From cfbcf468454ab4b20f0b4b62da51920b99fdb19e Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Thu, 28 Apr 2016 12:30:53 -0300 Subject: perf core: Pass max stack as a perf_callchain_entry context This makes perf_callchain_{user,kernel}() receive the max stack as context for the perf_callchain_entry, instead of accessing the global sysctl_perf_event_max_stack. Cc: Adrian Hunter Cc: Alexander Shishkin Cc: Alexei Starovoitov Cc: Brendan Gregg Cc: David Ahern Cc: Frederic Weisbecker Cc: He Kuang Cc: Jiri Olsa Cc: Linus Torvalds Cc: Masami Hiramatsu Cc: Milian Wolff Cc: Namhyung Kim Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Thomas Gleixner Cc: Vince Weaver Cc: Wang Nan Cc: Zefan Li Link: http://lkml.kernel.org/n/tip-kolmn1yo40p7jhswxwrc7rrd@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- include/linux/perf_event.h | 16 +++++++++++----- 1 file changed, 11 insertions(+), 5 deletions(-) (limited to 'include') diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index 9e1c3ada91c4..dbd18246b36e 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -61,6 +61,11 @@ struct perf_callchain_entry { __u64 ip[0]; /* /proc/sys/kernel/perf_event_max_stack */ }; +struct perf_callchain_entry_ctx { + struct perf_callchain_entry *entry; + u32 max_stack; +}; + struct perf_raw_record { u32 size; void *data; @@ -1063,19 +1068,20 @@ extern void perf_event_fork(struct task_struct *tsk); /* Callchains */ DECLARE_PER_CPU(struct perf_callchain_entry, perf_callchain_entry); -extern void perf_callchain_user(struct perf_callchain_entry *entry, struct pt_regs *regs); -extern void perf_callchain_kernel(struct perf_callchain_entry *entry, struct pt_regs *regs); +extern void perf_callchain_user(struct perf_callchain_entry_ctx *entry, struct pt_regs *regs); +extern void perf_callchain_kernel(struct perf_callchain_entry_ctx *entry, struct pt_regs *regs); extern struct perf_callchain_entry * get_perf_callchain(struct pt_regs *regs, u32 init_nr, bool kernel, bool user, - bool crosstask, bool add_mark); + u32 max_stack, bool crosstask, bool add_mark); extern int get_callchain_buffers(void); extern void put_callchain_buffers(void); extern int sysctl_perf_event_max_stack; -static inline int perf_callchain_store(struct perf_callchain_entry *entry, u64 ip) +static inline int perf_callchain_store(struct perf_callchain_entry_ctx *ctx, u64 ip) { - if (entry->nr < sysctl_perf_event_max_stack) { + struct perf_callchain_entry *entry = ctx->entry; + if (entry->nr < ctx->max_stack) { entry->ip[entry->nr++] = ip; return 0; } else { -- cgit v1.2.3 From 3b1fff08038bd0792b1aa1e9703b2dd0512a3fd0 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 10 May 2016 18:08:32 -0300 Subject: perf core: Add a 'nr' field to perf_event_callchain_context We will use it to count how many addresses are in the entry->ip[] array, excluding PERF_CONTEXT_{KERNEL,USER,etc} entries, so that we can really return the number of entries specified by the user via the relevant sysctl, kernel.perf_event_max_contexts, or via the per event perf_event_attr.sample_max_stack knob. This way we keep the perf_sample->ip_callchain->nr meaning, that is the number of entries, be it real addresses or PERF_CONTEXT_ entries, while honouring the max_stack knobs, i.e. the end result will be max_stack entries if we have at least that many entries in a given stack trace. Cc: David Ahern Cc: Frederic Weisbecker Cc: Jiri Olsa Cc: Namhyung Kim Cc: Peter Zijlstra Link: http://lkml.kernel.org/n/tip-s8teto51tdqvlfhefndtat9r@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- arch/arm/kernel/perf_callchain.c | 2 +- arch/arm64/kernel/perf_callchain.c | 4 ++-- arch/metag/kernel/perf_callchain.c | 2 +- arch/mips/kernel/perf_event.c | 4 ++-- arch/powerpc/perf/callchain.c | 4 ++-- arch/sparc/kernel/perf_event.c | 6 +++--- arch/x86/events/core.c | 4 ++-- include/linux/perf_event.h | 6 ++++-- kernel/events/callchain.c | 3 +-- 9 files changed, 18 insertions(+), 17 deletions(-) (limited to 'include') diff --git a/arch/arm/kernel/perf_callchain.c b/arch/arm/kernel/perf_callchain.c index bc552e813e7b..22bf1f64d99a 100644 --- a/arch/arm/kernel/perf_callchain.c +++ b/arch/arm/kernel/perf_callchain.c @@ -75,7 +75,7 @@ perf_callchain_user(struct perf_callchain_entry_ctx *entry, struct pt_regs *regs tail = (struct frame_tail __user *)regs->ARM_fp - 1; - while ((entry->entry->nr < entry->max_stack) && + while ((entry->nr < entry->max_stack) && tail && !((unsigned long)tail & 0x3)) tail = user_backtrace(tail, entry); } diff --git a/arch/arm64/kernel/perf_callchain.c b/arch/arm64/kernel/perf_callchain.c index 0d60150057cf..713ca824f266 100644 --- a/arch/arm64/kernel/perf_callchain.c +++ b/arch/arm64/kernel/perf_callchain.c @@ -122,7 +122,7 @@ void perf_callchain_user(struct perf_callchain_entry_ctx *entry, tail = (struct frame_tail __user *)regs->regs[29]; - while (entry->entry->nr < entry->max_stack && + while (entry->nr < entry->max_stack && tail && !((unsigned long)tail & 0xf)) tail = user_backtrace(tail, entry); } else { @@ -132,7 +132,7 @@ void perf_callchain_user(struct perf_callchain_entry_ctx *entry, tail = (struct compat_frame_tail __user *)regs->compat_fp - 1; - while ((entry->entry->nr < entry->max_stack) && + while ((entry->nr < entry->max_stack) && tail && !((unsigned long)tail & 0x3)) tail = compat_user_backtrace(tail, entry); #endif diff --git a/arch/metag/kernel/perf_callchain.c b/arch/metag/kernel/perf_callchain.c index b3261a98b15b..3e8e048040df 100644 --- a/arch/metag/kernel/perf_callchain.c +++ b/arch/metag/kernel/perf_callchain.c @@ -65,7 +65,7 @@ perf_callchain_user(struct perf_callchain_entry_ctx *entry, struct pt_regs *regs --frame; - while ((entry->entry->nr < entry->max_stack) && frame) + while ((entry->nr < entry->max_stack) && frame) frame = user_backtrace(frame, entry); } diff --git a/arch/mips/kernel/perf_event.c b/arch/mips/kernel/perf_event.c index 22395c7d7030..d64056e0bb56 100644 --- a/arch/mips/kernel/perf_event.c +++ b/arch/mips/kernel/perf_event.c @@ -35,7 +35,7 @@ static void save_raw_perf_callchain(struct perf_callchain_entry_ctx *entry, addr = *sp++; if (__kernel_text_address(addr)) { perf_callchain_store(entry, addr); - if (entry->entry->nr >= entry->max_stack) + if (entry->nr >= entry->max_stack) break; } } @@ -59,7 +59,7 @@ void perf_callchain_kernel(struct perf_callchain_entry_ctx *entry, } do { perf_callchain_store(entry, pc); - if (entry->entry->nr >= entry->max_stack) + if (entry->nr >= entry->max_stack) break; pc = unwind_stack(current, &sp, pc, &ra); } while (pc); diff --git a/arch/powerpc/perf/callchain.c b/arch/powerpc/perf/callchain.c index c9260c1dfdbc..f68f213dc36c 100644 --- a/arch/powerpc/perf/callchain.c +++ b/arch/powerpc/perf/callchain.c @@ -247,7 +247,7 @@ static void perf_callchain_user_64(struct perf_callchain_entry_ctx *entry, sp = regs->gpr[1]; perf_callchain_store(entry, next_ip); - while (entry->entry->nr < entry->max_stack) { + while (entry->nr < entry->max_stack) { fp = (unsigned long __user *) sp; if (!valid_user_sp(sp, 1) || read_user_stack_64(fp, &next_sp)) return; @@ -453,7 +453,7 @@ static void perf_callchain_user_32(struct perf_callchain_entry_ctx *entry, sp = regs->gpr[1]; perf_callchain_store(entry, next_ip); - while (entry->entry->nr < entry->max_stack) { + while (entry->nr < entry->max_stack) { fp = (unsigned int __user *) (unsigned long) sp; if (!valid_user_sp(sp, 0) || read_user_stack_32(fp, &next_sp)) return; diff --git a/arch/sparc/kernel/perf_event.c b/arch/sparc/kernel/perf_event.c index bcc5376db74b..710f3278d448 100644 --- a/arch/sparc/kernel/perf_event.c +++ b/arch/sparc/kernel/perf_event.c @@ -1756,7 +1756,7 @@ void perf_callchain_kernel(struct perf_callchain_entry_ctx *entry, } } #endif - } while (entry->entry->nr < entry->max_stack); + } while (entry->nr < entry->max_stack); } static inline int @@ -1790,7 +1790,7 @@ static void perf_callchain_user_64(struct perf_callchain_entry_ctx *entry, pc = sf.callers_pc; ufp = (unsigned long)sf.fp + STACK_BIAS; perf_callchain_store(entry, pc); - } while (entry->entry->nr < entry->max_stack); + } while (entry->nr < entry->max_stack); } static void perf_callchain_user_32(struct perf_callchain_entry_ctx *entry, @@ -1822,7 +1822,7 @@ static void perf_callchain_user_32(struct perf_callchain_entry_ctx *entry, ufp = (unsigned long)sf.fp; } perf_callchain_store(entry, pc); - } while (entry->entry->nr < entry->max_stack); + } while (entry->nr < entry->max_stack); } void diff --git a/arch/x86/events/core.c b/arch/x86/events/core.c index 07f2b01cfb72..5de96a18cd9c 100644 --- a/arch/x86/events/core.c +++ b/arch/x86/events/core.c @@ -2283,7 +2283,7 @@ perf_callchain_user32(struct pt_regs *regs, struct perf_callchain_entry_ctx *ent fp = compat_ptr(ss_base + regs->bp); pagefault_disable(); - while (entry->entry->nr < entry->max_stack) { + while (entry->nr < entry->max_stack) { unsigned long bytes; frame.next_frame = 0; frame.return_address = 0; @@ -2343,7 +2343,7 @@ perf_callchain_user(struct perf_callchain_entry_ctx *entry, struct pt_regs *regs return; pagefault_disable(); - while (entry->entry->nr < entry->max_stack) { + while (entry->nr < entry->max_stack) { unsigned long bytes; frame.next_frame = NULL; frame.return_address = 0; diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index dbd18246b36e..3803bb1a862b 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -64,6 +64,7 @@ struct perf_callchain_entry { struct perf_callchain_entry_ctx { struct perf_callchain_entry *entry; u32 max_stack; + u32 nr; }; struct perf_raw_record { @@ -1080,9 +1081,10 @@ extern int sysctl_perf_event_max_stack; static inline int perf_callchain_store(struct perf_callchain_entry_ctx *ctx, u64 ip) { - struct perf_callchain_entry *entry = ctx->entry; - if (entry->nr < ctx->max_stack) { + if (ctx->nr < ctx->max_stack) { + struct perf_callchain_entry *entry = ctx->entry; entry->ip[entry->nr++] = ip; + ++ctx->nr; return 0; } else { return -1; /* no more room, stop walking the stack */ diff --git a/kernel/events/callchain.c b/kernel/events/callchain.c index af95ad92893a..8774ff86debb 100644 --- a/kernel/events/callchain.c +++ b/kernel/events/callchain.c @@ -196,8 +196,7 @@ get_perf_callchain(struct pt_regs *regs, u32 init_nr, bool kernel, bool user, ctx.entry = entry; ctx.max_stack = max_stack; - - entry->nr = init_nr; + ctx.nr = entry->nr = init_nr; if (kernel && !user_mode(regs)) { if (add_mark) -- cgit v1.2.3 From 3e4de4ec4cfea40994b47a79767610153edbf45b Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Thu, 12 May 2016 13:01:50 -0300 Subject: perf core: Add perf_callchain_store_context() helper We need have different helpers to account how many contexts we have in the sample and for real addresses, so do it now as a prep patch, to ease review. Cc: David Ahern Cc: Frederic Weisbecker Cc: Jiri Olsa Cc: Namhyung Kim Cc: Peter Zijlstra Link: http://lkml.kernel.org/n/tip-q964tnyuqrxw5gld18vizs3c@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- arch/powerpc/perf/callchain.c | 6 +++--- include/linux/perf_event.h | 2 ++ kernel/events/callchain.c | 4 ++-- 3 files changed, 7 insertions(+), 5 deletions(-) (limited to 'include') diff --git a/arch/powerpc/perf/callchain.c b/arch/powerpc/perf/callchain.c index f68f213dc36c..f62597dbd757 100644 --- a/arch/powerpc/perf/callchain.c +++ b/arch/powerpc/perf/callchain.c @@ -76,7 +76,7 @@ perf_callchain_kernel(struct perf_callchain_entry_ctx *entry, struct pt_regs *re next_ip = regs->nip; lr = regs->link; level = 0; - perf_callchain_store(entry, PERF_CONTEXT_KERNEL); + perf_callchain_store_context(entry, PERF_CONTEXT_KERNEL); } else { if (level == 0) @@ -274,7 +274,7 @@ static void perf_callchain_user_64(struct perf_callchain_entry_ctx *entry, read_user_stack_64(&uregs[PT_R1], &sp)) return; level = 0; - perf_callchain_store(entry, PERF_CONTEXT_USER); + perf_callchain_store_context(entry, PERF_CONTEXT_USER); perf_callchain_store(entry, next_ip); continue; } @@ -473,7 +473,7 @@ static void perf_callchain_user_32(struct perf_callchain_entry_ctx *entry, read_user_stack_32(&uregs[PT_R1], &sp)) return; level = 0; - perf_callchain_store(entry, PERF_CONTEXT_USER); + perf_callchain_store_context(entry, PERF_CONTEXT_USER); perf_callchain_store(entry, next_ip); continue; } diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index 3803bb1a862b..2024b14cc2b1 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -1079,6 +1079,8 @@ extern void put_callchain_buffers(void); extern int sysctl_perf_event_max_stack; +#define perf_callchain_store_context(ctx, context) perf_callchain_store(ctx, context) + static inline int perf_callchain_store(struct perf_callchain_entry_ctx *ctx, u64 ip) { if (ctx->nr < ctx->max_stack) { diff --git a/kernel/events/callchain.c b/kernel/events/callchain.c index 8774ff86debb..ca645736a983 100644 --- a/kernel/events/callchain.c +++ b/kernel/events/callchain.c @@ -200,7 +200,7 @@ get_perf_callchain(struct pt_regs *regs, u32 init_nr, bool kernel, bool user, if (kernel && !user_mode(regs)) { if (add_mark) - perf_callchain_store(&ctx, PERF_CONTEXT_KERNEL); + perf_callchain_store_context(&ctx, PERF_CONTEXT_KERNEL); perf_callchain_kernel(&ctx, regs); } @@ -217,7 +217,7 @@ get_perf_callchain(struct pt_regs *regs, u32 init_nr, bool kernel, bool user, goto exit_put; if (add_mark) - perf_callchain_store(&ctx, PERF_CONTEXT_USER); + perf_callchain_store_context(&ctx, PERF_CONTEXT_USER); perf_callchain_user(&ctx, regs); } } -- cgit v1.2.3 From c85b03349640b34f3545503c8429fc43005e9a92 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Thu, 12 May 2016 13:06:21 -0300 Subject: perf core: Separate accounting of contexts and real addresses in a stack trace The perf_sample->ip_callchain->nr value includes all the entries in the ip_callchain->ip[] array, real addresses and PERF_CONTEXT_{KERNEL,USER,etc}, while what the user expects is that what is in the kernel.perf_event_max_stack sysctl or in the upcoming per event perf_event_attr.sample_max_stack knob be honoured in terms of IP addresses in the stack trace. So allocate a bunch of extra entries for contexts, and do the accounting via perf_callchain_entry_ctx struct members. A new sysctl, kernel.perf_event_max_contexts_per_stack is also introduced for investigating possible bugs in the callchain implementation by some arch. Cc: Adrian Hunter Cc: Alexander Shishkin Cc: Alexei Starovoitov Cc: Brendan Gregg Cc: David Ahern Cc: Frederic Weisbecker Cc: He Kuang Cc: Jiri Olsa Cc: Masami Hiramatsu Cc: Milian Wolff Cc: Namhyung Kim Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Thomas Gleixner Cc: Vince Weaver Cc: Wang Nan Cc: Zefan Li Link: http://lkml.kernel.org/n/tip-3b4wnqk340c4sg4gwkfdi9yk@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- Documentation/sysctl/kernel.txt | 14 ++++++++++++++ include/linux/perf_event.h | 18 ++++++++++++++++-- include/uapi/linux/perf_event.h | 1 + kernel/events/callchain.c | 10 +++++++++- kernel/sysctl.c | 9 +++++++++ 5 files changed, 49 insertions(+), 3 deletions(-) (limited to 'include') diff --git a/Documentation/sysctl/kernel.txt b/Documentation/sysctl/kernel.txt index daabdd7ee543..a3683ce2a2f3 100644 --- a/Documentation/sysctl/kernel.txt +++ b/Documentation/sysctl/kernel.txt @@ -61,6 +61,7 @@ show up in /proc/sys/kernel: - perf_cpu_time_max_percent - perf_event_paranoid - perf_event_max_stack +- perf_event_max_contexts_per_stack - pid_max - powersave-nap [ PPC only ] - printk @@ -668,6 +669,19 @@ The default value is 127. ============================================================== +perf_event_max_contexts_per_stack: + +Controls maximum number of stack frame context entries for +(attr.sample_type & PERF_SAMPLE_CALLCHAIN) configured events, for +instance, when using 'perf record -g' or 'perf trace --call-graph fp'. + +This can only be done when no events are in use that have callchains +enabled, otherwise writing to this file will return -EBUSY. + +The default value is 8. + +============================================================== + pid_max: PID allocation wrap value. When the kernel's next PID value diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index 2024b14cc2b1..6b87be908790 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -65,6 +65,8 @@ struct perf_callchain_entry_ctx { struct perf_callchain_entry *entry; u32 max_stack; u32 nr; + short contexts; + bool contexts_maxed; }; struct perf_raw_record { @@ -1078,12 +1080,24 @@ extern int get_callchain_buffers(void); extern void put_callchain_buffers(void); extern int sysctl_perf_event_max_stack; +extern int sysctl_perf_event_max_contexts_per_stack; -#define perf_callchain_store_context(ctx, context) perf_callchain_store(ctx, context) +static inline int perf_callchain_store_context(struct perf_callchain_entry_ctx *ctx, u64 ip) +{ + if (ctx->contexts < sysctl_perf_event_max_contexts_per_stack) { + struct perf_callchain_entry *entry = ctx->entry; + entry->ip[entry->nr++] = ip; + ++ctx->contexts; + return 0; + } else { + ctx->contexts_maxed = true; + return -1; /* no more room, stop walking the stack */ + } +} static inline int perf_callchain_store(struct perf_callchain_entry_ctx *ctx, u64 ip) { - if (ctx->nr < ctx->max_stack) { + if (ctx->nr < ctx->max_stack && !ctx->contexts_maxed) { struct perf_callchain_entry *entry = ctx->entry; entry->ip[entry->nr++] = ip; ++ctx->nr; diff --git a/include/uapi/linux/perf_event.h b/include/uapi/linux/perf_event.h index 43fc8d213472..36ce552cf6a9 100644 --- a/include/uapi/linux/perf_event.h +++ b/include/uapi/linux/perf_event.h @@ -862,6 +862,7 @@ enum perf_event_type { }; #define PERF_MAX_STACK_DEPTH 127 +#define PERF_MAX_CONTEXTS_PER_STACK 8 enum perf_callchain_context { PERF_CONTEXT_HV = (__u64)-32, diff --git a/kernel/events/callchain.c b/kernel/events/callchain.c index ca645736a983..179ef4640964 100644 --- a/kernel/events/callchain.c +++ b/kernel/events/callchain.c @@ -19,11 +19,13 @@ struct callchain_cpus_entries { }; int sysctl_perf_event_max_stack __read_mostly = PERF_MAX_STACK_DEPTH; +int sysctl_perf_event_max_contexts_per_stack __read_mostly = PERF_MAX_CONTEXTS_PER_STACK; static inline size_t perf_callchain_entry__sizeof(void) { return (sizeof(struct perf_callchain_entry) + - sizeof(__u64) * sysctl_perf_event_max_stack); + sizeof(__u64) * (sysctl_perf_event_max_stack + + sysctl_perf_event_max_contexts_per_stack)); } static DEFINE_PER_CPU(int, callchain_recursion[PERF_NR_CONTEXTS]); @@ -197,6 +199,8 @@ get_perf_callchain(struct pt_regs *regs, u32 init_nr, bool kernel, bool user, ctx.entry = entry; ctx.max_stack = max_stack; ctx.nr = entry->nr = init_nr; + ctx.contexts = 0; + ctx.contexts_maxed = false; if (kernel && !user_mode(regs)) { if (add_mark) @@ -228,6 +232,10 @@ exit_put: return entry; } +/* + * Used for sysctl_perf_event_max_stack and + * sysctl_perf_event_max_contexts_per_stack. + */ int perf_event_max_stack_handler(struct ctl_table *table, int write, void __user *buffer, size_t *lenp, loff_t *ppos) { diff --git a/kernel/sysctl.c b/kernel/sysctl.c index 0ec6907a16b3..bec4c11c47d6 100644 --- a/kernel/sysctl.c +++ b/kernel/sysctl.c @@ -1156,6 +1156,15 @@ static struct ctl_table kern_table[] = { .extra1 = &zero, .extra2 = &six_hundred_forty_kb, }, + { + .procname = "perf_event_max_contexts_per_stack", + .data = &sysctl_perf_event_max_contexts_per_stack, + .maxlen = sizeof(sysctl_perf_event_max_contexts_per_stack), + .mode = 0644, + .proc_handler = perf_event_max_stack_handler, + .extra1 = &zero, + .extra2 = &one_thousand, + }, #endif #ifdef CONFIG_KMEMCHECK { -- cgit v1.2.3