1Documentation for /proc/sys/fs/* kernel version 2.2.10 2 (c) 1998, 1999, Rik van Riel <riel@nl.linux.org> 3 (c) 2009, Shen Feng<shen@cn.fujitsu.com> 4 5For general info and legal blurb, please look in README. 6 7============================================================== 8 9This file contains documentation for the sysctl files in 10/proc/sys/fs/ and is valid for Linux kernel version 2.2. 11 12The files in this directory can be used to tune and monitor 13miscellaneous and general things in the operation of the Linux 14kernel. Since some of the files _can_ be used to screw up your 15system, it is advisable to read both documentation and source 16before actually making adjustments. 17 181. /proc/sys/fs 19---------------------------------------------------------- 20 21Currently, these files are in /proc/sys/fs: 22- aio-max-nr 23- aio-nr 24- dentry-state 25- dquot-max 26- dquot-nr 27- file-max 28- file-nr 29- inode-max 30- inode-nr 31- inode-state 32- nr_open 33- overflowuid 34- overflowgid 35- pipe-user-pages-hard 36- pipe-user-pages-soft 37- protected_fifos 38- protected_hardlinks 39- protected_regular 40- protected_symlinks 41- suid_dumpable 42- super-max 43- super-nr 44 45============================================================== 46 47aio-nr & aio-max-nr: 48 49aio-nr is the running total of the number of events specified on the 50io_setup system call for all currently active aio contexts. If aio-nr 51reaches aio-max-nr then io_setup will fail with EAGAIN. Note that 52raising aio-max-nr does not result in the pre-allocation or re-sizing 53of any kernel data structures. 54 55============================================================== 56 57dentry-state: 58 59From linux/fs/dentry.c: 60-------------------------------------------------------------- 61struct { 62 int nr_dentry; 63 int nr_unused; 64 int age_limit; /* age in seconds */ 65 int want_pages; /* pages requested by system */ 66 int dummy[2]; 67} dentry_stat = {0, 0, 45, 0,}; 68-------------------------------------------------------------- 69 70Dentries are dynamically allocated and deallocated, and 71nr_dentry seems to be 0 all the time. Hence it's safe to 72assume that only nr_unused, age_limit and want_pages are 73used. Nr_unused seems to be exactly what its name says. 74Age_limit is the age in seconds after which dcache entries 75can be reclaimed when memory is short and want_pages is 76nonzero when shrink_dcache_pages() has been called and the 77dcache isn't pruned yet. 78 79============================================================== 80 81dquot-max & dquot-nr: 82 83The file dquot-max shows the maximum number of cached disk 84quota entries. 85 86The file dquot-nr shows the number of allocated disk quota 87entries and the number of free disk quota entries. 88 89If the number of free cached disk quotas is very low and 90you have some awesome number of simultaneous system users, 91you might want to raise the limit. 92 93============================================================== 94 95file-max & file-nr: 96 97The value in file-max denotes the maximum number of file- 98handles that the Linux kernel will allocate. When you get lots 99of error messages about running out of file handles, you might 100want to increase this limit. 101 102Historically,the kernel was able to allocate file handles 103dynamically, but not to free them again. The three values in 104file-nr denote the number of allocated file handles, the number 105of allocated but unused file handles, and the maximum number of 106file handles. Linux 2.6 always reports 0 as the number of free 107file handles -- this is not an error, it just means that the 108number of allocated file handles exactly matches the number of 109used file handles. 110 111Attempts to allocate more file descriptors than file-max are 112reported with printk, look for "VFS: file-max limit <number> 113reached". 114============================================================== 115 116nr_open: 117 118This denotes the maximum number of file-handles a process can 119allocate. Default value is 1024*1024 (1048576) which should be 120enough for most machines. Actual limit depends on RLIMIT_NOFILE 121resource limit. 122 123============================================================== 124 125inode-max, inode-nr & inode-state: 126 127As with file handles, the kernel allocates the inode structures 128dynamically, but can't free them yet. 129 130The value in inode-max denotes the maximum number of inode 131handlers. This value should be 3-4 times larger than the value 132in file-max, since stdin, stdout and network sockets also 133need an inode struct to handle them. When you regularly run 134out of inodes, you need to increase this value. 135 136The file inode-nr contains the first two items from 137inode-state, so we'll skip to that file... 138 139Inode-state contains three actual numbers and four dummies. 140The actual numbers are, in order of appearance, nr_inodes, 141nr_free_inodes and preshrink. 142 143Nr_inodes stands for the number of inodes the system has 144allocated, this can be slightly more than inode-max because 145Linux allocates them one pageful at a time. 146 147Nr_free_inodes represents the number of free inodes (?) and 148preshrink is nonzero when the nr_inodes > inode-max and the 149system needs to prune the inode list instead of allocating 150more. 151 152============================================================== 153 154overflowgid & overflowuid: 155 156Some filesystems only support 16-bit UIDs and GIDs, although in Linux 157UIDs and GIDs are 32 bits. When one of these filesystems is mounted 158with writes enabled, any UID or GID that would exceed 65535 is translated 159to a fixed value before being written to disk. 160 161These sysctls allow you to change the value of the fixed UID and GID. 162The default is 65534. 163 164============================================================== 165 166pipe-user-pages-hard: 167 168Maximum total number of pages a non-privileged user may allocate for pipes. 169Once this limit is reached, no new pipes may be allocated until usage goes 170below the limit again. When set to 0, no limit is applied, which is the default 171setting. 172 173============================================================== 174 175pipe-user-pages-soft: 176 177Maximum total number of pages a non-privileged user may allocate for pipes 178before the pipe size gets limited to a single page. Once this limit is reached, 179new pipes will be limited to a single page in size for this user in order to 180limit total memory usage, and trying to increase them using fcntl() will be 181denied until usage goes below the limit again. The default value allows to 182allocate up to 1024 pipes at their default size. When set to 0, no limit is 183applied. 184 185============================================================== 186 187protected_fifos: 188 189The intent of this protection is to avoid unintentional writes to 190an attacker-controlled FIFO, where a program expected to create a regular 191file. 192 193When set to "0", writing to FIFOs is unrestricted. 194 195When set to "1" don't allow O_CREAT open on FIFOs that we don't own 196in world writable sticky directories, unless they are owned by the 197owner of the directory. 198 199When set to "2" it also applies to group writable sticky directories. 200 201This protection is based on the restrictions in Openwall. 202 203============================================================== 204 205protected_hardlinks: 206 207A long-standing class of security issues is the hardlink-based 208time-of-check-time-of-use race, most commonly seen in world-writable 209directories like /tmp. The common method of exploitation of this flaw 210is to cross privilege boundaries when following a given hardlink (i.e. a 211root process follows a hardlink created by another user). Additionally, 212on systems without separated partitions, this stops unauthorized users 213from "pinning" vulnerable setuid/setgid files against being upgraded by 214the administrator, or linking to special files. 215 216When set to "0", hardlink creation behavior is unrestricted. 217 218When set to "1" hardlinks cannot be created by users if they do not 219already own the source file, or do not have read/write access to it. 220 221This protection is based on the restrictions in Openwall and grsecurity. 222 223============================================================== 224 225protected_regular: 226 227This protection is similar to protected_fifos, but it 228avoids writes to an attacker-controlled regular file, where a program 229expected to create one. 230 231When set to "0", writing to regular files is unrestricted. 232 233When set to "1" don't allow O_CREAT open on regular files that we 234don't own in world writable sticky directories, unless they are 235owned by the owner of the directory. 236 237When set to "2" it also applies to group writable sticky directories. 238 239============================================================== 240 241protected_symlinks: 242 243A long-standing class of security issues is the symlink-based 244time-of-check-time-of-use race, most commonly seen in world-writable 245directories like /tmp. The common method of exploitation of this flaw 246is to cross privilege boundaries when following a given symlink (i.e. a 247root process follows a symlink belonging to another user). For a likely 248incomplete list of hundreds of examples across the years, please see: 249http://cve.mitre.org/cgi-bin/cvekey.cgi?keyword=/tmp 250 251When set to "0", symlink following behavior is unrestricted. 252 253When set to "1" symlinks are permitted to be followed only when outside 254a sticky world-writable directory, or when the uid of the symlink and 255follower match, or when the directory owner matches the symlink's owner. 256 257This protection is based on the restrictions in Openwall and grsecurity. 258 259============================================================== 260 261suid_dumpable: 262 263This value can be used to query and set the core dump mode for setuid 264or otherwise protected/tainted binaries. The modes are 265 2660 - (default) - traditional behaviour. Any process which has changed 267 privilege levels or is execute only will not be dumped. 2681 - (debug) - all processes dump core when possible. The core dump is 269 owned by the current user and no security is applied. This is 270 intended for system debugging situations only. Ptrace is unchecked. 271 This is insecure as it allows regular users to examine the memory 272 contents of privileged processes. 2732 - (suidsafe) - any binary which normally would not be dumped is dumped 274 anyway, but only if the "core_pattern" kernel sysctl is set to 275 either a pipe handler or a fully qualified path. (For more details 276 on this limitation, see CVE-2006-2451.) This mode is appropriate 277 when administrators are attempting to debug problems in a normal 278 environment, and either have a core dump pipe handler that knows 279 to treat privileged core dumps with care, or specific directory 280 defined for catching core dumps. If a core dump happens without 281 a pipe handler or fully qualifid path, a message will be emitted 282 to syslog warning about the lack of a correct setting. 283 284============================================================== 285 286super-max & super-nr: 287 288These numbers control the maximum number of superblocks, and 289thus the maximum number of mounted filesystems the kernel 290can have. You only need to increase super-max if you need to 291mount more filesystems than the current value in super-max 292allows you to. 293 294============================================================== 295 296aio-nr & aio-max-nr: 297 298aio-nr shows the current system-wide number of asynchronous io 299requests. aio-max-nr allows you to change the maximum value 300aio-nr can grow to. 301 302============================================================== 303 304mount-max: 305 306This denotes the maximum number of mounts that may exist 307in a mount namespace. 308 309============================================================== 310 311 3122. /proc/sys/fs/binfmt_misc 313---------------------------------------------------------- 314 315Documentation for the files in /proc/sys/fs/binfmt_misc is 316in Documentation/binfmt_misc.txt. 317 318 3193. /proc/sys/fs/mqueue - POSIX message queues filesystem 320---------------------------------------------------------- 321 322The "mqueue" filesystem provides the necessary kernel features to enable the 323creation of a user space library that implements the POSIX message queues 324API (as noted by the MSG tag in the POSIX 1003.1-2001 version of the System 325Interfaces specification.) 326 327The "mqueue" filesystem contains values for determining/setting the amount of 328resources used by the file system. 329 330/proc/sys/fs/mqueue/queues_max is a read/write file for setting/getting the 331maximum number of message queues allowed on the system. 332 333/proc/sys/fs/mqueue/msg_max is a read/write file for setting/getting the 334maximum number of messages in a queue value. In fact it is the limiting value 335for another (user) limit which is set in mq_open invocation. This attribute of 336a queue must be less or equal then msg_max. 337 338/proc/sys/fs/mqueue/msgsize_max is a read/write file for setting/getting the 339maximum message size value (it is every message queue's attribute set during 340its creation). 341 342/proc/sys/fs/mqueue/msg_default is a read/write file for setting/getting the 343default number of messages in a queue value if attr parameter of mq_open(2) is 344NULL. If it exceed msg_max, the default value is initialized msg_max. 345 346/proc/sys/fs/mqueue/msgsize_default is a read/write file for setting/getting 347the default message size value if attr parameter of mq_open(2) is NULL. If it 348exceed msgsize_max, the default value is initialized msgsize_max. 349 3504. /proc/sys/fs/epoll - Configuration options for the epoll interface 351-------------------------------------------------------- 352 353This directory contains configuration options for the epoll(7) interface. 354 355max_user_watches 356---------------- 357 358Every epoll file descriptor can store a number of files to be monitored 359for event readiness. Each one of these monitored files constitutes a "watch". 360This configuration option sets the maximum number of "watches" that are 361allowed for each user. 362Each "watch" costs roughly 90 bytes on a 32bit kernel, and roughly 160 bytes 363on a 64bit one. 364The current default value for max_user_watches is the 1/32 of the available 365low memory, divided for the "watch" cost in bytes. 366 367