]> arthur.barton.de Git - netdata.git/blobdiff - src/proc_diskstats.c
dns_query_time plugin: replace "." with "_" in dimensions
[netdata.git] / src / proc_diskstats.c
index 78c8092dfe769f769e24bd90d67f9f9c8b21da52..a1b4072dd93ec81a6805f70f2b4b2cc0a28125f1 100644 (file)
@@ -1,25 +1,4 @@
-#ifdef HAVE_CONFIG_H
-#include <config.h>
-#endif
-#include <dirent.h>
-#include <fcntl.h>
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
-#include <sys/stat.h>
-#include <sys/statvfs.h>
-#include <sys/types.h>
-#include <unistd.h>
-
-
 #include "common.h"
-#include "log.h"
-#include "appconfig.h"
-#include "procfile.h"
-#include "rrd.h"
-#include "plugin_proc.h"
-
-#include "proc_self_mountinfo.h"
 
 #define RRD_TYPE_DISK "disk"
 
 #define DISK_TYPE_PARTITION 2
 #define DISK_TYPE_CONTAINER 3
 
+#define CONFIG_SECTION_DISKSTATS "plugin:proc:/proc/diskstats"
+#define DELAULT_EXLUDED_DISKS "loop* ram*"
+
 static struct disk {
-       char *disk;                             // the name of the disk (sda, sdb, etc)
-       unsigned long major;
-       unsigned long minor;
-       int sector_size;
-       int type;
-       char *mount_point;
-
-       // disk options caching
-       int configured;
-       int do_io;
-       int do_ops;
-       int do_mops;
-       int do_iotime;
-       int do_qops;
-       int do_util;
-       int do_backlog;
-       int do_space;
-       int do_inodes;
-
-       struct disk *next;
+    char *disk;             // the name of the disk (sda, sdb, etc)
+    unsigned long major;
+    unsigned long minor;
+    int sector_size;
+    int type;
+
+    char *mount_point;
+
+    // disk options caching
+    int configured;
+    int do_io;
+    int do_ops;
+    int do_mops;
+    int do_iotime;
+    int do_qops;
+    int do_util;
+    int do_backlog;
+
+    int updated;
+
+    RRDSET *st_avgsz;
+    RRDSET *st_await;
+    RRDSET *st_backlog;
+    RRDSET *st_io;
+    RRDSET *st_iotime;
+    RRDSET *st_mops;
+    RRDSET *st_ops;
+    RRDSET *st_qops;
+    RRDSET *st_svctm;
+    RRDSET *st_util;
+
+    struct disk *next;
 } *disk_root = NULL;
 
-static struct mountinfo *disk_mountinfo_root = NULL;
+#define rrdset_obsolete_and_pointer_null(st) do { if(st) { rrdset_flag_set(st, RRDSET_FLAG_OBSOLETE); st = NULL; } } while(st)
 
 static struct disk *get_disk(unsigned long major, unsigned long minor, char *disk) {
-       static char path_to_get_hw_sector_size[FILENAME_MAX + 1] = "";
-       static char path_to_get_hw_sector_size_partitions[FILENAME_MAX + 1] = "";
-       static char path_find_block_device[FILENAME_MAX + 1] = "";
-       struct disk *d;
-
-       // search for it in our RAM list.
-       // this is sequential, but since we just walk through
-       // and the number of disks / partitions in a system
-       // should not be that many, it should be acceptable
-       for(d = disk_root; d ; d = d->next)
-               if(unlikely(d->major == major && d->minor == minor))
-                       break;
-
-       // if we found it, return it
-       if(likely(d))
-               return d;
-
-       // not found
-       // create a new disk structure
-       d = (struct disk *)malloc(sizeof(struct disk));
-       if(!d) fatal("Cannot allocate memory for struct disk in proc_diskstats.");
-
-       d->disk = strdup(disk);
-       d->major = major;
-       d->minor = minor;
-       d->type = DISK_TYPE_PHYSICAL; // Default type. Changed later if not correct.
-       d->configured = 0;
-       d->sector_size = 512; // the default, will be changed below
-       d->next = NULL;
-
-       // append it to the list
-       if(!disk_root)
-               disk_root = d;
-       else {
-               struct disk *last;
-               for(last = disk_root; last->next ;last = last->next);
-               last->next = d;
-       }
-
-       // ------------------------------------------------------------------------
-       // find the type of the device
-
-       char buffer[FILENAME_MAX + 1];
-
-       // get the default path for finding info about the block device
-       if(unlikely(!path_find_block_device[0])) {
-               snprintfz(buffer, FILENAME_MAX, "%s%s", global_host_prefix, "/sys/dev/block/%lu:%lu/%s");
-               snprintfz(path_find_block_device, FILENAME_MAX, "%s", config_get("plugin:proc:/proc/diskstats", "path to get block device infos", buffer));
-       }
-
-       // find if it is a partition
-       // by checking if /sys/dev/block/MAJOR:MINOR/partition is readable.
-       snprintfz(buffer, FILENAME_MAX, path_find_block_device, major, minor, "partition");
-       if(access(buffer, R_OK) == 0) {
-               d->type = DISK_TYPE_PARTITION;
-       } else {
-               // find if it is a container
-               // by checking if /sys/dev/block/MAJOR:MINOR/slaves has entries
-               snprintfz(buffer, FILENAME_MAX, path_find_block_device, major, minor, "slaves/");
-               DIR *dirp = opendir(buffer);    
-               if (dirp != NULL) {
-                       struct dirent *dp;
-                       while( (dp = readdir(dirp)) ) {
-                               // . and .. are also files in empty folders.
-                               if(strcmp(dp->d_name, ".") == 0 || strcmp(dp->d_name, "..") == 0) {
-                                       continue;
-                               }
-
-                               d->type = DISK_TYPE_CONTAINER;
-
-                               // Stop the loop after we found one file.
-                               break;
-                       }
-                       if(closedir(dirp) == -1)
-                               error("Unable to close dir %s", buffer);
-               }
-       }
-
-       // ------------------------------------------------------------------------
-       // check if we can find its mount point
-
-       // mountinfo_find() can be called with NULL disk_mountinfo_root
-       struct mountinfo *mi = mountinfo_find(disk_mountinfo_root, d->major, d->minor);
-       if(unlikely(!mi)) {
-               // mountinfo_free() can be called with NULL disk_mountinfo_root
-               mountinfo_free(disk_mountinfo_root);
-
-               // re-read mountinfo in case something changed
-               disk_mountinfo_root = mountinfo_read();
-
-               // search again for this disk
-               mi = mountinfo_find(disk_mountinfo_root, d->major, d->minor);
-       }
-
-       if(mi)
-               d->mount_point = strdup(mi->mount_point);
-               // no need to check for NULL
-       else
-               d->mount_point = NULL;
-
-       // ------------------------------------------------------------------------
-       // find the disk sector size
-
-       if(!path_to_get_hw_sector_size[0]) {
-               snprintfz(buffer, FILENAME_MAX, "%s%s", global_host_prefix, "/sys/block/%s/queue/hw_sector_size");
-               snprintfz(path_to_get_hw_sector_size, FILENAME_MAX, "%s", config_get("plugin:proc:/proc/diskstats", "path to get h/w sector size", buffer));
-       }
-       if(!path_to_get_hw_sector_size_partitions[0]) {
-               snprintfz(buffer, FILENAME_MAX, "%s%s", global_host_prefix, "/sys/dev/block/%lu:%lu/subsystem/%s/../queue/hw_sector_size");
-               snprintfz(path_to_get_hw_sector_size_partitions, FILENAME_MAX, "%s", config_get("plugin:proc:/proc/diskstats", "path to get h/w sector size for partitions", buffer));
-       }
-
-       {
-               char tf[FILENAME_MAX + 1], *t;
-               strncpyz(tf, d->disk, FILENAME_MAX);
-
-               // replace all / with !
-               for(t = tf; *t ;t++)
-                       if(*t == '/') *t = '!';
-
-               if(d->type == DISK_TYPE_PARTITION)
-                       snprintfz(buffer, FILENAME_MAX, path_to_get_hw_sector_size_partitions, d->major, d->minor, tf);
-               else
-                       snprintfz(buffer, FILENAME_MAX, path_to_get_hw_sector_size, tf);
-
-               FILE *fpss = fopen(buffer, "r");
-               if(fpss) {
-                       char buffer2[1024 + 1];
-                       char *tmp = fgets(buffer2, 1024, fpss);
-
-                       if(tmp) {
-                               d->sector_size = atoi(tmp);
-                               if(d->sector_size <= 0) {
-                                       error("Invalid sector size %d for device %s in %s. Assuming 512.", d->sector_size, d->disk, buffer);
-                                       d->sector_size = 512;
-                               }
-                       }
-                       else error("Cannot read data for sector size for device %s from %s. Assuming 512.", d->disk, buffer);
-
-                       fclose(fpss);
-               }
-               else error("Cannot read sector size for device %s from %s. Assuming 512.", d->disk, buffer);
-       }
-
-       return d;
+    static char path_to_get_hw_sector_size[FILENAME_MAX + 1] = "";
+    static char path_to_get_hw_sector_size_partitions[FILENAME_MAX + 1] = "";
+    static char path_find_block_device[FILENAME_MAX + 1] = "";
+    static struct mountinfo *disk_mountinfo_root = NULL;
+
+    struct disk *d;
+
+    // search for it in our RAM list.
+    // this is sequential, but since we just walk through
+    // and the number of disks / partitions in a system
+    // should not be that many, it should be acceptable
+    for(d = disk_root; d ; d = d->next)
+        if(unlikely(d->major == major && d->minor == minor))
+            return d;
+
+    // not found
+    // create a new disk structure
+    d = (struct disk *)callocz(1, sizeof(struct disk));
+
+    d->disk = strdupz(disk);
+    d->major = major;
+    d->minor = minor;
+    d->type = DISK_TYPE_PHYSICAL; // Default type. Changed later if not correct.
+    d->configured = 0;
+    d->sector_size = 512; // the default, will be changed below
+    d->next = NULL;
+
+    // append it to the list
+    if(unlikely(!disk_root))
+        disk_root = d;
+    else {
+        struct disk *last;
+        for(last = disk_root; last->next ;last = last->next);
+        last->next = d;
+    }
+
+    // ------------------------------------------------------------------------
+    // find the type of the device
+
+    char buffer[FILENAME_MAX + 1];
+
+    // get the default path for finding info about the block device
+    if(unlikely(!path_find_block_device[0])) {
+        snprintfz(buffer, FILENAME_MAX, "%s%s", netdata_configured_host_prefix, "/sys/dev/block/%lu:%lu/%s");
+        snprintfz(path_find_block_device, FILENAME_MAX, "%s", config_get(CONFIG_SECTION_DISKSTATS, "path to get block device infos", buffer));
+    }
+
+    // find if it is a partition
+    // by checking if /sys/dev/block/MAJOR:MINOR/partition is readable.
+    snprintfz(buffer, FILENAME_MAX, path_find_block_device, major, minor, "partition");
+    if(likely(access(buffer, R_OK) == 0)) {
+        d->type = DISK_TYPE_PARTITION;
+    }
+    else {
+        // find if it is a container
+        // by checking if /sys/dev/block/MAJOR:MINOR/slaves has entries
+        snprintfz(buffer, FILENAME_MAX, path_find_block_device, major, minor, "slaves/");
+        DIR *dirp = opendir(buffer);
+        if(likely(dirp != NULL)) {
+            struct dirent *dp;
+            while( (dp = readdir(dirp)) ) {
+                // . and .. are also files in empty folders.
+                if(unlikely(strcmp(dp->d_name, ".") == 0 || strcmp(dp->d_name, "..") == 0)) {
+                    continue;
+                }
+
+                d->type = DISK_TYPE_CONTAINER;
+
+                // Stop the loop after we found one file.
+                break;
+            }
+            if(unlikely(closedir(dirp) == -1))
+                error("Unable to close dir %s", buffer);
+        }
+    }
+
+    // ------------------------------------------------------------------------
+    // check if we can find its mount point
+
+    // mountinfo_find() can be called with NULL disk_mountinfo_root
+    struct mountinfo *mi = mountinfo_find(disk_mountinfo_root, d->major, d->minor);
+    if(unlikely(!mi)) {
+        // mountinfo_free can be called with NULL
+        mountinfo_free(disk_mountinfo_root);
+        disk_mountinfo_root = mountinfo_read(0);
+        mi = mountinfo_find(disk_mountinfo_root, d->major, d->minor);
+    }
+
+    if(unlikely(mi))
+        d->mount_point = strdupz(mi->mount_point);
+    else
+        d->mount_point = NULL;
+
+    // ------------------------------------------------------------------------
+    // find the disk sector size
+
+    if(unlikely(!path_to_get_hw_sector_size[0])) {
+        snprintfz(buffer, FILENAME_MAX, "%s%s", netdata_configured_host_prefix, "/sys/block/%s/queue/hw_sector_size");
+        snprintfz(path_to_get_hw_sector_size, FILENAME_MAX, "%s", config_get(CONFIG_SECTION_DISKSTATS, "path to get h/w sector size", buffer));
+    }
+    if(unlikely(!path_to_get_hw_sector_size_partitions[0])) {
+        snprintfz(buffer, FILENAME_MAX, "%s%s", netdata_configured_host_prefix, "/sys/dev/block/%lu:%lu/subsystem/%s/../queue/hw_sector_size");
+        snprintfz(path_to_get_hw_sector_size_partitions, FILENAME_MAX, "%s", config_get(CONFIG_SECTION_DISKSTATS, "path to get h/w sector size for partitions", buffer));
+    }
+
+    {
+        char tf[FILENAME_MAX + 1], *t;
+        strncpyz(tf, d->disk, FILENAME_MAX);
+
+        // replace all / with !
+        for(t = tf; *t ;t++)
+            if(unlikely(*t == '/')) *t = '!';
+
+        if(likely(d->type == DISK_TYPE_PARTITION))
+            snprintfz(buffer, FILENAME_MAX, path_to_get_hw_sector_size_partitions, d->major, d->minor, tf);
+        else
+            snprintfz(buffer, FILENAME_MAX, path_to_get_hw_sector_size, tf);
+
+        FILE *fpss = fopen(buffer, "r");
+        if(likely(fpss)) {
+            char buffer2[1024 + 1];
+            char *tmp = fgets(buffer2, 1024, fpss);
+
+            if(likely(tmp)) {
+                d->sector_size = str2i(tmp);
+                if(unlikely(d->sector_size <= 0)) {
+                    error("Invalid sector size %d for device %s in %s. Assuming 512.", d->sector_size, d->disk, buffer);
+                    d->sector_size = 512;
+                }
+            }
+            else error("Cannot read data for sector size for device %s from %s. Assuming 512.", d->disk, buffer);
+
+            fclose(fpss);
+        }
+        else error("Cannot read sector size for device %s from %s. Assuming 512.", d->disk, buffer);
+    }
+
+    return d;
 }
 
-static inline int select_positive_option(int option1, int option2) {
-       if(option1 == CONFIG_ONDEMAND_YES || option2 == CONFIG_ONDEMAND_YES)
-               return CONFIG_ONDEMAND_YES;
-       else if(option1 == CONFIG_ONDEMAND_ONDEMAND || option2 == CONFIG_ONDEMAND_ONDEMAND)
-               return CONFIG_ONDEMAND_ONDEMAND;
+static inline int is_major_enabled(int major) {
+    static int8_t *major_configs = NULL;
+    static size_t major_size = 0;
+
+    if(major < 0) return 1;
+
+    size_t wanted_size = (size_t)major + 1;
+
+    if(major_size < wanted_size) {
+        major_configs = reallocz(major_configs, wanted_size * sizeof(int8_t));
+
+        size_t i;
+        for(i = major_size; i < wanted_size ; i++)
+            major_configs[i] = -1;
+
+        major_size = wanted_size;
+    }
+
+    if(major_configs[major] == -1) {
+        char buffer[CONFIG_MAX_NAME + 1];
+        snprintfz(buffer, CONFIG_MAX_NAME, "performance metrics for disks with major %d", major);
+        major_configs[major] = (char)config_get_boolean(CONFIG_SECTION_DISKSTATS, buffer, 1);
+    }
 
-       return CONFIG_ONDEMAND_NO;
+    return (int)major_configs[major];
 }
 
-int do_proc_diskstats(int update_every, unsigned long long dt) {
-       static procfile *ff = NULL;
-       static struct statvfs buff_statvfs;
-       static struct stat buff_stat;
-       static int      global_enable_new_disks_detected_at_runtime = CONFIG_ONDEMAND_YES,
-                               global_enable_performance_for_physical_disks = CONFIG_ONDEMAND_ONDEMAND,
-                               global_enable_performance_for_virtual_disks = CONFIG_ONDEMAND_NO,
-                               global_enable_performance_for_partitions = CONFIG_ONDEMAND_NO,
-                               global_enable_performance_for_mountpoints = CONFIG_ONDEMAND_NO,
-                               global_enable_performance_for_virtual_mountpoints = CONFIG_ONDEMAND_ONDEMAND,
-                               global_enable_space_for_mountpoints = CONFIG_ONDEMAND_ONDEMAND,
-                               global_do_io = CONFIG_ONDEMAND_ONDEMAND,
-                               global_do_ops = CONFIG_ONDEMAND_ONDEMAND,
-                               global_do_mops = CONFIG_ONDEMAND_ONDEMAND,
-                               global_do_iotime = CONFIG_ONDEMAND_ONDEMAND,
-                               global_do_qops = CONFIG_ONDEMAND_ONDEMAND,
-                               global_do_util = CONFIG_ONDEMAND_ONDEMAND,
-                               global_do_backlog = CONFIG_ONDEMAND_ONDEMAND,
-                               global_do_space = CONFIG_ONDEMAND_ONDEMAND,
-                               global_do_inodes = CONFIG_ONDEMAND_ONDEMAND,
-                               globals_initialized = 0;
-
-       if(unlikely(!globals_initialized)) {
-               global_enable_new_disks_detected_at_runtime = config_get_boolean("plugin:proc:/proc/diskstats", "enable new disks detected at runtime", global_enable_new_disks_detected_at_runtime);
-
-               global_enable_performance_for_physical_disks = config_get_boolean_ondemand("plugin:proc:/proc/diskstats", "performance metrics for physical disks", global_enable_performance_for_physical_disks);
-               global_enable_performance_for_virtual_disks = config_get_boolean_ondemand("plugin:proc:/proc/diskstats", "performance metrics for virtual disks", global_enable_performance_for_virtual_disks);
-               global_enable_performance_for_partitions = config_get_boolean_ondemand("plugin:proc:/proc/diskstats", "performance metrics for partitions", global_enable_performance_for_partitions);
-               global_enable_performance_for_mountpoints = config_get_boolean_ondemand("plugin:proc:/proc/diskstats", "performance metrics for mounted filesystems", global_enable_performance_for_mountpoints);
-               global_enable_performance_for_virtual_mountpoints = config_get_boolean_ondemand("plugin:proc:/proc/diskstats", "performance metrics for mounted virtual disks", global_enable_performance_for_virtual_mountpoints);
-               global_enable_space_for_mountpoints = config_get_boolean_ondemand("plugin:proc:/proc/diskstats", "space metrics for mounted filesystems", global_enable_space_for_mountpoints);
-
-               global_do_io      = config_get_boolean_ondemand("plugin:proc:/proc/diskstats", "bandwidth for all disks", global_do_io);
-               global_do_ops     = config_get_boolean_ondemand("plugin:proc:/proc/diskstats", "operations for all disks", global_do_ops);
-               global_do_mops    = config_get_boolean_ondemand("plugin:proc:/proc/diskstats", "merged operations for all disks", global_do_mops);
-               global_do_iotime  = config_get_boolean_ondemand("plugin:proc:/proc/diskstats", "i/o time for all disks", global_do_iotime);
-               global_do_qops    = config_get_boolean_ondemand("plugin:proc:/proc/diskstats", "queued operations for all disks", global_do_qops);
-               global_do_util    = config_get_boolean_ondemand("plugin:proc:/proc/diskstats", "utilization percentage for all disks", global_do_util);
-               global_do_backlog = config_get_boolean_ondemand("plugin:proc:/proc/diskstats", "backlog for all disks", global_do_backlog);
-               global_do_space   = config_get_boolean_ondemand("plugin:proc:/proc/diskstats", "space usage for all disks", global_do_space);
-               global_do_inodes  = config_get_boolean_ondemand("plugin:proc:/proc/diskstats", "inodes usage for all disks", global_do_inodes);
-
-               globals_initialized = 1;
-       }
-
-       if(!ff) {
-               char filename[FILENAME_MAX + 1];
-               snprintfz(filename, FILENAME_MAX, "%s%s", global_host_prefix, "/proc/diskstats");
-               ff = procfile_open(config_get("plugin:proc:/proc/diskstats", "filename to monitor", filename), " \t", PROCFILE_FLAG_DEFAULT);
-       }
-       if(!ff) return 1;
-
-       ff = procfile_readall(ff);
-       if(!ff) return 0; // we return 0, so that we will retry to open it next time
-
-       uint32_t lines = procfile_lines(ff), l;
-       uint32_t words;
-
-       for(l = 0; l < lines ;l++) {
-               // --------------------------------------------------------------------------
-               // Read parameters
-
-               char *disk;
-               unsigned long long      major = 0, minor = 0,
-                                                       reads = 0,  mreads = 0,  readsectors = 0,  readms = 0,
-                                                       writes = 0, mwrites = 0, writesectors = 0, writems = 0,
-                                                       queued_ios = 0, busy_ms = 0, backlog_ms = 0,
-                                                       space_avail = 0, space_avail_root = 0, space_used = 0,
-                                                       inodes_avail = 0, inodes_avail_root = 0, inodes_used = 0;
-
-               unsigned long long      last_reads = 0,  last_readsectors = 0,  last_readms = 0,
-                                                       last_writes = 0, last_writesectors = 0, last_writems = 0,
-                                                       last_busy_ms = 0;
-
-               words = procfile_linewords(ff, l);
-               if(words < 14) continue;
-
-               major                   = strtoull(procfile_lineword(ff, l, 0), NULL, 10);
-               minor                   = strtoull(procfile_lineword(ff, l, 1), NULL, 10);
-               disk                    = procfile_lineword(ff, l, 2);
-
-               // # of reads completed # of writes completed
-               // This is the total number of reads or writes completed successfully.
-               reads                   = strtoull(procfile_lineword(ff, l, 3), NULL, 10);      // rd_ios
-               writes                  = strtoull(procfile_lineword(ff, l, 7), NULL, 10);      // wr_ios
-
-               // # of reads merged # of writes merged
-               // Reads and writes which are adjacent to each other may be merged for
-           // efficiency.  Thus two 4K reads may become one 8K read before it is
-           // ultimately handed to the disk, and so it will be counted (and queued)
-               mreads                  = strtoull(procfile_lineword(ff, l, 4), NULL, 10);      // rd_merges_or_rd_sec
-               mwrites                 = strtoull(procfile_lineword(ff, l, 8), NULL, 10);      // wr_merges
-
-               // # of sectors read # of sectors written
-               // This is the total number of sectors read or written successfully.
-               readsectors     = strtoull(procfile_lineword(ff, l, 5), NULL, 10);      // rd_sec_or_wr_ios
-               writesectors    = strtoull(procfile_lineword(ff, l, 9), NULL, 10);      // wr_sec
-
-               // # of milliseconds spent reading # of milliseconds spent writing
-               // This is the total number of milliseconds spent by all reads or writes (as
-               // measured from __make_request() to end_that_request_last()).
-               readms                  = strtoull(procfile_lineword(ff, l, 6), NULL, 10);      // rd_ticks_or_wr_sec
-               writems                 = strtoull(procfile_lineword(ff, l, 10), NULL, 10);     // wr_ticks
-
-               // # of I/Os currently in progress
-               // The only field that should go to zero. Incremented as requests are
-               // given to appropriate struct request_queue and decremented as they finish.
-               queued_ios              = strtoull(procfile_lineword(ff, l, 11), NULL, 10);     // ios_pgr
-
-               // # of milliseconds spent doing I/Os
-               // This field increases so long as field queued_ios is nonzero.
-               busy_ms                 = strtoull(procfile_lineword(ff, l, 12), NULL, 10);     // tot_ticks
-
-               // weighted # of milliseconds spent doing I/Os
-               // This field is incremented at each I/O start, I/O completion, I/O
-               // merge, or read of these stats by the number of I/Os in progress
-               // (field queued_ios) times the number of milliseconds spent doing I/O since the
-               // last update of this field.  This can provide an easy measure of both
-               // I/O completion time and the backlog that may be accumulating.
-               backlog_ms              = strtoull(procfile_lineword(ff, l, 13), NULL, 10);     // rq_ticks
-
-
-               // --------------------------------------------------------------------------
-               // remove slashes from disk names
-               char *s;
-               for(s = disk; *s ;s++)
-                       if(*s == '/') *s = '_';
-
-               // --------------------------------------------------------------------------
-               // get a disk structure for the disk
-
-               struct disk *d = get_disk(major, minor, disk);
-
-
-               // --------------------------------------------------------------------------
-               // Set its family based on mount point
-
-               char *family = d->mount_point;
-               if(!family) family = disk;
-
-
-               // --------------------------------------------------------------------------
-               // Check the configuration for the device
-
-               if(unlikely(!d->configured)) {
-                       char var_name[4096 + 1];
-                       snprintfz(var_name, 4096, "plugin:proc:/proc/diskstats:%s", disk);
-
-                       int def_enable = config_get_boolean_ondemand(var_name, "enable", global_enable_new_disks_detected_at_runtime);
-                       if(def_enable == CONFIG_ONDEMAND_NO) {
-                               // the user does not want any metrics for this disk
-                               d->do_io = CONFIG_ONDEMAND_NO;
-                               d->do_ops = CONFIG_ONDEMAND_NO;
-                               d->do_mops = CONFIG_ONDEMAND_NO;
-                               d->do_iotime = CONFIG_ONDEMAND_NO;
-                               d->do_qops = CONFIG_ONDEMAND_NO;
-                               d->do_util = CONFIG_ONDEMAND_NO;
-                               d->do_backlog = CONFIG_ONDEMAND_NO;
-                               d->do_space = CONFIG_ONDEMAND_NO;
-                               d->do_inodes = CONFIG_ONDEMAND_NO;
-                       }
-                       else {
-                               // this disk is enabled
-                               // check its direct settings
-
-                               int def_performance = CONFIG_ONDEMAND_ONDEMAND;
-                               int def_space = (d->mount_point)?CONFIG_ONDEMAND_ONDEMAND:CONFIG_ONDEMAND_NO;
-
-                               // since this is 'on demand' we can figure the performance settings
-                               // based on the type of disk
-
-                               switch(d->type) {
-                                       case DISK_TYPE_PHYSICAL:
-                                               def_performance = global_enable_performance_for_physical_disks;
-                                               break;
-
-                                       case DISK_TYPE_PARTITION:
-                                               def_performance = global_enable_performance_for_partitions;
-                                               break;
-
-                                       case DISK_TYPE_CONTAINER:
-                                               def_performance = global_enable_performance_for_virtual_disks;
-
-                                               if(d->mount_point)
-                                                       def_performance = select_positive_option(def_performance, global_enable_performance_for_virtual_mountpoints);
-
-                                               break;
-                               }
-
-                               if(d->mount_point)
-                                       def_performance = select_positive_option(def_performance, global_enable_performance_for_mountpoints);
-
-                               // ------------------------------------------------------------
-                               // now we have def_performance and def_space
-                               // to work further
-
-                               // def_performance
-                               // check the user configuration (this will also show our 'on demand' decision)
-                               def_performance = config_get_boolean_ondemand(var_name, "enable performance metrics", def_performance);
-
-                               int ddo_io = CONFIG_ONDEMAND_NO,
-                                       ddo_ops = CONFIG_ONDEMAND_NO,
-                                       ddo_mops = CONFIG_ONDEMAND_NO,
-                                       ddo_iotime = CONFIG_ONDEMAND_NO,
-                                       ddo_qops = CONFIG_ONDEMAND_NO,
-                                       ddo_util = CONFIG_ONDEMAND_NO,
-                                       ddo_backlog = CONFIG_ONDEMAND_NO;
-
-                               // we enable individual performance charts only when def_performance is not disabled
-                               if(def_performance != CONFIG_ONDEMAND_NO) {
-                                       ddo_io = global_do_io,
-                                       ddo_ops = global_do_ops,
-                                       ddo_mops = global_do_mops,
-                                       ddo_iotime = global_do_iotime,
-                                       ddo_qops = global_do_qops,
-                                       ddo_util = global_do_util,
-                                       ddo_backlog = global_do_backlog;
-                               }
-
-                               d->do_io      = config_get_boolean_ondemand(var_name, "bandwidth", ddo_io);
-                               d->do_ops     = config_get_boolean_ondemand(var_name, "operations", ddo_ops);
-                               d->do_mops    = config_get_boolean_ondemand(var_name, "merged operations", ddo_mops);
-                               d->do_iotime  = config_get_boolean_ondemand(var_name, "i/o time", ddo_iotime);
-                               d->do_qops    = config_get_boolean_ondemand(var_name, "queued operations", ddo_qops);
-                               d->do_util    = config_get_boolean_ondemand(var_name, "utilization percentage", ddo_util);
-                               d->do_backlog = config_get_boolean_ondemand(var_name, "backlog", ddo_backlog);
-
-                               // def_space
-                               if(d->mount_point) {
-                                       // check the user configuration (this will also show our 'on demand' decision)
-                                       def_space = config_get_boolean_ondemand(var_name, "enable space metrics", def_space);
-
-                                       int ddo_space = def_space,
-                                               ddo_inodes = def_space;
-
-                                       d->do_space = config_get_boolean_ondemand(var_name, "space usage", ddo_space);
-                                       d->do_inodes = config_get_boolean_ondemand(var_name, "inodes usage", ddo_inodes);
-                               }
-                               else {
-                                       // don't show settings for this disk
-                                       d->do_space = CONFIG_ONDEMAND_NO;
-                                       d->do_inodes = CONFIG_ONDEMAND_NO;
-                               }
-                       }
-
-                       d->configured = 1;
-               }
-
-               RRDSET *st;
-
-               // --------------------------------------------------------------------------
-               // Do performance metrics
-
-               if(d->do_io == CONFIG_ONDEMAND_YES || (d->do_io == CONFIG_ONDEMAND_ONDEMAND && (readsectors || writesectors))) {
-                       d->do_io = CONFIG_ONDEMAND_YES;
-
-                       st = rrdset_find_bytype(RRD_TYPE_DISK, disk);
-                       if(!st) {
-                               st = rrdset_create(RRD_TYPE_DISK, disk, NULL, family, "disk.io", "Disk I/O Bandwidth", "kilobytes/s", 2000, update_every, RRDSET_TYPE_AREA);
-
-                               rrddim_add(st, "reads", NULL, d->sector_size, 1024, RRDDIM_INCREMENTAL);
-                               rrddim_add(st, "writes", NULL, d->sector_size * -1, 1024, RRDDIM_INCREMENTAL);
-                       }
-                       else rrdset_next_usec(st, dt);
-
-                       last_readsectors  = rrddim_set(st, "reads", readsectors);
-                       last_writesectors = rrddim_set(st, "writes", writesectors);
-                       rrdset_done(st);
-               }
-
-               // --------------------------------------------------------------------
-
-               if(d->do_ops == CONFIG_ONDEMAND_YES || (d->do_ops == CONFIG_ONDEMAND_ONDEMAND && (reads || writes))) {
-                       d->do_ops = CONFIG_ONDEMAND_YES;
-
-                       st = rrdset_find_bytype("disk_ops", disk);
-                       if(!st) {
-                               st = rrdset_create("disk_ops", disk, NULL, family, "disk.ops", "Disk Completed I/O Operations", "operations/s", 2001, update_every, RRDSET_TYPE_LINE);
-                               st->isdetail = 1;
-
-                               rrddim_add(st, "reads", NULL, 1, 1, RRDDIM_INCREMENTAL);
-                               rrddim_add(st, "writes", NULL, -1, 1, RRDDIM_INCREMENTAL);
-                       }
-                       else rrdset_next_usec(st, dt);
-
-                       last_reads  = rrddim_set(st, "reads", reads);
-                       last_writes = rrddim_set(st, "writes", writes);
-                       rrdset_done(st);
-               }
-
-               // --------------------------------------------------------------------
-
-               if(d->do_qops == CONFIG_ONDEMAND_YES || (d->do_qops == CONFIG_ONDEMAND_ONDEMAND && queued_ios)) {
-                       d->do_qops = CONFIG_ONDEMAND_YES;
-
-                       st = rrdset_find_bytype("disk_qops", disk);
-                       if(!st) {
-                               st = rrdset_create("disk_qops", disk, NULL, family, "disk.qops", "Disk Current I/O Operations", "operations", 2002, update_every, RRDSET_TYPE_LINE);
-                               st->isdetail = 1;
-
-                               rrddim_add(st, "operations", NULL, 1, 1, RRDDIM_ABSOLUTE);
-                       }
-                       else rrdset_next_usec(st, dt);
-
-                       rrddim_set(st, "operations", queued_ios);
-                       rrdset_done(st);
-               }
-
-               // --------------------------------------------------------------------
-
-               if(d->do_backlog == CONFIG_ONDEMAND_YES || (d->do_backlog == CONFIG_ONDEMAND_ONDEMAND && backlog_ms)) {
-                       d->do_backlog = CONFIG_ONDEMAND_YES;
-
-                       st = rrdset_find_bytype("disk_backlog", disk);
-                       if(!st) {
-                               st = rrdset_create("disk_backlog", disk, NULL, family, "disk.backlog", "Disk Backlog", "backlog (ms)", 2003, update_every, RRDSET_TYPE_AREA);
-                               st->isdetail = 1;
-
-                               rrddim_add(st, "backlog", NULL, 1, 10, RRDDIM_INCREMENTAL);
-                       }
-                       else rrdset_next_usec(st, dt);
-
-                       rrddim_set(st, "backlog", backlog_ms);
-                       rrdset_done(st);
-               }
-
-               // --------------------------------------------------------------------
-
-               if(d->do_util == CONFIG_ONDEMAND_YES || (d->do_util == CONFIG_ONDEMAND_ONDEMAND && busy_ms)) {
-                       d->do_util = CONFIG_ONDEMAND_YES;
-
-                       st = rrdset_find_bytype("disk_util", disk);
-                       if(!st) {
-                               st = rrdset_create("disk_util", disk, NULL, family, "disk.util", "Disk Utilization Time", "% of time working", 2004, update_every, RRDSET_TYPE_AREA);
-                               st->isdetail = 1;
-
-                               rrddim_add(st, "utilization", NULL, 1, 10, RRDDIM_INCREMENTAL);
-                       }
-                       else rrdset_next_usec(st, dt);
-
-                       last_busy_ms = rrddim_set(st, "utilization", busy_ms);
-                       rrdset_done(st);
-               }
-
-               // --------------------------------------------------------------------
-
-               if(d->do_mops == CONFIG_ONDEMAND_YES || (d->do_mops == CONFIG_ONDEMAND_ONDEMAND && (mreads || mwrites))) {
-                       d->do_mops = CONFIG_ONDEMAND_YES;
-
-                       st = rrdset_find_bytype("disk_mops", disk);
-                       if(!st) {
-                               st = rrdset_create("disk_mops", disk, NULL, family, "disk.mops", "Disk Merged Operations", "merged operations/s", 2021, update_every, RRDSET_TYPE_LINE);
-                               st->isdetail = 1;
-
-                               rrddim_add(st, "reads", NULL, 1, 1, RRDDIM_INCREMENTAL);
-                               rrddim_add(st, "writes", NULL, -1, 1, RRDDIM_INCREMENTAL);
-                       }
-                       else rrdset_next_usec(st, dt);
-
-                       rrddim_set(st, "reads", mreads);
-                       rrddim_set(st, "writes", mwrites);
-                       rrdset_done(st);
-               }
-
-               // --------------------------------------------------------------------
-
-               if(d->do_iotime == CONFIG_ONDEMAND_YES || (d->do_iotime == CONFIG_ONDEMAND_ONDEMAND && (readms || writems))) {
-                       d->do_iotime = CONFIG_ONDEMAND_YES;
-
-                       st = rrdset_find_bytype("disk_iotime", disk);
-                       if(!st) {
-                               st = rrdset_create("disk_iotime", disk, NULL, family, "disk.iotime", "Disk Total I/O Time", "milliseconds/s", 2022, update_every, RRDSET_TYPE_LINE);
-                               st->isdetail = 1;
-
-                               rrddim_add(st, "reads", NULL, 1, 1, RRDDIM_INCREMENTAL);
-                               rrddim_add(st, "writes", NULL, -1, 1, RRDDIM_INCREMENTAL);
-                       }
-                       else rrdset_next_usec(st, dt);
-
-                       last_readms  = rrddim_set(st, "reads", readms);
-                       last_writems = rrddim_set(st, "writes", writems);
-                       rrdset_done(st);
-               }
-
-               // --------------------------------------------------------------------
-               // calculate differential charts
-               // only if this is not the first time we run
-
-               if(dt) {
-                       if( (d->do_iotime == CONFIG_ONDEMAND_YES || (d->do_iotime == CONFIG_ONDEMAND_ONDEMAND && (readms || writems))) &&
-                               (d->do_ops    == CONFIG_ONDEMAND_YES || (d->do_ops    == CONFIG_ONDEMAND_ONDEMAND && (reads || writes)))) {
-                               st = rrdset_find_bytype("disk_await", disk);
-                               if(!st) {
-                                       st = rrdset_create("disk_await", disk, NULL, family, "disk.await", "Average Completed I/O Operation Time", "ms per operation", 2005, update_every, RRDSET_TYPE_LINE);
-                                       st->isdetail = 1;
-
-                                       rrddim_add(st, "reads", NULL, 1, 1, RRDDIM_ABSOLUTE);
-                                       rrddim_add(st, "writes", NULL, -1, 1, RRDDIM_ABSOLUTE);
-                               }
-                               else rrdset_next_usec(st, dt);
-
-                               rrddim_set(st, "reads", (reads - last_reads) ? (readms - last_readms) / (reads - last_reads) : 0);
-                               rrddim_set(st, "writes", (writes - last_writes) ? (writems - last_writems) / (writes - last_writes) : 0);
-                               rrdset_done(st);
-                       }
-
-                       if( (d->do_io  == CONFIG_ONDEMAND_YES || (d->do_io  == CONFIG_ONDEMAND_ONDEMAND && (readsectors || writesectors))) &&
-                               (d->do_ops == CONFIG_ONDEMAND_YES || (d->do_ops == CONFIG_ONDEMAND_ONDEMAND && (reads || writes)))) {
-                               st = rrdset_find_bytype("disk_avgsz", disk);
-                               if(!st) {
-                                       st = rrdset_create("disk_avgsz", disk, NULL, family, "disk.avgsz", "Average Completed I/O Operation Bandwidth", "kilobytes per operation", 2006, update_every, RRDSET_TYPE_AREA);
-                                       st->isdetail = 1;
-
-                                       rrddim_add(st, "reads", NULL, d->sector_size, 1024, RRDDIM_ABSOLUTE);
-                                       rrddim_add(st, "writes", NULL, d->sector_size * -1, 1024, RRDDIM_ABSOLUTE);
-                               }
-                               else rrdset_next_usec(st, dt);
-
-                               rrddim_set(st, "reads", (reads - last_reads) ? (readsectors - last_readsectors) / (reads - last_reads) : 0);
-                               rrddim_set(st, "writes", (writes - last_writes) ? (writesectors - last_writesectors) / (writes - last_writes) : 0);
-                               rrdset_done(st);
-                       }
-
-                       if( (d->do_util == CONFIG_ONDEMAND_YES || (d->do_util == CONFIG_ONDEMAND_ONDEMAND && busy_ms)) &&
-                               (d->do_ops  == CONFIG_ONDEMAND_YES || (d->do_ops  == CONFIG_ONDEMAND_ONDEMAND && (reads || writes)))) {
-                               st = rrdset_find_bytype("disk_svctm", disk);
-                               if(!st) {
-                                       st = rrdset_create("disk_svctm", disk, NULL, family, "disk.svctm", "Average Service Time", "ms per operation", 2007, update_every, RRDSET_TYPE_LINE);
-                                       st->isdetail = 1;
-
-                                       rrddim_add(st, "svctm", NULL, 1, 1, RRDDIM_ABSOLUTE);
-                               }
-                               else rrdset_next_usec(st, dt);
-
-                               rrddim_set(st, "svctm", ((reads - last_reads) + (writes - last_writes)) ? (busy_ms - last_busy_ms) / ((reads - last_reads) + (writes - last_writes)) : 0);
-                               rrdset_done(st);
-                       }
-               }
-
-               // --------------------------------------------------------------------------
-               // space metrics
-
-               if(d->mount_point && (d->do_space || d->do_inodes) ) {
-                       // collect space metrics using statvfs
-
-                       if (statvfs(d->mount_point, &buff_statvfs) < 0)
-                               error("Failed statvfs() for '%s' (disk '%s')", d->mount_point, d->disk);
-                       else {
-                               space_avail = buff_statvfs.f_bavail * buff_statvfs.f_bsize;
-                               space_avail_root = (buff_statvfs.f_bfree - buff_statvfs.f_bavail) * buff_statvfs.f_bsize;
-                               space_used = (buff_statvfs.f_blocks - buff_statvfs.f_bfree) * buff_statvfs.f_bsize;
-
-                               inodes_avail = buff_statvfs.f_favail;
-                               inodes_avail_root = buff_statvfs.f_ffree - buff_statvfs.f_favail;
-                               inodes_used = buff_statvfs.f_files - buff_statvfs.f_ffree;
-
-                               // verify we collected the metrics for the right disk.
-                               // if not the mountpoint has changed.
-
-                               if(stat(d->mount_point, &buff_stat) == -1)
-                                       error("Failed to stat() for '%s' (disk '%s')", d->mount_point, d->disk);
-                               else {
-                                       if(major(buff_stat.st_dev) == major && minor(buff_stat.st_dev) == minor) {
-
-                                               // --------------------------------------------------------------------------
-
-                                               if(d->do_space == CONFIG_ONDEMAND_YES || (d->do_space == CONFIG_ONDEMAND_ONDEMAND && (space_avail || space_avail_root || space_used))) {
-                                                       st = rrdset_find_bytype("disk_space", disk);
-                                                       if(!st) {
-                                                               st = rrdset_create("disk_space", disk, NULL, family, "disk.space", "Disk Space Usage", "GB", 2023, update_every, RRDSET_TYPE_STACKED);
-                                                               st->isdetail = 1;
-
-                                                               rrddim_add(st, "avail", NULL, 1, 1024*1024*1024, RRDDIM_ABSOLUTE);
-                                                               rrddim_add(st, "reserved_for_root", "reserved for root", 1, 1024*1024*1024, RRDDIM_ABSOLUTE);
-                                                               rrddim_add(st, "used" , NULL, 1, 1024*1024*1024, RRDDIM_ABSOLUTE);
-                                                       }
-                                                       else rrdset_next_usec(st, dt);
-
-                                                       rrddim_set(st, "avail", space_avail);
-                                                       rrddim_set(st, "reserved_for_root", space_avail_root);
-                                                       rrddim_set(st, "used", space_used);
-                                                       rrdset_done(st);
-                                               }
-
-                                               // --------------------------------------------------------------------------
-
-                                               if(d->do_inodes == CONFIG_ONDEMAND_YES || (d->do_inodes == CONFIG_ONDEMAND_ONDEMAND && (inodes_avail || inodes_avail_root || inodes_used))) {
-                                                       st = rrdset_find_bytype("disk_inodes", disk);
-                                                       if(!st) {
-                                                               st = rrdset_create("disk_inodes", disk, NULL, family, "disk.inodes", "Disk Inodes Usage", "Inodes", 2024, update_every, RRDSET_TYPE_STACKED);
-                                                               st->isdetail = 1;
-
-                                                               rrddim_add(st, "avail", NULL, 1, 1, RRDDIM_ABSOLUTE);
-                                                               rrddim_add(st, "reserved_for_root", "reserved for root", 1, 1, RRDDIM_ABSOLUTE);
-                                                               rrddim_add(st, "used" , NULL, 1, 1, RRDDIM_ABSOLUTE);
-                                                       }
-                                                       else rrdset_next_usec(st, dt);
-
-                                                       rrddim_set(st, "avail", inodes_avail);
-                                                       rrddim_set(st, "reserved_for_root", inodes_avail_root);
-                                                       rrddim_set(st, "used", inodes_used);
-                                                       rrdset_done(st);
-                                               }
-                                       }
-                               }
-                       }
-               }
-       }
-
-       return 0;
+int do_proc_diskstats(int update_every, usec_t dt) {
+    static procfile *ff = NULL;
+    static int  global_enable_new_disks_detected_at_runtime = CONFIG_BOOLEAN_YES,
+                global_enable_performance_for_physical_disks = CONFIG_BOOLEAN_AUTO,
+                global_enable_performance_for_virtual_disks = CONFIG_BOOLEAN_AUTO,
+                global_enable_performance_for_partitions = CONFIG_BOOLEAN_NO,
+                global_do_io = CONFIG_BOOLEAN_AUTO,
+                global_do_ops = CONFIG_BOOLEAN_AUTO,
+                global_do_mops = CONFIG_BOOLEAN_AUTO,
+                global_do_iotime = CONFIG_BOOLEAN_AUTO,
+                global_do_qops = CONFIG_BOOLEAN_AUTO,
+                global_do_util = CONFIG_BOOLEAN_AUTO,
+                global_do_backlog = CONFIG_BOOLEAN_AUTO,
+                globals_initialized = 0;
+
+    if(unlikely(!globals_initialized)) {
+        global_enable_new_disks_detected_at_runtime = config_get_boolean(CONFIG_SECTION_DISKSTATS, "enable new disks detected at runtime", global_enable_new_disks_detected_at_runtime);
+        global_enable_performance_for_physical_disks = config_get_boolean_ondemand(CONFIG_SECTION_DISKSTATS, "performance metrics for physical disks", global_enable_performance_for_physical_disks);
+        global_enable_performance_for_virtual_disks = config_get_boolean_ondemand(CONFIG_SECTION_DISKSTATS, "performance metrics for virtual disks", global_enable_performance_for_virtual_disks);
+        global_enable_performance_for_partitions = config_get_boolean_ondemand(CONFIG_SECTION_DISKSTATS, "performance metrics for partitions", global_enable_performance_for_partitions);
+
+        global_do_io      = config_get_boolean_ondemand(CONFIG_SECTION_DISKSTATS, "bandwidth for all disks", global_do_io);
+        global_do_ops     = config_get_boolean_ondemand(CONFIG_SECTION_DISKSTATS, "operations for all disks", global_do_ops);
+        global_do_mops    = config_get_boolean_ondemand(CONFIG_SECTION_DISKSTATS, "merged operations for all disks", global_do_mops);
+        global_do_iotime  = config_get_boolean_ondemand(CONFIG_SECTION_DISKSTATS, "i/o time for all disks", global_do_iotime);
+        global_do_qops    = config_get_boolean_ondemand(CONFIG_SECTION_DISKSTATS, "queued operations for all disks", global_do_qops);
+        global_do_util    = config_get_boolean_ondemand(CONFIG_SECTION_DISKSTATS, "utilization percentage for all disks", global_do_util);
+        global_do_backlog = config_get_boolean_ondemand(CONFIG_SECTION_DISKSTATS, "backlog for all disks", global_do_backlog);
+
+        globals_initialized = 1;
+    }
+
+    // --------------------------------------------------------------------------
+
+    if(unlikely(!ff)) {
+        char filename[FILENAME_MAX + 1];
+        snprintfz(filename, FILENAME_MAX, "%s%s", netdata_configured_host_prefix, "/proc/diskstats");
+        ff = procfile_open(config_get(CONFIG_SECTION_DISKSTATS, "filename to monitor", filename), " \t", PROCFILE_FLAG_DEFAULT);
+    }
+    if(unlikely(!ff)) return 0;
+
+    ff = procfile_readall(ff);
+    if(unlikely(!ff)) return 0; // we return 0, so that we will retry to open it next time
+
+    size_t lines = procfile_lines(ff), l;
+
+    for(l = 0; l < lines ;l++) {
+        // --------------------------------------------------------------------------
+        // Read parameters
+
+        char *disk;
+        unsigned long       major = 0, minor = 0;
+
+        collected_number    reads = 0,  mreads = 0,  readsectors = 0,  readms = 0,
+                            writes = 0, mwrites = 0, writesectors = 0, writems = 0,
+                            queued_ios = 0, busy_ms = 0, backlog_ms = 0;
+
+        collected_number    last_reads = 0,  last_readsectors = 0,  last_readms = 0,
+                            last_writes = 0, last_writesectors = 0, last_writems = 0,
+                            last_busy_ms = 0;
+
+        size_t words = procfile_linewords(ff, l);
+        if(unlikely(words < 14)) continue;
+
+        major           = str2ul(procfile_lineword(ff, l, 0));
+        minor           = str2ul(procfile_lineword(ff, l, 1));
+        disk            = procfile_lineword(ff, l, 2);
+
+        // # of reads completed # of writes completed
+        // This is the total number of reads or writes completed successfully.
+        reads           = str2ull(procfile_lineword(ff, l, 3));  // rd_ios
+        writes          = str2ull(procfile_lineword(ff, l, 7));  // wr_ios
+
+        // # of reads merged # of writes merged
+        // Reads and writes which are adjacent to each other may be merged for
+        // efficiency.  Thus two 4K reads may become one 8K read before it is
+        // ultimately handed to the disk, and so it will be counted (and queued)
+        mreads          = str2ull(procfile_lineword(ff, l, 4));  // rd_merges_or_rd_sec
+        mwrites         = str2ull(procfile_lineword(ff, l, 8));  // wr_merges
+
+        // # of sectors read # of sectors written
+        // This is the total number of sectors read or written successfully.
+        readsectors     = str2ull(procfile_lineword(ff, l, 5));  // rd_sec_or_wr_ios
+        writesectors    = str2ull(procfile_lineword(ff, l, 9));  // wr_sec
+
+        // # of milliseconds spent reading # of milliseconds spent writing
+        // This is the total number of milliseconds spent by all reads or writes (as
+        // measured from __make_request() to end_that_request_last()).
+        readms          = str2ull(procfile_lineword(ff, l, 6));  // rd_ticks_or_wr_sec
+        writems         = str2ull(procfile_lineword(ff, l, 10)); // wr_ticks
+
+        // # of I/Os currently in progress
+        // The only field that should go to zero. Incremented as requests are
+        // given to appropriate struct request_queue and decremented as they finish.
+        queued_ios      = str2ull(procfile_lineword(ff, l, 11)); // ios_pgr
+
+        // # of milliseconds spent doing I/Os
+        // This field increases so long as field queued_ios is nonzero.
+        busy_ms         = str2ull(procfile_lineword(ff, l, 12)); // tot_ticks
+
+        // weighted # of milliseconds spent doing I/Os
+        // This field is incremented at each I/O start, I/O completion, I/O
+        // merge, or read of these stats by the number of I/Os in progress
+        // (field queued_ios) times the number of milliseconds spent doing I/O since the
+        // last update of this field.  This can provide an easy measure of both
+        // I/O completion time and the backlog that may be accumulating.
+        backlog_ms      = str2ull(procfile_lineword(ff, l, 13)); // rq_ticks
+
+
+        // --------------------------------------------------------------------------
+        // remove slashes from disk names
+        char *s;
+        for(s = disk; *s ;s++)
+            if(*s == '/') *s = '_';
+
+        // --------------------------------------------------------------------------
+        // get a disk structure for the disk
+
+        struct disk *d = get_disk(major, minor, disk);
+        d->updated = 1;
+
+        // --------------------------------------------------------------------------
+        // Set its family based on mount point
+
+        char *family = d->mount_point;
+        if(!family) family = disk;
+
+
+        // --------------------------------------------------------------------------
+        // Check the configuration for the device
+
+        if(unlikely(!d->configured)) {
+            d->configured = 1;
+
+            static SIMPLE_PATTERN *excluded_disks = NULL;
+
+            if(unlikely(!excluded_disks)) {
+                excluded_disks = simple_pattern_create(
+                        config_get(CONFIG_SECTION_DISKSTATS, "exclude disks", DELAULT_EXLUDED_DISKS),
+                        SIMPLE_PATTERN_EXACT
+                );
+            }
+
+            int def_enable = global_enable_new_disks_detected_at_runtime;
+
+            if(def_enable != CONFIG_BOOLEAN_NO && simple_pattern_matches(excluded_disks, disk))
+                def_enable = CONFIG_BOOLEAN_NO;
+
+            char var_name[4096 + 1];
+            snprintfz(var_name, 4096, "plugin:proc:/proc/diskstats:%s", disk);
+
+            def_enable = config_get_boolean_ondemand(var_name, "enable", def_enable);
+            if(unlikely(def_enable == CONFIG_BOOLEAN_NO)) {
+                // the user does not want any metrics for this disk
+                d->do_io = CONFIG_BOOLEAN_NO;
+                d->do_ops = CONFIG_BOOLEAN_NO;
+                d->do_mops = CONFIG_BOOLEAN_NO;
+                d->do_iotime = CONFIG_BOOLEAN_NO;
+                d->do_qops = CONFIG_BOOLEAN_NO;
+                d->do_util = CONFIG_BOOLEAN_NO;
+                d->do_backlog = CONFIG_BOOLEAN_NO;
+            }
+            else {
+                // this disk is enabled
+                // check its direct settings
+
+                int def_performance = CONFIG_BOOLEAN_AUTO;
+
+                // since this is 'on demand' we can figure the performance settings
+                // based on the type of disk
+
+                switch(d->type) {
+                    case DISK_TYPE_PHYSICAL:
+                        def_performance = global_enable_performance_for_physical_disks;
+                        break;
+
+                    case DISK_TYPE_PARTITION:
+                        def_performance = global_enable_performance_for_partitions;
+                        break;
+
+                    case DISK_TYPE_CONTAINER:
+                        def_performance = global_enable_performance_for_virtual_disks;
+                        break;
+                }
+
+                // check if we have to disable performance for this disk
+                if(def_performance)
+                    def_performance = is_major_enabled((int)major);
+
+                // ------------------------------------------------------------
+                // now we have def_performance and def_space
+                // to work further
+
+                // def_performance
+                // check the user configuration (this will also show our 'on demand' decision)
+                def_performance = config_get_boolean_ondemand(var_name, "enable performance metrics", def_performance);
+
+                int ddo_io = CONFIG_BOOLEAN_NO,
+                    ddo_ops = CONFIG_BOOLEAN_NO,
+                    ddo_mops = CONFIG_BOOLEAN_NO,
+                    ddo_iotime = CONFIG_BOOLEAN_NO,
+                    ddo_qops = CONFIG_BOOLEAN_NO,
+                    ddo_util = CONFIG_BOOLEAN_NO,
+                    ddo_backlog = CONFIG_BOOLEAN_NO;
+
+                // we enable individual performance charts only when def_performance is not disabled
+                if(unlikely(def_performance != CONFIG_BOOLEAN_NO)) {
+                    ddo_io = global_do_io,
+                    ddo_ops = global_do_ops,
+                    ddo_mops = global_do_mops,
+                    ddo_iotime = global_do_iotime,
+                    ddo_qops = global_do_qops,
+                    ddo_util = global_do_util,
+                    ddo_backlog = global_do_backlog;
+                }
+
+                d->do_io      = config_get_boolean_ondemand(var_name, "bandwidth", ddo_io);
+                d->do_ops     = config_get_boolean_ondemand(var_name, "operations", ddo_ops);
+                d->do_mops    = config_get_boolean_ondemand(var_name, "merged operations", ddo_mops);
+                d->do_iotime  = config_get_boolean_ondemand(var_name, "i/o time", ddo_iotime);
+                d->do_qops    = config_get_boolean_ondemand(var_name, "queued operations", ddo_qops);
+                d->do_util    = config_get_boolean_ondemand(var_name, "utilization percentage", ddo_util);
+                d->do_backlog = config_get_boolean_ondemand(var_name, "backlog", ddo_backlog);
+            }
+        }
+
+        // --------------------------------------------------------------------------
+        // Do performance metrics
+
+        if(d->do_io == CONFIG_BOOLEAN_YES || (d->do_io == CONFIG_BOOLEAN_AUTO && (readsectors || writesectors))) {
+            d->do_io = CONFIG_BOOLEAN_YES;
+
+            if(unlikely(!d->st_io)) {
+                d->st_io = rrdset_create_localhost(
+                        RRD_TYPE_DISK
+                        , disk
+                        , NULL
+                        , family
+                        , "disk.io"
+                        , "Disk I/O Bandwidth"
+                        , "kilobytes/s"
+                        , 2000
+                        , update_every
+                        , RRDSET_TYPE_AREA
+                );
+
+                rrddim_add(d->st_io, "reads", NULL, d->sector_size, 1024, RRD_ALGORITHM_INCREMENTAL);
+                rrddim_add(d->st_io, "writes", NULL, d->sector_size * -1, 1024, RRD_ALGORITHM_INCREMENTAL);
+            }
+            else rrdset_next(d->st_io);
+
+            last_readsectors  = rrddim_set(d->st_io, "reads", readsectors);
+            last_writesectors = rrddim_set(d->st_io, "writes", writesectors);
+            rrdset_done(d->st_io);
+        }
+
+        // --------------------------------------------------------------------
+
+        if(d->do_ops == CONFIG_BOOLEAN_YES || (d->do_ops == CONFIG_BOOLEAN_AUTO && (reads || writes))) {
+            d->do_ops = CONFIG_BOOLEAN_YES;
+
+            if(unlikely(!d->st_ops)) {
+                d->st_ops = rrdset_create_localhost(
+                        "disk_ops"
+                        , disk
+                        , NULL
+                        , family
+                        , "disk.ops"
+                        , "Disk Completed I/O Operations"
+                        , "operations/s"
+                        , 2001
+                        , update_every
+                        , RRDSET_TYPE_LINE
+                );
+
+                rrdset_flag_set(d->st_ops, RRDSET_FLAG_DETAIL);
+
+                rrddim_add(d->st_ops, "reads", NULL, 1, 1, RRD_ALGORITHM_INCREMENTAL);
+                rrddim_add(d->st_ops, "writes", NULL, -1, 1, RRD_ALGORITHM_INCREMENTAL);
+            }
+            else rrdset_next(d->st_ops);
+
+            last_reads  = rrddim_set(d->st_ops, "reads", reads);
+            last_writes = rrddim_set(d->st_ops, "writes", writes);
+            rrdset_done(d->st_ops);
+        }
+
+        // --------------------------------------------------------------------
+
+        if(d->do_qops == CONFIG_BOOLEAN_YES || (d->do_qops == CONFIG_BOOLEAN_AUTO && queued_ios)) {
+            d->do_qops = CONFIG_BOOLEAN_YES;
+
+            if(unlikely(!d->st_qops)) {
+                d->st_qops = rrdset_create_localhost(
+                        "disk_qops"
+                        , disk
+                        , NULL
+                        , family
+                        , "disk.qops"
+                        , "Disk Current I/O Operations"
+                        , "operations"
+                        , 2002
+                        , update_every
+                        , RRDSET_TYPE_LINE
+                );
+
+                rrdset_flag_set(d->st_qops, RRDSET_FLAG_DETAIL);
+
+                rrddim_add(d->st_qops, "operations", NULL, 1, 1, RRD_ALGORITHM_ABSOLUTE);
+            }
+            else rrdset_next(d->st_qops);
+
+            rrddim_set(d->st_qops, "operations", queued_ios);
+            rrdset_done(d->st_qops);
+        }
+
+        // --------------------------------------------------------------------
+
+        if(d->do_backlog == CONFIG_BOOLEAN_YES || (d->do_backlog == CONFIG_BOOLEAN_AUTO && backlog_ms)) {
+            d->do_backlog = CONFIG_BOOLEAN_YES;
+
+            if(unlikely(!d->st_backlog)) {
+                d->st_backlog = rrdset_create_localhost(
+                        "disk_backlog"
+                        , disk
+                        , NULL
+                        , family
+                        , "disk.backlog"
+                        , "Disk Backlog"
+                        , "backlog (ms)"
+                        , 2003
+                        , update_every
+                        , RRDSET_TYPE_AREA
+                );
+
+                rrdset_flag_set(d->st_backlog, RRDSET_FLAG_DETAIL);
+
+                rrddim_add(d->st_backlog, "backlog", NULL, 1, 10, RRD_ALGORITHM_INCREMENTAL);
+            }
+            else rrdset_next(d->st_backlog);
+
+            rrddim_set(d->st_backlog, "backlog", backlog_ms);
+            rrdset_done(d->st_backlog);
+        }
+
+        // --------------------------------------------------------------------
+
+        if(d->do_util == CONFIG_BOOLEAN_YES || (d->do_util == CONFIG_BOOLEAN_AUTO && busy_ms)) {
+            d->do_util = CONFIG_BOOLEAN_YES;
+
+            if(unlikely(!d->st_util)) {
+                d->st_util = rrdset_create_localhost(
+                        "disk_util"
+                        , disk
+                        , NULL
+                        , family
+                        , "disk.util"
+                        , "Disk Utilization Time"
+                        , "% of time working"
+                        , 2004
+                        , update_every
+                        , RRDSET_TYPE_AREA
+                );
+
+                rrdset_flag_set(d->st_util, RRDSET_FLAG_DETAIL);
+
+                rrddim_add(d->st_util, "utilization", NULL, 1, 10, RRD_ALGORITHM_INCREMENTAL);
+            }
+            else rrdset_next(d->st_util);
+
+            last_busy_ms = rrddim_set(d->st_util, "utilization", busy_ms);
+            rrdset_done(d->st_util);
+        }
+
+        // --------------------------------------------------------------------
+
+        if(d->do_mops == CONFIG_BOOLEAN_YES || (d->do_mops == CONFIG_BOOLEAN_AUTO && (mreads || mwrites))) {
+            d->do_mops = CONFIG_BOOLEAN_YES;
+
+            if(unlikely(!d->st_mops)) {
+                d->st_mops = rrdset_create_localhost(
+                        "disk_mops"
+                        , disk
+                        , NULL
+                        , family
+                        , "disk.mops"
+                        , "Disk Merged Operations"
+                        , "merged operations/s"
+                        , 2021
+                        , update_every
+                        , RRDSET_TYPE_LINE
+                );
+
+                rrdset_flag_set(d->st_mops, RRDSET_FLAG_DETAIL);
+
+                rrddim_add(d->st_mops, "reads", NULL, 1, 1, RRD_ALGORITHM_INCREMENTAL);
+                rrddim_add(d->st_mops, "writes", NULL, -1, 1, RRD_ALGORITHM_INCREMENTAL);
+            }
+            else rrdset_next(d->st_mops);
+
+            rrddim_set(d->st_mops, "reads", mreads);
+            rrddim_set(d->st_mops, "writes", mwrites);
+            rrdset_done(d->st_mops);
+        }
+
+        // --------------------------------------------------------------------
+
+        if(d->do_iotime == CONFIG_BOOLEAN_YES || (d->do_iotime == CONFIG_BOOLEAN_AUTO && (readms || writems))) {
+            d->do_iotime = CONFIG_BOOLEAN_YES;
+
+            if(unlikely(!d->st_iotime)) {
+                d->st_iotime = rrdset_create_localhost(
+                        "disk_iotime"
+                        , disk
+                        , NULL
+                        , family
+                        , "disk.iotime"
+                        , "Disk Total I/O Time"
+                        , "milliseconds/s"
+                        , 2022
+                        , update_every
+                        , RRDSET_TYPE_LINE
+                );
+
+                rrdset_flag_set(d->st_iotime, RRDSET_FLAG_DETAIL);
+
+                rrddim_add(d->st_iotime, "reads", NULL, 1, 1, RRD_ALGORITHM_INCREMENTAL);
+                rrddim_add(d->st_iotime, "writes", NULL, -1, 1, RRD_ALGORITHM_INCREMENTAL);
+            }
+            else rrdset_next(d->st_iotime);
+
+            last_readms  = rrddim_set(d->st_iotime, "reads", readms);
+            last_writems = rrddim_set(d->st_iotime, "writes", writems);
+            rrdset_done(d->st_iotime);
+        }
+
+        // --------------------------------------------------------------------
+        // calculate differential charts
+        // only if this is not the first time we run
+
+        if(likely(dt)) {
+            if( (d->do_iotime == CONFIG_BOOLEAN_YES || (d->do_iotime == CONFIG_BOOLEAN_AUTO && (readms || writems))) &&
+                (d->do_ops    == CONFIG_BOOLEAN_YES || (d->do_ops    == CONFIG_BOOLEAN_AUTO && (reads || writes)))) {
+
+                if(unlikely(!d->st_await)) {
+                    d->st_await = rrdset_create_localhost(
+                            "disk_await"
+                            , disk
+                            , NULL
+                            , family
+                            , "disk.await"
+                            , "Average Completed I/O Operation Time"
+                            , "ms per operation"
+                            , 2005
+                            , update_every
+                            , RRDSET_TYPE_LINE
+                    );
+
+                    rrdset_flag_set(d->st_await, RRDSET_FLAG_DETAIL);
+
+                    rrddim_add(d->st_await, "reads", NULL, 1, 1, RRD_ALGORITHM_ABSOLUTE);
+                    rrddim_add(d->st_await, "writes", NULL, -1, 1, RRD_ALGORITHM_ABSOLUTE);
+                }
+                else rrdset_next(d->st_await);
+
+                rrddim_set(d->st_await, "reads", (reads - last_reads) ? (readms - last_readms) / (reads - last_reads) : 0);
+                rrddim_set(d->st_await, "writes", (writes - last_writes) ? (writems - last_writems) / (writes - last_writes) : 0);
+                rrdset_done(d->st_await);
+            }
+
+            if( (d->do_io  == CONFIG_BOOLEAN_YES || (d->do_io  == CONFIG_BOOLEAN_AUTO && (readsectors || writesectors))) &&
+                (d->do_ops == CONFIG_BOOLEAN_YES || (d->do_ops == CONFIG_BOOLEAN_AUTO && (reads || writes)))) {
+
+                if(unlikely(!d->st_avgsz)) {
+                    d->st_avgsz = rrdset_create_localhost(
+                            "disk_avgsz"
+                            , disk
+                            , NULL
+                            , family
+                            , "disk.avgsz"
+                            , "Average Completed I/O Operation Bandwidth"
+                            , "kilobytes per operation"
+                            , 2006
+                            , update_every
+                            , RRDSET_TYPE_AREA
+                    );
+
+                    rrdset_flag_set(d->st_avgsz, RRDSET_FLAG_DETAIL);
+
+                    rrddim_add(d->st_avgsz, "reads", NULL, d->sector_size, 1024, RRD_ALGORITHM_ABSOLUTE);
+                    rrddim_add(d->st_avgsz, "writes", NULL, d->sector_size * -1, 1024, RRD_ALGORITHM_ABSOLUTE);
+                }
+                else rrdset_next(d->st_avgsz);
+
+                rrddim_set(d->st_avgsz, "reads", (reads - last_reads) ? (readsectors - last_readsectors) / (reads - last_reads) : 0);
+                rrddim_set(d->st_avgsz, "writes", (writes - last_writes) ? (writesectors - last_writesectors) / (writes - last_writes) : 0);
+                rrdset_done(d->st_avgsz);
+            }
+
+            if( (d->do_util == CONFIG_BOOLEAN_YES || (d->do_util == CONFIG_BOOLEAN_AUTO && busy_ms)) &&
+                (d->do_ops  == CONFIG_BOOLEAN_YES || (d->do_ops  == CONFIG_BOOLEAN_AUTO && (reads || writes)))) {
+
+                if(unlikely(!d->st_svctm)) {
+                    d->st_svctm = rrdset_create_localhost(
+                            "disk_svctm"
+                            , disk
+                            , NULL
+                            , family
+                            , "disk.svctm"
+                            , "Average Service Time"
+                            , "ms per operation"
+                            , 2007
+                            , update_every
+                            , RRDSET_TYPE_LINE
+                    );
+
+                    rrdset_flag_set(d->st_svctm, RRDSET_FLAG_DETAIL);
+
+                    rrddim_add(d->st_svctm, "svctm", NULL, 1, 1, RRD_ALGORITHM_ABSOLUTE);
+                }
+                else rrdset_next(d->st_svctm);
+
+                rrddim_set(d->st_svctm, "svctm", ((reads - last_reads) + (writes - last_writes)) ? (busy_ms - last_busy_ms) / ((reads - last_reads) + (writes - last_writes)) : 0);
+                rrdset_done(d->st_svctm);
+            }
+        }
+    }
+
+    // cleanup removed disks
+
+    struct disk *d = disk_root, *last = NULL;
+    while(d) {
+        if(unlikely(!d->updated)) {
+            struct disk *t = d;
+
+            rrdset_obsolete_and_pointer_null(d->st_avgsz);
+            rrdset_obsolete_and_pointer_null(d->st_await);
+            rrdset_obsolete_and_pointer_null(d->st_backlog);
+            rrdset_obsolete_and_pointer_null(d->st_io);
+            rrdset_obsolete_and_pointer_null(d->st_iotime);
+            rrdset_obsolete_and_pointer_null(d->st_mops);
+            rrdset_obsolete_and_pointer_null(d->st_ops);
+            rrdset_obsolete_and_pointer_null(d->st_qops);
+            rrdset_obsolete_and_pointer_null(d->st_svctm);
+            rrdset_obsolete_and_pointer_null(d->st_util);
+
+            if(d == disk_root) {
+                disk_root = d = d->next;
+                last = NULL;
+            }
+            else if(last) {
+                last->next = d = d->next;
+            }
+
+            freez(t->disk);
+            freez(t->mount_point);
+            freez(t);
+        }
+        else {
+            d->updated = 0;
+            last = d;
+            d = d->next;
+        }
+    }
+
+    return 0;
 }