static int show_pxstat_by_cpuid(xc_interface *xc_handle, int cpuid)
{
int ret = 0;
struct xc_px_stat pxstatinfo;
ret = get_pxstat_by_cpuid(xc_handle, cpuid, &pxstatinfo);
if ( ret )
return ret;
print_pxstat(cpuid, &pxstatinfo);
free(pxstatinfo.trans_pt);
free(pxstatinfo.pt);
return 0;
}
void start_gather_func(int argc, char *argv[])
{
int i;
struct timeval tv;
int timeout = 0;
if ( argc == 1 )
{
sscanf(argv[0], "%d", &timeout);
if ( timeout <= 0 )
fprintf(stderr, "failed to set timeout seconds, falling back...\n");
else
printf("Timeout set to %d seconds\n", timeout);
}
if ( gettimeofday(&tv, NULL) == -1 )
{
fprintf(stderr, "failed to get timeofday\n");
return ;
}
usec_start = tv.tv_sec * 1000000UL + tv.tv_usec;
sum = calloc(2 * max_cpu_nr, sizeof(*sum));
if ( sum == NULL )
return ;
cxstat = calloc(2 * max_cpu_nr, sizeof(*cxstat));
if ( cxstat == NULL )
{
free(sum);
return ;
}
pxstat = calloc(2 * max_cpu_nr, sizeof(*pxstat));
if ( pxstat == NULL )
{
free(sum);
free(cxstat);
return ;
}
avgfreq = calloc(max_cpu_nr, sizeof(*avgfreq));
if ( avgfreq == NULL )
{
free(sum);
free(cxstat);
free(pxstat);
return ;
}
sum_cx = sum;
sum_px = sum + max_cpu_nr;
cxstat_start = cxstat;
cxstat_end = cxstat + max_cpu_nr;
pxstat_start = pxstat;
pxstat_end = pxstat + max_cpu_nr;
if ( get_cxstat_by_cpuid(xc_handle, 0, NULL) == -ENODEV &&
get_pxstat_by_cpuid(xc_handle, 0, NULL) == -ENODEV )
{
fprintf(stderr, "Xen cpu idle and frequency is disabled!\n");
return ;
}
for ( i = 0; i < max_cpu_nr; i++ )
{
get_cxstat_by_cpuid(xc_handle, i, &cxstat_start[i]);
get_pxstat_by_cpuid(xc_handle, i, &pxstat_start[i]);
get_avgfreq_by_cpuid(xc_handle, i, &avgfreq[i]);
}
if (signal(SIGINT, signal_int_handler) == SIG_ERR)
{
fprintf(stderr, "failed to set signal int handler\n");
free(sum);
free(pxstat);
free(cxstat);
free(avgfreq);
return ;
}
if ( timeout > 0 )
{
if ( signal(SIGALRM, signal_int_handler) == SIG_ERR )
{
fprintf(stderr, "failed to set signal alarm handler\n");
free(sum);
free(pxstat);
free(cxstat);
free(avgfreq);
return ;
}
alarm(timeout);
}
printf("Start sampling, waiting for CTRL-C or SIGINT or SIGALARM signal ...\n");
pause();
}
static void signal_int_handler(int signo)
{
int i, j, k;
struct timeval tv;
int cx_cap = 0, px_cap = 0;
DECLARE_HYPERCALL_BUFFER(uint32_t, cpu_to_core);
DECLARE_HYPERCALL_BUFFER(uint32_t, cpu_to_socket);
DECLARE_HYPERCALL_BUFFER(uint32_t, cpu_to_node);
xc_topologyinfo_t info = { 0 };
cpu_to_core = xc_hypercall_buffer_alloc(xc_handle, cpu_to_core, sizeof(*cpu_to_core) * MAX_NR_CPU);
cpu_to_socket = xc_hypercall_buffer_alloc(xc_handle, cpu_to_socket, sizeof(*cpu_to_socket) * MAX_NR_CPU);
cpu_to_node = xc_hypercall_buffer_alloc(xc_handle, cpu_to_node, sizeof(*cpu_to_node) * MAX_NR_CPU);
if ( cpu_to_core == NULL || cpu_to_socket == NULL || cpu_to_node == NULL )
{
fprintf(stderr, "failed to allocate hypercall buffers\n");
goto out;
}
if ( gettimeofday(&tv, NULL) == -1 )
{
fprintf(stderr, "failed to get timeofday\n");
goto out ;
}
usec_end = tv.tv_sec * 1000000UL + tv.tv_usec;
if ( get_cxstat_by_cpuid(xc_handle, 0, NULL) != -ENODEV )
{
cx_cap = 1;
for ( i = 0; i < max_cpu_nr; i++ )
if ( !get_cxstat_by_cpuid(xc_handle, i, &cxstat_end[i]) )
for ( j = 0; j < cxstat_end[i].nr; j++ )
{
int64_t diff = (int64_t)cxstat_end[i].residencies[j] -
(int64_t)cxstat_start[i].residencies[j];
if ( diff >=0 )
sum_cx[i] += diff;
}
}
if ( get_pxstat_by_cpuid(xc_handle, 0, NULL) != -ENODEV )
{
px_cap = 1;
for ( i = 0; i < max_cpu_nr; i++ )
if ( !get_pxstat_by_cpuid(xc_handle, i , &pxstat_end[i]) )
for ( j = 0; j < pxstat_end[i].total; j++ )
sum_px[i] += pxstat_end[i].pt[j].residency -
pxstat_start[i].pt[j].residency;
}
for ( i = 0; i < max_cpu_nr; i++ )
get_avgfreq_by_cpuid(xc_handle, i, &avgfreq[i]);
printf("Elapsed time (ms): %"PRIu64"\n", (usec_end - usec_start) / 1000UL);
for ( i = 0; i < max_cpu_nr; i++ )
{
uint64_t res, triggers;
double avg_res;
printf("\nCPU%d:\tResidency(ms)\t\tAvg Res(ms)\n",i);
if ( cx_cap && sum_cx[i] > 0 )
{
for ( j = 0; j < cxstat_end[i].nr; j++ )
{
int64_t diff = (int64_t)cxstat_end[i].residencies[j] -
(int64_t)cxstat_start[i].residencies[j];
res = ( diff >= 0 ) ? diff : 0;
triggers = cxstat_end[i].triggers[j] -
cxstat_start[i].triggers[j];
/*
* triggers may be zero if the CPU has been in this state for
* the whole sample or if it never entered the state
*/
if ( triggers == 0 && cxstat_end[i].last == j )
avg_res = (double)sum_cx[i]/1000000.0;
else
avg_res = (triggers==0) ? 0: (double)res/triggers/1000000.0;
printf(" C%d\t%"PRIu64"\t(%5.2f%%)\t%.2f\n", j, res/1000000UL,
100 * res / (double)sum_cx[i], avg_res );
}
printf("\n");
}
if ( px_cap && sum_px[i]>0 )
{
for ( j = 0; j < pxstat_end[i].total; j++ )
{
res = pxstat_end[i].pt[j].residency -
pxstat_start[i].pt[j].residency;
printf(" P%d\t%"PRIu64"\t(%5.2f%%)\n", j,
res / 1000000UL, 100UL * res / (double)sum_px[i]);
}
}
if ( px_cap && avgfreq[i] )
printf(" Avg freq\t%d\tKHz\n", avgfreq[i]);
}
set_xen_guest_handle(info.cpu_to_core, cpu_to_core);
set_xen_guest_handle(info.cpu_to_socket, cpu_to_socket);
set_xen_guest_handle(info.cpu_to_node, cpu_to_node);
info.max_cpu_index = MAX_NR_CPU - 1;
if ( cx_cap && !xc_topologyinfo(xc_handle, &info) )
{
uint32_t socket_ids[MAX_NR_CPU];
uint32_t core_ids[MAX_NR_CPU];
uint32_t socket_nr = 0;
uint32_t core_nr = 0;
if ( info.max_cpu_index > MAX_NR_CPU - 1 )
info.max_cpu_index = MAX_NR_CPU - 1;
/* check validity */
for ( i = 0; i <= info.max_cpu_index; i++ )
{
if ( cpu_to_core[i] == INVALID_TOPOLOGY_ID ||
cpu_to_socket[i] == INVALID_TOPOLOGY_ID )
break;
}
if ( i > info.max_cpu_index )
{
/* find socket nr & core nr per socket */
for ( i = 0; i <= info.max_cpu_index; i++ )
{
for ( j = 0; j < socket_nr; j++ )
if ( cpu_to_socket[i] == socket_ids[j] )
break;
if ( j == socket_nr )
{
socket_ids[j] = cpu_to_socket[i];
socket_nr++;
}
for ( j = 0; j < core_nr; j++ )
if ( cpu_to_core[i] == core_ids[j] )
break;
if ( j == core_nr )
{
core_ids[j] = cpu_to_core[i];
core_nr++;
}
}
/* print out CC? and PC? */
for ( i = 0; i < socket_nr; i++ )
{
uint64_t res;
for ( j = 0; j <= info.max_cpu_index; j++ )
{
if ( cpu_to_socket[j] == socket_ids[i] )
break;
}
printf("\nSocket %d\n", socket_ids[i]);
res = cxstat_end[j].pc2 - cxstat_start[j].pc2;
printf("\tPC2\t%"PRIu64" ms\t%.2f%%\n", res / 1000000UL,
100UL * res / (double)sum_cx[j]);
res = cxstat_end[j].pc3 - cxstat_start[j].pc3;
printf("\tPC3\t%"PRIu64" ms\t%.2f%%\n", res / 1000000UL,
100UL * res / (double)sum_cx[j]);
res = cxstat_end[j].pc6 - cxstat_start[j].pc6;
printf("\tPC6\t%"PRIu64" ms\t%.2f%%\n", res / 1000000UL,
100UL * res / (double)sum_cx[j]);
res = cxstat_end[j].pc7 - cxstat_start[j].pc7;
printf("\tPC7\t%"PRIu64" ms\t%.2f%%\n", res / 1000000UL,
100UL * res / (double)sum_cx[j]);
for ( k = 0; k < core_nr; k++ )
{
for ( j = 0; j <= info.max_cpu_index; j++ )
{
if ( cpu_to_socket[j] == socket_ids[i] &&
cpu_to_core[j] == core_ids[k] )
break;
}
printf("\t Core %d CPU %d\n", core_ids[k], j);
res = cxstat_end[j].cc3 - cxstat_start[j].cc3;
printf("\t\tCC3\t%"PRIu64" ms\t%.2f%%\n", res / 1000000UL,
100UL * res / (double)sum_cx[j]);
res = cxstat_end[j].cc6 - cxstat_start[j].cc6;
printf("\t\tCC6\t%"PRIu64" ms\t%.2f%%\n", res / 1000000UL,
100UL * res / (double)sum_cx[j]);
res = cxstat_end[j].cc7 - cxstat_start[j].cc7;
printf("\t\tCC7\t%"PRIu64" ms\t%.2f%%\n", res / 1000000UL,
100UL * res / (double)sum_cx[j]);
}
}
}
}
/* some clean up and then exits */
for ( i = 0; i < 2 * max_cpu_nr; i++ )
{
free(cxstat[i].triggers);
free(cxstat[i].residencies);
free(pxstat[i].trans_pt);
free(pxstat[i].pt);
}
free(cxstat);
free(pxstat);
free(sum);
free(avgfreq);
out:
xc_hypercall_buffer_free(xc_handle, cpu_to_core);
xc_hypercall_buffer_free(xc_handle, cpu_to_socket);
xc_hypercall_buffer_free(xc_handle, cpu_to_node);
xc_interface_close(xc_handle);
exit(0);
}
static void signal_int_handler(int signo)
{
int i, j, k;
struct timeval tv;
int cx_cap = 0, px_cap = 0;
xc_cputopo_t *cputopo = NULL;
unsigned max_cpus = 0;
if ( xc_cputopoinfo(xc_handle, &max_cpus, NULL) != 0 )
{
fprintf(stderr, "failed to discover number of CPUs: %s\n",
strerror(errno));
goto out;
}
cputopo = calloc(max_cpus, sizeof(*cputopo));
if ( cputopo == NULL )
{
fprintf(stderr, "failed to allocate hypercall buffers\n");
goto out;
}
if ( gettimeofday(&tv, NULL) )
{
fprintf(stderr, "failed to get timeofday\n");
goto out;
}
usec_end = tv.tv_sec * 1000000UL + tv.tv_usec;
if ( get_cxstat_by_cpuid(xc_handle, 0, NULL) != -ENODEV )
{
cx_cap = 1;
for ( i = 0; i < max_cpu_nr; i++ )
if ( !get_cxstat_by_cpuid(xc_handle, i, &cxstat_end[i]) )
for ( j = 0; j < cxstat_end[i].nr; j++ )
{
int64_t diff = (int64_t)cxstat_end[i].residencies[j] -
(int64_t)cxstat_start[i].residencies[j];
if ( diff >=0 )
sum_cx[i] += diff;
}
}
if ( get_pxstat_by_cpuid(xc_handle, 0, NULL) != -ENODEV )
{
px_cap = 1;
for ( i = 0; i < max_cpu_nr; i++ )
if ( !get_pxstat_by_cpuid(xc_handle, i , &pxstat_end[i]) )
for ( j = 0; j < pxstat_end[i].total; j++ )
sum_px[i] += pxstat_end[i].pt[j].residency -
pxstat_start[i].pt[j].residency;
}
for ( i = 0; i < max_cpu_nr; i++ )
get_avgfreq_by_cpuid(xc_handle, i, &avgfreq[i]);
printf("Elapsed time (ms): %"PRIu64"\n", (usec_end - usec_start) / 1000UL);
for ( i = 0; i < max_cpu_nr; i++ )
{
uint64_t res, triggers;
double avg_res;
printf("\nCPU%d:\tResidency(ms)\t\tAvg Res(ms)\n",i);
if ( cx_cap && sum_cx[i] > 0 )
{
for ( j = 0; j < cxstat_end[i].nr; j++ )
{
int64_t diff = (int64_t)cxstat_end[i].residencies[j] -
(int64_t)cxstat_start[i].residencies[j];
res = ( diff >= 0 ) ? diff : 0;
triggers = cxstat_end[i].triggers[j] -
cxstat_start[i].triggers[j];
/*
* triggers may be zero if the CPU has been in this state for
* the whole sample or if it never entered the state
*/
if ( triggers == 0 && cxstat_end[i].last == j )
avg_res = (double)sum_cx[i]/1000000.0;
else
avg_res = (triggers==0) ? 0: (double)res/triggers/1000000.0;
printf(" C%d\t%"PRIu64"\t(%5.2f%%)\t%.2f\n", j, res/1000000UL,
100 * res / (double)sum_cx[i], avg_res );
}
printf("\n");
}
if ( px_cap && sum_px[i]>0 )
{
for ( j = 0; j < pxstat_end[i].total; j++ )
{
res = pxstat_end[i].pt[j].residency -
pxstat_start[i].pt[j].residency;
printf(" P%d\t%"PRIu64"\t(%5.2f%%)\n", j,
res / 1000000UL, 100UL * res / (double)sum_px[i]);
}
}
if ( px_cap && avgfreq[i] )
printf(" Avg freq\t%d\tKHz\n", avgfreq[i]);
}
if ( cx_cap && !xc_cputopoinfo(xc_handle, &max_cpus, cputopo) )
{
uint32_t socket_ids[MAX_NR_CPU];
uint32_t core_ids[MAX_NR_CPU];
uint32_t socket_nr = 0;
uint32_t core_nr = 0;
if ( max_cpus > MAX_NR_CPU )
max_cpus = MAX_NR_CPU;
/* check validity */
for ( i = 0; i < max_cpus; i++ )
{
if ( cputopo[i].core == XEN_INVALID_CORE_ID ||
cputopo[i].socket == XEN_INVALID_SOCKET_ID )
break;
}
if ( i >= max_cpus )
{
/* find socket nr & core nr per socket */
for ( i = 0; i < max_cpus; i++ )
{
for ( j = 0; j < socket_nr; j++ )
if ( cputopo[i].socket == socket_ids[j] )
break;
if ( j == socket_nr )
{
socket_ids[j] = cputopo[i].socket;
socket_nr++;
}
for ( j = 0; j < core_nr; j++ )
if ( cputopo[i].core == core_ids[j] )
break;
if ( j == core_nr )
{
core_ids[j] = cputopo[i].core;
core_nr++;
}
}
/* print out CC? and PC? */
for ( i = 0; i < socket_nr; i++ )
{
unsigned int n;
uint64_t res;
for ( j = 0; j < max_cpus; j++ )
{
if ( cputopo[j].socket == socket_ids[i] )
break;
}
printf("\nSocket %d\n", socket_ids[i]);
for ( n = 0; n < MAX_PKG_RESIDENCIES; ++n )
{
if ( n >= cxstat_end[j].nr_pc )
continue;
res = cxstat_end[j].pc[n];
if ( n < cxstat_start[j].nr_pc )
res -= cxstat_start[j].pc[n];
printf("\tPC%u\t%"PRIu64" ms\t%.2f%%\n",
n + 1, res / 1000000UL,
100UL * res / (double)sum_cx[j]);
}
for ( k = 0; k < core_nr; k++ )
{
for ( j = 0; j < max_cpus; j++ )
{
if ( cputopo[j].socket == socket_ids[i] &&
cputopo[j].core == core_ids[k] )
break;
}
printf("\t Core %d CPU %d\n", core_ids[k], j);
for ( n = 0; n < MAX_CORE_RESIDENCIES; ++n )
{
if ( n >= cxstat_end[j].nr_cc )
continue;
res = cxstat_end[j].cc[n];
if ( n < cxstat_start[j].nr_cc )
res -= cxstat_start[j].cc[n];
printf("\t\tCC%u\t%"PRIu64" ms\t%.2f%%\n",
n + 1, res / 1000000UL,
100UL * res / (double)sum_cx[j]);
}
}
}
}
}
/* some clean up and then exits */
for ( i = 0; i < 2 * max_cpu_nr; i++ )
{
free(cxstat[i].triggers);
free(cxstat[i].residencies);
free(cxstat[i].pc);
free(cxstat[i].cc);
free(pxstat[i].trans_pt);
free(pxstat[i].pt);
}
free(cxstat);
free(pxstat);
free(sum);
free(avgfreq);
out:
free(cputopo);
xc_interface_close(xc_handle);
exit(0);
}
