source: mainline/kernel/generic/src/sysinfo/stats.c@ 70e2b2d

/*
 * Copyright (c) 2010 Stanislav Kozina
 * Copyright (c) 2010 Martin Decky
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 *
 * - Redistributions of source code must retain the above copyright
 *   notice, this list of conditions and the following disclaimer.
 * - Redistributions in binary form must reproduce the above copyright
 *   notice, this list of conditions and the following disclaimer in the
 *   documentation and/or other materials provided with the distribution.
 * - The name of the author may not be used to endorse or promote products
 *   derived from this software without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 */

/** @addtogroup generic
 * @{
 */
/** @file
 */

#include <typedefs.h>
#include <sysinfo/abi.h>
#include <sysinfo/stats.h>
#include <sysinfo/sysinfo.h>
#include <time/clock.h>
#include <mm/frame.h>
#include <proc/thread.h>
#include <str.h>
#include <errno.h>
#include <cpu.h>
#include <arch.h>

/** Bits of fixed-point precision for load */
#define LOAD_FIXED_SHIFT  11

/** 1.0 as fixed-point for load */
#define LOAD_FIXED_1  (1 << LOAD_FIXED_SHIFT)

/** Compute load in 5 second intervals */
#define LOAD_INTERVAL  5

/** Fixed-point representation of
 *
 * 1 / exp(5 sec / 1 min)
 * 1 / exp(5 sec / 5 min)
 * 1 / exp(5 sec / 15 min)
 *
 */
static load_t load_exp[LOAD_STEPS] = {1884, 2014, 2037};

/** Running average of the number of ready threads */
static load_t avenrdy[LOAD_STEPS] = {0, 0, 0};

/** Load calculation spinlock */
SPINLOCK_STATIC_INITIALIZE_NAME(load_lock, "load_lock");

/** Get system uptime
 *
 * @param item Sysinfo item (unused).
 *
 * @return System uptime (in secords).
 *
 */
static unative_t get_stats_uptime(struct sysinfo_item *item)
{
        /* This doesn't have to be very accurate */
        return uptime->seconds1;
}

/** Get statistics of all CPUs
 *
 * @param item    Sysinfo item (unused).
 * @param size    Size of the returned data.
 * @param dry_run Do not get the data, just calculate the size.
 *
 * @return Data containing several stats_cpu_t structures.
 *         If the return value is not NULL, it should be freed
 *         in the context of the sysinfo request.
 */
static void *get_stats_cpus(struct sysinfo_item *item, size_t *size,
    bool dry_run)
{
        *size = sizeof(stats_cpu_t) * config.cpu_count;
        if (dry_run)
                return NULL;
        
        /* Assumption: config.cpu_count is constant */
        stats_cpu_t *stats_cpus = (stats_cpu_t *) malloc(*size, FRAME_ATOMIC);
        if (stats_cpus == NULL) {
                *size = 0;
                return NULL;
        }
        
        /* Each CPU structure is locked separatelly */
        ipl_t ipl = interrupts_disable();
        
        size_t i;
        for (i = 0; i < config.cpu_count; i++) {
                spinlock_lock(&cpus[i].lock);
                
                stats_cpus[i].id = cpus[i].id;
                stats_cpus[i].frequency_mhz = cpus[i].frequency_mhz;
                stats_cpus[i].busy_ticks = cpus[i].busy_ticks;
                stats_cpus[i].idle_ticks = cpus[i].idle_ticks;
                
                spinlock_unlock(&cpus[i].lock);
        }
        
        interrupts_restore(ipl);
        
        return ((void *) stats_cpus);
}

/** Count number of tasks
 *
 * AVL task tree walker for counting tasks.
 *
 * @param node AVL task tree node (unused).
 * @param arg  Pointer to the counter.
 *
 * @param Always true (continue the walk).
 *
 */
static bool task_count_walker(avltree_node_t *node, void *arg)
{
        size_t *count = (size_t *) arg;
        (*count)++;
        
        return true;
}

/** Gather tasks
 *
 * AVL task tree walker for gathering task IDs. Interrupts should
 * be already disabled while walking the tree.
 *
 * @param node AVL task tree node.
 * @param arg  Pointer to the iterator into the array of task IDs.
 *
 * @param Always true (continue the walk).
 *
 */
static bool task_serialize_walker(avltree_node_t *node, void *arg)
{
        task_id_t **ids = (task_id_t **) arg;
        task_t *task = avltree_get_instance(node, task_t, tasks_tree_node);
        
        /* Interrupts are already disabled */
        spinlock_lock(&(task->lock));
        
        /* Record the ID and increment the iterator */
        **ids = task->taskid;
        (*ids)++;
        
        spinlock_unlock(&(task->lock));
        
        return true;
}

/** Get task IDs
 *
 * @param item    Sysinfo item (unused).
 * @param size    Size of the returned data.
 * @param dry_run Do not get the data, just calculate the size.
 *
 * @return Data containing task IDs of all tasks.
 *         If the return value is not NULL, it should be freed
 *         in the context of the sysinfo request.
 */
static void *get_stats_tasks(struct sysinfo_item *item, size_t *size,
    bool dry_run)
{
        /* Messing with task structures, avoid deadlock */
        ipl_t ipl = interrupts_disable();
        spinlock_lock(&tasks_lock);
        
        /* First walk the task tree to count the tasks */
        size_t count = 0;
        avltree_walk(&tasks_tree, task_count_walker, (void *) &count);
        
        if (count == 0) {
                /* No tasks found (strange) */
                spinlock_unlock(&tasks_lock);
                interrupts_restore(ipl);
                
                *size = 0;
                return NULL;
        }
        
        *size = sizeof(task_id_t) * count;
        if (dry_run) {
                spinlock_unlock(&tasks_lock);
                interrupts_restore(ipl);
                return NULL;
        }
        
        task_id_t *task_ids = (task_id_t *) malloc(*size, FRAME_ATOMIC);
        if (task_ids == NULL) {
                /* No free space for allocation */
                spinlock_unlock(&tasks_lock);
                interrupts_restore(ipl);
                
                *size = 0;
                return NULL;
        }
        
        /* Walk tha task tree again to gather the IDs */
        task_id_t *iterator = task_ids;
        avltree_walk(&tasks_tree, task_serialize_walker, (void *) &iterator);
        
        spinlock_unlock(&tasks_lock);
        interrupts_restore(ipl);
        
        return ((void *) task_ids);
}

/** Get the size of a virtual address space
 *
 * @param as Address space.
 *
 * @return Size of the mapped virtual address space (bytes).
 *
 */
static size_t get_task_virtmem(as_t *as)
{
        mutex_lock(&as->lock);
        
        size_t result = 0;
        
        /* Walk the B+ tree and count pages */
        link_t *cur;
        for (cur = as->as_area_btree.leaf_head.next;
            cur != &as->as_area_btree.leaf_head; cur = cur->next) {
                btree_node_t *node =
                    list_get_instance(cur, btree_node_t, leaf_link);
                
                unsigned int i;
                for (i = 0; i < node->keys; i++) {
                        as_area_t *area = node->value[i];
                        
                        mutex_lock(&area->lock);
                        result += area->pages;
                        mutex_unlock(&area->lock);
                }
        }
        
        mutex_unlock(&as->lock);
        
        return result * PAGE_SIZE;
}

/** Get task statistics
 *
 * Get statistics of a given task. The task ID is passed
 * as a string (current limitation of the sysinfo interface,
 * but it is still reasonable for the given purpose).
 *
 * @param name Task ID (string-encoded number).
 *
 * @return Sysinfo return holder. The type of the returned
 *         data is either SYSINFO_VAL_UNDEFINED (unknown
 *         task ID or memory allocation error) or
 *         SYSINFO_VAL_FUNCTION_DATA (in that case the
 *         generated data should be freed within the
 *         sysinfo request context).
 *
 */
static sysinfo_return_t get_stats_task(const char *name)
{
        /* Initially no return value */
        sysinfo_return_t ret;
        ret.tag = SYSINFO_VAL_UNDEFINED;
        
        /* Parse the task ID */
        task_id_t task_id;
        if (str_uint64(name, NULL, 0, true, &task_id) != EOK)
                return ret;
        
        /* Allocate stats_task_t structure */
        stats_task_t *stats_task =
            (stats_task_t *) malloc(sizeof(stats_task_t), FRAME_ATOMIC);
        if (stats_task == NULL)
                return ret;
        
        /* Messing with task structures, avoid deadlock */
        ipl_t ipl = interrupts_disable();
        spinlock_lock(&tasks_lock);
        
        task_t *task = task_find_by_id(task_id);
        if (task == NULL) {
                /* No task with this ID */
                spinlock_unlock(&tasks_lock);
                interrupts_restore(ipl);
                free(stats_task);
                return ret;
        }
        
        /* Hand-over-hand locking */
        spinlock_lock(&task->lock);
        spinlock_unlock(&tasks_lock);
        
        /* Copy task's statistics */
        str_cpy(stats_task->name, TASK_NAME_BUFLEN, task->name);
        stats_task->virtmem = get_task_virtmem(task->as);
        stats_task->threads = atomic_get(&task->refcount);
        task_get_accounting(task, &(stats_task->ucycles),
            &(stats_task->kcycles));
        stats_task->ipc_info = task->ipc_info;
        
        spinlock_unlock(&task->lock);
        interrupts_restore(ipl);
        
        /* Correct return value */
        ret.tag = SYSINFO_VAL_FUNCTION_DATA;
        ret.data.data = (void *) stats_task;
        ret.data.size = sizeof(stats_task_t);
        
        return ret;
}

/** Get physical memory statistics
 *
 * @param item    Sysinfo item (unused).
 * @param size    Size of the returned data.
 * @param dry_run Do not get the data, just calculate the size.
 *
 * @return Data containing stats_physmem_t.
 *         If the return value is not NULL, it should be freed
 *         in the context of the sysinfo request.
 */
static void *get_stats_physmem(struct sysinfo_item *item, size_t *size,
    bool dry_run)
{
        *size = sizeof(stats_physmem_t);
        if (dry_run)
                return NULL;
        
        stats_physmem_t *stats_physmem =
            (stats_physmem_t *) malloc(*size, FRAME_ATOMIC);
        if (stats_physmem == NULL) {
                *size = 0;
                return NULL;
        }
        
        zones_stats(&(stats_physmem->total), &(stats_physmem->unavail),
            &(stats_physmem->used), &(stats_physmem->free));
        
        return ((void *) stats_physmem);
}

/** Get system load
 *
 * @param item    Sysinfo item (unused).
 * @param size    Size of the returned data.
 * @param dry_run Do not get the data, just calculate the size.
 *
 * @return Data several load_t values.
 *         If the return value is not NULL, it should be freed
 *         in the context of the sysinfo request.
 */
static void *get_stats_load(struct sysinfo_item *item, size_t *size,
    bool dry_run)
{
        *size = sizeof(load_t) * LOAD_STEPS;
        if (dry_run)
                return NULL;
        
        load_t *stats_load = (load_t *) malloc(*size, FRAME_ATOMIC);
        if (stats_load == NULL) {
                *size = 0;
                return NULL;
        }
        
        /* To always get consistent values acquire the spinlock */
        ipl_t ipl = interrupts_disable();
        spinlock_lock(&load_lock);
        
        unsigned int i;
        for (i = 0; i < LOAD_STEPS; i++)
                stats_load[i] = avenrdy[i] << LOAD_FIXED_SHIFT;
        
        spinlock_unlock(&load_lock);
        interrupts_restore(ipl);
        
        return ((void *) stats_load);
}

/** Calculate load
 *
 */
static inline load_t load_calc(load_t load, load_t exp, size_t ready)
{
        load *= exp;
        load += ready * (LOAD_FIXED_1 - exp);
        
        return (load >> LOAD_FIXED_SHIFT);
}

/** Count threads in ready queues
 *
 * Should be called with interrupts disabled.
 *
 */
static inline size_t get_ready_count(void)
{
        size_t i;
        size_t count = 0;
        
        for (i = 0; i < config.cpu_count; i++) {
                spinlock_lock(&cpus[i].lock);
                
                size_t j;
                for (j = 0; j < RQ_COUNT; j++) {
                        spinlock_lock(&cpus[i].rq[j].lock);
                        count += cpus[i].rq[j].n;
                        spinlock_unlock(&cpus[i].rq[j].lock);
                }
                
                spinlock_unlock(&cpus[i].lock);
        }
        
        return count;
}

/** Load computation thread.
 *
 * Compute system load every few seconds.
 *
 * @param arg Unused.
 *
 */
void kload(void *arg)
{
        thread_detach(THREAD);
        
        while (true) {
                /* Mutually exclude with get_stats_load() */
                ipl_t ipl = interrupts_disable();
                spinlock_lock(&load_lock);
                
                size_t ready = get_ready_count() * LOAD_FIXED_1;
                
                unsigned int i;
                for (i = 0; i < LOAD_STEPS; i++)
                        avenrdy[i] = load_calc(avenrdy[i], load_exp[i], ready);
                
                spinlock_unlock(&load_lock);
                interrupts_restore(ipl);
                
                thread_sleep(LOAD_INTERVAL);
        }
}

/** Register sysinfo statistical items
 *
 */
void stats_init(void)
{
        sysinfo_set_item_fn_val("system.uptime", NULL, get_stats_uptime);
        sysinfo_set_item_fn_data("system.cpus", NULL, get_stats_cpus);
        sysinfo_set_item_fn_data("system.physmem", NULL, get_stats_physmem);
        sysinfo_set_item_fn_data("system.load", NULL, get_stats_load);
        sysinfo_set_item_fn_data("system.tasks", NULL, get_stats_tasks);
        sysinfo_set_subtree_fn("system.tasks", NULL, get_stats_task);
}

/** @}
 */