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 51 4 Configuring procfs 52 4.1 Mount options 53 54 5 Filesystem behavior 55 56Preface 57======= 58 590.1 Introduction/Credits 60------------------------ 61 62This documentation is part of a soon (or so we hope) to be released book on 63the SuSE Linux distribution. As there is no complete documentation for the 64/proc file system and we've used many freely available sources to write these 65chapters, it seems only fair to give the work back to the Linux community. 66This work is based on the 2.2.* kernel version and the upcoming 2.4.*. I'm 67afraid it's still far from complete, but we hope it will be useful. As far as 68we know, it is the first 'all-in-one' document about the /proc file system. It 69is focused on the Intel x86 hardware, so if you are looking for PPC, ARM, 70SPARC, AXP, etc., features, you probably won't find what you are looking for. 71It also only covers IPv4 networking, not IPv6 nor other protocols - sorry. But 72additions and patches are welcome and will be added to this document if you 73mail them to Bodo. 74 75We'd like to thank Alan Cox, Rik van Riel, and Alexey Kuznetsov and a lot of 76other people for help compiling this documentation. We'd also like to extend a 77special thank you to Andi Kleen for documentation, which we relied on heavily 78to create this document, as well as the additional information he provided. 79Thanks to everybody else who contributed source or docs to the Linux kernel 80and helped create a great piece of software... :) 81 82If you have any comments, corrections or additions, please don't hesitate to 83contact Bodo Bauer at bb@ricochet.net. We'll be happy to add them to this 84document. 85 86The latest version of this document is available online at 87http://tldp.org/LDP/Linux-Filesystem-Hierarchy/html/proc.html 88 89If the above direction does not works for you, you could try the kernel 90mailing list at linux-kernel@vger.kernel.org and/or try to reach me at 91comandante@zaralinux.com. 92 930.2 Legal Stuff 94--------------- 95 96We don't guarantee the correctness of this document, and if you come to us 97complaining about how you screwed up your system because of incorrect 98documentation, we won't feel responsible... 99 100Chapter 1: Collecting System Information 101======================================== 102 103In This Chapter 104--------------- 105* Investigating the properties of the pseudo file system /proc and its 106 ability to provide information on the running Linux system 107* Examining /proc's structure 108* Uncovering various information about the kernel and the processes running 109 on the system 110 111------------------------------------------------------------------------------ 112 113The proc file system acts as an interface to internal data structures in the 114kernel. It can be used to obtain information about the system and to change 115certain kernel parameters at runtime (sysctl). 116 117First, we'll take a look at the read-only parts of /proc. In Chapter 2, we 118show you how you can use /proc/sys to change settings. 119 1201.1 Process-Specific Subdirectories 121----------------------------------- 122 123The directory /proc contains (among other things) one subdirectory for each 124process running on the system, which is named after the process ID (PID). 125 126The link 'self' points to the process reading the file system. Each process 127subdirectory has the entries listed in Table 1-1. 128 129Note that an open file descriptor to /proc/<pid> or to any of its 130contained files or subdirectories does not prevent <pid> being reused 131for some other process in the event that <pid> exits. Operations on 132open /proc/<pid> file descriptors corresponding to dead processes 133never act on any new process that the kernel may, through chance, have 134also assigned the process ID <pid>. Instead, operations on these FDs 135usually fail with ESRCH. 136 137.. table:: Table 1-1: Process specific entries in /proc 138 139 ============= =============================================================== 140 File Content 141 ============= =============================================================== 142 clear_refs Clears page referenced bits shown in smaps output 143 cmdline Command line arguments 144 cpu Current and last cpu in which it was executed (2.4)(smp) 145 cwd Link to the current working directory 146 environ Values of environment variables 147 exe Link to the executable of this process 148 fd Directory, which contains all file descriptors 149 maps Memory maps to executables and library files (2.4) 150 mem Memory held by this process 151 root Link to the root directory of this process 152 stat Process status 153 statm Process memory status information 154 status Process status in human readable form 155 wchan Present with CONFIG_KALLSYMS=y: it shows the kernel function 156 symbol the task is blocked in - or "0" if not blocked. 157 pagemap Page table 158 stack Report full stack trace, enable via CONFIG_STACKTRACE 159 smaps An extension based on maps, showing the memory consumption of 160 each mapping and flags associated with it 161 smaps_rollup Accumulated smaps stats for all mappings of the process. This 162 can be derived from smaps, but is faster and more convenient 163 numa_maps An extension based on maps, showing the memory locality and 164 binding policy as well as mem usage (in pages) of each mapping. 165 ============= =============================================================== 166 167For example, to get the status information of a process, all you have to do is 168read the file /proc/PID/status:: 169 170 >cat /proc/self/status 171 Name: cat 172 State: R (running) 173 Tgid: 5452 174 Pid: 5452 175 PPid: 743 176 TracerPid: 0 (2.4) 177 Uid: 501 501 501 501 178 Gid: 100 100 100 100 179 FDSize: 256 180 Groups: 100 14 16 181 VmPeak: 5004 kB 182 VmSize: 5004 kB 183 VmLck: 0 kB 184 VmHWM: 476 kB 185 VmRSS: 476 kB 186 RssAnon: 352 kB 187 RssFile: 120 kB 188 RssShmem: 4 kB 189 VmData: 156 kB 190 VmStk: 88 kB 191 VmExe: 68 kB 192 VmLib: 1412 kB 193 VmPTE: 20 kb 194 VmSwap: 0 kB 195 HugetlbPages: 0 kB 196 CoreDumping: 0 197 THP_enabled: 1 198 Threads: 1 199 SigQ: 0/28578 200 SigPnd: 0000000000000000 201 ShdPnd: 0000000000000000 202 SigBlk: 0000000000000000 203 SigIgn: 0000000000000000 204 SigCgt: 0000000000000000 205 CapInh: 00000000fffffeff 206 CapPrm: 0000000000000000 207 CapEff: 0000000000000000 208 CapBnd: ffffffffffffffff 209 CapAmb: 0000000000000000 210 NoNewPrivs: 0 211 Seccomp: 0 212 Speculation_Store_Bypass: thread vulnerable 213 SpeculationIndirectBranch: conditional enabled 214 voluntary_ctxt_switches: 0 215 nonvoluntary_ctxt_switches: 1 216 217This shows you nearly the same information you would get if you viewed it with 218the ps command. In fact, ps uses the proc file system to obtain its 219information. But you get a more detailed view of the process by reading the 220file /proc/PID/status. It fields are described in table 1-2. 221 222The statm file contains more detailed information about the process 223memory usage. Its seven fields are explained in Table 1-3. The stat file 224contains detailed information about the process itself. Its fields are 225explained in Table 1-4. 226 227(for SMP CONFIG users) 228 229For making accounting scalable, RSS related information are handled in an 230asynchronous manner and the value may not be very precise. To see a precise 231snapshot of a moment, you can see /proc/<pid>/smaps file and scan page table. 232It's slow but very precise. 233 234.. table:: Table 1-2: Contents of the status files (as of 4.19) 235 236 ========================== =================================================== 237 Field Content 238 ========================== =================================================== 239 Name filename of the executable 240 Umask file mode creation mask 241 State state (R is running, S is sleeping, D is sleeping 242 in an uninterruptible wait, Z is zombie, 243 T is traced or stopped) 244 Tgid thread group ID 245 Ngid NUMA group ID (0 if none) 246 Pid process id 247 PPid process id of the parent process 248 TracerPid PID of process tracing this process (0 if not) 249 Uid Real, effective, saved set, and file system UIDs 250 Gid Real, effective, saved set, and file system GIDs 251 FDSize number of file descriptor slots currently allocated 252 Groups supplementary group list 253 NStgid descendant namespace thread group ID hierarchy 254 NSpid descendant namespace process ID hierarchy 255 NSpgid descendant namespace process group ID hierarchy 256 NSsid descendant namespace session ID hierarchy 257 VmPeak peak virtual memory size 258 VmSize total program size 259 VmLck locked memory size 260 VmPin pinned memory size 261 VmHWM peak resident set size ("high water mark") 262 VmRSS size of memory portions. It contains the three 263 following parts 264 (VmRSS = RssAnon + RssFile + RssShmem) 265 RssAnon size of resident anonymous memory 266 RssFile size of resident file mappings 267 RssShmem size of resident shmem memory (includes SysV shm, 268 mapping of tmpfs and shared anonymous mappings) 269 VmData size of private data segments 270 VmStk size of stack segments 271 VmExe size of text segment 272 VmLib size of shared library code 273 VmPTE size of page table entries 274 VmSwap amount of swap used by anonymous private data 275 (shmem swap usage is not included) 276 HugetlbPages size of hugetlb memory portions 277 CoreDumping process's memory is currently being dumped 278 (killing the process may lead to a corrupted core) 279 THP_enabled process is allowed to use THP (returns 0 when 280 PR_SET_THP_DISABLE is set on the process 281 Threads number of threads 282 SigQ number of signals queued/max. number for queue 283 SigPnd bitmap of pending signals for the thread 284 ShdPnd bitmap of shared pending signals for the process 285 SigBlk bitmap of blocked signals 286 SigIgn bitmap of ignored signals 287 SigCgt bitmap of caught signals 288 CapInh bitmap of inheritable capabilities 289 CapPrm bitmap of permitted capabilities 290 CapEff bitmap of effective capabilities 291 CapBnd bitmap of capabilities bounding set 292 CapAmb bitmap of ambient capabilities 293 NoNewPrivs no_new_privs, like prctl(PR_GET_NO_NEW_PRIV, ...) 294 Seccomp seccomp mode, like prctl(PR_GET_SECCOMP, ...) 295 Speculation_Store_Bypass speculative store bypass mitigation status 296 SpeculationIndirectBranch indirect branch speculation mode 297 Cpus_allowed mask of CPUs on which this process may run 298 Cpus_allowed_list Same as previous, but in "list format" 299 Mems_allowed mask of memory nodes allowed to this process 300 Mems_allowed_list Same as previous, but in "list format" 301 voluntary_ctxt_switches number of voluntary context switches 302 nonvoluntary_ctxt_switches number of non voluntary context switches 303 ========================== =================================================== 304 305 306.. table:: Table 1-3: Contents of the statm files (as of 2.6.8-rc3) 307 308 ======== =============================== ============================== 309 Field Content 310 ======== =============================== ============================== 311 size total program size (pages) (same as VmSize in status) 312 resident size of memory portions (pages) (same as VmRSS in status) 313 shared number of pages that are shared (i.e. backed by a file, same 314 as RssFile+RssShmem in status) 315 trs number of pages that are 'code' (not including libs; broken, 316 includes data segment) 317 lrs number of pages of library (always 0 on 2.6) 318 drs number of pages of data/stack (including libs; broken, 319 includes library text) 320 dt number of dirty pages (always 0 on 2.6) 321 ======== =============================== ============================== 322 323 324.. table:: Table 1-4: Contents of the stat files (as of 2.6.30-rc7) 325 326 ============= =============================================================== 327 Field Content 328 ============= =============================================================== 329 pid process id 330 tcomm filename of the executable 331 state state (R is running, S is sleeping, D is sleeping in an 332 uninterruptible wait, Z is zombie, T is traced or stopped) 333 ppid process id of the parent process 334 pgrp pgrp of the process 335 sid session id 336 tty_nr tty the process uses 337 tty_pgrp pgrp of the tty 338 flags task flags 339 min_flt number of minor faults 340 cmin_flt number of minor faults with child's 341 maj_flt number of major faults 342 cmaj_flt number of major faults with child's 343 utime user mode jiffies 344 stime kernel mode jiffies 345 cutime user mode jiffies with child's 346 cstime kernel mode jiffies with child's 347 priority priority level 348 nice nice level 349 num_threads number of threads 350 it_real_value (obsolete, always 0) 351 start_time time the process started after system boot 352 vsize virtual memory size 353 rss resident set memory size 354 rsslim current limit in bytes on the rss 355 start_code address above which program text can run 356 end_code address below which program text can run 357 start_stack address of the start of the main process stack 358 esp current value of ESP 359 eip current value of EIP 360 pending bitmap of pending signals 361 blocked bitmap of blocked signals 362 sigign bitmap of ignored signals 363 sigcatch bitmap of caught signals 364 0 (place holder, used to be the wchan address, 365 use /proc/PID/wchan instead) 366 0 (place holder) 367 0 (place holder) 368 exit_signal signal to send to parent thread on exit 369 task_cpu which CPU the task is scheduled on 370 rt_priority realtime priority 371 policy scheduling policy (man sched_setscheduler) 372 blkio_ticks time spent waiting for block IO 373 gtime guest time of the task in jiffies 374 cgtime guest time of the task children in jiffies 375 start_data address above which program data+bss is placed 376 end_data address below which program data+bss is placed 377 start_brk address above which program heap can be expanded with brk() 378 arg_start address above which program command line is placed 379 arg_end address below which program command line is placed 380 env_start address above which program environment is placed 381 env_end address below which program environment is placed 382 exit_code the thread's exit_code in the form reported by the waitpid 383 system call 384 ============= =============================================================== 385 386The /proc/PID/maps file contains the currently mapped memory regions and 387their access permissions. 388 389The format is:: 390 391 address perms offset dev inode pathname 392 393 08048000-08049000 r-xp 00000000 03:00 8312 /opt/test 394 08049000-0804a000 rw-p 00001000 03:00 8312 /opt/test 395 0804a000-0806b000 rw-p 00000000 00:00 0 [heap] 396 a7cb1000-a7cb2000 ---p 00000000 00:00 0 397 a7cb2000-a7eb2000 rw-p 00000000 00:00 0 398 a7eb2000-a7eb3000 ---p 00000000 00:00 0 399 a7eb3000-a7ed5000 rw-p 00000000 00:00 0 400 a7ed5000-a8008000 r-xp 00000000 03:00 4222 /lib/libc.so.6 401 a8008000-a800a000 r--p 00133000 03:00 4222 /lib/libc.so.6 402 a800a000-a800b000 rw-p 00135000 03:00 4222 /lib/libc.so.6 403 a800b000-a800e000 rw-p 00000000 00:00 0 404 a800e000-a8022000 r-xp 00000000 03:00 14462 /lib/libpthread.so.0 405 a8022000-a8023000 r--p 00013000 03:00 14462 /lib/libpthread.so.0 406 a8023000-a8024000 rw-p 00014000 03:00 14462 /lib/libpthread.so.0 407 a8024000-a8027000 rw-p 00000000 00:00 0 408 a8027000-a8043000 r-xp 00000000 03:00 8317 /lib/ld-linux.so.2 409 a8043000-a8044000 r--p 0001b000 03:00 8317 /lib/ld-linux.so.2 410 a8044000-a8045000 rw-p 0001c000 03:00 8317 /lib/ld-linux.so.2 411 aff35000-aff4a000 rw-p 00000000 00:00 0 [stack] 412 ffffe000-fffff000 r-xp 00000000 00:00 0 [vdso] 413 414where "address" is the address space in the process that it occupies, "perms" 415is a set of permissions:: 416 417 r = read 418 w = write 419 x = execute 420 s = shared 421 p = private (copy on write) 422 423"offset" is the offset into the mapping, "dev" is the device (major:minor), and 424"inode" is the inode on that device. 0 indicates that no inode is associated 425with the memory region, as the case would be with BSS (uninitialized data). 426The "pathname" shows the name associated file for this mapping. If the mapping 427is not associated with a file: 428 429 =================== =========================================== 430 [heap] the heap of the program 431 [stack] the stack of the main process 432 [vdso] the "virtual dynamic shared object", 433 the kernel system call handler 434 [anon:<name>] a private anonymous mapping that has been 435 named by userspace 436 [anon_shmem:<name>] an anonymous shared memory mapping that has 437 been named by userspace 438 =================== =========================================== 439 440 or if empty, the mapping is anonymous. 441 442The /proc/PID/smaps is an extension based on maps, showing the memory 443consumption for each of the process's mappings. For each mapping (aka Virtual 444Memory Area, or VMA) there is a series of lines such as the following:: 445 446 08048000-080bc000 r-xp 00000000 03:02 13130 /bin/bash 447 448 Size: 1084 kB 449 KernelPageSize: 4 kB 450 MMUPageSize: 4 kB 451 Rss: 892 kB 452 Pss: 374 kB 453 Pss_Dirty: 0 kB 454 Shared_Clean: 892 kB 455 Shared_Dirty: 0 kB 456 Private_Clean: 0 kB 457 Private_Dirty: 0 kB 458 Referenced: 892 kB 459 Anonymous: 0 kB 460 LazyFree: 0 kB 461 AnonHugePages: 0 kB 462 ShmemPmdMapped: 0 kB 463 Shared_Hugetlb: 0 kB 464 Private_Hugetlb: 0 kB 465 Swap: 0 kB 466 SwapPss: 0 kB 467 KernelPageSize: 4 kB 468 MMUPageSize: 4 kB 469 Locked: 0 kB 470 THPeligible: 0 471 VmFlags: rd ex mr mw me dw 472 473The first of these lines shows the same information as is displayed for the 474mapping in /proc/PID/maps. Following lines show the size of the mapping 475(size); the size of each page allocated when backing a VMA (KernelPageSize), 476which is usually the same as the size in the page table entries; the page size 477used by the MMU when backing a VMA (in most cases, the same as KernelPageSize); 478the amount of the mapping that is currently resident in RAM (RSS); the 479process' proportional share of this mapping (PSS); and the number of clean and 480dirty shared and private pages in the mapping. 481 482The "proportional set size" (PSS) of a process is the count of pages it has 483in memory, where each page is divided by the number of processes sharing it. 484So if a process has 1000 pages all to itself, and 1000 shared with one other 485process, its PSS will be 1500. "Pss_Dirty" is the portion of PSS which 486consists of dirty pages. ("Pss_Clean" is not included, but it can be 487calculated by subtracting "Pss_Dirty" from "Pss".) 488 489Note that even a page which is part of a MAP_SHARED mapping, but has only 490a single pte mapped, i.e. is currently used by only one process, is accounted 491as private and not as shared. 492 493"Referenced" indicates the amount of memory currently marked as referenced or 494accessed. 495 496"Anonymous" shows the amount of memory that does not belong to any file. Even 497a mapping associated with a file may contain anonymous pages: when MAP_PRIVATE 498and a page is modified, the file page is replaced by a private anonymous copy. 499 500"LazyFree" shows the amount of memory which is marked by madvise(MADV_FREE). 501The memory isn't freed immediately with madvise(). It's freed in memory 502pressure if the memory is clean. Please note that the printed value might 503be lower than the real value due to optimizations used in the current 504implementation. If this is not desirable please file a bug report. 505 506"AnonHugePages" shows the ammount of memory backed by transparent hugepage. 507 508"ShmemPmdMapped" shows the ammount of shared (shmem/tmpfs) memory backed by 509huge pages. 510 511"Shared_Hugetlb" and "Private_Hugetlb" show the ammounts of memory backed by 512hugetlbfs page which is *not* counted in "RSS" or "PSS" field for historical 513reasons. And these are not included in {Shared,Private}_{Clean,Dirty} field. 514 515"Swap" shows how much would-be-anonymous memory is also used, but out on swap. 516 517For shmem mappings, "Swap" includes also the size of the mapped (and not 518replaced by copy-on-write) part of the underlying shmem object out on swap. 519"SwapPss" shows proportional swap share of this mapping. Unlike "Swap", this 520does not take into account swapped out page of underlying shmem objects. 521"Locked" indicates whether the mapping is locked in memory or not. 522 523"THPeligible" indicates whether the mapping is eligible for allocating THP 524pages as well as the THP is PMD mappable or not - 1 if true, 0 otherwise. 525It just shows the current status. 526 527"VmFlags" field deserves a separate description. This member represents the 528kernel flags associated with the particular virtual memory area in two letter 529encoded manner. The codes are the following: 530 531 == ======================================= 532 rd readable 533 wr writeable 534 ex executable 535 sh shared 536 mr may read 537 mw may write 538 me may execute 539 ms may share 540 gd stack segment growns down 541 pf pure PFN range 542 dw disabled write to the mapped file 543 lo pages are locked in memory 544 io memory mapped I/O area 545 sr sequential read advise provided 546 rr random read advise provided 547 dc do not copy area on fork 548 de do not expand area on remapping 549 ac area is accountable 550 nr swap space is not reserved for the area 551 ht area uses huge tlb pages 552 sf synchronous page fault 553 ar architecture specific flag 554 wf wipe on fork 555 dd do not include area into core dump 556 sd soft dirty flag 557 mm mixed map area 558 hg huge page advise flag 559 nh no huge page advise flag 560 mg mergable advise flag 561 bt arm64 BTI guarded page 562 mt arm64 MTE allocation tags are enabled 563 um userfaultfd missing tracking 564 uw userfaultfd wr-protect tracking 565 == ======================================= 566 567Note that there is no guarantee that every flag and associated mnemonic will 568be present in all further kernel releases. Things get changed, the flags may 569be vanished or the reverse -- new added. Interpretation of their meaning 570might change in future as well. So each consumer of these flags has to 571follow each specific kernel version for the exact semantic. 572 573This file is only present if the CONFIG_MMU kernel configuration option is 574enabled. 575 576Note: reading /proc/PID/maps or /proc/PID/smaps is inherently racy (consistent 577output can be achieved only in the single read call). 578 579This typically manifests when doing partial reads of these files while the 580memory map is being modified. Despite the races, we do provide the following 581guarantees: 582 5831) The mapped addresses never go backwards, which implies no two 584 regions will ever overlap. 5852) If there is something at a given vaddr during the entirety of the 586 life of the smaps/maps walk, there will be some output for it. 587 588The /proc/PID/smaps_rollup file includes the same fields as /proc/PID/smaps, 589but their values are the sums of the corresponding values for all mappings of 590the process. Additionally, it contains these fields: 591 592- Pss_Anon 593- Pss_File 594- Pss_Shmem 595 596They represent the proportional shares of anonymous, file, and shmem pages, as 597described for smaps above. These fields are omitted in smaps since each 598mapping identifies the type (anon, file, or shmem) of all pages it contains. 599Thus all information in smaps_rollup can be derived from smaps, but at a 600significantly higher cost. 601 602The /proc/PID/clear_refs is used to reset the PG_Referenced and ACCESSED/YOUNG 603bits on both physical and virtual pages associated with a process, and the 604soft-dirty bit on pte (see Documentation/admin-guide/mm/soft-dirty.rst 605for details). 606To clear the bits for all the pages associated with the process:: 607 608 > echo 1 > /proc/PID/clear_refs 609 610To clear the bits for the anonymous pages associated with the process:: 611 612 > echo 2 > /proc/PID/clear_refs 613 614To clear the bits for the file mapped pages associated with the process:: 615 616 > echo 3 > /proc/PID/clear_refs 617 618To clear the soft-dirty bit:: 619 620 > echo 4 > /proc/PID/clear_refs 621 622To reset the peak resident set size ("high water mark") to the process's 623current value:: 624 625 > echo 5 > /proc/PID/clear_refs 626 627Any other value written to /proc/PID/clear_refs will have no effect. 628 629The /proc/pid/pagemap gives the PFN, which can be used to find the pageflags 630using /proc/kpageflags and number of times a page is mapped using 631/proc/kpagecount. For detailed explanation, see 632Documentation/admin-guide/mm/pagemap.rst. 633 634The /proc/pid/numa_maps is an extension based on maps, showing the memory 635locality and binding policy, as well as the memory usage (in pages) of 636each mapping. The output follows a general format where mapping details get 637summarized separated by blank spaces, one mapping per each file line:: 638 639 address policy mapping details 640 641 00400000 default file=/usr/local/bin/app mapped=1 active=0 N3=1 kernelpagesize_kB=4 642 00600000 default file=/usr/local/bin/app anon=1 dirty=1 N3=1 kernelpagesize_kB=4 643 3206000000 default file=/lib64/ld-2.12.so mapped=26 mapmax=6 N0=24 N3=2 kernelpagesize_kB=4 644 320621f000 default file=/lib64/ld-2.12.so anon=1 dirty=1 N3=1 kernelpagesize_kB=4 645 3206220000 default file=/lib64/ld-2.12.so anon=1 dirty=1 N3=1 kernelpagesize_kB=4 646 3206221000 default anon=1 dirty=1 N3=1 kernelpagesize_kB=4 647 3206800000 default file=/lib64/libc-2.12.so mapped=59 mapmax=21 active=55 N0=41 N3=18 kernelpagesize_kB=4 648 320698b000 default file=/lib64/libc-2.12.so 649 3206b8a000 default file=/lib64/libc-2.12.so anon=2 dirty=2 N3=2 kernelpagesize_kB=4 650 3206b8e000 default file=/lib64/libc-2.12.so anon=1 dirty=1 N3=1 kernelpagesize_kB=4 651 3206b8f000 default anon=3 dirty=3 active=1 N3=3 kernelpagesize_kB=4 652 7f4dc10a2000 default anon=3 dirty=3 N3=3 kernelpagesize_kB=4 653 7f4dc10b4000 default anon=2 dirty=2 active=1 N3=2 kernelpagesize_kB=4 654 7f4dc1200000 default file=/anon_hugepage\040(deleted) huge anon=1 dirty=1 N3=1 kernelpagesize_kB=2048 655 7fff335f0000 default stack anon=3 dirty=3 N3=3 kernelpagesize_kB=4 656 7fff3369d000 default mapped=1 mapmax=35 active=0 N3=1 kernelpagesize_kB=4 657 658Where: 659 660"address" is the starting address for the mapping; 661 662"policy" reports the NUMA memory policy set for the mapping (see Documentation/admin-guide/mm/numa_memory_policy.rst); 663 664"mapping details" summarizes mapping data such as mapping type, page usage counters, 665node locality page counters (N0 == node0, N1 == node1, ...) and the kernel page 666size, in KB, that is backing the mapping up. 667 6681.2 Kernel data 669--------------- 670 671Similar to the process entries, the kernel data files give information about 672the running kernel. The files used to obtain this information are contained in 673/proc and are listed in Table 1-5. Not all of these will be present in your 674system. It depends on the kernel configuration and the loaded modules, which 675files are there, and which are missing. 676 677.. table:: Table 1-5: Kernel info in /proc 678 679 ============ =============================================================== 680 File Content 681 ============ =============================================================== 682 apm Advanced power management info 683 buddyinfo Kernel memory allocator information (see text) (2.5) 684 bus Directory containing bus specific information 685 cmdline Kernel command line 686 cpuinfo Info about the CPU 687 devices Available devices (block and character) 688 dma Used DMS channels 689 filesystems Supported filesystems 690 driver Various drivers grouped here, currently rtc (2.4) 691 execdomains Execdomains, related to security (2.4) 692 fb Frame Buffer devices (2.4) 693 fs File system parameters, currently nfs/exports (2.4) 694 ide Directory containing info about the IDE subsystem 695 interrupts Interrupt usage 696 iomem Memory map (2.4) 697 ioports I/O port usage 698 irq Masks for irq to cpu affinity (2.4)(smp?) 699 isapnp ISA PnP (Plug&Play) Info (2.4) 700 kcore Kernel core image (can be ELF or A.OUT(deprecated in 2.4)) 701 kmsg Kernel messages 702 ksyms Kernel symbol table 703 loadavg Load average of last 1, 5 & 15 minutes; 704 number of processes currently runnable (running or on ready queue); 705 total number of processes in system; 706 last pid created. 707 All fields are separated by one space except "number of 708 processes currently runnable" and "total number of processes 709 in system", which are separated by a slash ('/'). Example: 710 0.61 0.61 0.55 3/828 22084 711 locks Kernel locks 712 meminfo Memory info 713 misc Miscellaneous 714 modules List of loaded modules 715 mounts Mounted filesystems 716 net Networking info (see text) 717 pagetypeinfo Additional page allocator information (see text) (2.5) 718 partitions Table of partitions known to the system 719 pci Deprecated info of PCI bus (new way -> /proc/bus/pci/, 720 decoupled by lspci (2.4) 721 rtc Real time clock 722 scsi SCSI info (see text) 723 slabinfo Slab pool info 724 softirqs softirq usage 725 stat Overall statistics 726 swaps Swap space utilization 727 sys See chapter 2 728 sysvipc Info of SysVIPC Resources (msg, sem, shm) (2.4) 729 tty Info of tty drivers 730 uptime Wall clock since boot, combined idle time of all cpus 731 version Kernel version 732 video bttv info of video resources (2.4) 733 vmallocinfo Show vmalloced areas 734 ============ =============================================================== 735 736You can, for example, check which interrupts are currently in use and what 737they are used for by looking in the file /proc/interrupts:: 738 739 > cat /proc/interrupts 740 CPU0 741 0: 8728810 XT-PIC timer 742 1: 895 XT-PIC keyboard 743 2: 0 XT-PIC cascade 744 3: 531695 XT-PIC aha152x 745 4: 2014133 XT-PIC serial 746 5: 44401 XT-PIC pcnet_cs 747 8: 2 XT-PIC rtc 748 11: 8 XT-PIC i82365 749 12: 182918 XT-PIC PS/2 Mouse 750 13: 1 XT-PIC fpu 751 14: 1232265 XT-PIC ide0 752 15: 7 XT-PIC ide1 753 NMI: 0 754 755In 2.4.* a couple of lines where added to this file LOC & ERR (this time is the 756output of a SMP machine):: 757 758 > cat /proc/interrupts 759 760 CPU0 CPU1 761 0: 1243498 1214548 IO-APIC-edge timer 762 1: 8949 8958 IO-APIC-edge keyboard 763 2: 0 0 XT-PIC cascade 764 5: 11286 10161 IO-APIC-edge soundblaster 765 8: 1 0 IO-APIC-edge rtc 766 9: 27422 27407 IO-APIC-edge 3c503 767 12: 113645 113873 IO-APIC-edge PS/2 Mouse 768 13: 0 0 XT-PIC fpu 769 14: 22491 24012 IO-APIC-edge ide0 770 15: 2183 2415 IO-APIC-edge ide1 771 17: 30564 30414 IO-APIC-level eth0 772 18: 177 164 IO-APIC-level bttv 773 NMI: 2457961 2457959 774 LOC: 2457882 2457881 775 ERR: 2155 776 777NMI is incremented in this case because every timer interrupt generates a NMI 778(Non Maskable Interrupt) which is used by the NMI Watchdog to detect lockups. 779 780LOC is the local interrupt counter of the internal APIC of every CPU. 781 782ERR is incremented in the case of errors in the IO-APIC bus (the bus that 783connects the CPUs in a SMP system. This means that an error has been detected, 784the IO-APIC automatically retry the transmission, so it should not be a big 785problem, but you should read the SMP-FAQ. 786 787In 2.6.2* /proc/interrupts was expanded again. This time the goal was for 788/proc/interrupts to display every IRQ vector in use by the system, not 789just those considered 'most important'. The new vectors are: 790 791THR 792 interrupt raised when a machine check threshold counter 793 (typically counting ECC corrected errors of memory or cache) exceeds 794 a configurable threshold. Only available on some systems. 795 796TRM 797 a thermal event interrupt occurs when a temperature threshold 798 has been exceeded for the CPU. This interrupt may also be generated 799 when the temperature drops back to normal. 800 801SPU 802 a spurious interrupt is some interrupt that was raised then lowered 803 by some IO device before it could be fully processed by the APIC. Hence 804 the APIC sees the interrupt but does not know what device it came from. 805 For this case the APIC will generate the interrupt with a IRQ vector 806 of 0xff. This might also be generated by chipset bugs. 807 808RES, CAL, TLB 809 rescheduling, call and TLB flush interrupts are 810 sent from one CPU to another per the needs of the OS. Typically, 811 their statistics are used by kernel developers and interested users to 812 determine the occurrence of interrupts of the given type. 813 814The above IRQ vectors are displayed only when relevant. For example, 815the threshold vector does not exist on x86_64 platforms. Others are 816suppressed when the system is a uniprocessor. As of this writing, only 817i386 and x86_64 platforms support the new IRQ vector displays. 818 819Of some interest is the introduction of the /proc/irq directory to 2.4. 820It could be used to set IRQ to CPU affinity. This means that you can "hook" an 821IRQ to only one CPU, or to exclude a CPU of handling IRQs. The contents of the 822irq subdir is one subdir for each IRQ, and two files; default_smp_affinity and 823prof_cpu_mask. 824 825For example:: 826 827 > ls /proc/irq/ 828 0 10 12 14 16 18 2 4 6 8 prof_cpu_mask 829 1 11 13 15 17 19 3 5 7 9 default_smp_affinity 830 > ls /proc/irq/0/ 831 smp_affinity 832 833smp_affinity is a bitmask, in which you can specify which CPUs can handle the 834IRQ. You can set it by doing:: 835 836 > echo 1 > /proc/irq/10/smp_affinity 837 838This means that only the first CPU will handle the IRQ, but you can also echo 8395 which means that only the first and third CPU can handle the IRQ. 840 841The contents of each smp_affinity file is the same by default:: 842 843 > cat /proc/irq/0/smp_affinity 844 ffffffff 845 846There is an alternate interface, smp_affinity_list which allows specifying 847a CPU range instead of a bitmask:: 848 849 > cat /proc/irq/0/smp_affinity_list 850 1024-1031 851 852The default_smp_affinity mask applies to all non-active IRQs, which are the 853IRQs which have not yet been allocated/activated, and hence which lack a 854/proc/irq/[0-9]* directory. 855 856The node file on an SMP system shows the node to which the device using the IRQ 857reports itself as being attached. This hardware locality information does not 858include information about any possible driver locality preference. 859 860prof_cpu_mask specifies which CPUs are to be profiled by the system wide 861profiler. Default value is ffffffff (all CPUs if there are only 32 of them). 862 863The way IRQs are routed is handled by the IO-APIC, and it's Round Robin 864between all the CPUs which are allowed to handle it. As usual the kernel has 865more info than you and does a better job than you, so the defaults are the 866best choice for almost everyone. [Note this applies only to those IO-APIC's 867that support "Round Robin" interrupt distribution.] 868 869There are three more important subdirectories in /proc: net, scsi, and sys. 870The general rule is that the contents, or even the existence of these 871directories, depend on your kernel configuration. If SCSI is not enabled, the 872directory scsi may not exist. The same is true with the net, which is there 873only when networking support is present in the running kernel. 874 875The slabinfo file gives information about memory usage at the slab level. 876Linux uses slab pools for memory management above page level in version 2.2. 877Commonly used objects have their own slab pool (such as network buffers, 878directory cache, and so on). 879 880:: 881 882 > cat /proc/buddyinfo 883 884 Node 0, zone DMA 0 4 5 4 4 3 ... 885 Node 0, zone Normal 1 0 0 1 101 8 ... 886 Node 0, zone HighMem 2 0 0 1 1 0 ... 887 888External fragmentation is a problem under some workloads, and buddyinfo is a 889useful tool for helping diagnose these problems. Buddyinfo will give you a 890clue as to how big an area you can safely allocate, or why a previous 891allocation failed. 892 893Each column represents the number of pages of a certain order which are 894available. In this case, there are 0 chunks of 2^0*PAGE_SIZE available in 895ZONE_DMA, 4 chunks of 2^1*PAGE_SIZE in ZONE_DMA, 101 chunks of 2^4*PAGE_SIZE 896available in ZONE_NORMAL, etc... 897 898More information relevant to external fragmentation can be found in 899pagetypeinfo:: 900 901 > cat /proc/pagetypeinfo 902 Page block order: 9 903 Pages per block: 512 904 905 Free pages count per migrate type at order 0 1 2 3 4 5 6 7 8 9 10 906 Node 0, zone DMA, type Unmovable 0 0 0 1 1 1 1 1 1 1 0 907 Node 0, zone DMA, type Reclaimable 0 0 0 0 0 0 0 0 0 0 0 908 Node 0, zone DMA, type Movable 1 1 2 1 2 1 1 0 1 0 2 909 Node 0, zone DMA, type Reserve 0 0 0 0 0 0 0 0 0 1 0 910 Node 0, zone DMA, type Isolate 0 0 0 0 0 0 0 0 0 0 0 911 Node 0, zone DMA32, type Unmovable 103 54 77 1 1 1 11 8 7 1 9 912 Node 0, zone DMA32, type Reclaimable 0 0 2 1 0 0 0 0 1 0 0 913 Node 0, zone DMA32, type Movable 169 152 113 91 77 54 39 13 6 1 452 914 Node 0, zone DMA32, type Reserve 1 2 2 2 2 0 1 1 1 1 0 915 Node 0, zone DMA32, type Isolate 0 0 0 0 0 0 0 0 0 0 0 916 917 Number of blocks type Unmovable Reclaimable Movable Reserve Isolate 918 Node 0, zone DMA 2 0 5 1 0 919 Node 0, zone DMA32 41 6 967 2 0 920 921Fragmentation avoidance in the kernel works by grouping pages of different 922migrate types into the same contiguous regions of memory called page blocks. 923A page block is typically the size of the default hugepage size, e.g. 2MB on 924X86-64. By keeping pages grouped based on their ability to move, the kernel 925can reclaim pages within a page block to satisfy a high-order allocation. 926 927The pagetypinfo begins with information on the size of a page block. It 928then gives the same type of information as buddyinfo except broken down 929by migrate-type and finishes with details on how many page blocks of each 930type exist. 931 932If min_free_kbytes has been tuned correctly (recommendations made by hugeadm 933from libhugetlbfs https://github.com/libhugetlbfs/libhugetlbfs/), one can 934make an estimate of the likely number of huge pages that can be allocated 935at a given point in time. All the "Movable" blocks should be allocatable 936unless memory has been mlock()'d. Some of the Reclaimable blocks should 937also be allocatable although a lot of filesystem metadata may have to be 938reclaimed to achieve this. 939 940 941meminfo 942~~~~~~~ 943 944Provides information about distribution and utilization of memory. This 945varies by architecture and compile options. Some of the counters reported 946here overlap. The memory reported by the non overlapping counters may not 947add up to the overall memory usage and the difference for some workloads 948can be substantial. In many cases there are other means to find out 949additional memory using subsystem specific interfaces, for instance 950/proc/net/sockstat for TCP memory allocations. 951 952Example output. You may not have all of these fields. 953 954:: 955 956 > cat /proc/meminfo 957 958 MemTotal: 32858820 kB 959 MemFree: 21001236 kB 960 MemAvailable: 27214312 kB 961 Buffers: 581092 kB 962 Cached: 5587612 kB 963 SwapCached: 0 kB 964 Active: 3237152 kB 965 Inactive: 7586256 kB 966 Active(anon): 94064 kB 967 Inactive(anon): 4570616 kB 968 Active(file): 3143088 kB 969 Inactive(file): 3015640 kB 970 Unevictable: 0 kB 971 Mlocked: 0 kB 972 SwapTotal: 0 kB 973 SwapFree: 0 kB 974 Zswap: 1904 kB 975 Zswapped: 7792 kB 976 Dirty: 12 kB 977 Writeback: 0 kB 978 AnonPages: 4654780 kB 979 Mapped: 266244 kB 980 Shmem: 9976 kB 981 KReclaimable: 517708 kB 982 Slab: 660044 kB 983 SReclaimable: 517708 kB 984 SUnreclaim: 142336 kB 985 KernelStack: 11168 kB 986 PageTables: 20540 kB 987 SecPageTables: 0 kB 988 NFS_Unstable: 0 kB 989 Bounce: 0 kB 990 WritebackTmp: 0 kB 991 CommitLimit: 16429408 kB 992 Committed_AS: 7715148 kB 993 VmallocTotal: 34359738367 kB 994 VmallocUsed: 40444 kB 995 VmallocChunk: 0 kB 996 Percpu: 29312 kB 997 HardwareCorrupted: 0 kB 998 AnonHugePages: 4149248 kB 999 ShmemHugePages: 0 kB 1000 ShmemPmdMapped: 0 kB 1001 FileHugePages: 0 kB 1002 FilePmdMapped: 0 kB 1003 CmaTotal: 0 kB 1004 CmaFree: 0 kB 1005 HugePages_Total: 0 1006 HugePages_Free: 0 1007 HugePages_Rsvd: 0 1008 HugePages_Surp: 0 1009 Hugepagesize: 2048 kB 1010 Hugetlb: 0 kB 1011 DirectMap4k: 401152 kB 1012 DirectMap2M: 10008576 kB 1013 DirectMap1G: 24117248 kB 1014 1015MemTotal 1016 Total usable RAM (i.e. physical RAM minus a few reserved 1017 bits and the kernel binary code) 1018MemFree 1019 Total free RAM. On highmem systems, the sum of LowFree+HighFree 1020MemAvailable 1021 An estimate of how much memory is available for starting new 1022 applications, without swapping. Calculated from MemFree, 1023 SReclaimable, the size of the file LRU lists, and the low 1024 watermarks in each zone. 1025 The estimate takes into account that the system needs some 1026 page cache to function well, and that not all reclaimable 1027 slab will be reclaimable, due to items being in use. The 1028 impact of those factors will vary from system to system. 1029Buffers 1030 Relatively temporary storage for raw disk blocks 1031 shouldn't get tremendously large (20MB or so) 1032Cached 1033 In-memory cache for files read from the disk (the 1034 pagecache) as well as tmpfs & shmem. 1035 Doesn't include SwapCached. 1036SwapCached 1037 Memory that once was swapped out, is swapped back in but 1038 still also is in the swapfile (if memory is needed it 1039 doesn't need to be swapped out AGAIN because it is already 1040 in the swapfile. This saves I/O) 1041Active 1042 Memory that has been used more recently and usually not 1043 reclaimed unless absolutely necessary. 1044Inactive 1045 Memory which has been less recently used. It is more 1046 eligible to be reclaimed for other purposes 1047Unevictable 1048 Memory allocated for userspace which cannot be reclaimed, such 1049 as mlocked pages, ramfs backing pages, secret memfd pages etc. 1050Mlocked 1051 Memory locked with mlock(). 1052HighTotal, HighFree 1053 Highmem is all memory above ~860MB of physical memory. 1054 Highmem areas are for use by userspace programs, or 1055 for the pagecache. The kernel must use tricks to access 1056 this memory, making it slower to access than lowmem. 1057LowTotal, LowFree 1058 Lowmem is memory which can be used for everything that 1059 highmem can be used for, but it is also available for the 1060 kernel's use for its own data structures. Among many 1061 other things, it is where everything from the Slab is 1062 allocated. Bad things happen when you're out of lowmem. 1063SwapTotal 1064 total amount of swap space available 1065SwapFree 1066 Memory which has been evicted from RAM, and is temporarily 1067 on the disk 1068Zswap 1069 Memory consumed by the zswap backend (compressed size) 1070Zswapped 1071 Amount of anonymous memory stored in zswap (original size) 1072Dirty 1073 Memory which is waiting to get written back to the disk 1074Writeback 1075 Memory which is actively being written back to the disk 1076AnonPages 1077 Non-file backed pages mapped into userspace page tables 1078Mapped 1079 files which have been mmaped, such as libraries 1080Shmem 1081 Total memory used by shared memory (shmem) and tmpfs 1082KReclaimable 1083 Kernel allocations that the kernel will attempt to reclaim 1084 under memory pressure. Includes SReclaimable (below), and other 1085 direct allocations with a shrinker. 1086Slab 1087 in-kernel data structures cache 1088SReclaimable 1089 Part of Slab, that might be reclaimed, such as caches 1090SUnreclaim 1091 Part of Slab, that cannot be reclaimed on memory pressure 1092KernelStack 1093 Memory consumed by the kernel stacks of all tasks 1094PageTables 1095 Memory consumed by userspace page tables 1096SecPageTables 1097 Memory consumed by secondary page tables, this currently 1098 currently includes KVM mmu allocations on x86 and arm64. 1099NFS_Unstable 1100 Always zero. Previous counted pages which had been written to 1101 the server, but has not been committed to stable storage. 1102Bounce 1103 Memory used for block device "bounce buffers" 1104WritebackTmp 1105 Memory used by FUSE for temporary writeback buffers 1106CommitLimit 1107 Based on the overcommit ratio ('vm.overcommit_ratio'), 1108 this is the total amount of memory currently available to 1109 be allocated on the system. This limit is only adhered to 1110 if strict overcommit accounting is enabled (mode 2 in 1111 'vm.overcommit_memory'). 1112 1113 The CommitLimit is calculated with the following formula:: 1114 1115 CommitLimit = ([total RAM pages] - [total huge TLB pages]) * 1116 overcommit_ratio / 100 + [total swap pages] 1117 1118 For example, on a system with 1G of physical RAM and 7G 1119 of swap with a `vm.overcommit_ratio` of 30 it would 1120 yield a CommitLimit of 7.3G. 1121 1122 For more details, see the memory overcommit documentation 1123 in mm/overcommit-accounting. 1124Committed_AS 1125 The amount of memory presently allocated on the system. 1126 The committed memory is a sum of all of the memory which 1127 has been allocated by processes, even if it has not been 1128 "used" by them as of yet. A process which malloc()'s 1G 1129 of memory, but only touches 300M of it will show up as 1130 using 1G. This 1G is memory which has been "committed" to 1131 by the VM and can be used at any time by the allocating 1132 application. With strict overcommit enabled on the system 1133 (mode 2 in 'vm.overcommit_memory'), allocations which would 1134 exceed the CommitLimit (detailed above) will not be permitted. 1135 This is useful if one needs to guarantee that processes will 1136 not fail due to lack of memory once that memory has been 1137 successfully allocated. 1138VmallocTotal 1139 total size of vmalloc virtual address space 1140VmallocUsed 1141 amount of vmalloc area which is used 1142VmallocChunk 1143 largest contiguous block of vmalloc area which is free 1144Percpu 1145 Memory allocated to the percpu allocator used to back percpu 1146 allocations. This stat excludes the cost of metadata. 1147HardwareCorrupted 1148 The amount of RAM/memory in KB, the kernel identifies as 1149 corrupted. 1150AnonHugePages 1151 Non-file backed huge pages mapped into userspace page tables 1152ShmemHugePages 1153 Memory used by shared memory (shmem) and tmpfs allocated 1154 with huge pages 1155ShmemPmdMapped 1156 Shared memory mapped into userspace with huge pages 1157FileHugePages 1158 Memory used for filesystem data (page cache) allocated 1159 with huge pages 1160FilePmdMapped 1161 Page cache mapped into userspace with huge pages 1162CmaTotal 1163 Memory reserved for the Contiguous Memory Allocator (CMA) 1164CmaFree 1165 Free remaining memory in the CMA reserves 1166HugePages_Total, HugePages_Free, HugePages_Rsvd, HugePages_Surp, Hugepagesize, Hugetlb 1167 See Documentation/admin-guide/mm/hugetlbpage.rst. 1168DirectMap4k, DirectMap2M, DirectMap1G 1169 Breakdown of page table sizes used in the kernel's 1170 identity mapping of RAM 1171 1172vmallocinfo 1173~~~~~~~~~~~ 1174 1175Provides information about vmalloced/vmaped areas. One line per area, 1176containing the virtual address range of the area, size in bytes, 1177caller information of the creator, and optional information depending 1178on the kind of area: 1179 1180 ========== =================================================== 1181 pages=nr number of pages 1182 phys=addr if a physical address was specified 1183 ioremap I/O mapping (ioremap() and friends) 1184 vmalloc vmalloc() area 1185 vmap vmap()ed pages 1186 user VM_USERMAP area 1187 vpages buffer for pages pointers was vmalloced (huge area) 1188 N<node>=nr (Only on NUMA kernels) 1189 Number of pages allocated on memory node <node> 1190 ========== =================================================== 1191 1192:: 1193 1194 > cat /proc/vmallocinfo 1195 0xffffc20000000000-0xffffc20000201000 2101248 alloc_large_system_hash+0x204 ... 1196 /0x2c0 pages=512 vmalloc N0=128 N1=128 N2=128 N3=128 1197 0xffffc20000201000-0xffffc20000302000 1052672 alloc_large_system_hash+0x204 ... 1198 /0x2c0 pages=256 vmalloc N0=64 N1=64 N2=64 N3=64 1199 0xffffc20000302000-0xffffc20000304000 8192 acpi_tb_verify_table+0x21/0x4f... 1200 phys=7fee8000 ioremap 1201 0xffffc20000304000-0xffffc20000307000 12288 acpi_tb_verify_table+0x21/0x4f... 1202 phys=7fee7000 ioremap 1203 0xffffc2000031d000-0xffffc2000031f000 8192 init_vdso_vars+0x112/0x210 1204 0xffffc2000031f000-0xffffc2000032b000 49152 cramfs_uncompress_init+0x2e ... 1205 /0x80 pages=11 vmalloc N0=3 N1=3 N2=2 N3=3 1206 0xffffc2000033a000-0xffffc2000033d000 12288 sys_swapon+0x640/0xac0 ... 1207 pages=2 vmalloc N1=2 1208 0xffffc20000347000-0xffffc2000034c000 20480 xt_alloc_table_info+0xfe ... 1209 /0x130 [x_tables] pages=4 vmalloc N0=4 1210 0xffffffffa0000000-0xffffffffa000f000 61440 sys_init_module+0xc27/0x1d00 ... 1211 pages=14 vmalloc N2=14 1212 0xffffffffa000f000-0xffffffffa0014000 20480 sys_init_module+0xc27/0x1d00 ... 1213 pages=4 vmalloc N1=4 1214 0xffffffffa0014000-0xffffffffa0017000 12288 sys_init_module+0xc27/0x1d00 ... 1215 pages=2 vmalloc N1=2 1216 0xffffffffa0017000-0xffffffffa0022000 45056 sys_init_module+0xc27/0x1d00 ... 1217 pages=10 vmalloc N0=10 1218 1219 1220softirqs 1221~~~~~~~~ 1222 1223Provides counts of softirq handlers serviced since boot time, for each CPU. 1224 1225:: 1226 1227 > cat /proc/softirqs 1228 CPU0 CPU1 CPU2 CPU3 1229 HI: 0 0 0 0 1230 TIMER: 27166 27120 27097 27034 1231 NET_TX: 0 0 0 17 1232 NET_RX: 42 0 0 39 1233 BLOCK: 0 0 107 1121 1234 TASKLET: 0 0 0 290 1235 SCHED: 27035 26983 26971 26746 1236 HRTIMER: 0 0 0 0 1237 RCU: 1678 1769 2178 2250 1238 12391.3 Networking info in /proc/net 1240-------------------------------- 1241 1242The subdirectory /proc/net follows the usual pattern. Table 1-8 shows the 1243additional values you get for IP version 6 if you configure the kernel to 1244support this. Table 1-9 lists the files and their meaning. 1245 1246 1247.. table:: Table 1-8: IPv6 info in /proc/net 1248 1249 ========== ===================================================== 1250 File Content 1251 ========== ===================================================== 1252 udp6 UDP sockets (IPv6) 1253 tcp6 TCP sockets (IPv6) 1254 raw6 Raw device statistics (IPv6) 1255 igmp6 IP multicast addresses, which this host joined (IPv6) 1256 if_inet6 List of IPv6 interface addresses 1257 ipv6_route Kernel routing table for IPv6 1258 rt6_stats Global IPv6 routing tables statistics 1259 sockstat6 Socket statistics (IPv6) 1260 snmp6 Snmp data (IPv6) 1261 ========== ===================================================== 1262 1263.. table:: Table 1-9: Network info in /proc/net 1264 1265 ============= ================================================================ 1266 File Content 1267 ============= ================================================================ 1268 arp Kernel ARP table 1269 dev network devices with statistics 1270 dev_mcast the Layer2 multicast groups a device is listening too 1271 (interface index, label, number of references, number of bound 1272 addresses). 1273 dev_stat network device status 1274 ip_fwchains Firewall chain linkage 1275 ip_fwnames Firewall chain names 1276 ip_masq Directory containing the masquerading tables 1277 ip_masquerade Major masquerading table 1278 netstat Network statistics 1279 raw raw device statistics 1280 route Kernel routing table 1281 rpc Directory containing rpc info 1282 rt_cache Routing cache 1283 snmp SNMP data 1284 sockstat Socket statistics 1285 tcp TCP sockets 1286 udp UDP sockets 1287 unix UNIX domain sockets 1288 wireless Wireless interface data (Wavelan etc) 1289 igmp IP multicast addresses, which this host joined 1290 psched Global packet scheduler parameters. 1291 netlink List of PF_NETLINK sockets 1292 ip_mr_vifs List of multicast virtual interfaces 1293 ip_mr_cache List of multicast routing cache 1294 ============= ================================================================ 1295 1296You can use this information to see which network devices are available in 1297your system and how much traffic was routed over those devices:: 1298 1299 > cat /proc/net/dev 1300 Inter-|Receive |[... 1301 face |bytes packets errs drop fifo frame compressed multicast|[... 1302 lo: 908188 5596 0 0 0 0 0 0 [... 1303 ppp0:15475140 20721 410 0 0 410 0 0 [... 1304 eth0: 614530 7085 0 0 0 0 0 1 [... 1305 1306 ...] Transmit 1307 ...] bytes packets errs drop fifo colls carrier compressed 1308 ...] 908188 5596 0 0 0 0 0 0 1309 ...] 1375103 17405 0 0 0 0 0 0 1310 ...] 1703981 5535 0 0 0 3 0 0 1311 1312In addition, each Channel Bond interface has its own directory. For 1313example, the bond0 device will have a directory called /proc/net/bond0/. 1314It will contain information that is specific to that bond, such as the 1315current slaves of the bond, the link status of the slaves, and how 1316many times the slaves link has failed. 1317 13181.4 SCSI info 1319------------- 1320 1321If you have a SCSI host adapter in your system, you'll find a subdirectory 1322named after the driver for this adapter in /proc/scsi. You'll also see a list 1323of all recognized SCSI devices in /proc/scsi:: 1324 1325 >cat /proc/scsi/scsi 1326 Attached devices: 1327 Host: scsi0 Channel: 00 Id: 00 Lun: 00 1328 Vendor: IBM Model: DGHS09U Rev: 03E0 1329 Type: Direct-Access ANSI SCSI revision: 03 1330 Host: scsi0 Channel: 00 Id: 06 Lun: 00 1331 Vendor: PIONEER Model: CD-ROM DR-U06S Rev: 1.04 1332 Type: CD-ROM ANSI SCSI revision: 02 1333 1334 1335The directory named after the driver has one file for each adapter found in 1336the system. These files contain information about the controller, including 1337the used IRQ and the IO address range. The amount of information shown is 1338dependent on the adapter you use. The example shows the output for an Adaptec 1339AHA-2940 SCSI adapter:: 1340 1341 > cat /proc/scsi/aic7xxx/0 1342 1343 Adaptec AIC7xxx driver version: 5.1.19/3.2.4 1344 Compile Options: 1345 TCQ Enabled By Default : Disabled 1346 AIC7XXX_PROC_STATS : Disabled 1347 AIC7XXX_RESET_DELAY : 5 1348 Adapter Configuration: 1349 SCSI Adapter: Adaptec AHA-294X Ultra SCSI host adapter 1350 Ultra Wide Controller 1351 PCI MMAPed I/O Base: 0xeb001000 1352 Adapter SEEPROM Config: SEEPROM found and used. 1353 Adaptec SCSI BIOS: Enabled 1354 IRQ: 10 1355 SCBs: Active 0, Max Active 2, 1356 Allocated 15, HW 16, Page 255 1357 Interrupts: 160328 1358 BIOS Control Word: 0x18b6 1359 Adapter Control Word: 0x005b 1360 Extended Translation: Enabled 1361 Disconnect Enable Flags: 0xffff 1362 Ultra Enable Flags: 0x0001 1363 Tag Queue Enable Flags: 0x0000 1364 Ordered Queue Tag Flags: 0x0000 1365 Default Tag Queue Depth: 8 1366 Tagged Queue By Device array for aic7xxx host instance 0: 1367 {255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255} 1368 Actual queue depth per device for aic7xxx host instance 0: 1369 {1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1} 1370 Statistics: 1371 (scsi0:0:0:0) 1372 Device using Wide/Sync transfers at 40.0 MByte/sec, offset 8 1373 Transinfo settings: current(12/8/1/0), goal(12/8/1/0), user(12/15/1/0) 1374 Total transfers 160151 (74577 reads and 85574 writes) 1375 (scsi0:0:6:0) 1376 Device using Narrow/Sync transfers at 5.0 MByte/sec, offset 15 1377 Transinfo settings: current(50/15/0/0), goal(50/15/0/0), user(50/15/0/0) 1378 Total transfers 0 (0 reads and 0 writes) 1379 1380 13811.5 Parallel port info in /proc/parport 1382--------------------------------------- 1383 1384The directory /proc/parport contains information about the parallel ports of 1385your system. It has one subdirectory for each port, named after the port 1386number (0,1,2,...). 1387 1388These directories contain the four files shown in Table 1-10. 1389 1390 1391.. table:: Table 1-10: Files in /proc/parport 1392 1393 ========= ==================================================================== 1394 File Content 1395 ========= ==================================================================== 1396 autoprobe Any IEEE-1284 device ID information that has been acquired. 1397 devices list of the device drivers using that port. A + will appear by the 1398 name of the device currently using the port (it might not appear 1399 against any). 1400 hardware Parallel port's base address, IRQ line and DMA channel. 1401 irq IRQ that parport is using for that port. This is in a separate 1402 file to allow you to alter it by writing a new value in (IRQ 1403 number or none). 1404 ========= ==================================================================== 1405 14061.6 TTY info in /proc/tty 1407------------------------- 1408 1409Information about the available and actually used tty's can be found in the 1410directory /proc/tty. You'll find entries for drivers and line disciplines in 1411this directory, as shown in Table 1-11. 1412 1413 1414.. table:: Table 1-11: Files in /proc/tty 1415 1416 ============= ============================================== 1417 File Content 1418 ============= ============================================== 1419 drivers list of drivers and their usage 1420 ldiscs registered line disciplines 1421 driver/serial usage statistic and status of single tty lines 1422 ============= ============================================== 1423 1424To see which tty's are currently in use, you can simply look into the file 1425/proc/tty/drivers:: 1426 1427 > cat /proc/tty/drivers 1428 pty_slave /dev/pts 136 0-255 pty:slave 1429 pty_master /dev/ptm 128 0-255 pty:master 1430 pty_slave /dev/ttyp 3 0-255 pty:slave 1431 pty_master /dev/pty 2 0-255 pty:master 1432 serial /dev/cua 5 64-67 serial:callout 1433 serial /dev/ttyS 4 64-67 serial 1434 /dev/tty0 /dev/tty0 4 0 system:vtmaster 1435 /dev/ptmx /dev/ptmx 5 2 system 1436 /dev/console /dev/console 5 1 system:console 1437 /dev/tty /dev/tty 5 0 system:/dev/tty 1438 unknown /dev/tty 4 1-63 console 1439 1440 14411.7 Miscellaneous kernel statistics in /proc/stat 1442------------------------------------------------- 1443 1444Various pieces of information about kernel activity are available in the 1445/proc/stat file. All of the numbers reported in this file are aggregates 1446since the system first booted. For a quick look, simply cat the file:: 1447 1448 > cat /proc/stat 1449 cpu 2255 34 2290 22625563 6290 127 456 0 0 0 1450 cpu0 1132 34 1441 11311718 3675 127 438 0 0 0 1451 cpu1 1123 0 849 11313845 2614 0 18 0 0 0 1452 intr 114930548 113199788 3 0 5 263 0 4 [... lots more numbers ...] 1453 ctxt 1990473 1454 btime 1062191376 1455 processes 2915 1456 procs_running 1 1457 procs_blocked 0 1458 softirq 183433 0 21755 12 39 1137 231 21459 2263 1459 1460The very first "cpu" line aggregates the numbers in all of the other "cpuN" 1461lines. These numbers identify the amount of time the CPU has spent performing 1462different kinds of work. Time units are in USER_HZ (typically hundredths of a 1463second). The meanings of the columns are as follows, from left to right: 1464 1465- user: normal processes executing in user mode 1466- nice: niced processes executing in user mode 1467- system: processes executing in kernel mode 1468- idle: twiddling thumbs 1469- iowait: In a word, iowait stands for waiting for I/O to complete. But there 1470 are several problems: 1471 1472 1. CPU will not wait for I/O to complete, iowait is the time that a task is 1473 waiting for I/O to complete. When CPU goes into idle state for 1474 outstanding task I/O, another task will be scheduled on this CPU. 1475 2. In a multi-core CPU, the task waiting for I/O to complete is not running 1476 on any CPU, so the iowait of each CPU is difficult to calculate. 1477 3. The value of iowait field in /proc/stat will decrease in certain 1478 conditions. 1479 1480 So, the iowait is not reliable by reading from /proc/stat. 1481- irq: servicing interrupts 1482- softirq: servicing softirqs 1483- steal: involuntary wait 1484- guest: running a normal guest 1485- guest_nice: running a niced guest 1486 1487The "intr" line gives counts of interrupts serviced since boot time, for each 1488of the possible system interrupts. The first column is the total of all 1489interrupts serviced including unnumbered architecture specific interrupts; 1490each subsequent column is the total for that particular numbered interrupt. 1491Unnumbered interrupts are not shown, only summed into the total. 1492 1493The "ctxt" line gives the total number of context switches across all CPUs. 1494 1495The "btime" line gives the time at which the system booted, in seconds since 1496the Unix epoch. 1497 1498The "processes" line gives the number of processes and threads created, which 1499includes (but is not limited to) those created by calls to the fork() and 1500clone() system calls. 1501 1502The "procs_running" line gives the total number of threads that are 1503running or ready to run (i.e., the total number of runnable threads). 1504 1505The "procs_blocked" line gives the number of processes currently blocked, 1506waiting for I/O to complete. 1507 1508The "softirq" line gives counts of softirqs serviced since boot time, for each 1509of the possible system softirqs. The first column is the total of all 1510softirqs serviced; each subsequent column is the total for that particular 1511softirq. 1512 1513 15141.8 Ext4 file system parameters 1515------------------------------- 1516 1517Information about mounted ext4 file systems can be found in 1518/proc/fs/ext4. Each mounted filesystem will have a directory in 1519/proc/fs/ext4 based on its device name (i.e., /proc/fs/ext4/hdc or 1520/proc/fs/ext4/dm-0). The files in each per-device directory are shown 1521in Table 1-12, below. 1522 1523.. table:: Table 1-12: Files in /proc/fs/ext4/<devname> 1524 1525 ============== ========================================================== 1526 File Content 1527 mb_groups details of multiblock allocator buddy cache of free blocks 1528 ============== ========================================================== 1529 15301.9 /proc/consoles 1531------------------- 1532Shows registered system console lines. 1533 1534To see which character device lines are currently used for the system console 1535/dev/console, you may simply look into the file /proc/consoles:: 1536 1537 > cat /proc/consoles 1538 tty0 -WU (ECp) 4:7 1539 ttyS0 -W- (Ep) 4:64 1540 1541The columns are: 1542 1543+--------------------+-------------------------------------------------------+ 1544| device | name of the device | 1545+====================+=======================================================+ 1546| operations | * R = can do read operations | 1547| | * W = can do write operations | 1548| | * U = can do unblank | 1549+--------------------+-------------------------------------------------------+ 1550| flags | * E = it is enabled | 1551| | * C = it is preferred console | 1552| | * B = it is primary boot console | 1553| | * p = it is used for printk buffer | 1554| | * b = it is not a TTY but a Braille device | 1555| | * a = it is safe to use when cpu is offline | 1556+--------------------+-------------------------------------------------------+ 1557| major:minor | major and minor number of the device separated by a | 1558| | colon | 1559+--------------------+-------------------------------------------------------+ 1560 1561Summary 1562------- 1563 1564The /proc file system serves information about the running system. It not only 1565allows access to process data but also allows you to request the kernel status 1566by reading files in the hierarchy. 1567 1568The directory structure of /proc reflects the types of information and makes 1569it easy, if not obvious, where to look for specific data. 1570 1571Chapter 2: Modifying System Parameters 1572====================================== 1573 1574In This Chapter 1575--------------- 1576 1577* Modifying kernel parameters by writing into files found in /proc/sys 1578* Exploring the files which modify certain parameters 1579* Review of the /proc/sys file tree 1580 1581------------------------------------------------------------------------------ 1582 1583A very interesting part of /proc is the directory /proc/sys. This is not only 1584a source of information, it also allows you to change parameters within the 1585kernel. Be very careful when attempting this. You can optimize your system, 1586but you can also cause it to crash. Never alter kernel parameters on a 1587production system. Set up a development machine and test to make sure that 1588everything works the way you want it to. You may have no alternative but to 1589reboot the machine once an error has been made. 1590 1591To change a value, simply echo the new value into the file. 1592You need to be root to do this. You can create your own boot script 1593to perform this every time your system boots. 1594 1595The files in /proc/sys can be used to fine tune and monitor miscellaneous and 1596general things in the operation of the Linux kernel. Since some of the files 1597can inadvertently disrupt your system, it is advisable to read both 1598documentation and source before actually making adjustments. In any case, be 1599very careful when writing to any of these files. The entries in /proc may 1600change slightly between the 2.1.* and the 2.2 kernel, so if there is any doubt 1601review the kernel documentation in the directory /usr/src/linux/Documentation. 1602This chapter is heavily based on the documentation included in the pre 2.2 1603kernels, and became part of it in version 2.2.1 of the Linux kernel. 1604 1605Please see: Documentation/admin-guide/sysctl/ directory for descriptions of these 1606entries. 1607 1608Summary 1609------- 1610 1611Certain aspects of kernel behavior can be modified at runtime, without the 1612need to recompile the kernel, or even to reboot the system. The files in the 1613/proc/sys tree can not only be read, but also modified. You can use the echo 1614command to write value into these files, thereby changing the default settings 1615of the kernel. 1616 1617 1618Chapter 3: Per-process Parameters 1619================================= 1620 16213.1 /proc/<pid>/oom_adj & /proc/<pid>/oom_score_adj- Adjust the oom-killer score 1622-------------------------------------------------------------------------------- 1623 1624These files can be used to adjust the badness heuristic used to select which 1625process gets killed in out of memory (oom) conditions. 1626 1627The badness heuristic assigns a value to each candidate task ranging from 0 1628(never kill) to 1000 (always kill) to determine which process is targeted. The 1629units are roughly a proportion along that range of allowed memory the process 1630may allocate from based on an estimation of its current memory and swap use. 1631For example, if a task is using all allowed memory, its badness score will be 16321000. If it is using half of its allowed memory, its score will be 500. 1633 1634The amount of "allowed" memory depends on the context in which the oom killer 1635was called. If it is due to the memory assigned to the allocating task's cpuset 1636being exhausted, the allowed memory represents the set of mems assigned to that 1637cpuset. If it is due to a mempolicy's node(s) being exhausted, the allowed 1638memory represents the set of mempolicy nodes. If it is due to a memory 1639limit (or swap limit) being reached, the allowed memory is that configured 1640limit. Finally, if it is due to the entire system being out of memory, the 1641allowed memory represents all allocatable resources. 1642 1643The value of /proc/<pid>/oom_score_adj is added to the badness score before it 1644is used to determine which task to kill. Acceptable values range from -1000 1645(OOM_SCORE_ADJ_MIN) to +1000 (OOM_SCORE_ADJ_MAX). This allows userspace to 1646polarize the preference for oom killing either by always preferring a certain 1647task or completely disabling it. The lowest possible value, -1000, is 1648equivalent to disabling oom killing entirely for that task since it will always 1649report a badness score of 0. 1650 1651Consequently, it is very simple for userspace to define the amount of memory to 1652consider for each task. Setting a /proc/<pid>/oom_score_adj value of +500, for 1653example, is roughly equivalent to allowing the remainder of tasks sharing the 1654same system, cpuset, mempolicy, or memory controller resources to use at least 165550% more memory. A value of -500, on the other hand, would be roughly 1656equivalent to discounting 50% of the task's allowed memory from being considered 1657as scoring against the task. 1658 1659For backwards compatibility with previous kernels, /proc/<pid>/oom_adj may also 1660be used to tune the badness score. Its acceptable values range from -16 1661(OOM_ADJUST_MIN) to +15 (OOM_ADJUST_MAX) and a special value of -17 1662(OOM_DISABLE) to disable oom killing entirely for that task. Its value is 1663scaled linearly with /proc/<pid>/oom_score_adj. 1664 1665The value of /proc/<pid>/oom_score_adj may be reduced no lower than the last 1666value set by a CAP_SYS_RESOURCE process. To reduce the value any lower 1667requires CAP_SYS_RESOURCE. 1668 1669 16703.2 /proc/<pid>/oom_score - Display current oom-killer score 1671------------------------------------------------------------- 1672 1673This file can be used to check the current score used by the oom-killer for 1674any given <pid>. Use it together with /proc/<pid>/oom_score_adj to tune which 1675process should be killed in an out-of-memory situation. 1676 1677Please note that the exported value includes oom_score_adj so it is 1678effectively in range [0,2000]. 1679 1680 16813.3 /proc/<pid>/io - Display the IO accounting fields 1682------------------------------------------------------- 1683 1684This file contains IO statistics for each running process. 1685 1686Example 1687~~~~~~~ 1688 1689:: 1690 1691 test:/tmp # dd if=/dev/zero of=/tmp/test.dat & 1692 [1] 3828 1693 1694 test:/tmp # cat /proc/3828/io 1695 rchar: 323934931 1696 wchar: 323929600 1697 syscr: 632687 1698 syscw: 632675 1699 read_bytes: 0 1700 write_bytes: 323932160 1701 cancelled_write_bytes: 0 1702 1703 1704Description 1705~~~~~~~~~~~ 1706 1707rchar 1708^^^^^ 1709 1710I/O counter: chars read 1711The number of bytes which this task has caused to be read from storage. This 1712is simply the sum of bytes which this process passed to read() and pread(). 1713It includes things like tty IO and it is unaffected by whether or not actual 1714physical disk IO was required (the read might have been satisfied from 1715pagecache). 1716 1717 1718wchar 1719^^^^^ 1720 1721I/O counter: chars written 1722The number of bytes which this task has caused, or shall cause to be written 1723to disk. Similar caveats apply here as with rchar. 1724 1725 1726syscr 1727^^^^^ 1728 1729I/O counter: read syscalls 1730Attempt to count the number of read I/O operations, i.e. syscalls like read() 1731and pread(). 1732 1733 1734syscw 1735^^^^^ 1736 1737I/O counter: write syscalls 1738Attempt to count the number of write I/O operations, i.e. syscalls like 1739write() and pwrite(). 1740 1741 1742read_bytes 1743^^^^^^^^^^ 1744 1745I/O counter: bytes read 1746Attempt to count the number of bytes which this process really did cause to 1747be fetched from the storage layer. Done at the submit_bio() level, so it is 1748accurate for block-backed filesystems. <please add status regarding NFS and 1749CIFS at a later time> 1750 1751 1752write_bytes 1753^^^^^^^^^^^ 1754 1755I/O counter: bytes written 1756Attempt to count the number of bytes which this process caused to be sent to 1757the storage layer. This is done at page-dirtying time. 1758 1759 1760cancelled_write_bytes 1761^^^^^^^^^^^^^^^^^^^^^ 1762 1763The big inaccuracy here is truncate. If a process writes 1MB to a file and 1764then deletes the file, it will in fact perform no writeout. But it will have 1765been accounted as having caused 1MB of write. 1766In other words: The number of bytes which this process caused to not happen, 1767by truncating pagecache. A task can cause "negative" IO too. If this task 1768truncates some dirty pagecache, some IO which another task has been accounted 1769for (in its write_bytes) will not be happening. We _could_ just subtract that 1770from the truncating task's write_bytes, but there is information loss in doing 1771that. 1772 1773 1774.. Note:: 1775 1776 At its current implementation state, this is a bit racy on 32-bit machines: 1777 if process A reads process B's /proc/pid/io while process B is updating one 1778 of those 64-bit counters, process A could see an intermediate result. 1779 1780 1781More information about this can be found within the taskstats documentation in 1782Documentation/accounting. 1783 17843.4 /proc/<pid>/coredump_filter - Core dump filtering settings 1785--------------------------------------------------------------- 1786When a process is dumped, all anonymous memory is written to a core file as 1787long as the size of the core file isn't limited. But sometimes we don't want 1788to dump some memory segments, for example, huge shared memory or DAX. 1789Conversely, sometimes we want to save file-backed memory segments into a core 1790file, not only the individual files. 1791 1792/proc/<pid>/coredump_filter allows you to customize which memory segments 1793will be dumped when the <pid> process is dumped. coredump_filter is a bitmask 1794of memory types. If a bit of the bitmask is set, memory segments of the 1795corresponding memory type are dumped, otherwise they are not dumped. 1796 1797The following 9 memory types are supported: 1798 1799 - (bit 0) anonymous private memory 1800 - (bit 1) anonymous shared memory 1801 - (bit 2) file-backed private memory 1802 - (bit 3) file-backed shared memory 1803 - (bit 4) ELF header pages in file-backed private memory areas (it is 1804 effective only if the bit 2 is cleared) 1805 - (bit 5) hugetlb private memory 1806 - (bit 6) hugetlb shared memory 1807 - (bit 7) DAX private memory 1808 - (bit 8) DAX shared memory 1809 1810 Note that MMIO pages such as frame buffer are never dumped and vDSO pages 1811 are always dumped regardless of the bitmask status. 1812 1813 Note that bits 0-4 don't affect hugetlb or DAX memory. hugetlb memory is 1814 only affected by bit 5-6, and DAX is only affected by bits 7-8. 1815 1816The default value of coredump_filter is 0x33; this means all anonymous memory 1817segments, ELF header pages and hugetlb private memory are dumped. 1818 1819If you don't want to dump all shared memory segments attached to pid 1234, 1820write 0x31 to the process's proc file:: 1821 1822 $ echo 0x31 > /proc/1234/coredump_filter 1823 1824When a new process is created, the process inherits the bitmask status from its 1825parent. It is useful to set up coredump_filter before the program runs. 1826For example:: 1827 1828 $ echo 0x7 > /proc/self/coredump_filter 1829 $ ./some_program 1830 18313.5 /proc/<pid>/mountinfo - Information about mounts 1832-------------------------------------------------------- 1833 1834This file contains lines of the form:: 1835 1836 36 35 98:0 /mnt1 /mnt2 rw,noatime master:1 - ext3 /dev/root rw,errors=continue 1837 (1)(2)(3) (4) (5) (6) (n…m) (m+1)(m+2) (m+3) (m+4) 1838 1839 (1) mount ID: unique identifier of the mount (may be reused after umount) 1840 (2) parent ID: ID of parent (or of self for the top of the mount tree) 1841 (3) major:minor: value of st_dev for files on filesystem 1842 (4) root: root of the mount within the filesystem 1843 (5) mount point: mount point relative to the process's root 1844 (6) mount options: per mount options 1845 (n…m) optional fields: zero or more fields of the form "tag[:value]" 1846 (m+1) separator: marks the end of the optional fields 1847 (m+2) filesystem type: name of filesystem of the form "type[.subtype]" 1848 (m+3) mount source: filesystem specific information or "none" 1849 (m+4) super options: per super block options 1850 1851Parsers should ignore all unrecognised optional fields. Currently the 1852possible optional fields are: 1853 1854================ ============================================================== 1855shared:X mount is shared in peer group X 1856master:X mount is slave to peer group X 1857propagate_from:X mount is slave and receives propagation from peer group X [#]_ 1858unbindable mount is unbindable 1859================ ============================================================== 1860 1861.. [#] X is the closest dominant peer group under the process's root. If 1862 X is the immediate master of the mount, or if there's no dominant peer 1863 group under the same root, then only the "master:X" field is present 1864 and not the "propagate_from:X" field. 1865 1866For more information on mount propagation see: 1867 1868 Documentation/filesystems/sharedsubtree.rst 1869 1870 18713.6 /proc/<pid>/comm & /proc/<pid>/task/<tid>/comm 1872-------------------------------------------------------- 1873These files provide a method to access a task's comm value. It also allows for 1874a task to set its own or one of its thread siblings comm value. The comm value 1875is limited in size compared to the cmdline value, so writing anything longer 1876then the kernel's TASK_COMM_LEN (currently 16 chars) will result in a truncated 1877comm value. 1878 1879 18803.7 /proc/<pid>/task/<tid>/children - Information about task children 1881------------------------------------------------------------------------- 1882This file provides a fast way to retrieve first level children pids 1883of a task pointed by <pid>/<tid> pair. The format is a space separated 1884stream of pids. 1885 1886Note the "first level" here -- if a child has its own children they will 1887not be listed here; one needs to read /proc/<children-pid>/task/<tid>/children 1888to obtain the descendants. 1889 1890Since this interface is intended to be fast and cheap it doesn't 1891guarantee to provide precise results and some children might be 1892skipped, especially if they've exited right after we printed their 1893pids, so one needs to either stop or freeze processes being inspected 1894if precise results are needed. 1895 1896 18973.8 /proc/<pid>/fdinfo/<fd> - Information about opened file 1898--------------------------------------------------------------- 1899This file provides information associated with an opened file. The regular 1900files have at least four fields -- 'pos', 'flags', 'mnt_id' and 'ino'. 1901The 'pos' represents the current offset of the opened file in decimal 1902form [see lseek(2) for details], 'flags' denotes the octal O_xxx mask the 1903file has been created with [see open(2) for details] and 'mnt_id' represents 1904mount ID of the file system containing the opened file [see 3.5 1905/proc/<pid>/mountinfo for details]. 'ino' represents the inode number of 1906the file. 1907 1908A typical output is:: 1909 1910 pos: 0 1911 flags: 0100002 1912 mnt_id: 19 1913 ino: 63107 1914 1915All locks associated with a file descriptor are shown in its fdinfo too:: 1916 1917 lock: 1: FLOCK ADVISORY WRITE 359 00:13:11691 0 EOF 1918 1919The files such as eventfd, fsnotify, signalfd, epoll among the regular pos/flags 1920pair provide additional information particular to the objects they represent. 1921 1922Eventfd files 1923~~~~~~~~~~~~~ 1924 1925:: 1926 1927 pos: 0 1928 flags: 04002 1929 mnt_id: 9 1930 ino: 63107 1931 eventfd-count: 5a 1932 1933where 'eventfd-count' is hex value of a counter. 1934 1935Signalfd files 1936~~~~~~~~~~~~~~ 1937 1938:: 1939 1940 pos: 0 1941 flags: 04002 1942 mnt_id: 9 1943 ino: 63107 1944 sigmask: 0000000000000200 1945 1946where 'sigmask' is hex value of the signal mask associated 1947with a file. 1948 1949Epoll files 1950~~~~~~~~~~~ 1951 1952:: 1953 1954 pos: 0 1955 flags: 02 1956 mnt_id: 9 1957 ino: 63107 1958 tfd: 5 events: 1d data: ffffffffffffffff pos:0 ino:61af sdev:7 1959 1960where 'tfd' is a target file descriptor number in decimal form, 1961'events' is events mask being watched and the 'data' is data 1962associated with a target [see epoll(7) for more details]. 1963 1964The 'pos' is current offset of the target file in decimal form 1965[see lseek(2)], 'ino' and 'sdev' are inode and device numbers 1966where target file resides, all in hex format. 1967 1968Fsnotify files 1969~~~~~~~~~~~~~~ 1970For inotify files the format is the following:: 1971 1972 pos: 0 1973 flags: 02000000 1974 mnt_id: 9 1975 ino: 63107 1976 inotify wd:3 ino:9e7e sdev:800013 mask:800afce ignored_mask:0 fhandle-bytes:8 fhandle-type:1 f_handle:7e9e0000640d1b6d 1977 1978where 'wd' is a watch descriptor in decimal form, i.e. a target file 1979descriptor number, 'ino' and 'sdev' are inode and device where the 1980target file resides and the 'mask' is the mask of events, all in hex 1981form [see inotify(7) for more details]. 1982 1983If the kernel was built with exportfs support, the path to the target 1984file is encoded as a file handle. The file handle is provided by three 1985fields 'fhandle-bytes', 'fhandle-type' and 'f_handle', all in hex 1986format. 1987 1988If the kernel is built without exportfs support the file handle won't be 1989printed out. 1990 1991If there is no inotify mark attached yet the 'inotify' line will be omitted. 1992 1993For fanotify files the format is:: 1994 1995 pos: 0 1996 flags: 02 1997 mnt_id: 9 1998 ino: 63107 1999 fanotify flags:10 event-flags:0 2000 fanotify mnt_id:12 mflags:40 mask:38 ignored_mask:40000003 2001 fanotify ino:4f969 sdev:800013 mflags:0 mask:3b ignored_mask:40000000 fhandle-bytes:8 fhandle-type:1 f_handle:69f90400c275b5b4 2002 2003where fanotify 'flags' and 'event-flags' are values used in fanotify_init 2004call, 'mnt_id' is the mount point identifier, 'mflags' is the value of 2005flags associated with mark which are tracked separately from events 2006mask. 'ino' and 'sdev' are target inode and device, 'mask' is the events 2007mask and 'ignored_mask' is the mask of events which are to be ignored. 2008All are in hex format. Incorporation of 'mflags', 'mask' and 'ignored_mask' 2009provide information about flags and mask used in fanotify_mark 2010call [see fsnotify manpage for details]. 2011 2012While the first three lines are mandatory and always printed, the rest is 2013optional and may be omitted if no marks created yet. 2014 2015Timerfd files 2016~~~~~~~~~~~~~ 2017 2018:: 2019 2020 pos: 0 2021 flags: 02 2022 mnt_id: 9 2023 ino: 63107 2024 clockid: 0 2025 ticks: 0 2026 settime flags: 01 2027 it_value: (0, 49406829) 2028 it_interval: (1, 0) 2029 2030where 'clockid' is the clock type and 'ticks' is the number of the timer expirations 2031that have occurred [see timerfd_create(2) for details]. 'settime flags' are 2032flags in octal form been used to setup the timer [see timerfd_settime(2) for 2033details]. 'it_value' is remaining time until the timer expiration. 2034'it_interval' is the interval for the timer. Note the timer might be set up 2035with TIMER_ABSTIME option which will be shown in 'settime flags', but 'it_value' 2036still exhibits timer's remaining time. 2037 2038DMA Buffer files 2039~~~~~~~~~~~~~~~~ 2040 2041:: 2042 2043 pos: 0 2044 flags: 04002 2045 mnt_id: 9 2046 ino: 63107 2047 size: 32768 2048 count: 2 2049 exp_name: system-heap 2050 2051where 'size' is the size of the DMA buffer in bytes. 'count' is the file count of 2052the DMA buffer file. 'exp_name' is the name of the DMA buffer exporter. 2053 20543.9 /proc/<pid>/map_files - Information about memory mapped files 2055--------------------------------------------------------------------- 2056This directory contains symbolic links which represent memory mapped files 2057the process is maintaining. Example output:: 2058 2059 | lr-------- 1 root root 64 Jan 27 11:24 333c600000-333c620000 -> /usr/lib64/ld-2.18.so 2060 | lr-------- 1 root root 64 Jan 27 11:24 333c81f000-333c820000 -> /usr/lib64/ld-2.18.so 2061 | lr-------- 1 root root 64 Jan 27 11:24 333c820000-333c821000 -> /usr/lib64/ld-2.18.so 2062 | ... 2063 | lr-------- 1 root root 64 Jan 27 11:24 35d0421000-35d0422000 -> /usr/lib64/libselinux.so.1 2064 | lr-------- 1 root root 64 Jan 27 11:24 400000-41a000 -> /usr/bin/ls 2065 2066The name of a link represents the virtual memory bounds of a mapping, i.e. 2067vm_area_struct::vm_start-vm_area_struct::vm_end. 2068 2069The main purpose of the map_files is to retrieve a set of memory mapped 2070files in a fast way instead of parsing /proc/<pid>/maps or 2071/proc/<pid>/smaps, both of which contain many more records. At the same 2072time one can open(2) mappings from the listings of two processes and 2073comparing their inode numbers to figure out which anonymous memory areas 2074are actually shared. 2075 20763.10 /proc/<pid>/timerslack_ns - Task timerslack value 2077--------------------------------------------------------- 2078This file provides the value of the task's timerslack value in nanoseconds. 2079This value specifies an amount of time that normal timers may be deferred 2080in order to coalesce timers and avoid unnecessary wakeups. 2081 2082This allows a task's interactivity vs power consumption tradeoff to be 2083adjusted. 2084 2085Writing 0 to the file will set the task's timerslack to the default value. 2086 2087Valid values are from 0 - ULLONG_MAX 2088 2089An application setting the value must have PTRACE_MODE_ATTACH_FSCREDS level 2090permissions on the task specified to change its timerslack_ns value. 2091 20923.11 /proc/<pid>/patch_state - Livepatch patch operation state 2093----------------------------------------------------------------- 2094When CONFIG_LIVEPATCH is enabled, this file displays the value of the 2095patch state for the task. 2096 2097A value of '-1' indicates that no patch is in transition. 2098 2099A value of '0' indicates that a patch is in transition and the task is 2100unpatched. If the patch is being enabled, then the task hasn't been 2101patched yet. If the patch is being disabled, then the task has already 2102been unpatched. 2103 2104A value of '1' indicates that a patch is in transition and the task is 2105patched. If the patch is being enabled, then the task has already been 2106patched. If the patch is being disabled, then the task hasn't been 2107unpatched yet. 2108 21093.12 /proc/<pid>/arch_status - task architecture specific status 2110------------------------------------------------------------------- 2111When CONFIG_PROC_PID_ARCH_STATUS is enabled, this file displays the 2112architecture specific status of the task. 2113 2114Example 2115~~~~~~~ 2116 2117:: 2118 2119 $ cat /proc/6753/arch_status 2120 AVX512_elapsed_ms: 8 2121 2122Description 2123~~~~~~~~~~~ 2124 2125x86 specific entries 2126~~~~~~~~~~~~~~~~~~~~~ 2127 2128AVX512_elapsed_ms 2129^^^^^^^^^^^^^^^^^^ 2130 2131 If AVX512 is supported on the machine, this entry shows the milliseconds 2132 elapsed since the last time AVX512 usage was recorded. The recording 2133 happens on a best effort basis when a task is scheduled out. This means 2134 that the value depends on two factors: 2135 2136 1) The time which the task spent on the CPU without being scheduled 2137 out. With CPU isolation and a single runnable task this can take 2138 several seconds. 2139 2140 2) The time since the task was scheduled out last. Depending on the 2141 reason for being scheduled out (time slice exhausted, syscall ...) 2142 this can be arbitrary long time. 2143 2144 As a consequence the value cannot be considered precise and authoritative 2145 information. The application which uses this information has to be aware 2146 of the overall scenario on the system in order to determine whether a 2147 task is a real AVX512 user or not. Precise information can be obtained 2148 with performance counters. 2149 2150 A special value of '-1' indicates that no AVX512 usage was recorded, thus 2151 the task is unlikely an AVX512 user, but depends on the workload and the 2152 scheduling scenario, it also could be a false negative mentioned above. 2153 2154Chapter 4: Configuring procfs 2155============================= 2156 21574.1 Mount options 2158--------------------- 2159 2160The following mount options are supported: 2161 2162 ========= ======================================================== 2163 hidepid= Set /proc/<pid>/ access mode. 2164 gid= Set the group authorized to learn processes information. 2165 subset= Show only the specified subset of procfs. 2166 ========= ======================================================== 2167 2168hidepid=off or hidepid=0 means classic mode - everybody may access all 2169/proc/<pid>/ directories (default). 2170 2171hidepid=noaccess or hidepid=1 means users may not access any /proc/<pid>/ 2172directories but their own. Sensitive files like cmdline, sched*, status are now 2173protected against other users. This makes it impossible to learn whether any 2174user runs specific program (given the program doesn't reveal itself by its 2175behaviour). As an additional bonus, as /proc/<pid>/cmdline is unaccessible for 2176other users, poorly written programs passing sensitive information via program 2177arguments are now protected against local eavesdroppers. 2178 2179hidepid=invisible or hidepid=2 means hidepid=1 plus all /proc/<pid>/ will be 2180fully invisible to other users. It doesn't mean that it hides a fact whether a 2181process with a specific pid value exists (it can be learned by other means, e.g. 2182by "kill -0 $PID"), but it hides process' uid and gid, which may be learned by 2183stat()'ing /proc/<pid>/ otherwise. It greatly complicates an intruder's task of 2184gathering information about running processes, whether some daemon runs with 2185elevated privileges, whether other user runs some sensitive program, whether 2186other users run any program at all, etc. 2187 2188hidepid=ptraceable or hidepid=4 means that procfs should only contain 2189/proc/<pid>/ directories that the caller can ptrace. 2190 2191gid= defines a group authorized to learn processes information otherwise 2192prohibited by hidepid=. If you use some daemon like identd which needs to learn 2193information about processes information, just add identd to this group. 2194 2195subset=pid hides all top level files and directories in the procfs that 2196are not related to tasks. 2197 2198Chapter 5: Filesystem behavior 2199============================== 2200 2201Originally, before the advent of pid namepsace, procfs was a global file 2202system. It means that there was only one procfs instance in the system. 2203 2204When pid namespace was added, a separate procfs instance was mounted in 2205each pid namespace. So, procfs mount options are global among all 2206mountpoints within the same namespace:: 2207 2208 # grep ^proc /proc/mounts 2209 proc /proc proc rw,relatime,hidepid=2 0 0 2210 2211 # strace -e mount mount -o hidepid=1 -t proc proc /tmp/proc 2212 mount("proc", "/tmp/proc", "proc", 0, "hidepid=1") = 0 2213 +++ exited with 0 +++ 2214 2215 # grep ^proc /proc/mounts 2216 proc /proc proc rw,relatime,hidepid=2 0 0 2217 proc /tmp/proc proc rw,relatime,hidepid=2 0 0 2218 2219and only after remounting procfs mount options will change at all 2220mountpoints:: 2221 2222 # mount -o remount,hidepid=1 -t proc proc /tmp/proc 2223 2224 # grep ^proc /proc/mounts 2225 proc /proc proc rw,relatime,hidepid=1 0 0 2226 proc /tmp/proc proc rw,relatime,hidepid=1 0 0 2227 2228This behavior is different from the behavior of other filesystems. 2229 2230The new procfs behavior is more like other filesystems. Each procfs mount 2231creates a new procfs instance. Mount options affect own procfs instance. 2232It means that it became possible to have several procfs instances 2233displaying tasks with different filtering options in one pid namespace:: 2234 2235 # mount -o hidepid=invisible -t proc proc /proc 2236 # mount -o hidepid=noaccess -t proc proc /tmp/proc 2237 # grep ^proc /proc/mounts 2238 proc /proc proc rw,relatime,hidepid=invisible 0 0 2239 proc /tmp/proc proc rw,relatime,hidepid=noaccess 0 0 2240