.\" Copyright (c) 1996 Free Software Foundation, Inc. .\" This program is distributed according to the Gnu General Public License. .\" See the file COPYING in the kernel source directory. .\" .TH INSMOD 8 "January 30, 2002" Linux "Linux Module Support" .SH NAME insmod \- install loadable kernel module .hy 0 .SH SYNOPSIS .B insmod [\-fhkLmnpqrsSvVxXyYN] [\-e\ \fIpersist_name\fR] [\-o\ \fImodule_name\fR] [\-O\ \fIblob_name\fR] [\-P\ \fIprefix\fR] \fImodule\fR [\ \fBsymbol\fR=\fIvalue\fR\ ...\ ] .SH DESCRIPTION .B insmod installs a loadable module in the running kernel. .PP .B insmod tries to link a module into the running kernel by resolving all symbols from the kernel's exported symbol table. .PP If the module file name is given without directories or extension, .B insmod will search for the module in some common default directories. The environment variable \fBMODPATH\fR can be used to override this default. If a module configuration file such as \fI/etc/modules.conf\fR exists, it will override the paths defined in \fBMODPATH\fR. .PP The environment variable \fBMODULECONF\fR can also be used to select a different configuration file from the default \fI/etc/modules.conf\fR (or \fI/etc/conf.modules\fR (deprecated)). This environment variable will override all the definitions above. .PP When environment variable .B UNAME_MACHINE is set, modutils will use its value instead of the machine field from the uname() syscall. This is mainly of use when you are compiling 64 bit modules in 32 bit user space or vice versa, set .B UNAME_MACHINE to the type of the modules. Current modutils does not support full cross build mode for modules, it is limited to choosing between 32 and 64 bit versions of the host architecture. .SS OPTIONS .TP \fB\-e\fR \fIpersist_name\fR, \fB\-\-persist\fR=\fIpersist_name\fR Specifies where any persistent data for the module is read from on load and written to when this instantiantion of the module is unloaded. This option is silently ignored if the module has no persistent data. Persistent data is only read by \fBinsmod\fR if this option is present, by default \fBinsmod\fR does not process persistent data. .TP .B "" As a shorthand form, \fB\-e\ ""\fR\ (an empty string) is interpreted by \fBinsmod\fR as the value of \fBpersistdir\fR as defined in \fImodules.conf\fR, followed by the filename of the module relative to the module search path it was found in, minus any trailing ".gz", ".o" or ".mod". If \fBmodules.conf\fR specifies "\fBpersistdir\ =\fR" (i.e. \fBpersistdir\fR is an empty field) then this shorthand form is silently ignored. (See \fBmodules.conf\fR (5).) .TP .BR \-f ", " \-\-force Attempt load the module even if the version of the running kernel and the version of the kernel for which the module was compiled do not match. This only overrides the kernel version check, it has no effect on symbol name checks. If the symbol names in the module do not match the kernel then there is no way to force \fBinsmod\fR to load the module. .TP .BR \-h ", " \-\-help Display a summary of options and immediately exit. .TP .BR \-k ", " \-\-autoclean Set the auto-clean flag on the module. This flag will be used by \fBkerneld\fR(8) to remove modules that have not been used in some period of time \(em usually one minute. .TP .BR \-L ", " \-\-lock Use \fBflock\fR(2) to prevent simultaneous loads of the same module. .TP .BR \-m ", " \-\-map Output a load map on stdout, making it easier to debug the module in the event of a kernel panic. .TP .BR \-n ", " \-\-noload Dummy run, do everything except load the module into the kernel. If requested by an \fB\-m\fR or \fB\-O\fR, the run will produce a map or blob file. Since the module is not loaded, the real kernel load address is unknown so the map and blob file are based on an arbitrary load address of 0x12340000. .TP \fB\-o\fR \fImodule_name\fR, \fB\-\-name\fR=\fImodule_name\fR Explicitly name the module, rather than deriving the name from the base name of the source object file. .TP \fB\-O\fR \fIblob_name\fR, \fB\-\-blob\fR=\fIblob_name\fR Save the binary object in \fIblob_name\fR. The result is a binary blob (no ELF headers) showing exactly what is loaded into the kernel after section manipulation and relocation. Option \fB\-m\fR is recommended to get a map of the object. .TP .BR \-p ", " \-\-probe Probe the module to see if it could be successfully loaded. This includes locating the object file in the module path, checking version numbers, and resolving symbols. It does not check the relocations nor does it produce a map or blob file. .TP \fB\-P\fR \fIprefix\fR, \fB\-\-prefix\fR=\fIprefix\fR This option can be used with versioned modules for an SMP or bigmem kernel, since such modules have an extra prefix added in their symbol names. If the kernel was built with symbol versions then \fBinsmod\fR will automatically extract the prefix from the definition of "get_module_symbol" or "inter_module_get", one of which must exist in any kernel that supports modules. If the kernel has no symbol versions but the module was built with symbol versions then the user must supply .BR \-P . .TP .BR \-q ", " \-\-quiet Do not print a list of any unresolved symbols. Do not complain about version mismatch. The problem will only be reflected in the exit status of \fBinsmod\fR. .TP .BR \-r ", " \-\-root Some users compile modules under a non-root userid then install the modules as root. This process can leave the modules owned by the non-root userid, even though the modules directory is owned by root. If the non-root userid is compromised, an intruder can overwrite existing modules owned by that userid and use this exposure to bootstrap up to root access. .TP .B "" By default, modutils will reject attempts to use a module that is not owned by root. Specifying -r will toggle the check and allow root to load modules that are not owned by root. \fBNote:\fR the default value for root check can be changed when modutils is configured. .TP .B "" \fBUse of \-r to disable root checking or setting the default to "no root check" at configuration time is a major security exposure and is not recommended.\fR .TP .BR \-s ", " \-\-syslog Output everything to \fBsyslog\fR(3) instead of the terminal. .TP .BR \-S ", " \-\-kallsyms Force the loaded module to have \fBkallsyms\fR data, even if the kernel does not support it. This option is for small systems where the kernel is loaded without \fBkallsyms\fR data but selected modules need \fBkallsyms\fR for debugging. .TP .BR \-v ", " \-\-verbose Be verbose. .TP .BR \-V ", " \-\-version Display the version of \fBinsmod\fR. .TP .BR \-X ", " \-\-export "; " \-x ", " \-\-noexport Do and do not export all of the module's external symbols, respectively. The default is for the symbols to be exported. This option is only effective if the module does not explicitly export its own controlled symbol table, and thus is deprecated. .TP .BR \-Y ", " \-\-ksymoops "; " \-y ", " \-\-noksymoops Do and do not add \fBksymoops\fR symbols to ksyms. These symbols are used by \fBksymoops\fR to provide better debugging if there is an Oops in this module. The default is for the \fBksymoops\fR symbols to be defined. This option is independent of the \fB\-X\fR/\fB\-x\fR options. .TP .B "" \fBksymoops\fR symbols add approximately 260 bytes per loaded module. Unless you are really short on kernel space and are trying to reduce ksyms to its minimum size, take the default and get more accurate Oops debugging. \fBksymoops\fR symbols are required to save persistent module data. .TP .BR \-N ", " \-\-numeric-only Only check the numeric part of the module version against the kernel version, i.e. ignore EXTRAVERSION when deciding if a module belongs to a kernel. This flag is automatically set for kernel 2.5 onwards, it is optional for earlier kernels. .SS MODULE PARAMETERS Some modules accept load-time parameters to customize their operation. These parameters are often I/O port and IRQ numbers that vary from machine to machine and cannot be determined from the hardware. .PP In modules built for 2.0 series kernels, any integer or character pointer symbol may be treated as a parameter and modified. Beginning in the 2.1 series kernels, symbols are explicitly marked as parameters so that only specific values may be changed. Furthermore type information is provided for checking the values provided at load time. .PP In the case of integers, all values may be in decimal, octal or hexadecimal a la C: 17, 021 or 0x11. Array elements are specified sequence separated by commas. Elements can be skipped by omitting the value. .PP In 2.0 series modules, values that do not begin with a number are considered strings. Beginning in 2.1, the parameter's type information indicates whether to interpret the value as a string. If the value begins with double-quotes (\fB"\fR), .\" (syntax highlighting kludge: This comment closes previous quote. the string is interpreted as in C, escape sequences and all. Do note that from the shell prompt, the quotes themselves may need to be protected from shell interpretation. .SS GPL LICENSED MODULES AND SYMBOLS Starting with kernel 2.4.10, modules should have a license string, defined using \fBMODULE_LICENSE()\fR. Several strings are recognised as being GPL compatible; any other license string or no license at all means that the module is treated as proprietary. See \fIinclude/linux/module.h\fR for a list of GPL compatible license strings. .PP If the kernel supports the \fI/proc/sys/kernel/tainted\fR flag then \fBinsmod\fR will OR the tainted flag with '1' when loading a module without a GPL license. A warning is issued if the kernel supports tainting and a module is loaded without a license. A warning is always issued for modules which have a \fBMODULE_LICENSE()\fR that is not GPL compatible, even on older kernels that do not support tainting. This minimizes warnings when a new modutils is used on older kernels. .PP \fBinsmod\ \-f\fR (force) mode will OR the tainted flag with '2' on kernels that support tainting. It always issues a warning. .PP Some kernel developers require that symbols exported by their code must only be used by modules with a GPL compatible license. These symbols are exported by \fBEXPORT_SYMBOL_GPL\fR instead of the normal \fBEXPORT_SYMBOL\fR. GPL-only symbols exported by the kernel and by other modules are only visible to modules with a GPL-compatible license, these symbols appear in \fI/proc/ksyms\fR with a prefix of '\fBGPLONLY_\fR'. \fBinsmod\fR ignores the \fBGPLONLY_\fR prefix on symbols while loading a GPL licensed module so the module just refers to the normal symbol name, without the prefix. GPL only symbols are not made available to modules without a GPL compatible license, this includes modules with no license at all. .SS KSYMOOPS ASSISTANCE To assist with debugging of kernel Oops when using modules, \fBinsmod\fR defaults to adding some symbols to ksyms, see the \fB\-Y\fR option. These symbols start with \fB__insmod_modulename_\fR. The \fImodulename\fR is required to make the symbols unique. It is legal to load the same object more than once under different module names. Currently defined symbols are: .TP .B __insmod_modulename_Oobjectfile_Mmtime_Vversion \fIobjectfile\fR is the name of the file that the object was loaded from. This ensures that ksymoops can match the code to the correct object. \fImtime\fR is the last modified timestamp on that file in hex, zero if stat failed. \fIversion\fR is the kernel version that the module was compiled for, -1 if no version is available. The \fB_O\fR symbol has the same start address as the module header. .TP .B __insmod_modulename_Ssectionname_Llength This symbol appears at the start of selected ELF sections, currently .text, .rodata, .data, .bss and .sbss. It only appears if the section has a non-zero size. \fIsectionname\fR is the name of the ELF section, \fIlength\fR is the length of the section in decimal. These symbols help ksymoops map addresses to sections when no symbols are available. .TP .B __insmod_modulename_Ppersistent_filename Only created by \fBinsmod\fR if the module has one or more parameters that are marked as persistent data and a filename to save persistent data (see \fB\-e\fR, above) is available. .PP The other problem with debugging kernel Oops in modules is that the contents of \fI/proc/ksyms\fR and \fI/proc/modules\fR can change between the Oops and when you process the log file. To help overcome this problem, if the directory \fI/var/log/ksymoops\fR exists then \fBinsmod\fR and \fBrmmod\fR will automatically copy \fI/proc/ksyms\fR and \fI/proc/modules\fR to \fI/var/log/ksymoops\fR with a prefix of `date\ +%Y%m%d%H%M%S`. The system administrator can tell \fBksymoops\fR which snapshot files to use when debugging an Oops. There is no switch to disable this automatic copy. If you do not want it to occur, do not create \fI/var/log/ksymoops\fR. If that directory exists, it should be owned by root and be mode 644 or 600 and you should run this script every day or so. The script below is installed as \fBinsmod_ksymoops_clean\fR. .PP .ne 8 .nf #!/bin/sh # Delete saved ksyms and modules not accessed in 2 days if [ -d /var/log/ksymoops ] then set -e # Make sure there is always at least one version d=`date +%Y%m%d%H%M%S` cp -a /proc/ksyms /var/log/ksymoops/${d}.ksyms cp -a /proc/modules /var/log/ksymoops/${d}.modules find /var/log/ksymoops -type f -atime +2 -exec rm {} \\; fi .SH SEE ALSO \fBrmmod\fP(8), \fBmodprobe\fP(8), \fBdepmod\fP(8), \fBlsmod\fP(8), \fBksyms\fP(8), \fBmodules\fP(2), \fBgenksyms\fP(8), \fBkerneld\fP(8), \fBksymoops\fP(kernel). .SH BUGS \fBinsmod\fR [\fB\-V\fR\ |\ \fB\-\-version\fR] should display version information and then exit immediately. Instead, it prints the version information and behaves as if no options were given. .SH HISTORY Module support was first conceived by Anonymous .br Initial Linux version by Bas Laarhoven .br Version 0.99.14 by Jon Tombs .br Extended by Bjorn Ekwall .br Original ELF help from Eric Youngdale .br Rewritten for 2.1.17 by Richard Henderson .br Extended by Bjorn Ekwall for modutils-2.2.*, March 1999 .br Assistance for ksymoops by Keith Owens , May 1999 .br Maintainer: Keith Owens .