Python classes to extract information from the Linux kernel /proc files.
python/python-linux-procfs/python-linux-procfs.git - Python classes to extract information from the Linux kernel /proc files. https://git.kernel.org/pub/scm/libs/python/python-linux-procfs/python-linux-procfs.git/
#! /usr/bin/python
# -*- python -*-
# -*- coding: utf-8 -*-
#
# Copyright (C) 2007-2015 Red Hat, Inc.
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation; version 2 of the License.
#
# 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., 675 Mass Ave, Cambridge, MA 02139, USA.
#
from __future__ import absolute_import
from __future__ import print_function
import os, time
from functools import reduce
from .utilist import bitmasklist
from six.moves import range
VERSION="0.5"
def process_cmdline(pid_info):
"""
Returns the process command line, if available in the given `process' class, if
not available, falls back to using the comm (short process name) in its pidstat key.
"""
if pid_info["cmdline"]:
return reduce(lambda a, b: a + " %s" % b, pid_info["cmdline"]).strip()
return pid_info["stat"]["comm"]
class pidstat:
"""Provides a dictionary to access the fields in the per process /proc/PID/stat
files.
One can obtain the available fields asking for the keys of the dictionary, e.g.:
>>> p = procfs.pidstat(1)
>>> print p.keys()
['majflt', 'rss', 'cnswap', 'cstime', 'pid', 'session', 'startstack', 'startcode', 'cmajflt', 'blocked', 'exit_signal', 'minflt', 'nswap', 'environ', 'priority', 'state', 'delayacct_blkio_ticks', 'policy', 'rt_priority', 'ppid', 'nice', 'cutime', 'endcode', 'wchan', 'num_threads', 'sigcatch', 'comm', 'stime', 'sigignore', 'tty_nr', 'kstkeip', 'utime', 'tpgid', 'itrealvalue', 'kstkesp', 'rlim', 'signal', 'pgrp', 'flags', 'starttime', 'cminflt', 'vsize', 'processor']
And then access the various process properties using it as a dictionary:
>>> print p['comm']
systemd
>>> print p['priority']
20
>>> print p['state']
S
Please refer to the 'procfs(5)' man page, by using:
$ man 5 procfs
To see information for each of the above fields, it is part of the
'man-pages' RPM package.
"""
# Entries with the same value, the one with a comment after it is the
# more recent, having replaced the other name in v4.1-rc kernel times.
PF_ALIGNWARN = 0x00000001
PF_STARTING = 0x00000002
PF_EXITING = 0x00000004
PF_EXITPIDONE = 0x00000008
PF_VCPU = 0x00000010
PF_WQ_WORKER = 0x00000020 # /* I'm a workqueue worker */
PF_FORKNOEXEC = 0x00000040
PF_MCE_PROCESS = 0x00000080 # /* process policy on mce errors */
PF_SUPERPRIV = 0x00000100
PF_DUMPCORE = 0x00000200
PF_SIGNALED = 0x00000400
PF_MEMALLOC = 0x00000800
PF_NPROC_EXCEEDED= 0x00001000 # /* set_user noticed that RLIMIT_NPROC was exceeded */
PF_FLUSHER = 0x00001000
PF_USED_MATH = 0x00002000
PF_USED_ASYNC = 0x00004000 # /* used async_schedule*(), used by module init */
PF_NOFREEZE = 0x00008000
PF_FROZEN = 0x00010000
PF_FSTRANS = 0x00020000
PF_KSWAPD = 0x00040000
PF_MEMALLOC_NOIO = 0x00080000 # /* Allocating memory without IO involved */
PF_SWAPOFF = 0x00080000
PF_LESS_THROTTLE = 0x00100000
PF_KTHREAD = 0x00200000
PF_RANDOMIZE = 0x00400000
PF_SWAPWRITE = 0x00800000
PF_SPREAD_PAGE = 0x01000000
PF_SPREAD_SLAB = 0x02000000
PF_THREAD_BOUND = 0x04000000
PF_NO_SETAFFINITY = 0x04000000 # /* Userland is not allowed to meddle with cpus_allowed */
PF_MCE_EARLY = 0x08000000 # /* Early kill for mce process policy */
PF_MEMPOLICY = 0x10000000
PF_MUTEX_TESTER = 0x20000000
PF_FREEZER_SKIP = 0x40000000
PF_FREEZER_NOSIG = 0x80000000
PF_SUSPEND_TASK = 0x80000000 # /* this thread called freeze_processes and should not be frozen */
proc_stat_fields = [ "pid", "comm", "state", "ppid", "pgrp", "session",
"tty_nr", "tpgid", "flags", "minflt", "cminflt",
"majflt", "cmajflt", "utime", "stime", "cutime",
"cstime", "priority", "nice", "num_threads",
"itrealvalue", "starttime", "vsize", "rss",
"rlim", "startcode", "endcode", "startstack",
"kstkesp", "kstkeip", "signal", "blocked",
"sigignore", "sigcatch", "wchan", "nswap",
"cnswap", "exit_signal", "processor",
"rt_priority", "policy",
"delayacct_blkio_ticks", "environ" ]
def __init__(self, pid, basedir = "/proc"):
self.pid = pid
self.load(basedir)
def __getitem__(self, fieldname):
return self.fields[fieldname]
def keys(self):
return list(self.fields.keys())
def values(self):
return list(self.fields.values())
def has_key(self, fieldname):
return fieldname in self.fields
def items(self):
return self.fields
def __contains__(self, fieldname):
return fieldname in self.fields
def load(self, basedir = "/proc"):
f = open("%s/%d/stat" % (basedir, self.pid))
fields = f.readline().strip().split(') ')
f.close()
fields = fields[0].split(' (') + fields[1].split()
self.fields = {}
nr_fields = min(len(fields), len(self.proc_stat_fields))
for i in range(nr_fields):
attrname = self.proc_stat_fields[i]
value = fields[i]
if attrname == "comm":
self.fields["comm"] = value.strip('()')
else:
try:
self.fields[attrname] = int(value)
except:
self.fields[attrname] = value
def is_bound_to_cpu(self):
"""
Returns true if this process has a fixed smp affinity mask,
not allowing it to be moved to a different set of CPUs.
"""
return self.fields["flags"] & self.PF_THREAD_BOUND and \
True or False
def process_flags(self):
"""
Returns a list with all the process flags known, details depend
on kernel version, declared in the file include/linux/sched.h in
the kernel sources.
As of v4.2-rc7 these include (from include/linux/sched.h comments):
PF_EXITING Getting shut down
PF_EXITPIDONE Pi exit done on shut down
PF_VCPU I'm a virtual CPU
PF_WQ_WORKER I'm a workqueue worker
PF_FORKNOEXEC Forked but didn't exec
PF_MCE_PROCESS Process policy on mce errors
PF_SUPERPRIV Used super-user privileges
PF_DUMPCORE Dumped core
PF_SIGNALED Killed by a signal
PF_MEMALLOC Allocating memory
PF_NPROC_EXCEEDED Set_user noticed that RLIMIT_NPROC was exceeded
PF_USED_MATH If unset the fpu must be initialized before use
PF_USED_ASYNC Used async_schedule*(), used by module init
PF_NOFREEZE This thread should not be frozen
PF_FROZEN Frozen for system suspend
PF_FSTRANS Inside a filesystem transaction
PF_KSWAPD I am kswapd
PF_MEMALLOC_NOIO Allocating memory without IO involved
PF_LESS_THROTTLE Throttle me less: I clean memory
PF_KTHREAD I am a kernel thread
PF_RANDOMIZE Randomize virtual address space
PF_SWAPWRITE Allowed to write to swap
PF_NO_SETAFFINITY Userland is not allowed to meddle with cpus_allowed
PF_MCE_EARLY Early kill for mce process policy
PF_MUTEX_TESTER Thread belongs to the rt mutex tester
PF_FREEZER_SKIP Freezer should not count it as freezable
PF_SUSPEND_TASK This thread called freeze_processes and should not be frozen
"""
sflags = []
for attr in dir(self):
if attr[:3] != "PF_":
continue
value = getattr(self, attr)
if value & self.fields["flags"]:
sflags.append(attr)
return sflags
def cannot_set_affinity(self, pid):
PF_NO_SETAFFINITY = 0x04000000
try:
return int(self.processes[pid]["stat"]["flags"]) & \
PF_NO_SETAFFINITY and True or False
except:
return True
def cannot_set_thread_affinity(self, pid, tid):
PF_NO_SETAFFINITY = 0x04000000
try:
return int(self.processes[pid].threads[tid]["stat"]["flags"]) & \
PF_NO_SETAFFINITY and True or False
except:
return True
class pidstatus:
"""
Provides a dictionary to access the fields in the per process /proc/PID/status
files. This provides additional information about processes and threads to what
can be obtained with the procfs.pidstat() class.
One can obtain the available fields asking for the keys of the dictionary, e.g.:
>>> import procfs
>>> p = procfs.pidstatus(1)
>>> print p.keys()
['VmExe', 'CapBnd', 'NSpgid', 'Tgid', 'NSpid', 'VmSize', 'VmPMD', 'ShdPnd', 'State', 'Gid', 'nonvoluntary_ctxt_switches', 'SigIgn', 'VmStk', 'VmData', 'SigCgt', 'CapEff', 'VmPTE', 'Groups', 'NStgid', 'Threads', 'PPid', 'VmHWM', 'NSsid', 'VmSwap', 'Name', 'SigBlk', 'Mems_allowed_list', 'VmPeak', 'Ngid', 'VmLck', 'SigQ', 'VmPin', 'Mems_allowed', 'CapPrm', 'Seccomp', 'VmLib', 'Cpus_allowed', 'Uid', 'SigPnd', 'Pid', 'Cpus_allowed_list', 'TracerPid', 'CapInh', 'voluntary_ctxt_switches', 'VmRSS', 'FDSize']
>>> print p["Pid"]
1
>>> print p["Threads"]
1
>>> print p["VmExe"]
1248 kB
>>> print p["Cpus_allowed"]
f
>>> print p["SigQ"]
0/30698
>>> print p["VmPeak"]
320300 kB
>>>
Please refer to the 'procfs(5)' man page, by using:
$ man 5 procfs
To see information for each of the above fields, it is part of the
'man-pages' RPM package.
In the man page there will be references to further documentation, like
referring to the "getrlimit(2)" man page when explaining the "SigQ"
line/field.
"""
def __init__(self, pid, basedir = "/proc"):
self.pid = pid
self.load(basedir)
def __getitem__(self, fieldname):
return self.fields[fieldname]
def keys(self):
return list(self.fields.keys())
def values(self):
return list(self.fields.values())
def has_key(self, fieldname):
return fieldname in self.fields
def items(self):
return self.fields
def __contains__(self, fieldname):
return fieldname in self.fields
def load(self, basedir = "/proc"):
f = open("%s/%d/status" % (basedir, self.pid))
self.fields = {}
for line in f.readlines():
fields = line.split(":")
if len(fields) != 2:
continue
name = fields[0]
value = fields[1].strip()
try:
self.fields[name] = int(value)
except:
self.fields[name] = value
f.close()
class process:
"""
Information about a process with a given pid, provides a dictionary with
two entries, instances of different wrappers for /proc/ process related
meta files: "stat" and "status", see the documentation for procfs.pidstat
and procfs.pidstatus for further info about those classes.
"""
def __init__(self, pid, basedir = "/proc"):
self.pid = pid
self.basedir = basedir
def __getitem__(self, attr):
if not hasattr(self, attr):
if attr in ("stat", "status"):
if attr == "stat":
sclass = pidstat
else:
sclass = pidstatus
setattr(self, attr, sclass(self.pid, self.basedir))
elif attr == "cmdline":
self.load_cmdline()
elif attr == "threads":
self.load_threads()
elif attr == "cgroups":
self.load_cgroups()
elif attr == "environ":
self.load_environ()
return getattr(self, attr)
def has_key(self, attr):
return hasattr(self, attr)
def __contains__(self, attr):
return hasattr(self, attr)
def load_cmdline(self):
f = open("/proc/%d/cmdline" % self.pid)
self.cmdline = f.readline().strip().split('\0')[:-1]
f.close()
def load_threads(self):
self.threads = pidstats("/proc/%d/task/" % self.pid)
# remove thread leader
del self.threads[self.pid]
def load_cgroups(self):
f = open("/proc/%d/cgroup" % self.pid)
self.cgroups = ""
for line in reversed(f.readlines()):
if len(self.cgroups):
self.cgroups = self.cgroups + "," + line[:-1]
else:
self.cgroups = line[:-1]
f.close()
def load_environ(self):
"""
Loads the environment variables for this process. The entries then
become available via the 'environ' member, or via the 'environ'
dict key when accessing as p["environ"].
E.g.:
>>> all_processes = procfs.pidstats()
>>> firefox_pid = all_processes.find_by_name("firefox")
>>> firefox_process = all_processes[firefox_pid[0]]
>>> print firefox_process["environ"]["PWD"]
/home/acme
>>> print len(firefox_process.environ.keys())
66
>>> print firefox_process["environ"]["SHELL"]
/bin/bash
>>> print firefox_process["environ"]["USERNAME"]
acme
>>> print firefox_process["environ"]["HOME"]
/home/acme
>>> print firefox_process["environ"]["MAIL"]
/var/spool/mail/acme
>>>
"""
self.environ = {}
f = open("/proc/%d/environ" % self.pid)
for x in f.readline().split('\0'):
if len(x) > 0:
y = x.split('=')
self.environ[y[0]] = y[1]
f.close()
class pidstats:
"""
Provides access to all the processes in the system, to get a picture of
how many processes there are at any given moment.
The entries can be accessed as a dictionary, keyed by pid. Also there are
methods to find processes that match a given COMM or regular expression.
"""
def __init__(self, basedir = "/proc"):
self.basedir = basedir
self.processes = {}
self.reload()
def __getitem__(self, key):
return self.processes[key]
def __delitem__(self, key):
# not clear on why this can fail, but it can
try:
del self.processes[key]
except:
pass
def keys(self):
return list(self.processes.keys())
def values(self):
return list(self.processes.values())
def has_key(self, key):
return key in self.processes
def items(self):
return self.processes
def __contains__(self, key):
return key in self.processes
def reload(self):
"""
This operation will throw away the current dictionary contents, if any, and
read all the pid files from /proc/, instantiating a 'process' instance for
each of them.
This is a high overhead operation, and should be avoided if the perf python
binding can be used to detect when new threads appear and existing ones
terminate.
In RHEL it is found in the python-perf rpm package.
More information about the perf facilities can be found in the 'perf_event_open'
man page.
"""
del self.processes
self.processes = {}
pids = os.listdir(self.basedir)
for spid in pids:
try:
pid = int(spid)
except:
continue
self.processes[pid] = process(pid, self.basedir)
def reload_threads(self):
to_remove = []
for pid in self.processes.keys():
try:
self.processes[pid].load_threads()
except OSError:
# process vanished, remove it
to_remove.append(pid)
for pid in to_remove:
del self.processes[pid]
def find_by_name(self, name):
name = name[:15]
pids = []
for pid in self.processes.keys():
try:
if name == self.processes[pid]["stat"]["comm"]:
pids.append(pid)
except IOError:
# We're doing lazy loading of /proc files
# So if we get this exception is because the
# process vanished, remove it
del self.processes[pid]
return pids
def find_by_regex(self, regex):
pids = []
for pid in self.processes.keys():
try:
if regex.match(self.processes[pid]["stat"]["comm"]):
pids.append(pid)
except IOError:
# We're doing lazy loading of /proc files
# So if we get this exception is because the
# process vanished, remove it
del self.processes[pid]
return pids
def find_by_cmdline_regex(self, regex):
pids = []
for pid in self.processes.keys():
try:
if regex.match(process_cmdline(self.processes[pid])):
pids.append(pid)
except IOError:
# We're doing lazy loading of /proc files
# So if we get this exception is because the
# process vanished, remove it
del self.processes[pid]
return pids
def get_per_cpu_rtprios(self, basename):
cpu = 0
priorities=""
processed_pids = []
while True:
name = "%s/%d" % (basename, cpu)
pids = self.find_by_name(name)
if not pids or len([n for n in pids if n not in processed_pids]) == 0:
break
for pid in pids:
try:
priorities += "%s," % self.processes[pid]["stat"]["rt_priority"]
except IOError:
# We're doing lazy loading of /proc files
# So if we get this exception is because the
# process vanished, remove it
del self.processes[pid]
processed_pids += pids
cpu += 1
priorities = priorities.strip(',')
return priorities
def get_rtprios(self, name):
cpu = 0
priorities=""
processed_pids = []
while True:
pids = self.find_by_name(name)
if not pids or len([n for n in pids if n not in processed_pids]) == 0:
break
for pid in pids:
try:
priorities += "%s," % self.processes[pid]["stat"]["rt_priority"]
except IOError:
# We're doing lazy loading of /proc files
# So if we get this exception is because the
# process vanished, remove it
del self.processes[pid]
processed_pids += pids
cpu += 1
priorities = priorities.strip(',')
return priorities
def is_bound_to_cpu(self, pid):
"""
Checks if a given pid can't have its SMP affinity mask changed.
"""
return self.processes[pid]["stat"].is_bound_to_cpu()
class interrupts:
"""
Information about IRQs in the system. A dictionary keyed by IRQ number
will have as its value another dictionary with "cpu", "type" and "users"
keys, with the SMP affinity mask, type of IRQ and the drivers associated
with each interrupt.
The information comes from the /proc/interrupts file, documented in
'man procfs(5)', for instance, the 'cpu' dict is an array with one entry
per CPU present in the sistem, each value being the number of interrupts
that took place per CPU.
E.g.:
>>> import procfs
>>> interrupts = procfs.interrupts()
>>> thunderbolt_irq = interrupts.find_by_user("thunderbolt")
>>> print thunderbolt_irq
34
>>> thunderbolt = interrupts[thunderbolt_irq]
>>> print thunderbolt
{'affinity': [0, 1, 2, 3], 'type': 'PCI-MSI', 'cpu': [3495, 0, 81, 0], 'users': ['thunderbolt']}
>>>
"""
def __init__(self):
self.interrupts = {}
self.reload()
def __getitem__(self, key):
return self.interrupts[str(key)]
def keys(self):
return list(self.interrupts.keys())
def values(self):
return list(self.interrupts.values())
def has_key(self, key):
return str(key) in self.interrupts
def items(self):
return self.interrupts
def __contains__(self, key):
return str(key) in self.interrupts
def reload(self):
del self.interrupts
self.interrupts = {}
f = open("/proc/interrupts")
for line in f.readlines():
line = line.strip()
fields = line.split()
if fields[0][:3] == "CPU":
self.nr_cpus = len(fields)
continue
irq = fields[0].strip(":")
self.interrupts[irq] = {}
self.interrupts[irq] = self.parse_entry(fields[1:], line)
try:
nirq = int(irq)
except:
continue
self.interrupts[irq]["affinity"] = self.parse_affinity(nirq)
f.close()
def parse_entry(self, fields, line):
dict = {}
dict["cpu"] = []
dict["cpu"].append(int(fields[0]))
nr_fields = len(fields)
if nr_fields >= self.nr_cpus:
dict["cpu"] += [int(i) for i in fields[1:self.nr_cpus]]
if nr_fields > self.nr_cpus:
dict["type"] = fields[self.nr_cpus]
# look if there are users (interrupts 3 and 4 haven't)
if nr_fields > self.nr_cpus + 1:
dict["users"] = [a.strip() for a in fields[nr_fields - 1].split(',')]
else:
dict["users"] = []
return dict
def parse_affinity(self, irq):
try:
f = open("/proc/irq/%s/smp_affinity" % irq)
line = f.readline()
f.close()
return bitmasklist(line, self.nr_cpus)
except IOError:
return [ 0, ]
def find_by_user(self, user):
"""
Looks up a interrupt number by the name of one of its users"
E.g.:
>>> import procfs
>>> interrupts = procfs.interrupts()
>>> thunderbolt_irq = interrupts.find_by_user("thunderbolt")
>>> print thunderbolt_irq
34
>>> thunderbolt = interrupts[thunderbolt_irq]
>>> print thunderbolt
{'affinity': [0, 1, 2, 3], 'type': 'PCI-MSI', 'cpu': [3495, 0, 81, 0], 'users': ['thunderbolt']}
>>>
"""
for i in self.interrupts.keys():
if "users" in self.interrupts[i] and \
user in self.interrupts[i]["users"]:
return i
return None
def find_by_user_regex(self, regex):
"""
Looks up a interrupt number by a regex that matches names of its users"
E.g.:
>>> import procfs
>>> import re
>>> interrupts = procfs.interrupts()
>>> usb_controllers = interrupts.find_by_user_regex(re.compile(".*hcd"))
>>> print usb_controllers
['22', '23', '31']
>>> print [ interrupts[irq]["users"] for irq in usb_controllers ]
[['ehci_hcd:usb4'], ['ehci_hcd:usb3'], ['xhci_hcd']]
>>>
"""
irqs = []
for i in self.interrupts.keys():
if "users" not in self.interrupts[i]:
continue
for user in self.interrupts[i]["users"]:
if regex.match(user):
irqs.append(i)
break
return irqs
class cmdline:
"""
Parses the kernel command line (/proc/cmdline), turning it into a dictionary."
Useful to figure out if some kernel boolean knob has been turned on, as well
as to find the value associated to other kernel knobs.
It can also be used to find out about parameters passed to the init process,
such as 'BOOT_IMAGE', etc.
E.g.:
>>> import procfs
>>> kcmd = procfs.cmdline()
>>> print kcmd.keys()
['LANG', 'BOOT_IMAGE', 'quiet', 'rhgb', 'rd.lvm.lv', 'ro', 'root']
>>> print kcmd["BOOT_IMAGE"]
/vmlinuz-4.3.0-rc1+
>>>
"""
def __init__(self):
self.options = {}
self.parse()
def parse(self):
f = open("/proc/cmdline")
for option in f.readline().strip().split():
fields = option.split("=")
if len(fields) == 1:
self.options[fields[0]] = True
else:
self.options[fields[0]] = fields[1]
f.close()
def __getitem__(self, key):
return self.options[key]
def keys(self):
return list(self.options.keys())
def values(self):
return list(self.options.values())
def items(self):
return self.options
class cpuinfo:
"""
Dictionary with information about CPUs in the system.
Please refer to 'man procfs(5)' for further information about the
'/proc/cpuinfo' file, that is the source of the information provided by this
class. The 'man lscpu(1)' also has information about a program that uses
the '/proc/cpuinfo' file.
Using this class one can obtain the number of CPUs in a system:
>>> cpus = procfs.cpuinfo()
>>> print cpus.nr_cpus
4
It is also possible to figure out aspects of the CPU topology, such as
how many CPU physical sockets exists, i.e. groups of CPUs sharing components
such as CPU memory caches:
>>> print len(cpus.sockets)
1
Additionally dictionary with information common to all CPUs in the system is
available:
>>> print cpus["model name"]
Intel(R) Core(TM) i7-3667U CPU @ 2.00GHz
>>> print cpus["cache size"]
4096 KB
>>>
"""
def __init__(self, filename="/proc/cpuinfo"):
self.tags = {}
self.nr_cpus = 0
self.sockets = []
self.parse(filename)
def __getitem__(self, key):
return self.tags[key.lower()]
def keys(self):
return list(self.tags.keys())
def values(self):
return list(self.tags.values())
def items(self):
return self.tags
def parse(self, filename):
f = open(filename)
for line in f.readlines():
line = line.strip()
if len(line) == 0:
continue
fields = line.split(":")
tagname = fields[0].strip().lower()
if tagname == "processor":
self.nr_cpus += 1
continue
elif tagname == "core id":
continue
self.tags[tagname] = fields[1].strip()
if tagname == "physical id":
socket_id = self.tags[tagname]
if socket_id not in self.sockets:
self.sockets.append(socket_id)
f.close()
self.nr_sockets = self.sockets and len(self.sockets) or \
(self.nr_cpus / ("siblings" in self.tags and int(self.tags["siblings"]) or 1))
self.nr_cores = ("cpu cores" in self.tags and int(self.tags["cpu cores"]) or 1) * self.nr_sockets
class smaps_lib:
"""
Representation of an mmap in place for a process. Can be used to figure
out which processes have an library mapped, etc.
The 'perm' member can be used to figure out executable mmaps, i.e. libraries.
The 'vm_start' and 'vm_end' in turn can be used when trying to resolve
processor instruction pointer addresses to a symbol name in a library.
"""
def __init__(self, lines):
fields = lines[0].split()
self.vm_start, self.vm_end = [int(a, 16) for a in fields[0].split("-")]
self.perms = fields[1]
self.offset = int(fields[2], 16)
self.major, self.minor = fields[3].split(":")
self.inode = int(fields[4])
if len(fields) > 5:
self.name = fields[5]
else:
self.name = None
self.tags = {}
for line in lines[1:]:
fields = line.split()
tag = fields[0][:-1].lower()
try:
self.tags[tag] = int(fields[1])
except:
# VmFlags are strings
self.tags[tag] = fields
def __getitem__(self, key):
return self.tags[key.lower()]
def keys(self):
return list(self.tags.keys())
def values(self):
return list(self.tags.values())
def items(self):
return self.tags
class smaps:
"""
List of libraries mapped by a process. Parses the lines in
the /proc/PID/smaps file, that is further documented in the
procfs(5) man page.
Example: Listing the executable maps for the 'sshd' process:
>>> import procfs
>>> processes = procfs.pidstats()
>>> sshd = processes.find_by_name("sshd")
>>> sshd_maps = procfs.smaps(sshd[0])
>>> for i in range(len(sshd_maps)):
... if 'x' in sshd_maps[i].perms:
... print "%s: %s" % (sshd_maps[i].name, sshd_maps[i].perms)
...
/usr/sbin/sshd: r-xp
/usr/lib64/libnss_files-2.20.so: r-xp
/usr/lib64/librt-2.20.so: r-xp
/usr/lib64/libkeyutils.so.1.5: r-xp
/usr/lib64/libkrb5support.so.0.1: r-xp
/usr/lib64/libfreebl3.so: r-xp
/usr/lib64/libpthread-2.20.so: r-xp
...
"""
def __init__(self, pid):
self.pid = pid
self.entries = []
self.reload()
def parse_entry(self, f, line):
lines = []
if not line:
line = f.readline().strip()
if not line:
return
lines.append(line)
while True:
line = f.readline()
if not line:
break
line = line.strip()
if line.split()[0][-1] == ':':
lines.append(line)
else:
break
self.entries.append(smaps_lib(lines))
return line
def __len__(self):
return len(self.entries)
def __getitem__(self, index):
return self.entries[index]
def reload(self):
f = open("/proc/%d/smaps" % self.pid)
line = None
while True:
line = self.parse_entry(f, line)
if not line:
break
f.close()
self.nr_entries = len(self.entries)
def find_by_name_fragment(self, fragment):
result = []
for i in range(self.nr_entries):
if self.entries[i].name and \
self.entries[i].name.find(fragment) >= 0:
result.append(self.entries[i])
return result
class cpustat:
"""
CPU statistics, obtained from a line in the '/proc/stat' file, Please
refer to 'man procfs(5)' for further information about the '/proc/stat'
file, that is the source of the information provided by this class.
"""
def __init__(self, fields):
self.name = fields[0]
(self.user,
self.nice,
self.system,
self.idle,
self.iowait,
self.irq,
self.softirq) = [int(i) for i in fields[1:8]]
if len(fields) > 7:
self.steal = int(fields[7])
if len(fields) > 8:
self.guest = int(fields[8])
def __repr__(self):
s = "< user: %s, nice: %s, system: %s, idle: %s, iowait: %s, irq: %s, softirq: %s" % \
(self.user, self.nice, self.system, self.idle, self.iowait, self.irq, self.softirq)
if hasattr(self, 'steal'):
s += ", steal: %d" % self.steal
if hasattr(self, 'guest'):
s += ", guest: %d" % self.guest
return s + ">"
class cpusstats:
"""
Dictionary with information about CPUs in the system. First entry in the
dictionary gives an aggregate view of all CPUs, each other entry is about
separate CPUs. Please refer to 'man procfs(5)' for further information
about the '/proc/stat' file, that is the source of the information provided
by this class.
"""
def __init__(self, filename = "/proc/stat"):
self.entries = {}
self.time = None
self.hertz = os.sysconf(2)
self.filename = filename
self.reload()
def __iter__(self):
return iter(self.entries)
def __getitem__(self, key):
return self.entries[key]
def __len__(self):
return len(list(self.entries.keys()))
def keys(self):
return list(self.entries.keys())
def values(self):
return list(self.entries.values())
def items(self):
return self.entries
def reload(self):
last_entries = self.entries
self.entries = {}
f = open(self.filename)
for line in f.readlines():
fields = line.strip().split()
if fields[0][:3].lower() != "cpu":
continue
c = cpustat(fields)
if c.name == "cpu":
idx = 0
else:
idx = int(c.name[3:]) + 1
self.entries[idx] = c
f.close()
last_time = self.time
self.time = time.time()
if last_entries:
delta_sec = self.time - last_time
interval_hz = delta_sec * self.hertz
for cpu in self.entries.keys():
if cpu not in last_entries:
curr.usage = 0
continue
curr = self.entries[cpu]
prev = last_entries[cpu]
delta = (curr.user - prev.user) + \
(curr.nice - prev.nice) + \
(curr.system - prev.system)
curr.usage = (delta / interval_hz) * 100
if curr.usage > 100:
curr.usage = 100
if __name__ == '__main__':
import sys
ints = interrupts()
for i in ints.interrupts.keys():
print("%s: %s" % (i, ints.interrupts[i]))
options = cmdline()
for o in options.options.keys():
print("%s: %s" % (o, options.options[o]))
cpu = cpuinfo()
print("\ncpuinfo data: %d processors" % cpu.nr_cpus)
for tag in cpu.keys():
print("%s=%s" % (tag, cpu[tag]))
print("smaps:\n" + ("-" * 40))
s = smaps(int(sys.argv[1]))
for i in range(s.nr_entries):
print("%#x %s" % (s.entries[i].vm_start, s.entries[i].name))
print("-" * 40)
for a in s.find_by_name_fragment(sys.argv[2]):
print(a["Size"])
ps = pidstats()
print(ps[1])
cs = cpusstats()
while True:
time.sleep(1)
cs.reload()
for cpu in cs:
print("%s: %d" % (cpu.name, cpu.usage))
print("-" * 10)
/*
* =====================================================================================
*
* Filename: proc_more.c
*
* Description:
*
* Version: 1.0
* Created: 2013年08月07日 11时34分40秒
* Revision: none
* Compiler: gcc
*
* Author: linkscue (scue), linkscue@gmail.com
* Organization:
*
* =====================================================================================
*/
#include <stdio.h>
#include <sys/time.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <unistd.h>
#include <fcntl.h>
#include <stdlib.h>
#include <string.h>
void sampleLoadAvg(){
int f=0;
char buffer[80]=""; /* 定义字符串并初始化为'\0' */
char buf[5][10];
char *file="/proc/loadavg";
f = open(file, O_RDONLY);
if (f == 0)
{
printf("error to open: %s\n", file);
exit(EXIT_FAILURE);
}
read(f, (void *)buffer, 80);
sscanf(buffer, "%s %s %s %s %s", /* sscanf()拆分成多个字符串 */
&buf[0],&buf[1],&buf[2],&buf[3],&buf[4]);
printf("一分钟平均负载：%s\n", buf[0]);
printf("五分钟平均负载：%s\n", buf[1]);
printf("一刻钟平均负载：%s\n", buf[2]);
printf("采样时刻的间隔：%s\n", buf[3]);
printf("最大线程的数目：%s\n", buf[4]);
close(f);
}
void sampleTime(){
int f=0;
char buffer[80]="";
char buf[2][10];
float seconds=0;
float secondr=0;
char *file="/proc/uptime";
f = open(file, O_RDONLY);
if (f == 0)
{
printf("error to open: %s\n", file);
exit(EXIT_FAILURE);
}
read(f, (void *)buffer, 80);
sscanf(buffer, "%s %s", &buf[0], &buf[1]);
close(f);
printf("系统运行时间：\t%s秒\n", buf[0]);
printf("系统空闲时间：\t%s秒\n", buf[1]);
close(f);
seconds=strtof(buf[0],NULL);
secondr=strtof(buf[1],NULL);
printf("系统运行时间：\t%03d天%02d时%02d分%02.6f秒\n",
(int)seconds/(24*3600), /* 天 */
((int)seconds/(3600))%(24), /* 时 */
((int)seconds/60)%60, /* 分 */
((int)seconds%60)+(seconds-(int)seconds)); /* 秒(精确至毫秒) */
printf("系统空闲时间：\t%03d天%02d时%02d分%02.6f秒\n",
(int)secondr/(24*3600), /* 天 */
((int)secondr/(3600))%(24), /* 时 */
((int)secondr/60)%60, /* 分 */
((int)secondr%60)+(secondr-(int)secondr)); /* 秒(精确至毫秒) */
}
void sampleKernelVersion(){
int f=0;
char buffer[80]="";
char *file="/proc/sys/kernel/version";
f = open(file, O_RDONLY);
if (f == 0)
{
printf("error to open: %s\n", file);
exit(EXIT_FAILURE);
}
read(f, (void *)buffer, 80);
buffer[strlen(buffer)-1]=0; /* 简单实现tr()函数的功能 */
printf("当前内核版本：\t%s\n", buffer);
close(f);
}
void sampleOsRelease(){
int f=0;
char buffer[80]="";
char *file="/proc/sys/kernel/osrelease";
f = open(file, O_RDONLY);
if (f == 0)
{
printf("error to open: %s\n", file);
exit(EXIT_FAILURE);
}
read(f, (void *)buffer, 80);
buffer[strlen(buffer)-1]=0;
printf("当前发行版本：\t%s\n", buffer);
close(f);
}
void sampleOsType(){
int f=0;
char buffer[80]="";
char *file="/proc/sys/kernel/ostype";
f = open(file, O_RDONLY);
if (f == 0)
{
printf("error to open: %s\n", file);
exit(EXIT_FAILURE);
}
read(f, (void *)buffer, 80);
buffer[strlen(buffer)-1]=0;
printf("当前操作系统：\t%s\n", buffer);
close(f);
}
void sampleDiskStat(){
int i;
FILE *fp;
int nread=0;
ssize_t len = 0;
char *buffer=NULL;
char buf[20][32];
char *file="/proc/diskstats";
char *p;
fp = fopen(file, "rb");
if (fp == NULL)
{
printf("error to open: %s\n", file);
exit(EXIT_FAILURE);
}
printf(" 磁盘 读次数 写次数\n");
while((nread = getline(&buffer, &len, fp)) != -1) { /* 简单实现读行的功能 */
sscanf(buffer, "%04s%08s%s %s %s %s %s %s %s %s %s %s %s %s",
&buf[0],&buf[1],&buf[2],&buf[3],&buf[4],&buf[5],&buf[6],
&buf[7],&buf[8],&buf[9],&buf[10],&buf[11],&buf[12],&buf[13]);
if ((p=strstr(buf[2], "loop"))!=NULL)
{
; /* loop本地回路不作操作 */
}
else {
printf("%06s%08s%08s\n",
&buf[2],&buf[3], &buf[7]);
}
}
}
void sampleProcesses(void)
{
FILE *fp;
int nread=0;
ssize_t len = 0;
char *buf=NULL;
char *buffer=NULL;
char *file="/proc/stat";
fp = fopen(file, "rb");
if (fp == NULL)
{
printf("error to open: %s\n", file);
exit(EXIT_FAILURE);
}
while((nread = getline(&buffer, &len, fp)) != -1) {
if((buf=strstr(buffer, "processes"))!=NULL) /* 简单实现grep的功能 */
break;
}
buffer[strlen(buffer)-1]=0; /* 简单实现tr()函数的功能 */
char count[16]="";
sscanf(buffer, "%s%s", count, count);
printf("执行线程数目:\t%s\n", count);
}
void sampleContext(void)
{
FILE *fp;
int nread=0;
ssize_t len = 0;
char *buf=NULL;
char *buffer=NULL;
char *file="/proc/stat";
fp = fopen(file, "rb");
if (fp == NULL)
{
printf("error to open: %s\n", file);
exit(EXIT_FAILURE);
}
while((nread = getline(&buffer, &len, fp)) != -1) {
if((buf=strstr(buffer, "ctxt"))!=NULL) /* 简单实现grep的功能 */
break;
}
buffer[strlen(buffer)-1]=0; /* 简单实现tr()函数的功能 */
char count[16]="";
sscanf(buffer, "%s%s", count, count);
printf("上下文切换次数:\t%s\n", count);
}
void sampleMeminfo()
{
FILE *fp;
int nread=0;
ssize_t len = 0;
char *buf=NULL;
char *buffer=NULL;
char *file="/proc/meminfo";
char content[16]="";
fp = fopen(file, "rb");
if (fp == NULL)
{
printf("error to open: %s\n", file);
exit(EXIT_FAILURE);
}
while((nread = getline(&buffer, &len, fp)) != -1) {
if((buf=strstr(buffer, "MemTotal"))!=NULL) /* 简单实现grep的功能 */
{
buffer[strlen(buffer)-1]=0; /* 简单实现tr()函数的功能 */
sscanf(buffer, "%s%s", content, content);
int memtotal_kb=(int)(strtof(content, NULL));
printf("内存总容量:\t%dG%4dM %4dK\n",
memtotal_kb/(1024*1024), /* Gb */
(memtotal_kb/(1024))%1024, /* Mb */
(memtotal_kb%(1024*1024))%1024); /* Kb */
}
if((buf=strstr(buffer, "MemFree"))!=NULL) /* 简单实现grep的功能 */
{
buffer[strlen(buffer)-1]=0; /* 简单实现tr()函数的功能 */
sscanf(buffer, "%s%s", content, content);
int memfree_kb=(int)(strtof(content, NULL));
printf("内存可用容量:\t%dG%4dM %4dK\n",
memfree_kb/(1024*1024), /* Gb */
(memfree_kb/(1024))%1024, /* Mb */
(memfree_kb%(1024*1024))%1024); /* Kb */
}
if((buf=strstr(buffer, "SwapTotal"))!=NULL) /* 简单实现grep的功能 */
{
buffer[strlen(buffer)-1]=0; /* 简单实现tr()函数的功能 */
sscanf(buffer, "%s%s", content, content);
int swaptotal_kb=(int)(strtof(content, NULL));
printf("SWAP总容量:\t%dG%4dM %4dK\n",
swaptotal_kb/(1024*1024), /* Gb */
(swaptotal_kb/(1024))%1024, /* Mb */
(swaptotal_kb%(1024*1024))%1024); /* Kb */
}
if((buf=strstr(buffer, "SwapFree"))!=NULL) /* 简单实现grep的功能 */
{
buffer[strlen(buffer)-1]=0; /* 简单实现tr()函数的功能 */
sscanf(buffer, "%s%s", content, content);
int swapfree_kb=(int)(strtof(content, NULL));
printf("SWAP可用容量:\t%dG%4dM %4dK\n",
swapfree_kb/(1024*1024), /* Gb */
(swapfree_kb/(1024))%1024, /* Mb */
(swapfree_kb%(1024*1024))%1024); /* Kb */
break; /* 所需的信息已满足，退出循环 */
}
}
}
/*
* === FUNCTION ======================================================================
* Name: main
* Description:
* =====================================================================================
*/
int main ( int argc, char *argv[] )
{
printf(">>>系统负载<<<\n");
sampleLoadAvg();
printf("----------------------------------------\n");
printf("\n");
printf(">>>运行时间<<<\n");
sampleTime();
printf("\n");
printf(">>>版本信息<<<\n");
sampleOsType();
sampleOsRelease();
sampleKernelVersion();
printf("----------------------------------------\n");
printf("\n");
printf(">>>磁盘信息<<<\n");
sampleDiskStat();
printf("----------------------------------------\n");
printf("\n");
printf(">>>上下文切换<<<\n");
sampleContext();
printf("----------------------------------------\n");
printf("\n");
printf(">>>线程数目<<<\n");
sampleProcesses();
printf("----------------------------------------\n");
printf("\n");
printf(">>>内存信息<<<\n");
sampleMeminfo();
printf("----------------------------------------\n");
return EXIT_SUCCESS;
}
>>>系统负载<<< 一分钟平均负载：0.00 五分钟平均负载：0.02 一刻钟平均负载：0.05 采样时刻的间隔：2/196 最大线程的数目：26046 ---------------------------------------- >>>运行时间<<< 系统运行时间： 5206300.48427660.03秒 系统空闲时间： 427660.03秒 系统运行时间： 060天06时11分40.500000秒 系统空闲时间： 004天22时47分40.031250秒 >>>版本信息<<< 当前操作系统： Linux 当前发行版本： 3.10.0-957.21.3.el7.x86_64 当前内核版本： #1 SMP Tue Jun 18 16:35:19 UTC 2019 ---------------------------------------- >>>磁盘信息<<< 磁盘 读次数 写次数 vda 195840 4341388 vda1 195716 4306060 ---------------------------------------- >>>上下文切换<<< 上下文切换次数: 10188051629 ---------------------------------------- >>>线程数目<<< 执行线程数目: 5001714 ---------------------------------------- >>>内存信息<<< 内存总容量: 3G 717M 468K 内存可用容量: 0G 718M 920K SWAP总容量: 0G 0M 0K SWAP可用容量: 0G 0M 0K ----------------------------------------
〖Linux〗proc的一些文件读取操作 - scue - 博客园 https://www.cnblogs.com/scue/p/3244753.html
http://www.linuxhowtos.org/System/procstat.htm
#include <stdio.h>
#include <stdlib.h>
#define _GNU_SOURCE
int main(void)
{
long double res1[4], res2[4], load_avg;
FILE *file_pointer;
for(;;)
{
file_pointer = fopen("/proc/stat","r");
fscanf(file_pointer,"%*s %Lf %Lf %Lf %Lf",&res1[0],&res1[1],&res1[2],&res1[3]);
fclose(file_pointer);
sleep(1);
file_pointer = fopen("/proc/stat","r");
fscanf(file_pointer,"%*s %Lf %Lf %Lf %Lf",&res2[0],&res2[1],&res2[2],&res2[3]);
fclose(file_pointer);
load_avg = ((res2[0]+res2[1]+res2[2]) - (res1[0]+res1[1]+res1[2])) / ((res2[0]+res2[1]+res2[2]+res2[3]) - (res1[0]+res1[1]+res1[2]+res1[3]));
fprintf(stderr,"%2.3Lf %%\n", 100*load_avg);
}
return(0);
}
[root@offical perfUcan]# ./getcpuusage
1.005 %
1.500 %
1.508 %
1.005 %
1.500 %
1.508 %
1.500 %
top命令源码分析 - panwanning的博客 - CSDN博客 https://blog.csdn.net/panwanning/article/details/86766494
