_emit(buf, "%s", ppmu->name); + + return 0; +} + +static DEVICE_ATTR_RO(pmu_name); + +static struct attribute *pmu_caps_attrs[] = { + &dev_attr_pmu_name.attr, + NULL +}; + +static const struct attribute_group pmu_caps_group = { + .name = "caps", + .attrs = pmu_caps_attrs, +}; + +static const struct attribute_group *pmu_caps_groups[] = { + &pmu_caps_group, + NULL, +}; + +int __init register_power_pmu(struct power_pmu *pmu) { if (ppmu) return -EBUSY; /* something's already registered */ @@ -2281,6 +2540,11 @@ int register_power_pmu(struct power_pmu *pmu) power_pmu.attr_groups = ppmu->attr_groups; + if (ppmu->flags & PPMU_ARCH_207S) + power_pmu.attr_update = pmu_caps_groups; + + power_pmu.capabilities |= (ppmu->capabilities & PERF_PMU_CAP_EXTENDED_REGS); + #ifdef MSR_HV /* * Use FCHV to ignore kernel events if MSR.HV is set. @@ -2296,8 +2560,24 @@ int register_power_pmu(struct power_pmu *pmu) } #ifdef CONFIG_PPC64 +static bool pmu_override = false; +static unsigned long pmu_override_val; +static void do_pmu_override(void *data) +{ + ppc_set_pmu_inuse(1); + if (pmu_override_val) + mtspr(SPRN_MMCR1, pmu_override_val); + mtspr(SPRN_MMCR0, mfspr(SPRN_MMCR0) & ~MMCR0_FC); +} + static int __init init_ppc64_pmu(void) { + if (cpu_has_feature(CPU_FTR_HVMODE) && pmu_override) { + pr_warn("disabling perf due to pmu_override= command line option.\n"); + on_each_cpu(do_pmu_override, NULL, 1); + return 0; + } + /* run through all the pmu drivers one at a time */ if (!init_power5_pmu()) return 0; @@ -2311,10 +2591,33 @@ static int __init init_ppc64_pmu(void) return 0; else if (!init_power9_pmu()) return 0; + else if (!init_power10_pmu()) + return 0; + else if (!init_power11_pmu()) + return 0; else if (!init_ppc970_pmu()) return 0; else return init_generic_compat_pmu(); } early_initcall(init_ppc64_pmu); + +static int __init pmu_setup(char *str) +{ + unsigned long val; + + if (!early_cpu_has_feature(CPU_FTR_HVMODE)) + return 0; + + pmu_override = true; + + if (kstrtoul(str, 0, &val)) + val = 0; + + pmu_override_val = val; + + return 1; +} +__setup("pmu_override=", pmu_setup); + #endif diff --git a/arch/powerpc/perf/core-fsl-emb.c b/arch/powerpc/perf/core-fsl-emb.c index e0e7e276bfd2..1a53ab08447c 100644 --- a/arch/powerpc/perf/core-fsl-emb.c +++ b/arch/powerpc/perf/core-fsl-emb.c @@ -31,19 +31,6 @@ static atomic_t num_events; /* Used to avoid races in calling reserve/release_pmc_hardware */ static DEFINE_MUTEX(pmc_reserve_mutex); -/* - * If interrupts were soft-disabled when a PMU interrupt occurs, treat - * it as an NMI. - */ -static inline int perf_intr_is_nmi(struct pt_regs *regs) -{ -#ifdef __powerpc64__ - return (regs->softe & IRQS_DISABLED); -#else - return 0; -#endif -} - static void perf_event_interrupt(struct pt_regs *regs); /* @@ -658,14 +645,6 @@ static void perf_event_interrupt(struct pt_regs *regs) struct cpu_hw_events *cpuhw = this_cpu_ptr(&cpu_hw_events); struct perf_event *event; unsigned long val; - int found = 0; - int nmi; - - nmi = perf_intr_is_nmi(regs); - if (nmi) - nmi_enter(); - else - irq_enter(); for (i = 0; i < ppmu->n_counter; ++i) { event = cpuhw->event[i]; @@ -674,7 +653,6 @@ static void perf_event_interrupt(struct pt_regs *regs) if ((int)val < 0) { if (event) { /* event has overflowed */ - found = 1; record_and_restart(event, val, regs); } else { /* @@ -690,18 +668,15 @@ static void perf_event_interrupt(struct pt_regs *regs) mtmsr(mfmsr() | MSR_PMM); mtpmr(PMRN_PMGC0, PMGC0_PMIE | PMGC0_FCECE); isync(); - - if (nmi) - nmi_exit(); - else - irq_exit(); } -void hw_perf_event_setup(int cpu) +static int fsl_emb_pmu_prepare_cpu(unsigned int cpu) { struct cpu_hw_events *cpuhw = &per_cpu(cpu_hw_events, cpu); memset(cpuhw, 0, sizeof(*cpuhw)); + + return 0; } int register_fsl_emb_pmu(struct fsl_emb_pmu *pmu) @@ -714,6 +689,8 @@ int register_fsl_emb_pmu(struct fsl_emb_pmu *pmu) pmu->name); perf_pmu_register(&fsl_emb_pmu, "cpu", PERF_TYPE_RAW); + cpuhp_setup_state(CPUHP_PERF_POWER, "perf/powerpc:prepare", + fsl_emb_pmu_prepare_cpu, NULL); return 0; } diff --git a/arch/powerpc/perf/e500-pmu.c b/arch/powerpc/perf/e500-pmu.c index a59c33bed32a..e3e1a68eb1d5 100644 --- a/arch/powerpc/perf/e500-pmu.c +++ b/arch/powerpc/perf/e500-pmu.c @@ -118,12 +118,13 @@ static struct fsl_emb_pmu e500_pmu = { static int init_e500_pmu(void) { - if (!cur_cpu_spec->oprofile_cpu_type) - return -ENODEV; + unsigned int pvr = mfspr(SPRN_PVR); - if (!strcmp(cur_cpu_spec->oprofile_cpu_type, "ppc/e500mc")) + /* ec500mc */ + if (PVR_VER(pvr) == PVR_VER_E500MC || PVR_VER(pvr) == PVR_VER_E5500) num_events = 256; - else if (strcmp(cur_cpu_spec->oprofile_cpu_type, "ppc/e500")) + /* e500 */ + else if (PVR_VER(pvr) != PVR_VER_E500V1 && PVR_VER(pvr) != PVR_VER_E500V2) return -ENODEV; return register_fsl_emb_pmu(&e500_pmu); diff --git a/arch/powerpc/perf/e6500-pmu.c b/arch/powerpc/perf/e6500-pmu.c index 44ad65da82ed..bd779a2338f8 100644 --- a/arch/powerpc/perf/e6500-pmu.c +++ b/arch/powerpc/perf/e6500-pmu.c @@ -107,8 +107,9 @@ static struct fsl_emb_pmu e6500_pmu = { static int init_e6500_pmu(void) { - if (!cur_cpu_spec->oprofile_cpu_type || - strcmp(cur_cpu_spec->oprofile_cpu_type, "ppc/e6500")) + unsigned int pvr = mfspr(SPRN_PVR); + + if (PVR_VER(pvr) != PVR_VER_E6500) return -ENODEV; return register_fsl_emb_pmu(&e6500_pmu); diff --git a/arch/powerpc/perf/generic-compat-pmu.c b/arch/powerpc/perf/generic-compat-pmu.c index 5e5a54d5588e..b5c414876ed5 100644 --- a/arch/powerpc/perf/generic-compat-pmu.c +++ b/arch/powerpc/perf/generic-compat-pmu.c @@ -14,84 +14,165 @@ * * 28 24 20 16 12 8 4 0 * | - - - - | - - - - | - - - - | - - - - | - - - - | - - - - | - - - - | - - - - | - * [ pmc ] [unit ] [ ] m [ pmcxsel ] - * | | - * | *- mark - * | - * | - * *- combine - * - * Below uses IBM bit numbering. - * - * MMCR1[x:y] = unit (PMCxUNIT) - * MMCR1[24] = pmc1combine[0] - * MMCR1[25] = pmc1combine[1] - * MMCR1[26] = pmc2combine[0] - * MMCR1[27] = pmc2combine[1] - * MMCR1[28] = pmc3combine[0] - * MMCR1[29] = pmc3combine[1] - * MMCR1[30] = pmc4combine[0] - * MMCR1[31] = pmc4combine[1] - * + * [ pmc ] [ pmcxsel ] */ /* - * Some power9 event codes. + * Event codes defined in ISA v3.0B */ #define EVENT(_name, _code) _name = _code, enum { -EVENT(PM_CYC, 0x0001e) -EVENT(PM_INST_CMPL, 0x00002) + /* Cycles, alternate code */ + EVENT(PM_CYC_ALT, 0x100f0) + /* One or more instructions completed in a cycle */ + EVENT(PM_CYC_INST_CMPL, 0x100f2) + /* Floating-point instruction completed */ + EVENT(PM_FLOP_CMPL, 0x100f4) + /* Instruction ERAT/L1-TLB miss */ + EVENT(PM_L1_ITLB_MISS, 0x100f6) + /* All instructions completed and none available */ + EVENT(PM_NO_INST_AVAIL, 0x100f8) + /* A load-type instruction completed (ISA v3.0+) */ + EVENT(PM_LD_CMPL, 0x100fc) + /* Instruction completed, alternate code (ISA v3.0+) */ + EVENT(PM_INST_CMPL_ALT, 0x100fe) + /* A store-type instruction completed */ + EVENT(PM_ST_CMPL, 0x200f0) + /* Instruction Dispatched */ + EVENT(PM_INST_DISP, 0x200f2) + /* Run_cycles */ + EVENT(PM_RUN_CYC, 0x200f4) + /* Data ERAT/L1-TLB miss/reload */ + EVENT(PM_L1_DTLB_RELOAD, 0x200f6) + /* Taken branch completed */ + EVENT(PM_BR_TAKEN_CMPL, 0x200fa) + /* Demand iCache Miss */ + EVENT(PM_L1_ICACHE_MISS, 0x200fc) + /* L1 Dcache reload from memory */ + EVENT(PM_L1_RELOAD_FROM_MEM, 0x200fe) + /* L1 Dcache store miss */ + EVENT(PM_ST_MISS_L1, 0x300f0) + /* Alternate code for PM_INST_DISP */ + EVENT(PM_INST_DISP_ALT, 0x300f2) + /* Branch direction or target mispredicted */ + EVENT(PM_BR_MISPREDICT, 0x300f6) + /* Data TLB miss/reload */ + EVENT(PM_DTLB_MISS, 0x300fc) + /* Demand LD - L3 Miss (not L2 hit and not L3 hit) */ + EVENT(PM_DATA_FROM_L3MISS, 0x300fe) + /* L1 Dcache load miss */ + EVENT(PM_LD_MISS_L1, 0x400f0) + /* Cycle when instruction(s) dispatched */ + EVENT(PM_CYC_INST_DISP, 0x400f2) + /* Branch or branch target mispredicted */ + EVENT(PM_BR_MPRED_CMPL, 0x400f6) + /* Instructions completed with run latch set */ + EVENT(PM_RUN_INST_CMPL, 0x400fa) + /* Instruction TLB miss/reload */ + EVENT(PM_ITLB_MISS, 0x400fc) + /* Load data not cached */ + EVENT(PM_LD_NOT_CACHED, 0x400fe) + /* Instructions */ + EVENT(PM_INST_CMPL, 0x500fa) + /* Cycles */ + EVENT(PM_CYC, 0x600f4) }; #undef EVENT +/* Table of alternatives, sorted in increasing order of column 0 */ +/* Note that in each row, column 0 must be the smallest */ +static const unsigned int generic_event_alternatives[][MAX_ALT] = { + { PM_CYC_ALT, PM_CYC }, + { PM_INST_CMPL_ALT, PM_INST_CMPL }, + { PM_INST_DISP, PM_INST_DISP_ALT }, +}; + +static int generic_get_alternatives(u64 event, unsigned int flags, u64 alt[]) +{ + int num_alt = 0; + + num_alt = isa207_get_alternatives(event, alt, + ARRAY_SIZE(generic_event_alternatives), flags, + generic_event_alternatives); + + return num_alt; +} + GENERIC_EVENT_ATTR(cpu-cycles, PM_CYC); GENERIC_EVENT_ATTR(instructions, PM_INST_CMPL); +GENERIC_EVENT_ATTR(stalled-cycles-frontend, PM_NO_INST_AVAIL); +GENERIC_EVENT_ATTR(branch-misses, PM_BR_MPRED_CMPL); +GENERIC_EVENT_ATTR(cache-misses, PM_LD_MISS_L1); + +CACHE_EVENT_ATTR(L1-dcache-load-misses, PM_LD_MISS_L1); +CACHE_EVENT_ATTR(L1-dcache-store-misses, PM_ST_MISS_L1); +CACHE_EVENT_ATTR(L1-icache-load-misses, PM_L1_ICACHE_MISS); +CACHE_EVENT_ATTR(LLC-load-misses, PM_DATA_FROM_L3MISS); +CACHE_EVENT_ATTR(branch-load-misses, PM_BR_MPRED_CMPL); +CACHE_EVENT_ATTR(dTLB-load-misses, PM_DTLB_MISS); +CACHE_EVENT_ATTR(iTLB-load-misses, PM_ITLB_MISS); static struct attribute *generic_compat_events_attr[] = { GENERIC_EVENT_PTR(PM_CYC), GENERIC_EVENT_PTR(PM_INST_CMPL), + GENERIC_EVENT_PTR(PM_NO_INST_AVAIL), + GENERIC_EVENT_PTR(PM_BR_MPRED_CMPL), + GENERIC_EVENT_PTR(PM_LD_MISS_L1), + CACHE_EVENT_PTR(PM_LD_MISS_L1), + CACHE_EVENT_PTR(PM_ST_MISS_L1), + CACHE_EVENT_PTR(PM_L1_ICACHE_MISS), + CACHE_EVENT_PTR(PM_DATA_FROM_L3MISS), + CACHE_EVENT_PTR(PM_BR_MPRED_CMPL), + CACHE_EVENT_PTR(PM_DTLB_MISS), + CACHE_EVENT_PTR(PM_ITLB_MISS), NULL }; -static struct attribute_group generic_compat_pmu_events_group = { +static const struct attribute_group generic_compat_pmu_events_group = { .name = "events", .attrs = generic_compat_events_attr, }; PMU_FORMAT_ATTR(event, "config:0-19"); PMU_FORMAT_ATTR(pmcxsel, "config:0-7"); -PMU_FORMAT_ATTR(mark, "config:8"); -PMU_FORMAT_ATTR(combine, "config:10-11"); -PMU_FORMAT_ATTR(unit, "config:12-15"); PMU_FORMAT_ATTR(pmc, "config:16-19"); static struct attribute *generic_compat_pmu_format_attr[] = { &format_attr_event.attr, &format_attr_pmcxsel.attr, - &format_attr_mark.attr, - &format_attr_combine.attr, - &format_attr_unit.attr, &format_attr_pmc.attr, NULL, }; -static struct attribute_group generic_compat_pmu_format_group = { +static const struct attribute_group generic_compat_pmu_format_group = { .name = "format", .attrs = generic_compat_pmu_format_attr, }; +static struct attribute *generic_compat_pmu_caps_attrs[] = { + NULL +}; + +static struct attribute_group generic_compat_pmu_caps_group = { + .name = "caps", + .attrs = generic_compat_pmu_caps_attrs, +}; + static const struct attribute_group *generic_compat_pmu_attr_groups[] = { &generic_compat_pmu_format_group, &generic_compat_pmu_events_group, + &generic_compat_pmu_caps_group, NULL, }; static int compat_generic_events[] = { [PERF_COUNT_HW_CPU_CYCLES] = PM_CYC, [PERF_COUNT_HW_INSTRUCTIONS] = PM_INST_CMPL, + [PERF_COUNT_HW_STALLED_CYCLES_FRONTEND] = PM_NO_INST_AVAIL, + [PERF_COUNT_HW_BRANCH_MISSES] = PM_BR_MPRED_CMPL, + [PERF_COUNT_HW_CACHE_MISSES] = PM_LD_MISS_L1, }; #define C(x) PERF_COUNT_HW_CACHE_##x @@ -101,15 +182,15 @@ static int compat_generic_events[] = { * 0 means not supported, -1 means nonsensical, other values * are event codes. */ -static int generic_compat_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { +static u64 generic_compat_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { [ C(L1D) ] = { [ C(OP_READ) ] = { [ C(RESULT_ACCESS) ] = 0, - [ C(RESULT_MISS) ] = 0, + [ C(RESULT_MISS) ] = PM_LD_MISS_L1, }, [ C(OP_WRITE) ] = { [ C(RESULT_ACCESS) ] = 0, - [ C(RESULT_MISS) ] = 0, + [ C(RESULT_MISS) ] = PM_ST_MISS_L1, }, [ C(OP_PREFETCH) ] = { [ C(RESULT_ACCESS) ] = 0, @@ -119,7 +200,7 @@ static int generic_compat_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { [ C(L1I) ] = { [ C(OP_READ) ] = { [ C(RESULT_ACCESS) ] = 0, - [ C(RESULT_MISS) ] = 0, + [ C(RESULT_MISS) ] = PM_L1_ICACHE_MISS, }, [ C(OP_WRITE) ] = { [ C(RESULT_ACCESS) ] = 0, @@ -133,7 +214,7 @@ static int generic_compat_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { [ C(LL) ] = { [ C(OP_READ) ] = { [ C(RESULT_ACCESS) ] = 0, - [ C(RESULT_MISS) ] = 0, + [ C(RESULT_MISS) ] = PM_DATA_FROM_L3MISS, }, [ C(OP_WRITE) ] = { [ C(RESULT_ACCESS) ] = 0, @@ -147,7 +228,7 @@ static int generic_compat_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { [ C(DTLB) ] = { [ C(OP_READ) ] = { [ C(RESULT_ACCESS) ] = 0, - [ C(RESULT_MISS) ] = 0, + [ C(RESULT_MISS) ] = PM_DTLB_MISS, }, [ C(OP_WRITE) ] = { [ C(RESULT_ACCESS) ] = -1, @@ -161,7 +242,7 @@ static int generic_compat_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { [ C(ITLB) ] = { [ C(OP_READ) ] = { [ C(RESULT_ACCESS) ] = 0, - [ C(RESULT_MISS) ] = 0, + [ C(RESULT_MISS) ] = PM_ITLB_MISS, }, [ C(OP_WRITE) ] = { [ C(RESULT_ACCESS) ] = -1, @@ -175,7 +256,7 @@ static int generic_compat_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { [ C(BPU) ] = { [ C(OP_READ) ] = { [ C(RESULT_ACCESS) ] = 0, - [ C(RESULT_MISS) ] = 0, + [ C(RESULT_MISS) ] = PM_BR_MPRED_CMPL, }, [ C(OP_WRITE) ] = { [ C(RESULT_ACCESS) ] = -1, @@ -204,13 +285,30 @@ static int generic_compat_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { #undef C +/* + * We set MMCR0[CC5-6RUN] so we can use counters 5 and 6 for + * PM_INST_CMPL and PM_CYC. + */ +static int generic_compute_mmcr(u64 event[], int n_ev, + unsigned int hwc[], struct mmcr_regs *mmcr, + struct perf_event *pevents[], u32 flags) +{ + int ret; + + ret = isa207_compute_mmcr(event, n_ev, hwc, mmcr, pevents, flags); + if (!ret) + mmcr->mmcr0 |= MMCR0_C56RUN; + return ret; +} + static struct power_pmu generic_compat_pmu = { - .name = "GENERIC_COMPAT", + .name = "ISAv3", .n_counter = MAX_PMU_COUNTERS, .add_fields = ISA207_ADD_FIELDS, .test_adder = ISA207_TEST_ADDER, - .compute_mmcr = isa207_compute_mmcr, + .compute_mmcr = generic_compute_mmcr, .get_constraint = isa207_get_constraint, + .get_alternatives = generic_get_alternatives, .disable_pmc = isa207_disable_pmc, .flags = PPMU_HAS_SIER | PPMU_ARCH_207S, .n_generic = ARRAY_SIZE(compat_generic_events), @@ -219,10 +317,20 @@ static struct power_pmu generic_compat_pmu = { .attr_groups = generic_compat_pmu_attr_groups, }; -int init_generic_compat_pmu(void) +int __init init_generic_compat_pmu(void) { int rc = 0; + /* + * From ISA v2.07 on, PMU features are architected; + * we require >= v3.0 because (a) that has PM_LD_CMPL and + * PM_INST_CMPL_ALT, which v2.07 doesn't have, and + * (b) we don't expect any non-IBM Power ISA + * implementations that conform to v2.07 but not v3.0. + */ + if (!cpu_has_feature(CPU_FTR_ARCH_300)) + return -ENODEV; + rc = register_power_pmu(&generic_compat_pmu); if (rc) return rc; diff --git a/arch/powerpc/perf/hv-24x7.c b/arch/powerpc/perf/hv-24x7.c index 573e0b309c0c..057ec2e3451d 100644 --- a/arch/powerpc/perf/hv-24x7.c +++ b/arch/powerpc/perf/hv-24x7.c @@ -18,8 +18,10 @@ #include <asm/firmware.h> #include <asm/hvcall.h> #include <asm/io.h> +#include <asm/papr-sysparm.h> #include <linux/byteorder/generic.h> +#include <asm/rtas.h> #include "hv-24x7.h" #include "hv-24x7-catalog.h" #include "hv-common.h" @@ -30,7 +32,9 @@ static int interface_version; /* Whether we have to aggregate result data for some domains. */ static bool aggregate_result_elements; -static bool domain_is_valid(unsigned domain) +static cpumask_t hv_24x7_cpumask; + +static bool domain_is_valid(unsigned int domain) { switch (domain) { #define DOMAIN(n, v, x, c) \ @@ -44,7 +48,7 @@ static bool domain_is_valid(unsigned domain) } } -static bool is_physical_domain(unsigned domain) +static bool is_physical_domain(unsigned int domain) { switch (domain) { #define DOMAIN(n, v, x, c) \ @@ -57,6 +61,51 @@ static bool is_physical_domain(unsigned domain) } } +/* + * The Processor Module Information system parameter allows transferring + * of certain processor module information from the platform to the OS. + * Refer PAPR+ document to get parameter token value as '43'. + */ + +static u32 phys_sockets; /* Physical sockets */ +static u32 phys_chipspersocket; /* Physical chips per socket*/ +static u32 phys_coresperchip; /* Physical cores per chip */ + +/* + * read_24x7_sys_info() + * Retrieve the number of sockets and chips per socket and cores per + * chip details through the get-system-parameter rtas call. + */ +void read_24x7_sys_info(void) +{ + struct papr_sysparm_buf *buf; + + /* + * Making system parameter: chips and sockets and cores per chip + * default to 1. + */ + phys_sockets = 1; + phys_chipspersocket = 1; + phys_coresperchip = 1; + + buf = papr_sysparm_buf_alloc(); + if (!buf) + return; + + if (!papr_sysparm_get(PAPR_SYSPARM_PROC_MODULE_INFO, buf)) { + int ntypes = be16_to_cpup((__be16 *)&buf->val[0]); + int len = be16_to_cpu(buf->len); + + if (len >= 8 && ntypes != 0) { + phys_sockets = be16_to_cpup((__be16 *)&buf->val[2]); + phys_chipspersocket = be16_to_cpup((__be16 *)&buf->val[4]); + phys_coresperchip = be16_to_cpup((__be16 *)&buf->val[6]); + } + } + + papr_sysparm_buf_free(buf); +} + /* Domains for which more than one result element are returned for each event. */ static bool domain_needs_aggregation(unsigned int domain) { @@ -66,7 +115,7 @@ static bool domain_needs_aggregation(unsigned int domain) domain <= HV_PERF_DOMAIN_VCPU_REMOTE_NODE)); } -static const char *domain_name(unsigned domain) +static const char *domain_name(unsigned int domain) { if (!domain_is_valid(domain)) return NULL; @@ -84,7 +133,7 @@ static const char *domain_name(unsigned domain) return NULL; } -static bool catalog_entry_domain_is_valid(unsigned domain) +static bool catalog_entry_domain_is_valid(unsigned int domain) { /* POWER8 doesn't support virtual domains. */ if (interface_version == 1) @@ -142,7 +191,7 @@ static struct attribute *format_attrs[] = { NULL, }; -static struct attribute_group format_group = { +static const struct attribute_group format_group = { .name = "format", .attrs = format_attrs, }; @@ -164,14 +213,14 @@ static struct attribute_group event_long_desc_group = { static struct kmem_cache *hv_page_cache; -DEFINE_PER_CPU(int, hv_24x7_txn_flags); -DEFINE_PER_CPU(int, hv_24x7_txn_err); +static DEFINE_PER_CPU(int, hv_24x7_txn_flags); +static DEFINE_PER_CPU(int, hv_24x7_txn_err); struct hv_24x7_hw { struct perf_event *events[255]; }; -DEFINE_PER_CPU(struct hv_24x7_hw, hv_24x7_hw); +static DEFINE_PER_CPU(struct hv_24x7_hw, hv_24x7_hw); /* * request_buffer and result_buffer are not required to be 4k aligned, @@ -179,8 +228,8 @@ DEFINE_PER_CPU(struct hv_24x7_hw, hv_24x7_hw); * the simplest way to ensure that. */ #define H24x7_DATA_BUFFER_SIZE 4096 -DEFINE_PER_CPU(char, hv_24x7_reqb[H24x7_DATA_BUFFER_SIZE]) __aligned(4096); -DEFINE_PER_CPU(char, hv_24x7_resb[H24x7_DATA_BUFFER_SIZE]) __aligned(4096); +static DEFINE_PER_CPU(char, hv_24x7_reqb[H24x7_DATA_BUFFER_SIZE]) __aligned(4096); +static DEFINE_PER_CPU(char, hv_24x7_resb[H24x7_DATA_BUFFER_SIZE]) __aligned(4096); static unsigned int max_num_requests(int interface_version) { @@ -196,7 +245,7 @@ static char *event_name(struct hv_24x7_event_data *ev, int *len) static char *event_desc(struct hv_24x7_event_data *ev, int *len) { - unsigned nl = be16_to_cpu(ev->event_name_len); + unsigned int nl = be16_to_cpu(ev->event_name_len); __be16 *desc_len = (__be16 *)(ev->remainder + nl - 2); *len = be16_to_cpu(*desc_len) - 2; @@ -205,9 +254,9 @@ static char *event_desc(struct hv_24x7_event_data *ev, int *len) static char *event_long_desc(struct hv_24x7_event_data *ev, int *len) { - unsigned nl = be16_to_cpu(ev->event_name_len); + unsigned int nl = be16_to_cpu(ev->event_name_len); __be16 *desc_len_ = (__be16 *)(ev->remainder + nl - 2); - unsigned desc_len = be16_to_cpu(*desc_len_); + unsigned int desc_len = be16_to_cpu(*desc_len_); __be16 *long_desc_len = (__be16 *)(ev->remainder + nl + desc_len - 2); *len = be16_to_cpu(*long_desc_len) - 2; @@ -234,8 +283,8 @@ static void *event_end(struct hv_24x7_event_data *ev, void *end) { void *start = ev; __be16 *dl_, *ldl_; - unsigned dl, ldl; - unsigned nl = be16_to_cpu(ev->event_name_len); + unsigned int dl, ldl; + unsigned int nl = be16_to_cpu(ev->event_name_len); if (nl < 2) { pr_debug("%s: name length too short: %d", __func__, nl); @@ -336,7 +385,7 @@ static long h_get_24x7_catalog_page(char page[], u64 version, u32 index) * - Specifying (i.e overriding) values for other parameters * is undefined. */ -static char *event_fmt(struct hv_24x7_event_data *event, unsigned domain) +static char *event_fmt(struct hv_24x7_event_data *event, unsigned int domain) { const char *sindex; const char *lpar; @@ -386,6 +435,30 @@ static ssize_t device_show_string(struct device *dev, return sprintf(buf, "%s\n", (char *)d->var); } +static ssize_t cpumask_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + return cpumap_print_to_pagebuf(true, buf, &hv_24x7_cpumask); +} + +static ssize_t sockets_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + return sprintf(buf, "%d\n", phys_sockets); +} + +static ssize_t chipspersocket_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + return sprintf(buf, "%d\n", phys_chipspersocket); +} + +static ssize_t coresperchip_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + return sprintf(buf, "%d\n", phys_coresperchip); +} + static struct attribute *device_str_attr_create_(char *name, char *str) { struct dev_ext_attribute *attr = kzalloc(sizeof(*attr), GFP_KERNEL); @@ -443,9 +516,9 @@ out_s: return NULL; } -static struct attribute *event_to_attr(unsigned ix, +static struct attribute *event_to_attr(unsigned int ix, struct hv_24x7_event_data *event, - unsigned domain, + unsigned int domain, int nonce) { int event_name_len; @@ -513,8 +586,8 @@ event_to_long_desc_attr(struct hv_24x7_event_data *event, int nonce) return device_str_attr_create(name, nl, nonce, desc, dl); } -static int event_data_to_attrs(unsigned ix, struct attribute **attrs, - struct hv_24x7_event_data *event, int nonce) +static int event_data_to_attrs(unsigned int ix, struct attribute **attrs, + struct hv_24x7_event_data *event, int nonce) { *attrs = event_to_attr(ix, event, event->domain, nonce); if (!*attrs) @@ -528,8 +601,8 @@ struct event_uniq { struct rb_node node; const char *name; int nl; - unsigned ct; - unsigned domain; + unsigned int ct; + unsigned int domain; }; static int memord(const void *d1, size_t s1, const void *d2, size_t s2) @@ -542,8 +615,8 @@ static int memord(const void *d1, size_t s1, const void *d2, size_t s2) return memcmp(d1, d2, s1); } -static int ev_uniq_ord(const void *v1, size_t s1, unsigned d1, const void *v2, - size_t s2, unsigned d2) +static int ev_uniq_ord(const void *v1, size_t s1, unsigned int d1, + const void *v2, size_t s2, unsigned int d2) { int r = memord(v1, s1, v2, s2); @@ -557,7 +630,7 @@ static int ev_uniq_ord(const void *v1, size_t s1, unsigned d1, const void *v2, } static int event_uniq_add(struct rb_root *root, const char *name, int nl, - unsigned domain) + unsigned int domain) { struct rb_node **new = &(root->rb_node), *parent = NULL; struct event_uniq *data; @@ -670,7 +743,7 @@ static ssize_t catalog_event_len_validate(struct hv_24x7_event_data *event, } if (calc_ev_end > ev_end) { - pr_warn("event %zu exceeds it's own length: event=%pK, end=%pK, offset=%zu, calc_ev_end=%pK\n", + pr_warn("event %zu exceeds its own length: event=%pK, end=%pK, offset=%zu, calc_ev_end=%pK\n", event_idx, event, ev_end, offset, calc_ev_end); return -1; } @@ -678,6 +751,14 @@ static ssize_t catalog_event_len_validate(struct hv_24x7_event_data *event, return ev_len; } +/* + * Return true incase of invalid or dummy events with names like RESERVED* + */ +static bool ignore_event(const char *name) +{ + return strncmp(name, "RESERVED", 8) == 0; +} + #define MAX_4K (SIZE_MAX / 4096) static int create_events_from_catalog(struct attribute ***events_, @@ -808,6 +889,10 @@ static int create_events_from_catalog(struct attribute ***events_, name = event_name(event, &nl); + if (ignore_event(name)) { + junk_events++; + continue; + } if (event->event_group_record_len == 0) { pr_devel("invalid event %zu (%.*s): group_record_len == 0, skipping\n", event_idx, nl, name); @@ -869,6 +954,9 @@ static int create_events_from_catalog(struct attribute ***events_, continue; name = event_name(event, &nl); + if (ignore_event(name)) + continue; + nonce = event_uniq_add(&ev_uniq, name, nl, event->domain); ct = event_data_to_attrs(event_idx, events + event_attr_ct, event, nonce); @@ -1032,20 +1120,36 @@ PAGE_0_ATTR(catalog_len, "%lld\n", (unsigned long long)be32_to_cpu(page_0->length) * 4096); static BIN_ATTR_RO(catalog, 0/* real length varies */); static DEVICE_ATTR_RO(domains); +static DEVICE_ATTR_RO(sockets); +static DEVICE_ATTR_RO(chipspersocket); +static DEVICE_ATTR_RO(coresperchip); +static DEVICE_ATTR_RO(cpumask); static struct bin_attribute *if_bin_attrs[] = { &bin_attr_catalog, NULL, }; +static struct attribute *cpumask_attrs[] = { + &dev_attr_cpumask.attr, + NULL, +}; + +static const struct attribute_group cpumask_attr_group = { + .attrs = cpumask_attrs, +}; + static struct attribute *if_attrs[] = { &dev_attr_catalog_len.attr, &dev_attr_catalog_version.attr, &dev_attr_domains.attr, + &dev_attr_sockets.attr, + &dev_attr_chipspersocket.attr, + &dev_attr_coresperchip.attr, NULL, }; -static struct attribute_group if_group = { +static const struct attribute_group if_group = { .name = "interface", .bin_attrs = if_bin_attrs, .attrs = if_attrs, @@ -1057,6 +1161,7 @@ static const struct attribute_group *attr_groups[] = { &event_desc_group, &event_long_desc_group, &if_group, + &cpumask_attr_group, NULL, }; @@ -1233,7 +1338,7 @@ static int get_count_from_result(struct perf_event *event, for (i = count = 0, element_data = res->elements + data_offset; i < num_elements; i++, element_data += data_size + data_offset) - count += be64_to_cpu(*((u64 *) element_data)); + count += be64_to_cpu(*((__be64 *)element_data)); *countp = count; @@ -1280,7 +1385,7 @@ out: static int h_24x7_event_init(struct perf_event *event) { struct hv_perf_caps caps; - unsigned domain; + unsigned int domain; unsigned long hret; u64 ct; @@ -1313,7 +1418,7 @@ static int h_24x7_event_init(struct perf_event *event) } domain = event_get_domain(event); - if (domain >= HV_PERF_DOMAIN_MAX) { + if (domain == 0 || domain >= HV_PERF_DOMAIN_MAX) { pr_devel("invalid domain %d\n", domain); return -EINVAL; } @@ -1400,16 +1505,6 @@ static void h_24x7_event_read(struct perf_event *event) h24x7hw = &get_cpu_var(hv_24x7_hw); h24x7hw->events[i] = event; put_cpu_var(h24x7hw); - /* - * Clear the event count so we can compute the _change_ - * in the 24x7 raw counter value at the end of the txn. - * - * Note that we could alternatively read the 24x7 value - * now and save its value in event->hw.prev_count. But - * that would require issuing a hcall, which would then - * defeat the purpose of using the txn interface. - */ - local64_set(&event->count, 0); } put_cpu_var(hv_24x7_reqb); @@ -1567,20 +1662,60 @@ static struct pmu h_24x7_pmu = { .capabilities = PERF_PMU_CAP_NO_EXCLUDE, }; +static int ppc_hv_24x7_cpu_online(unsigned int cpu) +{ + if (cpumask_empty(&hv_24x7_cpumask)) + cpumask_set_cpu(cpu, &hv_24x7_cpumask); + + return 0; +} + +static int ppc_hv_24x7_cpu_offline(unsigned int cpu) +{ + int target; + + /* Check if exiting cpu is used for collecting 24x7 events */ + if (!cpumask_test_and_clear_cpu(cpu, &hv_24x7_cpumask)) + return 0; + + /* Find a new cpu to collect 24x7 events */ + target = cpumask_last(cpu_active_mask); + + if (target < 0 || target >= nr_cpu_ids) { + pr_err("hv_24x7: CPU hotplug init failed\n"); + return -1; + } + + /* Migrate 24x7 events to the new target */ + cpumask_set_cpu(target, &hv_24x7_cpumask); + perf_pmu_migrate_context(&h_24x7_pmu, cpu, target); + + return 0; +} + +static int hv_24x7_cpu_hotplug_init(void) +{ + return cpuhp_setup_state(CPUHP_AP_PERF_POWERPC_HV_24x7_ONLINE, + "perf/powerpc/hv_24x7:online", + ppc_hv_24x7_cpu_online, + ppc_hv_24x7_cpu_offline); +} + static int hv_24x7_init(void) { int r; unsigned long hret; + unsigned int pvr = mfspr(SPRN_PVR); struct hv_perf_caps caps; if (!firmware_has_feature(FW_FEATURE_LPAR)) { pr_debug("not a virtualized system, not enabling\n"); return -ENODEV; - } else if (!cur_cpu_spec->oprofile_cpu_type) - return -ENODEV; + } /* POWER8 only supports v1, while POWER9 only supports v2. */ - if (!strcmp(cur_cpu_spec->oprofile_cpu_type, "ppc64/power8")) + if (PVR_VER(pvr) == PVR_POWER8 || PVR_VER(pvr) == PVR_POWER8E || + PVR_VER(pvr) == PVR_POWER8NVL) interface_version = 1; else { interface_version = 2; @@ -1611,10 +1746,17 @@ static int hv_24x7_init(void) if (r) return r; + /* init cpuhotplug */ + r = hv_24x7_cpu_hotplug_init(); + if (r) + return r; + r = perf_pmu_register(&h_24x7_pmu, h_24x7_pmu.name, -1); if (r) return r; + read_24x7_sys_info(); + return 0; } diff --git a/arch/powerpc/perf/hv-gpci-requests.h b/arch/powerpc/perf/hv-gpci-requests.h index e608f9db12dd..5e86371a20c7 100644 --- a/arch/powerpc/perf/hv-gpci-requests.h +++ b/arch/powerpc/perf/hv-gpci-requests.h @@ -79,6 +79,7 @@ REQUEST(__field(0, 8, partition_id) ) #include I(REQUEST_END) +#ifdef ENABLE_EVENTS_COUNTERINFO_V6 /* * Not available for counter_info_version >= 0x8, use * run_instruction_cycles_by_partition(0x100) instead. @@ -92,10 +93,11 @@ REQUEST(__field(0, 8, partition_id) __count(0x10, 8, cycles) ) #include I(REQUEST_END) +#endif #define REQUEST_NAME system_performance_capabilities #define REQUEST_NUM 0x40 -#define REQUEST_IDX_KIND "starting_index=0xffffffffffffffff" +#define REQUEST_IDX_KIND "starting_index=0xffffffff" #include I(REQUEST_BEGIN) REQUEST(__field(0, 1, perf_collect_privileged) __field(0x1, 1, capability_mask) @@ -103,6 +105,7 @@ REQUEST(__field(0, 1, perf_collect_privileged) ) #include I(REQUEST_END) +#ifdef ENABLE_EVENTS_COUNTERINFO_V6 #define REQUEST_NAME processor_bus_utilization_abc_links #define REQUEST_NUM 0x50 #define REQUEST_IDX_KIND "hw_chip_id=?" @@ -194,6 +197,7 @@ REQUEST(__field(0, 4, phys_processor_idx) __count(0x28, 8, instructions_completed) ) #include I(REQUEST_END) +#endif /* Processor_core_power_mode (0x95) skipped, no counters */ /* Affinity_domain_information_by_virtual_processor (0xA0) skipped, @@ -223,7 +227,7 @@ REQUEST(__field(0, 2, partition_id) #define REQUEST_NAME system_hypervisor_times #define REQUEST_NUM 0xF0 -#define REQUEST_IDX_KIND "starting_index=0xffffffffffffffff" +#define REQUEST_IDX_KIND "starting_index=0xffffffff" #include I(REQUEST_BEGIN) REQUEST(__count(0, 8, time_spent_to_dispatch_virtual_processors) __count(0x8, 8, time_spent_processing_virtual_processor_timers) @@ -234,7 +238,7 @@ REQUEST(__count(0, 8, time_spent_to_dispatch_virtual_processors) #define REQUEST_NAME system_tlbie_count_and_time #define REQUEST_NUM 0xF4 -#define REQUEST_IDX_KIND "starting_index=0xffffffffffffffff" +#define REQUEST_IDX_KIND "starting_index=0xffffffff" #include I(REQUEST_BEGIN) REQUEST(__count(0, 8, tlbie_instructions_issued) /* diff --git a/arch/powerpc/perf/hv-gpci.c b/arch/powerpc/perf/hv-gpci.c index 6884d16ec19b..241551d1282f 100644 --- a/arch/powerpc/perf/hv-gpci.c +++ b/arch/powerpc/perf/hv-gpci.c @@ -48,6 +48,8 @@ EVENT_DEFINE_RANGE_FORMAT(length, config1, 24, 31); /* u32, byte offset */ EVENT_DEFINE_RANGE_FORMAT(offset, config1, 32, 63); +static cpumask_t hv_gpci_cpumask; + static struct attribute *format_attrs[] = { &format_attr_request.attr, &format_attr_starting_index.attr, @@ -63,14 +65,14 @@ static struct attribute *format_attrs[] = { NULL, }; -static struct attribute_group format_group = { +static const struct attribute_group format_group = { .name = "format", .attrs = format_attrs, }; static struct attribute_group event_group = { .name = "events", - .attrs = hv_gpci_event_attrs, + /* .attrs is set in init */ }; #define HV_CAPS_ATTR(_name, _format) \ @@ -94,7 +96,523 @@ static ssize_t kernel_version_show(struct device *dev, return sprintf(page, "0x%x\n", COUNTER_INFO_VERSION_CURRENT); } +static ssize_t cpumask_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + return cpumap_print_to_pagebuf(true, buf, &hv_gpci_cpumask); +} + +/* Interface attribute array index to store system information */ +#define INTERFACE_PROCESSOR_BUS_TOPOLOGY_ATTR 6 +#define INTERFACE_PROCESSOR_CONFIG_ATTR 7 +#define INTERFACE_AFFINITY_DOMAIN_VIA_VP_ATTR 8 +#define INTERFACE_AFFINITY_DOMAIN_VIA_DOM_ATTR 9 +#define INTERFACE_AFFINITY_DOMAIN_VIA_PAR_ATTR 10 +#define INTERFACE_NULL_ATTR 11 + +/* Counter request value to retrieve system information */ +enum { + PROCESSOR_BUS_TOPOLOGY, + PROCESSOR_CONFIG, + AFFINITY_DOMAIN_VIA_VP, /* affinity domain via virtual processor */ + AFFINITY_DOMAIN_VIA_DOM, /* affinity domain via domain */ + AFFINITY_DOMAIN_VIA_PAR, /* affinity domain via partition */ +}; + +static int sysinfo_counter_request[] = { + [PROCESSOR_BUS_TOPOLOGY] = 0xD0, + [PROCESSOR_CONFIG] = 0x90, + [AFFINITY_DOMAIN_VIA_VP] = 0xA0, + [AFFINITY_DOMAIN_VIA_DOM] = 0xB0, + [AFFINITY_DOMAIN_VIA_PAR] = 0xB1, +}; + +static DEFINE_PER_CPU(char, hv_gpci_reqb[HGPCI_REQ_BUFFER_SIZE]) __aligned(sizeof(uint64_t)); + +static unsigned long systeminfo_gpci_request(u32 req, u32 starting_index, + u16 secondary_index, char *buf, + size_t *n, struct hv_gpci_request_buffer *arg) +{ + unsigned long ret; + size_t i, j; + + arg->params.counter_request = cpu_to_be32(req); + arg->params.starting_index = cpu_to_be32(starting_index); + arg->params.secondary_index = cpu_to_be16(secondary_index); + + ret = plpar_hcall_norets(H_GET_PERF_COUNTER_INFO, + virt_to_phys(arg), HGPCI_REQ_BUFFER_SIZE); + + /* + * ret value as 'H_PARAMETER' corresponds to 'GEN_BUF_TOO_SMALL', + * which means that the current buffer size cannot accommodate + * all the information and a partial buffer returned. + * hcall fails incase of ret value other than H_SUCCESS or H_PARAMETER. + * + * ret value as H_AUTHORITY implies that partition is not permitted to retrieve + * performance information, and required to set + * "Enable Performance Information Collection" option. + */ + if (ret == H_AUTHORITY) + return -EPERM; + + /* + * hcall can fail with other possible ret value like H_PRIVILEGE/H_HARDWARE + * because of invalid buffer-length/address or due to some hardware + * error. + */ + if (ret && (ret != H_PARAMETER)) + return -EIO; + + /* + * hcall H_GET_PERF_COUNTER_INFO populates the 'returned_values' + * to show the total number of counter_value array elements + * returned via hcall. + * hcall also populates 'cv_element_size' corresponds to individual + * counter_value array element size. Below loop go through all + * counter_value array elements as per their size and add it to + * the output buffer. + */ + for (i = 0; i < be16_to_cpu(arg->params.returned_values); i++) { + j = i * be16_to_cpu(arg->params.cv_element_size); + + for (; j < (i + 1) * be16_to_cpu(arg->params.cv_element_size); j++) + *n += sprintf(buf + *n, "%02x", (u8)arg->bytes[j]); + *n += sprintf(buf + *n, "\n"); + } + + if (*n >= PAGE_SIZE) { + pr_info("System information exceeds PAGE_SIZE\n"); + return -EFBIG; + } + + return ret; +} + +static ssize_t processor_bus_topology_show(struct device *dev, struct device_attribute *attr, + char *buf) +{ + struct hv_gpci_request_buffer *arg; + unsigned long ret; + size_t n = 0; + + arg = (void *)get_cpu_var(hv_gpci_reqb); + memset(arg, 0, HGPCI_REQ_BUFFER_SIZE); + + /* + * Pass the counter request value 0xD0 corresponds to request + * type 'Processor_bus_topology', to retrieve + * the system topology information. + * starting_index value implies the starting hardware + * chip id. + */ + ret = systeminfo_gpci_request(sysinfo_counter_request[PROCESSOR_BUS_TOPOLOGY], + 0, 0, buf, &n, arg); + + if (!ret) + return n; + + if (ret != H_PARAMETER) + goto out; + + /* + * ret value as 'H_PARAMETER' corresponds to 'GEN_BUF_TOO_SMALL', which + * implies that buffer can't accommodate all information, and a partial buffer + * returned. To handle that, we need to make subsequent requests + * with next starting index to retrieve additional (missing) data. + * Below loop do subsequent hcalls with next starting index and add it + * to buffer util we get all the information. + */ + while (ret == H_PARAMETER) { + int returned_values = be16_to_cpu(arg->params.returned_values); + int elementsize = be16_to_cpu(arg->params.cv_element_size); + int last_element = (returned_values - 1) * elementsize; + + /* + * Since the starting index value is part of counter_value + * buffer elements, use the starting index value in the last + * element and add 1 to make subsequent hcalls. + */ + u32 starting_index = arg->bytes[last_element + 3] + + (arg->bytes[last_element + 2] << 8) + + (arg->bytes[last_element + 1] << 16) + + (arg->bytes[last_element] << 24) + 1; + + memset(arg, 0, HGPCI_REQ_BUFFER_SIZE); + + ret = systeminfo_gpci_request(sysinfo_counter_request[PROCESSOR_BUS_TOPOLOGY], + starting_index, 0, buf, &n, arg); + + if (!ret) + return n; + + if (ret != H_PARAMETER) + goto out; + } + + return n; + +out: + put_cpu_var(hv_gpci_reqb); + return ret; +} + +static ssize_t processor_config_show(struct device *dev, struct device_attribute *attr, + char *buf) +{ + struct hv_gpci_request_buffer *arg; + unsigned long ret; + size_t n = 0; + + arg = (void *)get_cpu_var(hv_gpci_reqb); + memset(arg, 0, HGPCI_REQ_BUFFER_SIZE); + + /* + * Pass the counter request value 0x90 corresponds to request + * type 'Processor_config', to retrieve + * the system processor information. + * starting_index value implies the starting hardware + * processor index. + */ + ret = systeminfo_gpci_request(sysinfo_counter_request[PROCESSOR_CONFIG], + 0, 0, buf, &n, arg); + + if (!ret) + return n; + + if (ret != H_PARAMETER) + goto out; + + /* + * ret value as 'H_PARAMETER' corresponds to 'GEN_BUF_TOO_SMALL', which + * implies that buffer can't accommodate all information, and a partial buffer + * returned. To handle that, we need to take subsequent requests + * with next starting index to retrieve additional (missing) data. + * Below loop do subsequent hcalls with next starting index and add it + * to buffer util we get all the information. + */ + while (ret == H_PARAMETER) { + int returned_values = be16_to_cpu(arg->params.returned_values); + int elementsize = be16_to_cpu(arg->params.cv_element_size); + int last_element = (returned_values - 1) * elementsize; + + /* + * Since the starting index is part of counter_value + * buffer elements, use the starting index value in the last + * element and add 1 to subsequent hcalls. + */ + u32 starting_index = arg->bytes[last_element + 3] + + (arg->bytes[last_element + 2] << 8) + + (arg->bytes[last_element + 1] << 16) + + (arg->bytes[last_element] << 24) + 1; + + memset(arg, 0, HGPCI_REQ_BUFFER_SIZE); + + ret = systeminfo_gpci_request(sysinfo_counter_request[PROCESSOR_CONFIG], + starting_index, 0, buf, &n, arg); + + if (!ret) + return n; + + if (ret != H_PARAMETER) + goto out; + } + + return n; + +out: + put_cpu_var(hv_gpci_reqb); + return ret; +} + +static ssize_t affinity_domain_via_virtual_processor_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct hv_gpci_request_buffer *arg; + unsigned long ret; + size_t n = 0; + + arg = (void *)get_cpu_var(hv_gpci_reqb); + memset(arg, 0, HGPCI_REQ_BUFFER_SIZE); + + /* + * Pass the counter request 0xA0 corresponds to request + * type 'Affinity_domain_information_by_virutal_processor', + * to retrieve the system affinity domain information. + * starting_index value refers to the starting hardware + * processor index. + */ + ret = systeminfo_gpci_request(sysinfo_counter_request[AFFINITY_DOMAIN_VIA_VP], + 0, 0, buf, &n, arg); + + if (!ret) + return n; + + if (ret != H_PARAMETER) + goto out; + + /* + * ret value as 'H_PARAMETER' corresponds to 'GEN_BUF_TOO_SMALL', which + * implies that buffer can't accommodate all information, and a partial buffer + * returned. To handle that, we need to take subsequent requests + * with next secondary index to retrieve additional (missing) data. + * Below loop do subsequent hcalls with next secondary index and add it + * to buffer util we get all the information. + */ + while (ret == H_PARAMETER) { + int returned_values = be16_to_cpu(arg->params.returned_values); + int elementsize = be16_to_cpu(arg->params.cv_element_size); + int last_element = (returned_values - 1) * elementsize; + + /* + * Since the starting index and secondary index type is part of the + * counter_value buffer elements, use the starting index value in the + * last array element as subsequent starting index, and use secondary index + * value in the last array element plus 1 as subsequent secondary index. + * For counter request '0xA0', starting index points to partition id + * and secondary index points to corresponding virtual processor index. + */ + u32 starting_index = arg->bytes[last_element + 1] + (arg->bytes[last_element] << 8); + u16 secondary_index = arg->bytes[last_element + 3] + + (arg->bytes[last_element + 2] << 8) + 1; + + memset(arg, 0, HGPCI_REQ_BUFFER_SIZE); + + ret = systeminfo_gpci_request(sysinfo_counter_request[AFFINITY_DOMAIN_VIA_VP], + starting_index, secondary_index, buf, &n, arg); + + if (!ret) + return n; + + if (ret != H_PARAMETER) + goto out; + } + + return n; + +out: + put_cpu_var(hv_gpci_reqb); + return ret; +} + +static ssize_t affinity_domain_via_domain_show(struct device *dev, struct device_attribute *attr, + char *buf) +{ + struct hv_gpci_request_buffer *arg; + unsigned long ret; + size_t n = 0; + + arg = (void *)get_cpu_var(hv_gpci_reqb); + memset(arg, 0, HGPCI_REQ_BUFFER_SIZE); + + /* + * Pass the counter request 0xB0 corresponds to request + * type 'Affinity_domain_information_by_domain', + * to retrieve the system affinity domain information. + * starting_index value refers to the starting hardware + * processor index. + */ + ret = systeminfo_gpci_request(sysinfo_counter_request[AFFINITY_DOMAIN_VIA_DOM], + 0, 0, buf, &n, arg); + + if (!ret) + return n; + + if (ret != H_PARAMETER) + goto out; + + /* + * ret value as 'H_PARAMETER' corresponds to 'GEN_BUF_TOO_SMALL', which + * implies that buffer can't accommodate all information, and a partial buffer + * returned. To handle that, we need to take subsequent requests + * with next starting index to retrieve additional (missing) data. + * Below loop do subsequent hcalls with next starting index and add it + * to buffer util we get all the information. + */ + while (ret == H_PARAMETER) { + int returned_values = be16_to_cpu(arg->params.returned_values); + int elementsize = be16_to_cpu(arg->params.cv_element_size); + int last_element = (returned_values - 1) * elementsize; + + /* + * Since the starting index value is part of counter_value + * buffer elements, use the starting index value in the last + * element and add 1 to make subsequent hcalls. + */ + u32 starting_index = arg->bytes[last_element + 1] + + (arg->bytes[last_element] << 8) + 1; + + memset(arg, 0, HGPCI_REQ_BUFFER_SIZE); + + ret = systeminfo_gpci_request(sysinfo_counter_request[AFFINITY_DOMAIN_VIA_DOM], + starting_index, 0, buf, &n, arg); + + if (!ret) + return n; + + if (ret != H_PARAMETER) + goto out; + } + + return n; + +out: + put_cpu_var(hv_gpci_reqb); + return ret; +} + +static void affinity_domain_via_partition_result_parse(int returned_values, + int element_size, char *buf, size_t *last_element, + size_t *n, struct hv_gpci_request_buffer *arg) +{ + size_t i = 0, j = 0; + size_t k, l, m; + uint16_t total_affinity_domain_ele, size_of_each_affinity_domain_ele; + + /* + * hcall H_GET_PERF_COUNTER_INFO populates the 'returned_values' + * to show the total number of counter_value array elements + * returned via hcall. + * Unlike other request types, the data structure returned by this + * request is variable-size. For this counter request type, + * hcall populates 'cv_element_size' corresponds to minimum size of + * the structure returned i.e; the size of the structure with no domain + * information. Below loop go through all counter_value array + * to determine the number and size of each domain array element and + * add it to the output buffer. + */ + while (i < returned_values) { + k = j; + for (; k < j + element_size; k++) + *n += sprintf(buf + *n, "%02x", (u8)arg->bytes[k]); + *n += sprintf(buf + *n, "\n"); + + total_affinity_domain_ele = (u8)arg->bytes[k - 2] << 8 | (u8)arg->bytes[k - 3]; + size_of_each_affinity_domain_ele = (u8)arg->bytes[k] << 8 | (u8)arg->bytes[k - 1]; + + for (l = 0; l < total_affinity_domain_ele; l++) { + for (m = 0; m < size_of_each_affinity_domain_ele; m++) { + *n += sprintf(buf + *n, "%02x", (u8)arg->bytes[k]); + k++; + } + *n += sprintf(buf + *n, "\n"); + } + + *n += sprintf(buf + *n, "\n"); + i++; + j = k; + } + + *last_element = k; +} + +static ssize_t affinity_domain_via_partition_show(struct device *dev, struct device_attribute *attr, + char *buf) +{ + struct hv_gpci_request_buffer *arg; + unsigned long ret; + size_t n = 0; + size_t last_element = 0; + u32 starting_index; + + arg = (void *)get_cpu_var(hv_gpci_reqb); + memset(arg, 0, HGPCI_REQ_BUFFER_SIZE); + + /* + * Pass the counter request value 0xB1 corresponds to counter request + * type 'Affinity_domain_information_by_partition', + * to retrieve the system affinity domain by partition information. + * starting_index value refers to the starting hardware + * processor index. + */ + arg->params.counter_request = cpu_to_be32(sysinfo_counter_request[AFFINITY_DOMAIN_VIA_PAR]); + arg->params.starting_index = cpu_to_be32(0); + + ret = plpar_hcall_norets(H_GET_PERF_COUNTER_INFO, + virt_to_phys(arg), HGPCI_REQ_BUFFER_SIZE); + + if (!ret) + goto parse_result; + + if (ret && (ret != H_PARAMETER)) + goto out; + + /* + * ret value as 'H_PARAMETER' implies that the current buffer size + * can't accommodate all the information, and a partial buffer + * returned. To handle that, we need to make subsequent requests + * with next starting index to retrieve additional (missing) data. + * Below loop do subsequent hcalls with next starting index and add it + * to buffer util we get all the information. + */ + while (ret == H_PARAMETER) { + affinity_domain_via_partition_result_parse( + be16_to_cpu(arg->params.returned_values) - 1, + be16_to_cpu(arg->params.cv_element_size), buf, + &last_element, &n, arg); + + if (n >= PAGE_SIZE) { + put_cpu_var(hv_gpci_reqb); + pr_debug("System information exceeds PAGE_SIZE\n"); + return -EFBIG; + } + + /* + * Since the starting index value is part of counter_value + * buffer elements, use the starting_index value in the last + * element and add 1 to make subsequent hcalls. + */ + starting_index = (u8)arg->bytes[last_element] << 8 | + (u8)arg->bytes[last_element + 1]; + + memset(arg, 0, HGPCI_REQ_BUFFER_SIZE); + arg->params.counter_request = cpu_to_be32( + sysinfo_counter_request[AFFINITY_DOMAIN_VIA_PAR]); + arg->params.starting_index = cpu_to_be32(starting_index); + + ret = plpar_hcall_norets(H_GET_PERF_COUNTER_INFO, + virt_to_phys(arg), HGPCI_REQ_BUFFER_SIZE); + + if (ret && (ret != H_PARAMETER)) + goto out; + } + +parse_result: + affinity_domain_via_partition_result_parse( + be16_to_cpu(arg->params.returned_values), + be16_to_cpu(arg->params.cv_element_size), + buf, &last_element, &n, arg); + + put_cpu_var(hv_gpci_reqb); + return n; + +out: + put_cpu_var(hv_gpci_reqb); + + /* + * ret value as 'H_PARAMETER' corresponds to 'GEN_BUF_TOO_SMALL', + * which means that the current buffer size cannot accommodate + * all the information and a partial buffer returned. + * hcall fails incase of ret value other than H_SUCCESS or H_PARAMETER. + * + * ret value as H_AUTHORITY implies that partition is not permitted to retrieve + * performance information, and required to set + * "Enable Performance Information Collection" option. + */ + if (ret == H_AUTHORITY) + return -EPERM; + + /* + * hcall can fail with other possible ret value like H_PRIVILEGE/H_HARDWARE + * because of invalid buffer-length/address or due to some hardware + * error. + */ + return -EIO; +} + static DEVICE_ATTR_RO(kernel_version); +static DEVICE_ATTR_RO(cpumask); + HV_CAPS_ATTR(version, "0x%x\n"); HV_CAPS_ATTR(ga, "%d\n"); HV_CAPS_ATTR(expanded, "%d\n"); @@ -108,10 +626,44 @@ static struct attribute *interface_attrs[] = { &hv_caps_attr_expanded.attr, &hv_caps_attr_lab.attr, &hv_caps_attr_collect_privileged.attr, + /* + * This NULL is a placeholder for the processor_bus_topology + * attribute, set in init function if applicable. + */ + NULL, + /* + * This NULL is a placeholder for the processor_config + * attribute, set in init function if applicable. + */ + NULL, + /* + * This NULL is a placeholder for the affinity_domain_via_virtual_processor + * attribute, set in init function if applicable. + */ + NULL, + /* + * This NULL is a placeholder for the affinity_domain_via_domain + * attribute, set in init function if applicable. + */ + NULL, + /* + * This NULL is a placeholder for the affinity_domain_via_partition + * attribute, set in init function if applicable. + */ + NULL, + NULL, +}; + +static struct attribute *cpumask_attrs[] = { + &dev_attr_cpumask.attr, NULL, }; -static struct attribute_group interface_group = { +static const struct attribute_group cpumask_attr_group = { + .attrs = cpumask_attrs, +}; + +static const struct attribute_group interface_group = { .name = "interface", .attrs = interface_attrs, }; @@ -120,20 +672,10 @@ static const struct attribute_group *attr_groups[] = { &format_group, &event_group, &interface_group, + &cpumask_attr_group, NULL, }; -#define HGPCI_REQ_BUFFER_SIZE 4096 -#define HGPCI_MAX_DATA_BYTES \ - (HGPCI_REQ_BUFFER_SIZE - sizeof(struct hv_get_perf_counter_info_params)) - -static DEFINE_PER_CPU(char, hv_gpci_reqb[HGPCI_REQ_BUFFER_SIZE]) __aligned(sizeof(uint64_t)); - -struct hv_gpci_request_buffer { - struct hv_get_perf_counter_info_params params; - uint8_t bytes[HGPCI_MAX_DATA_BYTES]; -} __packed; - static unsigned long single_gpci_request(u32 req, u32 starting_index, u16 secondary_index, u8 version_in, u32 offset, u8 length, u64 *value) @@ -153,6 +695,20 @@ static unsigned long single_gpci_request(u32 req, u32 starting_index, ret = plpar_hcall_norets(H_GET_PERF_COUNTER_INFO, virt_to_phys(arg), HGPCI_REQ_BUFFER_SIZE); + + /* + * ret value as 'H_PARAMETER' with detail_rc as 'GEN_BUF_TOO_SMALL', + * specifies that the current buffer size cannot accommodate + * all the information and a partial buffer returned. + * Since in this function we are only accessing data for a given starting index, + * we don't need to accommodate whole data and can get required count by + * accessing first entry data. + * Hence hcall fails only incase the ret value is other than H_SUCCESS or + * H_PARAMETER with detail_rc value as GEN_BUF_TOO_SMALL(0x1B). + */ + if (ret == H_PARAMETER && be32_to_cpu(arg->params.detail_rc) == 0x1B) + ret = 0; + if (ret) { pr_devel("hcall failed: 0x%lx\n", ret); goto out; @@ -164,7 +720,7 @@ static unsigned long single_gpci_request(u32 req, u32 starting_index, */ count = 0; for (i = offset; i < offset + length; i++) - count |= arg->bytes[i] << (i - offset); + count |= (u64)(arg->bytes[i]) << ((length - 1 - (i - offset)) * 8); *value = count; out: @@ -217,6 +773,7 @@ static int h_gpci_event_init(struct perf_event *event) { u64 count; u8 length; + unsigned long ret; /* Not our event */ if (event->attr.type != event->pmu->type) @@ -247,13 +804,23 @@ static int h_gpci_event_init(struct perf_event *event) } /* check if the request works... */ - if (single_gpci_request(event_get_request(event), + ret = single_gpci_request(event_get_request(event), event_get_starting_index(event), event_get_secondary_index(event), event_get_counter_info_version(event), event_get_offset(event), length, - &count)) { + &count); + + /* + * ret value as H_AUTHORITY implies that partition is not permitted to retrieve + * performance information, and required to set + * "Enable Performance Information Collection" option. + */ + if (ret == H_AUTHORITY) + return -EPERM; + + if (ret) { pr_devel("gpci hcall failed\n"); return -EINVAL; } @@ -275,11 +842,152 @@ static struct pmu h_gpci_pmu = { .capabilities = PERF_PMU_CAP_NO_EXCLUDE, }; +static int ppc_hv_gpci_cpu_online(unsigned int cpu) +{ + if (cpumask_empty(&hv_gpci_cpumask)) + cpumask_set_cpu(cpu, &hv_gpci_cpumask); + + return 0; +} + +static int ppc_hv_gpci_cpu_offline(unsigned int cpu) +{ + int target; + + /* Check if exiting cpu is used for collecting gpci events */ + if (!cpumask_test_and_clear_cpu(cpu, &hv_gpci_cpumask)) + return 0; + + /* Find a new cpu to collect gpci events */ + target = cpumask_last(cpu_active_mask); + + if (target < 0 || target >= nr_cpu_ids) { + pr_err("hv_gpci: CPU hotplug init failed\n"); + return -1; + } + + /* Migrate gpci events to the new target */ + cpumask_set_cpu(target, &hv_gpci_cpumask); + perf_pmu_migrate_context(&h_gpci_pmu, cpu, target); + + return 0; +} + +static int hv_gpci_cpu_hotplug_init(void) +{ + return cpuhp_setup_state(CPUHP_AP_PERF_POWERPC_HV_GPCI_ONLINE, + "perf/powerpc/hv_gcpi:online", + ppc_hv_gpci_cpu_online, + ppc_hv_gpci_cpu_offline); +} + +static struct device_attribute *sysinfo_device_attr_create(int + sysinfo_interface_group_index, u32 req) +{ + struct device_attribute *attr = NULL; + unsigned long ret; + struct hv_gpci_request_buffer *arg; + + if (sysinfo_interface_group_index < INTERFACE_PROCESSOR_BUS_TOPOLOGY_ATTR || + sysinfo_interface_group_index >= INTERFACE_NULL_ATTR) { + pr_info("Wrong interface group index for system information\n"); + return NULL; + } + + /* Check for given counter request value support */ + arg = (void *)get_cpu_var(hv_gpci_reqb); + memset(arg, 0, HGPCI_REQ_BUFFER_SIZE); + + arg->params.counter_request = cpu_to_be32(req); + + ret = plpar_hcall_norets(H_GET_PERF_COUNTER_INFO, + virt_to_phys(arg), HGPCI_REQ_BUFFER_SIZE); + + put_cpu_var(hv_gpci_reqb); + + /* + * Add given counter request value attribute in the interface_attrs + * attribute array, only for valid return types. + */ + if (!ret || ret == H_AUTHORITY || ret == H_PARAMETER) { + attr = kzalloc(sizeof(*attr), GFP_KERNEL); + if (!attr) + return NULL; + + sysfs_attr_init(&attr->attr); + attr->attr.mode = 0444; + + switch (sysinfo_interface_group_index) { + case INTERFACE_PROCESSOR_BUS_TOPOLOGY_ATTR: + attr->attr.name = "processor_bus_topology"; + attr->show = processor_bus_topology_show; + break; + case INTERFACE_PROCESSOR_CONFIG_ATTR: + attr->attr.name = "processor_config"; + attr->show = processor_config_show; + break; + case INTERFACE_AFFINITY_DOMAIN_VIA_VP_ATTR: + attr->attr.name = "affinity_domain_via_virtual_processor"; + attr->show = affinity_domain_via_virtual_processor_show; + break; + case INTERFACE_AFFINITY_DOMAIN_VIA_DOM_ATTR: + attr->attr.name = "affinity_domain_via_domain"; + attr->show = affinity_domain_via_domain_show; + break; + case INTERFACE_AFFINITY_DOMAIN_VIA_PAR_ATTR: + attr->attr.name = "affinity_domain_via_partition"; + attr->show = affinity_domain_via_partition_show; + break; + } + } else + pr_devel("hcall failed, with error: 0x%lx\n", ret); + + return attr; +} + +static void add_sysinfo_interface_files(void) +{ + int sysfs_count; + struct device_attribute *attr[INTERFACE_NULL_ATTR - INTERFACE_PROCESSOR_BUS_TOPOLOGY_ATTR]; + int i; + + sysfs_count = INTERFACE_NULL_ATTR - INTERFACE_PROCESSOR_BUS_TOPOLOGY_ATTR; + + /* Get device attribute for a given counter request value */ + for (i = 0; i < sysfs_count; i++) { + attr[i] = sysinfo_device_attr_create(i + INTERFACE_PROCESSOR_BUS_TOPOLOGY_ATTR, + sysinfo_counter_request[i]); + + if (!attr[i]) + goto out; + } + + /* Add sysinfo interface attributes in the interface_attrs attribute array */ + for (i = 0; i < sysfs_count; i++) + interface_attrs[i + INTERFACE_PROCESSOR_BUS_TOPOLOGY_ATTR] = &attr[i]->attr; + + return; + +out: + /* + * The sysinfo interface attributes will be added, only if hcall passed for + * all the counter request values. Free the device attribute array incase + * of any hcall failure. + */ + if (i > 0) { + while (i >= 0) { + kfree(attr[i]); + i--; + } + } +} + static int hv_gpci_init(void) { int r; unsigned long hret; struct hv_perf_caps caps; + struct hv_gpci_request_buffer *arg; hv_gpci_assert_offsets_correct(); @@ -295,13 +1003,52 @@ static int hv_gpci_init(void) return -ENODEV; } + /* init cpuhotplug */ + r = hv_gpci_cpu_hotplug_init(); + if (r) + return r; + /* sampling not supported */ h_gpci_pmu.capabilities |= PERF_PMU_CAP_NO_INTERRUPT; + arg = (void *)get_cpu_var(hv_gpci_reqb); + memset(arg, 0, HGPCI_REQ_BUFFER_SIZE); + + /* + * hcall H_GET_PERF_COUNTER_INFO populates the output + * counter_info_version value based on the system hypervisor. + * Pass the counter request 0x10 corresponds to request type + * 'Dispatch_timebase_by_processor', to get the supported + * counter_info_version. + */ + arg->params.counter_request = cpu_to_be32(0x10); + + r = plpar_hcall_norets(H_GET_PERF_COUNTER_INFO, + virt_to_phys(arg), HGPCI_REQ_BUFFER_SIZE); + if (r) { + pr_devel("hcall failed, can't get supported counter_info_version: 0x%x\n", r); + arg->params.counter_info_version_out = 0x8; + } + + /* + * Use counter_info_version_out value to assign + * required hv-gpci event list. + */ + if (arg->params.counter_info_version_out >= 0x8) + event_group.attrs = hv_gpci_event_attrs; + else + event_group.attrs = hv_gpci_event_attrs_v6; + + put_cpu_var(hv_gpci_reqb); + r = perf_pmu_register(&h_gpci_pmu, h_gpci_pmu.name, -1); if (r) return r; + /* sysinfo interface files are only available for power10 and above platforms */ + if (PVR_VER(mfspr(SPRN_PVR)) >= PVR_POWER10) + add_sysinfo_interface_files(); + return 0; } diff --git a/arch/powerpc/perf/hv-gpci.h b/arch/powerpc/perf/hv-gpci.h index a3053eda5dcc..c72020912dea 100644 --- a/arch/powerpc/perf/hv-gpci.h +++ b/arch/powerpc/perf/hv-gpci.h @@ -2,33 +2,6 @@ #ifndef LINUX_POWERPC_PERF_HV_GPCI_H_ #define LINUX_POWERPC_PERF_HV_GPCI_H_ -#include <linux/types.h> - -/* From the document "H_GetPerformanceCounterInfo Interface" v1.07 */ - -/* H_GET_PERF_COUNTER_INFO argument */ -struct hv_get_perf_counter_info_params { - __be32 counter_request; /* I */ - __be32 starting_index; /* IO */ - __be16 secondary_index; /* IO */ - __be16 returned_values; /* O */ - __be32 detail_rc; /* O, only needed when called via *_norets() */ - - /* - * O, size each of counter_value element in bytes, only set for version - * >= 0x3 - */ - __be16 cv_element_size; - - /* I, 0 (zero) for versions < 0x3 */ - __u8 counter_info_version_in; - - /* O, 0 (zero) if version < 0x3. Must be set to 0 when making hcall */ - __u8 counter_info_version_out; - __u8 reserved[0xC]; - __u8 counter_value[]; -} __packed; - /* * counter info version => fw version/reference (spec version) * @@ -53,6 +26,7 @@ enum { #define REQUEST_FILE "../hv-gpci-requests.h" #define NAME_LOWER hv_gpci #define NAME_UPPER HV_GPCI +#define ENABLE_EVENTS_COUNTERINFO_V6 #include "req-gen/perf.h" #undef REQUEST_FILE #undef NAME_LOWER diff --git a/arch/powerpc/perf/imc-pmu.c b/arch/powerpc/perf/imc-pmu.c index cb50a9e1fd2d..8664a7d297ad 100644 --- a/arch/powerpc/perf/imc-pmu.c +++ b/arch/powerpc/perf/imc-pmu.c @@ -6,6 +6,7 @@ * (C) 2017 Anju T Sudhakar, IBM Corporation. * (C) 2017 Hemant K Shaw, IBM Corporation. */ +#include <linux/of.h> #include <linux/perf_event.h> #include <linux/slab.h> #include <asm/opal.h> @@ -13,6 +14,7 @@ #include <asm/cputhreads.h> #include <asm/smp.h> #include <linux/string.h> +#include <linux/spinlock.h> /* Nest IMC data structures and variables */ @@ -44,6 +46,16 @@ static DEFINE_PER_CPU(u64 *, trace_imc_mem); static struct imc_pmu_ref *trace_imc_refc; static int trace_imc_mem_size; +/* + * Global data structure used to avoid races between thread, + * core and trace-imc + */ +static struct imc_pmu_ref imc_global_refc = { + .lock = __SPIN_LOCK_UNLOCKED(imc_global_refc.lock), + .id = 0, + .refc = 0, +}; + static struct imc_pmu *imc_event_to_pmu(struct perf_event *event) { return container_of(event->pmu, struct imc_pmu, pmu); @@ -61,7 +73,7 @@ static struct attribute *imc_format_attrs[] = { NULL, }; -static struct attribute_group imc_format_group = { +static const struct attribute_group imc_format_group = { .name = "format", .attrs = imc_format_attrs, }; @@ -80,7 +92,7 @@ static struct attribute *trace_imc_format_attrs[] = { NULL, }; -static struct attribute_group trace_imc_format_group = { +static const struct attribute_group trace_imc_format_group = { .name = "format", .attrs = trace_imc_format_attrs, }; @@ -115,7 +127,7 @@ static struct attribute *imc_pmu_cpumask_attrs[] = { NULL, }; -static struct attribute_group imc_pmu_cpumask_attr_group = { +static const struct attribute_group imc_pmu_cpumask_attr_group = { .attrs = imc_pmu_cpumask_attrs, }; @@ -229,8 +241,10 @@ static int update_events_in_group(struct device_node *node, struct imc_pmu *pmu) ct = of_get_child_count(pmu_events); /* Get the event prefix */ - if (of_property_read_string(node, "events-prefix", &prefix)) + if (of_property_read_string(node, "events-prefix", &prefix)) { + of_node_put(pmu_events); return 0; + } /* Get a global unit and scale data if available */ if (of_property_read_string(node, "scale", &g_scale)) @@ -244,8 +258,10 @@ static int update_events_in_group(struct device_node *node, struct imc_pmu *pmu) /* Allocate memory for the events */ pmu->events = kcalloc(ct, sizeof(struct imc_events), GFP_KERNEL); - if (!pmu->events) + if (!pmu->events) { + of_node_put(pmu_events); return -ENOMEM; + } ct = 0; /* Parse the events and update the struct */ @@ -255,6 +271,8 @@ static int update_events_in_group(struct device_node *node, struct imc_pmu *pmu) ct++; } + of_node_put(pmu_events); + /* Allocate memory for attribute group */ attr_group = kzalloc(sizeof(*attr_group), GFP_KERNEL); if (!attr_group) { @@ -281,6 +299,8 @@ static int update_events_in_group(struct device_node *node, struct imc_pmu *pmu) attr_group->attrs = attrs; do { ev_val_str = kasprintf(GFP_KERNEL, "event=0x%x", pmu->events[i].value); + if (!ev_val_str) + continue; dev_str = device_str_attr_create(pmu->events[i].name, ev_val_str); if (!dev_str) continue; @@ -288,6 +308,8 @@ static int update_events_in_group(struct device_node *node, struct imc_pmu *pmu) attrs[j++] = dev_str; if (pmu->events[i].scale) { ev_scale_str = kasprintf(GFP_KERNEL, "%s.scale", pmu->events[i].name); + if (!ev_scale_str) + continue; dev_str = device_str_attr_create(ev_scale_str, pmu->events[i].scale); if (!dev_str) continue; @@ -297,6 +319,8 @@ static int update_events_in_group(struct device_node *node, struct imc_pmu *pmu) if (pmu->events[i].unit) { ev_unit_str = kasprintf(GFP_KERNEL, "%s.unit", pmu->events[i].name); + if (!ev_unit_str) + continue; dev_str = device_str_attr_create(ev_unit_str, pmu->events[i].unit); if (!dev_str) continue; @@ -383,7 +407,7 @@ static int ppc_nest_imc_cpu_offline(unsigned int cpu) get_hard_smp_processor_id(cpu)); /* * If this is the last cpu in this chip then, skip the reference - * count mutex lock and make the reference count on this chip zero. + * count lock and make the reference count on this chip zero. */ ref = get_nest_pmu_ref(cpu); if (!ref) @@ -445,15 +469,15 @@ static void nest_imc_counters_release(struct perf_event *event) /* * See if we need to disable the nest PMU. * If no events are currently in use, then we have to take a - * mutex to ensure that we don't race with another task doing + * lock to ensure that we don't race with another task doing * enable or disable the nest counters. */ ref = get_nest_pmu_ref(event->cpu); if (!ref) return; - /* Take the mutex lock for this node and then decrement the reference count */ - mutex_lock(&ref->lock); + /* Take the lock for this node and then decrement the reference count */ + spin_lock(&ref->lock); if (ref->refc == 0) { /* * The scenario where this is true is, when perf session is @@ -465,7 +489,7 @@ static void nest_imc_counters_release(struct perf_event *event) * an OPAL call to disable the engine in that node. * */ - mutex_unlock(&ref->lock); + spin_unlock(&ref->lock); return; } ref->refc--; @@ -473,7 +497,7 @@ static void nest_imc_counters_release(struct perf_event *event) rc = opal_imc_counters_stop(OPAL_IMC_COUNTERS_NEST, get_hard_smp_processor_id(event->cpu)); if (rc) { - mutex_unlock(&ref->lock); + spin_unlock(&ref->lock); pr_err("nest-imc: Unable to stop the counters for core %d\n", node_id); return; } @@ -481,7 +505,7 @@ static void nest_imc_counters_release(struct perf_event *event) WARN(1, "nest-imc: Invalid event reference count\n"); ref->refc = 0; } - mutex_unlock(&ref->lock); + spin_unlock(&ref->lock); } static int nest_imc_event_init(struct perf_event *event) @@ -511,7 +535,7 @@ static int nest_imc_event_init(struct perf_event *event) /* * Nest HW counter memory resides in a per-chip reserve-memory (HOMER). - * Get the base memory addresss for this cpu. + * Get the base memory address for this cpu. */ chip_id = cpu_to_chip_id(event->cpu); @@ -526,7 +550,7 @@ static int nest_imc_event_init(struct perf_event *event) break; } pcni++; - } while (pcni->vbase != 0); + } while (pcni->vbase); if (!flag) return -ENODEV; @@ -540,26 +564,25 @@ static int nest_imc_event_init(struct perf_event *event) /* * Get the imc_pmu_ref struct for this node. - * Take the mutex lock and then increment the count of nest pmu events - * inited. + * Take the lock and then increment the count of nest pmu events inited. */ ref = get_nest_pmu_ref(event->cpu); if (!ref) return -EINVAL; - mutex_lock(&ref->lock); + spin_lock(&ref->lock); if (ref->refc == 0) { rc = opal_imc_counters_start(OPAL_IMC_COUNTERS_NEST, get_hard_smp_processor_id(event->cpu)); if (rc) { - mutex_unlock(&ref->lock); + spin_unlock(&ref->lock); pr_err("nest-imc: Unable to start the counters for node %d\n", node_id); return rc; } } ++ref->refc; - mutex_unlock(&ref->lock); + spin_unlock(&ref->lock); event->destroy = nest_imc_counters_release; return 0; @@ -595,9 +618,8 @@ static int core_imc_mem_init(int cpu, int size) return -ENOMEM; mem_info->vbase = page_address(page); - /* Init the mutex */ core_imc_refc[core_id].id = core_id; - mutex_init(&core_imc_refc[core_id].lock); + spin_lock_init(&core_imc_refc[core_id].lock); rc = opal_imc_counters_init(OPAL_IMC_COUNTERS_CORE, __pa((void *)mem_info->vbase), @@ -664,7 +686,7 @@ static int ppc_core_imc_cpu_offline(unsigned int cpu) /* * Check whether core_imc is registered. We could end up here * if the cpuhotplug callback registration fails. i.e, callback - * invokes the offline path for all sucessfully registered cpus. + * invokes the offline path for all successfully registered cpus. * At this stage, core_imc pmu will not be registered and we * should return here. * @@ -686,9 +708,8 @@ static int ppc_core_imc_cpu_offline(unsigned int cpu) perf_pmu_migrate_context(&core_imc_pmu->pmu, cpu, ncpu); } else { /* - * If this is the last cpu in this core then, skip taking refernce - * count mutex lock for this core and directly zero "refc" for - * this core. + * If this is the last cpu in this core then skip taking reference + * count lock for this core and directly zero "refc" for this core. */ opal_imc_counters_stop(OPAL_IMC_COUNTERS_CORE, get_hard_smp_processor_id(cpu)); @@ -698,6 +719,16 @@ static int ppc_core_imc_cpu_offline(unsigned int cpu) return -EINVAL; ref->refc = 0; + /* + * Reduce the global reference count, if this is the + * last cpu in this core and core-imc event running + * in this cpu. + */ + spin_lock(&imc_global_refc.lock); + if (imc_global_refc.id == IMC_DOMAIN_CORE) + imc_global_refc.refc--; + + spin_unlock(&imc_global_refc.lock); } return 0; } @@ -710,6 +741,23 @@ static int core_imc_pmu_cpumask_init(void) ppc_core_imc_cpu_offline); } +static void reset_global_refc(struct perf_event *event) +{ + spin_lock(&imc_global_refc.lock); + imc_global_refc.refc--; + + /* + * If no other thread is running any + * event for this domain(thread/core/trace), + * set the global id to zero. + */ + if (imc_global_refc.refc <= 0) { + imc_global_refc.refc = 0; + imc_global_refc.id = 0; + } + spin_unlock(&imc_global_refc.lock); +} + static void core_imc_counters_release(struct perf_event *event) { int rc, core_id; @@ -720,17 +768,17 @@ static void core_imc_counters_release(struct perf_event *event) /* * See if we need to disable the IMC PMU. * If no events are currently in use, then we have to take a - * mutex to ensure that we don't race with another task doing + * lock to ensure that we don't race with another task doing * enable or disable the core counters. */ core_id = event->cpu / threads_per_core; - /* Take the mutex lock and decrement the refernce count for this core */ + /* Take the lock and decrement the refernce count for this core */ ref = &core_imc_refc[core_id]; if (!ref) return; - mutex_lock(&ref->lock); + spin_lock(&ref->lock); if (ref->refc == 0) { /* * The scenario where this is true is, when perf session is @@ -742,7 +790,7 @@ static void core_imc_counters_release(struct perf_event *event) * an OPAL call to disable the engine in that core. * */ - mutex_unlock(&ref->lock); + spin_unlock(&ref->lock); return; } ref->refc--; @@ -750,7 +798,7 @@ static void core_imc_counters_release(struct perf_event *event) rc = opal_imc_counters_stop(OPAL_IMC_COUNTERS_CORE, get_hard_smp_processor_id(event->cpu)); if (rc) { - mutex_unlock(&ref->lock); + spin_unlock(&ref->lock); pr_err("IMC: Unable to stop the counters for core %d\n", core_id); return; } @@ -758,7 +806,9 @@ static void core_imc_counters_release(struct perf_event *event) WARN(1, "core-imc: Invalid event reference count\n"); ref->refc = 0; } - mutex_unlock(&ref->lock); + spin_unlock(&ref->lock); + + reset_global_refc(event); } static int core_imc_event_init(struct perf_event *event) @@ -794,7 +844,6 @@ static int core_imc_event_init(struct perf_event *event) if ((!pcmi->vbase)) return -ENODEV; - /* Get the core_imc mutex for this core */ ref = &core_imc_refc[core_id]; if (!ref) return -EINVAL; @@ -802,22 +851,45 @@ static int core_imc_event_init(struct perf_event *event) /* * Core pmu units are enabled only when it is used. * See if this is triggered for the first time. - * If yes, take the mutex lock and enable the core counters. + * If yes, take the lock and enable the core counters. * If not, just increment the count in core_imc_refc struct. */ - mutex_lock(&ref->lock); + spin_lock(&ref->lock); if (ref->refc == 0) { rc = opal_imc_counters_start(OPAL_IMC_COUNTERS_CORE, get_hard_smp_processor_id(event->cpu)); if (rc) { - mutex_unlock(&ref->lock); + spin_unlock(&ref->lock); pr_err("core-imc: Unable to start the counters for core %d\n", core_id); return rc; } } ++ref->refc; - mutex_unlock(&ref->lock); + spin_unlock(&ref->lock); + + /* + * Since the system can run either in accumulation or trace-mode + * of IMC at a time, core-imc events are allowed only if no other + * trace/thread imc events are enabled/monitored. + * + * Take the global lock, and check the refc.id + * to know whether any other trace/thread imc + * events are running. + */ + spin_lock(&imc_global_refc.lock); + if (imc_global_refc.id == 0 || imc_global_refc.id == IMC_DOMAIN_CORE) { + /* + * No other trace/thread imc events are running in + * the system, so set the refc.id to core-imc. + */ + imc_global_refc.id = IMC_DOMAIN_CORE; + imc_global_refc.refc++; + } else { + spin_unlock(&imc_global_refc.lock); + return -EBUSY; + } + spin_unlock(&imc_global_refc.lock); event->hw.event_base = (u64)pcmi->vbase + (config & IMC_EVENT_OFFSET_MASK); event->destroy = core_imc_counters_release; @@ -877,7 +949,23 @@ static int ppc_thread_imc_cpu_online(unsigned int cpu) static int ppc_thread_imc_cpu_offline(unsigned int cpu) { - mtspr(SPRN_LDBAR, 0); + /* + * Set the bit 0 of LDBAR to zero. + * + * If bit 0 of LDBAR is unset, it will stop posting + * the counter data to memory. + * For thread-imc, bit 0 of LDBAR will be set to 1 in the + * event_add function. So reset this bit here, to stop the updates + * to memory in the cpu_offline path. + */ + mtspr(SPRN_LDBAR, (mfspr(SPRN_LDBAR) & (~(1UL << 63)))); + + /* Reduce the refc if thread-imc event running on this cpu */ + spin_lock(&imc_global_refc.lock); + if (imc_global_refc.id == IMC_DOMAIN_THREAD) + imc_global_refc.refc--; + spin_unlock(&imc_global_refc.lock); + return 0; } @@ -898,7 +986,7 @@ static int thread_imc_event_init(struct perf_event *event) if (event->attr.type != event->pmu->type) return -ENOENT; - if (!capable(CAP_SYS_ADMIN)) + if (!perfmon_capable()) return -EACCES; /* Sampling not supported */ @@ -916,7 +1004,22 @@ static int thread_imc_event_init(struct perf_event *event) if (!target) return -EINVAL; + spin_lock(&imc_global_refc.lock); + /* + * Check if any other trace/core imc events are running in the + * system, if not set the global id to thread-imc. + */ + if (imc_global_refc.id == 0 || imc_global_refc.id == IMC_DOMAIN_THREAD) { + imc_global_refc.id = IMC_DOMAIN_THREAD; + imc_global_refc.refc++; + } else { + spin_unlock(&imc_global_refc.lock); + return -EBUSY; + } + spin_unlock(&imc_global_refc.lock); + event->pmu->task_ctx_nr = perf_sw_context; + event->destroy = reset_global_refc; return 0; } @@ -928,16 +1031,16 @@ static bool is_thread_imc_pmu(struct perf_event *event) return false; } -static u64 * get_event_base_addr(struct perf_event *event) +static __be64 *get_event_base_addr(struct perf_event *event) { u64 addr; if (is_thread_imc_pmu(event)) { addr = (u64)per_cpu(thread_imc_mem, smp_processor_id()); - return (u64 *)(addr + (event->attr.config & IMC_EVENT_OFFSET_MASK)); + return (__be64 *)(addr + (event->attr.config & IMC_EVENT_OFFSET_MASK)); } - return (u64 *)event->hw.event_base; + return (__be64 *)event->hw.event_base; } static void thread_imc_pmu_start_txn(struct pmu *pmu, @@ -961,7 +1064,8 @@ static int thread_imc_pmu_commit_txn(struct pmu *pmu) static u64 imc_read_counter(struct perf_event *event) { - u64 *addr, data; + __be64 *addr; + u64 data; /* * In-Memory Collection (IMC) counters are free flowing counters. @@ -1035,25 +1139,25 @@ static int thread_imc_event_add(struct perf_event *event, int flags) /* * imc pmus are enabled only when it is used. * See if this is triggered for the first time. - * If yes, take the mutex lock and enable the counters. + * If yes, take the lock and enable the counters. * If not, just increment the count in ref count struct. */ ref = &core_imc_refc[core_id]; if (!ref) return -EINVAL; - mutex_lock(&ref->lock); + spin_lock(&ref->lock); if (ref->refc == 0) { if (opal_imc_counters_start(OPAL_IMC_COUNTERS_CORE, get_hard_smp_processor_id(smp_processor_id()))) { - mutex_unlock(&ref->lock); + spin_unlock(&ref->lock); pr_err("thread-imc: Unable to start the counter\ for core %d\n", core_id); return -EINVAL; } } ++ref->refc; - mutex_unlock(&ref->lock); + spin_unlock(&ref->lock); return 0; } @@ -1063,17 +1167,19 @@ static void thread_imc_event_del(struct perf_event *event, int flags) int core_id; struct imc_pmu_ref *ref; - mtspr(SPRN_LDBAR, 0); - core_id = smp_processor_id() / threads_per_core; ref = &core_imc_refc[core_id]; + if (!ref) { + pr_debug("imc: Failed to get event reference count\n"); + return; + } - mutex_lock(&ref->lock); + spin_lock(&ref->lock); ref->refc--; if (ref->refc == 0) { if (opal_imc_counters_stop(OPAL_IMC_COUNTERS_CORE, get_hard_smp_processor_id(smp_processor_id()))) { - mutex_unlock(&ref->lock); + spin_unlock(&ref->lock); pr_err("thread-imc: Unable to stop the counters\ for core %d\n", core_id); return; @@ -1081,7 +1187,11 @@ static void thread_imc_event_del(struct perf_event *event, int flags) } else if (ref->refc < 0) { ref->refc = 0; } - mutex_unlock(&ref->lock); + spin_unlock(&ref->lock); + + /* Set bit 0 of LDBAR to zero, to stop posting updates to memory */ + mtspr(SPRN_LDBAR, (mfspr(SPRN_LDBAR) & (~(1UL << 63)))); + /* * Take a snapshot and calculate the delta and update * the event counter values. @@ -1118,9 +1228,8 @@ static int trace_imc_mem_alloc(int cpu_id, int size) } } - /* Init the mutex, if not already */ trace_imc_refc[core_id].id = core_id; - mutex_init(&trace_imc_refc[core_id].lock); + spin_lock_init(&trace_imc_refc[core_id].lock); mtspr(SPRN_LDBAR, 0); return 0; @@ -1133,7 +1242,18 @@ static int ppc_trace_imc_cpu_online(unsigned int cpu) static int ppc_trace_imc_cpu_offline(unsigned int cpu) { - mtspr(SPRN_LDBAR, 0); + /* + * No need to set bit 0 of LDBAR to zero, as + * it is set to zero for imc trace-mode + * + * Reduce the refc if any trace-imc event running + * on this cpu. + */ + spin_lock(&imc_global_refc.lock); + if (imc_global_refc.id == IMC_DOMAIN_TRACE) + imc_global_refc.refc--; + spin_unlock(&imc_global_refc.lock); + return 0; } @@ -1178,11 +1298,30 @@ static int trace_imc_prepare_sample(struct trace_imc_data *mem, header->size = sizeof(*header) + event->header_size; header->misc = 0; - if (is_kernel_addr(data->ip)) - header->misc |= PERF_RECORD_MISC_KERNEL; - else - header->misc |= PERF_RECORD_MISC_USER; - + if (cpu_has_feature(CPU_FTR_ARCH_31)) { + switch (IMC_TRACE_RECORD_VAL_HVPR(be64_to_cpu(READ_ONCE(mem->val)))) { + case 0:/* when MSR HV and PR not set in the trace-record */ + header->misc |= PERF_RECORD_MISC_GUEST_KERNEL; + break; + case 1: /* MSR HV is 0 and PR is 1 */ + header->misc |= PERF_RECORD_MISC_GUEST_USER; + break; + case 2: /* MSR HV is 1 and PR is 0 */ + header->misc |= PERF_RECORD_MISC_KERNEL; + break; + case 3: /* MSR HV is 1 and PR is 1 */ + header->misc |= PERF_RECORD_MISC_USER; + break; + default: + pr_info("IMC: Unable to set the flag based on MSR bits\n"); + break; + } + } else { + if (is_kernel_addr(data->ip)) + header->misc |= PERF_RECORD_MISC_KERNEL; + else + header->misc |= PERF_RECORD_MISC_USER; + } perf_event_header__init_id(header, data, event); return 0; @@ -1207,7 +1346,7 @@ static void dump_trace_imc_data(struct perf_event *event) /* If this is a valid record, create the sample */ struct perf_output_handle handle; - if (perf_output_begin(&handle, event, header.size)) + if (perf_output_begin(&handle, &data, event, header.size)) return; perf_output_sample(&handle, &header, &data, event); @@ -1226,29 +1365,26 @@ static int trace_imc_event_add(struct perf_event *event, int flags) local_mem = get_trace_imc_event_base_addr(); ldbar_value = ((u64)local_mem & THREAD_IMC_LDBAR_MASK) | TRACE_IMC_ENABLE; - if (core_imc_refc) - ref = &core_imc_refc[core_id]; + /* trace-imc reference count */ + if (trace_imc_refc) + ref = &trace_imc_refc[core_id]; if (!ref) { - /* If core-imc is not enabled, use trace-imc reference count */ - if (trace_imc_refc) - ref = &trace_imc_refc[core_id]; - if (!ref) - return -EINVAL; + pr_debug("imc: Failed to get the event reference count\n"); + return -EINVAL; } + mtspr(SPRN_LDBAR, ldbar_value); - mutex_lock(&ref->lock); + spin_lock(&ref->lock); if (ref->refc == 0) { if (opal_imc_counters_start(OPAL_IMC_COUNTERS_TRACE, get_hard_smp_processor_id(smp_processor_id()))) { - mutex_unlock(&ref->lock); + spin_unlock(&ref->lock); pr_err("trace-imc: Unable to start the counters for core %d\n", core_id); - mtspr(SPRN_LDBAR, 0); return -EINVAL; } } ++ref->refc; - mutex_unlock(&ref->lock); - + spin_unlock(&ref->lock); return 0; } @@ -1274,50 +1410,69 @@ static void trace_imc_event_del(struct perf_event *event, int flags) int core_id = smp_processor_id() / threads_per_core; struct imc_pmu_ref *ref = NULL; - if (core_imc_refc) - ref = &core_imc_refc[core_id]; + if (trace_imc_refc) + ref = &trace_imc_refc[core_id]; if (!ref) { - /* If core-imc is not enabled, use trace-imc reference count */ - if (trace_imc_refc) - ref = &trace_imc_refc[core_id]; - if (!ref) - return; + pr_debug("imc: Failed to get event reference count\n"); + return; } - mtspr(SPRN_LDBAR, 0); - mutex_lock(&ref->lock); + + spin_lock(&ref->lock); ref->refc--; if (ref->refc == 0) { if (opal_imc_counters_stop(OPAL_IMC_COUNTERS_TRACE, get_hard_smp_processor_id(smp_processor_id()))) { - mutex_unlock(&ref->lock); + spin_unlock(&ref->lock); pr_err("trace-imc: Unable to stop the counters for core %d\n", core_id); return; } } else if (ref->refc < 0) { ref->refc = 0; } - mutex_unlock(&ref->lock); + spin_unlock(&ref->lock); + trace_imc_event_stop(event, flags); } static int trace_imc_event_init(struct perf_event *event) { - struct task_struct *target; - if (event->attr.type != event->pmu->type) return -ENOENT; - if (!capable(CAP_SYS_ADMIN)) + if (!perfmon_capable()) return -EACCES; /* Return if this is a couting event */ if (event->attr.sample_period == 0) return -ENOENT; + /* + * Take the global lock, and make sure + * no other thread is running any core/thread imc + * events + */ + spin_lock(&imc_global_refc.lock); + if (imc_global_refc.id == 0 || imc_global_refc.id == IMC_DOMAIN_TRACE) { + /* + * No core/thread imc events are running in the + * system, so set the refc.id to trace-imc. + */ + imc_global_refc.id = IMC_DOMAIN_TRACE; + imc_global_refc.refc++; + } else { + spin_unlock(&imc_global_refc.lock); + return -EBUSY; + } + spin_unlock(&imc_global_refc.lock); + event->hw.idx = -1; - target = event->hw.target; - event->pmu->task_ctx_nr = perf_hw_context; + /* + * There can only be a single PMU for perf_hw_context events which is assigned to + * core PMU. Hence use "perf_sw_context" for trace_imc. + */ + event->pmu->task_ctx_nr = perf_sw_context; + event->destroy = reset_global_refc; return 0; } @@ -1359,6 +1514,7 @@ static int update_pmu_ops(struct imc_pmu *pmu) pmu->pmu.stop = trace_imc_event_stop; pmu->pmu.read = trace_imc_event_read; pmu->attr_groups[IMC_FORMAT_ATTR] = &trace_imc_format_group; + break; default: break; } @@ -1380,10 +1536,10 @@ static int init_nest_pmu_ref(void) i = 0; for_each_node(nid) { /* - * Mutex lock to avoid races while tracking the number of + * Take the lock to avoid races while tracking the number of * sessions using the chip's nest pmu units. */ - mutex_init(&nest_imc_refc[i].lock); + spin_lock_init(&nest_imc_refc[i].lock); /* * Loop to init the "id" with the node_id. Variable "i" initialized to @@ -1429,10 +1585,10 @@ static void cleanup_all_core_imc_memory(void) static void thread_imc_ldbar_disable(void *dummy) { /* - * By Zeroing LDBAR, we disable thread-imc - * updates. + * By setting 0th bit of LDBAR to zero, we disable thread-imc + * updates to memory. */ - mtspr(SPRN_LDBAR, 0); + mtspr(SPRN_LDBAR, (mfspr(SPRN_LDBAR) & (~(1UL << 63)))); } void thread_imc_disable(void) diff --git a/arch/powerpc/perf/internal.h b/arch/powerpc/perf/internal.h index f755c64da137..a70ac471a5a5 100644 --- a/arch/powerpc/perf/internal.h +++ b/arch/powerpc/perf/internal.h @@ -2,11 +2,13 @@ // // Copyright 2019 Madhavan Srinivasan, IBM Corporation. -extern int init_ppc970_pmu(void); -extern int init_power5_pmu(void); -extern int init_power5p_pmu(void); -extern int init_power6_pmu(void); -extern int init_power7_pmu(void); -extern int init_power8_pmu(void); -extern int init_power9_pmu(void); -extern int init_generic_compat_pmu(void); +int __init init_ppc970_pmu(void); +int __init init_power5_pmu(void); +int __init init_power5p_pmu(void); +int __init init_power6_pmu(void); +int __init init_power7_pmu(void); +int __init init_power8_pmu(void); +int __init init_power9_pmu(void); +int __init init_power10_pmu(void); +int __init init_power11_pmu(void); +int __init init_generic_compat_pmu(void); diff --git a/arch/powerpc/perf/isa207-common.c b/arch/powerpc/perf/isa207-common.c index 4c86da5eb28a..56301b2bc8ae 100644 --- a/arch/powerpc/perf/isa207-common.c +++ b/arch/powerpc/perf/isa207-common.c @@ -21,7 +21,7 @@ PMU_FORMAT_ATTR(thresh_stop, "config:32-35"); PMU_FORMAT_ATTR(thresh_start, "config:36-39"); PMU_FORMAT_ATTR(thresh_cmp, "config:40-49"); -struct attribute *isa207_pmu_format_attr[] = { +static struct attribute *isa207_pmu_format_attr[] = { &format_attr_event.attr, &format_attr_pmcxsel.attr, &format_attr_mark.attr, @@ -37,7 +37,7 @@ struct attribute *isa207_pmu_format_attr[] = { NULL, }; -struct attribute_group isa207_pmu_format_group = { +const struct attribute_group isa207_pmu_format_group = { .name = "format", .attrs = isa207_pmu_format_attr, }; @@ -55,7 +55,9 @@ static bool is_event_valid(u64 event) { u64 valid_mask = EVENT_VALID_MASK; - if (cpu_has_feature(CPU_FTR_ARCH_300)) + if (cpu_has_feature(CPU_FTR_ARCH_31)) + valid_mask = p10_EVENT_VALID_MASK; + else if (cpu_has_feature(CPU_FTR_ARCH_300)) valid_mask = p9_EVENT_VALID_MASK; return !(event & ~valid_mask); @@ -69,17 +71,25 @@ static inline bool is_event_marked(u64 event) return false; } +static unsigned long sdar_mod_val(u64 event) +{ + if (cpu_has_feature(CPU_FTR_ARCH_31)) + return p10_SDAR_MODE(event); + + return p9_SDAR_MODE(event); +} + static void mmcra_sdar_mode(u64 event, unsigned long *mmcra) { /* - * MMCRA[SDAR_MODE] specifices how the SDAR should be updated in - * continous sampling mode. + * MMCRA[SDAR_MODE] specifies how the SDAR should be updated in + * continuous sampling mode. * * Incase of Power8: - * MMCRA[SDAR_MODE] will be programmed as "0b01" for continous sampling + * MMCRA[SDAR_MODE] will be programmed as "0b01" for continuous sampling * mode and will be un-changed when setting MMCRA[63] (Marked events). * - * Incase of Power9: + * Incase of Power9/power10: * Marked event: MMCRA[SDAR_MODE] will be set to 0b00 ('No Updates'), * or if group already have any marked events. * For rest @@ -90,20 +100,65 @@ static void mmcra_sdar_mode(u64 event, unsigned long *mmcra) if (cpu_has_feature(CPU_FTR_ARCH_300)) { if (is_event_marked(event) || (*mmcra & MMCRA_SAMPLE_ENABLE)) *mmcra &= MMCRA_SDAR_MODE_NO_UPDATES; - else if (p9_SDAR_MODE(event)) - *mmcra |= p9_SDAR_MODE(event) << MMCRA_SDAR_MODE_SHIFT; + else if (sdar_mod_val(event)) + *mmcra |= sdar_mod_val(event) << MMCRA_SDAR_MODE_SHIFT; else *mmcra |= MMCRA_SDAR_MODE_DCACHE; } else *mmcra |= MMCRA_SDAR_MODE_TLB; } +static int p10_thresh_cmp_val(u64 value) +{ + int exp = 0; + u64 result = value; + + if (!value) + return value; + + /* + * Incase of P10, thresh_cmp value is not part of raw event code + * and provided via attr.config1 parameter. To program threshold in MMCRA, + * take a 18 bit number N and shift right 2 places and increment + * the exponent E by 1 until the upper 10 bits of N are zero. + * Write E to the threshold exponent and write the lower 8 bits of N + * to the threshold mantissa. + * The max threshold that can be written is 261120. + */ + if (cpu_has_feature(CPU_FTR_ARCH_31)) { + if (value > 261120) + value = 261120; + while ((64 - __builtin_clzl(value)) > 8) { + exp++; + value >>= 2; + } + + /* + * Note that it is invalid to write a mantissa with the + * upper 2 bits of mantissa being zero, unless the + * exponent is also zero. + */ + if (!(value & 0xC0) && exp) + result = -1; + else + result = (exp << 8) | value; + } + return result; +} + static u64 thresh_cmp_val(u64 value) { + if (cpu_has_feature(CPU_FTR_ARCH_31)) + value = p10_thresh_cmp_val(value); + + /* + * Since location of threshold compare bits in MMCRA + * is different for p8, using different shift value. + */ if (cpu_has_feature(CPU_FTR_ARCH_300)) return value << p9_MMCRA_THR_CMP_SHIFT; - - return value << MMCRA_THR_CMP_SHIFT; + else + return value << MMCRA_THR_CMP_SHIFT; } static unsigned long combine_from_event(u64 event) @@ -131,10 +186,14 @@ static bool is_thresh_cmp_valid(u64 event) { unsigned int cmp, exp; + if (cpu_has_feature(CPU_FTR_ARCH_31)) + return p10_thresh_cmp_val(event) >= 0; + /* * Check the mantissa upper two bits are not zero, unless the * exponent is also zero. See the THRESH_CMP_MANTISSA doc. */ + cmp = (event >> EVENT_THR_CMP_SHIFT) & EVENT_THR_CMP_MASK; exp = cmp >> 7; @@ -161,36 +220,82 @@ static inline u64 isa207_find_source(u64 idx, u32 sub_idx) /* Nothing to do */ break; case 1: - ret = PH(LVL, L1); + ret = PH(LVL, L1) | LEVEL(L1) | P(SNOOP, HIT); break; case 2: - ret = PH(LVL, L2); + ret = PH(LVL, L2) | LEVEL(L2) | P(SNOOP, HIT); break; case 3: - ret = PH(LVL, L3); + ret = PH(LVL, L3) | LEVEL(L3) | P(SNOOP, HIT); break; case 4: - if (sub_idx <= 1) - ret = PH(LVL, LOC_RAM); - else if (sub_idx > 1 && sub_idx <= 2) - ret = PH(LVL, REM_RAM1); - else - ret = PH(LVL, REM_RAM2); - ret |= P(SNOOP, HIT); + if (cpu_has_feature(CPU_FTR_ARCH_31)) { + ret = P(SNOOP, HIT); + + if (sub_idx == 1) + ret |= PH(LVL, LOC_RAM) | LEVEL(RAM); + else if (sub_idx == 2 || sub_idx == 3) + ret |= P(LVL, HIT) | LEVEL(PMEM); + else if (sub_idx == 4) + ret |= PH(LVL, REM_RAM1) | REM | LEVEL(RAM) | P(HOPS, 2); + else if (sub_idx == 5 || sub_idx == 7) + ret |= P(LVL, HIT) | LEVEL(PMEM) | REM; + else if (sub_idx == 6) + ret |= PH(LVL, REM_RAM2) | REM | LEVEL(RAM) | P(HOPS, 3); + } else { + if (sub_idx <= 1) + ret = PH(LVL, LOC_RAM); + else if (sub_idx > 1 && sub_idx <= 2) + ret = PH(LVL, REM_RAM1); + else + ret = PH(LVL, REM_RAM2); + ret |= P(SNOOP, HIT); + } break; case 5: - ret = PH(LVL, REM_CCE1); - if ((sub_idx == 0) || (sub_idx == 2) || (sub_idx == 4)) - ret |= P(SNOOP, HIT); - else if ((sub_idx == 1) || (sub_idx == 3) || (sub_idx == 5)) - ret |= P(SNOOP, HITM); + if (cpu_has_feature(CPU_FTR_ARCH_31)) { + ret = REM | P(HOPS, 0); + + if (sub_idx == 0 || sub_idx == 4) + ret |= PH(LVL, L2) | LEVEL(L2) | P(SNOOP, HIT); + else if (sub_idx == 1 || sub_idx == 5) + ret |= PH(LVL, L2) | LEVEL(L2) | P(SNOOP, HITM); + else if (sub_idx == 2 || sub_idx == 6) + ret |= PH(LVL, L3) | LEVEL(L3) | P(SNOOP, HIT); + else if (sub_idx == 3 || sub_idx == 7) + ret |= PH(LVL, L3) | LEVEL(L3) | P(SNOOP, HITM); + } else { + if (sub_idx == 0) + ret = PH(LVL, L2) | LEVEL(L2) | REM | P(SNOOP, HIT) | P(HOPS, 0); + else if (sub_idx == 1) + ret = PH(LVL, L2) | LEVEL(L2) | REM | P(SNOOP, HITM) | P(HOPS, 0); + else if (sub_idx == 2 || sub_idx == 4) + ret = PH(LVL, L3) | LEVEL(L3) | REM | P(SNOOP, HIT) | P(HOPS, 0); + else if (sub_idx == 3 || sub_idx == 5) + ret = PH(LVL, L3) | LEVEL(L3) | REM | P(SNOOP, HITM) | P(HOPS, 0); + } break; case 6: - ret = PH(LVL, REM_CCE2); - if ((sub_idx == 0) || (sub_idx == 2)) - ret |= P(SNOOP, HIT); - else if ((sub_idx == 1) || (sub_idx == 3)) - ret |= P(SNOOP, HITM); + if (cpu_has_feature(CPU_FTR_ARCH_31)) { + if (sub_idx == 0) + ret = PH(LVL, REM_CCE1) | LEVEL(ANY_CACHE) | REM | + P(SNOOP, HIT) | P(HOPS, 2); + else if (sub_idx == 1) + ret = PH(LVL, REM_CCE1) | LEVEL(ANY_CACHE) | REM | + P(SNOOP, HITM) | P(HOPS, 2); + else if (sub_idx == 2) + ret = PH(LVL, REM_CCE2) | LEVEL(ANY_CACHE) | REM | + P(SNOOP, HIT) | P(HOPS, 3); + else if (sub_idx == 3) + ret = PH(LVL, REM_CCE2) | LEVEL(ANY_CACHE) | REM | + P(SNOOP, HITM) | P(HOPS, 3); + } else { + ret = PH(LVL, REM_CCE2); + if (sub_idx == 0 || sub_idx == 2) + ret |= P(SNOOP, HIT); + else if (sub_idx == 1 || sub_idx == 3) + ret |= P(SNOOP, HITM); + } break; case 7: ret = PM(LVL, L1); @@ -216,30 +321,84 @@ void isa207_get_mem_data_src(union perf_mem_data_src *dsrc, u32 flags, sier = mfspr(SPRN_SIER); val = (sier & ISA207_SIER_TYPE_MASK) >> ISA207_SIER_TYPE_SHIFT; - if (val == 1 || val == 2) { - idx = (sier & ISA207_SIER_LDST_MASK) >> ISA207_SIER_LDST_SHIFT; - sub_idx = (sier & ISA207_SIER_DATA_SRC_MASK) >> ISA207_SIER_DATA_SRC_SHIFT; + if (val != 1 && val != 2 && !(val == 7 && cpu_has_feature(CPU_FTR_ARCH_31))) + return; - dsrc->val = isa207_find_source(idx, sub_idx); + idx = (sier & ISA207_SIER_LDST_MASK) >> ISA207_SIER_LDST_SHIFT; + sub_idx = (sier & ISA207_SIER_DATA_SRC_MASK) >> ISA207_SIER_DATA_SRC_SHIFT; + + dsrc->val = isa207_find_source(idx, sub_idx); + if (val == 7) { + u64 mmcra; + u32 op_type; + + /* + * Type 0b111 denotes either larx or stcx instruction. Use the + * MMCRA sampling bits [57:59] along with the type value + * to determine the exact instruction type. If the sampling + * criteria is neither load or store, set the type as default + * to NA. + */ + mmcra = mfspr(SPRN_MMCRA); + + op_type = (mmcra >> MMCRA_SAMP_ELIG_SHIFT) & MMCRA_SAMP_ELIG_MASK; + switch (op_type) { + case 5: + dsrc->val |= P(OP, LOAD); + break; + case 7: + dsrc->val |= P(OP, STORE); + break; + default: + dsrc->val |= P(OP, NA); + break; + } + } else { dsrc->val |= (val == 1) ? P(OP, LOAD) : P(OP, STORE); } } -void isa207_get_mem_weight(u64 *weight) +void isa207_get_mem_weight(u64 *weight, u64 type) { + union perf_sample_weight *weight_fields; + u64 weight_lat; u64 mmcra = mfspr(SPRN_MMCRA); u64 exp = MMCRA_THR_CTR_EXP(mmcra); u64 mantissa = MMCRA_THR_CTR_MANT(mmcra); u64 sier = mfspr(SPRN_SIER); u64 val = (sier & ISA207_SIER_TYPE_MASK) >> ISA207_SIER_TYPE_SHIFT; - if (val == 0 || val == 7) - *weight = 0; + if (cpu_has_feature(CPU_FTR_ARCH_31)) + mantissa = P10_MMCRA_THR_CTR_MANT(mmcra); + + if (val == 0 || (val == 7 && !cpu_has_feature(CPU_FTR_ARCH_31))) + weight_lat = 0; else - *weight = mantissa << (2 * exp); + weight_lat = mantissa << (2 * exp); + + /* + * Use 64 bit weight field (full) if sample type is + * WEIGHT. + * + * if sample type is WEIGHT_STRUCT: + * - store memory latency in the lower 32 bits. + * - For ISA v3.1, use remaining two 16 bit fields of + * perf_sample_weight to store cycle counter values + * from sier2. + */ + weight_fields = (union perf_sample_weight *)weight; + if (type & PERF_SAMPLE_WEIGHT) + weight_fields->full = weight_lat; + else { + weight_fields->var1_dw = (u32)weight_lat; + if (cpu_has_feature(CPU_FTR_ARCH_31)) { + weight_fields->var2_w = P10_SIER2_FINISH_CYC(mfspr(SPRN_SIER2)); + weight_fields->var3_w = P10_SIER2_DISPATCH_CYC(mfspr(SPRN_SIER2)); + } + } } -int isa207_get_constraint(u64 event, unsigned long *maskp, unsigned long *valp) +int isa207_get_constraint(u64 event, unsigned long *maskp, unsigned long *valp, u64 event_config1) { unsigned int unit, pmc, cache, ebb; unsigned long mask, value; @@ -251,7 +410,12 @@ int isa207_get_constraint(u64 event, unsigned long *maskp, unsigned long *valp) pmc = (event >> EVENT_PMC_SHIFT) & EVENT_PMC_MASK; unit = (event >> EVENT_UNIT_SHIFT) & EVENT_UNIT_MASK; - cache = (event >> EVENT_CACHE_SEL_SHIFT) & EVENT_CACHE_SEL_MASK; + if (cpu_has_feature(CPU_FTR_ARCH_31)) + cache = (event >> EVENT_CACHE_SEL_SHIFT) & + p10_EVENT_CACHE_SEL_MASK; + else + cache = (event >> EVENT_CACHE_SEL_SHIFT) & + EVENT_CACHE_SEL_MASK; ebb = (event >> EVENT_EBB_SHIFT) & EVENT_EBB_MASK; if (pmc) { @@ -269,6 +433,15 @@ int isa207_get_constraint(u64 event, unsigned long *maskp, unsigned long *valp) mask |= CNST_PMC_MASK(pmc); value |= CNST_PMC_VAL(pmc); + + /* + * PMC5 and PMC6 are used to count cycles and instructions and + * they do not support most of the constraint bits. Add a check + * to exclude PMC5/6 from most of the constraints except for + * EBB/BHRB. + */ + if (pmc >= 5) + goto ebb_bhrb; } if (pmc <= 4) { @@ -283,7 +456,12 @@ int isa207_get_constraint(u64 event, unsigned long *maskp, unsigned long *valp) } if (unit >= 6 && unit <= 9) { - if (cpu_has_feature(CPU_FTR_ARCH_300)) { + if (cpu_has_feature(CPU_FTR_ARCH_31)) { + if (unit == 6) { + mask |= CNST_L2L3_GROUP_MASK; + value |= CNST_L2L3_GROUP_VAL(event >> p10_L2L3_EVENT_SHIFT); + } + } else if (cpu_has_feature(CPU_FTR_ARCH_300)) { mask |= CNST_CACHE_GROUP_MASK; value |= CNST_CACHE_GROUP_VAL(event & 0xff); @@ -308,16 +486,30 @@ int isa207_get_constraint(u64 event, unsigned long *maskp, unsigned long *valp) value |= CNST_L1_QUAL_VAL(cache); } + if (cpu_has_feature(CPU_FTR_ARCH_31)) { + mask |= CNST_RADIX_SCOPE_GROUP_MASK; + value |= CNST_RADIX_SCOPE_GROUP_VAL(event >> p10_EVENT_RADIX_SCOPE_QUAL_SHIFT); + } + if (is_event_marked(event)) { mask |= CNST_SAMPLE_MASK; value |= CNST_SAMPLE_VAL(event >> EVENT_SAMPLE_SHIFT); } - if (cpu_has_feature(CPU_FTR_ARCH_300)) { + if (cpu_has_feature(CPU_FTR_ARCH_31)) { + if (event_is_threshold(event) && is_thresh_cmp_valid(event_config1)) { + mask |= CNST_THRESH_CTL_SEL_MASK; + value |= CNST_THRESH_CTL_SEL_VAL(event >> EVENT_THRESH_SHIFT); + mask |= p10_CNST_THRESH_CMP_MASK; + value |= p10_CNST_THRESH_CMP_VAL(p10_thresh_cmp_val(event_config1)); + } else if (event_is_threshold(event)) + return -1; + } else if (cpu_has_feature(CPU_FTR_ARCH_300)) { if (event_is_threshold(event) && is_thresh_cmp_valid(event)) { mask |= CNST_THRESH_MASK; value |= CNST_THRESH_VAL(event >> EVENT_THRESH_SHIFT); - } + } else if (event_is_threshold(event)) + return -1; } else { /* * Special case for PM_MRK_FAB_RSP_MATCH and PM_MRK_FAB_RSP_MATCH_CYC, @@ -335,6 +527,7 @@ int isa207_get_constraint(u64 event, unsigned long *maskp, unsigned long *valp) } } +ebb_bhrb: if (!pmc && ebb) /* EBB events must specify the PMC */ return -1; @@ -353,8 +546,8 @@ int isa207_get_constraint(u64 event, unsigned long *maskp, unsigned long *valp) * EBB events are pinned & exclusive, so this should never actually * hit, but we leave it as a fallback in case. */ - mask |= CNST_EBB_VAL(ebb); - value |= CNST_EBB_MASK; + mask |= CNST_EBB_MASK; + value |= CNST_EBB_VAL(ebb); *maskp = mask; *valp = value; @@ -363,10 +556,11 @@ int isa207_get_constraint(u64 event, unsigned long *maskp, unsigned long *valp) } int isa207_compute_mmcr(u64 event[], int n_ev, - unsigned int hwc[], unsigned long mmcr[], - struct perf_event *pevents[]) + unsigned int hwc[], struct mmcr_regs *mmcr, + struct perf_event *pevents[], u32 flags) { unsigned long mmcra, mmcr1, mmcr2, unit, combine, psel, cache, val; + unsigned long mmcr3; unsigned int pmc, pmc_inuse; int i; @@ -379,7 +573,14 @@ int isa207_compute_mmcr(u64 event[], int n_ev, pmc_inuse |= 1 << pmc; } - mmcra = mmcr1 = mmcr2 = 0; + mmcra = mmcr1 = mmcr2 = mmcr3 = 0; + + /* + * Disable bhrb unless explicitly requested + * by setting MMCRA (BHRBRD) bit. + */ + if (cpu_has_feature(CPU_FTR_ARCH_31)) + mmcra |= MMCRA_BHRB_DISABLE; /* Second pass: assign PMCs, set all MMCR1 fields */ for (i = 0; i < n_ev; ++i) { @@ -416,6 +617,13 @@ int isa207_compute_mmcr(u64 event[], int n_ev, } } + /* Set RADIX_SCOPE_QUAL bit */ + if (cpu_has_feature(CPU_FTR_ARCH_31)) { + val = (event[i] >> p10_EVENT_RADIX_SCOPE_QUAL_SHIFT) & + p10_EVENT_RADIX_SCOPE_QUAL_MASK; + mmcr1 |= val << p10_MMCR1_RADIX_SCOPE_QUAL_SHIFT; + } + if (is_event_marked(event[i])) { mmcra |= MMCRA_SAMPLE_ENABLE; @@ -438,8 +646,21 @@ int isa207_compute_mmcr(u64 event[], int n_ev, mmcra |= val << MMCRA_THR_CTL_SHIFT; val = (event[i] >> EVENT_THR_SEL_SHIFT) & EVENT_THR_SEL_MASK; mmcra |= val << MMCRA_THR_SEL_SHIFT; - val = (event[i] >> EVENT_THR_CMP_SHIFT) & EVENT_THR_CMP_MASK; - mmcra |= thresh_cmp_val(val); + if (!cpu_has_feature(CPU_FTR_ARCH_31)) { + val = (event[i] >> EVENT_THR_CMP_SHIFT) & + EVENT_THR_CMP_MASK; + mmcra |= thresh_cmp_val(val); + } else if (flags & PPMU_HAS_ATTR_CONFIG1) { + val = (pevents[i]->attr.config1 >> p10_EVENT_THR_CMP_SHIFT) & + p10_EVENT_THR_CMP_MASK; + mmcra |= thresh_cmp_val(val); + } + } + + if (cpu_has_feature(CPU_FTR_ARCH_31) && (unit == 6)) { + val = (event[i] >> p10_L2L3_EVENT_SHIFT) & + p10_EVENT_L2L3_SEL_MASK; + mmcr2 |= val << p10_L2L3_SEL_SHIFT; } if (event[i] & EVENT_WANTS_BHRB) { @@ -447,6 +668,11 @@ int isa207_compute_mmcr(u64 event[], int n_ev, mmcra |= val << MMCRA_IFM_SHIFT; } + /* set MMCRA (BHRBRD) to 0 if there is user request for BHRB */ + if (cpu_has_feature(CPU_FTR_ARCH_31) && + (has_branch_stack(pevents[i]) || (event[i] & EVENT_WANTS_BHRB))) + mmcra &= ~MMCRA_BHRB_DISABLE; + if (pevents[i]->attr.exclude_user) mmcr2 |= MMCR2_FCP(pmc); @@ -460,34 +686,54 @@ int isa207_compute_mmcr(u64 event[], int n_ev, mmcr2 |= MMCR2_FCS(pmc); } + if (pevents[i]->attr.exclude_idle) + mmcr2 |= MMCR2_FCWAIT(pmc); + + if (cpu_has_feature(CPU_FTR_ARCH_31)) { + if (pmc <= 4) { + val = (event[i] >> p10_EVENT_MMCR3_SHIFT) & + p10_EVENT_MMCR3_MASK; + mmcr3 |= val << MMCR3_SHIFT(pmc); + } + } + hwc[i] = pmc - 1; } /* Return MMCRx values */ - mmcr[0] = 0; + mmcr->mmcr0 = 0; /* pmc_inuse is 1-based */ if (pmc_inuse & 2) - mmcr[0] = MMCR0_PMC1CE; + mmcr->mmcr0 = MMCR0_PMC1CE; if (pmc_inuse & 0x7c) - mmcr[0] |= MMCR0_PMCjCE; + mmcr->mmcr0 |= MMCR0_PMCjCE; /* If we're not using PMC 5 or 6, freeze them */ if (!(pmc_inuse & 0x60)) - mmcr[0] |= MMCR0_FC56; + mmcr->mmcr0 |= MMCR0_FC56; + + /* + * Set mmcr0 (PMCCEXT) for p10 which + * will restrict access to group B registers + * when MMCR0 PMCC=0b00. + */ + if (cpu_has_feature(CPU_FTR_ARCH_31)) + mmcr->mmcr0 |= MMCR0_PMCCEXT; - mmcr[1] = mmcr1; - mmcr[2] = mmcra; - mmcr[3] = mmcr2; + mmcr->mmcr1 = mmcr1; + mmcr->mmcra = mmcra; + mmcr->mmcr2 = mmcr2; + mmcr->mmcr3 = mmcr3; return 0; } -void isa207_disable_pmc(unsigned int pmc, unsigned long mmcr[]) +void isa207_disable_pmc(unsigned int pmc, struct mmcr_regs *mmcr) { if (pmc <= 3) - mmcr[1] &= ~(0xffUL << MMCR1_PMCSEL_SHIFT(pmc + 1)); + mmcr->mmcr1 &= ~(0xffUL << MMCR1_PMCSEL_SHIFT(pmc + 1)); } static int find_alternative(u64 event, const unsigned int ev_alt[][MAX_ALT], int size) @@ -550,3 +796,45 @@ int isa207_get_alternatives(u64 event, u64 alt[], int size, unsigned int flags, return num_alt; } + +int isa3XX_check_attr_config(struct perf_event *ev) +{ + u64 val, sample_mode; + u64 event = ev->attr.config; + + val = (event >> EVENT_SAMPLE_SHIFT) & EVENT_SAMPLE_MASK; + sample_mode = val & 0x3; + + /* + * MMCRA[61:62] is Random Sampling Mode (SM). + * value of 0b11 is reserved. + */ + if (sample_mode == 0x3) + return -EINVAL; + + /* + * Check for all reserved value + * Source: Performance Monitoring Unit User Guide + */ + switch (val) { + case 0x5: + case 0x9: + case 0xD: + case 0x19: + case 0x1D: + case 0x1A: + case 0x1E: + return -EINVAL; + } + + /* + * MMCRA[48:51]/[52:55]) Threshold Start/Stop + * Events Selection. + * 0b11110000/0b00001111 is reserved. + */ + val = (event >> EVENT_THR_CTL_SHIFT) & EVENT_THR_CTL_MASK; + if (((val & 0xF0) == 0xF0) || ((val & 0xF) == 0xF)) + return -EINVAL; + + return 0; +} diff --git a/arch/powerpc/perf/isa207-common.h b/arch/powerpc/perf/isa207-common.h index 63fd4f3f6013..f594fa6580d1 100644 --- a/arch/powerpc/perf/isa207-common.h +++ b/arch/powerpc/perf/isa207-common.h @@ -13,6 +13,8 @@ #include <asm/firmware.h> #include <asm/cputable.h> +#include "internal.h" + #define EVENT_EBB_MASK 1ull #define EVENT_EBB_SHIFT PERF_EVENT_CONFIG_EBB_SHIFT #define EVENT_BHRB_MASK 1ull @@ -87,20 +89,53 @@ EVENT_LINUX_MASK | \ EVENT_PSEL_MASK)) +/* Contants to support power10 raw encoding format */ +#define p10_SDAR_MODE_SHIFT 22 +#define p10_SDAR_MODE_MASK 0x3ull +#define p10_SDAR_MODE(v) (((v) >> p10_SDAR_MODE_SHIFT) & \ + p10_SDAR_MODE_MASK) +#define p10_EVENT_L2L3_SEL_MASK 0x1f +#define p10_L2L3_SEL_SHIFT 3 +#define p10_L2L3_EVENT_SHIFT 40 +#define p10_EVENT_THRESH_MASK 0xffffull +#define p10_EVENT_CACHE_SEL_MASK 0x3ull +#define p10_EVENT_MMCR3_MASK 0x7fffull +#define p10_EVENT_MMCR3_SHIFT 45 +#define p10_EVENT_RADIX_SCOPE_QUAL_SHIFT 9 +#define p10_EVENT_RADIX_SCOPE_QUAL_MASK 0x1 +#define p10_MMCR1_RADIX_SCOPE_QUAL_SHIFT 45 + +/* Event Threshold Compare bit constant for power10 in config1 attribute */ +#define p10_EVENT_THR_CMP_SHIFT 0 +#define p10_EVENT_THR_CMP_MASK 0x3FFFFull + +#define p10_EVENT_VALID_MASK \ + ((p10_SDAR_MODE_MASK << p10_SDAR_MODE_SHIFT | \ + (p10_EVENT_THRESH_MASK << EVENT_THRESH_SHIFT) | \ + (EVENT_SAMPLE_MASK << EVENT_SAMPLE_SHIFT) | \ + (p10_EVENT_CACHE_SEL_MASK << EVENT_CACHE_SEL_SHIFT) | \ + (EVENT_PMC_MASK << EVENT_PMC_SHIFT) | \ + (EVENT_UNIT_MASK << EVENT_UNIT_SHIFT) | \ + (p9_EVENT_COMBINE_MASK << p9_EVENT_COMBINE_SHIFT) | \ + (p10_EVENT_MMCR3_MASK << p10_EVENT_MMCR3_SHIFT) | \ + (EVENT_MARKED_MASK << EVENT_MARKED_SHIFT) | \ + (p10_EVENT_RADIX_SCOPE_QUAL_MASK << p10_EVENT_RADIX_SCOPE_QUAL_SHIFT) | \ + EVENT_LINUX_MASK | \ + EVENT_PSEL_MASK)) /* * Layout of constraint bits: * * 60 56 52 48 44 40 36 32 * | - - - - | - - - - | - - - - | - - - - | - - - - | - - - - | - - - - | - - - - | * [ fab_match ] [ thresh_cmp ] [ thresh_ctl ] [ ] - * | - * thresh_sel -* + * | | + * [ thresh_cmp bits for p10] thresh_sel -* * * 28 24 20 16 12 8 4 0 * | - - - - | - - - - | - - - - | - - - - | - - - - | - - - - | - - - - | - - - - | - * [ ] | [ ] [ sample ] [ ] [6] [5] [4] [3] [2] [1] - * | | | | - * BHRB IFM -* | | | Count of events for each PMC. + * [ ] | [ ] | [ sample ] [ ] [6] [5] [4] [3] [2] [1] + * | | | | | + * BHRB IFM -* | | |*radix_scope | Count of events for each PMC. * EBB -* | | p1, p2, p3, p4, p5, p6. * L1 I/D qualifier -* | * nc - number of counters -* @@ -118,6 +153,12 @@ #define CNST_THRESH_VAL(v) (((v) & EVENT_THRESH_MASK) << 32) #define CNST_THRESH_MASK CNST_THRESH_VAL(EVENT_THRESH_MASK) +#define CNST_THRESH_CTL_SEL_VAL(v) (((v) & 0x7ffull) << 32) +#define CNST_THRESH_CTL_SEL_MASK CNST_THRESH_CTL_SEL_VAL(0x7ff) + +#define p10_CNST_THRESH_CMP_VAL(v) (((v) & 0x7ffull) << 43) +#define p10_CNST_THRESH_CMP_MASK p10_CNST_THRESH_CMP_VAL(0x7ff) + #define CNST_EBB_VAL(v) (((v) & EVENT_EBB_MASK) << 24) #define CNST_EBB_MASK CNST_EBB_VAL(EVENT_EBB_MASK) @@ -135,6 +176,12 @@ #define CNST_CACHE_PMC4_VAL (1ull << 54) #define CNST_CACHE_PMC4_MASK CNST_CACHE_PMC4_VAL +#define CNST_L2L3_GROUP_VAL(v) (((v) & 0x1full) << 55) +#define CNST_L2L3_GROUP_MASK CNST_L2L3_GROUP_VAL(0x1f) + +#define CNST_RADIX_SCOPE_GROUP_VAL(v) (((v) & 0x1ull) << 21) +#define CNST_RADIX_SCOPE_GROUP_MASK CNST_RADIX_SCOPE_GROUP_VAL(1) + /* * For NC we are counting up to 4 events. This requires three bits, and we need * the fifth event to overflow and set the 4th bit. To achieve that we bias the @@ -173,6 +220,7 @@ /* Bits in MMCRA for PowerISA v2.07 */ #define MMCRA_SAMP_MODE_SHIFT 1 #define MMCRA_SAMP_ELIG_SHIFT 4 +#define MMCRA_SAMP_ELIG_MASK 7 #define MMCRA_THR_CTL_SHIFT 8 #define MMCRA_THR_SEL_SHIFT 16 #define MMCRA_THR_CMP_SHIFT 32 @@ -191,17 +239,25 @@ #define MMCRA_THR_CTR_EXP(v) (((v) >> MMCRA_THR_CTR_EXP_SHIFT) &\ MMCRA_THR_CTR_EXP_MASK) -/* MMCR1 Threshold Compare bit constant for power9 */ +#define P10_MMCRA_THR_CTR_MANT_MASK 0xFFul +#define P10_MMCRA_THR_CTR_MANT(v) (((v) >> MMCRA_THR_CTR_MANT_SHIFT) &\ + P10_MMCRA_THR_CTR_MANT_MASK) + +/* MMCRA Threshold Compare bit constant for power9 */ #define p9_MMCRA_THR_CMP_SHIFT 45 /* Bits in MMCR2 for PowerISA v2.07 */ #define MMCR2_FCS(pmc) (1ull << (63 - (((pmc) - 1) * 9))) #define MMCR2_FCP(pmc) (1ull << (62 - (((pmc) - 1) * 9))) +#define MMCR2_FCWAIT(pmc) (1ull << (58 - (((pmc) - 1) * 9))) #define MMCR2_FCH(pmc) (1ull << (57 - (((pmc) - 1) * 9))) #define MAX_ALT 2 #define MAX_PMU_COUNTERS 6 +/* Bits in MMCR3 for PowerISA v3.10 */ +#define MMCR3_SHIFT(pmc) (49 - (15 * ((pmc) - 1))) + #define ISA207_SIER_TYPE_SHIFT 15 #define ISA207_SIER_TYPE_MASK (0x7ull << ISA207_SIER_TYPE_SHIFT) @@ -211,19 +267,27 @@ #define ISA207_SIER_DATA_SRC_SHIFT 53 #define ISA207_SIER_DATA_SRC_MASK (0x7ull << ISA207_SIER_DATA_SRC_SHIFT) +/* Bits in SIER2/SIER3 for Power10 */ +#define P10_SIER2_FINISH_CYC(sier2) (((sier2) >> (63 - 37)) & 0x7fful) +#define P10_SIER2_DISPATCH_CYC(sier2) (((sier2) >> (63 - 13)) & 0x7fful) + #define P(a, b) PERF_MEM_S(a, b) #define PH(a, b) (P(LVL, HIT) | P(a, b)) #define PM(a, b) (P(LVL, MISS) | P(a, b)) +#define LEVEL(x) P(LVLNUM, x) +#define REM P(REMOTE, REMOTE) -int isa207_get_constraint(u64 event, unsigned long *maskp, unsigned long *valp); +int isa207_get_constraint(u64 event, unsigned long *maskp, unsigned long *valp, u64 event_config1); int isa207_compute_mmcr(u64 event[], int n_ev, - unsigned int hwc[], unsigned long mmcr[], - struct perf_event *pevents[]); -void isa207_disable_pmc(unsigned int pmc, unsigned long mmcr[]); + unsigned int hwc[], struct mmcr_regs *mmcr, + struct perf_event *pevents[], u32 flags); +void isa207_disable_pmc(unsigned int pmc, struct mmcr_regs *mmcr); int isa207_get_alternatives(u64 event, u64 alt[], int size, unsigned int flags, const unsigned int ev_alt[][MAX_ALT]); void isa207_get_mem_data_src(union perf_mem_data_src *dsrc, u32 flags, struct pt_regs *regs); -void isa207_get_mem_weight(u64 *weight); +void isa207_get_mem_weight(u64 *weight, u64 type); + +int isa3XX_check_attr_config(struct perf_event *ev); #endif diff --git a/arch/powerpc/perf/mpc7450-pmu.c b/arch/powerpc/perf/mpc7450-pmu.c index 4d5ef92511d1..db451b9aac35 100644 --- a/arch/powerpc/perf/mpc7450-pmu.c +++ b/arch/powerpc/perf/mpc7450-pmu.c @@ -148,7 +148,7 @@ static u32 classbits[N_CLASSES - 1][2] = { }; static int mpc7450_get_constraint(u64 event, unsigned long *maskp, - unsigned long *valp) + unsigned long *valp, u64 event_config1 __maybe_unused) { int pmc, class; u32 mask, value; @@ -257,8 +257,9 @@ static const u32 pmcsel_mask[N_COUNTER] = { * Compute MMCR0/1/2 values for a set of events. */ static int mpc7450_compute_mmcr(u64 event[], int n_ev, unsigned int hwc[], - unsigned long mmcr[], - struct perf_event *pevents[]) + struct mmcr_regs *mmcr, + struct perf_event *pevents[], + u32 flags __maybe_unused) { u8 event_index[N_CLASSES][N_COUNTER]; int n_classevent[N_CLASSES]; @@ -321,9 +322,16 @@ static int mpc7450_compute_mmcr(u64 event[], int n_ev, unsigned int hwc[], mmcr0 |= MMCR0_PMCnCE; /* Return MMCRx values */ - mmcr[0] = mmcr0; - mmcr[1] = mmcr1; - mmcr[2] = mmcr2; + mmcr->mmcr0 = mmcr0; + mmcr->mmcr1 = mmcr1; + mmcr->mmcr2 = mmcr2; + /* + * 32-bit doesn't have an MMCRA and uses SPRN_MMCR2 to define + * SPRN_MMCRA. So assign mmcra of cpu_hw_events with `mmcr2` + * value to ensure that any write to this SPRN_MMCRA will + * use mmcr2 value. + */ + mmcr->mmcra = mmcr2; return 0; } @@ -331,12 +339,12 @@ static int mpc7450_compute_mmcr(u64 event[], int n_ev, unsigned int hwc[], * Disable counting by a PMC. * Note that the pmc argument is 0-based here, not 1-based. */ -static void mpc7450_disable_pmc(unsigned int pmc, unsigned long mmcr[]) +static void mpc7450_disable_pmc(unsigned int pmc, struct mmcr_regs *mmcr) { if (pmc <= 1) - mmcr[0] &= ~(pmcsel_mask[pmc] << pmcsel_shift[pmc]); + mmcr->mmcr0 &= ~(pmcsel_mask[pmc] << pmcsel_shift[pmc]); else - mmcr[1] &= ~(pmcsel_mask[pmc] << pmcsel_shift[pmc]); + mmcr->mmcr1 &= ~(pmcsel_mask[pmc] << pmcsel_shift[pmc]); } static int mpc7450_generic_events[] = { @@ -354,7 +362,7 @@ static int mpc7450_generic_events[] = { * 0 means not supported, -1 means nonsensical, other values * are event codes. */ -static int mpc7450_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { +static u64 mpc7450_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { [C(L1D)] = { /* RESULT_ACCESS RESULT_MISS */ [C(OP_READ)] = { 0, 0x225 }, [C(OP_WRITE)] = { 0, 0x227 }, @@ -409,8 +417,9 @@ struct power_pmu mpc7450_pmu = { static int __init init_mpc7450_pmu(void) { - if (!cur_cpu_spec->oprofile_cpu_type || - strcmp(cur_cpu_spec->oprofile_cpu_type, "ppc/7450")) + if (!pvr_version_is(PVR_VER_7450) && !pvr_version_is(PVR_VER_7455) && + !pvr_version_is(PVR_VER_7447) && !pvr_version_is(PVR_VER_7447A) && + !pvr_version_is(PVR_VER_7448)) return -ENODEV; return register_power_pmu(&mpc7450_pmu); diff --git a/arch/powerpc/perf/perf_regs.c b/arch/powerpc/perf/perf_regs.c index a213a0aa5d25..350dccb0143c 100644 --- a/arch/powerpc/perf/perf_regs.c +++ b/arch/powerpc/perf/perf_regs.c @@ -13,9 +13,11 @@ #include <asm/ptrace.h> #include <asm/perf_regs.h> +u64 PERF_REG_EXTENDED_MASK; + #define PT_REGS_OFFSET(id, r) [id] = offsetof(struct pt_regs, r) -#define REG_RESERVED (~((1ULL << PERF_REG_POWERPC_MAX) - 1)) +#define REG_RESERVED (~(PERF_REG_EXTENDED_MASK | PERF_REG_PMU_MASK)) static unsigned int pt_regs_offset[PERF_REG_POWERPC_MAX] = { PT_REGS_OFFSET(PERF_REG_POWERPC_R0, gpr[0]), @@ -69,11 +71,36 @@ static unsigned int pt_regs_offset[PERF_REG_POWERPC_MAX] = { PT_REGS_OFFSET(PERF_REG_POWERPC_MMCRA, dsisr), }; -u64 perf_reg_value(struct pt_regs *regs, int idx) +/* Function to return the extended register values */ +static u64 get_ext_regs_value(int idx) { - if (WARN_ON_ONCE(idx >= PERF_REG_POWERPC_MAX)) - return 0; + switch (idx) { + case PERF_REG_POWERPC_PMC1 ... PERF_REG_POWERPC_PMC6: + return get_pmcs_ext_regs(idx - PERF_REG_POWERPC_PMC1); + case PERF_REG_POWERPC_MMCR0: + return mfspr(SPRN_MMCR0); + case PERF_REG_POWERPC_MMCR1: + return mfspr(SPRN_MMCR1); + case PERF_REG_POWERPC_MMCR2: + return mfspr(SPRN_MMCR2); +#ifdef CONFIG_PPC64 + case PERF_REG_POWERPC_MMCR3: + return mfspr(SPRN_MMCR3); + case PERF_REG_POWERPC_SIER2: + return mfspr(SPRN_SIER2); + case PERF_REG_POWERPC_SIER3: + return mfspr(SPRN_SIER3); + case PERF_REG_POWERPC_SDAR: + return mfspr(SPRN_SDAR); +#endif + case PERF_REG_POWERPC_SIAR: + return mfspr(SPRN_SIAR); + default: return 0; + } +} +u64 perf_reg_value(struct pt_regs *regs, int idx) +{ if (idx == PERF_REG_POWERPC_SIER && (IS_ENABLED(CONFIG_FSL_EMB_PERF_EVENT) || IS_ENABLED(CONFIG_PPC32) || @@ -85,6 +112,16 @@ u64 perf_reg_value(struct pt_regs *regs, int idx) IS_ENABLED(CONFIG_PPC32))) return 0; + if (idx >= PERF_REG_POWERPC_MAX && idx < PERF_REG_EXTENDED_MAX) + return get_ext_regs_value(idx); + + /* + * If the idx is referring to value beyond the + * supported registers, return 0 with a warning + */ + if (WARN_ON_ONCE(idx >= PERF_REG_EXTENDED_MAX)) + return 0; + return regs_get_register(regs, pt_regs_offset[idx]); } @@ -97,17 +134,14 @@ int perf_reg_validate(u64 mask) u64 perf_reg_abi(struct task_struct *task) { -#ifdef CONFIG_PPC64 - if (!test_tsk_thread_flag(task, TIF_32BIT)) - return PERF_SAMPLE_REGS_ABI_64; + if (is_tsk_32bit_task(task)) + return PERF_SAMPLE_REGS_ABI_32; else -#endif - return PERF_SAMPLE_REGS_ABI_32; + return PERF_SAMPLE_REGS_ABI_64; } void perf_get_regs_user(struct perf_regs *regs_user, - struct pt_regs *regs, - struct pt_regs *regs_user_copy) + struct pt_regs *regs) { regs_user->regs = task_pt_regs(current); regs_user->abi = (regs_user->regs) ? perf_reg_abi(current) : diff --git a/arch/powerpc/perf/power10-events-list.h b/arch/powerpc/perf/power10-events-list.h new file mode 100644 index 000000000000..564f14097f07 --- /dev/null +++ b/arch/powerpc/perf/power10-events-list.h @@ -0,0 +1,79 @@ +/* SPDX-License-Identifier: GPL-2.0-or-later */ +/* + * Performance counter support for POWER10 processors. + * + * Copyright 2020 Madhavan Srinivasan, IBM Corporation. + * Copyright 2020 Athira Rajeev, IBM Corporation. + */ + +/* + * Power10 event codes. + */ +EVENT(PM_CYC, 0x600f4); +EVENT(PM_DISP_STALL_CYC, 0x100f8); +EVENT(PM_EXEC_STALL, 0x30008); +EVENT(PM_INST_CMPL, 0x500fa); +EVENT(PM_BR_CMPL, 0x4d05e); +EVENT(PM_BR_MPRED_CMPL, 0x400f6); +EVENT(PM_BR_FIN, 0x2f04a); +EVENT(PM_MPRED_BR_FIN, 0x3e098); +EVENT(PM_LD_DEMAND_MISS_L1_FIN, 0x400f0); + +/* All L1 D cache load references counted at finish, gated by reject */ +EVENT(PM_LD_REF_L1, 0x100fc); +/* Load Missed L1 */ +EVENT(PM_LD_MISS_L1, 0x3e054); +/* Store Missed L1 */ +EVENT(PM_ST_MISS_L1, 0x300f0); +/* L1 cache data prefetches */ +EVENT(PM_LD_PREFETCH_CACHE_LINE_MISS, 0x1002c); +/* Demand iCache Miss */ +EVENT(PM_L1_ICACHE_MISS, 0x200fc); +/* Instruction fetches from L1 */ +EVENT(PM_INST_FROM_L1, 0x04080); +/* Instruction Demand sectors wriittent into IL1 */ +EVENT(PM_INST_FROM_L1MISS, 0x03f00000001c040); +/* Instruction prefetch written into IL1 */ +EVENT(PM_IC_PREF_REQ, 0x040a0); +/* The data cache was reloaded from local core's L3 due to a demand load */ +EVENT(PM_DATA_FROM_L3, 0x01340000001c040); +/* Demand LD - L3 Miss (not L2 hit and not L3 hit) */ +EVENT(PM_DATA_FROM_L3MISS, 0x300fe); +/* All successful D-side store dispatches for this thread */ +EVENT(PM_L2_ST, 0x010000046080); +/* All successful D-side store dispatches for this thread that were L2 Miss */ +EVENT(PM_L2_ST_MISS, 0x26880); +/* Total HW L3 prefetches(Load+store) */ +EVENT(PM_L3_PF_MISS_L3, 0x100000016080); +/* Data PTEG reload */ +EVENT(PM_DTLB_MISS, 0x300fc); +/* ITLB Reloaded */ +EVENT(PM_ITLB_MISS, 0x400fc); + +EVENT(PM_CYC_ALT, 0x0001e); +EVENT(PM_INST_CMPL_ALT, 0x00002); + +/* + * Memory Access Events + * + * Primary PMU event used here is PM_MRK_INST_CMPL (0x401e0) + * To enable capturing of memory profiling, these MMCRA bits + * needs to be programmed and corresponding raw event format + * encoding. + * + * MMCRA bits encoding needed are + * SM (Sampling Mode) + * EM (Eligibility for Random Sampling) + * TECE (Threshold Event Counter Event) + * TS (Threshold Start Event) + * TE (Threshold End Event) + * + * Corresponding Raw Encoding bits: + * sample [EM,SM] + * thresh_sel (TECE) + * thresh start (TS) + * thresh end (TE) + */ + +EVENT(MEM_LOADS, 0x35340401e0); +EVENT(MEM_STORES, 0x353c0401e0); diff --git a/arch/powerpc/perf/power10-pmu.c b/arch/powerpc/perf/power10-pmu.c new file mode 100644 index 000000000000..62a68b6b2d4b --- /dev/null +++ b/arch/powerpc/perf/power10-pmu.c @@ -0,0 +1,663 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * Performance counter support for POWER10 processors. + * + * Copyright 2020 Madhavan Srinivasan, IBM Corporation. + * Copyright 2020 Athira Rajeev, IBM Corporation. + */ + +#define pr_fmt(fmt) "power10-pmu: " fmt + +#include "isa207-common.h" + +/* + * Raw event encoding for Power10: + * + * 60 56 52 48 44 40 36 32 + * | - - - - | - - - - | - - - - | - - - - | - - - - | - - - - | - - - - | - - - - | + * | | [ ] [ src_match ] [ src_mask ] | [ ] [ l2l3_sel ] [ thresh_ctl ] + * | | | | | | + * | | *- IFM (Linux) | | thresh start/stop -* + * | *- BHRB (Linux) | src_sel + * *- EBB (Linux) *invert_bit + * + * 28 24 20 16 12 8 4 0 + * | - - - - | - - - - | - - - - | - - - - | - - - - | - - - - | - - - - | - - - - | + * [ ] [ sample ] [ ] [ ] [ pmc ] [unit ] [ ] | m [ pmcxsel ] + * | | | | | | | + * | | | | | | *- mark + * | | | *- L1/L2/L3 cache_sel | |*-radix_scope_qual + * | | sdar_mode | + * | *- sampling mode for marked events *- combine + * | + * *- thresh_sel + * + * Below uses IBM bit numbering. + * + * MMCR1[x:y] = unit (PMCxUNIT) + * MMCR1[24] = pmc1combine[0] + * MMCR1[25] = pmc1combine[1] + * MMCR1[26] = pmc2combine[0] + * MMCR1[27] = pmc2combine[1] + * MMCR1[28] = pmc3combine[0] + * MMCR1[29] = pmc3combine[1] + * MMCR1[30] = pmc4combine[0] + * MMCR1[31] = pmc4combine[1] + * + * if pmc == 3 and unit == 0 and pmcxsel[0:6] == 0b0101011 + * MMCR1[20:27] = thresh_ctl + * else if pmc == 4 and unit == 0xf and pmcxsel[0:6] == 0b0101001 + * MMCR1[20:27] = thresh_ctl + * else + * MMCRA[48:55] = thresh_ctl (THRESH START/END) + * + * if thresh_sel: + * MMCRA[45:47] = thresh_sel + * + * if l2l3_sel: + * MMCR2[56:60] = l2l3_sel[0:4] + * + * MMCR1[16] = cache_sel[0] + * MMCR1[17] = cache_sel[1] + * MMCR1[18] = radix_scope_qual + * + * if mark: + * MMCRA[63] = 1 (SAMPLE_ENABLE) + * MMCRA[57:59] = sample[0:2] (RAND_SAMP_ELIG) + * MMCRA[61:62] = sample[3:4] (RAND_SAMP_MODE) + * + * if EBB and BHRB: + * MMCRA[32:33] = IFM + * + * MMCRA[SDAR_MODE] = sdar_mode[0:1] + */ + +/* + * Some power10 event codes. + */ +#define EVENT(_name, _code) enum{_name = _code} + +#include "power10-events-list.h" + +#undef EVENT + +/* MMCRA IFM bits - POWER10 */ +#define POWER10_MMCRA_IFM1 0x0000000040000000UL +#define POWER10_MMCRA_IFM2 0x0000000080000000UL +#define POWER10_MMCRA_IFM3 0x00000000C0000000UL +#define POWER10_MMCRA_BHRB_MASK 0x00000000C0000000UL + +extern u64 PERF_REG_EXTENDED_MASK; + +/* Table of alternatives, sorted by column 0 */ +static const unsigned int power10_event_alternatives[][MAX_ALT] = { + { PM_INST_CMPL_ALT, PM_INST_CMPL }, + { PM_CYC_ALT, PM_CYC }, +}; + +static int power10_get_alternatives(u64 event, unsigned int flags, u64 alt[]) +{ + int num_alt = 0; + + num_alt = isa207_get_alternatives(event, alt, + ARRAY_SIZE(power10_event_alternatives), flags, + power10_event_alternatives); + + return num_alt; +} + +static int power10_check_attr_config(struct perf_event *ev) +{ + u64 val; + u64 event = ev->attr.config; + + val = (event >> EVENT_SAMPLE_SHIFT) & EVENT_SAMPLE_MASK; + if (val == 0x10 || isa3XX_check_attr_config(ev)) + return -EINVAL; + + return 0; +} + +GENERIC_EVENT_ATTR(cpu-cycles, PM_CYC); +GENERIC_EVENT_ATTR(instructions, PM_INST_CMPL); +GENERIC_EVENT_ATTR(branch-instructions, PM_BR_CMPL); +GENERIC_EVENT_ATTR(branch-misses, PM_BR_MPRED_CMPL); +GENERIC_EVENT_ATTR(cache-references, PM_LD_REF_L1); +GENERIC_EVENT_ATTR(cache-misses, PM_LD_MISS_L1); +GENERIC_EVENT_ATTR(mem-loads, MEM_LOADS); +GENERIC_EVENT_ATTR(mem-stores, MEM_STORES); +GENERIC_EVENT_ATTR(branch-instructions, PM_BR_FIN); +GENERIC_EVENT_ATTR(branch-misses, PM_MPRED_BR_FIN); +GENERIC_EVENT_ATTR(cache-misses, PM_LD_DEMAND_MISS_L1_FIN); + +CACHE_EVENT_ATTR(L1-dcache-load-misses, PM_LD_MISS_L1); +CACHE_EVENT_ATTR(L1-dcache-loads, PM_LD_REF_L1); +CACHE_EVENT_ATTR(L1-dcache-prefetches, PM_LD_PREFETCH_CACHE_LINE_MISS); +CACHE_EVENT_ATTR(L1-dcache-store-misses, PM_ST_MISS_L1); +CACHE_EVENT_ATTR(L1-icache-load-misses, PM_L1_ICACHE_MISS); +CACHE_EVENT_ATTR(L1-icache-loads, PM_INST_FROM_L1); +CACHE_EVENT_ATTR(L1-icache-prefetches, PM_IC_PREF_REQ); +CACHE_EVENT_ATTR(LLC-load-misses, PM_DATA_FROM_L3MISS); +CACHE_EVENT_ATTR(LLC-loads, PM_DATA_FROM_L3); +CACHE_EVENT_ATTR(LLC-prefetches, PM_L3_PF_MISS_L3); +CACHE_EVENT_ATTR(LLC-store-misses, PM_L2_ST_MISS); +CACHE_EVENT_ATTR(LLC-stores, PM_L2_ST); +CACHE_EVENT_ATTR(branch-load-misses, PM_BR_MPRED_CMPL); +CACHE_EVENT_ATTR(branch-loads, PM_BR_CMPL); +CACHE_EVENT_ATTR(dTLB-load-misses, PM_DTLB_MISS); +CACHE_EVENT_ATTR(iTLB-load-misses, PM_ITLB_MISS); + +static struct attribute *power10_events_attr_dd1[] = { + GENERIC_EVENT_PTR(PM_CYC), + GENERIC_EVENT_PTR(PM_INST_CMPL), + GENERIC_EVENT_PTR(PM_BR_CMPL), + GENERIC_EVENT_PTR(PM_BR_MPRED_CMPL), + GENERIC_EVENT_PTR(PM_LD_REF_L1), + GENERIC_EVENT_PTR(PM_LD_MISS_L1), + GENERIC_EVENT_PTR(MEM_LOADS), + GENERIC_EVENT_PTR(MEM_STORES), + CACHE_EVENT_PTR(PM_LD_MISS_L1), + CACHE_EVENT_PTR(PM_LD_REF_L1), + CACHE_EVENT_PTR(PM_LD_PREFETCH_CACHE_LINE_MISS), + CACHE_EVENT_PTR(PM_ST_MISS_L1), + CACHE_EVENT_PTR(PM_L1_ICACHE_MISS), + CACHE_EVENT_PTR(PM_INST_FROM_L1), + CACHE_EVENT_PTR(PM_IC_PREF_REQ), + CACHE_EVENT_PTR(PM_DATA_FROM_L3MISS), + CACHE_EVENT_PTR(PM_DATA_FROM_L3), + CACHE_EVENT_PTR(PM_BR_MPRED_CMPL), + CACHE_EVENT_PTR(PM_BR_CMPL), + CACHE_EVENT_PTR(PM_DTLB_MISS), + CACHE_EVENT_PTR(PM_ITLB_MISS), + NULL +}; + +static struct attribute *power10_events_attr[] = { + GENERIC_EVENT_PTR(PM_CYC), + GENERIC_EVENT_PTR(PM_INST_CMPL), + GENERIC_EVENT_PTR(PM_BR_FIN), + GENERIC_EVENT_PTR(PM_MPRED_BR_FIN), + GENERIC_EVENT_PTR(PM_LD_REF_L1), + GENERIC_EVENT_PTR(PM_LD_DEMAND_MISS_L1_FIN), + GENERIC_EVENT_PTR(MEM_LOADS), + GENERIC_EVENT_PTR(MEM_STORES), + CACHE_EVENT_PTR(PM_LD_MISS_L1), + CACHE_EVENT_PTR(PM_LD_REF_L1), + CACHE_EVENT_PTR(PM_LD_PREFETCH_CACHE_LINE_MISS), + CACHE_EVENT_PTR(PM_ST_MISS_L1), + CACHE_EVENT_PTR(PM_L1_ICACHE_MISS), + CACHE_EVENT_PTR(PM_INST_FROM_L1), + CACHE_EVENT_PTR(PM_IC_PREF_REQ), + CACHE_EVENT_PTR(PM_DATA_FROM_L3MISS), + CACHE_EVENT_PTR(PM_DATA_FROM_L3), + CACHE_EVENT_PTR(PM_L3_PF_MISS_L3), + CACHE_EVENT_PTR(PM_L2_ST_MISS), + CACHE_EVENT_PTR(PM_L2_ST), + CACHE_EVENT_PTR(PM_BR_MPRED_CMPL), + CACHE_EVENT_PTR(PM_BR_CMPL), + CACHE_EVENT_PTR(PM_DTLB_MISS), + CACHE_EVENT_PTR(PM_ITLB_MISS), + NULL +}; + +static const struct attribute_group power10_pmu_events_group_dd1 = { + .name = "events", + .attrs = power10_events_attr_dd1, +}; + +static const struct attribute_group power10_pmu_events_group = { + .name = "events", + .attrs = power10_events_attr, +}; + +PMU_FORMAT_ATTR(event, "config:0-59"); +PMU_FORMAT_ATTR(pmcxsel, "config:0-7"); +PMU_FORMAT_ATTR(mark, "config:8"); +PMU_FORMAT_ATTR(combine, "config:10-11"); +PMU_FORMAT_ATTR(unit, "config:12-15"); +PMU_FORMAT_ATTR(pmc, "config:16-19"); +PMU_FORMAT_ATTR(cache_sel, "config:20-21"); +PMU_FORMAT_ATTR(sdar_mode, "config:22-23"); +PMU_FORMAT_ATTR(sample_mode, "config:24-28"); +PMU_FORMAT_ATTR(thresh_sel, "config:29-31"); +PMU_FORMAT_ATTR(thresh_stop, "config:32-35"); +PMU_FORMAT_ATTR(thresh_start, "config:36-39"); +PMU_FORMAT_ATTR(l2l3_sel, "config:40-44"); +PMU_FORMAT_ATTR(src_sel, "config:45-46"); +PMU_FORMAT_ATTR(invert_bit, "config:47"); +PMU_FORMAT_ATTR(src_mask, "config:48-53"); +PMU_FORMAT_ATTR(src_match, "config:54-59"); +PMU_FORMAT_ATTR(radix_scope, "config:9"); +PMU_FORMAT_ATTR(thresh_cmp, "config1:0-17"); + +static struct attribute *power10_pmu_format_attr[] = { + &format_attr_event.attr, + &format_attr_pmcxsel.attr, + &format_attr_mark.attr, + &format_attr_combine.attr, + &format_attr_unit.attr, + &format_attr_pmc.attr, + &format_attr_cache_sel.attr, + &format_attr_sdar_mode.attr, + &format_attr_sample_mode.attr, + &format_attr_thresh_sel.attr, + &format_attr_thresh_stop.attr, + &format_attr_thresh_start.attr, + &format_attr_l2l3_sel.attr, + &format_attr_src_sel.attr, + &format_attr_invert_bit.attr, + &format_attr_src_mask.attr, + &format_attr_src_match.attr, + &format_attr_radix_scope.attr, + &format_attr_thresh_cmp.attr, + NULL, +}; + +static const struct attribute_group power10_pmu_format_group = { + .name = "format", + .attrs = power10_pmu_format_attr, +}; + +static struct attribute *power10_pmu_caps_attrs[] = { + NULL +}; + +static struct attribute_group power10_pmu_caps_group = { + .name = "caps", + .attrs = power10_pmu_caps_attrs, +}; + +static const struct attribute_group *power10_pmu_attr_groups_dd1[] = { + &power10_pmu_format_group, + &power10_pmu_events_group_dd1, + &power10_pmu_caps_group, + NULL, +}; + +static const struct attribute_group *power10_pmu_attr_groups[] = { + &power10_pmu_format_group, + &power10_pmu_events_group, + &power10_pmu_caps_group, + NULL, +}; + +static int power10_generic_events_dd1[] = { + [PERF_COUNT_HW_CPU_CYCLES] = PM_CYC, + [PERF_COUNT_HW_INSTRUCTIONS] = PM_INST_CMPL, + [PERF_COUNT_HW_BRANCH_INSTRUCTIONS] = PM_BR_CMPL, + [PERF_COUNT_HW_BRANCH_MISSES] = PM_BR_MPRED_CMPL, + [PERF_COUNT_HW_CACHE_REFERENCES] = PM_LD_REF_L1, + [PERF_COUNT_HW_CACHE_MISSES] = PM_LD_MISS_L1, +}; + +static int power10_generic_events[] = { + [PERF_COUNT_HW_CPU_CYCLES] = PM_CYC, + [PERF_COUNT_HW_INSTRUCTIONS] = PM_INST_CMPL, + [PERF_COUNT_HW_BRANCH_INSTRUCTIONS] = PM_BR_FIN, + [PERF_COUNT_HW_BRANCH_MISSES] = PM_MPRED_BR_FIN, + [PERF_COUNT_HW_CACHE_REFERENCES] = PM_LD_REF_L1, + [PERF_COUNT_HW_CACHE_MISSES] = PM_LD_DEMAND_MISS_L1_FIN, +}; + +static u64 power10_bhrb_filter_map(u64 branch_sample_type) +{ + u64 pmu_bhrb_filter = 0; + + /* BHRB and regular PMU events share the same privilege state + * filter configuration. BHRB is always recorded along with a + * regular PMU event. As the privilege state filter is handled + * in the basic PMC configuration of the accompanying regular + * PMU event, we ignore any separate BHRB specific request. + */ + + /* No branch filter requested */ + if (branch_sample_type & PERF_SAMPLE_BRANCH_ANY) + return pmu_bhrb_filter; + + /* Invalid branch filter options - HW does not support */ + if (branch_sample_type & PERF_SAMPLE_BRANCH_ANY_RETURN) + return -1; + + if (branch_sample_type & PERF_SAMPLE_BRANCH_IND_CALL) { + pmu_bhrb_filter |= POWER10_MMCRA_IFM2; + return pmu_bhrb_filter; + } + + if (branch_sample_type & PERF_SAMPLE_BRANCH_COND) { + pmu_bhrb_filter |= POWER10_MMCRA_IFM3; + return pmu_bhrb_filter; + } + + if (branch_sample_type & PERF_SAMPLE_BRANCH_CALL) + return -1; + + if (branch_sample_type & PERF_SAMPLE_BRANCH_ANY_CALL) { + pmu_bhrb_filter |= POWER10_MMCRA_IFM1; + return pmu_bhrb_filter; + } + + /* Every thing else is unsupported */ + return -1; +} + +static void power10_config_bhrb(u64 pmu_bhrb_filter) +{ + pmu_bhrb_filter &= POWER10_MMCRA_BHRB_MASK; + + /* Enable BHRB filter in PMU */ + mtspr(SPRN_MMCRA, (mfspr(SPRN_MMCRA) | pmu_bhrb_filter)); +} + +#define C(x) PERF_COUNT_HW_CACHE_##x + +/* + * Table of generalized cache-related events. + * 0 means not supported, -1 means nonsensical, other values + * are event codes. + */ +static u64 power10_cache_events_dd1[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { + [C(L1D)] = { + [C(OP_READ)] = { + [C(RESULT_ACCESS)] = PM_LD_REF_L1, + [C(RESULT_MISS)] = PM_LD_MISS_L1, + }, + [C(OP_WRITE)] = { + [C(RESULT_ACCESS)] = 0, + [C(RESULT_MISS)] = PM_ST_MISS_L1, + }, + [C(OP_PREFETCH)] = { + [C(RESULT_ACCESS)] = PM_LD_PREFETCH_CACHE_LINE_MISS, + [C(RESULT_MISS)] = 0, + }, + }, + [C(L1I)] = { + [C(OP_READ)] = { + [C(RESULT_ACCESS)] = PM_INST_FROM_L1, + [C(RESULT_MISS)] = PM_L1_ICACHE_MISS, + }, + [C(OP_WRITE)] = { + [C(RESULT_ACCESS)] = PM_INST_FROM_L1MISS, + [C(RESULT_MISS)] = -1, + }, + [C(OP_PREFETCH)] = { + [C(RESULT_ACCESS)] = PM_IC_PREF_REQ, + [C(RESULT_MISS)] = 0, + }, + }, + [C(LL)] = { + [C(OP_READ)] = { + [C(RESULT_ACCESS)] = PM_DATA_FROM_L3, + [C(RESULT_MISS)] = PM_DATA_FROM_L3MISS, + }, + [C(OP_WRITE)] = { + [C(RESULT_ACCESS)] = -1, + [C(RESULT_MISS)] = -1, + }, + [C(OP_PREFETCH)] = { + [C(RESULT_ACCESS)] = -1, + [C(RESULT_MISS)] = 0, + }, + }, + [C(DTLB)] = { + [C(OP_READ)] = { + [C(RESULT_ACCESS)] = 0, + [C(RESULT_MISS)] = PM_DTLB_MISS, + }, + [C(OP_WRITE)] = { + [C(RESULT_ACCESS)] = -1, + [C(RESULT_MISS)] = -1, + }, + [C(OP_PREFETCH)] = { + [C(RESULT_ACCESS)] = -1, + [C(RESULT_MISS)] = -1, + }, + }, + [C(ITLB)] = { + [C(OP_READ)] = { + [C(RESULT_ACCESS)] = 0, + [C(RESULT_MISS)] = PM_ITLB_MISS, + }, + [C(OP_WRITE)] = { + [C(RESULT_ACCESS)] = -1, + [C(RESULT_MISS)] = -1, + }, + [C(OP_PREFETCH)] = { + [C(RESULT_ACCESS)] = -1, + [C(RESULT_MISS)] = -1, + }, + }, + [C(BPU)] = { + [C(OP_READ)] = { + [C(RESULT_ACCESS)] = PM_BR_CMPL, + [C(RESULT_MISS)] = PM_BR_MPRED_CMPL, + }, + [C(OP_WRITE)] = { + [C(RESULT_ACCESS)] = -1, + [C(RESULT_MISS)] = -1, + }, + [C(OP_PREFETCH)] = { + [C(RESULT_ACCESS)] = -1, + [C(RESULT_MISS)] = -1, + }, + }, + [C(NODE)] = { + [C(OP_READ)] = { + [C(RESULT_ACCESS)] = -1, + [C(RESULT_MISS)] = -1, + }, + [C(OP_WRITE)] = { + [C(RESULT_ACCESS)] = -1, + [C(RESULT_MISS)] = -1, + }, + [C(OP_PREFETCH)] = { + [C(RESULT_ACCESS)] = -1, + [C(RESULT_MISS)] = -1, + }, + }, +}; + +static u64 power10_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { + [C(L1D)] = { + [C(OP_READ)] = { + [C(RESULT_ACCESS)] = PM_LD_REF_L1, + [C(RESULT_MISS)] = PM_LD_MISS_L1, + }, + [C(OP_WRITE)] = { + [C(RESULT_ACCESS)] = 0, + [C(RESULT_MISS)] = PM_ST_MISS_L1, + }, + [C(OP_PREFETCH)] = { + [C(RESULT_ACCESS)] = PM_LD_PREFETCH_CACHE_LINE_MISS, + [C(RESULT_MISS)] = 0, + }, + }, + [C(L1I)] = { + [C(OP_READ)] = { + [C(RESULT_ACCESS)] = PM_INST_FROM_L1, + [C(RESULT_MISS)] = PM_L1_ICACHE_MISS, + }, + [C(OP_WRITE)] = { + [C(RESULT_ACCESS)] = PM_INST_FROM_L1MISS, + [C(RESULT_MISS)] = -1, + }, + [C(OP_PREFETCH)] = { + [C(RESULT_ACCESS)] = PM_IC_PREF_REQ, + [C(RESULT_MISS)] = 0, + }, + }, + [C(LL)] = { + [C(OP_READ)] = { + [C(RESULT_ACCESS)] = PM_DATA_FROM_L3, + [C(RESULT_MISS)] = PM_DATA_FROM_L3MISS, + }, + [C(OP_WRITE)] = { + [C(RESULT_ACCESS)] = PM_L2_ST, + [C(RESULT_MISS)] = PM_L2_ST_MISS, + }, + [C(OP_PREFETCH)] = { + [C(RESULT_ACCESS)] = PM_L3_PF_MISS_L3, + [C(RESULT_MISS)] = 0, + }, + }, + [C(DTLB)] = { + [C(OP_READ)] = { + [C(RESULT_ACCESS)] = 0, + [C(RESULT_MISS)] = PM_DTLB_MISS, + }, + [C(OP_WRITE)] = { + [C(RESULT_ACCESS)] = -1, + [C(RESULT_MISS)] = -1, + }, + [C(OP_PREFETCH)] = { + [C(RESULT_ACCESS)] = -1, + [C(RESULT_MISS)] = -1, + }, + }, + [C(ITLB)] = { + [C(OP_READ)] = { + [C(RESULT_ACCESS)] = 0, + [C(RESULT_MISS)] = PM_ITLB_MISS, + }, + [C(OP_WRITE)] = { + [C(RESULT_ACCESS)] = -1, + [C(RESULT_MISS)] = -1, + }, + [C(OP_PREFETCH)] = { + [C(RESULT_ACCESS)] = -1, + [C(RESULT_MISS)] = -1, + }, + }, + [C(BPU)] = { + [C(OP_READ)] = { + [C(RESULT_ACCESS)] = PM_BR_CMPL, + [C(RESULT_MISS)] = PM_BR_MPRED_CMPL, + }, + [C(OP_WRITE)] = { + [C(RESULT_ACCESS)] = -1, + [C(RESULT_MISS)] = -1, + }, + [C(OP_PREFETCH)] = { + [C(RESULT_ACCESS)] = -1, + [C(RESULT_MISS)] = -1, + }, + }, + [C(NODE)] = { + [C(OP_READ)] = { + [C(RESULT_ACCESS)] = -1, + [C(RESULT_MISS)] = -1, + }, + [C(OP_WRITE)] = { + [C(RESULT_ACCESS)] = -1, + [C(RESULT_MISS)] = -1, + }, + [C(OP_PREFETCH)] = { + [C(RESULT_ACCESS)] = -1, + [C(RESULT_MISS)] = -1, + }, + }, +}; + +#undef C + +/* + * Set the MMCR0[CC56RUN] bit to enable counting for + * PMC5 and PMC6 regardless of the state of CTRL[RUN], + * so that we can use counters 5 and 6 as PM_INST_CMPL and + * PM_CYC. + */ +static int power10_compute_mmcr(u64 event[], int n_ev, + unsigned int hwc[], struct mmcr_regs *mmcr, + struct perf_event *pevents[], u32 flags) +{ + int ret; + + ret = isa207_compute_mmcr(event, n_ev, hwc, mmcr, pevents, flags); + if (!ret) + mmcr->mmcr0 |= MMCR0_C56RUN; + return ret; +} + +static struct power_pmu power10_pmu = { + .name = "POWER10", + .n_counter = MAX_PMU_COUNTERS, + .add_fields = ISA207_ADD_FIELDS, + .test_adder = ISA207_TEST_ADDER, + .group_constraint_mask = CNST_CACHE_PMC4_MASK, + .group_constraint_val = CNST_CACHE_PMC4_VAL, + .compute_mmcr = power10_compute_mmcr, + .config_bhrb = power10_config_bhrb, + .bhrb_filter_map = power10_bhrb_filter_map, + .get_constraint = isa207_get_constraint, + .get_alternatives = power10_get_alternatives, + .get_mem_data_src = isa207_get_mem_data_src, + .get_mem_weight = isa207_get_mem_weight, + .disable_pmc = isa207_disable_pmc, + .flags = PPMU_HAS_SIER | PPMU_ARCH_207S | + PPMU_ARCH_31 | PPMU_HAS_ATTR_CONFIG1, + .n_generic = ARRAY_SIZE(power10_generic_events), + .generic_events = power10_generic_events, + .cache_events = &power10_cache_events, + .attr_groups = power10_pmu_attr_groups, + .bhrb_nr = 32, + .capabilities = PERF_PMU_CAP_EXTENDED_REGS, + .check_attr_config = power10_check_attr_config, +}; + +int __init init_power10_pmu(void) +{ + unsigned int pvr; + int rc; + + pvr = mfspr(SPRN_PVR); + if (PVR_VER(pvr) != PVR_POWER10) + return -ENODEV; + + /* Add the ppmu flag for power10 DD1 */ + if ((PVR_CFG(pvr) == 1)) + power10_pmu.flags |= PPMU_P10_DD1; + + /* Set the PERF_REG_EXTENDED_MASK here */ + PERF_REG_EXTENDED_MASK = PERF_REG_PMU_MASK_31; + + if ((PVR_CFG(pvr) == 1)) { + power10_pmu.generic_events = power10_generic_events_dd1; + power10_pmu.attr_groups = power10_pmu_attr_groups_dd1; + power10_pmu.cache_events = &power10_cache_events_dd1; + } + + rc = register_power_pmu(&power10_pmu); + if (rc) + return rc; + + /* Tell userspace that EBB is supported */ + cur_cpu_spec->cpu_user_features2 |= PPC_FEATURE2_EBB; + + return 0; +} + +static struct power_pmu power11_pmu; + +int __init init_power11_pmu(void) +{ + unsigned int pvr; + int rc; + + pvr = mfspr(SPRN_PVR); + if (PVR_VER(pvr) != PVR_POWER11) + return -ENODEV; + + /* Set the PERF_REG_EXTENDED_MASK here */ + PERF_REG_EXTENDED_MASK = PERF_REG_PMU_MASK_31; + + power11_pmu = power10_pmu; + power11_pmu.name = "Power11"; + + rc = register_power_pmu(&power11_pmu); + if (rc) + return rc; + + /* Tell userspace that EBB is supported */ + cur_cpu_spec->cpu_user_features2 |= PPC_FEATURE2_EBB; + + return 0; +} diff --git a/arch/powerpc/perf/power5+-pmu.c b/arch/powerpc/perf/power5+-pmu.c index f8574547fc6b..b4708ab73145 100644 --- a/arch/powerpc/perf/power5+-pmu.c +++ b/arch/powerpc/perf/power5+-pmu.c @@ -10,6 +10,8 @@ #include <asm/reg.h> #include <asm/cputable.h> +#include "internal.h" + /* * Bits in event code for POWER5+ (POWER5 GS) and POWER5++ (POWER5 GS DD3) */ @@ -130,7 +132,7 @@ static unsigned long unit_cons[PM_LASTUNIT+1][2] = { }; static int power5p_get_constraint(u64 event, unsigned long *maskp, - unsigned long *valp) + unsigned long *valp, u64 event_config1 __maybe_unused) { int pmc, byte, unit, sh; int bit, fmask; @@ -448,7 +450,9 @@ static int power5p_marked_instr_event(u64 event) } static int power5p_compute_mmcr(u64 event[], int n_ev, - unsigned int hwc[], unsigned long mmcr[], struct perf_event *pevents[]) + unsigned int hwc[], struct mmcr_regs *mmcr, + struct perf_event *pevents[], + u32 flags __maybe_unused) { unsigned long mmcr1 = 0; unsigned long mmcra = 0; @@ -586,20 +590,20 @@ static int power5p_compute_mmcr(u64 event[], int n_ev, } /* Return MMCRx values */ - mmcr[0] = 0; + mmcr->mmcr0 = 0; if (pmc_inuse & 1) - mmcr[0] = MMCR0_PMC1CE; + mmcr->mmcr0 = MMCR0_PMC1CE; if (pmc_inuse & 0x3e) - mmcr[0] |= MMCR0_PMCjCE; - mmcr[1] = mmcr1; - mmcr[2] = mmcra; + mmcr->mmcr0 |= MMCR0_PMCjCE; + mmcr->mmcr1 = mmcr1; + mmcr->mmcra = mmcra; return 0; } -static void power5p_disable_pmc(unsigned int pmc, unsigned long mmcr[]) +static void power5p_disable_pmc(unsigned int pmc, struct mmcr_regs *mmcr) { if (pmc <= 3) - mmcr[1] &= ~(0x7fUL << MMCR1_PMCSEL_SH(pmc)); + mmcr->mmcr1 &= ~(0x7fUL << MMCR1_PMCSEL_SH(pmc)); } static int power5p_generic_events[] = { @@ -618,7 +622,7 @@ static int power5p_generic_events[] = { * 0 means not supported, -1 means nonsensical, other values * are event codes. */ -static int power5p_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { +static u64 power5p_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { [C(L1D)] = { /* RESULT_ACCESS RESULT_MISS */ [C(OP_READ)] = { 0x1c10a8, 0x3c1088 }, [C(OP_WRITE)] = { 0x2c10a8, 0xc10c3 }, @@ -673,11 +677,11 @@ static struct power_pmu power5p_pmu = { .cache_events = &power5p_cache_events, }; -int init_power5p_pmu(void) +int __init init_power5p_pmu(void) { - if (!cur_cpu_spec->oprofile_cpu_type || - (strcmp(cur_cpu_spec->oprofile_cpu_type, "ppc64/power5+") - && strcmp(cur_cpu_spec->oprofile_cpu_type, "ppc64/power5++"))) + unsigned int pvr = mfspr(SPRN_PVR); + + if (PVR_VER(pvr) != PVR_POWER5p) return -ENODEV; return register_power_pmu(&power5p_pmu); diff --git a/arch/powerpc/perf/power5-pmu.c b/arch/powerpc/perf/power5-pmu.c index da52ecab7c9a..c6aefd0a1cc8 100644 --- a/arch/powerpc/perf/power5-pmu.c +++ b/arch/powerpc/perf/power5-pmu.c @@ -10,6 +10,8 @@ #include <asm/reg.h> #include <asm/cputable.h> +#include "internal.h" + /* * Bits in event code for POWER5 (not POWER5++) */ @@ -134,7 +136,7 @@ static unsigned long unit_cons[PM_LASTUNIT+1][2] = { }; static int power5_get_constraint(u64 event, unsigned long *maskp, - unsigned long *valp) + unsigned long *valp, u64 event_config1 __maybe_unused) { int pmc, byte, unit, sh; int bit, fmask; @@ -379,7 +381,9 @@ static int power5_marked_instr_event(u64 event) } static int power5_compute_mmcr(u64 event[], int n_ev, - unsigned int hwc[], unsigned long mmcr[], struct perf_event *pevents[]) + unsigned int hwc[], struct mmcr_regs *mmcr, + struct perf_event *pevents[], + u32 flags __maybe_unused) { unsigned long mmcr1 = 0; unsigned long mmcra = MMCRA_SDAR_DCACHE_MISS | MMCRA_SDAR_ERAT_MISS; @@ -528,20 +532,20 @@ static int power5_compute_mmcr(u64 event[], int n_ev, } /* Return MMCRx values */ - mmcr[0] = 0; + mmcr->mmcr0 = 0; if (pmc_inuse & 1) - mmcr[0] = MMCR0_PMC1CE; + mmcr->mmcr0 = MMCR0_PMC1CE; if (pmc_inuse & 0x3e) - mmcr[0] |= MMCR0_PMCjCE; - mmcr[1] = mmcr1; - mmcr[2] = mmcra; + mmcr->mmcr0 |= MMCR0_PMCjCE; + mmcr->mmcr1 = mmcr1; + mmcr->mmcra = mmcra; return 0; } -static void power5_disable_pmc(unsigned int pmc, unsigned long mmcr[]) +static void power5_disable_pmc(unsigned int pmc, struct mmcr_regs *mmcr) { if (pmc <= 3) - mmcr[1] &= ~(0x7fUL << MMCR1_PMCSEL_SH(pmc)); + mmcr->mmcr1 &= ~(0x7fUL << MMCR1_PMCSEL_SH(pmc)); } static int power5_generic_events[] = { @@ -560,7 +564,7 @@ static int power5_generic_events[] = { * 0 means not supported, -1 means nonsensical, other values * are event codes. */ -static int power5_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { +static u64 power5_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { [C(L1D)] = { /* RESULT_ACCESS RESULT_MISS */ [C(OP_READ)] = { 0x4c1090, 0x3c1088 }, [C(OP_WRITE)] = { 0x3c1090, 0xc10c3 }, @@ -614,10 +618,11 @@ static struct power_pmu power5_pmu = { .flags = PPMU_HAS_SSLOT, }; -int init_power5_pmu(void) +int __init init_power5_pmu(void) { - if (!cur_cpu_spec->oprofile_cpu_type || - strcmp(cur_cpu_spec->oprofile_cpu_type, "ppc64/power5")) + unsigned int pvr = mfspr(SPRN_PVR); + + if (PVR_VER(pvr) != PVR_POWER5) return -ENODEV; return register_power_pmu(&power5_pmu); diff --git a/arch/powerpc/perf/power6-pmu.c b/arch/powerpc/perf/power6-pmu.c index 3929cacf72ed..9f720b522e17 100644 --- a/arch/powerpc/perf/power6-pmu.c +++ b/arch/powerpc/perf/power6-pmu.c @@ -10,6 +10,8 @@ #include <asm/reg.h> #include <asm/cputable.h> +#include "internal.h" + /* * Bits in event code for POWER6 */ @@ -171,7 +173,8 @@ static int power6_marked_instr_event(u64 event) * Assign PMC numbers and compute MMCR1 value for a set of events */ static int p6_compute_mmcr(u64 event[], int n_ev, - unsigned int hwc[], unsigned long mmcr[], struct perf_event *pevents[]) + unsigned int hwc[], struct mmcr_regs *mmcr, struct perf_event *pevents[], + u32 flags __maybe_unused) { unsigned long mmcr1 = 0; unsigned long mmcra = MMCRA_SDAR_DCACHE_MISS | MMCRA_SDAR_ERAT_MISS; @@ -243,13 +246,13 @@ static int p6_compute_mmcr(u64 event[], int n_ev, if (pmc < 4) mmcr1 |= (unsigned long)psel << MMCR1_PMCSEL_SH(pmc); } - mmcr[0] = 0; + mmcr->mmcr0 = 0; if (pmc_inuse & 1) - mmcr[0] = MMCR0_PMC1CE; + mmcr->mmcr0 = MMCR0_PMC1CE; if (pmc_inuse & 0xe) - mmcr[0] |= MMCR0_PMCjCE; - mmcr[1] = mmcr1; - mmcr[2] = mmcra; + mmcr->mmcr0 |= MMCR0_PMCjCE; + mmcr->mmcr1 = mmcr1; + mmcr->mmcra = mmcra; return 0; } @@ -264,7 +267,7 @@ static int p6_compute_mmcr(u64 event[], int n_ev, * 32-34 select field: nest (subunit) event selector */ static int p6_get_constraint(u64 event, unsigned long *maskp, - unsigned long *valp) + unsigned long *valp, u64 event_config1 __maybe_unused) { int pmc, byte, sh, subunit; unsigned long mask = 0, value = 0; @@ -332,26 +335,38 @@ static const unsigned int event_alternatives[][MAX_ALT] = { { 0x3000fe, 0x400056 }, /* PM_DATA_FROM_L3MISS */ }; -/* - * This could be made more efficient with a binary search on - * a presorted list, if necessary - */ static int find_alternatives_list(u64 event) { - int i, j; - unsigned int alt; - - for (i = 0; i < ARRAY_SIZE(event_alternatives); ++i) { - if (event < event_alternatives[i][0]) - return -1; - for (j = 0; j < MAX_ALT; ++j) { - alt = event_alternatives[i][j]; - if (!alt || event < alt) - break; - if (event == alt) - return i; - } + const unsigned int presorted_event_table[] = { + 0x0130e8, 0x080080, 0x080088, 0x10000a, 0x10000b, 0x10000d, 0x10000e, + 0x100010, 0x10001a, 0x100026, 0x100054, 0x100056, 0x1000f0, 0x1000f8, + 0x1000fc, 0x200008, 0x20000e, 0x200010, 0x200012, 0x200054, 0x2000f0, + 0x2000f2, 0x2000f4, 0x2000f5, 0x2000f6, 0x2000f8, 0x2000fc, 0x2000fe, + 0x2d0030, 0x30000a, 0x30000c, 0x300010, 0x300012, 0x30001a, 0x300056, + 0x3000f0, 0x3000f2, 0x3000f6, 0x3000f8, 0x3000fc, 0x3000fe, 0x400006, + 0x400007, 0x40000a, 0x40000e, 0x400010, 0x400018, 0x400056, 0x4000f0, + 0x4000f8, 0x600005 + }; + const unsigned int event_index_table[] = { + 0, 1, 2, 3, 4, 1, 5, 6, 7, 8, 9, 10, 11, 12, 13, 12, 14, + 7, 15, 2, 9, 16, 3, 4, 0, 17, 10, 18, 19, 20, 1, 17, 15, 19, + 18, 2, 16, 21, 8, 0, 22, 13, 14, 11, 21, 5, 20, 22, 1, 6, 3 + }; + int hi = ARRAY_SIZE(presorted_event_table) - 1; + int lo = 0; + + while (lo <= hi) { + int mid = lo + (hi - lo) / 2; + unsigned int alt = presorted_event_table[mid]; + + if (alt < event) + lo = mid + 1; + else if (alt > event) + hi = mid - 1; + else + return event_index_table[mid]; } + return -1; } @@ -457,11 +472,11 @@ static int p6_get_alternatives(u64 event, unsigned int flags, u64 alt[]) return nalt; } -static void p6_disable_pmc(unsigned int pmc, unsigned long mmcr[]) +static void p6_disable_pmc(unsigned int pmc, struct mmcr_regs *mmcr) { /* Set PMCxSEL to 0 to disable PMCx */ if (pmc <= 3) - mmcr[1] &= ~(0xffUL << MMCR1_PMCSEL_SH(pmc)); + mmcr->mmcr1 &= ~(0xffUL << MMCR1_PMCSEL_SH(pmc)); } static int power6_generic_events[] = { @@ -481,7 +496,7 @@ static int power6_generic_events[] = { * are event codes. * The "DTLB" and "ITLB" events relate to the DERAT and IERAT. */ -static int power6_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { +static u64 power6_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { [C(L1D)] = { /* RESULT_ACCESS RESULT_MISS */ [C(OP_READ)] = { 0x280030, 0x80080 }, [C(OP_WRITE)] = { 0x180032, 0x80088 }, @@ -536,10 +551,11 @@ static struct power_pmu power6_pmu = { .cache_events = &power6_cache_events, }; -int init_power6_pmu(void) +int __init init_power6_pmu(void) { - if (!cur_cpu_spec->oprofile_cpu_type || - strcmp(cur_cpu_spec->oprofile_cpu_type, "ppc64/power6")) + unsigned int pvr = mfspr(SPRN_PVR); + + if (PVR_VER(pvr) != PVR_POWER6) return -ENODEV; return register_power_pmu(&power6_pmu); diff --git a/arch/powerpc/perf/power7-pmu.c b/arch/powerpc/perf/power7-pmu.c index a137813a3076..c95ccf2e28da 100644 --- a/arch/powerpc/perf/power7-pmu.c +++ b/arch/powerpc/perf/power7-pmu.c @@ -10,6 +10,8 @@ #include <asm/reg.h> #include <asm/cputable.h> +#include "internal.h" + /* * Bits in event code for POWER7 */ @@ -79,7 +81,7 @@ enum { */ static int power7_get_constraint(u64 event, unsigned long *maskp, - unsigned long *valp) + unsigned long *valp, u64 event_config1 __maybe_unused) { int pmc, sh, unit; unsigned long mask = 0, value = 0; @@ -242,7 +244,9 @@ static int power7_marked_instr_event(u64 event) } static int power7_compute_mmcr(u64 event[], int n_ev, - unsigned int hwc[], unsigned long mmcr[], struct perf_event *pevents[]) + unsigned int hwc[], struct mmcr_regs *mmcr, + struct perf_event *pevents[], + u32 flags __maybe_unused) { unsigned long mmcr1 = 0; unsigned long mmcra = MMCRA_SDAR_DCACHE_MISS | MMCRA_SDAR_ERAT_MISS; @@ -298,20 +302,20 @@ static int power7_compute_mmcr(u64 event[], int n_ev, } /* Return MMCRx values */ - mmcr[0] = 0; + mmcr->mmcr0 = 0; if (pmc_inuse & 1) - mmcr[0] = MMCR0_PMC1CE; + mmcr->mmcr0 = MMCR0_PMC1CE; if (pmc_inuse & 0x3e) - mmcr[0] |= MMCR0_PMCjCE; - mmcr[1] = mmcr1; - mmcr[2] = mmcra; + mmcr->mmcr0 |= MMCR0_PMCjCE; + mmcr->mmcr1 = mmcr1; + mmcr->mmcra = mmcra; return 0; } -static void power7_disable_pmc(unsigned int pmc, unsigned long mmcr[]) +static void power7_disable_pmc(unsigned int pmc, struct mmcr_regs *mmcr) { if (pmc <= 3) - mmcr[1] &= ~(0xffUL << MMCR1_PMCSEL_SH(pmc)); + mmcr->mmcr1 &= ~(0xffUL << MMCR1_PMCSEL_SH(pmc)); } static int power7_generic_events[] = { @@ -332,7 +336,7 @@ static int power7_generic_events[] = { * 0 means not supported, -1 means nonsensical, other values * are event codes. */ -static int power7_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { +static u64 power7_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { [C(L1D)] = { /* RESULT_ACCESS RESULT_MISS */ [C(OP_READ)] = { 0xc880, 0x400f0 }, [C(OP_WRITE)] = { 0, 0x300f0 }, @@ -401,7 +405,7 @@ static struct attribute *power7_events_attr[] = { NULL }; -static struct attribute_group power7_pmu_events_group = { +static const struct attribute_group power7_pmu_events_group = { .name = "events", .attrs = power7_events_attr, }; @@ -413,7 +417,7 @@ static struct attribute *power7_pmu_format_attr[] = { NULL, }; -static struct attribute_group power7_pmu_format_group = { +static const struct attribute_group power7_pmu_format_group = { .name = "format", .attrs = power7_pmu_format_attr, }; @@ -441,13 +445,14 @@ static struct power_pmu power7_pmu = { .cache_events = &power7_cache_events, }; -int init_power7_pmu(void) +int __init init_power7_pmu(void) { - if (!cur_cpu_spec->oprofile_cpu_type || - strcmp(cur_cpu_spec->oprofile_cpu_type, "ppc64/power7")) + unsigned int pvr = mfspr(SPRN_PVR); + + if (PVR_VER(pvr) != PVR_POWER7 && PVR_VER(pvr) != PVR_POWER7p) return -ENODEV; - if (pvr_version_is(PVR_POWER7p)) + if (PVR_VER(pvr) == PVR_POWER7p) power7_pmu.flags |= PPMU_SIAR_VALID; return register_power_pmu(&power7_pmu); diff --git a/arch/powerpc/perf/power8-pmu.c b/arch/powerpc/perf/power8-pmu.c index 3a5fcc20ff31..ef9685065aaf 100644 --- a/arch/powerpc/perf/power8-pmu.c +++ b/arch/powerpc/perf/power8-pmu.c @@ -92,7 +92,7 @@ enum { */ /* PowerISA v2.07 format attribute structure*/ -extern struct attribute_group isa207_pmu_format_group; +extern const struct attribute_group isa207_pmu_format_group; /* Table of alternatives, sorted by column 0 */ static const unsigned int event_alternatives[][MAX_ALT] = { @@ -182,14 +182,24 @@ static struct attribute *power8_events_attr[] = { NULL }; -static struct attribute_group power8_pmu_events_group = { +static const struct attribute_group power8_pmu_events_group = { .name = "events", .attrs = power8_events_attr, }; +static struct attribute *power8_pmu_caps_attrs[] = { + NULL +}; + +static struct attribute_group power8_pmu_caps_group = { + .name = "caps", + .attrs = power8_pmu_caps_attrs, +}; + static const struct attribute_group *power8_pmu_attr_groups[] = { &isa207_pmu_format_group, &power8_pmu_events_group, + &power8_pmu_caps_group, NULL, }; @@ -253,7 +263,7 @@ static void power8_config_bhrb(u64 pmu_bhrb_filter) * 0 means not supported, -1 means nonsensical, other values * are event codes. */ -static int power8_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { +static u64 power8_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { [ C(L1D) ] = { [ C(OP_READ) ] = { [ C(RESULT_ACCESS) ] = PM_LD_REF_L1, @@ -378,12 +388,13 @@ static struct power_pmu power8_pmu = { .bhrb_nr = 32, }; -int init_power8_pmu(void) +int __init init_power8_pmu(void) { int rc; + unsigned int pvr = mfspr(SPRN_PVR); - if (!cur_cpu_spec->oprofile_cpu_type || - strcmp(cur_cpu_spec->oprofile_cpu_type, "ppc64/power8")) + if (PVR_VER(pvr) != PVR_POWER8E && PVR_VER(pvr) != PVR_POWER8NVL && + PVR_VER(pvr) != PVR_POWER8) return -ENODEV; rc = register_power_pmu(&power8_pmu); diff --git a/arch/powerpc/perf/power9-pmu.c b/arch/powerpc/perf/power9-pmu.c index 08c3ef796198..cb6a7dc02dd7 100644 --- a/arch/powerpc/perf/power9-pmu.c +++ b/arch/powerpc/perf/power9-pmu.c @@ -90,13 +90,15 @@ enum { #define POWER9_MMCRA_IFM3 0x00000000C0000000UL #define POWER9_MMCRA_BHRB_MASK 0x00000000C0000000UL +extern u64 PERF_REG_EXTENDED_MASK; + /* Nasty Power9 specific hack */ #define PVR_POWER9_CUMULUS 0x00002000 /* PowerISA v2.07 format attribute structure*/ -extern struct attribute_group isa207_pmu_format_group; +extern const struct attribute_group isa207_pmu_format_group; -int p9_dd21_bl_ev[] = { +static int p9_dd21_bl_ev[] = { PM_MRK_ST_DONE_L2, PM_RADIX_PWC_L1_HIT, PM_FLOP_CMPL, @@ -110,7 +112,7 @@ int p9_dd21_bl_ev[] = { PM_DISP_HELD_SYNC_HOLD, }; -int p9_dd22_bl_ev[] = { +static int p9_dd22_bl_ev[] = { PM_DTLB_MISS_16G, PM_DERAT_MISS_2M, PM_DTLB_MISS_2M, @@ -131,11 +133,11 @@ int p9_dd22_bl_ev[] = { /* Table of alternatives, sorted by column 0 */ static const unsigned int power9_event_alternatives[][MAX_ALT] = { - { PM_INST_DISP, PM_INST_DISP_ALT }, - { PM_RUN_CYC_ALT, PM_RUN_CYC }, - { PM_RUN_INST_CMPL_ALT, PM_RUN_INST_CMPL }, - { PM_LD_MISS_L1, PM_LD_MISS_L1_ALT }, { PM_BR_2PATH, PM_BR_2PATH_ALT }, + { PM_INST_DISP, PM_INST_DISP_ALT }, + { PM_RUN_CYC_ALT, PM_RUN_CYC }, + { PM_LD_MISS_L1, PM_LD_MISS_L1_ALT }, + { PM_RUN_INST_CMPL_ALT, PM_RUN_INST_CMPL }, }; static int power9_get_alternatives(u64 event, unsigned int flags, u64 alt[]) @@ -149,6 +151,18 @@ static int power9_get_alternatives(u64 event, unsigned int flags, u64 alt[]) return num_alt; } +static int power9_check_attr_config(struct perf_event *ev) +{ + u64 val; + u64 event = ev->attr.config; + + val = (event >> EVENT_SAMPLE_SHIFT) & EVENT_SAMPLE_MASK; + if (val == 0xC || isa3XX_check_attr_config(ev)) + return -EINVAL; + + return 0; +} + GENERIC_EVENT_ATTR(cpu-cycles, PM_CYC); GENERIC_EVENT_ATTR(stalled-cycles-frontend, PM_ICT_NOSLOT_CYC); GENERIC_EVENT_ATTR(stalled-cycles-backend, PM_CMPLU_STALL); @@ -203,7 +217,7 @@ static struct attribute *power9_events_attr[] = { NULL }; -static struct attribute_group power9_pmu_events_group = { +static const struct attribute_group power9_pmu_events_group = { .name = "events", .attrs = power9_events_attr, }; @@ -239,14 +253,24 @@ static struct attribute *power9_pmu_format_attr[] = { NULL, }; -static struct attribute_group power9_pmu_format_group = { +static const struct attribute_group power9_pmu_format_group = { .name = "format", .attrs = power9_pmu_format_attr, }; +static struct attribute *power9_pmu_caps_attrs[] = { + NULL +}; + +static struct attribute_group power9_pmu_caps_group = { + .name = "caps", + .attrs = power9_pmu_caps_attrs, +}; + static const struct attribute_group *power9_pmu_attr_groups[] = { &power9_pmu_format_group, &power9_pmu_events_group, + &power9_pmu_caps_group, NULL, }; @@ -310,7 +334,7 @@ static void power9_config_bhrb(u64 pmu_bhrb_filter) * 0 means not supported, -1 means nonsensical, other values * are event codes. */ -static int power9_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { +static u64 power9_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { [ C(L1D) ] = { [ C(OP_READ) ] = { [ C(RESULT_ACCESS) ] = PM_LD_REF_L1, @@ -434,16 +458,16 @@ static struct power_pmu power9_pmu = { .cache_events = &power9_cache_events, .attr_groups = power9_pmu_attr_groups, .bhrb_nr = 32, + .capabilities = PERF_PMU_CAP_EXTENDED_REGS, + .check_attr_config = power9_check_attr_config, }; -int init_power9_pmu(void) +int __init init_power9_pmu(void) { int rc = 0; unsigned int pvr = mfspr(SPRN_PVR); - /* Comes from cpu_specs[] */ - if (!cur_cpu_spec->oprofile_cpu_type || - strcmp(cur_cpu_spec->oprofile_cpu_type, "ppc64/power9")) + if (PVR_VER(pvr) != PVR_POWER9) return -ENODEV; /* Blacklist events */ @@ -457,6 +481,9 @@ int init_power9_pmu(void) } } + /* Set the PERF_REG_EXTENDED_MASK here */ + PERF_REG_EXTENDED_MASK = PERF_REG_PMU_MASK_300; + rc = register_power_pmu(&power9_pmu); if (rc) return rc; diff --git a/arch/powerpc/perf/ppc970-pmu.c b/arch/powerpc/perf/ppc970-pmu.c index 4035d93d87ab..762676fb839e 100644 --- a/arch/powerpc/perf/ppc970-pmu.c +++ b/arch/powerpc/perf/ppc970-pmu.c @@ -9,6 +9,8 @@ #include <asm/reg.h> #include <asm/cputable.h> +#include "internal.h" + /* * Bits in event code for PPC970 */ @@ -188,7 +190,7 @@ static unsigned long unit_cons[PM_LASTUNIT+1][2] = { }; static int p970_get_constraint(u64 event, unsigned long *maskp, - unsigned long *valp) + unsigned long *valp, u64 event_config1 __maybe_unused) { int pmc, byte, unit, sh, spcsel; unsigned long mask = 0, value = 0; @@ -253,7 +255,9 @@ static int p970_get_alternatives(u64 event, unsigned int flags, u64 alt[]) } static int p970_compute_mmcr(u64 event[], int n_ev, - unsigned int hwc[], unsigned long mmcr[], struct perf_event *pevents[]) + unsigned int hwc[], struct mmcr_regs *mmcr, + struct perf_event *pevents[], + u32 flags __maybe_unused) { unsigned long mmcr0 = 0, mmcr1 = 0, mmcra = 0; unsigned int pmc, unit, byte, psel; @@ -393,27 +397,26 @@ static int p970_compute_mmcr(u64 event[], int n_ev, mmcra |= 0x2000; /* mark only one IOP per PPC instruction */ /* Return MMCRx values */ - mmcr[0] = mmcr0; - mmcr[1] = mmcr1; - mmcr[2] = mmcra; + mmcr->mmcr0 = mmcr0; + mmcr->mmcr1 = mmcr1; + mmcr->mmcra = mmcra; return 0; } -static void p970_disable_pmc(unsigned int pmc, unsigned long mmcr[]) +static void p970_disable_pmc(unsigned int pmc, struct mmcr_regs *mmcr) { - int shift, i; + int shift; + /* + * Setting the PMCxSEL field to 0x08 disables PMC x. + */ if (pmc <= 1) { shift = MMCR0_PMC1SEL_SH - 7 * pmc; - i = 0; + mmcr->mmcr0 = (mmcr->mmcr0 & ~(0x1fUL << shift)) | (0x08UL << shift); } else { shift = MMCR1_PMC3SEL_SH - 5 * (pmc - 2); - i = 1; + mmcr->mmcr1 = (mmcr->mmcr1 & ~(0x1fUL << shift)) | (0x08UL << shift); } - /* - * Setting the PMCxSEL field to 0x08 disables PMC x. - */ - mmcr[i] = (mmcr[i] & ~(0x1fUL << shift)) | (0x08UL << shift); } static int ppc970_generic_events[] = { @@ -432,7 +435,7 @@ static int ppc970_generic_events[] = { * 0 means not supported, -1 means nonsensical, other values * are event codes. */ -static int ppc970_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { +static u64 ppc970_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { [C(L1D)] = { /* RESULT_ACCESS RESULT_MISS */ [C(OP_READ)] = { 0x8810, 0x3810 }, [C(OP_WRITE)] = { 0x7810, 0x813 }, @@ -486,11 +489,12 @@ static struct power_pmu ppc970_pmu = { .flags = PPMU_NO_SIPR | PPMU_NO_CONT_SAMPLING, }; -int init_ppc970_pmu(void) +int __init init_ppc970_pmu(void) { - if (!cur_cpu_spec->oprofile_cpu_type || - (strcmp(cur_cpu_spec->oprofile_cpu_type, "ppc64/970") - && strcmp(cur_cpu_spec->oprofile_cpu_type, "ppc64/970MP"))) + unsigned int pvr = mfspr(SPRN_PVR); + + if (PVR_VER(pvr) != PVR_970 && PVR_VER(pvr) != PVR_970MP && + PVR_VER(pvr) != PVR_970FX && PVR_VER(pvr) != PVR_970GX) return -ENODEV; return register_power_pmu(&ppc970_pmu); diff --git a/arch/powerpc/perf/req-gen/perf.h b/arch/powerpc/perf/req-gen/perf.h index fa9bc804e67a..6b2a59fefffa 100644 --- a/arch/powerpc/perf/req-gen/perf.h +++ b/arch/powerpc/perf/req-gen/perf.h @@ -139,6 +139,26 @@ PMU_EVENT_ATTR_STRING( \ #define REQUEST_(r_name, r_value, r_idx_1, r_fields) \ r_fields +/* Generate event list for platforms with counter_info_version 0x6 or below */ +static __maybe_unused struct attribute *hv_gpci_event_attrs_v6[] = { +#include REQUEST_FILE + NULL +}; + +/* + * Based on getPerfCountInfo v1.018 documentation, some of the hv-gpci + * events were deprecated for platform firmware that supports + * counter_info_version 0x8 or above. + * Those deprecated events are still part of platform firmware that + * support counter_info_version 0x6 and below. As per the getPerfCountInfo + * v1.018 documentation there is no counter_info_version 0x7. + * Undefining macro ENABLE_EVENTS_COUNTERINFO_V6, to disable the addition of + * deprecated events in "hv_gpci_event_attrs" attribute group, for platforms + * that supports counter_info_version 0x8 or above. + */ +#undef ENABLE_EVENTS_COUNTERINFO_V6 + +/* Generate event list for platforms with counter_info_version 0x8 or above*/ static __maybe_unused struct attribute *hv_gpci_event_attrs[] = { #include REQUEST_FILE NULL |