368 #define stackshot_ctx (stackshot_contexts[cur_stackshot_ctx_idx])  macro
369 #define stackshot_args (stackshot_ctx.sc_args)
602 		os_atomic_dec(&stackshot_ctx.sc_cpus_working, acquire);  in stackshot_panic_spin()
604 	if (stackshot_ctx.sc_calling_cpuid == cpu_number()) {  in stackshot_panic_spin()
605 		while (os_atomic_load(&stackshot_ctx.sc_cpus_working, acquire) != 0) {  in stackshot_panic_spin()
621 	if (__improbable(os_atomic_load(&stackshot_ctx.sc_state, relaxed) == SS_PANICKED)) {  in stackshot_panic_guard()
642 		o_state = os_atomic_xchg(&stackshot_ctx.sc_state, SS_PANICKED, seq_cst);  in stackshot_cpu_signal_panic()
650 			os_atomic_dec(&stackshot_ctx.sc_cpus_working, acquire);  in stackshot_cpu_signal_panic()
652 		while (os_atomic_load(&stackshot_ctx.sc_cpus_working, seq_cst) != 0) {  in stackshot_cpu_signal_panic()
667 …if (__improbable(!os_atomic_cmpxchgv(&stackshot_ctx.sc_state, SS_RUNNING, err_state, &cur_state, s…  in stackshot_set_error()
669 			os_atomic_cmpxchg(&stackshot_ctx.sc_state, SS_SETUP, err_state, seq_cst);  in stackshot_set_error()
684 	stackshot_state_t state = os_atomic_load(&stackshot_ctx.sc_state, relaxed);  in stackshot_status_check()
799 	assert(!stackshot_ctx.sc_is_singlethreaded);  in stackshot_buffer_alloc()
838 	ret = stackshot_buffer_alloc(size, &stackshot_ctx.sc_buffers[my_cluster], &err);  in stackshot_best_buffer_alloc()
842 		for (size_t buf_idx = 0; buf_idx < stackshot_ctx.sc_num_buffers; buf_idx++) {  in stackshot_best_buffer_alloc()
847 			ret = stackshot_buffer_alloc(size, &stackshot_ctx.sc_buffers[buf_idx], &err);  in stackshot_best_buffer_alloc()
863 	return stackshot_buffer_alloc(size, &stackshot_ctx.sc_buffers[0], error);  in stackshot_best_buffer_alloc()
879 	assert(!stackshot_ctx.sc_is_singlethreaded);  in stackshot_buffer_free()
915 	if (stackshot_ctx.sc_is_singlethreaded) {  in stackshot_alloc_with_size()
948 	size_t             buf_size = MAX(min_size, stackshot_ctx.sc_min_kcdata_size);  in linked_kcdata_init()
985 …stackshot_buffer_free((void*) descriptor->kcd_addr_end, &stackshot_ctx.sc_buffers[buffer], free_si…  in stackshot_kcdata_free_unused()
1062 	assert(!stackshot_ctx.sc_panic_stackshot);  in stackshot_new_linked_kcdata()
1088 	assert(!stackshot_ctx.sc_panic_stackshot);  in stackshot_finalize_linked_kcdata()
1278 	os_atomic_store(&stackshot_ctx.sc_state, SS_INACTIVE, release);  in stack_snapshot_from_kernel()
1284 	if (error == KERN_SUCCESS && !stackshot_ctx.sc_is_singlethreaded) {  in stack_snapshot_from_kernel()
1295 		if (!stackshot_ctx.sc_is_singlethreaded) {  in stack_snapshot_from_kernel()
1402 	queue = &stackshot_ctx.sc_workqueues[queue_idx];  in stackshot_put_workitem()
1614 		for (size_t buf_idx = 0; buf_idx < stackshot_ctx.sc_num_buffers; buf_idx++) {  in stackshot_push_duration_and_latency()
1615 			struct stackshot_buffer *buf = &stackshot_ctx.sc_buffers[buf_idx];  in stackshot_push_duration_and_latency()
1619 		stackshot_ctx.sc_latency.buffer_size = stackshot_ctx.sc_args.buffer_size;  in stackshot_push_duration_and_latency()
1620 		stackshot_ctx.sc_latency.buffer_overhead = buffer_overhead;  in stackshot_push_duration_and_latency()
1621 		stackshot_ctx.sc_latency.buffer_used = buffer_used;  in stackshot_push_duration_and_latency()
1622 		stackshot_ctx.sc_latency.buffer_count = stackshot_ctx.sc_num_buffers;  in stackshot_push_duration_and_latency()
1627 			    sizeof(stackshot_ctx.sc_latency), &stackshot_ctx.sc_latency));  in stackshot_push_duration_and_latency()
1632 		assert(!stackshot_ctx.sc_panic_stackshot);  in stackshot_push_duration_and_latency()
1636 		memcpy(duration_p, &stackshot_ctx.sc_duration, sizeof(*duration_p));  in stackshot_push_duration_and_latency()
1640 …ta(kcdata, STACKSHOT_KCTYPE_STACKSHOT_DURATION, sizeof(stackshot_ctx.sc_duration), &stackshot_ctx.…  in stackshot_push_duration_and_latency()
1662 	if (stackshot_ctx.sc_is_singlethreaded) {  in stackshot_alloc_final_kcdata()
1672 	stackshot_ctx.sc_finalized_kcdata = kcdata_memory_alloc_init(final_kcdata_buffer, hdr_tag,  in stackshot_alloc_final_kcdata()
1675 	if (stackshot_ctx.sc_finalized_kcdata == NULL) {  in stackshot_alloc_final_kcdata()
1689 	if (stackshot_ctx.sc_is_singlethreaded || (stackshot_ctx.sc_finalized_kcdata == NULL)) {  in stackshot_free_final_kcdata()
1693 …kmem_free(kernel_map, stackshot_ctx.sc_finalized_kcdata->kcd_addr_begin, stackshot_args.buffer_siz…  in stackshot_free_final_kcdata()
1694 	kcdata_memory_destroy(stackshot_ctx.sc_finalized_kcdata);  in stackshot_free_final_kcdata()
1695 	stackshot_ctx.sc_finalized_kcdata = NULL;  in stackshot_free_final_kcdata()
1714 	assert(!stackshot_ctx.sc_panic_stackshot && !stackshot_ctx.sc_is_singlethreaded);  in stackshot_collect_kcdata()
1719 	assert(stackshot_ctx.sc_finalized_kcdata != NULL);  in stackshot_collect_kcdata()
1725 …kcd_exit_on_error(kcdata_init_compress(stackshot_ctx.sc_finalized_kcdata, hdr_tag, kdp_memcpy, KCD…  in stackshot_collect_kcdata()
1729 …kcd_exit_on_error(stackshot_copy_linked_kcdata(stackshot_ctx.sc_finalized_kcdata, stackshot_ctx.sc…  in stackshot_collect_kcdata()
1733 		os_atomic_store(&stackshot_ctx.sc_workqueues[i].sswq_cur_item, 0, relaxed);  in stackshot_collect_kcdata()
1743 			struct stackshot_workqueue *queue = &stackshot_ctx.sc_workqueues[i];  in stackshot_collect_kcdata()
1768 …kcd_exit_on_error(stackshot_copy_linked_kcdata(stackshot_ctx.sc_finalized_kcdata, next_item->sswi_…  in stackshot_collect_kcdata()
1772 …kcd_exit_on_error(stackshot_copy_linked_kcdata(stackshot_ctx.sc_finalized_kcdata, stackshot_ctx.sc…  in stackshot_collect_kcdata()
1795 	assert(!stackshot_ctx.sc_panic_stackshot && !stackshot_ctx.sc_is_singlethreaded);  in stackshot_finalize_kcdata()
1798 	assert(stackshot_ctx.sc_finalized_kcdata != NULL);  in stackshot_finalize_kcdata()
1801 	kcd_exit_on_error(stackshot_push_duration_and_latency(stackshot_ctx.sc_finalized_kcdata));  in stackshot_finalize_kcdata()
1804 	kcd_finalize_compression(stackshot_ctx.sc_finalized_kcdata);  in stackshot_finalize_kcdata()
1805 …kcd_exit_on_error(kcdata_add_uint64_with_description(stackshot_ctx.sc_finalized_kcdata, stackshot_…  in stackshot_finalize_kcdata()
1806 	kcd_exit_on_error(kcdata_write_buffer_end(stackshot_ctx.sc_finalized_kcdata));  in stackshot_finalize_kcdata()
1808 …stackshot_ctx.sc_bytes_traced = (uint32_t) kcdata_memory_get_used_bytes(stackshot_ctx.sc_finalized…  in stackshot_finalize_kcdata()
1809 …stackshot_ctx.sc_bytes_uncompressed = (uint32_t) kcdata_memory_get_uncompressed_bytes(stackshot_ct…  in stackshot_finalize_kcdata()
1811 	if (os_atomic_load(&stackshot_ctx.sc_retval, relaxed) == KERN_SUCCESS) {  in stackshot_finalize_kcdata()
1813 		kcd_exit_on_error(kcdata_finish(stackshot_ctx.sc_finalized_kcdata));  in stackshot_finalize_kcdata()
1816 …memcpy(stackshot_args.buffer, (void*) stackshot_ctx.sc_finalized_kcdata->kcd_addr_begin, stackshot…  in stackshot_finalize_kcdata()
1820 …ackshot_args.buffer + ((mach_vm_address_t) stackshot_duration_outer - stackshot_ctx.sc_finalized_k…  in stackshot_finalize_kcdata()
1839 	assert(stackshot_ctx.sc_is_singlethreaded);  in stackshot_finalize_singlethreaded_kcdata()
1841 	kcd_exit_on_error(stackshot_push_duration_and_latency(stackshot_ctx.sc_finalized_kcdata));  in stackshot_finalize_singlethreaded_kcdata()
1843 	kcd_finalize_compression(stackshot_ctx.sc_finalized_kcdata);  in stackshot_finalize_singlethreaded_kcdata()
1844 …kcd_exit_on_error(kcdata_add_uint64_with_description(stackshot_ctx.sc_finalized_kcdata, stackshot_…  in stackshot_finalize_singlethreaded_kcdata()
1845 	kcd_exit_on_error(kcdata_write_buffer_end(stackshot_ctx.sc_finalized_kcdata));  in stackshot_finalize_singlethreaded_kcdata()
1847 …stackshot_ctx.sc_bytes_traced = (uint32_t) kcdata_memory_get_used_bytes(stackshot_ctx.sc_finalized…  in stackshot_finalize_singlethreaded_kcdata()
1848 …stackshot_ctx.sc_bytes_uncompressed = (uint32_t) kcdata_memory_get_uncompressed_bytes(stackshot_ct…  in stackshot_finalize_singlethreaded_kcdata()
1850 	kcd_exit_on_error(kcdata_finish(stackshot_ctx.sc_finalized_kcdata));  in stackshot_finalize_singlethreaded_kcdata()
1852 	if (stackshot_ctx.sc_panic_stackshot) {  in stackshot_finalize_singlethreaded_kcdata()
1853 		*stackshot_args.descriptor = *stackshot_ctx.sc_finalized_kcdata;  in stackshot_finalize_singlethreaded_kcdata()
1969 	kcdata_add_container_marker(stackshot_ctx.sc_finalized_kcdata, KCDATA_TYPE_CONTAINER_BEGIN,  in collect_exclave_threads()
1985 	kr = kcdata_add_container_marker(stackshot_ctx.sc_finalized_kcdata, KCDATA_TYPE_CONTAINER_END,  in collect_exclave_threads()
2251 …kr = stackshot_exclaves_process_stackshot(result, stackshot_ctx.sc_finalized_kcdata, want_raw_addr…  in stackshot_exclaves_process_result()
2529 		os_atomic_store(&stackshot_ctx.sc_state, SS_INACTIVE, release);  in kern_stack_snapshot_internal()
2539 		if (error == KERN_SUCCESS && !stackshot_ctx.sc_is_singlethreaded) {  in kern_stack_snapshot_internal()
2553 			if (stackshot_ctx.sc_is_singlethreaded) {  in kern_stack_snapshot_internal()
2687 	stackshot_ctx = (struct stackshot_context) {  in kdp_snapshot_preflight_internal()
2703 	if (!stackshot_ctx.sc_panic_stackshot) {  in kdp_snapshot_preflight_internal()
2706 		stackshot_ctx.sc_num_buffers = is_singlethreaded ? 1 : ml_get_cluster_count();  in kdp_snapshot_preflight_internal()
2708 		stackshot_ctx.sc_num_buffers = 1;  in kdp_snapshot_preflight_internal()
2710 		size_t bufsz = args.buffer_size / stackshot_ctx.sc_num_buffers;  in kdp_snapshot_preflight_internal()
2711 		for (int buf_idx = 0; buf_idx < stackshot_ctx.sc_num_buffers; buf_idx++) {  in kdp_snapshot_preflight_internal()
2712 			stackshot_ctx.sc_buffers[buf_idx] = (struct stackshot_buffer) {  in kdp_snapshot_preflight_internal()
2791 	stackshot_ctx = (struct stackshot_context) { 0 };  in stackshot_reset_state()
2803 	return os_atomic_load(&stackshot_ctx.sc_state, relaxed) != SS_INACTIVE;  in stackshot_active()
2815 	return stackshot_ctx.sc_bytes_traced;  in kdp_stack_snapshot_bytes_traced()
2821 	return stackshot_ctx.sc_bytes_uncompressed;  in kdp_stack_snapshot_bytes_uncompressed()
2882 	struct port_label_hash *plh = &stackshot_ctx.sc_plh;  in stackshot_plh_est_size()
2899 	stackshot_ctx.sc_plh = (struct port_label_hash){.plh_size = 0};  /* structure assignment */  in stackshot_plh_reset()
2947 		PLH_STAT_OP(os_atomic_inc(&stackshot_ctx.sc_plh.plh_bad, relaxed));  in stackshot_plh_setup()
2957 	stackshot_ctx.sc_plh = plh;  /* structure assignment */  in stackshot_plh_setup()
2987 	uint16_t plh_size = stackshot_ctx.sc_plh.plh_size;  in stackshot_plh_resetgen()
3006 	struct port_label_hash *plh = &stackshot_ctx.sc_plh;  in stackshot_plh_lookup_locked()
3076 	struct port_label_hash *plh = &stackshot_ctx.sc_plh;  in kdp_stackshot_plh_record_locked()
3132 	plh_lock(&stackshot_ctx.sc_plh);  in kdp_stackshot_plh_record()
3134 	plh_unlock(&stackshot_ctx.sc_plh);  in kdp_stackshot_plh_record()
3142 	plh_lock(&stackshot_ctx.sc_plh);  in stackshot_plh_lookup()
3144 	plh_unlock(&stackshot_ctx.sc_plh);  in stackshot_plh_lookup()
3153 	struct port_label_hash *plh = &stackshot_ctx.sc_plh;  in kdp_stackshot_plh_stats()
3898 		assert(stackshot_ctx.sc_is_singlethreaded);  in kcdata_record_task_snapshot()
4233 …laves_state & TH_EXCLAVES_RPC) && stackshot_exclave_inspect_ctids && !stackshot_ctx.sc_panic_stack…  in kcdata_record_thread_snapshot()
4331 			.sbc_allow_faulting = stackshot_ctx.sc_enable_faulting,  in kcdata_record_thread_snapshot()
4611 	if (!stackshot_ctx.sc_panic_stackshot) {  in kdp_stackshot_record_task()
4915 		if (!stackshot_ctx.sc_panic_stackshot) {  in kdp_stackshot_record_task()
5048 	if (!stackshot_ctx.sc_panic_stackshot) {  in kdp_stackshot_kcdata_format()
5062 	stackshot_ctx.sc_enable_faulting = (stackshot_flags & (STACKSHOT_ENABLE_BT_FAULTING));  in kdp_stackshot_kcdata_format()
5075 	if (!stackshot_ctx.sc_panic_stackshot && collect_exclaves) {  in kdp_stackshot_kcdata_format()
5086 	if (stackshot_ctx.sc_bytes_uncompressed) {  in kdp_stackshot_kcdata_format()
5087 		stackshot_ctx.sc_bytes_uncompressed = 0;  in kdp_stackshot_kcdata_format()
5096 	if (!stackshot_ctx.sc_is_singlethreaded) {  in kdp_stackshot_kcdata_format()
5098 		stackshot_ctx.sc_pretask_kcdata = stackshot_cpu_ctx.scc_kcdata_head;  in kdp_stackshot_kcdata_format()
5113 	stackshot_ctx.sc_latency.setup_latency_mt = mach_absolute_time();  in kdp_stackshot_kcdata_format()
5145 …stackshot_kcdata_p, KCDATA_TYPE_USECS_SINCE_EPOCH, sizeof(uint64_t), &stackshot_ctx.sc_microsecs));  in kdp_stackshot_kcdata_format()
5163 	if (!stackshot_ctx.sc_is_singlethreaded && (stackshot_flags & STACKSHOT_THREAD_GROUP)) {  in kdp_stackshot_kcdata_format()
5191 	if (!stackshot_ctx.sc_panic_stackshot && (thread_group_begin_cpu_cycle_count != 0)) {  in kdp_stackshot_kcdata_format()
5202 …stackshot_ctx.sc_latency.setup_latency_mt = mach_absolute_time() - stackshot_ctx.sc_latency.setup_…  in kdp_stackshot_kcdata_format()
5203 	if (stackshot_ctx.sc_is_singlethreaded) {  in kdp_stackshot_kcdata_format()
5204 		stackshot_ctx.sc_latency.total_task_iteration_latency_mt = mach_absolute_time();  in kdp_stackshot_kcdata_format()
5206 		stackshot_ctx.sc_latency.task_queue_building_latency_mt = mach_absolute_time();  in kdp_stackshot_kcdata_format()
5240 		if (stackshot_ctx.sc_is_singlethreaded) {  in kdp_stackshot_kcdata_format()
5257 	if (stackshot_ctx.sc_is_singlethreaded) {  in kdp_stackshot_kcdata_format()
5258 …stackshot_ctx.sc_latency.total_task_iteration_latency_mt = mach_absolute_time() - stackshot_ctx.sc…  in kdp_stackshot_kcdata_format()
5260 …stackshot_ctx.sc_latency.task_queue_building_latency_mt = mach_absolute_time() - stackshot_ctx.sc_…  in kdp_stackshot_kcdata_format()
5265 	if (!stackshot_ctx.sc_is_singlethreaded) {  in kdp_stackshot_kcdata_format()
5268 		stackshot_ctx.sc_posttask_kcdata = stackshot_cpu_ctx.scc_kcdata_head;  in kdp_stackshot_kcdata_format()
5280 		if (!stackshot_ctx.sc_panic_stackshot && (stackshot_flags & STACKSHOT_SAVE_JETSAM_COALITIONS)) {  in kdp_stackshot_kcdata_format()
5307 		if (!stackshot_ctx.sc_panic_stackshot && (coalition_begin_cpu_cycle_count != 0)) {  in kdp_stackshot_kcdata_format()
5320 	if (stackshot_ctx.sc_is_singlethreaded) {  in kdp_stackshot_kcdata_format()
5321 		stackshot_ctx.sc_latency.total_terminated_task_iteration_latency_mt = mach_absolute_time();  in kdp_stackshot_kcdata_format()
5323 		stackshot_ctx.sc_latency.terminated_task_queue_building_latency_mt = mach_absolute_time();  in kdp_stackshot_kcdata_format()
5346 		if (stackshot_ctx.sc_is_singlethreaded) {  in kdp_stackshot_kcdata_format()
5359 		os_atomic_store(&stackshot_ctx.sc_workqueues[i].sswq_populated, true, release);  in kdp_stackshot_kcdata_format()
5367 	if (stackshot_ctx.sc_is_singlethreaded) {  in kdp_stackshot_kcdata_format()
5368 …stackshot_ctx.sc_latency.total_terminated_task_iteration_latency_mt = mach_absolute_time() - stack…  in kdp_stackshot_kcdata_format()
5370 …stackshot_ctx.sc_latency.terminated_task_queue_building_latency_mt = mach_absolute_time() - stacks…  in kdp_stackshot_kcdata_format()
5376 		stackshot_ctx.sc_latency.latency_version = 2;  in kdp_stackshot_kcdata_format()
5377 		stackshot_ctx.sc_latency.main_cpu_number = stackshot_ctx.sc_main_cpuid;  in kdp_stackshot_kcdata_format()
5378 		stackshot_ctx.sc_latency.calling_cpu_number = stackshot_ctx.sc_calling_cpuid;  in kdp_stackshot_kcdata_format()
5383 	if (!stackshot_ctx.sc_panic_stackshot) {  in kdp_stackshot_kcdata_format()
5394 	if (!stackshot_ctx.sc_is_singlethreaded) {  in kdp_stackshot_kcdata_format()
5403 	if (stackshot_ctx.sc_panic_stackshot) {  in kdp_stackshot_kcdata_format()
5516 	if (!stackshot_ctx.sc_panic_stackshot) {  in stackshot_find_phys()
5529 …_fault_stats.sfs_time_spent_faulting >= stackshot_max_fault_time) && !stackshot_ctx.sc_panic_stack…  in stackshot_find_phys()
5591 	if (!stackshot_ctx.sc_is_singlethreaded) {  in do_stackshot()
5595 			stackshot_ctx.sc_workqueues[i] = (struct stackshot_workqueue) {  in do_stackshot()
5619 	if (stackshot_ctx.sc_main_cpuid == -1) {  in do_stackshot()
5620 		os_atomic_cmpxchg(&stackshot_ctx.sc_main_cpuid, -1, cpu_number(), acquire);  in do_stackshot()
5625 	os_atomic_store(&stackshot_ctx.sc_state, SS_RUNNING, release);  in do_stackshot()
5628 	if (stackshot_ctx.sc_panic_stackshot || (stackshot_ctx.sc_main_cpuid == cpu_number())) {  in do_stackshot()
5629 		stackshot_ctx.sc_retval = kdp_stackshot_kcdata_format();  in do_stackshot()
5636 	stackshot_ctx.sc_latency.cpu_wait_latency_mt = mach_absolute_time();  in do_stackshot()
5639 		os_atomic_dec(&stackshot_ctx.sc_cpus_working, seq_cst);  in do_stackshot()
5642 	while (os_atomic_load(&stackshot_ctx.sc_cpus_working, seq_cst) != 0) {  in do_stackshot()
5647 …stackshot_ctx.sc_latency.cpu_wait_latency_mt = mach_absolute_time() - stackshot_ctx.sc_latency.cpu…  in do_stackshot()
5652 	stackshot_ctx.sc_duration = (struct stackshot_duration_v2) {  in do_stackshot()
5674 	if (!stackshot_ctx.sc_panic_stackshot && interrupt_masked_debug_mode) {  in do_stackshot()
5686 	if (stackshot_ctx.sc_retval == KERN_SUCCESS) {  in do_stackshot()
5687 		stackshot_ctx.sc_retval = stackshot_status_check();  in do_stackshot()
5692 	if (!stackshot_ctx.sc_panic_stackshot && stackshot_ctx.sc_retval == KERN_SUCCESS) {  in do_stackshot()
5695 	if (stackshot_ctx.sc_retval != KERN_SUCCESS && stackshot_exclave_inspect_ctids) {  in do_stackshot()
5704 	if (stackshot_ctx.sc_is_singlethreaded) {  in do_stackshot()
5705 		stackshot_ctx.sc_finalized_kcdata = stackshot_kcdata_p;  in do_stackshot()
5708 	return stackshot_ctx.sc_retval;  in do_stackshot()
5762 		os_atomic_cmpxchg(&stackshot_ctx.sc_main_cpuid, -1, my_cpu_no, acquire);  in stackshot_cpu_preflight()
5765 	is_calling_cpu = stackshot_ctx.sc_calling_cpuid == my_cpu_no;  in stackshot_cpu_preflight()
5768 …stackshot_cpu_ctx.scc_can_work = is_calling_cpu || (is_recommended && !stackshot_ctx.sc_is_singlet…  in stackshot_cpu_preflight()
5771 		os_atomic_inc(&stackshot_ctx.sc_cpus_working, relaxed);  in stackshot_cpu_preflight()
5833 			kcd_exit_on_error(stackshot_cpu_work_on_queue(&stackshot_ctx.sc_workqueues[i - 1]));  in stackshot_cpu_do_work()
5838 			kcd_exit_on_error(stackshot_cpu_work_on_queue(&stackshot_ctx.sc_workqueues[i]));  in stackshot_cpu_do_work()
5878 	while (os_atomic_load(&stackshot_ctx.sc_state, acquire) == SS_SETUP) {  in stackshot_aux_cpu_entry()
5890 	if (stackshot_ctx.sc_main_cpuid == cpu_number()) {  in stackshot_aux_cpu_entry()
5891 		stackshot_ctx.sc_retval = kdp_stackshot_kcdata_format();  in stackshot_aux_cpu_entry()
5897 	os_atomic_dec(&stackshot_ctx.sc_cpus_working, release);  in stackshot_aux_cpu_entry()