xref: /linux/kernel/module/Kconfig (revision 6f81a446f86106c68630032e114024ec7a557077)
1# SPDX-License-Identifier: GPL-2.0-only
2menuconfig MODULES
3	bool "Enable loadable module support"
4	modules
5	select EXECMEM
6	help
7	  Kernel modules are small pieces of compiled code which can
8	  be inserted in the running kernel, rather than being
9	  permanently built into the kernel.  You use the "modprobe"
10	  tool to add (and sometimes remove) them.  If you say Y here,
11	  many parts of the kernel can be built as modules (by
12	  answering M instead of Y where indicated): this is most
13	  useful for infrequently used options which are not required
14	  for booting.  For more information, see the man pages for
15	  modprobe, lsmod, modinfo, insmod and rmmod.
16
17	  If you say Y here, you will need to run "make
18	  modules_install" to put the modules under /lib/modules/
19	  where modprobe can find them (you may need to be root to do
20	  this).
21
22	  If unsure, say Y.
23
24if MODULES
25
26config MODULE_DEBUGFS
27	bool
28
29config MODULE_DEBUG
30	bool "Module debugging"
31	depends on DEBUG_FS
32	help
33	  Allows you to enable / disable features which can help you debug
34	  modules. You don't need these options on production systems.
35
36if MODULE_DEBUG
37
38config MODULE_STATS
39	bool "Module statistics"
40	depends on DEBUG_FS
41	select MODULE_DEBUGFS
42	help
43	  This option allows you to maintain a record of module statistics.
44	  For example, size of all modules, average size, text size, a list
45	  of failed modules and the size for each of those. For failed
46	  modules we keep track of modules which failed due to either the
47	  existing module taking too long to load or that module was already
48	  loaded.
49
50	  You should enable this if you are debugging production loads
51	  and want to see if userspace or the kernel is doing stupid things
52	  with loading modules when it shouldn't or if you want to help
53	  optimize userspace / kernel space module autoloading schemes.
54	  You might want to do this because failed modules tend to use
55	  up significant amount of memory, and so you'd be doing everyone a
56	  favor in avoiding these failures proactively.
57
58	  This functionality is also useful for those experimenting with
59	  module .text ELF section optimization.
60
61	  If unsure, say N.
62
63config MODULE_DEBUG_AUTOLOAD_DUPS
64	bool "Debug duplicate modules with auto-loading"
65	help
66	  Module autoloading allows in-kernel code to request modules through
67	  the *request_module*() API calls. This in turn just calls userspace
68	  modprobe. Although modprobe checks to see if a module is already
69	  loaded before trying to load a module there is a small time window in
70	  which multiple duplicate requests can end up in userspace and multiple
71	  modprobe calls race calling finit_module() around the same time for
72	  duplicate modules. The finit_module() system call can consume in the
73	  worst case more than twice the respective module size in virtual
74	  memory for each duplicate module requests. Although duplicate module
75	  requests are non-fatal virtual memory is a limited resource and each
76	  duplicate module request ends up just unnecessarily straining virtual
77	  memory.
78
79	  This debugging facility will create pr_warn() splats for duplicate
80	  module requests to help identify if module auto-loading may be the
81	  culprit to your early boot virtual memory pressure. Since virtual
82	  memory abuse caused by duplicate module requests could render a
83	  system unusable this functionality will also converge races in
84	  requests for the same module to a single request. You can boot with
85	  the module.enable_dups_trace=1 kernel parameter to use WARN_ON()
86	  instead of the pr_warn().
87
88	  If the first module request used request_module_nowait() we cannot
89	  use that as the anchor to wait for duplicate module requests, since
90	  users of request_module() do want a proper return value. If a call
91	  for the same module happened earlier with request_module() though,
92	  then a duplicate request_module_nowait() would be detected. The
93	  non-wait request_module() call is synchronous and waits until modprobe
94	  completes. Subsequent auto-loading requests for the same module do
95	  not trigger a new finit_module() calls and do not strain virtual
96	  memory, and so as soon as modprobe successfully completes we remove
97	  tracking for duplicates for that module.
98
99	  Enable this functionality to try to debug virtual memory abuse during
100	  boot on systems which are failing to boot or if you suspect you may be
101	  straining virtual memory during boot, and you want to identify if the
102	  abuse was due to module auto-loading. These issues are currently only
103	  known to occur on systems with many CPUs (over 400) and is likely the
104	  result of udev issuing duplicate module requests for each CPU, and so
105	  module auto-loading is not the culprit. There may very well still be
106	  many duplicate module auto-loading requests which could be optimized
107	  for and this debugging facility can be used to help identify them.
108
109	  Only enable this for debugging system functionality, never have it
110	  enabled on real systems.
111
112config MODULE_DEBUG_AUTOLOAD_DUPS_TRACE
113	bool "Force full stack trace when duplicates are found"
114	depends on MODULE_DEBUG_AUTOLOAD_DUPS
115	help
116	  Enabling this will force a full stack trace for duplicate module
117	  auto-loading requests using WARN_ON() instead of pr_warn(). You
118	  should keep this disabled at all times unless you are a developer
119	  and are doing a manual inspection and want to debug exactly why
120	  these duplicates occur.
121
122endif # MODULE_DEBUG
123
124config MODULE_FORCE_LOAD
125	bool "Forced module loading"
126	default n
127	help
128	  Allow loading of modules without version information (ie. modprobe
129	  --force).  Forced module loading sets the 'F' (forced) taint flag and
130	  is usually a really bad idea.
131
132config MODULE_UNLOAD
133	bool "Module unloading"
134	help
135	  Without this option you will not be able to unload any
136	  modules (note that some modules may not be unloadable
137	  anyway), which makes your kernel smaller, faster
138	  and simpler.  If unsure, say Y.
139
140config MODULE_FORCE_UNLOAD
141	bool "Forced module unloading"
142	depends on MODULE_UNLOAD
143	help
144	  This option allows you to force a module to unload, even if the
145	  kernel believes it is unsafe: the kernel will remove the module
146	  without waiting for anyone to stop using it (using the -f option to
147	  rmmod).  This is mainly for kernel developers and desperate users.
148	  If unsure, say N.
149
150config MODULE_UNLOAD_TAINT_TRACKING
151	bool "Tainted module unload tracking"
152	depends on MODULE_UNLOAD
153	select MODULE_DEBUGFS
154	help
155	  This option allows you to maintain a record of each unloaded
156	  module that tainted the kernel. In addition to displaying a
157	  list of linked (or loaded) modules e.g. on detection of a bad
158	  page (see bad_page()), the aforementioned details are also
159	  shown. If unsure, say N.
160
161config MODVERSIONS
162	bool "Module versioning support"
163	depends on !COMPILE_TEST
164	help
165	  Usually, you have to use modules compiled with your kernel.
166	  Saying Y here makes it sometimes possible to use modules
167	  compiled for different kernels, by adding enough information
168	  to the modules to (hopefully) spot any changes which would
169	  make them incompatible with the kernel you are running.  If
170	  unsure, say N.
171
172config ASM_MODVERSIONS
173	bool
174	default HAVE_ASM_MODVERSIONS && MODVERSIONS
175	help
176	  This enables module versioning for exported symbols also from
177	  assembly. This can be enabled only when the target architecture
178	  supports it.
179
180config MODULE_SRCVERSION_ALL
181	bool "Source checksum for all modules"
182	help
183	  Modules which contain a MODULE_VERSION get an extra "srcversion"
184	  field inserted into their modinfo section, which contains a
185	  sum of the source files which made it.  This helps maintainers
186	  see exactly which source was used to build a module (since
187	  others sometimes change the module source without updating
188	  the version).  With this option, such a "srcversion" field
189	  will be created for all modules.  If unsure, say N.
190
191config MODULE_SIG
192	bool "Module signature verification"
193	select MODULE_SIG_FORMAT
194	help
195	  Check modules for valid signatures upon load: the signature
196	  is simply appended to the module. For more information see
197	  <file:Documentation/admin-guide/module-signing.rst>.
198
199	  Note that this option adds the OpenSSL development packages as a
200	  kernel build dependency so that the signing tool can use its crypto
201	  library.
202
203	  You should enable this option if you wish to use either
204	  CONFIG_SECURITY_LOCKDOWN_LSM or lockdown functionality imposed via
205	  another LSM - otherwise unsigned modules will be loadable regardless
206	  of the lockdown policy.
207
208	  !!!WARNING!!!  If you enable this option, you MUST make sure that the
209	  module DOES NOT get stripped after being signed.  This includes the
210	  debuginfo strip done by some packagers (such as rpmbuild) and
211	  inclusion into an initramfs that wants the module size reduced.
212
213config MODULE_SIG_FORCE
214	bool "Require modules to be validly signed"
215	depends on MODULE_SIG
216	help
217	  Reject unsigned modules or signed modules for which we don't have a
218	  key.  Without this, such modules will simply taint the kernel.
219
220config MODULE_SIG_ALL
221	bool "Automatically sign all modules"
222	default y
223	depends on MODULE_SIG || IMA_APPRAISE_MODSIG
224	help
225	  Sign all modules during make modules_install. Without this option,
226	  modules must be signed manually, using the scripts/sign-file tool.
227
228comment "Do not forget to sign required modules with scripts/sign-file"
229	depends on MODULE_SIG_FORCE && !MODULE_SIG_ALL
230
231choice
232	prompt "Hash algorithm to sign modules"
233	depends on MODULE_SIG || IMA_APPRAISE_MODSIG
234	help
235	  This determines which sort of hashing algorithm will be used during
236	  signature generation.  This algorithm _must_ be built into the kernel
237	  directly so that signature verification can take place.  It is not
238	  possible to load a signed module containing the algorithm to check
239	  the signature on that module.
240
241config MODULE_SIG_SHA1
242	bool "SHA-1"
243	select CRYPTO_SHA1
244
245config MODULE_SIG_SHA256
246	bool "SHA-256"
247	select CRYPTO_SHA256
248
249config MODULE_SIG_SHA384
250	bool "SHA-384"
251	select CRYPTO_SHA512
252
253config MODULE_SIG_SHA512
254	bool "SHA-512"
255	select CRYPTO_SHA512
256
257config MODULE_SIG_SHA3_256
258	bool "SHA3-256"
259	select CRYPTO_SHA3
260
261config MODULE_SIG_SHA3_384
262	bool "SHA3-384"
263	select CRYPTO_SHA3
264
265config MODULE_SIG_SHA3_512
266	bool "SHA3-512"
267	select CRYPTO_SHA3
268
269endchoice
270
271config MODULE_SIG_HASH
272	string
273	depends on MODULE_SIG || IMA_APPRAISE_MODSIG
274	default "sha1" if MODULE_SIG_SHA1
275	default "sha256" if MODULE_SIG_SHA256
276	default "sha384" if MODULE_SIG_SHA384
277	default "sha512" if MODULE_SIG_SHA512
278	default "sha3-256" if MODULE_SIG_SHA3_256
279	default "sha3-384" if MODULE_SIG_SHA3_384
280	default "sha3-512" if MODULE_SIG_SHA3_512
281
282config MODULE_COMPRESS
283	bool "Module compression"
284	help
285	  Enable module compression to reduce on-disk size of module binaries.
286	  This is fully compatible with signed modules.
287
288	  The tool used to work with modules needs to support the selected
289	  compression type. kmod MAY support gzip, xz and zstd. Other tools
290	  might have a limited selection of the supported types.
291
292	  Note that for modules inside an initrd or initramfs, it's more
293	  efficient to compress the whole ramdisk instead.
294
295	  If unsure, say N.
296
297choice
298	prompt "Module compression type"
299	depends on MODULE_COMPRESS
300	help
301	  Choose the supported algorithm for module compression.
302
303config MODULE_COMPRESS_GZIP
304	bool "GZIP"
305	help
306	  Support modules compressed with GZIP. The installed modules are
307	  suffixed with .ko.gz.
308
309config MODULE_COMPRESS_XZ
310	bool "XZ"
311	help
312	  Support modules compressed with XZ. The installed modules are
313	  suffixed with .ko.xz.
314
315config MODULE_COMPRESS_ZSTD
316	bool "ZSTD"
317	help
318	  Support modules compressed with ZSTD. The installed modules are
319	  suffixed with .ko.zst.
320
321endchoice
322
323config MODULE_COMPRESS_ALL
324	bool "Automatically compress all modules"
325	default y
326	depends on MODULE_COMPRESS
327	help
328	  Compress all modules during 'make modules_install'.
329
330	  Your build system needs to provide the appropriate compression tool
331	  for the selected compression type. External modules will also be
332	  compressed in the same way during the installation.
333
334config MODULE_DECOMPRESS
335	bool "Support in-kernel module decompression"
336	depends on MODULE_COMPRESS
337	select ZLIB_INFLATE if MODULE_COMPRESS_GZIP
338	select XZ_DEC if MODULE_COMPRESS_XZ
339	select ZSTD_DECOMPRESS if MODULE_COMPRESS_ZSTD
340	help
341	  Support for decompressing kernel modules by the kernel itself
342	  instead of relying on userspace to perform this task. Useful when
343	  load pinning security policy is enabled.
344
345	  If unsure, say N.
346
347config MODULE_ALLOW_MISSING_NAMESPACE_IMPORTS
348	bool "Allow loading of modules with missing namespace imports"
349	help
350	  Symbols exported with EXPORT_SYMBOL_NS*() are considered exported in
351	  a namespace. A module that makes use of a symbol exported with such a
352	  namespace is required to import the namespace via MODULE_IMPORT_NS().
353	  There is no technical reason to enforce correct namespace imports,
354	  but it creates consistency between symbols defining namespaces and
355	  users importing namespaces they make use of. This option relaxes this
356	  requirement and lifts the enforcement when loading a module.
357
358	  If unsure, say N.
359
360config MODPROBE_PATH
361	string "Path to modprobe binary"
362	default "/sbin/modprobe"
363	help
364	  When kernel code requests a module, it does so by calling
365	  the "modprobe" userspace utility. This option allows you to
366	  set the path where that binary is found. This can be changed
367	  at runtime via the sysctl file
368	  /proc/sys/kernel/modprobe. Setting this to the empty string
369	  removes the kernel's ability to request modules (but
370	  userspace can still load modules explicitly).
371
372config TRIM_UNUSED_KSYMS
373	bool "Trim unused exported kernel symbols"
374	help
375	  The kernel and some modules make many symbols available for
376	  other modules to use via EXPORT_SYMBOL() and variants. Depending
377	  on the set of modules being selected in your kernel configuration,
378	  many of those exported symbols might never be used.
379
380	  This option allows for unused exported symbols to be dropped from
381	  the build. In turn, this provides the compiler more opportunities
382	  (especially when using LTO) for optimizing the code and reducing
383	  binary size.  This might have some security advantages as well.
384
385	  If unsure, or if you need to build out-of-tree modules, say N.
386
387config UNUSED_KSYMS_WHITELIST
388	string "Whitelist of symbols to keep in ksymtab"
389	depends on TRIM_UNUSED_KSYMS
390	help
391	  By default, all unused exported symbols will be un-exported from the
392	  build when TRIM_UNUSED_KSYMS is selected.
393
394	  UNUSED_KSYMS_WHITELIST allows to whitelist symbols that must be kept
395	  exported at all times, even in absence of in-tree users. The value to
396	  set here is the path to a text file containing the list of symbols,
397	  one per line. The path can be absolute, or relative to the kernel
398	  source or obj tree.
399
400config MODULES_TREE_LOOKUP
401	def_bool y
402	depends on PERF_EVENTS || TRACING || CFI_CLANG
403
404endif # MODULES
405