/*
- * kerneltop.c: show top kernel functions - performance counters showcase
-
- Build with:
-
- make -C Documentation/perf_counter/
-
- Sample output:
-
-------------------------------------------------------------------------------
- KernelTop: 2669 irqs/sec [NMI, cache-misses/cache-refs], (all, cpu: 2)
-------------------------------------------------------------------------------
-
- weight RIP kernel function
- ______ ________________ _______________
-
- 35.20 - ffffffff804ce74b : skb_copy_and_csum_dev
- 33.00 - ffffffff804cb740 : sock_alloc_send_skb
- 31.26 - ffffffff804ce808 : skb_push
- 22.43 - ffffffff80510004 : tcp_established_options
- 19.00 - ffffffff8027d250 : find_get_page
- 15.76 - ffffffff804e4fc9 : eth_type_trans
- 15.20 - ffffffff804d8baa : dst_release
- 14.86 - ffffffff804cf5d8 : skb_release_head_state
- 14.00 - ffffffff802217d5 : read_hpet
- 12.00 - ffffffff804ffb7f : __ip_local_out
- 11.97 - ffffffff804fc0c8 : ip_local_deliver_finish
- 8.54 - ffffffff805001a3 : ip_queue_xmit
+ * builtin-top.c
+ *
+ * Builtin top command: Display a continuously updated profile of
+ * any workload, CPU or specific PID.
+ *
+ * Copyright (C) 2008, Red Hat Inc, Ingo Molnar <mingo@redhat.com>
+ *
+ * Improvements and fixes by:
+ *
+ * Arjan van de Ven <arjan@linux.intel.com>
+ * Yanmin Zhang <yanmin.zhang@intel.com>
+ * Wu Fengguang <fengguang.wu@intel.com>
+ * Mike Galbraith <efault@gmx.de>
+ * Paul Mackerras <paulus@samba.org>
+ *
+ * Released under the GPL v2. (and only v2, not any later version)
*/
+#include "builtin.h"
- /*
- * Copyright (C) 2008, Red Hat Inc, Ingo Molnar <mingo@redhat.com>
- *
- * Improvements and fixes by:
- *
- * Arjan van de Ven <arjan@linux.intel.com>
- * Yanmin Zhang <yanmin.zhang@intel.com>
- * Wu Fengguang <fengguang.wu@intel.com>
- * Mike Galbraith <efault@gmx.de>
- * Paul Mackerras <paulus@samba.org>
- *
- * Released under the GPL v2. (and only v2, not any later version)
- */
+#include "perf.h"
+#include "util/symbol.h"
+#include "util/color.h"
#include "util/util.h"
+#include "util/rbtree.h"
+#include "util/parse-options.h"
+#include "util/parse-events.h"
-#include <getopt.h>
#include <assert.h>
#include <fcntl.h>
+
#include <stdio.h>
+
#include <errno.h>
#include <time.h>
#include <sched.h>
#include <linux/unistd.h>
#include <linux/types.h>
-#include "../../include/linux/perf_counter.h"
-
-#include "perf.h"
-
static int system_wide = 0;
-static int nr_counters = 0;
-static __u64 event_id[MAX_COUNTERS] = {
+static __u64 default_event_id[MAX_COUNTERS] = {
EID(PERF_TYPE_SOFTWARE, PERF_COUNT_TASK_CLOCK),
EID(PERF_TYPE_SOFTWARE, PERF_COUNT_CONTEXT_SWITCHES),
EID(PERF_TYPE_SOFTWARE, PERF_COUNT_CPU_MIGRATIONS),
static int event_count[MAX_COUNTERS];
static int fd[MAX_NR_CPUS][MAX_COUNTERS];
-static __u64 count_filter = 100;
+static __u64 count_filter = 5;
+static int print_entries = 15;
-static int tid = -1;
+static int target_pid = -1;
static int profile_cpu = -1;
static int nr_cpus = 0;
-static int nmi = 1;
static unsigned int realtime_prio = 0;
static int group = 0;
static unsigned int page_size;
static unsigned int mmap_pages = 16;
-static int use_mmap = 0;
-static int use_munmap = 0;
static int freq = 0;
-static char *vmlinux;
-
static char *sym_filter;
static unsigned long filter_start;
static unsigned long filter_end;
static int zero;
static int dump_symtab;
-static int scale;
-
-struct source_line {
- uint64_t EIP;
- unsigned long count;
- char *line;
- struct source_line *next;
-};
-
-static struct source_line *lines;
-static struct source_line **lines_tail;
-
static const unsigned int default_count[] = {
1000000,
1000000,
10000,
};
-static char *hw_event_names[] = {
- "CPU cycles",
- "instructions",
- "cache references",
- "cache misses",
- "branches",
- "branch misses",
- "bus cycles",
-};
-
-static char *sw_event_names[] = {
- "cpu clock ticks",
- "task clock ticks",
- "pagefaults",
- "context switches",
- "CPU migrations",
- "minor faults",
- "major faults",
-};
-
-struct event_symbol {
- __u64 event;
- char *symbol;
-};
-
-static struct event_symbol event_symbols[] = {
- {EID(PERF_TYPE_HARDWARE, PERF_COUNT_CPU_CYCLES), "cpu-cycles", },
- {EID(PERF_TYPE_HARDWARE, PERF_COUNT_CPU_CYCLES), "cycles", },
- {EID(PERF_TYPE_HARDWARE, PERF_COUNT_INSTRUCTIONS), "instructions", },
- {EID(PERF_TYPE_HARDWARE, PERF_COUNT_CACHE_REFERENCES), "cache-references", },
- {EID(PERF_TYPE_HARDWARE, PERF_COUNT_CACHE_MISSES), "cache-misses", },
- {EID(PERF_TYPE_HARDWARE, PERF_COUNT_BRANCH_INSTRUCTIONS), "branch-instructions", },
- {EID(PERF_TYPE_HARDWARE, PERF_COUNT_BRANCH_INSTRUCTIONS), "branches", },
- {EID(PERF_TYPE_HARDWARE, PERF_COUNT_BRANCH_MISSES), "branch-misses", },
- {EID(PERF_TYPE_HARDWARE, PERF_COUNT_BUS_CYCLES), "bus-cycles", },
-
- {EID(PERF_TYPE_SOFTWARE, PERF_COUNT_CPU_CLOCK), "cpu-clock", },
- {EID(PERF_TYPE_SOFTWARE, PERF_COUNT_TASK_CLOCK), "task-clock", },
- {EID(PERF_TYPE_SOFTWARE, PERF_COUNT_PAGE_FAULTS), "page-faults", },
- {EID(PERF_TYPE_SOFTWARE, PERF_COUNT_PAGE_FAULTS), "faults", },
- {EID(PERF_TYPE_SOFTWARE, PERF_COUNT_PAGE_FAULTS_MIN), "minor-faults", },
- {EID(PERF_TYPE_SOFTWARE, PERF_COUNT_PAGE_FAULTS_MAJ), "major-faults", },
- {EID(PERF_TYPE_SOFTWARE, PERF_COUNT_CONTEXT_SWITCHES), "context-switches", },
- {EID(PERF_TYPE_SOFTWARE, PERF_COUNT_CONTEXT_SWITCHES), "cs", },
- {EID(PERF_TYPE_SOFTWARE, PERF_COUNT_CPU_MIGRATIONS), "cpu-migrations", },
- {EID(PERF_TYPE_SOFTWARE, PERF_COUNT_CPU_MIGRATIONS), "migrations", },
-};
-
-#define __PERF_COUNTER_FIELD(config, name) \
- ((config & PERF_COUNTER_##name##_MASK) >> PERF_COUNTER_##name##_SHIFT)
-
-#define PERF_COUNTER_RAW(config) __PERF_COUNTER_FIELD(config, RAW)
-#define PERF_COUNTER_CONFIG(config) __PERF_COUNTER_FIELD(config, CONFIG)
-#define PERF_COUNTER_TYPE(config) __PERF_COUNTER_FIELD(config, TYPE)
-#define PERF_COUNTER_ID(config) __PERF_COUNTER_FIELD(config, EVENT)
-
-static void display_events_help(void)
-{
- unsigned int i;
- __u64 e;
-
- printf(
- " -e EVENT --event=EVENT # symbolic-name abbreviations");
-
- for (i = 0; i < ARRAY_SIZE(event_symbols); i++) {
- int type, id;
-
- e = event_symbols[i].event;
- type = PERF_COUNTER_TYPE(e);
- id = PERF_COUNTER_ID(e);
-
- printf("\n %d:%d: %-20s",
- type, id, event_symbols[i].symbol);
- }
-
- printf("\n"
- " rNNN: raw PMU events (eventsel+umask)\n\n");
-}
-
-static void display_help(void)
-{
- printf(
- "Usage: kerneltop [<options>]\n"
- " Or: kerneltop -S [<options>] COMMAND [ARGS]\n\n"
- "KernelTop Options (up to %d event types can be specified at once):\n\n",
- MAX_COUNTERS);
-
- display_events_help();
-
- printf(
- " -c CNT --count=CNT # event period to sample\n\n"
- " -C CPU --cpu=CPU # CPU (-1 for all) [default: -1]\n"
- " -p PID --pid=PID # PID of sampled task (-1 for all) [default: -1]\n\n"
- " -l # show scale factor for RR events\n"
- " -d delay --delay=<seconds> # sampling/display delay [default: 2]\n"
- " -f CNT --filter=CNT # min-event-count filter [default: 100]\n\n"
- " -r prio --realtime=<prio> # event acquisition runs with SCHED_FIFO policy\n"
- " -s symbol --symbol=<symbol> # function to be showed annotated one-shot\n"
- " -x path --vmlinux=<path> # the vmlinux binary, required for -s use\n"
- " -z --zero # zero counts after display\n"
- " -D --dump_symtab # dump symbol table to stderr on startup\n"
- " -m pages --mmap_pages=<pages> # number of mmap data pages\n"
- " -M --mmap_info # print mmap info stream\n"
- " -U --munmap_info # print munmap info stream\n"
- );
-
- exit(0);
-}
-
-static char *event_name(int ctr)
-{
- __u64 config = event_id[ctr];
- int type = PERF_COUNTER_TYPE(config);
- int id = PERF_COUNTER_ID(config);
- static char buf[32];
-
- if (PERF_COUNTER_RAW(config)) {
- sprintf(buf, "raw 0x%llx", PERF_COUNTER_CONFIG(config));
- return buf;
- }
-
- switch (type) {
- case PERF_TYPE_HARDWARE:
- if (id < PERF_HW_EVENTS_MAX)
- return hw_event_names[id];
- return "unknown-hardware";
-
- case PERF_TYPE_SOFTWARE:
- if (id < PERF_SW_EVENTS_MAX)
- return sw_event_names[id];
- return "unknown-software";
-
- default:
- break;
- }
-
- return "unknown";
-}
-
-/*
- * Each event can have multiple symbolic names.
- * Symbolic names are (almost) exactly matched.
- */
-static __u64 match_event_symbols(char *str)
-{
- __u64 config, id;
- int type;
- unsigned int i;
-
- if (sscanf(str, "r%llx", &config) == 1)
- return config | PERF_COUNTER_RAW_MASK;
-
- if (sscanf(str, "%d:%llu", &type, &id) == 2)
- return EID(type, id);
-
- for (i = 0; i < ARRAY_SIZE(event_symbols); i++) {
- if (!strncmp(str, event_symbols[i].symbol,
- strlen(event_symbols[i].symbol)))
- return event_symbols[i].event;
- }
-
- return ~0ULL;
-}
-
-static int parse_events(char *str)
-{
- __u64 config;
-
-again:
- if (nr_counters == MAX_COUNTERS)
- return -1;
-
- config = match_event_symbols(str);
- if (config == ~0ULL)
- return -1;
-
- event_id[nr_counters] = config;
- nr_counters++;
-
- str = strstr(str, ",");
- if (str) {
- str++;
- goto again;
- }
-
- return 0;
-}
-
/*
* Symbols
*/
static uint64_t max_ip = -1ll;
struct sym_entry {
- unsigned long long addr;
- char *sym;
+ struct rb_node rb_node;
+ struct list_head node;
unsigned long count[MAX_COUNTERS];
+ unsigned long snap_count;
+ double weight;
int skip;
- struct source_line *source;
};
-#define MAX_SYMS 100000
-
-static int sym_table_count;
-
struct sym_entry *sym_filter_entry;
-static struct sym_entry sym_table[MAX_SYMS];
+struct dso *kernel_dso;
-static void show_details(struct sym_entry *sym);
+/*
+ * Symbols will be added here in record_ip and will get out
+ * after decayed.
+ */
+static LIST_HEAD(active_symbols);
+static pthread_mutex_t active_symbols_lock = PTHREAD_MUTEX_INITIALIZER;
/*
* Ordering weight: count-1 * count-2 * ... / count-n
*/
static double sym_weight(const struct sym_entry *sym)
{
- double weight;
+ double weight = sym->snap_count;
int counter;
- weight = sym->count[0];
-
for (counter = 1; counter < nr_counters-1; counter++)
weight *= sym->count[counter];
return weight;
}
-static int compare(const void *__sym1, const void *__sym2)
+static long events;
+static long userspace_events;
+static const char CONSOLE_CLEAR[] = "\e[H\e[2J";
+
+static void __list_insert_active_sym(struct sym_entry *syme)
{
- const struct sym_entry *sym1 = __sym1, *sym2 = __sym2;
+ list_add(&syme->node, &active_symbols);
+}
- return sym_weight(sym1) < sym_weight(sym2);
+static void list_remove_active_sym(struct sym_entry *syme)
+{
+ pthread_mutex_lock(&active_symbols_lock);
+ list_del_init(&syme->node);
+ pthread_mutex_unlock(&active_symbols_lock);
}
-static long events;
-static long userspace_events;
-static const char CONSOLE_CLEAR[] = "\e[H\e[2J";
+static void rb_insert_active_sym(struct rb_root *tree, struct sym_entry *se)
+{
+ struct rb_node **p = &tree->rb_node;
+ struct rb_node *parent = NULL;
+ struct sym_entry *iter;
-static struct sym_entry tmp[MAX_SYMS];
+ while (*p != NULL) {
+ parent = *p;
+ iter = rb_entry(parent, struct sym_entry, rb_node);
+
+ if (se->weight > iter->weight)
+ p = &(*p)->rb_left;
+ else
+ p = &(*p)->rb_right;
+ }
+
+ rb_link_node(&se->rb_node, parent, p);
+ rb_insert_color(&se->rb_node, tree);
+}
static void print_sym_table(void)
{
- int i, printed;
+ int printed = 0, j;
int counter;
float events_per_sec = events/delay_secs;
float kevents_per_sec = (events-userspace_events)/delay_secs;
float sum_kevents = 0.0;
+ struct sym_entry *syme, *n;
+ struct rb_root tmp = RB_ROOT;
+ struct rb_node *nd;
events = userspace_events = 0;
- memcpy(tmp, sym_table, sizeof(sym_table[0])*sym_table_count);
- qsort(tmp, sym_table_count, sizeof(tmp[0]), compare);
- for (i = 0; i < sym_table_count && tmp[i].count[0]; i++)
- sum_kevents += tmp[i].count[0];
+ /* Sort the active symbols */
+ pthread_mutex_lock(&active_symbols_lock);
+ syme = list_entry(active_symbols.next, struct sym_entry, node);
+ pthread_mutex_unlock(&active_symbols_lock);
+
+ list_for_each_entry_safe_from(syme, n, &active_symbols, node) {
+ syme->snap_count = syme->count[0];
+ if (syme->snap_count != 0) {
+ syme->weight = sym_weight(syme);
+ rb_insert_active_sym(&tmp, syme);
+ sum_kevents += syme->snap_count;
+
+ for (j = 0; j < nr_counters; j++)
+ syme->count[j] = zero ? 0 : syme->count[j] * 7 / 8;
+ } else
+ list_remove_active_sym(syme);
+ }
write(1, CONSOLE_CLEAR, strlen(CONSOLE_CLEAR));
printf(
"------------------------------------------------------------------------------\n");
- printf( " KernelTop:%8.0f irqs/sec kernel:%4.1f%% [%s, ",
+ printf( " PerfTop:%8.0f irqs/sec kernel:%4.1f%% [",
events_per_sec,
- 100.0 - (100.0*((events_per_sec-kevents_per_sec)/events_per_sec)),
- nmi ? "NMI" : "IRQ");
+ 100.0 - (100.0*((events_per_sec-kevents_per_sec)/events_per_sec)));
if (nr_counters == 1)
printf("%d ", event_count[0]);
printf( "], ");
- if (tid != -1)
- printf(" (tid: %d", tid);
+ if (target_pid != -1)
+ printf(" (target_pid: %d", target_pid);
else
printf(" (all");
if (profile_cpu != -1)
printf(", cpu: %d)\n", profile_cpu);
else {
- if (tid != -1)
+ if (target_pid != -1)
printf(")\n");
else
printf(", %d CPUs)\n", nr_cpus);
" ______ ______ _____ ________________ _______________\n\n"
);
- for (i = 0, printed = 0; i < sym_table_count; i++) {
- float pcnt;
- int count;
-
- if (printed <= 18 && tmp[i].count[0] >= count_filter) {
- pcnt = 100.0 - (100.0*((sum_kevents-tmp[i].count[0])/sum_kevents));
-
- if (nr_counters == 1)
- printf("%19.2f - %4.1f%% - %016llx : %s\n",
- sym_weight(tmp + i),
- pcnt, tmp[i].addr, tmp[i].sym);
- else
- printf("%8.1f %10ld - %4.1f%% - %016llx : %s\n",
- sym_weight(tmp + i),
- tmp[i].count[0],
- pcnt, tmp[i].addr, tmp[i].sym);
- printed++;
- }
+ for (nd = rb_first(&tmp); nd; nd = rb_next(nd)) {
+ struct sym_entry *syme = rb_entry(nd, struct sym_entry, rb_node);
+ struct symbol *sym = (struct symbol *)(syme + 1);
+ char *color = PERF_COLOR_NORMAL;
+ double pcnt;
+
+ if (++printed > print_entries || syme->snap_count < count_filter)
+ continue;
+
+ pcnt = 100.0 - (100.0 * ((sum_kevents - syme->snap_count) /
+ sum_kevents));
+
/*
- * Add decay to the counts:
+ * We color high-overhead entries in red, low-overhead
+ * entries in green - and keep the middle ground normal:
*/
- for (count = 0; count < nr_counters; count++)
- sym_table[i].count[count] = zero ? 0 : sym_table[i].count[count] * 7 / 8;
- }
+ if (pcnt >= 5.0)
+ color = PERF_COLOR_RED;
+ if (pcnt < 0.5)
+ color = PERF_COLOR_GREEN;
- if (sym_filter_entry)
- show_details(sym_filter_entry);
+ if (nr_counters == 1)
+ printf("%19.2f - ", syme->weight);
+ else
+ printf("%8.1f %10ld - ", syme->weight, syme->snap_count);
+
+ color_fprintf(stdout, color, "%4.1f%%", pcnt);
+ printf(" - %016llx : %s\n", sym->start, sym->name);
+ }
{
struct pollfd stdin_poll = { .fd = 0, .events = POLLIN };
static void *display_thread(void *arg)
{
- printf("KernelTop refresh period: %d seconds\n", delay_secs);
+ printf("PerfTop refresh period: %d seconds\n", delay_secs);
while (!sleep(delay_secs))
print_sym_table();
return NULL;
}
-static int read_symbol(FILE *in, struct sym_entry *s)
+static int symbol_filter(struct dso *self, struct symbol *sym)
{
- static int filter_match = 0;
- char *sym, stype;
- char str[500];
- int rc, pos;
-
- rc = fscanf(in, "%llx %c %499s", &s->addr, &stype, str);
- if (rc == EOF)
- return -1;
-
- assert(rc == 3);
-
- /* skip until end of line: */
- pos = strlen(str);
- do {
- rc = fgetc(in);
- if (rc == '\n' || rc == EOF || pos >= 499)
- break;
- str[pos] = rc;
- pos++;
- } while (1);
- str[pos] = 0;
-
- sym = str;
-
- /* Filter out known duplicates and non-text symbols. */
- if (!strcmp(sym, "_text"))
- return 1;
- if (!min_ip && !strcmp(sym, "_stext"))
- return 1;
- if (!strcmp(sym, "_etext") || !strcmp(sym, "_sinittext"))
- return 1;
- if (stype != 'T' && stype != 't')
- return 1;
- if (!strncmp("init_module", sym, 11) || !strncmp("cleanup_module", sym, 14))
+ static int filter_match;
+ struct sym_entry *syme;
+ const char *name = sym->name;
+
+ if (!strcmp(name, "_text") ||
+ !strcmp(name, "_etext") ||
+ !strcmp(name, "_sinittext") ||
+ !strncmp("init_module", name, 11) ||
+ !strncmp("cleanup_module", name, 14) ||
+ strstr(name, "_text_start") ||
+ strstr(name, "_text_end"))
return 1;
- if (strstr(sym, "_text_start") || strstr(sym, "_text_end"))
- return 1;
-
- s->sym = malloc(strlen(str)+1);
- assert(s->sym);
-
- strcpy((char *)s->sym, str);
- s->skip = 0;
+ syme = dso__sym_priv(self, sym);
/* Tag events to be skipped. */
- if (!strcmp("default_idle", s->sym) || !strcmp("cpu_idle", s->sym))
- s->skip = 1;
- else if (!strcmp("enter_idle", s->sym) || !strcmp("exit_idle", s->sym))
- s->skip = 1;
- else if (!strcmp("mwait_idle", s->sym))
- s->skip = 1;
+ if (!strcmp("default_idle", name) ||
+ !strcmp("cpu_idle", name) ||
+ !strcmp("enter_idle", name) ||
+ !strcmp("exit_idle", name) ||
+ !strcmp("mwait_idle", name))
+ syme->skip = 1;
if (filter_match == 1) {
- filter_end = s->addr;
+ filter_end = sym->start;
filter_match = -1;
if (filter_end - filter_start > 10000) {
- printf("hm, too large filter symbol <%s> - skipping.\n",
+ fprintf(stderr,
+ "hm, too large filter symbol <%s> - skipping.\n",
sym_filter);
- printf("symbol filter start: %016lx\n", filter_start);
- printf(" end: %016lx\n", filter_end);
+ fprintf(stderr, "symbol filter start: %016lx\n",
+ filter_start);
+ fprintf(stderr, " end: %016lx\n",
+ filter_end);
filter_end = filter_start = 0;
sym_filter = NULL;
sleep(1);
}
}
- if (filter_match == 0 && sym_filter && !strcmp(s->sym, sym_filter)) {
- filter_match = 1;
- filter_start = s->addr;
- }
-
- return 0;
-}
-static int compare_addr(const void *__sym1, const void *__sym2)
-{
- const struct sym_entry *sym1 = __sym1, *sym2 = __sym2;
-
- return sym1->addr > sym2->addr;
-}
-
-static void sort_symbol_table(void)
-{
- int i, dups;
-
- do {
- qsort(sym_table, sym_table_count, sizeof(sym_table[0]), compare_addr);
- for (i = 0, dups = 0; i < sym_table_count; i++) {
- if (sym_table[i].addr == sym_table[i+1].addr) {
- sym_table[i+1].addr = -1ll;
- dups++;
- }
- }
- sym_table_count -= dups;
- } while(dups);
-}
-
-static void parse_symbols(void)
-{
- struct sym_entry *last;
-
- FILE *kallsyms = fopen("/proc/kallsyms", "r");
-
- if (!kallsyms) {
- printf("Could not open /proc/kallsyms - no CONFIG_KALLSYMS_ALL=y?\n");
- exit(-1);
- }
-
- while (!feof(kallsyms)) {
- if (read_symbol(kallsyms, &sym_table[sym_table_count]) == 0) {
- sym_table_count++;
- assert(sym_table_count <= MAX_SYMS);
- }
+ if (filter_match == 0 && sym_filter && !strcmp(name, sym_filter)) {
+ filter_match = 1;
+ filter_start = sym->start;
}
- sort_symbol_table();
- min_ip = sym_table[0].addr;
- max_ip = sym_table[sym_table_count-1].addr;
- last = sym_table + sym_table_count++;
- last->addr = -1ll;
- last->sym = "<end>";
-
- if (filter_end) {
- int count;
- for (count=0; count < sym_table_count; count ++) {
- if (!strcmp(sym_table[count].sym, sym_filter)) {
- sym_filter_entry = &sym_table[count];
- break;
- }
- }
- }
- if (dump_symtab) {
- int i;
-
- for (i = 0; i < sym_table_count; i++)
- fprintf(stderr, "%llx %s\n",
- sym_table[i].addr, sym_table[i].sym);
- }
+ return 0;
}
-/*
- * Source lines
- */
-
-static void parse_vmlinux(char *filename)
+static int parse_symbols(void)
{
- FILE *file;
- char command[PATH_MAX*2];
- if (!filename)
- return;
-
- sprintf(command, "objdump --start-address=0x%016lx --stop-address=0x%016lx -dS %s", filter_start, filter_end, filename);
+ struct rb_node *node;
+ struct symbol *sym;
- file = popen(command, "r");
- if (!file)
- return;
+ kernel_dso = dso__new("[kernel]", sizeof(struct sym_entry));
+ if (kernel_dso == NULL)
+ return -1;
- lines_tail = &lines;
- while (!feof(file)) {
- struct source_line *src;
- size_t dummy = 0;
- char *c;
-
- src = malloc(sizeof(struct source_line));
- assert(src != NULL);
- memset(src, 0, sizeof(struct source_line));
-
- if (getline(&src->line, &dummy, file) < 0)
- break;
- if (!src->line)
- break;
-
- c = strchr(src->line, '\n');
- if (c)
- *c = 0;
-
- src->next = NULL;
- *lines_tail = src;
- lines_tail = &src->next;
-
- if (strlen(src->line)>8 && src->line[8] == ':')
- src->EIP = strtoull(src->line, NULL, 16);
- if (strlen(src->line)>8 && src->line[16] == ':')
- src->EIP = strtoull(src->line, NULL, 16);
- }
- pclose(file);
-}
+ if (dso__load_kernel(kernel_dso, NULL, symbol_filter, 1) != 0)
+ goto out_delete_dso;
-static void record_precise_ip(uint64_t ip)
-{
- struct source_line *line;
+ node = rb_first(&kernel_dso->syms);
+ sym = rb_entry(node, struct symbol, rb_node);
+ min_ip = sym->start;
- for (line = lines; line; line = line->next) {
- if (line->EIP == ip)
- line->count++;
- if (line->EIP > ip)
- break;
- }
-}
+ node = rb_last(&kernel_dso->syms);
+ sym = rb_entry(node, struct symbol, rb_node);
+ max_ip = sym->end;
-static void lookup_sym_in_vmlinux(struct sym_entry *sym)
-{
- struct source_line *line;
- char pattern[PATH_MAX];
- sprintf(pattern, "<%s>:", sym->sym);
-
- for (line = lines; line; line = line->next) {
- if (strstr(line->line, pattern)) {
- sym->source = line;
- break;
- }
- }
-}
+ if (dump_symtab)
+ dso__fprintf(kernel_dso, stderr);
-static void show_lines(struct source_line *line_queue, int line_queue_count)
-{
- int i;
- struct source_line *line;
+ return 0;
- line = line_queue;
- for (i = 0; i < line_queue_count; i++) {
- printf("%8li\t%s\n", line->count, line->line);
- line = line->next;
- }
+out_delete_dso:
+ dso__delete(kernel_dso);
+ kernel_dso = NULL;
+ return -1;
}
#define TRACE_COUNT 3
-static void show_details(struct sym_entry *sym)
-{
- struct source_line *line;
- struct source_line *line_queue = NULL;
- int displayed = 0;
- int line_queue_count = 0;
-
- if (!sym->source)
- lookup_sym_in_vmlinux(sym);
- if (!sym->source)
- return;
-
- printf("Showing details for %s\n", sym->sym);
-
- line = sym->source;
- while (line) {
- if (displayed && strstr(line->line, ">:"))
- break;
-
- if (!line_queue_count)
- line_queue = line;
- line_queue_count ++;
-
- if (line->count >= count_filter) {
- show_lines(line_queue, line_queue_count);
- line_queue_count = 0;
- line_queue = NULL;
- } else if (line_queue_count > TRACE_COUNT) {
- line_queue = line_queue->next;
- line_queue_count --;
- }
-
- line->count = 0;
- displayed++;
- if (displayed > 300)
- break;
- line = line->next;
- }
-}
-
/*
* Binary search in the histogram table and record the hit:
*/
static void record_ip(uint64_t ip, int counter)
{
- int left_idx, middle_idx, right_idx, idx;
- unsigned long left, middle, right;
-
- record_precise_ip(ip);
-
- left_idx = 0;
- right_idx = sym_table_count-1;
- assert(ip <= max_ip && ip >= min_ip);
-
- while (left_idx + 1 < right_idx) {
- middle_idx = (left_idx + right_idx) / 2;
-
- left = sym_table[ left_idx].addr;
- middle = sym_table[middle_idx].addr;
- right = sym_table[ right_idx].addr;
-
- if (!(left <= middle && middle <= right)) {
- printf("%016lx...\n%016lx...\n%016lx\n", left, middle, right);
- printf("%d %d %d\n", left_idx, middle_idx, right_idx);
- }
- assert(left <= middle && middle <= right);
- if (!(left <= ip && ip <= right)) {
- printf(" left: %016lx\n", left);
- printf(" ip: %016lx\n", (unsigned long)ip);
- printf("right: %016lx\n", right);
- }
- assert(left <= ip && ip <= right);
- /*
- * [ left .... target .... middle .... right ]
- * => right := middle
- */
- if (ip < middle) {
- right_idx = middle_idx;
- continue;
+ struct symbol *sym = dso__find_symbol(kernel_dso, ip);
+
+ if (sym != NULL) {
+ struct sym_entry *syme = dso__sym_priv(kernel_dso, sym);
+
+ if (!syme->skip) {
+ syme->count[counter]++;
+ pthread_mutex_lock(&active_symbols_lock);
+ if (list_empty(&syme->node) || !syme->node.next)
+ __list_insert_active_sym(syme);
+ pthread_mutex_unlock(&active_symbols_lock);
+ return;
}
- /*
- * [ left .... middle ... target ... right ]
- * => left := middle
- */
- left_idx = middle_idx;
}
- idx = left_idx;
-
- if (!sym_table[idx].skip)
- sym_table[idx].count[counter]++;
- else events--;
+ events--;
}
static void process_event(uint64_t ip, int counter)
record_ip(ip, counter);
}
-static void process_options(int argc, char **argv)
-{
- int error = 0, counter;
-
- for (;;) {
- int option_index = 0;
- /** Options for getopt */
- static struct option long_options[] = {
- {"count", required_argument, NULL, 'c'},
- {"cpu", required_argument, NULL, 'C'},
- {"delay", required_argument, NULL, 'd'},
- {"dump_symtab", no_argument, NULL, 'D'},
- {"event", required_argument, NULL, 'e'},
- {"filter", required_argument, NULL, 'f'},
- {"group", required_argument, NULL, 'g'},
- {"help", no_argument, NULL, 'h'},
- {"nmi", required_argument, NULL, 'n'},
- {"mmap_info", no_argument, NULL, 'M'},
- {"mmap_pages", required_argument, NULL, 'm'},
- {"munmap_info", no_argument, NULL, 'U'},
- {"pid", required_argument, NULL, 'p'},
- {"realtime", required_argument, NULL, 'r'},
- {"scale", no_argument, NULL, 'l'},
- {"symbol", required_argument, NULL, 's'},
- {"stat", no_argument, NULL, 'S'},
- {"vmlinux", required_argument, NULL, 'x'},
- {"zero", no_argument, NULL, 'z'},
- {"freq", required_argument, NULL, 'F'},
- {NULL, 0, NULL, 0 }
- };
- int c = getopt_long(argc, argv, "+:ac:C:d:De:f:g:hln:m:p:r:s:Sx:zMUF:",
- long_options, &option_index);
- if (c == -1)
- break;
-
- switch (c) {
- case 'a': system_wide = 1; break;
- case 'c': default_interval = atoi(optarg); break;
- case 'C':
- /* CPU and PID are mutually exclusive */
- if (tid != -1) {
- printf("WARNING: CPU switch overriding PID\n");
- sleep(1);
- tid = -1;
- }
- profile_cpu = atoi(optarg); break;
- case 'd': delay_secs = atoi(optarg); break;
- case 'D': dump_symtab = 1; break;
-
- case 'e': error = parse_events(optarg); break;
-
- case 'f': count_filter = atoi(optarg); break;
- case 'g': group = atoi(optarg); break;
- case 'h': display_help(); break;
- case 'l': scale = 1; break;
- case 'n': nmi = atoi(optarg); break;
- case 'p':
- /* CPU and PID are mutually exclusive */
- if (profile_cpu != -1) {
- printf("WARNING: PID switch overriding CPU\n");
- sleep(1);
- profile_cpu = -1;
- }
- tid = atoi(optarg); break;
- case 'r': realtime_prio = atoi(optarg); break;
- case 's': sym_filter = strdup(optarg); break;
- case 'x': vmlinux = strdup(optarg); break;
- case 'z': zero = 1; break;
- case 'm': mmap_pages = atoi(optarg); break;
- case 'M': use_mmap = 1; break;
- case 'U': use_munmap = 1; break;
- case 'F': freq = 1; default_interval = atoi(optarg); break;
- default: error = 1; break;
- }
- }
- if (error)
- display_help();
-
- if (!nr_counters) {
- nr_counters = 1;
- event_id[0] = 0;
- }
-
- for (counter = 0; counter < nr_counters; counter++) {
- if (event_count[counter])
- continue;
-
- event_count[counter] = default_interval;
- }
-}
-
struct mmap_data {
int counter;
void *base;
struct ip_event {
struct perf_event_header header;
__u64 ip;
- __u32 pid, tid;
+ __u32 pid, target_pid;
};
struct mmap_event {
struct perf_event_header header;
- __u32 pid, tid;
+ __u32 pid, target_pid;
__u64 start;
__u64 len;
__u64 pgoff;
old += size;
if (event->header.misc & PERF_EVENT_MISC_OVERFLOW) {
- if (event->header.type & PERF_RECORD_IP)
+ if (event->header.type & PERF_SAMPLE_IP)
process_event(event->ip.ip, md->counter);
- } else {
- switch (event->header.type) {
- case PERF_EVENT_MMAP:
- case PERF_EVENT_MUNMAP:
- printf("%s: %Lu %Lu %Lu %s\n",
- event->header.type == PERF_EVENT_MMAP
- ? "mmap" : "munmap",
- event->mmap.start,
- event->mmap.len,
- event->mmap.pgoff,
- event->mmap.filename);
- break;
- }
}
}
md->prev = old;
}
-int cmd_top(int argc, char **argv, const char *prefix)
+static struct pollfd event_array[MAX_NR_CPUS * MAX_COUNTERS];
+static struct mmap_data mmap_array[MAX_NR_CPUS][MAX_COUNTERS];
+
+static int __cmd_top(void)
{
- struct pollfd event_array[MAX_NR_CPUS * MAX_COUNTERS];
- struct mmap_data mmap_array[MAX_NR_CPUS][MAX_COUNTERS];
- struct perf_counter_hw_event hw_event;
+ struct perf_counter_attr attr;
pthread_t thread;
int i, counter, group_fd, nr_poll = 0;
unsigned int cpu;
int ret;
- page_size = sysconf(_SC_PAGE_SIZE);
-
- process_options(argc, argv);
-
- nr_cpus = sysconf(_SC_NPROCESSORS_ONLN);
- assert(nr_cpus <= MAX_NR_CPUS);
- assert(nr_cpus >= 0);
-
- if (tid != -1 || profile_cpu != -1)
- nr_cpus = 1;
-
- parse_symbols();
- if (vmlinux && sym_filter_entry)
- parse_vmlinux(vmlinux);
-
for (i = 0; i < nr_cpus; i++) {
group_fd = -1;
for (counter = 0; counter < nr_counters; counter++) {
cpu = profile_cpu;
- if (tid == -1 && profile_cpu == -1)
+ if (target_pid == -1 && profile_cpu == -1)
cpu = i;
- memset(&hw_event, 0, sizeof(hw_event));
- hw_event.config = event_id[counter];
- hw_event.irq_period = event_count[counter];
- hw_event.record_type = PERF_RECORD_IP | PERF_RECORD_TID;
- hw_event.nmi = nmi;
- hw_event.mmap = use_mmap;
- hw_event.munmap = use_munmap;
- hw_event.freq = freq;
+ memset(&attr, 0, sizeof(attr));
+ attr.config = event_id[counter];
+ attr.sample_period = event_count[counter];
+ attr.sample_type = PERF_SAMPLE_IP | PERF_SAMPLE_TID;
+ attr.freq = freq;
- fd[i][counter] = sys_perf_counter_open(&hw_event, tid, cpu, group_fd, 0);
+ fd[i][counter] = sys_perf_counter_open(&attr, target_pid, cpu, group_fd, 0);
if (fd[i][counter] < 0) {
int err = errno;
- printf("kerneltop error: syscall returned with %d (%s)\n",
+
+ error("syscall returned with %d (%s)\n",
fd[i][counter], strerror(err));
if (err == EPERM)
printf("Are you root?\n");
mmap_array[i][counter].mask = mmap_pages*page_size - 1;
mmap_array[i][counter].base = mmap(NULL, (mmap_pages+1)*page_size,
PROT_READ, MAP_SHARED, fd[i][counter], 0);
- if (mmap_array[i][counter].base == MAP_FAILED) {
- printf("kerneltop error: failed to mmap with %d (%s)\n",
- errno, strerror(errno));
- exit(-1);
- }
+ if (mmap_array[i][counter].base == MAP_FAILED)
+ die("failed to mmap with %d (%s)\n", errno, strerror(errno));
}
}
return 0;
}
+
+static const char * const top_usage[] = {
+ "perf top [<options>]",
+ NULL
+};
+
+static char events_help_msg[EVENTS_HELP_MAX];
+
+static const struct option options[] = {
+ OPT_CALLBACK('e', "event", NULL, "event",
+ events_help_msg, parse_events),
+ OPT_INTEGER('c', "count", &default_interval,
+ "event period to sample"),
+ OPT_INTEGER('p', "pid", &target_pid,
+ "profile events on existing pid"),
+ OPT_BOOLEAN('a', "all-cpus", &system_wide,
+ "system-wide collection from all CPUs"),
+ OPT_INTEGER('C', "CPU", &profile_cpu,
+ "CPU to profile on"),
+ OPT_INTEGER('m', "mmap-pages", &mmap_pages,
+ "number of mmap data pages"),
+ OPT_INTEGER('r', "realtime", &realtime_prio,
+ "collect data with this RT SCHED_FIFO priority"),
+ OPT_INTEGER('d', "delay", &delay_secs,
+ "number of seconds to delay between refreshes"),
+ OPT_BOOLEAN('D', "dump-symtab", &dump_symtab,
+ "dump the symbol table used for profiling"),
+ OPT_INTEGER('f', "count-filter", &count_filter,
+ "only display functions with more events than this"),
+ OPT_BOOLEAN('g', "group", &group,
+ "put the counters into a counter group"),
+ OPT_STRING('s', "sym-filter", &sym_filter, "pattern",
+ "only display symbols matchig this pattern"),
+ OPT_BOOLEAN('z', "zero", &group,
+ "zero history across updates"),
+ OPT_INTEGER('F', "freq", &freq,
+ "profile at this frequency"),
+ OPT_INTEGER('E', "entries", &print_entries,
+ "display this many functions"),
+ OPT_END()
+};
+
+int cmd_top(int argc, const char **argv, const char *prefix)
+{
+ int counter;
+
+ page_size = sysconf(_SC_PAGE_SIZE);
+
+ create_events_help(events_help_msg);
+ memcpy(event_id, default_event_id, sizeof(default_event_id));
+
+ argc = parse_options(argc, argv, options, top_usage, 0);
+ if (argc)
+ usage_with_options(top_usage, options);
+
+ if (freq) {
+ default_interval = freq;
+ freq = 1;
+ }
+
+ /* CPU and PID are mutually exclusive */
+ if (target_pid != -1 && profile_cpu != -1) {
+ printf("WARNING: PID switch overriding CPU\n");
+ sleep(1);
+ profile_cpu = -1;
+ }
+
+ if (!nr_counters) {
+ nr_counters = 1;
+ event_id[0] = 0;
+ }
+
+ for (counter = 0; counter < nr_counters; counter++) {
+ if (event_count[counter])
+ continue;
+
+ event_count[counter] = default_interval;
+ }
+
+ nr_cpus = sysconf(_SC_NPROCESSORS_ONLN);
+ assert(nr_cpus <= MAX_NR_CPUS);
+ assert(nr_cpus >= 0);
+
+ if (target_pid != -1 || profile_cpu != -1)
+ nr_cpus = 1;
+
+ parse_symbols();
+
+ return __cmd_top();
+}