SYSCTL(3) | Library Functions Manual | SYSCTL(3) |
int
sysctl(const int *name, u_int namelen, void *oldp, size_t *oldlenp, const void *newp, size_t newlen);
int
sysctlbyname(const char *sname, void *oldp, size_t *oldlenp, const void *newp, size_t newlen);
int
sysctlgetmibinfo(const char *sname, int *name, u_int *namelenp, char *cname, size_t *csz, struct sysctlnode **rnode, int v);
int
sysctlnametomib(const char *sname, int *name, size_t *namelenp);
Unless explicitly noted below, sysctl returns a consistent snapshot of the data requested. Consistency is obtained by locking the destination buffer into memory so that the data may be copied out without blocking. Calls to sysctl are serialized to avoid deadlock.
The state is described using a ``Management Information Base'' (MIB) style name, listed in name, which is a namelen length array of integers.
The sysctlbyname() function accepts a string representation of a MIB entry and internally maps it to the appropriate numeric MIB representation. Its semantics are otherwise no different from sysctl().
The information is copied into the buffer specified by oldp. The size of the buffer is given by the location specified by oldlenp before the call, and that location gives the amount of data copied after a successful call. If the amount of data available is greater than the size of the buffer supplied, the call supplies as much data as fits in the buffer provided and returns with the error code ENOMEM. If the old value is not desired, oldp and oldlenp should be set to NULL.
The size of the available data can be determined by calling sysctl with a NULL parameter for oldp. The size of the available data will be returned in the location pointed to by oldlenp. For some operations, the amount of space may change often. For these operations, the system attempts to round up so that the returned size is large enough for a call to return the data shortly thereafter.
To set a new value, newp is set to point to a buffer of length newlen from which the requested value is to be taken. If a new value is not to be set, newp should be set to NULL and newlen set to 0.
The sysctlnametomib() function can be used to map the string representation of a MIB entry to the numeric version. The name argument should point to an array of integers large enough to hold the MIB, and namelenp should indicate the number of integer slots available. Following a successful translation, the size_t indicated by namelenp will be changed to show the number of slots consumed.
The sysctlgetmibinfo() function performs name translation similar to sysctlnametomib(), but also canonicalizes the name (or returns the first erroneous token from the string being parsed) into the space indicated by cname and csz. csz should indicate the size of the buffer pointed to by cname and on return, will indicate the size of the returned string including the trailing ‘nul' character.
The rnode and v arguments to sysctlgetmibinfo() are used to provide a tree for it to parse into, and to get back either a pointer to, or a copy of, the terminal node. If rnode is NULL, sysctlgetmibinfo() uses its own internal tree for parsing, and checks it against the kernel at each call, to make sure that the name-to-number mapping is kept up to date. The v argument is ignored in this case. If rnode is not NULL but the pointer it references is, on a successful return, rnode will be adjusted to point to a copy of the terminal node. The v argument indicates which version of the sysctl node structure the caller wants. The application must later free() this copy. If neither rnode nor the pointer it references are NULL, the pointer is used as the address of a tree over which the parsing is done. In this last case, the tree is not checked against the kernel, no refreshing of the mappings is performed, and the value given by v must agree with the version indicated by the tree. It is recommended that applications always use SYSCTL_VERSION as the value for v, as defined in the include file sys/sysctl.h.
The numeric and text names of sysctl variables are described in sysctl(7). The numeric names are defined as preprocessor macros. The top level names are defined with a CTL_ prefix in <sys/sysctl.h>. The next and subsequent levels down have different prefixes for each subtree.
For example, the following retrieves the maximum number of processes allowed in the system - the kern.maxproc
variable:
int mib[2], maxproc; size_t len;mib[0] = CTL_KERN; mib[1] = KERN_MAXPROC; len = sizeof(maxproc); sysctl(mib, 2, &maxproc, &len, NULL, 0);
user.cs_path
:int mib[2]; size_t len; char *p;mib[0] = CTL_USER; mib[1] = USER_CS_PATH; sysctl(mib, 2, NULL, &len, NULL, 0); p = malloc(len); sysctl(mib, 2, p, &len, NULL, 0);
Name | Description |
CTL_QUERY | Retrieve a mapping of names to numbers below a given node |
CTL_CREATE | Create a new node |
CTL_CREATESYM | Create a new node by its kernel symbol |
CTL_DESTROY | Destroy a node |
CTL_DESCRIBE | Retrieve node descriptions |
The core interface to all of these meta-functions is the structure that the kernel uses to describe the tree internally, as defined in <sys/sysctl.h> as:
struct sysctlnode { uint32_t sysctl_flags; /* flags and type */ int32_t sysctl_num; /* mib number */ char sysctl_name[SYSCTL_NAMELEN]; /* node name */ uint32_t sysctl_ver; /* node's version vs. rest of tree */ uint32_t __rsvd; union { struct { uint32_t suc_csize; /* size of child node array */ uint32_t suc_clen; /* number of valid children */ struct sysctlnode* suc_child; /* array of child nodes */ } scu_child; struct { void *sud_data; /* pointer to external data */ size_t sud_offset; /* offset to data */ } scu_data; int32_t scu_alias; /* node this node refers to */ int32_t scu_idata; /* immediate "int" data */ u_quad_t scu_qdata; /* immediate "u_quad_t" data */ } sysctl_un; size_t _sysctl_size; /* size of instrumented data */ sysctlfn _sysctl_func; /* access helper function */ struct sysctlnode *sysctl_parent; /* parent of this node */ const char *sysctl_desc; /* description of node */ }; #define sysctl_csize sysctl_un.scu_child.suc_csize #define sysctl_clen sysctl_un.scu_child.suc_clen #define sysctl_child sysctl_un.scu_child.suc_child #define sysctl_data sysctl_un.scu_data.sud_data #define sysctl_offset sysctl_un.scu_data.sud_offset #define sysctl_alias sysctl_un.scu_alias #define sysctl_idata sysctl_un.scu_idata #define sysctl_qdata sysctl_un.scu_qdata
Querying the tree to discover the name to number mapping permits dynamic discovery of all the data that the tree currently has instrumented. For example, to discover all the nodes below the CTL_VFS node:
struct sysctlnode query, vfs[128]; int mib[2]; size_t len;mib[0] = CTL_VFS; mib[1] = CTL_QUERY; memset(&query, 0, sizeof(query)); query.sysctl_flags = SYSCTL_VERSION; len = sizeof(vfs); sysctl(mib, 2, &vfs[0], &len, &query, sizeof(query));
Note that a reference to an empty node with sysctl_flags set to SYSCTL_VERSION is passed to sysctl in order to indicate the version that the program is using. All dynamic operations passing nodes into sysctl require that the version be explicitly specified.
Creation and destruction of nodes works by constructing part of a new node description (or a description of the existing node) and invoking CTL_CREATE (or CTL_CREATESYM) or CTL_DESTROY at the parent of the new node, with a pointer to the new node passed via the new and newlen arguments. If valid values for old and oldlenp are passed, a copy of the new node once in the tree will be returned. If the create operation fails because a node with the same name or MIB number exists, a copy of the conflicting node will be returned.
The minimum requirements for creating a node are setting the sysctl_flags to indicate the new node's type, sysctl_num to either the new node's number (or CTL_CREATE or CTL_CREATESYM if a dynamically allocated MIB number is acceptable), sysctl_size to the size of the data to be instrumented (which must agree with the given type), and sysctl_name must be set to the new node's name. Nodes that are not of type “node” must also have some description of the data to be instrumented, which will vary depending on what is to be instrumented.
If existing kernel data is to be covered by this new node, its address should be given in sysctl_data or, if CTL_CREATESYM is used, sysctl_data should be set to a string containing its name from the kernel's symbol table. If new data is to be instrumented and an initial value is available, the new integer or quad type data should be placed into either sysctl_idata or sysctl_qdata, respectively, along with the SYSCTL_IMMEDIATE flag being set, or sysctl_data should be set to point to a copy of the new data, and the SYSCTL_OWNDATA flag must be set. This latter method is the only way that new string and struct type nodes can be initialized. Invalid kernel addresses are accepted, but any attempt to access those nodes will return an error.
The sysctl_csize, sysctl_clen, sysctl_child, sysctl_parent, and sysctl_alias members are used by the kernel to link the tree together and must be NULL or 0. Nodes created in this manner cannot have helper functions, so sysctl_func must also be NULL. If the sysctl_ver member is non-zero, it must match either the version of the parent or the version at the root of the MIB or an error is returned. This can be used to ensure that nodes are only added or removed from a known state of the tree. Note: It may not be possible to determine the version at the root of the tree.
This example creates a new subtree and adds a node to it that controls the audiodebug kernel variable, thereby making it tunable at at any time, without needing to use ddb(4) or kvm(3) to alter the kernel's memory directly.
struct sysctlnode node; int mib[2]; size_t len;mib[0] = CTL_CREATE; /* create at top-level */ len = sizeof(node); memset(&node, 0, len); node.sysctl_flags = SYSCTL_VERSION|CTLFLAG_READWRITE|CTLTYPE_NODE; snprintf(node.sysctl_name, sizeof(node.sysctl_name), "local"); node.sysctl_num = CTL_CREATE; /* request dynamic MIB number */ sysctl(&mib[0], 1, &node, &len, &node, len);mib[0] = node.sysctl_num; /* use new MIB number */ mib[1] = CTL_CREATESYM; /* create at second level */ len = sizeof(node); memset(&node, 0, len); node.sysctl_flags = SYSCTL_VERSION|CTLFLAG_READWRITE|CTLTYPE_INT; snprintf(node.sysctl_name, sizeof(node.sysctl_name), "audiodebug"); node.sysctl_num = CTL_CREATE; node.sysctl_data = "audiodebug"; /* kernel symbol to be used */ sysctl(&mib[0], 2, NULL, NULL, &node, len);
The process for deleting nodes is similar, but less data needs to be supplied. Only the sysctl_num field needs to be filled in; almost all other fields must be left blank. The sysctl_name and/or sysctl_ver fields can be filled in with the name and version of the existing node as additional checks on what will be deleted. If all the given data fail to match any node, nothing will be deleted. If valid values for old and oldlenp are supplied and a node is deleted, a copy of what was in the MIB tree will be returned.
This sample code shows the deletion of the two nodes created in the above example:
int mib[2];len = sizeof(node); memset(&node, 0, len); node.sysctl_flags = SYSCTL_VERSION;mib[0] = 3214; /* assumed number for "local" */ mib[1] = CTL_DESTROY; node.sysctl_num = 3215; /* assumed number for "audiodebug" */ sysctl(&mib[0], 2, NULL, NULL, &node, len);mib[0] = CTL_DESTROY; node.sysctl_num = 3214; /* now deleting "local" */ sysctl(&mib[0], 1, NULL, NULL, &node, len);
Descriptions of each of the nodes can also be retrieved, if they are available. Descriptions can be retrieved in bulk at each level or on a per-node basis. The layout of the buffer into which the descriptions are returned is a series of variable length structures, each of which describes its own size. The length indicated includes the terminating ‘nul' character. Nodes that have no description or where the description is not available are indicated by an empty string. The descr_ver will match the sysctl_ver value for a given node, so that descriptions for nodes whose number have been recycled can be detected and ignored or discarded.
struct sysctldesc { int32_t descr_num; /* mib number of node */ uint32_t descr_ver; /* version of node */ uint32_t descr_len; /* length of description string */ char descr_str[1]; /* not really 1...see above */ };
The NEXT_DESCR() macro can be used to skip to the next description in the retrieved list.
struct sysctlnode desc; struct sysctldesc *d; char buf[1024]; int mib[2]; size_t len;/* retrieve kern-level descriptions */ mib[0] = CTL_KERN; mib[1] = CTL_DESCRIBE; d = (struct sysctldesc *)&buf[0]; len = sizeof(buf); sysctl(mib, 2, d, &len, NULL, 0); while ((caddr_t)d < (caddr_t)&buf[len]) { printf("node %d: %.*s\n", d->descr_num, d->descr_len, d->descr_str); d = NEXT_DESCR(d); }/* retrieve description for kern.securelevel */ memset(&desc, 0, sizeof(desc)); desc.sysctl_flags = SYSCTL_VERSION; desc.sysctl_num = KERN_SECURELEVEL; d = (struct sysctldesc *)&buf[0]; len = sizeof(buf); sysctl(mib, 2, d, &len, &desc, sizeof(desc)); printf("kern.securelevel: %.*s\n", d->descr_len, d->descr_str);
Descriptions can also be set as follows, subject to the following rules:
struct sysctlnode desc; int mib[2];/* presuming the given top-level node was just added... */ mib[0] = 3214; /* mib numbers taken from previous examples */ mib[1] = CTL_DESCRIBE; memset(&desc, 0, sizeof(desc)); desc.sysctl_flags = SYSCTL_VERSION; desc.sysctl_num = 3215; desc.sysctl_desc = "audio debug control knob"; sysctl(mib, 2, NULL, NULL, &desc, sizeof(desc));
Upon successfully setting a description, the new description will be returned in the space indicated by the oldp and oldlenp arguments.
The sysctl_flags field in the struct sysctlnode contains the sysctl version, node type information, and a number of flags. The macros SYSCTL_VERS(), SYSCTL_TYPE(), and SYSCTL_FLAGS() can be used to access the different fields. Valid flags are:
Name | Description |
CTLFLAG_READONLY | Node is read-only |
CTLFLAG_READWRITE | Node is writable by the superuser |
CTLFLAG_ANYWRITE | Node is writable by anyone |
CTLFLAG_PRIVATE | Node is readable only by the superuser |
CTLFLAG_PERMANENT | Node cannot be removed (cannot be set by processes) |
CTLFLAG_OWNDATA | Node owns data and does not instrument existing data |
CTLFLAG_IMMEDIATE | Node contains instrumented data and does not instrument existing data |
CTLFLAG_HEX | Node's contents should be displayed in a hexadecimal form |
CTLFLAG_ROOT | Node is the root of a tree (cannot be set at any time) |
CTLFLAG_ANYNUMBER | Node matches any MIB number (cannot be set by processes) |
CTLFLAG_HIDDEN | Node not displayed by default |
CTLFLAG_ALIAS | Node refers to a sibling node (cannot be set by processes) |
CTLFLAG_OWNDESC | Node owns its own description string space |
September 26, 2009 | NetBSD 6.1 |