From ca037701a025334e724e5c61b3b1082940c8b981 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Tue, 2 Mar 2010 19:52:12 +0100 Subject: perf, x86: Add PEBS infrastructure This patch implements support for Intel Precise Event Based Sampling, which is an alternative counter mode in which the counter triggers a hardware assist to collect information on events. The hardware assist takes a trap like snapshot of a subset of the machine registers. This data is written to the Intel Debug-Store, which can be programmed with a data threshold at which to raise a PMI. With the PEBS hardware assist being trap like, the reported IP is always one instruction after the actual instruction that triggered the event. This implements a simple PEBS model that always takes a single PEBS event at a time. This is done so that the interaction with the rest of the system is as expected (freq adjust, period randomization, lbr, callchains, etc.). It adds an ABI element: perf_event_attr::precise, which indicates that we wish to use this (constrained, but precise) mode. Signed-off-by: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: paulus@samba.org Cc: eranian@google.com Cc: robert.richter@amd.com Cc: fweisbec@gmail.com LKML-Reference: <20100304140100.392111285@chello.nl> Signed-off-by: Ingo Molnar --- include/linux/perf_event.h | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'include') diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index 80acbf3d5de1..42307b50c787 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -203,8 +203,9 @@ struct perf_event_attr { enable_on_exec : 1, /* next exec enables */ task : 1, /* trace fork/exit */ watermark : 1, /* wakeup_watermark */ + precise : 1, /* OoO invariant counter */ - __reserved_1 : 49; + __reserved_1 : 48; union { __u32 wakeup_events; /* wakeup every n events */ -- cgit 1.4.1 From caff2befffe899e63df5cc760b7ed01cfd902685 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Wed, 3 Mar 2010 12:02:30 +0100 Subject: perf, x86: Implement simple LBR support Implement simple suport Intel Last-Branch-Record, it supports all hardware that implements FREEZE_LBRS_ON_PMI, but does not (yet) implement the LBR config register. The Intel LBR is a FIFO of From,To addresses describing the last few branches the hardware took. This patch does not add perf interface to the LBR, but merely provides an interface for internal use. Signed-off-by: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: paulus@samba.org Cc: eranian@google.com Cc: robert.richter@amd.com Cc: fweisbec@gmail.com LKML-Reference: <20100304140100.544191154@chello.nl> Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_event.c | 18 +++ arch/x86/kernel/cpu/perf_event_intel.c | 13 ++ arch/x86/kernel/cpu/perf_event_intel_lbr.c | 228 +++++++++++++++++++++++++++++ include/linux/perf_event.h | 11 ++ 4 files changed, 270 insertions(+) create mode 100644 arch/x86/kernel/cpu/perf_event_intel_lbr.c (limited to 'include') diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c index 0c03d5c1671f..1badff6b6b28 100644 --- a/arch/x86/kernel/cpu/perf_event.c +++ b/arch/x86/kernel/cpu/perf_event.c @@ -48,6 +48,8 @@ struct amd_nb { struct event_constraint event_constraints[X86_PMC_IDX_MAX]; }; +#define MAX_LBR_ENTRIES 16 + struct cpu_hw_events { /* * Generic x86 PMC bits @@ -69,6 +71,14 @@ struct cpu_hw_events { struct debug_store *ds; u64 pebs_enabled; + /* + * Intel LBR bits + */ + int lbr_users; + void *lbr_context; + struct perf_branch_stack lbr_stack; + struct perf_branch_entry lbr_entries[MAX_LBR_ENTRIES]; + /* * AMD specific bits */ @@ -159,6 +169,13 @@ struct x86_pmu { int pebs_record_size; void (*drain_pebs)(struct pt_regs *regs); struct event_constraint *pebs_constraints; + + /* + * Intel LBR + */ + unsigned long lbr_tos, lbr_from, lbr_to; /* MSR base regs */ + int lbr_nr; /* hardware stack size */ + int lbr_format; /* hardware format */ }; static struct x86_pmu x86_pmu __read_mostly; @@ -1237,6 +1254,7 @@ undo: #include "perf_event_amd.c" #include "perf_event_p6.c" +#include "perf_event_intel_lbr.c" #include "perf_event_intel_ds.c" #include "perf_event_intel.c" diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c index 11446412e4c7..44f6ed42a934 100644 --- a/arch/x86/kernel/cpu/perf_event_intel.c +++ b/arch/x86/kernel/cpu/perf_event_intel.c @@ -480,6 +480,7 @@ static void intel_pmu_disable_all(void) intel_pmu_disable_bts(); intel_pmu_pebs_disable_all(); + intel_pmu_lbr_disable_all(); } static void intel_pmu_enable_all(void) @@ -499,6 +500,7 @@ static void intel_pmu_enable_all(void) } intel_pmu_pebs_enable_all(); + intel_pmu_lbr_enable_all(); } static inline u64 intel_pmu_get_status(void) @@ -674,6 +676,8 @@ again: inc_irq_stat(apic_perf_irqs); ack = status; + intel_pmu_lbr_read(); + /* * PEBS overflow sets bit 62 in the global status register */ @@ -848,6 +852,8 @@ static __init int intel_pmu_init(void) memcpy(hw_cache_event_ids, core2_hw_cache_event_ids, sizeof(hw_cache_event_ids)); + intel_pmu_lbr_init_core(); + x86_pmu.event_constraints = intel_core2_event_constraints; pr_cont("Core2 events, "); break; @@ -857,13 +863,18 @@ static __init int intel_pmu_init(void) memcpy(hw_cache_event_ids, nehalem_hw_cache_event_ids, sizeof(hw_cache_event_ids)); + intel_pmu_lbr_init_nhm(); + x86_pmu.event_constraints = intel_nehalem_event_constraints; pr_cont("Nehalem/Corei7 events, "); break; + case 28: /* Atom */ memcpy(hw_cache_event_ids, atom_hw_cache_event_ids, sizeof(hw_cache_event_ids)); + intel_pmu_lbr_init_atom(); + x86_pmu.event_constraints = intel_gen_event_constraints; pr_cont("Atom events, "); break; @@ -873,6 +884,8 @@ static __init int intel_pmu_init(void) memcpy(hw_cache_event_ids, westmere_hw_cache_event_ids, sizeof(hw_cache_event_ids)); + intel_pmu_lbr_init_nhm(); + x86_pmu.event_constraints = intel_westmere_event_constraints; pr_cont("Westmere events, "); break; diff --git a/arch/x86/kernel/cpu/perf_event_intel_lbr.c b/arch/x86/kernel/cpu/perf_event_intel_lbr.c new file mode 100644 index 000000000000..ea3e99ed82ce --- /dev/null +++ b/arch/x86/kernel/cpu/perf_event_intel_lbr.c @@ -0,0 +1,228 @@ +#ifdef CONFIG_CPU_SUP_INTEL + +enum { + LBR_FORMAT_32 = 0x00, + LBR_FORMAT_LIP = 0x01, + LBR_FORMAT_EIP = 0x02, + LBR_FORMAT_EIP_FLAGS = 0x03, +}; + +/* + * We only support LBR implementations that have FREEZE_LBRS_ON_PMI + * otherwise it becomes near impossible to get a reliable stack. + */ + +#define X86_DEBUGCTL_LBR (1 << 0) +#define X86_DEBUGCTL_FREEZE_LBRS_ON_PMI (1 << 11) + +static void __intel_pmu_lbr_enable(void) +{ + u64 debugctl; + + rdmsrl(MSR_IA32_DEBUGCTLMSR, debugctl); + debugctl |= (X86_DEBUGCTL_LBR | X86_DEBUGCTL_FREEZE_LBRS_ON_PMI); + wrmsrl(MSR_IA32_DEBUGCTLMSR, debugctl); +} + +static void __intel_pmu_lbr_disable(void) +{ + u64 debugctl; + + rdmsrl(MSR_IA32_DEBUGCTLMSR, debugctl); + debugctl &= ~(X86_DEBUGCTL_LBR | X86_DEBUGCTL_FREEZE_LBRS_ON_PMI); + wrmsrl(MSR_IA32_DEBUGCTLMSR, debugctl); +} + +static void intel_pmu_lbr_reset_32(void) +{ + int i; + + for (i = 0; i < x86_pmu.lbr_nr; i++) + wrmsrl(x86_pmu.lbr_from + i, 0); +} + +static void intel_pmu_lbr_reset_64(void) +{ + int i; + + for (i = 0; i < x86_pmu.lbr_nr; i++) { + wrmsrl(x86_pmu.lbr_from + i, 0); + wrmsrl(x86_pmu.lbr_to + i, 0); + } +} + +static void intel_pmu_lbr_reset(void) +{ + if (x86_pmu.lbr_format == LBR_FORMAT_32) + intel_pmu_lbr_reset_32(); + else + intel_pmu_lbr_reset_64(); +} + +static void intel_pmu_lbr_enable(struct perf_event *event) +{ + struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); + + if (!x86_pmu.lbr_nr) + return; + + WARN_ON(cpuc->enabled); + + /* + * Reset the LBR stack if this is the first LBR user or + * we changed task context so as to avoid data leaks. + */ + + if (!cpuc->lbr_users || + (event->ctx->task && cpuc->lbr_context != event->ctx)) { + intel_pmu_lbr_reset(); + cpuc->lbr_context = event->ctx; + } + + cpuc->lbr_users++; +} + +static void intel_pmu_lbr_disable(struct perf_event *event) +{ + struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); + + if (!x86_pmu.lbr_nr) + return; + + cpuc->lbr_users--; + + BUG_ON(cpuc->lbr_users < 0); + WARN_ON(cpuc->enabled); +} + +static void intel_pmu_lbr_enable_all(void) +{ + struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); + + if (cpuc->lbr_users) + __intel_pmu_lbr_enable(); +} + +static void intel_pmu_lbr_disable_all(void) +{ + struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); + + if (cpuc->lbr_users) + __intel_pmu_lbr_disable(); +} + +static inline u64 intel_pmu_lbr_tos(void) +{ + u64 tos; + + rdmsrl(x86_pmu.lbr_tos, tos); + + return tos; +} + +static void intel_pmu_lbr_read_32(struct cpu_hw_events *cpuc) +{ + unsigned long mask = x86_pmu.lbr_nr - 1; + u64 tos = intel_pmu_lbr_tos(); + int i; + + for (i = 0; i < x86_pmu.lbr_nr; i++, tos--) { + unsigned long lbr_idx = (tos - i) & mask; + union { + struct { + u32 from; + u32 to; + }; + u64 lbr; + } msr_lastbranch; + + rdmsrl(x86_pmu.lbr_from + lbr_idx, msr_lastbranch.lbr); + + cpuc->lbr_entries[i].from = msr_lastbranch.from; + cpuc->lbr_entries[i].to = msr_lastbranch.to; + cpuc->lbr_entries[i].flags = 0; + } + cpuc->lbr_stack.nr = i; +} + +#define LBR_FROM_FLAG_MISPRED (1ULL << 63) + +/* + * Due to lack of segmentation in Linux the effective address (offset) + * is the same as the linear address, allowing us to merge the LIP and EIP + * LBR formats. + */ +static void intel_pmu_lbr_read_64(struct cpu_hw_events *cpuc) +{ + unsigned long mask = x86_pmu.lbr_nr - 1; + u64 tos = intel_pmu_lbr_tos(); + int i; + + for (i = 0; i < x86_pmu.lbr_nr; i++, tos--) { + unsigned long lbr_idx = (tos - i) & mask; + u64 from, to, flags = 0; + + rdmsrl(x86_pmu.lbr_from + lbr_idx, from); + rdmsrl(x86_pmu.lbr_to + lbr_idx, to); + + if (x86_pmu.lbr_format == LBR_FORMAT_EIP_FLAGS) { + flags = !!(from & LBR_FROM_FLAG_MISPRED); + from = (u64)((((s64)from) << 1) >> 1); + } + + cpuc->lbr_entries[i].from = from; + cpuc->lbr_entries[i].to = to; + cpuc->lbr_entries[i].flags = flags; + } + cpuc->lbr_stack.nr = i; +} + +static void intel_pmu_lbr_read(void) +{ + struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); + + if (!cpuc->lbr_users) + return; + + if (x86_pmu.lbr_format == LBR_FORMAT_32) + intel_pmu_lbr_read_32(cpuc); + else + intel_pmu_lbr_read_64(cpuc); +} + +static int intel_pmu_lbr_format(void) +{ + u64 capabilities; + + rdmsrl(MSR_IA32_PERF_CAPABILITIES, capabilities); + return capabilities & 0x1f; +} + +static void intel_pmu_lbr_init_core(void) +{ + x86_pmu.lbr_format = intel_pmu_lbr_format(); + x86_pmu.lbr_nr = 4; + x86_pmu.lbr_tos = 0x01c9; + x86_pmu.lbr_from = 0x40; + x86_pmu.lbr_to = 0x60; +} + +static void intel_pmu_lbr_init_nhm(void) +{ + x86_pmu.lbr_format = intel_pmu_lbr_format(); + x86_pmu.lbr_nr = 16; + x86_pmu.lbr_tos = 0x01c9; + x86_pmu.lbr_from = 0x680; + x86_pmu.lbr_to = 0x6c0; +} + +static void intel_pmu_lbr_init_atom(void) +{ + x86_pmu.lbr_format = intel_pmu_lbr_format(); + x86_pmu.lbr_nr = 8; + x86_pmu.lbr_tos = 0x01c9; + x86_pmu.lbr_from = 0x40; + x86_pmu.lbr_to = 0x60; +} + +#endif /* CONFIG_CPU_SUP_INTEL */ diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index 42307b50c787..ab4fd9ede264 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -467,6 +467,17 @@ struct perf_raw_record { void *data; }; +struct perf_branch_entry { + __u64 from; + __u64 to; + __u64 flags; +}; + +struct perf_branch_stack { + __u64 nr; + struct perf_branch_entry entries[0]; +}; + struct task_struct; /** -- cgit 1.4.1 From ef21f683a045a79b6aa86ad81e5fdfc0d5ddd250 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Wed, 3 Mar 2010 13:12:23 +0100 Subject: perf, x86: use LBR for PEBS IP+1 fixup Use the LBR to fix up the PEBS IP+1 issue. As said, PEBS reports the next instruction, here we use the LBR to find the last branch and from that construct the actual IP. If the IP matches the LBR-TO, we use LBR-FROM, otherwise we use the LBR-TO address as the beginning of the last basic block and decode forward. Once we find a match to the current IP, we use the previous location. This patch introduces a new ABI element: PERF_RECORD_MISC_EXACT, which conveys that the reported IP (PERF_SAMPLE_IP) is the exact instruction that caused the event (barring CPU errata). The fixup can fail due to various reasons: 1) LBR contains invalid data (quite possible) 2) part of the basic block got paged out 3) the reported IP isn't part of the basic block (see 1) Signed-off-by: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Masami Hiramatsu Cc: "Zhang, Yanmin" Cc: paulus@samba.org Cc: eranian@google.com Cc: robert.richter@amd.com Cc: fweisbec@gmail.com LKML-Reference: <20100304140100.619375431@chello.nl> Signed-off-by: Ingo Molnar --- arch/x86/include/asm/perf_event.h | 19 +++++++ arch/x86/kernel/cpu/perf_event.c | 70 +++++++++++++------------- arch/x86/kernel/cpu/perf_event_intel.c | 4 +- arch/x86/kernel/cpu/perf_event_intel_ds.c | 84 ++++++++++++++++++++++++++++++- include/linux/perf_event.h | 6 +++ 5 files changed, 144 insertions(+), 39 deletions(-) (limited to 'include') diff --git a/arch/x86/include/asm/perf_event.h b/arch/x86/include/asm/perf_event.h index db6109a885a7..a9038c951619 100644 --- a/arch/x86/include/asm/perf_event.h +++ b/arch/x86/include/asm/perf_event.h @@ -136,6 +136,25 @@ extern void perf_events_lapic_init(void); #define PERF_EVENT_INDEX_OFFSET 0 +/* + * Abuse bit 3 of the cpu eflags register to indicate proper PEBS IP fixups. + * This flag is otherwise unused and ABI specified to be 0, so nobody should + * care what we do with it. + */ +#define PERF_EFLAGS_EXACT (1UL << 3) + +#define perf_misc_flags(regs) \ +({ int misc = 0; \ + if (user_mode(regs)) \ + misc |= PERF_RECORD_MISC_USER; \ + else \ + misc |= PERF_RECORD_MISC_KERNEL; \ + if (regs->flags & PERF_EFLAGS_EXACT) \ + misc |= PERF_RECORD_MISC_EXACT; \ + misc; }) + +#define perf_instruction_pointer(regs) ((regs)->ip) + #else static inline void init_hw_perf_events(void) { } static inline void perf_events_lapic_init(void) { } diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c index 1badff6b6b28..5cb4e8dcee4b 100644 --- a/arch/x86/kernel/cpu/perf_event.c +++ b/arch/x86/kernel/cpu/perf_event.c @@ -29,6 +29,41 @@ #include #include +/* + * best effort, GUP based copy_from_user() that assumes IRQ or NMI context + */ +static unsigned long +copy_from_user_nmi(void *to, const void __user *from, unsigned long n) +{ + unsigned long offset, addr = (unsigned long)from; + int type = in_nmi() ? KM_NMI : KM_IRQ0; + unsigned long size, len = 0; + struct page *page; + void *map; + int ret; + + do { + ret = __get_user_pages_fast(addr, 1, 0, &page); + if (!ret) + break; + + offset = addr & (PAGE_SIZE - 1); + size = min(PAGE_SIZE - offset, n - len); + + map = kmap_atomic(page, type); + memcpy(to, map+offset, size); + kunmap_atomic(map, type); + put_page(page); + + len += size; + to += size; + addr += size; + + } while (len < n); + + return len; +} + static u64 perf_event_mask __read_mostly; struct event_constraint { @@ -1550,41 +1585,6 @@ perf_callchain_kernel(struct pt_regs *regs, struct perf_callchain_entry *entry) dump_trace(NULL, regs, NULL, regs->bp, &backtrace_ops, entry); } -/* - * best effort, GUP based copy_from_user() that assumes IRQ or NMI context - */ -static unsigned long -copy_from_user_nmi(void *to, const void __user *from, unsigned long n) -{ - unsigned long offset, addr = (unsigned long)from; - int type = in_nmi() ? KM_NMI : KM_IRQ0; - unsigned long size, len = 0; - struct page *page; - void *map; - int ret; - - do { - ret = __get_user_pages_fast(addr, 1, 0, &page); - if (!ret) - break; - - offset = addr & (PAGE_SIZE - 1); - size = min(PAGE_SIZE - offset, n - len); - - map = kmap_atomic(page, type); - memcpy(to, map+offset, size); - kunmap_atomic(map, type); - put_page(page); - - len += size; - to += size; - addr += size; - - } while (len < n); - - return len; -} - static int copy_stack_frame(const void __user *fp, struct stack_frame *frame) { unsigned long bytes; diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c index 44f6ed42a934..7eb78be3b229 100644 --- a/arch/x86/kernel/cpu/perf_event_intel.c +++ b/arch/x86/kernel/cpu/perf_event_intel.c @@ -547,7 +547,7 @@ static void intel_pmu_disable_event(struct perf_event *event) x86_pmu_disable_event(event); if (unlikely(event->attr.precise)) - intel_pmu_pebs_disable(hwc); + intel_pmu_pebs_disable(event); } static void intel_pmu_enable_fixed(struct hw_perf_event *hwc) @@ -600,7 +600,7 @@ static void intel_pmu_enable_event(struct perf_event *event) } if (unlikely(event->attr.precise)) - intel_pmu_pebs_enable(hwc); + intel_pmu_pebs_enable(event); __x86_pmu_enable_event(hwc); } diff --git a/arch/x86/kernel/cpu/perf_event_intel_ds.c b/arch/x86/kernel/cpu/perf_event_intel_ds.c index 0d994ef213b9..50e6ff3281fc 100644 --- a/arch/x86/kernel/cpu/perf_event_intel_ds.c +++ b/arch/x86/kernel/cpu/perf_event_intel_ds.c @@ -331,26 +331,32 @@ intel_pebs_constraints(struct perf_event *event) return &emptyconstraint; } -static void intel_pmu_pebs_enable(struct hw_perf_event *hwc) +static void intel_pmu_pebs_enable(struct perf_event *event) { struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); + struct hw_perf_event *hwc = &event->hw; u64 val = cpuc->pebs_enabled; hwc->config &= ~ARCH_PERFMON_EVENTSEL_INT; val |= 1ULL << hwc->idx; wrmsrl(MSR_IA32_PEBS_ENABLE, val); + + intel_pmu_lbr_enable(event); } -static void intel_pmu_pebs_disable(struct hw_perf_event *hwc) +static void intel_pmu_pebs_disable(struct perf_event *event) { struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); + struct hw_perf_event *hwc = &event->hw; u64 val = cpuc->pebs_enabled; val &= ~(1ULL << hwc->idx); wrmsrl(MSR_IA32_PEBS_ENABLE, val); hwc->config |= ARCH_PERFMON_EVENTSEL_INT; + + intel_pmu_lbr_disable(event); } static void intel_pmu_pebs_enable_all(void) @@ -369,6 +375,70 @@ static void intel_pmu_pebs_disable_all(void) wrmsrl(MSR_IA32_PEBS_ENABLE, 0); } +#include + +#define MAX_INSN_SIZE 16 + +static inline bool kernel_ip(unsigned long ip) +{ +#ifdef CONFIG_X86_32 + return ip > PAGE_OFFSET; +#else + return (long)ip < 0; +#endif +} + +static int intel_pmu_pebs_fixup_ip(struct pt_regs *regs) +{ + struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); + unsigned long from = cpuc->lbr_entries[0].from; + unsigned long old_to, to = cpuc->lbr_entries[0].to; + unsigned long ip = regs->ip; + + if (!cpuc->lbr_stack.nr || !from || !to) + return 0; + + if (ip < to) + return 0; + + /* + * We sampled a branch insn, rewind using the LBR stack + */ + if (ip == to) { + regs->ip = from; + return 1; + } + + do { + struct insn insn; + u8 buf[MAX_INSN_SIZE]; + void *kaddr; + + old_to = to; + if (!kernel_ip(ip)) { + int bytes, size = min_t(int, MAX_INSN_SIZE, ip - to); + + bytes = copy_from_user_nmi(buf, (void __user *)to, size); + if (bytes != size) + return 0; + + kaddr = buf; + } else + kaddr = (void *)to; + + kernel_insn_init(&insn, kaddr); + insn_get_length(&insn); + to += insn.length; + } while (to < ip); + + if (to == ip) { + regs->ip = old_to; + return 1; + } + + return 0; +} + static int intel_pmu_save_and_restart(struct perf_event *event); static void intel_pmu_disable_event(struct perf_event *event); @@ -424,6 +494,11 @@ static void intel_pmu_drain_pebs_core(struct pt_regs *iregs) regs.bp = at->bp; regs.sp = at->sp; + if (intel_pmu_pebs_fixup_ip(®s)) + regs.flags |= PERF_EFLAGS_EXACT; + else + regs.flags &= ~PERF_EFLAGS_EXACT; + if (perf_event_overflow(event, 1, &data, ®s)) intel_pmu_disable_event(event); @@ -487,6 +562,11 @@ static void intel_pmu_drain_pebs_nhm(struct pt_regs *iregs) regs.bp = at->bp; regs.sp = at->sp; + if (intel_pmu_pebs_fixup_ip(®s)) + regs.flags |= PERF_EFLAGS_EXACT; + else + regs.flags &= ~PERF_EFLAGS_EXACT; + if (perf_event_overflow(event, 1, &data, ®s)) intel_pmu_disable_event(event); } diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index ab4fd9ede264..be85f7c4a94f 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -294,6 +294,12 @@ struct perf_event_mmap_page { #define PERF_RECORD_MISC_USER (2 << 0) #define PERF_RECORD_MISC_HYPERVISOR (3 << 0) +#define PERF_RECORD_MISC_EXACT (1 << 14) +/* + * Reserve the last bit to indicate some extended misc field + */ +#define PERF_RECORD_MISC_EXT_RESERVED (1 << 15) + struct perf_event_header { __u32 type; __u16 misc; -- cgit 1.4.1 From faa4602e47690fb11221e00f9b9697c8dc0d4b19 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Thu, 25 Mar 2010 14:51:50 +0100 Subject: x86, perf, bts, mm: Delete the never used BTS-ptrace code Support for the PMU's BTS features has been upstreamed in v2.6.32, but we still have the old and disabled ptrace-BTS, as Linus noticed it not so long ago. It's buggy: TIF_DEBUGCTLMSR is trampling all over that MSR without regard for other uses (perf) and doesn't provide the flexibility needed for perf either. Its users are ptrace-block-step and ptrace-bts, since ptrace-bts was never used and ptrace-block-step can be implemented using a much simpler approach. So axe all 3000 lines of it. That includes the *locked_memory*() APIs in mm/mlock.c as well. Reported-by: Linus Torvalds Signed-off-by: Peter Zijlstra Cc: Roland McGrath Cc: Oleg Nesterov Cc: Markus Metzger Cc: Steven Rostedt Cc: Andrew Morton LKML-Reference: <20100325135413.938004390@chello.nl> Signed-off-by: Ingo Molnar --- arch/x86/Kconfig.cpu | 20 - arch/x86/Kconfig.debug | 9 - arch/x86/include/asm/ds.h | 302 -------- arch/x86/include/asm/processor.h | 33 +- arch/x86/include/asm/ptrace-abi.h | 57 +- arch/x86/include/asm/ptrace.h | 6 - arch/x86/include/asm/thread_info.h | 6 +- arch/x86/kernel/Makefile | 2 - arch/x86/kernel/cpu/intel.c | 2 - arch/x86/kernel/ds.c | 1437 ------------------------------------ arch/x86/kernel/ds_selftest.c | 408 ---------- arch/x86/kernel/ds_selftest.h | 15 - arch/x86/kernel/dumpstack.c | 5 - arch/x86/kernel/kprobes.c | 6 +- arch/x86/kernel/process.c | 9 - arch/x86/kernel/process_32.c | 8 - arch/x86/kernel/process_64.c | 8 - arch/x86/kernel/ptrace.c | 382 ---------- arch/x86/kernel/step.c | 36 +- arch/x86/kernel/traps.c | 5 - include/linux/ftrace.h | 12 - include/linux/mm.h | 4 - include/linux/ptrace.h | 12 - include/linux/sched.h | 9 - kernel/fork.c | 3 - kernel/ptrace.c | 1 - kernel/sched.c | 43 -- kernel/trace/Kconfig | 11 - kernel/trace/Makefile | 1 - kernel/trace/trace.h | 4 - kernel/trace/trace_entries.h | 12 - kernel/trace/trace_hw_branches.c | 312 -------- kernel/trace/trace_selftest.c | 57 -- mm/mlock.c | 41 - 34 files changed, 9 insertions(+), 3269 deletions(-) delete mode 100644 arch/x86/include/asm/ds.h delete mode 100644 arch/x86/kernel/ds.c delete mode 100644 arch/x86/kernel/ds_selftest.c delete mode 100644 arch/x86/kernel/ds_selftest.h delete mode 100644 kernel/trace/trace_hw_branches.c (limited to 'include') diff --git a/arch/x86/Kconfig.cpu b/arch/x86/Kconfig.cpu index a19829374e6a..918fbb1855cc 100644 --- a/arch/x86/Kconfig.cpu +++ b/arch/x86/Kconfig.cpu @@ -502,23 +502,3 @@ config CPU_SUP_UMC_32 CPU might render the kernel unbootable. If unsure, say N. - -config X86_DS - def_bool X86_PTRACE_BTS - depends on X86_DEBUGCTLMSR - select HAVE_HW_BRANCH_TRACER - -config X86_PTRACE_BTS - bool "Branch Trace Store" - default y - depends on X86_DEBUGCTLMSR - depends on BROKEN - ---help--- - This adds a ptrace interface to the hardware's branch trace store. - - Debuggers may use it to collect an execution trace of the debugged - application in order to answer the question 'how did I get here?'. - Debuggers may trace user mode as well as kernel mode. - - Say Y unless there is no application development on this machine - and you want to save a small amount of code size. diff --git a/arch/x86/Kconfig.debug b/arch/x86/Kconfig.debug index bc01e3ebfeb2..bd58c8abbfbd 100644 --- a/arch/x86/Kconfig.debug +++ b/arch/x86/Kconfig.debug @@ -174,15 +174,6 @@ config IOMMU_LEAK Add a simple leak tracer to the IOMMU code. This is useful when you are debugging a buggy device driver that leaks IOMMU mappings. -config X86_DS_SELFTEST - bool "DS selftest" - default y - depends on DEBUG_KERNEL - depends on X86_DS - ---help--- - Perform Debug Store selftests at boot time. - If in doubt, say "N". - config HAVE_MMIOTRACE_SUPPORT def_bool y diff --git a/arch/x86/include/asm/ds.h b/arch/x86/include/asm/ds.h deleted file mode 100644 index 70dac199b093..000000000000 --- a/arch/x86/include/asm/ds.h +++ /dev/null @@ -1,302 +0,0 @@ -/* - * Debug Store (DS) support - * - * This provides a low-level interface to the hardware's Debug Store - * feature that is used for branch trace store (BTS) and - * precise-event based sampling (PEBS). - * - * It manages: - * - DS and BTS hardware configuration - * - buffer overflow handling (to be done) - * - buffer access - * - * It does not do: - * - security checking (is the caller allowed to trace the task) - * - buffer allocation (memory accounting) - * - * - * Copyright (C) 2007-2009 Intel Corporation. - * Markus Metzger , 2007-2009 - */ - -#ifndef _ASM_X86_DS_H -#define _ASM_X86_DS_H - - -#include -#include -#include - - -#ifdef CONFIG_X86_DS - -struct task_struct; -struct ds_context; -struct ds_tracer; -struct bts_tracer; -struct pebs_tracer; - -typedef void (*bts_ovfl_callback_t)(struct bts_tracer *); -typedef void (*pebs_ovfl_callback_t)(struct pebs_tracer *); - - -/* - * A list of features plus corresponding macros to talk about them in - * the ds_request function's flags parameter. - * - * We use the enum to index an array of corresponding control bits; - * we use the macro to index a flags bit-vector. - */ -enum ds_feature { - dsf_bts = 0, - dsf_bts_kernel, -#define BTS_KERNEL (1 << dsf_bts_kernel) - /* trace kernel-mode branches */ - - dsf_bts_user, -#define BTS_USER (1 << dsf_bts_user) - /* trace user-mode branches */ - - dsf_bts_overflow, - dsf_bts_max, - dsf_pebs = dsf_bts_max, - - dsf_pebs_max, - dsf_ctl_max = dsf_pebs_max, - dsf_bts_timestamps = dsf_ctl_max, -#define BTS_TIMESTAMPS (1 << dsf_bts_timestamps) - /* add timestamps into BTS trace */ - -#define BTS_USER_FLAGS (BTS_KERNEL | BTS_USER | BTS_TIMESTAMPS) -}; - - -/* - * Request BTS or PEBS - * - * Due to alignement constraints, the actual buffer may be slightly - * smaller than the requested or provided buffer. - * - * Returns a pointer to a tracer structure on success, or - * ERR_PTR(errcode) on failure. - * - * The interrupt threshold is independent from the overflow callback - * to allow users to use their own overflow interrupt handling mechanism. - * - * The function might sleep. - * - * task: the task to request recording for - * cpu: the cpu to request recording for - * base: the base pointer for the (non-pageable) buffer; - * size: the size of the provided buffer in bytes - * ovfl: pointer to a function to be called on buffer overflow; - * NULL if cyclic buffer requested - * th: the interrupt threshold in records from the end of the buffer; - * -1 if no interrupt threshold is requested. - * flags: a bit-mask of the above flags - */ -extern struct bts_tracer *ds_request_bts_task(struct task_struct *task, - void *base, size_t size, - bts_ovfl_callback_t ovfl, - size_t th, unsigned int flags); -extern struct bts_tracer *ds_request_bts_cpu(int cpu, void *base, size_t size, - bts_ovfl_callback_t ovfl, - size_t th, unsigned int flags); -extern struct pebs_tracer *ds_request_pebs_task(struct task_struct *task, - void *base, size_t size, - pebs_ovfl_callback_t ovfl, - size_t th, unsigned int flags); -extern struct pebs_tracer *ds_request_pebs_cpu(int cpu, - void *base, size_t size, - pebs_ovfl_callback_t ovfl, - size_t th, unsigned int flags); - -/* - * Release BTS or PEBS resources - * Suspend and resume BTS or PEBS tracing - * - * Must be called with irq's enabled. - * - * tracer: the tracer handle returned from ds_request_~() - */ -extern void ds_release_bts(struct bts_tracer *tracer); -extern void ds_suspend_bts(struct bts_tracer *tracer); -extern void ds_resume_bts(struct bts_tracer *tracer); -extern void ds_release_pebs(struct pebs_tracer *tracer); -extern void ds_suspend_pebs(struct pebs_tracer *tracer); -extern void ds_resume_pebs(struct pebs_tracer *tracer); - -/* - * Release BTS or PEBS resources - * Suspend and resume BTS or PEBS tracing - * - * Cpu tracers must call this on the traced cpu. - * Task tracers must call ds_release_~_noirq() for themselves. - * - * May be called with irq's disabled. - * - * Returns 0 if successful; - * -EPERM if the cpu tracer does not trace the current cpu. - * -EPERM if the task tracer does not trace itself. - * - * tracer: the tracer handle returned from ds_request_~() - */ -extern int ds_release_bts_noirq(struct bts_tracer *tracer); -extern int ds_suspend_bts_noirq(struct bts_tracer *tracer); -extern int ds_resume_bts_noirq(struct bts_tracer *tracer); -extern int ds_release_pebs_noirq(struct pebs_tracer *tracer); -extern int ds_suspend_pebs_noirq(struct pebs_tracer *tracer); -extern int ds_resume_pebs_noirq(struct pebs_tracer *tracer); - - -/* - * The raw DS buffer state as it is used for BTS and PEBS recording. - * - * This is the low-level, arch-dependent interface for working - * directly on the raw trace data. - */ -struct ds_trace { - /* the number of bts/pebs records */ - size_t n; - /* the size of a bts/pebs record in bytes */ - size_t size; - /* pointers into the raw buffer: - - to the first entry */ - void *begin; - /* - one beyond the last entry */ - void *end; - /* - one beyond the newest entry */ - void *top; - /* - the interrupt threshold */ - void *ith; - /* flags given on ds_request() */ - unsigned int flags; -}; - -/* - * An arch-independent view on branch trace data. - */ -enum bts_qualifier { - bts_invalid, -#define BTS_INVALID bts_invalid - - bts_branch, -#define BTS_BRANCH bts_branch - - bts_task_arrives, -#define BTS_TASK_ARRIVES bts_task_arrives - - bts_task_departs, -#define BTS_TASK_DEPARTS bts_task_departs - - bts_qual_bit_size = 4, - bts_qual_max = (1 << bts_qual_bit_size), -}; - -struct bts_struct { - __u64 qualifier; - union { - /* BTS_BRANCH */ - struct { - __u64 from; - __u64 to; - } lbr; - /* BTS_TASK_ARRIVES or BTS_TASK_DEPARTS */ - struct { - __u64 clock; - pid_t pid; - } event; - } variant; -}; - - -/* - * The BTS state. - * - * This gives access to the raw DS state and adds functions to provide - * an arch-independent view of the BTS data. - */ -struct bts_trace { - struct ds_trace ds; - - int (*read)(struct bts_tracer *tracer, const void *at, - struct bts_struct *out); - int (*write)(struct bts_tracer *tracer, const struct bts_struct *in); -}; - - -/* - * The PEBS state. - * - * This gives access to the raw DS state and the PEBS-specific counter - * reset value. - */ -struct pebs_trace { - struct ds_trace ds; - - /* the number of valid counters in the below array */ - unsigned int counters; - -#define MAX_PEBS_COUNTERS 4 - /* the counter reset value */ - unsigned long long counter_reset[MAX_PEBS_COUNTERS]; -}; - - -/* - * Read the BTS or PEBS trace. - * - * Returns a view on the trace collected for the parameter tracer. - * - * The view remains valid as long as the traced task is not running or - * the tracer is suspended. - * Writes into the trace buffer are not reflected. - * - * tracer: the tracer handle returned from ds_request_~() - */ -extern const struct bts_trace *ds_read_bts(struct bts_tracer *tracer); -extern const struct pebs_trace *ds_read_pebs(struct pebs_tracer *tracer); - - -/* - * Reset the write pointer of the BTS/PEBS buffer. - * - * Returns 0 on success; -Eerrno on error - * - * tracer: the tracer handle returned from ds_request_~() - */ -extern int ds_reset_bts(struct bts_tracer *tracer); -extern int ds_reset_pebs(struct pebs_tracer *tracer); - -/* - * Set the PEBS counter reset value. - * - * Returns 0 on success; -Eerrno on error - * - * tracer: the tracer handle returned from ds_request_pebs() - * counter: the index of the counter - * value: the new counter reset value - */ -extern int ds_set_pebs_reset(struct pebs_tracer *tracer, - unsigned int counter, u64 value); - -/* - * Initialization - */ -struct cpuinfo_x86; -extern void __cpuinit ds_init_intel(struct cpuinfo_x86 *); - -/* - * Context switch work - */ -extern void ds_switch_to(struct task_struct *prev, struct task_struct *next); - -#else /* CONFIG_X86_DS */ - -struct cpuinfo_x86; -static inline void __cpuinit ds_init_intel(struct cpuinfo_x86 *ignored) {} -static inline void ds_switch_to(struct task_struct *prev, - struct task_struct *next) {} - -#endif /* CONFIG_X86_DS */ -#endif /* _ASM_X86_DS_H */ diff --git a/arch/x86/include/asm/processor.h b/arch/x86/include/asm/processor.h index b753ea59703a..5bec21a66dc5 100644 --- a/arch/x86/include/asm/processor.h +++ b/arch/x86/include/asm/processor.h @@ -21,7 +21,6 @@ struct mm_struct; #include #include #include -#include #include #include @@ -29,6 +28,7 @@ struct mm_struct; #include #include #include +#include #define HBP_NUM 4 /* @@ -473,10 +473,6 @@ struct thread_struct { unsigned long iopl; /* Max allowed port in the bitmap, in bytes: */ unsigned io_bitmap_max; -/* MSR_IA32_DEBUGCTLMSR value to switch in if TIF_DEBUGCTLMSR is set. */ - unsigned long debugctlmsr; - /* Debug Store context; see asm/ds.h */ - struct ds_context *ds_ctx; }; static inline unsigned long native_get_debugreg(int regno) @@ -814,21 +810,6 @@ static inline unsigned long get_debugctlmsr(void) return debugctlmsr; } -static inline unsigned long get_debugctlmsr_on_cpu(int cpu) -{ - u64 debugctlmsr = 0; - u32 val1, val2; - -#ifndef CONFIG_X86_DEBUGCTLMSR - if (boot_cpu_data.x86 < 6) - return 0; -#endif - rdmsr_on_cpu(cpu, MSR_IA32_DEBUGCTLMSR, &val1, &val2); - debugctlmsr = val1 | ((u64)val2 << 32); - - return debugctlmsr; -} - static inline void update_debugctlmsr(unsigned long debugctlmsr) { #ifndef CONFIG_X86_DEBUGCTLMSR @@ -838,18 +819,6 @@ static inline void update_debugctlmsr(unsigned long debugctlmsr) wrmsrl(MSR_IA32_DEBUGCTLMSR, debugctlmsr); } -static inline void update_debugctlmsr_on_cpu(int cpu, - unsigned long debugctlmsr) -{ -#ifndef CONFIG_X86_DEBUGCTLMSR - if (boot_cpu_data.x86 < 6) - return; -#endif - wrmsr_on_cpu(cpu, MSR_IA32_DEBUGCTLMSR, - (u32)((u64)debugctlmsr), - (u32)((u64)debugctlmsr >> 32)); -} - /* * from system description table in BIOS. Mostly for MCA use, but * others may find it useful: diff --git a/arch/x86/include/asm/ptrace-abi.h b/arch/x86/include/asm/ptrace-abi.h index 86723035a515..52b098a6eebb 100644 --- a/arch/x86/include/asm/ptrace-abi.h +++ b/arch/x86/include/asm/ptrace-abi.h @@ -82,61 +82,6 @@ #ifndef __ASSEMBLY__ #include - -/* configuration/status structure used in PTRACE_BTS_CONFIG and - PTRACE_BTS_STATUS commands. -*/ -struct ptrace_bts_config { - /* requested or actual size of BTS buffer in bytes */ - __u32 size; - /* bitmask of below flags */ - __u32 flags; - /* buffer overflow signal */ - __u32 signal; - /* actual size of bts_struct in bytes */ - __u32 bts_size; -}; -#endif /* __ASSEMBLY__ */ - -#define PTRACE_BTS_O_TRACE 0x1 /* branch trace */ -#define PTRACE_BTS_O_SCHED 0x2 /* scheduling events w/ jiffies */ -#define PTRACE_BTS_O_SIGNAL 0x4 /* send SIG on buffer overflow - instead of wrapping around */ -#define PTRACE_BTS_O_ALLOC 0x8 /* (re)allocate buffer */ - -#define PTRACE_BTS_CONFIG 40 -/* Configure branch trace recording. - ADDR points to a struct ptrace_bts_config. - DATA gives the size of that buffer. - A new buffer is allocated, if requested in the flags. - An overflow signal may only be requested for new buffers. - Returns the number of bytes read. -*/ -#define PTRACE_BTS_STATUS 41 -/* Return the current configuration in a struct ptrace_bts_config - pointed to by ADDR; DATA gives the size of that buffer. - Returns the number of bytes written. -*/ -#define PTRACE_BTS_SIZE 42 -/* Return the number of available BTS records for draining. - DATA and ADDR are ignored. -*/ -#define PTRACE_BTS_GET 43 -/* Get a single BTS record. - DATA defines the index into the BTS array, where 0 is the newest - entry, and higher indices refer to older entries. - ADDR is pointing to struct bts_struct (see asm/ds.h). -*/ -#define PTRACE_BTS_CLEAR 44 -/* Clear the BTS buffer. - DATA and ADDR are ignored. -*/ -#define PTRACE_BTS_DRAIN 45 -/* Read all available BTS records and clear the buffer. - ADDR points to an array of struct bts_struct. - DATA gives the size of that buffer. - BTS records are read from oldest to newest. - Returns number of BTS records drained. -*/ +#endif #endif /* _ASM_X86_PTRACE_ABI_H */ diff --git a/arch/x86/include/asm/ptrace.h b/arch/x86/include/asm/ptrace.h index 69a686a7dff0..78cd1ea94500 100644 --- a/arch/x86/include/asm/ptrace.h +++ b/arch/x86/include/asm/ptrace.h @@ -289,12 +289,6 @@ extern int do_get_thread_area(struct task_struct *p, int idx, extern int do_set_thread_area(struct task_struct *p, int idx, struct user_desc __user *info, int can_allocate); -#ifdef CONFIG_X86_PTRACE_BTS -extern void ptrace_bts_untrace(struct task_struct *tsk); - -#define arch_ptrace_untrace(tsk) ptrace_bts_untrace(tsk) -#endif /* CONFIG_X86_PTRACE_BTS */ - #endif /* __KERNEL__ */ #endif /* !__ASSEMBLY__ */ diff --git a/arch/x86/include/asm/thread_info.h b/arch/x86/include/asm/thread_info.h index e0d28901e969..dc85e12d1405 100644 --- a/arch/x86/include/asm/thread_info.h +++ b/arch/x86/include/asm/thread_info.h @@ -92,8 +92,6 @@ struct thread_info { #define TIF_IO_BITMAP 22 /* uses I/O bitmap */ #define TIF_FREEZE 23 /* is freezing for suspend */ #define TIF_FORCED_TF 24 /* true if TF in eflags artificially */ -#define TIF_DEBUGCTLMSR 25 /* uses thread_struct.debugctlmsr */ -#define TIF_DS_AREA_MSR 26 /* uses thread_struct.ds_area_msr */ #define TIF_LAZY_MMU_UPDATES 27 /* task is updating the mmu lazily */ #define TIF_SYSCALL_TRACEPOINT 28 /* syscall tracepoint instrumentation */ @@ -115,8 +113,6 @@ struct thread_info { #define _TIF_IO_BITMAP (1 << TIF_IO_BITMAP) #define _TIF_FREEZE (1 << TIF_FREEZE) #define _TIF_FORCED_TF (1 << TIF_FORCED_TF) -#define _TIF_DEBUGCTLMSR (1 << TIF_DEBUGCTLMSR) -#define _TIF_DS_AREA_MSR (1 << TIF_DS_AREA_MSR) #define _TIF_LAZY_MMU_UPDATES (1 << TIF_LAZY_MMU_UPDATES) #define _TIF_SYSCALL_TRACEPOINT (1 << TIF_SYSCALL_TRACEPOINT) @@ -147,7 +143,7 @@ struct thread_info { /* flags to check in __switch_to() */ #define _TIF_WORK_CTXSW \ - (_TIF_IO_BITMAP|_TIF_DEBUGCTLMSR|_TIF_DS_AREA_MSR|_TIF_NOTSC) + (_TIF_IO_BITMAP|_TIF_NOTSC) #define _TIF_WORK_CTXSW_PREV (_TIF_WORK_CTXSW|_TIF_USER_RETURN_NOTIFY) #define _TIF_WORK_CTXSW_NEXT (_TIF_WORK_CTXSW|_TIF_DEBUG) diff --git a/arch/x86/kernel/Makefile b/arch/x86/kernel/Makefile index 4c58352209e0..e77b22083721 100644 --- a/arch/x86/kernel/Makefile +++ b/arch/x86/kernel/Makefile @@ -47,8 +47,6 @@ obj-$(CONFIG_X86_TRAMPOLINE) += trampoline.o obj-y += process.o obj-y += i387.o xsave.o obj-y += ptrace.o -obj-$(CONFIG_X86_DS) += ds.o -obj-$(CONFIG_X86_DS_SELFTEST) += ds_selftest.o obj-$(CONFIG_X86_32) += tls.o obj-$(CONFIG_IA32_EMULATION) += tls.o obj-y += step.o diff --git a/arch/x86/kernel/cpu/intel.c b/arch/x86/kernel/cpu/intel.c index 7e1cca13af35..d72377c41c76 100644 --- a/arch/x86/kernel/cpu/intel.c +++ b/arch/x86/kernel/cpu/intel.c @@ -12,7 +12,6 @@ #include #include #include -#include #include #include @@ -367,7 +366,6 @@ static void __cpuinit init_intel(struct cpuinfo_x86 *c) set_cpu_cap(c, X86_FEATURE_BTS); if (!(l1 & (1<<12))) set_cpu_cap(c, X86_FEATURE_PEBS); - ds_init_intel(c); } if (c->x86 == 6 && c->x86_model == 29 && cpu_has_clflush) diff --git a/arch/x86/kernel/ds.c b/arch/x86/kernel/ds.c deleted file mode 100644 index 1c47390dd0e5..000000000000 --- a/arch/x86/kernel/ds.c +++ /dev/null @@ -1,1437 +0,0 @@ -/* - * Debug Store support - * - * This provides a low-level interface to the hardware's Debug Store - * feature that is used for branch trace store (BTS) and - * precise-event based sampling (PEBS). - * - * It manages: - * - DS and BTS hardware configuration - * - buffer overflow handling (to be done) - * - buffer access - * - * It does not do: - * - security checking (is the caller allowed to trace the task) - * - buffer allocation (memory accounting) - * - * - * Copyright (C) 2007-2009 Intel Corporation. - * Markus Metzger , 2007-2009 - */ - -#include -#include -#include -#include -#include -#include -#include - -#include - -#include "ds_selftest.h" - -/* - * The configuration for a particular DS hardware implementation: - */ -struct ds_configuration { - /* The name of the configuration: */ - const char *name; - - /* The size of pointer-typed fields in DS, BTS, and PEBS: */ - unsigned char sizeof_ptr_field; - - /* The size of a BTS/PEBS record in bytes: */ - unsigned char sizeof_rec[2]; - - /* The number of pebs counter reset values in the DS structure. */ - unsigned char nr_counter_reset; - - /* Control bit-masks indexed by enum ds_feature: */ - unsigned long ctl[dsf_ctl_max]; -}; -static struct ds_configuration ds_cfg __read_mostly; - - -/* Maximal size of a DS configuration: */ -#define MAX_SIZEOF_DS 0x80 - -/* Maximal size of a BTS record: */ -#define MAX_SIZEOF_BTS (3 * 8) - -/* BTS and PEBS buffer alignment: */ -#define DS_ALIGNMENT (1 << 3) - -/* Number of buffer pointers in DS: */ -#define NUM_DS_PTR_FIELDS 8 - -/* Size of a pebs reset value in DS: */ -#define PEBS_RESET_FIELD_SIZE 8 - -/* Mask of control bits in the DS MSR register: */ -#define BTS_CONTROL \ - ( ds_cfg.ctl[dsf_bts] | \ - ds_cfg.ctl[dsf_bts_kernel] | \ - ds_cfg.ctl[dsf_bts_user] | \ - ds_cfg.ctl[dsf_bts_overflow] ) - -/* - * A BTS or PEBS tracer. - * - * This holds the configuration of the tracer and serves as a handle - * to identify tracers. - */ -struct ds_tracer { - /* The DS context (partially) owned by this tracer. */ - struct ds_context *context; - /* The buffer provided on ds_request() and its size in bytes. */ - void *buffer; - size_t size; -}; - -struct bts_tracer { - /* The common DS part: */ - struct ds_tracer ds; - - /* The trace including the DS configuration: */ - struct bts_trace trace; - - /* Buffer overflow notification function: */ - bts_ovfl_callback_t ovfl; - - /* Active flags affecting trace collection. */ - unsigned int flags; -}; - -struct pebs_tracer { - /* The common DS part: */ - struct ds_tracer ds; - - /* The trace including the DS configuration: */ - struct pebs_trace trace; - - /* Buffer overflow notification function: */ - pebs_ovfl_callback_t ovfl; -}; - -/* - * Debug Store (DS) save area configuration (see Intel64 and IA32 - * Architectures Software Developer's Manual, section 18.5) - * - * The DS configuration consists of the following fields; different - * architetures vary in the size of those fields. - * - * - double-word aligned base linear address of the BTS buffer - * - write pointer into the BTS buffer - * - end linear address of the BTS buffer (one byte beyond the end of - * the buffer) - * - interrupt pointer into BTS buffer - * (interrupt occurs when write pointer passes interrupt pointer) - * - double-word aligned base linear address of the PEBS buffer - * - write pointer into the PEBS buffer - * - end linear address of the PEBS buffer (one byte beyond the end of - * the buffer) - * - interrupt pointer into PEBS buffer - * (interrupt occurs when write pointer passes interrupt pointer) - * - value to which counter is reset following counter overflow - * - * Later architectures use 64bit pointers throughout, whereas earlier - * architectures use 32bit pointers in 32bit mode. - * - * - * We compute the base address for the first 8 fields based on: - * - the field size stored in the DS configuration - * - the relative field position - * - an offset giving the start of the respective region - * - * This offset is further used to index various arrays holding - * information for BTS and PEBS at the respective index. - * - * On later 32bit processors, we only access the lower 32bit of the - * 64bit pointer fields. The upper halves will be zeroed out. - */ - -enum ds_field { - ds_buffer_base = 0, - ds_index, - ds_absolute_maximum, - ds_interrupt_threshold, -}; - -enum ds_qualifier { - ds_bts = 0, - ds_pebs -}; - -static inline unsigned long -ds_get(const unsigned char *base, enum ds_qualifier qual, enum ds_field field) -{ - base += (ds_cfg.sizeof_ptr_field * (field + (4 * qual))); - return *(unsigned long *)base; -} - -static inline void -ds_set(unsigned char *base, enum ds_qualifier qual, enum ds_field field, - unsigned long value) -{ - base += (ds_cfg.sizeof_ptr_field * (field + (4 * qual))); - (*(unsigned long *)base) = value; -} - - -/* - * Locking is done only for allocating BTS or PEBS resources. - */ -static DEFINE_SPINLOCK(ds_lock); - -/* - * We either support (system-wide) per-cpu or per-thread allocation. - * We distinguish the two based on the task_struct pointer, where a - * NULL pointer indicates per-cpu allocation for the current cpu. - * - * Allocations are use-counted. As soon as resources are allocated, - * further allocations must be of the same type (per-cpu or - * per-thread). We model this by counting allocations (i.e. the number - * of tracers of a certain type) for one type negatively: - * =0 no tracers - * >0 number of per-thread tracers - * <0 number of per-cpu tracers - * - * Tracers essentially gives the number of ds contexts for a certain - * type of allocation. - */ -static atomic_t tracers = ATOMIC_INIT(0); - -static inline int get_tracer(struct task_struct *task) -{ - int error; - - spin_lock_irq(&ds_lock); - - if (task) { - error = -EPERM; - if (atomic_read(&tracers) < 0) - goto out; - atomic_inc(&tracers); - } else { - error = -EPERM; - if (atomic_read(&tracers) > 0) - goto out; - atomic_dec(&tracers); - } - - error = 0; -out: - spin_unlock_irq(&ds_lock); - return error; -} - -static inline void put_tracer(struct task_struct *task) -{ - if (task) - atomic_dec(&tracers); - else - atomic_inc(&tracers); -} - -/* - * The DS context is either attached to a thread or to a cpu: - * - in the former case, the thread_struct contains a pointer to the - * attached context. - * - in the latter case, we use a static array of per-cpu context - * pointers. - * - * Contexts are use-counted. They are allocated on first access and - * deallocated when the last user puts the context. - */ -struct ds_context { - /* The DS configuration; goes into MSR_IA32_DS_AREA: */ - unsigned char ds[MAX_SIZEOF_DS]; - - /* The owner of the BTS and PEBS configuration, respectively: */ - struct bts_tracer *bts_master; - struct pebs_tracer *pebs_master; - - /* Use count: */ - unsigned long count; - - /* Pointer to the context pointer field: */ - struct ds_context **this; - - /* The traced task; NULL for cpu tracing: */ - struct task_struct *task; - - /* The traced cpu; only valid if task is NULL: */ - int cpu; -}; - -static DEFINE_PER_CPU(struct ds_context *, cpu_ds_context); - - -static struct ds_context *ds_get_context(struct task_struct *task, int cpu) -{ - struct ds_context **p_context = - (task ? &task->thread.ds_ctx : &per_cpu(cpu_ds_context, cpu)); - struct ds_context *context = NULL; - struct ds_context *new_context = NULL; - - /* Chances are small that we already have a context. */ - new_context = kzalloc(sizeof(*new_context), GFP_KERNEL); - if (!new_context) - return NULL; - - spin_lock_irq(&ds_lock); - - context = *p_context; - if (likely(!context)) { - context = new_context; - - context->this = p_context; - context->task = task; - context->cpu = cpu; - context->count = 0; - - *p_context = context; - } - - context->count++; - - spin_unlock_irq(&ds_lock); - - if (context != new_context) - kfree(new_context); - - return context; -} - -static void ds_put_context(struct ds_context *context) -{ - struct task_struct *task; - unsigned long irq; - - if (!context) - return; - - spin_lock_irqsave(&ds_lock, irq); - - if (--context->count) { - spin_unlock_irqrestore(&ds_lock, irq); - return; - } - - *(context->this) = NULL; - - task = context->task; - - if (task) - clear_tsk_thread_flag(task, TIF_DS_AREA_MSR); - - /* - * We leave the (now dangling) pointer to the DS configuration in - * the DS_AREA msr. This is as good or as bad as replacing it with - * NULL - the hardware would crash if we enabled tracing. - * - * This saves us some problems with having to write an msr on a - * different cpu while preventing others from doing the same for the - * next context for that same cpu. - */ - - spin_unlock_irqrestore(&ds_lock, irq); - - /* The context might still be in use for context switching. */ - if (task && (task != current)) - wait_task_context_switch(task); - - kfree(context); -} - -static void ds_install_ds_area(struct ds_context *context) -{ - unsigned long ds; - - ds = (unsigned long)context->ds; - - /* - * There is a race between the bts master and the pebs master. - * - * The thread/cpu access is synchronized via get/put_cpu() for - * task tracing and via wrmsr_on_cpu for cpu tracing. - * - * If bts and pebs are collected for the same task or same cpu, - * the same confiuration is written twice. - */ - if (context->task) { - get_cpu(); - if (context->task == current) - wrmsrl(MSR_IA32_DS_AREA, ds); - set_tsk_thread_flag(context->task, TIF_DS_AREA_MSR); - put_cpu(); - } else - wrmsr_on_cpu(context->cpu, MSR_IA32_DS_AREA, - (u32)((u64)ds), (u32)((u64)ds >> 32)); -} - -/* - * Call the tracer's callback on a buffer overflow. - * - * context: the ds context - * qual: the buffer type - */ -static void ds_overflow(struct ds_context *context, enum ds_qualifier qual) -{ - switch (qual) { - case ds_bts: - if (context->bts_master && - context->bts_master->ovfl) - context->bts_master->ovfl(context->bts_master); - break; - case ds_pebs: - if (context->pebs_master && - context->pebs_master->ovfl) - context->pebs_master->ovfl(context->pebs_master); - break; - } -} - - -/* - * Write raw data into the BTS or PEBS buffer. - * - * The remainder of any partially written record is zeroed out. - * - * context: the DS context - * qual: the buffer type - * record: the data to write - * size: the size of the data - */ -static int ds_write(struct ds_context *context, enum ds_qualifier qual, - const void *record, size_t size) -{ - int bytes_written = 0; - - if (!record) - return -EINVAL; - - while (size) { - unsigned long base, index, end, write_end, int_th; - unsigned long write_size, adj_write_size; - - /* - * Write as much as possible without producing an - * overflow interrupt. - * - * Interrupt_threshold must either be - * - bigger than absolute_maximum or - * - point to a record between buffer_base and absolute_maximum - * - * Index points to a valid record. - */ - base = ds_get(context->ds, qual, ds_buffer_base); - index = ds_get(context->ds, qual, ds_index); - end = ds_get(context->ds, qual, ds_absolute_maximum); - int_th = ds_get(context->ds, qual, ds_interrupt_threshold); - - write_end = min(end, int_th); - - /* - * If we are already beyond the interrupt threshold, - * we fill the entire buffer. - */ - if (write_end <= index) - write_end = end; - - if (write_end <= index) - break; - - write_size = min((unsigned long) size, write_end - index); - memcpy((void *)index, record, write_size); - - record = (const char *)record + write_size; - size -= write_size; - bytes_written += write_size; - - adj_write_size = write_size / ds_cfg.sizeof_rec[qual]; - adj_write_size *= ds_cfg.sizeof_rec[qual]; - - /* Zero out trailing bytes. */ - memset((char *)index + write_size, 0, - adj_write_size - write_size); - index += adj_write_size; - - if (index >= end) - index = base; - ds_set(context->ds, qual, ds_index, index); - - if (index >= int_th) - ds_overflow(context, qual); - } - - return bytes_written; -} - - -/* - * Branch Trace Store (BTS) uses the following format. Different - * architectures vary in the size of those fields. - * - source linear address - * - destination linear address - * - flags - * - * Later architectures use 64bit pointers throughout, whereas earlier - * architectures use 32bit pointers in 32bit mode. - * - * We compute the base address for the fields based on: - * - the field size stored in the DS configuration - * - the relative field position - * - * In order to store additional information in the BTS buffer, we use - * a special source address to indicate that the record requires - * special interpretation. - * - * Netburst indicated via a bit in the flags field whether the branch - * was predicted; this is ignored. - * - * We use two levels of abstraction: - * - the raw data level defined here - * - an arch-independent level defined in ds.h - */ - -enum bts_field { - bts_from, - bts_to, - bts_flags, - - bts_qual = bts_from, - bts_clock = bts_to, - bts_pid = bts_flags, - - bts_qual_mask = (bts_qual_max - 1), - bts_escape = ((unsigned long)-1 & ~bts_qual_mask) -}; - -static inline unsigned long bts_get(const char *base, unsigned long field) -{ - base += (ds_cfg.sizeof_ptr_field * field); - return *(unsigned long *)base; -} - -static inline void bts_set(char *base, unsigned long field, unsigned long val) -{ - base += (ds_cfg.sizeof_ptr_field * field); - (*(unsigned long *)base) = val; -} - - -/* - * The raw BTS data is architecture dependent. - * - * For higher-level users, we give an arch-independent view. - * - ds.h defines struct bts_struct - * - bts_read translates one raw bts record into a bts_struct - * - bts_write translates one bts_struct into the raw format and - * writes it into the top of the parameter tracer's buffer. - * - * return: bytes read/written on success; -Eerrno, otherwise - */ -static int -bts_read(struct bts_tracer *tracer, const void *at, struct bts_struct *out) -{ - if (!tracer) - return -EINVAL; - - if (at < tracer->trace.ds.begin) - return -EINVAL; - - if (tracer->trace.ds.end < (at + tracer->trace.ds.size)) - return -EINVAL; - - memset(out, 0, sizeof(*out)); - if ((bts_get(at, bts_qual) & ~bts_qual_mask) == bts_escape) { - out->qualifier = (bts_get(at, bts_qual) & bts_qual_mask); - out->variant.event.clock = bts_get(at, bts_clock); - out->variant.event.pid = bts_get(at, bts_pid); - } else { - out->qualifier = bts_branch; - out->variant.lbr.from = bts_get(at, bts_from); - out->variant.lbr.to = bts_get(at, bts_to); - - if (!out->variant.lbr.from && !out->variant.lbr.to) - out->qualifier = bts_invalid; - } - - return ds_cfg.sizeof_rec[ds_bts]; -} - -static int bts_write(struct bts_tracer *tracer, const struct bts_struct *in) -{ - unsigned char raw[MAX_SIZEOF_BTS]; - - if (!tracer) - return -EINVAL; - - if (MAX_SIZEOF_BTS < ds_cfg.sizeof_rec[ds_bts]) - return -EOVERFLOW; - - switch (in->qualifier) { - case bts_invalid: - bts_set(raw, bts_from, 0); - bts_set(raw, bts_to, 0); - bts_set(raw, bts_flags, 0); - break; - case bts_branch: - bts_set(raw, bts_from, in->variant.lbr.from); - bts_set(raw, bts_to, in->variant.lbr.to); - bts_set(raw, bts_flags, 0); - break; - case bts_task_arrives: - case bts_task_departs: - bts_set(raw, bts_qual, (bts_escape | in->qualifier)); - bts_set(raw, bts_clock, in->variant.event.clock); - bts_set(raw, bts_pid, in->variant.event.pid); - break; - default: - return -EINVAL; - } - - return ds_write(tracer->ds.context, ds_bts, raw, - ds_cfg.sizeof_rec[ds_bts]); -} - - -static void ds_write_config(struct ds_context *context, - struct ds_trace *cfg, enum ds_qualifier qual) -{ - unsigned char *ds = context->ds; - - ds_set(ds, qual, ds_buffer_base, (unsigned long)cfg->begin); - ds_set(ds, qual, ds_index, (unsigned long)cfg->top); - ds_set(ds, qual, ds_absolute_maximum, (unsigned long)cfg->end); - ds_set(ds, qual, ds_interrupt_threshold, (unsigned long)cfg->ith); -} - -static void ds_read_config(struct ds_context *context, - struct ds_trace *cfg, enum ds_qualifier qual) -{ - unsigned char *ds = context->ds; - - cfg->begin = (void *)ds_get(ds, qual, ds_buffer_base); - cfg->top = (void *)ds_get(ds, qual, ds_index); - cfg->end = (void *)ds_get(ds, qual, ds_absolute_maximum); - cfg->ith = (void *)ds_get(ds, qual, ds_interrupt_threshold); -} - -static void ds_init_ds_trace(struct ds_trace *trace, enum ds_qualifier qual, - void *base, size_t size, size_t ith, - unsigned int flags) { - unsigned long buffer, adj; - - /* - * Adjust the buffer address and size to meet alignment - * constraints: - * - buffer is double-word aligned - * - size is multiple of record size - * - * We checked the size at the very beginning; we have enough - * space to do the adjustment. - */ - buffer = (unsigned long)base; - - adj = ALIGN(buffer, DS_ALIGNMENT) - buffer; - buffer += adj; - size -= adj; - - trace->n = size / ds_cfg.sizeof_rec[qual]; - trace->size = ds_cfg.sizeof_rec[qual]; - - size = (trace->n * trace->size); - - trace->begin = (void *)buffer; - trace->top = trace->begin; - trace->end = (void *)(buffer + size); - /* - * The value for 'no threshold' is -1, which will set the - * threshold outside of the buffer, just like we want it. - */ - ith *= ds_cfg.sizeof_rec[qual]; - trace->ith = (void *)(buffer + size - ith); - - trace->flags = flags; -} - - -static int ds_request(struct ds_tracer *tracer, struct ds_trace *trace, - enum ds_qualifier qual, struct task_struct *task, - int cpu, void *base, size_t size, size_t th) -{ - struct ds_context *context; - int error; - size_t req_size; - - error = -EOPNOTSUPP; - if (!ds_cfg.sizeof_rec[qual]) - goto out; - - error = -EINVAL; - if (!base) - goto out; - - req_size = ds_cfg.sizeof_rec[qual]; - /* We might need space for alignment adjustments. */ - if (!IS_ALIGNED((unsigned long)base, DS_ALIGNMENT)) - req_size += DS_ALIGNMENT; - - error = -EINVAL; - if (size < req_size) - goto out; - - if (th != (size_t)-1) { - th *= ds_cfg.sizeof_rec[qual]; - - error = -EINVAL; - if (size <= th) - goto out; - } - - tracer->buffer = base; - tracer->size = size; - - error = -ENOMEM; - context = ds_get_context(task, cpu); - if (!context) - goto out; - tracer->context = context; - - /* - * Defer any tracer-specific initialization work for the context until - * context ownership has been clarified. - */ - - error = 0; - out: - return error; -} - -static struct bts_tracer *ds_request_bts(struct task_struct *task, int cpu, - void *base, size_t size, - bts_ovfl_callback_t ovfl, size_t th, - unsigned int flags) -{ - struct bts_tracer *tracer; - int error; - - /* Buffer overflow notification is not yet implemented. */ - error = -EOPNOTSUPP; - if (ovfl) - goto out; - - error = get_tracer(task); - if (error < 0) - goto out; - - error = -ENOMEM; - tracer = kzalloc(sizeof(*tracer), GFP_KERNEL); - if (!tracer) - goto out_put_tracer; - tracer->ovfl = ovfl; - - /* Do some more error checking and acquire a tracing context. */ - error = ds_request(&tracer->ds, &tracer->trace.ds, - ds_bts, task, cpu, base, size, th); - if (error < 0) - goto out_tracer; - - /* Claim the bts part of the tracing context we acquired above. */ - spin_lock_irq(&ds_lock); - - error = -EPERM; - if (tracer->ds.context->bts_master) - goto out_unlock; - tracer->ds.context->bts_master = tracer; - - spin_unlock_irq(&ds_lock); - - /* - * Now that we own the bts part of the context, let's complete the - * initialization for that part. - */ - ds_init_ds_trace(&tracer->trace.ds, ds_bts, base, size, th, flags); - ds_write_config(tracer->ds.context, &tracer->trace.ds, ds_bts); - ds_install_ds_area(tracer->ds.context); - - tracer->trace.read = bts_read; - tracer->trace.write = bts_write; - - /* Start tracing. */ - ds_resume_bts(tracer); - - return tracer; - - out_unlock: - spin_unlock_irq(&ds_lock); - ds_put_context(tracer->ds.context); - out_tracer: - kfree(tracer); - out_put_tracer: - put_tracer(task); - out: - return ERR_PTR(error); -} - -struct bts_tracer *ds_request_bts_task(struct task_struct *task, - void *base, size_t size, - bts_ovfl_callback_t ovfl, - size_t th, unsigned int flags) -{ - return ds_request_bts(task, 0, base, size, ovfl, th, flags); -} - -struct bts_tracer *ds_request_bts_cpu(int cpu, void *base, size_t size, - bts_ovfl_callback_t ovfl, - size_t th, unsigned int flags) -{ - return ds_request_bts(NULL, cpu, base, size, ovfl, th, flags); -} - -static struct pebs_tracer *ds_request_pebs(struct task_struct *task, int cpu, - void *base, size_t size, - pebs_ovfl_callback_t ovfl, size_t th, - unsigned int flags) -{ - struct pebs_tracer *tracer; - int error; - - /* Buffer overflow notification is not yet implemented. */ - error = -EOPNOTSUPP; - if (ovfl) - goto out; - - error = get_tracer(task); - if (error < 0) - goto out; - - error = -ENOMEM; - tracer = kzalloc(sizeof(*tracer), GFP_KERNEL); - if (!tracer) - goto out_put_tracer; - tracer->ovfl = ovfl; - - /* Do some more error checking and acquire a tracing context. */ - error = ds_request(&tracer->ds, &tracer->trace.ds, - ds_pebs, task, cpu, base, size, th); - if (error < 0) - goto out_tracer; - - /* Claim the pebs part of the tracing context we acquired above. */ - spin_lock_irq(&ds_lock); - - error = -EPERM; - if (tracer->ds.context->pebs_master) - goto out_unlock; - tracer->ds.context->pebs_master = tracer; - - spin_unlock_irq(&ds_lock); - - /* - * Now that we own the pebs part of the context, let's complete the - * initialization for that part. - */ - ds_init_ds_trace(&tracer->trace.ds, ds_pebs, base, size, th, flags); - ds_write_config(tracer->ds.context, &tracer->trace.ds, ds_pebs); - ds_install_ds_area(tracer->ds.context); - - /* Start tracing. */ - ds_resume_pebs(tracer); - - return tracer; - - out_unlock: - spin_unlock_irq(&ds_lock); - ds_put_context(tracer->ds.context); - out_tracer: - kfree(tracer); - out_put_tracer: - put_tracer(task); - out: - return ERR_PTR(error); -} - -struct pebs_tracer *ds_request_pebs_task(struct task_struct *task, - void *base, size_t size, - pebs_ovfl_callback_t ovfl, - size_t th, unsigned int flags) -{ - return ds_request_pebs(task, 0, base, size, ovfl, th, flags); -} - -struct pebs_tracer *ds_request_pebs_cpu(int cpu, void *base, size_t size, - pebs_ovfl_callback_t ovfl, - size_t th, unsigned int flags) -{ - return ds_request_pebs(NULL, cpu, base, size, ovfl, th, flags); -} - -static void ds_free_bts(struct bts_tracer *tracer) -{ - struct task_struct *task; - - task = tracer->ds.context->task; - - WARN_ON_ONCE(tracer->ds.context->bts_master != tracer); - tracer->ds.context->bts_master = NULL; - - /* Make sure tracing stopped and the tracer is not in use. */ - if (task && (task != current)) - wait_task_context_switch(task); - - ds_put_context(tracer->ds.context); - put_tracer(task); - - kfree(tracer); -} - -void ds_release_bts(struct bts_tracer *tracer) -{ - might_sleep(); - - if (!tracer) - return; - - ds_suspend_bts(tracer); - ds_free_bts(tracer); -} - -int ds_release_bts_noirq(struct bts_tracer *tracer) -{ - struct task_struct *task; - unsigned long irq; - int error; - - if (!tracer) - return 0; - - task = tracer->ds.context->task; - - local_irq_save(irq); - - error = -EPERM; - if (!task && - (tracer->ds.context->cpu != smp_processor_id())) - goto out; - - error = -EPERM; - if (task && (task != current)) - goto out; - - ds_suspend_bts_noirq(tracer); - ds_free_bts(tracer); - - error = 0; - out: - local_irq_restore(irq); - return error; -} - -static void update_task_debugctlmsr(struct task_struct *task, - unsigned long debugctlmsr) -{ - task->thread.debugctlmsr = debugctlmsr; - - get_cpu(); - if (task == current) - update_debugctlmsr(debugctlmsr); - put_cpu(); -} - -void ds_suspend_bts(struct bts_tracer *tracer) -{ - struct task_struct *task; - unsigned long debugctlmsr; - int cpu; - - if (!tracer) - return; - - tracer->flags = 0; - - task = tracer->ds.context->task; - cpu = tracer->ds.context->cpu; - - WARN_ON(!task && irqs_disabled()); - - debugctlmsr = (task ? - task->thread.debugctlmsr : - get_debugctlmsr_on_cpu(cpu)); - debugctlmsr &= ~BTS_CONTROL; - - if (task) - update_task_debugctlmsr(task, debugctlmsr); - else - update_debugctlmsr_on_cpu(cpu, debugctlmsr); -} - -int ds_suspend_bts_noirq(struct bts_tracer *tracer) -{ - struct task_struct *task; - unsigned long debugctlmsr, irq; - int cpu, error = 0; - - if (!tracer) - return 0; - - tracer->flags = 0; - - task = tracer->ds.context->task; - cpu = tracer->ds.context->cpu; - - local_irq_save(irq); - - error = -EPERM; - if (!task && (cpu != smp_processor_id())) - goto out; - - debugctlmsr = (task ? - task->thread.debugctlmsr : - get_debugctlmsr()); - debugctlmsr &= ~BTS_CONTROL; - - if (task) - update_task_debugctlmsr(task, debugctlmsr); - else - update_debugctlmsr(debugctlmsr); - - error = 0; - out: - local_irq_restore(irq); - return error; -} - -static unsigned long ds_bts_control(struct bts_tracer *tracer) -{ - unsigned long control; - - control = ds_cfg.ctl[dsf_bts]; - if (!(tracer->trace.ds.flags & BTS_KERNEL)) - control |= ds_cfg.ctl[dsf_bts_kernel]; - if (!(tracer->trace.ds.flags & BTS_USER)) - control |= ds_cfg.ctl[dsf_bts_user]; - - return control; -} - -void ds_resume_bts(struct bts_tracer *tracer) -{ - struct task_struct *task; - unsigned long debugctlmsr; - int cpu; - - if (!tracer) - return; - - tracer->flags = tracer->trace.ds.flags; - - task = tracer->ds.context->task; - cpu = tracer->ds.context->cpu; - - WARN_ON(!task && irqs_disabled()); - - debugctlmsr = (task ? - task->thread.debugctlmsr : - get_debugctlmsr_on_cpu(cpu)); - debugctlmsr |= ds_bts_control(tracer); - - if (task) - update_task_debugctlmsr(task, debugctlmsr); - else - update_debugctlmsr_on_cpu(cpu, debugctlmsr); -} - -int ds_resume_bts_noirq(struct bts_tracer *tracer) -{ - struct task_struct *task; - unsigned long debugctlmsr, irq; - int cpu, error = 0; - - if (!tracer) - return 0; - - tracer->flags = tracer->trace.ds.flags; - - task = tracer->ds.context->task; - cpu = tracer->ds.context->cpu; - - local_irq_save(irq); - - error = -EPERM; - if (!task && (cpu != smp_processor_id())) - goto out; - - debugctlmsr = (task ? - task->thread.debugctlmsr : - get_debugctlmsr()); - debugctlmsr |= ds_bts_control(tracer); - - if (task) - update_task_debugctlmsr(task, debugctlmsr); - else - update_debugctlmsr(debugctlmsr); - - error = 0; - out: - local_irq_restore(irq); - return error; -} - -static void ds_free_pebs(struct pebs_tracer *tracer) -{ - struct task_struct *task; - - task = tracer->ds.context->task; - - WARN_ON_ONCE(tracer->ds.context->pebs_master != tracer); - tracer->ds.context->pebs_master = NULL; - - ds_put_context(tracer->ds.context); - put_tracer(task); - - kfree(tracer); -} - -void ds_release_pebs(struct pebs_tracer *tracer) -{ - might_sleep(); - - if (!tracer) - return; - - ds_suspend_pebs(tracer); - ds_free_pebs(tracer); -} - -int ds_release_pebs_noirq(struct pebs_tracer *tracer) -{ - struct task_struct *task; - unsigned long irq; - int error; - - if (!tracer) - return 0; - - task = tracer->ds.context->task; - - local_irq_save(irq); - - error = -EPERM; - if (!task && - (tracer->ds.context->cpu != smp_processor_id())) - goto out; - - error = -EPERM; - if (task && (task != current)) - goto out; - - ds_suspend_pebs_noirq(tracer); - ds_free_pebs(tracer); - - error = 0; - out: - local_irq_restore(irq); - return error; -} - -void ds_suspend_pebs(struct pebs_tracer *tracer) -{ - -} - -int ds_suspend_pebs_noirq(struct pebs_tracer *tracer) -{ - return 0; -} - -void ds_resume_pebs(struct pebs_tracer *tracer) -{ - -} - -int ds_resume_pebs_noirq(struct pebs_tracer *tracer) -{ - return 0; -} - -const struct bts_trace *ds_read_bts(struct bts_tracer *tracer) -{ - if (!tracer) - return NULL; - - ds_read_config(tracer->ds.context, &tracer->trace.ds, ds_bts); - return &tracer->trace; -} - -const struct pebs_trace *ds_read_pebs(struct pebs_tracer *tracer) -{ - if (!tracer) - return NULL; - - ds_read_config(tracer->ds.context, &tracer->trace.ds, ds_pebs); - - tracer->trace.counters = ds_cfg.nr_counter_reset; - memcpy(tracer->trace.counter_reset, - tracer->ds.context->ds + - (NUM_DS_PTR_FIELDS * ds_cfg.sizeof_ptr_field), - ds_cfg.nr_counter_reset * PEBS_RESET_FIELD_SIZE); - - return &tracer->trace; -} - -int ds_reset_bts(struct bts_tracer *tracer) -{ - if (!tracer) - return -EINVAL; - - tracer->trace.ds.top = tracer->trace.ds.begin; - - ds_set(tracer->ds.context->ds, ds_bts, ds_index, - (unsigned long)tracer->trace.ds.top); - - return 0; -} - -int ds_reset_pebs(struct pebs_tracer *tracer) -{ - if (!tracer) - return -EINVAL; - - tracer->trace.ds.top = tracer->trace.ds.begin; - - ds_set(tracer->ds.context->ds, ds_pebs, ds_index, - (unsigned long)tracer->trace.ds.top); - - return 0; -} - -int ds_set_pebs_reset(struct pebs_tracer *tracer, - unsigned int counter, u64 value) -{ - if (!tracer) - return -EINVAL; - - if (ds_cfg.nr_counter_reset < counter) - return -EINVAL; - - *(u64 *)(tracer->ds.context->ds + - (NUM_DS_PTR_FIELDS * ds_cfg.sizeof_ptr_field) + - (counter * PEBS_RESET_FIELD_SIZE)) = value; - - return 0; -} - -static const struct ds_configuration ds_cfg_netburst = { - .name = "Netburst", - .ctl[dsf_bts] = (1 << 2) | (1 << 3), - .ctl[dsf_bts_kernel] = (1 << 5), - .ctl[dsf_bts_user] = (1 << 6), - .nr_counter_reset = 1, -}; -static const struct ds_configuration ds_cfg_pentium_m = { - .name = "Pentium M", - .ctl[dsf_bts] = (1 << 6) | (1 << 7), - .nr_counter_reset = 1, -}; -static const struct ds_configuration ds_cfg_core2_atom = { - .name = "Core 2/Atom", - .ctl[dsf_bts] = (1 << 6) | (1 << 7), - .ctl[dsf_bts_kernel] = (1 << 9), - .ctl[dsf_bts_user] = (1 << 10), - .nr_counter_reset = 1, -}; -static const struct ds_configuration ds_cfg_core_i7 = { - .name = "Core i7", - .ctl[dsf_bts] = (1 << 6) | (1 << 7), - .ctl[dsf_bts_kernel] = (1 << 9), - .ctl[dsf_bts_user] = (1 << 10), - .nr_counter_reset = 4, -}; - -static void -ds_configure(const struct ds_configuration *cfg, - struct cpuinfo_x86 *cpu) -{ - unsigned long nr_pebs_fields = 0; - - printk(KERN_INFO "[ds] using %s configuration\n", cfg->name); - -#ifdef __i386__ - nr_pebs_fields = 10; -#else - nr_pebs_fields = 18; -#endif - - /* - * Starting with version 2, architectural performance - * monitoring supports a format specifier. - */ - if ((cpuid_eax(0xa) & 0xff) > 1) { - unsigned long perf_capabilities, format; - - rdmsrl(MSR_IA32_PERF_CAPABILITIES, perf_capabilities); - - format = (perf_capabilities >> 8) & 0xf; - - switch (format) { - case 0: - nr_pebs_fields = 18; - break; - case 1: - nr_pebs_fields = 22; - break; - default: - printk(KERN_INFO - "[ds] unknown PEBS format: %lu\n", format); - nr_pebs_fields = 0; - break; - } - } - - memset(&ds_cfg, 0, sizeof(ds_cfg)); - ds_cfg = *cfg; - - ds_cfg.sizeof_ptr_field = - (cpu_has(cpu, X86_FEATURE_DTES64) ? 8 : 4); - - ds_cfg.sizeof_rec[ds_bts] = ds_cfg.sizeof_ptr_field * 3; - ds_cfg.sizeof_rec[ds_pebs] = ds_cfg.sizeof_ptr_field * nr_pebs_fields; - - if (!cpu_has(cpu, X86_FEATURE_BTS)) { - ds_cfg.sizeof_rec[ds_bts] = 0; - printk(KERN_INFO "[ds] bts not available\n"); - } - if (!cpu_has(cpu, X86_FEATURE_PEBS)) { - ds_cfg.sizeof_rec[ds_pebs] = 0; - printk(KERN_INFO "[ds] pebs not available\n"); - } - - printk(KERN_INFO "[ds] sizes: address: %u bit, ", - 8 * ds_cfg.sizeof_ptr_field); - printk("bts/pebs record: %u/%u bytes\n", - ds_cfg.sizeof_rec[ds_bts], ds_cfg.sizeof_rec[ds_pebs]); - - WARN_ON_ONCE(MAX_PEBS_COUNTERS < ds_cfg.nr_counter_reset); -} - -void __cpuinit ds_init_intel(struct cpuinfo_x86 *c) -{ - /* Only configure the first cpu. Others are identical. */ - if (ds_cfg.name) - return; - - switch (c->x86) { - case 0x6: - switch (c->x86_model) { - case 0x9: - case 0xd: /* Pentium M */ - ds_configure(&ds_cfg_pentium_m, c); - break; - case 0xf: - case 0x17: /* Core2 */ - case 0x1c: /* Atom */ - ds_configure(&ds_cfg_core2_atom, c); - break; - case 0x1a: /* Core i7 */ - ds_configure(&ds_cfg_core_i7, c); - break; - default: - /* Sorry, don't know about them. */ - break; - } - break; - case 0xf: - switch (c->x86_model) { - case 0x0: - case 0x1: - case 0x2: /* Netburst */ - ds_configure(&ds_cfg_netburst, c); - break; - default: - /* Sorry, don't know about them. */ - break; - } - break; - default: - /* Sorry, don't know about them. */ - break; - } -} - -static inline void ds_take_timestamp(struct ds_context *context, - enum bts_qualifier qualifier, - struct task_struct *task) -{ - struct bts_tracer *tracer = context->bts_master; - struct bts_struct ts; - - /* Prevent compilers from reading the tracer pointer twice. */ - barrier(); - - if (!tracer || !(tracer->flags & BTS_TIMESTAMPS)) - return; - - memset(&ts, 0, sizeof(ts)); - ts.qualifier = qualifier; - ts.variant.event.clock = trace_clock_global(); - ts.variant.event.pid = task->pid; - - bts_write(tracer, &ts); -} - -/* - * Change the DS configuration from tracing prev to tracing next. - */ -void ds_switch_to(struct task_struct *prev, struct task_struct *next) -{ - struct ds_context *prev_ctx = prev->thread.ds_ctx; - struct ds_context *next_ctx = next->thread.ds_ctx; - unsigned long debugctlmsr = next->thread.debugctlmsr; - - /* Make sure all data is read before we start. */ - barrier(); - - if (prev_ctx) { - update_debugctlmsr(0); - - ds_take_timestamp(prev_ctx, bts_task_departs, prev); - } - - if (next_ctx) { - ds_take_timestamp(next_ctx, bts_task_arrives, next); - - wrmsrl(MSR_IA32_DS_AREA, (unsigned long)next_ctx->ds); - } - - update_debugctlmsr(debugctlmsr); -} - -static __init int ds_selftest(void) -{ - if (ds_cfg.sizeof_rec[ds_bts]) { - int error; - - error = ds_selftest_bts(); - if (error) { - WARN(1, "[ds] selftest failed. disabling bts.\n"); - ds_cfg.sizeof_rec[ds_bts] = 0; - } - } - - if (ds_cfg.sizeof_rec[ds_pebs]) { - int error; - - error = ds_selftest_pebs(); - if (error) { - WARN(1, "[ds] selftest failed. disabling pebs.\n"); - ds_cfg.sizeof_rec[ds_pebs] = 0; - } - } - - return 0; -} -device_initcall(ds_selftest); diff --git a/arch/x86/kernel/ds_selftest.c b/arch/x86/kernel/ds_selftest.c deleted file mode 100644 index 6bc7c199ab99..000000000000 --- a/arch/x86/kernel/ds_selftest.c +++ /dev/null @@ -1,408 +0,0 @@ -/* - * Debug Store support - selftest - * - * - * Copyright (C) 2009 Intel Corporation. - * Markus Metzger , 2009 - */ - -#include "ds_selftest.h" - -#include -#include -#include -#include - -#include - - -#define BUFFER_SIZE 521 /* Intentionally chose an odd size. */ -#define SMALL_BUFFER_SIZE 24 /* A single bts entry. */ - -struct ds_selftest_bts_conf { - struct bts_tracer *tracer; - int error; - int (*suspend)(struct bts_tracer *); - int (*resume)(struct bts_tracer *); -}; - -static int ds_selftest_bts_consistency(const struct bts_trace *trace) -{ - int error = 0; - - if (!trace) { - printk(KERN_CONT "failed to access trace..."); - /* Bail out. Other tests are pointless. */ - return -1; - } - - if (!trace->read) { - printk(KERN_CONT "bts read not available..."); - error = -1; - } - - /* Do some sanity checks on the trace configuration. */ - if (!trace->ds.n) { - printk(KERN_CONT "empty bts buffer..."); - error = -1; - } - if (!trace->ds.size) { - printk(KERN_CONT "bad bts trace setup..."); - error = -1; - } - if (trace->ds.end != - (char *)trace->ds.begin + (trace->ds.n * trace->ds.size)) { - printk(KERN_CONT "bad bts buffer setup..."); - error = -1; - } - /* - * We allow top in [begin; end], since its not clear when the - * overflow adjustment happens: after the increment or before the - * write. - */ - if ((trace->ds.top < trace->ds.begin) || - (trace->ds.end < trace->ds.top)) { - printk(KERN_CONT "bts top out of bounds..."); - error = -1; - } - - return error; -} - -static int ds_selftest_bts_read(struct bts_tracer *tracer, - const struct bts_trace *trace, - const void *from, const void *to) -{ - const unsigned char *at; - - /* - * Check a few things which do not belong to this test. - * They should be covered by other tests. - */ - if (!trace) - return -1; - - if (!trace->read) - return -1; - - if (to < from) - return -1; - - if (from < trace->ds.begin) - return -1; - - if (trace->ds.end < to) - return -1; - - if (!trace->ds.size) - return -1; - - /* Now to the test itself. */ - for (at = from; (void *)at < to; at += trace->ds.size) { - struct bts_struct bts; - unsigned long index; - int error; - - if (((void *)at - trace->ds.begin) % trace->ds.size) { - printk(KERN_CONT - "read from non-integer index..."); - return -1; - } - index = ((void *)at - trace->ds.begin) / trace->ds.size; - - memset(&bts, 0, sizeof(bts)); - error = trace->read(tracer, at, &bts); - if (error < 0) { - printk(KERN_CONT - "error reading bts trace at [%lu] (0x%p)...", - index, at); - return error; - } - - switch (bts.qualifier) { - case BTS_BRANCH: - break; - default: - printk(KERN_CONT - "unexpected bts entry %llu at [%lu] (0x%p)...", - bts.qualifier, index, at); - return -1; - } - } - - return 0; -} - -static void ds_selftest_bts_cpu(void *arg) -{ - struct ds_selftest_bts_conf *conf = arg; - const struct bts_trace *trace; - void *top; - - if (IS_ERR(conf->tracer)) { - conf->error = PTR_ERR(conf->tracer); - conf->tracer = NULL; - - printk(KERN_CONT - "initialization failed (err: %d)...", conf->error); - return; - } - - /* We should meanwhile have enough trace. */ - conf->error = conf->suspend(conf->tracer); - if (conf->error < 0) - return; - - /* Let's see if we can access the trace. */ - trace = ds_read_bts(conf->tracer); - - conf->error = ds_selftest_bts_consistency(trace); - if (conf->error < 0) - return; - - /* If everything went well, we should have a few trace entries. */ - if (trace->ds.top == trace->ds.begin) { - /* - * It is possible but highly unlikely that we got a - * buffer overflow and end up at exactly the same - * position we started from. - * Let's issue a warning, but continue. - */ - printk(KERN_CONT "no trace/overflow..."); - } - - /* Let's try to read the trace we collected. */ - conf->error = - ds_selftest_bts_read(conf->tracer, trace, - trace->ds.begin, trace->ds.top); - if (conf->error < 0) - return; - - /* - * Let's read the trace again. - * Since we suspended tracing, we should get the same result. - */ - top = trace->ds.top; - - trace = ds_read_bts(conf->tracer); - conf->error = ds_selftest_bts_consistency(trace); - if (conf->error < 0) - return; - - if (top != trace->ds.top) { - printk(KERN_CONT "suspend not working..."); - conf->error = -1; - return; - } - - /* Let's collect some more trace - see if resume is working. */ - conf->error = conf->resume(conf->tracer); - if (conf->error < 0) - return; - - conf->error = conf->suspend(conf->tracer); - if (conf->error < 0) - return; - - trace = ds_read_bts(conf->tracer); - - conf->error = ds_selftest_bts_consistency(trace); - if (conf->error < 0) - return; - - if (trace->ds.top == top) { - /* - * It is possible but highly unlikely that we got a - * buffer overflow and end up at exactly the same - * position we started from. - * Let's issue a warning and check the full trace. - */ - printk(KERN_CONT - "no resume progress/overflow..."); - - conf->error = - ds_selftest_bts_read(conf->tracer, trace, - trace->ds.begin, trace->ds.end); - } else if (trace->ds.top < top) { - /* - * We had a buffer overflow - the entire buffer should - * contain trace records. - */ - conf->error = - ds_selftest_bts_read(conf->tracer, trace, - trace->ds.begin, trace->ds.end); - } else { - /* - * It is quite likely that the buffer did not overflow. - * Let's just check the delta trace. - */ - conf->error = - ds_selftest_bts_read(conf->tracer, trace, top, - trace->ds.top); - } - if (conf->error < 0) - return; - - conf->error = 0; -} - -static int ds_suspend_bts_wrap(struct bts_tracer *tracer) -{ - ds_suspend_bts(tracer); - return 0; -} - -static int ds_resume_bts_wrap(struct bts_tracer *tracer) -{ - ds_resume_bts(tracer); - return 0; -} - -static void ds_release_bts_noirq_wrap(void *tracer) -{ - (void)ds_release_bts_noirq(tracer); -} - -static int ds_selftest_bts_bad_release_noirq(int cpu, - struct bts_tracer *tracer) -{ - int error = -EPERM; - - /* Try to release the tracer on the wrong cpu. */ - get_cpu(); - if (cpu != smp_processor_id()) { - error = ds_release_bts_noirq(tracer); - if (error != -EPERM) - printk(KERN_CONT "release on wrong cpu..."); - } - put_cpu(); - - return error ? 0 : -1; -} - -static int ds_selftest_bts_bad_request_cpu(int cpu, void *buffer) -{ - struct bts_tracer *tracer; - int error; - - /* Try to request cpu tracing while task tracing is active. */ - tracer = ds_request_bts_cpu(cpu, buffer, BUFFER_SIZE, NULL, - (size_t)-1, BTS_KERNEL); - error = PTR_ERR(tracer); - if (!IS_ERR(tracer)) { - ds_release_bts(tracer); - error = 0; - } - - if (error != -EPERM) - printk(KERN_CONT "cpu/task tracing overlap..."); - - return error ? 0 : -1; -} - -static int ds_selftest_bts_bad_request_task(void *buffer) -{ - struct bts_tracer *tracer; - int error; - - /* Try to request cpu tracing while task tracing is active. */ - tracer = ds_request_bts_task(current, buffer, BUFFER_SIZE, NULL, - (size_t)-1, BTS_KERNEL); - error = PTR_ERR(tracer); - if (!IS_ERR(tracer)) { - error = 0; - ds_release_bts(tracer); - } - - if (error != -EPERM) - printk(KERN_CONT "task/cpu tracing overlap..."); - - return error ? 0 : -1; -} - -int ds_selftest_bts(void) -{ - struct ds_selftest_bts_conf conf; - unsigned char buffer[BUFFER_SIZE], *small_buffer; - unsigned long irq; - int cpu; - - printk(KERN_INFO "[ds] bts selftest..."); - conf.error = 0; - - small_buffer = (unsigned char *)ALIGN((unsigned long)buffer, 8) + 8; - - get_online_cpus(); - for_each_online_cpu(cpu) { - conf.suspend = ds_suspend_bts_wrap; - conf.resume = ds_resume_bts_wrap; - conf.tracer = - ds_request_bts_cpu(cpu, buffer, BUFFER_SIZE, - NULL, (size_t)-1, BTS_KERNEL); - ds_selftest_bts_cpu(&conf); - if (conf.error >= 0) - conf.error = ds_selftest_bts_bad_request_task(buffer); - ds_release_bts(conf.tracer); - if (conf.error < 0) - goto out; - - conf.suspend = ds_suspend_bts_noirq; - conf.resume = ds_resume_bts_noirq; - conf.tracer = - ds_request_bts_cpu(cpu, buffer, BUFFER_SIZE, - NULL, (size_t)-1, BTS_KERNEL); - smp_call_function_single(cpu, ds_selftest_bts_cpu, &conf, 1); - if (conf.error >= 0) { - conf.error = - ds_selftest_bts_bad_release_noirq(cpu, - conf.tracer); - /* We must not release the tracer twice. */ - if (conf.error < 0) - conf.tracer = NULL; - } - if (conf.error >= 0) - conf.error = ds_selftest_bts_bad_request_task(buffer); - smp_call_function_single(cpu, ds_release_bts_noirq_wrap, - conf.tracer, 1); - if (conf.error < 0) - goto out; - } - - conf.suspend = ds_suspend_bts_wrap; - conf.resume = ds_resume_bts_wrap; - conf.tracer = - ds_request_bts_task(current, buffer, BUFFER_SIZE, - NULL, (size_t)-1, BTS_KERNEL); - ds_selftest_bts_cpu(&conf); - if (conf.error >= 0) - conf.error = ds_selftest_bts_bad_request_cpu(0, buffer); - ds_release_bts(conf.tracer); - if (conf.error < 0) - goto out; - - conf.suspend = ds_suspend_bts_noirq; - conf.resume = ds_resume_bts_noirq; - conf.tracer = - ds_request_bts_task(current, small_buffer, SMALL_BUFFER_SIZE, - NULL, (size_t)-1, BTS_KERNEL); - local_irq_save(irq); - ds_selftest_bts_cpu(&conf); - if (conf.error >= 0) - conf.error = ds_selftest_bts_bad_request_cpu(0, buffer); - ds_release_bts_noirq(conf.tracer); - local_irq_restore(irq); - if (conf.error < 0) - goto out; - - conf.error = 0; - out: - put_online_cpus(); - printk(KERN_CONT "%s.\n", (conf.error ? "failed" : "passed")); - - return conf.error; -} - -int ds_selftest_pebs(void) -{ - return 0; -} diff --git a/arch/x86/kernel/ds_selftest.h b/arch/x86/kernel/ds_selftest.h deleted file mode 100644 index 2ba8745c6663..000000000000 --- a/arch/x86/kernel/ds_selftest.h +++ /dev/null @@ -1,15 +0,0 @@ -/* - * Debug Store support - selftest - * - * - * Copyright (C) 2009 Intel Corporation. - * Markus Metzger , 2009 - */ - -#ifdef CONFIG_X86_DS_SELFTEST -extern int ds_selftest_bts(void); -extern int ds_selftest_pebs(void); -#else -static inline int ds_selftest_bts(void) { return 0; } -static inline int ds_selftest_pebs(void) { return 0; } -#endif diff --git a/arch/x86/kernel/dumpstack.c b/arch/x86/kernel/dumpstack.c index 6d817554780a..c89a386930b7 100644 --- a/arch/x86/kernel/dumpstack.c +++ b/arch/x86/kernel/dumpstack.c @@ -224,11 +224,6 @@ unsigned __kprobes long oops_begin(void) int cpu; unsigned long flags; - /* notify the hw-branch tracer so it may disable tracing and - add the last trace to the trace buffer - - the earlier this happens, the more useful the trace. */ - trace_hw_branch_oops(); - oops_enter(); /* racy, but better than risking deadlock. */ diff --git a/arch/x86/kernel/kprobes.c b/arch/x86/kernel/kprobes.c index b43bbaebe2c0..7a880ad3a208 100644 --- a/arch/x86/kernel/kprobes.c +++ b/arch/x86/kernel/kprobes.c @@ -422,14 +422,12 @@ static void __kprobes set_current_kprobe(struct kprobe *p, struct pt_regs *regs, static void __kprobes clear_btf(void) { - if (test_thread_flag(TIF_DEBUGCTLMSR)) - update_debugctlmsr(0); + /* XXX */ } static void __kprobes restore_btf(void) { - if (test_thread_flag(TIF_DEBUGCTLMSR)) - update_debugctlmsr(current->thread.debugctlmsr); + /* XXX */ } void __kprobes arch_prepare_kretprobe(struct kretprobe_instance *ri, diff --git a/arch/x86/kernel/process.c b/arch/x86/kernel/process.c index ad9540676fcc..1a60beb32ede 100644 --- a/arch/x86/kernel/process.c +++ b/arch/x86/kernel/process.c @@ -20,7 +20,6 @@ #include #include #include -#include #include unsigned long idle_halt; @@ -50,8 +49,6 @@ void free_thread_xstate(struct task_struct *tsk) kmem_cache_free(task_xstate_cachep, tsk->thread.xstate); tsk->thread.xstate = NULL; } - - WARN(tsk->thread.ds_ctx, "leaking DS context\n"); } void free_thread_info(struct thread_info *ti) @@ -198,12 +195,6 @@ void __switch_to_xtra(struct task_struct *prev_p, struct task_struct *next_p, prev = &prev_p->thread; next = &next_p->thread; - if (test_tsk_thread_flag(next_p, TIF_DS_AREA_MSR) || - test_tsk_thread_flag(prev_p, TIF_DS_AREA_MSR)) - ds_switch_to(prev_p, next_p); - else if (next->debugctlmsr != prev->debugctlmsr) - update_debugctlmsr(next->debugctlmsr); - if (test_tsk_thread_flag(prev_p, TIF_NOTSC) ^ test_tsk_thread_flag(next_p, TIF_NOTSC)) { /* prev and next are different */ diff --git a/arch/x86/kernel/process_32.c b/arch/x86/kernel/process_32.c index f6c62667e30c..75090c589b7a 100644 --- a/arch/x86/kernel/process_32.c +++ b/arch/x86/kernel/process_32.c @@ -55,7 +55,6 @@ #include #include #include -#include #include asmlinkage void ret_from_fork(void) __asm__("ret_from_fork"); @@ -238,13 +237,6 @@ int copy_thread(unsigned long clone_flags, unsigned long sp, kfree(p->thread.io_bitmap_ptr); p->thread.io_bitmap_max = 0; } - - clear_tsk_thread_flag(p, TIF_DS_AREA_MSR); - p->thread.ds_ctx = NULL; - - clear_tsk_thread_flag(p, TIF_DEBUGCTLMSR); - p->thread.debugctlmsr = 0; - return err; } diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c index dc9690b4c4cc..cc4258f2beb5 100644 --- a/arch/x86/kernel/process_64.c +++ b/arch/x86/kernel/process_64.c @@ -49,7 +49,6 @@ #include #include #include -#include #include asmlinkage extern void ret_from_fork(void); @@ -313,13 +312,6 @@ int copy_thread(unsigned long clone_flags, unsigned long sp, if (err) goto out; } - - clear_tsk_thread_flag(p, TIF_DS_AREA_MSR); - p->thread.ds_ctx = NULL; - - clear_tsk_thread_flag(p, TIF_DEBUGCTLMSR); - p->thread.debugctlmsr = 0; - err = 0; out: if (err && p->thread.io_bitmap_ptr) { diff --git a/arch/x86/kernel/ptrace.c b/arch/x86/kernel/ptrace.c index a503b1fd04e5..f2fd3b80e565 100644 --- a/arch/x86/kernel/ptrace.c +++ b/arch/x86/kernel/ptrace.c @@ -2,9 +2,6 @@ /* * Pentium III FXSR, SSE support * Gareth Hughes , May 2000 - * - * BTS tracing - * Markus Metzger , Dec 2007 */ #include @@ -21,7 +18,6 @@ #include #include #include -#include #include #include @@ -35,7 +31,6 @@ #include #include #include -#include #include #include "tls.h" @@ -788,342 +783,6 @@ static int ioperm_get(struct task_struct *target, 0, IO_BITMAP_BYTES); } -#ifdef CONFIG_X86_PTRACE_BTS -/* - * A branch trace store context. - * - * Contexts may only be installed by ptrace_bts_config() and only for - * ptraced tasks. - * - * Contexts are destroyed when the tracee is detached from the tracer. - * The actual destruction work requires interrupts enabled, so the - * work is deferred and will be scheduled during __ptrace_unlink(). - * - * Contexts hold an additional task_struct reference on the traced - * task, as well as a reference on the tracer's mm. - * - * Ptrace already holds a task_struct for the duration of ptrace operations, - * but since destruction is deferred, it may be executed after both - * tracer and tracee exited. - */ -struct bts_context { - /* The branch trace handle. */ - struct bts_tracer *tracer; - - /* The buffer used to store the branch trace and its size. */ - void *buffer; - unsigned int size; - - /* The mm that paid for the above buffer. */ - struct mm_struct *mm; - - /* The task this context belongs to. */ - struct task_struct *task; - - /* The signal to send on a bts buffer overflow. */ - unsigned int bts_ovfl_signal; - - /* The work struct to destroy a context. */ - struct work_struct work; -}; - -static int alloc_bts_buffer(struct bts_context *context, unsigned int size) -{ - void *buffer = NULL; - int err = -ENOMEM; - - err = account_locked_memory(current->mm, current->signal->rlim, size); - if (err < 0) - return err; - - buffer = kzalloc(size, GFP_KERNEL); - if (!buffer) - goto out_refund; - - context->buffer = buffer; - context->size = size; - context->mm = get_task_mm(current); - - return 0; - - out_refund: - refund_locked_memory(current->mm, size); - return err; -} - -static inline void free_bts_buffer(struct bts_context *context) -{ - if (!context->buffer) - return; - - kfree(context->buffer); - context->buffer = NULL; - - refund_locked_memory(context->mm, context->size); - context->size = 0; - - mmput(context->mm); - context->mm = NULL; -} - -static void free_bts_context_work(struct work_struct *w) -{ - struct bts_context *context; - - context = container_of(w, struct bts_context, work); - - ds_release_bts(context->tracer); - put_task_struct(context->task); - free_bts_buffer(context); - kfree(context); -} - -static inline void free_bts_context(struct bts_context *context) -{ - INIT_WORK(&context->work, free_bts_context_work); - schedule_work(&context->work); -} - -static inline struct bts_context *alloc_bts_context(struct task_struct *task) -{ - struct bts_context *context = kzalloc(sizeof(*context), GFP_KERNEL); - if (context) { - context->task = task; - task->bts = context; - - get_task_struct(task); - } - - return context; -} - -static int ptrace_bts_read_record(struct task_struct *child, size_t index, - struct bts_struct __user *out) -{ - struct bts_context *context; - const struct bts_trace *trace; - struct bts_struct bts; - const unsigned char *at; - int error; - - context = child->bts; - if (!context) - return -ESRCH; - - trace = ds_read_bts(context->tracer); - if (!trace) - return -ESRCH; - - at = trace->ds.top - ((index + 1) * trace->ds.size); - if ((void *)at < trace->ds.begin) - at += (trace->ds.n * trace->ds.size); - - if (!trace->read) - return -EOPNOTSUPP; - - error = trace->read(context->tracer, at, &bts); - if (error < 0) - return error; - - if (copy_to_user(out, &bts, sizeof(bts))) - return -EFAULT; - - return sizeof(bts); -} - -static int ptrace_bts_drain(struct task_struct *child, - long size, - struct bts_struct __user *out) -{ - struct bts_context *context; - const struct bts_trace *trace; - const unsigned char *at; - int error, drained = 0; - - context = child->bts; - if (!context) - return -ESRCH; - - trace = ds_read_bts(context->tracer); - if (!trace) - return -ESRCH; - - if (!trace->read) - return -EOPNOTSUPP; - - if (size < (trace->ds.top - trace->ds.begin)) - return -EIO; - - for (at = trace->ds.begin; (void *)at < trace->ds.top; - out++, drained++, at += trace->ds.size) { - struct bts_struct bts; - - error = trace->read(context->tracer, at, &bts); - if (error < 0) - return error; - - if (copy_to_user(out, &bts, sizeof(bts))) - return -EFAULT; - } - - memset(trace->ds.begin, 0, trace->ds.n * trace->ds.size); - - error = ds_reset_bts(context->tracer); - if (error < 0) - return error; - - return drained; -} - -static int ptrace_bts_config(struct task_struct *child, - long cfg_size, - const struct ptrace_bts_config __user *ucfg) -{ - struct bts_context *context; - struct ptrace_bts_config cfg; - unsigned int flags = 0; - - if (cfg_size < sizeof(cfg)) - return -EIO; - - if (copy_from_user(&cfg, ucfg, sizeof(cfg))) - return -EFAULT; - - context = child->bts; - if (!context) - context = alloc_bts_context(child); - if (!context) - return -ENOMEM; - - if (cfg.flags & PTRACE_BTS_O_SIGNAL) { - if (!cfg.signal) - return -EINVAL; - - return -EOPNOTSUPP; - context->bts_ovfl_signal = cfg.signal; - } - - ds_release_bts(context->tracer); - context->tracer = NULL; - - if ((cfg.flags & PTRACE_BTS_O_ALLOC) && (cfg.size != context->size)) { - int err; - - free_bts_buffer(context); - if (!cfg.size) - return 0; - - err = alloc_bts_buffer(context, cfg.size); - if (err < 0) - return err; - } - - if (cfg.flags & PTRACE_BTS_O_TRACE) - flags |= BTS_USER; - - if (cfg.flags & PTRACE_BTS_O_SCHED) - flags |= BTS_TIMESTAMPS; - - context->tracer = - ds_request_bts_task(child, context->buffer, context->size, - NULL, (size_t)-1, flags); - if (unlikely(IS_ERR(context->tracer))) { - int error = PTR_ERR(context->tracer); - - free_bts_buffer(context); - context->tracer = NULL; - return error; - } - - return sizeof(cfg); -} - -static int ptrace_bts_status(struct task_struct *child, - long cfg_size, - struct ptrace_bts_config __user *ucfg) -{ - struct bts_context *context; - const struct bts_trace *trace; - struct ptrace_bts_config cfg; - - context = child->bts; - if (!context) - return -ESRCH; - - if (cfg_size < sizeof(cfg)) - return -EIO; - - trace = ds_read_bts(context->tracer); - if (!trace) - return -ESRCH; - - memset(&cfg, 0, sizeof(cfg)); - cfg.size = trace->ds.end - trace->ds.begin; - cfg.signal = context->bts_ovfl_signal; - cfg.bts_size = sizeof(struct bts_struct); - - if (cfg.signal) - cfg.flags |= PTRACE_BTS_O_SIGNAL; - - if (trace->ds.flags & BTS_USER) - cfg.flags |= PTRACE_BTS_O_TRACE; - - if (trace->ds.flags & BTS_TIMESTAMPS) - cfg.flags |= PTRACE_BTS_O_SCHED; - - if (copy_to_user(ucfg, &cfg, sizeof(cfg))) - return -EFAULT; - - return sizeof(cfg); -} - -static int ptrace_bts_clear(struct task_struct *child) -{ - struct bts_context *context; - const struct bts_trace *trace; - - context = child->bts; - if (!context) - return -ESRCH; - - trace = ds_read_bts(context->tracer); - if (!trace) - return -ESRCH; - - memset(trace->ds.begin, 0, trace->ds.n * trace->ds.size); - - return ds_reset_bts(context->tracer); -} - -static int ptrace_bts_size(struct task_struct *child) -{ - struct bts_context *context; - const struct bts_trace *trace; - - context = child->bts; - if (!context) - return -ESRCH; - - trace = ds_read_bts(context->tracer); - if (!trace) - return -ESRCH; - - return (trace->ds.top - trace->ds.begin) / trace->ds.size; -} - -/* - * Called from __ptrace_unlink() after the child has been moved back - * to its original parent. - */ -void ptrace_bts_untrace(struct task_struct *child) -{ - if (unlikely(child->bts)) { - free_bts_context(child->bts); - child->bts = NULL; - } -} -#endif /* CONFIG_X86_PTRACE_BTS */ - /* * Called by kernel/ptrace.c when detaching.. * @@ -1251,39 +910,6 @@ long arch_ptrace(struct task_struct *child, long request, long addr, long data) break; #endif - /* - * These bits need more cooking - not enabled yet: - */ -#ifdef CONFIG_X86_PTRACE_BTS - case PTRACE_BTS_CONFIG: - ret = ptrace_bts_config - (child, data, (struct ptrace_bts_config __user *)addr); - break; - - case PTRACE_BTS_STATUS: - ret = ptrace_bts_status - (child, data, (struct ptrace_bts_config __user *)addr); - break; - - case PTRACE_BTS_SIZE: - ret = ptrace_bts_size(child); - break; - - case PTRACE_BTS_GET: - ret = ptrace_bts_read_record - (child, data, (struct bts_struct __user *) addr); - break; - - case PTRACE_BTS_CLEAR: - ret = ptrace_bts_clear(child); - break; - - case PTRACE_BTS_DRAIN: - ret = ptrace_bts_drain - (child, data, (struct bts_struct __user *) addr); - break; -#endif /* CONFIG_X86_PTRACE_BTS */ - default: ret = ptrace_request(child, request, addr, data); break; @@ -1543,14 +1169,6 @@ long compat_arch_ptrace(struct task_struct *child, compat_long_t request, case PTRACE_GET_THREAD_AREA: case PTRACE_SET_THREAD_AREA: -#ifdef CONFIG_X86_PTRACE_BTS - case PTRACE_BTS_CONFIG: - case PTRACE_BTS_STATUS: - case PTRACE_BTS_SIZE: - case PTRACE_BTS_GET: - case PTRACE_BTS_CLEAR: - case PTRACE_BTS_DRAIN: -#endif /* CONFIG_X86_PTRACE_BTS */ return arch_ptrace(child, request, addr, data); default: diff --git a/arch/x86/kernel/step.c b/arch/x86/kernel/step.c index 3149032ff107..7beba0769a8c 100644 --- a/arch/x86/kernel/step.c +++ b/arch/x86/kernel/step.c @@ -157,22 +157,6 @@ static int enable_single_step(struct task_struct *child) return 1; } -/* - * Install this value in MSR_IA32_DEBUGCTLMSR whenever child is running. - */ -static void write_debugctlmsr(struct task_struct *child, unsigned long val) -{ - if (child->thread.debugctlmsr == val) - return; - - child->thread.debugctlmsr = val; - - if (child != current) - return; - - update_debugctlmsr(val); -} - /* * Enable single or block step. */ @@ -185,17 +169,9 @@ static void enable_step(struct task_struct *child, bool block) * So noone should try to use debugger block stepping in a program * that uses user-mode single stepping itself. */ - if (enable_single_step(child) && block) { - set_tsk_thread_flag(child, TIF_DEBUGCTLMSR); - write_debugctlmsr(child, - child->thread.debugctlmsr | DEBUGCTLMSR_BTF); - } else { - write_debugctlmsr(child, - child->thread.debugctlmsr & ~DEBUGCTLMSR_BTF); - - if (!child->thread.debugctlmsr) - clear_tsk_thread_flag(child, TIF_DEBUGCTLMSR); - } + if (!enable_single_step(child)) + return; + /* XXX */ } void user_enable_single_step(struct task_struct *child) @@ -213,11 +189,7 @@ void user_disable_single_step(struct task_struct *child) /* * Make sure block stepping (BTF) is disabled. */ - write_debugctlmsr(child, - child->thread.debugctlmsr & ~DEBUGCTLMSR_BTF); - - if (!child->thread.debugctlmsr) - clear_tsk_thread_flag(child, TIF_DEBUGCTLMSR); + /* XXX */ /* Always clear TIF_SINGLESTEP... */ clear_tsk_thread_flag(child, TIF_SINGLESTEP); diff --git a/arch/x86/kernel/traps.c b/arch/x86/kernel/traps.c index 1168e4454188..e3da5d726a37 100644 --- a/arch/x86/kernel/traps.c +++ b/arch/x86/kernel/traps.c @@ -543,11 +543,6 @@ dotraplinkage void __kprobes do_debug(struct pt_regs *regs, long error_code) /* DR6 may or may not be cleared by the CPU */ set_debugreg(0, 6); - /* - * The processor cleared BTF, so don't mark that we need it set. - */ - clear_tsk_thread_flag(tsk, TIF_DEBUGCTLMSR); - tsk->thread.debugctlmsr = 0; /* Store the virtualized DR6 value */ tsk->thread.debugreg6 = dr6; diff --git a/include/linux/ftrace.h b/include/linux/ftrace.h index 01e6adea07ec..cc12b3c556b3 100644 --- a/include/linux/ftrace.h +++ b/include/linux/ftrace.h @@ -504,18 +504,6 @@ extern int ftrace_dump_on_oops; #define INIT_TRACE_RECURSION #endif -#ifdef CONFIG_HW_BRANCH_TRACER - -void trace_hw_branch(u64 from, u64 to); -void trace_hw_branch_oops(void); - -#else /* CONFIG_HW_BRANCH_TRACER */ - -static inline void trace_hw_branch(u64 from, u64 to) {} -static inline void trace_hw_branch_oops(void) {} - -#endif /* CONFIG_HW_BRANCH_TRACER */ - #ifdef CONFIG_FTRACE_SYSCALLS unsigned long arch_syscall_addr(int nr); diff --git a/include/linux/mm.h b/include/linux/mm.h index e70f21beb4b4..c8442b655111 100644 --- a/include/linux/mm.h +++ b/include/linux/mm.h @@ -19,7 +19,6 @@ struct anon_vma; struct file_ra_state; struct user_struct; struct writeback_control; -struct rlimit; #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */ extern unsigned long max_mapnr; @@ -1449,9 +1448,6 @@ int vmemmap_populate_basepages(struct page *start_page, int vmemmap_populate(struct page *start_page, unsigned long pages, int node); void vmemmap_populate_print_last(void); -extern int account_locked_memory(struct mm_struct *mm, struct rlimit *rlim, - size_t size); -extern void refund_locked_memory(struct mm_struct *mm, size_t size); enum mf_flags { MF_COUNT_INCREASED = 1 << 0, diff --git a/include/linux/ptrace.h b/include/linux/ptrace.h index e1fb60729979..4272521e29e9 100644 --- a/include/linux/ptrace.h +++ b/include/linux/ptrace.h @@ -345,18 +345,6 @@ static inline void user_single_step_siginfo(struct task_struct *tsk, #define arch_ptrace_stop(code, info) do { } while (0) #endif -#ifndef arch_ptrace_untrace -/* - * Do machine-specific work before untracing child. - * - * This is called for a normal detach as well as from ptrace_exit() - * when the tracing task dies. - * - * Called with write_lock(&tasklist_lock) held. - */ -#define arch_ptrace_untrace(task) do { } while (0) -#endif - extern int task_current_syscall(struct task_struct *target, long *callno, unsigned long args[6], unsigned int maxargs, unsigned long *sp, unsigned long *pc); diff --git a/include/linux/sched.h b/include/linux/sched.h index dad7f668ebf7..e0447c64af6a 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -99,7 +99,6 @@ struct futex_pi_state; struct robust_list_head; struct bio_list; struct fs_struct; -struct bts_context; struct perf_event_context; /* @@ -1272,12 +1271,6 @@ struct task_struct { struct list_head ptraced; struct list_head ptrace_entry; - /* - * This is the tracer handle for the ptrace BTS extension. - * This field actually belongs to the ptracer task. - */ - struct bts_context *bts; - /* PID/PID hash table linkage. */ struct pid_link pids[PIDTYPE_MAX]; struct list_head thread_group; @@ -2123,10 +2116,8 @@ extern void set_task_comm(struct task_struct *tsk, char *from); extern char *get_task_comm(char *to, struct task_struct *tsk); #ifdef CONFIG_SMP -extern void wait_task_context_switch(struct task_struct *p); extern unsigned long wait_task_inactive(struct task_struct *, long match_state); #else -static inline void wait_task_context_switch(struct task_struct *p) {} static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state) { diff --git a/kernel/fork.c b/kernel/fork.c index 4799c5f0e6d0..d67f1dbfbe03 100644 --- a/kernel/fork.c +++ b/kernel/fork.c @@ -1108,9 +1108,6 @@ static struct task_struct *copy_process(unsigned long clone_flags, p->memcg_batch.do_batch = 0; p->memcg_batch.memcg = NULL; #endif - - p->bts = NULL; - p->stack_start = stack_start; /* Perform scheduler related setup. Assign this task to a CPU. */ diff --git a/kernel/ptrace.c b/kernel/ptrace.c index 42ad8ae729a0..9fb51237b18c 100644 --- a/kernel/ptrace.c +++ b/kernel/ptrace.c @@ -76,7 +76,6 @@ void __ptrace_unlink(struct task_struct *child) child->parent = child->real_parent; list_del_init(&child->ptrace_entry); - arch_ptrace_untrace(child); if (task_is_traced(child)) ptrace_untrace(child); } diff --git a/kernel/sched.c b/kernel/sched.c index 9ab3cd7858d3..117b7cad31b3 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -2076,49 +2076,6 @@ migrate_task(struct task_struct *p, int dest_cpu, struct migration_req *req) return 1; } -/* - * wait_task_context_switch - wait for a thread to complete at least one - * context switch. - * - * @p must not be current. - */ -void wait_task_context_switch(struct task_struct *p) -{ - unsigned long nvcsw, nivcsw, flags; - int running; - struct rq *rq; - - nvcsw = p->nvcsw; - nivcsw = p->nivcsw; - for (;;) { - /* - * The runqueue is assigned before the actual context - * switch. We need to take the runqueue lock. - * - * We could check initially without the lock but it is - * very likely that we need to take the lock in every - * iteration. - */ - rq = task_rq_lock(p, &flags); - running = task_running(rq, p); - task_rq_unlock(rq, &flags); - - if (likely(!running)) - break; - /* - * The switch count is incremented before the actual - * context switch. We thus wait for two switches to be - * sure at least one completed. - */ - if ((p->nvcsw - nvcsw) > 1) - break; - if ((p->nivcsw - nivcsw) > 1) - break; - - cpu_relax(); - } -} - /* * wait_task_inactive - wait for a thread to unschedule. * diff --git a/kernel/trace/Kconfig b/kernel/trace/Kconfig index 13e13d428cd3..8b1797c4545b 100644 --- a/kernel/trace/Kconfig +++ b/kernel/trace/Kconfig @@ -44,9 +44,6 @@ config HAVE_FTRACE_MCOUNT_RECORD help See Documentation/trace/ftrace-design.txt -config HAVE_HW_BRANCH_TRACER - bool - config HAVE_SYSCALL_TRACEPOINTS bool help @@ -374,14 +371,6 @@ config STACK_TRACER Say N if unsure. -config HW_BRANCH_TRACER - depends on HAVE_HW_BRANCH_TRACER - bool "Trace hw branches" - select GENERIC_TRACER - help - This tracer records all branches on the system in a circular - buffer, giving access to the last N branches for each cpu. - config KMEMTRACE bool "Trace SLAB allocations" select GENERIC_TRACER diff --git a/kernel/trace/Makefile b/kernel/trace/Makefile index 78edc6490038..ffb1a5b0550e 100644 --- a/kernel/trace/Makefile +++ b/kernel/trace/Makefile @@ -41,7 +41,6 @@ obj-$(CONFIG_MMIOTRACE) += trace_mmiotrace.o obj-$(CONFIG_BOOT_TRACER) += trace_boot.o obj-$(CONFIG_FUNCTION_GRAPH_TRACER) += trace_functions_graph.o obj-$(CONFIG_TRACE_BRANCH_PROFILING) += trace_branch.o -obj-$(CONFIG_HW_BRANCH_TRACER) += trace_hw_branches.o obj-$(CONFIG_KMEMTRACE) += kmemtrace.o obj-$(CONFIG_WORKQUEUE_TRACER) += trace_workqueue.o obj-$(CONFIG_BLK_DEV_IO_TRACE) += blktrace.o diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h index 2825ef2c0b15..bec2c973ff0c 100644 --- a/kernel/trace/trace.h +++ b/kernel/trace/trace.h @@ -34,7 +34,6 @@ enum trace_type { TRACE_GRAPH_RET, TRACE_GRAPH_ENT, TRACE_USER_STACK, - TRACE_HW_BRANCHES, TRACE_KMEM_ALLOC, TRACE_KMEM_FREE, TRACE_BLK, @@ -229,7 +228,6 @@ extern void __ftrace_bad_type(void); TRACE_GRAPH_ENT); \ IF_ASSIGN(var, ent, struct ftrace_graph_ret_entry, \ TRACE_GRAPH_RET); \ - IF_ASSIGN(var, ent, struct hw_branch_entry, TRACE_HW_BRANCHES);\ IF_ASSIGN(var, ent, struct kmemtrace_alloc_entry, \ TRACE_KMEM_ALLOC); \ IF_ASSIGN(var, ent, struct kmemtrace_free_entry, \ @@ -467,8 +465,6 @@ extern int trace_selftest_startup_sysprof(struct tracer *trace, struct trace_array *tr); extern int trace_selftest_startup_branch(struct tracer *trace, struct trace_array *tr); -extern int trace_selftest_startup_hw_branches(struct tracer *trace, - struct trace_array *tr); extern int trace_selftest_startup_ksym(struct tracer *trace, struct trace_array *tr); #endif /* CONFIG_FTRACE_STARTUP_TEST */ diff --git a/kernel/trace/trace_entries.h b/kernel/trace/trace_entries.h index c16a08f399df..dc008c1240da 100644 --- a/kernel/trace/trace_entries.h +++ b/kernel/trace/trace_entries.h @@ -318,18 +318,6 @@ FTRACE_ENTRY(branch, trace_branch, __entry->func, __entry->file, __entry->correct) ); -FTRACE_ENTRY(hw_branch, hw_branch_entry, - - TRACE_HW_BRANCHES, - - F_STRUCT( - __field( u64, from ) - __field( u64, to ) - ), - - F_printk("from: %llx to: %llx", __entry->from, __entry->to) -); - FTRACE_ENTRY(kmem_alloc, kmemtrace_alloc_entry, TRACE_KMEM_ALLOC, diff --git a/kernel/trace/trace_hw_branches.c b/kernel/trace/trace_hw_branches.c deleted file mode 100644 index 7b97000745f5..000000000000 --- a/kernel/trace/trace_hw_branches.c +++ /dev/null @@ -1,312 +0,0 @@ -/* - * h/w branch tracer for x86 based on BTS - * - * Copyright (C) 2008-2009 Intel Corporation. - * Markus Metzger , 2008-2009 - */ -#include -#include -#include -#include -#include -#include -#include - -#include - -#include "trace_output.h" -#include "trace.h" - - -#define BTS_BUFFER_SIZE (1 << 13) - -static DEFINE_PER_CPU(struct bts_tracer *, hwb_tracer); -static DEFINE_PER_CPU(unsigned char[BTS_BUFFER_SIZE], hwb_buffer); - -#define this_tracer per_cpu(hwb_tracer, smp_processor_id()) - -static int trace_hw_branches_enabled __read_mostly; -static int trace_hw_branches_suspended __read_mostly; -static struct trace_array *hw_branch_trace __read_mostly; - - -static void bts_trace_init_cpu(int cpu) -{ - per_cpu(hwb_tracer, cpu) = - ds_request_bts_cpu(cpu, per_cpu(hwb_buffer, cpu), - BTS_BUFFER_SIZE, NULL, (size_t)-1, - BTS_KERNEL); - - if (IS_ERR(per_cpu(hwb_tracer, cpu))) - per_cpu(hwb_tracer, cpu) = NULL; -} - -static int bts_trace_init(struct trace_array *tr) -{ - int cpu; - - hw_branch_trace = tr; - trace_hw_branches_enabled = 0; - - get_online_cpus(); - for_each_online_cpu(cpu) { - bts_trace_init_cpu(cpu); - - if (likely(per_cpu(hwb_tracer, cpu))) - trace_hw_branches_enabled = 1; - } - trace_hw_branches_suspended = 0; - put_online_cpus(); - - /* If we could not enable tracing on a single cpu, we fail. */ - return trace_hw_branches_enabled ? 0 : -EOPNOTSUPP; -} - -static void bts_trace_reset(struct trace_array *tr) -{ - int cpu; - - get_online_cpus(); - for_each_online_cpu(cpu) { - if (likely(per_cpu(hwb_tracer, cpu))) { - ds_release_bts(per_cpu(hwb_tracer, cpu)); - per_cpu(hwb_tracer, cpu) = NULL; - } - } - trace_hw_branches_enabled = 0; - trace_hw_branches_suspended = 0; - put_online_cpus(); -} - -static void bts_trace_start(struct trace_array *tr) -{ - int cpu; - - get_online_cpus(); - for_each_online_cpu(cpu) - if (likely(per_cpu(hwb_tracer, cpu))) - ds_resume_bts(per_cpu(hwb_tracer, cpu)); - trace_hw_branches_suspended = 0; - put_online_cpus(); -} - -static void bts_trace_stop(struct trace_array *tr) -{ - int cpu; - - get_online_cpus(); - for_each_online_cpu(cpu) - if (likely(per_cpu(hwb_tracer, cpu))) - ds_suspend_bts(per_cpu(hwb_tracer, cpu)); - trace_hw_branches_suspended = 1; - put_online_cpus(); -} - -static int __cpuinit bts_hotcpu_handler(struct notifier_block *nfb, - unsigned long action, void *hcpu) -{ - int cpu = (long)hcpu; - - switch (action) { - case CPU_ONLINE: - case CPU_DOWN_FAILED: - /* The notification is sent with interrupts enabled. */ - if (trace_hw_branches_enabled) { - bts_trace_init_cpu(cpu); - - if (trace_hw_branches_suspended && - likely(per_cpu(hwb_tracer, cpu))) - ds_suspend_bts(per_cpu(hwb_tracer, cpu)); - } - break; - - case CPU_DOWN_PREPARE: - /* The notification is sent with interrupts enabled. */ - if (likely(per_cpu(hwb_tracer, cpu))) { - ds_release_bts(per_cpu(hwb_tracer, cpu)); - per_cpu(hwb_tracer, cpu) = NULL; - } - } - - return NOTIFY_DONE; -} - -static struct notifier_block bts_hotcpu_notifier __cpuinitdata = { - .notifier_call = bts_hotcpu_handler -}; - -static void bts_trace_print_header(struct seq_file *m) -{ - seq_puts(m, "# CPU# TO <- FROM\n"); -} - -static enum print_line_t bts_trace_print_line(struct trace_iterator *iter) -{ - unsigned long symflags = TRACE_ITER_SYM_OFFSET; - struct trace_entry *entry = iter->ent; - struct trace_seq *seq = &iter->seq; - struct hw_branch_entry *it; - - trace_assign_type(it, entry); - - if (entry->type == TRACE_HW_BRANCHES) { - if (trace_seq_printf(seq, "%4d ", iter->cpu) && - seq_print_ip_sym(seq, it->to, symflags) && - trace_seq_printf(seq, "\t <- ") && - seq_print_ip_sym(seq, it->from, symflags) && - trace_seq_printf(seq, "\n")) - return TRACE_TYPE_HANDLED; - return TRACE_TYPE_PARTIAL_LINE; - } - return TRACE_TYPE_UNHANDLED; -} - -void trace_hw_branch(u64 from, u64 to) -{ - struct ftrace_event_call *call = &event_hw_branch; - struct trace_array *tr = hw_branch_trace; - struct ring_buffer_event *event; - struct ring_buffer *buf; - struct hw_branch_entry *entry; - unsigned long irq1; - int cpu; - - if (unlikely(!tr)) - return; - - if (unlikely(!trace_hw_branches_enabled)) - return; - - local_irq_save(irq1); - cpu = raw_smp_processor_id(); - if (atomic_inc_return(&tr->data[cpu]->disabled) != 1) - goto out; - - buf = tr->buffer; - event = trace_buffer_lock_reserve(buf, TRACE_HW_BRANCHES, - sizeof(*entry), 0, 0); - if (!event) - goto out; - entry = ring_buffer_event_data(event); - tracing_generic_entry_update(&entry->ent, 0, from); - entry->ent.type = TRACE_HW_BRANCHES; - entry->from = from; - entry->to = to; - if (!filter_check_discard(call, entry, buf, event)) - trace_buffer_unlock_commit(buf, event, 0, 0); - - out: - atomic_dec(&tr->data[cpu]->disabled); - local_irq_restore(irq1); -} - -static void trace_bts_at(const struct bts_trace *trace, void *at) -{ - struct bts_struct bts; - int err = 0; - - WARN_ON_ONCE(!trace->read); - if (!trace->read) - return; - - err = trace->read(this_tracer, at, &bts); - if (err < 0) - return; - - switch (bts.qualifier) { - case BTS_BRANCH: - trace_hw_branch(bts.variant.lbr.from, bts.variant.lbr.to); - break; - } -} - -/* - * Collect the trace on the current cpu and write it into the ftrace buffer. - * - * pre: tracing must be suspended on the current cpu - */ -static void trace_bts_cpu(void *arg) -{ - struct trace_array *tr = (struct trace_array *)arg; - const struct bts_trace *trace; - unsigned char *at; - - if (unlikely(!tr)) - return; - - if (unlikely(atomic_read(&tr->data[raw_smp_processor_id()]->disabled))) - return; - - if (unlikely(!this_tracer)) - return; - - trace = ds_read_bts(this_tracer); - if (!trace) - return; - - for (at = trace->ds.top; (void *)at < trace->ds.end; - at += trace->ds.size) - trace_bts_at(trace, at); - - for (at = trace->ds.begin; (void *)at < trace->ds.top; - at += trace->ds.size) - trace_bts_at(trace, at); -} - -static void trace_bts_prepare(struct trace_iterator *iter) -{ - int cpu; - - get_online_cpus(); - for_each_online_cpu(cpu) - if (likely(per_cpu(hwb_tracer, cpu))) - ds_suspend_bts(per_cpu(hwb_tracer, cpu)); - /* - * We need to collect the trace on the respective cpu since ftrace - * implicitly adds the record for the current cpu. - * Once that is more flexible, we could collect the data from any cpu. - */ - on_each_cpu(trace_bts_cpu, iter->tr, 1); - - for_each_online_cpu(cpu) - if (likely(per_cpu(hwb_tracer, cpu))) - ds_resume_bts(per_cpu(hwb_tracer, cpu)); - put_online_cpus(); -} - -static void trace_bts_close(struct trace_iterator *iter) -{ - tracing_reset_online_cpus(iter->tr); -} - -void trace_hw_branch_oops(void) -{ - if (this_tracer) { - ds_suspend_bts_noirq(this_tracer); - trace_bts_cpu(hw_branch_trace); - ds_resume_bts_noirq(this_tracer); - } -} - -struct tracer bts_tracer __read_mostly = -{ - .name = "hw-branch-tracer", - .init = bts_trace_init, - .reset = bts_trace_reset, - .print_header = bts_trace_print_header, - .print_line = bts_trace_print_line, - .start = bts_trace_start, - .stop = bts_trace_stop, - .open = trace_bts_prepare, - .close = trace_bts_close, -#ifdef CONFIG_FTRACE_SELFTEST - .selftest = trace_selftest_startup_hw_branches, -#endif /* CONFIG_FTRACE_SELFTEST */ -}; - -__init static int init_bts_trace(void) -{ - register_hotcpu_notifier(&bts_hotcpu_notifier); - return register_tracer(&bts_tracer); -} -device_initcall(init_bts_trace); diff --git a/kernel/trace/trace_selftest.c b/kernel/trace/trace_selftest.c index 280fea470d67..a7084e7c0427 100644 --- a/kernel/trace/trace_selftest.c +++ b/kernel/trace/trace_selftest.c @@ -16,7 +16,6 @@ static inline int trace_valid_entry(struct trace_entry *entry) case TRACE_BRANCH: case TRACE_GRAPH_ENT: case TRACE_GRAPH_RET: - case TRACE_HW_BRANCHES: case TRACE_KSYM: return 1; } @@ -754,62 +753,6 @@ trace_selftest_startup_branch(struct tracer *trace, struct trace_array *tr) } #endif /* CONFIG_BRANCH_TRACER */ -#ifdef CONFIG_HW_BRANCH_TRACER -int -trace_selftest_startup_hw_branches(struct tracer *trace, - struct trace_array *tr) -{ - struct trace_iterator *iter; - struct tracer tracer; - unsigned long count; - int ret; - - if (!trace->open) { - printk(KERN_CONT "missing open function..."); - return -1; - } - - ret = tracer_init(trace, tr); - if (ret) { - warn_failed_init_tracer(trace, ret); - return ret; - } - - /* - * The hw-branch tracer needs to collect the trace from the various - * cpu trace buffers - before tracing is stopped. - */ - iter = kzalloc(sizeof(*iter), GFP_KERNEL); - if (!iter) - return -ENOMEM; - - memcpy(&tracer, trace, sizeof(tracer)); - - iter->trace = &tracer; - iter->tr = tr; - iter->pos = -1; - mutex_init(&iter->mutex); - - trace->open(iter); - - mutex_destroy(&iter->mutex); - kfree(iter); - - tracing_stop(); - - ret = trace_test_buffer(tr, &count); - trace->reset(tr); - tracing_start(); - - if (!ret && !count) { - printk(KERN_CONT "no entries found.."); - ret = -1; - } - - return ret; -} -#endif /* CONFIG_HW_BRANCH_TRACER */ - #ifdef CONFIG_KSYM_TRACER static int ksym_selftest_dummy; diff --git a/mm/mlock.c b/mm/mlock.c index 8f4e2dfceec1..3f82720e0515 100644 --- a/mm/mlock.c +++ b/mm/mlock.c @@ -607,44 +607,3 @@ void user_shm_unlock(size_t size, struct user_struct *user) spin_unlock(&shmlock_user_lock); free_uid(user); } - -int account_locked_memory(struct mm_struct *mm, struct rlimit *rlim, - size_t size) -{ - unsigned long lim, vm, pgsz; - int error = -ENOMEM; - - pgsz = PAGE_ALIGN(size) >> PAGE_SHIFT; - - down_write(&mm->mmap_sem); - - lim = ACCESS_ONCE(rlim[RLIMIT_AS].rlim_cur) >> PAGE_SHIFT; - vm = mm->total_vm + pgsz; - if (lim < vm) - goto out; - - lim = ACCESS_ONCE(rlim[RLIMIT_MEMLOCK].rlim_cur) >> PAGE_SHIFT; - vm = mm->locked_vm + pgsz; - if (lim < vm) - goto out; - - mm->total_vm += pgsz; - mm->locked_vm += pgsz; - - error = 0; - out: - up_write(&mm->mmap_sem); - return error; -} - -void refund_locked_memory(struct mm_struct *mm, size_t size) -{ - unsigned long pgsz = PAGE_ALIGN(size) >> PAGE_SHIFT; - - down_write(&mm->mmap_sem); - - mm->total_vm -= pgsz; - mm->locked_vm -= pgsz; - - up_write(&mm->mmap_sem); -} -- cgit 1.4.1 From 6cc8a7c1d8560c042f486b23318a6291569ab96b Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Fri, 19 Mar 2010 01:23:53 +0100 Subject: perf: Fetch hot regs from the template caller Trace events can be defined from a template using DECLARE_EVENT_CLASS/DEFINE_EVENT or directly with TRACE_EVENT. In both cases we have a template tracepoint handler, used to record the trace, to which we pass our ftrace event instance. In the function level, if the class is named "foo" and the event is named "blah", we have the following chain of calls: perf_trace_blah() -> perf_trace_templ_foo() In the case we have several events sharing the class "blah", we'll have multiple users of perf_trace_templ_foo(), and it won't be inlined by the compiler. This is usually what happens with the DECLARE_EVENT_CLASS/DEFINE_EVENT based definition. But if perf_trace_blah() is the only caller of perf_trace_templ_foo() there are fair chances that it will be inlined. The problem is that we fetch the regs from perf_trace_templ_foo() after we rewinded the frame pointer to the second caller, we want to reach the caller of perf_trace_blah() to get the right source of the event. And we do this by always assuming that perf_trace_templ_foo() is not inlined. But as shown above this is not always true. And if it is inlined we miss the first caller, losing the most important level of precision. We get: 61.31% ls [kernel.kallsyms] [k] do_softirq | --- do_softirq irq_exit do_IRQ common_interrupt | |--25.00%-- tty_buffer_request_room Instead of: 61.31% ls [kernel.kallsyms] [k] __do_softirq | --- __do_softirq do_softirq irq_exit do_IRQ common_interrupt | |--25.00%-- tty_buffer_request_room To fix this, we fetch the regs from perf_trace_blah() rather than perf_trace_templ_foo() so that we don't have to deal with inlining surprises. That also bring us the advantage of having the true source of the event even if we don't have frame pointers. Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Paul Mackerras Cc: Ingo Molnar --- include/trace/ftrace.h | 23 ++++++++++++----------- 1 file changed, 12 insertions(+), 11 deletions(-) (limited to 'include') diff --git a/include/trace/ftrace.h b/include/trace/ftrace.h index ea6f9d4a20e9..882c64832ffe 100644 --- a/include/trace/ftrace.h +++ b/include/trace/ftrace.h @@ -758,13 +758,12 @@ __attribute__((section("_ftrace_events"))) event_##call = { \ #define DECLARE_EVENT_CLASS(call, proto, args, tstruct, assign, print) \ static notrace void \ perf_trace_templ_##call(struct ftrace_event_call *event_call, \ - proto) \ + struct pt_regs *__regs, proto) \ { \ struct ftrace_data_offsets_##call __maybe_unused __data_offsets;\ struct ftrace_raw_##call *entry; \ u64 __addr = 0, __count = 1; \ unsigned long irq_flags; \ - struct pt_regs *__regs; \ int __entry_size; \ int __data_size; \ int rctx; \ @@ -785,20 +784,22 @@ perf_trace_templ_##call(struct ftrace_event_call *event_call, \ \ { assign; } \ \ - __regs = &__get_cpu_var(perf_trace_regs); \ - perf_fetch_caller_regs(__regs, 2); \ - \ perf_trace_buf_submit(entry, __entry_size, rctx, __addr, \ __count, irq_flags, __regs); \ } #undef DEFINE_EVENT -#define DEFINE_EVENT(template, call, proto, args) \ -static notrace void perf_trace_##call(proto) \ -{ \ - struct ftrace_event_call *event_call = &event_##call; \ - \ - perf_trace_templ_##template(event_call, args); \ +#define DEFINE_EVENT(template, call, proto, args) \ +static notrace void perf_trace_##call(proto) \ +{ \ + struct ftrace_event_call *event_call = &event_##call; \ + struct pt_regs *__regs = &get_cpu_var(perf_trace_regs); \ + \ + perf_fetch_caller_regs(__regs, 1); \ + \ + perf_trace_templ_##template(event_call, __regs, args); \ + \ + put_cpu_var(perf_trace_regs); \ } #undef DEFINE_EVENT_PRINT -- cgit 1.4.1 From 76e1d9047e4edefb8ada20aa90d5762306082bd6 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Mon, 5 Apr 2010 15:35:57 +0200 Subject: perf: Store active software events in a hashlist Each time a software event triggers, we need to walk through the entire list of events from the current cpu and task contexts to retrieve a running perf event that matches. We also need to check a matching perf event is actually counting. This walk is wasteful and makes the event fast path scaling down with a growing number of events running on the same contexts. To solve this, we store the running perf events in a hashlist to get an immediate access to them against their type:event_id when they trigger. v2: - Fix SWEVENT_HLIST_SIZE definition (and re-learn some basic maths along the way) - Only allocate hlist for online cpus, but keep track of the refcount on offline possible cpus too, so that we allocate it if needed when it becomes online. - Drop the kref use as it's not adapted to our tricks anymore. v3: - Fix bad refcount check (address instead of value). Thanks to Eric Dumazet who spotted this. - While exiting cpu, move the hlist release out of the IPI path to lock the hlist mutex sanely. Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Paul Mackerras Cc: Ingo Molnar --- include/linux/perf_event.h | 12 +++ kernel/perf_event.c | 246 +++++++++++++++++++++++++++++++++------------ 2 files changed, 195 insertions(+), 63 deletions(-) (limited to 'include') diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index 6e96cc8225d4..bf896d0b2e9c 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -589,6 +589,14 @@ enum perf_group_flag { PERF_GROUP_SOFTWARE = 0x1, }; +#define SWEVENT_HLIST_BITS 8 +#define SWEVENT_HLIST_SIZE (1 << SWEVENT_HLIST_BITS) + +struct swevent_hlist { + struct hlist_head heads[SWEVENT_HLIST_SIZE]; + struct rcu_head rcu_head; +}; + /** * struct perf_event - performance event kernel representation: */ @@ -597,6 +605,7 @@ struct perf_event { struct list_head group_entry; struct list_head event_entry; struct list_head sibling_list; + struct hlist_node hlist_entry; int nr_siblings; int group_flags; struct perf_event *group_leader; @@ -744,6 +753,9 @@ struct perf_cpu_context { int active_oncpu; int max_pertask; int exclusive; + struct swevent_hlist *swevent_hlist; + struct mutex hlist_mutex; + int hlist_refcount; /* * Recursion avoidance: diff --git a/kernel/perf_event.c b/kernel/perf_event.c index fcf42dcd6089..9efdfe5b8d3b 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -16,6 +16,7 @@ #include #include #include +#include #include #include #include @@ -3966,36 +3967,6 @@ static void perf_swevent_add(struct perf_event *event, u64 nr, perf_swevent_overflow(event, 0, nmi, data, regs); } -static int perf_swevent_is_counting(struct perf_event *event) -{ - /* - * The event is active, we're good! - */ - if (event->state == PERF_EVENT_STATE_ACTIVE) - return 1; - - /* - * The event is off/error, not counting. - */ - if (event->state != PERF_EVENT_STATE_INACTIVE) - return 0; - - /* - * The event is inactive, if the context is active - * we're part of a group that didn't make it on the 'pmu', - * not counting. - */ - if (event->ctx->is_active) - return 0; - - /* - * We're inactive and the context is too, this means the - * task is scheduled out, we're counting events that happen - * to us, like migration events. - */ - return 1; -} - static int perf_tp_event_match(struct perf_event *event, struct perf_sample_data *data); @@ -4019,12 +3990,6 @@ static int perf_swevent_match(struct perf_event *event, struct perf_sample_data *data, struct pt_regs *regs) { - if (event->cpu != -1 && event->cpu != smp_processor_id()) - return 0; - - if (!perf_swevent_is_counting(event)) - return 0; - if (event->attr.type != type) return 0; @@ -4041,18 +4006,53 @@ static int perf_swevent_match(struct perf_event *event, return 1; } -static void perf_swevent_ctx_event(struct perf_event_context *ctx, - enum perf_type_id type, - u32 event_id, u64 nr, int nmi, - struct perf_sample_data *data, - struct pt_regs *regs) +static inline u64 swevent_hash(u64 type, u32 event_id) +{ + u64 val = event_id | (type << 32); + + return hash_64(val, SWEVENT_HLIST_BITS); +} + +static struct hlist_head * +find_swevent_head(struct perf_cpu_context *ctx, u64 type, u32 event_id) +{ + u64 hash; + struct swevent_hlist *hlist; + + hash = swevent_hash(type, event_id); + + hlist = rcu_dereference(ctx->swevent_hlist); + if (!hlist) + return NULL; + + return &hlist->heads[hash]; +} + +static void do_perf_sw_event(enum perf_type_id type, u32 event_id, + u64 nr, int nmi, + struct perf_sample_data *data, + struct pt_regs *regs) { + struct perf_cpu_context *cpuctx; struct perf_event *event; + struct hlist_node *node; + struct hlist_head *head; - list_for_each_entry_rcu(event, &ctx->event_list, event_entry) { + cpuctx = &__get_cpu_var(perf_cpu_context); + + rcu_read_lock(); + + head = find_swevent_head(cpuctx, type, event_id); + + if (!head) + goto end; + + hlist_for_each_entry_rcu(event, node, head, hlist_entry) { if (perf_swevent_match(event, type, event_id, data, regs)) perf_swevent_add(event, nr, nmi, data, regs); } +end: + rcu_read_unlock(); } int perf_swevent_get_recursion_context(void) @@ -4090,27 +4090,6 @@ void perf_swevent_put_recursion_context(int rctx) } EXPORT_SYMBOL_GPL(perf_swevent_put_recursion_context); -static void do_perf_sw_event(enum perf_type_id type, u32 event_id, - u64 nr, int nmi, - struct perf_sample_data *data, - struct pt_regs *regs) -{ - struct perf_cpu_context *cpuctx; - struct perf_event_context *ctx; - - cpuctx = &__get_cpu_var(perf_cpu_context); - rcu_read_lock(); - perf_swevent_ctx_event(&cpuctx->ctx, type, event_id, - nr, nmi, data, regs); - /* - * doesn't really matter which of the child contexts the - * events ends up in. - */ - ctx = rcu_dereference(current->perf_event_ctxp); - if (ctx) - perf_swevent_ctx_event(ctx, type, event_id, nr, nmi, data, regs); - rcu_read_unlock(); -} void __perf_sw_event(u32 event_id, u64 nr, int nmi, struct pt_regs *regs, u64 addr) @@ -4136,16 +4115,28 @@ static void perf_swevent_read(struct perf_event *event) static int perf_swevent_enable(struct perf_event *event) { struct hw_perf_event *hwc = &event->hw; + struct perf_cpu_context *cpuctx; + struct hlist_head *head; + + cpuctx = &__get_cpu_var(perf_cpu_context); if (hwc->sample_period) { hwc->last_period = hwc->sample_period; perf_swevent_set_period(event); } + + head = find_swevent_head(cpuctx, event->attr.type, event->attr.config); + if (WARN_ON_ONCE(!head)) + return -EINVAL; + + hlist_add_head_rcu(&event->hlist_entry, head); + return 0; } static void perf_swevent_disable(struct perf_event *event) { + hlist_del_rcu(&event->hlist_entry); } static const struct pmu perf_ops_generic = { @@ -4359,13 +4350,115 @@ static int perf_tp_event_match(struct perf_event *event, return 0; } +static void swevent_hlist_release_rcu(struct rcu_head *rcu_head) +{ + struct swevent_hlist *hlist; + + hlist = container_of(rcu_head, struct swevent_hlist, rcu_head); + kfree(hlist); +} + +static void swevent_hlist_release(struct perf_cpu_context *cpuctx) +{ + struct swevent_hlist *hlist; + + if (!cpuctx->swevent_hlist) + return; + + hlist = cpuctx->swevent_hlist; + rcu_assign_pointer(cpuctx->swevent_hlist, NULL); + call_rcu(&hlist->rcu_head, swevent_hlist_release_rcu); +} + +static void swevent_hlist_put_cpu(struct perf_event *event, int cpu) +{ + struct perf_cpu_context *cpuctx = &per_cpu(perf_cpu_context, cpu); + + mutex_lock(&cpuctx->hlist_mutex); + + if (!--cpuctx->hlist_refcount) + swevent_hlist_release(cpuctx); + + mutex_unlock(&cpuctx->hlist_mutex); +} + +static void swevent_hlist_put(struct perf_event *event) +{ + int cpu; + + if (event->cpu != -1) { + swevent_hlist_put_cpu(event, event->cpu); + return; + } + + for_each_possible_cpu(cpu) + swevent_hlist_put_cpu(event, cpu); +} + +static int swevent_hlist_get_cpu(struct perf_event *event, int cpu) +{ + struct perf_cpu_context *cpuctx = &per_cpu(perf_cpu_context, cpu); + int err = 0; + + mutex_lock(&cpuctx->hlist_mutex); + + if (!cpuctx->swevent_hlist && cpu_online(cpu)) { + struct swevent_hlist *hlist; + + hlist = kzalloc(sizeof(*hlist), GFP_KERNEL); + if (!hlist) { + err = -ENOMEM; + goto exit; + } + rcu_assign_pointer(cpuctx->swevent_hlist, hlist); + } + cpuctx->hlist_refcount++; + exit: + mutex_unlock(&cpuctx->hlist_mutex); + + return err; +} + +static int swevent_hlist_get(struct perf_event *event) +{ + int err; + int cpu, failed_cpu; + + if (event->cpu != -1) + return swevent_hlist_get_cpu(event, event->cpu); + + get_online_cpus(); + for_each_possible_cpu(cpu) { + err = swevent_hlist_get_cpu(event, cpu); + if (err) { + failed_cpu = cpu; + goto fail; + } + } + put_online_cpus(); + + return 0; + fail: + for_each_possible_cpu(cpu) { + if (cpu == failed_cpu) + break; + swevent_hlist_put_cpu(event, cpu); + } + + put_online_cpus(); + return err; +} + static void tp_perf_event_destroy(struct perf_event *event) { perf_trace_disable(event->attr.config); + swevent_hlist_put(event); } static const struct pmu *tp_perf_event_init(struct perf_event *event) { + int err; + /* * Raw tracepoint data is a severe data leak, only allow root to * have these. @@ -4379,6 +4472,11 @@ static const struct pmu *tp_perf_event_init(struct perf_event *event) return NULL; event->destroy = tp_perf_event_destroy; + err = swevent_hlist_get(event); + if (err) { + perf_trace_disable(event->attr.config); + return ERR_PTR(err); + } return &perf_ops_generic; } @@ -4479,6 +4577,7 @@ static void sw_perf_event_destroy(struct perf_event *event) WARN_ON(event->parent); atomic_dec(&perf_swevent_enabled[event_id]); + swevent_hlist_put(event); } static const struct pmu *sw_perf_event_init(struct perf_event *event) @@ -4517,6 +4616,12 @@ static const struct pmu *sw_perf_event_init(struct perf_event *event) case PERF_COUNT_SW_ALIGNMENT_FAULTS: case PERF_COUNT_SW_EMULATION_FAULTS: if (!event->parent) { + int err; + + err = swevent_hlist_get(event); + if (err) + return ERR_PTR(err); + atomic_inc(&perf_swevent_enabled[event_id]); event->destroy = sw_perf_event_destroy; } @@ -5389,6 +5494,7 @@ static void __init perf_event_init_all_cpus(void) for_each_possible_cpu(cpu) { cpuctx = &per_cpu(perf_cpu_context, cpu); + mutex_init(&cpuctx->hlist_mutex); __perf_event_init_context(&cpuctx->ctx, NULL); } } @@ -5402,6 +5508,16 @@ static void __cpuinit perf_event_init_cpu(int cpu) spin_lock(&perf_resource_lock); cpuctx->max_pertask = perf_max_events - perf_reserved_percpu; spin_unlock(&perf_resource_lock); + + mutex_lock(&cpuctx->hlist_mutex); + if (cpuctx->hlist_refcount > 0) { + struct swevent_hlist *hlist; + + hlist = kzalloc(sizeof(*hlist), GFP_KERNEL); + WARN_ON_ONCE(!hlist); + rcu_assign_pointer(cpuctx->swevent_hlist, hlist); + } + mutex_unlock(&cpuctx->hlist_mutex); } #ifdef CONFIG_HOTPLUG_CPU @@ -5421,6 +5537,10 @@ static void perf_event_exit_cpu(int cpu) struct perf_cpu_context *cpuctx = &per_cpu(perf_cpu_context, cpu); struct perf_event_context *ctx = &cpuctx->ctx; + mutex_lock(&cpuctx->hlist_mutex); + swevent_hlist_release(cpuctx); + mutex_unlock(&cpuctx->hlist_mutex); + mutex_lock(&ctx->mutex); smp_call_function_single(cpu, __perf_event_exit_cpu, NULL, 1); mutex_unlock(&ctx->mutex); -- cgit 1.4.1 From 39447b386c846bbf1c56f6403c5282837486200f Mon Sep 17 00:00:00 2001 From: "Zhang, Yanmin" Date: Mon, 19 Apr 2010 13:32:41 +0800 Subject: perf: Enhance perf to allow for guest statistic collection from host Below patch introduces perf_guest_info_callbacks and related register/unregister functions. Add more PERF_RECORD_MISC_XXX bits meaning guest kernel and guest user space. Signed-off-by: Zhang Yanmin Signed-off-by: Avi Kivity --- arch/x86/include/asm/perf_event.h | 15 ++++----------- arch/x86/kernel/cpu/perf_event.c | 31 +++++++++++++++++++++++++++++++ include/linux/perf_event.h | 21 ++++++++++++++++++++- kernel/perf_event.c | 23 ++++++++++++++++++++++- 4 files changed, 77 insertions(+), 13 deletions(-) (limited to 'include') diff --git a/arch/x86/include/asm/perf_event.h b/arch/x86/include/asm/perf_event.h index f6d43dbfd8e7..254883d0c7e0 100644 --- a/arch/x86/include/asm/perf_event.h +++ b/arch/x86/include/asm/perf_event.h @@ -135,17 +135,10 @@ extern void perf_events_lapic_init(void); */ #define PERF_EFLAGS_EXACT (1UL << 3) -#define perf_misc_flags(regs) \ -({ int misc = 0; \ - if (user_mode(regs)) \ - misc |= PERF_RECORD_MISC_USER; \ - else \ - misc |= PERF_RECORD_MISC_KERNEL; \ - if (regs->flags & PERF_EFLAGS_EXACT) \ - misc |= PERF_RECORD_MISC_EXACT; \ - misc; }) - -#define perf_instruction_pointer(regs) ((regs)->ip) +struct pt_regs; +extern unsigned long perf_instruction_pointer(struct pt_regs *regs); +extern unsigned long perf_misc_flags(struct pt_regs *regs); +#define perf_misc_flags(regs) perf_misc_flags(regs) #else static inline void init_hw_perf_events(void) { } diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c index 626154a9f535..2ea78abf69d9 100644 --- a/arch/x86/kernel/cpu/perf_event.c +++ b/arch/x86/kernel/cpu/perf_event.c @@ -1720,6 +1720,11 @@ struct perf_callchain_entry *perf_callchain(struct pt_regs *regs) { struct perf_callchain_entry *entry; + if (perf_guest_cbs && perf_guest_cbs->is_in_guest()) { + /* TODO: We don't support guest os callchain now */ + return NULL; + } + if (in_nmi()) entry = &__get_cpu_var(pmc_nmi_entry); else @@ -1743,3 +1748,29 @@ void perf_arch_fetch_caller_regs(struct pt_regs *regs, unsigned long ip, int ski regs->cs = __KERNEL_CS; local_save_flags(regs->flags); } + +unsigned long perf_instruction_pointer(struct pt_regs *regs) +{ + unsigned long ip; + if (perf_guest_cbs && perf_guest_cbs->is_in_guest()) + ip = perf_guest_cbs->get_guest_ip(); + else + ip = instruction_pointer(regs); + return ip; +} + +unsigned long perf_misc_flags(struct pt_regs *regs) +{ + int misc = 0; + if (perf_guest_cbs && perf_guest_cbs->is_in_guest()) { + misc |= perf_guest_cbs->is_user_mode() ? + PERF_RECORD_MISC_GUEST_USER : + PERF_RECORD_MISC_GUEST_KERNEL; + } else + misc |= user_mode(regs) ? PERF_RECORD_MISC_USER : + PERF_RECORD_MISC_KERNEL; + if (regs->flags & PERF_EFLAGS_EXACT) + misc |= PERF_RECORD_MISC_EXACT; + + return misc; +} diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index bf896d0b2e9c..24de5f181a41 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -288,11 +288,13 @@ struct perf_event_mmap_page { __u64 data_tail; /* user-space written tail */ }; -#define PERF_RECORD_MISC_CPUMODE_MASK (3 << 0) +#define PERF_RECORD_MISC_CPUMODE_MASK (7 << 0) #define PERF_RECORD_MISC_CPUMODE_UNKNOWN (0 << 0) #define PERF_RECORD_MISC_KERNEL (1 << 0) #define PERF_RECORD_MISC_USER (2 << 0) #define PERF_RECORD_MISC_HYPERVISOR (3 << 0) +#define PERF_RECORD_MISC_GUEST_KERNEL (4 << 0) +#define PERF_RECORD_MISC_GUEST_USER (5 << 0) #define PERF_RECORD_MISC_EXACT (1 << 14) /* @@ -446,6 +448,12 @@ enum perf_callchain_context { # include #endif +struct perf_guest_info_callbacks { + int (*is_in_guest) (void); + int (*is_user_mode) (void); + unsigned long (*get_guest_ip) (void); +}; + #ifdef CONFIG_HAVE_HW_BREAKPOINT #include #endif @@ -932,6 +940,12 @@ static inline void perf_event_mmap(struct vm_area_struct *vma) __perf_event_mmap(vma); } +extern struct perf_guest_info_callbacks *perf_guest_cbs; +extern int perf_register_guest_info_callbacks( + struct perf_guest_info_callbacks *); +extern int perf_unregister_guest_info_callbacks( + struct perf_guest_info_callbacks *); + extern void perf_event_comm(struct task_struct *tsk); extern void perf_event_fork(struct task_struct *tsk); @@ -1001,6 +1015,11 @@ perf_sw_event(u32 event_id, u64 nr, int nmi, static inline void perf_bp_event(struct perf_event *event, void *data) { } +static inline int perf_register_guest_info_callbacks +(struct perf_guest_info_callbacks *) {return 0; } +static inline int perf_unregister_guest_info_callbacks +(struct perf_guest_info_callbacks *) {return 0; } + static inline void perf_event_mmap(struct vm_area_struct *vma) { } static inline void perf_event_comm(struct task_struct *tsk) { } static inline void perf_event_fork(struct task_struct *tsk) { } diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 07b7a435bf03..9dbe8cdaf145 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -2797,6 +2797,27 @@ void perf_arch_fetch_caller_regs(struct pt_regs *regs, unsigned long ip, int ski } +/* + * We assume there is only KVM supporting the callbacks. + * Later on, we might change it to a list if there is + * another virtualization implementation supporting the callbacks. + */ +struct perf_guest_info_callbacks *perf_guest_cbs; + +int perf_register_guest_info_callbacks(struct perf_guest_info_callbacks *cbs) +{ + perf_guest_cbs = cbs; + return 0; +} +EXPORT_SYMBOL_GPL(perf_register_guest_info_callbacks); + +int perf_unregister_guest_info_callbacks(struct perf_guest_info_callbacks *cbs) +{ + perf_guest_cbs = NULL; + return 0; +} +EXPORT_SYMBOL_GPL(perf_unregister_guest_info_callbacks); + /* * Output */ @@ -3749,7 +3770,7 @@ void __perf_event_mmap(struct vm_area_struct *vma) .event_id = { .header = { .type = PERF_RECORD_MMAP, - .misc = 0, + .misc = PERF_RECORD_MISC_USER, /* .size */ }, /* .pid */ -- cgit 1.4.1