1.. SPDX-License-Identifier: GPL-2.0
2
3===================================
4Cache on Already Mounted Filesystem
5===================================
6
7.. Contents:
8
9 (*) Overview.
10
11 (*) Requirements.
12
13 (*) Configuration.
14
15 (*) Starting the cache.
16
17 (*) Things to avoid.
18
19 (*) Cache culling.
20
21 (*) Cache structure.
22
23 (*) Security model and SELinux.
24
25 (*) A note on security.
26
27 (*) Statistical information.
28
29 (*) Debugging.
30
31 (*) On-demand Read.
32
33
34Overview
35========
36
37CacheFiles is a caching backend that's meant to use as a cache a directory on
38an already mounted filesystem of a local type (such as Ext3).
39
40CacheFiles uses a userspace daemon to do some of the cache management - such as
41reaping stale nodes and culling.  This is called cachefilesd and lives in
42/sbin.
43
44The filesystem and data integrity of the cache are only as good as those of the
45filesystem providing the backing services.  Note that CacheFiles does not
46attempt to journal anything since the journalling interfaces of the various
47filesystems are very specific in nature.
48
49CacheFiles creates a misc character device - "/dev/cachefiles" - that is used
50to communication with the daemon.  Only one thing may have this open at once,
51and while it is open, a cache is at least partially in existence.  The daemon
52opens this and sends commands down it to control the cache.
53
54CacheFiles is currently limited to a single cache.
55
56CacheFiles attempts to maintain at least a certain percentage of free space on
57the filesystem, shrinking the cache by culling the objects it contains to make
58space if necessary - see the "Cache Culling" section.  This means it can be
59placed on the same medium as a live set of data, and will expand to make use of
60spare space and automatically contract when the set of data requires more
61space.
62
63
64
65Requirements
66============
67
68The use of CacheFiles and its daemon requires the following features to be
69available in the system and in the cache filesystem:
70
71	- dnotify.
72
73	- extended attributes (xattrs).
74
75	- openat() and friends.
76
77	- bmap() support on files in the filesystem (FIBMAP ioctl).
78
79	- The use of bmap() to detect a partial page at the end of the file.
80
81It is strongly recommended that the "dir_index" option is enabled on Ext3
82filesystems being used as a cache.
83
84
85Configuration
86=============
87
88The cache is configured by a script in /etc/cachefilesd.conf.  These commands
89set up cache ready for use.  The following script commands are available:
90
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.
94
95	The commands beginning with a 'b' are file space (block) limits, those
96	beginning with an 'f' are file count limits.
97
98 dir <path>
99	Specify the directory containing the root of the cache.  Mandatory.
100
101 tag <name>
102	Specify a tag to FS-Cache to use in distinguishing multiple caches.
103	Optional.  The default is "CacheFiles".
104
105 debug <mask>
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:
109
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		==	=================================================
115
116	This mask can also be set through sysfs, eg::
117
118		echo 5 >/sys/modules/cachefiles/parameters/debug
119
120
121Starting the Cache
122==================
123
124The cache is started by running the daemon.  The daemon opens the cache device,
125configures the cache and tells it to begin caching.  At that point the cache
126binds to fscache and the cache becomes live.
127
128The daemon is run as follows::
129
130	/sbin/cachefilesd [-d]* [-s] [-n] [-f <configfile>]
131
132The flags are:
133
134 ``-d``
135	Increase the debugging level.  This can be specified multiple times and
136	is cumulative with itself.
137
138 ``-s``
139	Send messages to stderr instead of syslog.
140
141 ``-n``
142	Don't daemonise and go into background.
143
144 ``-f <configfile>``
145	Use an alternative configuration file rather than the default one.
146
147
148Things to Avoid
149===============
150
151Do not mount other things within the cache as this will cause problems.  The
152kernel module contains its own very cut-down path walking facility that ignores
153mountpoints, but the daemon can't avoid them.
154
155Do not create, rename or unlink files and directories in the cache while the
156cache is active, as this may cause the state to become uncertain.
157
158Renaming files in the cache might make objects appear to be other objects (the
159filename is part of the lookup key).
160
161Do not change or remove the extended attributes attached to cache files by the
162cache as this will cause the cache state management to get confused.
163
164Do not create files or directories in the cache, lest the cache get confused or
165serve incorrect data.
166
167Do not chmod files in the cache.  The module creates things with minimal
168permissions to prevent random users being able to access them directly.
169
170
171Cache Culling
172=============
173
174The cache may need culling occasionally to make space.  This involves
175discarding objects from the cache that have been used less recently than
176anything else.  Culling is based on the access time of data objects.  Empty
177directories are culled if not in use.
178
179Cache culling is done on the basis of the percentage of blocks and the
180percentage of files available in the underlying filesystem.  There are six
181"limits":
182
183 brun, frun
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.
186
187 bcull, fcull
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.
190
191 bstop, fstop
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
195     these limits again.
196
197These must be configured thusly::
198
199	0 <= bstop < bcull < brun < 100
200	0 <= fstop < fcull < frun < 100
201
202Note 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.
204
205The userspace daemon scans the cache to build up a table of cullable objects.
206These are then culled in least recently used order.  A new scan of the cache is
207started as soon as space is made in the table.  Objects will be skipped if
208their atimes have changed or if the kernel module says it is still using them.
209
210
211Cache Structure
212===============
213
214The CacheFiles module will create two directories in the directory it was
215given:
216
217 * cache/
218 * graveyard/
219
220The active cache objects all reside in the first directory.  The CacheFiles
221kernel module moves any retired or culled objects that it can't simply unlink
222to the graveyard from which the daemon will actually delete them.
223
224The daemon uses dnotify to monitor the graveyard directory, and will delete
225anything that appears therein.
226
227
228The module represents index objects as directories with the filename "I..." or
229"J...".  Note that the "cache/" directory is itself a special index.
230
231Data objects are represented as files if they have no children, or directories
232if they do.  Their filenames all begin "D..." or "E...".  If represented as a
233directory, data objects will have a file in the directory called "data" that
234actually holds the data.
235
236Special objects are similar to data objects, except their filenames begin
237"S..." or "T...".
238
239
240If an object has children, then it will be represented as a directory.
241Immediately in the representative directory are a collection of directories
242named for hash values of the child object keys with an '@' prepended.  Into
243this directory, if possible, will be placed the representations of the child
244objects::
245
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
252
253
254If the key is so long that it exceeds NAME_MAX with the decorations added on to
255it, then it will be cut into pieces, the first few of which will be used to
256make a nest of directories, and the last one of which will be the objects
257inside the last directory.  The names of the intermediate directories will have
258'+' prepended::
259
260	J1223/@23/+xy...z/+kl...m/Epqr
261
262
263Note that keys are raw data, and not only may they exceed NAME_MAX in size,
264they may also contain things like '/' and NUL characters, and so they may not
265be suitable for turning directly into a filename.
266
267To 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
269object filenames indicate the encoding:
270
271	===============	===============	===============
272	OBJECT TYPE	PRINTABLE	ENCODED
273	===============	===============	===============
274	Index		"I..."		"J..."
275	Data		"D..."		"E..."
276	Special		"S..."		"T..."
277	===============	===============	===============
278
279Intermediate directories are always "@" or "+" as appropriate.
280
281
282Each object in the cache has an extended attribute label that holds the object
283type ID (required to distinguish special objects) and the auxiliary data from
284the netfs.  The latter is used to detect stale objects in the cache and update
285or retire them.
286
287
288Note that CacheFiles will erase from the cache any file it doesn't recognise or
289any file of an incorrect type (such as a FIFO file or a device file).
290
291
292Security Model and SELinux
293==========================
294
295CacheFiles is implemented to deal properly with the LSM security features of
296the Linux kernel and the SELinux facility.
297
298One of the problems that CacheFiles faces is that it is generally acting on
299behalf of a process, and running in that process's context, and that includes a
300security context that is not appropriate for accessing the cache - either
301because the files in the cache are inaccessible to that process, or because if
302the process creates a file in the cache, that file may be inaccessible to other
303processes.
304
305The way CacheFiles works is to temporarily change the security context (fsuid,
306fsgid and actor security label) that the process acts as - without changing the
307security context of the process when it the target of an operation performed by
308some other process (so signalling and suchlike still work correctly).
309
310
311When the CacheFiles module is asked to bind to its cache, it:
312
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,
315     this is::
316
317	cachefiles_var_t
318
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::
321
322	cachefilesd_t
323
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::
326
327	cachefiles_kernel_t
328
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::
331
332	type_transition <daemon's-ID> kernel_t : process <module's-ID>;
333
334     For instance::
335
336	type_transition cachefilesd_t kernel_t : process cachefiles_kernel_t;
337
338
339The module's security ID gives it permission to create, move and remove files
340and directories in the cache, to find and access directories and files in the
341cache, to set and access extended attributes on cache objects, and to read and
342write files in the cache.
343
344The daemon's security ID gives it only a very restricted set of permissions: it
345may scan directories, stat files and erase files and directories.  It may
346not read or write files in the cache, and so it is precluded from accessing the
347data cached therein; nor is it permitted to create new files in the cache.
348
349
350There are policy source files available in:
351
352	https://people.redhat.com/~dhowells/fscache/cachefilesd-0.8.tar.bz2
353
354and later versions.  In that tarball, see the files::
355
356	cachefilesd.te
357	cachefilesd.fc
358	cachefilesd.if
359
360They are built and installed directly by the RPM.
361
362If a non-RPM based system is being used, then copy the above files to their own
363directory and run::
364
365	make -f /usr/share/selinux/devel/Makefile
366	semodule -i cachefilesd.pp
367
368You will need checkpolicy and selinux-policy-devel installed prior to the
369build.
370
371
372By default, the cache is located in /var/fscache, but if it is desirable that
373it should be elsewhere, than either the above policy files must be altered, or
374an auxiliary policy must be installed to label the alternate location of the
375cache.
376
377For instructions on how to add an auxiliary policy to enable the cache to be
378located elsewhere when SELinux is in enforcing mode, please see::
379
380	/usr/share/doc/cachefilesd-*/move-cache.txt
381
382When the cachefilesd rpm is installed; alternatively, the document can be found
383in the sources.
384
385
386A Note on Security
387==================
388
389CacheFiles makes use of the split security in the task_struct.  It allocates
390its own task_security structure, and redirects current->cred to point to it
391when it acts on behalf of another process, in that process's context.
392
393The reason it does this is that it calls vfs_mkdir() and suchlike rather than
394bypassing security and calling inode ops directly.  Therefore the VFS and LSM
395may deny the CacheFiles access to the cache data because under some
396circumstances the caching code is running in the security context of whatever
397process issued the original syscall on the netfs.
398
399Furthermore, should CacheFiles create a file or directory, the security
400parameters with that object is created (UID, GID, security label) would be
401derived from that process that issued the system call, thus potentially
402preventing other processes from accessing the cache - including CacheFiles's
403cache management daemon (cachefilesd).
404
405What is required is to temporarily override the security of the process that
406issued the system call.  We can't, however, just do an in-place change of the
407security data as that affects the process as an object, not just as a subject.
408This means it may lose signals or ptrace events for example, and affects what
409the process looks like in /proc.
410
411So CacheFiles makes use of a logical split in the security between the
412objective security (task->real_cred) and the subjective security (task->cred).
413The objective security holds the intrinsic security properties of a process and
414is never overridden.  This is what appears in /proc, and is what is used when a
415process is the target of an operation by some other process (SIGKILL for
416example).
417
418The subjective security holds the active security properties of a process, and
419may be overridden.  This is not seen externally, and is used whan a process
420acts upon another object, for example SIGKILLing another process or opening a
421file.
422
423LSM hooks exist that allow SELinux (or Smack or whatever) to reject a request
424for CacheFiles to run in a context of a specific security label, or to create
425files and directories with another security label.
426
427
428Statistical Information
429=======================
430
431If FS-Cache is compiled with the following option enabled::
432
433	CONFIG_CACHEFILES_HISTOGRAM=y
434
435then it will gather certain statistics and display them through a proc file.
436
437 /proc/fs/cachefiles/histogram
438
439     ::
440
441	cat /proc/fs/cachefiles/histogram
442	JIFS  SECS  LOOKUPS   MKDIRS    CREATES
443	===== ===== ========= ========= =========
444
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:
448
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	=======		=======================================================
456
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.
460
461
462Debugging
463=========
464
465If CONFIG_CACHEFILES_DEBUG is enabled, the CacheFiles facility can have runtime
466debugging enabled by adjusting the value in::
467
468	/sys/module/cachefiles/parameters/debug
469
470This is a bitmask of debugging streams to enable:
471
472	=======	=======	===============================	=======================
473	BIT	VALUE	STREAM				POINT
474	=======	=======	===============================	=======================
475	0	1	General				Function entry trace
476	1	2					Function exit trace
477	2	4					General
478	=======	=======	===============================	=======================
479
480The appropriate set of values should be OR'd together and the result written to
481the control file.  For example::
482
483	echo $((1|4|8)) >/sys/module/cachefiles/parameters/debug
484
485will turn on all function entry debugging.
486
487
488On-demand Read
489==============
490
491When working in its original mode, CacheFiles serves as a local cache for a
492remote networking fs - while in on-demand read mode, CacheFiles can boost the
493scenario where on-demand read semantics are needed, e.g. container image
494distribution.
495
496The essential difference between these two modes is seen when a cache miss
497occurs: In the original mode, the netfs will fetch the data from the remote
498server and then write it to the cache file; in on-demand read mode, fetching
499the data and writing it into the cache is delegated to a user daemon.
500
501``CONFIG_CACHEFILES_ONDEMAND`` should be enabled to support on-demand read mode.
502
503
504Protocol Communication
505----------------------
506
507The on-demand read mode uses a simple protocol for communication between kernel
508and user daemon. The protocol can be modeled as::
509
510	kernel --[request]--> user daemon --[reply]--> kernel
511
512CacheFiles will send requests to the user daemon when needed.  The user daemon
513should poll the devnode ('/dev/cachefiles') to check if there's a pending
514request to be processed.  A POLLIN event will be returned when there's a pending
515request.
516
517The user daemon then reads the devnode to fetch a request to process.  It should
518be noted that each read only gets one request. When it has finished processing
519the request, the user daemon should write the reply to the devnode.
520
521Each request starts with a message header of the form::
522
523	struct cachefiles_msg {
524		__u32 msg_id;
525		__u32 opcode;
526		__u32 len;
527		__u32 object_id;
528		__u8  data[];
529	};
530
531where:
532
533	* ``msg_id`` is a unique ID identifying this request among all pending
534	  requests.
535
536	* ``opcode`` indicates the type of this request.
537
538	* ``object_id`` is a unique ID identifying the cache file operated on.
539
540	* ``data`` indicates the payload of this request.
541
542	* ``len`` indicates the whole length of this request, including the
543	  header and following type-specific payload.
544
545
546Turning on On-demand Mode
547-------------------------
548
549An optional parameter becomes available to the "bind" command::
550
551	bind [ondemand]
552
553When the "bind" command is given no argument, it defaults to the original mode.
554When it is given the "ondemand" argument, i.e. "bind ondemand", on-demand read
555mode will be enabled.
556
557
558The OPEN Request
559----------------
560
561When the netfs opens a cache file for the first time, a request with the
562CACHEFILES_OP_OPEN opcode, a.k.a an OPEN request will be sent to the user
563daemon.  The payload format is of the form::
564
565	struct cachefiles_open {
566		__u32 volume_key_size;
567		__u32 cookie_key_size;
568		__u32 fd;
569		__u32 flags;
570		__u8  data[];
571	};
572
573where:
574
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
577	  data.
578
579	* ``volume_key_size`` indicates the size of the volume key in bytes.
580
581	* ``cookie_key_size`` indicates the size of the cookie key in bytes.
582
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
585	  cache file.
586
587
588The user daemon can use the given (volume_key, cookie_key) pair to distinguish
589the requested cache file.  With the given anonymous fd, the user daemon can
590fetch the data and write it to the cache file in the background, even when
591kernel has not triggered a cache miss yet.
592
593Be noted that each cache file has a unique object_id, while it may have multiple
594anonymous fds.  The user daemon may duplicate anonymous fds from the initial
595anonymous fd indicated by the @fd field through dup().  Thus each object_id can
596be mapped to multiple anonymous fds, while the usr daemon itself needs to
597maintain the mapping.
598
599When 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
601be huge since they're related to the number of open device blobs rather than
602open files of each individual filesystem.
603
604The user daemon should reply the OPEN request by issuing a "copen" (complete
605open) command on the devnode::
606
607	copen <msg_id>,<cache_size>
608
609where:
610
611	* ``msg_id`` must match the msg_id field of the OPEN request.
612
613	* When >= 0, ``cache_size`` indicates the size of the cache file;
614	  when < 0, ``cache_size`` indicates any error code encountered by the
615	  user daemon.
616
617
618The CLOSE Request
619-----------------
620
621When a cookie withdrawn, a CLOSE request (opcode CACHEFILES_OP_CLOSE) will be
622sent to the user daemon.  This tells the user daemon to close all anonymous fds
623associated with the given object_id.  The CLOSE request has no extra payload,
624and shouldn't be replied.
625
626
627The READ Request
628----------------
629
630When a cache miss is encountered in on-demand read mode, CacheFiles will send a
631READ request (opcode CACHEFILES_OP_READ) to the user daemon. This tells the user
632daemon to fetch the contents of the requested file range.  The payload is of the
633form::
634
635	struct cachefiles_read {
636		__u64 off;
637		__u64 len;
638	};
639
640where:
641
642	* ``off`` indicates the starting offset of the requested file range.
643
644	* ``len`` indicates the length of the requested file range.
645
646
647When it receives a READ request, the user daemon should fetch the requested data
648and write it to the cache file identified by object_id.
649
650When it has finished processing the READ request, the user daemon should reply
651by using the CACHEFILES_IOC_READ_COMPLETE ioctl on one of the anonymous fds
652associated with the object_id given in the READ request.  The ioctl is of the
653form::
654
655	ioctl(fd, CACHEFILES_IOC_READ_COMPLETE, msg_id);
656
657where:
658
659	* ``fd`` is one of the anonymous fds associated with the object_id
660	  given.
661
662	* ``msg_id`` must match the msg_id field of the READ request.
663