1 ------------------------------------------------------------------------------
2 T H E /proc F I L E S Y S T E M
3 ------------------------------------------------------------------------------
9 ------------------------------------------------------------------------------
10 Version 1.3 Kernel version 2.2.12
11 Kernel version 2.4.0-test11-pre4
12 ------------------------------------------------------------------------------
19 0.1 Introduction/Credits
22 1 Collecting System Information
23 1.1 Process-Specific Subdirectories
25 1.3 IDE devices in /proc/ide
26 1.4 Networking info in /proc/net
28 1.6 Parallel port info in /proc/parport
29 1.7 TTY info in /proc/tty
30 1.8 Miscellaneous kernel statistics in /proc/stat
31 1.9 Ext4 file system parameters
33 2 Modifying System Parameters
35 3 Per-Process Parameters
36 3.1 /proc/<pid>/oom_adj & /proc/<pid>/oom_score_adj - Adjust the oom-killer
38 3.2 /proc/<pid>/oom_score - Display current oom-killer score
39 3.3 /proc/<pid>/io - Display the IO accounting fields
40 3.4 /proc/<pid>/coredump_filter - Core dump filtering settings
41 3.5 /proc/<pid>/mountinfo - Information about mounts
42 3.6 /proc/<pid>/comm & /proc/<pid>/task/<tid>/comm
43 3.7 /proc/<pid>/task/<tid>/children - Information about task children
48 ------------------------------------------------------------------------------
50 ------------------------------------------------------------------------------
52 0.1 Introduction/Credits
53 ------------------------
55 This documentation is part of a soon (or so we hope) to be released book on
56 the SuSE Linux distribution. As there is no complete documentation for the
57 /proc file system and we've used many freely available sources to write these
58 chapters, it seems only fair to give the work back to the Linux community.
59 This work is based on the 2.2.* kernel version and the upcoming 2.4.*. I'm
60 afraid it's still far from complete, but we hope it will be useful. As far as
61 we know, it is the first 'all-in-one' document about the /proc file system. It
62 is focused on the Intel x86 hardware, so if you are looking for PPC, ARM,
63 SPARC, AXP, etc., features, you probably won't find what you are looking for.
64 It also only covers IPv4 networking, not IPv6 nor other protocols - sorry. But
65 additions and patches are welcome and will be added to this document if you
68 We'd like to thank Alan Cox, Rik van Riel, and Alexey Kuznetsov and a lot of
69 other people for help compiling this documentation. We'd also like to extend a
70 special thank you to Andi Kleen for documentation, which we relied on heavily
71 to create this document, as well as the additional information he provided.
72 Thanks to everybody else who contributed source or docs to the Linux kernel
73 and helped create a great piece of software... :)
75 If you have any comments, corrections or additions, please don't hesitate to
79 The latest version of this document is available online at
80 http://tldp.org/LDP/Linux-Filesystem-Hierarchy/html/proc.html
82 If the above direction does not works for you, you could try the kernel
89 We don't guarantee the correctness of this document, and if you come to us
90 complaining about how you screwed up your system because of incorrect
91 documentation, we won't feel responsible...
93 ------------------------------------------------------------------------------
94 CHAPTER 1: COLLECTING SYSTEM INFORMATION
95 ------------------------------------------------------------------------------
97 ------------------------------------------------------------------------------
99 ------------------------------------------------------------------------------
100 * Investigating the properties of the pseudo file system /proc and its
101 ability to provide information on the running Linux system
102 * Examining /proc's structure
103 * Uncovering various information about the kernel and the processes running
105 ------------------------------------------------------------------------------
108 The proc file system acts as an interface to internal data structures in the
109 kernel. It can be used to obtain information about the system and to change
110 certain kernel parameters at runtime (sysctl).
112 First, we'll take a look at the read-only parts of /proc. In Chapter 2, we
113 show you how you can use /proc/sys to change settings.
115 1.1 Process-Specific Subdirectories
116 -----------------------------------
118 The directory /proc contains (among other things) one subdirectory for each
119 process running on the system, which is named after the process ID (PID).
121 The link self points to the process reading the file system. Each process
122 subdirectory has the entries listed in Table 1-1.
125 Table 1-1: Process specific entries in /proc
126 ..............................................................................
128 clear_refs Clears page referenced bits shown in smaps output
129 cmdline Command line arguments
130 cpu Current and last cpu in which it was executed (2.4)(smp)
131 cwd Link to the current working directory
132 environ Values of environment variables
133 exe Link to the executable of this process
134 fd Directory, which contains all file descriptors
135 maps Memory maps to executables and library files (2.4)
136 mem Memory held by this process
137 root Link to the root directory of this process
139 statm Process memory status information
140 status Process status in human readable form
141 wchan If CONFIG_KALLSYMS is set, a pre-decoded wchan
143 stack Report full stack trace, enable via CONFIG_STACKTRACE
144 smaps a extension based on maps, showing the memory consumption of
146 ..............................................................................
148 For example, to get the status information of a process, all you have to do is
149 read the file /proc/PID/status:
151 >cat /proc/self/status
175 SigPnd: 0000000000000000
176 ShdPnd: 0000000000000000
177 SigBlk: 0000000000000000
178 SigIgn: 0000000000000000
179 SigCgt: 0000000000000000
180 CapInh: 00000000fffffeff
181 CapPrm: 0000000000000000
182 CapEff: 0000000000000000
183 CapBnd: ffffffffffffffff
184 voluntary_ctxt_switches: 0
185 nonvoluntary_ctxt_switches: 1
187 This shows you nearly the same information you would get if you viewed it with
188 the ps command. In fact, ps uses the proc file system to obtain its
189 information. But you get a more detailed view of the process by reading the
190 file /proc/PID/status. It fields are described in table 1-2.
192 The statm file contains more detailed information about the process
193 memory usage. Its seven fields are explained in Table 1-3. The stat file
194 contains details information about the process itself. Its fields are
195 explained in Table 1-4.
197 (for SMP CONFIG users)
198 For making accounting scalable, RSS related information are handled in
199 asynchronous manner and the vaule may not be very precise. To see a precise
200 snapshot of a moment, you can see /proc/<pid>/smaps file and scan page table.
201 It's slow but very precise.
203 Table 1-2: Contents of the status files (as of 2.6.30-rc7)
204 ..............................................................................
206 Name filename of the executable
207 State state (R is running, S is sleeping, D is sleeping
208 in an uninterruptible wait, Z is zombie,
209 T is traced or stopped)
212 PPid process id of the parent process
213 TracerPid PID of process tracing this process (0 if not)
214 Uid Real, effective, saved set, and file system UIDs
215 Gid Real, effective, saved set, and file system GIDs
216 FDSize number of file descriptor slots currently allocated
217 Groups supplementary group list
218 VmPeak peak virtual memory size
219 VmSize total program size
220 VmLck locked memory size
221 VmHWM peak resident set size ("high water mark")
222 VmRSS size of memory portions
223 VmData size of data, stack, and text segments
224 VmStk size of data, stack, and text segments
225 VmExe size of text segment
226 VmLib size of shared library code
227 VmPTE size of page table entries
228 VmSwap size of swap usage (the number of referred swapents)
229 Threads number of threads
230 SigQ number of signals queued/max. number for queue
231 SigPnd bitmap of pending signals for the thread
232 ShdPnd bitmap of shared pending signals for the process
233 SigBlk bitmap of blocked signals
234 SigIgn bitmap of ignored signals
235 SigCgt bitmap of catched signals
236 CapInh bitmap of inheritable capabilities
237 CapPrm bitmap of permitted capabilities
238 CapEff bitmap of effective capabilities
239 CapBnd bitmap of capabilities bounding set
240 Cpus_allowed mask of CPUs on which this process may run
241 Cpus_allowed_list Same as previous, but in "list format"
242 Mems_allowed mask of memory nodes allowed to this process
243 Mems_allowed_list Same as previous, but in "list format"
244 voluntary_ctxt_switches number of voluntary context switches
245 nonvoluntary_ctxt_switches number of non voluntary context switches
246 ..............................................................................
248 Table 1-3: Contents of the statm files (as of 2.6.8-rc3)
249 ..............................................................................
251 size total program size (pages) (same as VmSize in status)
252 resident size of memory portions (pages) (same as VmRSS in status)
253 shared number of pages that are shared (i.e. backed by a file)
254 trs number of pages that are 'code' (not including libs; broken,
255 includes data segment)
256 lrs number of pages of library (always 0 on 2.6)
257 drs number of pages of data/stack (including libs; broken,
258 includes library text)
259 dt number of dirty pages (always 0 on 2.6)
260 ..............................................................................
263 Table 1-4: Contents of the stat files (as of 2.6.30-rc7)
264 ..............................................................................
267 tcomm filename of the executable
268 state state (R is running, S is sleeping, D is sleeping in an
269 uninterruptible wait, Z is zombie, T is traced or stopped)
270 ppid process id of the parent process
271 pgrp pgrp of the process
273 tty_nr tty the process uses
274 tty_pgrp pgrp of the tty
276 min_flt number of minor faults
277 cmin_flt number of minor faults with child's
278 maj_flt number of major faults
279 cmaj_flt number of major faults with child's
280 utime user mode jiffies
281 stime kernel mode jiffies
282 cutime user mode jiffies with child's
283 cstime kernel mode jiffies with child's
284 priority priority level
286 num_threads number of threads
287 it_real_value (obsolete, always 0)
288 start_time time the process started after system boot
289 vsize virtual memory size
290 rss resident set memory size
291 rsslim current limit in bytes on the rss
292 start_code address above which program text can run
293 end_code address below which program text can run
294 start_stack address of the start of the main process stack
295 esp current value of ESP
296 eip current value of EIP
297 pending bitmap of pending signals
298 blocked bitmap of blocked signals
299 sigign bitmap of ignored signals
300 sigcatch bitmap of catched signals
301 wchan address where process went to sleep
304 exit_signal signal to send to parent thread on exit
305 task_cpu which CPU the task is scheduled on
306 rt_priority realtime priority
307 policy scheduling policy (man sched_setscheduler)
308 blkio_ticks time spent waiting for block IO
309 gtime guest time of the task in jiffies
310 cgtime guest time of the task children in jiffies
311 start_data address above which program data+bss is placed
312 end_data address below which program data+bss is placed
313 start_brk address above which program heap can be expanded with brk()
314 arg_start address above which program command line is placed
315 arg_end address below which program command line is placed
316 env_start address above which program environment is placed
317 env_end address below which program environment is placed
318 exit_code the thread's exit_code in the form reported by the waitpid system call
319 ..............................................................................
321 The /proc/PID/maps file containing the currently mapped memory regions and
322 their access permissions.
326 address perms offset dev inode pathname
328 08048000-08049000 r-xp 00000000 03:00 8312 /opt/test
329 08049000-0804a000 rw-p 00001000 03:00 8312 /opt/test
330 0804a000-0806b000 rw-p 00000000 00:00 0 [heap]
331 a7cb1000-a7cb2000 ---p 00000000 00:00 0
332 a7cb2000-a7eb2000 rw-p 00000000 00:00 0
333 a7eb2000-a7eb3000 ---p 00000000 00:00 0
334 a7eb3000-a7ed5000 rw-p 00000000 00:00 0 [stack:1001]
335 a7ed5000-a8008000 r-xp 00000000 03:00 4222 /lib/libc.so.6
336 a8008000-a800a000 r--p 00133000 03:00 4222 /lib/libc.so.6
337 a800a000-a800b000 rw-p 00135000 03:00 4222 /lib/libc.so.6
338 a800b000-a800e000 rw-p 00000000 00:00 0
339 a800e000-a8022000 r-xp 00000000 03:00 14462 /lib/libpthread.so.0
340 a8022000-a8023000 r--p 00013000 03:00 14462 /lib/libpthread.so.0
341 a8023000-a8024000 rw-p 00014000 03:00 14462 /lib/libpthread.so.0
342 a8024000-a8027000 rw-p 00000000 00:00 0
343 a8027000-a8043000 r-xp 00000000 03:00 8317 /lib/ld-linux.so.2
344 a8043000-a8044000 r--p 0001b000 03:00 8317 /lib/ld-linux.so.2
345 a8044000-a8045000 rw-p 0001c000 03:00 8317 /lib/ld-linux.so.2
346 aff35000-aff4a000 rw-p 00000000 00:00 0 [stack]
347 ffffe000-fffff000 r-xp 00000000 00:00 0 [vdso]
349 where "address" is the address space in the process that it occupies, "perms"
350 is a set of permissions:
356 p = private (copy on write)
358 "offset" is the offset into the mapping, "dev" is the device (major:minor), and
359 "inode" is the inode on that device. 0 indicates that no inode is associated
360 with the memory region, as the case would be with BSS (uninitialized data).
361 The "pathname" shows the name associated file for this mapping. If the mapping
362 is not associated with a file:
364 [heap] = the heap of the program
365 [stack] = the stack of the main process
366 [stack:1001] = the stack of the thread with tid 1001
367 [vdso] = the "virtual dynamic shared object",
368 the kernel system call handler
370 or if empty, the mapping is anonymous.
372 The /proc/PID/task/TID/maps is a view of the virtual memory from the viewpoint
373 of the individual tasks of a process. In this file you will see a mapping marked
374 as [stack] if that task sees it as a stack. This is a key difference from the
375 content of /proc/PID/maps, where you will see all mappings that are being used
376 as stack by all of those tasks. Hence, for the example above, the task-level
377 map, i.e. /proc/PID/task/TID/maps for thread 1001 will look like this:
379 08048000-08049000 r-xp 00000000 03:00 8312 /opt/test
380 08049000-0804a000 rw-p 00001000 03:00 8312 /opt/test
381 0804a000-0806b000 rw-p 00000000 00:00 0 [heap]
382 a7cb1000-a7cb2000 ---p 00000000 00:00 0
383 a7cb2000-a7eb2000 rw-p 00000000 00:00 0
384 a7eb2000-a7eb3000 ---p 00000000 00:00 0
385 a7eb3000-a7ed5000 rw-p 00000000 00:00 0 [stack]
386 a7ed5000-a8008000 r-xp 00000000 03:00 4222 /lib/libc.so.6
387 a8008000-a800a000 r--p 00133000 03:00 4222 /lib/libc.so.6
388 a800a000-a800b000 rw-p 00135000 03:00 4222 /lib/libc.so.6
389 a800b000-a800e000 rw-p 00000000 00:00 0
390 a800e000-a8022000 r-xp 00000000 03:00 14462 /lib/libpthread.so.0
391 a8022000-a8023000 r--p 00013000 03:00 14462 /lib/libpthread.so.0
392 a8023000-a8024000 rw-p 00014000 03:00 14462 /lib/libpthread.so.0
393 a8024000-a8027000 rw-p 00000000 00:00 0
394 a8027000-a8043000 r-xp 00000000 03:00 8317 /lib/ld-linux.so.2
395 a8043000-a8044000 r--p 0001b000 03:00 8317 /lib/ld-linux.so.2
396 a8044000-a8045000 rw-p 0001c000 03:00 8317 /lib/ld-linux.so.2
397 aff35000-aff4a000 rw-p 00000000 00:00 0
398 ffffe000-fffff000 r-xp 00000000 00:00 0 [vdso]
400 The /proc/PID/smaps is an extension based on maps, showing the memory
401 consumption for each of the process's mappings. For each of mappings there
402 is a series of lines such as the following:
404 08048000-080bc000 r-xp 00000000 03:02 13130 /bin/bash
419 The first of these lines shows the same information as is displayed for the
420 mapping in /proc/PID/maps. The remaining lines show the size of the mapping
421 (size), the amount of the mapping that is currently resident in RAM (RSS), the
422 process' proportional share of this mapping (PSS), the number of clean and
423 dirty private pages in the mapping. Note that even a page which is part of a
424 MAP_SHARED mapping, but has only a single pte mapped, i.e. is currently used
425 by only one process, is accounted as private and not as shared. "Referenced"
426 indicates the amount of memory currently marked as referenced or accessed.
427 "Anonymous" shows the amount of memory that does not belong to any file. Even
428 a mapping associated with a file may contain anonymous pages: when MAP_PRIVATE
429 and a page is modified, the file page is replaced by a private anonymous copy.
430 "Swap" shows how much would-be-anonymous memory is also used, but out on
433 This file is only present if the CONFIG_MMU kernel configuration option is
436 The /proc/PID/clear_refs is used to reset the PG_Referenced and ACCESSED/YOUNG
437 bits on both physical and virtual pages associated with a process.
438 To clear the bits for all the pages associated with the process
439 > echo 1 > /proc/PID/clear_refs
441 To clear the bits for the anonymous pages associated with the process
442 > echo 2 > /proc/PID/clear_refs
444 To clear the bits for the file mapped pages associated with the process
445 > echo 3 > /proc/PID/clear_refs
446 Any other value written to /proc/PID/clear_refs will have no effect.
448 The /proc/pid/pagemap gives the PFN, which can be used to find the pageflags
449 using /proc/kpageflags and number of times a page is mapped using
450 /proc/kpagecount. For detailed explanation, see Documentation/vm/pagemap.txt.
455 Similar to the process entries, the kernel data files give information about
456 the running kernel. The files used to obtain this information are contained in
457 /proc and are listed in Table 1-5. Not all of these will be present in your
458 system. It depends on the kernel configuration and the loaded modules, which
459 files are there, and which are missing.
461 Table 1-5: Kernel info in /proc
462 ..............................................................................
464 apm Advanced power management info
465 buddyinfo Kernel memory allocator information (see text) (2.5)
466 bus Directory containing bus specific information
467 cmdline Kernel command line
468 cpuinfo Info about the CPU
469 devices Available devices (block and character)
470 dma Used DMS channels
471 filesystems Supported filesystems
472 driver Various drivers grouped here, currently rtc (2.4)
473 execdomains Execdomains, related to security (2.4)
474 fb Frame Buffer devices (2.4)
475 fs File system parameters, currently nfs/exports (2.4)
476 ide Directory containing info about the IDE subsystem
477 interrupts Interrupt usage
478 iomem Memory map (2.4)
479 ioports I/O port usage
480 irq Masks for irq to cpu affinity (2.4)(smp?)
481 isapnp ISA PnP (Plug&Play) Info (2.4)
482 kcore Kernel core image (can be ELF or A.OUT(deprecated in 2.4))
484 ksyms Kernel symbol table
485 loadavg Load average of last 1, 5 & 15 minutes
489 modules List of loaded modules
490 mounts Mounted filesystems
491 net Networking info (see text)
492 pagetypeinfo Additional page allocator information (see text) (2.5)
493 partitions Table of partitions known to the system
494 pci Deprecated info of PCI bus (new way -> /proc/bus/pci/,
495 decoupled by lspci (2.4)
497 scsi SCSI info (see text)
498 slabinfo Slab pool info
499 softirqs softirq usage
500 stat Overall statistics
501 swaps Swap space utilization
503 sysvipc Info of SysVIPC Resources (msg, sem, shm) (2.4)
504 tty Info of tty drivers
506 version Kernel version
507 video bttv info of video resources (2.4)
508 vmallocinfo Show vmalloced areas
509 ..............................................................................
511 You can, for example, check which interrupts are currently in use and what
512 they are used for by looking in the file /proc/interrupts:
514 > cat /proc/interrupts
516 0: 8728810 XT-PIC timer
517 1: 895 XT-PIC keyboard
519 3: 531695 XT-PIC aha152x
520 4: 2014133 XT-PIC serial
521 5: 44401 XT-PIC pcnet_cs
524 12: 182918 XT-PIC PS/2 Mouse
526 14: 1232265 XT-PIC ide0
530 In 2.4.* a couple of lines where added to this file LOC & ERR (this time is the
531 output of a SMP machine):
533 > cat /proc/interrupts
536 0: 1243498 1214548 IO-APIC-edge timer
537 1: 8949 8958 IO-APIC-edge keyboard
538 2: 0 0 XT-PIC cascade
539 5: 11286 10161 IO-APIC-edge soundblaster
540 8: 1 0 IO-APIC-edge rtc
541 9: 27422 27407 IO-APIC-edge 3c503
542 12: 113645 113873 IO-APIC-edge PS/2 Mouse
544 14: 22491 24012 IO-APIC-edge ide0
545 15: 2183 2415 IO-APIC-edge ide1
546 17: 30564 30414 IO-APIC-level eth0
547 18: 177 164 IO-APIC-level bttv
552 NMI is incremented in this case because every timer interrupt generates a NMI
553 (Non Maskable Interrupt) which is used by the NMI Watchdog to detect lockups.
555 LOC is the local interrupt counter of the internal APIC of every CPU.
557 ERR is incremented in the case of errors in the IO-APIC bus (the bus that
558 connects the CPUs in a SMP system. This means that an error has been detected,
559 the IO-APIC automatically retry the transmission, so it should not be a big
560 problem, but you should read the SMP-FAQ.
562 In 2.6.2* /proc/interrupts was expanded again. This time the goal was for
563 /proc/interrupts to display every IRQ vector in use by the system, not
564 just those considered 'most important'. The new vectors are:
566 THR -- interrupt raised when a machine check threshold counter
567 (typically counting ECC corrected errors of memory or cache) exceeds
568 a configurable threshold. Only available on some systems.
570 TRM -- a thermal event interrupt occurs when a temperature threshold
571 has been exceeded for the CPU. This interrupt may also be generated
572 when the temperature drops back to normal.
574 SPU -- a spurious interrupt is some interrupt that was raised then lowered
575 by some IO device before it could be fully processed by the APIC. Hence
576 the APIC sees the interrupt but does not know what device it came from.
577 For this case the APIC will generate the interrupt with a IRQ vector
578 of 0xff. This might also be generated by chipset bugs.
580 RES, CAL, TLB -- rescheduling, call and TLB flush interrupts are
581 sent from one CPU to another per the needs of the OS. Typically,
582 their statistics are used by kernel developers and interested users to
583 determine the occurrence of interrupts of the given type.
585 The above IRQ vectors are displayed only when relevant. For example,
586 the threshold vector does not exist on x86_64 platforms. Others are
587 suppressed when the system is a uniprocessor. As of this writing, only
588 i386 and x86_64 platforms support the new IRQ vector displays.
590 Of some interest is the introduction of the /proc/irq directory to 2.4.
591 It could be used to set IRQ to CPU affinity, this means that you can "hook" an
592 IRQ to only one CPU, or to exclude a CPU of handling IRQs. The contents of the
593 irq subdir is one subdir for each IRQ, and two files; default_smp_affinity and
598 0 10 12 14 16 18 2 4 6 8 prof_cpu_mask
599 1 11 13 15 17 19 3 5 7 9 default_smp_affinity
603 smp_affinity is a bitmask, in which you can specify which CPUs can handle the
604 IRQ, you can set it by doing:
606 > echo 1 > /proc/irq/10/smp_affinity
608 This means that only the first CPU will handle the IRQ, but you can also echo
609 5 which means that only the first and fourth CPU can handle the IRQ.
611 The contents of each smp_affinity file is the same by default:
613 > cat /proc/irq/0/smp_affinity
616 There is an alternate interface, smp_affinity_list which allows specifying
617 a cpu range instead of a bitmask:
619 > cat /proc/irq/0/smp_affinity_list
622 The default_smp_affinity mask applies to all non-active IRQs, which are the
623 IRQs which have not yet been allocated/activated, and hence which lack a
624 /proc/irq/[0-9]* directory.
626 The node file on an SMP system shows the node to which the device using the IRQ
627 reports itself as being attached. This hardware locality information does not
628 include information about any possible driver locality preference.
630 prof_cpu_mask specifies which CPUs are to be profiled by the system wide
631 profiler. Default value is ffffffff (all cpus if there are only 32 of them).
633 The way IRQs are routed is handled by the IO-APIC, and it's Round Robin
634 between all the CPUs which are allowed to handle it. As usual the kernel has
635 more info than you and does a better job than you, so the defaults are the
636 best choice for almost everyone. [Note this applies only to those IO-APIC's
637 that support "Round Robin" interrupt distribution.]
639 There are three more important subdirectories in /proc: net, scsi, and sys.
640 The general rule is that the contents, or even the existence of these
641 directories, depend on your kernel configuration. If SCSI is not enabled, the
642 directory scsi may not exist. The same is true with the net, which is there
643 only when networking support is present in the running kernel.
645 The slabinfo file gives information about memory usage at the slab level.
646 Linux uses slab pools for memory management above page level in version 2.2.
647 Commonly used objects have their own slab pool (such as network buffers,
648 directory cache, and so on).
650 ..............................................................................
652 > cat /proc/buddyinfo
654 Node 0, zone DMA 0 4 5 4 4 3 ...
655 Node 0, zone Normal 1 0 0 1 101 8 ...
656 Node 0, zone HighMem 2 0 0 1 1 0 ...
658 External fragmentation is a problem under some workloads, and buddyinfo is a
659 useful tool for helping diagnose these problems. Buddyinfo will give you a
660 clue as to how big an area you can safely allocate, or why a previous
663 Each column represents the number of pages of a certain order which are
664 available. In this case, there are 0 chunks of 2^0*PAGE_SIZE available in
665 ZONE_DMA, 4 chunks of 2^1*PAGE_SIZE in ZONE_DMA, 101 chunks of 2^4*PAGE_SIZE
666 available in ZONE_NORMAL, etc...
668 More information relevant to external fragmentation can be found in
671 > cat /proc/pagetypeinfo
675 Free pages count per migrate type at order 0 1 2 3 4 5 6 7 8 9 10
676 Node 0, zone DMA, type Unmovable 0 0 0 1 1 1 1 1 1 1 0
677 Node 0, zone DMA, type Reclaimable 0 0 0 0 0 0 0 0 0 0 0
678 Node 0, zone DMA, type Movable 1 1 2 1 2 1 1 0 1 0 2
679 Node 0, zone DMA, type Reserve 0 0 0 0 0 0 0 0 0 1 0
680 Node 0, zone DMA, type Isolate 0 0 0 0 0 0 0 0 0 0 0
681 Node 0, zone DMA32, type Unmovable 103 54 77 1 1 1 11 8 7 1 9
682 Node 0, zone DMA32, type Reclaimable 0 0 2 1 0 0 0 0 1 0 0
683 Node 0, zone DMA32, type Movable 169 152 113 91 77 54 39 13 6 1 452
684 Node 0, zone DMA32, type Reserve 1 2 2 2 2 0 1 1 1 1 0
685 Node 0, zone DMA32, type Isolate 0 0 0 0 0 0 0 0 0 0 0
687 Number of blocks type Unmovable Reclaimable Movable Reserve Isolate
688 Node 0, zone DMA 2 0 5 1 0
689 Node 0, zone DMA32 41 6 967 2 0
691 Fragmentation avoidance in the kernel works by grouping pages of different
692 migrate types into the same contiguous regions of memory called page blocks.
693 A page block is typically the size of the default hugepage size e.g. 2MB on
694 X86-64. By keeping pages grouped based on their ability to move, the kernel
695 can reclaim pages within a page block to satisfy a high-order allocation.
697 The pagetypinfo begins with information on the size of a page block. It
698 then gives the same type of information as buddyinfo except broken down
699 by migrate-type and finishes with details on how many page blocks of each
702 If min_free_kbytes has been tuned correctly (recommendations made by hugeadm
703 from libhugetlbfs http://sourceforge.net/projects/libhugetlbfs/), one can
704 make an estimate of the likely number of huge pages that can be allocated
705 at a given point in time. All the "Movable" blocks should be allocatable
706 unless memory has been mlock()'d. Some of the Reclaimable blocks should
707 also be allocatable although a lot of filesystem metadata may have to be
708 reclaimed to achieve this.
710 ..............................................................................
714 Provides information about distribution and utilization of memory. This
715 varies by architecture and compile options. The following is from a
716 16GB PIII, which has highmem enabled. You may not have all of these fields.
720 The "Locked" indicates whether the mapping is locked in memory or not.
723 MemTotal: 16344972 kB
730 HighTotal: 15597528 kB
731 HighFree: 13629632 kB
741 SReclaimable: 159856 kB
742 SUnreclaim: 124508 kB
747 CommitLimit: 7669796 kB
748 Committed_AS: 100056 kB
749 VmallocTotal: 112216 kB
751 VmallocChunk: 111088 kB
752 AnonHugePages: 49152 kB
754 MemTotal: Total usable ram (i.e. physical ram minus a few reserved
755 bits and the kernel binary code)
756 MemFree: The sum of LowFree+HighFree
757 Buffers: Relatively temporary storage for raw disk blocks
758 shouldn't get tremendously large (20MB or so)
759 Cached: in-memory cache for files read from the disk (the
760 pagecache). Doesn't include SwapCached
761 SwapCached: Memory that once was swapped out, is swapped back in but
762 still also is in the swapfile (if memory is needed it
763 doesn't need to be swapped out AGAIN because it is already
764 in the swapfile. This saves I/O)
765 Active: Memory that has been used more recently and usually not
766 reclaimed unless absolutely necessary.
767 Inactive: Memory which has been less recently used. It is more
768 eligible to be reclaimed for other purposes
770 HighFree: Highmem is all memory above ~860MB of physical memory
771 Highmem areas are for use by userspace programs, or
772 for the pagecache. The kernel must use tricks to access
773 this memory, making it slower to access than lowmem.
775 LowFree: Lowmem is memory which can be used for everything that
776 highmem can be used for, but it is also available for the
777 kernel's use for its own data structures. Among many
778 other things, it is where everything from the Slab is
779 allocated. Bad things happen when you're out of lowmem.
780 SwapTotal: total amount of swap space available
781 SwapFree: Memory which has been evicted from RAM, and is temporarily
783 Dirty: Memory which is waiting to get written back to the disk
784 Writeback: Memory which is actively being written back to the disk
785 AnonPages: Non-file backed pages mapped into userspace page tables
786 AnonHugePages: Non-file backed huge pages mapped into userspace page tables
787 Mapped: files which have been mmaped, such as libraries
788 Slab: in-kernel data structures cache
789 SReclaimable: Part of Slab, that might be reclaimed, such as caches
790 SUnreclaim: Part of Slab, that cannot be reclaimed on memory pressure
791 PageTables: amount of memory dedicated to the lowest level of page
793 NFS_Unstable: NFS pages sent to the server, but not yet committed to stable
795 Bounce: Memory used for block device "bounce buffers"
796 WritebackTmp: Memory used by FUSE for temporary writeback buffers
797 CommitLimit: Based on the overcommit ratio ('vm.overcommit_ratio'),
798 this is the total amount of memory currently available to
799 be allocated on the system. This limit is only adhered to
800 if strict overcommit accounting is enabled (mode 2 in
801 'vm.overcommit_memory').
802 The CommitLimit is calculated with the following formula:
803 CommitLimit = ('vm.overcommit_ratio' * Physical RAM) + Swap
804 For example, on a system with 1G of physical RAM and 7G
805 of swap with a `vm.overcommit_ratio` of 30 it would
806 yield a CommitLimit of 7.3G.
807 For more details, see the memory overcommit documentation
808 in vm/overcommit-accounting.
809 Committed_AS: The amount of memory presently allocated on the system.
810 The committed memory is a sum of all of the memory which
811 has been allocated by processes, even if it has not been
812 "used" by them as of yet. A process which malloc()'s 1G
813 of memory, but only touches 300M of it will only show up
814 as using 300M of memory even if it has the address space
815 allocated for the entire 1G. This 1G is memory which has
816 been "committed" to by the VM and can be used at any time
817 by the allocating application. With strict overcommit
818 enabled on the system (mode 2 in 'vm.overcommit_memory'),
819 allocations which would exceed the CommitLimit (detailed
820 above) will not be permitted. This is useful if one needs
821 to guarantee that processes will not fail due to lack of
822 memory once that memory has been successfully allocated.
823 VmallocTotal: total size of vmalloc memory area
824 VmallocUsed: amount of vmalloc area which is used
825 VmallocChunk: largest contiguous block of vmalloc area which is free
827 ..............................................................................
831 Provides information about vmalloced/vmaped areas. One line per area,
832 containing the virtual address range of the area, size in bytes,
833 caller information of the creator, and optional information depending
834 on the kind of area :
836 pages=nr number of pages
837 phys=addr if a physical address was specified
838 ioremap I/O mapping (ioremap() and friends)
839 vmalloc vmalloc() area
842 vpages buffer for pages pointers was vmalloced (huge area)
843 N<node>=nr (Only on NUMA kernels)
844 Number of pages allocated on memory node <node>
846 > cat /proc/vmallocinfo
847 0xffffc20000000000-0xffffc20000201000 2101248 alloc_large_system_hash+0x204 ...
848 /0x2c0 pages=512 vmalloc N0=128 N1=128 N2=128 N3=128
849 0xffffc20000201000-0xffffc20000302000 1052672 alloc_large_system_hash+0x204 ...
850 /0x2c0 pages=256 vmalloc N0=64 N1=64 N2=64 N3=64
851 0xffffc20000302000-0xffffc20000304000 8192 acpi_tb_verify_table+0x21/0x4f...
852 phys=7fee8000 ioremap
853 0xffffc20000304000-0xffffc20000307000 12288 acpi_tb_verify_table+0x21/0x4f...
854 phys=7fee7000 ioremap
855 0xffffc2000031d000-0xffffc2000031f000 8192 init_vdso_vars+0x112/0x210
856 0xffffc2000031f000-0xffffc2000032b000 49152 cramfs_uncompress_init+0x2e ...
857 /0x80 pages=11 vmalloc N0=3 N1=3 N2=2 N3=3
858 0xffffc2000033a000-0xffffc2000033d000 12288 sys_swapon+0x640/0xac0 ...
860 0xffffc20000347000-0xffffc2000034c000 20480 xt_alloc_table_info+0xfe ...
861 /0x130 [x_tables] pages=4 vmalloc N0=4
862 0xffffffffa0000000-0xffffffffa000f000 61440 sys_init_module+0xc27/0x1d00 ...
863 pages=14 vmalloc N2=14
864 0xffffffffa000f000-0xffffffffa0014000 20480 sys_init_module+0xc27/0x1d00 ...
866 0xffffffffa0014000-0xffffffffa0017000 12288 sys_init_module+0xc27/0x1d00 ...
868 0xffffffffa0017000-0xffffffffa0022000 45056 sys_init_module+0xc27/0x1d00 ...
869 pages=10 vmalloc N0=10
871 ..............................................................................
875 Provides counts of softirq handlers serviced since boot time, for each cpu.
880 TIMER: 27166 27120 27097 27034
885 SCHED: 27035 26983 26971 26746
887 RCU: 1678 1769 2178 2250
890 1.3 IDE devices in /proc/ide
891 ----------------------------
893 The subdirectory /proc/ide contains information about all IDE devices of which
894 the kernel is aware. There is one subdirectory for each IDE controller, the
895 file drivers and a link for each IDE device, pointing to the device directory
896 in the controller specific subtree.
898 The file drivers contains general information about the drivers used for the
901 > cat /proc/ide/drivers
902 ide-cdrom version 4.53
903 ide-disk version 1.08
905 More detailed information can be found in the controller specific
906 subdirectories. These are named ide0, ide1 and so on. Each of these
907 directories contains the files shown in table 1-6.
910 Table 1-6: IDE controller info in /proc/ide/ide?
911 ..............................................................................
913 channel IDE channel (0 or 1)
914 config Configuration (only for PCI/IDE bridge)
916 model Type/Chipset of IDE controller
917 ..............................................................................
919 Each device connected to a controller has a separate subdirectory in the
920 controllers directory. The files listed in table 1-7 are contained in these
924 Table 1-7: IDE device information
925 ..............................................................................
928 capacity Capacity of the medium (in 512Byte blocks)
929 driver driver and version
930 geometry physical and logical geometry
931 identify device identify block
933 model device identifier
934 settings device setup
935 smart_thresholds IDE disk management thresholds
936 smart_values IDE disk management values
937 ..............................................................................
939 The most interesting file is settings. This file contains a nice overview of
940 the drive parameters:
942 # cat /proc/ide/ide0/hda/settings
943 name value min max mode
944 ---- ----- --- --- ----
945 bios_cyl 526 0 65535 rw
946 bios_head 255 0 255 rw
948 breada_readahead 4 0 127 rw
950 file_readahead 72 0 2097151 rw
952 keepsettings 0 0 1 rw
953 max_kb_per_request 122 1 127 rw
957 pio_mode write-only 0 255 w
963 1.4 Networking info in /proc/net
964 --------------------------------
966 The subdirectory /proc/net follows the usual pattern. Table 1-8 shows the
967 additional values you get for IP version 6 if you configure the kernel to
968 support this. Table 1-9 lists the files and their meaning.
971 Table 1-8: IPv6 info in /proc/net
972 ..............................................................................
974 udp6 UDP sockets (IPv6)
975 tcp6 TCP sockets (IPv6)
976 raw6 Raw device statistics (IPv6)
977 igmp6 IP multicast addresses, which this host joined (IPv6)
978 if_inet6 List of IPv6 interface addresses
979 ipv6_route Kernel routing table for IPv6
980 rt6_stats Global IPv6 routing tables statistics
981 sockstat6 Socket statistics (IPv6)
982 snmp6 Snmp data (IPv6)
983 ..............................................................................
986 Table 1-9: Network info in /proc/net
987 ..............................................................................
990 dev network devices with statistics
991 dev_mcast the Layer2 multicast groups a device is listening too
992 (interface index, label, number of references, number of bound
994 dev_stat network device status
995 ip_fwchains Firewall chain linkage
996 ip_fwnames Firewall chain names
997 ip_masq Directory containing the masquerading tables
998 ip_masquerade Major masquerading table
999 netstat Network statistics
1000 raw raw device statistics
1001 route Kernel routing table
1002 rpc Directory containing rpc info
1003 rt_cache Routing cache
1005 sockstat Socket statistics
1008 unix UNIX domain sockets
1009 wireless Wireless interface data (Wavelan etc)
1010 igmp IP multicast addresses, which this host joined
1011 psched Global packet scheduler parameters.
1012 netlink List of PF_NETLINK sockets
1013 ip_mr_vifs List of multicast virtual interfaces
1014 ip_mr_cache List of multicast routing cache
1015 ..............................................................................
1017 You can use this information to see which network devices are available in
1018 your system and how much traffic was routed over those devices:
1021 Inter-|Receive |[...
1022 face |bytes packets errs drop fifo frame compressed multicast|[...
1023 lo: 908188 5596 0 0 0 0 0 0 [...
1024 ppp0:15475140 20721 410 0 0 410 0 0 [...
1025 eth0: 614530 7085 0 0 0 0 0 1 [...
1028 ...] bytes packets errs drop fifo colls carrier compressed
1029 ...] 908188 5596 0 0 0 0 0 0
1030 ...] 1375103 17405 0 0 0 0 0 0
1031 ...] 1703981 5535 0 0 0 3 0 0
1033 In addition, each Channel Bond interface has its own directory. For
1034 example, the bond0 device will have a directory called /proc/net/bond0/.
1035 It will contain information that is specific to that bond, such as the
1036 current slaves of the bond, the link status of the slaves, and how
1037 many times the slaves link has failed.
1042 If you have a SCSI host adapter in your system, you'll find a subdirectory
1043 named after the driver for this adapter in /proc/scsi. You'll also see a list
1044 of all recognized SCSI devices in /proc/scsi:
1046 >cat /proc/scsi/scsi
1048 Host: scsi0 Channel: 00 Id: 00 Lun: 00
1049 Vendor: IBM Model: DGHS09U Rev: 03E0
1050 Type: Direct-Access ANSI SCSI revision: 03
1051 Host: scsi0 Channel: 00 Id: 06 Lun: 00
1052 Vendor: PIONEER Model: CD-ROM DR-U06S Rev: 1.04
1053 Type: CD-ROM ANSI SCSI revision: 02
1056 The directory named after the driver has one file for each adapter found in
1057 the system. These files contain information about the controller, including
1058 the used IRQ and the IO address range. The amount of information shown is
1059 dependent on the adapter you use. The example shows the output for an Adaptec
1060 AHA-2940 SCSI adapter:
1062 > cat /proc/scsi/aic7xxx/0
1064 Adaptec AIC7xxx driver version: 5.1.19/3.2.4
1066 TCQ Enabled By Default : Disabled
1067 AIC7XXX_PROC_STATS : Disabled
1068 AIC7XXX_RESET_DELAY : 5
1069 Adapter Configuration:
1070 SCSI Adapter: Adaptec AHA-294X Ultra SCSI host adapter
1071 Ultra Wide Controller
1072 PCI MMAPed I/O Base: 0xeb001000
1073 Adapter SEEPROM Config: SEEPROM found and used.
1074 Adaptec SCSI BIOS: Enabled
1076 SCBs: Active 0, Max Active 2,
1077 Allocated 15, HW 16, Page 255
1079 BIOS Control Word: 0x18b6
1080 Adapter Control Word: 0x005b
1081 Extended Translation: Enabled
1082 Disconnect Enable Flags: 0xffff
1083 Ultra Enable Flags: 0x0001
1084 Tag Queue Enable Flags: 0x0000
1085 Ordered Queue Tag Flags: 0x0000
1086 Default Tag Queue Depth: 8
1087 Tagged Queue By Device array for aic7xxx host instance 0:
1088 {255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255}
1089 Actual queue depth per device for aic7xxx host instance 0:
1090 {1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}
1093 Device using Wide/Sync transfers at 40.0 MByte/sec, offset 8
1094 Transinfo settings: current(12/8/1/0), goal(12/8/1/0), user(12/15/1/0)
1095 Total transfers 160151 (74577 reads and 85574 writes)
1097 Device using Narrow/Sync transfers at 5.0 MByte/sec, offset 15
1098 Transinfo settings: current(50/15/0/0), goal(50/15/0/0), user(50/15/0/0)
1099 Total transfers 0 (0 reads and 0 writes)
1102 1.6 Parallel port info in /proc/parport
1103 ---------------------------------------
1105 The directory /proc/parport contains information about the parallel ports of
1106 your system. It has one subdirectory for each port, named after the port
1109 These directories contain the four files shown in Table 1-10.
1112 Table 1-10: Files in /proc/parport
1113 ..............................................................................
1115 autoprobe Any IEEE-1284 device ID information that has been acquired.
1116 devices list of the device drivers using that port. A + will appear by the
1117 name of the device currently using the port (it might not appear
1119 hardware Parallel port's base address, IRQ line and DMA channel.
1120 irq IRQ that parport is using for that port. This is in a separate
1121 file to allow you to alter it by writing a new value in (IRQ
1123 ..............................................................................
1125 1.7 TTY info in /proc/tty
1126 -------------------------
1128 Information about the available and actually used tty's can be found in the
1129 directory /proc/tty.You'll find entries for drivers and line disciplines in
1130 this directory, as shown in Table 1-11.
1133 Table 1-11: Files in /proc/tty
1134 ..............................................................................
1136 drivers list of drivers and their usage
1137 ldiscs registered line disciplines
1138 driver/serial usage statistic and status of single tty lines
1139 ..............................................................................
1141 To see which tty's are currently in use, you can simply look into the file
1144 > cat /proc/tty/drivers
1145 pty_slave /dev/pts 136 0-255 pty:slave
1146 pty_master /dev/ptm 128 0-255 pty:master
1147 pty_slave /dev/ttyp 3 0-255 pty:slave
1148 pty_master /dev/pty 2 0-255 pty:master
1149 serial /dev/cua 5 64-67 serial:callout
1150 serial /dev/ttyS 4 64-67 serial
1151 /dev/tty0 /dev/tty0 4 0 system:vtmaster
1152 /dev/ptmx /dev/ptmx 5 2 system
1153 /dev/console /dev/console 5 1 system:console
1154 /dev/tty /dev/tty 5 0 system:/dev/tty
1155 unknown /dev/tty 4 1-63 console
1158 1.8 Miscellaneous kernel statistics in /proc/stat
1159 -------------------------------------------------
1161 Various pieces of information about kernel activity are available in the
1162 /proc/stat file. All of the numbers reported in this file are aggregates
1163 since the system first booted. For a quick look, simply cat the file:
1166 cpu 2255 34 2290 22625563 6290 127 456 0 0
1167 cpu0 1132 34 1441 11311718 3675 127 438 0 0
1168 cpu1 1123 0 849 11313845 2614 0 18 0 0
1169 intr 114930548 113199788 3 0 5 263 0 4 [... lots more numbers ...]
1175 softirq 183433 0 21755 12 39 1137 231 21459 2263
1177 The very first "cpu" line aggregates the numbers in all of the other "cpuN"
1178 lines. These numbers identify the amount of time the CPU has spent performing
1179 different kinds of work. Time units are in USER_HZ (typically hundredths of a
1180 second). The meanings of the columns are as follows, from left to right:
1182 - user: normal processes executing in user mode
1183 - nice: niced processes executing in user mode
1184 - system: processes executing in kernel mode
1185 - idle: twiddling thumbs
1186 - iowait: waiting for I/O to complete
1187 - irq: servicing interrupts
1188 - softirq: servicing softirqs
1189 - steal: involuntary wait
1190 - guest: running a normal guest
1191 - guest_nice: running a niced guest
1193 The "intr" line gives counts of interrupts serviced since boot time, for each
1194 of the possible system interrupts. The first column is the total of all
1195 interrupts serviced; each subsequent column is the total for that particular
1198 The "ctxt" line gives the total number of context switches across all CPUs.
1200 The "btime" line gives the time at which the system booted, in seconds since
1203 The "processes" line gives the number of processes and threads created, which
1204 includes (but is not limited to) those created by calls to the fork() and
1205 clone() system calls.
1207 The "procs_running" line gives the total number of threads that are
1208 running or ready to run (i.e., the total number of runnable threads).
1210 The "procs_blocked" line gives the number of processes currently blocked,
1211 waiting for I/O to complete.
1213 The "softirq" line gives counts of softirqs serviced since boot time, for each
1214 of the possible system softirqs. The first column is the total of all
1215 softirqs serviced; each subsequent column is the total for that particular
1219 1.9 Ext4 file system parameters
1220 ------------------------------
1222 Information about mounted ext4 file systems can be found in
1223 /proc/fs/ext4. Each mounted filesystem will have a directory in
1224 /proc/fs/ext4 based on its device name (i.e., /proc/fs/ext4/hdc or
1225 /proc/fs/ext4/dm-0). The files in each per-device directory are shown
1226 in Table 1-12, below.
1228 Table 1-12: Files in /proc/fs/ext4/<devname>
1229 ..............................................................................
1231 mb_groups details of multiblock allocator buddy cache of free blocks
1232 ..............................................................................
1236 Shows registered system console lines.
1238 To see which character device lines are currently used for the system console
1239 /dev/console, you may simply look into the file /proc/consoles:
1241 > cat /proc/consoles
1247 device name of the device
1248 operations R = can do read operations
1249 W = can do write operations
1251 flags E = it is enabled
1252 C = it is preferred console
1253 B = it is primary boot console
1254 p = it is used for printk buffer
1255 b = it is not a TTY but a Braille device
1256 a = it is safe to use when cpu is offline
1257 major:minor major and minor number of the device separated by a colon
1259 ------------------------------------------------------------------------------
1261 ------------------------------------------------------------------------------
1262 The /proc file system serves information about the running system. It not only
1263 allows access to process data but also allows you to request the kernel status
1264 by reading files in the hierarchy.
1266 The directory structure of /proc reflects the types of information and makes
1267 it easy, if not obvious, where to look for specific data.
1268 ------------------------------------------------------------------------------
1270 ------------------------------------------------------------------------------
1271 CHAPTER 2: MODIFYING SYSTEM PARAMETERS
1272 ------------------------------------------------------------------------------
1274 ------------------------------------------------------------------------------
1276 ------------------------------------------------------------------------------
1277 * Modifying kernel parameters by writing into files found in /proc/sys
1278 * Exploring the files which modify certain parameters
1279 * Review of the /proc/sys file tree
1280 ------------------------------------------------------------------------------
1283 A very interesting part of /proc is the directory /proc/sys. This is not only
1284 a source of information, it also allows you to change parameters within the
1285 kernel. Be very careful when attempting this. You can optimize your system,
1286 but you can also cause it to crash. Never alter kernel parameters on a
1287 production system. Set up a development machine and test to make sure that
1288 everything works the way you want it to. You may have no alternative but to
1289 reboot the machine once an error has been made.
1291 To change a value, simply echo the new value into the file. An example is
1292 given below in the section on the file system data. You need to be root to do
1293 this. You can create your own boot script to perform this every time your
1296 The files in /proc/sys can be used to fine tune and monitor miscellaneous and
1297 general things in the operation of the Linux kernel. Since some of the files
1298 can inadvertently disrupt your system, it is advisable to read both
1299 documentation and source before actually making adjustments. In any case, be
1300 very careful when writing to any of these files. The entries in /proc may
1301 change slightly between the 2.1.* and the 2.2 kernel, so if there is any doubt
1302 review the kernel documentation in the directory /usr/src/linux/Documentation.
1303 This chapter is heavily based on the documentation included in the pre 2.2
1304 kernels, and became part of it in version 2.2.1 of the Linux kernel.
1306 Please see: Documentation/sysctl/ directory for descriptions of these
1309 ------------------------------------------------------------------------------
1311 ------------------------------------------------------------------------------
1312 Certain aspects of kernel behavior can be modified at runtime, without the
1313 need to recompile the kernel, or even to reboot the system. The files in the
1314 /proc/sys tree can not only be read, but also modified. You can use the echo
1315 command to write value into these files, thereby changing the default settings
1317 ------------------------------------------------------------------------------
1319 ------------------------------------------------------------------------------
1320 CHAPTER 3: PER-PROCESS PARAMETERS
1321 ------------------------------------------------------------------------------
1323 3.1 /proc/<pid>/oom_adj & /proc/<pid>/oom_score_adj- Adjust the oom-killer score
1324 --------------------------------------------------------------------------------
1326 These file can be used to adjust the badness heuristic used to select which
1327 process gets killed in out of memory conditions.
1329 The badness heuristic assigns a value to each candidate task ranging from 0
1330 (never kill) to 1000 (always kill) to determine which process is targeted. The
1331 units are roughly a proportion along that range of allowed memory the process
1332 may allocate from based on an estimation of its current memory and swap use.
1333 For example, if a task is using all allowed memory, its badness score will be
1334 1000. If it is using half of its allowed memory, its score will be 500.
1336 There is an additional factor included in the badness score: root
1337 processes are given 3% extra memory over other tasks.
1339 The amount of "allowed" memory depends on the context in which the oom killer
1340 was called. If it is due to the memory assigned to the allocating task's cpuset
1341 being exhausted, the allowed memory represents the set of mems assigned to that
1342 cpuset. If it is due to a mempolicy's node(s) being exhausted, the allowed
1343 memory represents the set of mempolicy nodes. If it is due to a memory
1344 limit (or swap limit) being reached, the allowed memory is that configured
1345 limit. Finally, if it is due to the entire system being out of memory, the
1346 allowed memory represents all allocatable resources.
1348 The value of /proc/<pid>/oom_score_adj is added to the badness score before it
1349 is used to determine which task to kill. Acceptable values range from -1000
1350 (OOM_SCORE_ADJ_MIN) to +1000 (OOM_SCORE_ADJ_MAX). This allows userspace to
1351 polarize the preference for oom killing either by always preferring a certain
1352 task or completely disabling it. The lowest possible value, -1000, is
1353 equivalent to disabling oom killing entirely for that task since it will always
1354 report a badness score of 0.
1356 Consequently, it is very simple for userspace to define the amount of memory to
1357 consider for each task. Setting a /proc/<pid>/oom_score_adj value of +500, for
1358 example, is roughly equivalent to allowing the remainder of tasks sharing the
1359 same system, cpuset, mempolicy, or memory controller resources to use at least
1360 50% more memory. A value of -500, on the other hand, would be roughly
1361 equivalent to discounting 50% of the task's allowed memory from being considered
1362 as scoring against the task.
1364 For backwards compatibility with previous kernels, /proc/<pid>/oom_adj may also
1365 be used to tune the badness score. Its acceptable values range from -16
1366 (OOM_ADJUST_MIN) to +15 (OOM_ADJUST_MAX) and a special value of -17
1367 (OOM_DISABLE) to disable oom killing entirely for that task. Its value is
1368 scaled linearly with /proc/<pid>/oom_score_adj.
1370 Writing to /proc/<pid>/oom_score_adj or /proc/<pid>/oom_adj will change the
1371 other with its scaled value.
1373 The value of /proc/<pid>/oom_score_adj may be reduced no lower than the last
1374 value set by a CAP_SYS_RESOURCE process. To reduce the value any lower
1375 requires CAP_SYS_RESOURCE.
1377 NOTICE: /proc/<pid>/oom_adj is deprecated and will be removed, please see
1378 Documentation/feature-removal-schedule.txt.
1380 Caveat: when a parent task is selected, the oom killer will sacrifice any first
1381 generation children with separate address spaces instead, if possible. This
1382 avoids servers and important system daemons from being killed and loses the
1383 minimal amount of work.
1386 3.2 /proc/<pid>/oom_score - Display current oom-killer score
1387 -------------------------------------------------------------
1389 This file can be used to check the current score used by the oom-killer is for
1390 any given <pid>. Use it together with /proc/<pid>/oom_adj to tune which
1391 process should be killed in an out-of-memory situation.
1394 3.3 /proc/<pid>/io - Display the IO accounting fields
1395 -------------------------------------------------------
1397 This file contains IO statistics for each running process
1402 test:/tmp # dd if=/dev/zero of=/tmp/test.dat &
1405 test:/tmp # cat /proc/3828/io
1411 write_bytes: 323932160
1412 cancelled_write_bytes: 0
1421 I/O counter: chars read
1422 The number of bytes which this task has caused to be read from storage. This
1423 is simply the sum of bytes which this process passed to read() and pread().
1424 It includes things like tty IO and it is unaffected by whether or not actual
1425 physical disk IO was required (the read might have been satisfied from
1432 I/O counter: chars written
1433 The number of bytes which this task has caused, or shall cause to be written
1434 to disk. Similar caveats apply here as with rchar.
1440 I/O counter: read syscalls
1441 Attempt to count the number of read I/O operations, i.e. syscalls like read()
1448 I/O counter: write syscalls
1449 Attempt to count the number of write I/O operations, i.e. syscalls like
1450 write() and pwrite().
1456 I/O counter: bytes read
1457 Attempt to count the number of bytes which this process really did cause to
1458 be fetched from the storage layer. Done at the submit_bio() level, so it is
1459 accurate for block-backed filesystems. <please add status regarding NFS and
1460 CIFS at a later time>
1466 I/O counter: bytes written
1467 Attempt to count the number of bytes which this process caused to be sent to
1468 the storage layer. This is done at page-dirtying time.
1471 cancelled_write_bytes
1472 ---------------------
1474 The big inaccuracy here is truncate. If a process writes 1MB to a file and
1475 then deletes the file, it will in fact perform no writeout. But it will have
1476 been accounted as having caused 1MB of write.
1477 In other words: The number of bytes which this process caused to not happen,
1478 by truncating pagecache. A task can cause "negative" IO too. If this task
1479 truncates some dirty pagecache, some IO which another task has been accounted
1480 for (in its write_bytes) will not be happening. We _could_ just subtract that
1481 from the truncating task's write_bytes, but there is information loss in doing
1488 At its current implementation state, this is a bit racy on 32-bit machines: if
1489 process A reads process B's /proc/pid/io while process B is updating one of
1490 those 64-bit counters, process A could see an intermediate result.
1493 More information about this can be found within the taskstats documentation in
1494 Documentation/accounting.
1496 3.4 /proc/<pid>/coredump_filter - Core dump filtering settings
1497 ---------------------------------------------------------------
1498 When a process is dumped, all anonymous memory is written to a core file as
1499 long as the size of the core file isn't limited. But sometimes we don't want
1500 to dump some memory segments, for example, huge shared memory. Conversely,
1501 sometimes we want to save file-backed memory segments into a core file, not
1502 only the individual files.
1504 /proc/<pid>/coredump_filter allows you to customize which memory segments
1505 will be dumped when the <pid> process is dumped. coredump_filter is a bitmask
1506 of memory types. If a bit of the bitmask is set, memory segments of the
1507 corresponding memory type are dumped, otherwise they are not dumped.
1509 The following 7 memory types are supported:
1510 - (bit 0) anonymous private memory
1511 - (bit 1) anonymous shared memory
1512 - (bit 2) file-backed private memory
1513 - (bit 3) file-backed shared memory
1514 - (bit 4) ELF header pages in file-backed private memory areas (it is
1515 effective only if the bit 2 is cleared)
1516 - (bit 5) hugetlb private memory
1517 - (bit 6) hugetlb shared memory
1519 Note that MMIO pages such as frame buffer are never dumped and vDSO pages
1520 are always dumped regardless of the bitmask status.
1522 Note bit 0-4 doesn't effect any hugetlb memory. hugetlb memory are only
1523 effected by bit 5-6.
1525 Default value of coredump_filter is 0x23; this means all anonymous memory
1526 segments and hugetlb private memory are dumped.
1528 If you don't want to dump all shared memory segments attached to pid 1234,
1529 write 0x21 to the process's proc file.
1531 $ echo 0x21 > /proc/1234/coredump_filter
1533 When a new process is created, the process inherits the bitmask status from its
1534 parent. It is useful to set up coredump_filter before the program runs.
1537 $ echo 0x7 > /proc/self/coredump_filter
1540 3.5 /proc/<pid>/mountinfo - Information about mounts
1541 --------------------------------------------------------
1543 This file contains lines of the form:
1545 36 35 98:0 /mnt1 /mnt2 rw,noatime master:1 - ext3 /dev/root rw,errors=continue
1546 (1)(2)(3) (4) (5) (6) (7) (8) (9) (10) (11)
1548 (1) mount ID: unique identifier of the mount (may be reused after umount)
1549 (2) parent ID: ID of parent (or of self for the top of the mount tree)
1550 (3) major:minor: value of st_dev for files on filesystem
1551 (4) root: root of the mount within the filesystem
1552 (5) mount point: mount point relative to the process's root
1553 (6) mount options: per mount options
1554 (7) optional fields: zero or more fields of the form "tag[:value]"
1555 (8) separator: marks the end of the optional fields
1556 (9) filesystem type: name of filesystem of the form "type[.subtype]"
1557 (10) mount source: filesystem specific information or "none"
1558 (11) super options: per super block options
1560 Parsers should ignore all unrecognised optional fields. Currently the
1561 possible optional fields are:
1563 shared:X mount is shared in peer group X
1564 master:X mount is slave to peer group X
1565 propagate_from:X mount is slave and receives propagation from peer group X (*)
1566 unbindable mount is unbindable
1568 (*) X is the closest dominant peer group under the process's root. If
1569 X is the immediate master of the mount, or if there's no dominant peer
1570 group under the same root, then only the "master:X" field is present
1571 and not the "propagate_from:X" field.
1573 For more information on mount propagation see:
1575 Documentation/filesystems/sharedsubtree.txt
1578 3.6 /proc/<pid>/comm & /proc/<pid>/task/<tid>/comm
1579 --------------------------------------------------------
1580 These files provide a method to access a tasks comm value. It also allows for
1581 a task to set its own or one of its thread siblings comm value. The comm value
1582 is limited in size compared to the cmdline value, so writing anything longer
1583 then the kernel's TASK_COMM_LEN (currently 16 chars) will result in a truncated
1587 3.7 /proc/<pid>/task/<tid>/children - Information about task children
1588 -------------------------------------------------------------------------
1589 This file provides a fast way to retrieve first level children pids
1590 of a task pointed by <pid>/<tid> pair. The format is a space separated
1593 Note the "first level" here -- if a child has own children they will
1594 not be listed here, one needs to read /proc/<children-pid>/task/<tid>/children
1595 to obtain the descendants.
1597 Since this interface is intended to be fast and cheap it doesn't
1598 guarantee to provide precise results and some children might be
1599 skipped, especially if they've exited right after we printed their
1600 pids, so one need to either stop or freeze processes being inspected
1601 if precise results are needed.
1604 ------------------------------------------------------------------------------
1606 ------------------------------------------------------------------------------
1609 ---------------------
1611 The following mount options are supported:
1613 hidepid= Set /proc/<pid>/ access mode.
1614 gid= Set the group authorized to learn processes information.
1616 hidepid=0 means classic mode - everybody may access all /proc/<pid>/ directories
1619 hidepid=1 means users may not access any /proc/<pid>/ directories but their
1620 own. Sensitive files like cmdline, sched*, status are now protected against
1621 other users. This makes it impossible to learn whether any user runs
1622 specific program (given the program doesn't reveal itself by its behaviour).
1623 As an additional bonus, as /proc/<pid>/cmdline is unaccessible for other users,
1624 poorly written programs passing sensitive information via program arguments are
1625 now protected against local eavesdroppers.
1627 hidepid=2 means hidepid=1 plus all /proc/<pid>/ will be fully invisible to other
1628 users. It doesn't mean that it hides a fact whether a process with a specific
1629 pid value exists (it can be learned by other means, e.g. by "kill -0 $PID"),
1630 but it hides process' uid and gid, which may be learned by stat()'ing
1631 /proc/<pid>/ otherwise. It greatly complicates an intruder's task of gathering
1632 information about running processes, whether some daemon runs with elevated
1633 privileges, whether other user runs some sensitive program, whether other users
1634 run any program at all, etc.
1636 gid= defines a group authorized to learn processes information otherwise
1637 prohibited by hidepid=. If you use some daemon like identd which needs to learn
1638 information about processes information, just add identd to this group.
projects / linux.git / blob