hyperic-sigar 1.7.0

data/src/os/hpux/sigar_os.h

@@ -0,0 +1,51 @@
+/*
+ * Copyright (C) [2004, 2005, 2006], Hyperic, Inc.
+ * This file is part of SIGAR.
+ * 
+ * SIGAR is free software; you can redistribute it and/or modify
+ * it under the terms version 2 of the GNU General Public License as
+ * published by the Free Software Foundation. This program is distributed
+ * in the hope that it will be useful, but WITHOUT ANY WARRANTY; without
+ * even the implied warranty of MERCHANTABILITY or FITNESS FOR A
+ * PARTICULAR PURPOSE. See the GNU General Public License for more
+ * details.
+ * 
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307
+ * USA.
+ */
+
+#ifndef SIGAR_OS_H
+#define SIGAR_OS_H
+
+#if defined(__ia64) && !defined(__ia64__)
+#define __ia64__
+#endif
+
+#ifdef __ia64__
+#ifndef _LP64
+#define _LP64
+#endif
+#endif
+
+#define _PSTAT64
+
+#include <sys/pstat.h>
+#include <sys/mib.h>
+#include <stdlib.h>
+#include <fcntl.h>
+
+struct sigar_t {
+    SIGAR_T_BASE;
+    struct pst_static pstatic;
+    time_t last_getprocs;
+    sigar_pid_t last_pid;
+    struct pst_status *pinfo;
+
+    int mib;
+};
+
+int hpux_get_mib_ifentry(int ppa, mib_ifEntry *mib);
+
+#endif /* SIGAR_OS_H */

data/src/os/linux/linux_sigar.c

@@ -0,0 +1,2595 @@
+/*
+ * Copyright (C) [2004, 2005, 2006], Hyperic, Inc.
+ * This file is part of SIGAR.
+ * 
+ * SIGAR is free software; you can redistribute it and/or modify
+ * it under the terms version 2 of the GNU General Public License as
+ * published by the Free Software Foundation. This program is distributed
+ * in the hope that it will be useful, but WITHOUT ANY WARRANTY; without
+ * even the implied warranty of MERCHANTABILITY or FITNESS FOR A
+ * PARTICULAR PURPOSE. See the GNU General Public License for more
+ * details.
+ * 
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307
+ * USA.
+ */
+
+#include <dirent.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <errno.h>
+#include <sys/param.h>
+#include <sys/stat.h>
+#include <sys/times.h>
+#include <sys/utsname.h>
+
+#include "sigar.h"
+#include "sigar_private.h"
+#include "sigar_util.h"
+#include "sigar_os.h"
+
+#define pageshift(x) ((x) << sigar->pagesize)
+
+#define PROC_MEMINFO PROC_FS_ROOT "meminfo"
+#define PROC_VMSTAT  PROC_FS_ROOT "vmstat"
+#define PROC_MTRR    PROC_FS_ROOT "mtrr"
+#define PROC_STAT    PROC_FS_ROOT "stat"
+#define PROC_UPTIME  PROC_FS_ROOT "uptime"
+#define PROC_LOADAVG PROC_FS_ROOT "loadavg"
+
+#define PROC_PSTAT   "/stat"
+#define PROC_PSTATUS "/status"
+
+#define SYS_BLOCK "/sys/block"
+#define PROC_PARTITIONS PROC_FS_ROOT "partitions"
+#define PROC_DISKSTATS  PROC_FS_ROOT "diskstats"
+
+/*
+ * /proc/self/stat fields:
+ * 1 - pid
+ * 2 - comm
+ * 3 - state
+ * 4 - ppid
+ * 5 - pgrp
+ * 6 - session
+ * 7 - tty_nr
+ * 8 - tpgid
+ * 9 - flags
+ * 10 - minflt
+ * 11 - cminflt
+ * 12 - majflt
+ * 13 - cmajflt
+ * 14 - utime
+ * 15 - stime
+ * 16 - cutime
+ * 17 - cstime
+ * 18 - priority
+ * 19 - nice
+ * 20 - 0 (removed field)
+ * 21 - itrealvalue
+ * 22 - starttime
+ * 23 - vsize
+ * 24 - rss
+ * 25 - rlim
+ * 26 - startcode
+ * 27 - endcode
+ * 28 - startstack
+ * 29 - kstkesp
+ * 30 - kstkeip
+ * 31 - signal
+ * 32 - blocked
+ * 33 - sigignore
+ * 34 - sigcache
+ * 35 - wchan
+ * 36 - nswap
+ * 37 - cnswap
+ * 38 - exit_signal <-- looking for this.
+ * 39 - processor
+ * ... more for newer RH
+ */
+
+#define PROC_SIGNAL_IX 38
+
+static int get_proc_signal_offset(void)
+{
+    char buffer[BUFSIZ], *ptr=buffer;
+    int fields = 0;
+    int status = sigar_file2str(PROCP_FS_ROOT "self/stat",
+                                buffer, sizeof(buffer));
+
+    if (status != SIGAR_OK) {
+        return 1;
+    }
+
+    while (*ptr) {
+        if (*ptr++ == ' ') {
+            fields++;
+        }
+    }
+
+    return (fields - PROC_SIGNAL_IX) + 1;
+}
+
+sigar_pid_t sigar_pid_get(sigar_t *sigar)
+{
+    /* XXX cannot safely cache getpid unless using nptl */
+    /* we can however, cache it for optimizations in the
+     * case of proc_env_get for example.
+     */
+    sigar->pid = getpid();
+    return sigar->pid;
+}
+
+static int sigar_boot_time_get(sigar_t *sigar)
+{
+    FILE *fp;
+    char buffer[BUFSIZ], *ptr;
+    int found = 0;
+
+    if (!(fp = fopen(PROC_STAT, "r"))) {
+        return errno;
+    }
+
+    while ((ptr = fgets(buffer, sizeof(buffer), fp))) {
+        if (strnEQ(ptr, "btime", 5)) {
+            if ((ptr = sigar_skip_token(ptr))) {
+                sigar->boot_time = sigar_strtoul(ptr);
+                found = 1;
+            }
+            break;
+        }
+    }
+
+    fclose(fp);
+
+    if (!found) {
+        /* should never happen */
+        sigar->boot_time = time(NULL);
+    }
+
+    return SIGAR_OK;
+}
+
+int sigar_os_open(sigar_t **sigar)
+{
+    int i, status;
+    int kernel_rev, has_nptl;
+    struct stat sb;
+    struct utsname name;
+
+    *sigar = malloc(sizeof(**sigar));
+
+    (*sigar)->pagesize = 0;
+    i = getpagesize();
+    while ((i >>= 1) > 0) {
+        (*sigar)->pagesize++;
+    }
+
+    status = sigar_boot_time_get(*sigar);
+    if (status != SIGAR_OK) {
+        return status;
+    }
+    
+    (*sigar)->ticks = sysconf(_SC_CLK_TCK);
+
+    (*sigar)->ram = -1;
+
+    (*sigar)->proc_signal_offset = -1;
+
+    (*sigar)->last_proc_stat.pid = -1;
+
+    (*sigar)->lcpu = -1;
+
+    if (stat(PROC_DISKSTATS, &sb) == 0) {
+        (*sigar)->iostat = IOSTAT_DISKSTATS;
+    }
+    else if (stat(SYS_BLOCK, &sb) == 0) {
+        (*sigar)->iostat = IOSTAT_SYS;
+    }
+    else if (stat(PROC_PARTITIONS, &sb) == 0) {
+        /* XXX file exists does not mean is has the fields */
+        (*sigar)->iostat = IOSTAT_PARTITIONS;
+    }
+    else {
+        (*sigar)->iostat = IOSTAT_NONE;
+    }
+
+    /* hook for using mirrored /proc/net/tcp file */
+    (*sigar)->proc_net = getenv("SIGAR_PROC_NET");
+
+    uname(&name);
+    /* 2.X.y.z -> just need X (unless there is ever a kernel version 3!) */
+    kernel_rev = atoi(&name.release[2]);
+    if (kernel_rev >= 6) {
+        has_nptl = 1;
+    }
+    else {
+        has_nptl = getenv("SIGAR_HAS_NPTL") ? 1 : 0;
+    }
+    (*sigar)->has_nptl = has_nptl;
+
+    return SIGAR_OK;
+}
+
+int sigar_os_close(sigar_t *sigar)
+{
+    free(sigar);
+    return SIGAR_OK;
+}
+
+char *sigar_os_error_string(sigar_t *sigar, int err)
+{
+    return NULL;
+}
+
+static int sigar_cpu_total_count(sigar_t *sigar)
+{
+    sigar->ncpu = (int)sysconf(_SC_NPROCESSORS_CONF);
+    sigar_log_printf(sigar, SIGAR_LOG_DEBUG, "[cpu] ncpu=%d\n",
+                     sigar->ncpu);
+    return sigar->ncpu;
+}
+
+static int get_ram(sigar_t *sigar, sigar_mem_t *mem)
+{
+    char buffer[BUFSIZ], *ptr;
+    FILE *fp;
+    int total = 0;
+    sigar_uint64_t sys_total = (mem->total / (1024 * 1024));
+
+    if (sigar->ram > 0) {
+        /* return cached value */
+        mem->ram = sigar->ram;
+        return SIGAR_OK;
+    }
+
+    if (sigar->ram == 0) {
+        return ENOENT;
+    }
+
+    /*
+     * Memory Type Range Registers
+     * write-back registers add up to the total.
+     * Well, they are supposed to add up, but seen
+     * at least one configuration where that is not the
+     * case.
+     */
+    if (!(fp = fopen(PROC_MTRR, "r"))) {
+        return errno;
+    }
+
+    while ((ptr = fgets(buffer, sizeof(buffer), fp))) {
+        if (!(ptr = strstr(ptr, "size="))) {
+            continue;
+        }
+
+        if (!strstr(ptr, "write-back")) {
+            continue;
+        }
+
+        ptr += 5;
+        while (sigar_isspace(*ptr)) {
+            ++ptr;
+        }
+
+        total += atoi(ptr);
+    }
+
+    fclose(fp);
+
+    if ((total - sys_total) > 256) {
+        /* mtrr write-back registers are way off
+         * kernel should not be using more that 256MB of mem
+         */
+        total = 0; /* punt */
+    }
+
+    if (total == 0) {
+        return ENOENT;
+    }
+ 
+    mem->ram = sigar->ram = total;
+
+    return SIGAR_OK;
+}
+
+#define MEMINFO_PARAM(a) a ":", SSTRLEN(a ":")
+
+static SIGAR_INLINE sigar_uint64_t sigar_meminfo(char *buffer,
+                                                 char *attr, int len)
+{
+    sigar_uint64_t val = 0;
+    char *ptr, *tok;
+
+    if ((ptr = strstr(buffer, attr))) {
+        ptr += len;
+        val = strtoull(ptr, &tok, 0);
+        while (*tok == ' ') {
+            ++tok;
+        }
+        if (*tok == 'k') {
+            val *= 1024;
+        }
+        else if (*tok == 'M') {
+            val *= (1024 * 1024);
+        }
+    }
+
+    return val;
+}
+
+int sigar_mem_get(sigar_t *sigar, sigar_mem_t *mem)
+{
+    sigar_uint64_t buffers, cached, kern;
+    char buffer[BUFSIZ];
+
+    int status = sigar_file2str(PROC_MEMINFO,
+                                buffer, sizeof(buffer));
+
+    if (status != SIGAR_OK) {
+        return status;
+    }
+
+    mem->total  = sigar_meminfo(buffer, MEMINFO_PARAM("MemTotal"));
+    mem->free   = sigar_meminfo(buffer, MEMINFO_PARAM("MemFree"));
+    mem->used   = mem->total - mem->free;
+
+    buffers = sigar_meminfo(buffer, MEMINFO_PARAM("Buffers"));
+    cached  = sigar_meminfo(buffer, MEMINFO_PARAM("Cached"));
+
+    kern = buffers + cached;
+    mem->actual_free = mem->free + kern;
+    mem->actual_used = mem->used - kern;
+
+    sigar_mem_calc_ram(sigar, mem);
+
+    if (get_ram(sigar, mem) != SIGAR_OK) {
+        /* XXX other options on failure? */
+    }
+
+    return SIGAR_OK;
+}
+
+int sigar_swap_get(sigar_t *sigar, sigar_swap_t *swap)
+{
+    char buffer[BUFSIZ], *ptr;
+
+    /* XXX: we open/parse the same file here as sigar_mem_get */
+    int status = sigar_file2str(PROC_MEMINFO,
+                                buffer, sizeof(buffer));
+
+    if (status != SIGAR_OK) {
+        return status;
+    }
+
+    swap->total  = sigar_meminfo(buffer, MEMINFO_PARAM("SwapTotal"));
+    swap->free   = sigar_meminfo(buffer, MEMINFO_PARAM("SwapFree"));
+    swap->used   = swap->total - swap->free;
+
+    swap->page_in = swap->page_out = -1;
+
+    status = sigar_file2str(PROC_VMSTAT,
+                            buffer, sizeof(buffer));
+
+    if (status == SIGAR_OK) {
+        /* 2.6+ kernel */
+        if ((ptr = strstr(buffer, "\npswpin"))) {
+            ptr = sigar_skip_token(ptr);
+            swap->page_in = sigar_strtoull(ptr);
+            ptr = sigar_skip_token(ptr);
+            swap->page_out = sigar_strtoull(ptr);
+        }
+    }
+    else {
+        /* 2.2, 2.4 kernels */
+        status = sigar_file2str(PROC_STAT,
+                                buffer, sizeof(buffer));
+        if (status != SIGAR_OK) {
+            return status;
+        }
+
+        if ((ptr = strstr(buffer, "\nswap"))) {
+            ptr = sigar_skip_token(ptr);
+            swap->page_in = sigar_strtoull(ptr);
+            swap->page_out = sigar_strtoull(ptr);
+        }
+    }
+
+    return SIGAR_OK;
+}
+
+static void get_cpu_metrics(sigar_t *sigar, sigar_cpu_t *cpu, char *line)
+{
+    char *ptr = sigar_skip_token(line); /* "cpu%d" */
+
+    cpu->user += SIGAR_TICK2MSEC(sigar_strtoull(ptr));
+    cpu->nice += SIGAR_TICK2MSEC(sigar_strtoull(ptr));
+    cpu->sys  += SIGAR_TICK2MSEC(sigar_strtoull(ptr));
+    cpu->idle += SIGAR_TICK2MSEC(sigar_strtoull(ptr));
+    if (*ptr == ' ') {
+        /* 2.6+ kernels only */
+        cpu->wait += SIGAR_TICK2MSEC(sigar_strtoull(ptr));
+        cpu->irq += SIGAR_TICK2MSEC(sigar_strtoull(ptr));
+        cpu->soft_irq += SIGAR_TICK2MSEC(sigar_strtoull(ptr));
+    }
+    if (*ptr == ' ') {
+        /* 2.6.11+ kernels only */
+        cpu->stolen += SIGAR_TICK2MSEC(sigar_strtoull(ptr));
+    }
+    cpu->total =
+        cpu->user + cpu->nice + cpu->sys + cpu->idle +
+        cpu->wait + cpu->irq + cpu->soft_irq + cpu->stolen;
+}
+
+int sigar_cpu_get(sigar_t *sigar, sigar_cpu_t *cpu)
+{
+    char buffer[BUFSIZ];
+    int status = sigar_file2str(PROC_STAT, buffer, sizeof(buffer));
+
+    if (status != SIGAR_OK) {
+        return status;
+    }
+
+    SIGAR_ZERO(cpu);
+    get_cpu_metrics(sigar, cpu, buffer);
+
+    return SIGAR_OK;
+}
+
+int sigar_cpu_list_get(sigar_t *sigar, sigar_cpu_list_t *cpulist)
+{
+    FILE *fp;
+    char buffer[BUFSIZ], cpu_total[BUFSIZ], *ptr;
+    int core_rollup = sigar_cpu_core_rollup(sigar), i=0;
+    sigar_cpu_t *cpu;
+
+    if (!(fp = fopen(PROC_STAT, "r"))) {
+        return errno;
+    }
+
+    /* skip first line */
+    (void)fgets(cpu_total, sizeof(cpu_total), fp);
+
+    sigar_cpu_list_create(cpulist);
+
+    /* XXX: merge times of logical processors if hyperthreading */
+    while ((ptr = fgets(buffer, sizeof(buffer), fp))) {
+        if (!strnEQ(ptr, "cpu", 3)) {
+            break;
+        }
+
+        if (core_rollup && (i % sigar->lcpu)) {
+            /* merge times of logical processors */
+            cpu = &cpulist->data[cpulist->number-1];
+        }
+        else {
+            SIGAR_CPU_LIST_GROW(cpulist);
+            cpu = &cpulist->data[cpulist->number++];
+            SIGAR_ZERO(cpu);
+        }
+
+        get_cpu_metrics(sigar, cpu, ptr);
+
+        i++;
+    }
+
+    fclose(fp);
+
+    if (cpulist->number == 0) {
+        /* likely older kernel where cpu\d is not present */
+        cpu = &cpulist->data[cpulist->number++];
+        SIGAR_ZERO(cpu);
+        get_cpu_metrics(sigar, cpu, cpu_total);
+    }
+
+    return SIGAR_OK;
+}
+
+int sigar_uptime_get(sigar_t *sigar,
+                     sigar_uptime_t *uptime)
+{
+    char buffer[BUFSIZ], *ptr = buffer;
+    int status = sigar_file2str(PROC_UPTIME, buffer, sizeof(buffer));
+
+    if (status != SIGAR_OK) {
+        return status;
+    }
+
+    uptime->uptime   = strtod(buffer, &ptr);
+
+    return SIGAR_OK;
+}
+
+int sigar_loadavg_get(sigar_t *sigar,
+                      sigar_loadavg_t *loadavg)
+{
+    char buffer[BUFSIZ], *ptr = buffer;
+    int status = sigar_file2str(PROC_LOADAVG, buffer, sizeof(buffer));
+
+    if (status != SIGAR_OK) {
+        return status;
+    }
+
+    loadavg->loadavg[0] = strtod(buffer, &ptr);
+    loadavg->loadavg[1] = strtod(ptr, &ptr);
+    loadavg->loadavg[2] = strtod(ptr, &ptr);
+
+    return SIGAR_OK;
+}
+
+/*
+ * seems the easiest/fastest way to tell if a process listed in /proc
+ * is a thread is to check the "exit signal" flag in /proc/num/stat.
+ * any value other than SIGCHLD seems to be a thread.  this make hulk mad.
+ * redhat's procps patch (named "threadbadhack.pat") does not use
+ * this flag to filter out threads.  instead does much more expensive
+ * comparisions.  their patch also bubbles up thread cpu times to the main
+ * process.  functionality we currently lack.
+ * when nptl is in use, this is not the case and all threads spawned from
+ * a process have the same pid.  however, it seems both old-style linux
+ * threads and nptl threads can be run on the same machine.
+ * there is also the "Tgid" field in /proc/self/status which could be used
+ * to detect threads, but this is not available in older kernels.
+ */
+static SIGAR_INLINE int proc_isthread(sigar_t *sigar, char *pidstr, int len)
+{
+    char buffer[BUFSIZ], *ptr=buffer;
+    int fd, n, offset=sigar->proc_signal_offset;
+
+    /* sprintf(buffer, "/proc/%s/stat", pidstr) */
+    memcpy(ptr, PROCP_FS_ROOT, SSTRLEN(PROCP_FS_ROOT));
+    ptr += SSTRLEN(PROCP_FS_ROOT);
+
+    memcpy(ptr, pidstr, len);
+    ptr += len;
+
+    memcpy(ptr, PROC_PSTAT, SSTRLEN(PROC_PSTAT));
+    ptr += SSTRLEN(PROC_PSTAT);
+
+    *ptr = '\0';
+
+    if ((fd = open(buffer, O_RDONLY)) < 0) {
+        /* unlikely if pid was from readdir proc */
+        return 0;
+    }
+
+    n = read(fd, buffer, sizeof(buffer));
+    close(fd);
+
+    if (n < 0) {
+        return 0; /* chances: slim..none */
+    }
+
+    buffer[n--] = '\0';
+
+    /* exit_signal is the second to last field so we look backwards.
+     * XXX if newer kernels drop more turds in this file we'll need
+     * to go the other way.  luckily linux has no real api for this shit.
+     */
+
+    /* skip trailing crap */
+    while ((n > 0) && !isdigit(buffer[n--])) ;
+
+    while (offset-- > 0) {
+        /* skip last field */
+        while ((n > 0) && isdigit(buffer[n--])) ;
+
+        /* skip whitespace */
+        while ((n > 0) && !isdigit(buffer[n--])) ;
+    }
+
+    if (n < 3) {
+        return 0; /* hulk smashed /proc? */
+    }
+
+    ptr = &buffer[n];
+    /*
+     * '17' == SIGCHLD == real process.
+     * '33' and '0' are threads
+     */
+    if ((*ptr++ == '1') &&
+        (*ptr++ == '7') &&
+        (*ptr++ == ' '))
+    {
+        return 0;
+    }
+
+    return 1;
+}
+
+int sigar_os_proc_list_get(sigar_t *sigar,
+                           sigar_proc_list_t *proclist)
+{
+    DIR *dirp = opendir(PROCP_FS_ROOT);
+    struct dirent *ent, dbuf;
+    register const int threadbadhack = !sigar->has_nptl;
+
+    if (!dirp) {
+        return errno;
+    }
+
+    if (threadbadhack && (sigar->proc_signal_offset == -1)) {
+        sigar->proc_signal_offset = get_proc_signal_offset();
+    }
+
+    while (readdir_r(dirp, &dbuf, &ent) == 0) {
+        if (!ent) {
+            break;
+        }
+
+        if (!sigar_isdigit(*ent->d_name)) {
+            continue;
+        }
+
+        if (threadbadhack &&
+            proc_isthread(sigar, ent->d_name, strlen(ent->d_name)))
+        {
+            continue;
+        }
+
+        /* XXX: more sanity checking */
+
+        SIGAR_PROC_LIST_GROW(proclist);
+
+        proclist->data[proclist->number++] =
+            strtoul(ent->d_name, NULL, 10);
+    }
+
+    closedir(dirp);
+
+    return SIGAR_OK;
+}
+
+static int proc_stat_read(sigar_t *sigar, sigar_pid_t pid)
+{
+    char buffer[BUFSIZ], *ptr=buffer, *tmp;
+    unsigned int len;
+    linux_proc_stat_t *pstat = &sigar->last_proc_stat;
+    int status;
+
+    time_t timenow = time(NULL);
+
+    /* 
+     * short-lived cache read/parse of last /proc/pid/stat
+     * as this info is spread out across a few functions.
+     */
+    if (pstat->pid == pid) {
+        if ((timenow - pstat->mtime) < SIGAR_LAST_PROC_EXPIRE) {
+            return SIGAR_OK;
+        }
+    }
+
+    pstat->pid = pid;
+    pstat->mtime = timenow;
+
+    status = SIGAR_PROC_FILE2STR(buffer, pid, PROC_PSTAT);
+
+    if (status != SIGAR_OK) {
+        return status;
+    }
+
+    if (!(ptr = strchr(ptr, '('))) {
+        return EINVAL;
+    }
+    if (!(tmp = strrchr(++ptr, ')'))) {
+        return EINVAL;
+    }
+    len = tmp-ptr;
+
+    if (len >= sizeof(pstat->name)) {
+        len = sizeof(pstat->name)-1;
+    }
+
+    /* (1,2) */
+    memcpy(pstat->name, ptr, len);
+    pstat->name[len] = '\0';
+    ptr = tmp+1;
+
+    SIGAR_SKIP_SPACE(ptr);
+    pstat->state = *ptr++; /* (3) */
+    SIGAR_SKIP_SPACE(ptr);
+
+    pstat->ppid = sigar_strtoul(ptr); /* (4) */
+    ptr = sigar_skip_token(ptr); /* (5) pgrp */
+    ptr = sigar_skip_token(ptr); /* (6) session */
+    pstat->tty = sigar_strtoul(ptr); /* (7) */
+    ptr = sigar_skip_token(ptr); /* (8) tty pgrp */
+
+    ptr = sigar_skip_token(ptr); /* (9) flags */
+    pstat->minor_faults = sigar_strtoull(ptr); /* (10) */
+    ptr = sigar_skip_token(ptr); /* (11) cmin flt */
+    pstat->major_faults = sigar_strtoull(ptr); /* (12) */
+    ptr = sigar_skip_token(ptr); /* (13) cmaj flt */
+
+    pstat->utime = SIGAR_TICK2MSEC(sigar_strtoull(ptr)); /* (14) */
+    pstat->stime = SIGAR_TICK2MSEC(sigar_strtoull(ptr)); /* (15) */
+
+    ptr = sigar_skip_token(ptr); /* (16) cutime */
+    ptr = sigar_skip_token(ptr); /* (17) cstime */
+
+    pstat->priority = sigar_strtoul(ptr); /* (18) */
+    pstat->nice     = sigar_strtoul(ptr); /* (19) */
+
+    ptr = sigar_skip_token(ptr); /* (20) timeout */
+    ptr = sigar_skip_token(ptr); /* (21) it_real_value */
+
+    pstat->start_time  = sigar_strtoul(ptr); /* (22) */
+    pstat->start_time /= sigar->ticks;
+    pstat->start_time += sigar->boot_time; /* seconds */
+    pstat->start_time *= 1000; /* milliseconds */
+
+    pstat->vsize = sigar_strtoull(ptr); /* (23) */
+    pstat->rss   = pageshift(sigar_strtoull(ptr)); /* (24) */
+
+    ptr = sigar_skip_token(ptr); /* (25) rlim */
+    ptr = sigar_skip_token(ptr); /* (26) startcode */
+    ptr = sigar_skip_token(ptr); /* (27) endcode */
+    ptr = sigar_skip_token(ptr); /* (28) startstack */
+    ptr = sigar_skip_token(ptr); /* (29) kstkesp */
+    ptr = sigar_skip_token(ptr); /* (30) kstkeip */
+    ptr = sigar_skip_token(ptr); /* (31) signal */
+    ptr = sigar_skip_token(ptr); /* (32) blocked */
+    ptr = sigar_skip_token(ptr); /* (33) sigignore */
+    ptr = sigar_skip_token(ptr); /* (34) sigcache */
+    ptr = sigar_skip_token(ptr); /* (35) wchan */
+    ptr = sigar_skip_token(ptr); /* (36) nswap */
+    ptr = sigar_skip_token(ptr); /* (37) cnswap */
+    ptr = sigar_skip_token(ptr); /* (38) exit_signal */
+
+    pstat->processor = sigar_strtoul(ptr); /* (39) */
+
+    return SIGAR_OK;
+}
+
+int sigar_proc_mem_get(sigar_t *sigar, sigar_pid_t pid,
+                       sigar_proc_mem_t *procmem)
+{
+    char buffer[BUFSIZ], *ptr=buffer;
+    int status = proc_stat_read(sigar, pid);
+    linux_proc_stat_t *pstat = &sigar->last_proc_stat;
+
+    procmem->minor_faults = pstat->minor_faults;
+    procmem->major_faults = pstat->major_faults;
+    procmem->page_faults =
+        procmem->minor_faults + procmem->major_faults;
+    
+    status = SIGAR_PROC_FILE2STR(buffer, pid, "/statm");
+
+    if (status != SIGAR_OK) {
+        return status;
+    }
+
+    procmem->size     = pageshift(sigar_strtoull(ptr));
+    procmem->resident = pageshift(sigar_strtoull(ptr));
+    procmem->share    = pageshift(sigar_strtoull(ptr));
+
+    return SIGAR_OK;
+}
+
+#define NO_ID_MSG "[proc_cred] /proc/%lu" PROC_PSTATUS " missing "
+
+int sigar_proc_cred_get(sigar_t *sigar, sigar_pid_t pid,
+                        sigar_proc_cred_t *proccred)
+{
+    char buffer[BUFSIZ], *ptr;
+    int status = SIGAR_PROC_FILE2STR(buffer, pid, PROC_PSTATUS);
+
+    if (status != SIGAR_OK) {
+        return status;
+    }
+
+    if ((ptr = strstr(buffer, "\nUid:"))) {
+        ptr = sigar_skip_token(ptr);
+
+        proccred->uid  = sigar_strtoul(ptr);
+        proccred->euid = sigar_strtoul(ptr);
+    }
+    else {
+        sigar_log_printf(sigar, SIGAR_LOG_WARN,
+                         NO_ID_MSG "Uid", pid);
+        return ENOENT;
+    }
+
+    if ((ptr = strstr(ptr, "\nGid:"))) {
+        ptr = sigar_skip_token(ptr);
+
+        proccred->gid  = sigar_strtoul(ptr);
+        proccred->egid = sigar_strtoul(ptr);
+    }
+    else {
+        sigar_log_printf(sigar, SIGAR_LOG_WARN,
+                         NO_ID_MSG "Gid", pid);
+        return ENOENT;
+    }
+
+    return SIGAR_OK;
+}
+
+int sigar_proc_time_get(sigar_t *sigar, sigar_pid_t pid,
+                        sigar_proc_time_t *proctime)
+{
+    int status = proc_stat_read(sigar, pid);
+    linux_proc_stat_t *pstat = &sigar->last_proc_stat;
+
+    if (status != SIGAR_OK) {
+        return status;
+    }
+
+    proctime->user = pstat->utime;
+    proctime->sys  = pstat->stime;
+    proctime->total = proctime->user + proctime->sys;
+    proctime->start_time = pstat->start_time;
+
+    return SIGAR_OK;
+}
+
+static int proc_status_get(sigar_t *sigar, sigar_pid_t pid,
+                           sigar_proc_state_t *procstate)
+{
+    char buffer[BUFSIZ], *ptr;
+    int status = SIGAR_PROC_FILE2STR(buffer, pid, PROC_PSTATUS);
+
+    if (status != SIGAR_OK) {
+        return status;
+    }
+
+    ptr = strstr(buffer, "\nThreads:");
+    if (ptr) {
+        /* 2.6+ kernel only */
+        ptr = sigar_skip_token(ptr);
+        procstate->threads = sigar_strtoul(ptr);
+    }
+    else {
+        procstate->threads = SIGAR_FIELD_NOTIMPL;
+    }
+
+    return SIGAR_OK;
+}
+
+int sigar_proc_state_get(sigar_t *sigar, sigar_pid_t pid,
+                         sigar_proc_state_t *procstate)
+{
+    int status = proc_stat_read(sigar, pid);
+    linux_proc_stat_t *pstat = &sigar->last_proc_stat;
+
+    if (status != SIGAR_OK) {
+        return status;
+    }
+
+    memcpy(procstate->name, pstat->name, sizeof(procstate->name));
+    procstate->state = pstat->state;
+
+    procstate->ppid     = pstat->ppid;
+    procstate->tty      = pstat->tty;
+    procstate->priority = pstat->priority;
+    procstate->nice     = pstat->nice;
+    procstate->processor = pstat->processor;
+
+    if (sigar_cpu_core_rollup(sigar)) {
+        procstate->processor /= sigar->lcpu;
+    }
+
+    proc_status_get(sigar, pid, procstate);
+
+    return SIGAR_OK;
+}
+
+int sigar_os_proc_args_get(sigar_t *sigar, sigar_pid_t pid,
+                           sigar_proc_args_t *procargs)
+{
+    return sigar_procfs_args_get(sigar, pid, procargs);
+}
+
+/* glibc 2.8 XXX use sysconf(_SC_ARG_MAX) */
+#ifndef ARG_MAX
+#define ARG_MAX 131072
+#endif
+
+int sigar_proc_env_get(sigar_t *sigar, sigar_pid_t pid,
+                       sigar_proc_env_t *procenv)
+{
+    int fd;
+    char buffer[ARG_MAX]; /* XXX: ARG_MAX == 130k */
+    char name[BUFSIZ];
+    size_t len;
+    char *ptr, *end;
+
+    /* optimize if pid == $$ and type == ENV_KEY */
+    SIGAR_PROC_ENV_KEY_LOOKUP();
+
+    (void)SIGAR_PROC_FILENAME(name, pid, "/environ");
+
+    if ((fd = open(name, O_RDONLY)) < 0) {
+        if (errno == ENOENT) {
+            return ESRCH;
+        }
+        return errno;
+    }
+
+    len = read(fd, buffer, sizeof(buffer));
+
+    close(fd);
+
+    buffer[len] = '\0';
+    ptr = buffer;
+
+    end = buffer + len;
+    while (ptr < end) {
+        char *val = strchr(ptr, '=');
+        int klen, vlen, status;
+        char key[128]; /* XXX is there a max key size? */
+
+        if (val == NULL) {
+            /* not key=val format */
+            break;
+        }
+
+        klen = val - ptr;
+        SIGAR_SSTRCPY(key, ptr);
+        key[klen] = '\0';
+        ++val;
+
+        vlen = strlen(val);
+        status = procenv->env_getter(procenv->data,
+                                     key, klen, val, vlen);
+
+        if (status != SIGAR_OK) {
+            /* not an error; just stop iterating */
+            break;
+        }
+
+        ptr += (klen + 1 + vlen + 1);
+    }
+
+    return SIGAR_OK;
+}
+
+int sigar_proc_fd_get(sigar_t *sigar, sigar_pid_t pid,
+                      sigar_proc_fd_t *procfd)
+{
+    int status =
+        sigar_proc_fd_count(sigar, pid, &procfd->total);
+
+    return status;
+}
+
+int sigar_proc_exe_get(sigar_t *sigar, sigar_pid_t pid,
+                       sigar_proc_exe_t *procexe)
+{
+    int len;
+    char name[BUFSIZ];
+
+    (void)SIGAR_PROC_FILENAME(name, pid, "/cwd");
+
+    if ((len = readlink(name, procexe->cwd,
+                        sizeof(procexe->cwd)-1)) < 0)
+    {
+        return errno;
+    }
+
+    procexe->cwd[len] = '\0';
+
+    (void)SIGAR_PROC_FILENAME(name, pid, "/exe");
+
+    if ((len = readlink(name, procexe->name,
+                        sizeof(procexe->name)-1)) < 0)
+    {
+        return errno;
+    }
+
+    procexe->name[len] = '\0';
+
+    (void)SIGAR_PROC_FILENAME(name, pid, "/root");
+
+    if ((len = readlink(name, procexe->root,
+                        sizeof(procexe->root)-1)) < 0)
+    {
+        return errno;
+    }
+
+    procexe->root[len] = '\0';
+
+    return SIGAR_OK;
+}
+
+int sigar_proc_modules_get(sigar_t *sigar, sigar_pid_t pid,
+                           sigar_proc_modules_t *procmods)
+{
+    FILE *fp;
+    char buffer[BUFSIZ], *ptr;
+    unsigned long inode, last_inode = 0;
+
+    (void)SIGAR_PROC_FILENAME(buffer, pid, "/maps");
+
+    if (!(fp = fopen(buffer, "r"))) {
+        return errno;
+    }
+
+    while ((ptr = fgets(buffer, sizeof(buffer), fp))) {
+        int len, status;
+        /* skip region, flags, offset, dev */
+        ptr = sigar_skip_multiple_token(ptr, 4);
+        inode = sigar_strtoul(ptr);
+
+        if ((inode == 0) || (inode == last_inode)) {
+            last_inode = 0;
+            continue;
+        }
+
+        last_inode = inode;
+        SIGAR_SKIP_SPACE(ptr);
+        len = strlen(ptr);
+        ptr[len-1] = '\0'; /* chop \n */
+
+        status =
+            procmods->module_getter(procmods->data,
+                                    ptr, len-1);
+
+        if (status != SIGAR_OK) {
+            /* not an error; just stop iterating */
+            break;
+        }
+    }
+    
+    fclose(fp);
+
+    return SIGAR_OK;
+}
+
+int sigar_thread_cpu_get(sigar_t *sigar,
+                         sigar_uint64_t id,
+                         sigar_thread_cpu_t *cpu)
+{
+    struct tms now;
+
+    if (id != 0) {
+        return SIGAR_ENOTIMPL;
+    }
+
+    times(&now);
+
+    cpu->user  = SIGAR_TICK2NSEC(now.tms_utime);
+    cpu->sys   = SIGAR_TICK2NSEC(now.tms_stime);
+    cpu->total = SIGAR_TICK2NSEC(now.tms_utime + now.tms_stime);
+
+    return SIGAR_OK;
+}
+
+#include <mntent.h>
+
+int sigar_os_fs_type_get(sigar_file_system_t *fsp)
+{
+    char *type = fsp->sys_type_name;
+
+    switch (*type) {
+      case 'e':
+        if (strnEQ(type, "ext", 3)) {
+            fsp->type = SIGAR_FSTYPE_LOCAL_DISK;
+        }
+        break;
+      case 'g':
+        if (strEQ(type, "gfs")) {
+            fsp->type = SIGAR_FSTYPE_LOCAL_DISK;
+        }
+        break;
+      case 'h':
+        if (strEQ(type, "hpfs")) {
+            fsp->type = SIGAR_FSTYPE_LOCAL_DISK;
+        }
+        break;
+      case 'j':
+        if (strnEQ(type, "jfs", 3)) {
+            fsp->type = SIGAR_FSTYPE_LOCAL_DISK;
+        }
+        break;
+      case 'o':
+        if (strnEQ(type, "ocfs", 4)) {
+            fsp->type = SIGAR_FSTYPE_LOCAL_DISK;
+        }
+        break;
+      case 'p':
+        if (strnEQ(type, "psfs", 4)) {
+            fsp->type = SIGAR_FSTYPE_LOCAL_DISK;
+        }
+        break;
+      case 'r':
+        if (strEQ(type, "reiserfs")) {
+            fsp->type = SIGAR_FSTYPE_LOCAL_DISK;
+        }
+        break;
+      case 'v':
+        if (strEQ(type, "vzfs")) {
+            fsp->type = SIGAR_FSTYPE_LOCAL_DISK;
+        }
+        break;
+      case 'x':
+        if (strEQ(type, "xfs") || strEQ(type, "xiafs")) {
+            fsp->type = SIGAR_FSTYPE_LOCAL_DISK;
+        }
+        break;
+    }
+
+    return fsp->type;
+}
+
+int sigar_file_system_list_get(sigar_t *sigar,
+                               sigar_file_system_list_t *fslist)
+{
+    struct mntent ent;
+    char buf[1025]; /* buffer for strings within ent */
+    FILE *fp;
+    sigar_file_system_t *fsp;
+
+    if (!(fp = setmntent(MOUNTED, "r"))) {
+        return errno;
+    }
+
+    sigar_file_system_list_create(fslist);
+
+    while (getmntent_r(fp, &ent, buf, sizeof(buf))) {
+        SIGAR_FILE_SYSTEM_LIST_GROW(fslist);
+
+        fsp = &fslist->data[fslist->number++];
+
+        fsp->type = SIGAR_FSTYPE_UNKNOWN; /* unknown, will be set later */
+        SIGAR_SSTRCPY(fsp->dir_name, ent.mnt_dir);
+        SIGAR_SSTRCPY(fsp->dev_name, ent.mnt_fsname);
+        SIGAR_SSTRCPY(fsp->sys_type_name, ent.mnt_type);
+        SIGAR_SSTRCPY(fsp->options, ent.mnt_opts);
+        sigar_fs_type_get(fsp);
+    }
+
+    endmntent(fp);
+
+    return SIGAR_OK;
+}
+
+#define ST_MAJOR(sb) major((sb).st_rdev)
+#define ST_MINOR(sb) minor((sb).st_rdev)
+
+static int get_iostat_sys(sigar_t *sigar,
+                          const char *dirname,
+                          sigar_disk_usage_t *disk,
+                          sigar_iodev_t **iodev)
+{
+    char stat[1025], dev[1025];
+    char *name, *ptr, *fsdev;
+    int partition, status;
+
+    if (!(*iodev = sigar_iodev_get(sigar, dirname))) {
+        return ENXIO;
+    }
+
+    name = fsdev = (*iodev)->name;
+
+    if (SIGAR_NAME_IS_DEV(name)) {
+        name += SSTRLEN(SIGAR_DEV_PREFIX); /* strip "/dev/" */
+    }
+
+    while (!sigar_isdigit(*fsdev)) {
+        fsdev++;
+    }
+
+    partition = strtoul(fsdev, NULL, 0);
+    *fsdev = '\0';
+
+    snprintf(stat, sizeof(stat),
+             SYS_BLOCK "/%s/%s%d/stat", name, name, partition);
+
+    status = sigar_file2str(stat, dev, sizeof(dev));
+    if (status != SIGAR_OK) {
+        return status;
+    }
+
+    ptr = dev;
+    ptr = sigar_skip_token(ptr);
+    disk->reads = sigar_strtoull(ptr);
+    ptr = sigar_skip_token(ptr);
+    disk->writes = sigar_strtoull(ptr);
+
+    disk->read_bytes  = SIGAR_FIELD_NOTIMPL;
+    disk->write_bytes = SIGAR_FIELD_NOTIMPL;
+    disk->queue       = SIGAR_FIELD_NOTIMPL;
+
+    return SIGAR_OK;
+}
+
+static int get_iostat_proc_dstat(sigar_t *sigar,
+                                 const char *dirname,
+                                 sigar_disk_usage_t *disk,
+                                 sigar_iodev_t **iodev,
+                                 sigar_disk_usage_t *device_usage)
+{
+    FILE *fp;
+    char buffer[1025];
+    char *ptr;
+    struct stat sb;
+    int status=ENOENT;
+
+    SIGAR_DISK_STATS_INIT(device_usage);
+
+    if (!(*iodev = sigar_iodev_get(sigar, dirname))) {
+        return ENXIO;
+    }
+
+    if (stat((*iodev)->name, &sb) < 0) {
+        return errno;
+    }
+
+    if (SIGAR_LOG_IS_DEBUG(sigar)) {
+        sigar_log_printf(sigar, SIGAR_LOG_DEBUG,
+                         PROC_DISKSTATS " %s -> %s [%d,%d]",
+                         dirname, (*iodev)->name,
+                         ST_MAJOR(sb), ST_MINOR(sb));
+    }
+
+    if (!(fp = fopen(PROC_DISKSTATS, "r"))) {
+        return errno;
+    }
+
+    while ((ptr = fgets(buffer, sizeof(buffer), fp))) {
+        unsigned long major, minor;
+
+        major = sigar_strtoul(ptr);
+        minor = sigar_strtoul(ptr);
+
+        if ((major == ST_MAJOR(sb)) &&
+            ((minor == ST_MINOR(sb)) || (minor == 0)))
+        {
+            int num;
+            unsigned long
+                rio, rmerge, rsect, ruse,
+                wio, wmerge, wsect, wuse,
+                running, use, aveq;
+
+            ptr = sigar_skip_token(ptr); /* name */
+
+            num = sscanf(ptr,
+                         "%lu %lu %lu %lu "
+                         "%lu %lu %lu %lu "
+                         "%lu %lu %lu",
+                         &rio,     /* 1  # reads issued */
+                         &rmerge,  /* 2  # reads merged */
+                         &rsect,   /* 3  # sectors read */
+                         &ruse,    /* 4  # millis spent reading */
+                         &wio,     /* 5  # writes completed */
+                         &wmerge,  /* 6  # writes merged */
+                         &wsect,   /* 7  # sectors written */
+                         &wuse,    /* 8  # millis spent writing */
+                         &running, /* 9  # I/Os currently in progress */
+                         &use,     /* 10 # millis spent doing I/Os */
+                         &aveq);   /* 11 # of millis spent doing I/Os (weighted) */
+
+            if (num == 11) {
+                disk->rtime = ruse;
+                disk->wtime = wuse;
+                disk->time = use;
+                disk->qtime = aveq;
+            }
+            else if (num == 4) {
+                wio = rsect;
+                rsect = rmerge;
+                wsect = ruse;
+                disk->time = disk->qtime = SIGAR_FIELD_NOTIMPL;
+            }
+            else {
+                status = ENOENT;
+            }
+
+            disk->reads = rio;
+            disk->writes = wio;
+            disk->read_bytes  = rsect;
+            disk->write_bytes = wsect;
+
+            /* convert sectors to bytes (512 is fixed size in 2.6 kernels) */
+            disk->read_bytes  *= 512;
+            disk->write_bytes *= 512;
+
+            if (minor == ST_MINOR(sb)) {
+                status = SIGAR_OK;
+                break;
+            }
+            else if (minor == 0) {
+                memcpy(device_usage, disk, sizeof(*device_usage));
+            }
+        }
+    }
+
+    fclose(fp);
+
+    return status;
+}
+
+static int get_iostat_procp(sigar_t *sigar,
+                            const char *dirname,
+                            sigar_disk_usage_t *disk,
+                            sigar_iodev_t **iodev)
+{
+    FILE *fp;
+    char buffer[1025];
+    char *ptr;
+    struct stat sb;
+
+    if (!(*iodev = sigar_iodev_get(sigar, dirname))) {
+        return ENXIO;
+    }
+
+    if (stat((*iodev)->name, &sb) < 0) {
+        return errno;
+    }
+
+    if (SIGAR_LOG_IS_DEBUG(sigar)) {
+        sigar_log_printf(sigar, SIGAR_LOG_DEBUG,
+                         PROC_PARTITIONS " %s -> %s [%d,%d]",
+                         dirname, (*iodev)->name,
+                         ST_MAJOR(sb), ST_MINOR(sb));
+    }
+
+    if (!(fp = fopen(PROC_PARTITIONS, "r"))) {
+        return errno;
+    }
+
+    (void)fgets(buffer, sizeof(buffer), fp); /* skip header */
+    while ((ptr = fgets(buffer, sizeof(buffer), fp))) {
+        unsigned long major, minor;
+
+        major = sigar_strtoul(ptr);
+        minor = sigar_strtoul(ptr);
+
+        if ((major == ST_MAJOR(sb)) && (minor == ST_MINOR(sb))) {
+            ptr = sigar_skip_token(ptr); /* blocks */
+            ptr = sigar_skip_token(ptr); /* name */
+            disk->reads = sigar_strtoull(ptr); /* rio */
+            ptr = sigar_skip_token(ptr);  /* rmerge */ 
+            disk->read_bytes  = sigar_strtoull(ptr); /* rsect */
+            disk->rtime = sigar_strtoull(ptr); /* ruse */
+            disk->writes = sigar_strtoull(ptr); /* wio */
+            ptr = sigar_skip_token(ptr);  /* wmerge */ 
+            disk->write_bytes = sigar_strtoull(ptr); /* wsect */
+            disk->wtime = sigar_strtoull(ptr); /* wuse */
+            ptr = sigar_skip_token(ptr); /* running */
+            disk->time = sigar_strtoull(ptr); /* use */
+            disk->qtime  = sigar_strtoull(ptr); /* aveq */
+
+            /* convert sectors to bytes (512 is fixed size in 2.6 kernels) */
+            disk->read_bytes  *= 512;
+            disk->write_bytes *= 512;
+
+            fclose(fp);
+            return SIGAR_OK;
+        }
+    }
+
+    fclose(fp);
+
+    return ENOENT;
+}
+
+int sigar_disk_usage_get(sigar_t *sigar, const char *name,
+                         sigar_disk_usage_t *disk)
+{
+    int status;
+    sigar_iodev_t *iodev = NULL;
+    sigar_disk_usage_t device_usage;
+    SIGAR_DISK_STATS_INIT(disk);
+
+    /*
+     * 2.2 has metrics /proc/stat, but wtf is the device mapping?
+     * 2.4 has /proc/partitions w/ the metrics.
+     * 2.6 has /proc/partitions w/o the metrics.
+     *     instead the metrics are within the /proc-like /sys filesystem.
+     *     also has /proc/diskstats
+     */
+    switch (sigar->iostat) {
+      case IOSTAT_SYS:
+        status = get_iostat_sys(sigar, name, disk, &iodev);
+        break;
+      case IOSTAT_DISKSTATS:
+        status = get_iostat_proc_dstat(sigar, name, disk, &iodev, &device_usage);
+        break;
+      case IOSTAT_PARTITIONS:
+        status = get_iostat_procp(sigar, name, disk, &iodev);
+        break;
+      /*
+       * case IOSTAT_SOME_OTHER_WIERD_THING:
+       * break;
+       */
+      case IOSTAT_NONE:
+      default:
+        status = ENOENT;
+        break;
+    }
+
+    if ((status == SIGAR_OK) && iodev) {
+        sigar_uptime_t uptime;
+        sigar_uint64_t interval, ios;
+        double tput, util;
+        sigar_disk_usage_t *partition_usage=NULL;
+
+        sigar_uptime_get(sigar, &uptime);
+
+        if (iodev->is_partition &&
+            (sigar->iostat == IOSTAT_DISKSTATS))
+        {
+            /* 2.6 kernels do not have per-partition times */
+            partition_usage = disk;
+            disk = &device_usage;
+        }
+
+        disk->snaptime = uptime.uptime;
+
+        if (iodev->disk.snaptime) {
+            interval = disk->snaptime - iodev->disk.snaptime;
+        }
+        else {
+            interval = disk->snaptime;
+        }
+
+        ios =
+            (disk->reads - iodev->disk.reads) +
+            (disk->writes - iodev->disk.writes);
+
+        if (disk->time == SIGAR_FIELD_NOTIMPL) {
+            disk->service_time = SIGAR_FIELD_NOTIMPL;
+        }
+        else {
+            tput = ((double)ios) * HZ / interval;
+            util = ((double)(disk->time - iodev->disk.time)) / interval * HZ;
+            disk->service_time = tput ? util / tput : 0.0;
+        }
+        if (disk->qtime == SIGAR_FIELD_NOTIMPL) {
+            disk->queue = SIGAR_FIELD_NOTIMPL;
+        }
+        else {
+            util = ((double)(disk->qtime - iodev->disk.qtime)) / interval;
+            disk->queue = util / 1000.0;
+        }
+
+        memcpy(&iodev->disk, disk, sizeof(iodev->disk));
+        if (partition_usage) {
+            partition_usage->service_time = disk->service_time;
+            partition_usage->queue = disk->queue;
+        }
+    }
+
+    return status;
+}
+
+int sigar_file_system_usage_get(sigar_t *sigar,
+                                const char *dirname,
+                                sigar_file_system_usage_t *fsusage)
+{
+    int status = sigar_statvfs(sigar, dirname, fsusage);
+
+    if (status != SIGAR_OK) {
+        return status;
+    }
+
+    fsusage->use_percent = sigar_file_system_usage_calc_used(sigar, fsusage);
+
+    (void)sigar_disk_usage_get(sigar, dirname, &fsusage->disk);
+
+    return SIGAR_OK;
+}
+
+static SIGAR_INLINE char *cpu_info_strval(char *ptr)
+{
+    if ((ptr = strchr(ptr, ':'))) {
+        ptr++;
+        while (isspace (*ptr)) ptr++;
+        return ptr;
+    }
+    return NULL;
+}
+
+static SIGAR_INLINE void cpu_info_strcpy(char *ptr, char *buf, int len)
+{
+    int slen;
+    ptr = cpu_info_strval(ptr);
+    if (!ptr) {
+        return;
+    }
+    slen = strlen(ptr);
+    strncpy(buf, ptr, len);
+    buf[len] = '\0';
+    if (slen < len) {
+        buf[slen-1] = '\0'; /* rid \n */
+    }
+}
+
+static int get_cpu_info(sigar_t *sigar, sigar_cpu_info_t *info,
+                        FILE *fp)
+{
+    char buffer[BUFSIZ], *ptr;
+
+    int found = 0;
+
+    /* UML vm wont have "cpu MHz" or "cache size" fields */
+    info->mhz        = 0;
+    info->cache_size = 0;
+
+#ifdef __powerpc64__
+    SIGAR_SSTRCPY(info->vendor, "IBM");
+#endif
+
+    while ((ptr = fgets(buffer, sizeof(buffer), fp))) {
+        switch (*ptr) {
+          case 'p': /* processor	: 0 */
+            if (strnEQ(ptr, "processor", 9)) {
+                found = 1;
+            }
+            break;
+          case 'v':
+            /* "vendor_id" or "vendor" */
+            if (strnEQ(ptr, "vendor", 6)) {
+                cpu_info_strcpy(ptr, info->vendor, sizeof(info->vendor));
+                if (strEQ(info->vendor, "GenuineIntel")) {
+                    SIGAR_SSTRCPY(info->vendor, "Intel");
+                }
+                else if (strEQ(info->vendor, "AuthenticAMD")) {
+                    SIGAR_SSTRCPY(info->vendor, "AMD");
+                }
+            }
+            break;
+          case 'f':
+            if (strnEQ(ptr, "family", 6)) {
+                /* IA64 version of "model name" */
+                cpu_info_strcpy(ptr, info->model, sizeof(info->model));
+                sigar_cpu_model_adjust(sigar, info);
+            }
+            break;
+          case 'm':
+            if (strnEQ(ptr, "model name", 10)) {
+                cpu_info_strcpy(ptr, info->model, sizeof(info->model));
+                sigar_cpu_model_adjust(sigar, info);
+            }
+            break;
+          case 'c':
+            if (strnEQ(ptr, "cpu MHz", 7)) {
+                ptr = cpu_info_strval(ptr);
+                info->mhz = atoi(ptr);
+            }
+            else if (strnEQ(ptr, "cache size", 10)) {
+                ptr = cpu_info_strval(ptr);
+                info->cache_size = sigar_strtoul(ptr);
+            }
+#ifdef __powerpc64__
+            /* each /proc/cpuinfo entry looks like so:
+             *   processor       : 0
+             *   cpu             : POWER5 (gr)
+             *   clock           : 1656.392000MHz
+             *   revision        : 2.2
+             */
+            else if (strnEQ(ptr, "clock", 5)) {
+                ptr = cpu_info_strval(ptr);
+                info->mhz = atoi(ptr);
+            }
+            else if (strnEQ(ptr, "cpu", 3)) {
+                cpu_info_strcpy(ptr, info->model, sizeof(info->model));
+
+                if ((ptr = strchr(info->model, ' '))) {
+                    /* "POWER5 (gr)" -> "POWER5" */
+                    *ptr = '\0';
+                }
+            }
+#endif
+            break;
+           /* lone \n means end of info for this processor */
+          case '\n':
+            return found;
+        }
+    }
+
+    return found;
+}
+
+/* /proc/cpuinfo MHz will change w/ AMD + PowerNow */
+static void get_cpuinfo_max_freq(sigar_cpu_info_t *cpu_info, int num)
+{
+    int status;
+    char max_freq[PATH_MAX];
+    snprintf(max_freq, sizeof(max_freq),
+             "/sys/devices/system/cpu/cpu%d"
+             "/cpufreq/cpuinfo_max_freq", num);
+
+    status =
+        sigar_file2str(max_freq, max_freq, sizeof(max_freq)-1);
+
+    if (status == SIGAR_OK) {
+        cpu_info->mhz = atoi(max_freq) / 1000;
+    }
+}
+
+int sigar_cpu_info_list_get(sigar_t *sigar,
+                            sigar_cpu_info_list_t *cpu_infos)
+{
+    FILE *fp;
+    int core_rollup = sigar_cpu_core_rollup(sigar), i=0;
+
+    if (!(fp = fopen(PROC_FS_ROOT "cpuinfo", "r"))) {
+        return errno;
+    }
+
+    (void)sigar_cpu_total_count(sigar);
+    sigar_cpu_info_list_create(cpu_infos);
+
+    while (get_cpu_info(sigar, &cpu_infos->data[cpu_infos->number], fp)) {
+        sigar_cpu_info_t *info;
+
+        if (core_rollup && (i++ % sigar->lcpu)) {
+            continue; /* fold logical processors */
+        }
+
+        info = &cpu_infos->data[cpu_infos->number];
+        get_cpuinfo_max_freq(info, cpu_infos->number);
+
+        info->total_cores = sigar->ncpu;
+        info->cores_per_socket = sigar->lcpu;
+        info->total_sockets = sigar_cpu_socket_count(sigar);
+
+        ++cpu_infos->number;
+        SIGAR_CPU_INFO_LIST_GROW(cpu_infos);
+    }
+
+    fclose(fp);
+
+    return SIGAR_OK;
+}
+
+static SIGAR_INLINE unsigned int hex2int(const char *x, int len)
+{
+    int i;
+    unsigned int j;
+
+    for (i=0, j=0; i<len; i++) {
+        register int ch = x[i];
+        j <<= 4;
+        if (isdigit(ch)) {
+            j |= ch - '0';
+        }
+        else if (isupper(ch)) {
+            j |= ch - ('A' - 10);
+        }
+        else {
+            j |= ch - ('a' - 10);
+        }
+    }
+
+    return j;
+}
+
+#define HEX_ENT_LEN 8
+
+#ifdef SIGAR_64BIT
+#define ROUTE_FMT "%16s %128s %128s %X %ld %ld %ld %128s %ld %ld %ld\n"
+#else
+#define ROUTE_FMT "%16s %128s %128s %X %lld %lld %lld %128s %lld %lld %lld\n"
+#endif
+#define RTF_UP 0x0001
+
+int sigar_net_route_list_get(sigar_t *sigar,
+                             sigar_net_route_list_t *routelist)
+{
+    FILE *fp;
+    char buffer[1024];
+    char net_addr[128], gate_addr[128], mask_addr[128];
+    int flags;
+    sigar_net_route_t *route;
+
+    routelist->size = routelist->number = 0;
+
+    if (!(fp = fopen(PROC_FS_ROOT "net/route", "r"))) {
+        return errno;
+    }
+
+    sigar_net_route_list_create(routelist);
+
+    (void)fgets(buffer, sizeof(buffer), fp); /* skip header */
+    while (fgets(buffer, sizeof(buffer), fp)) {
+        int num;
+
+        SIGAR_NET_ROUTE_LIST_GROW(routelist);
+        route = &routelist->data[routelist->number++];
+
+        /* XXX rid sscanf */
+        num = sscanf(buffer, ROUTE_FMT,
+                     route->ifname, net_addr, gate_addr,
+                     &flags, &route->refcnt, &route->use,
+                     &route->metric, mask_addr,
+                     &route->mtu, &route->window, &route->irtt);
+
+        if ((num < 10) || !(flags & RTF_UP)) {
+            --routelist->number;
+            continue;
+        }
+
+        route->flags = flags;
+        
+        sigar_net_address_set(route->destination, hex2int(net_addr, HEX_ENT_LEN));
+        sigar_net_address_set(route->gateway, hex2int(gate_addr, HEX_ENT_LEN));
+        sigar_net_address_set(route->mask, hex2int(mask_addr, HEX_ENT_LEN));
+    }
+
+    fclose(fp);
+
+    return SIGAR_OK;
+}
+
+int sigar_net_interface_stat_get(sigar_t *sigar, const char *name,
+                                 sigar_net_interface_stat_t *ifstat)
+{
+    int found = 0;
+    char buffer[BUFSIZ];
+    FILE *fp = fopen(PROC_FS_ROOT "net/dev", "r");
+    
+    if (!fp) {
+        return errno;
+    }
+
+    /* skip header */
+    fgets(buffer, sizeof(buffer), fp);
+    fgets(buffer, sizeof(buffer), fp);
+
+    while (fgets(buffer, sizeof(buffer), fp)) {
+        char *ptr, *dev;
+
+        dev = buffer;
+        while (isspace(*dev)) {
+            dev++;
+        }
+
+        if (!(ptr = strchr(dev, ':'))) {
+            continue;
+        }
+
+        *ptr++ = 0;
+
+        if (!strEQ(dev, name)) {
+            continue;
+        }
+
+        found = 1;
+        ifstat->rx_bytes    = sigar_strtoull(ptr);
+        ifstat->rx_packets  = sigar_strtoull(ptr);
+        ifstat->rx_errors   = sigar_strtoull(ptr);
+        ifstat->rx_dropped  = sigar_strtoull(ptr);
+        ifstat->rx_overruns = sigar_strtoull(ptr);
+        ifstat->rx_frame    = sigar_strtoull(ptr);
+
+        /* skip: compressed multicast */
+        ptr = sigar_skip_multiple_token(ptr, 2);
+
+        ifstat->tx_bytes      = sigar_strtoull(ptr);
+        ifstat->tx_packets    = sigar_strtoull(ptr);
+        ifstat->tx_errors     = sigar_strtoull(ptr);
+        ifstat->tx_dropped    = sigar_strtoull(ptr);
+        ifstat->tx_overruns   = sigar_strtoull(ptr);
+        ifstat->tx_collisions = sigar_strtoull(ptr);
+        ifstat->tx_carrier    = sigar_strtoull(ptr);
+
+        ifstat->speed         = SIGAR_FIELD_NOTIMPL;
+
+        break;
+    }
+
+    fclose(fp);
+
+    return found ? SIGAR_OK : ENXIO;
+}
+
+static SIGAR_INLINE void convert_hex_address(sigar_net_address_t *address,
+                                             char *ptr, int len)
+{
+    if (len > HEX_ENT_LEN) {
+        int i;
+        for (i=0; i<=3; i++, ptr+=HEX_ENT_LEN) {
+            address->addr.in6[i] = hex2int(ptr, HEX_ENT_LEN);
+        }
+
+        address->family = SIGAR_AF_INET6;
+    }
+    else {
+        address->addr.in =
+            (len == HEX_ENT_LEN) ? hex2int(ptr, HEX_ENT_LEN) : 0;
+
+        address->family = SIGAR_AF_INET;
+    }
+}
+
+typedef struct {
+    sigar_net_connection_list_t *connlist;
+    sigar_net_connection_t *conn;
+    unsigned long port;
+} net_conn_getter_t;
+
+static int proc_net_walker(sigar_net_connection_walker_t *walker,
+                           sigar_net_connection_t *conn)
+{
+    net_conn_getter_t *getter =
+        (net_conn_getter_t *)walker->data;
+
+    if (getter->connlist) {
+        SIGAR_NET_CONNLIST_GROW(getter->connlist);
+        memcpy(&getter->connlist->data[getter->connlist->number++],
+               conn, sizeof(*conn));
+    }
+    else {
+        if ((getter->port == conn->local_port) &&
+            (conn->remote_port == 0))
+        {
+            memcpy(getter->conn, conn, sizeof(*conn));
+            return !SIGAR_OK; /* break loop */
+        }
+    }
+
+    return SIGAR_OK; /* continue loop */
+}
+
+#define SKIP_WHILE(p, c) while (*p == c) p++
+#define SKIP_PAST(p, c) \
+   while(*p && (*p != c)) p++; \
+   SKIP_WHILE(p, c)
+
+typedef struct {
+    FILE *fp;
+    int (*close)(FILE *);
+} xproc_t;
+
+static FILE *xproc_open(const char *command, xproc_t *xproc)
+{
+    struct stat sb;
+    if (stat(command, &sb) == 0) {
+        if (sb.st_mode & S_IXUSR) {
+            /* executable script for testing large
+             * conn table where we can sleep() to better
+             * simulate /proc/net/tcp behavior
+             */
+            xproc->fp = popen(command, "r");
+            xproc->close = pclose;
+        }
+        else {
+            xproc->fp = fopen(command, "r");
+            xproc->close = fclose;
+        }
+        return xproc->fp;
+    }
+    else {
+        return NULL;
+    }
+}
+
+static int proc_net_read(sigar_net_connection_walker_t *walker,
+                         const char *fname,
+                         int type)
+{
+    FILE *fp = NULL;
+    char buffer[8192];
+    sigar_t *sigar = walker->sigar;
+    char *ptr = sigar->proc_net;
+    int flags = walker->flags;
+    xproc_t xproc = { NULL, fclose };
+
+    if (ptr) {
+        snprintf(buffer, sizeof(buffer),
+                 "%s/%s", ptr,
+                 fname + sizeof(PROC_FS_ROOT)-1);
+
+        if ((fp = xproc_open(buffer, &xproc))) {
+            if (SIGAR_LOG_IS_DEBUG(sigar)) {
+                sigar_log_printf(sigar, SIGAR_LOG_DEBUG,
+                                 "[proc_net] using %s",
+                                 buffer);
+            }
+        }
+        else if (SIGAR_LOG_IS_DEBUG(sigar)) {
+            sigar_log_printf(sigar, SIGAR_LOG_DEBUG,
+                             "[proc_net] cannot open %s",
+                             buffer);
+        }
+    }
+
+    if (!(fp || (fp = fopen(fname, "r")))) {
+        return errno;
+    }
+
+    fgets(buffer, sizeof(buffer), fp); /* skip header */
+
+    while ((ptr = fgets(buffer, sizeof(buffer), fp))) {
+        sigar_net_connection_t conn;
+        char *laddr, *raddr;
+        int laddr_len=0, raddr_len=0;
+        int more;
+
+        /* skip leading space */
+        SKIP_WHILE(ptr, ' ');
+
+        /* skip "%d: " */
+        SKIP_PAST(ptr, ' ');
+
+        laddr = ptr;
+        while (*ptr && (*ptr != ':')) {
+            laddr_len++;
+            ptr++;
+        }
+        SKIP_WHILE(ptr, ':');
+
+        conn.local_port = (strtoul(ptr, &ptr, 16) & 0xffff);
+
+        SKIP_WHILE(ptr, ' ');
+
+        raddr = ptr;
+        while (*ptr && (*ptr != ':')) {
+            raddr_len++;
+            ptr++;
+        }
+        SKIP_WHILE(ptr, ':');
+
+        conn.remote_port = (strtoul(ptr, &ptr, 16) & 0xffff);
+
+        SKIP_WHILE(ptr, ' ');
+
+        if (!((conn.remote_port && (flags & SIGAR_NETCONN_CLIENT)) ||
+              (!conn.remote_port && (flags & SIGAR_NETCONN_SERVER))))
+        {
+            continue;
+        }
+
+        conn.type = type;
+
+        convert_hex_address(&conn.local_address,
+                            laddr, laddr_len);
+
+        convert_hex_address(&conn.remote_address,
+                            raddr, raddr_len);
+
+        /* SIGAR_TCP_* currently matches TCP_* in linux/tcp.h */
+        conn.state = hex2int(ptr, 2);
+        ptr += 2;
+        SKIP_WHILE(ptr, ' ');
+
+        conn.send_queue = hex2int(ptr, HEX_ENT_LEN);
+        ptr += HEX_ENT_LEN+1; /* tx + ':' */;
+
+        conn.receive_queue = hex2int(ptr, HEX_ENT_LEN);
+        ptr += HEX_ENT_LEN;
+        SKIP_WHILE(ptr, ' ');
+
+        SKIP_PAST(ptr, ' '); /* tr:tm->whem */
+        SKIP_PAST(ptr, ' '); /* retrnsmt */
+
+        conn.uid = sigar_strtoul(ptr);
+
+        SKIP_WHILE(ptr, ' ');
+        SKIP_PAST(ptr, ' '); /* timeout */
+
+        conn.inode = sigar_strtoul(ptr);
+
+        more = walker->add_connection(walker, &conn);
+        if (more != SIGAR_OK) {
+            xproc.close(fp);
+            return SIGAR_OK;
+        }
+    }
+
+    xproc.close(fp);
+
+    return SIGAR_OK;
+}
+
+int sigar_net_connection_walk(sigar_net_connection_walker_t *walker)
+{
+    int flags = walker->flags;
+    int status;
+
+    if (flags & SIGAR_NETCONN_TCP) {
+        status = proc_net_read(walker,
+                               PROC_FS_ROOT "net/tcp",
+                               SIGAR_NETCONN_TCP);
+
+        if (status != SIGAR_OK) {
+            return status;
+        }
+
+        status = proc_net_read(walker,
+                               PROC_FS_ROOT "net/tcp6",
+                               SIGAR_NETCONN_TCP);
+
+        if (!((status == SIGAR_OK) || (status == ENOENT))) {
+            return status;
+        }
+    }
+
+    if (flags & SIGAR_NETCONN_UDP) {
+        status = proc_net_read(walker,
+                               PROC_FS_ROOT "net/udp",
+                               SIGAR_NETCONN_UDP);
+
+        if (status != SIGAR_OK) {
+            return status;
+        }
+
+        status = proc_net_read(walker,
+                               PROC_FS_ROOT "net/udp6",
+                               SIGAR_NETCONN_UDP);
+
+        if (!((status == SIGAR_OK) || (status == ENOENT))) {
+            return status;
+        }
+    }
+
+    if (flags & SIGAR_NETCONN_RAW) {
+        status = proc_net_read(walker,
+                               PROC_FS_ROOT "net/raw",
+                               SIGAR_NETCONN_RAW);
+        
+        if (status != SIGAR_OK) {
+            return status;
+        }
+
+        status = proc_net_read(walker,
+                               PROC_FS_ROOT "net/raw6",
+                               SIGAR_NETCONN_RAW);
+
+        if (!((status == SIGAR_OK) || (status == ENOENT))) {
+            return status;
+        }
+    }
+
+    /* XXX /proc/net/unix */
+
+    return SIGAR_OK;
+}
+
+int sigar_net_connection_list_get(sigar_t *sigar,
+                                  sigar_net_connection_list_t *connlist,
+                                  int flags)
+{
+    int status;
+    sigar_net_connection_walker_t walker;
+    net_conn_getter_t getter;
+
+    sigar_net_connection_list_create(connlist);
+
+    getter.conn = NULL;
+    getter.connlist = connlist;
+
+    walker.sigar = sigar;
+    walker.flags = flags;
+    walker.data = &getter;
+    walker.add_connection = proc_net_walker;
+
+    status = sigar_net_connection_walk(&walker);
+
+    if (status != SIGAR_OK) {
+        sigar_net_connection_list_destroy(sigar, connlist);
+    }
+
+    return status;
+}
+
+static int sigar_net_connection_get(sigar_t *sigar,
+                                    sigar_net_connection_t *netconn,
+                                    unsigned long port,
+                                    int flags)
+{
+    int status;
+    sigar_net_connection_walker_t walker;
+    net_conn_getter_t getter;
+
+    getter.conn = netconn;
+    getter.connlist = NULL;
+    getter.port = port;
+
+    walker.sigar = sigar;
+    walker.flags = flags;
+    walker.data = &getter;
+    walker.add_connection = proc_net_walker;
+
+    status = sigar_net_connection_walk(&walker);
+
+    return status;
+}
+
+#define SNMP_TCP_PREFIX "Tcp: "
+
+SIGAR_DECLARE(int)
+sigar_tcp_get(sigar_t *sigar,
+              sigar_tcp_t *tcp)
+{
+    FILE *fp;
+    char buffer[1024], *ptr=buffer;
+    int status = SIGAR_ENOENT;
+
+    if (!(fp = fopen(PROC_FS_ROOT "net/snmp", "r"))) {
+        return errno;
+    }
+
+    while (fgets(buffer, sizeof(buffer), fp)) {
+        if (strnEQ(buffer, SNMP_TCP_PREFIX, sizeof(SNMP_TCP_PREFIX)-1)) {
+            if (fgets(buffer, sizeof(buffer), fp)) {
+                status = SIGAR_OK;
+                break;
+            }
+        }
+    }
+
+    fclose(fp);
+
+    if (status == SIGAR_OK) {
+        /* assuming field order, same in 2.2, 2.4 and 2.6 kernels */ 
+        /* Tcp: RtoAlgorithm RtoMin RtoMax MaxConn */
+        ptr = sigar_skip_multiple_token(ptr, 5);
+        tcp->active_opens = sigar_strtoull(ptr);
+        tcp->passive_opens = sigar_strtoull(ptr);
+        tcp->attempt_fails = sigar_strtoull(ptr);
+        tcp->estab_resets = sigar_strtoull(ptr);
+        tcp->curr_estab = sigar_strtoull(ptr);
+        tcp->in_segs = sigar_strtoull(ptr);
+        tcp->out_segs = sigar_strtoull(ptr);
+        tcp->retrans_segs = sigar_strtoull(ptr);
+        tcp->in_errs = sigar_strtoull(ptr);
+        tcp->out_rsts = sigar_strtoull(ptr);
+    }
+
+    return status;
+}
+
+static int sigar_proc_nfs_gets(char *file, char *tok,
+                               char *buffer, size_t size)
+{
+    int status = ENOENT;
+    int len = strlen(tok);
+    FILE *fp = fopen(file, "r");
+    
+    if (!fp) {
+        return SIGAR_ENOTIMPL;
+    }
+
+    while (fgets(buffer, size, fp)) {
+        if (strnEQ(buffer, tok, len)) {
+            status = SIGAR_OK;
+            break;
+        }
+    }
+
+    fclose(fp);
+
+    return status;
+}
+
+static int sigar_nfs_v2_get(char *file, sigar_nfs_v2_t *nfs)
+{
+    char buffer[BUFSIZ], *ptr=buffer;
+    int status =
+        sigar_proc_nfs_gets(file,
+                            "proc2", buffer, sizeof(buffer));
+
+    if (status != SIGAR_OK) {
+        return status;
+    }
+
+    ptr = sigar_skip_multiple_token(ptr, 2);
+
+    nfs->null = sigar_strtoull(ptr);
+    nfs->getattr = sigar_strtoull(ptr);
+    nfs->setattr = sigar_strtoull(ptr);
+    nfs->root = sigar_strtoull(ptr);
+    nfs->lookup = sigar_strtoull(ptr);
+    nfs->readlink = sigar_strtoull(ptr);
+    nfs->read = sigar_strtoull(ptr);
+    nfs->writecache = sigar_strtoull(ptr);
+    nfs->write = sigar_strtoull(ptr);
+    nfs->create = sigar_strtoull(ptr);
+    nfs->remove = sigar_strtoull(ptr);
+    nfs->rename = sigar_strtoull(ptr);
+    nfs->link = sigar_strtoull(ptr);
+    nfs->symlink = sigar_strtoull(ptr);
+    nfs->mkdir = sigar_strtoull(ptr);
+    nfs->rmdir = sigar_strtoull(ptr);
+    nfs->readdir = sigar_strtoull(ptr);
+    nfs->fsstat = sigar_strtoull(ptr);
+
+    return SIGAR_OK;
+}
+
+int sigar_nfs_client_v2_get(sigar_t *sigar,
+                            sigar_nfs_client_v2_t *nfs)
+{
+    return sigar_nfs_v2_get(PROC_FS_ROOT "net/rpc/nfs",
+                            (sigar_nfs_v2_t *)nfs);
+}
+
+int sigar_nfs_server_v2_get(sigar_t *sigar,
+                            sigar_nfs_server_v2_t *nfs)
+{
+    return sigar_nfs_v2_get(PROC_FS_ROOT "net/rpc/nfsd",
+                            (sigar_nfs_v2_t *)nfs);
+}
+
+static int sigar_nfs_v3_get(char *file, sigar_nfs_v3_t *nfs)
+{
+    char buffer[BUFSIZ], *ptr=buffer;
+    int status =
+        sigar_proc_nfs_gets(file,
+                            "proc3", buffer, sizeof(buffer));
+
+    if (status != SIGAR_OK) {
+        return status;
+    }
+
+    ptr = sigar_skip_multiple_token(ptr, 2);
+
+    nfs->null = sigar_strtoull(ptr);
+    nfs->getattr = sigar_strtoull(ptr);
+    nfs->setattr = sigar_strtoull(ptr);
+    nfs->lookup = sigar_strtoull(ptr);
+    nfs->access = sigar_strtoull(ptr);
+    nfs->readlink = sigar_strtoull(ptr);
+    nfs->read = sigar_strtoull(ptr);
+    nfs->write = sigar_strtoull(ptr);
+    nfs->create = sigar_strtoull(ptr);
+    nfs->mkdir = sigar_strtoull(ptr);
+    nfs->symlink = sigar_strtoull(ptr);
+    nfs->mknod = sigar_strtoull(ptr);
+    nfs->remove = sigar_strtoull(ptr);
+    nfs->rmdir = sigar_strtoull(ptr);
+    nfs->rename = sigar_strtoull(ptr);
+    nfs->link = sigar_strtoull(ptr);
+    nfs->readdir = sigar_strtoull(ptr);
+    nfs->readdirplus = sigar_strtoull(ptr);
+    nfs->fsstat = sigar_strtoull(ptr);
+    nfs->fsinfo = sigar_strtoull(ptr);
+    nfs->pathconf = sigar_strtoull(ptr);
+    nfs->commit = sigar_strtoull(ptr);
+
+    return SIGAR_OK;
+}
+
+int sigar_nfs_client_v3_get(sigar_t *sigar,
+                            sigar_nfs_client_v3_t *nfs)
+{
+    return sigar_nfs_v3_get(PROC_FS_ROOT "net/rpc/nfs",
+                            (sigar_nfs_v3_t *)nfs);
+}
+
+int sigar_nfs_server_v3_get(sigar_t *sigar,
+                            sigar_nfs_server_v3_t *nfs)
+{
+    return sigar_nfs_v3_get(PROC_FS_ROOT "net/rpc/nfsd",
+                            (sigar_nfs_v3_t *)nfs);
+}
+
+int sigar_proc_port_get(sigar_t *sigar, int protocol,
+                        unsigned long port, sigar_pid_t *pid)
+{
+    int status;
+    sigar_net_connection_t netconn;
+    DIR *dirp;
+    struct dirent *ent, dbuf;
+
+    SIGAR_ZERO(&netconn);
+    *pid = 0;
+
+    status = sigar_net_connection_get(sigar, &netconn, port,
+                                      SIGAR_NETCONN_SERVER|protocol);
+
+    if (status != SIGAR_OK) {
+        return status;
+    }
+
+    if (netconn.local_port != port) {
+        return SIGAR_OK; /* XXX or ENOENT? */
+    }
+
+    if (!(dirp = opendir(PROCP_FS_ROOT))) {
+        return errno;
+    }
+
+    while (readdir_r(dirp, &dbuf, &ent) == 0) {
+        DIR *fd_dirp;
+        struct dirent *fd_ent, fd_dbuf;
+        struct stat sb;
+        char fd_name[BUFSIZ], pid_name[BUFSIZ];
+        int len, slen;
+
+        if (ent == NULL) {
+            break;
+        }
+
+        if (!sigar_isdigit(*ent->d_name)) {
+            continue;
+        }
+
+        /* sprintf(pid_name, "/proc/%s", ent->d_name) */
+        memcpy(&pid_name[0], PROCP_FS_ROOT, SSTRLEN(PROCP_FS_ROOT));
+        len = SSTRLEN(PROCP_FS_ROOT);
+        pid_name[len++] = '/';
+
+        slen = strlen(ent->d_name);
+        memcpy(&pid_name[len], ent->d_name, slen);
+        len += slen;
+        pid_name[len] = '\0';
+        
+        if (stat(pid_name, &sb) < 0) {
+            continue;
+        }
+        if (sb.st_uid != netconn.uid) {
+            continue;
+        }
+
+        /* sprintf(fd_name, "%s/fd", pid_name) */
+        memcpy(&fd_name[0], pid_name, len);
+        memcpy(&fd_name[len], "/fd", 3);
+        fd_name[len+=3] = '\0';
+
+        if (!(fd_dirp = opendir(fd_name))) {
+            continue;
+        }
+
+        while (readdir_r(fd_dirp, &fd_dbuf, &fd_ent) == 0) {
+            char fd_ent_name[BUFSIZ];
+
+            if (fd_ent == NULL) {
+                break;
+            }
+
+            if (!sigar_isdigit(*fd_ent->d_name)) {
+                continue;
+            }
+
+            /* sprintf(fd_ent_name, "%s/%s", fd_name, fd_ent->d_name) */
+            slen = strlen(fd_ent->d_name);
+            memcpy(&fd_ent_name[0], fd_name, len);
+            fd_ent_name[len] = '/';
+            memcpy(&fd_ent_name[len+1], fd_ent->d_name, slen);
+            fd_ent_name[len+1+slen] = '\0';
+
+            if (stat(fd_ent_name, &sb) < 0) {
+                continue;
+            }
+
+            if (sb.st_ino == netconn.inode) {
+                closedir(fd_dirp);
+                closedir(dirp);
+                *pid = strtoul(ent->d_name, NULL, 10);
+                return SIGAR_OK;
+            }
+            
+        }
+
+        closedir(fd_dirp);
+    }
+
+    closedir(dirp);
+
+    return SIGAR_OK;
+}
+
+static void generic_vendor_parse(char *line, sigar_sys_info_t *info)
+{
+    char *ptr;
+    int len = 0;
+        
+    while (*line) {
+        SIGAR_SKIP_SPACE(line);
+        if (!isdigit(*line)) {
+            ++line;
+            continue;
+        }
+
+        ptr = line;
+        while ((isdigit(*ptr) || (*ptr == '.'))) {
+            ++ptr;
+            ++len;
+        }
+
+        if (len) {
+            /* sanity check */
+            if (len > sizeof(info->vendor_version)) {
+                continue;
+            }
+            memcpy(info->vendor_version, line, len);/*XXX*/
+            info->vendor_version[len] = '\0';
+            return;
+        }
+    }
+}
+
+static void redhat_vendor_parse(char *line, sigar_sys_info_t *info)
+{
+    char *start, *end;
+
+    generic_vendor_parse(line, info); /* super.parse */
+
+    if ((start = strchr(line, '('))) {
+        ++start;
+        if ((end = strchr(start, ')'))) {
+            int len = end-start;
+            memcpy(info->vendor_code_name, start, len);/*XXX*/
+            info->vendor_code_name[len] = '\0';
+        }
+    }
+
+#define RHEL_PREFIX "Red Hat Enterprise Linux "
+#define CENTOS_VENDOR "CentOS"
+#define SL_VENDOR "Scientific Linux"
+
+    if (strnEQ(line, RHEL_PREFIX, sizeof(RHEL_PREFIX)-1)) {
+        snprintf(info->vendor_version,
+                 sizeof(info->vendor_version),
+                 "Enterprise Linux %c",
+                 info->vendor_version[0]);
+    }
+    else if (strnEQ(line, CENTOS_VENDOR, sizeof(CENTOS_VENDOR)-1)) {
+        SIGAR_SSTRCPY(info->vendor, CENTOS_VENDOR);
+    }
+    else if (strnEQ(line, SL_VENDOR, sizeof(SL_VENDOR)-1)) {
+        SIGAR_SSTRCPY(info->vendor, SL_VENDOR);
+    }
+}
+
+#define is_quote(c) ((c == '\'') || (c == '"')) 
+
+static void kv_parse(char *data, sigar_sys_info_t *info,
+                     void (*func)(sigar_sys_info_t *, char *, char *))
+{
+    char *ptr = data;
+    int len = strlen(data);
+    char *end = data+len;
+
+    while (ptr < end) {
+        char *val = strchr(ptr, '=');
+        int klen, vlen;
+        char key[256], *ix;
+
+        if (!val) {
+            continue;
+        }
+        klen = val - ptr;
+        SIGAR_SSTRCPY(key, ptr);
+        key[klen] = '\0';
+        ++val;
+
+        if ((ix = strchr(val, '\n'))) {
+            *ix = '\0';
+        }
+        vlen = strlen(val);
+        if (is_quote(*val)) {
+            if (is_quote(val[vlen-1])) {
+                val[vlen-1] =  '\0';
+            }
+            ++val;
+        }
+
+        func(info, key, val);
+
+        ptr += (klen + 1 + vlen + 1);
+    }
+}
+
+static void lsb_parse(sigar_sys_info_t *info,
+                      char *key, char *val)
+{
+    if (strEQ(key, "DISTRIB_ID")) {
+        SIGAR_SSTRCPY(info->vendor, val);
+    }
+    else if (strEQ(key, "DISTRIB_RELEASE")) {
+        SIGAR_SSTRCPY(info->vendor_version, val);
+    }
+    else if (strEQ(key, "DISTRIB_CODENAME")) {
+        SIGAR_SSTRCPY(info->vendor_code_name, val);
+    }
+}
+
+static void lsb_vendor_parse(char *data, sigar_sys_info_t *info)
+{
+    kv_parse(data, info, lsb_parse);
+}
+
+static void xen_parse(sigar_sys_info_t *info,
+                      char *key, char *val)
+{
+    if (strEQ(key, "PRODUCT_VERSION")) {
+        SIGAR_SSTRCPY(info->vendor_version, val);
+    }
+    else if (strEQ(key, "KERNEL_VERSION")) {
+        SIGAR_SSTRCPY(info->version, val);
+    }
+}
+
+static void xen_vendor_parse(char *data, sigar_sys_info_t *info)
+{
+    kv_parse(data, info, xen_parse);
+
+    snprintf(info->description,
+             sizeof(info->description),
+             "XenServer %s",
+             info->vendor_version);
+}
+
+typedef struct {
+    const char *name;
+    const char *file;
+    void (*parse)(char *, sigar_sys_info_t *);
+} linux_vendor_info_t;
+
+static linux_vendor_info_t linux_vendors[] = {
+    { "Fedora",    "/etc/fedora-release", NULL },
+    { "SuSE",      "/etc/SuSE-release", NULL },
+    { "Gentoo",    "/etc/gentoo-release", NULL },
+    { "Slackware", "/etc/slackware-version", NULL },
+    { "Mandrake",  "/etc/mandrake-release", NULL },
+    { "VMware",    "/proc/vmware/version", NULL },
+    { "XenSource", "/etc/xensource-inventory", xen_vendor_parse },
+    { "Red Hat",   "/etc/redhat-release", redhat_vendor_parse },
+    { "lsb",       "/etc/lsb-release", lsb_vendor_parse },
+    { "Debian",    "/etc/debian_version", NULL },
+    { NULL }
+};
+
+static int get_linux_vendor_info(sigar_sys_info_t *info)
+{
+    int i, status = ENOENT;
+    /* env vars for testing */
+    const char *release_file = getenv("SIGAR_OS_RELEASE_FILE");
+    const char *vendor_name  = getenv("SIGAR_OS_VENDOR_NAME");
+    char buffer[8192], *data;
+    linux_vendor_info_t *vendor = NULL;
+
+    for (i=0; linux_vendors[i].name; i++) {
+        struct stat sb;
+        vendor = &linux_vendors[i];
+
+        if (release_file && vendor_name) {
+            if (!strEQ(vendor->name, vendor_name)) {
+                continue;
+            }
+        }
+        else {
+            if (stat(vendor->file, &sb) < 0) {
+                continue;
+            }
+            release_file = vendor->file;
+        }
+
+        status =
+            sigar_file2str(release_file, buffer, sizeof(buffer)-1);
+
+        break;
+    }
+
+    if (status != SIGAR_OK) {
+        return status;
+    }
+
+    data = buffer;
+
+    SIGAR_SSTRCPY(info->vendor, vendor->name);
+
+    if (vendor->parse) {
+        vendor->parse(data, info);
+    }
+    else {
+        generic_vendor_parse(data, info);
+    }
+
+    if (info->description[0] == '\0') {
+        snprintf(info->description,
+                 sizeof(info->description),
+                 "%s %s",
+                 info->vendor, info->vendor_version);
+    }
+
+    return SIGAR_OK;
+}
+
+int sigar_os_sys_info_get(sigar_t *sigar,
+                          sigar_sys_info_t *sysinfo)
+{
+
+    get_linux_vendor_info(sysinfo);
+
+    return SIGAR_OK;
+}