/*
- * kerneltop.c: show top kernel functions - performance counters showcase
-
- Build with:
-
- make -C Documentation/perf_counter/
-
- Sample output:
-
-------------------------------------------------------------------------------
- KernelTop: 2669 irqs/sec [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)
*/
-
- /*
- * 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"
#include "perf.h"
-#include "builtin.h"
+
#include "util/symbol.h"
+#include "util/color.h"
#include "util/util.h"
#include "util/rbtree.h"
#include "util/parse-options.h"
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 target_pid = -1;
+static int target_pid = -1;
static int profile_cpu = -1;
static int nr_cpus = 0;
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 *sym_filter;
struct rb_node rb_node;
struct list_head node;
unsigned long count[MAX_COUNTERS];
+ unsigned long snap_count;
+ double weight;
int skip;
};
* 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];
static long userspace_events;
static const char CONSOLE_CLEAR[] = "\e[H\e[2J";
-static void list_insert_active_sym(struct sym_entry *syme)
+static void __list_insert_active_sym(struct sym_entry *syme)
{
list_add(&syme->node, &active_symbols);
}
+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 void rb_insert_active_sym(struct rb_root *tree, struct sym_entry *se)
{
struct rb_node **p = &tree->rb_node;
parent = *p;
iter = rb_entry(parent, struct sym_entry, rb_node);
- if (sym_weight(se) > sym_weight(iter))
+ if (se->weight > iter->weight)
p = &(*p)->rb_left;
else
p = &(*p)->rb_right;
static void print_sym_table(void)
{
- int printed, j;
+ int printed = 0, j;
int counter;
float events_per_sec = events/delay_secs;
float kevents_per_sec = (events-userspace_events)/delay_secs;
events = userspace_events = 0;
/* Sort the active symbols */
- list_for_each_entry_safe(syme, n, &active_symbols, node) {
- if (syme->count[0] != 0) {
+ 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->count[0];
+ sum_kevents += syme->snap_count;
for (j = 0; j < nr_counters; j++)
syme->count[j] = zero ? 0 : syme->count[j] * 7 / 8;
} else
- list_del_init(&syme->node);
+ list_remove_active_sym(syme);
}
write(1, CONSOLE_CLEAR, strlen(CONSOLE_CLEAR));
printf(
"------------------------------------------------------------------------------\n");
- printf( " KernelTop:%8.0f irqs/sec kernel:%4.1f%% [",
+ 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)));
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);
- float pcnt;
+ char *color = PERF_COLOR_NORMAL;
+ double pcnt;
- if (++printed > 18 || syme->count[0] < count_filter)
- break;
+ if (++printed > print_entries || syme->snap_count < count_filter)
+ continue;
- pcnt = 100.0 - (100.0 * ((sum_kevents - syme->count[0]) /
+ pcnt = 100.0 - (100.0 * ((sum_kevents - syme->snap_count) /
sum_kevents));
+ /*
+ * We color high-overhead entries in red, low-overhead
+ * entries in green - and keep the middle ground normal:
+ */
+ if (pcnt >= 5.0)
+ color = PERF_COLOR_RED;
+ if (pcnt < 0.5)
+ color = PERF_COLOR_GREEN;
+
if (nr_counters == 1)
- printf("%19.2f - %4.1f%% - %016llx : %s\n",
- sym_weight(syme),
- pcnt, sym->start, sym->name);
+ printf("%19.2f - ", syme->weight);
else
- printf("%8.1f %10ld - %4.1f%% - %016llx : %s\n",
- sym_weight(syme), syme->count[0],
- pcnt, sym->start, sym->name);
+ printf("%8.1f %10ld - ", syme->weight, syme->snap_count);
+
+ color_fprintf(stdout, color, "%4.1f%%", pcnt);
+ printf(" - %016llx : %s\n", sym->start, sym->name);
}
{
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();
if (kernel_dso == NULL)
return -1;
- if (dso__load_kernel(kernel_dso, NULL, symbol_filter) != 0)
+ if (dso__load_kernel(kernel_dso, NULL, symbol_filter, 1) != 0)
goto out_delete_dso;
node = rb_first(&kernel_dso->syms);
node = rb_last(&kernel_dso->syms);
sym = rb_entry(node, struct symbol, rb_node);
- max_ip = sym->start;
+ max_ip = sym->end;
if (dump_symtab)
- dso__fprintf(kernel_dso, stdout);
+ dso__fprintf(kernel_dso, stderr);
return 0;
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);
+ __list_insert_active_sym(syme);
+ pthread_mutex_unlock(&active_symbols_lock);
return;
}
}
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;
- }
}
}
static int __cmd_top(void)
{
- 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;
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 = 1;
- 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, target_pid, 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));
}
}
"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,
+ 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"),
"only display symbols matchig this pattern"),
OPT_BOOLEAN('z', "zero", &group,
"zero history across updates"),
- OPT_BOOLEAN('M', "use-mmap", &use_mmap,
- "track mmap events"),
- OPT_BOOLEAN('U', "use-munmap", &use_munmap,
- "track munmap events"),
- OPT_INTEGER('F', "--freq", &freq,
+ OPT_INTEGER('F', "freq", &freq,
"profile at this frequency"),
+ OPT_INTEGER('E', "entries", &print_entries,
+ "display this many functions"),
OPT_END()
};