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 Networking info in /proc/net 27 1.4 SCSI info 28 1.5 Parallel port info in /proc/parport 29 1.6 TTY info in /proc/tty 30 1.7 Miscellaneous kernel statistics in /proc/stat 31 1.8 Ext4 file system parameters 32 1.9 /proc/consoles - Shows registered system consoles 33 34 2 Modifying System Parameters 35 36 3 Per-Process Parameters 37 3.1 /proc/<pid>/oom_adj & /proc/<pid>/oom_score_adj - Adjust the oom-killer 38 score 39 3.2 /proc/<pid>/oom_score - Display current oom-killer score 40 3.3 /proc/<pid>/io - Display the IO accounting fields 41 3.4 /proc/<pid>/coredump_filter - Core dump filtering settings 42 3.5 /proc/<pid>/mountinfo - Information about mounts 43 3.6 /proc/<pid>/comm & /proc/<pid>/task/<tid>/comm 44 3.7 /proc/<pid>/task/<tid>/children - Information about task children 45 3.8 /proc/<pid>/fdinfo/<fd> - Information about opened file 46 3.9 /proc/<pid>/map_files - Information about memory mapped files 47 3.10 /proc/<pid>/timerslack_ns - Task timerslack value 48 3.11 /proc/<pid>/patch_state - Livepatch patch operation state 49 3.12 /proc/<pid>/arch_status - Task architecture specific information 50 3.13 /proc/<pid>/fd - List of symlinks to open files 51 3.14 /proc/<pid>/ksm_stat - Information about the process's ksm status. 52 53 4 Configuring procfs 54 4.1 Mount options 55 4.2 Mount restrictions 56 57 5 Filesystem behavior 58 59Preface 60======= 61 620.1 Introduction/Credits 63------------------------ 64 65We'd like to thank Alan Cox, Rik van Riel, and Alexey Kuznetsov and a lot of 66other people for help compiling this documentation. We'd also like to extend a 67special thank you to Andi Kleen for documentation, which we relied on heavily 68to create this document, as well as the additional information he provided. 69Thanks to everybody else who contributed source or docs to the Linux kernel 70and helped create a great piece of software... :) 71 72The latest version of this document is available online at 73https://www.kernel.org/doc/html/latest/filesystems/proc.html 74 750.2 Legal Stuff 76--------------- 77 78We don't guarantee the correctness of this document, and if you come to us 79complaining about how you screwed up your system because of incorrect 80documentation, we won't feel responsible... 81 82Chapter 1: Collecting System Information 83======================================== 84 85In This Chapter 86--------------- 87* Investigating the properties of the pseudo file system /proc and its 88 ability to provide information on the running Linux system 89* Examining /proc's structure 90* Uncovering various information about the kernel and the processes running 91 on the system 92 93------------------------------------------------------------------------------ 94 95The proc file system acts as an interface to internal data structures in the 96kernel. It can be used to obtain information about the system and to change 97certain kernel parameters at runtime (sysctl). 98 99First, we'll take a look at the read-only parts of /proc. In Chapter 2, we 100show you how you can use /proc/sys to change settings. 101 1021.1 Process-Specific Subdirectories 103----------------------------------- 104 105The directory /proc contains (among other things) one subdirectory for each 106process running on the system, which is named after the process ID (PID). 107 108The link 'self' points to the process reading the file system. Each process 109subdirectory has the entries listed in Table 1-1. 110 111A process can read its own information from /proc/PID/* with no extra 112permissions. When reading /proc/PID/* information for other processes, reading 113process is required to have either CAP_SYS_PTRACE capability with 114PTRACE_MODE_READ access permissions, or, alternatively, CAP_PERFMON 115capability. This applies to all read-only information like `maps`, `environ`, 116`pagemap`, etc. The only exception is `mem` file due to its read-write nature, 117which requires CAP_SYS_PTRACE capabilities with more elevated 118PTRACE_MODE_ATTACH permissions; CAP_PERFMON capability does not grant access 119to /proc/PID/mem for other processes. 120 121Note that an open file descriptor to /proc/<pid> or to any of its 122contained files or subdirectories does not prevent <pid> from being reused 123for some other process in the event that <pid> exits. Operations on 124open /proc/<pid> file descriptors corresponding to dead processes 125never act on any new process that the kernel may, through chance, have 126also assigned the process ID <pid>. Instead, operations on these FDs 127usually fail with ESRCH. 128 129.. table:: Table 1-1: Process specific entries in /proc 130 131 ============= =============================================================== 132 File Content 133 ============= =============================================================== 134 clear_refs Clears page referenced bits shown in smaps output 135 cmdline Command line arguments 136 cpu Current and last cpu in which it was executed (2.4)(smp) 137 cwd Link to the current working directory 138 environ Values of environment variables 139 exe Link to the executable of this process 140 fd Directory, which contains all file descriptors 141 maps Memory maps to executables and library files (2.4) 142 mem Memory held by this process 143 root Link to the root directory of this process 144 stat Process status 145 statm Process memory status information 146 status Process status in human readable form 147 wchan Present with CONFIG_KALLSYMS=y: it shows the kernel function 148 symbol the task is blocked in - or "0" if not blocked. 149 pagemap Page table 150 stack Report full stack trace, enable via CONFIG_STACKTRACE 151 smaps An extension based on maps, showing the memory consumption of 152 each mapping and flags associated with it 153 smaps_rollup Accumulated smaps stats for all mappings of the process. This 154 can be derived from smaps, but is faster and more convenient 155 numa_maps An extension based on maps, showing the memory locality and 156 binding policy as well as mem usage (in pages) of each mapping. 157 ============= =============================================================== 158 159For example, to get the status information of a process, all you have to do is 160read the file /proc/PID/status:: 161 162 >cat /proc/self/status 163 Name: cat 164 State: R (running) 165 Tgid: 5452 166 Pid: 5452 167 PPid: 743 168 TracerPid: 0 (2.4) 169 Uid: 501 501 501 501 170 Gid: 100 100 100 100 171 FDSize: 256 172 Groups: 100 14 16 173 Kthread: 0 174 VmPeak: 5004 kB 175 VmSize: 5004 kB 176 VmLck: 0 kB 177 VmHWM: 476 kB 178 VmRSS: 476 kB 179 RssAnon: 352 kB 180 RssFile: 120 kB 181 RssShmem: 4 kB 182 VmData: 156 kB 183 VmStk: 88 kB 184 VmExe: 68 kB 185 VmLib: 1412 kB 186 VmPTE: 20 kb 187 VmSwap: 0 kB 188 HugetlbPages: 0 kB 189 CoreDumping: 0 190 THP_enabled: 1 191 Threads: 1 192 SigQ: 0/28578 193 SigPnd: 0000000000000000 194 ShdPnd: 0000000000000000 195 SigBlk: 0000000000000000 196 SigIgn: 0000000000000000 197 SigCgt: 0000000000000000 198 CapInh: 00000000fffffeff 199 CapPrm: 0000000000000000 200 CapEff: 0000000000000000 201 CapBnd: ffffffffffffffff 202 CapAmb: 0000000000000000 203 NoNewPrivs: 0 204 Seccomp: 0 205 Speculation_Store_Bypass: thread vulnerable 206 SpeculationIndirectBranch: conditional enabled 207 voluntary_ctxt_switches: 0 208 nonvoluntary_ctxt_switches: 1 209 210This shows you nearly the same information you would get if you viewed it with 211the ps command. In fact, ps uses the proc file system to obtain its 212information. But you get a more detailed view of the process by reading the 213file /proc/PID/status. It fields are described in table 1-2. 214 215The statm file contains more detailed information about the process 216memory usage. Its seven fields are explained in Table 1-3. The stat file 217contains detailed information about the process itself. Its fields are 218explained in Table 1-4. 219 220(for SMP CONFIG users) 221 222For making accounting scalable, RSS related information are handled in an 223asynchronous manner and the value may not be very precise. To see a precise 224snapshot of a moment, you can see /proc/<pid>/smaps file and scan page table. 225It's slow but very precise. 226 227.. table:: Table 1-2: Contents of the status fields (as of 4.19) 228 229 ========================== =================================================== 230 Field Content 231 ========================== =================================================== 232 Name filename of the executable 233 Umask file mode creation mask 234 State state (R is running, S is sleeping, D is sleeping 235 in an uninterruptible wait, Z is zombie, 236 T is traced or stopped) 237 Tgid thread group ID 238 Ngid NUMA group ID (0 if none) 239 Pid process id 240 PPid process id of the parent process 241 TracerPid PID of process tracing this process (0 if not, or 242 the tracer is outside of the current pid namespace) 243 Uid Real, effective, saved set, and file system UIDs 244 Gid Real, effective, saved set, and file system GIDs 245 FDSize number of file descriptor slots currently allocated 246 Groups supplementary group list 247 NStgid descendant namespace thread group ID hierarchy 248 NSpid descendant namespace process ID hierarchy 249 NSpgid descendant namespace process group ID hierarchy 250 NSsid descendant namespace session ID hierarchy 251 Kthread kernel thread flag, 1 is yes, 0 is no 252 VmPeak peak virtual memory size 253 VmSize total program size 254 VmLck locked memory size 255 VmPin pinned memory size 256 VmHWM peak resident set size ("high water mark") 257 VmRSS size of memory portions. It contains the three 258 following parts 259 (VmRSS = RssAnon + RssFile + RssShmem) 260 RssAnon size of resident anonymous memory 261 RssFile size of resident file mappings 262 RssShmem size of resident shmem memory (includes SysV shm, 263 mapping of tmpfs and shared anonymous mappings) 264 VmData size of private data segments 265 VmStk size of stack segments 266 VmExe size of text segment 267 VmLib size of shared library code 268 VmPTE size of page table entries 269 VmSwap amount of swap used by anonymous private data 270 (shmem swap usage is not included) 271 HugetlbPages size of hugetlb memory portions 272 CoreDumping process's memory is currently being dumped 273 (killing the process may lead to a corrupted core) 274 THP_enabled process is allowed to use THP (returns 0 when 275 PR_SET_THP_DISABLE is set on the process to disable 276 THP completely, not just partially) 277 Threads number of threads 278 SigQ number of signals queued/max. number for queue 279 SigPnd bitmap of pending signals for the thread 280 ShdPnd bitmap of shared pending signals for the process 281 SigBlk bitmap of blocked signals 282 SigIgn bitmap of ignored signals 283 SigCgt bitmap of caught signals 284 CapInh bitmap of inheritable capabilities 285 CapPrm bitmap of permitted capabilities 286 CapEff bitmap of effective capabilities 287 CapBnd bitmap of capabilities bounding set 288 CapAmb bitmap of ambient capabilities 289 NoNewPrivs no_new_privs, like prctl(PR_GET_NO_NEW_PRIV, ...) 290 Seccomp seccomp mode, like prctl(PR_GET_SECCOMP, ...) 291 Speculation_Store_Bypass speculative store bypass mitigation status 292 SpeculationIndirectBranch indirect branch speculation mode 293 Cpus_allowed mask of CPUs on which this process may run 294 Cpus_allowed_list Same as previous, but in "list format" 295 Mems_allowed mask of memory nodes allowed to this process 296 Mems_allowed_list Same as previous, but in "list format" 297 voluntary_ctxt_switches number of voluntary context switches 298 nonvoluntary_ctxt_switches number of non voluntary context switches 299 ========================== =================================================== 300 301 302.. table:: Table 1-3: Contents of the statm fields (as of 2.6.8-rc3) 303 304 ======== =============================== ============================== 305 Field Content 306 ======== =============================== ============================== 307 size total program size (pages) (same as VmSize in status) 308 resident size of memory portions (pages) (same as VmRSS in status) 309 shared number of pages that are shared (i.e. backed by a file, same 310 as RssFile+RssShmem in status) 311 trs number of pages that are 'code' (not including libs; broken, 312 includes data segment) 313 lrs number of pages of library (always 0 on 2.6) 314 drs number of pages of data/stack (including libs; broken, 315 includes library text) 316 dt number of dirty pages (always 0 on 2.6) 317 ======== =============================== ============================== 318 319 320.. table:: Table 1-4: Contents of the stat fields (as of 2.6.30-rc7) 321 322 ============= =============================================================== 323 Field Content 324 ============= =============================================================== 325 pid process id 326 tcomm filename of the executable 327 state state (R is running, S is sleeping, D is sleeping in an 328 uninterruptible wait, Z is zombie, T is traced or stopped) 329 ppid process id of the parent process 330 pgrp pgrp of the process 331 sid session id 332 tty_nr tty the process uses 333 tty_pgrp pgrp of the tty 334 flags task flags 335 min_flt number of minor faults 336 cmin_flt number of minor faults with child's 337 maj_flt number of major faults 338 cmaj_flt number of major faults with child's 339 utime user mode jiffies 340 stime kernel mode jiffies 341 cutime user mode jiffies with child's 342 cstime kernel mode jiffies with child's 343 priority priority level 344 nice nice level 345 num_threads number of threads 346 it_real_value (obsolete, always 0) 347 start_time time the process started after system boot 348 vsize virtual memory size 349 rss resident set memory size 350 rsslim current limit in bytes on the rss 351 start_code address above which program text can run 352 end_code address below which program text can run 353 start_stack address of the start of the main process stack 354 esp current value of ESP 355 eip current value of EIP 356 pending bitmap of pending signals 357 blocked bitmap of blocked signals 358 sigign bitmap of ignored signals 359 sigcatch bitmap of caught signals 360 0 (place holder, used to be the wchan address, 361 use /proc/PID/wchan instead) 362 0 (place holder) 363 0 (place holder) 364 exit_signal signal to send to parent thread on exit 365 task_cpu which CPU the task is scheduled on 366 rt_priority realtime priority 367 policy scheduling policy (man sched_setscheduler) 368 blkio_ticks time spent waiting for block IO 369 gtime guest time of the task in jiffies 370 cgtime guest time of the task children in jiffies 371 start_data address above which program data+bss is placed 372 end_data address below which program data+bss is placed 373 start_brk address above which program heap can be expanded with brk() 374 arg_start address above which program command line is placed 375 arg_end address below which program command line is placed 376 env_start address above which program environment is placed 377 env_end address below which program environment is placed 378 exit_code the thread's exit_code in the form reported by the waitpid 379 system call 380 ============= =============================================================== 381 382The /proc/PID/maps file contains the currently mapped memory regions and 383their access permissions. 384 385The format is:: 386 387 address perms offset dev inode pathname 388 389 08048000-08049000 r-xp 00000000 03:00 8312 /opt/test 390 08049000-0804a000 rw-p 00001000 03:00 8312 /opt/test 391 0804a000-0806b000 rw-p 00000000 00:00 0 [heap] 392 a7cb1000-a7cb2000 ---p 00000000 00:00 0 393 a7cb2000-a7eb2000 rw-p 00000000 00:00 0 394 a7eb2000-a7eb3000 ---p 00000000 00:00 0 395 a7eb3000-a7ed5000 rw-p 00000000 00:00 0 396 a7ed5000-a8008000 r-xp 00000000 03:00 4222 /lib/libc.so.6 397 a8008000-a800a000 r--p 00133000 03:00 4222 /lib/libc.so.6 398 a800a000-a800b000 rw-p 00135000 03:00 4222 /lib/libc.so.6 399 a800b000-a800e000 rw-p 00000000 00:00 0 400 a800e000-a8022000 r-xp 00000000 03:00 14462 /lib/libpthread.so.0 401 a8022000-a8023000 r--p 00013000 03:00 14462 /lib/libpthread.so.0 402 a8023000-a8024000 rw-p 00014000 03:00 14462 /lib/libpthread.so.0 403 a8024000-a8027000 rw-p 00000000 00:00 0 404 a8027000-a8043000 r-xp 00000000 03:00 8317 /lib/ld-linux.so.2 405 a8043000-a8044000 r--p 0001b000 03:00 8317 /lib/ld-linux.so.2 406 a8044000-a8045000 rw-p 0001c000 03:00 8317 /lib/ld-linux.so.2 407 aff35000-aff4a000 rw-p 00000000 00:00 0 [stack] 408 ffffe000-fffff000 r-xp 00000000 00:00 0 [vdso] 409 410where "address" is the address space in the process that it occupies, "perms" 411is a set of permissions:: 412 413 r = read 414 w = write 415 x = execute 416 s = shared 417 p = private (copy on write) 418 419"offset" is the offset into the mapping, "dev" is the device (major:minor), and 420"inode" is the inode on that device. 0 indicates that no inode is associated 421with the memory region, as the case would be with BSS (uninitialized data). 422The "pathname" shows the name associated file for this mapping. If the mapping 423is not associated with a file: 424 425 =================== =========================================== 426 [heap] the heap of the program 427 [stack] the stack of the main process 428 [vdso] the "virtual dynamic shared object", 429 the kernel system call handler 430 [anon:<name>] a private anonymous mapping that has been 431 named by userspace 432 [anon_shmem:<name>] an anonymous shared memory mapping that has 433 been named by userspace 434 =================== =========================================== 435 436 or if empty, the mapping is anonymous. 437 438Starting with 6.11 kernel, /proc/PID/maps provides an alternative 439ioctl()-based API that gives ability to flexibly and efficiently query and 440filter individual VMAs. This interface is binary and is meant for more 441efficient and easy programmatic use. `struct procmap_query`, defined in 442linux/fs.h UAPI header, serves as an input/output argument to the 443`PROCMAP_QUERY` ioctl() command. See comments in linus/fs.h UAPI header for 444details on query semantics, supported flags, data returned, and general API 445usage information. 446 447The /proc/PID/smaps is an extension based on maps, showing the memory 448consumption for each of the process's mappings. For each mapping (aka Virtual 449Memory Area, or VMA) there is a series of lines such as the following:: 450 451 08048000-080bc000 r-xp 00000000 03:02 13130 /bin/bash 452 453 Size: 1084 kB 454 KernelPageSize: 4 kB 455 MMUPageSize: 4 kB 456 Rss: 892 kB 457 Pss: 374 kB 458 Pss_Dirty: 0 kB 459 Shared_Clean: 892 kB 460 Shared_Dirty: 0 kB 461 Private_Clean: 0 kB 462 Private_Dirty: 0 kB 463 Referenced: 892 kB 464 Anonymous: 0 kB 465 KSM: 0 kB 466 LazyFree: 0 kB 467 AnonHugePages: 0 kB 468 FilePmdMapped: 0 kB 469 ShmemPmdMapped: 0 kB 470 Shared_Hugetlb: 0 kB 471 Private_Hugetlb: 0 kB 472 Swap: 0 kB 473 SwapPss: 0 kB 474 Locked: 0 kB 475 THPeligible: 0 476 VmFlags: rd ex mr mw me dw 477 478The first of these lines shows the same information as is displayed for 479the mapping in /proc/PID/maps. Following lines show the size of the 480mapping (size); the smallest possible page size allocated when backing a 481VMA (KernelPageSize), which is the granularity in which VMA modifications 482can be performed; the smallest possible page size that could be used by the 483MMU (MMUPageSize) when backing a VMA; the amount of the mapping that is 484currently resident in RAM (RSS); the process's proportional share of this 485mapping (PSS); and the number of clean and dirty shared and private pages 486in the mapping. 487 488"KernelPageSize" always corresponds to "MMUPageSize", except when a larger 489kernel page size is emulated on a system with a smaller page size used by the 490MMU, which is the case for some PPC64 setups with hugetlb. Furthermore, 491"KernelPageSize" and "MMUPageSize" always correspond to the smallest 492possible granularity (fallback) that can be encountered in a VMA throughout 493its lifetime. These values are not affected by Transparent Huge Pages 494being in effect, or any usage of larger MMU page sizes (either through 495architectural huge-page mappings or other explicit/implicit coalescing of 496virtual ranges performed by the MMU). "AnonHugePages", "ShmemPmdMapped" and 497"FilePmdMapped" provide insight into the usage of PMD-level architectural 498huge-page mappings. 499 500The "proportional set size" (PSS) of a process is the count of pages it has 501in memory, where each page is divided by the number of processes sharing it. 502So if a process has 1000 pages all to itself, and 1000 shared with one other 503process, its PSS will be 1500. "Pss_Dirty" is the portion of PSS which 504consists of dirty pages. ("Pss_Clean" is not included, but it can be 505calculated by subtracting "Pss_Dirty" from "Pss".) 506 507Traditionally, a page is accounted as "private" if it is mapped exactly once, 508and a page is accounted as "shared" when mapped multiple times, even when 509mapped in the same process multiple times. Note that this accounting is 510independent of MAP_SHARED. 511 512In some kernel configurations, the semantics of pages part of a larger 513allocation (e.g., THP) can differ: a page is accounted as "private" if all 514pages part of the corresponding large allocation are *certainly* mapped in the 515same process, even if the page is mapped multiple times in that process. A 516page is accounted as "shared" if any page of the larger allocation 517is *maybe* mapped in a different process. In some cases, a large allocation 518might be treated as "maybe mapped by multiple processes" even though this 519is no longer the case. 520 521Some kernel configurations do not track the precise number of times a page part 522of a larger allocation is mapped. In this case, when calculating the PSS, the 523average number of mappings per page in this larger allocation might be used 524as an approximation for the number of mappings of a page. The PSS calculation 525will be imprecise in this case. 526 527"Referenced" indicates the amount of memory currently marked as referenced or 528accessed. 529 530"Anonymous" shows the amount of memory that does not belong to any file. Even 531a mapping associated with a file may contain anonymous pages: when MAP_PRIVATE 532and a page is modified, the file page is replaced by a private anonymous copy. 533 534"KSM" reports how many of the pages are KSM pages. Note that KSM-placed zeropages 535are not included, only actual KSM pages. 536 537"LazyFree" shows the amount of memory which is marked by madvise(MADV_FREE). 538The memory isn't freed immediately with madvise(). It's freed in memory 539pressure if the memory is clean. Please note that the printed value might 540be lower than the real value due to optimizations used in the current 541implementation. If this is not desirable please file a bug report. 542 543"AnonHugePages", "ShmemPmdMapped" and "FilePmdMapped" show the amount of 544memory backed by Transparent Huge Pages that are currently mapped by 545architectural huge-page mappings at the PMD level. "AnonHugePages" 546corresponds to memory that does not belong to a file, "ShmemPmdMapped" to 547shared memory (shmem/tmpfs) and "FilePmdMapped" to file-backed memory 548(excluding shmem/tmpfs). 549 550There are no dedicated entries for Transparent Huge Pages (or similar concepts) 551that are not mapped by architectural huge-page mappings at the PMD level. 552 553"Shared_Hugetlb" and "Private_Hugetlb" show the amounts of memory backed by 554hugetlbfs page which is *not* counted in "RSS" or "PSS" field for historical 555reasons. And these are not included in {Shared,Private}_{Clean,Dirty} field. 556 557"Swap" shows how much would-be-anonymous memory is also used, but out on swap. 558 559For shmem mappings, "Swap" includes also the size of the mapped (and not 560replaced by copy-on-write) part of the underlying shmem object out on swap. 561"SwapPss" shows proportional swap share of this mapping. Unlike "Swap", this 562does not take into account swapped out page of underlying shmem objects. 563"Locked" indicates whether the mapping is locked in memory or not. 564 565"THPeligible" indicates whether the mapping is eligible for allocating 566naturally aligned THP pages of any currently enabled size. 1 if true, 0 567otherwise. 568 569If both the kernel and the CPU support protection keys (pkeys), 570"ProtectionKey" indicates the memory protection key associated with the 571virtual memory area. 572 573"VmFlags" field deserves a separate description. This member represents the 574kernel flags associated with the particular virtual memory area in two letter 575encoded manner. The codes are the following: 576 577 == ============================================================= 578 rd readable 579 wr writeable 580 ex executable 581 sh shared 582 mr may read 583 mw may write 584 me may execute 585 ms may share 586 gd stack segment growns down 587 pf pure PFN range 588 lo pages are locked in memory 589 io memory mapped I/O area 590 sr sequential read advise provided 591 rr random read advise provided 592 dc do not copy area on fork 593 de do not expand area on remapping 594 ac area is accountable 595 nr swap space is not reserved for the area 596 ht area uses huge tlb pages 597 sf synchronous page fault 598 ar architecture specific flag 599 wf wipe on fork 600 dd do not include area into core dump 601 sd soft dirty flag 602 mm mixed map area 603 hg huge page advise flag 604 nh no huge page advise flag 605 mg mergeable advise flag 606 bt arm64 BTI guarded page 607 mt arm64 MTE allocation tags are enabled 608 um userfaultfd missing tracking 609 uw userfaultfd wr-protect tracking 610 ui userfaultfd minor fault 611 ur userfaultfd read-write-protect tracking 612 ss shadow/guarded control stack page 613 sl sealed 614 lf lock on fault pages 615 dp always lazily freeable mapping 616 gu maybe contains guard regions (if not set, definitely doesn't) 617 == ============================================================= 618 619Note that there is no guarantee that every flag and associated mnemonic will 620be present in all further kernel releases. Things get changed, the flags may 621be vanished or the reverse -- new added. Interpretation of their meaning 622might change in future as well. So each consumer of these flags has to 623follow each specific kernel version for the exact semantic. 624 625This file is only present if the CONFIG_MMU kernel configuration option is 626enabled. 627 628Note: reading /proc/PID/maps or /proc/PID/smaps is inherently racy (consistent 629output can be achieved only in the single read call). 630 631This typically manifests when doing partial reads of these files while the 632memory map is being modified. Despite the races, we do provide the following 633guarantees: 634 6351) The mapped addresses never go backwards, which implies no two 636 regions will ever overlap. 6372) If there is something at a given vaddr during the entirety of the 638 life of the smaps/maps walk, there will be some output for it. 639 640The /proc/PID/smaps_rollup file includes the same fields as /proc/PID/smaps, 641but their values are the sums of the corresponding values for all mappings of 642the process. Additionally, it contains these fields: 643 644- Pss_Anon 645- Pss_File 646- Pss_Shmem 647 648They represent the proportional shares of anonymous, file, and shmem pages, as 649described for smaps above. These fields are omitted in smaps since each 650mapping identifies the type (anon, file, or shmem) of all pages it contains. 651Thus all information in smaps_rollup can be derived from smaps, but at a 652significantly higher cost. 653 654The /proc/PID/clear_refs is used to reset the PG_Referenced and ACCESSED/YOUNG 655bits on both physical and virtual pages associated with a process, and the 656soft-dirty bit on pte (see Documentation/admin-guide/mm/soft-dirty.rst 657for details). 658To clear the bits for all the pages associated with the process:: 659 660 > echo 1 > /proc/PID/clear_refs 661 662To clear the bits for the anonymous pages associated with the process:: 663 664 > echo 2 > /proc/PID/clear_refs 665 666To clear the bits for the file mapped pages associated with the process:: 667 668 > echo 3 > /proc/PID/clear_refs 669 670To clear the soft-dirty bit:: 671 672 > echo 4 > /proc/PID/clear_refs 673 674To reset the peak resident set size ("high water mark") to the process's 675current value:: 676 677 > echo 5 > /proc/PID/clear_refs 678 679Any other value written to /proc/PID/clear_refs will have no effect. 680 681The /proc/pid/pagemap gives the PFN, which can be used to find the pageflags 682using /proc/kpageflags and number of times a page is mapped using 683/proc/kpagecount. For detailed explanation, see 684Documentation/admin-guide/mm/pagemap.rst. 685 686The /proc/pid/numa_maps is an extension based on maps, showing the memory 687locality and binding policy, as well as the memory usage (in pages) of 688each mapping. The output follows a general format where mapping details get 689summarized separated by blank spaces, one mapping per each file line:: 690 691 address policy mapping details 692 693 00400000 default file=/usr/local/bin/app mapped=1 active=0 N3=1 kernelpagesize_kB=4 694 00600000 default file=/usr/local/bin/app anon=1 dirty=1 N3=1 kernelpagesize_kB=4 695 3206000000 default file=/lib64/ld-2.12.so mapped=26 mapmax=6 N0=24 N3=2 kernelpagesize_kB=4 696 320621f000 default file=/lib64/ld-2.12.so anon=1 dirty=1 N3=1 kernelpagesize_kB=4 697 3206220000 default file=/lib64/ld-2.12.so anon=1 dirty=1 N3=1 kernelpagesize_kB=4 698 3206221000 default anon=1 dirty=1 N3=1 kernelpagesize_kB=4 699 3206800000 default file=/lib64/libc-2.12.so mapped=59 mapmax=21 active=55 N0=41 N3=18 kernelpagesize_kB=4 700 320698b000 default file=/lib64/libc-2.12.so 701 3206b8a000 default file=/lib64/libc-2.12.so anon=2 dirty=2 N3=2 kernelpagesize_kB=4 702 3206b8e000 default file=/lib64/libc-2.12.so anon=1 dirty=1 N3=1 kernelpagesize_kB=4 703 3206b8f000 default anon=3 dirty=3 active=1 N3=3 kernelpagesize_kB=4 704 7f4dc10a2000 default anon=3 dirty=3 N3=3 kernelpagesize_kB=4 705 7f4dc10b4000 default anon=2 dirty=2 active=1 N3=2 kernelpagesize_kB=4 706 7f4dc1200000 default file=/anon_hugepage\040(deleted) huge anon=1 dirty=1 N3=1 kernelpagesize_kB=2048 707 7fff335f0000 default stack anon=3 dirty=3 N3=3 kernelpagesize_kB=4 708 7fff3369d000 default mapped=1 mapmax=35 active=0 N3=1 kernelpagesize_kB=4 709 710Where: 711 712"address" is the starting address for the mapping; 713 714"policy" reports the NUMA memory policy set for the mapping (see Documentation/admin-guide/mm/numa_memory_policy.rst); 715 716"mapping details" summarizes mapping data such as mapping type, page usage counters, 717node locality page counters (N0 == node0, N1 == node1, ...) and the kernel page 718size, in KB, that is backing the mapping up. 719 720Note that some kernel configurations do not track the precise number of times 721a page part of a larger allocation (e.g., THP) is mapped. In these 722configurations, "mapmax" might corresponds to the average number of mappings 723per page in such a larger allocation instead. 724 7251.2 Kernel data 726--------------- 727 728Similar to the process entries, the kernel data files give information about 729the running kernel. The files used to obtain this information are contained in 730/proc and are listed in Table 1-5. Not all of these will be present in your 731system. It depends on the kernel configuration and the loaded modules, which 732files are there, and which are missing. 733 734.. table:: Table 1-5: Kernel info in /proc 735 736 ============ =============================================================== 737 File Content 738 ============ =============================================================== 739 allocinfo Memory allocations profiling information 740 apm Advanced power management info 741 bootconfig Kernel command line obtained from boot config, 742 and, if there were kernel parameters from the 743 boot loader, a "# Parameters from bootloader:" 744 line followed by a line containing those 745 parameters prefixed by "# ". (5.5) 746 buddyinfo Kernel memory allocator information (see text) (2.5) 747 bus Directory containing bus specific information 748 cmdline Kernel command line, both from bootloader and embedded 749 in the kernel image 750 cpuinfo Info about the CPU 751 devices Available devices (block and character) 752 dma Used DMA channels 753 filesystems Supported filesystems 754 driver Various drivers grouped here, currently rtc (2.4) 755 execdomains Execdomains, related to security (2.4) 756 fb Frame Buffer devices (2.4) 757 fs File system parameters, currently nfs/exports (2.4) 758 ide Directory containing info about the IDE subsystem 759 interrupts Interrupt usage 760 iomem Memory map (2.4) 761 ioports I/O port usage 762 irq Masks for irq to cpu affinity (2.4)(smp?) 763 isapnp ISA PnP (Plug&Play) Info (2.4) 764 kcore Kernel core image (can be ELF or A.OUT(deprecated in 2.4)) 765 kmsg Kernel messages 766 ksyms Kernel symbol table 767 loadavg Load average of last 1, 5 & 15 minutes; 768 number of processes currently runnable (running or on ready queue); 769 total number of processes in system; 770 last pid created. 771 All fields are separated by one space except "number of 772 processes currently runnable" and "total number of processes 773 in system", which are separated by a slash ('/'). Example: 774 0.61 0.61 0.55 3/828 22084 775 locks Kernel locks 776 meminfo Memory info 777 misc Miscellaneous 778 modules List of loaded modules 779 mounts Mounted filesystems 780 net Networking info (see text) 781 pagetypeinfo Additional page allocator information (see text) (2.5) 782 partitions Table of partitions known to the system 783 pci Deprecated info of PCI bus (new way -> /proc/bus/pci/, 784 decoupled by lspci (2.4) 785 rtc Real time clock 786 scsi SCSI info (see text) 787 slabinfo Slab pool info 788 softirqs softirq usage 789 stat Overall statistics 790 swaps Swap space utilization 791 sys See chapter 2 792 sysvipc Info of SysVIPC Resources (msg, sem, shm) (2.4) 793 tty Info of tty drivers 794 uptime Wall clock since boot, combined idle time of all cpus 795 version Kernel version 796 video bttv info of video resources (2.4) 797 vmallocinfo Show vmalloced areas 798 ============ =============================================================== 799 800You can, for example, check which interrupts are currently in use and what 801they are used for by looking in the file /proc/interrupts:: 802 803 > cat /proc/interrupts 804 CPU0 805 0: 8728810 XT-PIC timer 806 1: 895 XT-PIC keyboard 807 2: 0 XT-PIC cascade 808 3: 531695 XT-PIC aha152x 809 4: 2014133 XT-PIC serial 810 5: 44401 XT-PIC pcnet_cs 811 8: 2 XT-PIC rtc 812 11: 8 XT-PIC i82365 813 12: 182918 XT-PIC PS/2 Mouse 814 13: 1 XT-PIC fpu 815 14: 1232265 XT-PIC ide0 816 15: 7 XT-PIC ide1 817 NMI: 0 818 819In 2.4.* a couple of lines where added to this file LOC & ERR (this time is the 820output of a SMP machine):: 821 822 > cat /proc/interrupts 823 824 CPU0 CPU1 825 0: 1243498 1214548 IO-APIC-edge timer 826 1: 8949 8958 IO-APIC-edge keyboard 827 2: 0 0 XT-PIC cascade 828 5: 11286 10161 IO-APIC-edge soundblaster 829 8: 1 0 IO-APIC-edge rtc 830 9: 27422 27407 IO-APIC-edge 3c503 831 12: 113645 113873 IO-APIC-edge PS/2 Mouse 832 13: 0 0 XT-PIC fpu 833 14: 22491 24012 IO-APIC-edge ide0 834 15: 2183 2415 IO-APIC-edge ide1 835 17: 30564 30414 IO-APIC-level eth0 836 18: 177 164 IO-APIC-level bttv 837 NMI: 2457961 2457959 838 LOC: 2457882 2457881 839 ERR: 2155 840 841NMI is incremented in this case because every timer interrupt generates a NMI 842(Non Maskable Interrupt) which is used by the NMI Watchdog to detect lockups. 843 844LOC is the local interrupt counter of the internal APIC of every CPU. 845 846ERR is incremented in the case of errors in the IO-APIC bus (the bus that 847connects the CPUs in a SMP system. This means that an error has been detected, 848the IO-APIC automatically retry the transmission, so it should not be a big 849problem, but you should read the SMP-FAQ. 850 851In 2.6.2* /proc/interrupts was expanded again. This time the goal was for 852/proc/interrupts to display every IRQ vector in use by the system, not 853just those considered 'most important'. The new vectors are: 854 855THR 856 interrupt raised when a machine check threshold counter 857 (typically counting ECC corrected errors of memory or cache) exceeds 858 a configurable threshold. Only available on some systems. 859 860TRM 861 a thermal event interrupt occurs when a temperature threshold 862 has been exceeded for the CPU. This interrupt may also be generated 863 when the temperature drops back to normal. 864 865SPU 866 a spurious interrupt is some interrupt that was raised then lowered 867 by some IO device before it could be fully processed by the APIC. Hence 868 the APIC sees the interrupt but does not know what device it came from. 869 For this case the APIC will generate the interrupt with a IRQ vector 870 of 0xff. This might also be generated by chipset bugs. 871 872RES, CAL, TLB 873 rescheduling, call and TLB flush interrupts are 874 sent from one CPU to another per the needs of the OS. Typically, 875 their statistics are used by kernel developers and interested users to 876 determine the occurrence of interrupts of the given type. 877 878The above IRQ vectors are displayed only when relevant. For example, 879the threshold vector does not exist on x86_64 platforms. Others are 880suppressed when the system is a uniprocessor. As of this writing, only 881i386 and x86_64 platforms support the new IRQ vector displays. 882 883Of some interest is the introduction of the /proc/irq directory to 2.4. 884It could be used to set IRQ to CPU affinity. This means that you can "hook" an 885IRQ to only one CPU, or to exclude a CPU of handling IRQs. The contents of the 886irq subdir is one subdir for each IRQ, and default_smp_affinity. 887 888For example:: 889 890 > ls /proc/irq/ 891 0 10 12 14 16 18 2 4 6 8 default_smp_affinity 892 1 11 13 15 17 19 3 5 7 9 893 > ls /proc/irq/0/ 894 smp_affinity 895 896smp_affinity is a bitmask, in which you can specify which CPUs can handle the 897IRQ. You can set it by doing:: 898 899 > echo 1 > /proc/irq/10/smp_affinity 900 901This means that only the first CPU will handle the IRQ, but you can also echo 9025 which means that only the first and third CPU can handle the IRQ. 903 904The contents of each smp_affinity file is the same by default:: 905 906 > cat /proc/irq/0/smp_affinity 907 ffffffff 908 909There is an alternate interface, smp_affinity_list which allows specifying 910a CPU range instead of a bitmask:: 911 912 > cat /proc/irq/0/smp_affinity_list 913 1024-1031 914 915The default_smp_affinity mask applies to all non-active IRQs, which are the 916IRQs which have not yet been allocated/activated, and hence which lack a 917/proc/irq/[0-9]* directory. 918 919The node file on an SMP system shows the node to which the device using the IRQ 920reports itself as being attached. This hardware locality information does not 921include information about any possible driver locality preference. 922 923The way IRQs are routed is handled by the IO-APIC, and it's Round Robin 924between all the CPUs which are allowed to handle it. As usual the kernel has 925more info than you and does a better job than you, so the defaults are the 926best choice for almost everyone. [Note this applies only to those IO-APIC's 927that support "Round Robin" interrupt distribution.] 928 929There are three more important subdirectories in /proc: net, scsi, and sys. 930The general rule is that the contents, or even the existence of these 931directories, depend on your kernel configuration. If SCSI is not enabled, the 932directory scsi may not exist. The same is true with the net, which is there 933only when networking support is present in the running kernel. 934 935The slabinfo file gives information about memory usage at the slab level. 936Linux uses slab pools for memory management above page level in version 2.2. 937Commonly used objects have their own slab pool (such as network buffers, 938directory cache, and so on). 939 940:: 941 942 > cat /proc/buddyinfo 943 944 Node 0, zone DMA 0 4 5 4 4 3 ... 945 Node 0, zone Normal 1 0 0 1 101 8 ... 946 Node 0, zone HighMem 2 0 0 1 1 0 ... 947 948External fragmentation is a problem under some workloads, and buddyinfo is a 949useful tool for helping diagnose these problems. Buddyinfo will give you a 950clue as to how big an area you can safely allocate, or why a previous 951allocation failed. 952 953Each column represents the number of pages of a certain order which are 954available. In this case, there are 0 chunks of 2^0*PAGE_SIZE available in 955ZONE_DMA, 4 chunks of 2^1*PAGE_SIZE in ZONE_DMA, 101 chunks of 2^4*PAGE_SIZE 956available in ZONE_NORMAL, etc... 957 958More information relevant to external fragmentation can be found in 959pagetypeinfo:: 960 961 > cat /proc/pagetypeinfo 962 Page block order: 9 963 Pages per block: 512 964 965 Free pages count per migrate type at order 0 1 2 3 4 5 6 7 8 9 10 966 Node 0, zone DMA, type Unmovable 0 0 0 1 1 1 1 1 1 1 0 967 Node 0, zone DMA, type Reclaimable 0 0 0 0 0 0 0 0 0 0 0 968 Node 0, zone DMA, type Movable 1 1 2 1 2 1 1 0 1 0 2 969 Node 0, zone DMA, type Reserve 0 0 0 0 0 0 0 0 0 1 0 970 Node 0, zone DMA, type Isolate 0 0 0 0 0 0 0 0 0 0 0 971 Node 0, zone DMA32, type Unmovable 103 54 77 1 1 1 11 8 7 1 9 972 Node 0, zone DMA32, type Reclaimable 0 0 2 1 0 0 0 0 1 0 0 973 Node 0, zone DMA32, type Movable 169 152 113 91 77 54 39 13 6 1 452 974 Node 0, zone DMA32, type Reserve 1 2 2 2 2 0 1 1 1 1 0 975 Node 0, zone DMA32, type Isolate 0 0 0 0 0 0 0 0 0 0 0 976 977 Number of blocks type Unmovable Reclaimable Movable Reserve Isolate 978 Node 0, zone DMA 2 0 5 1 0 979 Node 0, zone DMA32 41 6 967 2 0 980 981Fragmentation avoidance in the kernel works by grouping pages of different 982migrate types into the same contiguous regions of memory called page blocks. 983A page block is typically the size of the default hugepage size, e.g. 2MB on 984X86-64. By keeping pages grouped based on their ability to move, the kernel 985can reclaim pages within a page block to satisfy a high-order allocation. 986 987The pagetypinfo begins with information on the size of a page block. It 988then gives the same type of information as buddyinfo except broken down 989by migrate-type and finishes with details on how many page blocks of each 990type exist. 991 992If min_free_kbytes has been tuned correctly (recommendations made by hugeadm 993from libhugetlbfs https://github.com/libhugetlbfs/libhugetlbfs/), one can 994make an estimate of the likely number of huge pages that can be allocated 995at a given point in time. All the "Movable" blocks should be allocatable 996unless memory has been mlock()'d. Some of the Reclaimable blocks should 997also be allocatable although a lot of filesystem metadata may have to be 998reclaimed to achieve this. 999 1000 1001allocinfo 1002~~~~~~~~~ 1003 1004Provides information about memory allocations at all locations in the code 1005base. Each allocation in the code is identified by its source file, line 1006number, module (if originates from a loadable module) and the function calling 1007the allocation. The number of bytes allocated and number of calls at each 1008location are reported. The first line indicates the version of the file, the 1009second line is the header listing fields in the file. 1010If file version is 2.0 or higher then each line may contain additional 1011<key>:<value> pairs representing extra information about the call site. 1012For example if the counters are not accurate, the line will be appended with 1013"accurate:no" pair. 1014 1015Supported markers in v2: 1016accurate:no 1017 1018 Absolute values of the counters in this line are not accurate 1019 because of the failure to allocate memory to track some of the 1020 allocations made at this location. Deltas in these counters are 1021 accurate, therefore counters can be used to track allocation size 1022 and count changes. 1023 1024Example output. 1025 1026:: 1027 1028 > tail -n +3 /proc/allocinfo | sort -rn 1029 127664128 31168 mm/page_ext.c:270 func:alloc_page_ext 1030 56373248 4737 mm/slub.c:2259 func:alloc_slab_page 1031 14880768 3633 mm/readahead.c:247 func:page_cache_ra_unbounded 1032 14417920 3520 mm/mm_init.c:2530 func:alloc_large_system_hash 1033 13377536 234 block/blk-mq.c:3421 func:blk_mq_alloc_rqs 1034 11718656 2861 mm/filemap.c:1919 func:__filemap_get_folio 1035 9192960 2800 kernel/fork.c:307 func:alloc_thread_stack_node 1036 4206592 4 net/netfilter/nf_conntrack_core.c:2567 func:nf_ct_alloc_hashtable 1037 4136960 1010 drivers/staging/ctagmod/ctagmod.c:20 [ctagmod] func:ctagmod_start 1038 3940352 962 mm/memory.c:4214 func:alloc_anon_folio 1039 2894464 22613 fs/kernfs/dir.c:615 func:__kernfs_new_node 1040 ... 1041 1042 1043meminfo 1044~~~~~~~ 1045 1046Provides information about distribution and utilization of memory. This 1047varies by architecture and compile options. Some of the counters reported 1048here overlap. The memory reported by the non overlapping counters may not 1049add up to the overall memory usage and the difference for some workloads 1050can be substantial. In many cases there are other means to find out 1051additional memory using subsystem specific interfaces, for instance 1052/proc/net/sockstat for TCP memory allocations. 1053 1054Example output. You may not have all of these fields. 1055 1056:: 1057 1058 > cat /proc/meminfo 1059 1060 MemTotal: 32858820 kB 1061 MemFree: 21001236 kB 1062 MemAvailable: 27214312 kB 1063 Buffers: 581092 kB 1064 Cached: 5587612 kB 1065 SwapCached: 0 kB 1066 Active: 3237152 kB 1067 Inactive: 7586256 kB 1068 Active(anon): 94064 kB 1069 Inactive(anon): 4570616 kB 1070 Active(file): 3143088 kB 1071 Inactive(file): 3015640 kB 1072 Unevictable: 0 kB 1073 Mlocked: 0 kB 1074 SwapTotal: 0 kB 1075 SwapFree: 0 kB 1076 Zswap: 1904 kB 1077 Zswapped: 7792 kB 1078 Dirty: 12 kB 1079 Writeback: 0 kB 1080 AnonPages: 4654780 kB 1081 Mapped: 266244 kB 1082 Shmem: 9976 kB 1083 KReclaimable: 517708 kB 1084 Slab: 660044 kB 1085 SReclaimable: 517708 kB 1086 SUnreclaim: 142336 kB 1087 KernelStack: 11168 kB 1088 PageTables: 20540 kB 1089 SecPageTables: 0 kB 1090 NFS_Unstable: 0 kB 1091 Bounce: 0 kB 1092 WritebackTmp: 0 kB 1093 CommitLimit: 16429408 kB 1094 Committed_AS: 7715148 kB 1095 VmallocTotal: 34359738367 kB 1096 VmallocUsed: 40444 kB 1097 VmallocChunk: 0 kB 1098 Percpu: 29312 kB 1099 EarlyMemtestBad: 0 kB 1100 HardwareCorrupted: 0 kB 1101 AnonHugePages: 4149248 kB 1102 ShmemHugePages: 0 kB 1103 ShmemPmdMapped: 0 kB 1104 FileHugePages: 0 kB 1105 FilePmdMapped: 0 kB 1106 CmaTotal: 0 kB 1107 CmaFree: 0 kB 1108 Unaccepted: 0 kB 1109 Balloon: 0 kB 1110 GPUActive: 0 kB 1111 GPUReclaim: 0 kB 1112 HugePages_Total: 0 1113 HugePages_Free: 0 1114 HugePages_Rsvd: 0 1115 HugePages_Surp: 0 1116 Hugepagesize: 2048 kB 1117 Hugetlb: 0 kB 1118 DirectMap4k: 401152 kB 1119 DirectMap2M: 10008576 kB 1120 DirectMap1G: 24117248 kB 1121 1122MemTotal 1123 Total usable RAM (i.e. physical RAM minus a few reserved 1124 bits and the kernel binary code) 1125MemFree 1126 Total free RAM. On highmem systems, the sum of LowFree+HighFree 1127MemAvailable 1128 An estimate of how much memory is available for starting new 1129 applications, without swapping. Calculated from MemFree, 1130 SReclaimable, the size of the file LRU lists, and the low 1131 watermarks in each zone. 1132 The estimate takes into account that the system needs some 1133 page cache to function well, and that not all reclaimable 1134 slab will be reclaimable, due to items being in use. The 1135 impact of those factors will vary from system to system. 1136Buffers 1137 Relatively temporary storage for raw disk blocks 1138 shouldn't get tremendously large (20MB or so) 1139Cached 1140 In-memory cache for files read from the disk (the 1141 pagecache) as well as tmpfs & shmem. 1142 Doesn't include SwapCached. 1143SwapCached 1144 Memory that once was swapped out, is swapped back in but 1145 still also is in the swapfile (if memory is needed it 1146 doesn't need to be swapped out AGAIN because it is already 1147 in the swapfile. This saves I/O) 1148Active 1149 Memory that has been used more recently and usually not 1150 reclaimed unless absolutely necessary. 1151Inactive 1152 Memory which has been less recently used. It is more 1153 eligible to be reclaimed for other purposes 1154Unevictable 1155 Memory allocated for userspace which cannot be reclaimed, such 1156 as mlocked pages, ramfs backing pages, secret memfd pages etc. 1157Mlocked 1158 Memory locked with mlock(). 1159HighTotal, HighFree 1160 Highmem is all memory above ~860MB of physical memory. 1161 Highmem areas are for use by userspace programs, or 1162 for the pagecache. The kernel must use tricks to access 1163 this memory, making it slower to access than lowmem. 1164LowTotal, LowFree 1165 Lowmem is memory which can be used for everything that 1166 highmem can be used for, but it is also available for the 1167 kernel's use for its own data structures. Among many 1168 other things, it is where everything from the Slab is 1169 allocated. Bad things happen when you're out of lowmem. 1170SwapTotal 1171 total amount of swap space available 1172SwapFree 1173 Memory which has been evicted from RAM, and is temporarily 1174 on the disk 1175Zswap 1176 Memory consumed by the zswap backend (compressed size) 1177Zswapped 1178 Amount of anonymous memory stored in zswap (original size) 1179Dirty 1180 Memory which is waiting to get written back to the disk 1181Writeback 1182 Memory which is actively being written back to the disk 1183AnonPages 1184 Non-file backed pages mapped into userspace page tables. Note that 1185 some kernel configurations might consider all pages part of a 1186 larger allocation (e.g., THP) as "mapped", as soon as a single 1187 page is mapped. 1188Mapped 1189 files which have been mmapped, such as libraries. Note that some 1190 kernel configurations might consider all pages part of a larger 1191 allocation (e.g., THP) as "mapped", as soon as a single page is 1192 mapped. 1193Shmem 1194 Total memory used by shared memory (shmem) and tmpfs 1195KReclaimable 1196 Kernel allocations that the kernel will attempt to reclaim 1197 under memory pressure. Includes SReclaimable (below), and other 1198 direct allocations with a shrinker. 1199Slab 1200 in-kernel data structures cache 1201SReclaimable 1202 Part of Slab, that might be reclaimed, such as caches 1203SUnreclaim 1204 Part of Slab, that cannot be reclaimed on memory pressure 1205KernelStack 1206 Memory consumed by the kernel stacks of all tasks 1207PageTables 1208 Memory consumed by userspace page tables 1209SecPageTables 1210 Memory consumed by secondary page tables, this currently includes 1211 KVM mmu and IOMMU allocations on x86 and arm64. 1212NFS_Unstable 1213 Always zero. Previously counted pages which had been written to 1214 the server, but has not been committed to stable storage. 1215Bounce 1216 Always zero. Previously memory used for block device 1217 "bounce buffers". 1218WritebackTmp 1219 Always zero. Previously memory used by FUSE for temporary 1220 writeback buffers. 1221CommitLimit 1222 Based on the overcommit ratio ('vm.overcommit_ratio'), 1223 this is the total amount of memory currently available to 1224 be allocated on the system. This limit is only adhered to 1225 if strict overcommit accounting is enabled (mode 2 in 1226 'vm.overcommit_memory'). 1227 1228 The CommitLimit is calculated with the following formula:: 1229 1230 CommitLimit = ([total RAM pages] - [total huge TLB pages]) * 1231 overcommit_ratio / 100 + [total swap pages] 1232 1233 For example, on a system with 1G of physical RAM and 7G 1234 of swap with a `vm.overcommit_ratio` of 30 it would 1235 yield a CommitLimit of 7.3G. 1236 1237 For more details, see the memory overcommit documentation 1238 in mm/overcommit-accounting. 1239Committed_AS 1240 The amount of memory presently allocated on the system. 1241 The committed memory is a sum of all of the memory which 1242 has been allocated by processes, even if it has not been 1243 "used" by them as of yet. A process which malloc()'s 1G 1244 of memory, but only touches 300M of it will show up as 1245 using 1G. This 1G is memory which has been "committed" to 1246 by the VM and can be used at any time by the allocating 1247 application. With strict overcommit enabled on the system 1248 (mode 2 in 'vm.overcommit_memory'), allocations which would 1249 exceed the CommitLimit (detailed above) will not be permitted. 1250 This is useful if one needs to guarantee that processes will 1251 not fail due to lack of memory once that memory has been 1252 successfully allocated. 1253VmallocTotal 1254 total size of vmalloc virtual address space 1255VmallocUsed 1256 amount of vmalloc area which is used 1257VmallocChunk 1258 largest contiguous block of vmalloc area which is free 1259Percpu 1260 Memory allocated to the percpu allocator used to back percpu 1261 allocations. This stat excludes the cost of metadata. 1262EarlyMemtestBad 1263 The amount of RAM/memory in kB, that was identified as corrupted 1264 by early memtest. If memtest was not run, this field will not 1265 be displayed at all. Size is never rounded down to 0 kB. 1266 That means if 0 kB is reported, you can safely assume 1267 there was at least one pass of memtest and none of the passes 1268 found a single faulty byte of RAM. 1269HardwareCorrupted 1270 The amount of RAM/memory in KB, the kernel identifies as 1271 corrupted. 1272AnonHugePages 1273 Non-file backed huge pages mapped into userspace page tables 1274ShmemHugePages 1275 Memory used by shared memory (shmem) and tmpfs allocated 1276 with huge pages 1277ShmemPmdMapped 1278 Shared memory mapped into userspace with huge pages 1279FileHugePages 1280 Memory used for filesystem data (page cache) allocated 1281 with huge pages 1282FilePmdMapped 1283 Page cache mapped into userspace with huge pages 1284CmaTotal 1285 Memory reserved for the Contiguous Memory Allocator (CMA) 1286CmaFree 1287 Free remaining memory in the CMA reserves 1288Unaccepted 1289 Memory that has not been accepted by the guest 1290Balloon 1291 Memory returned to Host by VM Balloon Drivers 1292GPUActive 1293 System memory allocated to active GPU objects 1294GPUReclaim 1295 System memory stored in GPU pools for reuse. This memory is not 1296 counted in GPUActive. It is shrinker reclaimable memory kept in a reuse 1297 pool because it has non-standard page table attributes, like WC or UC. 1298HugePages_Total, HugePages_Free, HugePages_Rsvd, HugePages_Surp, Hugepagesize, Hugetlb 1299 See Documentation/admin-guide/mm/hugetlbpage.rst. 1300DirectMap4k, DirectMap2M, DirectMap1G 1301 Breakdown of page table sizes used in the kernel's 1302 identity mapping of RAM 1303 1304vmallocinfo 1305~~~~~~~~~~~ 1306 1307Provides information about vmalloced/vmaped areas. One line per area, 1308containing the virtual address range of the area, size in bytes, 1309caller information of the creator, and optional information depending 1310on the kind of area: 1311 1312 ========== =================================================== 1313 pages=nr number of pages 1314 phys=addr if a physical address was specified 1315 ioremap I/O mapping (ioremap() and friends) 1316 vmalloc vmalloc() area 1317 vmap vmap()ed pages 1318 user VM_USERMAP area 1319 vpages buffer for pages pointers was vmalloced (huge area) 1320 N<node>=nr (Only on NUMA kernels) 1321 Number of pages allocated on memory node <node> 1322 ========== =================================================== 1323 1324:: 1325 1326 > cat /proc/vmallocinfo 1327 0xffffc20000000000-0xffffc20000201000 2101248 alloc_large_system_hash+0x204 ... 1328 /0x2c0 pages=512 vmalloc N0=128 N1=128 N2=128 N3=128 1329 0xffffc20000201000-0xffffc20000302000 1052672 alloc_large_system_hash+0x204 ... 1330 /0x2c0 pages=256 vmalloc N0=64 N1=64 N2=64 N3=64 1331 0xffffc20000302000-0xffffc20000304000 8192 acpi_tb_verify_table+0x21/0x4f... 1332 phys=7fee8000 ioremap 1333 0xffffc20000304000-0xffffc20000307000 12288 acpi_tb_verify_table+0x21/0x4f... 1334 phys=7fee7000 ioremap 1335 0xffffc2000031d000-0xffffc2000031f000 8192 init_vdso_vars+0x112/0x210 1336 0xffffc2000031f000-0xffffc2000032b000 49152 cramfs_uncompress_init+0x2e ... 1337 /0x80 pages=11 vmalloc N0=3 N1=3 N2=2 N3=3 1338 0xffffc2000033a000-0xffffc2000033d000 12288 sys_swapon+0x640/0xac0 ... 1339 pages=2 vmalloc N1=2 1340 0xffffc20000347000-0xffffc2000034c000 20480 xt_alloc_table_info+0xfe ... 1341 /0x130 [x_tables] pages=4 vmalloc N0=4 1342 0xffffffffa0000000-0xffffffffa000f000 61440 sys_init_module+0xc27/0x1d00 ... 1343 pages=14 vmalloc N2=14 1344 0xffffffffa000f000-0xffffffffa0014000 20480 sys_init_module+0xc27/0x1d00 ... 1345 pages=4 vmalloc N1=4 1346 0xffffffffa0014000-0xffffffffa0017000 12288 sys_init_module+0xc27/0x1d00 ... 1347 pages=2 vmalloc N1=2 1348 0xffffffffa0017000-0xffffffffa0022000 45056 sys_init_module+0xc27/0x1d00 ... 1349 pages=10 vmalloc N0=10 1350 1351 1352softirqs 1353~~~~~~~~ 1354 1355Provides counts of softirq handlers serviced since boot time, for each CPU. 1356 1357:: 1358 1359 > cat /proc/softirqs 1360 CPU0 CPU1 CPU2 CPU3 1361 HI: 0 0 0 0 1362 TIMER: 27166 27120 27097 27034 1363 NET_TX: 0 0 0 17 1364 NET_RX: 42 0 0 39 1365 BLOCK: 0 0 107 1121 1366 TASKLET: 0 0 0 290 1367 SCHED: 27035 26983 26971 26746 1368 HRTIMER: 0 0 0 0 1369 RCU: 1678 1769 2178 2250 1370 13711.3 Networking info in /proc/net 1372-------------------------------- 1373 1374The subdirectory /proc/net follows the usual pattern. Table 1-8 shows the 1375additional values you get for IP version 6 if you configure the kernel to 1376support this. Table 1-9 lists the files and their meaning. 1377 1378 1379.. table:: Table 1-8: IPv6 info in /proc/net 1380 1381 ========== ===================================================== 1382 File Content 1383 ========== ===================================================== 1384 udp6 UDP sockets (IPv6) 1385 tcp6 TCP sockets (IPv6) 1386 raw6 Raw device statistics (IPv6) 1387 igmp6 IP multicast addresses, which this host joined (IPv6) 1388 if_inet6 List of IPv6 interface addresses 1389 ipv6_route Kernel routing table for IPv6 1390 rt6_stats Global IPv6 routing tables statistics 1391 sockstat6 Socket statistics (IPv6) 1392 snmp6 Snmp data (IPv6) 1393 ========== ===================================================== 1394 1395.. table:: Table 1-9: Network info in /proc/net 1396 1397 ============= ================================================================ 1398 File Content 1399 ============= ================================================================ 1400 arp Kernel ARP table 1401 dev network devices with statistics 1402 dev_mcast the Layer2 multicast groups a device is listening too 1403 (interface index, label, number of references, number of bound 1404 addresses). 1405 dev_stat network device status 1406 ip_fwchains Firewall chain linkage 1407 ip_fwnames Firewall chain names 1408 ip_masq Directory containing the masquerading tables 1409 ip_masquerade Major masquerading table 1410 netstat Network statistics 1411 raw raw device statistics 1412 route Kernel routing table 1413 rpc Directory containing rpc info 1414 rt_cache Routing cache 1415 snmp SNMP data 1416 sockstat Socket statistics 1417 softnet_stat Per-CPU incoming packets queues statistics of online CPUs 1418 tcp TCP sockets 1419 udp UDP sockets 1420 unix UNIX domain sockets 1421 wireless Wireless interface data (Wavelan etc) 1422 igmp IP multicast addresses, which this host joined 1423 psched Global packet scheduler parameters. 1424 netlink List of PF_NETLINK sockets 1425 ip_mr_vifs List of multicast virtual interfaces 1426 ip_mr_cache List of multicast routing cache 1427 ============= ================================================================ 1428 1429You can use this information to see which network devices are available in 1430your system and how much traffic was routed over those devices:: 1431 1432 > cat /proc/net/dev 1433 Inter-|Receive |[... 1434 face |bytes packets errs drop fifo frame compressed multicast|[... 1435 lo: 908188 5596 0 0 0 0 0 0 [... 1436 ppp0:15475140 20721 410 0 0 410 0 0 [... 1437 eth0: 614530 7085 0 0 0 0 0 1 [... 1438 1439 ...] Transmit 1440 ...] bytes packets errs drop fifo colls carrier compressed 1441 ...] 908188 5596 0 0 0 0 0 0 1442 ...] 1375103 17405 0 0 0 0 0 0 1443 ...] 1703981 5535 0 0 0 3 0 0 1444 1445In addition, each Channel Bond interface has its own directory. For 1446example, the bond0 device will have a directory called /proc/net/bond0/. 1447It will contain information that is specific to that bond, such as the 1448current slaves of the bond, the link status of the slaves, and how 1449many times the slaves link has failed. 1450 14511.4 SCSI info 1452------------- 1453 1454If you have a SCSI or ATA host adapter in your system, you'll find a 1455subdirectory named after the driver for this adapter in /proc/scsi. 1456You'll also see a list of all recognized SCSI devices in /proc/scsi:: 1457 1458 >cat /proc/scsi/scsi 1459 Attached devices: 1460 Host: scsi0 Channel: 00 Id: 00 Lun: 00 1461 Vendor: IBM Model: DGHS09U Rev: 03E0 1462 Type: Direct-Access ANSI SCSI revision: 03 1463 Host: scsi0 Channel: 00 Id: 06 Lun: 00 1464 Vendor: PIONEER Model: CD-ROM DR-U06S Rev: 1.04 1465 Type: CD-ROM ANSI SCSI revision: 02 1466 1467 1468The directory named after the driver has one file for each adapter found in 1469the system. These files contain information about the controller, including 1470the used IRQ and the IO address range. The amount of information shown is 1471dependent on the adapter you use. The example shows the output for an Adaptec 1472AHA-2940 SCSI adapter:: 1473 1474 > cat /proc/scsi/aic7xxx/0 1475 1476 Adaptec AIC7xxx driver version: 5.1.19/3.2.4 1477 Compile Options: 1478 TCQ Enabled By Default : Disabled 1479 AIC7XXX_PROC_STATS : Disabled 1480 AIC7XXX_RESET_DELAY : 5 1481 Adapter Configuration: 1482 SCSI Adapter: Adaptec AHA-294X Ultra SCSI host adapter 1483 Ultra Wide Controller 1484 PCI MMAPed I/O Base: 0xeb001000 1485 Adapter SEEPROM Config: SEEPROM found and used. 1486 Adaptec SCSI BIOS: Enabled 1487 IRQ: 10 1488 SCBs: Active 0, Max Active 2, 1489 Allocated 15, HW 16, Page 255 1490 Interrupts: 160328 1491 BIOS Control Word: 0x18b6 1492 Adapter Control Word: 0x005b 1493 Extended Translation: Enabled 1494 Disconnect Enable Flags: 0xffff 1495 Ultra Enable Flags: 0x0001 1496 Tag Queue Enable Flags: 0x0000 1497 Ordered Queue Tag Flags: 0x0000 1498 Default Tag Queue Depth: 8 1499 Tagged Queue By Device array for aic7xxx host instance 0: 1500 {255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255} 1501 Actual queue depth per device for aic7xxx host instance 0: 1502 {1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1} 1503 Statistics: 1504 (scsi0:0:0:0) 1505 Device using Wide/Sync transfers at 40.0 MByte/sec, offset 8 1506 Transinfo settings: current(12/8/1/0), goal(12/8/1/0), user(12/15/1/0) 1507 Total transfers 160151 (74577 reads and 85574 writes) 1508 (scsi0:0:6:0) 1509 Device using Narrow/Sync transfers at 5.0 MByte/sec, offset 15 1510 Transinfo settings: current(50/15/0/0), goal(50/15/0/0), user(50/15/0/0) 1511 Total transfers 0 (0 reads and 0 writes) 1512 1513 15141.5 Parallel port info in /proc/parport 1515--------------------------------------- 1516 1517The directory /proc/parport contains information about the parallel ports of 1518your system. It has one subdirectory for each port, named after the port 1519number (0,1,2,...). 1520 1521These directories contain the four files shown in Table 1-10. 1522 1523 1524.. table:: Table 1-10: Files in /proc/parport 1525 1526 ========= ==================================================================== 1527 File Content 1528 ========= ==================================================================== 1529 autoprobe Any IEEE-1284 device ID information that has been acquired. 1530 devices list of the device drivers using that port. A + will appear by the 1531 name of the device currently using the port (it might not appear 1532 against any). 1533 hardware Parallel port's base address, IRQ line and DMA channel. 1534 irq IRQ that parport is using for that port. This is in a separate 1535 file to allow you to alter it by writing a new value in (IRQ 1536 number or none). 1537 ========= ==================================================================== 1538 15391.6 TTY info in /proc/tty 1540------------------------- 1541 1542Information about the available and actually used tty's can be found in the 1543directory /proc/tty. You'll find entries for drivers and line disciplines in 1544this directory, as shown in Table 1-11. 1545 1546 1547.. table:: Table 1-11: Files in /proc/tty 1548 1549 ============= ============================================== 1550 File Content 1551 ============= ============================================== 1552 drivers list of drivers and their usage 1553 ldiscs registered line disciplines 1554 driver/serial usage statistic and status of single tty lines 1555 ============= ============================================== 1556 1557To see which tty's are currently in use, you can simply look into the file 1558/proc/tty/drivers:: 1559 1560 > cat /proc/tty/drivers 1561 pty_slave /dev/pts 136 0-255 pty:slave 1562 pty_master /dev/ptm 128 0-255 pty:master 1563 pty_slave /dev/ttyp 3 0-255 pty:slave 1564 pty_master /dev/pty 2 0-255 pty:master 1565 serial /dev/cua 5 64-67 serial:callout 1566 serial /dev/ttyS 4 64-67 serial 1567 /dev/tty0 /dev/tty0 4 0 system:vtmaster 1568 /dev/ptmx /dev/ptmx 5 2 system 1569 /dev/console /dev/console 5 1 system:console 1570 /dev/tty /dev/tty 5 0 system:/dev/tty 1571 unknown /dev/tty 4 1-63 console 1572 1573 15741.7 Miscellaneous kernel statistics in /proc/stat 1575------------------------------------------------- 1576 1577Various pieces of information about kernel activity are available in the 1578/proc/stat file. All of the numbers reported in this file are aggregates 1579since the system first booted. For a quick look, simply cat the file:: 1580 1581 > cat /proc/stat 1582 cpu 237902850 368826709 106375398 1873517540 1135548 0 14507935 0 0 0 1583 cpu0 60045249 91891769 26331539 468411416 495718 0 5739640 0 0 0 1584 cpu1 59746288 91759249 26609887 468860630 312281 0 4384817 0 0 0 1585 cpu2 59489247 92985423 26904446 467808813 171668 0 2268998 0 0 0 1586 cpu3 58622065 92190267 26529524 468436680 155879 0 2114478 0 0 0 1587 intr 8688370575 8 3373 0 0 0 0 0 0 1 40791 0 0 353317 0 0 0 0 224789828 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 190974333 41958554 123983334 43 0 224593 0 0 0 <more 0's deleted> 1588 ctxt 22848221062 1589 btime 1605316999 1590 processes 746787147 1591 procs_running 2 1592 procs_blocked 0 1593 softirq 12121874454 100099120 3938138295 127375644 2795979 187870761 0 173808342 3072582055 52608 224184354 1594 1595The very first "cpu" line aggregates the numbers in all of the other "cpuN" 1596lines. These numbers identify the amount of time the CPU has spent performing 1597different kinds of work. Time units are in USER_HZ (typically hundredths of a 1598second). The meanings of the columns are as follows, from left to right: 1599 1600- user: normal processes executing in user mode 1601- nice: niced processes executing in user mode 1602- system: processes executing in kernel mode 1603- idle: twiddling thumbs 1604- iowait: In a word, iowait stands for waiting for I/O to complete. But there 1605 are several problems: 1606 1607 1. CPU will not wait for I/O to complete, iowait is the time that a task is 1608 waiting for I/O to complete. When CPU goes into idle state for 1609 outstanding task I/O, another task will be scheduled on this CPU. 1610 2. In a multi-core CPU, the task waiting for I/O to complete is not running 1611 on any CPU, so the iowait of each CPU is difficult to calculate. 1612 3. The value of iowait field in /proc/stat will decrease in certain 1613 conditions. 1614 1615 So, the iowait is not reliable by reading from /proc/stat. 1616- irq: servicing interrupts 1617- softirq: servicing softirqs 1618- steal: involuntary wait 1619- guest: running a normal guest 1620- guest_nice: running a niced guest 1621 1622The "intr" line gives counts of interrupts serviced since boot time, for each 1623of the possible system interrupts. The first column is the total of all 1624interrupts serviced including unnumbered architecture specific interrupts; 1625each subsequent column is the total for that particular numbered interrupt. 1626Unnumbered interrupts are not shown, only summed into the total. 1627 1628The "ctxt" line gives the total number of context switches across all CPUs. 1629 1630The "btime" line gives the time at which the system booted, in seconds since 1631the Unix epoch. 1632 1633The "processes" line gives the number of processes and threads created, which 1634includes (but is not limited to) those created by calls to the fork() and 1635clone() system calls. 1636 1637The "procs_running" line gives the total number of threads that are 1638running or ready to run (i.e., the total number of runnable threads). 1639 1640The "procs_blocked" line gives the number of processes currently blocked, 1641waiting for I/O to complete. 1642 1643The "softirq" line gives counts of softirqs serviced since boot time, for each 1644of the possible system softirqs. The first column is the total of all 1645softirqs serviced; each subsequent column is the total for that particular 1646softirq. 1647 1648 16491.8 Ext4 file system parameters 1650------------------------------- 1651 1652Information about mounted ext4 file systems can be found in 1653/proc/fs/ext4. Each mounted filesystem will have a directory in 1654/proc/fs/ext4 based on its device name (i.e., /proc/fs/ext4/hdc or 1655/proc/fs/ext4/sda9 or /proc/fs/ext4/dm-0). The files in each per-device 1656directory are shown in Table 1-12, below. 1657 1658.. table:: Table 1-12: Files in /proc/fs/ext4/<devname> 1659 1660 ============== ========================================================== 1661 File Content 1662 mb_groups details of multiblock allocator buddy cache of free blocks 1663 ============== ========================================================== 1664 16651.9 /proc/consoles 1666------------------- 1667Shows registered system console lines. 1668 1669To see which character device lines are currently used for the system console 1670/dev/console, you may simply look into the file /proc/consoles:: 1671 1672 > cat /proc/consoles 1673 tty0 -WU (ECp) 4:7 1674 ttyS0 -W- (Ep) 4:64 1675 1676The columns are: 1677 1678+--------------------+-------------------------------------------------------+ 1679| device | name of the device | 1680+====================+=======================================================+ 1681| operations | * R = can do read operations | 1682| | * W = can do write operations | 1683| | * U = can do unblank | 1684+--------------------+-------------------------------------------------------+ 1685| flags | * E = it is enabled | 1686| | * C = it is preferred console | 1687| | * B = it is primary boot console | 1688| | * p = it is used for printk buffer | 1689| | * b = it is not a TTY but a Braille device | 1690| | * a = it is safe to use when cpu is offline | 1691+--------------------+-------------------------------------------------------+ 1692| major:minor | major and minor number of the device separated by a | 1693| | colon | 1694+--------------------+-------------------------------------------------------+ 1695 1696Summary 1697------- 1698 1699The /proc file system serves information about the running system. It not only 1700allows access to process data but also allows you to request the kernel status 1701by reading files in the hierarchy. 1702 1703The directory structure of /proc reflects the types of information and makes 1704it easy, if not obvious, where to look for specific data. 1705 1706Chapter 2: Modifying System Parameters 1707====================================== 1708 1709In This Chapter 1710--------------- 1711 1712* Modifying kernel parameters by writing into files found in /proc/sys 1713* Exploring the files which modify certain parameters 1714* Review of the /proc/sys file tree 1715 1716------------------------------------------------------------------------------ 1717 1718A very interesting part of /proc is the directory /proc/sys. This is not only 1719a source of information, it also allows you to change parameters within the 1720kernel. Be very careful when attempting this. You can optimize your system, 1721but you can also cause it to crash. Never alter kernel parameters on a 1722production system. Set up a development machine and test to make sure that 1723everything works the way you want it to. You may have no alternative but to 1724reboot the machine once an error has been made. 1725 1726To change a value, simply echo the new value into the file. 1727You need to be root to do this. You can create your own boot script 1728to perform this every time your system boots. 1729 1730The files in /proc/sys can be used to fine tune and monitor miscellaneous and 1731general things in the operation of the Linux kernel. Since some of the files 1732can inadvertently disrupt your system, it is advisable to read both 1733documentation and source before actually making adjustments. In any case, be 1734very careful when writing to any of these files. The entries in /proc may 1735change slightly between the 2.1.* and the 2.2 kernel, so if there is any doubt 1736review the kernel documentation in the directory linux/Documentation. 1737This chapter is heavily based on the documentation included in the pre 2.2 1738kernels, and became part of it in version 2.2.1 of the Linux kernel. 1739 1740Please see: Documentation/admin-guide/sysctl/ directory for descriptions of 1741these entries. 1742 1743Summary 1744------- 1745 1746Certain aspects of kernel behavior can be modified at runtime, without the 1747need to recompile the kernel, or even to reboot the system. The files in the 1748/proc/sys tree can not only be read, but also modified. You can use the echo 1749command to write value into these files, thereby changing the default settings 1750of the kernel. 1751 1752 1753Chapter 3: Per-process Parameters 1754================================= 1755 17563.1 /proc/<pid>/oom_adj & /proc/<pid>/oom_score_adj- Adjust the oom-killer score 1757-------------------------------------------------------------------------------- 1758 1759These files can be used to adjust the badness heuristic used to select which 1760process gets killed in out of memory (oom) conditions. 1761 1762The badness heuristic assigns a value to each candidate task ranging from 0 1763(never kill) to 1000 (always kill) to determine which process is targeted. The 1764units are roughly a proportion along that range of allowed memory the process 1765may allocate from based on an estimation of its current memory and swap use. 1766For example, if a task is using all allowed memory, its badness score will be 17671000. If it is using half of its allowed memory, its score will be 500. 1768 1769The amount of "allowed" memory depends on the context in which the oom killer 1770was called. If it is due to the memory assigned to the allocating task's cpuset 1771being exhausted, the allowed memory represents the set of mems assigned to that 1772cpuset. If it is due to a mempolicy's node(s) being exhausted, the allowed 1773memory represents the set of mempolicy nodes. If it is due to a memory 1774limit (or swap limit) being reached, the allowed memory is that configured 1775limit. Finally, if it is due to the entire system being out of memory, the 1776allowed memory represents all allocatable resources. 1777 1778The value of /proc/<pid>/oom_score_adj is added to the badness score before it 1779is used to determine which task to kill. Acceptable values range from -1000 1780(OOM_SCORE_ADJ_MIN) to +1000 (OOM_SCORE_ADJ_MAX). This allows userspace to 1781polarize the preference for oom killing either by always preferring a certain 1782task or completely disabling it. The lowest possible value, -1000, is 1783equivalent to disabling oom killing entirely for that task since it will always 1784report a badness score of 0. 1785 1786Consequently, it is very simple for userspace to define the amount of memory to 1787consider for each task. Setting a /proc/<pid>/oom_score_adj value of +500, for 1788example, is roughly equivalent to allowing the remainder of tasks sharing the 1789same system, cpuset, mempolicy, or memory controller resources to use at least 179050% more memory. A value of -500, on the other hand, would be roughly 1791equivalent to discounting 50% of the task's allowed memory from being considered 1792as scoring against the task. 1793 1794For backwards compatibility with previous kernels, /proc/<pid>/oom_adj may also 1795be used to tune the badness score. Its acceptable values range from -16 1796(OOM_ADJUST_MIN) to +15 (OOM_ADJUST_MAX) and a special value of -17 1797(OOM_DISABLE) to disable oom killing entirely for that task. Its value is 1798scaled linearly with /proc/<pid>/oom_score_adj. 1799 1800The value of /proc/<pid>/oom_score_adj may be reduced no lower than the last 1801value set by a CAP_SYS_RESOURCE process. To reduce the value any lower 1802requires CAP_SYS_RESOURCE. 1803 1804 18053.2 /proc/<pid>/oom_score - Display current oom-killer score 1806------------------------------------------------------------- 1807 1808This file can be used to check the current score used by the oom-killer for 1809any given <pid>. Use it together with /proc/<pid>/oom_score_adj to tune which 1810process should be killed in an out-of-memory situation. 1811 1812Please note that the exported value includes oom_score_adj so it is 1813effectively in range [0,2000]. 1814 1815 18163.3 /proc/<pid>/io - Display the IO accounting fields 1817------------------------------------------------------- 1818 1819This file contains IO statistics for each running process. 1820 1821Example 1822~~~~~~~ 1823 1824:: 1825 1826 test:/tmp # dd if=/dev/zero of=/tmp/test.dat & 1827 [1] 3828 1828 1829 test:/tmp # cat /proc/3828/io 1830 rchar: 323934931 1831 wchar: 323929600 1832 syscr: 632687 1833 syscw: 632675 1834 read_bytes: 0 1835 write_bytes: 323932160 1836 cancelled_write_bytes: 0 1837 1838 1839Description 1840~~~~~~~~~~~ 1841 1842rchar 1843^^^^^ 1844 1845I/O counter: chars read 1846The number of bytes which this task has caused to be read from storage. This 1847is simply the sum of bytes which this process passed to read() and pread(). 1848It includes things like tty IO and it is unaffected by whether or not actual 1849physical disk IO was required (the read might have been satisfied from 1850pagecache). 1851 1852 1853wchar 1854^^^^^ 1855 1856I/O counter: chars written 1857The number of bytes which this task has caused, or shall cause to be written 1858to disk. Similar caveats apply here as with rchar. 1859 1860 1861syscr 1862^^^^^ 1863 1864I/O counter: read syscalls 1865Attempt to count the number of read I/O operations, i.e. syscalls like read() 1866and pread(). 1867 1868 1869syscw 1870^^^^^ 1871 1872I/O counter: write syscalls 1873Attempt to count the number of write I/O operations, i.e. syscalls like 1874write() and pwrite(). 1875 1876 1877read_bytes 1878^^^^^^^^^^ 1879 1880I/O counter: bytes read 1881Attempt to count the number of bytes which this process really did cause to 1882be fetched from the storage layer. Done at the submit_bio() level, so it is 1883accurate for block-backed filesystems. <please add status regarding NFS and 1884CIFS at a later time> 1885 1886 1887write_bytes 1888^^^^^^^^^^^ 1889 1890I/O counter: bytes written 1891Attempt to count the number of bytes which this process caused to be sent to 1892the storage layer. This is done at page-dirtying time. 1893 1894 1895cancelled_write_bytes 1896^^^^^^^^^^^^^^^^^^^^^ 1897 1898The big inaccuracy here is truncate. If a process writes 1MB to a file and 1899then deletes the file, it will in fact perform no writeout. But it will have 1900been accounted as having caused 1MB of write. 1901In other words: The number of bytes which this process caused to not happen, 1902by truncating pagecache. A task can cause "negative" IO too. If this task 1903truncates some dirty pagecache, some IO which another task has been accounted 1904for (in its write_bytes) will not be happening. We _could_ just subtract that 1905from the truncating task's write_bytes, but there is information loss in doing 1906that. 1907 1908 1909.. Note:: 1910 1911 At its current implementation state, this is a bit racy on 32-bit machines: 1912 if process A reads process B's /proc/pid/io while process B is updating one 1913 of those 64-bit counters, process A could see an intermediate result. 1914 1915 1916More information about this can be found within the taskstats documentation in 1917Documentation/accounting. 1918 19193.4 /proc/<pid>/coredump_filter - Core dump filtering settings 1920--------------------------------------------------------------- 1921When a process is dumped, all anonymous memory is written to a core file as 1922long as the size of the core file isn't limited. But sometimes we don't want 1923to dump some memory segments, for example, huge shared memory or DAX. 1924Conversely, sometimes we want to save file-backed memory segments into a core 1925file, not only the individual files. 1926 1927/proc/<pid>/coredump_filter allows you to customize which memory segments 1928will be dumped when the <pid> process is dumped. coredump_filter is a bitmask 1929of memory types. If a bit of the bitmask is set, memory segments of the 1930corresponding memory type are dumped, otherwise they are not dumped. 1931 1932The following 9 memory types are supported: 1933 1934 - (bit 0) anonymous private memory 1935 - (bit 1) anonymous shared memory 1936 - (bit 2) file-backed private memory 1937 - (bit 3) file-backed shared memory 1938 - (bit 4) ELF header pages in file-backed private memory areas (it is 1939 effective only if the bit 2 is cleared) 1940 - (bit 5) hugetlb private memory 1941 - (bit 6) hugetlb shared memory 1942 - (bit 7) DAX private memory 1943 - (bit 8) DAX shared memory 1944 1945 Note that MMIO pages such as frame buffer are never dumped and vDSO pages 1946 are always dumped regardless of the bitmask status. 1947 1948 Note that bits 0-4 don't affect hugetlb or DAX memory. hugetlb memory is 1949 only affected by bit 5-6, and DAX is only affected by bits 7-8. 1950 1951The default value of coredump_filter is 0x33; this means all anonymous memory 1952segments, ELF header pages and hugetlb private memory are dumped. 1953 1954If you don't want to dump all shared memory segments attached to pid 1234, 1955write 0x31 to the process's proc file:: 1956 1957 $ echo 0x31 > /proc/1234/coredump_filter 1958 1959When a new process is created, the process inherits the bitmask status from its 1960parent. It is useful to set up coredump_filter before the program runs. 1961For example:: 1962 1963 $ echo 0x7 > /proc/self/coredump_filter 1964 $ ./some_program 1965 19663.5 /proc/<pid>/mountinfo - Information about mounts 1967-------------------------------------------------------- 1968 1969This file contains lines of the form:: 1970 1971 36 35 98:0 /mnt1 /mnt2 rw,noatime master:1 - ext3 /dev/root rw,errors=continue 1972 (1)(2)(3) (4) (5) (6) (n…m) (m+1)(m+2) (m+3) (m+4) 1973 1974 (1) mount ID: unique identifier of the mount (may be reused after umount) 1975 (2) parent ID: ID of parent (or of self for the top of the mount tree) 1976 (3) major:minor: value of st_dev for files on filesystem 1977 (4) root: root of the mount within the filesystem 1978 (5) mount point: mount point relative to the process's root 1979 (6) mount options: per mount options 1980 (n…m) optional fields: zero or more fields of the form "tag[:value]" 1981 (m+1) separator: marks the end of the optional fields 1982 (m+2) filesystem type: name of filesystem of the form "type[.subtype]" 1983 (m+3) mount source: filesystem specific information or "none" 1984 (m+4) super options: per super block options 1985 1986Parsers should ignore all unrecognised optional fields. Currently the 1987possible optional fields are: 1988 1989================ ============================================================== 1990shared:X mount is shared in peer group X 1991master:X mount is slave to peer group X 1992propagate_from:X mount is slave and receives propagation from peer group X [#]_ 1993unbindable mount is unbindable 1994================ ============================================================== 1995 1996.. [#] X is the closest dominant peer group under the process's root. If 1997 X is the immediate master of the mount, or if there's no dominant peer 1998 group under the same root, then only the "master:X" field is present 1999 and not the "propagate_from:X" field. 2000 2001For more information on mount propagation see: 2002 2003 Documentation/filesystems/sharedsubtree.rst 2004 2005 20063.6 /proc/<pid>/comm & /proc/<pid>/task/<tid>/comm 2007-------------------------------------------------------- 2008These files provide a method to access a task's comm value. It also allows for 2009a task to set its own or one of its thread siblings comm value. The comm value 2010is limited in size compared to the cmdline value, so writing anything longer 2011then the kernel's TASK_COMM_LEN (currently 16 chars, including the NUL 2012terminator) will result in a truncated comm value. 2013 2014 20153.7 /proc/<pid>/task/<tid>/children - Information about task children 2016------------------------------------------------------------------------- 2017This file provides a fast way to retrieve first level children pids 2018of a task pointed by <pid>/<tid> pair. The format is a space separated 2019stream of pids. 2020 2021Note the "first level" here -- if a child has its own children they will 2022not be listed here; one needs to read /proc/<children-pid>/task/<tid>/children 2023to obtain the descendants. 2024 2025Since this interface is intended to be fast and cheap it doesn't 2026guarantee to provide precise results and some children might be 2027skipped, especially if they've exited right after we printed their 2028pids, so one needs to either stop or freeze processes being inspected 2029if precise results are needed. 2030 2031 20323.8 /proc/<pid>/fdinfo/<fd> - Information about opened file 2033--------------------------------------------------------------- 2034This file provides information associated with an opened file. The regular 2035files have at least four fields -- 'pos', 'flags', 'mnt_id' and 'ino'. 2036The 'pos' represents the current offset of the opened file in decimal 2037form [see lseek(2) for details], 'flags' denotes the octal O_xxx mask the 2038file has been created with [see open(2) for details] and 'mnt_id' represents 2039mount ID of the file system containing the opened file [see 3.5 2040/proc/<pid>/mountinfo for details]. 'ino' represents the inode number of 2041the file. 2042 2043A typical output is:: 2044 2045 pos: 0 2046 flags: 0100002 2047 mnt_id: 19 2048 ino: 63107 2049 2050All locks associated with a file descriptor are shown in its fdinfo too:: 2051 2052 lock: 1: FLOCK ADVISORY WRITE 359 00:13:11691 0 EOF 2053 2054The files such as eventfd, fsnotify, signalfd, epoll among the regular pos/flags 2055pair provide additional information particular to the objects they represent. 2056 2057Eventfd files 2058~~~~~~~~~~~~~ 2059 2060:: 2061 2062 pos: 0 2063 flags: 04002 2064 mnt_id: 9 2065 ino: 63107 2066 eventfd-count: 5a 2067 2068where 'eventfd-count' is hex value of a counter. 2069 2070Signalfd files 2071~~~~~~~~~~~~~~ 2072 2073:: 2074 2075 pos: 0 2076 flags: 04002 2077 mnt_id: 9 2078 ino: 63107 2079 sigmask: 0000000000000200 2080 2081where 'sigmask' is hex value of the signal mask associated 2082with a file. 2083 2084Epoll files 2085~~~~~~~~~~~ 2086 2087:: 2088 2089 pos: 0 2090 flags: 02 2091 mnt_id: 9 2092 ino: 63107 2093 tfd: 5 events: 1d data: ffffffffffffffff pos:0 ino:61af sdev:7 2094 2095where 'tfd' is a target file descriptor number in decimal form, 2096'events' is events mask being watched and the 'data' is data 2097associated with a target [see epoll(7) for more details]. 2098 2099The 'pos' is current offset of the target file in decimal form 2100[see lseek(2)], 'ino' and 'sdev' are inode and device numbers 2101where target file resides, all in hex format. 2102 2103Fsnotify files 2104~~~~~~~~~~~~~~ 2105For inotify files the format is the following:: 2106 2107 pos: 0 2108 flags: 02000000 2109 mnt_id: 9 2110 ino: 63107 2111 inotify wd:3 ino:9e7e sdev:800013 mask:800afce ignored_mask:0 fhandle-bytes:8 fhandle-type:1 f_handle:7e9e0000640d1b6d 2112 2113where 'wd' is a watch descriptor in decimal form, i.e. a target file 2114descriptor number, 'ino' and 'sdev' are inode and device where the 2115target file resides and the 'mask' is the mask of events, all in hex 2116form [see inotify(7) for more details]. 2117 2118If the kernel was built with exportfs support, the path to the target 2119file is encoded as a file handle. The file handle is provided by three 2120fields 'fhandle-bytes', 'fhandle-type' and 'f_handle', all in hex 2121format. 2122 2123If the kernel is built without exportfs support the file handle won't be 2124printed out. 2125 2126If there is no inotify mark attached yet the 'inotify' line will be omitted. 2127 2128For fanotify files the format is:: 2129 2130 pos: 0 2131 flags: 02 2132 mnt_id: 9 2133 ino: 63107 2134 fanotify flags:10 event-flags:0 2135 fanotify mnt_id:12 mflags:40 mask:38 ignored_mask:40000003 2136 fanotify ino:4f969 sdev:800013 mflags:0 mask:3b ignored_mask:40000000 fhandle-bytes:8 fhandle-type:1 f_handle:69f90400c275b5b4 2137 2138where fanotify 'flags' and 'event-flags' are values used in fanotify_init 2139call, 'mnt_id' is the mount point identifier, 'mflags' is the value of 2140flags associated with mark which are tracked separately from events 2141mask. 'ino' and 'sdev' are target inode and device, 'mask' is the events 2142mask and 'ignored_mask' is the mask of events which are to be ignored. 2143All are in hex format. Incorporation of 'mflags', 'mask' and 'ignored_mask' 2144provide information about flags and mask used in fanotify_mark 2145call [see fsnotify manpage for details]. 2146 2147While the first three lines are mandatory and always printed, the rest is 2148optional and may be omitted if no marks created yet. 2149 2150Timerfd files 2151~~~~~~~~~~~~~ 2152 2153:: 2154 2155 pos: 0 2156 flags: 02 2157 mnt_id: 9 2158 ino: 63107 2159 clockid: 0 2160 ticks: 0 2161 settime flags: 01 2162 it_value: (0, 49406829) 2163 it_interval: (1, 0) 2164 2165where 'clockid' is the clock type and 'ticks' is the number of the timer expirations 2166that have occurred [see timerfd_create(2) for details]. 'settime flags' are 2167flags in octal form been used to setup the timer [see timerfd_settime(2) for 2168details]. 'it_value' is remaining time until the timer expiration. 2169'it_interval' is the interval for the timer. Note the timer might be set up 2170with TIMER_ABSTIME option which will be shown in 'settime flags', but 'it_value' 2171still exhibits timer's remaining time. 2172 2173DMA Buffer files 2174~~~~~~~~~~~~~~~~ 2175 2176:: 2177 2178 pos: 0 2179 flags: 04002 2180 mnt_id: 9 2181 ino: 63107 2182 size: 32768 2183 count: 2 2184 exp_name: system-heap 2185 2186where 'size' is the size of the DMA buffer in bytes. 'count' is the file count of 2187the DMA buffer file. 'exp_name' is the name of the DMA buffer exporter. 2188 2189VFIO Device files 2190~~~~~~~~~~~~~~~~~ 2191 2192:: 2193 2194 pos: 0 2195 flags: 02000002 2196 mnt_id: 17 2197 ino: 5122 2198 vfio-device-syspath: /sys/devices/pci0000:e0/0000:e0:01.1/0000:e1:00.0/0000:e2:05.0/0000:e8:00.0 2199 2200where 'vfio-device-syspath' is the sysfs path corresponding to the VFIO device 2201file. 2202 22033.9 /proc/<pid>/map_files - Information about memory mapped files 2204--------------------------------------------------------------------- 2205This directory contains symbolic links which represent memory mapped files 2206the process is maintaining. Example output:: 2207 2208 | lr-------- 1 root root 64 Jan 27 11:24 333c600000-333c620000 -> /usr/lib64/ld-2.18.so 2209 | lr-------- 1 root root 64 Jan 27 11:24 333c81f000-333c820000 -> /usr/lib64/ld-2.18.so 2210 | lr-------- 1 root root 64 Jan 27 11:24 333c820000-333c821000 -> /usr/lib64/ld-2.18.so 2211 | ... 2212 | lr-------- 1 root root 64 Jan 27 11:24 35d0421000-35d0422000 -> /usr/lib64/libselinux.so.1 2213 | lr-------- 1 root root 64 Jan 27 11:24 400000-41a000 -> /usr/bin/ls 2214 2215The name of a link represents the virtual memory bounds of a mapping, i.e. 2216vm_area_struct::vm_start - vm_area_struct::vm_end. 2217 2218The main purpose of the map_files is to retrieve a set of memory mapped 2219files in a fast way instead of parsing /proc/<pid>/maps or 2220/proc/<pid>/smaps, both of which contain many more records. At the same 2221time one can open(2) mappings from the listings of two processes and 2222comparing their inode numbers to figure out which anonymous memory areas 2223are actually shared. 2224 22253.10 /proc/<pid>/timerslack_ns - Task timerslack value 2226--------------------------------------------------------- 2227This file provides the value of the task's timerslack value in nanoseconds. 2228This value specifies an amount of time that normal timers may be deferred 2229in order to coalesce timers and avoid unnecessary wakeups. 2230 2231This allows a task's interactivity vs power consumption tradeoff to be 2232adjusted. 2233 2234Writing 0 to the file will set the task's timerslack to the default value. 2235 2236Valid values are from 0 - ULLONG_MAX 2237 2238An application setting the value must have PTRACE_MODE_ATTACH_FSCREDS level 2239permissions on the task specified to change its timerslack_ns value. 2240 22413.11 /proc/<pid>/patch_state - Livepatch patch operation state 2242----------------------------------------------------------------- 2243When CONFIG_LIVEPATCH is enabled, this file displays the value of the 2244patch state for the task. 2245 2246A value of '-1' indicates that no patch is in transition. 2247 2248A value of '0' indicates that a patch is in transition and the task is 2249unpatched. If the patch is being enabled, then the task hasn't been 2250patched yet. If the patch is being disabled, then the task has already 2251been unpatched. 2252 2253A value of '1' indicates that a patch is in transition and the task is 2254patched. If the patch is being enabled, then the task has already been 2255patched. If the patch is being disabled, then the task hasn't been 2256unpatched yet. 2257 22583.12 /proc/<pid>/arch_status - task architecture specific status 2259------------------------------------------------------------------- 2260When CONFIG_PROC_PID_ARCH_STATUS is enabled, this file displays the 2261architecture specific status of the task. 2262 2263Example 2264~~~~~~~ 2265 2266:: 2267 2268 $ cat /proc/6753/arch_status 2269 AVX512_elapsed_ms: 8 2270 2271Description 2272~~~~~~~~~~~ 2273 2274x86 specific entries 2275~~~~~~~~~~~~~~~~~~~~~ 2276 2277AVX512_elapsed_ms 2278^^^^^^^^^^^^^^^^^^ 2279 2280 If AVX512 is supported on the machine, this entry shows the milliseconds 2281 elapsed since the last time AVX512 usage was recorded. The recording 2282 happens on a best effort basis when a task is scheduled out. This means 2283 that the value depends on two factors: 2284 2285 1) The time which the task spent on the CPU without being scheduled 2286 out. With CPU isolation and a single runnable task this can take 2287 several seconds. 2288 2289 2) The time since the task was scheduled out last. Depending on the 2290 reason for being scheduled out (time slice exhausted, syscall ...) 2291 this can be arbitrary long time. 2292 2293 As a consequence the value cannot be considered precise and authoritative 2294 information. The application which uses this information has to be aware 2295 of the overall scenario on the system in order to determine whether a 2296 task is a real AVX512 user or not. Precise information can be obtained 2297 with performance counters. 2298 2299 A special value of '-1' indicates that no AVX512 usage was recorded, thus 2300 the task is unlikely an AVX512 user, but depends on the workload and the 2301 scheduling scenario, it also could be a false negative mentioned above. 2302 23033.13 /proc/<pid>/fd - List of symlinks to open files 2304------------------------------------------------------- 2305This directory contains symbolic links which represent open files 2306the process is maintaining. Example output:: 2307 2308 lr-x------ 1 root root 64 Sep 20 17:53 0 -> /dev/null 2309 l-wx------ 1 root root 64 Sep 20 17:53 1 -> /dev/null 2310 lrwx------ 1 root root 64 Sep 20 17:53 10 -> 'socket:[12539]' 2311 lrwx------ 1 root root 64 Sep 20 17:53 11 -> 'socket:[12540]' 2312 lrwx------ 1 root root 64 Sep 20 17:53 12 -> 'socket:[12542]' 2313 2314The number of open files for the process is stored in 'size' member 2315of stat() output for /proc/<pid>/fd for fast access. 2316------------------------------------------------------- 2317 23183.14 /proc/<pid>/ksm_stat - Information about the process's ksm status 2319---------------------------------------------------------------------- 2320When CONFIG_KSM is enabled, each process has this file which displays 2321the information of ksm merging status. 2322 2323Example 2324~~~~~~~ 2325 2326:: 2327 2328 / # cat /proc/self/ksm_stat 2329 ksm_rmap_items 0 2330 ksm_zero_pages 0 2331 ksm_merging_pages 0 2332 ksm_process_profit 0 2333 ksm_merge_any: no 2334 ksm_mergeable: no 2335 2336Description 2337~~~~~~~~~~~ 2338 2339ksm_rmap_items 2340^^^^^^^^^^^^^^ 2341 2342The number of ksm_rmap_item structures in use. The structure 2343ksm_rmap_item stores the reverse mapping information for virtual 2344addresses. KSM will generate a ksm_rmap_item for each ksm-scanned page of 2345the process. 2346 2347ksm_zero_pages 2348^^^^^^^^^^^^^^ 2349 2350When /sys/kernel/mm/ksm/use_zero_pages is enabled, it represent how many 2351empty pages are merged with kernel zero pages by KSM. 2352 2353ksm_merging_pages 2354^^^^^^^^^^^^^^^^^ 2355 2356It represents how many pages of this process are involved in KSM merging 2357(not including ksm_zero_pages). It is the same with what 2358/proc/<pid>/ksm_merging_pages shows. 2359 2360ksm_process_profit 2361^^^^^^^^^^^^^^^^^^ 2362 2363The profit that KSM brings (Saved bytes). KSM can save memory by merging 2364identical pages, but also can consume additional memory, because it needs 2365to generate a number of rmap_items to save each scanned page's brief rmap 2366information. Some of these pages may be merged, but some may not be abled 2367to be merged after being checked several times, which are unprofitable 2368memory consumed. 2369 2370ksm_merge_any 2371^^^^^^^^^^^^^ 2372 2373It specifies whether the process's 'mm is added by prctl() into the 2374candidate list of KSM or not, and if KSM scanning is fully enabled at 2375process level. 2376 2377ksm_mergeable 2378^^^^^^^^^^^^^ 2379 2380It specifies whether any VMAs of the process''s mms are currently 2381applicable to KSM. 2382 2383More information about KSM can be found in 2384Documentation/admin-guide/mm/ksm.rst. 2385 2386 2387Chapter 4: Configuring procfs 2388============================= 2389 23904.1 Mount options 2391--------------------- 2392 2393The following mount options are supported: 2394 2395 ========= ======================================================== 2396 hidepid= Set /proc/<pid>/ access mode. 2397 gid= Set the group authorized to learn processes information. 2398 subset= Show only the specified subset of procfs. 2399 pidns= Specify a the namespace used by this procfs. 2400 ========= ======================================================== 2401 2402hidepid=off or hidepid=0 means classic mode - everybody may access all 2403/proc/<pid>/ directories (default). 2404 2405hidepid=noaccess or hidepid=1 means users may not access any /proc/<pid>/ 2406directories but their own. Sensitive files like cmdline, sched*, status are now 2407protected against other users. This makes it impossible to learn whether any 2408user runs specific program (given the program doesn't reveal itself by its 2409behaviour). As an additional bonus, as /proc/<pid>/cmdline is unaccessible for 2410other users, poorly written programs passing sensitive information via program 2411arguments are now protected against local eavesdroppers. 2412 2413hidepid=invisible or hidepid=2 means hidepid=1 plus all /proc/<pid>/ will be 2414fully invisible to other users. It doesn't mean that it hides a fact whether a 2415process with a specific pid value exists (it can be learned by other means, e.g. 2416by "kill -0 $PID"), but it hides process's uid and gid, which may be learned by 2417stat()'ing /proc/<pid>/ otherwise. It greatly complicates an intruder's task of 2418gathering information about running processes, whether some daemon runs with 2419elevated privileges, whether other user runs some sensitive program, whether 2420other users run any program at all, etc. 2421 2422hidepid=ptraceable or hidepid=4 means that procfs should only contain 2423/proc/<pid>/ directories that the caller can ptrace. 2424 2425gid= defines a group authorized to learn processes information otherwise 2426prohibited by hidepid=. If you use some daemon like identd which needs to learn 2427information about processes information, just add identd to this group. 2428 2429subset=pid hides all top level files and directories in the procfs that 2430are not related to tasks. This option cannot be changed on an existing 2431procfs instance because overmounts that existed before the change could 2432otherwise remain reachable after the top level procfs entries are hidden. 2433 2434pidns= specifies a pid namespace (either as a string path to something like 2435`/proc/$pid/ns/pid`, or a file descriptor when using `FSCONFIG_SET_FD`) that 2436will be used by the procfs instance when translating pids. By default, procfs 2437will use the calling process's active pid namespace. Note that the pid 2438namespace of an existing procfs instance cannot be modified (attempting to do 2439so will give an `-EBUSY` error). 2440 24414.2 Mount restrictions 2442-------------------------- 2443 2444If user namespaces are in use, the kernel additionally checks the instances of 2445procfs available to the mounter and will not allow procfs to be mounted if: 2446 2447 1. This mount is not fully visible unless the new procfs is going to be 2448 mounted with subset=pid option. 2449 2450 a. Its root directory is not the root directory of the filesystem. 2451 b. If any file or non-empty procfs directory is hidden by another mount. 2452 2453 2. A new mount overrides the readonly option or any option from atime family. 2454 2455Chapter 5: Filesystem behavior 2456============================== 2457 2458Originally, before the advent of pid namespace, procfs was a global file 2459system. It means that there was only one procfs instance in the system. 2460 2461When pid namespace was added, a separate procfs instance was mounted in 2462each pid namespace. So, procfs mount options are global among all 2463mountpoints within the same namespace:: 2464 2465 # grep ^proc /proc/mounts 2466 proc /proc proc rw,relatime,hidepid=2 0 0 2467 2468 # strace -e mount mount -o hidepid=1 -t proc proc /tmp/proc 2469 mount("proc", "/tmp/proc", "proc", 0, "hidepid=1") = 0 2470 +++ exited with 0 +++ 2471 2472 # grep ^proc /proc/mounts 2473 proc /proc proc rw,relatime,hidepid=2 0 0 2474 proc /tmp/proc proc rw,relatime,hidepid=2 0 0 2475 2476and only after remounting procfs mount options will change at all 2477mountpoints:: 2478 2479 # mount -o remount,hidepid=1 -t proc proc /tmp/proc 2480 2481 # grep ^proc /proc/mounts 2482 proc /proc proc rw,relatime,hidepid=1 0 0 2483 proc /tmp/proc proc rw,relatime,hidepid=1 0 0 2484 2485This behavior is different from the behavior of other filesystems. 2486 2487The new procfs behavior is more like other filesystems. Each procfs mount 2488creates a new procfs instance. Mount options affect own procfs instance. 2489It means that it became possible to have several procfs instances 2490displaying tasks with different filtering options in one pid namespace:: 2491 2492 # mount -o hidepid=invisible -t proc proc /proc 2493 # mount -o hidepid=noaccess -t proc proc /tmp/proc 2494 # grep ^proc /proc/mounts 2495 proc /proc proc rw,relatime,hidepid=invisible 0 0 2496 proc /tmp/proc proc rw,relatime,hidepid=noaccess 0 0 2497