{ "mode": "man", "parameter": "capabilities", "section": "7", "url": "https://www.chedong.com/phpMan.php/man/capabilities/7/json", "generated": "2026-05-30T05:09:54Z", "sections": { "NAME": { "content": "capabilities - overview of Linux capabilities\n", "subsections": [] }, "DESCRIPTION": { "content": "For the purpose of performing permission checks, traditional UNIX implementations distinguish\ntwo categories of processes: privileged processes (whose effective user ID is 0, referred to\nas superuser or root), and unprivileged processes (whose effective UID is nonzero). Privi‐\nleged processes bypass all kernel permission checks, while unprivileged processes are subject\nto full permission checking based on the process's credentials (usually: effective UID, ef‐\nfective GID, and supplementary group list).\n\nStarting with kernel 2.2, Linux divides the privileges traditionally associated with supe‐\nruser into distinct units, known as capabilities, which can be independently enabled and dis‐\nabled. Capabilities are a per-thread attribute.\n", "subsections": [ { "name": "Capabilities list", "content": "The following list shows the capabilities implemented on Linux, and the operations or behav‐\niors that each capability permits:\n\nCAPAUDITCONTROL (since Linux 2.6.11)\nEnable and disable kernel auditing; change auditing filter rules; retrieve auditing\nstatus and filtering rules.\n\nCAPAUDITREAD (since Linux 3.16)\nAllow reading the audit log via a multicast netlink socket.\n\nCAPAUDITWRITE (since Linux 2.6.11)\nWrite records to kernel auditing log.\n\nCAPBLOCKSUSPEND (since Linux 3.5)\nEmploy features that can block system suspend (epoll(7) EPOLLWAKEUP,\n/proc/sys/wakelock).\n\nCAPBPF (since Linux 5.8)\nEmploy privileged BPF operations; see bpf(2) and bpf-helpers(7).\n\nThis capability was added in Linux 5.8 to separate out BPF functionality from the\noverloaded CAPSYSADMIN capability.\n\nCAPCHECKPOINTRESTORE (since Linux 5.9)\n* Update /proc/sys/kernel/nslastpid (see pidnamespaces(7));\n* employ the settid feature of clone3(2);\n* read the contents of the symbolic links in /proc/[pid]/mapfiles for other pro‐\ncesses.\n\nThis capability was added in Linux 5.9 to separate out checkpoint/restore functional‐\nity from the overloaded CAPSYSADMIN capability.\n\nCAPCHOWN\nMake arbitrary changes to file UIDs and GIDs (see chown(2)).\n\nCAPDACOVERRIDE\nBypass file read, write, and execute permission checks. (DAC is an abbreviation of\n\"discretionary access control\".)\n\nCAPDACREADSEARCH\n* Bypass file read permission checks and directory read and execute permission checks;\n* invoke openbyhandleat(2);\n* use the linkat(2) ATEMPTYPATH flag to create a link to a file referred to by a\nfile descriptor.\n\nCAPFOWNER\n* Bypass permission checks on operations that normally require the filesystem UID of\nthe process to match the UID of the file (e.g., chmod(2), utime(2)), excluding those\noperations covered by CAPDACOVERRIDE and CAPDACREADSEARCH;\n* set inode flags (see ioctliflags(2)) on arbitrary files;\n* set Access Control Lists (ACLs) on arbitrary files;\n* ignore directory sticky bit on file deletion;\n* modify user extended attributes on sticky directory owned by any user;\n* specify ONOATIME for arbitrary files in open(2) and fcntl(2).\n\nCAPFSETID\n* Don't clear set-user-ID and set-group-ID mode bits when a file is modified;\n* set the set-group-ID bit for a file whose GID does not match the filesystem or any\nof the supplementary GIDs of the calling process.\n\nCAPIPCLOCK\nLock memory (mlock(2), mlockall(2), mmap(2), shmctl(2)).\n\nCAPIPCOWNER\nBypass permission checks for operations on System V IPC objects.\n\nCAPKILL\nBypass permission checks for sending signals (see kill(2)). This includes use of the\nioctl(2) KDSIGACCEPT operation.\n\nCAPLEASE (since Linux 2.4)\nEstablish leases on arbitrary files (see fcntl(2)).\n\nCAPLINUXIMMUTABLE\nSet the FSAPPENDFL and FSIMMUTABLEFL inode flags (see ioctliflags(2)).\n\nCAPMACADMIN (since Linux 2.6.25)\nAllow MAC configuration or state changes. Implemented for the Smack Linux Security\nModule (LSM).\n\nCAPMACOVERRIDE (since Linux 2.6.25)\nOverride Mandatory Access Control (MAC). Implemented for the Smack LSM.\n\nCAPMKNOD (since Linux 2.4)\nCreate special files using mknod(2).\n\nCAPNETADMIN\nPerform various network-related operations:\n* interface configuration;\n* administration of IP firewall, masquerading, and accounting;\n* modify routing tables;\n* bind to any address for transparent proxying;\n* set type-of-service (TOS);\n* clear driver statistics;\n* set promiscuous mode;\n* enabling multicasting;\n* use setsockopt(2) to set the following socket options: SODEBUG, SOMARK, SOPRIOR‐‐\nITY (for a priority outside the range 0 to 6), SORCVBUFFORCE, and SOSNDBUFFORCE.\n\nCAPNETBINDSERVICE\nBind a socket to Internet domain privileged ports (port numbers less than 1024).\n\nCAPNETBROADCAST\n(Unused) Make socket broadcasts, and listen to multicasts.\n\nCAPNETRAW\n* Use RAW and PACKET sockets;\n* bind to any address for transparent proxying.\n\nCAPPERFMON (since Linux 5.8)\nEmploy various performance-monitoring mechanisms, including:\n\n* call perfeventopen(2);\n* employ various BPF operations that have performance implications.\n\nThis capability was added in Linux 5.8 to separate out performance monitoring func‐\ntionality from the overloaded CAPSYSADMIN capability. See also the kernel source\nfile Documentation/admin-guide/perf-security.rst.\n\nCAPSETGID\n* Make arbitrary manipulations of process GIDs and supplementary GID list;\n* forge GID when passing socket credentials via UNIX domain sockets;\n* write a group ID mapping in a user namespace (see usernamespaces(7)).\n\nCAPSETFCAP (since Linux 2.6.24)\nSet arbitrary capabilities on a file.\n\nCAPSETPCAP\nIf file capabilities are supported (i.e., since Linux 2.6.24): add any capability from\nthe calling thread's bounding set to its inheritable set; drop capabilities from the\nbounding set (via prctl(2) PRCAPBSETDROP); make changes to the securebits flags.\n\nIf file capabilities are not supported (i.e., kernels before Linux 2.6.24): grant or\nremove any capability in the caller's permitted capability set to or from any other\nprocess. (This property of CAPSETPCAP is not available when the kernel is configured\nto support file capabilities, since CAPSETPCAP has entirely different semantics for\nsuch kernels.)\n\nCAPSETUID\n* Make arbitrary manipulations of process UIDs (setuid(2), setreuid(2), setresuid(2),\nsetfsuid(2));\n* forge UID when passing socket credentials via UNIX domain sockets;\n* write a user ID mapping in a user namespace (see usernamespaces(7)).\n\nCAPSYSADMIN\nNote: this capability is overloaded; see Notes to kernel developers, below.\n\n* Perform a range of system administration operations including: quotactl(2),\nmount(2), umount(2), pivotroot(2), swapon(2), swapoff(2), sethostname(2), and set‐‐\ndomainname(2);\n* perform privileged syslog(2) operations (since Linux 2.6.37, CAPSYSLOG should be\nused to permit such operations);\n* perform VM86REQUESTIRQ vm86(2) command;\n* access the same checkpoint/restore functionality that is governed by CAPCHECK‐‐\nPOINTRESTORE (but the latter, weaker capability is preferred for accessing that\nfunctionality).\n* perform the same BPF operations as are governed by CAPBPF (but the latter, weaker\ncapability is preferred for accessing that functionality).\n* employ the same performance monitoring mechanisms as are governed by CAPPERFMON\n(but the latter, weaker capability is preferred for accessing that functionality).\n* perform IPCSET and IPCRMID operations on arbitrary System V IPC objects;\n* override RLIMITNPROC resource limit;\n* perform operations on trusted and security extended attributes (see xattr(7));\n* use lookupdcookie(2);\n* use ioprioset(2) to assign IOPRIOCLASSRT and (before Linux 2.6.25) IO‐‐\nPRIOCLASSIDLE I/O scheduling classes;\n* forge PID when passing socket credentials via UNIX domain sockets;\n* exceed /proc/sys/fs/file-max, the system-wide limit on the number of open files, in\nsystem calls that open files (e.g., accept(2), execve(2), open(2), pipe(2));\n* employ CLONE* flags that create new namespaces with clone(2) and unshare(2) (but,\nsince Linux 3.8, creating user namespaces does not require any capability);\n* access privileged perf event information;\n* call setns(2) (requires CAPSYSADMIN in the target namespace);\n* call fanotifyinit(2);\n* perform privileged KEYCTLCHOWN and KEYCTLSETPERM keyctl(2) operations;\n* perform madvise(2) MADVHWPOISON operation;\n* employ the TIOCSTI ioctl(2) to insert characters into the input queue of a terminal\nother than the caller's controlling terminal;\n* employ the obsolete nfsservctl(2) system call;\n* employ the obsolete bdflush(2) system call;\n* perform various privileged block-device ioctl(2) operations;\n* perform various privileged filesystem ioctl(2) operations;\n* perform privileged ioctl(2) operations on the /dev/random device (see random(4));\n* install a seccomp(2) filter without first having to set the nonewprivs thread at‐\ntribute;\n* modify allow/deny rules for device control groups;\n* employ the ptrace(2) PTRACESECCOMPGETFILTER operation to dump tracee's seccomp\nfilters;\n* employ the ptrace(2) PTRACESETOPTIONS operation to suspend the tracee's seccomp\nprotections (i.e., the PTRACEOSUSPENDSECCOMP flag);\n* perform administrative operations on many device drivers;\n* modify autogroup nice values by writing to /proc/[pid]/autogroup (see sched(7)).\n\nCAPSYSBOOT\nUse reboot(2) and kexecload(2).\n\nCAPSYSCHROOT\n* Use chroot(2);\n* change mount namespaces using setns(2).\n\nCAPSYSMODULE\n* Load and unload kernel modules (see initmodule(2) and deletemodule(2));\n* in kernels before 2.6.25: drop capabilities from the system-wide capability bounding\nset.\n\nCAPSYSNICE\n* Lower the process nice value (nice(2), setpriority(2)) and change the nice value for\narbitrary processes;\n* set real-time scheduling policies for calling process, and set scheduling policies\nand priorities for arbitrary processes (schedsetscheduler(2), schedsetparam(2),\nschedsetattr(2));\n* set CPU affinity for arbitrary processes (schedsetaffinity(2));\n* set I/O scheduling class and priority for arbitrary processes (ioprioset(2));\n* apply migratepages(2) to arbitrary processes and allow processes to be migrated to\narbitrary nodes;\n* apply movepages(2) to arbitrary processes;\n* use the MPOLMFMOVEALL flag with mbind(2) and movepages(2).\n\nCAPSYSPACCT\nUse acct(2).\n\nCAPSYSPTRACE\n* Trace arbitrary processes using ptrace(2);\n* apply getrobustlist(2) to arbitrary processes;\n* transfer data to or from the memory of arbitrary processes using processvmreadv(2)\nand processvmwritev(2);\n* inspect processes using kcmp(2).\n\nCAPSYSRAWIO\n* Perform I/O port operations (iopl(2) and ioperm(2));\n* access /proc/kcore;\n* employ the FIBMAP ioctl(2) operation;\n* open devices for accessing x86 model-specific registers (MSRs, see msr(4));\n* update /proc/sys/vm/mmapminaddr;\n* create memory mappings at addresses below the value specified by\n/proc/sys/vm/mmapminaddr;\n* map files in /proc/bus/pci;\n* open /dev/mem and /dev/kmem;\n* perform various SCSI device commands;\n* perform certain operations on hpsa(4) and cciss(4) devices;\n* perform a range of device-specific operations on other devices.\n\nCAPSYSRESOURCE\n* Use reserved space on ext2 filesystems;\n* make ioctl(2) calls controlling ext3 journaling;\n* override disk quota limits;\n* increase resource limits (see setrlimit(2));\n* override RLIMITNPROC resource limit;\n* override maximum number of consoles on console allocation;\n* override maximum number of keymaps;\n* allow more than 64hz interrupts from the real-time clock;\n* raise msgqbytes limit for a System V message queue above the limit in\n/proc/sys/kernel/msgmnb (see msgop(2) and msgctl(2));\n* allow the RLIMITNOFILE resource limit on the number of \"in-flight\" file descriptors\nto be bypassed when passing file descriptors to another process via a UNIX domain\nsocket (see unix(7));\n* override the /proc/sys/fs/pipe-size-max limit when setting the capacity of a pipe\nusing the FSETPIPESZ fcntl(2) command;\n* use FSETPIPESZ to increase the capacity of a pipe above the limit specified by\n/proc/sys/fs/pipe-max-size;\n* override /proc/sys/fs/mqueue/queuesmax, /proc/sys/fs/mqueue/msgmax, and\n/proc/sys/fs/mqueue/msgsizemax limits when creating POSIX message queues (see\nmqoverview(7));\n* employ the prctl(2) PRSETMM operation;\n* set /proc/[pid]/oomscoreadj to a value lower than the value last set by a process\nwith CAPSYSRESOURCE.\n\nCAPSYSTIME\nSet system clock (settimeofday(2), stime(2), adjtimex(2)); set real-time (hardware)\nclock.\n\nCAPSYSTTYCONFIG\nUse vhangup(2); employ various privileged ioctl(2) operations on virtual terminals.\n\nCAPSYSLOG (since Linux 2.6.37)\n* Perform privileged syslog(2) operations. See syslog(2) for information on which op‐\nerations require privilege.\n* View kernel addresses exposed via /proc and other interfaces when /proc/sys/ker‐\nnel/kptrrestrict has the value 1. (See the discussion of the kptrrestrict in\nproc(5).)\n\nCAPWAKEALARM (since Linux 3.0)\nTrigger something that will wake up the system (set CLOCKREALTIMEALARM and\nCLOCKBOOTTIMEALARM timers).\n" }, { "name": "Past and current implementation", "content": "A full implementation of capabilities requires that:\n\n1. For all privileged operations, the kernel must check whether the thread has the required\ncapability in its effective set.\n\n2. The kernel must provide system calls allowing a thread's capability sets to be changed and\nretrieved.\n\n3. The filesystem must support attaching capabilities to an executable file, so that a\nprocess gains those capabilities when the file is executed.\n\nBefore kernel 2.6.24, only the first two of these requirements are met; since kernel 2.6.24,\nall three requirements are met.\n" }, { "name": "Notes to kernel developers", "content": "When adding a new kernel feature that should be governed by a capability, consider the fol‐\nlowing points.\n\n* The goal of capabilities is divide the power of superuser into pieces, such that if a pro‐\ngram that has one or more capabilities is compromised, its power to do damage to the sys‐\ntem would be less than the same program running with root privilege.\n\n* You have the choice of either creating a new capability for your new feature, or associat‐\ning the feature with one of the existing capabilities. In order to keep the set of capa‐\nbilities to a manageable size, the latter option is preferable, unless there are com‐\npelling reasons to take the former option. (There is also a technical limit: the size of\ncapability sets is currently limited to 64 bits.)\n\n* To determine which existing capability might best be associated with your new feature, re‐\nview the list of capabilities above in order to find a \"silo\" into which your new feature\nbest fits. One approach to take is to determine if there are other features requiring ca‐\npabilities that will always be used along with the new feature. If the new feature is\nuseless without these other features, you should use the same capability as the other fea‐\ntures.\n\n* Don't choose CAPSYSADMIN if you can possibly avoid it! A vast proportion of existing\ncapability checks are associated with this capability (see the partial list above). It\ncan plausibly be called \"the new root\", since on the one hand, it confers a wide range of\npowers, and on the other hand, its broad scope means that this is the capability that is\nrequired by many privileged programs. Don't make the problem worse. The only new fea‐\ntures that should be associated with CAPSYSADMIN are ones that closely match existing\nuses in that silo.\n\n* If you have determined that it really is necessary to create a new capability for your\nfeature, don't make or name it as a \"single-use\" capability. Thus, for example, the addi‐\ntion of the highly specific CAPSYSPACCT was probably a mistake. Instead, try to iden‐\ntify and name your new capability as a broader silo into which other related future use\ncases might fit.\n" }, { "name": "Thread capability sets", "content": "Each thread has the following capability sets containing zero or more of the above capabili‐\nties:\n\nPermitted\nThis is a limiting superset for the effective capabilities that the thread may assume.\nIt is also a limiting superset for the capabilities that may be added to the inherita‐\nble set by a thread that does not have the CAPSETPCAP capability in its effective\nset.\n\nIf a thread drops a capability from its permitted set, it can never reacquire that ca‐\npability (unless it execve(2)s either a set-user-ID-root program, or a program whose\nassociated file capabilities grant that capability).\n\nInheritable\nThis is a set of capabilities preserved across an execve(2). Inheritable capabilities\nremain inheritable when executing any program, and inheritable capabilities are added\nto the permitted set when executing a program that has the corresponding bits set in\nthe file inheritable set.\n\nBecause inheritable capabilities are not generally preserved across execve(2) when\nrunning as a non-root user, applications that wish to run helper programs with ele‐\nvated capabilities should consider using ambient capabilities, described below.\n\nEffective\nThis is the set of capabilities used by the kernel to perform permission checks for\nthe thread.\n\nBounding (per-thread since Linux 2.6.25)\nThe capability bounding set is a mechanism that can be used to limit the capabilities\nthat are gained during execve(2).\n\nSince Linux 2.6.25, this is a per-thread capability set. In older kernels, the capa‐\nbility bounding set was a system wide attribute shared by all threads on the system.\n\nFor more details on the capability bounding set, see below.\n\nAmbient (since Linux 4.3)\nThis is a set of capabilities that are preserved across an execve(2) of a program that\nis not privileged. The ambient capability set obeys the invariant that no capability\ncan ever be ambient if it is not both permitted and inheritable.\n\nThe ambient capability set can be directly modified using prctl(2). Ambient capabili‐\nties are automatically lowered if either of the corresponding permitted or inheritable\ncapabilities is lowered.\n\nExecuting a program that changes UID or GID due to the set-user-ID or set-group-ID\nbits or executing a program that has any file capabilities set will clear the ambient\nset. Ambient capabilities are added to the permitted set and assigned to the effec‐\ntive set when execve(2) is called. If ambient capabilities cause a process's permit‐\nted and effective capabilities to increase during an execve(2), this does not trigger\nthe secure-execution mode described in ld.so(8).\n\nA child created via fork(2) inherits copies of its parent's capability sets. See below for a\ndiscussion of the treatment of capabilities during execve(2).\n\nUsing capset(2), a thread may manipulate its own capability sets (see below).\n\nSince Linux 3.2, the file /proc/sys/kernel/caplastcap exposes the numerical value of the\nhighest capability supported by the running kernel; this can be used to determine the highest\nbit that may be set in a capability set.\n" }, { "name": "File capabilities", "content": "Since kernel 2.6.24, the kernel supports associating capability sets with an executable file\nusing setcap(8). The file capability sets are stored in an extended attribute (see setx‐‐\nattr(2) and xattr(7)) named security.capability. Writing to this extended attribute requires\nthe CAPSETFCAP capability. The file capability sets, in conjunction with the capability\nsets of the thread, determine the capabilities of a thread after an execve(2).\n\nThe three file capability sets are:\n\nPermitted (formerly known as forced):\nThese capabilities are automatically permitted to the thread, regardless of the\nthread's inheritable capabilities.\n\nInheritable (formerly known as allowed):\nThis set is ANDed with the thread's inheritable set to determine which inheritable ca‐\npabilities are enabled in the permitted set of the thread after the execve(2).\n\nEffective:\nThis is not a set, but rather just a single bit. If this bit is set, then during an\nexecve(2) all of the new permitted capabilities for the thread are also raised in the\neffective set. If this bit is not set, then after an execve(2), none of the new per‐\nmitted capabilities is in the new effective set.\n\nEnabling the file effective capability bit implies that any file permitted or inheri‐\ntable capability that causes a thread to acquire the corresponding permitted capabil‐\nity during an execve(2) (see the transformation rules described below) will also ac‐\nquire that capability in its effective set. Therefore, when assigning capabilities to\na file (setcap(8), capsetfile(3), capsetfd(3)), if we specify the effective flag\nas being enabled for any capability, then the effective flag must also be specified as\nenabled for all other capabilities for which the corresponding permitted or inherita‐\nble flags is enabled.\n" }, { "name": "File capability extended attribute versioning", "content": "To allow extensibility, the kernel supports a scheme to encode a version number inside the\nsecurity.capability extended attribute that is used to implement file capabilities. These\nversion numbers are internal to the implementation, and not directly visible to user-space\napplications. To date, the following versions are supported:\n\nVFSCAPREVISION1\nThis was the original file capability implementation, which supported 32-bit masks for\nfile capabilities.\n\nVFSCAPREVISION2 (since Linux 2.6.25)\nThis version allows for file capability masks that are 64 bits in size, and was neces‐\nsary as the number of supported capabilities grew beyond 32. The kernel transparently\ncontinues to support the execution of files that have 32-bit version 1 capability\nmasks, but when adding capabilities to files that did not previously have capabili‐\nties, or modifying the capabilities of existing files, it automatically uses the ver‐\nsion 2 scheme (or possibly the version 3 scheme, as described below).\n\nVFSCAPREVISION3 (since Linux 4.14)\nVersion 3 file capabilities are provided to support namespaced file capabilities (de‐\nscribed below).\n\nAs with version 2 file capabilities, version 3 capability masks are 64 bits in size.\nBut in addition, the root user ID of namespace is encoded in the security.capability\nextended attribute. (A namespace's root user ID is the value that user ID 0 inside\nthat namespace maps to in the initial user namespace.)\n\nVersion 3 file capabilities are designed to coexist with version 2 capabilities; that\nis, on a modern Linux system, there may be some files with version 2 capabilities\nwhile others have version 3 capabilities.\n\nBefore Linux 4.14, the only kind of file capability extended attribute that could be attached\nto a file was a VFSCAPREVISION2 attribute. Since Linux 4.14, the version of the secu‐\nrity.capability extended attribute that is attached to a file depends on the circumstances in\nwhich the attribute was created.\n\nStarting with Linux 4.14, a security.capability extended attribute is automatically created\nas (or converted to) a version 3 (VFSCAPREVISION3) attribute if both of the following are\ntrue:\n\n(1) The thread writing the attribute resides in a noninitial user namespace. (More pre‐\ncisely: the thread resides in a user namespace other than the one from which the underly‐\ning filesystem was mounted.)\n\n(2) The thread has the CAPSETFCAP capability over the file inode, meaning that (a) the\nthread has the CAPSETFCAP capability in its own user namespace; and (b) the UID and GID\nof the file inode have mappings in the writer's user namespace.\n\nWhen a VFSCAPREVISION3 security.capability extended attribute is created, the root user ID\nof the creating thread's user namespace is saved in the extended attribute.\n\nBy contrast, creating or modifying a security.capability extended attribute from a privileged\n(CAPSETFCAP) thread that resides in the namespace where the underlying filesystem was\nmounted (this normally means the initial user namespace) automatically results in the cre‐\nation of a version 2 (VFSCAPREVISION2) attribute.\n\nNote that the creation of a version 3 security.capability extended attribute is automatic.\nThat is to say, when a user-space application writes (setxattr(2)) a security.capability at‐\ntribute in the version 2 format, the kernel will automatically create a version 3 attribute\nif the attribute is created in the circumstances described above. Correspondingly, when a\nversion 3 security.capability attribute is retrieved (getxattr(2)) by a process that resides\ninside a user namespace that was created by the root user ID (or a descendant of that user\nnamespace), the returned attribute is (automatically) simplified to appear as a version 2 at‐\ntribute (i.e., the returned value is the size of a version 2 attribute and does not include\nthe root user ID). These automatic translations mean that no changes are required to user-\nspace tools (e.g., setcap(1) and getcap(1)) in order for those tools to be used to create and\nretrieve version 3 security.capability attributes.\n\nNote that a file can have either a version 2 or a version 3 security.capability extended at‐\ntribute associated with it, but not both: creation or modification of the security.capability\nextended attribute will automatically modify the version according to the circumstances in\nwhich the extended attribute is created or modified.\n" }, { "name": "Transformation of capabilities during execve()", "content": "During an execve(2), the kernel calculates the new capabilities of the process using the fol‐\nlowing algorithm:\n\nP'(ambient) = (file is privileged) ? 0 : P(ambient)\n\nP'(permitted) = (P(inheritable) & F(inheritable)) |\n(F(permitted) & P(bounding)) | P'(ambient)\n\nP'(effective) = F(effective) ? P'(permitted) : P'(ambient)\n\nP'(inheritable) = P(inheritable) [i.e., unchanged]\n\nP'(bounding) = P(bounding) [i.e., unchanged]\n\nwhere:\n\nP() denotes the value of a thread capability set before the execve(2)\n\nP'() denotes the value of a thread capability set after the execve(2)\n\nF() denotes a file capability set\n\nNote the following details relating to the above capability transformation rules:\n\n* The ambient capability set is present only since Linux 4.3. When determining the trans‐\nformation of the ambient set during execve(2), a privileged file is one that has capabili‐\nties or has the set-user-ID or set-group-ID bit set.\n\n* Prior to Linux 2.6.25, the bounding set was a system-wide attribute shared by all threads.\nThat system-wide value was employed to calculate the new permitted set during execve(2) in\nthe same manner as shown above for P(bounding).\n\nNote: during the capability transitions described above, file capabilities may be ignored\n(treated as empty) for the same reasons that the set-user-ID and set-group-ID bits are ig‐\nnored; see execve(2). File capabilities are similarly ignored if the kernel was booted with\nthe nofilecaps option.\n\nNote: according to the rules above, if a process with nonzero user IDs performs an execve(2)\nthen any capabilities that are present in its permitted and effective sets will be cleared.\nFor the treatment of capabilities when a process with a user ID of zero performs an ex‐‐\necve(2), see below under Capabilities and execution of programs by root.\n" }, { "name": "Safety checking for capability-dumb binaries", "content": "A capability-dumb binary is an application that has been marked to have file capabilities,\nbut has not been converted to use the libcap(3) API to manipulate its capabilities. (In\nother words, this is a traditional set-user-ID-root program that has been switched to use\nfile capabilities, but whose code has not been modified to understand capabilities.) For\nsuch applications, the effective capability bit is set on the file, so that the file permit‐\nted capabilities are automatically enabled in the process effective set when executing the\nfile. The kernel recognizes a file which has the effective capability bit set as capability-\ndumb for the purpose of the check described here.\n\nWhen executing a capability-dumb binary, the kernel checks if the process obtained all per‐\nmitted capabilities that were specified in the file permitted set, after the capability\ntransformations described above have been performed. (The typical reason why this might not\noccur is that the capability bounding set masked out some of the capabilities in the file\npermitted set.) If the process did not obtain the full set of file permitted capabilities,\nthen execve(2) fails with the error EPERM. This prevents possible security risks that could\narise when a capability-dumb application is executed with less privilege that it needs. Note\nthat, by definition, the application could not itself recognize this problem, since it does\nnot employ the libcap(3) API.\n" }, { "name": "Capabilities and execution of programs by root", "content": "In order to mirror traditional UNIX semantics, the kernel performs special treatment of file\ncapabilities when a process with UID 0 (root) executes a program and when a set-user-ID-root\nprogram is executed.\n\nAfter having performed any changes to the process effective ID that were triggered by the\nset-user-ID mode bit of the binary—e.g., switching the effective user ID to 0 (root) because\na set-user-ID-root program was executed—the kernel calculates the file capability sets as\nfollows:\n\n1. If the real or effective user ID of the process is 0 (root), then the file inheritable and\npermitted sets are ignored; instead they are notionally considered to be all ones (i.e.,\nall capabilities enabled). (There is one exception to this behavior, described below in\nSet-user-ID-root programs that have file capabilities.)\n\n2. If the effective user ID of the process is 0 (root) or the file effective bit is in fact\nenabled, then the file effective bit is notionally defined to be one (enabled).\n\nThese notional values for the file's capability sets are then used as described above to cal‐\nculate the transformation of the process's capabilities during execve(2).\n\nThus, when a process with nonzero UIDs execve(2)s a set-user-ID-root program that does not\nhave capabilities attached, or when a process whose real and effective UIDs are zero ex‐‐\necve(2)s a program, the calculation of the process's new permitted capabilities simplifies\nto:\n\nP'(permitted) = P(inheritable) | P(bounding)\n\nP'(effective) = P'(permitted)\n\nConsequently, the process gains all capabilities in its permitted and effective capability\nsets, except those masked out by the capability bounding set. (In the calculation of P'(per‐\nmitted), the P'(ambient) term can be simplified away because it is by definition a proper\nsubset of P(inheritable).)\n\nThe special treatments of user ID 0 (root) described in this subsection can be disabled using\nthe securebits mechanism described below.\n" }, { "name": "Set-user-ID-root programs that have file capabilities", "content": "There is one exception to the behavior described under Capabilities and execution of programs\nby root. If (a) the binary that is being executed has capabilities attached and (b) the real\nuser ID of the process is not 0 (root) and (c) the effective user ID of the process is 0\n(root), then the file capability bits are honored (i.e., they are not notionally considered\nto be all ones). The usual way in which this situation can arise is when executing a set-\nUID-root program that also has file capabilities. When such a program is executed, the\nprocess gains just the capabilities granted by the program (i.e., not all capabilities, as\nwould occur when executing a set-user-ID-root program that does not have any associated file\ncapabilities).\n\nNote that one can assign empty capability sets to a program file, and thus it is possible to\ncreate a set-user-ID-root program that changes the effective and saved set-user-ID of the\nprocess that executes the program to 0, but confers no capabilities to that process.\n" }, { "name": "Capability bounding set", "content": "The capability bounding set is a security mechanism that can be used to limit the capabili‐\nties that can be gained during an execve(2). The bounding set is used in the following ways:\n\n* During an execve(2), the capability bounding set is ANDed with the file permitted capabil‐\nity set, and the result of this operation is assigned to the thread's permitted capability\nset. The capability bounding set thus places a limit on the permitted capabilities that\nmay be granted by an executable file.\n\n* (Since Linux 2.6.25) The capability bounding set acts as a limiting superset for the capa‐\nbilities that a thread can add to its inheritable set using capset(2). This means that if\na capability is not in the bounding set, then a thread can't add this capability to its in‐\nheritable set, even if it was in its permitted capabilities, and thereby cannot have this\ncapability preserved in its permitted set when it execve(2)s a file that has the capability\nin its inheritable set.\n\nNote that the bounding set masks the file permitted capabilities, but not the inheritable ca‐\npabilities. If a thread maintains a capability in its inheritable set that is not in its\nbounding set, then it can still gain that capability in its permitted set by executing a file\nthat has the capability in its inheritable set.\n\nDepending on the kernel version, the capability bounding set is either a system-wide attri‐\nbute, or a per-process attribute.\n" }, { "name": "Capability bounding set from Linux 2.6.25 onward", "content": "From Linux 2.6.25, the capability bounding set is a per-thread attribute. (The system-wide\ncapability bounding set described below no longer exists.)\n\nThe bounding set is inherited at fork(2) from the thread's parent, and is preserved across an\nexecve(2).\n\nA thread may remove capabilities from its capability bounding set using the prctl(2) PRCAPB‐‐\nSETDROP operation, provided it has the CAPSETPCAP capability. Once a capability has been\ndropped from the bounding set, it cannot be restored to that set. A thread can determine if\na capability is in its bounding set using the prctl(2) PRCAPBSETREAD operation.\n\nRemoving capabilities from the bounding set is supported only if file capabilities are com‐\npiled into the kernel. In kernels before Linux 2.6.33, file capabilities were an optional\nfeature configurable via the CONFIGSECURITYFILECAPABILITIES option. Since Linux 2.6.33,\nthe configuration option has been removed and file capabilities are always part of the ker‐\nnel. When file capabilities are compiled into the kernel, the init process (the ancestor of\nall processes) begins with a full bounding set. If file capabilities are not compiled into\nthe kernel, then init begins with a full bounding set minus CAPSETPCAP, because this capa‐\nbility has a different meaning when there are no file capabilities.\n\nRemoving a capability from the bounding set does not remove it from the thread's inheritable\nset. However it does prevent the capability from being added back into the thread's inheri‐\ntable set in the future.\n" }, { "name": "Capability bounding set prior to Linux 2.6.25", "content": "In kernels before 2.6.25, the capability bounding set is a system-wide attribute that affects\nall threads on the system. The bounding set is accessible via the file /proc/sys/kernel/cap-\nbound. (Confusingly, this bit mask parameter is expressed as a signed decimal number in\n/proc/sys/kernel/cap-bound.)\n\nOnly the init process may set capabilities in the capability bounding set; other than that,\nthe superuser (more precisely: a process with the CAPSYSMODULE capability) may only clear\ncapabilities from this set.\n\nOn a standard system the capability bounding set always masks out the CAPSETPCAP capability.\nTo remove this restriction (dangerous!), modify the definition of CAPINITEFFSET in in‐\nclude/linux/capability.h and rebuild the kernel.\n\nThe system-wide capability bounding set feature was added to Linux starting with kernel ver‐\nsion 2.2.11.\n" }, { "name": "Effect of user ID changes on capabilities", "content": "To preserve the traditional semantics for transitions between 0 and nonzero user IDs, the\nkernel makes the following changes to a thread's capability sets on changes to the thread's\nreal, effective, saved set, and filesystem user IDs (using setuid(2), setresuid(2), or simi‐\nlar):\n\n1. If one or more of the real, effective or saved set user IDs was previously 0, and as a re‐\nsult of the UID changes all of these IDs have a nonzero value, then all capabilities are\ncleared from the permitted, effective, and ambient capability sets.\n\n2. If the effective user ID is changed from 0 to nonzero, then all capabilities are cleared\nfrom the effective set.\n\n3. If the effective user ID is changed from nonzero to 0, then the permitted set is copied to\nthe effective set.\n\n4. If the filesystem user ID is changed from 0 to nonzero (see setfsuid(2)), then the follow‐\ning capabilities are cleared from the effective set: CAPCHOWN, CAPDACOVERRIDE,\nCAPDACREADSEARCH, CAPFOWNER, CAPFSETID, CAPLINUXIMMUTABLE (since Linux 2.6.30),\nCAPMACOVERRIDE, and CAPMKNOD (since Linux 2.6.30). If the filesystem UID is changed\nfrom nonzero to 0, then any of these capabilities that are enabled in the permitted set\nare enabled in the effective set.\n\nIf a thread that has a 0 value for one or more of its user IDs wants to prevent its permitted\ncapability set being cleared when it resets all of its user IDs to nonzero values, it can do\nso using the SECBITKEEPCAPS securebits flag described below.\n" }, { "name": "Programmatically adjusting capability sets", "content": "A thread can retrieve and change its permitted, effective, and inheritable capability sets\nusing the capget(2) and capset(2) system calls. However, the use of capgetproc(3) and\ncapsetproc(3), both provided in the libcap package, is preferred for this purpose. The\nfollowing rules govern changes to the thread capability sets:\n\n1. If the caller does not have the CAPSETPCAP capability, the new inheritable set must be a\nsubset of the combination of the existing inheritable and permitted sets.\n\n2. (Since Linux 2.6.25) The new inheritable set must be a subset of the combination of the\nexisting inheritable set and the capability bounding set.\n\n3. The new permitted set must be a subset of the existing permitted set (i.e., it is not pos‐\nsible to acquire permitted capabilities that the thread does not currently have).\n\n4. The new effective set must be a subset of the new permitted set.\n" }, { "name": "The securebits flags: establishing a capabilities-only environment", "content": "Starting with kernel 2.6.26, and with a kernel in which file capabilities are enabled, Linux\nimplements a set of per-thread securebits flags that can be used to disable special handling\nof capabilities for UID 0 (root). These flags are as follows:\n\nSECBITKEEPCAPS\nSetting this flag allows a thread that has one or more 0 UIDs to retain capabilities\nin its permitted set when it switches all of its UIDs to nonzero values. If this flag\nis not set, then such a UID switch causes the thread to lose all permitted capabili‐\nties. This flag is always cleared on an execve(2).\n\nNote that even with the SECBITKEEPCAPS flag set, the effective capabilities of a\nthread are cleared when it switches its effective UID to a nonzero value. However, if\nthe thread has set this flag and its effective UID is already nonzero, and the thread\nsubsequently switches all other UIDs to nonzero values, then the effective capabili‐\nties will not be cleared.\n\nThe setting of the SECBITKEEPCAPS flag is ignored if the SECBITNOSETUIDFIXUP flag\nis set. (The latter flag provides a superset of the effect of the former flag.)\n\nThis flag provides the same functionality as the older prctl(2) PRSETKEEPCAPS opera‐\ntion.\n\nSECBITNOSETUIDFIXUP\nSetting this flag stops the kernel from adjusting the process's permitted, effective,\nand ambient capability sets when the thread's effective and filesystem UIDs are\nswitched between zero and nonzero values. (See the subsection Effect of user ID\nchanges on capabilities.)\n\nSECBITNOROOT\nIf this bit is set, then the kernel does not grant capabilities when a set-user-ID-\nroot program is executed, or when a process with an effective or real UID of 0 calls\nexecve(2). (See the subsection Capabilities and execution of programs by root.)\n\nSECBITNOCAPAMBIENTRAISE\nSetting this flag disallows raising ambient capabilities via the prctl(2) PRCAPAMBI‐‐\nENTRAISE operation.\n\nEach of the above \"base\" flags has a companion \"locked\" flag. Setting any of the \"locked\"\nflags is irreversible, and has the effect of preventing further changes to the corresponding\n\"base\" flag. The locked flags are: SECBITKEEPCAPSLOCKED, SECBITNOSETUIDFIXUPLOCKED,\nSECBITNOROOTLOCKED, and SECBITNOCAPAMBIENTRAISELOCKED.\n\nThe securebits flags can be modified and retrieved using the prctl(2) PRSETSECUREBITS and\nPRGETSECUREBITS operations. The CAPSETPCAP capability is required to modify the flags.\nNote that the SECBIT* constants are available only after including the \nheader file.\n\nThe securebits flags are inherited by child processes. During an execve(2), all of the flags\nare preserved, except SECBITKEEPCAPS which is always cleared.\n\nAn application can use the following call to lock itself, and all of its descendants, into an\nenvironment where the only way of gaining capabilities is by executing a program with associ‐\nated file capabilities:\n\nprctl(PRSETSECUREBITS,\n/* SECBITKEEPCAPS off */\nSECBITKEEPCAPSLOCKED |\nSECBITNOSETUIDFIXUP |\nSECBITNOSETUIDFIXUPLOCKED |\nSECBITNOROOT |\nSECBITNOROOTLOCKED);\n/* Setting/locking SECBITNOCAPAMBIENTRAISE\nis not required */\n" }, { "name": "Per-user-namespace \"set-user-ID-root\" programs", "content": "A set-user-ID program whose UID matches the UID that created a user namespace will confer ca‐\npabilities in the process's permitted and effective sets when executed by any process inside\nthat namespace or any descendant user namespace.\n\nThe rules about the transformation of the process's capabilities during the execve(2) are ex‐\nactly as described in the subsections Transformation of capabilities during execve() and Ca‐\npabilities and execution of programs by root, with the difference that, in the latter subsec‐\ntion, \"root\" is the UID of the creator of the user namespace.\n" }, { "name": "Namespaced file capabilities", "content": "Traditional (i.e., version 2) file capabilities associate only a set of capability masks with\na binary executable file. When a process executes a binary with such capabilities, it gains\nthe associated capabilities (within its user namespace) as per the rules described above in\n\"Transformation of capabilities during execve()\".\n\nBecause version 2 file capabilities confer capabilities to the executing process regardless\nof which user namespace it resides in, only privileged processes are permitted to associate\ncapabilities with a file. Here, \"privileged\" means a process that has the CAPSETFCAP capa‐\nbility in the user namespace where the filesystem was mounted (normally the initial user\nnamespace). This limitation renders file capabilities useless for certain use cases. For\nexample, in user-namespaced containers, it can be desirable to be able to create a binary\nthat confers capabilities only to processes executed inside that container, but not to pro‐\ncesses that are executed outside the container.\n\nLinux 4.14 added so-called namespaced file capabilities to support such use cases. Names‐\npaced file capabilities are recorded as version 3 (i.e., VFSCAPREVISION3) security.capa‐\nbility extended attributes. Such an attribute is automatically created in the circumstances\ndescribed above under \"File capability extended attribute versioning\". When a version 3 se‐\ncurity.capability extended attribute is created, the kernel records not just the capability\nmasks in the extended attribute, but also the namespace root user ID.\n\nAs with a binary that has VFSCAPREVISION2 file capabilities, a binary with VFSCAPREVI‐‐\nSION3 file capabilities confers capabilities to a process during execve(). However, capa‐\nbilities are conferred only if the binary is executed by a process that resides in a user\nnamespace whose UID 0 maps to the root user ID that is saved in the extended attribute, or\nwhen executed by a process that resides in a descendant of such a namespace.\n" }, { "name": "Interaction with user namespaces", "content": "For further information on the interaction of capabilities and user namespaces, see\nusernamespaces(7).\n" } ] }, "CONFORMING TO": { "content": "No standards govern capabilities, but the Linux capability implementation is based on the\nwithdrawn POSIX.1e draft standard; see ⟨https://archive.org/details/posix1003.1e-990310⟩.\n", "subsections": [] }, "NOTES": { "content": "When attempting to strace(1) binaries that have capabilities (or set-user-ID-root binaries),\nyou may find the -u option useful. Something like:\n\n$ sudo strace -o trace.log -u ceci ./myprivprog\n\nFrom kernel 2.5.27 to kernel 2.6.26, capabilities were an optional kernel component, and\ncould be enabled/disabled via the CONFIGSECURITYCAPABILITIES kernel configuration option.\n\nThe /proc/[pid]/task/TID/status file can be used to view the capability sets of a thread.\nThe /proc/[pid]/status file shows the capability sets of a process's main thread. Before\nLinux 3.8, nonexistent capabilities were shown as being enabled (1) in these sets. Since\nLinux 3.8, all nonexistent capabilities (above CAPLASTCAP) are shown as disabled (0).\n\nThe libcap package provides a suite of routines for setting and getting capabilities that is\nmore comfortable and less likely to change than the interface provided by capset(2) and\ncapget(2). This package also provides the setcap(8) and getcap(8) programs. It can be found\nat\n⟨https://git.kernel.org/pub/scm/libs/libcap/libcap.git/refs/⟩.\n\nBefore kernel 2.6.24, and from kernel 2.6.24 to kernel 2.6.32 if file capabilities are not\nenabled, a thread with the CAPSETPCAP capability can manipulate the capabilities of threads\nother than itself. However, this is only theoretically possible, since no thread ever has\nCAPSETPCAP in either of these cases:\n\n* In the pre-2.6.25 implementation the system-wide capability bounding set, /proc/sys/ker‐\nnel/cap-bound, always masks out the CAPSETPCAP capability, and this can not be changed\nwithout modifying the kernel source and rebuilding the kernel.\n\n* If file capabilities are disabled (i.e., the kernel CONFIGSECURITYFILECAPABILITIES op‐\ntion is disabled), then init starts out with the CAPSETPCAP capability removed from its\nper-process bounding set, and that bounding set is inherited by all other processes created\non the system.\n", "subsections": [] }, "SEE ALSO": { "content": "capsh(1), setpriv(1), prctl(2), setfsuid(2), capclear(3), capcopyext(3), capfromtext(3),\ncapgetfile(3), capgetproc(3), capinit(3), capgetp(3), capsetp(3), libcap(3), proc(5),\ncredentials(7), pthreads(7), usernamespaces(7), captest(8), filecap(8), getcap(8), getp‐‐\ncaps(8), netcap(8), pscap(8), setcap(8)\n\ninclude/linux/capability.h in the Linux kernel source tree\n", "subsections": [] }, "COLOPHON": { "content": "This page is part of release 5.10 of the Linux man-pages project. A description of the\nproject, information about reporting bugs, and the latest version of this page, can be found\nat https://www.kernel.org/doc/man-pages/.\n\n\n\nLinux 2020-08-13 CAPABILITIES(7)", "subsections": [] } }, "summary": "capabilities - overview of Linux capabilities", "flags": [], "examples": [], "see_also": [ { "name": "capsh", "section": "1", "url": "https://www.chedong.com/phpMan.php/man/capsh/1/json" }, { "name": "setpriv", "section": "1", "url": "https://www.chedong.com/phpMan.php/man/setpriv/1/json" }, { "name": "prctl", "section": "2", "url": "https://www.chedong.com/phpMan.php/man/prctl/2/json" }, { "name": "setfsuid", "section": "2", "url": "https://www.chedong.com/phpMan.php/man/setfsuid/2/json" }, { "name": "capclear", "section": "3", "url": "https://www.chedong.com/phpMan.php/man/capclear/3/json" }, { "name": "capcopyext", "section": "3", "url": "https://www.chedong.com/phpMan.php/man/capcopyext/3/json" }, { "name": "capfromtext", "section": "3", "url": "https://www.chedong.com/phpMan.php/man/capfromtext/3/json" }, { "name": "capgetfile", "section": "3", "url": "https://www.chedong.com/phpMan.php/man/capgetfile/3/json" }, { "name": "capgetproc", "section": "3", "url": "https://www.chedong.com/phpMan.php/man/capgetproc/3/json" }, { "name": "capinit", "section": "3", "url": "https://www.chedong.com/phpMan.php/man/capinit/3/json" }, { "name": "capgetp", "section": "3", "url": "https://www.chedong.com/phpMan.php/man/capgetp/3/json" }, { "name": "capsetp", "section": "3", "url": "https://www.chedong.com/phpMan.php/man/capsetp/3/json" }, { "name": "libcap", "section": "3", "url": "https://www.chedong.com/phpMan.php/man/libcap/3/json" }, { "name": "proc", "section": "5", "url": "https://www.chedong.com/phpMan.php/man/proc/5/json" }, { "name": "credentials", "section": "7", "url": "https://www.chedong.com/phpMan.php/man/credentials/7/json" }, { "name": "pthreads", "section": "7", "url": "https://www.chedong.com/phpMan.php/man/pthreads/7/json" }, { "name": "usernamespaces", "section": "7", "url": "https://www.chedong.com/phpMan.php/man/usernamespaces/7/json" }, { "name": "captest", "section": "8", "url": "https://www.chedong.com/phpMan.php/man/captest/8/json" }, { "name": "filecap", "section": "8", "url": "https://www.chedong.com/phpMan.php/man/filecap/8/json" }, { "name": "getcap", "section": "8", "url": "https://www.chedong.com/phpMan.php/man/getcap/8/json" }, { "name": "caps", "section": "8", "url": "https://www.chedong.com/phpMan.php/man/caps/8/json" }, { "name": "netcap", "section": "8", "url": "https://www.chedong.com/phpMan.php/man/netcap/8/json" }, { "name": "pscap", "section": "8", "url": "https://www.chedong.com/phpMan.php/man/pscap/8/json" }, { "name": "setcap", "section": "8", "url": "https://www.chedong.com/phpMan.php/man/setcap/8/json" } ] }