1------------------------------------------------------------------------------ 2 T H E /proc F I L E S Y S T E M 3------------------------------------------------------------------------------ 4/proc/sys Terrehon Bowden <terrehon@pacbell.net> October 7 1999 5 Bodo Bauer <bb@ricochet.net> 6 72.4.x update Jorge Nerin <comandante@zaralinux.com> November 14 2000 8------------------------------------------------------------------------------ 9Version 1.3 Kernel version 2.2.12 10 Kernel version 2.4.0-test11-pre4 11------------------------------------------------------------------------------ 12 13Table of Contents 14----------------- 15 16 0 Preface 17 0.1 Introduction/Credits 18 0.2 Legal Stuff 19 20 1 Collecting System Information 21 1.1 Process-Specific Subdirectories 22 1.2 Kernel data 23 1.3 IDE devices in /proc/ide 24 1.4 Networking info in /proc/net 25 1.5 SCSI info 26 1.6 Parallel port info in /proc/parport 27 1.7 TTY info in /proc/tty 28 1.8 Miscellaneous kernel statistics in /proc/stat 29 30 2 Modifying System Parameters 31 2.1 /proc/sys/fs - File system data 32 2.2 /proc/sys/fs/binfmt_misc - Miscellaneous binary formats 33 2.3 /proc/sys/kernel - general kernel parameters 34 2.4 /proc/sys/vm - The virtual memory subsystem 35 2.5 /proc/sys/dev - Device specific parameters 36 2.6 /proc/sys/sunrpc - Remote procedure calls 37 2.7 /proc/sys/net - Networking stuff 38 2.8 /proc/sys/net/ipv4 - IPV4 settings 39 2.9 Appletalk 40 2.10 IPX 41 2.11 /proc/sys/fs/mqueue - POSIX message queues filesystem 42 2.12 /proc/<pid>/oom_adj - Adjust the oom-killer score 43 2.13 /proc/<pid>/oom_score - Display current oom-killer score 44 2.14 /proc/<pid>/io - Display the IO accounting fields 45 46------------------------------------------------------------------------------ 47Preface 48------------------------------------------------------------------------------ 49 500.1 Introduction/Credits 51------------------------ 52 53This documentation is part of a soon (or so we hope) to be released book on 54the SuSE Linux distribution. As there is no complete documentation for the 55/proc file system and we've used many freely available sources to write these 56chapters, it seems only fair to give the work back to the Linux community. 57This work is based on the 2.2.* kernel version and the upcoming 2.4.*. I'm 58afraid it's still far from complete, but we hope it will be useful. As far as 59we know, it is the first 'all-in-one' document about the /proc file system. It 60is focused on the Intel x86 hardware, so if you are looking for PPC, ARM, 61SPARC, AXP, etc., features, you probably won't find what you are looking for. 62It also only covers IPv4 networking, not IPv6 nor other protocols - sorry. But 63additions and patches are welcome and will be added to this document if you 64mail them to Bodo. 65 66We'd like to thank Alan Cox, Rik van Riel, and Alexey Kuznetsov and a lot of 67other people for help compiling this documentation. We'd also like to extend a 68special thank you to Andi Kleen for documentation, which we relied on heavily 69to create this document, as well as the additional information he provided. 70Thanks to everybody else who contributed source or docs to the Linux kernel 71and helped create a great piece of software... :) 72 73If you have any comments, corrections or additions, please don't hesitate to 74contact Bodo Bauer at bb@ricochet.net. We'll be happy to add them to this 75document. 76 77The latest version of this document is available online at 78http://skaro.nightcrawler.com/~bb/Docs/Proc as HTML version. 79 80If the above direction does not works for you, ypu could try the kernel 81mailing list at linux-kernel@vger.kernel.org and/or try to reach me at 82comandante@zaralinux.com. 83 840.2 Legal Stuff 85--------------- 86 87We don't guarantee the correctness of this document, and if you come to us 88complaining about how you screwed up your system because of incorrect 89documentation, we won't feel responsible... 90 91------------------------------------------------------------------------------ 92CHAPTER 1: COLLECTING SYSTEM INFORMATION 93------------------------------------------------------------------------------ 94 95------------------------------------------------------------------------------ 96In This Chapter 97------------------------------------------------------------------------------ 98* Investigating the properties of the pseudo file system /proc and its 99 ability to provide information on the running Linux system 100* Examining /proc's structure 101* Uncovering various information about the kernel and the processes running 102 on the system 103------------------------------------------------------------------------------ 104 105 106The proc file system acts as an interface to internal data structures in the 107kernel. It can be used to obtain information about the system and to change 108certain kernel parameters at runtime (sysctl). 109 110First, we'll take a look at the read-only parts of /proc. In Chapter 2, we 111show you how you can use /proc/sys to change settings. 112 1131.1 Process-Specific Subdirectories 114----------------------------------- 115 116The directory /proc contains (among other things) one subdirectory for each 117process running on the system, which is named after the process ID (PID). 118 119The link self points to the process reading the file system. Each process 120subdirectory has the entries listed in Table 1-1. 121 122 123Table 1-1: Process specific entries in /proc 124.............................................................................. 125 File Content 126 clear_refs Clears page referenced bits shown in smaps output 127 cmdline Command line arguments 128 cpu Current and last cpu in which it was executed (2.4)(smp) 129 cwd Link to the current working directory 130 environ Values of environment variables 131 exe Link to the executable of this process 132 fd Directory, which contains all file descriptors 133 maps Memory maps to executables and library files (2.4) 134 mem Memory held by this process 135 root Link to the root directory of this process 136 stat Process status 137 statm Process memory status information 138 status Process status in human readable form 139 wchan If CONFIG_KALLSYMS is set, a pre-decoded wchan 140 smaps Extension based on maps, the rss size for each mapped file 141.............................................................................. 142 143For example, to get the status information of a process, all you have to do is 144read the file /proc/PID/status: 145 146 >cat /proc/self/status 147 Name: cat 148 State: R (running) 149 Pid: 5452 150 PPid: 743 151 TracerPid: 0 (2.4) 152 Uid: 501 501 501 501 153 Gid: 100 100 100 100 154 Groups: 100 14 16 155 VmSize: 1112 kB 156 VmLck: 0 kB 157 VmRSS: 348 kB 158 VmData: 24 kB 159 VmStk: 12 kB 160 VmExe: 8 kB 161 VmLib: 1044 kB 162 SigPnd: 0000000000000000 163 SigBlk: 0000000000000000 164 SigIgn: 0000000000000000 165 SigCgt: 0000000000000000 166 CapInh: 00000000fffffeff 167 CapPrm: 0000000000000000 168 CapEff: 0000000000000000 169 170 171This shows you nearly the same information you would get if you viewed it with 172the ps command. In fact, ps uses the proc file system to obtain its 173information. The statm file contains more detailed information about the 174process memory usage. Its seven fields are explained in Table 1-2. 175 176 177Table 1-2: Contents of the statm files (as of 2.6.8-rc3) 178.............................................................................. 179 Field Content 180 size total program size (pages) (same as VmSize in status) 181 resident size of memory portions (pages) (same as VmRSS in status) 182 shared number of pages that are shared (i.e. backed by a file) 183 trs number of pages that are 'code' (not including libs; broken, 184 includes data segment) 185 lrs number of pages of library (always 0 on 2.6) 186 drs number of pages of data/stack (including libs; broken, 187 includes library text) 188 dt number of dirty pages (always 0 on 2.6) 189.............................................................................. 190 1911.2 Kernel data 192--------------- 193 194Similar to the process entries, the kernel data files give information about 195the running kernel. The files used to obtain this information are contained in 196/proc and are listed in Table 1-3. Not all of these will be present in your 197system. It depends on the kernel configuration and the loaded modules, which 198files are there, and which are missing. 199 200Table 1-3: Kernel info in /proc 201.............................................................................. 202 File Content 203 apm Advanced power management info 204 buddyinfo Kernel memory allocator information (see text) (2.5) 205 bus Directory containing bus specific information 206 cmdline Kernel command line 207 cpuinfo Info about the CPU 208 devices Available devices (block and character) 209 dma Used DMS channels 210 filesystems Supported filesystems 211 driver Various drivers grouped here, currently rtc (2.4) 212 execdomains Execdomains, related to security (2.4) 213 fb Frame Buffer devices (2.4) 214 fs File system parameters, currently nfs/exports (2.4) 215 ide Directory containing info about the IDE subsystem 216 interrupts Interrupt usage 217 iomem Memory map (2.4) 218 ioports I/O port usage 219 irq Masks for irq to cpu affinity (2.4)(smp?) 220 isapnp ISA PnP (Plug&Play) Info (2.4) 221 kcore Kernel core image (can be ELF or A.OUT(deprecated in 2.4)) 222 kmsg Kernel messages 223 ksyms Kernel symbol table 224 loadavg Load average of last 1, 5 & 15 minutes 225 locks Kernel locks 226 meminfo Memory info 227 misc Miscellaneous 228 modules List of loaded modules 229 mounts Mounted filesystems 230 net Networking info (see text) 231 partitions Table of partitions known to the system 232 pci Deprecated info of PCI bus (new way -> /proc/bus/pci/, 233 decoupled by lspci (2.4) 234 rtc Real time clock 235 scsi SCSI info (see text) 236 slabinfo Slab pool info 237 stat Overall statistics 238 swaps Swap space utilization 239 sys See chapter 2 240 sysvipc Info of SysVIPC Resources (msg, sem, shm) (2.4) 241 tty Info of tty drivers 242 uptime System uptime 243 version Kernel version 244 video bttv info of video resources (2.4) 245.............................................................................. 246 247You can, for example, check which interrupts are currently in use and what 248they are used for by looking in the file /proc/interrupts: 249 250 > cat /proc/interrupts 251 CPU0 252 0: 8728810 XT-PIC timer 253 1: 895 XT-PIC keyboard 254 2: 0 XT-PIC cascade 255 3: 531695 XT-PIC aha152x 256 4: 2014133 XT-PIC serial 257 5: 44401 XT-PIC pcnet_cs 258 8: 2 XT-PIC rtc 259 11: 8 XT-PIC i82365 260 12: 182918 XT-PIC PS/2 Mouse 261 13: 1 XT-PIC fpu 262 14: 1232265 XT-PIC ide0 263 15: 7 XT-PIC ide1 264 NMI: 0 265 266In 2.4.* a couple of lines where added to this file LOC & ERR (this time is the 267output of a SMP machine): 268 269 > cat /proc/interrupts 270 271 CPU0 CPU1 272 0: 1243498 1214548 IO-APIC-edge timer 273 1: 8949 8958 IO-APIC-edge keyboard 274 2: 0 0 XT-PIC cascade 275 5: 11286 10161 IO-APIC-edge soundblaster 276 8: 1 0 IO-APIC-edge rtc 277 9: 27422 27407 IO-APIC-edge 3c503 278 12: 113645 113873 IO-APIC-edge PS/2 Mouse 279 13: 0 0 XT-PIC fpu 280 14: 22491 24012 IO-APIC-edge ide0 281 15: 2183 2415 IO-APIC-edge ide1 282 17: 30564 30414 IO-APIC-level eth0 283 18: 177 164 IO-APIC-level bttv 284 NMI: 2457961 2457959 285 LOC: 2457882 2457881 286 ERR: 2155 287 288NMI is incremented in this case because every timer interrupt generates a NMI 289(Non Maskable Interrupt) which is used by the NMI Watchdog to detect lockups. 290 291LOC is the local interrupt counter of the internal APIC of every CPU. 292 293ERR is incremented in the case of errors in the IO-APIC bus (the bus that 294connects the CPUs in a SMP system. This means that an error has been detected, 295the IO-APIC automatically retry the transmission, so it should not be a big 296problem, but you should read the SMP-FAQ. 297 298In this context it could be interesting to note the new irq directory in 2.4. 299It could be used to set IRQ to CPU affinity, this means that you can "hook" an 300IRQ to only one CPU, or to exclude a CPU of handling IRQs. The contents of the 301irq subdir is one subdir for each IRQ, and one file; prof_cpu_mask 302 303For example 304 > ls /proc/irq/ 305 0 10 12 14 16 18 2 4 6 8 prof_cpu_mask 306 1 11 13 15 17 19 3 5 7 9 307 > ls /proc/irq/0/ 308 smp_affinity 309 310The contents of the prof_cpu_mask file and each smp_affinity file for each IRQ 311is the same by default: 312 313 > cat /proc/irq/0/smp_affinity 314 ffffffff 315 316It's a bitmask, in which you can specify which CPUs can handle the IRQ, you can 317set it by doing: 318 319 > echo 1 > /proc/irq/prof_cpu_mask 320 321This means that only the first CPU will handle the IRQ, but you can also echo 5 322which means that only the first and fourth CPU can handle the IRQ. 323 324The way IRQs are routed is handled by the IO-APIC, and it's Round Robin 325between all the CPUs which are allowed to handle it. As usual the kernel has 326more info than you and does a better job than you, so the defaults are the 327best choice for almost everyone. 328 329There are three more important subdirectories in /proc: net, scsi, and sys. 330The general rule is that the contents, or even the existence of these 331directories, depend on your kernel configuration. If SCSI is not enabled, the 332directory scsi may not exist. The same is true with the net, which is there 333only when networking support is present in the running kernel. 334 335The slabinfo file gives information about memory usage at the slab level. 336Linux uses slab pools for memory management above page level in version 2.2. 337Commonly used objects have their own slab pool (such as network buffers, 338directory cache, and so on). 339 340.............................................................................. 341 342> cat /proc/buddyinfo 343 344Node 0, zone DMA 0 4 5 4 4 3 ... 345Node 0, zone Normal 1 0 0 1 101 8 ... 346Node 0, zone HighMem 2 0 0 1 1 0 ... 347 348Memory fragmentation is a problem under some workloads, and buddyinfo is a 349useful tool for helping diagnose these problems. Buddyinfo will give you a 350clue as to how big an area you can safely allocate, or why a previous 351allocation failed. 352 353Each column represents the number of pages of a certain order which are 354available. In this case, there are 0 chunks of 2^0*PAGE_SIZE available in 355ZONE_DMA, 4 chunks of 2^1*PAGE_SIZE in ZONE_DMA, 101 chunks of 2^4*PAGE_SIZE 356available in ZONE_NORMAL, etc... 357 358.............................................................................. 359 360meminfo: 361 362Provides information about distribution and utilization of memory. This 363varies by architecture and compile options. The following is from a 36416GB PIII, which has highmem enabled. You may not have all of these fields. 365 366> cat /proc/meminfo 367 368 369MemTotal: 16344972 kB 370MemFree: 13634064 kB 371Buffers: 3656 kB 372Cached: 1195708 kB 373SwapCached: 0 kB 374Active: 891636 kB 375Inactive: 1077224 kB 376HighTotal: 15597528 kB 377HighFree: 13629632 kB 378LowTotal: 747444 kB 379LowFree: 4432 kB 380SwapTotal: 0 kB 381SwapFree: 0 kB 382Dirty: 968 kB 383Writeback: 0 kB 384Mapped: 280372 kB 385Slab: 684068 kB 386CommitLimit: 7669796 kB 387Committed_AS: 100056 kB 388PageTables: 24448 kB 389VmallocTotal: 112216 kB 390VmallocUsed: 428 kB 391VmallocChunk: 111088 kB 392 393 MemTotal: Total usable ram (i.e. physical ram minus a few reserved 394 bits and the kernel binary code) 395 MemFree: The sum of LowFree+HighFree 396 Buffers: Relatively temporary storage for raw disk blocks 397 shouldn't get tremendously large (20MB or so) 398 Cached: in-memory cache for files read from the disk (the 399 pagecache). Doesn't include SwapCached 400 SwapCached: Memory that once was swapped out, is swapped back in but 401 still also is in the swapfile (if memory is needed it 402 doesn't need to be swapped out AGAIN because it is already 403 in the swapfile. This saves I/O) 404 Active: Memory that has been used more recently and usually not 405 reclaimed unless absolutely necessary. 406 Inactive: Memory which has been less recently used. It is more 407 eligible to be reclaimed for other purposes 408 HighTotal: 409 HighFree: Highmem is all memory above ~860MB of physical memory 410 Highmem areas are for use by userspace programs, or 411 for the pagecache. The kernel must use tricks to access 412 this memory, making it slower to access than lowmem. 413 LowTotal: 414 LowFree: Lowmem is memory which can be used for everything that 415 highmem can be used for, but it is also available for the 416 kernel's use for its own data structures. Among many 417 other things, it is where everything from the Slab is 418 allocated. Bad things happen when you're out of lowmem. 419 SwapTotal: total amount of swap space available 420 SwapFree: Memory which has been evicted from RAM, and is temporarily 421 on the disk 422 Dirty: Memory which is waiting to get written back to the disk 423 Writeback: Memory which is actively being written back to the disk 424 Mapped: files which have been mmaped, such as libraries 425 Slab: in-kernel data structures cache 426 CommitLimit: Based on the overcommit ratio ('vm.overcommit_ratio'), 427 this is the total amount of memory currently available to 428 be allocated on the system. This limit is only adhered to 429 if strict overcommit accounting is enabled (mode 2 in 430 'vm.overcommit_memory'). 431 The CommitLimit is calculated with the following formula: 432 CommitLimit = ('vm.overcommit_ratio' * Physical RAM) + Swap 433 For example, on a system with 1G of physical RAM and 7G 434 of swap with a `vm.overcommit_ratio` of 30 it would 435 yield a CommitLimit of 7.3G. 436 For more details, see the memory overcommit documentation 437 in vm/overcommit-accounting. 438Committed_AS: The amount of memory presently allocated on the system. 439 The committed memory is a sum of all of the memory which 440 has been allocated by processes, even if it has not been 441 "used" by them as of yet. A process which malloc()'s 1G 442 of memory, but only touches 300M of it will only show up 443 as using 300M of memory even if it has the address space 444 allocated for the entire 1G. This 1G is memory which has 445 been "committed" to by the VM and can be used at any time 446 by the allocating application. With strict overcommit 447 enabled on the system (mode 2 in 'vm.overcommit_memory'), 448 allocations which would exceed the CommitLimit (detailed 449 above) will not be permitted. This is useful if one needs 450 to guarantee that processes will not fail due to lack of 451 memory once that memory has been successfully allocated. 452 PageTables: amount of memory dedicated to the lowest level of page 453 tables. 454VmallocTotal: total size of vmalloc memory area 455 VmallocUsed: amount of vmalloc area which is used 456VmallocChunk: largest contigious block of vmalloc area which is free 457 458 4591.3 IDE devices in /proc/ide 460---------------------------- 461 462The subdirectory /proc/ide contains information about all IDE devices of which 463the kernel is aware. There is one subdirectory for each IDE controller, the 464file drivers and a link for each IDE device, pointing to the device directory 465in the controller specific subtree. 466 467The file drivers contains general information about the drivers used for the 468IDE devices: 469 470 > cat /proc/ide/drivers 471 ide-cdrom version 4.53 472 ide-disk version 1.08 473 474More detailed information can be found in the controller specific 475subdirectories. These are named ide0, ide1 and so on. Each of these 476directories contains the files shown in table 1-4. 477 478 479Table 1-4: IDE controller info in /proc/ide/ide? 480.............................................................................. 481 File Content 482 channel IDE channel (0 or 1) 483 config Configuration (only for PCI/IDE bridge) 484 mate Mate name 485 model Type/Chipset of IDE controller 486.............................................................................. 487 488Each device connected to a controller has a separate subdirectory in the 489controllers directory. The files listed in table 1-5 are contained in these 490directories. 491 492 493Table 1-5: IDE device information 494.............................................................................. 495 File Content 496 cache The cache 497 capacity Capacity of the medium (in 512Byte blocks) 498 driver driver and version 499 geometry physical and logical geometry 500 identify device identify block 501 media media type 502 model device identifier 503 settings device setup 504 smart_thresholds IDE disk management thresholds 505 smart_values IDE disk management values 506.............................................................................. 507 508The most interesting file is settings. This file contains a nice overview of 509the drive parameters: 510 511 # cat /proc/ide/ide0/hda/settings 512 name value min max mode 513 ---- ----- --- --- ---- 514 bios_cyl 526 0 65535 rw 515 bios_head 255 0 255 rw 516 bios_sect 63 0 63 rw 517 breada_readahead 4 0 127 rw 518 bswap 0 0 1 r 519 file_readahead 72 0 2097151 rw 520 io_32bit 0 0 3 rw 521 keepsettings 0 0 1 rw 522 max_kb_per_request 122 1 127 rw 523 multcount 0 0 8 rw 524 nice1 1 0 1 rw 525 nowerr 0 0 1 rw 526 pio_mode write-only 0 255 w 527 slow 0 0 1 rw 528 unmaskirq 0 0 1 rw 529 using_dma 0 0 1 rw 530 531 5321.4 Networking info in /proc/net 533-------------------------------- 534 535The subdirectory /proc/net follows the usual pattern. Table 1-6 shows the 536additional values you get for IP version 6 if you configure the kernel to 537support this. Table 1-7 lists the files and their meaning. 538 539 540Table 1-6: IPv6 info in /proc/net 541.............................................................................. 542 File Content 543 udp6 UDP sockets (IPv6) 544 tcp6 TCP sockets (IPv6) 545 raw6 Raw device statistics (IPv6) 546 igmp6 IP multicast addresses, which this host joined (IPv6) 547 if_inet6 List of IPv6 interface addresses 548 ipv6_route Kernel routing table for IPv6 549 rt6_stats Global IPv6 routing tables statistics 550 sockstat6 Socket statistics (IPv6) 551 snmp6 Snmp data (IPv6) 552.............................................................................. 553 554 555Table 1-7: Network info in /proc/net 556.............................................................................. 557 File Content 558 arp Kernel ARP table 559 dev network devices with statistics 560 dev_mcast the Layer2 multicast groups a device is listening too 561 (interface index, label, number of references, number of bound 562 addresses). 563 dev_stat network device status 564 ip_fwchains Firewall chain linkage 565 ip_fwnames Firewall chain names 566 ip_masq Directory containing the masquerading tables 567 ip_masquerade Major masquerading table 568 netstat Network statistics 569 raw raw device statistics 570 route Kernel routing table 571 rpc Directory containing rpc info 572 rt_cache Routing cache 573 snmp SNMP data 574 sockstat Socket statistics 575 tcp TCP sockets 576 tr_rif Token ring RIF routing table 577 udp UDP sockets 578 unix UNIX domain sockets 579 wireless Wireless interface data (Wavelan etc) 580 igmp IP multicast addresses, which this host joined 581 psched Global packet scheduler parameters. 582 netlink List of PF_NETLINK sockets 583 ip_mr_vifs List of multicast virtual interfaces 584 ip_mr_cache List of multicast routing cache 585.............................................................................. 586 587You can use this information to see which network devices are available in 588your system and how much traffic was routed over those devices: 589 590 > cat /proc/net/dev 591 Inter-|Receive |[... 592 face |bytes packets errs drop fifo frame compressed multicast|[... 593 lo: 908188 5596 0 0 0 0 0 0 [... 594 ppp0:15475140 20721 410 0 0 410 0 0 [... 595 eth0: 614530 7085 0 0 0 0 0 1 [... 596 597 ...] Transmit 598 ...] bytes packets errs drop fifo colls carrier compressed 599 ...] 908188 5596 0 0 0 0 0 0 600 ...] 1375103 17405 0 0 0 0 0 0 601 ...] 1703981 5535 0 0 0 3 0 0 602 603In addition, each Channel Bond interface has it's own directory. For 604example, the bond0 device will have a directory called /proc/net/bond0/. 605It will contain information that is specific to that bond, such as the 606current slaves of the bond, the link status of the slaves, and how 607many times the slaves link has failed. 608 6091.5 SCSI info 610------------- 611 612If you have a SCSI host adapter in your system, you'll find a subdirectory 613named after the driver for this adapter in /proc/scsi. You'll also see a list 614of all recognized SCSI devices in /proc/scsi: 615 616 >cat /proc/scsi/scsi 617 Attached devices: 618 Host: scsi0 Channel: 00 Id: 00 Lun: 00 619 Vendor: IBM Model: DGHS09U Rev: 03E0 620 Type: Direct-Access ANSI SCSI revision: 03 621 Host: scsi0 Channel: 00 Id: 06 Lun: 00 622 Vendor: PIONEER Model: CD-ROM DR-U06S Rev: 1.04 623 Type: CD-ROM ANSI SCSI revision: 02 624 625 626The directory named after the driver has one file for each adapter found in 627the system. These files contain information about the controller, including 628the used IRQ and the IO address range. The amount of information shown is 629dependent on the adapter you use. The example shows the output for an Adaptec 630AHA-2940 SCSI adapter: 631 632 > cat /proc/scsi/aic7xxx/0 633 634 Adaptec AIC7xxx driver version: 5.1.19/3.2.4 635 Compile Options: 636 TCQ Enabled By Default : Disabled 637 AIC7XXX_PROC_STATS : Disabled 638 AIC7XXX_RESET_DELAY : 5 639 Adapter Configuration: 640 SCSI Adapter: Adaptec AHA-294X Ultra SCSI host adapter 641 Ultra Wide Controller 642 PCI MMAPed I/O Base: 0xeb001000 643 Adapter SEEPROM Config: SEEPROM found and used. 644 Adaptec SCSI BIOS: Enabled 645 IRQ: 10 646 SCBs: Active 0, Max Active 2, 647 Allocated 15, HW 16, Page 255 648 Interrupts: 160328 649 BIOS Control Word: 0x18b6 650 Adapter Control Word: 0x005b 651 Extended Translation: Enabled 652 Disconnect Enable Flags: 0xffff 653 Ultra Enable Flags: 0x0001 654 Tag Queue Enable Flags: 0x0000 655 Ordered Queue Tag Flags: 0x0000 656 Default Tag Queue Depth: 8 657 Tagged Queue By Device array for aic7xxx host instance 0: 658 {255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255} 659 Actual queue depth per device for aic7xxx host instance 0: 660 {1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1} 661 Statistics: 662 (scsi0:0:0:0) 663 Device using Wide/Sync transfers at 40.0 MByte/sec, offset 8 664 Transinfo settings: current(12/8/1/0), goal(12/8/1/0), user(12/15/1/0) 665 Total transfers 160151 (74577 reads and 85574 writes) 666 (scsi0:0:6:0) 667 Device using Narrow/Sync transfers at 5.0 MByte/sec, offset 15 668 Transinfo settings: current(50/15/0/0), goal(50/15/0/0), user(50/15/0/0) 669 Total transfers 0 (0 reads and 0 writes) 670 671 6721.6 Parallel port info in /proc/parport 673--------------------------------------- 674 675The directory /proc/parport contains information about the parallel ports of 676your system. It has one subdirectory for each port, named after the port 677number (0,1,2,...). 678 679These directories contain the four files shown in Table 1-8. 680 681 682Table 1-8: Files in /proc/parport 683.............................................................................. 684 File Content 685 autoprobe Any IEEE-1284 device ID information that has been acquired. 686 devices list of the device drivers using that port. A + will appear by the 687 name of the device currently using the port (it might not appear 688 against any). 689 hardware Parallel port's base address, IRQ line and DMA channel. 690 irq IRQ that parport is using for that port. This is in a separate 691 file to allow you to alter it by writing a new value in (IRQ 692 number or none). 693.............................................................................. 694 6951.7 TTY info in /proc/tty 696------------------------- 697 698Information about the available and actually used tty's can be found in the 699directory /proc/tty.You'll find entries for drivers and line disciplines in 700this directory, as shown in Table 1-9. 701 702 703Table 1-9: Files in /proc/tty 704.............................................................................. 705 File Content 706 drivers list of drivers and their usage 707 ldiscs registered line disciplines 708 driver/serial usage statistic and status of single tty lines 709.............................................................................. 710 711To see which tty's are currently in use, you can simply look into the file 712/proc/tty/drivers: 713 714 > cat /proc/tty/drivers 715 pty_slave /dev/pts 136 0-255 pty:slave 716 pty_master /dev/ptm 128 0-255 pty:master 717 pty_slave /dev/ttyp 3 0-255 pty:slave 718 pty_master /dev/pty 2 0-255 pty:master 719 serial /dev/cua 5 64-67 serial:callout 720 serial /dev/ttyS 4 64-67 serial 721 /dev/tty0 /dev/tty0 4 0 system:vtmaster 722 /dev/ptmx /dev/ptmx 5 2 system 723 /dev/console /dev/console 5 1 system:console 724 /dev/tty /dev/tty 5 0 system:/dev/tty 725 unknown /dev/tty 4 1-63 console 726 727 7281.8 Miscellaneous kernel statistics in /proc/stat 729------------------------------------------------- 730 731Various pieces of information about kernel activity are available in the 732/proc/stat file. All of the numbers reported in this file are aggregates 733since the system first booted. For a quick look, simply cat the file: 734 735 > cat /proc/stat 736 cpu 2255 34 2290 22625563 6290 127 456 737 cpu0 1132 34 1441 11311718 3675 127 438 738 cpu1 1123 0 849 11313845 2614 0 18 739 intr 114930548 113199788 3 0 5 263 0 4 [... lots more numbers ...] 740 ctxt 1990473 741 btime 1062191376 742 processes 2915 743 procs_running 1 744 procs_blocked 0 745 746The very first "cpu" line aggregates the numbers in all of the other "cpuN" 747lines. These numbers identify the amount of time the CPU has spent performing 748different kinds of work. Time units are in USER_HZ (typically hundredths of a 749second). The meanings of the columns are as follows, from left to right: 750 751- user: normal processes executing in user mode 752- nice: niced processes executing in user mode 753- system: processes executing in kernel mode 754- idle: twiddling thumbs 755- iowait: waiting for I/O to complete 756- irq: servicing interrupts 757- softirq: servicing softirqs 758 759The "intr" line gives counts of interrupts serviced since boot time, for each 760of the possible system interrupts. The first column is the total of all 761interrupts serviced; each subsequent column is the total for that particular 762interrupt. 763 764The "ctxt" line gives the total number of context switches across all CPUs. 765 766The "btime" line gives the time at which the system booted, in seconds since 767the Unix epoch. 768 769The "processes" line gives the number of processes and threads created, which 770includes (but is not limited to) those created by calls to the fork() and 771clone() system calls. 772 773The "procs_running" line gives the number of processes currently running on 774CPUs. 775 776The "procs_blocked" line gives the number of processes currently blocked, 777waiting for I/O to complete. 778 779 780------------------------------------------------------------------------------ 781Summary 782------------------------------------------------------------------------------ 783The /proc file system serves information about the running system. It not only 784allows access to process data but also allows you to request the kernel status 785by reading files in the hierarchy. 786 787The directory structure of /proc reflects the types of information and makes 788it easy, if not obvious, where to look for specific data. 789------------------------------------------------------------------------------ 790 791------------------------------------------------------------------------------ 792CHAPTER 2: MODIFYING SYSTEM PARAMETERS 793------------------------------------------------------------------------------ 794 795------------------------------------------------------------------------------ 796In This Chapter 797------------------------------------------------------------------------------ 798* Modifying kernel parameters by writing into files found in /proc/sys 799* Exploring the files which modify certain parameters 800* Review of the /proc/sys file tree 801------------------------------------------------------------------------------ 802 803 804A very interesting part of /proc is the directory /proc/sys. This is not only 805a source of information, it also allows you to change parameters within the 806kernel. Be very careful when attempting this. You can optimize your system, 807but you can also cause it to crash. Never alter kernel parameters on a 808production system. Set up a development machine and test to make sure that 809everything works the way you want it to. You may have no alternative but to 810reboot the machine once an error has been made. 811 812To change a value, simply echo the new value into the file. An example is 813given below in the section on the file system data. You need to be root to do 814this. You can create your own boot script to perform this every time your 815system boots. 816 817The files in /proc/sys can be used to fine tune and monitor miscellaneous and 818general things in the operation of the Linux kernel. Since some of the files 819can inadvertently disrupt your system, it is advisable to read both 820documentation and source before actually making adjustments. In any case, be 821very careful when writing to any of these files. The entries in /proc may 822change slightly between the 2.1.* and the 2.2 kernel, so if there is any doubt 823review the kernel documentation in the directory /usr/src/linux/Documentation. 824This chapter is heavily based on the documentation included in the pre 2.2 825kernels, and became part of it in version 2.2.1 of the Linux kernel. 826 8272.1 /proc/sys/fs - File system data 828----------------------------------- 829 830This subdirectory contains specific file system, file handle, inode, dentry 831and quota information. 832 833Currently, these files are in /proc/sys/fs: 834 835dentry-state 836------------ 837 838Status of the directory cache. Since directory entries are dynamically 839allocated and deallocated, this file indicates the current status. It holds 840six values, in which the last two are not used and are always zero. The others 841are listed in table 2-1. 842 843 844Table 2-1: Status files of the directory cache 845.............................................................................. 846 File Content 847 nr_dentry Almost always zero 848 nr_unused Number of unused cache entries 849 age_limit 850 in seconds after the entry may be reclaimed, when memory is short 851 want_pages internally 852.............................................................................. 853 854dquot-nr and dquot-max 855---------------------- 856 857The file dquot-max shows the maximum number of cached disk quota entries. 858 859The file dquot-nr shows the number of allocated disk quota entries and the 860number of free disk quota entries. 861 862If the number of available cached disk quotas is very low and you have a large 863number of simultaneous system users, you might want to raise the limit. 864 865file-nr and file-max 866-------------------- 867 868The kernel allocates file handles dynamically, but doesn't free them again at 869this time. 870 871The value in file-max denotes the maximum number of file handles that the 872Linux kernel will allocate. When you get a lot of error messages about running 873out of file handles, you might want to raise this limit. The default value is 87410% of RAM in kilobytes. To change it, just write the new number into the 875file: 876 877 # cat /proc/sys/fs/file-max 878 4096 879 # echo 8192 > /proc/sys/fs/file-max 880 # cat /proc/sys/fs/file-max 881 8192 882 883 884This method of revision is useful for all customizable parameters of the 885kernel - simply echo the new value to the corresponding file. 886 887Historically, the three values in file-nr denoted the number of allocated file 888handles, the number of allocated but unused file handles, and the maximum 889number of file handles. Linux 2.6 always reports 0 as the number of free file 890handles -- this is not an error, it just means that the number of allocated 891file handles exactly matches the number of used file handles. 892 893Attempts to allocate more file descriptors than file-max are reported with 894printk, look for "VFS: file-max limit <number> reached". 895 896inode-state and inode-nr 897------------------------ 898 899The file inode-nr contains the first two items from inode-state, so we'll skip 900to that file... 901 902inode-state contains two actual numbers and five dummy values. The numbers 903are nr_inodes and nr_free_inodes (in order of appearance). 904 905nr_inodes 906~~~~~~~~~ 907 908Denotes the number of inodes the system has allocated. This number will 909grow and shrink dynamically. 910 911nr_free_inodes 912-------------- 913 914Represents the number of free inodes. Ie. The number of inuse inodes is 915(nr_inodes - nr_free_inodes). 916 917aio-nr and aio-max-nr 918--------------------- 919 920aio-nr is the running total of the number of events specified on the 921io_setup system call for all currently active aio contexts. If aio-nr 922reaches aio-max-nr then io_setup will fail with EAGAIN. Note that 923raising aio-max-nr does not result in the pre-allocation or re-sizing 924of any kernel data structures. 925 9262.2 /proc/sys/fs/binfmt_misc - Miscellaneous binary formats 927----------------------------------------------------------- 928 929Besides these files, there is the subdirectory /proc/sys/fs/binfmt_misc. This 930handles the kernel support for miscellaneous binary formats. 931 932Binfmt_misc provides the ability to register additional binary formats to the 933Kernel without compiling an additional module/kernel. Therefore, binfmt_misc 934needs to know magic numbers at the beginning or the filename extension of the 935binary. 936 937It works by maintaining a linked list of structs that contain a description of 938a binary format, including a magic with size (or the filename extension), 939offset and mask, and the interpreter name. On request it invokes the given 940interpreter with the original program as argument, as binfmt_java and 941binfmt_em86 and binfmt_mz do. Since binfmt_misc does not define any default 942binary-formats, you have to register an additional binary-format. 943 944There are two general files in binfmt_misc and one file per registered format. 945The two general files are register and status. 946 947Registering a new binary format 948------------------------------- 949 950To register a new binary format you have to issue the command 951 952 echo :name:type:offset:magic:mask:interpreter: > /proc/sys/fs/binfmt_misc/register 953 954 955 956with appropriate name (the name for the /proc-dir entry), offset (defaults to 9570, if omitted), magic, mask (which can be omitted, defaults to all 0xff) and 958last but not least, the interpreter that is to be invoked (for example and 959testing /bin/echo). Type can be M for usual magic matching or E for filename 960extension matching (give extension in place of magic). 961 962Check or reset the status of the binary format handler 963------------------------------------------------------ 964 965If you do a cat on the file /proc/sys/fs/binfmt_misc/status, you will get the 966current status (enabled/disabled) of binfmt_misc. Change the status by echoing 9670 (disables) or 1 (enables) or -1 (caution: this clears all previously 968registered binary formats) to status. For example echo 0 > status to disable 969binfmt_misc (temporarily). 970 971Status of a single handler 972-------------------------- 973 974Each registered handler has an entry in /proc/sys/fs/binfmt_misc. These files 975perform the same function as status, but their scope is limited to the actual 976binary format. By cating this file, you also receive all related information 977about the interpreter/magic of the binfmt. 978 979Example usage of binfmt_misc (emulate binfmt_java) 980-------------------------------------------------- 981 982 cd /proc/sys/fs/binfmt_misc 983 echo ':Java:M::\xca\xfe\xba\xbe::/usr/local/java/bin/javawrapper:' > register 984 echo ':HTML:E::html::/usr/local/java/bin/appletviewer:' > register 985 echo ':Applet:M::<!--applet::/usr/local/java/bin/appletviewer:' > register 986 echo ':DEXE:M::\x0eDEX::/usr/bin/dosexec:' > register 987 988 989These four lines add support for Java executables and Java applets (like 990binfmt_java, additionally recognizing the .html extension with no need to put 991<!--applet> to every applet file). You have to install the JDK and the 992shell-script /usr/local/java/bin/javawrapper too. It works around the 993brokenness of the Java filename handling. To add a Java binary, just create a 994link to the class-file somewhere in the path. 995 9962.3 /proc/sys/kernel - general kernel parameters 997------------------------------------------------ 998 999This directory reflects general kernel behaviors. As I've said before, the 1000contents depend on your configuration. Here you'll find the most important 1001files, along with descriptions of what they mean and how to use them. 1002 1003acct 1004---- 1005 1006The file contains three values; highwater, lowwater, and frequency. 1007 1008It exists only when BSD-style process accounting is enabled. These values 1009control its behavior. If the free space on the file system where the log lives 1010goes below lowwater percentage, accounting suspends. If it goes above 1011highwater percentage, accounting resumes. Frequency determines how often you 1012check the amount of free space (value is in seconds). Default settings are: 4, 10132, and 30. That is, suspend accounting if there is less than 2 percent free; 1014resume it if we have a value of 3 or more percent; consider information about 1015the amount of free space valid for 30 seconds 1016 1017ctrl-alt-del 1018------------ 1019 1020When the value in this file is 0, ctrl-alt-del is trapped and sent to the init 1021program to handle a graceful restart. However, when the value is greater that 1022zero, Linux's reaction to this key combination will be an immediate reboot, 1023without syncing its dirty buffers. 1024 1025[NOTE] 1026 When a program (like dosemu) has the keyboard in raw mode, the 1027 ctrl-alt-del is intercepted by the program before it ever reaches the 1028 kernel tty layer, and it is up to the program to decide what to do with 1029 it. 1030 1031domainname and hostname 1032----------------------- 1033 1034These files can be controlled to set the NIS domainname and hostname of your 1035box. For the classic darkstar.frop.org a simple: 1036 1037 # echo "darkstar" > /proc/sys/kernel/hostname 1038 # echo "frop.org" > /proc/sys/kernel/domainname 1039 1040 1041would suffice to set your hostname and NIS domainname. 1042 1043osrelease, ostype and version 1044----------------------------- 1045 1046The names make it pretty obvious what these fields contain: 1047 1048 > cat /proc/sys/kernel/osrelease 1049 2.2.12 1050 1051 > cat /proc/sys/kernel/ostype 1052 Linux 1053 1054 > cat /proc/sys/kernel/version 1055 #4 Fri Oct 1 12:41:14 PDT 1999 1056 1057 1058The files osrelease and ostype should be clear enough. Version needs a little 1059more clarification. The #4 means that this is the 4th kernel built from this 1060source base and the date after it indicates the time the kernel was built. The 1061only way to tune these values is to rebuild the kernel. 1062 1063panic 1064----- 1065 1066The value in this file represents the number of seconds the kernel waits 1067before rebooting on a panic. When you use the software watchdog, the 1068recommended setting is 60. If set to 0, the auto reboot after a kernel panic 1069is disabled, which is the default setting. 1070 1071printk 1072------ 1073 1074The four values in printk denote 1075* console_loglevel, 1076* default_message_loglevel, 1077* minimum_console_loglevel and 1078* default_console_loglevel 1079respectively. 1080 1081These values influence printk() behavior when printing or logging error 1082messages, which come from inside the kernel. See syslog(2) for more 1083information on the different log levels. 1084 1085console_loglevel 1086---------------- 1087 1088Messages with a higher priority than this will be printed to the console. 1089 1090default_message_level 1091--------------------- 1092 1093Messages without an explicit priority will be printed with this priority. 1094 1095minimum_console_loglevel 1096------------------------ 1097 1098Minimum (highest) value to which the console_loglevel can be set. 1099 1100default_console_loglevel 1101------------------------ 1102 1103Default value for console_loglevel. 1104 1105sg-big-buff 1106----------- 1107 1108This file shows the size of the generic SCSI (sg) buffer. At this point, you 1109can't tune it yet, but you can change it at compile time by editing 1110include/scsi/sg.h and changing the value of SG_BIG_BUFF. 1111 1112If you use a scanner with SANE (Scanner Access Now Easy) you might want to set 1113this to a higher value. Refer to the SANE documentation on this issue. 1114 1115modprobe 1116-------- 1117 1118The location where the modprobe binary is located. The kernel uses this 1119program to load modules on demand. 1120 1121unknown_nmi_panic 1122----------------- 1123 1124The value in this file affects behavior of handling NMI. When the value is 1125non-zero, unknown NMI is trapped and then panic occurs. At that time, kernel 1126debugging information is displayed on console. 1127 1128NMI switch that most IA32 servers have fires unknown NMI up, for example. 1129If a system hangs up, try pressing the NMI switch. 1130 1131nmi_watchdog 1132------------ 1133 1134Enables/Disables the NMI watchdog on x86 systems. When the value is non-zero 1135the NMI watchdog is enabled and will continuously test all online cpus to 1136determine whether or not they are still functioning properly. 1137 1138Because the NMI watchdog shares registers with oprofile, by disabling the NMI 1139watchdog, oprofile may have more registers to utilize. 1140 1141maps_protect 1142------------ 1143 1144Enables/Disables the protection of the per-process proc entries "maps" and 1145"smaps". When enabled, the contents of these files are visible only to 1146readers that are allowed to ptrace() the given process. 1147 1148 11492.4 /proc/sys/vm - The virtual memory subsystem 1150----------------------------------------------- 1151 1152The files in this directory can be used to tune the operation of the virtual 1153memory (VM) subsystem of the Linux kernel. 1154 1155vfs_cache_pressure 1156------------------ 1157 1158Controls the tendency of the kernel to reclaim the memory which is used for 1159caching of directory and inode objects. 1160 1161At the default value of vfs_cache_pressure=100 the kernel will attempt to 1162reclaim dentries and inodes at a "fair" rate with respect to pagecache and 1163swapcache reclaim. Decreasing vfs_cache_pressure causes the kernel to prefer 1164to retain dentry and inode caches. Increasing vfs_cache_pressure beyond 100 1165causes the kernel to prefer to reclaim dentries and inodes. 1166 1167dirty_background_ratio 1168---------------------- 1169 1170Contains, as a percentage of total system memory, the number of pages at which 1171the pdflush background writeback daemon will start writing out dirty data. 1172 1173dirty_ratio 1174----------------- 1175 1176Contains, as a percentage of total system memory, the number of pages at which 1177a process which is generating disk writes will itself start writing out dirty 1178data. 1179 1180dirty_writeback_centisecs 1181------------------------- 1182 1183The pdflush writeback daemons will periodically wake up and write `old' data 1184out to disk. This tunable expresses the interval between those wakeups, in 1185100'ths of a second. 1186 1187Setting this to zero disables periodic writeback altogether. 1188 1189dirty_expire_centisecs 1190---------------------- 1191 1192This tunable is used to define when dirty data is old enough to be eligible 1193for writeout by the pdflush daemons. It is expressed in 100'ths of a second. 1194Data which has been dirty in-memory for longer than this interval will be 1195written out next time a pdflush daemon wakes up. 1196 1197legacy_va_layout 1198---------------- 1199 1200If non-zero, this sysctl disables the new 32-bit mmap mmap layout - the kernel 1201will use the legacy (2.4) layout for all processes. 1202 1203lower_zone_protection 1204--------------------- 1205 1206For some specialised workloads on highmem machines it is dangerous for 1207the kernel to allow process memory to be allocated from the "lowmem" 1208zone. This is because that memory could then be pinned via the mlock() 1209system call, or by unavailability of swapspace. 1210 1211And on large highmem machines this lack of reclaimable lowmem memory 1212can be fatal. 1213 1214So the Linux page allocator has a mechanism which prevents allocations 1215which _could_ use highmem from using too much lowmem. This means that 1216a certain amount of lowmem is defended from the possibility of being 1217captured into pinned user memory. 1218 1219(The same argument applies to the old 16 megabyte ISA DMA region. This 1220mechanism will also defend that region from allocations which could use 1221highmem or lowmem). 1222 1223The `lower_zone_protection' tunable determines how aggressive the kernel is 1224in defending these lower zones. The default value is zero - no 1225protection at all. 1226 1227If you have a machine which uses highmem or ISA DMA and your 1228applications are using mlock(), or if you are running with no swap then 1229you probably should increase the lower_zone_protection setting. 1230 1231The units of this tunable are fairly vague. It is approximately equal 1232to "megabytes," so setting lower_zone_protection=100 will protect around 100 1233megabytes of the lowmem zone from user allocations. It will also make 1234those 100 megabytes unavailable for use by applications and by 1235pagecache, so there is a cost. 1236 1237The effects of this tunable may be observed by monitoring 1238/proc/meminfo:LowFree. Write a single huge file and observe the point 1239at which LowFree ceases to fall. 1240 1241A reasonable value for lower_zone_protection is 100. 1242 1243page-cluster 1244------------ 1245 1246page-cluster controls the number of pages which are written to swap in 1247a single attempt. The swap I/O size. 1248 1249It is a logarithmic value - setting it to zero means "1 page", setting 1250it to 1 means "2 pages", setting it to 2 means "4 pages", etc. 1251 1252The default value is three (eight pages at a time). There may be some 1253small benefits in tuning this to a different value if your workload is 1254swap-intensive. 1255 1256overcommit_memory 1257----------------- 1258 1259Controls overcommit of system memory, possibly allowing processes 1260to allocate (but not use) more memory than is actually available. 1261 1262 12630 - Heuristic overcommit handling. Obvious overcommits of 1264 address space are refused. Used for a typical system. It 1265 ensures a seriously wild allocation fails while allowing 1266 overcommit to reduce swap usage. root is allowed to 1267 allocate slightly more memory in this mode. This is the 1268 default. 1269 12701 - Always overcommit. Appropriate for some scientific 1271 applications. 1272 12732 - Don't overcommit. The total address space commit 1274 for the system is not permitted to exceed swap plus a 1275 configurable percentage (default is 50) of physical RAM. 1276 Depending on the percentage you use, in most situations 1277 this means a process will not be killed while attempting 1278 to use already-allocated memory but will receive errors 1279 on memory allocation as appropriate. 1280 1281overcommit_ratio 1282---------------- 1283 1284Percentage of physical memory size to include in overcommit calculations 1285(see above.) 1286 1287Memory allocation limit = swapspace + physmem * (overcommit_ratio / 100) 1288 1289 swapspace = total size of all swap areas 1290 physmem = size of physical memory in system 1291 1292nr_hugepages and hugetlb_shm_group 1293---------------------------------- 1294 1295nr_hugepages configures number of hugetlb page reserved for the system. 1296 1297hugetlb_shm_group contains group id that is allowed to create SysV shared 1298memory segment using hugetlb page. 1299 1300laptop_mode 1301----------- 1302 1303laptop_mode is a knob that controls "laptop mode". All the things that are 1304controlled by this knob are discussed in Documentation/laptop-mode.txt. 1305 1306block_dump 1307---------- 1308 1309block_dump enables block I/O debugging when set to a nonzero value. More 1310information on block I/O debugging is in Documentation/laptop-mode.txt. 1311 1312swap_token_timeout 1313------------------ 1314 1315This file contains valid hold time of swap out protection token. The Linux 1316VM has token based thrashing control mechanism and uses the token to prevent 1317unnecessary page faults in thrashing situation. The unit of the value is 1318second. The value would be useful to tune thrashing behavior. 1319 1320drop_caches 1321----------- 1322 1323Writing to this will cause the kernel to drop clean caches, dentries and 1324inodes from memory, causing that memory to become free. 1325 1326To free pagecache: 1327 echo 1 > /proc/sys/vm/drop_caches 1328To free dentries and inodes: 1329 echo 2 > /proc/sys/vm/drop_caches 1330To free pagecache, dentries and inodes: 1331 echo 3 > /proc/sys/vm/drop_caches 1332 1333As this is a non-destructive operation and dirty objects are not freeable, the 1334user should run `sync' first. 1335 1336 13372.5 /proc/sys/dev - Device specific parameters 1338---------------------------------------------- 1339 1340Currently there is only support for CDROM drives, and for those, there is only 1341one read-only file containing information about the CD-ROM drives attached to 1342the system: 1343 1344 >cat /proc/sys/dev/cdrom/info 1345 CD-ROM information, Id: cdrom.c 2.55 1999/04/25 1346 1347 drive name: sr0 hdb 1348 drive speed: 32 40 1349 drive # of slots: 1 0 1350 Can close tray: 1 1 1351 Can open tray: 1 1 1352 Can lock tray: 1 1 1353 Can change speed: 1 1 1354 Can select disk: 0 1 1355 Can read multisession: 1 1 1356 Can read MCN: 1 1 1357 Reports media changed: 1 1 1358 Can play audio: 1 1 1359 1360 1361You see two drives, sr0 and hdb, along with a list of their features. 1362 13632.6 /proc/sys/sunrpc - Remote procedure calls 1364--------------------------------------------- 1365 1366This directory contains four files, which enable or disable debugging for the 1367RPC functions NFS, NFS-daemon, RPC and NLM. The default values are 0. They can 1368be set to one to turn debugging on. (The default value is 0 for each) 1369 13702.7 /proc/sys/net - Networking stuff 1371------------------------------------ 1372 1373The interface to the networking parts of the kernel is located in 1374/proc/sys/net. Table 2-3 shows all possible subdirectories. You may see only 1375some of them, depending on your kernel's configuration. 1376 1377 1378Table 2-3: Subdirectories in /proc/sys/net 1379.............................................................................. 1380 Directory Content Directory Content 1381 core General parameter appletalk Appletalk protocol 1382 unix Unix domain sockets netrom NET/ROM 1383 802 E802 protocol ax25 AX25 1384 ethernet Ethernet protocol rose X.25 PLP layer 1385 ipv4 IP version 4 x25 X.25 protocol 1386 ipx IPX token-ring IBM token ring 1387 bridge Bridging decnet DEC net 1388 ipv6 IP version 6 1389.............................................................................. 1390 1391We will concentrate on IP networking here. Since AX15, X.25, and DEC Net are 1392only minor players in the Linux world, we'll skip them in this chapter. You'll 1393find some short info on Appletalk and IPX further on in this chapter. Review 1394the online documentation and the kernel source to get a detailed view of the 1395parameters for those protocols. In this section we'll discuss the 1396subdirectories printed in bold letters in the table above. As default values 1397are suitable for most needs, there is no need to change these values. 1398 1399/proc/sys/net/core - Network core options 1400----------------------------------------- 1401 1402rmem_default 1403------------ 1404 1405The default setting of the socket receive buffer in bytes. 1406 1407rmem_max 1408-------- 1409 1410The maximum receive socket buffer size in bytes. 1411 1412wmem_default 1413------------ 1414 1415The default setting (in bytes) of the socket send buffer. 1416 1417wmem_max 1418-------- 1419 1420The maximum send socket buffer size in bytes. 1421 1422message_burst and message_cost 1423------------------------------ 1424 1425These parameters are used to limit the warning messages written to the kernel 1426log from the networking code. They enforce a rate limit to make a 1427denial-of-service attack impossible. A higher message_cost factor, results in 1428fewer messages that will be written. Message_burst controls when messages will 1429be dropped. The default settings limit warning messages to one every five 1430seconds. 1431 1432warnings 1433-------- 1434 1435This controls console messages from the networking stack that can occur because 1436of problems on the network like duplicate address or bad checksums. Normally, 1437this should be enabled, but if the problem persists the messages can be 1438disabled. 1439 1440 1441netdev_max_backlog 1442------------------ 1443 1444Maximum number of packets, queued on the INPUT side, when the interface 1445receives packets faster than kernel can process them. 1446 1447optmem_max 1448---------- 1449 1450Maximum ancillary buffer size allowed per socket. Ancillary data is a sequence 1451of struct cmsghdr structures with appended data. 1452 1453/proc/sys/net/unix - Parameters for Unix domain sockets 1454------------------------------------------------------- 1455 1456There are only two files in this subdirectory. They control the delays for 1457deleting and destroying socket descriptors. 1458 14592.8 /proc/sys/net/ipv4 - IPV4 settings 1460-------------------------------------- 1461 1462IP version 4 is still the most used protocol in Unix networking. It will be 1463replaced by IP version 6 in the next couple of years, but for the moment it's 1464the de facto standard for the internet and is used in most networking 1465environments around the world. Because of the importance of this protocol, 1466we'll have a deeper look into the subtree controlling the behavior of the IPv4 1467subsystem of the Linux kernel. 1468 1469Let's start with the entries in /proc/sys/net/ipv4. 1470 1471ICMP settings 1472------------- 1473 1474icmp_echo_ignore_all and icmp_echo_ignore_broadcasts 1475---------------------------------------------------- 1476 1477Turn on (1) or off (0), if the kernel should ignore all ICMP ECHO requests, or 1478just those to broadcast and multicast addresses. 1479 1480Please note that if you accept ICMP echo requests with a broadcast/multi\-cast 1481destination address your network may be used as an exploder for denial of 1482service packet flooding attacks to other hosts. 1483 1484icmp_destunreach_rate, icmp_echoreply_rate, icmp_paramprob_rate and icmp_timeexeed_rate 1485--------------------------------------------------------------------------------------- 1486 1487Sets limits for sending ICMP packets to specific targets. A value of zero 1488disables all limiting. Any positive value sets the maximum package rate in 1489hundredth of a second (on Intel systems). 1490 1491IP settings 1492----------- 1493 1494ip_autoconfig 1495------------- 1496 1497This file contains the number one if the host received its IP configuration by 1498RARP, BOOTP, DHCP or a similar mechanism. Otherwise it is zero. 1499 1500ip_default_ttl 1501-------------- 1502 1503TTL (Time To Live) for IPv4 interfaces. This is simply the maximum number of 1504hops a packet may travel. 1505 1506ip_dynaddr 1507---------- 1508 1509Enable dynamic socket address rewriting on interface address change. This is 1510useful for dialup interface with changing IP addresses. 1511 1512ip_forward 1513---------- 1514 1515Enable or disable forwarding of IP packages between interfaces. Changing this 1516value resets all other parameters to their default values. They differ if the 1517kernel is configured as host or router. 1518 1519ip_local_port_range 1520------------------- 1521 1522Range of ports used by TCP and UDP to choose the local port. Contains two 1523numbers, the first number is the lowest port, the second number the highest 1524local port. Default is 1024-4999. Should be changed to 32768-61000 for 1525high-usage systems. 1526 1527ip_no_pmtu_disc 1528--------------- 1529 1530Global switch to turn path MTU discovery off. It can also be set on a per 1531socket basis by the applications or on a per route basis. 1532 1533ip_masq_debug 1534------------- 1535 1536Enable/disable debugging of IP masquerading. 1537 1538IP fragmentation settings 1539------------------------- 1540 1541ipfrag_high_trash and ipfrag_low_trash 1542-------------------------------------- 1543 1544Maximum memory used to reassemble IP fragments. When ipfrag_high_thresh bytes 1545of memory is allocated for this purpose, the fragment handler will toss 1546packets until ipfrag_low_thresh is reached. 1547 1548ipfrag_time 1549----------- 1550 1551Time in seconds to keep an IP fragment in memory. 1552 1553TCP settings 1554------------ 1555 1556tcp_ecn 1557------- 1558 1559This file controls the use of the ECN bit in the IPv4 headers. This is a new 1560feature about Explicit Congestion Notification, but some routers and firewalls 1561block traffic that has this bit set, so it could be necessary to echo 0 to 1562/proc/sys/net/ipv4/tcp_ecn if you want to talk to these sites. For more info 1563you could read RFC2481. 1564 1565tcp_retrans_collapse 1566-------------------- 1567 1568Bug-to-bug compatibility with some broken printers. On retransmit, try to send 1569larger packets to work around bugs in certain TCP stacks. Can be turned off by 1570setting it to zero. 1571 1572tcp_keepalive_probes 1573-------------------- 1574 1575Number of keep alive probes TCP sends out, until it decides that the 1576connection is broken. 1577 1578tcp_keepalive_time 1579------------------ 1580 1581How often TCP sends out keep alive messages, when keep alive is enabled. The 1582default is 2 hours. 1583 1584tcp_syn_retries 1585--------------- 1586 1587Number of times initial SYNs for a TCP connection attempt will be 1588retransmitted. Should not be higher than 255. This is only the timeout for 1589outgoing connections, for incoming connections the number of retransmits is 1590defined by tcp_retries1. 1591 1592tcp_sack 1593-------- 1594 1595Enable select acknowledgments after RFC2018. 1596 1597tcp_timestamps 1598-------------- 1599 1600Enable timestamps as defined in RFC1323. 1601 1602tcp_stdurg 1603---------- 1604 1605Enable the strict RFC793 interpretation of the TCP urgent pointer field. The 1606default is to use the BSD compatible interpretation of the urgent pointer 1607pointing to the first byte after the urgent data. The RFC793 interpretation is 1608to have it point to the last byte of urgent data. Enabling this option may 1609lead to interoperability problems. Disabled by default. 1610 1611tcp_syncookies 1612-------------- 1613 1614Only valid when the kernel was compiled with CONFIG_SYNCOOKIES. Send out 1615syncookies when the syn backlog queue of a socket overflows. This is to ward 1616off the common 'syn flood attack'. Disabled by default. 1617 1618Note that the concept of a socket backlog is abandoned. This means the peer 1619may not receive reliable error messages from an over loaded server with 1620syncookies enabled. 1621 1622tcp_window_scaling 1623------------------ 1624 1625Enable window scaling as defined in RFC1323. 1626 1627tcp_fin_timeout 1628--------------- 1629 1630The length of time in seconds it takes to receive a final FIN before the 1631socket is always closed. This is strictly a violation of the TCP 1632specification, but required to prevent denial-of-service attacks. 1633 1634tcp_max_ka_probes 1635----------------- 1636 1637Indicates how many keep alive probes are sent per slow timer run. Should not 1638be set too high to prevent bursts. 1639 1640tcp_max_syn_backlog 1641------------------- 1642 1643Length of the per socket backlog queue. Since Linux 2.2 the backlog specified 1644in listen(2) only specifies the length of the backlog queue of already 1645established sockets. When more connection requests arrive Linux starts to drop 1646packets. When syncookies are enabled the packets are still answered and the 1647maximum queue is effectively ignored. 1648 1649tcp_retries1 1650------------ 1651 1652Defines how often an answer to a TCP connection request is retransmitted 1653before giving up. 1654 1655tcp_retries2 1656------------ 1657 1658Defines how often a TCP packet is retransmitted before giving up. 1659 1660Interface specific settings 1661--------------------------- 1662 1663In the directory /proc/sys/net/ipv4/conf you'll find one subdirectory for each 1664interface the system knows about and one directory calls all. Changes in the 1665all subdirectory affect all interfaces, whereas changes in the other 1666subdirectories affect only one interface. All directories have the same 1667entries: 1668 1669accept_redirects 1670---------------- 1671 1672This switch decides if the kernel accepts ICMP redirect messages or not. The 1673default is 'yes' if the kernel is configured for a regular host and 'no' for a 1674router configuration. 1675 1676accept_source_route 1677------------------- 1678 1679Should source routed packages be accepted or declined. The default is 1680dependent on the kernel configuration. It's 'yes' for routers and 'no' for 1681hosts. 1682 1683bootp_relay 1684~~~~~~~~~~~ 1685 1686Accept packets with source address 0.b.c.d with destinations not to this host 1687as local ones. It is supposed that a BOOTP relay daemon will catch and forward 1688such packets. 1689 1690The default is 0, since this feature is not implemented yet (kernel version 16912.2.12). 1692 1693forwarding 1694---------- 1695 1696Enable or disable IP forwarding on this interface. 1697 1698log_martians 1699------------ 1700 1701Log packets with source addresses with no known route to kernel log. 1702 1703mc_forwarding 1704------------- 1705 1706Do multicast routing. The kernel needs to be compiled with CONFIG_MROUTE and a 1707multicast routing daemon is required. 1708 1709proxy_arp 1710--------- 1711 1712Does (1) or does not (0) perform proxy ARP. 1713 1714rp_filter 1715--------- 1716 1717Integer value determines if a source validation should be made. 1 means yes, 0 1718means no. Disabled by default, but local/broadcast address spoofing is always 1719on. 1720 1721If you set this to 1 on a router that is the only connection for a network to 1722the net, it will prevent spoofing attacks against your internal networks 1723(external addresses can still be spoofed), without the need for additional 1724firewall rules. 1725 1726secure_redirects 1727---------------- 1728 1729Accept ICMP redirect messages only for gateways, listed in default gateway 1730list. Enabled by default. 1731 1732shared_media 1733------------ 1734 1735If it is not set the kernel does not assume that different subnets on this 1736device can communicate directly. Default setting is 'yes'. 1737 1738send_redirects 1739-------------- 1740 1741Determines whether to send ICMP redirects to other hosts. 1742 1743Routing settings 1744---------------- 1745 1746The directory /proc/sys/net/ipv4/route contains several file to control 1747routing issues. 1748 1749error_burst and error_cost 1750-------------------------- 1751 1752These parameters are used to limit how many ICMP destination unreachable to 1753send from the host in question. ICMP destination unreachable messages are 1754sent when we cannot reach the next hop while trying to transmit a packet. 1755It will also print some error messages to kernel logs if someone is ignoring 1756our ICMP redirects. The higher the error_cost factor is, the fewer 1757destination unreachable and error messages will be let through. Error_burst 1758controls when destination unreachable messages and error messages will be 1759dropped. The default settings limit warning messages to five every second. 1760 1761flush 1762----- 1763 1764Writing to this file results in a flush of the routing cache. 1765 1766gc_elasticity, gc_interval, gc_min_interval_ms, gc_timeout, gc_thresh 1767--------------------------------------------------------------------- 1768 1769Values to control the frequency and behavior of the garbage collection 1770algorithm for the routing cache. gc_min_interval is deprecated and replaced 1771by gc_min_interval_ms. 1772 1773 1774max_size 1775-------- 1776 1777Maximum size of the routing cache. Old entries will be purged once the cache 1778reached has this size. 1779 1780max_delay, min_delay 1781-------------------- 1782 1783Delays for flushing the routing cache. 1784 1785redirect_load, redirect_number 1786------------------------------ 1787 1788Factors which determine if more ICPM redirects should be sent to a specific 1789host. No redirects will be sent once the load limit or the maximum number of 1790redirects has been reached. 1791 1792redirect_silence 1793---------------- 1794 1795Timeout for redirects. After this period redirects will be sent again, even if 1796this has been stopped, because the load or number limit has been reached. 1797 1798Network Neighbor handling 1799------------------------- 1800 1801Settings about how to handle connections with direct neighbors (nodes attached 1802to the same link) can be found in the directory /proc/sys/net/ipv4/neigh. 1803 1804As we saw it in the conf directory, there is a default subdirectory which 1805holds the default values, and one directory for each interface. The contents 1806of the directories are identical, with the single exception that the default 1807settings contain additional options to set garbage collection parameters. 1808 1809In the interface directories you'll find the following entries: 1810 1811base_reachable_time, base_reachable_time_ms 1812------------------------------------------- 1813 1814A base value used for computing the random reachable time value as specified 1815in RFC2461. 1816 1817Expression of base_reachable_time, which is deprecated, is in seconds. 1818Expression of base_reachable_time_ms is in milliseconds. 1819 1820retrans_time, retrans_time_ms 1821----------------------------- 1822 1823The time between retransmitted Neighbor Solicitation messages. 1824Used for address resolution and to determine if a neighbor is 1825unreachable. 1826 1827Expression of retrans_time, which is deprecated, is in 1/100 seconds (for 1828IPv4) or in jiffies (for IPv6). 1829Expression of retrans_time_ms is in milliseconds. 1830 1831unres_qlen 1832---------- 1833 1834Maximum queue length for a pending arp request - the number of packets which 1835are accepted from other layers while the ARP address is still resolved. 1836 1837anycast_delay 1838------------- 1839 1840Maximum for random delay of answers to neighbor solicitation messages in 1841jiffies (1/100 sec). Not yet implemented (Linux does not have anycast support 1842yet). 1843 1844ucast_solicit 1845------------- 1846 1847Maximum number of retries for unicast solicitation. 1848 1849mcast_solicit 1850------------- 1851 1852Maximum number of retries for multicast solicitation. 1853 1854delay_first_probe_time 1855---------------------- 1856 1857Delay for the first time probe if the neighbor is reachable. (see 1858gc_stale_time) 1859 1860locktime 1861-------- 1862 1863An ARP/neighbor entry is only replaced with a new one if the old is at least 1864locktime old. This prevents ARP cache thrashing. 1865 1866proxy_delay 1867----------- 1868 1869Maximum time (real time is random [0..proxytime]) before answering to an ARP 1870request for which we have an proxy ARP entry. In some cases, this is used to 1871prevent network flooding. 1872 1873proxy_qlen 1874---------- 1875 1876Maximum queue length of the delayed proxy arp timer. (see proxy_delay). 1877 1878app_solicit 1879---------- 1880 1881Determines the number of requests to send to the user level ARP daemon. Use 0 1882to turn off. 1883 1884gc_stale_time 1885------------- 1886 1887Determines how often to check for stale ARP entries. After an ARP entry is 1888stale it will be resolved again (which is useful when an IP address migrates 1889to another machine). When ucast_solicit is greater than 0 it first tries to 1890send an ARP packet directly to the known host When that fails and 1891mcast_solicit is greater than 0, an ARP request is broadcasted. 1892 18932.9 Appletalk 1894------------- 1895 1896The /proc/sys/net/appletalk directory holds the Appletalk configuration data 1897when Appletalk is loaded. The configurable parameters are: 1898 1899aarp-expiry-time 1900---------------- 1901 1902The amount of time we keep an ARP entry before expiring it. Used to age out 1903old hosts. 1904 1905aarp-resolve-time 1906----------------- 1907 1908The amount of time we will spend trying to resolve an Appletalk address. 1909 1910aarp-retransmit-limit 1911--------------------- 1912 1913The number of times we will retransmit a query before giving up. 1914 1915aarp-tick-time 1916-------------- 1917 1918Controls the rate at which expires are checked. 1919 1920The directory /proc/net/appletalk holds the list of active Appletalk sockets 1921on a machine. 1922 1923The fields indicate the DDP type, the local address (in network:node format) 1924the remote address, the size of the transmit pending queue, the size of the 1925received queue (bytes waiting for applications to read) the state and the uid 1926owning the socket. 1927 1928/proc/net/atalk_iface lists all the interfaces configured for appletalk.It 1929shows the name of the interface, its Appletalk address, the network range on 1930that address (or network number for phase 1 networks), and the status of the 1931interface. 1932 1933/proc/net/atalk_route lists each known network route. It lists the target 1934(network) that the route leads to, the router (may be directly connected), the 1935route flags, and the device the route is using. 1936 19372.10 IPX 1938-------- 1939 1940The IPX protocol has no tunable values in proc/sys/net. 1941 1942The IPX protocol does, however, provide proc/net/ipx. This lists each IPX 1943socket giving the local and remote addresses in Novell format (that is 1944network:node:port). In accordance with the strange Novell tradition, 1945everything but the port is in hex. Not_Connected is displayed for sockets that 1946are not tied to a specific remote address. The Tx and Rx queue sizes indicate 1947the number of bytes pending for transmission and reception. The state 1948indicates the state the socket is in and the uid is the owning uid of the 1949socket. 1950 1951The /proc/net/ipx_interface file lists all IPX interfaces. For each interface 1952it gives the network number, the node number, and indicates if the network is 1953the primary network. It also indicates which device it is bound to (or 1954Internal for internal networks) and the Frame Type if appropriate. Linux 1955supports 802.3, 802.2, 802.2 SNAP and DIX (Blue Book) ethernet framing for 1956IPX. 1957 1958The /proc/net/ipx_route table holds a list of IPX routes. For each route it 1959gives the destination network, the router node (or Directly) and the network 1960address of the router (or Connected) for internal networks. 1961 19622.11 /proc/sys/fs/mqueue - POSIX message queues filesystem 1963---------------------------------------------------------- 1964 1965The "mqueue" filesystem provides the necessary kernel features to enable the 1966creation of a user space library that implements the POSIX message queues 1967API (as noted by the MSG tag in the POSIX 1003.1-2001 version of the System 1968Interfaces specification.) 1969 1970The "mqueue" filesystem contains values for determining/setting the amount of 1971resources used by the file system. 1972 1973/proc/sys/fs/mqueue/queues_max is a read/write file for setting/getting the 1974maximum number of message queues allowed on the system. 1975 1976/proc/sys/fs/mqueue/msg_max is a read/write file for setting/getting the 1977maximum number of messages in a queue value. In fact it is the limiting value 1978for another (user) limit which is set in mq_open invocation. This attribute of 1979a queue must be less or equal then msg_max. 1980 1981/proc/sys/fs/mqueue/msgsize_max is a read/write file for setting/getting the 1982maximum message size value (it is every message queue's attribute set during 1983its creation). 1984 19852.12 /proc/<pid>/oom_adj - Adjust the oom-killer score 1986------------------------------------------------------ 1987 1988This file can be used to adjust the score used to select which processes 1989should be killed in an out-of-memory situation. Giving it a high score will 1990increase the likelihood of this process being killed by the oom-killer. Valid 1991values are in the range -16 to +15, plus the special value -17, which disables 1992oom-killing altogether for this process. 1993 19942.13 /proc/<pid>/oom_score - Display current oom-killer score 1995------------------------------------------------------------- 1996 1997------------------------------------------------------------------------------ 1998This file can be used to check the current score used by the oom-killer is for 1999any given <pid>. Use it together with /proc/<pid>/oom_adj to tune which 2000process should be killed in an out-of-memory situation. 2001 2002------------------------------------------------------------------------------ 2003Summary 2004------------------------------------------------------------------------------ 2005Certain aspects of kernel behavior can be modified at runtime, without the 2006need to recompile the kernel, or even to reboot the system. The files in the 2007/proc/sys tree can not only be read, but also modified. You can use the echo 2008command to write value into these files, thereby changing the default settings 2009of the kernel. 2010------------------------------------------------------------------------------ 2011 20122.14 /proc/<pid>/io - Display the IO accounting fields 2013------------------------------------------------------- 2014 2015This file contains IO statistics for each running process 2016 2017Example 2018------- 2019 2020test:/tmp # dd if=/dev/zero of=/tmp/test.dat & 2021[1] 3828 2022 2023test:/tmp # cat /proc/3828/io 2024rchar: 323934931 2025wchar: 323929600 2026syscr: 632687 2027syscw: 632675 2028read_bytes: 0 2029write_bytes: 323932160 2030cancelled_write_bytes: 0 2031 2032 2033Description 2034----------- 2035 2036rchar 2037----- 2038 2039I/O counter: chars read 2040The number of bytes which this task has caused to be read from storage. This 2041is simply the sum of bytes which this process passed to read() and pread(). 2042It includes things like tty IO and it is unaffected by whether or not actual 2043physical disk IO was required (the read might have been satisfied from 2044pagecache) 2045 2046 2047wchar 2048----- 2049 2050I/O counter: chars written 2051The number of bytes which this task has caused, or shall cause to be written 2052to disk. Similar caveats apply here as with rchar. 2053 2054 2055syscr 2056----- 2057 2058I/O counter: read syscalls 2059Attempt to count the number of read I/O operations, i.e. syscalls like read() 2060and pread(). 2061 2062 2063syscw 2064----- 2065 2066I/O counter: write syscalls 2067Attempt to count the number of write I/O operations, i.e. syscalls like 2068write() and pwrite(). 2069 2070 2071read_bytes 2072---------- 2073 2074I/O counter: bytes read 2075Attempt to count the number of bytes which this process really did cause to 2076be fetched from the storage layer. Done at the submit_bio() level, so it is 2077accurate for block-backed filesystems. <please add status regarding NFS and 2078CIFS at a later time> 2079 2080 2081write_bytes 2082----------- 2083 2084I/O counter: bytes written 2085Attempt to count the number of bytes which this process caused to be sent to 2086the storage layer. This is done at page-dirtying time. 2087 2088 2089cancelled_write_bytes 2090--------------------- 2091 2092The big inaccuracy here is truncate. If a process writes 1MB to a file and 2093then deletes the file, it will in fact perform no writeout. But it will have 2094been accounted as having caused 1MB of write. 2095In other words: The number of bytes which this process caused to not happen, 2096by truncating pagecache. A task can cause "negative" IO too. If this task 2097truncates some dirty pagecache, some IO which another task has been accounted 2098for (in it's write_bytes) will not be happening. We _could_ just subtract that 2099from the truncating task's write_bytes, but there is information loss in doing 2100that. 2101 2102 2103Note 2104---- 2105 2106At its current implementation state, this is a bit racy on 32-bit machines: if 2107process A reads process B's /proc/pid/io while process B is updating one of 2108those 64-bit counters, process A could see an intermediate result. 2109 2110 2111More information about this can be found within the taskstats documentation in 2112Documentation/accounting. 2113 2114------------------------------------------------------------------------------ 2115