1.. SPDX-License-Identifier: GPL-2.0 2 3==================== 4The /proc Filesystem 5==================== 6 7===================== ======================================= ================ 8/proc/sys Terrehon Bowden <terrehon@pacbell.net>, October 7 1999 9 Bodo Bauer <bb@ricochet.net> 102.4.x update Jorge Nerin <comandante@zaralinux.com> November 14 2000 11move /proc/sys Shen Feng <shen@cn.fujitsu.com> April 1 2009 12fixes/update part 1.1 Stefani Seibold <stefani@seibold.net> June 9 2009 13===================== ======================================= ================ 14 15 16 17.. Table of Contents 18 19 0 Preface 20 0.1 Introduction/Credits 21 0.2 Legal Stuff 22 23 1 Collecting System Information 24 1.1 Process-Specific Subdirectories 25 1.2 Kernel data 26 1.3 IDE devices in /proc/ide 27 1.4 Networking info in /proc/net 28 1.5 SCSI info 29 1.6 Parallel port info in /proc/parport 30 1.7 TTY info in /proc/tty 31 1.8 Miscellaneous kernel statistics in /proc/stat 32 1.9 Ext4 file system parameters 33 34 2 Modifying System Parameters 35 36 3 Per-Process Parameters 37 3.1 /proc/<pid>/oom_adj & /proc/<pid>/oom_score_adj - Adjust the oom-killer 38 score 39 3.2 /proc/<pid>/oom_score - Display current oom-killer score 40 3.3 /proc/<pid>/io - Display the IO accounting fields 41 3.4 /proc/<pid>/coredump_filter - Core dump filtering settings 42 3.5 /proc/<pid>/mountinfo - Information about mounts 43 3.6 /proc/<pid>/comm & /proc/<pid>/task/<tid>/comm 44 3.7 /proc/<pid>/task/<tid>/children - Information about task children 45 3.8 /proc/<pid>/fdinfo/<fd> - Information about opened file 46 3.9 /proc/<pid>/map_files - Information about memory mapped files 47 3.10 /proc/<pid>/timerslack_ns - Task timerslack value 48 3.11 /proc/<pid>/patch_state - Livepatch patch operation state 49 3.12 /proc/<pid>/arch_status - Task architecture specific information 50 3.13 /proc/<pid>/fd - List of symlinks to open files 51 52 4 Configuring procfs 53 4.1 Mount options 54 55 5 Filesystem behavior 56 57Preface 58======= 59 600.1 Introduction/Credits 61------------------------ 62 63This documentation is part of a soon (or so we hope) to be released book on 64the SuSE Linux distribution. As there is no complete documentation for the 65/proc file system and we've used many freely available sources to write these 66chapters, it seems only fair to give the work back to the Linux community. 67This work is based on the 2.2.* kernel version and the upcoming 2.4.*. I'm 68afraid it's still far from complete, but we hope it will be useful. As far as 69we know, it is the first 'all-in-one' document about the /proc file system. It 70is focused on the Intel x86 hardware, so if you are looking for PPC, ARM, 71SPARC, AXP, etc., features, you probably won't find what you are looking for. 72It also only covers IPv4 networking, not IPv6 nor other protocols - sorry. But 73additions and patches are welcome and will be added to this document if you 74mail them to Bodo. 75 76We'd like to thank Alan Cox, Rik van Riel, and Alexey Kuznetsov and a lot of 77other people for help compiling this documentation. We'd also like to extend a 78special thank you to Andi Kleen for documentation, which we relied on heavily 79to create this document, as well as the additional information he provided. 80Thanks to everybody else who contributed source or docs to the Linux kernel 81and helped create a great piece of software... :) 82 83If you have any comments, corrections or additions, please don't hesitate to 84contact Bodo Bauer at bb@ricochet.net. We'll be happy to add them to this 85document. 86 87The latest version of this document is available online at 88http://tldp.org/LDP/Linux-Filesystem-Hierarchy/html/proc.html 89 90If the above direction does not works for you, you could try the kernel 91mailing list at linux-kernel@vger.kernel.org and/or try to reach me at 92comandante@zaralinux.com. 93 940.2 Legal Stuff 95--------------- 96 97We don't guarantee the correctness of this document, and if you come to us 98complaining about how you screwed up your system because of incorrect 99documentation, we won't feel responsible... 100 101Chapter 1: Collecting System Information 102======================================== 103 104In This Chapter 105--------------- 106* Investigating the properties of the pseudo file system /proc and its 107 ability to provide information on the running Linux system 108* Examining /proc's structure 109* Uncovering various information about the kernel and the processes running 110 on the system 111 112------------------------------------------------------------------------------ 113 114The proc file system acts as an interface to internal data structures in the 115kernel. It can be used to obtain information about the system and to change 116certain kernel parameters at runtime (sysctl). 117 118First, we'll take a look at the read-only parts of /proc. In Chapter 2, we 119show you how you can use /proc/sys to change settings. 120 1211.1 Process-Specific Subdirectories 122----------------------------------- 123 124The directory /proc contains (among other things) one subdirectory for each 125process running on the system, which is named after the process ID (PID). 126 127The link 'self' points to the process reading the file system. Each process 128subdirectory has the entries listed in Table 1-1. 129 130Note that an open file descriptor to /proc/<pid> or to any of its 131contained files or subdirectories does not prevent <pid> being reused 132for some other process in the event that <pid> exits. Operations on 133open /proc/<pid> file descriptors corresponding to dead processes 134never act on any new process that the kernel may, through chance, have 135also assigned the process ID <pid>. Instead, operations on these FDs 136usually fail with ESRCH. 137 138.. table:: Table 1-1: Process specific entries in /proc 139 140 ============= =============================================================== 141 File Content 142 ============= =============================================================== 143 clear_refs Clears page referenced bits shown in smaps output 144 cmdline Command line arguments 145 cpu Current and last cpu in which it was executed (2.4)(smp) 146 cwd Link to the current working directory 147 environ Values of environment variables 148 exe Link to the executable of this process 149 fd Directory, which contains all file descriptors 150 maps Memory maps to executables and library files (2.4) 151 mem Memory held by this process 152 root Link to the root directory of this process 153 stat Process status 154 statm Process memory status information 155 status Process status in human readable form 156 wchan Present with CONFIG_KALLSYMS=y: it shows the kernel function 157 symbol the task is blocked in - or "0" if not blocked. 158 pagemap Page table 159 stack Report full stack trace, enable via CONFIG_STACKTRACE 160 smaps An extension based on maps, showing the memory consumption of 161 each mapping and flags associated with it 162 smaps_rollup Accumulated smaps stats for all mappings of the process. This 163 can be derived from smaps, but is faster and more convenient 164 numa_maps An extension based on maps, showing the memory locality and 165 binding policy as well as mem usage (in pages) of each mapping. 166 ============= =============================================================== 167 168For example, to get the status information of a process, all you have to do is 169read the file /proc/PID/status:: 170 171 >cat /proc/self/status 172 Name: cat 173 State: R (running) 174 Tgid: 5452 175 Pid: 5452 176 PPid: 743 177 TracerPid: 0 (2.4) 178 Uid: 501 501 501 501 179 Gid: 100 100 100 100 180 FDSize: 256 181 Groups: 100 14 16 182 Kthread: 0 183 VmPeak: 5004 kB 184 VmSize: 5004 kB 185 VmLck: 0 kB 186 VmHWM: 476 kB 187 VmRSS: 476 kB 188 RssAnon: 352 kB 189 RssFile: 120 kB 190 RssShmem: 4 kB 191 VmData: 156 kB 192 VmStk: 88 kB 193 VmExe: 68 kB 194 VmLib: 1412 kB 195 VmPTE: 20 kb 196 VmSwap: 0 kB 197 HugetlbPages: 0 kB 198 CoreDumping: 0 199 THP_enabled: 1 200 Threads: 1 201 SigQ: 0/28578 202 SigPnd: 0000000000000000 203 ShdPnd: 0000000000000000 204 SigBlk: 0000000000000000 205 SigIgn: 0000000000000000 206 SigCgt: 0000000000000000 207 CapInh: 00000000fffffeff 208 CapPrm: 0000000000000000 209 CapEff: 0000000000000000 210 CapBnd: ffffffffffffffff 211 CapAmb: 0000000000000000 212 NoNewPrivs: 0 213 Seccomp: 0 214 Speculation_Store_Bypass: thread vulnerable 215 SpeculationIndirectBranch: conditional enabled 216 voluntary_ctxt_switches: 0 217 nonvoluntary_ctxt_switches: 1 218 219This shows you nearly the same information you would get if you viewed it with 220the ps command. In fact, ps uses the proc file system to obtain its 221information. But you get a more detailed view of the process by reading the 222file /proc/PID/status. It fields are described in table 1-2. 223 224The statm file contains more detailed information about the process 225memory usage. Its seven fields are explained in Table 1-3. The stat file 226contains detailed information about the process itself. Its fields are 227explained in Table 1-4. 228 229(for SMP CONFIG users) 230 231For making accounting scalable, RSS related information are handled in an 232asynchronous manner and the value may not be very precise. To see a precise 233snapshot of a moment, you can see /proc/<pid>/smaps file and scan page table. 234It's slow but very precise. 235 236.. table:: Table 1-2: Contents of the status files (as of 4.19) 237 238 ========================== =================================================== 239 Field Content 240 ========================== =================================================== 241 Name filename of the executable 242 Umask file mode creation mask 243 State state (R is running, S is sleeping, D is sleeping 244 in an uninterruptible wait, Z is zombie, 245 T is traced or stopped) 246 Tgid thread group ID 247 Ngid NUMA group ID (0 if none) 248 Pid process id 249 PPid process id of the parent process 250 TracerPid PID of process tracing this process (0 if not, or 251 the tracer is outside of the current pid namespace) 252 Uid Real, effective, saved set, and file system UIDs 253 Gid Real, effective, saved set, and file system GIDs 254 FDSize number of file descriptor slots currently allocated 255 Groups supplementary group list 256 NStgid descendant namespace thread group ID hierarchy 257 NSpid descendant namespace process ID hierarchy 258 NSpgid descendant namespace process group ID hierarchy 259 NSsid descendant namespace session ID hierarchy 260 Kthread kernel thread flag, 1 is yes, 0 is no 261 VmPeak peak virtual memory size 262 VmSize total program size 263 VmLck locked memory size 264 VmPin pinned memory size 265 VmHWM peak resident set size ("high water mark") 266 VmRSS size of memory portions. It contains the three 267 following parts 268 (VmRSS = RssAnon + RssFile + RssShmem) 269 RssAnon size of resident anonymous memory 270 RssFile size of resident file mappings 271 RssShmem size of resident shmem memory (includes SysV shm, 272 mapping of tmpfs and shared anonymous mappings) 273 VmData size of private data segments 274 VmStk size of stack segments 275 VmExe size of text segment 276 VmLib size of shared library code 277 VmPTE size of page table entries 278 VmSwap amount of swap used by anonymous private data 279 (shmem swap usage is not included) 280 HugetlbPages size of hugetlb memory portions 281 CoreDumping process's memory is currently being dumped 282 (killing the process may lead to a corrupted core) 283 THP_enabled process is allowed to use THP (returns 0 when 284 PR_SET_THP_DISABLE is set on the process 285 Threads number of threads 286 SigQ number of signals queued/max. number for queue 287 SigPnd bitmap of pending signals for the thread 288 ShdPnd bitmap of shared pending signals for the process 289 SigBlk bitmap of blocked signals 290 SigIgn bitmap of ignored signals 291 SigCgt bitmap of caught signals 292 CapInh bitmap of inheritable capabilities 293 CapPrm bitmap of permitted capabilities 294 CapEff bitmap of effective capabilities 295 CapBnd bitmap of capabilities bounding set 296 CapAmb bitmap of ambient capabilities 297 NoNewPrivs no_new_privs, like prctl(PR_GET_NO_NEW_PRIV, ...) 298 Seccomp seccomp mode, like prctl(PR_GET_SECCOMP, ...) 299 Speculation_Store_Bypass speculative store bypass mitigation status 300 SpeculationIndirectBranch indirect branch speculation mode 301 Cpus_allowed mask of CPUs on which this process may run 302 Cpus_allowed_list Same as previous, but in "list format" 303 Mems_allowed mask of memory nodes allowed to this process 304 Mems_allowed_list Same as previous, but in "list format" 305 voluntary_ctxt_switches number of voluntary context switches 306 nonvoluntary_ctxt_switches number of non voluntary context switches 307 ========================== =================================================== 308 309 310.. table:: Table 1-3: Contents of the statm files (as of 2.6.8-rc3) 311 312 ======== =============================== ============================== 313 Field Content 314 ======== =============================== ============================== 315 size total program size (pages) (same as VmSize in status) 316 resident size of memory portions (pages) (same as VmRSS in status) 317 shared number of pages that are shared (i.e. backed by a file, same 318 as RssFile+RssShmem in status) 319 trs number of pages that are 'code' (not including libs; broken, 320 includes data segment) 321 lrs number of pages of library (always 0 on 2.6) 322 drs number of pages of data/stack (including libs; broken, 323 includes library text) 324 dt number of dirty pages (always 0 on 2.6) 325 ======== =============================== ============================== 326 327 328.. table:: Table 1-4: Contents of the stat files (as of 2.6.30-rc7) 329 330 ============= =============================================================== 331 Field Content 332 ============= =============================================================== 333 pid process id 334 tcomm filename of the executable 335 state state (R is running, S is sleeping, D is sleeping in an 336 uninterruptible wait, Z is zombie, T is traced or stopped) 337 ppid process id of the parent process 338 pgrp pgrp of the process 339 sid session id 340 tty_nr tty the process uses 341 tty_pgrp pgrp of the tty 342 flags task flags 343 min_flt number of minor faults 344 cmin_flt number of minor faults with child's 345 maj_flt number of major faults 346 cmaj_flt number of major faults with child's 347 utime user mode jiffies 348 stime kernel mode jiffies 349 cutime user mode jiffies with child's 350 cstime kernel mode jiffies with child's 351 priority priority level 352 nice nice level 353 num_threads number of threads 354 it_real_value (obsolete, always 0) 355 start_time time the process started after system boot 356 vsize virtual memory size 357 rss resident set memory size 358 rsslim current limit in bytes on the rss 359 start_code address above which program text can run 360 end_code address below which program text can run 361 start_stack address of the start of the main process stack 362 esp current value of ESP 363 eip current value of EIP 364 pending bitmap of pending signals 365 blocked bitmap of blocked signals 366 sigign bitmap of ignored signals 367 sigcatch bitmap of caught signals 368 0 (place holder, used to be the wchan address, 369 use /proc/PID/wchan instead) 370 0 (place holder) 371 0 (place holder) 372 exit_signal signal to send to parent thread on exit 373 task_cpu which CPU the task is scheduled on 374 rt_priority realtime priority 375 policy scheduling policy (man sched_setscheduler) 376 blkio_ticks time spent waiting for block IO 377 gtime guest time of the task in jiffies 378 cgtime guest time of the task children in jiffies 379 start_data address above which program data+bss is placed 380 end_data address below which program data+bss is placed 381 start_brk address above which program heap can be expanded with brk() 382 arg_start address above which program command line is placed 383 arg_end address below which program command line is placed 384 env_start address above which program environment is placed 385 env_end address below which program environment is placed 386 exit_code the thread's exit_code in the form reported by the waitpid 387 system call 388 ============= =============================================================== 389 390The /proc/PID/maps file contains the currently mapped memory regions and 391their access permissions. 392 393The format is:: 394 395 address perms offset dev inode pathname 396 397 08048000-08049000 r-xp 00000000 03:00 8312 /opt/test 398 08049000-0804a000 rw-p 00001000 03:00 8312 /opt/test 399 0804a000-0806b000 rw-p 00000000 00:00 0 [heap] 400 a7cb1000-a7cb2000 ---p 00000000 00:00 0 401 a7cb2000-a7eb2000 rw-p 00000000 00:00 0 402 a7eb2000-a7eb3000 ---p 00000000 00:00 0 403 a7eb3000-a7ed5000 rw-p 00000000 00:00 0 404 a7ed5000-a8008000 r-xp 00000000 03:00 4222 /lib/libc.so.6 405 a8008000-a800a000 r--p 00133000 03:00 4222 /lib/libc.so.6 406 a800a000-a800b000 rw-p 00135000 03:00 4222 /lib/libc.so.6 407 a800b000-a800e000 rw-p 00000000 00:00 0 408 a800e000-a8022000 r-xp 00000000 03:00 14462 /lib/libpthread.so.0 409 a8022000-a8023000 r--p 00013000 03:00 14462 /lib/libpthread.so.0 410 a8023000-a8024000 rw-p 00014000 03:00 14462 /lib/libpthread.so.0 411 a8024000-a8027000 rw-p 00000000 00:00 0 412 a8027000-a8043000 r-xp 00000000 03:00 8317 /lib/ld-linux.so.2 413 a8043000-a8044000 r--p 0001b000 03:00 8317 /lib/ld-linux.so.2 414 a8044000-a8045000 rw-p 0001c000 03:00 8317 /lib/ld-linux.so.2 415 aff35000-aff4a000 rw-p 00000000 00:00 0 [stack] 416 ffffe000-fffff000 r-xp 00000000 00:00 0 [vdso] 417 418where "address" is the address space in the process that it occupies, "perms" 419is a set of permissions:: 420 421 r = read 422 w = write 423 x = execute 424 s = shared 425 p = private (copy on write) 426 427"offset" is the offset into the mapping, "dev" is the device (major:minor), and 428"inode" is the inode on that device. 0 indicates that no inode is associated 429with the memory region, as the case would be with BSS (uninitialized data). 430The "pathname" shows the name associated file for this mapping. If the mapping 431is not associated with a file: 432 433 =================== =========================================== 434 [heap] the heap of the program 435 [stack] the stack of the main process 436 [vdso] the "virtual dynamic shared object", 437 the kernel system call handler 438 [anon:<name>] a private anonymous mapping that has been 439 named by userspace 440 [anon_shmem:<name>] an anonymous shared memory mapping that has 441 been named by userspace 442 =================== =========================================== 443 444 or if empty, the mapping is anonymous. 445 446The /proc/PID/smaps is an extension based on maps, showing the memory 447consumption for each of the process's mappings. For each mapping (aka Virtual 448Memory Area, or VMA) there is a series of lines such as the following:: 449 450 08048000-080bc000 r-xp 00000000 03:02 13130 /bin/bash 451 452 Size: 1084 kB 453 KernelPageSize: 4 kB 454 MMUPageSize: 4 kB 455 Rss: 892 kB 456 Pss: 374 kB 457 Pss_Dirty: 0 kB 458 Shared_Clean: 892 kB 459 Shared_Dirty: 0 kB 460 Private_Clean: 0 kB 461 Private_Dirty: 0 kB 462 Referenced: 892 kB 463 Anonymous: 0 kB 464 LazyFree: 0 kB 465 AnonHugePages: 0 kB 466 ShmemPmdMapped: 0 kB 467 Shared_Hugetlb: 0 kB 468 Private_Hugetlb: 0 kB 469 Swap: 0 kB 470 SwapPss: 0 kB 471 KernelPageSize: 4 kB 472 MMUPageSize: 4 kB 473 Locked: 0 kB 474 THPeligible: 0 475 VmFlags: rd ex mr mw me dw 476 477The first of these lines shows the same information as is displayed for the 478mapping in /proc/PID/maps. Following lines show the size of the mapping 479(size); the size of each page allocated when backing a VMA (KernelPageSize), 480which is usually the same as the size in the page table entries; the page size 481used by the MMU when backing a VMA (in most cases, the same as KernelPageSize); 482the amount of the mapping that is currently resident in RAM (RSS); the 483process' proportional share of this mapping (PSS); and the number of clean and 484dirty shared and private pages in the mapping. 485 486The "proportional set size" (PSS) of a process is the count of pages it has 487in memory, where each page is divided by the number of processes sharing it. 488So if a process has 1000 pages all to itself, and 1000 shared with one other 489process, its PSS will be 1500. "Pss_Dirty" is the portion of PSS which 490consists of dirty pages. ("Pss_Clean" is not included, but it can be 491calculated by subtracting "Pss_Dirty" from "Pss".) 492 493Note that even a page which is part of a MAP_SHARED mapping, but has only 494a single pte mapped, i.e. is currently used by only one process, is accounted 495as private and not as shared. 496 497"Referenced" indicates the amount of memory currently marked as referenced or 498accessed. 499 500"Anonymous" shows the amount of memory that does not belong to any file. Even 501a mapping associated with a file may contain anonymous pages: when MAP_PRIVATE 502and a page is modified, the file page is replaced by a private anonymous copy. 503 504"LazyFree" shows the amount of memory which is marked by madvise(MADV_FREE). 505The memory isn't freed immediately with madvise(). It's freed in memory 506pressure if the memory is clean. Please note that the printed value might 507be lower than the real value due to optimizations used in the current 508implementation. If this is not desirable please file a bug report. 509 510"AnonHugePages" shows the ammount of memory backed by transparent hugepage. 511 512"ShmemPmdMapped" shows the ammount of shared (shmem/tmpfs) memory backed by 513huge pages. 514 515"Shared_Hugetlb" and "Private_Hugetlb" show the ammounts of memory backed by 516hugetlbfs page which is *not* counted in "RSS" or "PSS" field for historical 517reasons. And these are not included in {Shared,Private}_{Clean,Dirty} field. 518 519"Swap" shows how much would-be-anonymous memory is also used, but out on swap. 520 521For shmem mappings, "Swap" includes also the size of the mapped (and not 522replaced by copy-on-write) part of the underlying shmem object out on swap. 523"SwapPss" shows proportional swap share of this mapping. Unlike "Swap", this 524does not take into account swapped out page of underlying shmem objects. 525"Locked" indicates whether the mapping is locked in memory or not. 526 527"THPeligible" indicates whether the mapping is eligible for allocating THP 528pages as well as the THP is PMD mappable or not - 1 if true, 0 otherwise. 529It just shows the current status. 530 531"VmFlags" field deserves a separate description. This member represents the 532kernel flags associated with the particular virtual memory area in two letter 533encoded manner. The codes are the following: 534 535 == ======================================= 536 rd readable 537 wr writeable 538 ex executable 539 sh shared 540 mr may read 541 mw may write 542 me may execute 543 ms may share 544 gd stack segment growns down 545 pf pure PFN range 546 dw disabled write to the mapped file 547 lo pages are locked in memory 548 io memory mapped I/O area 549 sr sequential read advise provided 550 rr random read advise provided 551 dc do not copy area on fork 552 de do not expand area on remapping 553 ac area is accountable 554 nr swap space is not reserved for the area 555 ht area uses huge tlb pages 556 sf synchronous page fault 557 ar architecture specific flag 558 wf wipe on fork 559 dd do not include area into core dump 560 sd soft dirty flag 561 mm mixed map area 562 hg huge page advise flag 563 nh no huge page advise flag 564 mg mergable advise flag 565 bt arm64 BTI guarded page 566 mt arm64 MTE allocation tags are enabled 567 um userfaultfd missing tracking 568 uw userfaultfd wr-protect tracking 569 == ======================================= 570 571Note that there is no guarantee that every flag and associated mnemonic will 572be present in all further kernel releases. Things get changed, the flags may 573be vanished or the reverse -- new added. Interpretation of their meaning 574might change in future as well. So each consumer of these flags has to 575follow each specific kernel version for the exact semantic. 576 577This file is only present if the CONFIG_MMU kernel configuration option is 578enabled. 579 580Note: reading /proc/PID/maps or /proc/PID/smaps is inherently racy (consistent 581output can be achieved only in the single read call). 582 583This typically manifests when doing partial reads of these files while the 584memory map is being modified. Despite the races, we do provide the following 585guarantees: 586 5871) The mapped addresses never go backwards, which implies no two 588 regions will ever overlap. 5892) If there is something at a given vaddr during the entirety of the 590 life of the smaps/maps walk, there will be some output for it. 591 592The /proc/PID/smaps_rollup file includes the same fields as /proc/PID/smaps, 593but their values are the sums of the corresponding values for all mappings of 594the process. Additionally, it contains these fields: 595 596- Pss_Anon 597- Pss_File 598- Pss_Shmem 599 600They represent the proportional shares of anonymous, file, and shmem pages, as 601described for smaps above. These fields are omitted in smaps since each 602mapping identifies the type (anon, file, or shmem) of all pages it contains. 603Thus all information in smaps_rollup can be derived from smaps, but at a 604significantly higher cost. 605 606The /proc/PID/clear_refs is used to reset the PG_Referenced and ACCESSED/YOUNG 607bits on both physical and virtual pages associated with a process, and the 608soft-dirty bit on pte (see Documentation/admin-guide/mm/soft-dirty.rst 609for details). 610To clear the bits for all the pages associated with the process:: 611 612 > echo 1 > /proc/PID/clear_refs 613 614To clear the bits for the anonymous pages associated with the process:: 615 616 > echo 2 > /proc/PID/clear_refs 617 618To clear the bits for the file mapped pages associated with the process:: 619 620 > echo 3 > /proc/PID/clear_refs 621 622To clear the soft-dirty bit:: 623 624 > echo 4 > /proc/PID/clear_refs 625 626To reset the peak resident set size ("high water mark") to the process's 627current value:: 628 629 > echo 5 > /proc/PID/clear_refs 630 631Any other value written to /proc/PID/clear_refs will have no effect. 632 633The /proc/pid/pagemap gives the PFN, which can be used to find the pageflags 634using /proc/kpageflags and number of times a page is mapped using 635/proc/kpagecount. For detailed explanation, see 636Documentation/admin-guide/mm/pagemap.rst. 637 638The /proc/pid/numa_maps is an extension based on maps, showing the memory 639locality and binding policy, as well as the memory usage (in pages) of 640each mapping. The output follows a general format where mapping details get 641summarized separated by blank spaces, one mapping per each file line:: 642 643 address policy mapping details 644 645 00400000 default file=/usr/local/bin/app mapped=1 active=0 N3=1 kernelpagesize_kB=4 646 00600000 default file=/usr/local/bin/app anon=1 dirty=1 N3=1 kernelpagesize_kB=4 647 3206000000 default file=/lib64/ld-2.12.so mapped=26 mapmax=6 N0=24 N3=2 kernelpagesize_kB=4 648 320621f000 default file=/lib64/ld-2.12.so anon=1 dirty=1 N3=1 kernelpagesize_kB=4 649 3206220000 default file=/lib64/ld-2.12.so anon=1 dirty=1 N3=1 kernelpagesize_kB=4 650 3206221000 default anon=1 dirty=1 N3=1 kernelpagesize_kB=4 651 3206800000 default file=/lib64/libc-2.12.so mapped=59 mapmax=21 active=55 N0=41 N3=18 kernelpagesize_kB=4 652 320698b000 default file=/lib64/libc-2.12.so 653 3206b8a000 default file=/lib64/libc-2.12.so anon=2 dirty=2 N3=2 kernelpagesize_kB=4 654 3206b8e000 default file=/lib64/libc-2.12.so anon=1 dirty=1 N3=1 kernelpagesize_kB=4 655 3206b8f000 default anon=3 dirty=3 active=1 N3=3 kernelpagesize_kB=4 656 7f4dc10a2000 default anon=3 dirty=3 N3=3 kernelpagesize_kB=4 657 7f4dc10b4000 default anon=2 dirty=2 active=1 N3=2 kernelpagesize_kB=4 658 7f4dc1200000 default file=/anon_hugepage\040(deleted) huge anon=1 dirty=1 N3=1 kernelpagesize_kB=2048 659 7fff335f0000 default stack anon=3 dirty=3 N3=3 kernelpagesize_kB=4 660 7fff3369d000 default mapped=1 mapmax=35 active=0 N3=1 kernelpagesize_kB=4 661 662Where: 663 664"address" is the starting address for the mapping; 665 666"policy" reports the NUMA memory policy set for the mapping (see Documentation/admin-guide/mm/numa_memory_policy.rst); 667 668"mapping details" summarizes mapping data such as mapping type, page usage counters, 669node locality page counters (N0 == node0, N1 == node1, ...) and the kernel page 670size, in KB, that is backing the mapping up. 671 6721.2 Kernel data 673--------------- 674 675Similar to the process entries, the kernel data files give information about 676the running kernel. The files used to obtain this information are contained in 677/proc and are listed in Table 1-5. Not all of these will be present in your 678system. It depends on the kernel configuration and the loaded modules, which 679files are there, and which are missing. 680 681.. table:: Table 1-5: Kernel info in /proc 682 683 ============ =============================================================== 684 File Content 685 ============ =============================================================== 686 apm Advanced power management info 687 buddyinfo Kernel memory allocator information (see text) (2.5) 688 bus Directory containing bus specific information 689 cmdline Kernel command line 690 cpuinfo Info about the CPU 691 devices Available devices (block and character) 692 dma Used DMS channels 693 filesystems Supported filesystems 694 driver Various drivers grouped here, currently rtc (2.4) 695 execdomains Execdomains, related to security (2.4) 696 fb Frame Buffer devices (2.4) 697 fs File system parameters, currently nfs/exports (2.4) 698 ide Directory containing info about the IDE subsystem 699 interrupts Interrupt usage 700 iomem Memory map (2.4) 701 ioports I/O port usage 702 irq Masks for irq to cpu affinity (2.4)(smp?) 703 isapnp ISA PnP (Plug&Play) Info (2.4) 704 kcore Kernel core image (can be ELF or A.OUT(deprecated in 2.4)) 705 kmsg Kernel messages 706 ksyms Kernel symbol table 707 loadavg Load average of last 1, 5 & 15 minutes; 708 number of processes currently runnable (running or on ready queue); 709 total number of processes in system; 710 last pid created. 711 All fields are separated by one space except "number of 712 processes currently runnable" and "total number of processes 713 in system", which are separated by a slash ('/'). Example: 714 0.61 0.61 0.55 3/828 22084 715 locks Kernel locks 716 meminfo Memory info 717 misc Miscellaneous 718 modules List of loaded modules 719 mounts Mounted filesystems 720 net Networking info (see text) 721 pagetypeinfo Additional page allocator information (see text) (2.5) 722 partitions Table of partitions known to the system 723 pci Deprecated info of PCI bus (new way -> /proc/bus/pci/, 724 decoupled by lspci (2.4) 725 rtc Real time clock 726 scsi SCSI info (see text) 727 slabinfo Slab pool info 728 softirqs softirq usage 729 stat Overall statistics 730 swaps Swap space utilization 731 sys See chapter 2 732 sysvipc Info of SysVIPC Resources (msg, sem, shm) (2.4) 733 tty Info of tty drivers 734 uptime Wall clock since boot, combined idle time of all cpus 735 version Kernel version 736 video bttv info of video resources (2.4) 737 vmallocinfo Show vmalloced areas 738 ============ =============================================================== 739 740You can, for example, check which interrupts are currently in use and what 741they are used for by looking in the file /proc/interrupts:: 742 743 > cat /proc/interrupts 744 CPU0 745 0: 8728810 XT-PIC timer 746 1: 895 XT-PIC keyboard 747 2: 0 XT-PIC cascade 748 3: 531695 XT-PIC aha152x 749 4: 2014133 XT-PIC serial 750 5: 44401 XT-PIC pcnet_cs 751 8: 2 XT-PIC rtc 752 11: 8 XT-PIC i82365 753 12: 182918 XT-PIC PS/2 Mouse 754 13: 1 XT-PIC fpu 755 14: 1232265 XT-PIC ide0 756 15: 7 XT-PIC ide1 757 NMI: 0 758 759In 2.4.* a couple of lines where added to this file LOC & ERR (this time is the 760output of a SMP machine):: 761 762 > cat /proc/interrupts 763 764 CPU0 CPU1 765 0: 1243498 1214548 IO-APIC-edge timer 766 1: 8949 8958 IO-APIC-edge keyboard 767 2: 0 0 XT-PIC cascade 768 5: 11286 10161 IO-APIC-edge soundblaster 769 8: 1 0 IO-APIC-edge rtc 770 9: 27422 27407 IO-APIC-edge 3c503 771 12: 113645 113873 IO-APIC-edge PS/2 Mouse 772 13: 0 0 XT-PIC fpu 773 14: 22491 24012 IO-APIC-edge ide0 774 15: 2183 2415 IO-APIC-edge ide1 775 17: 30564 30414 IO-APIC-level eth0 776 18: 177 164 IO-APIC-level bttv 777 NMI: 2457961 2457959 778 LOC: 2457882 2457881 779 ERR: 2155 780 781NMI is incremented in this case because every timer interrupt generates a NMI 782(Non Maskable Interrupt) which is used by the NMI Watchdog to detect lockups. 783 784LOC is the local interrupt counter of the internal APIC of every CPU. 785 786ERR is incremented in the case of errors in the IO-APIC bus (the bus that 787connects the CPUs in a SMP system. This means that an error has been detected, 788the IO-APIC automatically retry the transmission, so it should not be a big 789problem, but you should read the SMP-FAQ. 790 791In 2.6.2* /proc/interrupts was expanded again. This time the goal was for 792/proc/interrupts to display every IRQ vector in use by the system, not 793just those considered 'most important'. The new vectors are: 794 795THR 796 interrupt raised when a machine check threshold counter 797 (typically counting ECC corrected errors of memory or cache) exceeds 798 a configurable threshold. Only available on some systems. 799 800TRM 801 a thermal event interrupt occurs when a temperature threshold 802 has been exceeded for the CPU. This interrupt may also be generated 803 when the temperature drops back to normal. 804 805SPU 806 a spurious interrupt is some interrupt that was raised then lowered 807 by some IO device before it could be fully processed by the APIC. Hence 808 the APIC sees the interrupt but does not know what device it came from. 809 For this case the APIC will generate the interrupt with a IRQ vector 810 of 0xff. This might also be generated by chipset bugs. 811 812RES, CAL, TLB 813 rescheduling, call and TLB flush interrupts are 814 sent from one CPU to another per the needs of the OS. Typically, 815 their statistics are used by kernel developers and interested users to 816 determine the occurrence of interrupts of the given type. 817 818The above IRQ vectors are displayed only when relevant. For example, 819the threshold vector does not exist on x86_64 platforms. Others are 820suppressed when the system is a uniprocessor. As of this writing, only 821i386 and x86_64 platforms support the new IRQ vector displays. 822 823Of some interest is the introduction of the /proc/irq directory to 2.4. 824It could be used to set IRQ to CPU affinity. This means that you can "hook" an 825IRQ to only one CPU, or to exclude a CPU of handling IRQs. The contents of the 826irq subdir is one subdir for each IRQ, and two files; default_smp_affinity and 827prof_cpu_mask. 828 829For example:: 830 831 > ls /proc/irq/ 832 0 10 12 14 16 18 2 4 6 8 prof_cpu_mask 833 1 11 13 15 17 19 3 5 7 9 default_smp_affinity 834 > ls /proc/irq/0/ 835 smp_affinity 836 837smp_affinity is a bitmask, in which you can specify which CPUs can handle the 838IRQ. You can set it by doing:: 839 840 > echo 1 > /proc/irq/10/smp_affinity 841 842This means that only the first CPU will handle the IRQ, but you can also echo 8435 which means that only the first and third CPU can handle the IRQ. 844 845The contents of each smp_affinity file is the same by default:: 846 847 > cat /proc/irq/0/smp_affinity 848 ffffffff 849 850There is an alternate interface, smp_affinity_list which allows specifying 851a CPU range instead of a bitmask:: 852 853 > cat /proc/irq/0/smp_affinity_list 854 1024-1031 855 856The default_smp_affinity mask applies to all non-active IRQs, which are the 857IRQs which have not yet been allocated/activated, and hence which lack a 858/proc/irq/[0-9]* directory. 859 860The node file on an SMP system shows the node to which the device using the IRQ 861reports itself as being attached. This hardware locality information does not 862include information about any possible driver locality preference. 863 864prof_cpu_mask specifies which CPUs are to be profiled by the system wide 865profiler. Default value is ffffffff (all CPUs if there are only 32 of them). 866 867The way IRQs are routed is handled by the IO-APIC, and it's Round Robin 868between all the CPUs which are allowed to handle it. As usual the kernel has 869more info than you and does a better job than you, so the defaults are the 870best choice for almost everyone. [Note this applies only to those IO-APIC's 871that support "Round Robin" interrupt distribution.] 872 873There are three more important subdirectories in /proc: net, scsi, and sys. 874The general rule is that the contents, or even the existence of these 875directories, depend on your kernel configuration. If SCSI is not enabled, the 876directory scsi may not exist. The same is true with the net, which is there 877only when networking support is present in the running kernel. 878 879The slabinfo file gives information about memory usage at the slab level. 880Linux uses slab pools for memory management above page level in version 2.2. 881Commonly used objects have their own slab pool (such as network buffers, 882directory cache, and so on). 883 884:: 885 886 > cat /proc/buddyinfo 887 888 Node 0, zone DMA 0 4 5 4 4 3 ... 889 Node 0, zone Normal 1 0 0 1 101 8 ... 890 Node 0, zone HighMem 2 0 0 1 1 0 ... 891 892External fragmentation is a problem under some workloads, and buddyinfo is a 893useful tool for helping diagnose these problems. Buddyinfo will give you a 894clue as to how big an area you can safely allocate, or why a previous 895allocation failed. 896 897Each column represents the number of pages of a certain order which are 898available. In this case, there are 0 chunks of 2^0*PAGE_SIZE available in 899ZONE_DMA, 4 chunks of 2^1*PAGE_SIZE in ZONE_DMA, 101 chunks of 2^4*PAGE_SIZE 900available in ZONE_NORMAL, etc... 901 902More information relevant to external fragmentation can be found in 903pagetypeinfo:: 904 905 > cat /proc/pagetypeinfo 906 Page block order: 9 907 Pages per block: 512 908 909 Free pages count per migrate type at order 0 1 2 3 4 5 6 7 8 9 10 910 Node 0, zone DMA, type Unmovable 0 0 0 1 1 1 1 1 1 1 0 911 Node 0, zone DMA, type Reclaimable 0 0 0 0 0 0 0 0 0 0 0 912 Node 0, zone DMA, type Movable 1 1 2 1 2 1 1 0 1 0 2 913 Node 0, zone DMA, type Reserve 0 0 0 0 0 0 0 0 0 1 0 914 Node 0, zone DMA, type Isolate 0 0 0 0 0 0 0 0 0 0 0 915 Node 0, zone DMA32, type Unmovable 103 54 77 1 1 1 11 8 7 1 9 916 Node 0, zone DMA32, type Reclaimable 0 0 2 1 0 0 0 0 1 0 0 917 Node 0, zone DMA32, type Movable 169 152 113 91 77 54 39 13 6 1 452 918 Node 0, zone DMA32, type Reserve 1 2 2 2 2 0 1 1 1 1 0 919 Node 0, zone DMA32, type Isolate 0 0 0 0 0 0 0 0 0 0 0 920 921 Number of blocks type Unmovable Reclaimable Movable Reserve Isolate 922 Node 0, zone DMA 2 0 5 1 0 923 Node 0, zone DMA32 41 6 967 2 0 924 925Fragmentation avoidance in the kernel works by grouping pages of different 926migrate types into the same contiguous regions of memory called page blocks. 927A page block is typically the size of the default hugepage size, e.g. 2MB on 928X86-64. By keeping pages grouped based on their ability to move, the kernel 929can reclaim pages within a page block to satisfy a high-order allocation. 930 931The pagetypinfo begins with information on the size of a page block. It 932then gives the same type of information as buddyinfo except broken down 933by migrate-type and finishes with details on how many page blocks of each 934type exist. 935 936If min_free_kbytes has been tuned correctly (recommendations made by hugeadm 937from libhugetlbfs https://github.com/libhugetlbfs/libhugetlbfs/), one can 938make an estimate of the likely number of huge pages that can be allocated 939at a given point in time. All the "Movable" blocks should be allocatable 940unless memory has been mlock()'d. Some of the Reclaimable blocks should 941also be allocatable although a lot of filesystem metadata may have to be 942reclaimed to achieve this. 943 944 945meminfo 946~~~~~~~ 947 948Provides information about distribution and utilization of memory. This 949varies by architecture and compile options. Some of the counters reported 950here overlap. The memory reported by the non overlapping counters may not 951add up to the overall memory usage and the difference for some workloads 952can be substantial. In many cases there are other means to find out 953additional memory using subsystem specific interfaces, for instance 954/proc/net/sockstat for TCP memory allocations. 955 956Example output. You may not have all of these fields. 957 958:: 959 960 > cat /proc/meminfo 961 962 MemTotal: 32858820 kB 963 MemFree: 21001236 kB 964 MemAvailable: 27214312 kB 965 Buffers: 581092 kB 966 Cached: 5587612 kB 967 SwapCached: 0 kB 968 Active: 3237152 kB 969 Inactive: 7586256 kB 970 Active(anon): 94064 kB 971 Inactive(anon): 4570616 kB 972 Active(file): 3143088 kB 973 Inactive(file): 3015640 kB 974 Unevictable: 0 kB 975 Mlocked: 0 kB 976 SwapTotal: 0 kB 977 SwapFree: 0 kB 978 Zswap: 1904 kB 979 Zswapped: 7792 kB 980 Dirty: 12 kB 981 Writeback: 0 kB 982 AnonPages: 4654780 kB 983 Mapped: 266244 kB 984 Shmem: 9976 kB 985 KReclaimable: 517708 kB 986 Slab: 660044 kB 987 SReclaimable: 517708 kB 988 SUnreclaim: 142336 kB 989 KernelStack: 11168 kB 990 PageTables: 20540 kB 991 SecPageTables: 0 kB 992 NFS_Unstable: 0 kB 993 Bounce: 0 kB 994 WritebackTmp: 0 kB 995 CommitLimit: 16429408 kB 996 Committed_AS: 7715148 kB 997 VmallocTotal: 34359738367 kB 998 VmallocUsed: 40444 kB 999 VmallocChunk: 0 kB 1000 Percpu: 29312 kB 1001 HardwareCorrupted: 0 kB 1002 AnonHugePages: 4149248 kB 1003 ShmemHugePages: 0 kB 1004 ShmemPmdMapped: 0 kB 1005 FileHugePages: 0 kB 1006 FilePmdMapped: 0 kB 1007 CmaTotal: 0 kB 1008 CmaFree: 0 kB 1009 HugePages_Total: 0 1010 HugePages_Free: 0 1011 HugePages_Rsvd: 0 1012 HugePages_Surp: 0 1013 Hugepagesize: 2048 kB 1014 Hugetlb: 0 kB 1015 DirectMap4k: 401152 kB 1016 DirectMap2M: 10008576 kB 1017 DirectMap1G: 24117248 kB 1018 1019MemTotal 1020 Total usable RAM (i.e. physical RAM minus a few reserved 1021 bits and the kernel binary code) 1022MemFree 1023 Total free RAM. On highmem systems, the sum of LowFree+HighFree 1024MemAvailable 1025 An estimate of how much memory is available for starting new 1026 applications, without swapping. Calculated from MemFree, 1027 SReclaimable, the size of the file LRU lists, and the low 1028 watermarks in each zone. 1029 The estimate takes into account that the system needs some 1030 page cache to function well, and that not all reclaimable 1031 slab will be reclaimable, due to items being in use. The 1032 impact of those factors will vary from system to system. 1033Buffers 1034 Relatively temporary storage for raw disk blocks 1035 shouldn't get tremendously large (20MB or so) 1036Cached 1037 In-memory cache for files read from the disk (the 1038 pagecache) as well as tmpfs & shmem. 1039 Doesn't include SwapCached. 1040SwapCached 1041 Memory that once was swapped out, is swapped back in but 1042 still also is in the swapfile (if memory is needed it 1043 doesn't need to be swapped out AGAIN because it is already 1044 in the swapfile. This saves I/O) 1045Active 1046 Memory that has been used more recently and usually not 1047 reclaimed unless absolutely necessary. 1048Inactive 1049 Memory which has been less recently used. It is more 1050 eligible to be reclaimed for other purposes 1051Unevictable 1052 Memory allocated for userspace which cannot be reclaimed, such 1053 as mlocked pages, ramfs backing pages, secret memfd pages etc. 1054Mlocked 1055 Memory locked with mlock(). 1056HighTotal, HighFree 1057 Highmem is all memory above ~860MB of physical memory. 1058 Highmem areas are for use by userspace programs, or 1059 for the pagecache. The kernel must use tricks to access 1060 this memory, making it slower to access than lowmem. 1061LowTotal, LowFree 1062 Lowmem is memory which can be used for everything that 1063 highmem can be used for, but it is also available for the 1064 kernel's use for its own data structures. Among many 1065 other things, it is where everything from the Slab is 1066 allocated. Bad things happen when you're out of lowmem. 1067SwapTotal 1068 total amount of swap space available 1069SwapFree 1070 Memory which has been evicted from RAM, and is temporarily 1071 on the disk 1072Zswap 1073 Memory consumed by the zswap backend (compressed size) 1074Zswapped 1075 Amount of anonymous memory stored in zswap (original size) 1076Dirty 1077 Memory which is waiting to get written back to the disk 1078Writeback 1079 Memory which is actively being written back to the disk 1080AnonPages 1081 Non-file backed pages mapped into userspace page tables 1082Mapped 1083 files which have been mmaped, such as libraries 1084Shmem 1085 Total memory used by shared memory (shmem) and tmpfs 1086KReclaimable 1087 Kernel allocations that the kernel will attempt to reclaim 1088 under memory pressure. Includes SReclaimable (below), and other 1089 direct allocations with a shrinker. 1090Slab 1091 in-kernel data structures cache 1092SReclaimable 1093 Part of Slab, that might be reclaimed, such as caches 1094SUnreclaim 1095 Part of Slab, that cannot be reclaimed on memory pressure 1096KernelStack 1097 Memory consumed by the kernel stacks of all tasks 1098PageTables 1099 Memory consumed by userspace page tables 1100SecPageTables 1101 Memory consumed by secondary page tables, this currently 1102 currently includes KVM mmu allocations on x86 and arm64. 1103NFS_Unstable 1104 Always zero. Previous counted pages which had been written to 1105 the server, but has not been committed to stable storage. 1106Bounce 1107 Memory used for block device "bounce buffers" 1108WritebackTmp 1109 Memory used by FUSE for temporary writeback buffers 1110CommitLimit 1111 Based on the overcommit ratio ('vm.overcommit_ratio'), 1112 this is the total amount of memory currently available to 1113 be allocated on the system. This limit is only adhered to 1114 if strict overcommit accounting is enabled (mode 2 in 1115 'vm.overcommit_memory'). 1116 1117 The CommitLimit is calculated with the following formula:: 1118 1119 CommitLimit = ([total RAM pages] - [total huge TLB pages]) * 1120 overcommit_ratio / 100 + [total swap pages] 1121 1122 For example, on a system with 1G of physical RAM and 7G 1123 of swap with a `vm.overcommit_ratio` of 30 it would 1124 yield a CommitLimit of 7.3G. 1125 1126 For more details, see the memory overcommit documentation 1127 in mm/overcommit-accounting. 1128Committed_AS 1129 The amount of memory presently allocated on the system. 1130 The committed memory is a sum of all of the memory which 1131 has been allocated by processes, even if it has not been 1132 "used" by them as of yet. A process which malloc()'s 1G 1133 of memory, but only touches 300M of it will show up as 1134 using 1G. This 1G is memory which has been "committed" to 1135 by the VM and can be used at any time by the allocating 1136 application. With strict overcommit enabled on the system 1137 (mode 2 in 'vm.overcommit_memory'), allocations which would 1138 exceed the CommitLimit (detailed above) will not be permitted. 1139 This is useful if one needs to guarantee that processes will 1140 not fail due to lack of memory once that memory has been 1141 successfully allocated. 1142VmallocTotal 1143 total size of vmalloc virtual address space 1144VmallocUsed 1145 amount of vmalloc area which is used 1146VmallocChunk 1147 largest contiguous block of vmalloc area which is free 1148Percpu 1149 Memory allocated to the percpu allocator used to back percpu 1150 allocations. This stat excludes the cost of metadata. 1151HardwareCorrupted 1152 The amount of RAM/memory in KB, the kernel identifies as 1153 corrupted. 1154AnonHugePages 1155 Non-file backed huge pages mapped into userspace page tables 1156ShmemHugePages 1157 Memory used by shared memory (shmem) and tmpfs allocated 1158 with huge pages 1159ShmemPmdMapped 1160 Shared memory mapped into userspace with huge pages 1161FileHugePages 1162 Memory used for filesystem data (page cache) allocated 1163 with huge pages 1164FilePmdMapped 1165 Page cache mapped into userspace with huge pages 1166CmaTotal 1167 Memory reserved for the Contiguous Memory Allocator (CMA) 1168CmaFree 1169 Free remaining memory in the CMA reserves 1170HugePages_Total, HugePages_Free, HugePages_Rsvd, HugePages_Surp, Hugepagesize, Hugetlb 1171 See Documentation/admin-guide/mm/hugetlbpage.rst. 1172DirectMap4k, DirectMap2M, DirectMap1G 1173 Breakdown of page table sizes used in the kernel's 1174 identity mapping of RAM 1175 1176vmallocinfo 1177~~~~~~~~~~~ 1178 1179Provides information about vmalloced/vmaped areas. One line per area, 1180containing the virtual address range of the area, size in bytes, 1181caller information of the creator, and optional information depending 1182on the kind of area: 1183 1184 ========== =================================================== 1185 pages=nr number of pages 1186 phys=addr if a physical address was specified 1187 ioremap I/O mapping (ioremap() and friends) 1188 vmalloc vmalloc() area 1189 vmap vmap()ed pages 1190 user VM_USERMAP area 1191 vpages buffer for pages pointers was vmalloced (huge area) 1192 N<node>=nr (Only on NUMA kernels) 1193 Number of pages allocated on memory node <node> 1194 ========== =================================================== 1195 1196:: 1197 1198 > cat /proc/vmallocinfo 1199 0xffffc20000000000-0xffffc20000201000 2101248 alloc_large_system_hash+0x204 ... 1200 /0x2c0 pages=512 vmalloc N0=128 N1=128 N2=128 N3=128 1201 0xffffc20000201000-0xffffc20000302000 1052672 alloc_large_system_hash+0x204 ... 1202 /0x2c0 pages=256 vmalloc N0=64 N1=64 N2=64 N3=64 1203 0xffffc20000302000-0xffffc20000304000 8192 acpi_tb_verify_table+0x21/0x4f... 1204 phys=7fee8000 ioremap 1205 0xffffc20000304000-0xffffc20000307000 12288 acpi_tb_verify_table+0x21/0x4f... 1206 phys=7fee7000 ioremap 1207 0xffffc2000031d000-0xffffc2000031f000 8192 init_vdso_vars+0x112/0x210 1208 0xffffc2000031f000-0xffffc2000032b000 49152 cramfs_uncompress_init+0x2e ... 1209 /0x80 pages=11 vmalloc N0=3 N1=3 N2=2 N3=3 1210 0xffffc2000033a000-0xffffc2000033d000 12288 sys_swapon+0x640/0xac0 ... 1211 pages=2 vmalloc N1=2 1212 0xffffc20000347000-0xffffc2000034c000 20480 xt_alloc_table_info+0xfe ... 1213 /0x130 [x_tables] pages=4 vmalloc N0=4 1214 0xffffffffa0000000-0xffffffffa000f000 61440 sys_init_module+0xc27/0x1d00 ... 1215 pages=14 vmalloc N2=14 1216 0xffffffffa000f000-0xffffffffa0014000 20480 sys_init_module+0xc27/0x1d00 ... 1217 pages=4 vmalloc N1=4 1218 0xffffffffa0014000-0xffffffffa0017000 12288 sys_init_module+0xc27/0x1d00 ... 1219 pages=2 vmalloc N1=2 1220 0xffffffffa0017000-0xffffffffa0022000 45056 sys_init_module+0xc27/0x1d00 ... 1221 pages=10 vmalloc N0=10 1222 1223 1224softirqs 1225~~~~~~~~ 1226 1227Provides counts of softirq handlers serviced since boot time, for each CPU. 1228 1229:: 1230 1231 > cat /proc/softirqs 1232 CPU0 CPU1 CPU2 CPU3 1233 HI: 0 0 0 0 1234 TIMER: 27166 27120 27097 27034 1235 NET_TX: 0 0 0 17 1236 NET_RX: 42 0 0 39 1237 BLOCK: 0 0 107 1121 1238 TASKLET: 0 0 0 290 1239 SCHED: 27035 26983 26971 26746 1240 HRTIMER: 0 0 0 0 1241 RCU: 1678 1769 2178 2250 1242 12431.3 Networking info in /proc/net 1244-------------------------------- 1245 1246The subdirectory /proc/net follows the usual pattern. Table 1-8 shows the 1247additional values you get for IP version 6 if you configure the kernel to 1248support this. Table 1-9 lists the files and their meaning. 1249 1250 1251.. table:: Table 1-8: IPv6 info in /proc/net 1252 1253 ========== ===================================================== 1254 File Content 1255 ========== ===================================================== 1256 udp6 UDP sockets (IPv6) 1257 tcp6 TCP sockets (IPv6) 1258 raw6 Raw device statistics (IPv6) 1259 igmp6 IP multicast addresses, which this host joined (IPv6) 1260 if_inet6 List of IPv6 interface addresses 1261 ipv6_route Kernel routing table for IPv6 1262 rt6_stats Global IPv6 routing tables statistics 1263 sockstat6 Socket statistics (IPv6) 1264 snmp6 Snmp data (IPv6) 1265 ========== ===================================================== 1266 1267.. table:: Table 1-9: Network info in /proc/net 1268 1269 ============= ================================================================ 1270 File Content 1271 ============= ================================================================ 1272 arp Kernel ARP table 1273 dev network devices with statistics 1274 dev_mcast the Layer2 multicast groups a device is listening too 1275 (interface index, label, number of references, number of bound 1276 addresses). 1277 dev_stat network device status 1278 ip_fwchains Firewall chain linkage 1279 ip_fwnames Firewall chain names 1280 ip_masq Directory containing the masquerading tables 1281 ip_masquerade Major masquerading table 1282 netstat Network statistics 1283 raw raw device statistics 1284 route Kernel routing table 1285 rpc Directory containing rpc info 1286 rt_cache Routing cache 1287 snmp SNMP data 1288 sockstat Socket statistics 1289 softnet_stat Per-CPU incoming packets queues statistics of online CPUs 1290 tcp TCP sockets 1291 udp UDP sockets 1292 unix UNIX domain sockets 1293 wireless Wireless interface data (Wavelan etc) 1294 igmp IP multicast addresses, which this host joined 1295 psched Global packet scheduler parameters. 1296 netlink List of PF_NETLINK sockets 1297 ip_mr_vifs List of multicast virtual interfaces 1298 ip_mr_cache List of multicast routing cache 1299 ============= ================================================================ 1300 1301You can use this information to see which network devices are available in 1302your system and how much traffic was routed over those devices:: 1303 1304 > cat /proc/net/dev 1305 Inter-|Receive |[... 1306 face |bytes packets errs drop fifo frame compressed multicast|[... 1307 lo: 908188 5596 0 0 0 0 0 0 [... 1308 ppp0:15475140 20721 410 0 0 410 0 0 [... 1309 eth0: 614530 7085 0 0 0 0 0 1 [... 1310 1311 ...] Transmit 1312 ...] bytes packets errs drop fifo colls carrier compressed 1313 ...] 908188 5596 0 0 0 0 0 0 1314 ...] 1375103 17405 0 0 0 0 0 0 1315 ...] 1703981 5535 0 0 0 3 0 0 1316 1317In addition, each Channel Bond interface has its own directory. For 1318example, the bond0 device will have a directory called /proc/net/bond0/. 1319It will contain information that is specific to that bond, such as the 1320current slaves of the bond, the link status of the slaves, and how 1321many times the slaves link has failed. 1322 13231.4 SCSI info 1324------------- 1325 1326If you have a SCSI host adapter in your system, you'll find a subdirectory 1327named after the driver for this adapter in /proc/scsi. You'll also see a list 1328of all recognized SCSI devices in /proc/scsi:: 1329 1330 >cat /proc/scsi/scsi 1331 Attached devices: 1332 Host: scsi0 Channel: 00 Id: 00 Lun: 00 1333 Vendor: IBM Model: DGHS09U Rev: 03E0 1334 Type: Direct-Access ANSI SCSI revision: 03 1335 Host: scsi0 Channel: 00 Id: 06 Lun: 00 1336 Vendor: PIONEER Model: CD-ROM DR-U06S Rev: 1.04 1337 Type: CD-ROM ANSI SCSI revision: 02 1338 1339 1340The directory named after the driver has one file for each adapter found in 1341the system. These files contain information about the controller, including 1342the used IRQ and the IO address range. The amount of information shown is 1343dependent on the adapter you use. The example shows the output for an Adaptec 1344AHA-2940 SCSI adapter:: 1345 1346 > cat /proc/scsi/aic7xxx/0 1347 1348 Adaptec AIC7xxx driver version: 5.1.19/3.2.4 1349 Compile Options: 1350 TCQ Enabled By Default : Disabled 1351 AIC7XXX_PROC_STATS : Disabled 1352 AIC7XXX_RESET_DELAY : 5 1353 Adapter Configuration: 1354 SCSI Adapter: Adaptec AHA-294X Ultra SCSI host adapter 1355 Ultra Wide Controller 1356 PCI MMAPed I/O Base: 0xeb001000 1357 Adapter SEEPROM Config: SEEPROM found and used. 1358 Adaptec SCSI BIOS: Enabled 1359 IRQ: 10 1360 SCBs: Active 0, Max Active 2, 1361 Allocated 15, HW 16, Page 255 1362 Interrupts: 160328 1363 BIOS Control Word: 0x18b6 1364 Adapter Control Word: 0x005b 1365 Extended Translation: Enabled 1366 Disconnect Enable Flags: 0xffff 1367 Ultra Enable Flags: 0x0001 1368 Tag Queue Enable Flags: 0x0000 1369 Ordered Queue Tag Flags: 0x0000 1370 Default Tag Queue Depth: 8 1371 Tagged Queue By Device array for aic7xxx host instance 0: 1372 {255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255} 1373 Actual queue depth per device for aic7xxx host instance 0: 1374 {1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1} 1375 Statistics: 1376 (scsi0:0:0:0) 1377 Device using Wide/Sync transfers at 40.0 MByte/sec, offset 8 1378 Transinfo settings: current(12/8/1/0), goal(12/8/1/0), user(12/15/1/0) 1379 Total transfers 160151 (74577 reads and 85574 writes) 1380 (scsi0:0:6:0) 1381 Device using Narrow/Sync transfers at 5.0 MByte/sec, offset 15 1382 Transinfo settings: current(50/15/0/0), goal(50/15/0/0), user(50/15/0/0) 1383 Total transfers 0 (0 reads and 0 writes) 1384 1385 13861.5 Parallel port info in /proc/parport 1387--------------------------------------- 1388 1389The directory /proc/parport contains information about the parallel ports of 1390your system. It has one subdirectory for each port, named after the port 1391number (0,1,2,...). 1392 1393These directories contain the four files shown in Table 1-10. 1394 1395 1396.. table:: Table 1-10: Files in /proc/parport 1397 1398 ========= ==================================================================== 1399 File Content 1400 ========= ==================================================================== 1401 autoprobe Any IEEE-1284 device ID information that has been acquired. 1402 devices list of the device drivers using that port. A + will appear by the 1403 name of the device currently using the port (it might not appear 1404 against any). 1405 hardware Parallel port's base address, IRQ line and DMA channel. 1406 irq IRQ that parport is using for that port. This is in a separate 1407 file to allow you to alter it by writing a new value in (IRQ 1408 number or none). 1409 ========= ==================================================================== 1410 14111.6 TTY info in /proc/tty 1412------------------------- 1413 1414Information about the available and actually used tty's can be found in the 1415directory /proc/tty. You'll find entries for drivers and line disciplines in 1416this directory, as shown in Table 1-11. 1417 1418 1419.. table:: Table 1-11: Files in /proc/tty 1420 1421 ============= ============================================== 1422 File Content 1423 ============= ============================================== 1424 drivers list of drivers and their usage 1425 ldiscs registered line disciplines 1426 driver/serial usage statistic and status of single tty lines 1427 ============= ============================================== 1428 1429To see which tty's are currently in use, you can simply look into the file 1430/proc/tty/drivers:: 1431 1432 > cat /proc/tty/drivers 1433 pty_slave /dev/pts 136 0-255 pty:slave 1434 pty_master /dev/ptm 128 0-255 pty:master 1435 pty_slave /dev/ttyp 3 0-255 pty:slave 1436 pty_master /dev/pty 2 0-255 pty:master 1437 serial /dev/cua 5 64-67 serial:callout 1438 serial /dev/ttyS 4 64-67 serial 1439 /dev/tty0 /dev/tty0 4 0 system:vtmaster 1440 /dev/ptmx /dev/ptmx 5 2 system 1441 /dev/console /dev/console 5 1 system:console 1442 /dev/tty /dev/tty 5 0 system:/dev/tty 1443 unknown /dev/tty 4 1-63 console 1444 1445 14461.7 Miscellaneous kernel statistics in /proc/stat 1447------------------------------------------------- 1448 1449Various pieces of information about kernel activity are available in the 1450/proc/stat file. All of the numbers reported in this file are aggregates 1451since the system first booted. For a quick look, simply cat the file:: 1452 1453 > cat /proc/stat 1454 cpu 2255 34 2290 22625563 6290 127 456 0 0 0 1455 cpu0 1132 34 1441 11311718 3675 127 438 0 0 0 1456 cpu1 1123 0 849 11313845 2614 0 18 0 0 0 1457 intr 114930548 113199788 3 0 5 263 0 4 [... lots more numbers ...] 1458 ctxt 1990473 1459 btime 1062191376 1460 processes 2915 1461 procs_running 1 1462 procs_blocked 0 1463 softirq 183433 0 21755 12 39 1137 231 21459 2263 1464 1465The very first "cpu" line aggregates the numbers in all of the other "cpuN" 1466lines. These numbers identify the amount of time the CPU has spent performing 1467different kinds of work. Time units are in USER_HZ (typically hundredths of a 1468second). The meanings of the columns are as follows, from left to right: 1469 1470- user: normal processes executing in user mode 1471- nice: niced processes executing in user mode 1472- system: processes executing in kernel mode 1473- idle: twiddling thumbs 1474- iowait: In a word, iowait stands for waiting for I/O to complete. But there 1475 are several problems: 1476 1477 1. CPU will not wait for I/O to complete, iowait is the time that a task is 1478 waiting for I/O to complete. When CPU goes into idle state for 1479 outstanding task I/O, another task will be scheduled on this CPU. 1480 2. In a multi-core CPU, the task waiting for I/O to complete is not running 1481 on any CPU, so the iowait of each CPU is difficult to calculate. 1482 3. The value of iowait field in /proc/stat will decrease in certain 1483 conditions. 1484 1485 So, the iowait is not reliable by reading from /proc/stat. 1486- irq: servicing interrupts 1487- softirq: servicing softirqs 1488- steal: involuntary wait 1489- guest: running a normal guest 1490- guest_nice: running a niced guest 1491 1492The "intr" line gives counts of interrupts serviced since boot time, for each 1493of the possible system interrupts. The first column is the total of all 1494interrupts serviced including unnumbered architecture specific interrupts; 1495each subsequent column is the total for that particular numbered interrupt. 1496Unnumbered interrupts are not shown, only summed into the total. 1497 1498The "ctxt" line gives the total number of context switches across all CPUs. 1499 1500The "btime" line gives the time at which the system booted, in seconds since 1501the Unix epoch. 1502 1503The "processes" line gives the number of processes and threads created, which 1504includes (but is not limited to) those created by calls to the fork() and 1505clone() system calls. 1506 1507The "procs_running" line gives the total number of threads that are 1508running or ready to run (i.e., the total number of runnable threads). 1509 1510The "procs_blocked" line gives the number of processes currently blocked, 1511waiting for I/O to complete. 1512 1513The "softirq" line gives counts of softirqs serviced since boot time, for each 1514of the possible system softirqs. The first column is the total of all 1515softirqs serviced; each subsequent column is the total for that particular 1516softirq. 1517 1518 15191.8 Ext4 file system parameters 1520------------------------------- 1521 1522Information about mounted ext4 file systems can be found in 1523/proc/fs/ext4. Each mounted filesystem will have a directory in 1524/proc/fs/ext4 based on its device name (i.e., /proc/fs/ext4/hdc or 1525/proc/fs/ext4/dm-0). The files in each per-device directory are shown 1526in Table 1-12, below. 1527 1528.. table:: Table 1-12: Files in /proc/fs/ext4/<devname> 1529 1530 ============== ========================================================== 1531 File Content 1532 mb_groups details of multiblock allocator buddy cache of free blocks 1533 ============== ========================================================== 1534 15351.9 /proc/consoles 1536------------------- 1537Shows registered system console lines. 1538 1539To see which character device lines are currently used for the system console 1540/dev/console, you may simply look into the file /proc/consoles:: 1541 1542 > cat /proc/consoles 1543 tty0 -WU (ECp) 4:7 1544 ttyS0 -W- (Ep) 4:64 1545 1546The columns are: 1547 1548+--------------------+-------------------------------------------------------+ 1549| device | name of the device | 1550+====================+=======================================================+ 1551| operations | * R = can do read operations | 1552| | * W = can do write operations | 1553| | * U = can do unblank | 1554+--------------------+-------------------------------------------------------+ 1555| flags | * E = it is enabled | 1556| | * C = it is preferred console | 1557| | * B = it is primary boot console | 1558| | * p = it is used for printk buffer | 1559| | * b = it is not a TTY but a Braille device | 1560| | * a = it is safe to use when cpu is offline | 1561+--------------------+-------------------------------------------------------+ 1562| major:minor | major and minor number of the device separated by a | 1563| | colon | 1564+--------------------+-------------------------------------------------------+ 1565 1566Summary 1567------- 1568 1569The /proc file system serves information about the running system. It not only 1570allows access to process data but also allows you to request the kernel status 1571by reading files in the hierarchy. 1572 1573The directory structure of /proc reflects the types of information and makes 1574it easy, if not obvious, where to look for specific data. 1575 1576Chapter 2: Modifying System Parameters 1577====================================== 1578 1579In This Chapter 1580--------------- 1581 1582* Modifying kernel parameters by writing into files found in /proc/sys 1583* Exploring the files which modify certain parameters 1584* Review of the /proc/sys file tree 1585 1586------------------------------------------------------------------------------ 1587 1588A very interesting part of /proc is the directory /proc/sys. This is not only 1589a source of information, it also allows you to change parameters within the 1590kernel. Be very careful when attempting this. You can optimize your system, 1591but you can also cause it to crash. Never alter kernel parameters on a 1592production system. Set up a development machine and test to make sure that 1593everything works the way you want it to. You may have no alternative but to 1594reboot the machine once an error has been made. 1595 1596To change a value, simply echo the new value into the file. 1597You need to be root to do this. You can create your own boot script 1598to perform this every time your system boots. 1599 1600The files in /proc/sys can be used to fine tune and monitor miscellaneous and 1601general things in the operation of the Linux kernel. Since some of the files 1602can inadvertently disrupt your system, it is advisable to read both 1603documentation and source before actually making adjustments. In any case, be 1604very careful when writing to any of these files. The entries in /proc may 1605change slightly between the 2.1.* and the 2.2 kernel, so if there is any doubt 1606review the kernel documentation in the directory /usr/src/linux/Documentation. 1607This chapter is heavily based on the documentation included in the pre 2.2 1608kernels, and became part of it in version 2.2.1 of the Linux kernel. 1609 1610Please see: Documentation/admin-guide/sysctl/ directory for descriptions of these 1611entries. 1612 1613Summary 1614------- 1615 1616Certain aspects of kernel behavior can be modified at runtime, without the 1617need to recompile the kernel, or even to reboot the system. The files in the 1618/proc/sys tree can not only be read, but also modified. You can use the echo 1619command to write value into these files, thereby changing the default settings 1620of the kernel. 1621 1622 1623Chapter 3: Per-process Parameters 1624================================= 1625 16263.1 /proc/<pid>/oom_adj & /proc/<pid>/oom_score_adj- Adjust the oom-killer score 1627-------------------------------------------------------------------------------- 1628 1629These files can be used to adjust the badness heuristic used to select which 1630process gets killed in out of memory (oom) conditions. 1631 1632The badness heuristic assigns a value to each candidate task ranging from 0 1633(never kill) to 1000 (always kill) to determine which process is targeted. The 1634units are roughly a proportion along that range of allowed memory the process 1635may allocate from based on an estimation of its current memory and swap use. 1636For example, if a task is using all allowed memory, its badness score will be 16371000. If it is using half of its allowed memory, its score will be 500. 1638 1639The amount of "allowed" memory depends on the context in which the oom killer 1640was called. If it is due to the memory assigned to the allocating task's cpuset 1641being exhausted, the allowed memory represents the set of mems assigned to that 1642cpuset. If it is due to a mempolicy's node(s) being exhausted, the allowed 1643memory represents the set of mempolicy nodes. If it is due to a memory 1644limit (or swap limit) being reached, the allowed memory is that configured 1645limit. Finally, if it is due to the entire system being out of memory, the 1646allowed memory represents all allocatable resources. 1647 1648The value of /proc/<pid>/oom_score_adj is added to the badness score before it 1649is used to determine which task to kill. Acceptable values range from -1000 1650(OOM_SCORE_ADJ_MIN) to +1000 (OOM_SCORE_ADJ_MAX). This allows userspace to 1651polarize the preference for oom killing either by always preferring a certain 1652task or completely disabling it. The lowest possible value, -1000, is 1653equivalent to disabling oom killing entirely for that task since it will always 1654report a badness score of 0. 1655 1656Consequently, it is very simple for userspace to define the amount of memory to 1657consider for each task. Setting a /proc/<pid>/oom_score_adj value of +500, for 1658example, is roughly equivalent to allowing the remainder of tasks sharing the 1659same system, cpuset, mempolicy, or memory controller resources to use at least 166050% more memory. A value of -500, on the other hand, would be roughly 1661equivalent to discounting 50% of the task's allowed memory from being considered 1662as scoring against the task. 1663 1664For backwards compatibility with previous kernels, /proc/<pid>/oom_adj may also 1665be used to tune the badness score. Its acceptable values range from -16 1666(OOM_ADJUST_MIN) to +15 (OOM_ADJUST_MAX) and a special value of -17 1667(OOM_DISABLE) to disable oom killing entirely for that task. Its value is 1668scaled linearly with /proc/<pid>/oom_score_adj. 1669 1670The value of /proc/<pid>/oom_score_adj may be reduced no lower than the last 1671value set by a CAP_SYS_RESOURCE process. To reduce the value any lower 1672requires CAP_SYS_RESOURCE. 1673 1674 16753.2 /proc/<pid>/oom_score - Display current oom-killer score 1676------------------------------------------------------------- 1677 1678This file can be used to check the current score used by the oom-killer for 1679any given <pid>. Use it together with /proc/<pid>/oom_score_adj to tune which 1680process should be killed in an out-of-memory situation. 1681 1682Please note that the exported value includes oom_score_adj so it is 1683effectively in range [0,2000]. 1684 1685 16863.3 /proc/<pid>/io - Display the IO accounting fields 1687------------------------------------------------------- 1688 1689This file contains IO statistics for each running process. 1690 1691Example 1692~~~~~~~ 1693 1694:: 1695 1696 test:/tmp # dd if=/dev/zero of=/tmp/test.dat & 1697 [1] 3828 1698 1699 test:/tmp # cat /proc/3828/io 1700 rchar: 323934931 1701 wchar: 323929600 1702 syscr: 632687 1703 syscw: 632675 1704 read_bytes: 0 1705 write_bytes: 323932160 1706 cancelled_write_bytes: 0 1707 1708 1709Description 1710~~~~~~~~~~~ 1711 1712rchar 1713^^^^^ 1714 1715I/O counter: chars read 1716The number of bytes which this task has caused to be read from storage. This 1717is simply the sum of bytes which this process passed to read() and pread(). 1718It includes things like tty IO and it is unaffected by whether or not actual 1719physical disk IO was required (the read might have been satisfied from 1720pagecache). 1721 1722 1723wchar 1724^^^^^ 1725 1726I/O counter: chars written 1727The number of bytes which this task has caused, or shall cause to be written 1728to disk. Similar caveats apply here as with rchar. 1729 1730 1731syscr 1732^^^^^ 1733 1734I/O counter: read syscalls 1735Attempt to count the number of read I/O operations, i.e. syscalls like read() 1736and pread(). 1737 1738 1739syscw 1740^^^^^ 1741 1742I/O counter: write syscalls 1743Attempt to count the number of write I/O operations, i.e. syscalls like 1744write() and pwrite(). 1745 1746 1747read_bytes 1748^^^^^^^^^^ 1749 1750I/O counter: bytes read 1751Attempt to count the number of bytes which this process really did cause to 1752be fetched from the storage layer. Done at the submit_bio() level, so it is 1753accurate for block-backed filesystems. <please add status regarding NFS and 1754CIFS at a later time> 1755 1756 1757write_bytes 1758^^^^^^^^^^^ 1759 1760I/O counter: bytes written 1761Attempt to count the number of bytes which this process caused to be sent to 1762the storage layer. This is done at page-dirtying time. 1763 1764 1765cancelled_write_bytes 1766^^^^^^^^^^^^^^^^^^^^^ 1767 1768The big inaccuracy here is truncate. If a process writes 1MB to a file and 1769then deletes the file, it will in fact perform no writeout. But it will have 1770been accounted as having caused 1MB of write. 1771In other words: The number of bytes which this process caused to not happen, 1772by truncating pagecache. A task can cause "negative" IO too. If this task 1773truncates some dirty pagecache, some IO which another task has been accounted 1774for (in its write_bytes) will not be happening. We _could_ just subtract that 1775from the truncating task's write_bytes, but there is information loss in doing 1776that. 1777 1778 1779.. Note:: 1780 1781 At its current implementation state, this is a bit racy on 32-bit machines: 1782 if process A reads process B's /proc/pid/io while process B is updating one 1783 of those 64-bit counters, process A could see an intermediate result. 1784 1785 1786More information about this can be found within the taskstats documentation in 1787Documentation/accounting. 1788 17893.4 /proc/<pid>/coredump_filter - Core dump filtering settings 1790--------------------------------------------------------------- 1791When a process is dumped, all anonymous memory is written to a core file as 1792long as the size of the core file isn't limited. But sometimes we don't want 1793to dump some memory segments, for example, huge shared memory or DAX. 1794Conversely, sometimes we want to save file-backed memory segments into a core 1795file, not only the individual files. 1796 1797/proc/<pid>/coredump_filter allows you to customize which memory segments 1798will be dumped when the <pid> process is dumped. coredump_filter is a bitmask 1799of memory types. If a bit of the bitmask is set, memory segments of the 1800corresponding memory type are dumped, otherwise they are not dumped. 1801 1802The following 9 memory types are supported: 1803 1804 - (bit 0) anonymous private memory 1805 - (bit 1) anonymous shared memory 1806 - (bit 2) file-backed private memory 1807 - (bit 3) file-backed shared memory 1808 - (bit 4) ELF header pages in file-backed private memory areas (it is 1809 effective only if the bit 2 is cleared) 1810 - (bit 5) hugetlb private memory 1811 - (bit 6) hugetlb shared memory 1812 - (bit 7) DAX private memory 1813 - (bit 8) DAX shared memory 1814 1815 Note that MMIO pages such as frame buffer are never dumped and vDSO pages 1816 are always dumped regardless of the bitmask status. 1817 1818 Note that bits 0-4 don't affect hugetlb or DAX memory. hugetlb memory is 1819 only affected by bit 5-6, and DAX is only affected by bits 7-8. 1820 1821The default value of coredump_filter is 0x33; this means all anonymous memory 1822segments, ELF header pages and hugetlb private memory are dumped. 1823 1824If you don't want to dump all shared memory segments attached to pid 1234, 1825write 0x31 to the process's proc file:: 1826 1827 $ echo 0x31 > /proc/1234/coredump_filter 1828 1829When a new process is created, the process inherits the bitmask status from its 1830parent. It is useful to set up coredump_filter before the program runs. 1831For example:: 1832 1833 $ echo 0x7 > /proc/self/coredump_filter 1834 $ ./some_program 1835 18363.5 /proc/<pid>/mountinfo - Information about mounts 1837-------------------------------------------------------- 1838 1839This file contains lines of the form:: 1840 1841 36 35 98:0 /mnt1 /mnt2 rw,noatime master:1 - ext3 /dev/root rw,errors=continue 1842 (1)(2)(3) (4) (5) (6) (n…m) (m+1)(m+2) (m+3) (m+4) 1843 1844 (1) mount ID: unique identifier of the mount (may be reused after umount) 1845 (2) parent ID: ID of parent (or of self for the top of the mount tree) 1846 (3) major:minor: value of st_dev for files on filesystem 1847 (4) root: root of the mount within the filesystem 1848 (5) mount point: mount point relative to the process's root 1849 (6) mount options: per mount options 1850 (n…m) optional fields: zero or more fields of the form "tag[:value]" 1851 (m+1) separator: marks the end of the optional fields 1852 (m+2) filesystem type: name of filesystem of the form "type[.subtype]" 1853 (m+3) mount source: filesystem specific information or "none" 1854 (m+4) super options: per super block options 1855 1856Parsers should ignore all unrecognised optional fields. Currently the 1857possible optional fields are: 1858 1859================ ============================================================== 1860shared:X mount is shared in peer group X 1861master:X mount is slave to peer group X 1862propagate_from:X mount is slave and receives propagation from peer group X [#]_ 1863unbindable mount is unbindable 1864================ ============================================================== 1865 1866.. [#] X is the closest dominant peer group under the process's root. If 1867 X is the immediate master of the mount, or if there's no dominant peer 1868 group under the same root, then only the "master:X" field is present 1869 and not the "propagate_from:X" field. 1870 1871For more information on mount propagation see: 1872 1873 Documentation/filesystems/sharedsubtree.rst 1874 1875 18763.6 /proc/<pid>/comm & /proc/<pid>/task/<tid>/comm 1877-------------------------------------------------------- 1878These files provide a method to access a task's comm value. It also allows for 1879a task to set its own or one of its thread siblings comm value. The comm value 1880is limited in size compared to the cmdline value, so writing anything longer 1881then the kernel's TASK_COMM_LEN (currently 16 chars) will result in a truncated 1882comm value. 1883 1884 18853.7 /proc/<pid>/task/<tid>/children - Information about task children 1886------------------------------------------------------------------------- 1887This file provides a fast way to retrieve first level children pids 1888of a task pointed by <pid>/<tid> pair. The format is a space separated 1889stream of pids. 1890 1891Note the "first level" here -- if a child has its own children they will 1892not be listed here; one needs to read /proc/<children-pid>/task/<tid>/children 1893to obtain the descendants. 1894 1895Since this interface is intended to be fast and cheap it doesn't 1896guarantee to provide precise results and some children might be 1897skipped, especially if they've exited right after we printed their 1898pids, so one needs to either stop or freeze processes being inspected 1899if precise results are needed. 1900 1901 19023.8 /proc/<pid>/fdinfo/<fd> - Information about opened file 1903--------------------------------------------------------------- 1904This file provides information associated with an opened file. The regular 1905files have at least four fields -- 'pos', 'flags', 'mnt_id' and 'ino'. 1906The 'pos' represents the current offset of the opened file in decimal 1907form [see lseek(2) for details], 'flags' denotes the octal O_xxx mask the 1908file has been created with [see open(2) for details] and 'mnt_id' represents 1909mount ID of the file system containing the opened file [see 3.5 1910/proc/<pid>/mountinfo for details]. 'ino' represents the inode number of 1911the file. 1912 1913A typical output is:: 1914 1915 pos: 0 1916 flags: 0100002 1917 mnt_id: 19 1918 ino: 63107 1919 1920All locks associated with a file descriptor are shown in its fdinfo too:: 1921 1922 lock: 1: FLOCK ADVISORY WRITE 359 00:13:11691 0 EOF 1923 1924The files such as eventfd, fsnotify, signalfd, epoll among the regular pos/flags 1925pair provide additional information particular to the objects they represent. 1926 1927Eventfd files 1928~~~~~~~~~~~~~ 1929 1930:: 1931 1932 pos: 0 1933 flags: 04002 1934 mnt_id: 9 1935 ino: 63107 1936 eventfd-count: 5a 1937 1938where 'eventfd-count' is hex value of a counter. 1939 1940Signalfd files 1941~~~~~~~~~~~~~~ 1942 1943:: 1944 1945 pos: 0 1946 flags: 04002 1947 mnt_id: 9 1948 ino: 63107 1949 sigmask: 0000000000000200 1950 1951where 'sigmask' is hex value of the signal mask associated 1952with a file. 1953 1954Epoll files 1955~~~~~~~~~~~ 1956 1957:: 1958 1959 pos: 0 1960 flags: 02 1961 mnt_id: 9 1962 ino: 63107 1963 tfd: 5 events: 1d data: ffffffffffffffff pos:0 ino:61af sdev:7 1964 1965where 'tfd' is a target file descriptor number in decimal form, 1966'events' is events mask being watched and the 'data' is data 1967associated with a target [see epoll(7) for more details]. 1968 1969The 'pos' is current offset of the target file in decimal form 1970[see lseek(2)], 'ino' and 'sdev' are inode and device numbers 1971where target file resides, all in hex format. 1972 1973Fsnotify files 1974~~~~~~~~~~~~~~ 1975For inotify files the format is the following:: 1976 1977 pos: 0 1978 flags: 02000000 1979 mnt_id: 9 1980 ino: 63107 1981 inotify wd:3 ino:9e7e sdev:800013 mask:800afce ignored_mask:0 fhandle-bytes:8 fhandle-type:1 f_handle:7e9e0000640d1b6d 1982 1983where 'wd' is a watch descriptor in decimal form, i.e. a target file 1984descriptor number, 'ino' and 'sdev' are inode and device where the 1985target file resides and the 'mask' is the mask of events, all in hex 1986form [see inotify(7) for more details]. 1987 1988If the kernel was built with exportfs support, the path to the target 1989file is encoded as a file handle. The file handle is provided by three 1990fields 'fhandle-bytes', 'fhandle-type' and 'f_handle', all in hex 1991format. 1992 1993If the kernel is built without exportfs support the file handle won't be 1994printed out. 1995 1996If there is no inotify mark attached yet the 'inotify' line will be omitted. 1997 1998For fanotify files the format is:: 1999 2000 pos: 0 2001 flags: 02 2002 mnt_id: 9 2003 ino: 63107 2004 fanotify flags:10 event-flags:0 2005 fanotify mnt_id:12 mflags:40 mask:38 ignored_mask:40000003 2006 fanotify ino:4f969 sdev:800013 mflags:0 mask:3b ignored_mask:40000000 fhandle-bytes:8 fhandle-type:1 f_handle:69f90400c275b5b4 2007 2008where fanotify 'flags' and 'event-flags' are values used in fanotify_init 2009call, 'mnt_id' is the mount point identifier, 'mflags' is the value of 2010flags associated with mark which are tracked separately from events 2011mask. 'ino' and 'sdev' are target inode and device, 'mask' is the events 2012mask and 'ignored_mask' is the mask of events which are to be ignored. 2013All are in hex format. Incorporation of 'mflags', 'mask' and 'ignored_mask' 2014provide information about flags and mask used in fanotify_mark 2015call [see fsnotify manpage for details]. 2016 2017While the first three lines are mandatory and always printed, the rest is 2018optional and may be omitted if no marks created yet. 2019 2020Timerfd files 2021~~~~~~~~~~~~~ 2022 2023:: 2024 2025 pos: 0 2026 flags: 02 2027 mnt_id: 9 2028 ino: 63107 2029 clockid: 0 2030 ticks: 0 2031 settime flags: 01 2032 it_value: (0, 49406829) 2033 it_interval: (1, 0) 2034 2035where 'clockid' is the clock type and 'ticks' is the number of the timer expirations 2036that have occurred [see timerfd_create(2) for details]. 'settime flags' are 2037flags in octal form been used to setup the timer [see timerfd_settime(2) for 2038details]. 'it_value' is remaining time until the timer expiration. 2039'it_interval' is the interval for the timer. Note the timer might be set up 2040with TIMER_ABSTIME option which will be shown in 'settime flags', but 'it_value' 2041still exhibits timer's remaining time. 2042 2043DMA Buffer files 2044~~~~~~~~~~~~~~~~ 2045 2046:: 2047 2048 pos: 0 2049 flags: 04002 2050 mnt_id: 9 2051 ino: 63107 2052 size: 32768 2053 count: 2 2054 exp_name: system-heap 2055 2056where 'size' is the size of the DMA buffer in bytes. 'count' is the file count of 2057the DMA buffer file. 'exp_name' is the name of the DMA buffer exporter. 2058 20593.9 /proc/<pid>/map_files - Information about memory mapped files 2060--------------------------------------------------------------------- 2061This directory contains symbolic links which represent memory mapped files 2062the process is maintaining. Example output:: 2063 2064 | lr-------- 1 root root 64 Jan 27 11:24 333c600000-333c620000 -> /usr/lib64/ld-2.18.so 2065 | lr-------- 1 root root 64 Jan 27 11:24 333c81f000-333c820000 -> /usr/lib64/ld-2.18.so 2066 | lr-------- 1 root root 64 Jan 27 11:24 333c820000-333c821000 -> /usr/lib64/ld-2.18.so 2067 | ... 2068 | lr-------- 1 root root 64 Jan 27 11:24 35d0421000-35d0422000 -> /usr/lib64/libselinux.so.1 2069 | lr-------- 1 root root 64 Jan 27 11:24 400000-41a000 -> /usr/bin/ls 2070 2071The name of a link represents the virtual memory bounds of a mapping, i.e. 2072vm_area_struct::vm_start-vm_area_struct::vm_end. 2073 2074The main purpose of the map_files is to retrieve a set of memory mapped 2075files in a fast way instead of parsing /proc/<pid>/maps or 2076/proc/<pid>/smaps, both of which contain many more records. At the same 2077time one can open(2) mappings from the listings of two processes and 2078comparing their inode numbers to figure out which anonymous memory areas 2079are actually shared. 2080 20813.10 /proc/<pid>/timerslack_ns - Task timerslack value 2082--------------------------------------------------------- 2083This file provides the value of the task's timerslack value in nanoseconds. 2084This value specifies an amount of time that normal timers may be deferred 2085in order to coalesce timers and avoid unnecessary wakeups. 2086 2087This allows a task's interactivity vs power consumption tradeoff to be 2088adjusted. 2089 2090Writing 0 to the file will set the task's timerslack to the default value. 2091 2092Valid values are from 0 - ULLONG_MAX 2093 2094An application setting the value must have PTRACE_MODE_ATTACH_FSCREDS level 2095permissions on the task specified to change its timerslack_ns value. 2096 20973.11 /proc/<pid>/patch_state - Livepatch patch operation state 2098----------------------------------------------------------------- 2099When CONFIG_LIVEPATCH is enabled, this file displays the value of the 2100patch state for the task. 2101 2102A value of '-1' indicates that no patch is in transition. 2103 2104A value of '0' indicates that a patch is in transition and the task is 2105unpatched. If the patch is being enabled, then the task hasn't been 2106patched yet. If the patch is being disabled, then the task has already 2107been unpatched. 2108 2109A value of '1' indicates that a patch is in transition and the task is 2110patched. If the patch is being enabled, then the task has already been 2111patched. If the patch is being disabled, then the task hasn't been 2112unpatched yet. 2113 21143.12 /proc/<pid>/arch_status - task architecture specific status 2115------------------------------------------------------------------- 2116When CONFIG_PROC_PID_ARCH_STATUS is enabled, this file displays the 2117architecture specific status of the task. 2118 2119Example 2120~~~~~~~ 2121 2122:: 2123 2124 $ cat /proc/6753/arch_status 2125 AVX512_elapsed_ms: 8 2126 2127Description 2128~~~~~~~~~~~ 2129 2130x86 specific entries 2131~~~~~~~~~~~~~~~~~~~~~ 2132 2133AVX512_elapsed_ms 2134^^^^^^^^^^^^^^^^^^ 2135 2136 If AVX512 is supported on the machine, this entry shows the milliseconds 2137 elapsed since the last time AVX512 usage was recorded. The recording 2138 happens on a best effort basis when a task is scheduled out. This means 2139 that the value depends on two factors: 2140 2141 1) The time which the task spent on the CPU without being scheduled 2142 out. With CPU isolation and a single runnable task this can take 2143 several seconds. 2144 2145 2) The time since the task was scheduled out last. Depending on the 2146 reason for being scheduled out (time slice exhausted, syscall ...) 2147 this can be arbitrary long time. 2148 2149 As a consequence the value cannot be considered precise and authoritative 2150 information. The application which uses this information has to be aware 2151 of the overall scenario on the system in order to determine whether a 2152 task is a real AVX512 user or not. Precise information can be obtained 2153 with performance counters. 2154 2155 A special value of '-1' indicates that no AVX512 usage was recorded, thus 2156 the task is unlikely an AVX512 user, but depends on the workload and the 2157 scheduling scenario, it also could be a false negative mentioned above. 2158 21593.13 /proc/<pid>/fd - List of symlinks to open files 2160------------------------------------------------------- 2161This directory contains symbolic links which represent open files 2162the process is maintaining. Example output:: 2163 2164 lr-x------ 1 root root 64 Sep 20 17:53 0 -> /dev/null 2165 l-wx------ 1 root root 64 Sep 20 17:53 1 -> /dev/null 2166 lrwx------ 1 root root 64 Sep 20 17:53 10 -> 'socket:[12539]' 2167 lrwx------ 1 root root 64 Sep 20 17:53 11 -> 'socket:[12540]' 2168 lrwx------ 1 root root 64 Sep 20 17:53 12 -> 'socket:[12542]' 2169 2170The number of open files for the process is stored in 'size' member 2171of stat() output for /proc/<pid>/fd for fast access. 2172------------------------------------------------------- 2173 2174 2175Chapter 4: Configuring procfs 2176============================= 2177 21784.1 Mount options 2179--------------------- 2180 2181The following mount options are supported: 2182 2183 ========= ======================================================== 2184 hidepid= Set /proc/<pid>/ access mode. 2185 gid= Set the group authorized to learn processes information. 2186 subset= Show only the specified subset of procfs. 2187 ========= ======================================================== 2188 2189hidepid=off or hidepid=0 means classic mode - everybody may access all 2190/proc/<pid>/ directories (default). 2191 2192hidepid=noaccess or hidepid=1 means users may not access any /proc/<pid>/ 2193directories but their own. Sensitive files like cmdline, sched*, status are now 2194protected against other users. This makes it impossible to learn whether any 2195user runs specific program (given the program doesn't reveal itself by its 2196behaviour). As an additional bonus, as /proc/<pid>/cmdline is unaccessible for 2197other users, poorly written programs passing sensitive information via program 2198arguments are now protected against local eavesdroppers. 2199 2200hidepid=invisible or hidepid=2 means hidepid=1 plus all /proc/<pid>/ will be 2201fully invisible to other users. It doesn't mean that it hides a fact whether a 2202process with a specific pid value exists (it can be learned by other means, e.g. 2203by "kill -0 $PID"), but it hides process' uid and gid, which may be learned by 2204stat()'ing /proc/<pid>/ otherwise. It greatly complicates an intruder's task of 2205gathering information about running processes, whether some daemon runs with 2206elevated privileges, whether other user runs some sensitive program, whether 2207other users run any program at all, etc. 2208 2209hidepid=ptraceable or hidepid=4 means that procfs should only contain 2210/proc/<pid>/ directories that the caller can ptrace. 2211 2212gid= defines a group authorized to learn processes information otherwise 2213prohibited by hidepid=. If you use some daemon like identd which needs to learn 2214information about processes information, just add identd to this group. 2215 2216subset=pid hides all top level files and directories in the procfs that 2217are not related to tasks. 2218 2219Chapter 5: Filesystem behavior 2220============================== 2221 2222Originally, before the advent of pid namepsace, procfs was a global file 2223system. It means that there was only one procfs instance in the system. 2224 2225When pid namespace was added, a separate procfs instance was mounted in 2226each pid namespace. So, procfs mount options are global among all 2227mountpoints within the same namespace:: 2228 2229 # grep ^proc /proc/mounts 2230 proc /proc proc rw,relatime,hidepid=2 0 0 2231 2232 # strace -e mount mount -o hidepid=1 -t proc proc /tmp/proc 2233 mount("proc", "/tmp/proc", "proc", 0, "hidepid=1") = 0 2234 +++ exited with 0 +++ 2235 2236 # grep ^proc /proc/mounts 2237 proc /proc proc rw,relatime,hidepid=2 0 0 2238 proc /tmp/proc proc rw,relatime,hidepid=2 0 0 2239 2240and only after remounting procfs mount options will change at all 2241mountpoints:: 2242 2243 # mount -o remount,hidepid=1 -t proc proc /tmp/proc 2244 2245 # grep ^proc /proc/mounts 2246 proc /proc proc rw,relatime,hidepid=1 0 0 2247 proc /tmp/proc proc rw,relatime,hidepid=1 0 0 2248 2249This behavior is different from the behavior of other filesystems. 2250 2251The new procfs behavior is more like other filesystems. Each procfs mount 2252creates a new procfs instance. Mount options affect own procfs instance. 2253It means that it became possible to have several procfs instances 2254displaying tasks with different filtering options in one pid namespace:: 2255 2256 # mount -o hidepid=invisible -t proc proc /proc 2257 # mount -o hidepid=noaccess -t proc proc /tmp/proc 2258 # grep ^proc /proc/mounts 2259 proc /proc proc rw,relatime,hidepid=invisible 0 0 2260 proc /tmp/proc proc rw,relatime,hidepid=noaccess 0 0 2261