1 .. SPDX-License-Identifier: GPL-2.0
3 ===================================
4 Cache on Already Mounted Filesystem
5 ===================================
15 (*) Starting the cache.
23 (*) Security model and SELinux.
25 (*) A note on security.
27 (*) Statistical information.
37 CacheFiles is a caching backend that's meant to use as a cache a directory on
38 an already mounted filesystem of a local type (such as Ext3).
40 CacheFiles uses a userspace daemon to do some of the cache management - such as
41 reaping stale nodes and culling. This is called cachefilesd and lives in
44 The filesystem and data integrity of the cache are only as good as those of the
45 filesystem providing the backing services. Note that CacheFiles does not
46 attempt to journal anything since the journalling interfaces of the various
47 filesystems are very specific in nature.
49 CacheFiles creates a misc character device - "/dev/cachefiles" - that is used
50 to communication with the daemon. Only one thing may have this open at once,
51 and while it is open, a cache is at least partially in existence. The daemon
52 opens this and sends commands down it to control the cache.
54 CacheFiles is currently limited to a single cache.
56 CacheFiles attempts to maintain at least a certain percentage of free space on
57 the filesystem, shrinking the cache by culling the objects it contains to make
58 space if necessary - see the "Cache Culling" section. This means it can be
59 placed on the same medium as a live set of data, and will expand to make use of
60 spare space and automatically contract when the set of data requires more
68 The use of CacheFiles and its daemon requires the following features to be
69 available in the system and in the cache filesystem:
73 - extended attributes (xattrs).
75 - openat() and friends.
77 - bmap() support on files in the filesystem (FIBMAP ioctl).
79 - The use of bmap() to detect a partial page at the end of the file.
81 It is strongly recommended that the "dir_index" option is enabled on Ext3
82 filesystems being used as a cache.
88 The cache is configured by a script in /etc/cachefilesd.conf. These commands
89 set up cache ready for use. The following script commands are available:
91 brun <N>%, bcull <N>%, bstop <N>%, frun <N>%, fcull <N>%, fstop <N>%
92 Configure the culling limits. Optional. See the section on culling
93 The defaults are 7% (run), 5% (cull) and 1% (stop) respectively.
95 The commands beginning with a 'b' are file space (block) limits, those
96 beginning with an 'f' are file count limits.
99 Specify the directory containing the root of the cache. Mandatory.
102 Specify a tag to FS-Cache to use in distinguishing multiple caches.
103 Optional. The default is "CacheFiles".
106 Specify a numeric bitmask to control debugging in the kernel module.
107 Optional. The default is zero (all off). The following values can be
108 OR'd into the mask to collect various information:
110 == =================================================
111 1 Turn on trace of function entry (_enter() macros)
112 2 Turn on trace of function exit (_leave() macros)
113 4 Turn on trace of internal debug points (_debug())
114 == =================================================
116 This mask can also be set through sysfs, eg::
118 echo 5 >/sys/modules/cachefiles/parameters/debug
124 The cache is started by running the daemon. The daemon opens the cache device,
125 configures the cache and tells it to begin caching. At that point the cache
126 binds to fscache and the cache becomes live.
128 The daemon is run as follows::
130 /sbin/cachefilesd [-d]* [-s] [-n] [-f <configfile>]
135 Increase the debugging level. This can be specified multiple times and
136 is cumulative with itself.
139 Send messages to stderr instead of syslog.
142 Don't daemonise and go into background.
145 Use an alternative configuration file rather than the default one.
151 Do not mount other things within the cache as this will cause problems. The
152 kernel module contains its own very cut-down path walking facility that ignores
153 mountpoints, but the daemon can't avoid them.
155 Do not create, rename or unlink files and directories in the cache while the
156 cache is active, as this may cause the state to become uncertain.
158 Renaming files in the cache might make objects appear to be other objects (the
159 filename is part of the lookup key).
161 Do not change or remove the extended attributes attached to cache files by the
162 cache as this will cause the cache state management to get confused.
164 Do not create files or directories in the cache, lest the cache get confused or
165 serve incorrect data.
167 Do not chmod files in the cache. The module creates things with minimal
168 permissions to prevent random users being able to access them directly.
174 The cache may need culling occasionally to make space. This involves
175 discarding objects from the cache that have been used less recently than
176 anything else. Culling is based on the access time of data objects. Empty
177 directories are culled if not in use.
179 Cache culling is done on the basis of the percentage of blocks and the
180 percentage of files available in the underlying filesystem. There are six
184 If the amount of free space and the number of available files in the cache
185 rises above both these limits, then culling is turned off.
188 If the amount of available space or the number of available files in the
189 cache falls below either of these limits, then culling is started.
192 If the amount of available space or the number of available files in the
193 cache falls below either of these limits, then no further allocation of
194 disk space or files is permitted until culling has raised things above
197 These must be configured thusly::
199 0 <= bstop < bcull < brun < 100
200 0 <= fstop < fcull < frun < 100
202 Note that these are percentages of available space and available files, and do
203 _not_ appear as 100 minus the percentage displayed by the "df" program.
205 The userspace daemon scans the cache to build up a table of cullable objects.
206 These are then culled in least recently used order. A new scan of the cache is
207 started as soon as space is made in the table. Objects will be skipped if
208 their atimes have changed or if the kernel module says it is still using them.
214 The CacheFiles module will create two directories in the directory it was
220 The active cache objects all reside in the first directory. The CacheFiles
221 kernel module moves any retired or culled objects that it can't simply unlink
222 to the graveyard from which the daemon will actually delete them.
224 The daemon uses dnotify to monitor the graveyard directory, and will delete
225 anything that appears therein.
228 The module represents index objects as directories with the filename "I..." or
229 "J...". Note that the "cache/" directory is itself a special index.
231 Data objects are represented as files if they have no children, or directories
232 if they do. Their filenames all begin "D..." or "E...". If represented as a
233 directory, data objects will have a file in the directory called "data" that
234 actually holds the data.
236 Special objects are similar to data objects, except their filenames begin
240 If an object has children, then it will be represented as a directory.
241 Immediately in the representative directory are a collection of directories
242 named for hash values of the child object keys with an '@' prepended. Into
243 this directory, if possible, will be placed the representations of the child
246 /INDEX /INDEX /INDEX /DATA FILES
247 /=========/==========/=================================/================
248 cache/@4a/I03nfs/@30/Ji000000000000000--fHg8hi8400
249 cache/@4a/I03nfs/@30/Ji000000000000000--fHg8hi8400/@75/Es0g000w...DB1ry
250 cache/@4a/I03nfs/@30/Ji000000000000000--fHg8hi8400/@75/Es0g000w...N22ry
251 cache/@4a/I03nfs/@30/Ji000000000000000--fHg8hi8400/@75/Es0g000w...FP1ry
254 If the key is so long that it exceeds NAME_MAX with the decorations added on to
255 it, then it will be cut into pieces, the first few of which will be used to
256 make a nest of directories, and the last one of which will be the objects
257 inside the last directory. The names of the intermediate directories will have
260 J1223/@23/+xy...z/+kl...m/Epqr
263 Note that keys are raw data, and not only may they exceed NAME_MAX in size,
264 they may also contain things like '/' and NUL characters, and so they may not
265 be suitable for turning directly into a filename.
267 To handle this, CacheFiles will use a suitably printable filename directly and
268 "base-64" encode ones that aren't directly suitable. The two versions of
269 object filenames indicate the encoding:
271 =============== =============== ===============
272 OBJECT TYPE PRINTABLE ENCODED
273 =============== =============== ===============
276 Special "S..." "T..."
277 =============== =============== ===============
279 Intermediate directories are always "@" or "+" as appropriate.
282 Each object in the cache has an extended attribute label that holds the object
283 type ID (required to distinguish special objects) and the auxiliary data from
284 the netfs. The latter is used to detect stale objects in the cache and update
288 Note that CacheFiles will erase from the cache any file it doesn't recognise or
289 any file of an incorrect type (such as a FIFO file or a device file).
292 Security Model and SELinux
293 ==========================
295 CacheFiles is implemented to deal properly with the LSM security features of
296 the Linux kernel and the SELinux facility.
298 One of the problems that CacheFiles faces is that it is generally acting on
299 behalf of a process, and running in that process's context, and that includes a
300 security context that is not appropriate for accessing the cache - either
301 because the files in the cache are inaccessible to that process, or because if
302 the process creates a file in the cache, that file may be inaccessible to other
305 The way CacheFiles works is to temporarily change the security context (fsuid,
306 fsgid and actor security label) that the process acts as - without changing the
307 security context of the process when it the target of an operation performed by
308 some other process (so signalling and suchlike still work correctly).
311 When the CacheFiles module is asked to bind to its cache, it:
313 (1) Finds the security label attached to the root cache directory and uses
314 that as the security label with which it will create files. By default,
319 (2) Finds the security label of the process which issued the bind request
320 (presumed to be the cachefilesd daemon), which by default will be::
324 and asks LSM to supply a security ID as which it should act given the
325 daemon's label. By default, this will be::
329 SELinux transitions the daemon's security ID to the module's security ID
330 based on a rule of this form in the policy::
332 type_transition <daemon's-ID> kernel_t : process <module's-ID>;
336 type_transition cachefilesd_t kernel_t : process cachefiles_kernel_t;
339 The module's security ID gives it permission to create, move and remove files
340 and directories in the cache, to find and access directories and files in the
341 cache, to set and access extended attributes on cache objects, and to read and
342 write files in the cache.
344 The daemon's security ID gives it only a very restricted set of permissions: it
345 may scan directories, stat files and erase files and directories. It may
346 not read or write files in the cache, and so it is precluded from accessing the
347 data cached therein; nor is it permitted to create new files in the cache.
350 There are policy source files available in:
352 https://people.redhat.com/~dhowells/fscache/cachefilesd-0.8.tar.bz2
354 and later versions. In that tarball, see the files::
360 They are built and installed directly by the RPM.
362 If a non-RPM based system is being used, then copy the above files to their own
365 make -f /usr/share/selinux/devel/Makefile
366 semodule -i cachefilesd.pp
368 You will need checkpolicy and selinux-policy-devel installed prior to the
372 By default, the cache is located in /var/fscache, but if it is desirable that
373 it should be elsewhere, than either the above policy files must be altered, or
374 an auxiliary policy must be installed to label the alternate location of the
377 For instructions on how to add an auxiliary policy to enable the cache to be
378 located elsewhere when SELinux is in enforcing mode, please see::
380 /usr/share/doc/cachefilesd-*/move-cache.txt
382 When the cachefilesd rpm is installed; alternatively, the document can be found
389 CacheFiles makes use of the split security in the task_struct. It allocates
390 its own task_security structure, and redirects current->cred to point to it
391 when it acts on behalf of another process, in that process's context.
393 The reason it does this is that it calls vfs_mkdir() and suchlike rather than
394 bypassing security and calling inode ops directly. Therefore the VFS and LSM
395 may deny the CacheFiles access to the cache data because under some
396 circumstances the caching code is running in the security context of whatever
397 process issued the original syscall on the netfs.
399 Furthermore, should CacheFiles create a file or directory, the security
400 parameters with that object is created (UID, GID, security label) would be
401 derived from that process that issued the system call, thus potentially
402 preventing other processes from accessing the cache - including CacheFiles's
403 cache management daemon (cachefilesd).
405 What is required is to temporarily override the security of the process that
406 issued the system call. We can't, however, just do an in-place change of the
407 security data as that affects the process as an object, not just as a subject.
408 This means it may lose signals or ptrace events for example, and affects what
409 the process looks like in /proc.
411 So CacheFiles makes use of a logical split in the security between the
412 objective security (task->real_cred) and the subjective security (task->cred).
413 The objective security holds the intrinsic security properties of a process and
414 is never overridden. This is what appears in /proc, and is what is used when a
415 process is the target of an operation by some other process (SIGKILL for
418 The subjective security holds the active security properties of a process, and
419 may be overridden. This is not seen externally, and is used whan a process
420 acts upon another object, for example SIGKILLing another process or opening a
423 LSM hooks exist that allow SELinux (or Smack or whatever) to reject a request
424 for CacheFiles to run in a context of a specific security label, or to create
425 files and directories with another security label.
428 Statistical Information
429 =======================
431 If FS-Cache is compiled with the following option enabled::
433 CONFIG_CACHEFILES_HISTOGRAM=y
435 then it will gather certain statistics and display them through a proc file.
437 /proc/fs/cachefiles/histogram
441 cat /proc/fs/cachefiles/histogram
442 JIFS SECS LOOKUPS MKDIRS CREATES
443 ===== ===== ========= ========= =========
445 This shows the breakdown of the number of times each amount of time
446 between 0 jiffies and HZ-1 jiffies a variety of tasks took to run. The
447 columns are as follows:
449 ======= =======================================================
450 COLUMN TIME MEASUREMENT
451 ======= =======================================================
452 LOOKUPS Length of time to perform a lookup on the backing fs
453 MKDIRS Length of time to perform a mkdir on the backing fs
454 CREATES Length of time to perform a create on the backing fs
455 ======= =======================================================
457 Each row shows the number of events that took a particular range of times.
458 Each step is 1 jiffy in size. The JIFS column indicates the particular
459 jiffy range covered, and the SECS field the equivalent number of seconds.
465 If CONFIG_CACHEFILES_DEBUG is enabled, the CacheFiles facility can have runtime
466 debugging enabled by adjusting the value in::
468 /sys/module/cachefiles/parameters/debug
470 This is a bitmask of debugging streams to enable:
472 ======= ======= =============================== =======================
473 BIT VALUE STREAM POINT
474 ======= ======= =============================== =======================
475 0 1 General Function entry trace
476 1 2 Function exit trace
478 ======= ======= =============================== =======================
480 The appropriate set of values should be OR'd together and the result written to
481 the control file. For example::
483 echo $((1|4|8)) >/sys/module/cachefiles/parameters/debug
485 will turn on all function entry debugging.
491 When working in its original mode, CacheFiles serves as a local cache for a
492 remote networking fs - while in on-demand read mode, CacheFiles can boost the
493 scenario where on-demand read semantics are needed, e.g. container image
496 The essential difference between these two modes is seen when a cache miss
497 occurs: In the original mode, the netfs will fetch the data from the remote
498 server and then write it to the cache file; in on-demand read mode, fetching
499 the data and writing it into the cache is delegated to a user daemon.
501 ``CONFIG_CACHEFILES_ONDEMAND`` should be enabled to support on-demand read mode.
504 Protocol Communication
505 ----------------------
507 The on-demand read mode uses a simple protocol for communication between kernel
508 and user daemon. The protocol can be modeled as::
510 kernel --[request]--> user daemon --[reply]--> kernel
512 CacheFiles will send requests to the user daemon when needed. The user daemon
513 should poll the devnode ('/dev/cachefiles') to check if there's a pending
514 request to be processed. A POLLIN event will be returned when there's a pending
517 The user daemon then reads the devnode to fetch a request to process. It should
518 be noted that each read only gets one request. When it has finished processing
519 the request, the user daemon should write the reply to the devnode.
521 Each request starts with a message header of the form::
523 struct cachefiles_msg {
533 * ``msg_id`` is a unique ID identifying this request among all pending
536 * ``opcode`` indicates the type of this request.
538 * ``object_id`` is a unique ID identifying the cache file operated on.
540 * ``data`` indicates the payload of this request.
542 * ``len`` indicates the whole length of this request, including the
543 header and following type-specific payload.
546 Turning on On-demand Mode
547 -------------------------
549 An optional parameter becomes available to the "bind" command::
553 When the "bind" command is given no argument, it defaults to the original mode.
554 When it is given the "ondemand" argument, i.e. "bind ondemand", on-demand read
555 mode will be enabled.
561 When the netfs opens a cache file for the first time, a request with the
562 CACHEFILES_OP_OPEN opcode, a.k.a an OPEN request will be sent to the user
563 daemon. The payload format is of the form::
565 struct cachefiles_open {
566 __u32 volume_key_size;
567 __u32 cookie_key_size;
575 * ``data`` contains the volume_key followed directly by the cookie_key.
576 The volume key is a NUL-terminated string; the cookie key is binary
579 * ``volume_key_size`` indicates the size of the volume key in bytes.
581 * ``cookie_key_size`` indicates the size of the cookie key in bytes.
583 * ``fd`` indicates an anonymous fd referring to the cache file, through
584 which the user daemon can perform write/llseek file operations on the
588 The user daemon can use the given (volume_key, cookie_key) pair to distinguish
589 the requested cache file. With the given anonymous fd, the user daemon can
590 fetch the data and write it to the cache file in the background, even when
591 kernel has not triggered a cache miss yet.
593 Be noted that each cache file has a unique object_id, while it may have multiple
594 anonymous fds. The user daemon may duplicate anonymous fds from the initial
595 anonymous fd indicated by the @fd field through dup(). Thus each object_id can
596 be mapped to multiple anonymous fds, while the usr daemon itself needs to
597 maintain the mapping.
599 When implementing a user daemon, please be careful of RLIMIT_NOFILE,
600 ``/proc/sys/fs/nr_open`` and ``/proc/sys/fs/file-max``. Typically these needn't
601 be huge since they're related to the number of open device blobs rather than
602 open files of each individual filesystem.
604 The user daemon should reply the OPEN request by issuing a "copen" (complete
605 open) command on the devnode::
607 copen <msg_id>,<cache_size>
611 * ``msg_id`` must match the msg_id field of the OPEN request.
613 * When >= 0, ``cache_size`` indicates the size of the cache file;
614 when < 0, ``cache_size`` indicates any error code encountered by the
621 When a cookie withdrawn, a CLOSE request (opcode CACHEFILES_OP_CLOSE) will be
622 sent to the user daemon. This tells the user daemon to close all anonymous fds
623 associated with the given object_id. The CLOSE request has no extra payload,
624 and shouldn't be replied.
630 When a cache miss is encountered in on-demand read mode, CacheFiles will send a
631 READ request (opcode CACHEFILES_OP_READ) to the user daemon. This tells the user
632 daemon to fetch the contents of the requested file range. The payload is of the
635 struct cachefiles_read {
642 * ``off`` indicates the starting offset of the requested file range.
644 * ``len`` indicates the length of the requested file range.
647 When it receives a READ request, the user daemon should fetch the requested data
648 and write it to the cache file identified by object_id.
650 When it has finished processing the READ request, the user daemon should reply
651 by using the CACHEFILES_IOC_READ_COMPLETE ioctl on one of the anonymous fds
652 associated with the object_id given in the READ request. The ioctl is of the
655 ioctl(fd, CACHEFILES_IOC_READ_COMPLETE, msg_id);
659 * ``fd`` is one of the anonymous fds associated with the object_id
662 * ``msg_id`` must match the msg_id field of the READ request.