From 08280e6c4c2e8049ac61d9e8e3536ec1df629c0d Mon Sep 17 00:00:00 2001 From: "David S. Miller" Date: Sun, 14 Oct 2012 17:59:40 -0700 Subject: [PATCH 1/3] sparc64: Like x86 we should check current->mm during perf backtrace generation. If the MM is not active, only report the top-level PC. Do not try to access the address space. Signed-off-by: David S. Miller --- arch/sparc/kernel/perf_event.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/arch/sparc/kernel/perf_event.c b/arch/sparc/kernel/perf_event.c index e48651dace1b..9e96f849a744 100644 --- a/arch/sparc/kernel/perf_event.c +++ b/arch/sparc/kernel/perf_event.c @@ -1738,8 +1738,6 @@ static void perf_callchain_user_64(struct perf_callchain_entry *entry, { unsigned long ufp; - perf_callchain_store(entry, regs->tpc); - ufp = regs->u_regs[UREG_I6] + STACK_BIAS; do { struct sparc_stackf *usf, sf; @@ -1760,8 +1758,6 @@ static void perf_callchain_user_32(struct perf_callchain_entry *entry, { unsigned long ufp; - perf_callchain_store(entry, regs->tpc); - ufp = regs->u_regs[UREG_I6] & 0xffffffffUL; do { struct sparc_stackf32 *usf, sf; @@ -1780,6 +1776,11 @@ static void perf_callchain_user_32(struct perf_callchain_entry *entry, void perf_callchain_user(struct perf_callchain_entry *entry, struct pt_regs *regs) { + perf_callchain_store(entry, regs->tpc); + + if (!current->mm) + return; + flushw_user(); if (test_thread_flag(TIF_32BIT)) perf_callchain_user_32(entry, regs); From 916ca14aaf12a7191118adb51bb95e3c7866380d Mon Sep 17 00:00:00 2001 From: "David S. Miller" Date: Tue, 16 Oct 2012 09:34:01 -0700 Subject: [PATCH 2/3] sparc64: Add global PMU register dumping via sysrq. Signed-off-by: David S. Miller --- Documentation/sysrq.txt | 1 + arch/sparc/include/asm/ptrace.h | 13 +++- arch/sparc/include/asm/smp_64.h | 2 + arch/sparc/kernel/process_64.c | 120 ++++++++++++++++++++++++++------ arch/sparc/kernel/smp_64.c | 11 +++ arch/sparc/mm/ultra.S | 64 ++++++++++++++++- drivers/tty/sysrq.c | 1 + 7 files changed, 189 insertions(+), 23 deletions(-) diff --git a/Documentation/sysrq.txt b/Documentation/sysrq.txt index 642f84495b29..2a4cdda4828e 100644 --- a/Documentation/sysrq.txt +++ b/Documentation/sysrq.txt @@ -116,6 +116,7 @@ On all - write a character to /proc/sysrq-trigger. e.g.: 'w' - Dumps tasks that are in uninterruptable (blocked) state. 'x' - Used by xmon interface on ppc/powerpc platforms. + Show global PMU Registers on sparc64. 'y' - Show global CPU Registers [SPARC-64 specific] diff --git a/arch/sparc/include/asm/ptrace.h b/arch/sparc/include/asm/ptrace.h index 0c6f6b068289..da43bdc62294 100644 --- a/arch/sparc/include/asm/ptrace.h +++ b/arch/sparc/include/asm/ptrace.h @@ -42,7 +42,18 @@ struct global_reg_snapshot { struct thread_info *thread; unsigned long pad1; }; -extern struct global_reg_snapshot global_reg_snapshot[NR_CPUS]; + +struct global_pmu_snapshot { + unsigned long pcr[4]; + unsigned long pic[4]; +}; + +union global_cpu_snapshot { + struct global_reg_snapshot reg; + struct global_pmu_snapshot pmu; +}; + +extern union global_cpu_snapshot global_cpu_snapshot[NR_CPUS]; #define force_successful_syscall_return() \ do { current_thread_info()->syscall_noerror = 1; \ diff --git a/arch/sparc/include/asm/smp_64.h b/arch/sparc/include/asm/smp_64.h index 29862a9e9065..dd3bef4b9896 100644 --- a/arch/sparc/include/asm/smp_64.h +++ b/arch/sparc/include/asm/smp_64.h @@ -48,6 +48,7 @@ extern void smp_fill_in_sib_core_maps(void); extern void cpu_play_dead(void); extern void smp_fetch_global_regs(void); +extern void smp_fetch_global_pmu(void); struct seq_file; void smp_bogo(struct seq_file *); @@ -65,6 +66,7 @@ extern void __cpu_die(unsigned int cpu); #define hard_smp_processor_id() 0 #define smp_fill_in_sib_core_maps() do { } while (0) #define smp_fetch_global_regs() do { } while (0) +#define smp_fetch_global_pmu() do { } while (0) #endif /* !(CONFIG_SMP) */ diff --git a/arch/sparc/kernel/process_64.c b/arch/sparc/kernel/process_64.c index fcaa59421126..d778248ef3f8 100644 --- a/arch/sparc/kernel/process_64.c +++ b/arch/sparc/kernel/process_64.c @@ -27,6 +27,7 @@ #include #include #include +#include #include #include #include @@ -47,6 +48,7 @@ #include #include #include +#include #include "kstack.h" @@ -204,18 +206,22 @@ void show_regs(struct pt_regs *regs) show_stack(current, (unsigned long *) regs->u_regs[UREG_FP]); } -struct global_reg_snapshot global_reg_snapshot[NR_CPUS]; -static DEFINE_SPINLOCK(global_reg_snapshot_lock); +union global_cpu_snapshot global_cpu_snapshot[NR_CPUS]; +static DEFINE_SPINLOCK(global_cpu_snapshot_lock); static void __global_reg_self(struct thread_info *tp, struct pt_regs *regs, int this_cpu) { + struct global_reg_snapshot *rp; + flushw_all(); - global_reg_snapshot[this_cpu].tstate = regs->tstate; - global_reg_snapshot[this_cpu].tpc = regs->tpc; - global_reg_snapshot[this_cpu].tnpc = regs->tnpc; - global_reg_snapshot[this_cpu].o7 = regs->u_regs[UREG_I7]; + rp = &global_cpu_snapshot[this_cpu].reg; + + rp->tstate = regs->tstate; + rp->tpc = regs->tpc; + rp->tnpc = regs->tnpc; + rp->o7 = regs->u_regs[UREG_I7]; if (regs->tstate & TSTATE_PRIV) { struct reg_window *rw; @@ -223,17 +229,17 @@ static void __global_reg_self(struct thread_info *tp, struct pt_regs *regs, rw = (struct reg_window *) (regs->u_regs[UREG_FP] + STACK_BIAS); if (kstack_valid(tp, (unsigned long) rw)) { - global_reg_snapshot[this_cpu].i7 = rw->ins[7]; + rp->i7 = rw->ins[7]; rw = (struct reg_window *) (rw->ins[6] + STACK_BIAS); if (kstack_valid(tp, (unsigned long) rw)) - global_reg_snapshot[this_cpu].rpc = rw->ins[7]; + rp->rpc = rw->ins[7]; } } else { - global_reg_snapshot[this_cpu].i7 = 0; - global_reg_snapshot[this_cpu].rpc = 0; + rp->i7 = 0; + rp->rpc = 0; } - global_reg_snapshot[this_cpu].thread = tp; + rp->thread = tp; } /* In order to avoid hangs we do not try to synchronize with the @@ -261,9 +267,9 @@ void arch_trigger_all_cpu_backtrace(void) if (!regs) regs = tp->kregs; - spin_lock_irqsave(&global_reg_snapshot_lock, flags); + spin_lock_irqsave(&global_cpu_snapshot_lock, flags); - memset(global_reg_snapshot, 0, sizeof(global_reg_snapshot)); + memset(global_cpu_snapshot, 0, sizeof(global_cpu_snapshot)); this_cpu = raw_smp_processor_id(); @@ -272,7 +278,7 @@ void arch_trigger_all_cpu_backtrace(void) smp_fetch_global_regs(); for_each_online_cpu(cpu) { - struct global_reg_snapshot *gp = &global_reg_snapshot[cpu]; + struct global_reg_snapshot *gp = &global_cpu_snapshot[cpu].reg; __global_reg_poll(gp); @@ -295,9 +301,9 @@ void arch_trigger_all_cpu_backtrace(void) } } - memset(global_reg_snapshot, 0, sizeof(global_reg_snapshot)); + memset(global_cpu_snapshot, 0, sizeof(global_cpu_snapshot)); - spin_unlock_irqrestore(&global_reg_snapshot_lock, flags); + spin_unlock_irqrestore(&global_cpu_snapshot_lock, flags); } #ifdef CONFIG_MAGIC_SYSRQ @@ -309,16 +315,90 @@ static void sysrq_handle_globreg(int key) static struct sysrq_key_op sparc_globalreg_op = { .handler = sysrq_handle_globreg, - .help_msg = "Globalregs", + .help_msg = "global-regs(Y)", .action_msg = "Show Global CPU Regs", }; -static int __init sparc_globreg_init(void) +static void __global_pmu_self(int this_cpu) { - return register_sysrq_key('y', &sparc_globalreg_op); + struct global_pmu_snapshot *pp; + int i, num; + + pp = &global_cpu_snapshot[this_cpu].pmu; + + num = 1; + if (tlb_type == hypervisor && + sun4v_chip_type >= SUN4V_CHIP_NIAGARA4) + num = 4; + + for (i = 0; i < num; i++) { + pp->pcr[i] = pcr_ops->read_pcr(i); + pp->pic[i] = pcr_ops->read_pic(i); + } } -core_initcall(sparc_globreg_init); +static void __global_pmu_poll(struct global_pmu_snapshot *pp) +{ + int limit = 0; + + while (!pp->pcr[0] && ++limit < 100) { + barrier(); + udelay(1); + } +} + +static void pmu_snapshot_all_cpus(void) +{ + unsigned long flags; + int this_cpu, cpu; + + spin_lock_irqsave(&global_cpu_snapshot_lock, flags); + + memset(global_cpu_snapshot, 0, sizeof(global_cpu_snapshot)); + + this_cpu = raw_smp_processor_id(); + + __global_pmu_self(this_cpu); + + smp_fetch_global_pmu(); + + for_each_online_cpu(cpu) { + struct global_pmu_snapshot *pp = &global_cpu_snapshot[cpu].pmu; + + __global_pmu_poll(pp); + + printk("%c CPU[%3d]: PCR[%08lx:%08lx:%08lx:%08lx] PIC[%08lx:%08lx:%08lx:%08lx]\n", + (cpu == this_cpu ? '*' : ' '), cpu, + pp->pcr[0], pp->pcr[1], pp->pcr[2], pp->pcr[3], + pp->pic[0], pp->pic[1], pp->pic[2], pp->pic[3]); + } + + memset(global_cpu_snapshot, 0, sizeof(global_cpu_snapshot)); + + spin_unlock_irqrestore(&global_cpu_snapshot_lock, flags); +} + +static void sysrq_handle_globpmu(int key) +{ + pmu_snapshot_all_cpus(); +} + +static struct sysrq_key_op sparc_globalpmu_op = { + .handler = sysrq_handle_globpmu, + .help_msg = "global-pmu(X)", + .action_msg = "Show Global PMU Regs", +}; + +static int __init sparc_sysrq_init(void) +{ + int ret = register_sysrq_key('y', &sparc_globalreg_op); + + if (!ret) + ret = register_sysrq_key('x', &sparc_globalpmu_op); + return ret; +} + +core_initcall(sparc_sysrq_init); #endif diff --git a/arch/sparc/kernel/smp_64.c b/arch/sparc/kernel/smp_64.c index 781bcb10b8bd..d94b878577b7 100644 --- a/arch/sparc/kernel/smp_64.c +++ b/arch/sparc/kernel/smp_64.c @@ -852,6 +852,8 @@ extern unsigned long xcall_flush_tlb_mm; extern unsigned long xcall_flush_tlb_pending; extern unsigned long xcall_flush_tlb_kernel_range; extern unsigned long xcall_fetch_glob_regs; +extern unsigned long xcall_fetch_glob_pmu; +extern unsigned long xcall_fetch_glob_pmu_n4; extern unsigned long xcall_receive_signal; extern unsigned long xcall_new_mmu_context_version; #ifdef CONFIG_KGDB @@ -1000,6 +1002,15 @@ void smp_fetch_global_regs(void) smp_cross_call(&xcall_fetch_glob_regs, 0, 0, 0); } +void smp_fetch_global_pmu(void) +{ + if (tlb_type == hypervisor && + sun4v_chip_type >= SUN4V_CHIP_NIAGARA4) + smp_cross_call(&xcall_fetch_glob_pmu_n4, 0, 0, 0); + else + smp_cross_call(&xcall_fetch_glob_pmu, 0, 0, 0); +} + /* We know that the window frames of the user have been flushed * to the stack before we get here because all callers of us * are flush_tlb_*() routines, and these run after flush_cache_*() diff --git a/arch/sparc/mm/ultra.S b/arch/sparc/mm/ultra.S index 874162a11ceb..f8e13d421fcb 100644 --- a/arch/sparc/mm/ultra.S +++ b/arch/sparc/mm/ultra.S @@ -481,8 +481,8 @@ xcall_sync_tick: .globl xcall_fetch_glob_regs xcall_fetch_glob_regs: - sethi %hi(global_reg_snapshot), %g1 - or %g1, %lo(global_reg_snapshot), %g1 + sethi %hi(global_cpu_snapshot), %g1 + or %g1, %lo(global_cpu_snapshot), %g1 __GET_CPUID(%g2) sllx %g2, 6, %g3 add %g1, %g3, %g1 @@ -509,6 +509,66 @@ xcall_fetch_glob_regs: stx %g3, [%g1 + GR_SNAP_THREAD] retry + .globl xcall_fetch_glob_pmu +xcall_fetch_glob_pmu: + sethi %hi(global_cpu_snapshot), %g1 + or %g1, %lo(global_cpu_snapshot), %g1 + __GET_CPUID(%g2) + sllx %g2, 6, %g3 + add %g1, %g3, %g1 + rd %pic, %g7 + stx %g7, [%g1 + (4 * 8)] + rd %pcr, %g7 + stx %g7, [%g1 + (0 * 8)] + retry + + .globl xcall_fetch_glob_pmu_n4 +xcall_fetch_glob_pmu_n4: + sethi %hi(global_cpu_snapshot), %g1 + or %g1, %lo(global_cpu_snapshot), %g1 + __GET_CPUID(%g2) + sllx %g2, 6, %g3 + add %g1, %g3, %g1 + + ldxa [%g0] ASI_PIC, %g7 + stx %g7, [%g1 + (4 * 8)] + mov 0x08, %g3 + ldxa [%g3] ASI_PIC, %g7 + stx %g7, [%g1 + (5 * 8)] + mov 0x10, %g3 + ldxa [%g3] ASI_PIC, %g7 + stx %g7, [%g1 + (6 * 8)] + mov 0x18, %g3 + ldxa [%g3] ASI_PIC, %g7 + stx %g7, [%g1 + (7 * 8)] + + mov %o0, %g2 + mov %o1, %g3 + mov %o5, %g7 + + mov HV_FAST_VT_GET_PERFREG, %o5 + mov 3, %o0 + ta HV_FAST_TRAP + stx %o1, [%g1 + (3 * 8)] + mov HV_FAST_VT_GET_PERFREG, %o5 + mov 2, %o0 + ta HV_FAST_TRAP + stx %o1, [%g1 + (2 * 8)] + mov HV_FAST_VT_GET_PERFREG, %o5 + mov 1, %o0 + ta HV_FAST_TRAP + stx %o1, [%g1 + (1 * 8)] + mov HV_FAST_VT_GET_PERFREG, %o5 + mov 0, %o0 + ta HV_FAST_TRAP + stx %o1, [%g1 + (0 * 8)] + + mov %g2, %o0 + mov %g3, %o1 + mov %g7, %o5 + + retry + #ifdef DCACHE_ALIASING_POSSIBLE .align 32 .globl xcall_flush_dcache_page_cheetah diff --git a/drivers/tty/sysrq.c b/drivers/tty/sysrq.c index 05728894a88c..16ee6cee07da 100644 --- a/drivers/tty/sysrq.c +++ b/drivers/tty/sysrq.c @@ -452,6 +452,7 @@ static struct sysrq_key_op *sysrq_key_table[36] = { NULL, /* v */ &sysrq_showstate_blocked_op, /* w */ /* x: May be registered on ppc/powerpc for xmon */ + /* x: May be registered on sparc64 for global PMU dump */ NULL, /* x */ /* y: May be registered on sparc64 for global register dump */ NULL, /* y */ From e793d8c6740f8fe704fa216e95685f4d92c4c4b9 Mon Sep 17 00:00:00 2001 From: "David S. Miller" Date: Tue, 16 Oct 2012 13:05:25 -0700 Subject: [PATCH 3/3] sparc64: Fix bit twiddling in sparc_pmu_enable_event(). There was a serious disconnect in the logic happening in sparc_pmu_disable_event() vs. sparc_pmu_enable_event(). Event disable is implemented by programming a NOP event into the PCR. However, event enable was not reversing this operation. Instead, it was setting the User/Priv/Hypervisor trace enable bits. That's not sparc_pmu_enable_event()'s job, that's what sparc_pmu_enable() and sparc_pmu_disable() do . The intent of sparc_pmu_enable_event() is clear, since it first clear out the event type encoding field. So fix this by OR'ing in the event encoding rather than the trace enable bits. Signed-off-by: David S. Miller --- arch/sparc/kernel/perf_event.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/arch/sparc/kernel/perf_event.c b/arch/sparc/kernel/perf_event.c index 9e96f849a744..885a8af74064 100644 --- a/arch/sparc/kernel/perf_event.c +++ b/arch/sparc/kernel/perf_event.c @@ -817,15 +817,17 @@ static u64 nop_for_index(int idx) static inline void sparc_pmu_enable_event(struct cpu_hw_events *cpuc, struct hw_perf_event *hwc, int idx) { - u64 val, mask = mask_for_index(idx); + u64 enc, val, mask = mask_for_index(idx); int pcr_index = 0; if (sparc_pmu->num_pcrs > 1) pcr_index = idx; + enc = perf_event_get_enc(cpuc->events[idx]); + val = cpuc->pcr[pcr_index]; val &= ~mask; - val |= hwc->config; + val |= event_encoding(enc, idx); cpuc->pcr[pcr_index] = val; pcr_ops->write_pcr(pcr_index, cpuc->pcr[pcr_index]);